2008-03-24 20:01:56 +01:00
|
|
|
/*
|
|
|
|
* Copyright (C) 2007 Oracle. All rights reserved.
|
|
|
|
*
|
|
|
|
* This program is free software; you can redistribute it and/or
|
|
|
|
* modify it under the terms of the GNU General Public
|
|
|
|
* License v2 as published by the Free Software Foundation.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* General Public License for more details.
|
|
|
|
*
|
|
|
|
* You should have received a copy of the GNU General Public
|
|
|
|
* License along with this program; if not, write to the
|
|
|
|
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
|
|
|
|
* Boston, MA 021110-1307, USA.
|
|
|
|
*/
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/bio.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 09:04:11 +01:00
|
|
|
#include <linux/slab.h>
|
2008-03-24 20:02:07 +01:00
|
|
|
#include <linux/buffer_head.h>
|
2008-04-21 16:03:05 +02:00
|
|
|
#include <linux/blkdev.h>
|
2008-04-28 21:29:42 +02:00
|
|
|
#include <linux/random.h>
|
2009-04-03 16:27:10 +02:00
|
|
|
#include <linux/iocontext.h>
|
2010-12-29 15:55:03 +01:00
|
|
|
#include <linux/capability.h>
|
2008-03-25 21:50:33 +01:00
|
|
|
#include <asm/div64.h>
|
2008-11-20 16:22:27 +01:00
|
|
|
#include "compat.h"
|
2008-03-24 20:01:56 +01:00
|
|
|
#include "ctree.h"
|
|
|
|
#include "extent_map.h"
|
|
|
|
#include "disk-io.h"
|
|
|
|
#include "transaction.h"
|
|
|
|
#include "print-tree.h"
|
|
|
|
#include "volumes.h"
|
2008-06-11 22:50:36 +02:00
|
|
|
#include "async-thread.h"
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
static int init_first_rw_device(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_device *device);
|
|
|
|
static int btrfs_relocate_sys_chunks(struct btrfs_root *root);
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
static DEFINE_MUTEX(uuid_mutex);
|
|
|
|
static LIST_HEAD(fs_uuids);
|
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
static void lock_chunks(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
mutex_lock(&root->fs_info->chunk_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void unlock_chunks(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
mutex_unlock(&root->fs_info->chunk_mutex);
|
|
|
|
}
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
static void free_fs_devices(struct btrfs_fs_devices *fs_devices)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
WARN_ON(fs_devices->opened);
|
|
|
|
while (!list_empty(&fs_devices->devices)) {
|
|
|
|
device = list_entry(fs_devices->devices.next,
|
|
|
|
struct btrfs_device, dev_list);
|
|
|
|
list_del(&device->dev_list);
|
|
|
|
kfree(device->name);
|
|
|
|
kfree(device);
|
|
|
|
}
|
|
|
|
kfree(fs_devices);
|
|
|
|
}
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
int btrfs_cleanup_fs_uuids(void)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_devices *fs_devices;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
while (!list_empty(&fs_uuids)) {
|
|
|
|
fs_devices = list_entry(fs_uuids.next,
|
|
|
|
struct btrfs_fs_devices, list);
|
|
|
|
list_del(&fs_devices->list);
|
2008-12-12 16:03:26 +01:00
|
|
|
free_fs_devices(fs_devices);
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-09-05 22:09:51 +02:00
|
|
|
static noinline struct btrfs_device *__find_device(struct list_head *head,
|
|
|
|
u64 devid, u8 *uuid)
|
2008-03-24 20:02:07 +01:00
|
|
|
{
|
|
|
|
struct btrfs_device *dev;
|
|
|
|
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(dev, head, dev_list) {
|
2008-04-18 16:29:38 +02:00
|
|
|
if (dev->devid == devid &&
|
2008-04-25 22:53:30 +02:00
|
|
|
(!uuid || !memcmp(dev->uuid, uuid, BTRFS_UUID_SIZE))) {
|
2008-03-24 20:02:07 +01:00
|
|
|
return dev;
|
2008-04-18 16:29:38 +02:00
|
|
|
}
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-09-05 22:09:51 +02:00
|
|
|
static noinline struct btrfs_fs_devices *find_fsid(u8 *fsid)
|
2008-03-24 20:02:07 +01:00
|
|
|
{
|
|
|
|
struct btrfs_fs_devices *fs_devices;
|
|
|
|
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(fs_devices, &fs_uuids, list) {
|
2008-03-24 20:02:07 +01:00
|
|
|
if (memcmp(fsid, fs_devices->fsid, BTRFS_FSID_SIZE) == 0)
|
|
|
|
return fs_devices;
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2009-04-20 21:50:09 +02:00
|
|
|
static void requeue_list(struct btrfs_pending_bios *pending_bios,
|
|
|
|
struct bio *head, struct bio *tail)
|
|
|
|
{
|
|
|
|
|
|
|
|
struct bio *old_head;
|
|
|
|
|
|
|
|
old_head = pending_bios->head;
|
|
|
|
pending_bios->head = head;
|
|
|
|
if (pending_bios->tail)
|
|
|
|
tail->bi_next = old_head;
|
|
|
|
else
|
|
|
|
pending_bios->tail = tail;
|
|
|
|
}
|
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
/*
|
|
|
|
* we try to collect pending bios for a device so we don't get a large
|
|
|
|
* number of procs sending bios down to the same device. This greatly
|
|
|
|
* improves the schedulers ability to collect and merge the bios.
|
|
|
|
*
|
|
|
|
* But, it also turns into a long list of bios to process and that is sure
|
|
|
|
* to eventually make the worker thread block. The solution here is to
|
|
|
|
* make some progress and then put this work struct back at the end of
|
|
|
|
* the list if the block device is congested. This way, multiple devices
|
|
|
|
* can make progress from a single worker thread.
|
|
|
|
*/
|
2009-01-06 03:25:51 +01:00
|
|
|
static noinline int run_scheduled_bios(struct btrfs_device *device)
|
2008-06-11 22:50:36 +02:00
|
|
|
{
|
|
|
|
struct bio *pending;
|
|
|
|
struct backing_dev_info *bdi;
|
2008-08-20 19:39:41 +02:00
|
|
|
struct btrfs_fs_info *fs_info;
|
2009-04-20 21:50:09 +02:00
|
|
|
struct btrfs_pending_bios *pending_bios;
|
2008-06-11 22:50:36 +02:00
|
|
|
struct bio *tail;
|
|
|
|
struct bio *cur;
|
|
|
|
int again = 0;
|
2009-04-20 21:50:09 +02:00
|
|
|
unsigned long num_run;
|
2009-06-09 21:59:22 +02:00
|
|
|
unsigned long batch_run = 0;
|
2008-08-20 19:39:41 +02:00
|
|
|
unsigned long limit;
|
2009-04-03 16:27:10 +02:00
|
|
|
unsigned long last_waited = 0;
|
2009-06-09 21:39:08 +02:00
|
|
|
int force_reg = 0;
|
2011-08-05 11:32:37 +02:00
|
|
|
int sync_pending = 0;
|
2011-04-20 02:12:40 +02:00
|
|
|
struct blk_plug plug;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this function runs all the bios we've collected for
|
|
|
|
* a particular device. We don't want to wander off to
|
|
|
|
* another device without first sending all of these down.
|
|
|
|
* So, setup a plug here and finish it off before we return
|
|
|
|
*/
|
|
|
|
blk_start_plug(&plug);
|
2008-06-11 22:50:36 +02:00
|
|
|
|
2009-04-03 16:32:58 +02:00
|
|
|
bdi = blk_get_backing_dev_info(device->bdev);
|
2008-08-20 19:39:41 +02:00
|
|
|
fs_info = device->dev_root->fs_info;
|
|
|
|
limit = btrfs_async_submit_limit(fs_info);
|
|
|
|
limit = limit * 2 / 3;
|
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
loop:
|
|
|
|
spin_lock(&device->io_lock);
|
|
|
|
|
2009-02-04 15:19:41 +01:00
|
|
|
loop_lock:
|
2009-06-09 21:39:08 +02:00
|
|
|
num_run = 0;
|
2009-04-20 21:50:09 +02:00
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
/* take all the bios off the list at once and process them
|
|
|
|
* later on (without the lock held). But, remember the
|
|
|
|
* tail and other pointers so the bios can be properly reinserted
|
|
|
|
* into the list if we hit congestion
|
|
|
|
*/
|
2009-06-09 21:39:08 +02:00
|
|
|
if (!force_reg && device->pending_sync_bios.head) {
|
2009-04-20 21:50:09 +02:00
|
|
|
pending_bios = &device->pending_sync_bios;
|
2009-06-09 21:39:08 +02:00
|
|
|
force_reg = 1;
|
|
|
|
} else {
|
2009-04-20 21:50:09 +02:00
|
|
|
pending_bios = &device->pending_bios;
|
2009-06-09 21:39:08 +02:00
|
|
|
force_reg = 0;
|
|
|
|
}
|
2009-04-20 21:50:09 +02:00
|
|
|
|
|
|
|
pending = pending_bios->head;
|
|
|
|
tail = pending_bios->tail;
|
2008-06-11 22:50:36 +02:00
|
|
|
WARN_ON(pending && !tail);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if pending was null this time around, no bios need processing
|
|
|
|
* at all and we can stop. Otherwise it'll loop back up again
|
|
|
|
* and do an additional check so no bios are missed.
|
|
|
|
*
|
|
|
|
* device->running_pending is used to synchronize with the
|
|
|
|
* schedule_bio code.
|
|
|
|
*/
|
2009-04-20 21:50:09 +02:00
|
|
|
if (device->pending_sync_bios.head == NULL &&
|
|
|
|
device->pending_bios.head == NULL) {
|
2008-06-11 22:50:36 +02:00
|
|
|
again = 0;
|
|
|
|
device->running_pending = 0;
|
2009-04-20 21:50:09 +02:00
|
|
|
} else {
|
|
|
|
again = 1;
|
|
|
|
device->running_pending = 1;
|
2008-06-11 22:50:36 +02:00
|
|
|
}
|
2009-04-20 21:50:09 +02:00
|
|
|
|
|
|
|
pending_bios->head = NULL;
|
|
|
|
pending_bios->tail = NULL;
|
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
spin_unlock(&device->io_lock);
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
while (pending) {
|
2009-04-20 21:50:09 +02:00
|
|
|
|
|
|
|
rmb();
|
2009-06-09 21:39:08 +02:00
|
|
|
/* we want to work on both lists, but do more bios on the
|
|
|
|
* sync list than the regular list
|
|
|
|
*/
|
|
|
|
if ((num_run > 32 &&
|
|
|
|
pending_bios != &device->pending_sync_bios &&
|
|
|
|
device->pending_sync_bios.head) ||
|
|
|
|
(num_run > 64 && pending_bios == &device->pending_sync_bios &&
|
|
|
|
device->pending_bios.head)) {
|
2009-04-20 21:50:09 +02:00
|
|
|
spin_lock(&device->io_lock);
|
|
|
|
requeue_list(pending_bios, pending, tail);
|
|
|
|
goto loop_lock;
|
|
|
|
}
|
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
cur = pending;
|
|
|
|
pending = pending->bi_next;
|
|
|
|
cur->bi_next = NULL;
|
2008-08-20 19:39:41 +02:00
|
|
|
atomic_dec(&fs_info->nr_async_bios);
|
|
|
|
|
|
|
|
if (atomic_read(&fs_info->nr_async_bios) < limit &&
|
|
|
|
waitqueue_active(&fs_info->async_submit_wait))
|
|
|
|
wake_up(&fs_info->async_submit_wait);
|
2008-07-31 22:29:02 +02:00
|
|
|
|
|
|
|
BUG_ON(atomic_read(&cur->bi_cnt) == 0);
|
2009-06-09 21:59:22 +02:00
|
|
|
|
2011-08-04 20:28:36 +02:00
|
|
|
/*
|
|
|
|
* if we're doing the sync list, record that our
|
|
|
|
* plug has some sync requests on it
|
|
|
|
*
|
|
|
|
* If we're doing the regular list and there are
|
|
|
|
* sync requests sitting around, unplug before
|
|
|
|
* we add more
|
|
|
|
*/
|
|
|
|
if (pending_bios == &device->pending_sync_bios) {
|
|
|
|
sync_pending = 1;
|
|
|
|
} else if (sync_pending) {
|
|
|
|
blk_finish_plug(&plug);
|
|
|
|
blk_start_plug(&plug);
|
|
|
|
sync_pending = 0;
|
|
|
|
}
|
|
|
|
|
2010-03-15 15:21:30 +01:00
|
|
|
submit_bio(cur->bi_rw, cur);
|
|
|
|
num_run++;
|
|
|
|
batch_run++;
|
2011-03-10 08:52:07 +01:00
|
|
|
if (need_resched())
|
2009-04-20 21:50:09 +02:00
|
|
|
cond_resched();
|
2008-06-11 22:50:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* we made progress, there is more work to do and the bdi
|
|
|
|
* is now congested. Back off and let other work structs
|
|
|
|
* run instead
|
|
|
|
*/
|
2009-08-07 15:59:15 +02:00
|
|
|
if (pending && bdi_write_congested(bdi) && batch_run > 8 &&
|
2008-11-08 00:22:45 +01:00
|
|
|
fs_info->fs_devices->open_devices > 1) {
|
2009-04-03 16:27:10 +02:00
|
|
|
struct io_context *ioc;
|
2008-06-11 22:50:36 +02:00
|
|
|
|
2009-04-03 16:27:10 +02:00
|
|
|
ioc = current->io_context;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the main goal here is that we don't want to
|
|
|
|
* block if we're going to be able to submit
|
|
|
|
* more requests without blocking.
|
|
|
|
*
|
|
|
|
* This code does two great things, it pokes into
|
|
|
|
* the elevator code from a filesystem _and_
|
|
|
|
* it makes assumptions about how batching works.
|
|
|
|
*/
|
|
|
|
if (ioc && ioc->nr_batch_requests > 0 &&
|
|
|
|
time_before(jiffies, ioc->last_waited + HZ/50UL) &&
|
|
|
|
(last_waited == 0 ||
|
|
|
|
ioc->last_waited == last_waited)) {
|
|
|
|
/*
|
|
|
|
* we want to go through our batch of
|
|
|
|
* requests and stop. So, we copy out
|
|
|
|
* the ioc->last_waited time and test
|
|
|
|
* against it before looping
|
|
|
|
*/
|
|
|
|
last_waited = ioc->last_waited;
|
2011-03-10 08:52:07 +01:00
|
|
|
if (need_resched())
|
2009-04-20 21:50:09 +02:00
|
|
|
cond_resched();
|
2009-04-03 16:27:10 +02:00
|
|
|
continue;
|
|
|
|
}
|
2008-06-11 22:50:36 +02:00
|
|
|
spin_lock(&device->io_lock);
|
2009-04-20 21:50:09 +02:00
|
|
|
requeue_list(pending_bios, pending, tail);
|
2009-02-04 15:19:41 +01:00
|
|
|
device->running_pending = 1;
|
2008-06-11 22:50:36 +02:00
|
|
|
|
|
|
|
spin_unlock(&device->io_lock);
|
|
|
|
btrfs_requeue_work(&device->work);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
}
|
2009-04-20 21:50:09 +02:00
|
|
|
|
2010-03-10 21:33:32 +01:00
|
|
|
cond_resched();
|
|
|
|
if (again)
|
|
|
|
goto loop;
|
|
|
|
|
|
|
|
spin_lock(&device->io_lock);
|
|
|
|
if (device->pending_bios.head || device->pending_sync_bios.head)
|
|
|
|
goto loop_lock;
|
|
|
|
spin_unlock(&device->io_lock);
|
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
done:
|
2011-04-20 02:12:40 +02:00
|
|
|
blk_finish_plug(&plug);
|
2008-06-11 22:50:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-12-02 15:54:17 +01:00
|
|
|
static void pending_bios_fn(struct btrfs_work *work)
|
2008-06-11 22:50:36 +02:00
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
|
|
|
|
device = container_of(work, struct btrfs_device, work);
|
|
|
|
run_scheduled_bios(device);
|
|
|
|
}
|
|
|
|
|
2008-09-05 22:09:51 +02:00
|
|
|
static noinline int device_list_add(const char *path,
|
2008-03-24 20:02:07 +01:00
|
|
|
struct btrfs_super_block *disk_super,
|
|
|
|
u64 devid, struct btrfs_fs_devices **fs_devices_ret)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
struct btrfs_fs_devices *fs_devices;
|
|
|
|
u64 found_transid = btrfs_super_generation(disk_super);
|
2010-02-09 07:36:45 +01:00
|
|
|
char *name;
|
2008-03-24 20:02:07 +01:00
|
|
|
|
|
|
|
fs_devices = find_fsid(disk_super->fsid);
|
|
|
|
if (!fs_devices) {
|
2008-05-16 19:30:15 +02:00
|
|
|
fs_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
|
2008-03-24 20:02:07 +01:00
|
|
|
if (!fs_devices)
|
|
|
|
return -ENOMEM;
|
|
|
|
INIT_LIST_HEAD(&fs_devices->devices);
|
2008-04-22 15:22:07 +02:00
|
|
|
INIT_LIST_HEAD(&fs_devices->alloc_list);
|
2008-03-24 20:02:07 +01:00
|
|
|
list_add(&fs_devices->list, &fs_uuids);
|
|
|
|
memcpy(fs_devices->fsid, disk_super->fsid, BTRFS_FSID_SIZE);
|
|
|
|
fs_devices->latest_devid = devid;
|
|
|
|
fs_devices->latest_trans = found_transid;
|
2009-06-10 21:17:02 +02:00
|
|
|
mutex_init(&fs_devices->device_list_mutex);
|
2008-03-24 20:02:07 +01:00
|
|
|
device = NULL;
|
|
|
|
} else {
|
2008-04-18 16:29:38 +02:00
|
|
|
device = __find_device(&fs_devices->devices, devid,
|
|
|
|
disk_super->dev_item.uuid);
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
|
|
|
if (!device) {
|
2008-11-18 03:11:30 +01:00
|
|
|
if (fs_devices->opened)
|
|
|
|
return -EBUSY;
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
device = kzalloc(sizeof(*device), GFP_NOFS);
|
|
|
|
if (!device) {
|
|
|
|
/* we can safely leave the fs_devices entry around */
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
device->devid = devid;
|
2008-06-11 22:50:36 +02:00
|
|
|
device->work.func = pending_bios_fn;
|
2008-04-18 16:29:38 +02:00
|
|
|
memcpy(device->uuid, disk_super->dev_item.uuid,
|
|
|
|
BTRFS_UUID_SIZE);
|
2008-04-14 15:48:18 +02:00
|
|
|
spin_lock_init(&device->io_lock);
|
2008-03-24 20:02:07 +01:00
|
|
|
device->name = kstrdup(path, GFP_NOFS);
|
|
|
|
if (!device->name) {
|
|
|
|
kfree(device);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
INIT_LIST_HEAD(&device->dev_alloc_list);
|
2009-06-10 21:17:02 +02:00
|
|
|
|
|
|
|
mutex_lock(&fs_devices->device_list_mutex);
|
2011-04-20 12:09:16 +02:00
|
|
|
list_add_rcu(&device->dev_list, &fs_devices->devices);
|
2009-06-10 21:17:02 +02:00
|
|
|
mutex_unlock(&fs_devices->device_list_mutex);
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
device->fs_devices = fs_devices;
|
2008-03-24 20:02:07 +01:00
|
|
|
fs_devices->num_devices++;
|
2010-12-13 20:56:23 +01:00
|
|
|
} else if (!device->name || strcmp(device->name, path)) {
|
2010-02-09 07:36:45 +01:00
|
|
|
name = kstrdup(path, GFP_NOFS);
|
|
|
|
if (!name)
|
|
|
|
return -ENOMEM;
|
|
|
|
kfree(device->name);
|
|
|
|
device->name = name;
|
2010-12-13 20:56:23 +01:00
|
|
|
if (device->missing) {
|
|
|
|
fs_devices->missing_devices--;
|
|
|
|
device->missing = 0;
|
|
|
|
}
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (found_transid > fs_devices->latest_trans) {
|
|
|
|
fs_devices->latest_devid = devid;
|
|
|
|
fs_devices->latest_trans = found_transid;
|
|
|
|
}
|
|
|
|
*fs_devices_ret = fs_devices;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
static struct btrfs_fs_devices *clone_fs_devices(struct btrfs_fs_devices *orig)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_devices *fs_devices;
|
|
|
|
struct btrfs_device *device;
|
|
|
|
struct btrfs_device *orig_dev;
|
|
|
|
|
|
|
|
fs_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
|
|
|
|
if (!fs_devices)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&fs_devices->devices);
|
|
|
|
INIT_LIST_HEAD(&fs_devices->alloc_list);
|
|
|
|
INIT_LIST_HEAD(&fs_devices->list);
|
2009-06-10 21:17:02 +02:00
|
|
|
mutex_init(&fs_devices->device_list_mutex);
|
2008-12-12 16:03:26 +01:00
|
|
|
fs_devices->latest_devid = orig->latest_devid;
|
|
|
|
fs_devices->latest_trans = orig->latest_trans;
|
|
|
|
memcpy(fs_devices->fsid, orig->fsid, sizeof(fs_devices->fsid));
|
|
|
|
|
2011-04-20 12:08:47 +02:00
|
|
|
/* We have held the volume lock, it is safe to get the devices. */
|
2008-12-12 16:03:26 +01:00
|
|
|
list_for_each_entry(orig_dev, &orig->devices, dev_list) {
|
|
|
|
device = kzalloc(sizeof(*device), GFP_NOFS);
|
|
|
|
if (!device)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
device->name = kstrdup(orig_dev->name, GFP_NOFS);
|
2009-09-29 19:51:04 +02:00
|
|
|
if (!device->name) {
|
|
|
|
kfree(device);
|
2008-12-12 16:03:26 +01:00
|
|
|
goto error;
|
2009-09-29 19:51:04 +02:00
|
|
|
}
|
2008-12-12 16:03:26 +01:00
|
|
|
|
|
|
|
device->devid = orig_dev->devid;
|
|
|
|
device->work.func = pending_bios_fn;
|
|
|
|
memcpy(device->uuid, orig_dev->uuid, sizeof(device->uuid));
|
|
|
|
spin_lock_init(&device->io_lock);
|
|
|
|
INIT_LIST_HEAD(&device->dev_list);
|
|
|
|
INIT_LIST_HEAD(&device->dev_alloc_list);
|
|
|
|
|
|
|
|
list_add(&device->dev_list, &fs_devices->devices);
|
|
|
|
device->fs_devices = fs_devices;
|
|
|
|
fs_devices->num_devices++;
|
|
|
|
}
|
|
|
|
return fs_devices;
|
|
|
|
error:
|
|
|
|
free_fs_devices(fs_devices);
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
int btrfs_close_extra_devices(struct btrfs_fs_devices *fs_devices)
|
|
|
|
{
|
2009-01-21 16:59:08 +01:00
|
|
|
struct btrfs_device *device, *next;
|
2008-05-13 19:46:40 +02:00
|
|
|
|
|
|
|
mutex_lock(&uuid_mutex);
|
|
|
|
again:
|
2011-04-20 12:08:47 +02:00
|
|
|
/* This is the initialized path, it is safe to release the devices. */
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
|
2008-11-18 03:11:30 +01:00
|
|
|
if (device->in_fs_metadata)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
if (device->bdev) {
|
2010-11-13 11:55:18 +01:00
|
|
|
blkdev_put(device->bdev, device->mode);
|
2008-11-18 03:11:30 +01:00
|
|
|
device->bdev = NULL;
|
|
|
|
fs_devices->open_devices--;
|
|
|
|
}
|
|
|
|
if (device->writeable) {
|
|
|
|
list_del_init(&device->dev_alloc_list);
|
|
|
|
device->writeable = 0;
|
|
|
|
fs_devices->rw_devices--;
|
|
|
|
}
|
2008-12-12 16:03:26 +01:00
|
|
|
list_del_init(&device->dev_list);
|
|
|
|
fs_devices->num_devices--;
|
|
|
|
kfree(device->name);
|
|
|
|
kfree(device);
|
2008-05-13 19:46:40 +02:00
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
|
|
|
|
if (fs_devices->seed) {
|
|
|
|
fs_devices = fs_devices->seed;
|
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
return 0;
|
|
|
|
}
|
2008-05-13 22:03:06 +02:00
|
|
|
|
2011-04-20 12:09:16 +02:00
|
|
|
static void __free_device(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
|
|
|
|
device = container_of(work, struct btrfs_device, rcu_work);
|
|
|
|
|
|
|
|
if (device->bdev)
|
|
|
|
blkdev_put(device->bdev, device->mode);
|
|
|
|
|
|
|
|
kfree(device->name);
|
|
|
|
kfree(device);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void free_device(struct rcu_head *head)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
|
|
|
|
device = container_of(head, struct btrfs_device, rcu);
|
|
|
|
|
|
|
|
INIT_WORK(&device->rcu_work, __free_device);
|
|
|
|
schedule_work(&device->rcu_work);
