qemu-e2k/block/file-posix.c

3910 lines
112 KiB
C
Raw Normal View History

/*
* Block driver for RAW files (posix)
*
* Copyright (c) 2006 Fabrice Bellard
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "qemu/osdep.h"
#include "qemu-common.h"
2016-03-14 09:01:28 +01:00
#include "qapi/error.h"
#include "qemu/cutils.h"
#include "qemu/error-report.h"
#include "block/block_int.h"
#include "qemu/module.h"
#include "qemu/option.h"
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
#include "qemu/units.h"
#include "trace.h"
#include "block/thread-pool.h"
#include "qemu/iov.h"
#include "block/raw-aio.h"
#include "qapi/qmp/qdict.h"
#include "qapi/qmp/qstring.h"
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
#include "scsi/pr-manager.h"
#include "scsi/constants.h"
#if defined(__APPLE__) && (__MACH__)
#include <paths.h>
#include <sys/param.h>
#include <IOKit/IOKitLib.h>
#include <IOKit/IOBSD.h>
#include <IOKit/storage/IOMediaBSDClient.h>
#include <IOKit/storage/IOMedia.h>
#include <IOKit/storage/IOCDMedia.h>
//#include <IOKit/storage/IOCDTypes.h>
#include <IOKit/storage/IODVDMedia.h>
#include <CoreFoundation/CoreFoundation.h>
#endif
#ifdef __sun__
#define _POSIX_PTHREAD_SEMANTICS 1
#include <sys/dkio.h>
#endif
#ifdef __linux__
#include <sys/ioctl.h>
#include <sys/param.h>
#include <sys/syscall.h>
#include <sys/vfs.h>
#include <linux/cdrom.h>
#include <linux/fd.h>
#include <linux/fs.h>
#include <linux/hdreg.h>
#include <linux/magic.h>
#include <scsi/sg.h>
#ifdef __s390__
#include <asm/dasd.h>
#endif
#ifndef FS_NOCOW_FL
#define FS_NOCOW_FL 0x00800000 /* Do not cow file */
#endif
#endif
#if defined(CONFIG_FALLOCATE_PUNCH_HOLE) || defined(CONFIG_FALLOCATE_ZERO_RANGE)
#include <linux/falloc.h>
#endif
#if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
#include <sys/disk.h>
#include <sys/cdio.h>
#endif
#ifdef __OpenBSD__
#include <sys/ioctl.h>
#include <sys/disklabel.h>
#include <sys/dkio.h>
#endif
#ifdef __NetBSD__
#include <sys/ioctl.h>
#include <sys/disklabel.h>
#include <sys/dkio.h>
#include <sys/disk.h>
#endif
#ifdef __DragonFly__
#include <sys/ioctl.h>
#include <sys/diskslice.h>
#endif
#ifdef CONFIG_XFS
#include <xfs/xfs.h>
#endif
#include "trace.h"
/* OS X does not have O_DSYNC */
#ifndef O_DSYNC
#ifdef O_SYNC
#define O_DSYNC O_SYNC
#elif defined(O_FSYNC)
#define O_DSYNC O_FSYNC
#endif
#endif
/* Approximate O_DIRECT with O_DSYNC if O_DIRECT isn't available */
#ifndef O_DIRECT
#define O_DIRECT O_DSYNC
#endif
#define FTYPE_FILE 0
#define FTYPE_CD 1
#define MAX_BLOCKSIZE 4096
/* Posix file locking bytes. Libvirt takes byte 0, we start from higher bytes,
* leaving a few more bytes for its future use. */
#define RAW_LOCK_PERM_BASE 100
#define RAW_LOCK_SHARED_BASE 200
typedef struct BDRVRawState {
int fd;
bool use_lock;
int type;
int open_flags;
size_t buf_align;
/* The current permissions. */
uint64_t perm;
uint64_t shared_perm;
/* The perms bits whose corresponding bytes are already locked in
* s->fd. */
uint64_t locked_perm;
uint64_t locked_shared_perm;
int perm_change_fd;
int perm_change_flags;
BDRVReopenState *reopen_state;
#ifdef CONFIG_XFS
bool is_xfs:1;
#endif
bool has_discard:1;
bool has_write_zeroes:1;
bool discard_zeroes:1;
bool use_linux_aio:1;
bool use_linux_io_uring:1;
bool page_cache_inconsistent:1;
bool has_fallocate;
bool needs_alignment;
bool drop_cache;
bool check_cache_dropped;
struct {
uint64_t discard_nb_ok;
uint64_t discard_nb_failed;
uint64_t discard_bytes_ok;
} stats;
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
PRManager *pr_mgr;
} BDRVRawState;
typedef struct BDRVRawReopenState {
int fd;
int open_flags;
bool drop_cache;
bool check_cache_dropped;
} BDRVRawReopenState;
static int fd_open(BlockDriverState *bs);
static int64_t raw_getlength(BlockDriverState *bs);
typedef struct RawPosixAIOData {
BlockDriverState *bs;
int aio_type;
int aio_fildes;
off_t aio_offset;
uint64_t aio_nbytes;
union {
struct {
struct iovec *iov;
int niov;
} io;
struct {
uint64_t cmd;
void *buf;
} ioctl;
struct {
int aio_fd2;
off_t aio_offset2;
} copy_range;
struct {
PreallocMode prealloc;
Error **errp;
} truncate;
};
} RawPosixAIOData;
#if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
static int cdrom_reopen(BlockDriverState *bs);
#endif
/*
* Elide EAGAIN and EACCES details when failing to lock, as this
* indicates that the specified file region is already locked by
* another process, which is considered a common scenario.
*/
#define raw_lock_error_setg_errno(errp, err, fmt, ...) \
do { \
if ((err) == EAGAIN || (err) == EACCES) { \
error_setg((errp), (fmt), ## __VA_ARGS__); \
} else { \
error_setg_errno((errp), (err), (fmt), ## __VA_ARGS__); \
} \
} while (0)
#if defined(__NetBSD__)
static int raw_normalize_devicepath(const char **filename, Error **errp)
{
static char namebuf[PATH_MAX];
const char *dp, *fname;
struct stat sb;
fname = *filename;
dp = strrchr(fname, '/');
if (lstat(fname, &sb) < 0) {
error_setg_file_open(errp, errno, fname);
return -errno;
}
if (!S_ISBLK(sb.st_mode)) {
return 0;
}
if (dp == NULL) {
snprintf(namebuf, PATH_MAX, "r%s", fname);
} else {
snprintf(namebuf, PATH_MAX, "%.*s/r%s",
(int)(dp - fname), fname, dp + 1);
}
*filename = namebuf;
warn_report("%s is a block device, using %s", fname, *filename);
return 0;
}
#else
static int raw_normalize_devicepath(const char **filename, Error **errp)
{
return 0;
}
#endif
/*
* Get logical block size via ioctl. On success store it in @sector_size_p.
*/
static int probe_logical_blocksize(int fd, unsigned int *sector_size_p)
{
unsigned int sector_size;
bool success = false;
int i;
errno = ENOTSUP;
static const unsigned long ioctl_list[] = {
#ifdef BLKSSZGET
BLKSSZGET,
#endif
#ifdef DKIOCGETBLOCKSIZE
DKIOCGETBLOCKSIZE,
#endif
#ifdef DIOCGSECTORSIZE
DIOCGSECTORSIZE,
#endif
};
/* Try a few ioctls to get the right size */
for (i = 0; i < (int)ARRAY_SIZE(ioctl_list); i++) {
if (ioctl(fd, ioctl_list[i], &sector_size) >= 0) {
*sector_size_p = sector_size;
success = true;
}
}
return success ? 0 : -errno;
}
/**
* Get physical block size of @fd.
* On success, store it in @blk_size and return 0.
* On failure, return -errno.
*/
static int probe_physical_blocksize(int fd, unsigned int *blk_size)
{
#ifdef BLKPBSZGET
if (ioctl(fd, BLKPBSZGET, blk_size) < 0) {
return -errno;
}
return 0;
#else
return -ENOTSUP;
#endif
}
/*
* Returns true if no alignment restrictions are necessary even for files
* opened with O_DIRECT.
*
* raw_probe_alignment() probes the required alignment and assume that 1 means
* the probing failed, so it falls back to a safe default of 4k. This can be
* avoided if we know that byte alignment is okay for the file.
*/
static bool dio_byte_aligned(int fd)
{
#ifdef __linux__
struct statfs buf;
int ret;
ret = fstatfs(fd, &buf);
if (ret == 0 && buf.f_type == NFS_SUPER_MAGIC) {
return true;
}
#endif
return false;
}
/* Check if read is allowed with given memory buffer and length.
*
* This function is used to check O_DIRECT memory buffer and request alignment.
*/
static bool raw_is_io_aligned(int fd, void *buf, size_t len)
{
ssize_t ret = pread(fd, buf, len, 0);
if (ret >= 0) {
return true;
}
#ifdef __linux__
/* The Linux kernel returns EINVAL for misaligned O_DIRECT reads. Ignore
* other errors (e.g. real I/O error), which could happen on a failed
* drive, since we only care about probing alignment.
*/
if (errno != EINVAL) {
return true;
}
#endif
return false;
}
static void raw_probe_alignment(BlockDriverState *bs, int fd, Error **errp)
{
BDRVRawState *s = bs->opaque;
char *buf;
size_t max_align = MAX(MAX_BLOCKSIZE, qemu_real_host_page_size);
file-posix: Handle undetectable alignment In some cases buf_align or request_alignment cannot be detected: 1. With Gluster, buf_align cannot be detected since the actual I/O is done on Gluster server, and qemu buffer alignment does not matter. Since we don't have alignment requirement, buf_align=1 is the best value. 2. With local XFS filesystem, buf_align cannot be detected if reading from unallocated area. In this we must align the buffer, but we don't know what is the correct size. Using the wrong alignment results in I/O error. 3. With Gluster backed by XFS, request_alignment cannot be detected if reading from unallocated area. In this case we need to use the correct alignment, and failing to do so results in I/O errors. 4. With NFS, the server does not use direct I/O, so both buf_align cannot be detected. In this case we don't need any alignment so we can use buf_align=1 and request_alignment=1. These cases seems to work when storage sector size is 512 bytes, because the current code starts checking align=512. If the check succeeds because alignment cannot be detected we use 512. But this does not work for storage with 4k sector size. To determine if we can detect the alignment, we probe first with align=1. If probing succeeds, maybe there are no alignment requirement (cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we don't have any way to tell, we treat this as undetectable alignment. If probing with align=1 fails with EINVAL, but probing with one of the expected alignments succeeds, we know that we found a working alignment. Practically the alignment requirements are the same for buffer alignment, buffer length, and offset in file. So in case we cannot detect buf_align, we can use request alignment. If we cannot detect request alignment, we can fallback to a safe value. To use this logic, we probe first request alignment instead of buf_align. Here is a table showing the behaviour with current code (the value in parenthesis is the optimal value). Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 512 (1) 4096 (4096) FAIL ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 512 (4096) 4096 (4096) FAIL ---------------------------------------------------------------------- 3 512 512 (1) 512 (512) OK 3 4096 512 (1) 512 (4096) FAIL ---------------------------------------------------------------------- 4 512 512 (1) 512 (1) OK 4 4096 512 (1) 512 (1) OK Same cases with this change: Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 4096 (4096) 4096 (4096) OK ---------------------------------------------------------------------- 3 512 4096 (1) 4096 (512) OK 3 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 4 512 4096 (1) 4096 (1) OK 4 4096 4096 (1) 4096 (1) OK I tested that provisioning VMs and copying disks on local XFS and Gluster with 4k bytes sector size work now, resolving bugs [1],[2]. I tested also on XFS, NFS, Gluster with 512 bytes sector size. [1] https://bugzilla.redhat.com/1737256 [2] https://bugzilla.redhat.com/1738657 Signed-off-by: Nir Soffer <nsoffer@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-08-13 20:21:03 +02:00
size_t alignments[] = {1, 512, 1024, 2048, 4096};
/* For SCSI generic devices the alignment is not really used.
