Commit Graph

47 Commits

Author SHA1 Message Date
Mel Gorman d0164adc89 mm, page_alloc: distinguish between being unable to sleep, unwilling to sleep and avoiding waking kswapd
__GFP_WAIT has been used to identify atomic context in callers that hold
spinlocks or are in interrupts.  They are expected to be high priority and
have access one of two watermarks lower than "min" which can be referred
to as the "atomic reserve".  __GFP_HIGH users get access to the first
lower watermark and can be called the "high priority reserve".

Over time, callers had a requirement to not block when fallback options
were available.  Some have abused __GFP_WAIT leading to a situation where
an optimisitic allocation with a fallback option can access atomic
reserves.

This patch uses __GFP_ATOMIC to identify callers that are truely atomic,
cannot sleep and have no alternative.  High priority users continue to use
__GFP_HIGH.  __GFP_DIRECT_RECLAIM identifies callers that can sleep and
are willing to enter direct reclaim.  __GFP_KSWAPD_RECLAIM to identify
callers that want to wake kswapd for background reclaim.  __GFP_WAIT is
redefined as a caller that is willing to enter direct reclaim and wake
kswapd for background reclaim.

This patch then converts a number of sites

o __GFP_ATOMIC is used by callers that are high priority and have memory
  pools for those requests. GFP_ATOMIC uses this flag.

o Callers that have a limited mempool to guarantee forward progress clear
  __GFP_DIRECT_RECLAIM but keep __GFP_KSWAPD_RECLAIM. bio allocations fall
  into this category where kswapd will still be woken but atomic reserves
  are not used as there is a one-entry mempool to guarantee progress.

o Callers that are checking if they are non-blocking should use the
  helper gfpflags_allow_blocking() where possible. This is because
  checking for __GFP_WAIT as was done historically now can trigger false
  positives. Some exceptions like dm-crypt.c exist where the code intent
  is clearer if __GFP_DIRECT_RECLAIM is used instead of the helper due to
  flag manipulations.

o Callers that built their own GFP flags instead of starting with GFP_KERNEL
  and friends now also need to specify __GFP_KSWAPD_RECLAIM.

The first key hazard to watch out for is callers that removed __GFP_WAIT
and was depending on access to atomic reserves for inconspicuous reasons.
In some cases it may be appropriate for them to use __GFP_HIGH.

The second key hazard is callers that assembled their own combination of
GFP flags instead of starting with something like GFP_KERNEL.  They may
now wish to specify __GFP_KSWAPD_RECLAIM.  It's almost certainly harmless
if it's missed in most cases as other activity will wake kswapd.

Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-11-06 17:50:42 -08:00
Linus Torvalds d9734e0d1c Merge branch 'for-4.4/core' of git://git.kernel.dk/linux-block
Pull core block updates from Jens Axboe:
 "This is the core block pull request for 4.4.  I've got a few more
  topic branches this time around, some of them will layer on top of the
  core+drivers changes and will come in a separate round.  So not a huge
  chunk of changes in this round.

  This pull request contains:

   - Enable blk-mq page allocation tracking with kmemleak, from Catalin.

   - Unused prototype removal in blk-mq from Christoph.

   - Cleanup of the q->blk_trace exchange, using cmpxchg instead of two
     xchg()'s, from Davidlohr.

   - A plug flush fix from Jeff.

   - Also from Jeff, a fix that means we don't have to update shared tag
     sets at init time unless we do a state change.  This cuts down boot
     times on thousands of devices a lot with scsi/blk-mq.

   - blk-mq waitqueue barrier fix from Kosuke.

   - Various fixes from Ming:

        - Fixes for segment merging and splitting, and checks, for
          the old core and blk-mq.

        - Potential blk-mq speedup by marking ctx pending at the end
          of a plug insertion batch in blk-mq.

        - direct-io no page dirty on kernel direct reads.

   - A WRITE_SYNC fix for mpage from Roman"

* 'for-4.4/core' of git://git.kernel.dk/linux-block:
  blk-mq: avoid excessive boot delays with large lun counts
  blktrace: re-write setting q->blk_trace
  blk-mq: mark ctx as pending at batch in flush plug path
  blk-mq: fix for trace_block_plug()
  block: check bio_mergeable() early before merging
  blk-mq: check bio_mergeable() early before merging
  block: avoid to merge splitted bio
  block: setup bi_phys_segments after splitting
  block: fix plug list flushing for nomerge queues
  blk-mq: remove unused blk_mq_clone_flush_request prototype
  blk-mq: fix waitqueue_active without memory barrier in block/blk-mq-tag.c
  fs: direct-io: don't dirtying pages for ITER_BVEC/ITER_KVEC direct read
  fs/mpage.c: forgotten WRITE_SYNC in case of data integrity write
  block: kmemleak: Track the page allocations for struct request
2015-11-04 20:28:10 -08:00
Junichi Nomura f42d79ab67 blk-mq: fix use-after-free in blk_mq_free_tag_set()
tags is freed in blk_mq_free_rq_map() and should not be used after that.
The problem doesn't manifest if CONFIG_CPUMASK_OFFSTACK is false because
free_cpumask_var() is nop.

tags->cpumask is allocated in blk_mq_init_tags() so it's natural to
free cpumask in its counter part, blk_mq_free_tags().