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
static int __btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
|
2008-03-24 20:02:07 +01:00
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
2008-12-12 16:03:26 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (--fs_devices->opened > 0)
|
|
|
|
return 0;
|
2008-03-24 20:02:07 +01:00
|
|
|
|
2011-04-20 12:07:30 +02:00
|
|
|
mutex_lock(&fs_devices->device_list_mutex);
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(device, &fs_devices->devices, dev_list) {
|
2011-04-20 12:09:16 +02:00
|
|
|
struct btrfs_device *new_device;
|
|
|
|
|
|
|
|
if (device->bdev)
|
2008-05-13 22:03:06 +02:00
|
|
|
fs_devices->open_devices--;
|
2011-04-20 12:09:16 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (device->writeable) {
|
|
|
|
list_del_init(&device->dev_alloc_list);
|
|
|
|
fs_devices->rw_devices--;
|
|
|
|
}
|
|
|
|
|
2011-08-04 16:52:27 +02:00
|
|
|
if (device->can_discard)
|
|
|
|
fs_devices->num_can_discard--;
|
|
|
|
|
2011-04-20 12:09:16 +02:00
|
|
|
new_device = kmalloc(sizeof(*new_device), GFP_NOFS);
|
|
|
|
BUG_ON(!new_device);
|
|
|
|
memcpy(new_device, device, sizeof(*new_device));
|
|
|
|
new_device->name = kstrdup(device->name, GFP_NOFS);
|
2011-05-30 10:36:16 +02:00
|
|
|
BUG_ON(device->name && !new_device->name);
|
2011-04-20 12:09:16 +02:00
|
|
|
new_device->bdev = NULL;
|
|
|
|
new_device->writeable = 0;
|
|
|
|
new_device->in_fs_metadata = 0;
|
2011-08-04 16:52:27 +02:00
|
|
|
new_device->can_discard = 0;
|
2011-04-20 12:09:16 +02:00
|
|
|
list_replace_rcu(&device->dev_list, &new_device->dev_list);
|
|
|
|
|
|
|
|
call_rcu(&device->rcu, free_device);
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
2011-04-20 12:07:30 +02:00
|
|
|
mutex_unlock(&fs_devices->device_list_mutex);
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
WARN_ON(fs_devices->open_devices);
|
|
|
|
WARN_ON(fs_devices->rw_devices);
|
2008-11-18 03:11:30 +01:00
|
|
|
fs_devices->opened = 0;
|
|
|
|
fs_devices->seeding = 0;
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
int btrfs_close_devices(struct btrfs_fs_devices *fs_devices)
|
|
|
|
{
|
2008-12-12 16:03:26 +01:00
|
|
|
struct btrfs_fs_devices *seed_devices = NULL;
|
2008-11-18 03:11:30 +01:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
mutex_lock(&uuid_mutex);
|
|
|
|
ret = __btrfs_close_devices(fs_devices);
|
2008-12-12 16:03:26 +01:00
|
|
|
if (!fs_devices->opened) {
|
|
|
|
seed_devices = fs_devices->seed;
|
|
|
|
fs_devices->seed = NULL;
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
mutex_unlock(&uuid_mutex);
|
2008-12-12 16:03:26 +01:00
|
|
|
|
|
|
|
while (seed_devices) {
|
|
|
|
fs_devices = seed_devices;
|
|
|
|
seed_devices = fs_devices->seed;
|
|
|
|
__btrfs_close_devices(fs_devices);
|
|
|
|
free_fs_devices(fs_devices);
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
|
|
|
|
fmode_t flags, void *holder)
|
2008-03-24 20:02:07 +01:00
|
|
|
{
|
2011-08-04 16:52:27 +02:00
|
|
|
struct request_queue *q;
|
2008-03-24 20:02:07 +01:00
|
|
|
struct block_device *bdev;
|
|
|
|
struct list_head *head = &fs_devices->devices;
|
|
|
|
struct btrfs_device *device;
|
2008-05-13 22:03:06 +02:00
|
|
|
struct block_device *latest_bdev = NULL;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
struct btrfs_super_block *disk_super;
|
|
|
|
u64 latest_devid = 0;
|
|
|
|
u64 latest_transid = 0;
|
|
|
|
u64 devid;
|
2008-11-18 03:11:30 +01:00
|
|
|
int seeding = 1;
|
2008-05-13 22:03:06 +02:00
|
|
|
int ret = 0;
|
2008-03-24 20:02:07 +01:00
|
|
|
|
2010-11-13 11:55:18 +01:00
|
|
|
flags |= FMODE_EXCL;
|
|
|
|
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(device, head, dev_list) {
|
2008-05-08 21:05:58 +02:00
|
|
|
if (device->bdev)
|
|
|
|
continue;
|
2008-05-13 19:46:40 +02:00
|
|
|
if (!device->name)
|
|
|
|
continue;
|
|
|
|
|
2010-11-13 11:55:18 +01:00
|
|
|
bdev = blkdev_get_by_path(device->name, flags, holder);
|
2008-03-24 20:02:07 +01:00
|
|
|
if (IS_ERR(bdev)) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_INFO "open %s failed\n", device->name);
|
2008-05-13 22:03:06 +02:00
|
|
|
goto error;
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
2008-05-07 17:43:44 +02:00
|
|
|
set_blocksize(bdev, 4096);
|
2008-05-13 22:03:06 +02:00
|
|
|
|
2008-12-08 22:46:26 +01:00
|
|
|
bh = btrfs_read_dev_super(bdev);
|
2011-01-08 11:09:13 +01:00
|
|
|
if (!bh) {
|
|
|
|
ret = -EINVAL;
|
2008-05-13 22:03:06 +02:00
|
|
|
goto error_close;
|
2011-01-08 11:09:13 +01:00
|
|
|
}
|
2008-05-13 22:03:06 +02:00
|
|
|
|
|
|
|
disk_super = (struct btrfs_super_block *)bh->b_data;
|
2010-01-06 12:48:18 +01:00
|
|
|
devid = btrfs_stack_device_id(&disk_super->dev_item);
|
2008-05-13 22:03:06 +02:00
|
|
|
if (devid != device->devid)
|
|
|
|
goto error_brelse;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (memcmp(device->uuid, disk_super->dev_item.uuid,
|
|
|
|
BTRFS_UUID_SIZE))
|
|
|
|
goto error_brelse;
|
|
|
|
|
|
|
|
device->generation = btrfs_super_generation(disk_super);
|
|
|
|
if (!latest_transid || device->generation > latest_transid) {
|
2008-05-13 22:03:06 +02:00
|
|
|
latest_devid = devid;
|
2008-11-18 03:11:30 +01:00
|
|
|
latest_transid = device->generation;
|
2008-05-13 22:03:06 +02:00
|
|
|
latest_bdev = bdev;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_SEEDING) {
|
|
|
|
device->writeable = 0;
|
|
|
|
} else {
|
|
|
|
device->writeable = !bdev_read_only(bdev);
|
|
|
|
seeding = 0;
|
|
|
|
}
|
|
|
|
|
2011-08-04 16:52:27 +02:00
|
|
|
q = bdev_get_queue(bdev);
|
|
|
|
if (blk_queue_discard(q)) {
|
|
|
|
device->can_discard = 1;
|
|
|
|
fs_devices->num_can_discard++;
|
|
|
|
}
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
device->bdev = bdev;
|
2008-05-13 19:46:40 +02:00
|
|
|
device->in_fs_metadata = 0;
|
2008-11-20 03:17:22 +01:00
|
|
|
device->mode = flags;
|
|
|
|
|
2009-06-10 15:51:32 +02:00
|
|
|
if (!blk_queue_nonrot(bdev_get_queue(bdev)))
|
|
|
|
fs_devices->rotating = 1;
|
|
|
|
|
2008-05-13 22:03:06 +02:00
|
|
|
fs_devices->open_devices++;
|
2008-11-18 03:11:30 +01:00
|
|
|
if (device->writeable) {
|
|
|
|
fs_devices->rw_devices++;
|
|
|
|
list_add(&device->dev_alloc_list,
|
|
|
|
&fs_devices->alloc_list);
|
|
|
|
}
|
2011-04-20 12:06:40 +02:00
|
|
|
brelse(bh);
|
2008-05-13 22:03:06 +02:00
|
|
|
continue;
|
2008-05-07 17:43:44 +02:00
|
|
|
|
2008-05-13 22:03:06 +02:00
|
|
|
error_brelse:
|
|
|
|
brelse(bh);
|
|
|
|
error_close:
|
2010-11-13 11:55:18 +01:00
|
|
|
blkdev_put(bdev, flags);
|
2008-05-13 22:03:06 +02:00
|
|
|
error:
|
|
|
|
continue;
|
2008-03-24 20:02:07 +01:00
|
|
|
}
|
2008-05-13 22:03:06 +02:00
|
|
|
if (fs_devices->open_devices == 0) {
|
|
|
|
ret = -EIO;
|
|
|
|
goto out;
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
fs_devices->seeding = seeding;
|
|
|
|
fs_devices->opened = 1;
|
2008-05-13 22:03:06 +02:00
|
|
|
fs_devices->latest_bdev = latest_bdev;
|
|
|
|
fs_devices->latest_devid = latest_devid;
|
|
|
|
fs_devices->latest_trans = latest_transid;
|
2008-11-18 03:11:30 +01:00
|
|
|
fs_devices->total_rw_bytes = 0;
|
2008-05-13 22:03:06 +02:00
|
|
|
out:
|
2008-11-18 03:11:30 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
|
2008-12-02 12:36:09 +01:00
|
|
|
fmode_t flags, void *holder)
|
2008-11-18 03:11:30 +01:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
mutex_lock(&uuid_mutex);
|
|
|
|
if (fs_devices->opened) {
|
2008-12-12 16:03:26 +01:00
|
|
|
fs_devices->opened++;
|
|
|
|
ret = 0;
|
2008-11-18 03:11:30 +01:00
|
|
|
} else {
|
2008-11-20 03:17:22 +01:00
|
|
|
ret = __btrfs_open_devices(fs_devices, flags, holder);
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
2008-03-24 20:02:07 +01:00
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-02 12:36:09 +01:00
|
|
|
int btrfs_scan_one_device(const char *path, fmode_t flags, void *holder,
|
2008-03-24 20:02:07 +01:00
|
|
|
struct btrfs_fs_devices **fs_devices_ret)
|
|
|
|
{
|
|
|
|
struct btrfs_super_block *disk_super;
|
|
|
|
struct block_device *bdev;
|
|
|
|
struct buffer_head *bh;
|
|
|
|
int ret;
|
|
|
|
u64 devid;
|
2008-04-10 22:19:33 +02:00
|
|
|
u64 transid;
|
2008-03-24 20:02:07 +01:00
|
|
|
|
|
|
|
mutex_lock(&uuid_mutex);
|
|
|
|
|
2010-11-13 11:55:18 +01:00
|
|
|
flags |= FMODE_EXCL;
|
|
|
|
bdev = blkdev_get_by_path(path, flags, holder);
|
2008-03-24 20:02:07 +01:00
|
|
|
|
|
|
|
if (IS_ERR(bdev)) {
|
|
|
|
ret = PTR_ERR(bdev);
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = set_blocksize(bdev, 4096);
|
|
|
|
if (ret)
|
|
|
|
goto error_close;
|
2008-12-08 22:46:26 +01:00
|
|
|
bh = btrfs_read_dev_super(bdev);
|
2008-03-24 20:02:07 +01:00
|
|
|
if (!bh) {
|
2011-01-08 11:09:13 +01:00
|
|
|
ret = -EINVAL;
|
2008-03-24 20:02:07 +01:00
|
|
|
goto error_close;
|
|
|
|
}
|
|
|
|
disk_super = (struct btrfs_super_block *)bh->b_data;
|
2010-01-06 12:48:18 +01:00
|
|
|
devid = btrfs_stack_device_id(&disk_super->dev_item);
|
2008-04-10 22:19:33 +02:00
|
|
|
transid = btrfs_super_generation(disk_super);
|
2008-04-18 16:29:49 +02:00
|
|
|
if (disk_super->label[0])
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_INFO "device label %s ", disk_super->label);
|
2011-02-09 15:05:31 +01:00
|
|
|
else
|
|
|
|
printk(KERN_INFO "device fsid %pU ", disk_super->fsid);
|
2009-01-21 16:49:16 +01:00
|
|
|
printk(KERN_CONT "devid %llu transid %llu %s\n",
|
2009-01-06 03:25:51 +01:00
|
|
|
(unsigned long long)devid, (unsigned long long)transid, path);
|
2008-03-24 20:02:07 +01:00
|
|
|
ret = device_list_add(path, disk_super, devid, fs_devices_ret);
|
|
|
|
|
|
|
|
brelse(bh);
|
|
|
|
error_close:
|
2010-11-13 11:55:18 +01:00
|
|
|
blkdev_put(bdev, flags);
|
2008-03-24 20:02:07 +01:00
|
|
|
error:
|
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
|
btrfs: fix wrong free space information of btrfs
When we store data by raid profile in btrfs with two or more different size
disks, df command shows there is some free space in the filesystem, but the
user can not write any data in fact, df command shows the wrong free space
information of btrfs.
# mkfs.btrfs -d raid1 /dev/sda9 /dev/sda10
# btrfs-show
Label: none uuid: a95cd49e-6e33-45b8-8741-a36153ce4b64
Total devices 2 FS bytes used 28.00KB
devid 1 size 5.01GB used 2.03GB path /dev/sda9
devid 2 size 10.00GB used 2.01GB path /dev/sda10
# btrfs device scan /dev/sda9 /dev/sda10
# mount /dev/sda9 /mnt
# dd if=/dev/zero of=tmpfile0 bs=4K count=9999999999
(fill the filesystem)
# sync
# df -TH
Filesystem Type Size Used Avail Use% Mounted on
/dev/sda9 btrfs 17G 8.6G 5.4G 62% /mnt
# btrfs-show
Label: none uuid: a95cd49e-6e33-45b8-8741-a36153ce4b64
Total devices 2 FS bytes used 3.99GB
devid 1 size 5.01GB used 5.01GB path /dev/sda9
devid 2 size 10.00GB used 4.99GB path /dev/sda10
It is because btrfs cannot allocate chunks when one of the pairing disks has
no space, the free space on the other disks can not be used for ever, and should
be subtracted from the total space, but btrfs doesn't subtract this space from
the total. It is strange to the user.
This patch fixes it by calcing the free space that can be used to allocate
chunks.
Implementation:
1. get all the devices free space, and align them by stripe length.
2. sort the devices by the free space.
3. check the free space of the devices,
3.1. if it is not zero, and then check the number of the devices that has
more free space than this device,
if the number of the devices is beyond the min stripe number, the free
space can be used, and add into total free space.
if the number of the devices is below the min stripe number, we can not
use the free space, the check ends.
3.2. if the free space is zero, check the next devices, goto 3.1
This implementation is just likely fake chunk allocation.
After appling this patch, df can show correct space information:
# df -TH
Filesystem Type Size Used Avail Use% Mounted on
/dev/sda9 btrfs 17G 8.6G 0 100% /mnt
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-01-05 11:07:31 +01:00
|
|
|
/* helper to account the used device space in the range */
|
|
|
|
int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
|
|
|
|
u64 end, u64 *length)
|
|
|
|
{
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_root *root = device->dev_root;
|
|
|
|
struct btrfs_dev_extent *dev_extent;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
u64 extent_end;
|
|
|
|
int ret;
|
|
|
|
int slot;
|
|
|
|
struct extent_buffer *l;
|
|
|
|
|
|
|
|
*length = 0;
|
|
|
|
|
|
|
|
if (start >= device->total_bytes)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
path->reada = 2;
|
|
|
|
|
|
|
|
key.objectid = device->devid;
|
|
|
|
key.offset = start;
|
|
|
|
key.type = BTRFS_DEV_EXTENT_KEY;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
if (ret > 0) {
|
|
|
|
ret = btrfs_previous_item(root, path, key.objectid, key.type);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
l = path->nodes[0];
|
|
|
|
slot = path->slots[0];
|
|
|
|
if (slot >= btrfs_header_nritems(l)) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret == 0)
|
|
|
|
continue;
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
btrfs_item_key_to_cpu(l, &key, slot);
|
|
|
|
|
|
|
|
if (key.objectid < device->devid)
|
|
|
|
goto next;
|
|
|
|
|
|
|
|
if (key.objectid > device->devid)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY)
|
|
|
|
goto next;
|
|
|
|
|
|
|
|
dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
|
|
|
|
extent_end = key.offset + btrfs_dev_extent_length(l,
|
|
|
|
dev_extent);
|
|
|
|
if (key.offset <= start && extent_end > end) {
|
|
|
|
*length = end - start + 1;
|
|
|
|
break;
|
|
|
|
} else if (key.offset <= start && extent_end > start)
|
|
|
|
*length += extent_end - start;
|
|
|
|
else if (key.offset > start && extent_end <= end)
|
|
|
|
*length += extent_end - key.offset;
|
|
|
|
else if (key.offset > start && key.offset <= end) {
|
|
|
|
*length += end - key.offset + 1;
|
|
|
|
break;
|
|
|
|
} else if (key.offset > end)
|
|
|
|
break;
|
|
|
|
|
|
|
|
next:
|
|
|
|
path->slots[0]++;
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
/*
|
2011-01-05 11:07:26 +01:00
|
|
|
* find_free_dev_extent - find free space in the specified device
|
|
|
|
* @trans: transaction handler
|
|
|
|
* @device: the device which we search the free space in
|
|
|
|
* @num_bytes: the size of the free space that we need
|
|
|
|
* @start: store the start of the free space.
|
|
|
|
* @len: the size of the free space. that we find, or the size of the max
|
|
|
|
* free space if we don't find suitable free space
|
|
|
|
*
|
2008-03-24 20:01:56 +01:00
|
|
|
* this uses a pretty simple search, the expectation is that it is
|
|
|
|
* called very infrequently and that a given device has a small number
|
|
|
|
* of extents
|
2011-01-05 11:07:26 +01:00
|
|
|
*
|
|
|
|
* @start is used to store the start of the free space if we find. But if we
|
|
|
|
* don't find suitable free space, it will be used to store the start position
|
|
|
|
* of the max free space.
|
|
|
|
*
|
|
|
|
* @len is used to store the size of the free space that we find.
|
|
|
|
* But if we don't find suitable free space, it is used to store the size of
|
|
|
|
* the max free space.
|
2008-03-24 20:01:56 +01:00
|
|
|
*/
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
int find_free_dev_extent(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_device *device, u64 num_bytes,
|
2011-01-05 11:07:26 +01:00
|
|
|
u64 *start, u64 *len)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_root *root = device->dev_root;
|
2011-01-05 11:07:26 +01:00
|
|
|
struct btrfs_dev_extent *dev_extent;
|
2008-11-18 03:11:30 +01:00
|
|
|
struct btrfs_path *path;
|
2011-01-05 11:07:26 +01:00
|
|
|
u64 hole_size;
|
|
|
|
u64 max_hole_start;
|
|
|
|
u64 max_hole_size;
|
|
|
|
u64 extent_end;
|
|
|
|
u64 search_start;
|
2008-03-24 20:01:56 +01:00
|
|
|
u64 search_end = device->total_bytes;
|
|
|
|
int ret;
|
2011-01-05 11:07:26 +01:00
|
|
|
int slot;
|
2008-03-24 20:01:56 +01:00
|
|
|
struct extent_buffer *l;
|
|
|
|
|
|
|
|
/* FIXME use last free of some kind */
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
/* we don't want to overwrite the superblock on the drive,
|
|
|
|
* so we make sure to start at an offset of at least 1MB
|
|
|
|
*/
|
2011-04-12 11:01:20 +02:00
|
|
|
search_start = max(root->fs_info->alloc_start, 1024ull * 1024);
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
max_hole_start = search_start;
|
|
|
|
max_hole_size = 0;
|
Btrfs: fix a bug of balance on full multi-disk partitions
When balancing, we'll first try to shrink devices for some space,
but if it is working on a full multi-disk partition with raid protection,
we may encounter a bug, that is, while shrinking, total_bytes may be less
than bytes_used, and btrfs may allocate a dev extent that accesses out of
device's bounds.
Then we will not be able to write or read the data which stores at the end
of the device, and get the followings:
device fsid 0939f071-7ea3-46c8-95df-f176d773bfb6 devid 1 transid 10 /dev/sdb5
Btrfs detected SSD devices, enabling SSD mode
btrfs: relocating block group 476315648 flags 9
btrfs: found 4 extents
attempt to access beyond end of device
sdb5: rw=145, want=546176, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546304, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546432, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546560, limit=546147
attempt to access beyond end of device
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-08-02 04:39:03 +02:00
|
|
|
hole_size = 0;
|
2011-01-05 11:07:26 +01:00
|
|
|
|
|
|
|
if (search_start >= search_end) {
|
|
|
|
ret = -ENOSPC;
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
path->reada = 2;
|
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
key.objectid = device->devid;
|
|
|
|
key.offset = search_start;
|
|
|
|
key.type = BTRFS_DEV_EXTENT_KEY;
|
2011-01-05 11:07:26 +01:00
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
2011-01-05 11:07:26 +01:00
|
|
|
goto out;
|
2009-07-24 17:06:53 +02:00
|
|
|
if (ret > 0) {
|
|
|
|
ret = btrfs_previous_item(root, path, key.objectid, key.type);
|
|
|
|
if (ret < 0)
|
2011-01-05 11:07:26 +01:00
|
|
|
goto out;
|
2009-07-24 17:06:53 +02:00
|
|
|
}
|
2011-01-05 11:07:26 +01:00
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
while (1) {
|
|
|
|
l = path->nodes[0];
|
|
|
|
slot = path->slots[0];
|
|
|
|
if (slot >= btrfs_header_nritems(l)) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret == 0)
|
|
|
|
continue;
|
|
|
|
if (ret < 0)
|
2011-01-05 11:07:26 +01:00
|
|
|
goto out;
|
|
|
|
|
|
|
|
break;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
btrfs_item_key_to_cpu(l, &key, slot);
|
|
|
|
|
|
|
|
if (key.objectid < device->devid)
|
|
|
|
goto next;
|
|
|
|
|
|
|
|
if (key.objectid > device->devid)
|
2011-01-05 11:07:26 +01:00
|
|
|
break;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
if (btrfs_key_type(&key) != BTRFS_DEV_EXTENT_KEY)
|
|
|
|
goto next;
|
2009-07-24 22:41:41 +02:00
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
if (key.offset > search_start) {
|
|
|
|
hole_size = key.offset - search_start;
|
2009-07-24 22:41:41 +02:00
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
if (hole_size > max_hole_size) {
|
|
|
|
max_hole_start = search_start;
|
|
|
|
max_hole_size = hole_size;
|
|
|
|
}
|
2009-07-24 22:41:41 +02:00
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
/*
|
|
|
|
* If this free space is greater than which we need,
|
|
|
|
* it must be the max free space that we have found
|
|
|
|
* until now, so max_hole_start must point to the start
|
|
|
|
* of this free space and the length of this free space
|
|
|
|
* is stored in max_hole_size. Thus, we return
|
|
|
|
* max_hole_start and max_hole_size and go back to the
|
|
|
|
* caller.
|
|
|
|
*/
|
|
|
|
if (hole_size >= num_bytes) {
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
|
2011-01-05 11:07:26 +01:00
|
|
|
extent_end = key.offset + btrfs_dev_extent_length(l,
|
|
|
|
dev_extent);
|
|
|
|
if (extent_end > search_start)
|
|
|
|
search_start = extent_end;
|
2008-03-24 20:01:56 +01:00
|
|
|
next:
|
|
|
|
path->slots[0]++;
|
|
|
|
cond_resched();
|
|
|
|
}
|
|
|
|
|
Btrfs: fix a bug of balance on full multi-disk partitions
When balancing, we'll first try to shrink devices for some space,
but if it is working on a full multi-disk partition with raid protection,
we may encounter a bug, that is, while shrinking, total_bytes may be less
than bytes_used, and btrfs may allocate a dev extent that accesses out of
device's bounds.