With buffered I/O, we don't have any restrictions. */
if (bdrv_is_sg(bs) || !s->needs_alignment) {
bs->bl.request_alignment = 1;
s->buf_align = 1;
return;
}
bs->bl.request_alignment = 0;
s->buf_align = 0;
/* Let's try to use the logical blocksize for the alignment. */
if (probe_logical_blocksize(fd, &bs->bl.request_alignment) < 0) {
bs->bl.request_alignment = 0;
}
#ifdef CONFIG_XFS
if (s->is_xfs) {
struct dioattr da;
if (xfsctl(NULL, fd, XFS_IOC_DIOINFO, &da) >= 0) {
bs->bl.request_alignment = da.d_miniosz;
/* The kernel returns wrong information for d_mem */
/* s->buf_align = da.d_mem; */
}
}
#endif
file-posix: Handle undetectable alignment In some cases buf_align or request_alignment cannot be detected: 1. With Gluster, buf_align cannot be detected since the actual I/O is done on Gluster server, and qemu buffer alignment does not matter. Since we don't have alignment requirement, buf_align=1 is the best value. 2. With local XFS filesystem, buf_align cannot be detected if reading from unallocated area. In this we must align the buffer, but we don't know what is the correct size. Using the wrong alignment results in I/O error. 3. With Gluster backed by XFS, request_alignment cannot be detected if reading from unallocated area. In this case we need to use the correct alignment, and failing to do so results in I/O errors. 4. With NFS, the server does not use direct I/O, so both buf_align cannot be detected. In this case we don't need any alignment so we can use buf_align=1 and request_alignment=1. These cases seems to work when storage sector size is 512 bytes, because the current code starts checking align=512. If the check succeeds because alignment cannot be detected we use 512. But this does not work for storage with 4k sector size. To determine if we can detect the alignment, we probe first with align=1. If probing succeeds, maybe there are no alignment requirement (cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we don't have any way to tell, we treat this as undetectable alignment. If probing with align=1 fails with EINVAL, but probing with one of the expected alignments succeeds, we know that we found a working alignment. Practically the alignment requirements are the same for buffer alignment, buffer length, and offset in file. So in case we cannot detect buf_align, we can use request alignment. If we cannot detect request alignment, we can fallback to a safe value. To use this logic, we probe first request alignment instead of buf_align. Here is a table showing the behaviour with current code (the value in parenthesis is the optimal value). Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 512 (1) 4096 (4096) FAIL ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 512 (4096) 4096 (4096) FAIL ---------------------------------------------------------------------- 3 512 512 (1) 512 (512) OK 3 4096 512 (1) 512 (4096) FAIL ---------------------------------------------------------------------- 4 512 512 (1) 512 (1) OK 4 4096 512 (1) 512 (1) OK Same cases with this change: Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 4096 (4096) 4096 (4096) OK ---------------------------------------------------------------------- 3 512 4096 (1) 4096 (512) OK 3 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 4 512 4096 (1) 4096 (1) OK 4 4096 4096 (1) 4096 (1) OK I tested that provisioning VMs and copying disks on local XFS and Gluster with 4k bytes sector size work now, resolving bugs [1],[2]. I tested also on XFS, NFS, Gluster with 512 bytes sector size. [1] https://bugzilla.redhat.com/1737256 [2] https://bugzilla.redhat.com/1738657 Signed-off-by: Nir Soffer <nsoffer@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-08-13 20:21:03 +02:00
/*
* If we could not get the sizes so far, we can only guess them. First try
* to detect request alignment, since it is more likely to succeed. Then
* try to detect buf_align, which cannot be detected in some cases (e.g.
* Gluster). If buf_align cannot be detected, we fallback to the value of
* request_alignment.
*/
if (!bs->bl.request_alignment) {
int i;
size_t align;
file-posix: Handle undetectable alignment In some cases buf_align or request_alignment cannot be detected: 1. With Gluster, buf_align cannot be detected since the actual I/O is done on Gluster server, and qemu buffer alignment does not matter. Since we don't have alignment requirement, buf_align=1 is the best value. 2. With local XFS filesystem, buf_align cannot be detected if reading from unallocated area. In this we must align the buffer, but we don't know what is the correct size. Using the wrong alignment results in I/O error. 3. With Gluster backed by XFS, request_alignment cannot be detected if reading from unallocated area. In this case we need to use the correct alignment, and failing to do so results in I/O errors. 4. With NFS, the server does not use direct I/O, so both buf_align cannot be detected. In this case we don't need any alignment so we can use buf_align=1 and request_alignment=1. These cases seems to work when storage sector size is 512 bytes, because the current code starts checking align=512. If the check succeeds because alignment cannot be detected we use 512. But this does not work for storage with 4k sector size. To determine if we can detect the alignment, we probe first with align=1. If probing succeeds, maybe there are no alignment requirement (cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we don't have any way to tell, we treat this as undetectable alignment. If probing with align=1 fails with EINVAL, but probing with one of the expected alignments succeeds, we know that we found a working alignment. Practically the alignment requirements are the same for buffer alignment, buffer length, and offset in file. So in case we cannot detect buf_align, we can use request alignment. If we cannot detect request alignment, we can fallback to a safe value. To use this logic, we probe first request alignment instead of buf_align. Here is a table showing the behaviour with current code (the value in parenthesis is the optimal value). Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 512 (1) 4096 (4096) FAIL ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 512 (4096) 4096 (4096) FAIL ---------------------------------------------------------------------- 3 512 512 (1) 512 (512) OK 3 4096 512 (1) 512 (4096) FAIL ---------------------------------------------------------------------- 4 512 512 (1) 512 (1) OK 4 4096 512 (1) 512 (1) OK Same cases with this change: Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 4096 (4096) 4096 (4096) OK ---------------------------------------------------------------------- 3 512 4096 (1) 4096 (512) OK 3 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 4 512 4096 (1) 4096 (1) OK 4 4096 4096 (1) 4096 (1) OK I tested that provisioning VMs and copying disks on local XFS and Gluster with 4k bytes sector size work now, resolving bugs [1],[2]. I tested also on XFS, NFS, Gluster with 512 bytes sector size. [1] https://bugzilla.redhat.com/1737256 [2] https://bugzilla.redhat.com/1738657 Signed-off-by: Nir Soffer <nsoffer@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-08-13 20:21:03 +02:00
buf = qemu_memalign(max_align, max_align);
for (i = 0; i < ARRAY_SIZE(alignments); i++) {
align = alignments[i];
if (raw_is_io_aligned(fd, buf, align)) {
/* Fallback to safe value. */
bs->bl.request_alignment = (align != 1) ? align : max_align;
break;
}
}
qemu_vfree(buf);
}
file-posix: Handle undetectable alignment In some cases buf_align or request_alignment cannot be detected: 1. With Gluster, buf_align cannot be detected since the actual I/O is done on Gluster server, and qemu buffer alignment does not matter. Since we don't have alignment requirement, buf_align=1 is the best value. 2. With local XFS filesystem, buf_align cannot be detected if reading from unallocated area. In this we must align the buffer, but we don't know what is the correct size. Using the wrong alignment results in I/O error. 3. With Gluster backed by XFS, request_alignment cannot be detected if reading from unallocated area. In this case we need to use the correct alignment, and failing to do so results in I/O errors. 4. With NFS, the server does not use direct I/O, so both buf_align cannot be detected. In this case we don't need any alignment so we can use buf_align=1 and request_alignment=1. These cases seems to work when storage sector size is 512 bytes, because the current code starts checking align=512. If the check succeeds because alignment cannot be detected we use 512. But this does not work for storage with 4k sector size. To determine if we can detect the alignment, we probe first with align=1. If probing succeeds, maybe there are no alignment requirement (cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we don't have any way to tell, we treat this as undetectable alignment. If probing with align=1 fails with EINVAL, but probing with one of the expected alignments succeeds, we know that we found a working alignment. Practically the alignment requirements are the same for buffer alignment, buffer length, and offset in file. So in case we cannot detect buf_align, we can use request alignment. If we cannot detect request alignment, we can fallback to a safe value. To use this logic, we probe first request alignment instead of buf_align. Here is a table showing the behaviour with current code (the value in parenthesis is the optimal value). Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 512 (1) 4096 (4096) FAIL ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 512 (4096) 4096 (4096) FAIL ---------------------------------------------------------------------- 3 512 512 (1) 512 (512) OK 3 4096 512 (1) 512 (4096) FAIL ---------------------------------------------------------------------- 4 512 512 (1) 512 (1) OK 4 4096 512 (1) 512 (1) OK Same cases with this change: Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 4096 (4096) 4096 (4096) OK ---------------------------------------------------------------------- 3 512 4096 (1) 4096 (512) OK 3 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 4 512 4096 (1) 4096 (1) OK 4 4096 4096 (1) 4096 (1) OK I tested that provisioning VMs and copying disks on local XFS and Gluster with 4k bytes sector size work now, resolving bugs [1],[2]. I tested also on XFS, NFS, Gluster with 512 bytes sector size. [1] https://bugzilla.redhat.com/1737256 [2] https://bugzilla.redhat.com/1738657 Signed-off-by: Nir Soffer <nsoffer@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-08-13 20:21:03 +02:00
if (!s->buf_align) {
int i;
size_t align;
file-posix: Handle undetectable alignment In some cases buf_align or request_alignment cannot be detected: 1. With Gluster, buf_align cannot be detected since the actual I/O is done on Gluster server, and qemu buffer alignment does not matter. Since we don't have alignment requirement, buf_align=1 is the best value. 2. With local XFS filesystem, buf_align cannot be detected if reading from unallocated area. In this we must align the buffer, but we don't know what is the correct size. Using the wrong alignment results in I/O error. 3. With Gluster backed by XFS, request_alignment cannot be detected if reading from unallocated area. In this case we need to use the correct alignment, and failing to do so results in I/O errors. 4. With NFS, the server does not use direct I/O, so both buf_align cannot be detected. In this case we don't need any alignment so we can use buf_align=1 and request_alignment=1. These cases seems to work when storage sector size is 512 bytes, because the current code starts checking align=512. If the check succeeds because alignment cannot be detected we use 512. But this does not work for storage with 4k sector size. To determine if we can detect the alignment, we probe first with align=1. If probing succeeds, maybe there are no alignment requirement (cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we don't have any way to tell, we treat this as undetectable alignment. If probing with align=1 fails with EINVAL, but probing with one of the expected alignments succeeds, we know that we found a working alignment. Practically the alignment requirements are the same for buffer alignment, buffer length, and offset in file. So in case we cannot detect buf_align, we can use request alignment. If we cannot detect request alignment, we can fallback to a safe value. To use this logic, we probe first request alignment instead of buf_align. Here is a table showing the behaviour with current code (the value in parenthesis is the optimal value). Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 512 (1) 4096 (4096) FAIL ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 512 (4096) 4096 (4096) FAIL ---------------------------------------------------------------------- 3 512 512 (1) 512 (512) OK 3 4096 512 (1) 512 (4096) FAIL ---------------------------------------------------------------------- 4 512 512 (1) 512 (1) OK 4 4096 512 (1) 512 (1) OK Same cases with this change: Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 4096 (4096) 4096 (4096) OK ---------------------------------------------------------------------- 3 512 4096 (1) 4096 (512) OK 3 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 4 512 4096 (1) 4096 (1) OK 4 4096 4096 (1) 4096 (1) OK I tested that provisioning VMs and copying disks on local XFS and Gluster with 4k bytes sector size work now, resolving bugs [1],[2]. I tested also on XFS, NFS, Gluster with 512 bytes sector size. [1] https://bugzilla.redhat.com/1737256 [2] https://bugzilla.redhat.com/1738657 Signed-off-by: Nir Soffer <nsoffer@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-08-13 20:21:03 +02:00
buf = qemu_memalign(max_align, 2 * max_align);
for (i = 0; i < ARRAY_SIZE(alignments); i++) {
align = alignments[i];
if (raw_is_io_aligned(fd, buf + align, max_align)) {
/* Fallback to request_alignment. */