Fixes: f26cdc8536 ("blk-mq: Shared tag enhancements")
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: Keith Busch <keith.busch@intel.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-10-15 08:45:58 -06:00
Kosuke Tatsukawa 8ee1b7b9d8 blk-mq: fix waitqueue_active without memory barrier in block/blk-mq-tag.c
blk_mq_tag_update_depth() seems to be missing a memory barrier which
might cause the waker to not notice the waiter and fail to send a
wake_up as in the following figure.

	blk_mq_tag_update_depth			bt_get
------------------------------------------------------------------------
if (waitqueue_active(&bs->wait))
/* The CPU might reorder the test for
   the waitqueue up here, before
   prior writes complete */
					prepare_to_wait(&bs->wait, &wait,
					  TASK_UNINTERRUPTIBLE);
					tag = __bt_get(hctx, bt, last_tag,
					  tags);
					/* Value set in bt_update_count not
					   visible yet */
bt_update_count(&tags->bitmap_tags, tdepth);
/* blk_mq_tag_wakeup_all(tags, false); */
 bt = &tags->bitmap_tags;
 wake_index = atomic_read(&bt->wake_index);
					...
					io_schedule();
------------------------------------------------------------------------

This patch adds the missing memory barrier.

I found this issue when I was looking through the linux source code
for places calling waitqueue_active() before wake_up*(), but without
preceding memory barriers, after sending a patch to fix a similar
issue in drivers/tty/n_tty.c  (Details about the original issue can be
found here: https://lkml.org/lkml/2015/9/28/849).

Signed-off-by: Kosuke Tatsukawa <tatsu@ab.jp.nec.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-10-09 10:52:46 -06:00
Christoph Hellwig 0bf6cd5b95 blk-mq: factor out a helper to iterate all tags for a request_queue
And replace the blk_mq_tag_busy_iter with it - the driver use has been
replaced with a new helper a while ago, and internal to the block we
only need the new version.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-10-01 10:10:57 +02:00
Ming Lei 0048b4837a blk-mq: fix race between timeout and freeing request
Inside timeout handler, blk_mq_tag_to_rq() is called
to retrieve the request from one tag. This way is obviously
wrong because the request can be freed any time and some
fiedds of the request can't be trusted, then kernel oops
might be triggered[1].

Currently wrt. blk_mq_tag_to_rq(), the only special case is
that the flush request can share same tag with the request
cloned from, and the two requests can't be active at the same
time, so this patch fixes the above issue by updating tags->rqs[tag]
with the active request(either flush rq or the request cloned
from) of the tag.

Also blk_mq_tag_to_rq() gets much simplified with this patch.

Given blk_mq_tag_to_rq() is mainly for drivers and the caller must
make sure the request can't be freed, so in bt_for_each() this
helper is replaced with tags->rqs[tag].