Then we will not be able to write or read the data which stores at the end
of the device, and get the followings:
device fsid 0939f071-7ea3-46c8-95df-f176d773bfb6 devid 1 transid 10 /dev/sdb5
Btrfs detected SSD devices, enabling SSD mode
btrfs: relocating block group 476315648 flags 9
btrfs: found 4 extents
attempt to access beyond end of device
sdb5: rw=145, want=546176, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546304, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546432, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546560, limit=546147
attempt to access beyond end of device
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-08-02 04:39:03 +02:00
|
|
|
/*
|
|
|
|
* At this point, search_start should be the end of
|
|
|
|
* allocated dev extents, and when shrinking the device,
|
|
|
|
* search_end may be smaller than search_start.
|
|
|
|
*/
|
|
|
|
if (search_end > search_start)
|
|
|
|
hole_size = search_end - search_start;
|
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
if (hole_size > max_hole_size) {
|
|
|
|
max_hole_start = search_start;
|
|
|
|
max_hole_size = hole_size;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
|
2011-01-05 11:07:26 +01:00
|
|
|
/* See above. */
|
|
|
|
if (hole_size < num_bytes)
|
|
|
|
ret = -ENOSPC;
|
|
|
|
else
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
out:
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_free_path(path);
|
2011-01-05 11:07:26 +01:00
|
|
|
error:
|
|
|
|
*start = max_hole_start;
|
2011-01-05 11:07:28 +01:00
|
|
|
if (len)
|
2011-01-05 11:07:26 +01:00
|
|
|
*len = max_hole_size;
|
2008-03-24 20:01:56 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-02 15:54:17 +01:00
|
|
|
static int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
|
2008-04-25 22:53:30 +02:00
|
|
|
struct btrfs_device *device,
|
|
|
|
u64 start)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_root *root = device->dev_root;
|
|
|
|
struct btrfs_key key;
|
2008-05-07 17:43:44 +02:00
|
|
|
struct btrfs_key found_key;
|
|
|
|
struct extent_buffer *leaf = NULL;
|
|
|
|
struct btrfs_dev_extent *extent = NULL;
|
2008-04-25 22:53:30 +02:00
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
key.objectid = device->devid;
|
|
|
|
key.offset = start;
|
|
|
|
key.type = BTRFS_DEV_EXTENT_KEY;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
|
2008-05-07 17:43:44 +02:00
|
|
|
if (ret > 0) {
|
|
|
|
ret = btrfs_previous_item(root, path, key.objectid,
|
|
|
|
BTRFS_DEV_EXTENT_KEY);
|
2011-05-19 09:03:42 +02:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
2008-05-07 17:43:44 +02:00
|
|
|
leaf = path->nodes[0];
|
|
|
|
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
|
|
|
|
extent = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_dev_extent);
|
|
|
|
BUG_ON(found_key.offset > start || found_key.offset +
|
|
|
|
btrfs_dev_extent_length(leaf, extent) < start);
|
|
|
|
} else if (ret == 0) {
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
extent = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_dev_extent);
|
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
BUG_ON(ret);
|
|
|
|
|
2011-09-26 23:12:22 +02:00
|
|
|
if (device->bytes_used > 0) {
|
|
|
|
u64 len = btrfs_dev_extent_length(leaf, extent);
|
|
|
|
device->bytes_used -= len;
|
|
|
|
spin_lock(&root->fs_info->free_chunk_lock);
|
|
|
|
root->fs_info->free_chunk_space += len;
|
|
|
|
spin_unlock(&root->fs_info->free_chunk_lock);
|
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
ret = btrfs_del_item(trans, root, path);
|
|
|
|
|
2011-05-19 09:03:42 +02:00
|
|
|
out:
|
2008-04-25 22:53:30 +02:00
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
|
2008-03-24 20:01:56 +01:00
|
|
|
struct btrfs_device *device,
|
2008-04-15 21:41:47 +02:00
|
|
|
u64 chunk_tree, u64 chunk_objectid,
|
2008-11-18 03:11:30 +01:00
|
|
|
u64 chunk_offset, u64 start, u64 num_bytes)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_root *root = device->dev_root;
|
|
|
|
struct btrfs_dev_extent *extent;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_key key;
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
WARN_ON(!device->in_fs_metadata);
|
2008-03-24 20:01:56 +01:00
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
key.objectid = device->devid;
|
2008-11-18 03:11:30 +01:00
|
|
|
key.offset = start;
|
2008-03-24 20:01:56 +01:00
|
|
|
key.type = BTRFS_DEV_EXTENT_KEY;
|
|
|
|
ret = btrfs_insert_empty_item(trans, root, path, &key,
|
|
|
|
sizeof(*extent));
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
extent = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_dev_extent);
|
2008-04-15 21:41:47 +02:00
|
|
|
btrfs_set_dev_extent_chunk_tree(leaf, extent, chunk_tree);
|
|
|
|
btrfs_set_dev_extent_chunk_objectid(leaf, extent, chunk_objectid);
|
|
|
|
btrfs_set_dev_extent_chunk_offset(leaf, extent, chunk_offset);
|
|
|
|
|
|
|
|
write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
|
|
|
|
(unsigned long)btrfs_dev_extent_chunk_tree_uuid(extent),
|
|
|
|
BTRFS_UUID_SIZE);
|
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
btrfs_set_dev_extent_length(leaf, extent, num_bytes);
|
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-09-05 22:09:51 +02:00
|
|
|
static noinline int find_next_chunk(struct btrfs_root *root,
|
|
|
|
u64 objectid, u64 *offset)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
struct btrfs_path *path;
|
|
|
|
int ret;
|
|
|
|
struct btrfs_key key;
|
2008-04-15 21:41:47 +02:00
|
|
|
struct btrfs_chunk *chunk;
|
2008-03-24 20:01:56 +01:00
|
|
|
struct btrfs_key found_key;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
2011-07-12 19:57:59 +02:00
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-04-15 21:41:47 +02:00
|
|
|
key.objectid = objectid;
|
2008-03-24 20:01:56 +01:00
|
|
|
key.offset = (u64)-1;
|
|
|
|
key.type = BTRFS_CHUNK_ITEM_KEY;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
BUG_ON(ret == 0);
|
|
|
|
|
|
|
|
ret = btrfs_previous_item(root, path, 0, BTRFS_CHUNK_ITEM_KEY);
|
|
|
|
if (ret) {
|
2008-04-15 21:41:47 +02:00
|
|
|
*offset = 0;
|
2008-03-24 20:01:56 +01:00
|
|
|
} else {
|
|
|
|
btrfs_item_key_to_cpu(path->nodes[0], &found_key,
|
|
|
|
path->slots[0]);
|
2008-04-15 21:41:47 +02:00
|
|
|
if (found_key.objectid != objectid)
|
|
|
|
*offset = 0;
|
|
|
|
else {
|
|
|
|
chunk = btrfs_item_ptr(path->nodes[0], path->slots[0],
|
|
|
|
struct btrfs_chunk);
|
|
|
|
*offset = found_key.offset +
|
|
|
|
btrfs_chunk_length(path->nodes[0], chunk);
|
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
error:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
static noinline int find_next_devid(struct btrfs_root *root, u64 *objectid)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_key found_key;
|
2008-11-18 03:11:30 +01:00
|
|
|
struct btrfs_path *path;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
|
|
|
|
key.type = BTRFS_DEV_ITEM_KEY;
|
|
|
|
key.offset = (u64)-1;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
BUG_ON(ret == 0);
|
|
|
|
|
|
|
|
ret = btrfs_previous_item(root, path, BTRFS_DEV_ITEMS_OBJECTID,
|
|
|
|
BTRFS_DEV_ITEM_KEY);
|
|
|
|
if (ret) {
|
|
|
|
*objectid = 1;
|
|
|
|
} else {
|
|
|
|
btrfs_item_key_to_cpu(path->nodes[0], &found_key,
|
|
|
|
path->slots[0]);
|
|
|
|
*objectid = found_key.offset + 1;
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
error:
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_free_path(path);
|
2008-03-24 20:01:56 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* the device information is stored in the chunk root
|
|
|
|
* the btrfs_device struct should be fully filled in
|
|
|
|
*/
|
|
|
|
int btrfs_add_device(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_device *device)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_dev_item *dev_item;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_key key;
|
|
|
|
unsigned long ptr;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
|
|
|
|
key.type = BTRFS_DEV_ITEM_KEY;
|
2008-11-18 03:11:30 +01:00
|
|
|
key.offset = device->devid;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
ret = btrfs_insert_empty_item(trans, root, path, &key,
|
2008-03-24 20:02:07 +01:00
|
|
|
sizeof(*dev_item));
|
2008-03-24 20:01:56 +01:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
|
|
|
|
|
|
|
|
btrfs_set_device_id(leaf, dev_item, device->devid);
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_set_device_generation(leaf, dev_item, 0);
|
2008-03-24 20:01:56 +01:00
|
|
|
btrfs_set_device_type(leaf, dev_item, device->type);
|
|
|
|
btrfs_set_device_io_align(leaf, dev_item, device->io_align);
|
|
|
|
btrfs_set_device_io_width(leaf, dev_item, device->io_width);
|
|
|
|
btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
|
|
|
|
btrfs_set_device_total_bytes(leaf, dev_item, device->total_bytes);
|
|
|
|
btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
|
2008-04-15 21:41:47 +02:00
|
|
|
btrfs_set_device_group(leaf, dev_item, 0);
|
|
|
|
btrfs_set_device_seek_speed(leaf, dev_item, 0);
|
|
|
|
btrfs_set_device_bandwidth(leaf, dev_item, 0);
|
2008-12-08 22:40:21 +01:00
|
|
|
btrfs_set_device_start_offset(leaf, dev_item, 0);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
ptr = (unsigned long)btrfs_device_uuid(dev_item);
|
2008-04-15 21:41:47 +02:00
|
|
|
write_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
|
2008-11-18 03:11:30 +01:00
|
|
|
ptr = (unsigned long)btrfs_device_fsid(dev_item);
|
|
|
|
write_extent_buffer(leaf, root->fs_info->fsid, ptr, BTRFS_UUID_SIZE);
|
2008-03-24 20:01:56 +01:00
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = 0;
|
2008-03-24 20:01:56 +01:00
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2008-05-07 17:43:44 +02:00
|
|
|
static int btrfs_rm_dev_item(struct btrfs_root *root,
|
|
|
|
struct btrfs_device *device)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2010-05-16 16:48:46 +02:00
|
|
|
trans = btrfs_start_transaction(root, 0);
|
2011-01-20 07:19:37 +01:00
|
|
|
if (IS_ERR(trans)) {
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return PTR_ERR(trans);
|
|
|
|
}
|
2008-05-07 17:43:44 +02:00
|
|
|
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
|
|
|
|
key.type = BTRFS_DEV_ITEM_KEY;
|
|
|
|
key.offset = device->devid;
|
2008-07-08 20:19:17 +02:00
|
|
|
lock_chunks(root);
|
2008-05-07 17:43:44 +02:00
|
|
|
|
|
|
|
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (ret > 0) {
|
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
ret = btrfs_del_item(trans, root, path);
|
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
2008-07-08 20:19:17 +02:00
|
|
|
unlock_chunks(root);
|
2008-05-07 17:43:44 +02:00
|
|
|
btrfs_commit_transaction(trans, root);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_rm_device(struct btrfs_root *root, char *device_path)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
2008-11-18 03:11:30 +01:00
|
|
|
struct btrfs_device *next_device;
|
2008-05-07 17:43:44 +02:00
|
|
|
struct block_device *bdev;
|
2008-05-13 19:46:40 +02:00
|
|
|
struct buffer_head *bh = NULL;
|
2008-05-07 17:43:44 +02:00
|
|
|
struct btrfs_super_block *disk_super;
|
2011-04-20 12:09:16 +02:00
|
|
|
struct btrfs_fs_devices *cur_devices;
|
2008-05-07 17:43:44 +02:00
|
|
|
u64 all_avail;
|
|
|
|
u64 devid;
|
2008-11-18 03:11:30 +01:00
|
|
|
u64 num_devices;
|
|
|
|
u8 *dev_uuid;
|
2008-05-07 17:43:44 +02:00
|
|
|
int ret = 0;
|
2011-04-20 12:09:16 +02:00
|
|
|
bool clear_super = false;
|
2008-05-07 17:43:44 +02:00
|
|
|
|
|
|
|
mutex_lock(&uuid_mutex);
|
2008-07-08 20:19:17 +02:00
|
|
|
mutex_lock(&root->fs_info->volume_mutex);
|
2008-05-07 17:43:44 +02:00
|
|
|
|
|
|
|
all_avail = root->fs_info->avail_data_alloc_bits |
|
|
|
|
root->fs_info->avail_system_alloc_bits |
|
|
|
|
root->fs_info->avail_metadata_alloc_bits;
|
|
|
|
|
|
|
|
if ((all_avail & BTRFS_BLOCK_GROUP_RAID10) &&
|
2010-01-27 03:09:38 +01:00
|
|
|
root->fs_info->fs_devices->num_devices <= 4) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_ERR "btrfs: unable to go below four devices "
|
|
|
|
"on raid10\n");
|
2008-05-07 17:43:44 +02:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if ((all_avail & BTRFS_BLOCK_GROUP_RAID1) &&
|
2010-01-27 03:09:38 +01:00
|
|
|
root->fs_info->fs_devices->num_devices <= 2) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_ERR "btrfs: unable to go below two "
|
|
|
|
"devices on raid1\n");
|
2008-05-07 17:43:44 +02:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
if (strcmp(device_path, "missing") == 0) {
|
|
|
|
struct list_head *devices;
|
|
|
|
struct btrfs_device *tmp;
|
2008-05-07 17:43:44 +02:00
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
device = NULL;
|
|
|
|
devices = &root->fs_info->fs_devices->devices;
|
2011-04-20 12:08:47 +02:00
|
|
|
/*
|
|
|
|
* It is safe to read the devices since the volume_mutex
|
|
|
|
* is held.
|
|
|
|
*/
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(tmp, devices, dev_list) {
|
2008-05-13 19:46:40 +02:00
|
|
|
if (tmp->in_fs_metadata && !tmp->bdev) {
|
|
|
|
device = tmp;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
bdev = NULL;
|
|
|
|
bh = NULL;
|
|
|
|
disk_super = NULL;
|
|
|
|
if (!device) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_ERR "btrfs: no missing devices found to "
|
|
|
|
"remove\n");
|
2008-05-13 19:46:40 +02:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
} else {
|
2010-11-13 11:55:18 +01:00
|
|
|
bdev = blkdev_get_by_path(device_path, FMODE_READ | FMODE_EXCL,
|
|
|
|
root->fs_info->bdev_holder);
|
2008-05-13 19:46:40 +02:00
|
|
|
if (IS_ERR(bdev)) {
|
|
|
|
ret = PTR_ERR(bdev);
|
|
|
|
goto out;
|
|
|
|
}
|
2008-05-07 17:43:44 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
set_blocksize(bdev, 4096);
|
2008-12-08 22:46:26 +01:00
|
|
|
bh = btrfs_read_dev_super(bdev);
|
2008-05-13 19:46:40 +02:00
|
|
|
if (!bh) {
|
2011-01-08 11:09:13 +01:00
|
|
|
ret = -EINVAL;
|
2008-05-13 19:46:40 +02:00
|
|
|
goto error_close;
|
|
|
|
}
|
|
|
|
disk_super = (struct btrfs_super_block *)bh->b_data;
|
2010-01-06 12:48:18 +01:00
|
|
|
devid = btrfs_stack_device_id(&disk_super->dev_item);
|
2008-11-18 03:11:30 +01:00
|
|
|
dev_uuid = disk_super->dev_item.uuid;
|
|
|
|
device = btrfs_find_device(root, devid, dev_uuid,
|
|
|
|
disk_super->fsid);
|
2008-05-13 19:46:40 +02:00
|
|
|
if (!device) {
|
|
|
|
ret = -ENOENT;
|
|
|
|
goto error_brelse;
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
2008-05-13 19:46:40 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (device->writeable && root->fs_info->fs_devices->rw_devices == 1) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_ERR "btrfs: unable to remove the only writeable "
|
|
|
|
"device\n");
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto error_brelse;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (device->writeable) {
|
2011-04-20 12:08:16 +02:00
|
|
|
lock_chunks(root);
|
2008-11-18 03:11:30 +01:00
|
|
|
list_del_init(&device->dev_alloc_list);
|
2011-04-20 12:08:16 +02:00
|
|
|
unlock_chunks(root);
|
2008-11-18 03:11:30 +01:00
|
|
|
root->fs_info->fs_devices->rw_devices--;
|
2011-04-20 12:09:16 +02:00
|
|
|
clear_super = true;
|
2008-05-13 19:46:40 +02:00
|
|
|
}
|
2008-05-07 17:43:44 +02:00
|
|
|
|
|
|
|
ret = btrfs_shrink_device(device, 0);
|
|
|
|
if (ret)
|
2011-02-15 19:14:25 +01:00
|
|
|
goto error_undo;
|
2008-05-07 17:43:44 +02:00
|
|
|
|
|
|
|
ret = btrfs_rm_dev_item(root->fs_info->chunk_root, device);
|
|
|
|
if (ret)
|
2011-02-15 19:14:25 +01:00
|
|
|
goto error_undo;
|
2008-05-07 17:43:44 +02:00
|
|
|
|
2011-09-26 23:12:22 +02:00
|
|
|
spin_lock(&root->fs_info->free_chunk_lock);
|
|
|
|
root->fs_info->free_chunk_space = device->total_bytes -
|
|
|
|
device->bytes_used;
|
|
|
|
spin_unlock(&root->fs_info->free_chunk_lock);
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
device->in_fs_metadata = 0;
|
2011-03-08 14:14:00 +01:00
|
|
|
btrfs_scrub_cancel_dev(root, device);
|
2009-06-10 21:17:02 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* the device list mutex makes sure that we don't change
|
|
|
|
* the device list while someone else is writing out all
|
|
|
|
* the device supers.
|
|
|
|
*/
|
2011-04-20 12:09:16 +02:00
|
|
|
|
|
|
|
cur_devices = device->fs_devices;
|
2009-06-10 21:17:02 +02:00
|
|
|
mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
|
2011-04-20 12:09:16 +02:00
|
|
|
list_del_rcu(&device->dev_list);
|
2009-06-10 21:17:02 +02:00
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
device->fs_devices->num_devices--;
|
2008-11-18 03:11:30 +01:00
|
|
|
|
2010-12-13 20:56:23 +01:00
|
|
|
if (device->missing)
|
|
|
|
root->fs_info->fs_devices->missing_devices--;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
next_device = list_entry(root->fs_info->fs_devices->devices.next,
|
|
|
|
struct btrfs_device, dev_list);
|
|
|
|
if (device->bdev == root->fs_info->sb->s_bdev)
|
|
|
|
root->fs_info->sb->s_bdev = next_device->bdev;
|
|
|
|
if (device->bdev == root->fs_info->fs_devices->latest_bdev)
|
|
|
|
root->fs_info->fs_devices->latest_bdev = next_device->bdev;
|
|
|
|
|
2011-04-20 12:09:16 +02:00
|
|
|
if (device->bdev)
|
2008-12-12 16:03:26 +01:00
|
|
|
device->fs_devices->open_devices--;
|
2011-04-20 12:09:16 +02:00
|
|
|
|
|
|
|
call_rcu(&device->rcu, free_device);
|
|
|
|
mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
|
2008-12-12 16:03:26 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
num_devices = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
|
|
|
|
btrfs_set_super_num_devices(&root->fs_info->super_copy, num_devices);
|
|
|
|
|
2011-04-20 12:09:16 +02:00
|
|
|
if (cur_devices->open_devices == 0) {
|
2008-12-12 16:03:26 +01:00
|
|
|
struct btrfs_fs_devices *fs_devices;
|
|
|
|
fs_devices = root->fs_info->fs_devices;
|
|
|
|
while (fs_devices) {
|
2011-04-20 12:09:16 +02:00
|
|
|
if (fs_devices->seed == cur_devices)
|
2008-12-12 16:03:26 +01:00
|
|
|
break;
|
|
|
|
fs_devices = fs_devices->seed;
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
2011-04-20 12:09:16 +02:00
|
|
|
fs_devices->seed = cur_devices->seed;
|
|
|
|
cur_devices->seed = NULL;
|
2011-04-20 12:08:16 +02:00
|
|
|
lock_chunks(root);
|
2011-04-20 12:09:16 +02:00
|
|
|
__btrfs_close_devices(cur_devices);
|
2011-04-20 12:08:16 +02:00
|
|
|
unlock_chunks(root);
|
2011-04-20 12:09:16 +02:00
|
|
|
free_fs_devices(cur_devices);
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* at this point, the device is zero sized. We want to
|
|
|
|
* remove it from the devices list and zero out the old super
|
|
|
|
*/
|
2011-04-20 12:09:16 +02:00
|
|
|
if (clear_super) {
|
2008-05-13 19:46:40 +02:00
|
|
|
/* make sure this device isn't detected as part of
|
|
|
|
* the FS anymore
|
|
|
|
*/
|
|
|
|
memset(&disk_super->magic, 0, sizeof(disk_super->magic));
|
|
|
|
set_buffer_dirty(bh);
|
|
|
|
sync_dirty_buffer(bh);
|
|
|
|
}
|
2008-05-07 17:43:44 +02:00
|
|
|
|
|
|
|
ret = 0;
|
|
|
|
|
|
|
|
error_brelse:
|
|
|
|
brelse(bh);
|
|
|
|
error_close:
|
2008-05-13 19:46:40 +02:00
|
|
|
if (bdev)
|
block: make blkdev_get/put() handle exclusive access
Over time, block layer has accumulated a set of APIs dealing with bdev
open, close, claim and release.
* blkdev_get/put() are the primary open and close functions.
* bd_claim/release() deal with exclusive open.
* open/close_bdev_exclusive() are combination of open and claim and
the other way around, respectively.
* bd_link/unlink_disk_holder() to create and remove holder/slave
symlinks.
* open_by_devnum() wraps bdget() + blkdev_get().
The interface is a bit confusing and the decoupling of open and claim
makes it impossible to properly guarantee exclusive access as
in-kernel open + claim sequence can disturb the existing exclusive
open even before the block layer knows the current open if for another
exclusive access. Reorganize the interface such that,
* blkdev_get() is extended to include exclusive access management.
@holder argument is added and, if is @FMODE_EXCL specified, it will
gain exclusive access atomically w.r.t. other exclusive accesses.
* blkdev_put() is similarly extended. It now takes @mode argument and
if @FMODE_EXCL is set, it releases an exclusive access. Also, when
the last exclusive claim is released, the holder/slave symlinks are
removed automatically.
* bd_claim/release() and close_bdev_exclusive() are no longer
necessary and either made static or removed.
* bd_link_disk_holder() remains the same but bd_unlink_disk_holder()
is no longer necessary and removed.
* open_bdev_exclusive() becomes a simple wrapper around lookup_bdev()
and blkdev_get(). It also has an unexpected extra bdev_read_only()
test which probably should be moved into blkdev_get().
* open_by_devnum() is modified to take @holder argument and pass it to
blkdev_get().
Most of bdev open/close operations are unified into blkdev_get/put()
and most exclusive accesses are tested atomically at the open time (as
it should). This cleans up code and removes some, both valid and
invalid, but unnecessary all the same, corner cases.
open_bdev_exclusive() and open_by_devnum() can use further cleanup -
rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop
special features. Well, let's leave them for another day.
Most conversions are straight-forward. drbd conversion is a bit more
involved as there was some reordering, but the logic should stay the
same.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Neil Brown <neilb@suse.de>
Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Acked-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Philipp Reisner <philipp.reisner@linbit.com>
Cc: Peter Osterlund <petero2@telia.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <joel.becker@oracle.com>
Cc: Alex Elder <aelder@sgi.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: dm-devel@redhat.com
Cc: drbd-dev@lists.linbit.com
Cc: Leo Chen <leochen@broadcom.com>
Cc: Scott Branden <sbranden@broadcom.com>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Joern Engel <joern@logfs.org>
Cc: reiserfs-devel@vger.kernel.org
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
|
|
|
blkdev_put(bdev, FMODE_READ | FMODE_EXCL);
|
2008-05-07 17:43:44 +02:00
|
|
|
out:
|
2008-07-08 20:19:17 +02:00
|
|
|
mutex_unlock(&root->fs_info->volume_mutex);
|
2008-05-07 17:43:44 +02:00
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
return ret;
|
2011-02-15 19:14:25 +01:00
|
|
|
error_undo:
|
|
|
|
if (device->writeable) {
|
2011-04-20 12:08:16 +02:00
|
|
|
lock_chunks(root);
|
2011-02-15 19:14:25 +01:00
|
|
|
list_add(&device->dev_alloc_list,
|
|
|
|
&root->fs_info->fs_devices->alloc_list);
|
2011-04-20 12:08:16 +02:00
|
|
|
unlock_chunks(root);
|
2011-02-15 19:14:25 +01:00
|
|
|
root->fs_info->fs_devices->rw_devices++;
|
|
|
|
}
|
|
|
|
goto error_brelse;
|
2008-05-07 17:43:44 +02:00
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
/*
|
|
|
|
* does all the dirty work required for changing file system's UUID.
|
|
|
|
*/
|
|
|
|
static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
|
|
|
|
struct btrfs_fs_devices *old_devices;
|
2008-12-12 16:03:26 +01:00
|
|
|
struct btrfs_fs_devices *seed_devices;
|
2008-11-18 03:11:30 +01:00
|
|
|
struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
|
|
|
|
struct btrfs_device *device;
|
|
|
|
u64 super_flags;
|
|
|
|
|
|
|
|
BUG_ON(!mutex_is_locked(&uuid_mutex));
|
2008-12-12 16:03:26 +01:00
|
|
|
if (!fs_devices->seeding)
|
2008-11-18 03:11:30 +01:00
|
|
|
return -EINVAL;
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
seed_devices = kzalloc(sizeof(*fs_devices), GFP_NOFS);
|
|
|
|
if (!seed_devices)
|
2008-11-18 03:11:30 +01:00
|
|
|
return -ENOMEM;
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
old_devices = clone_fs_devices(fs_devices);
|
|
|
|
if (IS_ERR(old_devices)) {
|
|
|
|
kfree(seed_devices);
|
|
|
|
return PTR_ERR(old_devices);
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
2008-12-12 16:03:26 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
list_add(&old_devices->list, &fs_uuids);
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
memcpy(seed_devices, fs_devices, sizeof(*seed_devices));
|
|
|
|
seed_devices->opened = 1;
|
|
|
|
INIT_LIST_HEAD(&seed_devices->devices);
|
|
|
|
INIT_LIST_HEAD(&seed_devices->alloc_list);
|
2009-06-10 21:17:02 +02:00
|
|
|
mutex_init(&seed_devices->device_list_mutex);
|
2011-04-20 12:07:30 +02:00
|
|
|
|
|
|
|
mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
|
2011-04-20 12:09:16 +02:00
|
|
|
list_splice_init_rcu(&fs_devices->devices, &seed_devices->devices,
|
|
|
|
synchronize_rcu);
|
2011-04-20 12:07:30 +02:00
|
|
|
mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
list_splice_init(&fs_devices->alloc_list, &seed_devices->alloc_list);
|
|
|
|
list_for_each_entry(device, &seed_devices->devices, dev_list) {
|
|
|
|
device->fs_devices = seed_devices;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
fs_devices->seeding = 0;
|
|
|
|
fs_devices->num_devices = 0;
|
|
|
|
fs_devices->open_devices = 0;
|
2008-12-12 16:03:26 +01:00
|
|
|
fs_devices->seed = seed_devices;
|
2008-11-18 03:11:30 +01:00
|
|
|
|
|
|
|
generate_random_uuid(fs_devices->fsid);
|
|
|
|
memcpy(root->fs_info->fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
|
|
|
|
memcpy(disk_super->fsid, fs_devices->fsid, BTRFS_FSID_SIZE);
|
|
|
|
super_flags = btrfs_super_flags(disk_super) &
|
|
|
|
~BTRFS_SUPER_FLAG_SEEDING;
|
|
|
|
btrfs_set_super_flags(disk_super, super_flags);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* strore the expected generation for seed devices in device items.