file-posix: Handle undetectable alignment In some cases buf_align or request_alignment cannot be detected: 1. With Gluster, buf_align cannot be detected since the actual I/O is done on Gluster server, and qemu buffer alignment does not matter. Since we don't have alignment requirement, buf_align=1 is the best value. 2. With local XFS filesystem, buf_align cannot be detected if reading from unallocated area. In this we must align the buffer, but we don't know what is the correct size. Using the wrong alignment results in I/O error. 3. With Gluster backed by XFS, request_alignment cannot be detected if reading from unallocated area. In this case we need to use the correct alignment, and failing to do so results in I/O errors. 4. With NFS, the server does not use direct I/O, so both buf_align cannot be detected. In this case we don't need any alignment so we can use buf_align=1 and request_alignment=1. These cases seems to work when storage sector size is 512 bytes, because the current code starts checking align=512. If the check succeeds because alignment cannot be detected we use 512. But this does not work for storage with 4k sector size. To determine if we can detect the alignment, we probe first with align=1. If probing succeeds, maybe there are no alignment requirement (cases 1, 4) or we are probing unallocated area (cases 2, 3). Since we don't have any way to tell, we treat this as undetectable alignment. If probing with align=1 fails with EINVAL, but probing with one of the expected alignments succeeds, we know that we found a working alignment. Practically the alignment requirements are the same for buffer alignment, buffer length, and offset in file. So in case we cannot detect buf_align, we can use request alignment. If we cannot detect request alignment, we can fallback to a safe value. To use this logic, we probe first request alignment instead of buf_align. Here is a table showing the behaviour with current code (the value in parenthesis is the optimal value). Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 512 (1) 4096 (4096) FAIL ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 512 (4096) 4096 (4096) FAIL ---------------------------------------------------------------------- 3 512 512 (1) 512 (512) OK 3 4096 512 (1) 512 (4096) FAIL ---------------------------------------------------------------------- 4 512 512 (1) 512 (1) OK 4 4096 512 (1) 512 (1) OK Same cases with this change: Case Sector buf_align (opt) request_alignment (opt) result ====================================================================== 1 512 512 (1) 512 (512) OK 1 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 2 512 512 (512) 512 (512) OK 2 4096 4096 (4096) 4096 (4096) OK ---------------------------------------------------------------------- 3 512 4096 (1) 4096 (512) OK 3 4096 4096 (1) 4096 (4096) OK ---------------------------------------------------------------------- 4 512 4096 (1) 4096 (1) OK 4 4096 4096 (1) 4096 (1) OK I tested that provisioning VMs and copying disks on local XFS and Gluster with 4k bytes sector size work now, resolving bugs [1],[2]. I tested also on XFS, NFS, Gluster with 512 bytes sector size. [1] https://bugzilla.redhat.com/1737256 [2] https://bugzilla.redhat.com/1738657 Signed-off-by: Nir Soffer <nsoffer@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-08-13 20:21:03 +02:00
s->buf_align = (align != 1) ? align : bs->bl.request_alignment;
break;
}
}
qemu_vfree(buf);
}
if (!s->buf_align || !bs->bl.request_alignment) {
error_setg(errp, "Could not find working O_DIRECT alignment");
error_append_hint(errp, "Try cache.direct=off\n");
}
}
static int check_hdev_writable(int fd)
{
#if defined(BLKROGET)
/* Linux block devices can be configured "read-only" using blockdev(8).
* This is independent of device node permissions and therefore open(2)
* with O_RDWR succeeds. Actual writes fail with EPERM.
*
* bdrv_open() is supposed to fail if the disk is read-only. Explicitly
* check for read-only block devices so that Linux block devices behave
* properly.
*/
struct stat st;
int readonly = 0;
if (fstat(fd, &st)) {
return -errno;
}
if (!S_ISBLK(st.st_mode)) {
return 0;
}
if (ioctl(fd, BLKROGET, &readonly) < 0) {
return -errno;
}
if (readonly) {
return -EACCES;
}
#endif /* defined(BLKROGET) */
return 0;
}
static void raw_parse_flags(int bdrv_flags, int *open_flags, bool has_writers)
{
bool read_write = false;
assert(open_flags != NULL);
*open_flags |= O_BINARY;
*open_flags &= ~O_ACCMODE;
if (bdrv_flags & BDRV_O_AUTO_RDONLY) {
read_write = has_writers;
} else if (bdrv_flags & BDRV_O_RDWR) {
read_write = true;
}
if (read_write) {
*open_flags |= O_RDWR;
} else {
*open_flags |= O_RDONLY;
}
/* Use O_DSYNC for write-through caching, no flags for write-back caching,
* and O_DIRECT for no caching. */
if ((bdrv_flags & BDRV_O_NOCACHE)) {
*open_flags |= O_DIRECT;
}
}
static void raw_parse_filename(const char *filename, QDict *options,
Error **errp)
{
2017-05-22 21:52:16 +02:00
bdrv_parse_filename_strip_prefix(filename, "file:", options);
}
static QemuOptsList raw_runtime_opts = {
.name = "raw",
.head = QTAILQ_HEAD_INITIALIZER(raw_runtime_opts.head),
.desc = {
{
.name = "filename",
.type = QEMU_OPT_STRING,
.help = "File name of the image",
},
{
.name = "aio",
.type = QEMU_OPT_STRING,
.help = "host AIO implementation (threads, native, io_uring)",
},
{
.name = "locking",
.type = QEMU_OPT_STRING,
.help = "file locking mode (on/off/auto, default: auto)",
},
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
{
.name = "pr-manager",
.type = QEMU_OPT_STRING,
.help = "id of persistent reservation manager object (default: none)",
},
#if defined(__linux__)
{
.name = "drop-cache",
.type = QEMU_OPT_BOOL,
.help = "invalidate page cache during live migration (default: on)",
},
#endif
{
.name = "x-check-cache-dropped",
.type = QEMU_OPT_BOOL,
.help = "check that page cache was dropped on live migration (default: off)"
},
{ /* end of list */ }
},
};
static const char *const mutable_opts[] = { "x-check-cache-dropped", NULL };
static int raw_open_common(BlockDriverState *bs, QDict *options,
int bdrv_flags, int open_flags,
bool device, Error **errp)
{
BDRVRawState *s = bs->opaque;
QemuOpts *opts;
Error *local_err = NULL;
const char *filename = NULL;
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
const char *str;
BlockdevAioOptions aio, aio_default;
int fd, ret;
struct stat st;
OnOffAuto locking;
opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort);
if (!qemu_opts_absorb_qdict(opts, options, errp)) {
ret = -EINVAL;
goto fail;
}
filename = qemu_opt_get(opts, "filename");
ret = raw_normalize_devicepath(&filename, errp);
if (ret != 0) {
goto fail;
}
if (bdrv_flags & BDRV_O_NATIVE_AIO) {
aio_default = BLOCKDEV_AIO_OPTIONS_NATIVE;
#ifdef CONFIG_LINUX_IO_URING
} else if (bdrv_flags & BDRV_O_IO_URING) {
aio_default = BLOCKDEV_AIO_OPTIONS_IO_URING;
#endif
} else {
aio_default = BLOCKDEV_AIO_OPTIONS_THREADS;
}
aio = qapi_enum_parse(&BlockdevAioOptions_lookup,
qemu_opt_get(opts, "aio"),
aio_default, &local_err);
if (local_err) {
error_propagate(errp, local_err);
ret = -EINVAL;
goto fail;
}
s->use_linux_aio = (aio == BLOCKDEV_AIO_OPTIONS_NATIVE);
#ifdef CONFIG_LINUX_IO_URING
s->use_linux_io_uring = (aio == BLOCKDEV_AIO_OPTIONS_IO_URING);
#endif
locking = qapi_enum_parse(&OnOffAuto_lookup,
qemu_opt_get(opts, "locking"),
ON_OFF_AUTO_AUTO, &local_err);
if (local_err) {
error_propagate(errp, local_err);
ret = -EINVAL;
goto fail;
}
switch (locking) {
case ON_OFF_AUTO_ON:
s->use_lock = true;
if (!qemu_has_ofd_lock()) {
warn_report("File lock requested but OFD locking syscall is "
"unavailable, falling back to POSIX file locks");
error_printf("Due to the implementation, locks can be lost "
"unexpectedly.\n");
}
break;
case ON_OFF_AUTO_OFF:
s->use_lock = false;
break;
case ON_OFF_AUTO_AUTO:
s->use_lock = qemu_has_ofd_lock();
break;
default:
abort();
}
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
str = qemu_opt_get(opts, "pr-manager");
if (str) {
s->pr_mgr = pr_manager_lookup(str, &local_err);
if (local_err) {
error_propagate(errp, local_err);
ret = -EINVAL;
goto fail;
}
}
s->drop_cache = qemu_opt_get_bool(opts, "drop-cache", true);
s->check_cache_dropped = qemu_opt_get_bool(opts, "x-check-cache-dropped",
false);
s->open_flags = open_flags;
raw_parse_flags(bdrv_flags, &s->open_flags, false);
s->fd = -1;
fd = qemu_open(filename, s->open_flags, errp);
ret = fd < 0 ? -errno : 0;
if (ret < 0) {
if (ret == -EROFS) {
ret = -EACCES;
}
goto fail;
}
s->fd = fd;
raw-posix: refactor AIO support Currently the raw-posix.c code contains a lot of knowledge about the asynchronous I/O scheme that is mostly implemented in posix-aio-compat.c. All this code does not really belong here and is getting a bit in the way of implementing native AIO on Linux. So instead move all the guts of the AIO implementation into posix-aio-compat.c (which might need a better name, btw). There's now a very small interface between the AIO providers and raw-posix.c: - an init routine is called from raw_open_common to return an AIO context for this drive. An AIO implementation may either re-use one context for all drives, or use a different one for each as the Linux native AIO support will do. - an submit routine is called from the aio_reav/writev methods to submit an AIO request There are no indirect calls involved in this interface as we need to decide which one to call manually. We will only call the Linux AIO native init function if we were requested to by vl.c, and we will only call the native submit function if we are asked to and the request is properly aligned. That's also the reason why the alignment check actually does the inverse move and now goes into raw-posix.c. The old posix-aio-compat.h headers is removed now that most of it's content is private to posix-aio-compat.c, and instead we add a new block/raw-posix-aio.h headers is created containing only the tiny interface between raw-posix.c and the AIO implementation. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
2009-08-20 16:58:19 +02:00
/* Check s->open_flags rather than bdrv_flags due to auto-read-only */
if (s->open_flags & O_RDWR) {
ret = check_hdev_writable(s->fd);
if (ret < 0) {
error_setg_errno(errp, -ret, "The device is not writable");
goto fail;
}
}
s->perm = 0;
s->shared_perm = BLK_PERM_ALL;
#ifdef CONFIG_LINUX_AIO
/* Currently Linux does AIO only for files opened with O_DIRECT */
if (s->use_linux_aio) {
if (!(s->open_flags & O_DIRECT)) {
error_setg(errp, "aio=native was specified, but it requires "
"cache.direct=on, which was not specified.");
ret = -EINVAL;
goto fail;
}
if (!aio_setup_linux_aio(bdrv_get_aio_context(bs), errp)) {
error_prepend(errp, "Unable to use native AIO: ");
goto fail;
}
}
#else
if (s->use_linux_aio) {
error_setg(errp, "aio=native was specified, but is not supported "
"in this build.");
ret = -EINVAL;
goto fail;
}
#endif /* !defined(CONFIG_LINUX_AIO) */
raw-posix: refactor AIO support Currently the raw-posix.c code contains a lot of knowledge about the asynchronous I/O scheme that is mostly implemented in posix-aio-compat.c. All this code does not really belong here and is getting a bit in the way of implementing native AIO on Linux. So instead move all the guts of the AIO implementation into posix-aio-compat.c (which might need a better name, btw). There's now a very small interface between the AIO providers and raw-posix.c: - an init routine is called from raw_open_common to return an AIO context for this drive. An AIO implementation may either re-use one context for all drives, or use a different one for each as the Linux native AIO support will do. - an submit routine is called from the aio_reav/writev methods to submit an AIO request There are no indirect calls involved in this interface as we need to decide which one to call manually. We will only call the Linux AIO native init function if we were requested to by vl.c, and we will only call the native submit function if we are asked to and the request is properly aligned. That's also the reason why the alignment check actually does the inverse move and now goes into raw-posix.c. The old posix-aio-compat.h headers is removed now that most of it's content is private to posix-aio-compat.c, and instead we add a new block/raw-posix-aio.h headers is created containing only the tiny interface between raw-posix.c and the AIO implementation. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
2009-08-20 16:58:19 +02:00
#ifdef CONFIG_LINUX_IO_URING
if (s->use_linux_io_uring) {
if (!aio_setup_linux_io_uring(bdrv_get_aio_context(bs), errp)) {
error_prepend(errp, "Unable to use io_uring: ");
goto fail;
}
}
#else
if (s->use_linux_io_uring) {
error_setg(errp, "aio=io_uring was specified, but is not supported "
"in this build.");
ret = -EINVAL;
goto fail;
}
#endif /* !defined(CONFIG_LINUX_IO_URING) */
s->has_discard = true;
s->has_write_zeroes = true;
if ((bs->open_flags & BDRV_O_NOCACHE) != 0 && !dio_byte_aligned(s->fd)) {
s->needs_alignment = true;
}
if (fstat(s->fd, &st) < 0) {
ret = -errno;
error_setg_errno(errp, errno, "Could not stat file");
goto fail;
}
if (!device) {
if (S_ISBLK(st.st_mode)) {
warn_report("Opening a block device as a file using the '%s' "
"driver is deprecated", bs->drv->format_name);
} else if (S_ISCHR(st.st_mode)) {
warn_report("Opening a character device as a file using the '%s' "
"driver is deprecated", bs->drv->format_name);
} else if (!S_ISREG(st.st_mode)) {
error_setg(errp, "A regular file was expected by the '%s' driver, "
"but something else was given", bs->drv->format_name);
ret = -EINVAL;
goto fail;
} else {
s->discard_zeroes = true;
s->has_fallocate = true;
}
} else {
if (!(S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) {
error_setg(errp, "'%s' driver expects either "
"a character or block device", bs->drv->format_name);
ret = -EINVAL;
goto fail;
}
}
if (S_ISBLK(st.st_mode)) {
#ifdef BLKDISCARDZEROES
unsigned int arg;
if (ioctl(s->fd, BLKDISCARDZEROES, &arg) == 0 && arg) {
s->discard_zeroes = true;
}
#endif
#ifdef __linux__
/* On Linux 3.10, BLKDISCARD leaves stale data in the page cache. Do
* not rely on the contents of discarded blocks unless using O_DIRECT.