[1] kernel oops log
[  439.696220] BUG: unable to handle kernel NULL pointer dereference at 0000000000000158^M
[  439.697162] IP: [<ffffffff812d89ba>] blk_mq_tag_to_rq+0x21/0x6e^M
[  439.700653] PGD 7ef765067 PUD 7ef764067 PMD 0 ^M
[  439.700653] Oops: 0000 [#1] PREEMPT SMP DEBUG_PAGEALLOC ^M
[  439.700653] Dumping ftrace buffer:^M
[  439.700653]    (ftrace buffer empty)^M
[  439.700653] Modules linked in: nbd ipv6 kvm_intel kvm serio_raw^M
[  439.700653] CPU: 6 PID: 2779 Comm: stress-ng-sigfd Not tainted 4.2.0-rc5-next-20150805+ #265^M
[  439.730500] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011^M
[  439.730500] task: ffff880605308000 ti: ffff88060530c000 task.ti: ffff88060530c000^M
[  439.730500] RIP: 0010:[<ffffffff812d89ba>]  [<ffffffff812d89ba>] blk_mq_tag_to_rq+0x21/0x6e^M
[  439.730500] RSP: 0018:ffff880819203da0  EFLAGS: 00010283^M
[  439.730500] RAX: ffff880811b0e000 RBX: ffff8800bb465f00 RCX: 0000000000000002^M
[  439.730500] RDX: 0000000000000000 RSI: 0000000000000202 RDI: 0000000000000000^M
[  439.730500] RBP: ffff880819203db0 R08: 0000000000000002 R09: 0000000000000000^M
[  439.730500] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000202^M
[  439.730500] R13: ffff880814104800 R14: 0000000000000002 R15: ffff880811a2ea00^M
[  439.730500] FS:  00007f165b3f5740(0000) GS:ffff880819200000(0000) knlGS:0000000000000000^M
[  439.730500] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b^M
[  439.730500] CR2: 0000000000000158 CR3: 00000007ef766000 CR4: 00000000000006e0^M
[  439.730500] Stack:^M
[  439.730500]  0000000000000008 ffff8808114eed90 ffff880819203e00 ffffffff812dc104^M
[  439.755663]  ffff880819203e40 ffffffff812d9f5e 0000020000000000 ffff8808114eed80^M
[  439.755663] Call Trace:^M
[  439.755663]  <IRQ> ^M
[  439.755663]  [<ffffffff812dc104>] bt_for_each+0x6e/0xc8^M
[  439.755663]  [<ffffffff812d9f5e>] ? blk_mq_rq_timed_out+0x6a/0x6a^M
[  439.755663]  [<ffffffff812d9f5e>] ? blk_mq_rq_timed_out+0x6a/0x6a^M
[  439.755663]  [<ffffffff812dc1b3>] blk_mq_tag_busy_iter+0x55/0x5e^M
[  439.755663]  [<ffffffff812d88b4>] ? blk_mq_bio_to_request+0x38/0x38^M
[  439.755663]  [<ffffffff812d8911>] blk_mq_rq_timer+0x5d/0xd4^M
[  439.755663]  [<ffffffff810a3e10>] call_timer_fn+0xf7/0x284^M
[  439.755663]  [<ffffffff810a3d1e>] ? call_timer_fn+0x5/0x284^M
[  439.755663]  [<ffffffff812d88b4>] ? blk_mq_bio_to_request+0x38/0x38^M
[  439.755663]  [<ffffffff810a46d6>] run_timer_softirq+0x1ce/0x1f8^M
[  439.755663]  [<ffffffff8104c367>] __do_softirq+0x181/0x3a4^M
[  439.755663]  [<ffffffff8104c76e>] irq_exit+0x40/0x94^M
[  439.755663]  [<ffffffff81031482>] smp_apic_timer_interrupt+0x33/0x3e^M
[  439.755663]  [<ffffffff815559a4>] apic_timer_interrupt+0x84/0x90^M
[  439.755663]  <EOI> ^M
[  439.755663]  [<ffffffff81554350>] ? _raw_spin_unlock_irq+0x32/0x4a^M
[  439.755663]  [<ffffffff8106a98b>] finish_task_switch+0xe0/0x163^M
[  439.755663]  [<ffffffff8106a94d>] ? finish_task_switch+0xa2/0x163^M
[  439.755663]  [<ffffffff81550066>] __schedule+0x469/0x6cd^M
[  439.755663]  [<ffffffff8155039b>] schedule+0x82/0x9a^M
[  439.789267]  [<ffffffff8119b28b>] signalfd_read+0x186/0x49a^M
[  439.790911]  [<ffffffff8106d86a>] ? wake_up_q+0x47/0x47^M
[  439.790911]  [<ffffffff811618c2>] __vfs_read+0x28/0x9f^M
[  439.790911]  [<ffffffff8117a289>] ? __fget_light+0x4d/0x74^M
[  439.790911]  [<ffffffff811620a7>] vfs_read+0x7a/0xc6^M
[  439.790911]  [<ffffffff8116292b>] SyS_read+0x49/0x7f^M
[  439.790911]  [<ffffffff81554c17>] entry_SYSCALL_64_fastpath+0x12/0x6f^M
[  439.790911] Code: 48 89 e5 e8 a9 b8 e7 ff 5d c3 0f 1f 44 00 00 55 89
f2 48 89 e5 41 54 41 89 f4 53 48 8b 47 60 48 8b 1c d0 48 8b 7b 30 48 8b
53 38 <48> 8b 87 58 01 00 00 48 85 c0 75 09 48 8b 97 88 0c 00 00 eb 10
^M
[  439.790911] RIP  [<ffffffff812d89ba>] blk_mq_tag_to_rq+0x21/0x6e^M
[  439.790911]  RSP <ffff880819203da0>^M
[  439.790911] CR2: 0000000000000158^M
[  439.790911] ---[ end trace d40af58949325661 ]---^M

Cc: <stable@vger.kernel.org>
Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-08-15 09:45:21 -06:00
Keith Busch f26cdc8536 blk-mq: Shared tag enhancements
Storage controllers may expose multiple block devices that share hardware
resources managed by blk-mq. This patch enhances the shared tags so a
low-level driver can access the shared resources not tied to the unshared
h/w contexts. This way the LLD can dynamically add and delete disks and
request queues without having to track all the request_queue hctx's to
iterate outstanding tags.

Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-06-01 14:35:56 -06:00
Sam Bradshaw bc188d818e blkmq: Fix NULL pointer deref when all reserved tags in
When allocating from the reserved tags pool, bt_get() is called with
a NULL hctx.  If all tags are in use, the hw queue is kicked to push
out any pending IO, potentially freeing tags, and tag allocation is
retried.  The problem is that blk_mq_run_hw_queue() doesn't check for
a NULL hctx.  So we avoid it with a simple NULL hctx test.

Tested by hammering mtip32xx with concurrent smartctl/hdparm.

Signed-off-by: Sam Bradshaw <sbradshaw@micron.com>
Signed-off-by: Selvan Mani <smani@micron.com>
Fixes: b32232073e ("blk-mq: fix hang in bt_get()")
Cc: stable@kernel.org

Added appropriate comment.

Signed-off-by: Jens Axboe <axboe@fb.com>
2015-03-18 17:06:18 -06:00
Tony Battersby 564e559f2b blk-mq: fix double-free in error path
If the allocation of bt->bs fails, then bt->map can be freed twice, once
in blk_mq_init_bitmap_tags() -> bt_alloc(), and once in
blk_mq_init_bitmap_tags() -> bt_free().  Fix by setting the pointer to
NULL after the first free.