|
|
|
|
*/
|
|
|
|
static int btrfs_finish_sprout(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_dev_item *dev_item;
|
|
|
|
struct btrfs_device *device;
|
|
|
|
struct btrfs_key key;
|
|
|
|
u8 fs_uuid[BTRFS_UUID_SIZE];
|
|
|
|
u8 dev_uuid[BTRFS_UUID_SIZE];
|
|
|
|
u64 devid;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
|
|
|
|
key.offset = 0;
|
|
|
|
key.type = BTRFS_DEV_ITEM_KEY;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
next_slot:
|
|
|
|
if (path->slots[0] >= btrfs_header_nritems(leaf)) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret > 0)
|
|
|
|
break;
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2008-11-18 03:11:30 +01:00
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
|
|
|
|
if (key.objectid != BTRFS_DEV_ITEMS_OBJECTID ||
|
|
|
|
key.type != BTRFS_DEV_ITEM_KEY)
|
|
|
|
break;
|
|
|
|
|
|
|
|
dev_item = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_dev_item);
|
|
|
|
devid = btrfs_device_id(leaf, dev_item);
|
|
|
|
read_extent_buffer(leaf, dev_uuid,
|
|
|
|
(unsigned long)btrfs_device_uuid(dev_item),
|
|
|
|
BTRFS_UUID_SIZE);
|
|
|
|
read_extent_buffer(leaf, fs_uuid,
|
|
|
|
(unsigned long)btrfs_device_fsid(dev_item),
|
|
|
|
BTRFS_UUID_SIZE);
|
|
|
|
device = btrfs_find_device(root, devid, dev_uuid, fs_uuid);
|
|
|
|
BUG_ON(!device);
|
|
|
|
|
|
|
|
if (device->fs_devices->seeding) {
|
|
|
|
btrfs_set_device_generation(leaf, dev_item,
|
|
|
|
device->generation);
|
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
|
|
|
}
|
|
|
|
|
|
|
|
path->slots[0]++;
|
|
|
|
goto next_slot;
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
error:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-04-28 21:29:42 +02:00
|
|
|
int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
|
|
|
|
{
|
2011-08-04 16:52:27 +02:00
|
|
|
struct request_queue *q;
|
2008-04-28 21:29:42 +02:00
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
struct btrfs_device *device;
|
|
|
|
struct block_device *bdev;
|
|
|
|
struct list_head *devices;
|
2008-11-18 03:11:30 +01:00
|
|
|
struct super_block *sb = root->fs_info->sb;
|
2008-04-28 21:29:42 +02:00
|
|
|
u64 total_bytes;
|
2008-11-18 03:11:30 +01:00
|
|
|
int seeding_dev = 0;
|
2008-04-28 21:29:42 +02:00
|
|
|
int ret = 0;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if ((sb->s_flags & MS_RDONLY) && !root->fs_info->fs_devices->seeding)
|
|
|
|
return -EINVAL;
|
2008-04-28 21:29:42 +02:00
|
|
|
|
2010-11-13 11:55:18 +01:00
|
|
|
bdev = blkdev_get_by_path(device_path, FMODE_EXCL,
|
|
|
|
root->fs_info->bdev_holder);
|
2010-01-27 03:09:00 +01:00
|
|
|
if (IS_ERR(bdev))
|
|
|
|
return PTR_ERR(bdev);
|
2008-06-25 22:01:30 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (root->fs_info->fs_devices->seeding) {
|
|
|
|
seeding_dev = 1;
|
|
|
|
down_write(&sb->s_umount);
|
|
|
|
mutex_lock(&uuid_mutex);
|
|
|
|
}
|
|
|
|
|
2008-09-29 17:19:10 +02:00
|
|
|
filemap_write_and_wait(bdev->bd_inode->i_mapping);
|
2008-07-08 20:19:17 +02:00
|
|
|
mutex_lock(&root->fs_info->volume_mutex);
|
2008-06-25 22:01:30 +02:00
|
|
|
|
2008-04-28 21:29:42 +02:00
|
|
|
devices = &root->fs_info->fs_devices->devices;
|
2009-06-10 21:17:02 +02:00
|
|
|
/*
|
|
|
|
* we have the volume lock, so we don't need the extra
|
|
|
|
* device list mutex while reading the list here.
|
|
|
|
*/
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(device, devices, dev_list) {
|
2008-04-28 21:29:42 +02:00
|
|
|
if (device->bdev == bdev) {
|
|
|
|
ret = -EEXIST;
|
2008-11-18 03:11:30 +01:00
|
|
|
goto error;
|
2008-04-28 21:29:42 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
device = kzalloc(sizeof(*device), GFP_NOFS);
|
|
|
|
if (!device) {
|
|
|
|
/* we can safely leave the fs_devices entry around */
|
|
|
|
ret = -ENOMEM;
|
2008-11-18 03:11:30 +01:00
|
|
|
goto error;
|
2008-04-28 21:29:42 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
device->name = kstrdup(device_path, GFP_NOFS);
|
|
|
|
if (!device->name) {
|
|
|
|
kfree(device);
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = -ENOMEM;
|
|
|
|
goto error;
|
2008-04-28 21:29:42 +02:00
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
|
|
|
|
ret = find_next_devid(root, &device->devid);
|
|
|
|
if (ret) {
|
2011-02-06 20:58:21 +01:00
|
|
|
kfree(device->name);
|
2008-11-18 03:11:30 +01:00
|
|
|
kfree(device);
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2010-05-16 16:48:46 +02:00
|
|
|
trans = btrfs_start_transaction(root, 0);
|
2011-01-20 07:19:37 +01:00
|
|
|
if (IS_ERR(trans)) {
|
2011-02-06 20:58:21 +01:00
|
|
|
kfree(device->name);
|
2011-01-20 07:19:37 +01:00
|
|
|
kfree(device);
|
|
|
|
ret = PTR_ERR(trans);
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
lock_chunks(root);
|
|
|
|
|
2011-08-04 16:52:27 +02:00
|
|
|
q = bdev_get_queue(bdev);
|
|
|
|
if (blk_queue_discard(q))
|
|
|
|
device->can_discard = 1;
|
2008-11-18 03:11:30 +01:00
|
|
|
device->writeable = 1;
|
|
|
|
device->work.func = pending_bios_fn;
|
|
|
|
generate_random_uuid(device->uuid);
|
|
|
|
spin_lock_init(&device->io_lock);
|
|
|
|
device->generation = trans->transid;
|
2008-04-28 21:29:42 +02:00
|
|
|
device->io_width = root->sectorsize;
|
|
|
|
device->io_align = root->sectorsize;
|
|
|
|
device->sector_size = root->sectorsize;
|
|
|
|
device->total_bytes = i_size_read(bdev->bd_inode);
|
2009-06-04 15:23:50 +02:00
|
|
|
device->disk_total_bytes = device->total_bytes;
|
2008-04-28 21:29:42 +02:00
|
|
|
device->dev_root = root->fs_info->dev_root;
|
|
|
|
device->bdev = bdev;
|
2008-05-13 19:46:40 +02:00
|
|
|
device->in_fs_metadata = 1;
|
2011-02-15 19:12:57 +01:00
|
|
|
device->mode = FMODE_EXCL;
|
2008-11-18 03:11:30 +01:00
|
|
|
set_blocksize(device->bdev, 4096);
|
2008-04-28 21:29:42 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (seeding_dev) {
|
|
|
|
sb->s_flags &= ~MS_RDONLY;
|
|
|
|
ret = btrfs_prepare_sprout(trans, root);
|
|
|
|
BUG_ON(ret);
|
|
|
|
}
|
2008-04-28 21:29:42 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
device->fs_devices = root->fs_info->fs_devices;
|
2009-06-10 21:17:02 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* we don't want write_supers to jump in here with our device
|
|
|
|
* half setup
|
|
|
|
*/
|
|
|
|
mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
|
2011-04-20 12:09:16 +02:00
|
|
|
list_add_rcu(&device->dev_list, &root->fs_info->fs_devices->devices);
|
2008-11-18 03:11:30 +01:00
|
|
|
list_add(&device->dev_alloc_list,
|
|
|
|
&root->fs_info->fs_devices->alloc_list);
|
|
|
|
root->fs_info->fs_devices->num_devices++;
|
|
|
|
root->fs_info->fs_devices->open_devices++;
|
|
|
|
root->fs_info->fs_devices->rw_devices++;
|
2011-08-04 16:52:27 +02:00
|
|
|
if (device->can_discard)
|
|
|
|
root->fs_info->fs_devices->num_can_discard++;
|
2008-11-18 03:11:30 +01:00
|
|
|
root->fs_info->fs_devices->total_rw_bytes += device->total_bytes;
|
2008-09-05 22:43:54 +02:00
|
|
|
|
2011-09-26 23:12:22 +02:00
|
|
|
spin_lock(&root->fs_info->free_chunk_lock);
|
|
|
|
root->fs_info->free_chunk_space += device->total_bytes;
|
|
|
|
spin_unlock(&root->fs_info->free_chunk_lock);
|
|
|
|
|
2009-06-10 15:51:32 +02:00
|
|
|
if (!blk_queue_nonrot(bdev_get_queue(bdev)))
|
|
|
|
root->fs_info->fs_devices->rotating = 1;
|
|
|
|
|
2008-04-28 21:29:42 +02:00
|
|
|
total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
|
|
|
|
btrfs_set_super_total_bytes(&root->fs_info->super_copy,
|
|
|
|
total_bytes + device->total_bytes);
|
|
|
|
|
|
|
|
total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy);
|
|
|
|
btrfs_set_super_num_devices(&root->fs_info->super_copy,
|
|
|
|
total_bytes + 1);
|
2009-06-10 21:17:02 +02:00
|
|
|
mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
|
2008-04-28 21:29:42 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (seeding_dev) {
|
|
|
|
ret = init_first_rw_device(trans, root, device);
|
|
|
|
BUG_ON(ret);
|
|
|
|
ret = btrfs_finish_sprout(trans, root);
|
|
|
|
BUG_ON(ret);
|
|
|
|
} else {
|
|
|
|
ret = btrfs_add_device(trans, root, device);
|
|
|
|
}
|
|
|
|
|
2009-03-10 18:17:18 +01:00
|
|
|
/*
|
|
|
|
* we've got more storage, clear any full flags on the space
|
|
|
|
* infos
|
|
|
|
*/
|
|
|
|
btrfs_clear_space_info_full(root->fs_info);
|
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
unlock_chunks(root);
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_commit_transaction(trans, root);
|
2008-06-25 22:01:30 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (seeding_dev) {
|
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
up_write(&sb->s_umount);
|
2008-04-28 21:29:42 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = btrfs_relocate_sys_chunks(root);
|
|
|
|
BUG_ON(ret);
|
|
|
|
}
|
|
|
|
out:
|
|
|
|
mutex_unlock(&root->fs_info->volume_mutex);
|
|
|
|
return ret;
|
|
|
|
error:
|
block: make blkdev_get/put() handle exclusive access
Over time, block layer has accumulated a set of APIs dealing with bdev
open, close, claim and release.
* blkdev_get/put() are the primary open and close functions.
* bd_claim/release() deal with exclusive open.
* open/close_bdev_exclusive() are combination of open and claim and
the other way around, respectively.
* bd_link/unlink_disk_holder() to create and remove holder/slave
symlinks.
* open_by_devnum() wraps bdget() + blkdev_get().
The interface is a bit confusing and the decoupling of open and claim
makes it impossible to properly guarantee exclusive access as
in-kernel open + claim sequence can disturb the existing exclusive
open even before the block layer knows the current open if for another
exclusive access. Reorganize the interface such that,
* blkdev_get() is extended to include exclusive access management.
@holder argument is added and, if is @FMODE_EXCL specified, it will
gain exclusive access atomically w.r.t. other exclusive accesses.
* blkdev_put() is similarly extended. It now takes @mode argument and
if @FMODE_EXCL is set, it releases an exclusive access. Also, when
the last exclusive claim is released, the holder/slave symlinks are
removed automatically.
* bd_claim/release() and close_bdev_exclusive() are no longer
necessary and either made static or removed.
* bd_link_disk_holder() remains the same but bd_unlink_disk_holder()
is no longer necessary and removed.
* open_bdev_exclusive() becomes a simple wrapper around lookup_bdev()
and blkdev_get(). It also has an unexpected extra bdev_read_only()
test which probably should be moved into blkdev_get().
* open_by_devnum() is modified to take @holder argument and pass it to
blkdev_get().
Most of bdev open/close operations are unified into blkdev_get/put()
and most exclusive accesses are tested atomically at the open time (as
it should). This cleans up code and removes some, both valid and
invalid, but unnecessary all the same, corner cases.
open_bdev_exclusive() and open_by_devnum() can use further cleanup -
rename to blkdev_get_by_path() and blkdev_get_by_devt() and drop
special features. Well, let's leave them for another day.
Most conversions are straight-forward. drbd conversion is a bit more
involved as there was some reordering, but the logic should stay the
same.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Neil Brown <neilb@suse.de>
Acked-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Acked-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Philipp Reisner <philipp.reisner@linbit.com>
Cc: Peter Osterlund <petero2@telia.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <joel.becker@oracle.com>
Cc: Alex Elder <aelder@sgi.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: dm-devel@redhat.com
Cc: drbd-dev@lists.linbit.com
Cc: Leo Chen <leochen@broadcom.com>
Cc: Scott Branden <sbranden@broadcom.com>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Cc: Joern Engel <joern@logfs.org>
Cc: reiserfs-devel@vger.kernel.org
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
2010-11-13 11:55:17 +01:00
|
|
|
blkdev_put(bdev, FMODE_EXCL);
|
2008-11-18 03:11:30 +01:00
|
|
|
if (seeding_dev) {
|
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
up_write(&sb->s_umount);
|
|
|
|
}
|
2008-04-28 21:29:42 +02:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
static noinline int btrfs_update_device(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_device *device)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_root *root;
|
|
|
|
struct btrfs_dev_item *dev_item;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_key key;
|
|
|
|
|
|
|
|
root = device->dev_root->fs_info->chunk_root;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
|
|
|
|
key.type = BTRFS_DEV_ITEM_KEY;
|
|
|
|
key.offset = device->devid;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (ret > 0) {
|
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
leaf = path->nodes[0];
|
|
|
|
dev_item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_dev_item);
|
|
|
|
|
|
|
|
btrfs_set_device_id(leaf, dev_item, device->devid);
|
|
|
|
btrfs_set_device_type(leaf, dev_item, device->type);
|
|
|
|
btrfs_set_device_io_align(leaf, dev_item, device->io_align);
|
|
|
|
btrfs_set_device_io_width(leaf, dev_item, device->io_width);
|
|
|
|
btrfs_set_device_sector_size(leaf, dev_item, device->sector_size);
|
2009-04-27 13:29:03 +02:00
|
|
|
btrfs_set_device_total_bytes(leaf, dev_item, device->disk_total_bytes);
|
2008-03-24 20:01:56 +01:00
|
|
|
btrfs_set_device_bytes_used(leaf, dev_item, device->bytes_used);
|
|
|
|
btrfs_mark_buffer_dirty(leaf);
|
|
|
|
|
|
|
|
out:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
static int __btrfs_grow_device(struct btrfs_trans_handle *trans,
|
2008-04-25 22:53:30 +02:00
|
|
|
struct btrfs_device *device, u64 new_size)
|
|
|
|
{
|
|
|
|
struct btrfs_super_block *super_copy =
|
|
|
|
&device->dev_root->fs_info->super_copy;
|
|
|
|
u64 old_total = btrfs_super_total_bytes(super_copy);
|
|
|
|
u64 diff = new_size - device->total_bytes;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (!device->writeable)
|
|
|
|
return -EACCES;
|
|
|
|
if (new_size <= device->total_bytes)
|
|
|
|
return -EINVAL;
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
btrfs_set_super_total_bytes(super_copy, old_total + diff);
|
2008-11-18 03:11:30 +01:00
|
|
|
device->fs_devices->total_rw_bytes += diff;
|
|
|
|
|
|
|
|
device->total_bytes = new_size;
|
2009-07-24 22:41:41 +02:00
|
|
|
device->disk_total_bytes = new_size;
|
2009-03-10 17:39:20 +01:00
|
|
|
btrfs_clear_space_info_full(device->dev_root->fs_info);
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
return btrfs_update_device(trans, device);
|
|
|
|
}
|
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
int btrfs_grow_device(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_device *device, u64 new_size)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
lock_chunks(device->dev_root);
|
|
|
|
ret = __btrfs_grow_device(trans, device, new_size);
|
|
|
|
unlock_chunks(device->dev_root);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
static int btrfs_free_chunk(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *root,
|
|
|
|
u64 chunk_tree, u64 chunk_objectid,
|
|
|
|
u64 chunk_offset)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_key key;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
key.objectid = chunk_objectid;
|
|
|
|
key.offset = chunk_offset;
|
|
|
|
key.type = BTRFS_CHUNK_ITEM_KEY;
|
|
|
|
|
|
|
|
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
ret = btrfs_del_item(trans, root, path);
|
|
|
|
|
|
|
|
btrfs_free_path(path);
|
2011-05-19 06:37:44 +02:00
|
|
|
return ret;
|
2008-04-25 22:53:30 +02:00
|
|
|
}
|
|
|
|
|
2008-12-02 15:54:17 +01:00
|
|
|
static int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64
|
2008-04-25 22:53:30 +02:00
|
|
|
chunk_offset)
|
|
|
|
{
|
|
|
|
struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
|
|
|
|
struct btrfs_disk_key *disk_key;
|
|
|
|
struct btrfs_chunk *chunk;
|
|
|
|
u8 *ptr;
|
|
|
|
int ret = 0;
|
|
|
|
u32 num_stripes;
|
|
|
|
u32 array_size;
|
|
|
|
u32 len = 0;
|
|
|
|
u32 cur;
|
|
|
|
struct btrfs_key key;
|
|
|
|
|
|
|
|
array_size = btrfs_super_sys_array_size(super_copy);
|
|
|
|
|
|
|
|
ptr = super_copy->sys_chunk_array;
|
|
|
|
cur = 0;
|
|
|
|
|
|
|
|
while (cur < array_size) {
|
|
|
|
disk_key = (struct btrfs_disk_key *)ptr;
|
|
|
|
btrfs_disk_key_to_cpu(&key, disk_key);
|
|
|
|
|
|
|
|
len = sizeof(*disk_key);
|
|
|
|
|
|
|
|
if (key.type == BTRFS_CHUNK_ITEM_KEY) {
|
|
|
|
chunk = (struct btrfs_chunk *)(ptr + len);
|
|
|
|
num_stripes = btrfs_stack_chunk_num_stripes(chunk);
|
|
|
|
len += btrfs_chunk_item_size(num_stripes);
|
|
|
|
} else {
|
|
|
|
ret = -EIO;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
if (key.objectid == chunk_objectid &&
|
|
|
|
key.offset == chunk_offset) {
|
|
|
|
memmove(ptr, ptr + len, array_size - (cur + len));
|
|
|
|
array_size -= len;
|
|
|
|
btrfs_set_super_sys_array_size(super_copy, array_size);
|
|
|
|
} else {
|
|
|
|
ptr += len;
|
|
|
|
cur += len;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-02 15:54:17 +01:00
|
|
|
static int btrfs_relocate_chunk(struct btrfs_root *root,
|
2008-04-25 22:53:30 +02:00
|
|
|
u64 chunk_tree, u64 chunk_objectid,
|
|
|
|
u64 chunk_offset)
|
|
|
|
{
|
|
|
|
struct extent_map_tree *em_tree;
|
|
|
|
struct btrfs_root *extent_root;
|
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
struct extent_map *em;
|
|
|
|
struct map_lookup *map;
|
|
|
|
int ret;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
extent_root = root->fs_info->extent_root;
|
|
|
|
em_tree = &root->fs_info->mapping_tree.map_tree;
|
|
|
|
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
ret = btrfs_can_relocate(extent_root, chunk_offset);
|
|
|
|
if (ret)
|
|
|
|
return -ENOSPC;
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
/* step one, relocate all the extents inside this chunk */
|
Btrfs: update space balancing code
This patch updates the space balancing code to utilize the new
backref format. Before, btrfs-vol -b would break any COW links
on data blocks or metadata. This was slow and caused the amount
of space used to explode if a large number of snapshots were present.
The new code can keeps the sharing of all data extents and
most of the tree blocks.
To maintain the sharing of data extents, the space balance code uses
a seperate inode hold data extent pointers, then updates the references
to point to the new location.
To maintain the sharing of tree blocks, the space balance code uses
reloc trees to relocate tree blocks in reference counted roots.
There is one reloc tree for each subvol, and all reloc trees share
same root key objectid. Reloc trees are snapshots of the latest
committed roots of subvols (root->commit_root).
To relocate a tree block referenced by a subvol, there are two steps.
COW the block through subvol's reloc tree, then update block pointer in
the subvol to point to the new block. Since all reloc trees share
same root key objectid, doing special handing for tree blocks
owned by them is easy. Once a tree block has been COWed in one
reloc tree, we can use the resulting new block directly when the
same block is required to COW again through other reloc trees.
In this way, relocated tree blocks are shared between reloc trees,
so they are also shared between subvols.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-26 16:09:34 +02:00
|
|
|
ret = btrfs_relocate_block_group(extent_root, chunk_offset);
|
2010-05-16 16:48:46 +02:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2010-05-16 16:48:46 +02:00
|
|
|
trans = btrfs_start_transaction(root, 0);
|
2011-01-20 07:19:37 +01:00
|
|
|
BUG_ON(IS_ERR(trans));
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
lock_chunks(root);
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
/*
|
|
|
|
* step two, delete the device extents and the
|
|
|
|
* chunk tree entries
|
|
|
|
*/
|
2009-09-02 22:24:52 +02:00
|
|
|
read_lock(&em_tree->lock);
|
2008-04-25 22:53:30 +02:00
|
|
|
em = lookup_extent_mapping(em_tree, chunk_offset, 1);
|
2009-09-02 22:24:52 +02:00
|
|
|
read_unlock(&em_tree->lock);
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2008-05-07 17:43:44 +02:00
|
|
|
BUG_ON(em->start > chunk_offset ||
|
|
|
|
em->start + em->len < chunk_offset);
|
2008-04-25 22:53:30 +02:00
|
|
|
map = (struct map_lookup *)em->bdev;
|
|
|
|
|
|
|
|
for (i = 0; i < map->num_stripes; i++) {
|
|
|
|
ret = btrfs_free_dev_extent(trans, map->stripes[i].dev,
|
|
|
|
map->stripes[i].physical);
|
|
|
|
BUG_ON(ret);
|
2008-05-07 17:43:44 +02:00
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
if (map->stripes[i].dev) {
|
|
|
|
ret = btrfs_update_device(trans, map->stripes[i].dev);
|
|
|
|
BUG_ON(ret);
|
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
}
|
|
|
|
ret = btrfs_free_chunk(trans, root, chunk_tree, chunk_objectid,
|
|
|
|
chunk_offset);
|
|
|
|
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 12:18:59 +01:00
|
|
|
trace_btrfs_chunk_free(root, map, chunk_offset, em->len);
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
|
|
|
|
ret = btrfs_del_sys_chunk(root, chunk_objectid, chunk_offset);
|
|
|
|
BUG_ON(ret);
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = btrfs_remove_block_group(trans, extent_root, chunk_offset);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
write_lock(&em_tree->lock);
|
2008-11-18 03:11:30 +01:00
|
|
|
remove_extent_mapping(em_tree, em);
|
2009-09-02 22:24:52 +02:00
|
|
|
write_unlock(&em_tree->lock);
|
2008-11-18 03:11:30 +01:00
|
|
|
|
|
|
|
kfree(map);
|
|
|
|
em->bdev = NULL;
|
|
|
|
|
|
|
|
/* once for the tree */
|
|
|
|
free_extent_map(em);
|
|
|
|
/* once for us */
|
|
|
|
free_extent_map(em);
|
|
|
|
|
|
|
|
unlock_chunks(root);
|
|
|
|
btrfs_end_transaction(trans, root);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int btrfs_relocate_sys_chunks(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_root *chunk_root = root->fs_info->chunk_root;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_chunk *chunk;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_key found_key;
|
|
|
|
u64 chunk_tree = chunk_root->root_key.objectid;
|
|
|
|
u64 chunk_type;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
bool retried = false;
|
|
|
|
int failed = 0;
|
2008-11-18 03:11:30 +01:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
again:
|
2008-11-18 03:11:30 +01:00
|
|
|
key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
|
|
|
|
key.offset = (u64)-1;
|
|
|
|
key.type = BTRFS_CHUNK_ITEM_KEY;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
BUG_ON(ret == 0);
|
|
|
|
|
|
|
|
ret = btrfs_previous_item(chunk_root, path, key.objectid,
|
|
|
|
key.type);
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
if (ret > 0)
|
|
|
|
break;
|
Btrfs: update space balancing code
This patch updates the space balancing code to utilize the new
backref format. Before, btrfs-vol -b would break any COW links
on data blocks or metadata. This was slow and caused the amount
of space used to explode if a large number of snapshots were present.
The new code can keeps the sharing of all data extents and
most of the tree blocks.
To maintain the sharing of data extents, the space balance code uses
a seperate inode hold data extent pointers, then updates the references
to point to the new location.
To maintain the sharing of tree blocks, the space balance code uses
reloc trees to relocate tree blocks in reference counted roots.
There is one reloc tree for each subvol, and all reloc trees share
same root key objectid. Reloc trees are snapshots of the latest
committed roots of subvols (root->commit_root).
To relocate a tree block referenced by a subvol, there are two steps.
COW the block through subvol's reloc tree, then update block pointer in
the subvol to point to the new block. Since all reloc trees share
same root key objectid, doing special handing for tree blocks
owned by them is easy. Once a tree block has been COWed in one
reloc tree, we can use the resulting new block directly when the
same block is required to COW again through other reloc trees.