* Same for BLKZEROOUT.
*/
if (!(bs->open_flags & BDRV_O_NOCACHE)) {
s->discard_zeroes = false;
s->has_write_zeroes = false;
}
#endif
}
#ifdef __FreeBSD__
if (S_ISCHR(st.st_mode)) {
/*
* The file is a char device (disk), which on FreeBSD isn't behind
* a pager, so force all requests to be aligned. This is needed
* so QEMU makes sure all IO operations on the device are aligned
* to sector size, or else FreeBSD will reject them with EINVAL.
*/
s->needs_alignment = true;
}
#endif
#ifdef CONFIG_XFS
if (platform_test_xfs_fd(s->fd)) {
s->is_xfs = true;
}
#endif
bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK;
if (S_ISREG(st.st_mode)) {
/* When extending regular files, we get zeros from the OS */
bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
}
ret = 0;
fail:
if (ret < 0 && s->fd != -1) {
qemu_close(s->fd);
}
if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) {
unlink(filename);
}
qemu_opts_del(opts);
return ret;
}
static int raw_open(BlockDriverState *bs, QDict *options, int flags,
Error **errp)
{
BDRVRawState *s = bs->opaque;
s->type = FTYPE_FILE;
return raw_open_common(bs, options, flags, 0, false, errp);
}
typedef enum {
RAW_PL_PREPARE,
RAW_PL_COMMIT,
RAW_PL_ABORT,
} RawPermLockOp;
#define PERM_FOREACH(i) \
for ((i) = 0; (1ULL << (i)) <= BLK_PERM_ALL; i++)
/* Lock bytes indicated by @perm_lock_bits and @shared_perm_lock_bits in the
* file; if @unlock == true, also unlock the unneeded bytes.
* @shared_perm_lock_bits is the mask of all permissions that are NOT shared.
*/
static int raw_apply_lock_bytes(BDRVRawState *s, int fd,
uint64_t perm_lock_bits,
uint64_t shared_perm_lock_bits,
bool unlock, Error **errp)
{
int ret;
int i;
uint64_t locked_perm, locked_shared_perm;
if (s) {
locked_perm = s->locked_perm;
locked_shared_perm = s->locked_shared_perm;
} else {
/*
* We don't have the previous bits, just lock/unlock for each of the
* requested bits.
*/
if (unlock) {
locked_perm = BLK_PERM_ALL;
locked_shared_perm = BLK_PERM_ALL;
} else {
locked_perm = 0;
locked_shared_perm = 0;
}
}
PERM_FOREACH(i) {
int off = RAW_LOCK_PERM_BASE + i;
uint64_t bit = (1ULL << i);
if ((perm_lock_bits & bit) && !(locked_perm & bit)) {
ret = qemu_lock_fd(fd, off, 1, false);
if (ret) {
raw_lock_error_setg_errno(errp, -ret, "Failed to lock byte %d",
off);
return ret;
} else if (s) {
s->locked_perm |= bit;
}
} else if (unlock && (locked_perm & bit) && !(perm_lock_bits & bit)) {
ret = qemu_unlock_fd(fd, off, 1);
if (ret) {
error_setg_errno(errp, -ret, "Failed to unlock byte %d", off);
return ret;
} else if (s) {
s->locked_perm &= ~bit;
}
}
}
PERM_FOREACH(i) {
int off = RAW_LOCK_SHARED_BASE + i;
uint64_t bit = (1ULL << i);
if ((shared_perm_lock_bits & bit) && !(locked_shared_perm & bit)) {
ret = qemu_lock_fd(fd, off, 1, false);
if (ret) {
raw_lock_error_setg_errno(errp, -ret, "Failed to lock byte %d",
off);
return ret;
} else if (s) {
s->locked_shared_perm |= bit;
}
} else if (unlock && (locked_shared_perm & bit) &&
!(shared_perm_lock_bits & bit)) {
ret = qemu_unlock_fd(fd, off, 1);
if (ret) {
error_setg_errno(errp, -ret, "Failed to unlock byte %d", off);
return ret;
} else if (s) {
s->locked_shared_perm &= ~bit;
}
}
}
return 0;
}
/* Check "unshared" bytes implied by @perm and ~@shared_perm in the file. */
static int raw_check_lock_bytes(int fd, uint64_t perm, uint64_t shared_perm,
Error **errp)
{
int ret;
int i;
PERM_FOREACH(i) {
int off = RAW_LOCK_SHARED_BASE + i;
uint64_t p = 1ULL << i;
if (perm & p) {
ret = qemu_lock_fd_test(fd, off, 1, true);
if (ret) {
char *perm_name = bdrv_perm_names(p);
raw_lock_error_setg_errno(errp, -ret,
"Failed to get \"%s\" lock",
perm_name);
g_free(perm_name);
return ret;
}
}
}
PERM_FOREACH(i) {
int off = RAW_LOCK_PERM_BASE + i;
uint64_t p = 1ULL << i;
if (!(shared_perm & p)) {
ret = qemu_lock_fd_test(fd, off, 1, true);
if (ret) {
char *perm_name = bdrv_perm_names(p);
raw_lock_error_setg_errno(errp, -ret,
"Failed to get shared \"%s\" lock",
perm_name);
g_free(perm_name);
return ret;
}
}
}
return 0;
}
static int raw_handle_perm_lock(BlockDriverState *bs,
RawPermLockOp op,
uint64_t new_perm, uint64_t new_shared,
Error **errp)
{
BDRVRawState *s = bs->opaque;
int ret = 0;
Error *local_err = NULL;
if (!s->use_lock) {
return 0;
}
if (bdrv_get_flags(bs) & BDRV_O_INACTIVE) {
return 0;
}
switch (op) {
case RAW_PL_PREPARE:
if ((s->perm | new_perm) == s->perm &&
(s->shared_perm & new_shared) == s->shared_perm)
{
/*
* We are going to unlock bytes, it should not fail. If it fail due
* to some fs-dependent permission-unrelated reasons (which occurs
* sometimes on NFS and leads to abort in bdrv_replace_child) we
* can't prevent such errors by any check here. And we ignore them
* anyway in ABORT and COMMIT.
*/
return 0;
}
ret = raw_apply_lock_bytes(s, s->fd, s->perm | new_perm,
~s->shared_perm | ~new_shared,
false, errp);
if (!ret) {
ret = raw_check_lock_bytes(s->fd, new_perm, new_shared, errp);
if (!ret) {
return 0;
}
error_append_hint(errp,
"Is another process using the image [%s]?\n",
bs->filename);
}
/* fall through to unlock bytes. */
case RAW_PL_ABORT:
raw_apply_lock_bytes(s, s->fd, s->perm, ~s->shared_perm,
true, &local_err);
if (local_err) {
/* Theoretically the above call only unlocks bytes and it cannot
* fail. Something weird happened, report it.
*/
warn_report_err(local_err);
}
break;
case RAW_PL_COMMIT:
raw_apply_lock_bytes(s, s->fd, new_perm, ~new_shared,
true, &local_err);
if (local_err) {
/* Theoretically the above call only unlocks bytes and it cannot
* fail. Something weird happened, report it.
*/
warn_report_err(local_err);
}
break;
}
return ret;
}
static int raw_reconfigure_getfd(BlockDriverState *bs, int flags,
int *open_flags, uint64_t perm, bool force_dup,
Error **errp)
{
BDRVRawState *s = bs->opaque;
int fd = -1;
int ret;
bool has_writers = perm &
(BLK_PERM_WRITE | BLK_PERM_WRITE_UNCHANGED | BLK_PERM_RESIZE);
int fcntl_flags = O_APPEND | O_NONBLOCK;
#ifdef O_NOATIME
fcntl_flags |= O_NOATIME;
#endif
*open_flags = 0;
if (s->type == FTYPE_CD) {
*open_flags |= O_NONBLOCK;
}
raw_parse_flags(flags, open_flags, has_writers);
#ifdef O_ASYNC
/* Not all operating systems have O_ASYNC, and those that don't
* will not let us track the state into rs->open_flags (typically
* you achieve the same effect with an ioctl, for example I_SETSIG
* on Solaris). But we do not use O_ASYNC, so that's fine.