Cc: <stable@vger.kernel.org>
Signed-off-by: Tony Battersby <tonyb@cybernetics.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-02-11 09:35:21 -07:00
Shaohua Li 24391c0dc5 blk-mq: add tag allocation policy
This is the blk-mq part to support tag allocation policy. The default
allocation policy isn't changed (though it's not a strict FIFO). The new
policy is round-robin for libata. But it's a try-best implementation. If
multiple tasks are competing, the tags returned will be mixed (which is
unavoidable even with !mq, as requests from different tasks can be
mixed in queue)

Cc: Jens Axboe <axboe@fb.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2015-01-23 14:18:00 -07:00
Jens Axboe 0bf364984c blk-mq: fix false negative out-of-tags condition
The blk-mq tagging tries to maintain some locality between CPUs and
the tags issued. The tags are split into groups of words, and the
words may not be fully populated. When searching for a new free tag,
blk-mq may look at partial words, hence it passes in an offset/size
to find_next_zero_bit(). However, it does that wrong, the size must
always be the full length of the number of tags in that word,
otherwise we'll potentially miss some near the end.

Another issue is when __bt_get() goes from one word set to the next.
It bumps the index, but not the last_tag associated with the
previous index. Bump that to be in the range of the new word.

Finally, clean up __bt_get() and __bt_get_word() a bit and get
rid of the goto in there, and the unnecessary 'wrap' variable.

Signed-off-by: Jens Axboe <axboe@fb.com>
2015-01-14 08:49:55 -07:00
Jens Axboe aed3ea94bd block: wake up waiters when a queue is marked dying
If it's dying, we can't expect new request to complete and come
in an wake up other tasks waiting for requests. So after we
have marked it as dying, wake up everybody currently waiting
for a request. Once they wake, they will retry their allocation
and fail appropriately due to the state of the queue.

Tested-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-12-31 09:39:16 -07:00
Jens Axboe 35d37c6635 Revert "blk-mq: Micro-optimize bt_get()"
This reverts commit 52f7eb945f.

The optimization is only really safe for a single queue, otherwise
'bs' and 'bt' can indeed change, and if we don't do a finish_wait()
for each loop, we'll potentially change the wait structure and
corrupt task wait list.

Reported-by: Jan Kara <jack@suse.cz>
2014-12-15 08:30:26 -07:00
Linus Torvalds caf292ae5b Merge branch 'for-3.19/core' of git://git.kernel.dk/linux-block
Pull block driver core update from Jens Axboe:
 "This is the pull request for the core block IO changes for 3.19.  Not
  a huge round this time, mostly lots of little good fixes:

   - Fix a bug in sysfs blktrace interface causing a NULL pointer
     dereference, when enabled/disabled through that API.  From Arianna
     Avanzini.

   - Various updates/fixes/improvements for blk-mq:

        - A set of updates from Bart, mostly fixing buts in the tag
          handling.

        - Cleanup/code consolidation from Christoph.

        - Extend queue_rq API to be able to handle batching issues of IO
          requests. NVMe will utilize this shortly. From me.

        - A few tag and request handling updates from me.

        - Cleanup of the preempt handling for running queues from Paolo.

        - Prevent running of unmapped hardware queues from Ming Lei.

        - Move the kdump memory limiting check to be in the correct
          location, from Shaohua.

        - Initialize all software queues at init time from Takashi. This
          prevents a kobject warning when CPUs are brought online that
          weren't online when a queue was registered.

   - Single writeback fix for I_DIRTY clearing from Tejun.  Queued with
     the core IO changes, since it's just a single fix.

   - Version X of the __bio_add_page() segment addition retry from
     Maurizio.  Hope the Xth time is the charm.

   - Documentation fixup for IO scheduler merging from Jan.

   - Introduce (and use) generic IO stat accounting helpers for non-rq
     drivers, from Gu Zheng.

   - Kill off artificial limiting of max sectors in a request from
     Christoph"