In this way, relocated tree blocks are shared between reloc trees,
so they are also shared between subvols.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-26 16:09:34 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
leaf = path->nodes[0];
|
|
|
|
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
|
Btrfs: update space balancing code
This patch updates the space balancing code to utilize the new
backref format. Before, btrfs-vol -b would break any COW links
on data blocks or metadata. This was slow and caused the amount
of space used to explode if a large number of snapshots were present.
The new code can keeps the sharing of all data extents and
most of the tree blocks.
To maintain the sharing of data extents, the space balance code uses
a seperate inode hold data extent pointers, then updates the references
to point to the new location.
To maintain the sharing of tree blocks, the space balance code uses
reloc trees to relocate tree blocks in reference counted roots.
There is one reloc tree for each subvol, and all reloc trees share
same root key objectid. Reloc trees are snapshots of the latest
committed roots of subvols (root->commit_root).
To relocate a tree block referenced by a subvol, there are two steps.
COW the block through subvol's reloc tree, then update block pointer in
the subvol to point to the new block. Since all reloc trees share
same root key objectid, doing special handing for tree blocks
owned by them is easy. Once a tree block has been COWed in one
reloc tree, we can use the resulting new block directly when the
same block is required to COW again through other reloc trees.
In this way, relocated tree blocks are shared between reloc trees,
so they are also shared between subvols.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-26 16:09:34 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
chunk = btrfs_item_ptr(leaf, path->slots[0],
|
|
|
|
struct btrfs_chunk);
|
|
|
|
chunk_type = btrfs_chunk_type(leaf, chunk);
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (chunk_type & BTRFS_BLOCK_GROUP_SYSTEM) {
|
|
|
|
ret = btrfs_relocate_chunk(chunk_root, chunk_tree,
|
|
|
|
found_key.objectid,
|
|
|
|
found_key.offset);
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (ret == -ENOSPC)
|
|
|
|
failed++;
|
|
|
|
else if (ret)
|
|
|
|
BUG();
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (found_key.offset == 0)
|
|
|
|
break;
|
|
|
|
key.offset = found_key.offset - 1;
|
|
|
|
}
|
|
|
|
ret = 0;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (failed && !retried) {
|
|
|
|
failed = 0;
|
|
|
|
retried = true;
|
|
|
|
goto again;
|
|
|
|
} else if (failed && retried) {
|
|
|
|
WARN_ON(1);
|
|
|
|
ret = -ENOSPC;
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
error:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
2008-04-25 22:53:30 +02:00
|
|
|
}
|
|
|
|
|
2008-04-28 21:29:52 +02:00
|
|
|
static u64 div_factor(u64 num, int factor)
|
|
|
|
{
|
|
|
|
if (factor == 10)
|
|
|
|
return num;
|
|
|
|
num *= factor;
|
|
|
|
do_div(num, 10);
|
|
|
|
return num;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_balance(struct btrfs_root *dev_root)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
struct list_head *devices = &dev_root->fs_info->fs_devices->devices;
|
|
|
|
struct btrfs_device *device;
|
|
|
|
u64 old_size;
|
|
|
|
u64 size_to_free;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_root *chunk_root = dev_root->fs_info->chunk_root;
|
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
struct btrfs_key found_key;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (dev_root->fs_info->sb->s_flags & MS_RDONLY)
|
|
|
|
return -EROFS;
|
2008-04-28 21:29:52 +02:00
|
|
|
|
2010-12-29 15:55:03 +01:00
|
|
|
if (!capable(CAP_SYS_ADMIN))
|
|
|
|
return -EPERM;
|
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
mutex_lock(&dev_root->fs_info->volume_mutex);
|
2008-04-28 21:29:52 +02:00
|
|
|
dev_root = dev_root->fs_info->dev_root;
|
|
|
|
|
|
|
|
/* step one make some room on all the devices */
|
2009-01-21 16:59:08 +01:00
|
|
|
list_for_each_entry(device, devices, dev_list) {
|
2008-04-28 21:29:52 +02:00
|
|
|
old_size = device->total_bytes;
|
|
|
|
size_to_free = div_factor(old_size, 1);
|
|
|
|
size_to_free = min(size_to_free, (u64)1 * 1024 * 1024);
|
2008-11-18 03:11:30 +01:00
|
|
|
if (!device->writeable ||
|
|
|
|
device->total_bytes - device->bytes_used > size_to_free)
|
2008-04-28 21:29:52 +02:00
|
|
|
continue;
|
|
|
|
|
|
|
|
ret = btrfs_shrink_device(device, old_size - size_to_free);
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (ret == -ENOSPC)
|
|
|
|
break;
|
2008-04-28 21:29:52 +02:00
|
|
|
BUG_ON(ret);
|
|
|
|
|
2010-05-16 16:48:46 +02:00
|
|
|
trans = btrfs_start_transaction(dev_root, 0);
|
2011-01-20 07:19:37 +01:00
|
|
|
BUG_ON(IS_ERR(trans));
|
2008-04-28 21:29:52 +02:00
|
|
|
|
|
|
|
ret = btrfs_grow_device(trans, device, old_size);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
btrfs_end_transaction(trans, dev_root);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* step two, relocate all the chunks */
|
|
|
|
path = btrfs_alloc_path();
|
2011-07-12 20:10:23 +02:00
|
|
|
if (!path) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto error;
|
|
|
|
}
|
2008-04-28 21:29:52 +02:00
|
|
|
key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
|
|
|
|
key.offset = (u64)-1;
|
|
|
|
key.type = BTRFS_CHUNK_ITEM_KEY;
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
while (1) {
|
2008-04-28 21:29:52 +02:00
|
|
|
ret = btrfs_search_slot(NULL, chunk_root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* this shouldn't happen, it means the last relocate
|
|
|
|
* failed
|
|
|
|
*/
|
|
|
|
if (ret == 0)
|
|
|
|
break;
|
|
|
|
|
|
|
|
ret = btrfs_previous_item(chunk_root, path, 0,
|
|
|
|
BTRFS_CHUNK_ITEM_KEY);
|
2008-07-08 20:19:17 +02:00
|
|
|
if (ret)
|
2008-04-28 21:29:52 +02:00
|
|
|
break;
|
2008-07-08 20:19:17 +02:00
|
|
|
|
2008-04-28 21:29:52 +02:00
|
|
|
btrfs_item_key_to_cpu(path->nodes[0], &found_key,
|
|
|
|
path->slots[0]);
|
|
|
|
if (found_key.objectid != key.objectid)
|
|
|
|
break;
|
2008-07-08 20:19:17 +02:00
|
|
|
|
2008-04-28 21:29:52 +02:00
|
|
|
/* chunk zero is special */
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (found_key.offset == 0)
|
2008-04-28 21:29:52 +02:00
|
|
|
break;
|
|
|
|
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2008-04-28 21:29:52 +02:00
|
|
|
ret = btrfs_relocate_chunk(chunk_root,
|
|
|
|
chunk_root->root_key.objectid,
|
|
|
|
found_key.objectid,
|
|
|
|
found_key.offset);
|
2011-07-02 23:24:41 +02:00
|
|
|
if (ret && ret != -ENOSPC)
|
|
|
|
goto error;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
key.offset = found_key.offset - 1;
|
2008-04-28 21:29:52 +02:00
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
error:
|
|
|
|
btrfs_free_path(path);
|
2008-07-08 20:19:17 +02:00
|
|
|
mutex_unlock(&dev_root->fs_info->volume_mutex);
|
2008-04-28 21:29:52 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
/*
|
|
|
|
* shrinking a device means finding all of the device extents past
|
|
|
|
* the new size, and then following the back refs to the chunks.
|
|
|
|
* The chunk relocation code actually frees the device extent
|
|
|
|
*/
|
|
|
|
int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
|
|
|
|
{
|
|
|
|
struct btrfs_trans_handle *trans;
|
|
|
|
struct btrfs_root *root = device->dev_root;
|
|
|
|
struct btrfs_dev_extent *dev_extent = NULL;
|
|
|
|
struct btrfs_path *path;
|
|
|
|
u64 length;
|
|
|
|
u64 chunk_tree;
|
|
|
|
u64 chunk_objectid;
|
|
|
|
u64 chunk_offset;
|
|
|
|
int ret;
|
|
|
|
int slot;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
int failed = 0;
|
|
|
|
bool retried = false;
|
2008-04-25 22:53:30 +02:00
|
|
|
struct extent_buffer *l;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
|
|
|
|
u64 old_total = btrfs_super_total_bytes(super_copy);
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
u64 old_size = device->total_bytes;
|
2008-04-25 22:53:30 +02:00
|
|
|
u64 diff = device->total_bytes - new_size;
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (new_size >= device->total_bytes)
|
|
|
|
return -EINVAL;
|
2008-04-25 22:53:30 +02:00
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
path->reada = 2;
|
|
|
|
|
2008-07-08 20:19:17 +02:00
|
|
|
lock_chunks(root);
|
|
|
|
|
2008-04-25 22:53:30 +02:00
|
|
|
device->total_bytes = new_size;
|
2011-09-26 23:12:22 +02:00
|
|
|
if (device->writeable) {
|
2008-11-18 03:11:30 +01:00
|
|
|
device->fs_devices->total_rw_bytes -= diff;
|
2011-09-26 23:12:22 +02:00
|
|
|
spin_lock(&root->fs_info->free_chunk_lock);
|
|
|
|
root->fs_info->free_chunk_space -= diff;
|
|
|
|
spin_unlock(&root->fs_info->free_chunk_lock);
|
|
|
|
}
|
2008-07-08 20:19:17 +02:00
|
|
|
unlock_chunks(root);
|
2008-04-25 22:53:30 +02:00
|
|
|
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
again:
|
2008-04-25 22:53:30 +02:00
|
|
|
key.objectid = device->devid;
|
|
|
|
key.offset = (u64)-1;
|
|
|
|
key.type = BTRFS_DEV_EXTENT_KEY;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
|
|
|
if (ret < 0)
|
|
|
|
goto done;
|
|
|
|
|
|
|
|
ret = btrfs_previous_item(root, path, 0, key.type);
|
|
|
|
if (ret < 0)
|
|
|
|
goto done;
|
|
|
|
if (ret) {
|
|
|
|
ret = 0;
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2009-07-22 15:59:00 +02:00
|
|
|
break;
|
2008-04-25 22:53:30 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
l = path->nodes[0];
|
|
|
|
slot = path->slots[0];
|
|
|
|
btrfs_item_key_to_cpu(l, &key, path->slots[0]);
|
|
|
|
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (key.objectid != device->devid) {
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2009-07-22 15:59:00 +02:00
|
|
|
break;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
|
|
|
|
dev_extent = btrfs_item_ptr(l, slot, struct btrfs_dev_extent);
|
|
|
|
length = btrfs_dev_extent_length(l, dev_extent);
|
|
|
|
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (key.offset + length <= new_size) {
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2009-04-27 13:29:03 +02:00
|
|
|
break;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
}
|
2008-04-25 22:53:30 +02:00
|
|
|
|
|
|
|
chunk_tree = btrfs_dev_extent_chunk_tree(l, dev_extent);
|
|
|
|
chunk_objectid = btrfs_dev_extent_chunk_objectid(l, dev_extent);
|
|
|
|
chunk_offset = btrfs_dev_extent_chunk_offset(l, dev_extent);
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2008-04-25 22:53:30 +02:00
|
|
|
|
|
|
|
ret = btrfs_relocate_chunk(root, chunk_tree, chunk_objectid,
|
|
|
|
chunk_offset);
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (ret && ret != -ENOSPC)
|
2008-04-25 22:53:30 +02:00
|
|
|
goto done;
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
if (ret == -ENOSPC)
|
|
|
|
failed++;
|
|
|
|
key.offset -= 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (failed && !retried) {
|
|
|
|
failed = 0;
|
|
|
|
retried = true;
|
|
|
|
goto again;
|
|
|
|
} else if (failed && retried) {
|
|
|
|
ret = -ENOSPC;
|
|
|
|
lock_chunks(root);
|
|
|
|
|
|
|
|
device->total_bytes = old_size;
|
|
|
|
if (device->writeable)
|
|
|
|
device->fs_devices->total_rw_bytes += diff;
|
2011-09-26 23:12:22 +02:00
|
|
|
spin_lock(&root->fs_info->free_chunk_lock);
|
|
|
|
root->fs_info->free_chunk_space += diff;
|
|
|
|
spin_unlock(&root->fs_info->free_chunk_lock);
|
Btrfs: make balance code choose more wisely when relocating
Currently, we can panic the box if the first block group we go to move is of a
type where there is no space left to move those extents. For example, if we
fill the disk up with data, and then we try to balance and we have no room to
move the data nor room to allocate new chunks, we will panic. Change this by
checking to see if we have room to move this chunk around, and if not, return
-ENOSPC and move on to the next chunk. This will make sure we remove block
groups that are moveable, like if we have alot of empty metadata block groups,
and then that way we make room to be able to balance our data chunks as well.
Tested this with an fs that would panic on btrfs-vol -b normally, but no longer
panics with this patch.
V1->V2:
-actually search for a free extent on the device to make sure we can allocate a
chunk if need be.
-fix btrfs_shrink_device to make sure we actually try to relocate all the
chunks, and then if we can't return -ENOSPC so if we are doing a btrfs-vol -r
we don't remove the device with data still on it.
-check to make sure the block group we are going to relocate isn't the last one
in that particular space
-fix a bug in btrfs_shrink_device where we would change the device's size and
not fix it if we fail to do our relocate
Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 22:11:19 +02:00
|
|
|
unlock_chunks(root);
|
|
|
|
goto done;
|
2008-04-25 22:53:30 +02:00
|
|
|
}
|
|
|
|
|
2009-04-27 13:29:03 +02:00
|
|
|
/* Shrinking succeeded, else we would be at "done". */
|
2010-05-16 16:48:46 +02:00
|
|
|
trans = btrfs_start_transaction(root, 0);
|
2011-01-20 07:19:37 +01:00
|
|
|
if (IS_ERR(trans)) {
|
|
|
|
ret = PTR_ERR(trans);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
2009-04-27 13:29:03 +02:00
|
|
|
lock_chunks(root);
|
|
|
|
|
|
|
|
device->disk_total_bytes = new_size;
|
|
|
|
/* Now btrfs_update_device() will change the on-disk size. */
|
|
|
|
ret = btrfs_update_device(trans, device);
|
|
|
|
if (ret) {
|
|
|
|
unlock_chunks(root);
|
|
|
|
btrfs_end_transaction(trans, root);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
WARN_ON(diff > old_total);
|
|
|
|
btrfs_set_super_total_bytes(super_copy, old_total - diff);
|
|
|
|
unlock_chunks(root);
|
|
|
|
btrfs_end_transaction(trans, root);
|
2008-04-25 22:53:30 +02:00
|
|
|
done:
|
|
|
|
btrfs_free_path(path);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-02 15:54:17 +01:00
|
|
|
static int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
|
2008-03-24 20:01:56 +01:00
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_key *key,
|
|
|
|
struct btrfs_chunk *chunk, int item_size)
|
|
|
|
{
|
|
|
|
struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
|
|
|
|
struct btrfs_disk_key disk_key;
|
|
|
|
u32 array_size;
|
|
|
|
u8 *ptr;
|
|
|
|
|
|
|
|
array_size = btrfs_super_sys_array_size(super_copy);
|
|
|
|
if (array_size + item_size > BTRFS_SYSTEM_CHUNK_ARRAY_SIZE)
|
|
|
|
return -EFBIG;
|
|
|
|
|
|
|
|
ptr = super_copy->sys_chunk_array + array_size;
|
|
|
|
btrfs_cpu_key_to_disk(&disk_key, key);
|
|
|
|
memcpy(ptr, &disk_key, sizeof(disk_key));
|
|
|
|
ptr += sizeof(disk_key);
|
|
|
|
memcpy(ptr, chunk, item_size);
|
|
|
|
item_size += sizeof(disk_key);
|
|
|
|
btrfs_set_super_sys_array_size(super_copy, array_size + item_size);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
/*
|
|
|
|
* sort the devices in descending order by max_avail, total_avail
|
|
|
|
*/
|
|
|
|
static int btrfs_cmp_device_info(const void *a, const void *b)
|
2008-04-18 16:29:51 +02:00
|
|
|
{
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
const struct btrfs_device_info *di_a = a;
|
|
|
|
const struct btrfs_device_info *di_b = b;
|
2008-04-18 16:29:51 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (di_a->max_avail > di_b->max_avail)
|
2011-01-05 11:07:28 +01:00
|
|
|
return -1;
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (di_a->max_avail < di_b->max_avail)
|
2011-01-05 11:07:28 +01:00
|
|
|
return 1;
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (di_a->total_avail > di_b->total_avail)
|
|
|
|
return -1;
|
|
|
|
if (di_a->total_avail < di_b->total_avail)
|
|
|
|
return 1;
|
|
|
|
return 0;
|
2011-01-05 11:07:28 +01:00
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
static int __btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *extent_root,
|
|
|
|
struct map_lookup **map_ret,
|
|
|
|
u64 *num_bytes_out, u64 *stripe_size_out,
|
|
|
|
u64 start, u64 type)
|
2011-01-05 11:07:28 +01:00
|
|
|
{
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
struct btrfs_fs_info *info = extent_root->fs_info;
|
|
|
|
struct btrfs_fs_devices *fs_devices = info->fs_devices;
|
|
|
|
struct list_head *cur;
|
|
|
|
struct map_lookup *map = NULL;
|
|
|
|
struct extent_map_tree *em_tree;
|
|
|
|
struct extent_map *em;
|
|
|
|
struct btrfs_device_info *devices_info = NULL;
|
|
|
|
u64 total_avail;
|
|
|
|
int num_stripes; /* total number of stripes to allocate */
|
|
|
|
int sub_stripes; /* sub_stripes info for map */
|
|
|
|
int dev_stripes; /* stripes per dev */
|
|
|
|
int devs_max; /* max devs to use */
|
|
|
|
int devs_min; /* min devs needed */
|
|
|
|
int devs_increment; /* ndevs has to be a multiple of this */
|
|
|
|
int ncopies; /* how many copies to data has */
|
|
|
|
int ret;
|
|
|
|
u64 max_stripe_size;
|
|
|
|
u64 max_chunk_size;
|
|
|
|
u64 stripe_size;
|
|
|
|
u64 num_bytes;
|
|
|
|
int ndevs;
|
|
|
|
int i;
|
|
|
|
int j;
|
2008-03-25 21:50:33 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if ((type & BTRFS_BLOCK_GROUP_RAID1) &&
|
|
|
|
(type & BTRFS_BLOCK_GROUP_DUP)) {
|
|
|
|
WARN_ON(1);
|
|
|
|
type &= ~BTRFS_BLOCK_GROUP_DUP;
|
2008-04-16 16:49:51 +02:00
|
|
|
}
|
2008-04-18 16:29:51 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (list_empty(&fs_devices->alloc_list))
|
|
|
|
return -ENOSPC;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
sub_stripes = 1;
|
|
|
|
dev_stripes = 1;
|
|
|
|
devs_increment = 1;
|
|
|
|
ncopies = 1;
|
|
|
|
devs_max = 0; /* 0 == as many as possible */
|
|
|
|
devs_min = 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* define the properties of each RAID type.
|
|
|
|
* FIXME: move this to a global table and use it in all RAID
|
|
|
|
* calculation code
|
|
|
|
*/
|
|
|
|
if (type & (BTRFS_BLOCK_GROUP_DUP)) {
|
|
|
|
dev_stripes = 2;
|
2011-01-05 11:07:28 +01:00
|
|
|
ncopies = 2;
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
devs_max = 1;
|
|
|
|
} else if (type & (BTRFS_BLOCK_GROUP_RAID0)) {
|
|
|
|
devs_min = 2;
|
|
|
|
} else if (type & (BTRFS_BLOCK_GROUP_RAID1)) {
|
|
|
|
devs_increment = 2;
|
2011-01-05 11:07:28 +01:00
|
|
|
ncopies = 2;
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
devs_max = 2;
|
|
|
|
devs_min = 2;
|
|
|
|
} else if (type & (BTRFS_BLOCK_GROUP_RAID10)) {
|
|
|
|
sub_stripes = 2;
|
|
|
|
devs_increment = 2;
|
|
|
|
ncopies = 2;
|
|
|
|
devs_min = 4;
|
|
|
|
} else {
|
|
|
|
devs_max = 1;
|
|
|
|
}
|
2011-01-05 11:07:28 +01:00
|
|
|
|
2008-04-18 16:29:51 +02:00
|
|
|
if (type & BTRFS_BLOCK_GROUP_DATA) {
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
max_stripe_size = 1024 * 1024 * 1024;
|
|
|
|
max_chunk_size = 10 * max_stripe_size;
|
2008-04-18 16:29:51 +02:00
|
|
|
} else if (type & BTRFS_BLOCK_GROUP_METADATA) {
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
max_stripe_size = 256 * 1024 * 1024;
|
|
|
|
max_chunk_size = max_stripe_size;
|
2008-04-18 17:55:51 +02:00
|
|
|
} else if (type & BTRFS_BLOCK_GROUP_SYSTEM) {
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
max_stripe_size = 8 * 1024 * 1024;
|
|
|
|
max_chunk_size = 2 * max_stripe_size;
|
|
|
|
} else {
|
|
|
|
printk(KERN_ERR "btrfs: invalid chunk type 0x%llx requested\n",
|
|
|
|
type);
|
|
|
|
BUG_ON(1);
|
2008-04-18 16:29:51 +02:00
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
/* we don't want a chunk larger than 10% of writeable space */
|
|
|
|
max_chunk_size = min(div_factor(fs_devices->total_rw_bytes, 1),
|
|
|
|
max_chunk_size);
|
2008-04-18 16:29:51 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
devices_info = kzalloc(sizeof(*devices_info) * fs_devices->rw_devices,
|
|
|
|
GFP_NOFS);
|
|
|
|
if (!devices_info)
|
|
|
|
return -ENOMEM;
|
2010-03-17 21:45:56 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
cur = fs_devices->alloc_list.next;
|
2008-04-18 16:29:51 +02:00
|
|
|
|
2010-04-06 15:37:47 +02:00
|
|
|
/*
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
* in the first pass through the devices list, we gather information
|
|
|
|
* about the available holes on each device.
|
2010-04-06 15:37:47 +02:00
|
|
|
*/
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
ndevs = 0;
|
|
|
|
while (cur != &fs_devices->alloc_list) {
|
|
|
|
struct btrfs_device *device;
|
|
|
|
u64 max_avail;
|
|
|
|
u64 dev_offset;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
device = list_entry(cur, struct btrfs_device, dev_alloc_list);
|
2010-04-06 15:37:47 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
cur = cur->next;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (!device->writeable) {
|
|
|
|
printk(KERN_ERR
|
|
|
|
"btrfs: read-only device in alloc_list\n");
|
|
|
|
WARN_ON(1);
|
|
|
|
continue;
|
|
|
|
}
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (!device->in_fs_metadata)
|
|
|
|
continue;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (device->total_bytes > device->bytes_used)
|
|
|
|
total_avail = device->total_bytes - device->bytes_used;
|
|
|
|
else
|
|
|
|
total_avail = 0;
|
Btrfs: fix a bug of balance on full multi-disk partitions
When balancing, we'll first try to shrink devices for some space,
but if it is working on a full multi-disk partition with raid protection,
we may encounter a bug, that is, while shrinking, total_bytes may be less
than bytes_used, and btrfs may allocate a dev extent that accesses out of
device's bounds.