*/
assert((s->open_flags & O_ASYNC) == 0);
#endif
if (!force_dup && *open_flags == s->open_flags) {
/* We're lucky, the existing fd is fine */
return s->fd;
}
if ((*open_flags & ~fcntl_flags) == (s->open_flags & ~fcntl_flags)) {
/* dup the original fd */
fd = qemu_dup(s->fd);
if (fd >= 0) {
ret = fcntl_setfl(fd, *open_flags);
if (ret) {
qemu_close(fd);
fd = -1;
}
}
}
/* If we cannot use fcntl, or fcntl failed, fall back to qemu_open() */
if (fd == -1) {
const char *normalized_filename = bs->filename;
ret = raw_normalize_devicepath(&normalized_filename, errp);
if (ret >= 0) {
fd = qemu_open(normalized_filename, *open_flags, errp);
if (fd == -1) {
return -1;
}
}
}
if (fd != -1 && (*open_flags & O_RDWR)) {
ret = check_hdev_writable(fd);
if (ret < 0) {
qemu_close(fd);
error_setg_errno(errp, -ret, "The device is not writable");
return -1;
}
}
return fd;
}
static int raw_reopen_prepare(BDRVReopenState *state,
BlockReopenQueue *queue, Error **errp)
{
BDRVRawState *s;
BDRVRawReopenState *rs;
QemuOpts *opts;
int ret;
Error *local_err = NULL;
assert(state != NULL);
assert(state->bs != NULL);
s = state->bs->opaque;
state->opaque = g_new0(BDRVRawReopenState, 1);
rs = state->opaque;
/* Handle options changes */
opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort);
if (!qemu_opts_absorb_qdict(opts, state->options, errp)) {
ret = -EINVAL;
goto out;
}
rs->drop_cache = qemu_opt_get_bool_del(opts, "drop-cache", true);
rs->check_cache_dropped =
qemu_opt_get_bool_del(opts, "x-check-cache-dropped", false);
/* This driver's reopen function doesn't currently allow changing
* other options, so let's put them back in the original QDict and
* bdrv_reopen_prepare() will detect changes and complain. */
qemu_opts_to_qdict(opts, state->options);
rs->fd = raw_reconfigure_getfd(state->bs, state->flags, &rs->open_flags,
state->perm, true, &local_err);
if (local_err) {
error_propagate(errp, local_err);
ret = -1;
goto out;
}
/* Fail already reopen_prepare() if we can't get a working O_DIRECT
* alignment with the new fd. */
if (rs->fd != -1) {
raw_probe_alignment(state->bs, rs->fd, &local_err);
if (local_err) {
error_propagate(errp, local_err);
ret = -EINVAL;
goto out_fd;
}
}
s->reopen_state = state;
ret = 0;
out_fd:
if (ret < 0) {
qemu_close(rs->fd);
rs->fd = -1;
}
out:
qemu_opts_del(opts);
return ret;
}
static void raw_reopen_commit(BDRVReopenState *state)
{
BDRVRawReopenState *rs = state->opaque;
BDRVRawState *s = state->bs->opaque;
s->drop_cache = rs->drop_cache;
s->check_cache_dropped = rs->check_cache_dropped;
s->open_flags = rs->open_flags;
qemu_close(s->fd);
s->fd = rs->fd;
g_free(state->opaque);
state->opaque = NULL;
assert(s->reopen_state == state);
s->reopen_state = NULL;
}
static void raw_reopen_abort(BDRVReopenState *state)
{
BDRVRawReopenState *rs = state->opaque;
BDRVRawState *s = state->bs->opaque;
/* nothing to do if NULL, we didn't get far enough */
if (rs == NULL) {
return;
}
if (rs->fd >= 0) {
qemu_close(rs->fd);
rs->fd = -1;
}
g_free(state->opaque);
state->opaque = NULL;
assert(s->reopen_state == state);
s->reopen_state = NULL;
}
static int sg_get_max_transfer_length(int fd)
{
#ifdef BLKSECTGET
int max_bytes = 0;
if (ioctl(fd, BLKSECTGET, &max_bytes) == 0) {
return max_bytes;
} else {
return -errno;
}
#else
return -ENOSYS;
#endif
}
static int sg_get_max_segments(int fd)
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
{
#ifdef CONFIG_LINUX
char buf[32];
const char *end;
char *sysfspath = NULL;
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
int ret;
int sysfd = -1;
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
long max_segments;
struct stat st;
if (fstat(fd, &st)) {
ret = -errno;
goto out;
}
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
sysfspath = g_strdup_printf("/sys/dev/block/%u:%u/queue/max_segments",
major(st.st_rdev), minor(st.st_rdev));
sysfd = open(sysfspath, O_RDONLY);
if (sysfd == -1) {
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
ret = -errno;
goto out;
}
do {
ret = read(sysfd, buf, sizeof(buf) - 1);
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
} while (ret == -1 && errno == EINTR);
if (ret < 0) {
ret = -errno;
goto out;
} else if (ret == 0) {
ret = -EIO;
goto out;
}
buf[ret] = 0;
/* The file is ended with '\n', pass 'end' to accept that. */
ret = qemu_strtol(buf, &end, 10, &max_segments);
if (ret == 0 && end && *end == '\n') {
ret = max_segments;
}
out:
if (sysfd != -1) {
close(sysfd);
}
file-posix: Consider max_segments for BlockLimits.max_transfer BlockLimits.max_transfer can be too high without this fix, guest will encounter I/O error or even get paused with werror=stop or rerror=stop. The cause is explained below. Linux has a separate limit, /sys/block/.../queue/max_segments, which in the worst case can be more restrictive than the BLKSECTGET which we already consider (note that they are two different things). So, the failure scenario before this patch is: 1) host device has max_sectors_kb = 4096 and max_segments = 64; 2) guest learns max_sectors_kb limit from QEMU, but doesn't know max_segments; 3) guest issues e.g. a 512KB request thinking it's okay, but actually it's not, because it will be passed through to host device as an SG_IO req that has niov > 64; 4) host kernel doesn't like the segmenting of the request, and returns -EINVAL; This patch checks the max_segments sysfs entry for the host device and calculates a "conservative" bytes limit using the page size, which is then merged into the existing max_transfer limit. Guest will discover this from the usual virtual block device interfaces. (In the case of scsi-generic, it will be done in the INQUIRY reply interception in device model.) The other possibility is to actually propagate it as a separate limit, but it's not better. On the one hand, there is a big complication: the limit is per-LUN in QEMU PoV (because we can attach LUNs from different host HBAs to the same virtio-scsi bus), but the channel to communicate it in a per-LUN manner is missing down the stack; on the other hand, two limits versus one doesn't change much about the valid size of I/O (because guest has no control over host segmenting). Also, the idea to fall back to bounce buffering in QEMU, upon -EINVAL, was explored. Unfortunately there is no neat way to ensure the bounce buffer is less segmented (in terms of DMA addr) than the guest buffer. Practically, this bug is not very common. It is only reported on a Emulex (lpfc), so it's okay to get it fixed in the easier way. Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Fam Zheng <famz@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2017-03-08 13:08:14 +01:00
g_free(sysfspath);
return ret;
#else
return -ENOTSUP;
#endif
}
static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
{
BDRVRawState *s = bs->opaque;
if (bs->sg) {
int ret = sg_get_max_transfer_length(s->fd);
if (ret > 0 && ret <= BDRV_REQUEST_MAX_BYTES) {
bs->bl.max_transfer = pow2floor(ret);
}
ret = sg_get_max_segments(s->fd);
if (ret > 0) {
bs->bl.max_transfer = MIN(bs->bl.max_transfer,
ret * qemu_real_host_page_size);
}
}
raw_probe_alignment(bs, s->fd, errp);
bs->bl.min_mem_alignment = s->buf_align;
bs->bl.opt_mem_alignment = MAX(s->buf_align, qemu_real_host_page_size);
}
static int check_for_dasd(int fd)
{
#ifdef BIODASDINFO2
struct dasd_information2_t info = {0};
return ioctl(fd, BIODASDINFO2, &info);
#else
return -1;
#endif
}
/**
* Try to get @bs's logical and physical block size.
* On success, store them in @bsz and return zero.
* On failure, return negative errno.
*/
static int hdev_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
{
BDRVRawState *s = bs->opaque;
int ret;
/* If DASD, get blocksizes */
if (check_for_dasd(s->fd) < 0) {
return -ENOTSUP;
}
ret = probe_logical_blocksize(s->fd, &bsz->log);
if (ret < 0) {
return ret;
}
return probe_physical_blocksize(s->fd, &bsz->phys);
}
/**
* Try to get @bs's geometry: cyls, heads, sectors.
* On success, store them in @geo and return 0.
* On failure return -errno.
* (Allows block driver to assign default geometry values that guest sees)
*/
#ifdef __linux__
static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
{
BDRVRawState *s = bs->opaque;
struct hd_geometry ioctl_geo = {0};
/* If DASD, get its geometry */
if (check_for_dasd(s->fd) < 0) {
return -ENOTSUP;
}
if (ioctl(s->fd, HDIO_GETGEO, &ioctl_geo) < 0) {
return -errno;
}
/* HDIO_GETGEO may return success even though geo contains zeros
(e.g. certain multipath setups) */
if (!ioctl_geo.heads || !ioctl_geo.sectors || !ioctl_geo.cylinders) {
return -ENOTSUP;
}
/* Do not return a geometry for partition */
if (ioctl_geo.start != 0) {
return -ENOTSUP;
}
geo->heads = ioctl_geo.heads;
geo->sectors = ioctl_geo.sectors;
geo->cylinders = ioctl_geo.cylinders;
return 0;
}
#else /* __linux__ */
static int hdev_probe_geometry(BlockDriverState *bs, HDGeometry *geo)
{
return -ENOTSUP;
}
#endif
#if defined(__linux__)
static int handle_aiocb_ioctl(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
int ret;
ret = ioctl(aiocb->aio_fildes, aiocb->ioctl.cmd, aiocb->ioctl.buf);
if (ret == -1) {
return -errno;
}
return 0;
}
#endif /* linux */
static int handle_aiocb_flush(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
BDRVRawState *s = aiocb->bs->opaque;
int ret;
if (s->page_cache_inconsistent) {
return -EIO;
}
ret = qemu_fdatasync(aiocb->aio_fildes);
if (ret == -1) {
/* There is no clear definition of the semantics of a failing fsync(),
* so we may have to assume the worst. The sad truth is that this
* assumption is correct for Linux. Some pages are now probably marked
* clean in the page cache even though they are inconsistent with the
* on-disk contents. The next fdatasync() call would succeed, but no
* further writeback attempt will be made. We can't get back to a state
* in which we know what is on disk (we would have to rewrite
* everything that was touched since the last fdatasync() at least), so
* make bdrv_flush() fail permanently. Given that the behaviour isn't
* really defined, I have little hope that other OSes are doing better.
*
* Obviously, this doesn't affect O_DIRECT, which bypasses the page
* cache. */
if ((s->open_flags & O_DIRECT) == 0) {
s->page_cache_inconsistent = true;
}
return -errno;
}
return 0;
}
#ifdef CONFIG_PREADV
static bool preadv_present = true;
static ssize_t
qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
return preadv(fd, iov, nr_iov, offset);
}
static ssize_t
qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
return pwritev(fd, iov, nr_iov, offset);
}
#else
static bool preadv_present = false;
static ssize_t
qemu_preadv(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
return -ENOSYS;
}
static ssize_t
qemu_pwritev(int fd, const struct iovec *iov, int nr_iov, off_t offset)
{
return -ENOSYS;
}
#endif
static ssize_t handle_aiocb_rw_vector(RawPosixAIOData *aiocb)
{
ssize_t len;
do {
if (aiocb->aio_type & QEMU_AIO_WRITE)
len = qemu_pwritev(aiocb->aio_fildes,
aiocb->io.iov,
aiocb->io.niov,
aiocb->aio_offset);
else
len = qemu_preadv(aiocb->aio_fildes,
aiocb->io.iov,
aiocb->io.niov,
aiocb->aio_offset);
} while (len == -1 && errno == EINTR);
if (len == -1) {
return -errno;
}
return len;
}
/*
* Read/writes the data to/from a given linear buffer.