* 'for-3.19/core' of git://git.kernel.dk/linux-block: (26 commits)
  bio: modify __bio_add_page() to accept pages that don't start a new segment
  blk-mq: Fix uninitialized kobject at CPU hotplugging
  blktrace: don't let the sysfs interface remove trace from running list
  blk-mq: Use all available hardware queues
  blk-mq: Micro-optimize bt_get()
  blk-mq: Fix a race between bt_clear_tag() and bt_get()
  blk-mq: Avoid that __bt_get_word() wraps multiple times
  blk-mq: Fix a use-after-free
  blk-mq: prevent unmapped hw queue from being scheduled
  blk-mq: re-check for available tags after running the hardware queue
  blk-mq: fix hang in bt_get()
  blk-mq: move the kdump check to blk_mq_alloc_tag_set
  blk-mq: cleanup tag free handling
  blk-mq: use 'nr_cpu_ids' as highest CPU ID count for hwq <-> cpu map
  blk: introduce generic io stat accounting help function
  blk-mq: handle the single queue case in blk_mq_hctx_next_cpu
  genhd: check for int overflow in disk_expand_part_tbl()
  blk-mq: add blk_mq_free_hctx_request()
  blk-mq: export blk_mq_free_request()
  blk-mq: use get_cpu/put_cpu instead of preempt_disable/preempt_enable
  ...
2014-12-13 14:14:23 -08:00
Bart Van Assche 52f7eb945f blk-mq: Micro-optimize bt_get()
Remove a superfluous finish_wait() call. Convert the two bt_wait_ptr()
calls into a single call.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Robert Elliott <elliott@hp.com>
Cc: Ming Lei <ming.lei@canonical.com>
Cc: Alexander Gordeev <agordeev@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-12-09 09:07:28 -07:00
Bart Van Assche c38d185d4a blk-mq: Fix a race between bt_clear_tag() and bt_get()
What we need is the following two guarantees:
* Any thread that observes the effect of the test_and_set_bit() by
  __bt_get_word() also observes the preceding addition of 'current'
  to the appropriate wait list. This is guaranteed by the semantics
  of the spin_unlock() operation performed by prepare_and_wait().
  Hence the conversion of test_and_set_bit_lock() into
  test_and_set_bit().
* The wait lists are examined by bt_clear() after the tag bit has
  been cleared. clear_bit_unlock() guarantees that any thread that
  observes that the bit has been cleared also observes the store
  operations preceding clear_bit_unlock(). However,
  clear_bit_unlock() does not prevent that the wait lists are examined
  before that the tag bit is cleared. Hence the addition of a memory
  barrier between clear_bit() and the wait list examination.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Robert Elliott <elliott@hp.com>
Cc: Ming Lei <ming.lei@canonical.com>
Cc: Alexander Gordeev <agordeev@redhat.com>
Cc: <stable@vger.kernel.org> # v3.13+
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-12-09 09:07:16 -07:00
Bart Van Assche 9e98e9d7cf blk-mq: Avoid that __bt_get_word() wraps multiple times
If __bt_get_word() is called with last_tag != 0, if the first
find_next_zero_bit() fails, if after wrap-around the
test_and_set_bit() call fails and find_next_zero_bit() succeeds,
if the next test_and_set_bit() call fails and subsequently
find_next_zero_bit() does not find a zero bit, then another
wrap-around will occur. Avoid this by introducing an additional
local variable.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Robert Elliott <elliott@hp.com>
Cc: Ming Lei <ming.lei@canonical.com>
Cc: Alexander Gordeev <agordeev@redhat.com>
Cc: <stable@vger.kernel.org> # v3.13+
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-12-09 09:07:14 -07:00
Jens Axboe 080ff35114 blk-mq: re-check for available tags after running the hardware queue
If we run out of tags and have to sleep, we run the hardware queue
to kick pending IO into gear. During that run, we may have completed
requests, so re-check if we have free tags before going to sleep.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-12-08 08:49:06 -07:00
Bart Van Assche b32232073e blk-mq: fix hang in bt_get()
Avoid that if there are fewer hardware queues than CPU threads that
bt_get() can hang. The symptoms of the hang were as follows:

* All tags allocated for a particular hardware queue.
* (nr_tags) pending commands for that hardware queue.
* No pending commands for the software queues associated with that
  hardware queue.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-12-08 08:46:34 -07:00
Jens Axboe 70114c393c blk-mq: cleanup tag free handling
We only call __blk_mq_put_tag() and __blk_mq_put_reserved_tag()
from blk_mq_put_tag(), so just inline the two calls instead of
having them as separate functions.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-11-24 15:52:30 -07:00
Bart Van Assche 205fb5f5ba blk-mq: add blk_mq_unique_tag()
The queuecommand() callback functions in SCSI low-level drivers
need to know which hardware context has been selected by the
block layer. Since this information is not available in the
request structure, and since passing the hctx pointer directly to
the queuecommand callback function would require modification of
all SCSI LLDs, add a function to the block layer that allows to
query the hardware context index.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Acked-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
2014-11-12 11:16:09 +01:00
Bart Van Assche 9d8f0bcca6 blk-mq: Make bt_clear_tag() easier to read
Eliminate a backwards goto statement from bt_clear_tag().

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-10-07 08:45:21 -06:00
Jens Axboe abab13b5c4 blk-mq: fix potential hang if rolling wakeup depth is too high
We currently divide the queue depth by 4 as our batch wakeup
count, but we split the wakeups over BT_WAIT_QUEUES number of
wait queues. This defaults to 8. If the product of the resulting
batch wake count and BT_WAIT_QUEUES is higher than the device
queue depth, we can get into a situation where a task goes to
sleep waiting for a request, but never gets woken up.