Then we will not be able to write or read the data which stores at the end
of the device, and get the followings:
device fsid 0939f071-7ea3-46c8-95df-f176d773bfb6 devid 1 transid 10 /dev/sdb5
Btrfs detected SSD devices, enabling SSD mode
btrfs: relocating block group 476315648 flags 9
btrfs: found 4 extents
attempt to access beyond end of device
sdb5: rw=145, want=546176, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546304, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546432, limit=546147
attempt to access beyond end of device
sdb5: rw=145, want=546560, limit=546147
attempt to access beyond end of device
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-08-02 04:39:03 +02:00
|
|
|
|
|
|
|
/* If there is no space on this device, skip it. */
|
|
|
|
if (total_avail == 0)
|
|
|
|
continue;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
ret = find_free_dev_extent(trans, device,
|
|
|
|
max_stripe_size * dev_stripes,
|
|
|
|
&dev_offset, &max_avail);
|
|
|
|
if (ret && ret != -ENOSPC)
|
|
|
|
goto error;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (ret == 0)
|
|
|
|
max_avail = max_stripe_size * dev_stripes;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (max_avail < BTRFS_STRIPE_LEN * dev_stripes)
|
|
|
|
continue;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
devices_info[ndevs].dev_offset = dev_offset;
|
|
|
|
devices_info[ndevs].max_avail = max_avail;
|
|
|
|
devices_info[ndevs].total_avail = total_avail;
|
|
|
|
devices_info[ndevs].dev = device;
|
|
|
|
++ndevs;
|
|
|
|
}
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
/*
|
|
|
|
* now sort the devices by hole size / available space
|
|
|
|
*/
|
|
|
|
sort(devices_info, ndevs, sizeof(struct btrfs_device_info),
|
|
|
|
btrfs_cmp_device_info, NULL);
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
/* round down to number of usable stripes */
|
|
|
|
ndevs -= ndevs % devs_increment;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (ndevs < devs_increment * sub_stripes || ndevs < devs_min) {
|
|
|
|
ret = -ENOSPC;
|
|
|
|
goto error;
|
2011-01-05 11:07:28 +01:00
|
|
|
}
|
2010-04-06 15:37:47 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (devs_max && ndevs > devs_max)
|
|
|
|
ndevs = devs_max;
|
|
|
|
/*
|
|
|
|
* the primary goal is to maximize the number of stripes, so use as many
|
|
|
|
* devices as possible, even if the stripes are not maximum sized.
|
|
|
|
*/
|
|
|
|
stripe_size = devices_info[ndevs-1].max_avail;
|
|
|
|
num_stripes = ndevs * dev_stripes;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
if (stripe_size * num_stripes > max_chunk_size * ncopies) {
|
|
|
|
stripe_size = max_chunk_size * ncopies;
|
|
|
|
do_div(stripe_size, num_stripes);
|
2011-01-05 11:07:28 +01:00
|
|
|
}
|
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
do_div(stripe_size, dev_stripes);
|
|
|
|
do_div(stripe_size, BTRFS_STRIPE_LEN);
|
|
|
|
stripe_size *= BTRFS_STRIPE_LEN;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
|
|
|
map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
|
|
|
|
if (!map) {
|
|
|
|
ret = -ENOMEM;
|
|
|
|
goto error;
|
|
|
|
}
|
|
|
|
map->num_stripes = num_stripes;
|
2008-04-18 16:29:51 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
for (i = 0; i < ndevs; ++i) {
|
|
|
|
for (j = 0; j < dev_stripes; ++j) {
|
|
|
|
int s = i * dev_stripes + j;
|
|
|
|
map->stripes[s].dev = devices_info[i].dev;
|
|
|
|
map->stripes[s].physical = devices_info[i].dev_offset +
|
|
|
|
j * stripe_size;
|
2008-03-24 20:01:59 +01:00
|
|
|
}
|
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
map->sector_size = extent_root->sectorsize;
|
2011-01-05 11:07:28 +01:00
|
|
|
map->stripe_len = BTRFS_STRIPE_LEN;
|
|
|
|
map->io_align = BTRFS_STRIPE_LEN;
|
|
|
|
map->io_width = BTRFS_STRIPE_LEN;
|
2008-11-18 03:11:30 +01:00
|
|
|
map->type = type;
|
|
|
|
map->sub_stripes = sub_stripes;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
*map_ret = map;
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
num_bytes = stripe_size * (num_stripes / ncopies);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
*stripe_size_out = stripe_size;
|
|
|
|
*num_bytes_out = num_bytes;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
trace_btrfs_chunk_alloc(info->chunk_root, map, start, num_bytes);
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 12:18:59 +01:00
|
|
|
|
2011-04-21 00:48:27 +02:00
|
|
|
em = alloc_extent_map();
|
2008-11-18 03:11:30 +01:00
|
|
|
if (!em) {
|
2011-01-05 11:07:28 +01:00
|
|
|
ret = -ENOMEM;
|
|
|
|
goto error;
|
2008-03-25 21:50:33 +01:00
|
|
|
}
|
2008-11-18 03:11:30 +01:00
|
|
|
em->bdev = (struct block_device *)map;
|
|
|
|
em->start = start;
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
em->len = num_bytes;
|
2008-11-18 03:11:30 +01:00
|
|
|
em->block_start = 0;
|
|
|
|
em->block_len = em->len;
|
2008-03-25 21:50:33 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
em_tree = &extent_root->fs_info->mapping_tree.map_tree;
|
2009-09-02 22:24:52 +02:00
|
|
|
write_lock(&em_tree->lock);
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = add_extent_mapping(em_tree, em);
|
2009-09-02 22:24:52 +02:00
|
|
|
write_unlock(&em_tree->lock);
|
2008-11-18 03:11:30 +01:00
|
|
|
BUG_ON(ret);
|
|
|
|
free_extent_map(em);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = btrfs_make_block_group(trans, extent_root, 0, type,
|
|
|
|
BTRFS_FIRST_CHUNK_TREE_OBJECTID,
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
start, num_bytes);
|
2008-11-18 03:11:30 +01:00
|
|
|
BUG_ON(ret);
|
2008-04-03 22:29:03 +02:00
|
|
|
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
for (i = 0; i < map->num_stripes; ++i) {
|
|
|
|
struct btrfs_device *device;
|
|
|
|
u64 dev_offset;
|
|
|
|
|
|
|
|
device = map->stripes[i].dev;
|
|
|
|
dev_offset = map->stripes[i].physical;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
ret = btrfs_alloc_dev_extent(trans, device,
|
2008-11-18 03:11:30 +01:00
|
|
|
info->chunk_root->root_key.objectid,
|
|
|
|
BTRFS_FIRST_CHUNK_TREE_OBJECTID,
|
btrfs: quasi-round-robin for chunk allocation
In a multi device setup, the chunk allocator currently always allocates
chunks on the devices in the same order. This leads to a very uneven
distribution, especially with RAID1 or RAID10 and an uneven number of
devices.
This patch always sorts the devices before allocating, and allocates the
stripes on the devices with the most available space, as long as there
is enough space available. In a low space situation, it first tries to
maximize striping.
The patch also simplifies the allocator and reduces the checks for
corner cases.
The simplification is done by several means. First, it defines the
properties of each RAID type upfront. These properties are used afterwards
instead of differentiating cases in several places.
Second, the old allocator defined a minimum stripe size for each block
group type, tried to find a large enough chunk, and if this fails just
allocates a smaller one. This is now done in one step. The largest possible
chunk (up to max_chunk_size) is searched and allocated.
Because we now have only one pass, the allocation of the map (struct
map_lookup) is moved down to the point where the number of stripes is
already known. This way we avoid reallocation of the map.
We still avoid allocating stripes that are not a multiple of STRIPE_SIZE.
2011-04-12 12:07:57 +02:00
|
|
|
start, dev_offset, stripe_size);
|
2008-03-24 20:01:56 +01:00
|
|
|
BUG_ON(ret);
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
|
|
|
|
2011-01-05 11:07:28 +01:00
|
|
|
kfree(devices_info);
|
2008-11-18 03:11:30 +01:00
|
|
|
return 0;
|
2011-01-05 11:07:28 +01:00
|
|
|
|
|
|
|
error:
|
|
|
|
kfree(map);
|
|
|
|
kfree(devices_info);
|
|
|
|
return ret;
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static int __finish_chunk_alloc(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *extent_root,
|
|
|
|
struct map_lookup *map, u64 chunk_offset,
|
|
|
|
u64 chunk_size, u64 stripe_size)
|
|
|
|
{
|
|
|
|
u64 dev_offset;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
|
|
|
|
struct btrfs_device *device;
|
|
|
|
struct btrfs_chunk *chunk;
|
|
|
|
struct btrfs_stripe *stripe;
|
|
|
|
size_t item_size = btrfs_chunk_item_size(map->num_stripes);
|
|
|
|
int index = 0;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
chunk = kzalloc(item_size, GFP_NOFS);
|
|
|
|
if (!chunk)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
index = 0;
|
|
|
|
while (index < map->num_stripes) {
|
|
|
|
device = map->stripes[index].dev;
|
|
|
|
device->bytes_used += stripe_size;
|
2008-03-24 20:01:56 +01:00
|
|
|
ret = btrfs_update_device(trans, device);
|
|
|
|
BUG_ON(ret);
|
2008-11-18 03:11:30 +01:00
|
|
|
index++;
|
|
|
|
}
|
|
|
|
|
2011-09-26 23:12:22 +02:00
|
|
|
spin_lock(&extent_root->fs_info->free_chunk_lock);
|
|
|
|
extent_root->fs_info->free_chunk_space -= (stripe_size *
|
|
|
|
map->num_stripes);
|
|
|
|
spin_unlock(&extent_root->fs_info->free_chunk_lock);
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
index = 0;
|
|
|
|
stripe = &chunk->stripe;
|
|
|
|
while (index < map->num_stripes) {
|
|
|
|
device = map->stripes[index].dev;
|
|
|
|
dev_offset = map->stripes[index].physical;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-04-15 21:41:47 +02:00
|
|
|
btrfs_set_stack_stripe_devid(stripe, device->devid);
|
|
|
|
btrfs_set_stack_stripe_offset(stripe, dev_offset);
|
|
|
|
memcpy(stripe->dev_uuid, device->uuid, BTRFS_UUID_SIZE);
|
2008-11-18 03:11:30 +01:00
|
|
|
stripe++;
|
2008-03-24 20:01:56 +01:00
|
|
|
index++;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_set_stack_chunk_length(chunk, chunk_size);
|
2008-03-24 20:01:56 +01:00
|
|
|
btrfs_set_stack_chunk_owner(chunk, extent_root->root_key.objectid);
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_set_stack_chunk_stripe_len(chunk, map->stripe_len);
|
|
|
|
btrfs_set_stack_chunk_type(chunk, map->type);
|
|
|
|
btrfs_set_stack_chunk_num_stripes(chunk, map->num_stripes);
|
|
|
|
btrfs_set_stack_chunk_io_align(chunk, map->stripe_len);
|
|
|
|
btrfs_set_stack_chunk_io_width(chunk, map->stripe_len);
|
2008-03-24 20:01:56 +01:00
|
|
|
btrfs_set_stack_chunk_sector_size(chunk, extent_root->sectorsize);
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_set_stack_chunk_sub_stripes(chunk, map->sub_stripes);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
key.objectid = BTRFS_FIRST_CHUNK_TREE_OBJECTID;
|
|
|
|
key.type = BTRFS_CHUNK_ITEM_KEY;
|
|
|
|
key.offset = chunk_offset;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = btrfs_insert_item(trans, chunk_root, &key, chunk, item_size);
|
|
|
|
BUG_ON(ret);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_SYSTEM) {
|
|
|
|
ret = btrfs_add_system_chunk(trans, chunk_root, &key, chunk,
|
|
|
|
item_size);
|
2008-04-25 22:53:30 +02:00
|
|
|
BUG_ON(ret);
|
|
|
|
}
|
Btrfs: add initial tracepoint support for btrfs
Tracepoints can provide insight into why btrfs hits bugs and be greatly
helpful for debugging, e.g
dd-7822 [000] 2121.641088: btrfs_inode_request: root = 5(FS_TREE), gen = 4, ino = 256, blocks = 8, disk_i_size = 0, last_trans = 8, logged_trans = 0
dd-7822 [000] 2121.641100: btrfs_inode_new: root = 5(FS_TREE), gen = 8, ino = 257, blocks = 0, disk_i_size = 0, last_trans = 0, logged_trans = 0
btrfs-transacti-7804 [001] 2146.935420: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29368320 (orig_level = 0), cow_buf = 29388800 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.935473: btrfs_cow_block: root = 1(ROOT_TREE), refs = 2, orig_buf = 29364224 (orig_level = 0), cow_buf = 29392896 (cow_level = 0)
btrfs-transacti-7804 [001] 2146.972221: btrfs_transaction_commit: root = 1(ROOT_TREE), gen = 8
flush-btrfs-2-7821 [001] 2155.824210: btrfs_chunk_alloc: root = 3(CHUNK_TREE), offset = 1103101952, size = 1073741824, num_stripes = 1, sub_stripes = 0, type = DATA
flush-btrfs-2-7821 [001] 2155.824241: btrfs_cow_block: root = 2(EXTENT_TREE), refs = 2, orig_buf = 29388800 (orig_level = 0), cow_buf = 29396992 (cow_level = 0)
flush-btrfs-2-7821 [001] 2155.824255: btrfs_cow_block: root = 4(DEV_TREE), refs = 2, orig_buf = 29372416 (orig_level = 0), cow_buf = 29401088 (cow_level = 0)
flush-btrfs-2-7821 [000] 2155.824329: btrfs_cow_block: root = 3(CHUNK_TREE), refs = 2, orig_buf = 20971520 (orig_level = 0), cow_buf = 20975616 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898019: btrfs_cow_block: root = 5(FS_TREE), refs = 2, orig_buf = 29384704 (orig_level = 0), cow_buf = 29405184 (cow_level = 0)
btrfs-endio-wri-7800 [001] 2155.898043: btrfs_cow_block: root = 7(CSUM_TREE), refs = 2, orig_buf = 29376512 (orig_level = 0), cow_buf = 29409280 (cow_level = 0)
Here is what I have added:
1) ordere_extent:
btrfs_ordered_extent_add
btrfs_ordered_extent_remove
btrfs_ordered_extent_start
btrfs_ordered_extent_put
These provide critical information to understand how ordered_extents are
updated.
2) extent_map:
btrfs_get_extent
extent_map is used in both read and write cases, and it is useful for tracking
how btrfs specific IO is running.
3) writepage:
__extent_writepage
btrfs_writepage_end_io_hook
Pages are cirtical resourses and produce a lot of corner cases during writeback,
so it is valuable to know how page is written to disk.
4) inode:
btrfs_inode_new
btrfs_inode_request
btrfs_inode_evict
These can show where and when a inode is created, when a inode is evicted.
5) sync:
btrfs_sync_file
btrfs_sync_fs
These show sync arguments.
6) transaction:
btrfs_transaction_commit
In transaction based filesystem, it will be useful to know the generation and
who does commit.
7) back reference and cow:
btrfs_delayed_tree_ref
btrfs_delayed_data_ref
btrfs_delayed_ref_head
btrfs_cow_block
Btrfs natively supports back references, these tracepoints are helpful on
understanding btrfs's COW mechanism.
8) chunk:
btrfs_chunk_alloc
btrfs_chunk_free
Chunk is a link between physical offset and logical offset, and stands for space
infomation in btrfs, and these are helpful on tracing space things.
9) reserved_extent:
btrfs_reserved_extent_alloc
btrfs_reserved_extent_free
These can show how btrfs uses its space.
Signed-off-by: Liu Bo <liubo2009@cn.fujitsu.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2011-03-24 12:18:59 +01:00
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
kfree(chunk);
|
2008-11-18 03:11:30 +01:00
|
|
|
return 0;
|
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
/*
|
|
|
|
* Chunk allocation falls into two parts. The first part does works
|
|
|
|
* that make the new allocated chunk useable, but not do any operation
|
|
|
|
* that modifies the chunk tree. The second part does the works that
|
|
|
|
* require modifying the chunk tree. This division is important for the
|
|
|
|
* bootstrap process of adding storage to a seed btrfs.
|
|
|
|
*/
|
|
|
|
int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
|
|
|
|
struct btrfs_root *extent_root, u64 type)
|
|
|
|
{
|
|
|
|
u64 chunk_offset;
|
|
|
|
u64 chunk_size;
|
|
|
|
u64 stripe_size;
|
|
|
|
struct map_lookup *map;
|
|
|
|
struct btrfs_root *chunk_root = extent_root->fs_info->chunk_root;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = find_next_chunk(chunk_root, BTRFS_FIRST_CHUNK_TREE_OBJECTID,
|
|
|
|
&chunk_offset);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size,
|
|
|
|
&stripe_size, chunk_offset, type);
|
|
|
|
if (ret)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
ret = __finish_chunk_alloc(trans, extent_root, map, chunk_offset,
|
|
|
|
chunk_size, stripe_size);
|
|
|
|
BUG_ON(ret);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
static noinline int init_first_rw_device(struct btrfs_trans_handle *trans,
|
2008-11-18 03:11:30 +01:00
|
|
|
struct btrfs_root *root,
|
|
|
|
struct btrfs_device *device)
|
|
|
|
{
|
|
|
|
u64 chunk_offset;
|
|
|
|
u64 sys_chunk_offset;
|
|
|
|
u64 chunk_size;
|
|
|
|
u64 sys_chunk_size;
|
|
|
|
u64 stripe_size;
|
|
|
|
u64 sys_stripe_size;
|
|
|
|
u64 alloc_profile;
|
|
|
|
struct map_lookup *map;
|
|
|
|
struct map_lookup *sys_map;
|
|
|
|
struct btrfs_fs_info *fs_info = root->fs_info;
|
|
|
|
struct btrfs_root *extent_root = fs_info->extent_root;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = find_next_chunk(fs_info->chunk_root,
|
|
|
|
BTRFS_FIRST_CHUNK_TREE_OBJECTID, &chunk_offset);
|
2011-07-12 19:57:59 +02:00
|
|
|
if (ret)
|
|
|
|
return ret;
|
2008-11-18 03:11:30 +01:00
|
|
|
|
|
|
|
alloc_profile = BTRFS_BLOCK_GROUP_METADATA |
|
|
|
|
(fs_info->metadata_alloc_profile &
|
|
|
|
fs_info->avail_metadata_alloc_bits);
|
|
|
|
alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile);
|
|
|
|
|
|
|
|
ret = __btrfs_alloc_chunk(trans, extent_root, &map, &chunk_size,
|
|
|
|
&stripe_size, chunk_offset, alloc_profile);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
sys_chunk_offset = chunk_offset + chunk_size;
|
|
|
|
|
|
|
|
alloc_profile = BTRFS_BLOCK_GROUP_SYSTEM |
|
|
|
|
(fs_info->system_alloc_profile &
|
|
|
|
fs_info->avail_system_alloc_bits);
|
|
|
|
alloc_profile = btrfs_reduce_alloc_profile(root, alloc_profile);
|
|
|
|
|
|
|
|
ret = __btrfs_alloc_chunk(trans, extent_root, &sys_map,
|
|
|
|
&sys_chunk_size, &sys_stripe_size,
|
|
|
|
sys_chunk_offset, alloc_profile);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
ret = btrfs_add_device(trans, fs_info->chunk_root, device);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Modifying chunk tree needs allocating new blocks from both
|
|
|
|
* system block group and metadata block group. So we only can
|
|
|
|
* do operations require modifying the chunk tree after both
|
|
|
|
* block groups were created.
|
|
|
|
*/
|
|
|
|
ret = __finish_chunk_alloc(trans, extent_root, map, chunk_offset,
|
|
|
|
chunk_size, stripe_size);
|
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
ret = __finish_chunk_alloc(trans, extent_root, sys_map,
|
|
|
|
sys_chunk_offset, sys_chunk_size,
|
|
|
|
sys_stripe_size);
|
2008-04-14 15:48:18 +02:00
|
|
|
BUG_ON(ret);
|
2008-11-18 03:11:30 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_chunk_readonly(struct btrfs_root *root, u64 chunk_offset)
|
|
|
|
{
|
|
|
|
struct extent_map *em;
|
|
|
|
struct map_lookup *map;
|
|
|
|
struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
|
|
|
|
int readonly = 0;
|
|
|
|
int i;
|
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
read_lock(&map_tree->map_tree.lock);
|
2008-11-18 03:11:30 +01:00
|
|
|
em = lookup_extent_mapping(&map_tree->map_tree, chunk_offset, 1);
|
2009-09-02 22:24:52 +02:00
|
|
|
read_unlock(&map_tree->map_tree.lock);
|
2008-11-18 03:11:30 +01:00
|
|
|
if (!em)
|
|
|
|
return 1;
|
|
|
|
|
2010-01-27 03:07:59 +01:00
|
|
|
if (btrfs_test_opt(root, DEGRADED)) {
|
|
|
|
free_extent_map(em);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
map = (struct map_lookup *)em->bdev;
|
|
|
|
for (i = 0; i < map->num_stripes; i++) {
|
|
|
|
if (!map->stripes[i].dev->writeable) {
|
|
|
|
readonly = 1;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
free_extent_map(em);
|
2008-11-18 03:11:30 +01:00
|
|
|
return readonly;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_mapping_init(struct btrfs_mapping_tree *tree)
|
|
|
|
{
|
2011-04-21 00:34:43 +02:00
|
|
|
extent_map_tree_init(&tree->map_tree);
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void btrfs_mapping_tree_free(struct btrfs_mapping_tree *tree)
|
|
|
|
{
|
|
|
|
struct extent_map *em;
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
while (1) {
|
2009-09-02 22:24:52 +02:00
|
|
|
write_lock(&tree->map_tree.lock);
|
2008-03-24 20:01:56 +01:00
|
|
|
em = lookup_extent_mapping(&tree->map_tree, 0, (u64)-1);
|
|
|
|
if (em)
|
|
|
|
remove_extent_mapping(&tree->map_tree, em);
|
2009-09-02 22:24:52 +02:00
|
|
|
write_unlock(&tree->map_tree.lock);
|
2008-03-24 20:01:56 +01:00
|
|
|
if (!em)
|
|
|
|
break;
|
|
|
|
kfree(em->bdev);
|
|
|
|
/* once for us */
|
|
|
|
free_extent_map(em);
|
|
|
|
/* once for the tree */
|
|
|
|
free_extent_map(em);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-04-09 22:28:12 +02:00
|
|
|
int btrfs_num_copies(struct btrfs_mapping_tree *map_tree, u64 logical, u64 len)
|
|
|
|
{
|
|
|
|
struct extent_map *em;
|
|
|
|
struct map_lookup *map;
|
|
|
|
struct extent_map_tree *em_tree = &map_tree->map_tree;
|
|
|
|
int ret;
|
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
read_lock(&em_tree->lock);
|
2008-04-09 22:28:12 +02:00
|
|
|
em = lookup_extent_mapping(em_tree, logical, len);
|
2009-09-02 22:24:52 +02:00
|
|
|
read_unlock(&em_tree->lock);
|
2008-04-09 22:28:12 +02:00
|
|
|
BUG_ON(!em);
|
|
|
|
|
|
|
|
BUG_ON(em->start > logical || em->start + em->len < logical);
|
|
|
|
map = (struct map_lookup *)em->bdev;
|
|
|
|
if (map->type & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID1))
|
|
|
|
ret = map->num_stripes;
|
2008-04-16 16:49:51 +02:00
|
|
|
else if (map->type & BTRFS_BLOCK_GROUP_RAID10)
|
|
|
|
ret = map->sub_stripes;
|
2008-04-09 22:28:12 +02:00
|
|
|
else
|
|
|
|
ret = 1;
|
|
|
|
free_extent_map(em);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
static int find_live_mirror(struct map_lookup *map, int first, int num,
|
|
|
|
int optimal)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
if (map->stripes[optimal].dev->bdev)
|
|
|
|
return optimal;
|
|
|
|
for (i = first; i < first + num; i++) {
|
|
|
|
if (map->stripes[i].dev->bdev)
|
|
|
|
return i;
|
|
|
|
}
|
|
|
|
/* we couldn't find one that doesn't fail. Just return something
|
|
|
|
* and the io error handling code will clean up eventually
|
|
|
|
*/
|
|
|
|
return optimal;
|
|
|
|
}
|
|
|
|
|
2008-04-21 16:03:05 +02:00
|
|
|
static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
|
|
|
|
u64 logical, u64 *length,
|
|
|
|
struct btrfs_multi_bio **multi_ret,
|
2011-03-10 08:52:07 +01:00
|
|
|
int mirror_num)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
struct extent_map *em;
|
|
|
|
struct map_lookup *map;
|
|
|
|
struct extent_map_tree *em_tree = &map_tree->map_tree;
|
|
|
|
u64 offset;
|
2008-03-25 21:50:33 +01:00
|
|
|
u64 stripe_offset;
|
2011-03-24 11:24:26 +01:00
|
|
|
u64 stripe_end_offset;
|
2008-03-25 21:50:33 +01:00
|
|
|
u64 stripe_nr;
|
2011-03-24 11:24:26 +01:00
|
|
|
u64 stripe_nr_orig;
|
|
|
|
u64 stripe_nr_end;
|
2008-04-09 22:28:12 +02:00
|
|
|
int stripes_allocated = 8;
|
2008-04-16 16:49:51 +02:00
|
|
|
int stripes_required = 1;
|
2008-03-25 21:50:33 +01:00
|
|
|
int stripe_index;
|