*
* Returns the number of bytes handles or -errno in case of an error. Short
* reads are only returned if the end of the file is reached.
*/
static ssize_t handle_aiocb_rw_linear(RawPosixAIOData *aiocb, char *buf)
{
ssize_t offset = 0;
ssize_t len;
while (offset < aiocb->aio_nbytes) {
if (aiocb->aio_type & QEMU_AIO_WRITE) {
len = pwrite(aiocb->aio_fildes,
(const char *)buf + offset,
aiocb->aio_nbytes - offset,
aiocb->aio_offset + offset);
} else {
len = pread(aiocb->aio_fildes,
buf + offset,
aiocb->aio_nbytes - offset,
aiocb->aio_offset + offset);
}
if (len == -1 && errno == EINTR) {
continue;
} else if (len == -1 && errno == EINVAL &&
(aiocb->bs->open_flags & BDRV_O_NOCACHE) &&
!(aiocb->aio_type & QEMU_AIO_WRITE) &&
offset > 0) {
/* O_DIRECT pread() may fail with EINVAL when offset is unaligned
* after a short read. Assume that O_DIRECT short reads only occur
* at EOF. Therefore this is a short read, not an I/O error.
*/
break;
} else if (len == -1) {
offset = -errno;
break;
} else if (len == 0) {
break;
}
offset += len;
}
return offset;
}
static int handle_aiocb_rw(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
ssize_t nbytes;
char *buf;
if (!(aiocb->aio_type & QEMU_AIO_MISALIGNED)) {
/*
* If there is just a single buffer, and it is properly aligned
* we can just use plain pread/pwrite without any problems.
*/
if (aiocb->io.niov == 1) {
nbytes = handle_aiocb_rw_linear(aiocb, aiocb->io.iov->iov_base);
goto out;
}
/*
* We have more than one iovec, and all are properly aligned.
*
* Try preadv/pwritev first and fall back to linearizing the
* buffer if it's not supported.
*/
if (preadv_present) {
nbytes = handle_aiocb_rw_vector(aiocb);
if (nbytes == aiocb->aio_nbytes ||
(nbytes < 0 && nbytes != -ENOSYS)) {
goto out;
}
preadv_present = false;
}
/*
* XXX(hch): short read/write. no easy way to handle the reminder
* using these interfaces. For now retry using plain
* pread/pwrite?
*/
}
/*
* Ok, we have to do it the hard way, copy all segments into
* a single aligned buffer.
*/
buf = qemu_try_blockalign(aiocb->bs, aiocb->aio_nbytes);
if (buf == NULL) {
nbytes = -ENOMEM;
goto out;
}
if (aiocb->aio_type & QEMU_AIO_WRITE) {
char *p = buf;
int i;
for (i = 0; i < aiocb->io.niov; ++i) {
memcpy(p, aiocb->io.iov[i].iov_base, aiocb->io.iov[i].iov_len);
p += aiocb->io.iov[i].iov_len;
}
assert(p - buf == aiocb->aio_nbytes);
}
nbytes = handle_aiocb_rw_linear(aiocb, buf);
if (!(aiocb->aio_type & QEMU_AIO_WRITE)) {
char *p = buf;
size_t count = aiocb->aio_nbytes, copy;
int i;
for (i = 0; i < aiocb->io.niov && count; ++i) {
copy = count;
if (copy > aiocb->io.iov[i].iov_len) {
copy = aiocb->io.iov[i].iov_len;
}
memcpy(aiocb->io.iov[i].iov_base, p, copy);
assert(count >= copy);
p += copy;
count -= copy;
}
assert(count == 0);
}
qemu_vfree(buf);
out:
if (nbytes == aiocb->aio_nbytes) {
return 0;
} else if (nbytes >= 0 && nbytes < aiocb->aio_nbytes) {
if (aiocb->aio_type & QEMU_AIO_WRITE) {
return -EINVAL;
} else {
iov_memset(aiocb->io.iov, aiocb->io.niov, nbytes,
0, aiocb->aio_nbytes - nbytes);
return 0;
}
} else {
assert(nbytes < 0);
return nbytes;
}
}
static int translate_err(int err)
{
if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP ||
err == -ENOTTY) {
err = -ENOTSUP;
}
return err;
}
#ifdef CONFIG_FALLOCATE
static int do_fallocate(int fd, int mode, off_t offset, off_t len)
{
do {
if (fallocate(fd, mode, offset, len) == 0) {
return 0;
}
} while (errno == EINTR);
return translate_err(-errno);
}
#endif
static ssize_t handle_aiocb_write_zeroes_block(RawPosixAIOData *aiocb)
{
int ret = -ENOTSUP;
BDRVRawState *s = aiocb->bs->opaque;
if (!s->has_write_zeroes) {
return -ENOTSUP;
}
#ifdef BLKZEROOUT
/* The BLKZEROOUT implementation in the kernel doesn't set
* BLKDEV_ZERO_NOFALLBACK, so we can't call this if we have to avoid slow
* fallbacks. */
if (!(aiocb->aio_type & QEMU_AIO_NO_FALLBACK)) {
do {
uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes };
if (ioctl(aiocb->aio_fildes, BLKZEROOUT, range) == 0) {
return 0;
}
} while (errno == EINTR);
ret = translate_err(-errno);
if (ret == -ENOTSUP) {
s->has_write_zeroes = false;
}
}
#endif
return ret;
}
static int handle_aiocb_write_zeroes(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
#ifdef CONFIG_FALLOCATE
BDRVRawState *s = aiocb->bs->opaque;
int64_t len;
#endif
if (aiocb->aio_type & QEMU_AIO_BLKDEV) {
return handle_aiocb_write_zeroes_block(aiocb);
}
#ifdef CONFIG_FALLOCATE_ZERO_RANGE
if (s->has_write_zeroes) {
int ret = do_fallocate(s->fd, FALLOC_FL_ZERO_RANGE,
aiocb->aio_offset, aiocb->aio_nbytes);
if (ret == -EINVAL) {
/*
* Allow falling back to pwrite for file systems that
* do not support fallocate() for an unaligned byte range.
*/
return -ENOTSUP;
}
if (ret == 0 || ret != -ENOTSUP) {
return ret;
}
s->has_write_zeroes = false;
}
#endif
#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
if (s->has_discard && s->has_fallocate) {
int ret = do_fallocate(s->fd,
FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
aiocb->aio_offset, aiocb->aio_nbytes);
if (ret == 0) {
ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes);
if (ret == 0 || ret != -ENOTSUP) {
return ret;
}
s->has_fallocate = false;
} else if (ret != -ENOTSUP) {
return ret;
} else {
s->has_discard = false;
}
}
#endif
#ifdef CONFIG_FALLOCATE
/* Last resort: we are trying to extend the file with zeroed data. This
* can be done via fallocate(fd, 0) */
len = bdrv_getlength(aiocb->bs);
if (s->has_fallocate && len >= 0 && aiocb->aio_offset >= len) {
int ret = do_fallocate(s->fd, 0, aiocb->aio_offset, aiocb->aio_nbytes);
if (ret == 0 || ret != -ENOTSUP) {
return ret;
}
s->has_fallocate = false;
}
#endif
return -ENOTSUP;
}
static int handle_aiocb_write_zeroes_unmap(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
BDRVRawState *s G_GNUC_UNUSED = aiocb->bs->opaque;
/* First try to write zeros and unmap at the same time */
#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
int ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
aiocb->aio_offset, aiocb->aio_nbytes);
switch (ret) {
case -ENOTSUP:
case -EINVAL:
case -EBUSY:
break;
default:
return ret;
}
#endif
/* If we couldn't manage to unmap while guaranteed that the area reads as
* all-zero afterwards, just write zeroes without unmapping */
return handle_aiocb_write_zeroes(aiocb);
}
#ifndef HAVE_COPY_FILE_RANGE
static off_t copy_file_range(int in_fd, off_t *in_off, int out_fd,
off_t *out_off, size_t len, unsigned int flags)
{
#ifdef __NR_copy_file_range
return syscall(__NR_copy_file_range, in_fd, in_off, out_fd,
out_off, len, flags);
#else
errno = ENOSYS;
return -1;
#endif
}
#endif
static int handle_aiocb_copy_range(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
uint64_t bytes = aiocb->aio_nbytes;
off_t in_off = aiocb->aio_offset;
off_t out_off = aiocb->copy_range.aio_offset2;
while (bytes) {
ssize_t ret = copy_file_range(aiocb->aio_fildes, &in_off,
aiocb->copy_range.aio_fd2, &out_off,
bytes, 0);
trace_file_copy_file_range(aiocb->bs, aiocb->aio_fildes, in_off,
aiocb->copy_range.aio_fd2, out_off, bytes,
0, ret);
if (ret == 0) {
/* No progress (e.g. when beyond EOF), let the caller fall back to
* buffer I/O. */
return -ENOSPC;
}
if (ret < 0) {
switch (errno) {
case ENOSYS:
return -ENOTSUP;
case EINTR:
continue;
default:
return -errno;
}
}
bytes -= ret;
}
return 0;
}
static int handle_aiocb_discard(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
int ret = -EOPNOTSUPP;
BDRVRawState *s = aiocb->bs->opaque;
if (!s->has_discard) {
return -ENOTSUP;
}
if (aiocb->aio_type & QEMU_AIO_BLKDEV) {
#ifdef BLKDISCARD
do {
uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes };
if (ioctl(aiocb->aio_fildes, BLKDISCARD, range) == 0) {
return 0;
}
} while (errno == EINTR);
ret = -errno;
#endif
} else {
#ifdef CONFIG_FALLOCATE_PUNCH_HOLE
ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
aiocb->aio_offset, aiocb->aio_nbytes);
#endif
}
ret = translate_err(ret);
if (ret == -ENOTSUP) {
s->has_discard = false;
}
return ret;
}
block: posix: Always allocate the first block When creating an image with preallocation "off" or "falloc", the first block of the image is typically not allocated. When using Gluster storage backed by XFS filesystem, reading this block using direct I/O succeeds regardless of request length, fooling alignment detection. In this case we fallback to a safe value (4096) instead of the optimal value (512), which may lead to unneeded data copying when aligning requests. Allocating the first block avoids the fallback. Since we allocate the first block even with preallocation=off, we no longer create images with zero disk size: $ ./qemu-img create -f raw test.raw 1g Formatting 'test.raw', fmt=raw size=1073741824 $ ls -lhs test.raw 4.0K -rw-r--r--. 1 nsoffer nsoffer 1.0G Aug 16 23:48 test.raw And converting the image requires additional cluster: $ ./qemu-img measure -f raw -O qcow2 test.raw required size: 458752 fully allocated size: 1074135040 When using format like vmdk with multiple files per image, we allocate one block per file: $ ./qemu-img create -f vmdk -o subformat=twoGbMaxExtentFlat test.vmdk 4g Formatting 'test.vmdk', fmt=vmdk size=4294967296 compat6=off hwversion=undefined subformat=twoGbMaxExtentFlat $ ls -lhs test*.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f001.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f002.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 353 Aug 27 03:23 test.vmdk I did quick performance test for copying disks with qemu-img convert to new raw target image to Gluster storage with sector size of 512 bytes: for i in $(seq 10); do rm -f dst.raw sleep 10 time ./qemu-img convert -f raw -O raw -t none -T none src.raw dst.raw done Here is a table comparing the total time spent: Type Before(s) After(s) Diff(%) --------------------------------------- real 530.028 469.123 -11.4 user 17.204 10.768 -37.4 sys 17.881 7.011 -60.7 We can see very clear improvement in CPU usage. Signed-off-by: Nir Soffer <nsoffer@redhat.com> Message-id: 20190827010528.8818-2-nsoffer@redhat.com Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Max Reitz <mreitz@redhat.com>
2019-08-27 03:05:27 +02:00
/*
* Help alignment probing by allocating the first block.