Reported-by: Bart Van Assche <bvanassche@acm.org>
Fixes: 4bb659b156
Cc: stable@kernel.org
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-10-07 08:39:20 -06:00
Christoph Hellwig 81481eb423 blk-mq: fix and simplify tag iteration for the timeout handler
Don't do a kmalloc from timer to handle timeouts, chances are we could be
under heavy load or similar and thus just miss out on the timeouts.
Fortunately it is very easy to just iterate over all in use tags, and doing
this properly actually cleans up the blk_mq_busy_iter API as well, and
prepares us for the next patch by passing a reserved argument to the
iterator.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-09-22 12:00:07 -06:00
Alexander Gordeev 86fb5c56cf blk-mq: bitmap tag: fix races in bt_get() function
This update fixes few issues in bt_get() function:

- list_empty(&wait.task_list) check is not protected;

- was_empty check is always true which results in *every* thread
  entering the loop resets bt_wait_state::wait_cnt counter rather
  than every bt->wake_cnt'th thread;

- 'bt_wait_state::wait_cnt' counter update is redundant, since
  it also gets reset in bt_clear_tag() function;

Cc: Christoph Hellwig <hch@infradead.org>
Cc: Ming Lei <tom.leiming@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Alexander Gordeev <agordeev@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-06-17 22:13:08 -07:00
Alexander Gordeev 2971c35f35 blk-mq: bitmap tag: fix race on blk_mq_bitmap_tags::wake_cnt
This piece of code in bt_clear_tag() function is racy:

	bs = bt_wake_ptr(bt);
	if (bs && atomic_dec_and_test(&bs->wait_cnt)) {
		atomic_set(&bs->wait_cnt, bt->wake_cnt);
 		wake_up(&bs->wait);
	}

Since nothing prevents bt_wake_ptr() from returning the very
same 'bs' address on multiple CPUs, the following scenario is
possible:

    CPU1                                CPU2
    ----                                ----

0.  bs = bt_wake_ptr(bt);               bs = bt_wake_ptr(bt);
1.  atomic_dec_and_test(&bs->wait_cnt)
2.                                      atomic_dec_and_test(&bs->wait_cnt)
3.  atomic_set(&bs->wait_cnt, bt->wake_cnt);

If the decrement in [1] yields zero then for some amount of time
the decrement in [2] results in a negative/overflow value, which
is not expected. The follow-up assignment in [3] overwrites the
invalid value with the batch value (and likely prevents the issue
from being severe) which is still incorrect and should be a lesser.

Cc: Ming Lei <tom.leiming@gmail.com>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Alexander Gordeev <agordeev@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-06-17 22:13:05 -07:00
Alexander Gordeev 8537b12034 blk-mq: bitmap tag: fix races on shared ::wake_index fields
Fix racy updates of shared blk_mq_bitmap_tags::wake_index
and blk_mq_hw_ctx::wake_index fields.

Cc: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Alexander Gordeev <agordeev@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-06-17 22:12:35 -07:00
Ming Lei cb96a42cc1 blk-mq: fix schedule from atomic context
blk_mq_put_ctx() has to be called before io_schedule() in
bt_get().

This patch fixes the problem by taking similar approach from
percpu_ida allocation for the situation.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-06-03 21:04:39 -06:00
Jens Axboe 75bb4625bb blk-mq: add file comments and update copyright notices
None of the blk-mq files have an explanatory comment at the top
for what that particular file does. Add that and add appropriate
copyright notices as well.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-28 10:15:41 -06:00
Christoph Hellwig a3bd77567c blk-mq: remove blk_mq_wait_for_tags
The current logic for blocking tag allocation is rather confusing, as we
first allocated and then free again a tag in blk_mq_wait_for_tags, just
to attempt a non-blocking allocation and then repeat if someone else
managed to grab the tag before us.

Instead change blk_mq_alloc_request_pinned to simply do a blocking tag
allocation itself and use the request we get back from it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-28 09:49:23 -06:00
Sam Bradshaw edf866b380 blk-mq: export blk_mq_tag_busy_iter
Export the blk-mq in-flight tag iterator for driver consumption.
This is particularly useful in exception paths or SRSI where
in-flight IOs need to be cancelled and/or reissued. The NVMe driver
conversion will use this.

Signed-off-by: Sam Bradshaw <sbradshaw@micron.com>
Signed-off-by: Matias Bjørling <m@bjorling.me>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-23 13:30:16 -06:00
Jens Axboe e3a2b3f931 blk-mq: allow changing of queue depth through sysfs
For request_fn based devices, the block layer exports a 'nr_requests'
file through sysfs to allow adjusting of queue depth on the fly.
Currently this returns -EINVAL for blk-mq, since it's not wired up.
Wire this up for blk-mq, so that it now also always dynamic
adjustments of the allowed queue depth for any given block device
managed by blk-mq.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-20 11:49:02 -06:00
Jens Axboe 39a9f97e5e Merge branch 'for-3.16/blk-mq-tagging' into for-3.16/core
Signed-off-by: Jens Axboe <axboe@fb.com>

Conflicts:
	block/blk-mq-tag.c
2014-05-19 11:52:35 -06:00
Jens Axboe e93ecf602b blk-mq: move the cache friendly bitmap type of out blk-mq-tag
We will use it for the pending list in blk-mq core as well.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-19 11:02:47 -06:00
Jens Axboe 0d2602ca30 blk-mq: improve support for shared tags maps
This adds support for active queue tracking, meaning that the
blk-mq tagging maintains a count of active users of a tag set.
This allows us to maintain a notion of fairness between users,
so that we can distribute the tag depth evenly without starving
some users while allowing others to try unfair deep queues.

If sharing of a tag set is detected, each hardware queue will
track the depth of its own queue. And if this exceeds the total
depth divided by the number of active queues, the user is actively
throttled down.