2008-04-09 22:28:12 +02:00
|
|
|
int i;
|
2008-04-21 16:03:05 +02:00
|
|
|
int num_stripes;
|
2008-04-29 15:38:00 +02:00
|
|
|
int max_errors = 0;
|
2008-04-09 22:28:12 +02:00
|
|
|
struct btrfs_multi_bio *multi = NULL;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2011-03-24 11:24:26 +01:00
|
|
|
if (multi_ret && !(rw & (REQ_WRITE | REQ_DISCARD)))
|
2008-04-09 22:28:12 +02:00
|
|
|
stripes_allocated = 1;
|
|
|
|
again:
|
|
|
|
if (multi_ret) {
|
|
|
|
multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
|
|
|
|
GFP_NOFS);
|
|
|
|
if (!multi)
|
|
|
|
return -ENOMEM;
|
2008-04-29 15:38:00 +02:00
|
|
|
|
|
|
|
atomic_set(&multi->error, 0);
|
2008-04-09 22:28:12 +02:00
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
read_lock(&em_tree->lock);
|
2008-03-24 20:01:56 +01:00
|
|
|
em = lookup_extent_mapping(em_tree, logical, *length);
|
2009-09-02 22:24:52 +02:00
|
|
|
read_unlock(&em_tree->lock);
|
2008-04-21 16:03:05 +02:00
|
|
|
|
2008-04-17 17:29:12 +02:00
|
|
|
if (!em) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_CRIT "unable to find logical %llu len %llu\n",
|
|
|
|
(unsigned long long)logical,
|
|
|
|
(unsigned long long)*length);
|
2008-04-21 16:03:05 +02:00
|
|
|
BUG();
|
2008-04-17 17:29:12 +02:00
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
BUG_ON(em->start > logical || em->start + em->len < logical);
|
|
|
|
map = (struct map_lookup *)em->bdev;
|
|
|
|
offset = logical - em->start;
|
2008-03-25 21:50:33 +01:00
|
|
|
|
2008-04-09 22:28:12 +02:00
|
|
|
if (mirror_num > map->num_stripes)
|
|
|
|
mirror_num = 0;
|
|
|
|
|
2008-04-09 22:28:12 +02:00
|
|
|
/* if our multi bio struct is too small, back off and try again */
|
2010-08-07 18:20:39 +02:00
|
|
|
if (rw & REQ_WRITE) {
|
2008-04-16 16:49:51 +02:00
|
|
|
if (map->type & (BTRFS_BLOCK_GROUP_RAID1 |
|
|
|
|
BTRFS_BLOCK_GROUP_DUP)) {
|
|
|
|
stripes_required = map->num_stripes;
|
2008-04-29 15:38:00 +02:00
|
|
|
max_errors = 1;
|
2008-04-16 16:49:51 +02:00
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
|
|
|
|
stripes_required = map->sub_stripes;
|
2008-04-29 15:38:00 +02:00
|
|
|
max_errors = 1;
|
2008-04-16 16:49:51 +02:00
|
|
|
}
|
|
|
|
}
|
2011-03-24 11:24:26 +01:00
|
|
|
if (rw & REQ_DISCARD) {
|
|
|
|
if (map->type & (BTRFS_BLOCK_GROUP_RAID0 |
|
|
|
|
BTRFS_BLOCK_GROUP_RAID1 |
|
|
|
|
BTRFS_BLOCK_GROUP_DUP |
|
|
|
|
BTRFS_BLOCK_GROUP_RAID10)) {
|
|
|
|
stripes_required = map->num_stripes;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (multi_ret && (rw & (REQ_WRITE | REQ_DISCARD)) &&
|
2008-04-16 16:49:51 +02:00
|
|
|
stripes_allocated < stripes_required) {
|
2008-04-09 22:28:12 +02:00
|
|
|
stripes_allocated = map->num_stripes;
|
|
|
|
free_extent_map(em);
|
|
|
|
kfree(multi);
|
|
|
|
goto again;
|
|
|
|
}
|
2008-03-25 21:50:33 +01:00
|
|
|
stripe_nr = offset;
|
|
|
|
/*
|
|
|
|
* stripe_nr counts the total number of stripes we have to stride
|
|
|
|
* to get to this block
|
|
|
|
*/
|
|
|
|
do_div(stripe_nr, map->stripe_len);
|
|
|
|
|
|
|
|
stripe_offset = stripe_nr * map->stripe_len;
|
|
|
|
BUG_ON(offset < stripe_offset);
|
|
|
|
|
|
|
|
/* stripe_offset is the offset of this block in its stripe*/
|
|
|
|
stripe_offset = offset - stripe_offset;
|
|
|
|
|
2011-03-24 11:24:26 +01:00
|
|
|
if (rw & REQ_DISCARD)
|
|
|
|
*length = min_t(u64, em->len - offset, *length);
|
|
|
|
else if (map->type & (BTRFS_BLOCK_GROUP_RAID0 |
|
|
|
|
BTRFS_BLOCK_GROUP_RAID1 |
|
|
|
|
BTRFS_BLOCK_GROUP_RAID10 |
|
|
|
|
BTRFS_BLOCK_GROUP_DUP)) {
|
2008-04-09 22:28:12 +02:00
|
|
|
/* we limit the length of each bio to what fits in a stripe */
|
|
|
|
*length = min_t(u64, em->len - offset,
|
2011-03-24 11:24:26 +01:00
|
|
|
map->stripe_len - stripe_offset);
|
2008-04-09 22:28:12 +02:00
|
|
|
} else {
|
|
|
|
*length = em->len - offset;
|
|
|
|
}
|
2008-04-21 16:03:05 +02:00
|
|
|
|
2011-03-10 08:52:07 +01:00
|
|
|
if (!multi_ret)
|
2008-04-09 22:28:12 +02:00
|
|
|
goto out;
|
|
|
|
|
2008-04-21 16:03:05 +02:00
|
|
|
num_stripes = 1;
|
2008-04-09 22:28:12 +02:00
|
|
|
stripe_index = 0;
|
2011-03-24 11:24:26 +01:00
|
|
|
stripe_nr_orig = stripe_nr;
|
|
|
|
stripe_nr_end = (offset + *length + map->stripe_len - 1) &
|
|
|
|
(~(map->stripe_len - 1));
|
|
|
|
do_div(stripe_nr_end, map->stripe_len);
|
|
|
|
stripe_end_offset = stripe_nr_end * map->stripe_len -
|
|
|
|
(offset + *length);
|
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
|
|
|
|
if (rw & REQ_DISCARD)
|
|
|
|
num_stripes = min_t(u64, map->num_stripes,
|
|
|
|
stripe_nr_end - stripe_nr_orig);
|
|
|
|
stripe_index = do_div(stripe_nr, map->num_stripes);
|
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
|
2011-03-29 00:31:05 +02:00
|
|
|
if (rw & (REQ_WRITE | REQ_DISCARD))
|
2008-04-21 16:03:05 +02:00
|
|
|
num_stripes = map->num_stripes;
|
2008-04-29 20:12:09 +02:00
|
|
|
else if (mirror_num)
|
2008-04-09 22:28:12 +02:00
|
|
|
stripe_index = mirror_num - 1;
|
2008-05-13 19:46:40 +02:00
|
|
|
else {
|
|
|
|
stripe_index = find_live_mirror(map, 0,
|
|
|
|
map->num_stripes,
|
|
|
|
current->pid % map->num_stripes);
|
|
|
|
}
|
2008-04-29 20:12:09 +02:00
|
|
|
|
2008-04-03 22:29:03 +02:00
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
|
2011-03-24 11:24:26 +01:00
|
|
|
if (rw & (REQ_WRITE | REQ_DISCARD))
|
2008-04-21 16:03:05 +02:00
|
|
|
num_stripes = map->num_stripes;
|
2008-04-09 22:28:12 +02:00
|
|
|
else if (mirror_num)
|
|
|
|
stripe_index = mirror_num - 1;
|
2008-04-29 20:12:09 +02:00
|
|
|
|
2008-04-16 16:49:51 +02:00
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
|
|
|
|
int factor = map->num_stripes / map->sub_stripes;
|
|
|
|
|
|
|
|
stripe_index = do_div(stripe_nr, factor);
|
|
|
|
stripe_index *= map->sub_stripes;
|
|
|
|
|
2011-03-10 08:52:07 +01:00
|
|
|
if (rw & REQ_WRITE)
|
2008-04-21 16:03:05 +02:00
|
|
|
num_stripes = map->sub_stripes;
|
2011-03-24 11:24:26 +01:00
|
|
|
else if (rw & REQ_DISCARD)
|
|
|
|
num_stripes = min_t(u64, map->sub_stripes *
|
|
|
|
(stripe_nr_end - stripe_nr_orig),
|
|
|
|
map->num_stripes);
|
2008-04-16 16:49:51 +02:00
|
|
|
else if (mirror_num)
|
|
|
|
stripe_index += mirror_num - 1;
|
2008-05-13 19:46:40 +02:00
|
|
|
else {
|
|
|
|
stripe_index = find_live_mirror(map, stripe_index,
|
|
|
|
map->sub_stripes, stripe_index +
|
|
|
|
current->pid % map->sub_stripes);
|
|
|
|
}
|
2008-04-03 22:29:03 +02:00
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* after this do_div call, stripe_nr is the number of stripes
|
|
|
|
* on this device we have to walk to find the data, and
|
|
|
|
* stripe_index is the number of our device in the stripe array
|
|
|
|
*/
|
|
|
|
stripe_index = do_div(stripe_nr, map->num_stripes);
|
|
|
|
}
|
2008-03-25 21:50:33 +01:00
|
|
|
BUG_ON(stripe_index >= map->num_stripes);
|
2008-04-09 22:28:12 +02:00
|
|
|
|
2011-03-24 11:24:26 +01:00
|
|
|
if (rw & REQ_DISCARD) {
|
|
|
|
for (i = 0; i < num_stripes; i++) {
|
2008-04-21 16:03:05 +02:00
|
|
|
multi->stripes[i].physical =
|
|
|
|
map->stripes[stripe_index].physical +
|
|
|
|
stripe_offset + stripe_nr * map->stripe_len;
|
|
|
|
multi->stripes[i].dev = map->stripes[stripe_index].dev;
|
2011-03-24 11:24:26 +01:00
|
|
|
|
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
|
|
|
|
u64 stripes;
|
2011-03-28 03:23:21 +02:00
|
|
|
u32 last_stripe = 0;
|
2011-03-24 11:24:26 +01:00
|
|
|
int j;
|
|
|
|
|
2011-03-28 03:23:21 +02:00
|
|
|
div_u64_rem(stripe_nr_end - 1,
|
|
|
|
map->num_stripes,
|
|
|
|
&last_stripe);
|
|
|
|
|
2011-03-24 11:24:26 +01:00
|
|
|
for (j = 0; j < map->num_stripes; j++) {
|
2011-03-28 03:23:21 +02:00
|
|
|
u32 test;
|
|
|
|
|
|
|
|
div_u64_rem(stripe_nr_end - 1 - j,
|
|
|
|
map->num_stripes, &test);
|
|
|
|
if (test == stripe_index)
|
2011-03-24 11:24:26 +01:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
stripes = stripe_nr_end - 1 - j;
|
|
|
|
do_div(stripes, map->num_stripes);
|
|
|
|
multi->stripes[i].length = map->stripe_len *
|
|
|
|
(stripes - stripe_nr + 1);
|
|
|
|
|
|
|
|
if (i == 0) {
|
|
|
|
multi->stripes[i].length -=
|
|
|
|
stripe_offset;
|
|
|
|
stripe_offset = 0;
|
|
|
|
}
|
|
|
|
if (stripe_index == last_stripe)
|
|
|
|
multi->stripes[i].length -=
|
|
|
|
stripe_end_offset;
|
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
|
|
|
|
u64 stripes;
|
|
|
|
int j;
|
|
|
|
int factor = map->num_stripes /
|
|
|
|
map->sub_stripes;
|
2011-03-28 03:23:21 +02:00
|
|
|
u32 last_stripe = 0;
|
|
|
|
|
|
|
|
div_u64_rem(stripe_nr_end - 1,
|
|
|
|
factor, &last_stripe);
|
2011-03-24 11:24:26 +01:00
|
|
|
last_stripe *= map->sub_stripes;
|
|
|
|
|
|
|
|
for (j = 0; j < factor; j++) {
|
2011-03-28 03:23:21 +02:00
|
|
|
u32 test;
|
|
|
|
|
|
|
|
div_u64_rem(stripe_nr_end - 1 - j,
|
|
|
|
factor, &test);
|
|
|
|
|
|
|
|
if (test ==
|
2011-03-24 11:24:26 +01:00
|
|
|
stripe_index / map->sub_stripes)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
stripes = stripe_nr_end - 1 - j;
|
|
|
|
do_div(stripes, factor);
|
|
|
|
multi->stripes[i].length = map->stripe_len *
|
|
|
|
(stripes - stripe_nr + 1);
|
|
|
|
|
|
|
|
if (i < map->sub_stripes) {
|
|
|
|
multi->stripes[i].length -=
|
|
|
|
stripe_offset;
|
|
|
|
if (i == map->sub_stripes - 1)
|
|
|
|
stripe_offset = 0;
|
|
|
|
}
|
|
|
|
if (stripe_index >= last_stripe &&
|
|
|
|
stripe_index <= (last_stripe +
|
|
|
|
map->sub_stripes - 1)) {
|
|
|
|
multi->stripes[i].length -=
|
|
|
|
stripe_end_offset;
|
|
|
|
}
|
|
|
|
} else
|
|
|
|
multi->stripes[i].length = *length;
|
|
|
|
|
|
|
|
stripe_index++;
|
|
|
|
if (stripe_index == map->num_stripes) {
|
|
|
|
/* This could only happen for RAID0/10 */
|
|
|
|
stripe_index = 0;
|
|
|
|
stripe_nr++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
for (i = 0; i < num_stripes; i++) {
|
2011-03-29 00:31:05 +02:00
|
|
|
multi->stripes[i].physical =
|
|
|
|
map->stripes[stripe_index].physical +
|
|
|
|
stripe_offset +
|
|
|
|
stripe_nr * map->stripe_len;
|
|
|
|
multi->stripes[i].dev =
|
|
|
|
map->stripes[stripe_index].dev;
|
2011-03-24 11:24:26 +01:00
|
|
|
stripe_index++;
|
2008-04-21 16:03:05 +02:00
|
|
|
}
|
2008-03-25 21:50:33 +01:00
|
|
|
}
|
2008-04-21 16:03:05 +02:00
|
|
|
if (multi_ret) {
|
|
|
|
*multi_ret = multi;
|
|
|
|
multi->num_stripes = num_stripes;
|
2008-04-29 15:38:00 +02:00
|
|
|
multi->max_errors = max_errors;
|
2008-04-21 16:03:05 +02:00
|
|
|
}
|
2008-04-09 22:28:12 +02:00
|
|
|
out:
|
2008-03-24 20:01:56 +01:00
|
|
|
free_extent_map(em);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-04-21 16:03:05 +02:00
|
|
|
int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
|
|
|
|
u64 logical, u64 *length,
|
|
|
|
struct btrfs_multi_bio **multi_ret, int mirror_num)
|
|
|
|
{
|
|
|
|
return __btrfs_map_block(map_tree, rw, logical, length, multi_ret,
|
2011-03-10 08:52:07 +01:00
|
|
|
mirror_num);
|
2008-04-21 16:03:05 +02:00
|
|
|
}
|
|
|
|
|
2008-12-08 22:46:26 +01:00
|
|
|
int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
|
|
|
|
u64 chunk_start, u64 physical, u64 devid,
|
|
|
|
u64 **logical, int *naddrs, int *stripe_len)
|
|
|
|
{
|
|
|
|
struct extent_map_tree *em_tree = &map_tree->map_tree;
|
|
|
|
struct extent_map *em;
|
|
|
|
struct map_lookup *map;
|
|
|
|
u64 *buf;
|
|
|
|
u64 bytenr;
|
|
|
|
u64 length;
|
|
|
|
u64 stripe_nr;
|
|
|
|
int i, j, nr = 0;
|
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
read_lock(&em_tree->lock);
|
2008-12-08 22:46:26 +01:00
|
|
|
em = lookup_extent_mapping(em_tree, chunk_start, 1);
|
2009-09-02 22:24:52 +02:00
|
|
|
read_unlock(&em_tree->lock);
|
2008-12-08 22:46:26 +01:00
|
|
|
|
|
|
|
BUG_ON(!em || em->start != chunk_start);
|
|
|
|
map = (struct map_lookup *)em->bdev;
|
|
|
|
|
|
|
|
length = em->len;
|
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_RAID10)
|
|
|
|
do_div(length, map->num_stripes / map->sub_stripes);
|
|
|
|
else if (map->type & BTRFS_BLOCK_GROUP_RAID0)
|
|
|
|
do_div(length, map->num_stripes);
|
|
|
|
|
|
|
|
buf = kzalloc(sizeof(u64) * map->num_stripes, GFP_NOFS);
|
|
|
|
BUG_ON(!buf);
|
|
|
|
|
|
|
|
for (i = 0; i < map->num_stripes; i++) {
|
|
|
|
if (devid && map->stripes[i].dev->devid != devid)
|
|
|
|
continue;
|
|
|
|
if (map->stripes[i].physical > physical ||
|
|
|
|
map->stripes[i].physical + length <= physical)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
stripe_nr = physical - map->stripes[i].physical;
|
|
|
|
do_div(stripe_nr, map->stripe_len);
|
|
|
|
|
|
|
|
if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
|
|
|
|
stripe_nr = stripe_nr * map->num_stripes + i;
|
|
|
|
do_div(stripe_nr, map->sub_stripes);
|
|
|
|
} else if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
|
|
|
|
stripe_nr = stripe_nr * map->num_stripes + i;
|
|
|
|
}
|
|
|
|
bytenr = chunk_start + stripe_nr * map->stripe_len;
|
2008-12-08 22:43:10 +01:00
|
|
|
WARN_ON(nr >= map->num_stripes);
|
2008-12-08 22:46:26 +01:00
|
|
|
for (j = 0; j < nr; j++) {
|
|
|
|
if (buf[j] == bytenr)
|
|
|
|
break;
|
|
|
|
}
|
2008-12-08 22:43:10 +01:00
|
|
|
if (j == nr) {
|
|
|
|
WARN_ON(nr >= map->num_stripes);
|
2008-12-08 22:46:26 +01:00
|
|
|
buf[nr++] = bytenr;
|
2008-12-08 22:43:10 +01:00
|
|
|
}
|
2008-12-08 22:46:26 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
*logical = buf;
|
|
|
|
*naddrs = nr;
|
|
|
|
*stripe_len = map->stripe_len;
|
|
|
|
|
|
|
|
free_extent_map(em);
|
|
|
|
return 0;
|
2008-04-21 16:03:05 +02:00
|
|
|
}
|
|
|
|
|
2008-04-03 22:29:03 +02:00
|
|
|
static void end_bio_multi_stripe(struct bio *bio, int err)
|
|
|
|
{
|
2008-04-09 22:28:12 +02:00
|
|
|
struct btrfs_multi_bio *multi = bio->bi_private;
|
2008-08-05 16:13:57 +02:00
|
|
|
int is_orig_bio = 0;
|
2008-04-03 22:29:03 +02:00
|
|
|
|
|
|
|
if (err)
|
2008-04-29 15:38:00 +02:00
|
|
|
atomic_inc(&multi->error);
|
2008-04-03 22:29:03 +02:00
|
|
|
|
2008-08-05 16:13:57 +02:00
|
|
|
if (bio == multi->orig_bio)
|
|
|
|
is_orig_bio = 1;
|
|
|
|
|
2008-04-09 22:28:12 +02:00
|
|
|
if (atomic_dec_and_test(&multi->stripes_pending)) {
|
2008-08-05 16:13:57 +02:00
|
|
|
if (!is_orig_bio) {
|
|
|
|
bio_put(bio);
|
|
|
|
bio = multi->orig_bio;
|
|
|
|
}
|
2008-04-03 22:29:03 +02:00
|
|
|
bio->bi_private = multi->private;
|
|
|
|
bio->bi_end_io = multi->end_io;
|
2008-04-29 15:38:00 +02:00
|
|
|
/* only send an error to the higher layers if it is
|
|
|
|
* beyond the tolerance of the multi-bio
|
|
|
|
*/
|
2008-05-12 19:39:03 +02:00
|
|
|
if (atomic_read(&multi->error) > multi->max_errors) {
|
2008-04-29 15:38:00 +02:00
|
|
|
err = -EIO;
|
2008-05-12 19:39:03 +02:00
|
|
|
} else if (err) {
|
|
|
|
/*
|
|
|
|
* this bio is actually up to date, we didn't
|
|
|
|
* go over the max number of errors
|
|
|
|
*/
|
|
|
|
set_bit(BIO_UPTODATE, &bio->bi_flags);
|
2008-04-29 15:38:00 +02:00
|
|
|
err = 0;
|
2008-05-12 19:39:03 +02:00
|
|
|
}
|
2008-04-03 22:29:03 +02:00
|
|
|
kfree(multi);
|
|
|
|
|
|
|
|
bio_endio(bio, err);
|
2008-08-05 16:13:57 +02:00
|
|
|
} else if (!is_orig_bio) {
|
2008-04-03 22:29:03 +02:00
|
|
|
bio_put(bio);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-06-11 22:50:36 +02:00
|
|
|
struct async_sched {
|
|
|
|
struct bio *bio;
|
|
|
|
int rw;
|
|
|
|
struct btrfs_fs_info *info;
|
|
|
|
struct btrfs_work work;
|
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* see run_scheduled_bios for a description of why bios are collected for
|
|
|
|
* async submit.
|
|
|
|
*
|
|
|
|
* This will add one bio to the pending list for a device and make sure
|
|
|
|
* the work struct is scheduled.
|
|
|
|
*/
|
2009-01-06 03:25:51 +01:00
|
|
|
static noinline int schedule_bio(struct btrfs_root *root,
|
2008-09-05 22:09:51 +02:00
|
|
|
struct btrfs_device *device,
|
|
|
|
int rw, struct bio *bio)
|
2008-06-11 22:50:36 +02:00
|
|
|
{
|
|
|
|
int should_queue = 1;
|
2009-04-20 21:50:09 +02:00
|
|
|
struct btrfs_pending_bios *pending_bios;
|
2008-06-11 22:50:36 +02:00
|
|
|
|
|
|
|
/* don't bother with additional async steps for reads, right now */
|
2010-08-07 18:20:39 +02:00
|
|
|
if (!(rw & REQ_WRITE)) {
|
2008-07-31 22:29:02 +02:00
|
|
|
bio_get(bio);
|
2008-06-11 22:50:36 +02:00
|
|
|
submit_bio(rw, bio);
|
2008-07-31 22:29:02 +02:00
|
|
|
bio_put(bio);
|
2008-06-11 22:50:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2008-08-15 21:34:15 +02:00
|
|
|
* nr_async_bios allows us to reliably return congestion to the
|
2008-06-11 22:50:36 +02:00
|
|
|
* higher layers. Otherwise, the async bio makes it appear we have
|
|
|
|
* made progress against dirty pages when we've really just put it
|
|
|
|
* on a queue for later
|
|
|
|
*/
|
2008-08-15 21:34:15 +02:00
|
|
|
atomic_inc(&root->fs_info->nr_async_bios);
|
2008-07-31 22:29:02 +02:00
|
|
|
WARN_ON(bio->bi_next);
|
2008-06-11 22:50:36 +02:00
|
|
|
bio->bi_next = NULL;
|
|
|
|
bio->bi_rw |= rw;
|
|
|
|
|
|
|
|
spin_lock(&device->io_lock);
|
2010-08-07 18:20:39 +02:00
|
|
|
if (bio->bi_rw & REQ_SYNC)
|
2009-04-20 21:50:09 +02:00
|
|
|
pending_bios = &device->pending_sync_bios;
|
|
|
|
else
|
|
|
|
pending_bios = &device->pending_bios;
|
2008-06-11 22:50:36 +02:00
|
|
|
|
2009-04-20 21:50:09 +02:00
|
|
|
if (pending_bios->tail)
|
|
|
|
pending_bios->tail->bi_next = bio;
|
2008-06-11 22:50:36 +02:00
|
|
|
|
2009-04-20 21:50:09 +02:00
|
|
|
pending_bios->tail = bio;
|
|
|
|
if (!pending_bios->head)
|
|
|
|
pending_bios->head = bio;
|
2008-06-11 22:50:36 +02:00
|
|
|
if (device->running_pending)
|
|
|
|
should_queue = 0;
|
|
|
|
|
|
|
|
spin_unlock(&device->io_lock);
|
|
|
|
|
|
|
|
if (should_queue)
|
2008-06-12 20:46:17 +02:00
|
|
|
btrfs_queue_worker(&root->fs_info->submit_workers,
|
|
|
|
&device->work);
|
2008-06-11 22:50:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-04-09 22:28:12 +02:00
|
|
|
int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
|
2008-06-11 22:50:36 +02:00
|
|
|
int mirror_num, int async_submit)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
struct btrfs_mapping_tree *map_tree;
|
|
|
|
struct btrfs_device *dev;
|
2008-04-03 22:29:03 +02:00
|
|
|
struct bio *first_bio = bio;
|
2008-10-03 22:31:08 +02:00
|
|
|
u64 logical = (u64)bio->bi_sector << 9;
|
2008-03-24 20:01:56 +01:00
|
|
|
u64 length = 0;
|
|
|
|
u64 map_length;
|
2008-04-09 22:28:12 +02:00
|
|
|
struct btrfs_multi_bio *multi = NULL;
|
2008-03-24 20:01:56 +01:00
|
|
|
int ret;
|
2008-04-03 22:29:03 +02:00
|
|
|
int dev_nr = 0;
|
|
|
|
int total_devs = 1;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-04-21 16:03:05 +02:00
|
|
|
length = bio->bi_size;
|
2008-03-24 20:01:56 +01:00
|
|
|
map_tree = &root->fs_info->mapping_tree;
|
|
|
|
map_length = length;
|
2008-04-09 22:28:12 +02:00
|
|
|
|
2008-04-09 22:28:12 +02:00
|
|
|
ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi,
|
|
|
|
mirror_num);
|
2008-04-09 22:28:12 +02:00
|
|
|
BUG_ON(ret);
|
|
|
|
|
|
|
|
total_devs = multi->num_stripes;
|
|
|
|
if (map_length < length) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_CRIT "mapping failed logical %llu bio len %llu "
|
|
|
|
"len %llu\n", (unsigned long long)logical,
|
|
|
|
(unsigned long long)length,
|
|
|
|
(unsigned long long)map_length);
|
2008-04-09 22:28:12 +02:00
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
multi->end_io = first_bio->bi_end_io;
|
|
|
|
multi->private = first_bio->bi_private;
|
2008-08-05 16:13:57 +02:00
|
|
|
multi->orig_bio = first_bio;
|
2008-04-09 22:28:12 +02:00
|
|
|
atomic_set(&multi->stripes_pending, multi->num_stripes);
|
|
|
|
|
2009-01-06 03:25:51 +01:00
|
|
|
while (dev_nr < total_devs) {
|
2008-04-03 22:29:03 +02:00
|
|
|
if (total_devs > 1) {
|
|
|
|
if (dev_nr < total_devs - 1) {
|
|
|
|
bio = bio_clone(first_bio, GFP_NOFS);
|
|
|
|
BUG_ON(!bio);
|
|
|
|
} else {
|
|
|
|
bio = first_bio;
|
|
|
|
}
|
|
|
|
bio->bi_private = multi;
|
|
|
|
bio->bi_end_io = end_bio_multi_stripe;
|
|
|
|
}
|
2008-04-09 22:28:12 +02:00
|
|
|
bio->bi_sector = multi->stripes[dev_nr].physical >> 9;
|
|
|
|
dev = multi->stripes[dev_nr].dev;
|
2010-10-28 21:30:42 +02:00
|
|
|
if (dev && dev->bdev && (rw != WRITE || dev->writeable)) {
|
2008-05-13 19:46:40 +02:00
|
|
|
bio->bi_bdev = dev->bdev;
|
2008-06-11 22:50:36 +02:00
|
|
|
if (async_submit)
|
|
|
|
schedule_bio(root, dev, rw, bio);
|
|
|
|
else
|
|
|
|
submit_bio(rw, bio);
|
2008-05-13 19:46:40 +02:00
|
|
|
} else {
|
|
|
|
bio->bi_bdev = root->fs_info->fs_devices->latest_bdev;
|
|
|
|
bio->bi_sector = logical >> 9;
|
|
|
|
bio_endio(bio, -EIO);
|
|
|
|
}
|
2008-04-03 22:29:03 +02:00
|
|
|
dev_nr++;
|
|
|
|
}
|
2008-04-09 22:28:12 +02:00
|
|
|
if (total_devs == 1)
|
|
|
|
kfree(multi);
|
2008-03-24 20:01:56 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-04-18 16:29:38 +02:00
|
|
|
struct btrfs_device *btrfs_find_device(struct btrfs_root *root, u64 devid,
|
2008-11-18 03:11:30 +01:00
|
|
|
u8 *uuid, u8 *fsid)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
2008-11-18 03:11:30 +01:00
|
|
|
struct btrfs_device *device;
|
|
|
|
struct btrfs_fs_devices *cur_devices;
|
|
|
|
|
|
|
|
cur_devices = root->fs_info->fs_devices;
|
|
|
|
while (cur_devices) {
|
|
|
|
if (!fsid ||
|
|
|
|
!memcmp(cur_devices->fsid, fsid, BTRFS_UUID_SIZE)) {
|
|
|
|
device = __find_device(&cur_devices->devices,
|
|
|
|
devid, uuid);
|
|
|
|
if (device)
|
|
|
|
return device;
|
|
|
|
}
|
|
|
|
cur_devices = cur_devices->seed;
|
|
|
|
}
|
|
|
|
return NULL;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
|
2008-05-13 19:46:40 +02:00
|
|
|
static struct btrfs_device *add_missing_dev(struct btrfs_root *root,
|
|
|
|
u64 devid, u8 *dev_uuid)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
|
|
|
|
|
|
|
|
device = kzalloc(sizeof(*device), GFP_NOFS);
|
2008-11-12 20:38:54 +01:00
|
|
|
if (!device)
|
|
|
|
return NULL;
|
2008-05-13 19:46:40 +02:00
|
|
|
list_add(&device->dev_list,
|
|
|
|
&fs_devices->devices);
|
|
|
|
device->dev_root = root->fs_info->dev_root;
|
|
|
|
device->devid = devid;
|
2008-06-11 22:50:36 +02:00
|
|
|
device->work.func = pending_bios_fn;
|
2008-12-12 16:03:26 +01:00
|
|
|
device->fs_devices = fs_devices;
|
2010-12-13 20:56:23 +01:00
|
|
|
device->missing = 1;
|
2008-05-13 19:46:40 +02:00
|
|
|
fs_devices->num_devices++;
|
2010-12-13 20:56:23 +01:00
|
|
|
fs_devices->missing_devices++;
|
2008-05-13 19:46:40 +02:00
|
|
|
spin_lock_init(&device->io_lock);
|
Btrfs: move data checksumming into a dedicated tree
Btrfs stores checksums for each data block. Until now, they have
been stored in the subvolume trees, indexed by the inode that is
referencing the data block. This means that when we read the inode,
we've probably read in at least some checksums as well.