*
* When reading with direct I/O from unallocated area on Gluster backed by XFS,
* reading succeeds regardless of request length. In this case we fallback to
* safe alignment which is not optimal. Allocating the first block avoids this
* fallback.
*
* fd may be opened with O_DIRECT, but we don't know the buffer alignment or
* request alignment, so we use safe values.
*
* Returns: 0 on success, -errno on failure. Since this is an optimization,
* caller may ignore failures.
*/
static int allocate_first_block(int fd, size_t max_size)
{
size_t write_size = (max_size < MAX_BLOCKSIZE)
? BDRV_SECTOR_SIZE
: MAX_BLOCKSIZE;
size_t max_align = MAX(MAX_BLOCKSIZE, qemu_real_host_page_size);
block: posix: Always allocate the first block When creating an image with preallocation "off" or "falloc", the first block of the image is typically not allocated. When using Gluster storage backed by XFS filesystem, reading this block using direct I/O succeeds regardless of request length, fooling alignment detection. In this case we fallback to a safe value (4096) instead of the optimal value (512), which may lead to unneeded data copying when aligning requests. Allocating the first block avoids the fallback. Since we allocate the first block even with preallocation=off, we no longer create images with zero disk size: $ ./qemu-img create -f raw test.raw 1g Formatting 'test.raw', fmt=raw size=1073741824 $ ls -lhs test.raw 4.0K -rw-r--r--. 1 nsoffer nsoffer 1.0G Aug 16 23:48 test.raw And converting the image requires additional cluster: $ ./qemu-img measure -f raw -O qcow2 test.raw required size: 458752 fully allocated size: 1074135040 When using format like vmdk with multiple files per image, we allocate one block per file: $ ./qemu-img create -f vmdk -o subformat=twoGbMaxExtentFlat test.vmdk 4g Formatting 'test.vmdk', fmt=vmdk size=4294967296 compat6=off hwversion=undefined subformat=twoGbMaxExtentFlat $ ls -lhs test*.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f001.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f002.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 353 Aug 27 03:23 test.vmdk I did quick performance test for copying disks with qemu-img convert to new raw target image to Gluster storage with sector size of 512 bytes: for i in $(seq 10); do rm -f dst.raw sleep 10 time ./qemu-img convert -f raw -O raw -t none -T none src.raw dst.raw done Here is a table comparing the total time spent: Type Before(s) After(s) Diff(%) --------------------------------------- real 530.028 469.123 -11.4 user 17.204 10.768 -37.4 sys 17.881 7.011 -60.7 We can see very clear improvement in CPU usage. Signed-off-by: Nir Soffer <nsoffer@redhat.com> Message-id: 20190827010528.8818-2-nsoffer@redhat.com Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Max Reitz <mreitz@redhat.com>
2019-08-27 03:05:27 +02:00
void *buf;
ssize_t n;
int ret;
buf = qemu_memalign(max_align, write_size);
memset(buf, 0, write_size);
do {
n = pwrite(fd, buf, write_size, 0);
} while (n == -1 && errno == EINTR);
ret = (n == -1) ? -errno : 0;
qemu_vfree(buf);
return ret;
}
static int handle_aiocb_truncate(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
int result = 0;
int64_t current_length = 0;
char *buf = NULL;
struct stat st;
int fd = aiocb->aio_fildes;
int64_t offset = aiocb->aio_offset;
PreallocMode prealloc = aiocb->truncate.prealloc;
Error **errp = aiocb->truncate.errp;
if (fstat(fd, &st) < 0) {
result = -errno;
error_setg_errno(errp, -result, "Could not stat file");
return result;
}
current_length = st.st_size;
if (current_length > offset && prealloc != PREALLOC_MODE_OFF) {
error_setg(errp, "Cannot use preallocation for shrinking files");
return -ENOTSUP;
}
switch (prealloc) {
#ifdef CONFIG_POSIX_FALLOCATE
case PREALLOC_MODE_FALLOC:
/*
* Truncating before posix_fallocate() makes it about twice slower on
* file systems that do not support fallocate(), trying to check if a
* block is allocated before allocating it, so don't do that here.
*/
if (offset != current_length) {
result = -posix_fallocate(fd, current_length,
offset - current_length);
if (result != 0) {
/* posix_fallocate() doesn't set errno. */
error_setg_errno(errp, -result,
"Could not preallocate new data");
block: posix: Always allocate the first block When creating an image with preallocation "off" or "falloc", the first block of the image is typically not allocated. When using Gluster storage backed by XFS filesystem, reading this block using direct I/O succeeds regardless of request length, fooling alignment detection. In this case we fallback to a safe value (4096) instead of the optimal value (512), which may lead to unneeded data copying when aligning requests. Allocating the first block avoids the fallback. Since we allocate the first block even with preallocation=off, we no longer create images with zero disk size: $ ./qemu-img create -f raw test.raw 1g Formatting 'test.raw', fmt=raw size=1073741824 $ ls -lhs test.raw 4.0K -rw-r--r--. 1 nsoffer nsoffer 1.0G Aug 16 23:48 test.raw And converting the image requires additional cluster: $ ./qemu-img measure -f raw -O qcow2 test.raw required size: 458752 fully allocated size: 1074135040 When using format like vmdk with multiple files per image, we allocate one block per file: $ ./qemu-img create -f vmdk -o subformat=twoGbMaxExtentFlat test.vmdk 4g Formatting 'test.vmdk', fmt=vmdk size=4294967296 compat6=off hwversion=undefined subformat=twoGbMaxExtentFlat $ ls -lhs test*.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f001.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f002.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 353 Aug 27 03:23 test.vmdk I did quick performance test for copying disks with qemu-img convert to new raw target image to Gluster storage with sector size of 512 bytes: for i in $(seq 10); do rm -f dst.raw sleep 10 time ./qemu-img convert -f raw -O raw -t none -T none src.raw dst.raw done Here is a table comparing the total time spent: Type Before(s) After(s) Diff(%) --------------------------------------- real 530.028 469.123 -11.4 user 17.204 10.768 -37.4 sys 17.881 7.011 -60.7 We can see very clear improvement in CPU usage. Signed-off-by: Nir Soffer <nsoffer@redhat.com> Message-id: 20190827010528.8818-2-nsoffer@redhat.com Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Max Reitz <mreitz@redhat.com>
2019-08-27 03:05:27 +02:00
} else if (current_length == 0) {
/*
* posix_fallocate() uses fallocate() if the filesystem
* supports it, or fallback to manually writing zeroes. If
* fallocate() was used, unaligned reads from the fallocated
* area in raw_probe_alignment() will succeed, hence we need to
* allocate the first block.
*
* Optimize future alignment probing; ignore failures.
*/
allocate_first_block(fd, offset);
}
} else {
result = 0;
}
goto out;
#endif
case PREALLOC_MODE_FULL:
{
int64_t num = 0, left = offset - current_length;
off_t seek_result;
/*
* Knowing the final size from the beginning could allow the file
* system driver to do less allocations and possibly avoid
* fragmentation of the file.
*/
if (ftruncate(fd, offset) != 0) {
result = -errno;
error_setg_errno(errp, -result, "Could not resize file");
goto out;
}
buf = g_malloc0(65536);
seek_result = lseek(fd, current_length, SEEK_SET);
if (seek_result < 0) {
result = -errno;
error_setg_errno(errp, -result,
"Failed to seek to the old end of file");
goto out;
}
while (left > 0) {
num = MIN(left, 65536);
result = write(fd, buf, num);
if (result < 0) {
if (errno == EINTR) {
continue;
}
result = -errno;
error_setg_errno(errp, -result,
"Could not write zeros for preallocation");
goto out;
}
left -= result;
}
if (result >= 0) {
result = fsync(fd);
if (result < 0) {
result = -errno;
error_setg_errno(errp, -result,
"Could not flush file to disk");
goto out;
}
}
goto out;
}
case PREALLOC_MODE_OFF:
if (ftruncate(fd, offset) != 0) {
result = -errno;
error_setg_errno(errp, -result, "Could not resize file");
block: posix: Always allocate the first block When creating an image with preallocation "off" or "falloc", the first block of the image is typically not allocated. When using Gluster storage backed by XFS filesystem, reading this block using direct I/O succeeds regardless of request length, fooling alignment detection. In this case we fallback to a safe value (4096) instead of the optimal value (512), which may lead to unneeded data copying when aligning requests. Allocating the first block avoids the fallback. Since we allocate the first block even with preallocation=off, we no longer create images with zero disk size: $ ./qemu-img create -f raw test.raw 1g Formatting 'test.raw', fmt=raw size=1073741824 $ ls -lhs test.raw 4.0K -rw-r--r--. 1 nsoffer nsoffer 1.0G Aug 16 23:48 test.raw And converting the image requires additional cluster: $ ./qemu-img measure -f raw -O qcow2 test.raw required size: 458752 fully allocated size: 1074135040 When using format like vmdk with multiple files per image, we allocate one block per file: $ ./qemu-img create -f vmdk -o subformat=twoGbMaxExtentFlat test.vmdk 4g Formatting 'test.vmdk', fmt=vmdk size=4294967296 compat6=off hwversion=undefined subformat=twoGbMaxExtentFlat $ ls -lhs test*.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f001.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 2.0G Aug 27 03:23 test-f002.vmdk 4.0K -rw-r--r--. 1 nsoffer nsoffer 353 Aug 27 03:23 test.vmdk I did quick performance test for copying disks with qemu-img convert to new raw target image to Gluster storage with sector size of 512 bytes: for i in $(seq 10); do rm -f dst.raw sleep 10 time ./qemu-img convert -f raw -O raw -t none -T none src.raw dst.raw done Here is a table comparing the total time spent: Type Before(s) After(s) Diff(%) --------------------------------------- real 530.028 469.123 -11.4 user 17.204 10.768 -37.4 sys 17.881 7.011 -60.7 We can see very clear improvement in CPU usage. Signed-off-by: Nir Soffer <nsoffer@redhat.com> Message-id: 20190827010528.8818-2-nsoffer@redhat.com Reviewed-by: Max Reitz <mreitz@redhat.com> Signed-off-by: Max Reitz <mreitz@redhat.com>
2019-08-27 03:05:27 +02:00
} else if (current_length == 0 && offset > current_length) {
/* Optimize future alignment probing; ignore failures. */
allocate_first_block(fd, offset);
}
return result;
default:
result = -ENOTSUP;
error_setg(errp, "Unsupported preallocation mode: %s",
PreallocMode_str(prealloc));
return result;
}
out:
if (result < 0) {
if (ftruncate(fd, current_length) < 0) {
error_report("Failed to restore old file length: %s",
strerror(errno));
}
}
g_free(buf);
return result;
}
static int coroutine_fn raw_thread_pool_submit(BlockDriverState *bs,
ThreadPoolFunc func, void *arg)
{
/* @bs can be NULL, bdrv_get_aio_context() returns the main context then */
ThreadPool *pool = aio_get_thread_pool(bdrv_get_aio_context(bs));
return thread_pool_submit_co(pool, func, arg);
}
static int coroutine_fn raw_co_prw(BlockDriverState *bs, uint64_t offset,
uint64_t bytes, QEMUIOVector *qiov, int type)
{
BDRVRawState *s = bs->opaque;
RawPosixAIOData acb;
if (fd_open(bs) < 0)
return -EIO;
/*
* When using O_DIRECT, the request must be aligned to be able to use
* either libaio or io_uring interface. If not fail back to regular thread
* pool read/write code which emulates this for us if we
* set QEMU_AIO_MISALIGNED.