The active queue count is done lazily to avoid bouncing that data
between submitter and completer. Each hardware queue gets marked
active when it allocates its first tag, and gets marked inactive
when 1) the last tag is cleared, and 2) the queue timeout grace
period has passed.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-13 15:10:52 -06:00
Ming Lei 1f236ab22c blk-mq: bitmap tag: cleanup blk_mq_init_tags
Both nr_cache and nr_tags arn't needed for bitmap tag anymore.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-10 15:44:00 -06:00
Ming Lei 9d3d21aeb4 blk-mq: bitmap tag: select random tag betweet 0 and (depth - 1)
The selected tag should be selected at random between 0 and
(depth - 1) with probability 1/depth, instead between 0 and
(depth - 2) with probability 1/(depth - 1).

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-10 15:43:14 -06:00
Ming Lei 60f2df8a29 blk-mq: bitmap tag: remove barrier in bt_clear_tag()
The barrier isn't necessary because both atomic_dec_and_test()
and wake_up() implicate one barrier.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-10 15:42:13 -06:00
Ming Lei 0289b2e110 blk-mq: bitmap tag: use clear_bit_unlock in bt_clear_tag()
The unlock memory barrier need to order access to req in free
path and clearing tag bit, otherwise either request free path
may see a allocated request, or initialized request in allocate
path might be modified by the ongoing free path.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-10 15:41:42 -06:00
Jens Axboe 59d13bf5f5 blk-mq: use sparser tag layout for lower queue depth
For best performance, spreading tags over multiple cachelines
makes the tagging more efficient on multicore systems. But since
we have 8 * sizeof(unsigned long) tags per cacheline, we don't
always get a nice spread.

Attempt to spread the tags over at least 4 cachelines, using fewer
number of bits per unsigned long if we have to. This improves
tagging performance in setups with 32-128 tags. For higher depths,
the spread is the same as before (BITS_PER_LONG tags per cacheline).

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-09 13:41:15 -06:00
Jens Axboe 4bb659b156 blk-mq: implement new and more efficient tagging scheme
blk-mq currently uses percpu_ida for tag allocation. But that only
works well if the ratio between tag space and number of CPUs is
sufficiently high. For most devices and systems, that is not the
case. The end result if that we either only utilize the tag space
partially, or we end up attempting to fully exhaust it and run
into lots of lock contention with stealing between CPUs. This is
not optimal.

This new tagging scheme is a hybrid bitmap allocator. It uses
two tricks to both be SMP friendly and allow full exhaustion
of the space:

1) We cache the last allocated (or freed) tag on a per blk-mq
   software context basis. This allows us to limit the space
   we have to search. The key element here is not caching it
   in the shared tag structure, otherwise we end up dirtying
   more shared cache lines on each allocate/free operation.

2) The tag space is split into cache line sized groups, and
   each context will start off randomly in that space. Even up
   to full utilization of the space, this divides the tag users
   efficiently into cache line groups, avoiding dirtying the same
   one both between allocators and between allocator and freeer.

This scheme shows drastically better behaviour, both on small
tag spaces but on large ones as well. It has been tested extensively
to show better performance for all the cases blk-mq cares about.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-05-09 09:36:49 -06:00
Jens Axboe 5810d903fa blk-mq: fix waiting for reserved tags
blk_mq_wait_for_tags() is only able to wait for "normal" tags,
not reserved tags. Pass in which one we should attempt to get
a tag for, so that waiting for reserved tags will work.

Reserved tags are used for internal commands, which are usually
serialized. Hence no waiting generally takes place, but we should
ensure that it actually works if users need that functionality.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-04-29 20:49:48 -06:00
Christoph Hellwig 24d2f90309 blk-mq: split out tag initialization, support shared tags
Add a new blk_mq_tag_set structure that gets set up before we initialize
the queue.  A single blk_mq_tag_set structure can be shared by multiple
queues.

Signed-off-by: Christoph Hellwig <hch@lst.de>

Modular export of blk_mq_{alloc,free}_tagset added by me.

Signed-off-by: Jens Axboe <axboe@fb.com>
2014-04-15 14:18:02 -06:00
Linus Torvalds 5e57dc8110 Merge branch 'for-linus' of git://git.kernel.dk/linux-block
Pull block IO fixes from Jens Axboe:
 "Second round of updates and fixes for 3.14-rc2.  Most of this stuff
  has been queued up for a while.  The notable exception is the blk-mq
  changes, which are naturally a bit more in flux still.

  The pull request contains:

   - Two bug fixes for the new immutable vecs, causing crashes with raid
     or swap.  From Kent.

   - Various blk-mq tweaks and fixes from Christoph.  A fix for
     integrity bio's from Nic.

   - A few bcache fixes from Kent and Darrick Wong.

   - xen-blk{front,back} fixes from David Vrabel, Matt Rushton, Nicolas
     Swenson, and Roger Pau Monne.

   - Fix for a vec miscount with integrity vectors from Martin.

   - Minor annotations or fixes from Masanari Iida and Rashika Kheria.

   - Tweak to null_blk to do more normal FIFO processing of requests
     from Shlomo Pongratz.

   - Elevator switching bypass fix from Tejun.