But, this has a few problems:
* The checksums are indexed by logical offset in the file. When
compression is on, this means we have to do the expensive checksumming
on the uncompressed data. It would be faster if we could checksum
the compressed data instead.
* If we implement encryption, we'll be checksumming the plain text and
storing that on disk. This is significantly less secure.
* For either compression or encryption, we have to get the plain text
back before we can verify the checksum as correct. This makes the raid
layer balancing and extent moving much more expensive.
* It makes the front end caching code more complex, as we have touch
the subvolume and inodes as we cache extents.
* There is potentitally one copy of the checksum in each subvolume
referencing an extent.
The solution used here is to store the extent checksums in a dedicated
tree. This allows us to index the checksums by phyiscal extent
start and length. It means:
* The checksum is against the data stored on disk, after any compression
or encryption is done.
* The checksum is stored in a central location, and can be verified without
following back references, or reading inodes.
This makes compression significantly faster by reducing the amount of
data that needs to be checksummed. It will also allow much faster
raid management code in general.
The checksums are indexed by a key with a fixed objectid (a magic value
in ctree.h) and offset set to the starting byte of the extent. This
allows us to copy the checksum items into the fsync log tree directly (or
any other tree), without having to invent a second format for them.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-12-08 22:58:54 +01:00
|
|
|
INIT_LIST_HEAD(&device->dev_alloc_list);
|
2008-05-13 19:46:40 +02:00
|
|
|
memcpy(device->uuid, dev_uuid, BTRFS_UUID_SIZE);
|
|
|
|
return device;
|
|
|
|
}
|
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
static int read_one_chunk(struct btrfs_root *root, struct btrfs_key *key,
|
|
|
|
struct extent_buffer *leaf,
|
|
|
|
struct btrfs_chunk *chunk)
|
|
|
|
{
|
|
|
|
struct btrfs_mapping_tree *map_tree = &root->fs_info->mapping_tree;
|
|
|
|
struct map_lookup *map;
|
|
|
|
struct extent_map *em;
|
|
|
|
u64 logical;
|
|
|
|
u64 length;
|
|
|
|
u64 devid;
|
2008-04-18 16:29:38 +02:00
|
|
|
u8 uuid[BTRFS_UUID_SIZE];
|
2008-03-25 21:50:33 +01:00
|
|
|
int num_stripes;
|
2008-03-24 20:01:56 +01:00
|
|
|
int ret;
|
2008-03-25 21:50:33 +01:00
|
|
|
int i;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-04-15 21:41:47 +02:00
|
|
|
logical = key->offset;
|
|
|
|
length = btrfs_chunk_length(leaf, chunk);
|
2008-05-07 17:43:44 +02:00
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
read_lock(&map_tree->map_tree.lock);
|
2008-03-24 20:01:56 +01:00
|
|
|
em = lookup_extent_mapping(&map_tree->map_tree, logical, 1);
|
2009-09-02 22:24:52 +02:00
|
|
|
read_unlock(&map_tree->map_tree.lock);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
/* already mapped? */
|
|
|
|
if (em && em->start <= logical && em->start + em->len > logical) {
|
|
|
|
free_extent_map(em);
|
|
|
|
return 0;
|
|
|
|
} else if (em) {
|
|
|
|
free_extent_map(em);
|
|
|
|
}
|
|
|
|
|
2011-04-21 00:48:27 +02:00
|
|
|
em = alloc_extent_map();
|
2008-03-24 20:01:56 +01:00
|
|
|
if (!em)
|
|
|
|
return -ENOMEM;
|
2008-03-25 21:50:33 +01:00
|
|
|
num_stripes = btrfs_chunk_num_stripes(leaf, chunk);
|
|
|
|
map = kmalloc(map_lookup_size(num_stripes), GFP_NOFS);
|
2008-03-24 20:01:56 +01:00
|
|
|
if (!map) {
|
|
|
|
free_extent_map(em);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
em->bdev = (struct block_device *)map;
|
|
|
|
em->start = logical;
|
|
|
|
em->len = length;
|
|
|
|
em->block_start = 0;
|
Btrfs: Add zlib compression support
This is a large change for adding compression on reading and writing,
both for inline and regular extents. It does some fairly large
surgery to the writeback paths.
Compression is off by default and enabled by mount -o compress. Even
when the -o compress mount option is not used, it is possible to read
compressed extents off the disk.
If compression for a given set of pages fails to make them smaller, the
file is flagged to avoid future compression attempts later.
* While finding delalloc extents, the pages are locked before being sent down
to the delalloc handler. This allows the delalloc handler to do complex things
such as cleaning the pages, marking them writeback and starting IO on their
behalf.
* Inline extents are inserted at delalloc time now. This allows us to compress
the data before inserting the inline extent, and it allows us to insert
an inline extent that spans multiple pages.
* All of the in-memory extent representations (extent_map.c, ordered-data.c etc)
are changed to record both an in-memory size and an on disk size, as well
as a flag for compression.
From a disk format point of view, the extent pointers in the file are changed
to record the on disk size of a given extent and some encoding flags.
Space in the disk format is allocated for compression encoding, as well
as encryption and a generic 'other' field. Neither the encryption or the
'other' field are currently used.
In order to limit the amount of data read for a single random read in the
file, the size of a compressed extent is limited to 128k. This is a
software only limit, the disk format supports u64 sized compressed extents.
In order to limit the ram consumed while processing extents, the uncompressed
size of a compressed extent is limited to 256k. This is a software only limit
and will be subject to tuning later.
Checksumming is still done on compressed extents, and it is done on the
uncompressed version of the data. This way additional encodings can be
layered on without having to figure out which encoding to checksum.
Compression happens at delalloc time, which is basically singled threaded because
it is usually done by a single pdflush thread. This makes it tricky to
spread the compression load across all the cpus on the box. We'll have to
look at parallel pdflush walks of dirty inodes at a later time.
Decompression is hooked into readpages and it does spread across CPUs nicely.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 19:49:59 +01:00
|
|
|
em->block_len = em->len;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-03-25 21:50:33 +01:00
|
|
|
map->num_stripes = num_stripes;
|
|
|
|
map->io_width = btrfs_chunk_io_width(leaf, chunk);
|
|
|
|
map->io_align = btrfs_chunk_io_align(leaf, chunk);
|
|
|
|
map->sector_size = btrfs_chunk_sector_size(leaf, chunk);
|
|
|
|
map->stripe_len = btrfs_chunk_stripe_len(leaf, chunk);
|
|
|
|
map->type = btrfs_chunk_type(leaf, chunk);
|
2008-04-16 16:49:51 +02:00
|
|
|
map->sub_stripes = btrfs_chunk_sub_stripes(leaf, chunk);
|
2008-03-25 21:50:33 +01:00
|
|
|
for (i = 0; i < num_stripes; i++) {
|
|
|
|
map->stripes[i].physical =
|
|
|
|
btrfs_stripe_offset_nr(leaf, chunk, i);
|
|
|
|
devid = btrfs_stripe_devid_nr(leaf, chunk, i);
|
2008-04-18 16:29:38 +02:00
|
|
|
read_extent_buffer(leaf, uuid, (unsigned long)
|
|
|
|
btrfs_stripe_dev_uuid_nr(chunk, i),
|
|
|
|
BTRFS_UUID_SIZE);
|
2008-11-18 03:11:30 +01:00
|
|
|
map->stripes[i].dev = btrfs_find_device(root, devid, uuid,
|
|
|
|
NULL);
|
2008-05-13 19:46:40 +02:00
|
|
|
if (!map->stripes[i].dev && !btrfs_test_opt(root, DEGRADED)) {
|
2008-03-25 21:50:33 +01:00
|
|
|
kfree(map);
|
|
|
|
free_extent_map(em);
|
|
|
|
return -EIO;
|
|
|
|
}
|
2008-05-13 19:46:40 +02:00
|
|
|
if (!map->stripes[i].dev) {
|
|
|
|
map->stripes[i].dev =
|
|
|
|
add_missing_dev(root, devid, uuid);
|
|
|
|
if (!map->stripes[i].dev) {
|
|
|
|
kfree(map);
|
|
|
|
free_extent_map(em);
|
|
|
|
return -EIO;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
map->stripes[i].dev->in_fs_metadata = 1;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
|
2009-09-02 22:24:52 +02:00
|
|
|
write_lock(&map_tree->map_tree.lock);
|
2008-03-24 20:01:56 +01:00
|
|
|
ret = add_extent_mapping(&map_tree->map_tree, em);
|
2009-09-02 22:24:52 +02:00
|
|
|
write_unlock(&map_tree->map_tree.lock);
|
2008-04-14 15:48:18 +02:00
|
|
|
BUG_ON(ret);
|
2008-03-24 20:01:56 +01:00
|
|
|
free_extent_map(em);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int fill_device_from_item(struct extent_buffer *leaf,
|
|
|
|
struct btrfs_dev_item *dev_item,
|
|
|
|
struct btrfs_device *device)
|
|
|
|
{
|
|
|
|
unsigned long ptr;
|
|
|
|
|
|
|
|
device->devid = btrfs_device_id(leaf, dev_item);
|
2009-04-27 13:29:03 +02:00
|
|
|
device->disk_total_bytes = btrfs_device_total_bytes(leaf, dev_item);
|
|
|
|
device->total_bytes = device->disk_total_bytes;
|
2008-03-24 20:01:56 +01:00
|
|
|
device->bytes_used = btrfs_device_bytes_used(leaf, dev_item);
|
|
|
|
device->type = btrfs_device_type(leaf, dev_item);
|
|
|
|
device->io_align = btrfs_device_io_align(leaf, dev_item);
|
|
|
|
device->io_width = btrfs_device_io_width(leaf, dev_item);
|
|
|
|
device->sector_size = btrfs_device_sector_size(leaf, dev_item);
|
|
|
|
|
|
|
|
ptr = (unsigned long)btrfs_device_uuid(dev_item);
|
2008-04-15 21:41:47 +02:00
|
|
|
read_extent_buffer(leaf, device->uuid, ptr, BTRFS_UUID_SIZE);
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-11-18 03:11:30 +01:00
|
|
|
static int open_seed_devices(struct btrfs_root *root, u8 *fsid)
|
|
|
|
{
|
|
|
|
struct btrfs_fs_devices *fs_devices;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
mutex_lock(&uuid_mutex);
|
|
|
|
|
|
|
|
fs_devices = root->fs_info->fs_devices->seed;
|
|
|
|
while (fs_devices) {
|
|
|
|
if (!memcmp(fs_devices->fsid, fsid, BTRFS_UUID_SIZE)) {
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
fs_devices = fs_devices->seed;
|
|
|
|
}
|
|
|
|
|
|
|
|
fs_devices = find_fsid(fsid);
|
|
|
|
if (!fs_devices) {
|
|
|
|
ret = -ENOENT;
|
|
|
|
goto out;
|
|
|
|
}
|
2008-12-12 16:03:26 +01:00
|
|
|
|
|
|
|
fs_devices = clone_fs_devices(fs_devices);
|
|
|
|
if (IS_ERR(fs_devices)) {
|
|
|
|
ret = PTR_ERR(fs_devices);
|
2008-11-18 03:11:30 +01:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2008-12-02 12:36:09 +01:00
|
|
|
ret = __btrfs_open_devices(fs_devices, FMODE_READ,
|
2008-11-20 03:17:22 +01:00
|
|
|
root->fs_info->bdev_holder);
|
2008-11-18 03:11:30 +01:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (!fs_devices->seeding) {
|
|
|
|
__btrfs_close_devices(fs_devices);
|
2008-12-12 16:03:26 +01:00
|
|
|
free_fs_devices(fs_devices);
|
2008-11-18 03:11:30 +01:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
fs_devices->seed = root->fs_info->fs_devices->seed;
|
|
|
|
root->fs_info->fs_devices->seed = fs_devices;
|
|
|
|
out:
|
|
|
|
mutex_unlock(&uuid_mutex);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
static int read_one_dev(struct btrfs_root *root,
|
2008-03-24 20:01:56 +01:00
|
|
|
struct extent_buffer *leaf,
|
|
|
|
struct btrfs_dev_item *dev_item)
|
|
|
|
{
|
|
|
|
struct btrfs_device *device;
|
|
|
|
u64 devid;
|
|
|
|
int ret;
|
2008-11-18 03:11:30 +01:00
|
|
|
u8 fs_uuid[BTRFS_UUID_SIZE];
|
2008-04-18 16:29:38 +02:00
|
|
|
u8 dev_uuid[BTRFS_UUID_SIZE];
|
|
|
|
|
2008-03-24 20:01:56 +01:00
|
|
|
devid = btrfs_device_id(leaf, dev_item);
|
2008-04-18 16:29:38 +02:00
|
|
|
read_extent_buffer(leaf, dev_uuid,
|
|
|
|
(unsigned long)btrfs_device_uuid(dev_item),
|
|
|
|
BTRFS_UUID_SIZE);
|
2008-11-18 03:11:30 +01:00
|
|
|
read_extent_buffer(leaf, fs_uuid,
|
|
|
|
(unsigned long)btrfs_device_fsid(dev_item),
|
|
|
|
BTRFS_UUID_SIZE);
|
|
|
|
|
|
|
|
if (memcmp(fs_uuid, root->fs_info->fsid, BTRFS_UUID_SIZE)) {
|
|
|
|
ret = open_seed_devices(root, fs_uuid);
|
2008-12-12 16:03:26 +01:00
|
|
|
if (ret && !btrfs_test_opt(root, DEGRADED))
|
2008-11-18 03:11:30 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
device = btrfs_find_device(root, devid, dev_uuid, fs_uuid);
|
|
|
|
if (!device || !device->bdev) {
|
2008-12-12 16:03:26 +01:00
|
|
|
if (!btrfs_test_opt(root, DEGRADED))
|
2008-11-18 03:11:30 +01:00
|
|
|
return -EIO;
|
|
|
|
|
|
|
|
if (!device) {
|
2009-01-06 03:25:51 +01:00
|
|
|
printk(KERN_WARNING "warning devid %llu missing\n",
|
|
|
|
(unsigned long long)devid);
|
2008-11-18 03:11:30 +01:00
|
|
|
device = add_missing_dev(root, devid, dev_uuid);
|
|
|
|
if (!device)
|
|
|
|
return -ENOMEM;
|
2010-12-13 20:56:23 +01:00
|
|
|
} else if (!device->missing) {
|
|
|
|
/*
|
|
|
|
* this happens when a device that was properly setup
|
|
|
|
* in the device info lists suddenly goes bad.
|
|
|
|
* device->bdev is NULL, and so we have to set
|
|
|
|
* device->missing to one here
|
|
|
|
*/
|
|
|
|
root->fs_info->fs_devices->missing_devices++;
|
|
|
|
device->missing = 1;
|
2008-11-18 03:11:30 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (device->fs_devices != root->fs_info->fs_devices) {
|
|
|
|
BUG_ON(device->writeable);
|
|
|
|
if (device->generation !=
|
|
|
|
btrfs_device_generation(leaf, dev_item))
|
|
|
|
return -EINVAL;
|
2008-03-24 20:01:59 +01:00
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
|
|
|
|
fill_device_from_item(leaf, dev_item, device);
|
|
|
|
device->dev_root = root->fs_info->dev_root;
|
2008-05-13 19:46:40 +02:00
|
|
|
device->in_fs_metadata = 1;
|
2011-09-26 23:12:22 +02:00
|
|
|
if (device->writeable) {
|
2008-11-18 03:11:30 +01:00
|
|
|
device->fs_devices->total_rw_bytes += device->total_bytes;
|
2011-09-26 23:12:22 +02:00
|
|
|
spin_lock(&root->fs_info->free_chunk_lock);
|
|
|
|
root->fs_info->free_chunk_space += device->total_bytes -
|
|
|
|
device->bytes_used;
|
|
|
|
spin_unlock(&root->fs_info->free_chunk_lock);
|
|
|
|
}
|
2008-03-24 20:01:56 +01:00
|
|
|
ret = 0;
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
int btrfs_read_sys_array(struct btrfs_root *root)
|
2008-03-24 20:01:56 +01:00
|
|
|
{
|
|
|
|
struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
|
2008-05-07 17:43:44 +02:00
|
|
|
struct extent_buffer *sb;
|
2008-03-24 20:01:56 +01:00
|
|
|
struct btrfs_disk_key *disk_key;
|
|
|
|
struct btrfs_chunk *chunk;
|
2008-04-25 15:04:37 +02:00
|
|
|
u8 *ptr;
|
|
|
|
unsigned long sb_ptr;
|
|
|
|
int ret = 0;
|
2008-03-24 20:01:56 +01:00
|
|
|
u32 num_stripes;
|
|
|
|
u32 array_size;
|
|
|
|
u32 len = 0;
|
|
|
|
u32 cur;
|
2008-04-25 15:04:37 +02:00
|
|
|
struct btrfs_key key;
|
2008-03-24 20:01:56 +01:00
|
|
|
|
2008-12-12 16:03:26 +01:00
|
|
|
sb = btrfs_find_create_tree_block(root, BTRFS_SUPER_INFO_OFFSET,
|
2008-05-07 17:43:44 +02:00
|
|
|
BTRFS_SUPER_INFO_SIZE);
|
|
|
|
if (!sb)
|
|
|
|
return -ENOMEM;
|
|
|
|
btrfs_set_buffer_uptodate(sb);
|
2011-07-26 22:11:19 +02:00
|
|
|
btrfs_set_buffer_lockdep_class(root->root_key.objectid, sb, 0);
|
2009-02-12 20:09:45 +01:00
|
|
|
|
2008-05-07 17:43:44 +02:00
|
|
|
write_extent_buffer(sb, super_copy, 0, BTRFS_SUPER_INFO_SIZE);
|
2008-03-24 20:01:56 +01:00
|
|
|
array_size = btrfs_super_sys_array_size(super_copy);
|
|
|
|
|
|
|
|
ptr = super_copy->sys_chunk_array;
|
|
|
|
sb_ptr = offsetof(struct btrfs_super_block, sys_chunk_array);
|
|
|
|
cur = 0;
|
|
|
|
|
|
|
|
while (cur < array_size) {
|
|
|
|
disk_key = (struct btrfs_disk_key *)ptr;
|
|
|
|
btrfs_disk_key_to_cpu(&key, disk_key);
|
|
|
|
|
2008-05-07 17:43:44 +02:00
|
|
|
len = sizeof(*disk_key); ptr += len;
|
2008-03-24 20:01:56 +01:00
|
|
|
sb_ptr += len;
|
|
|
|
cur += len;
|
|
|
|
|
2008-03-24 20:02:07 +01:00
|
|
|
if (key.type == BTRFS_CHUNK_ITEM_KEY) {
|
2008-03-24 20:01:56 +01:00
|
|
|
chunk = (struct btrfs_chunk *)sb_ptr;
|
2008-03-24 20:02:07 +01:00
|
|
|
ret = read_one_chunk(root, &key, sb, chunk);
|
2008-04-25 15:04:37 +02:00
|
|
|
if (ret)
|
|
|
|
break;
|
2008-03-24 20:01:56 +01:00
|
|
|
num_stripes = btrfs_chunk_num_stripes(sb, chunk);
|
|
|
|
len = btrfs_chunk_item_size(num_stripes);
|
|
|
|
} else {
|
2008-04-25 15:04:37 +02:00
|
|
|
ret = -EIO;
|
|
|
|
break;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
ptr += len;
|
|
|
|
sb_ptr += len;
|
|
|
|
cur += len;
|
|
|
|
}
|
2008-05-07 17:43:44 +02:00
|
|
|
free_extent_buffer(sb);
|
2008-04-25 15:04:37 +02:00
|
|
|
return ret;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
int btrfs_read_chunk_tree(struct btrfs_root *root)
|
|
|
|
{
|
|
|
|
struct btrfs_path *path;
|
|
|
|
struct extent_buffer *leaf;
|
|
|
|
struct btrfs_key key;
|
|
|
|
struct btrfs_key found_key;
|
|
|
|
int ret;
|
|
|
|
int slot;
|
|
|
|
|
|
|
|
root = root->fs_info->chunk_root;
|
|
|
|
|
|
|
|
path = btrfs_alloc_path();
|
|
|
|
if (!path)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
/* first we search for all of the device items, and then we
|
|
|
|
* read in all of the chunk items. This way we can create chunk
|
|
|
|
* mappings that reference all of the devices that are afound
|
|
|
|
*/
|
|
|
|
key.objectid = BTRFS_DEV_ITEMS_OBJECTID;
|
|
|
|
key.offset = 0;
|
|
|
|
key.type = 0;
|
|
|
|
again:
|
|
|
|
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
|
2010-03-25 13:34:49 +01:00
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
2009-01-06 03:25:51 +01:00
|
|
|
while (1) {
|
2008-03-24 20:01:56 +01:00
|
|
|
leaf = path->nodes[0];
|
|
|
|
slot = path->slots[0];
|
|
|
|
if (slot >= btrfs_header_nritems(leaf)) {
|
|
|
|
ret = btrfs_next_leaf(root, path);
|
|
|
|
if (ret == 0)
|
|
|
|
continue;
|
|
|
|
if (ret < 0)
|
|
|
|
goto error;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
btrfs_item_key_to_cpu(leaf, &found_key, slot);
|
|
|
|
if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
|
|
|
|
if (found_key.objectid != BTRFS_DEV_ITEMS_OBJECTID)
|
|
|
|
break;
|
|
|
|
if (found_key.type == BTRFS_DEV_ITEM_KEY) {
|
|
|
|
struct btrfs_dev_item *dev_item;
|
|
|
|
dev_item = btrfs_item_ptr(leaf, slot,
|
|
|
|
struct btrfs_dev_item);
|
2008-03-24 20:02:07 +01:00
|
|
|
ret = read_one_dev(root, leaf, dev_item);
|
2008-11-18 03:11:30 +01:00
|
|
|
if (ret)
|
|
|
|
goto error;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
} else if (found_key.type == BTRFS_CHUNK_ITEM_KEY) {
|
|
|
|
struct btrfs_chunk *chunk;
|
|
|
|
chunk = btrfs_item_ptr(leaf, slot, struct btrfs_chunk);
|
|
|
|
ret = read_one_chunk(root, &found_key, leaf, chunk);
|
2008-11-18 03:11:30 +01:00
|
|
|
if (ret)
|
|
|
|
goto error;
|
2008-03-24 20:01:56 +01:00
|
|
|
}
|
|
|
|
path->slots[0]++;
|
|
|
|
}
|
|
|
|
if (key.objectid == BTRFS_DEV_ITEMS_OBJECTID) {
|
|
|
|
key.objectid = 0;
|
2011-04-21 01:20:15 +02:00
|
|
|
btrfs_release_path(path);
|
2008-03-24 20:01:56 +01:00
|
|
|
goto again;
|
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
error:
|
2008-11-18 03:11:30 +01:00
|
|
|
btrfs_free_path(path);
|
2008-03-24 20:01:56 +01:00
|
|
|
return ret;
|
|
|
|
}
|