   - Softlockup in blkdev_issue_discard() fix when !CONFIG_PREEMPT from
     me"

* 'for-linus' of git://git.kernel.dk/linux-block: (31 commits)
  block: add cond_resched() to potentially long running ioctl discard loop
  xen-blkback: init persistent_purge_work work_struct
  blk-mq: pair blk_mq_start_request / blk_mq_requeue_request
  blk-mq: dont assume rq->errors is set when returning an error from ->queue_rq
  block: Fix cloning of discard/write same bios
  block: Fix type mismatch in ssize_t_blk_mq_tag_sysfs_show
  blk-mq: rework flush sequencing logic
  null_blk: use blk_complete_request and blk_mq_complete_request
  virtio_blk: use blk_mq_complete_request
  blk-mq: rework I/O completions
  fs: Add prototype declaration to appropriate header file include/linux/bio.h
  fs: Mark function as static in fs/bio-integrity.c
  block/null_blk: Fix completion processing from LIFO to FIFO
  block: Explicitly handle discard/write same segments
  block: Fix nr_vecs for inline integrity vectors
  blk-mq: Add bio_integrity setup to blk_mq_make_request
  blk-mq: initialize sg_reserved_size
  blk-mq: handle dma_drain_size
  blk-mq: divert __blk_put_request for MQ ops
  blk-mq: support at_head inserations for blk_execute_rq
  ...
2014-02-14 10:45:18 -08:00
Masanari Iida 11c9444407 block: Fix type mismatch in ssize_t_blk_mq_tag_sysfs_show
cppcheck detected following format string mismatch.
[blk-mq-tag.c:201]: (warning) %u in format string (no. 1) requires
'unsigned int' but the argument type is 'int'.

Change "cpu" from int to unsigned int, because the cpu
never become minus value.

Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
2014-02-10 10:39:53 -07:00
Kent Overstreet 6f6b5d1ec5 percpu_ida: Make percpu_ida_alloc + callers accept task state bitmask
This patch changes percpu_ida_alloc() + callers to accept task state
bitmask for prepare_to_wait() for code like target/iscsi that needs
it for interruptible sleep, that is provided in a subsequent patch.

It now expects TASK_UNINTERRUPTIBLE when the caller is able to sleep
waiting for a new tag, or TASK_RUNNING when the caller cannot sleep,
and is forced to return a negative value when no tags are available.

v2 changes:
  - Include blk-mq + tcm_fc + vhost/scsi + target/iscsi changes
  - Drop signal_pending_state() call
v3 changes:
  - Only call prepare_to_wait() + finish_wait() when != TASK_RUNNING
    (PeterZ)

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: <stable@vger.kernel.org> #3.12+
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
2014-01-23 20:17:18 +00:00
Jens Axboe 320ae51fee blk-mq: new multi-queue block IO queueing mechanism
Linux currently has two models for block devices:

- The classic request_fn based approach, where drivers use struct
  request units for IO. The block layer provides various helper
  functionalities to let drivers share code, things like tag
  management, timeout handling, queueing, etc.

- The "stacked" approach, where a driver squeezes in between the
  block layer and IO submitter. Since this bypasses the IO stack,
  driver generally have to manage everything themselves.

With drivers being written for new high IOPS devices, the classic
request_fn based driver doesn't work well enough. The design dates
back to when both SMP and high IOPS was rare. It has problems with
scaling to bigger machines, and runs into scaling issues even on
smaller machines when you have IOPS in the hundreds of thousands
per device.

The stacked approach is then most often selected as the model
for the driver. But this means that everybody has to re-invent
everything, and along with that we get all the problems again
that the shared approach solved.

This commit introduces blk-mq, block multi queue support. The
design is centered around per-cpu queues for queueing IO, which
then funnel down into x number of hardware submission queues.
We might have a 1:1 mapping between the two, or it might be
an N:M mapping. That all depends on what the hardware supports.

blk-mq provides various helper functions, which include:

- Scalable support for request tagging. Most devices need to
  be able to uniquely identify a request both in the driver and
  to the hardware. The tagging uses per-cpu caches for freed
  tags, to enable cache hot reuse.

- Timeout handling without tracking request on a per-device
  basis. Basically the driver should be able to get a notification,
  if a request happens to fail.

- Optional support for non 1:1 mappings between issue and
  submission queues. blk-mq can redirect IO completions to the
  desired location.

- Support for per-request payloads. Drivers almost always need
  to associate a request structure with some driver private
  command structure. Drivers can tell blk-mq this at init time,
  and then any request handed to the driver will have the
  required size of memory associated with it.

- Support for merging of IO, and plugging. The stacked model
  gets neither of these. Even for high IOPS devices, merging
  sequential IO reduces per-command overhead and thus
  increases bandwidth.

For now, this is provided as a potential 3rd queueing model, with
the hope being that, as it matures, it can replace both the classic
and stacked model. That would get us back to having just 1 real
model for block devices, leaving the stacked approach to dm/md
devices (as it was originally intended).

Contributions in this patch from the following people:

Shaohua Li <shli@fusionio.com>
Alexander Gordeev <agordeev@redhat.com>
Christoph Hellwig <hch@infradead.org>
Mike Christie <michaelc@cs.wisc.edu>
Matias Bjorling <m@bjorling.me>
Jeff Moyer <jmoyer@redhat.com>

Acked-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2013-10-25 11:56:00 +01:00