*/
if (s->needs_alignment && !bdrv_qiov_is_aligned(bs, qiov)) {
type |= QEMU_AIO_MISALIGNED;
#ifdef CONFIG_LINUX_IO_URING
} else if (s->use_linux_io_uring) {
LuringState *aio = aio_get_linux_io_uring(bdrv_get_aio_context(bs));
assert(qiov->size == bytes);
return luring_co_submit(bs, aio, s->fd, offset, qiov, type);
#endif
#ifdef CONFIG_LINUX_AIO
} else if (s->use_linux_aio) {
LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs));
assert(qiov->size == bytes);
return laio_co_submit(bs, aio, s->fd, offset, qiov, type);
#endif
raw-posix: refactor AIO support Currently the raw-posix.c code contains a lot of knowledge about the asynchronous I/O scheme that is mostly implemented in posix-aio-compat.c. All this code does not really belong here and is getting a bit in the way of implementing native AIO on Linux. So instead move all the guts of the AIO implementation into posix-aio-compat.c (which might need a better name, btw). There's now a very small interface between the AIO providers and raw-posix.c: - an init routine is called from raw_open_common to return an AIO context for this drive. An AIO implementation may either re-use one context for all drives, or use a different one for each as the Linux native AIO support will do. - an submit routine is called from the aio_reav/writev methods to submit an AIO request There are no indirect calls involved in this interface as we need to decide which one to call manually. We will only call the Linux AIO native init function if we were requested to by vl.c, and we will only call the native submit function if we are asked to and the request is properly aligned. That's also the reason why the alignment check actually does the inverse move and now goes into raw-posix.c. The old posix-aio-compat.h headers is removed now that most of it's content is private to posix-aio-compat.c, and instead we add a new block/raw-posix-aio.h headers is created containing only the tiny interface between raw-posix.c and the AIO implementation. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
2009-08-20 16:58:19 +02:00
}
acb = (RawPosixAIOData) {
.bs = bs,
.aio_fildes = s->fd,
.aio_type = type,
.aio_offset = offset,
.aio_nbytes = bytes,
.io = {
.iov = qiov->iov,
.niov = qiov->niov,
},
};
assert(qiov->size == bytes);
return raw_thread_pool_submit(bs, handle_aiocb_rw, &acb);
}
static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset,
uint64_t bytes, QEMUIOVector *qiov,
int flags)
{
return raw_co_prw(bs, offset, bytes, qiov, QEMU_AIO_READ);
}
static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset,
uint64_t bytes, QEMUIOVector *qiov,
int flags)
{
assert(flags == 0);
return raw_co_prw(bs, offset, bytes, qiov, QEMU_AIO_WRITE);
}
static void raw_aio_plug(BlockDriverState *bs)
{
BDRVRawState __attribute__((unused)) *s = bs->opaque;
#ifdef CONFIG_LINUX_AIO
if (s->use_linux_aio) {
LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs));
laio_io_plug(bs, aio);
}
#endif
#ifdef CONFIG_LINUX_IO_URING
if (s->use_linux_io_uring) {
LuringState *aio = aio_get_linux_io_uring(bdrv_get_aio_context(bs));
luring_io_plug(bs, aio);
}
#endif
}
static void raw_aio_unplug(BlockDriverState *bs)
{
BDRVRawState __attribute__((unused)) *s = bs->opaque;
#ifdef CONFIG_LINUX_AIO
if (s->use_linux_aio) {
LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs));
laio_io_unplug(bs, aio);
}
#endif
#ifdef CONFIG_LINUX_IO_URING
if (s->use_linux_io_uring) {
LuringState *aio = aio_get_linux_io_uring(bdrv_get_aio_context(bs));
luring_io_unplug(bs, aio);
}
#endif
}
static int raw_co_flush_to_disk(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
RawPosixAIOData acb;
int ret;
ret = fd_open(bs);
if (ret < 0) {
return ret;
}
acb = (RawPosixAIOData) {
.bs = bs,
.aio_fildes = s->fd,
.aio_type = QEMU_AIO_FLUSH,
};
#ifdef CONFIG_LINUX_IO_URING
if (s->use_linux_io_uring) {
LuringState *aio = aio_get_linux_io_uring(bdrv_get_aio_context(bs));
return luring_co_submit(bs, aio, s->fd, 0, NULL, QEMU_AIO_FLUSH);
}
#endif
return raw_thread_pool_submit(bs, handle_aiocb_flush, &acb);
}
static void raw_aio_attach_aio_context(BlockDriverState *bs,
AioContext *new_context)
{
BDRVRawState __attribute__((unused)) *s = bs->opaque;
#ifdef CONFIG_LINUX_AIO
if (s->use_linux_aio) {
Error *local_err = NULL;
if (!aio_setup_linux_aio(new_context, &local_err)) {
error_reportf_err(local_err, "Unable to use native AIO, "
"falling back to thread pool: ");
s->use_linux_aio = false;
}
}
#endif
#ifdef CONFIG_LINUX_IO_URING
if (s->use_linux_io_uring) {
Error *local_err = NULL;
if (!aio_setup_linux_io_uring(new_context, &local_err)) {
error_reportf_err(local_err, "Unable to use linux io_uring, "
"falling back to thread pool: ");
s->use_linux_io_uring = false;
}
}
#endif
}
static void raw_close(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
if (s->fd >= 0) {
qemu_close(s->fd);
s->fd = -1;
}
}
/**
* Truncates the given regular file @fd to @offset and, when growing, fills the
* new space according to @prealloc.
*
* Returns: 0 on success, -errno on failure.
*/
static int coroutine_fn
raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset,
PreallocMode prealloc, Error **errp)
{
RawPosixAIOData acb;
acb = (RawPosixAIOData) {
.bs = bs,
.aio_fildes = fd,
.aio_type = QEMU_AIO_TRUNCATE,
.aio_offset = offset,
.truncate = {
.prealloc = prealloc,
.errp = errp,
},
};
return raw_thread_pool_submit(bs, handle_aiocb_truncate, &acb);
}
static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset,
bool exact, PreallocMode prealloc,
BdrvRequestFlags flags, Error **errp)
{
BDRVRawState *s = bs->opaque;
struct stat st;
int ret;
if (fstat(s->fd, &st)) {
ret = -errno;
error_setg_errno(errp, -ret, "Failed to fstat() the file");
return ret;
}
if (S_ISREG(st.st_mode)) {
/* Always resizes to the exact @offset */
return raw_regular_truncate(bs, s->fd, offset, prealloc, errp);
}
if (prealloc != PREALLOC_MODE_OFF) {
error_setg(errp, "Preallocation mode '%s' unsupported for this "
"non-regular file", PreallocMode_str(prealloc));
return -ENOTSUP;
}
if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
int64_t cur_length = raw_getlength(bs);
if (offset != cur_length && exact) {
error_setg(errp, "Cannot resize device files");
return -ENOTSUP;
} else if (offset > cur_length) {
error_setg(errp, "Cannot grow device files");
return -EINVAL;
}
} else {
error_setg(errp, "Resizing this file is not supported");
return -ENOTSUP;
}
return 0;
}
#ifdef __OpenBSD__
static int64_t raw_getlength(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
int fd = s->fd;
struct stat st;
if (fstat(fd, &st))
return -errno;
if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
struct disklabel dl;
if (ioctl(fd, DIOCGDINFO, &dl))
return -errno;
return (uint64_t)dl.d_secsize *
dl.d_partitions[DISKPART(st.st_rdev)].p_size;
} else
return st.st_size;
}
#elif defined(__NetBSD__)
static int64_t raw_getlength(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
int fd = s->fd;
struct stat st;
if (fstat(fd, &st))
return -errno;
if (S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode)) {
struct dkwedge_info dkw;
if (ioctl(fd, DIOCGWEDGEINFO, &dkw) != -1) {
return dkw.dkw_size * 512;
} else {
struct disklabel dl;
if (ioctl(fd, DIOCGDINFO, &dl))
return -errno;
return (uint64_t)dl.d_secsize *
dl.d_partitions[DISKPART(st.st_rdev)].p_size;
}
} else
return st.st_size;
}
#elif defined(__sun__)
static int64_t raw_getlength(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
struct dk_minfo minfo;
int ret;
int64_t size;
ret = fd_open(bs);
if (ret < 0) {
return ret;
}
/*
* Use the DKIOCGMEDIAINFO ioctl to read the size.
*/
ret = ioctl(s->fd, DKIOCGMEDIAINFO, &minfo);
if (ret != -1) {
return minfo.dki_lbsize * minfo.dki_capacity;
}
/*
* There are reports that lseek on some devices fails, but
* irc discussion said that contingency on contingency was overkill.
*/
size = lseek(s->fd, 0, SEEK_END);
if (size < 0) {
return -errno;
}
return size;
}
#elif defined(CONFIG_BSD)
static int64_t raw_getlength(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
int fd = s->fd;
int64_t size;
struct stat sb;
#if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
int reopened = 0;
#endif
int ret;
ret = fd_open(bs);
if (ret < 0)
return ret;
#if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
again:
#endif
if (!fstat(fd, &sb) && (S_IFCHR & sb.st_mode)) {
#ifdef DIOCGMEDIASIZE
avoid TABs in files that only contain a few Most files that have TABs only contain a handful of them. Change them to spaces so that we don't confuse people. disas, standard-headers, linux-headers and libdecnumber are imported from other projects and probably should be exempted from the check. Outside those, after this patch the following files still contain both 8-space and TAB sequences at the beginning of the line. Many of them have a majority of TABs, or were initially committed with all tabs. bsd-user/i386/target_syscall.h bsd-user/x86_64/target_syscall.h crypto/aes.c hw/audio/fmopl.c hw/audio/fmopl.h hw/block/tc58128.c hw/display/cirrus_vga.c hw/display/xenfb.c hw/dma/etraxfs_dma.c hw/intc/sh_intc.c hw/misc/mst_fpga.c hw/net/pcnet.c hw/sh4/sh7750.c hw/timer/m48t59.c hw/timer/sh_timer.c include/crypto/aes.h include/disas/bfd.h include/hw/sh4/sh.h libdecnumber/decNumber.c linux-headers/asm-generic/unistd.h linux-headers/linux/kvm.h linux-user/alpha/target_syscall.h linux-user/arm/nwfpe/double_cpdo.c linux-user/arm/nwfpe/fpa11_cpdt.c linux-user/arm/nwfpe/fpa11_cprt.c linux-user/arm/nwfpe/fpa11.h linux-user/flat.h linux-user/flatload.c linux-user/i386/target_syscall.h linux-user/ppc/target_syscall.h linux-user/sparc/target_syscall.h linux-user/syscall.c linux-user/syscall_defs.h linux-user/x86_64/target_syscall.h slirp/cksum.c slirp/if.c slirp/ip.h slirp/ip_icmp.c slirp/ip_icmp.h slirp/ip_input.c slirp/ip_output.c slirp/mbuf.c slirp/misc.c slirp/sbuf.c slirp/socket.c slirp/socket.h slirp/tcp_input.c slirp/tcpip.h slirp/tcp_output.c slirp/tcp_subr.c slirp/tcp_timer.c slirp/tftp.c slirp/udp.c slirp/udp.h target/cris/cpu.h target/cris/mmu.c target/cris/op_helper.c target/sh4/helper.c target/sh4/op_helper.c target/sh4/translate.c tcg/sparc/tcg-target.inc.c tests/tcg/cris/check_addo.c tests/tcg/cris/check_moveq.c tests/tcg/cris/check_swap.c tests/tcg/multiarch/test-mmap.c ui/vnc-enc-hextile-template.h ui/vnc-enc-zywrle.h util/envlist.c util/readline.c The following have only TABs: bsd-user/i386/target_signal.h bsd-user/sparc64/target_signal.h bsd-user/sparc64/target_syscall.h bsd-user/sparc/target_signal.h bsd-user/sparc/target_syscall.h bsd-user/x86_64/target_signal.h crypto/desrfb.c hw/audio/intel-hda-defs.h hw/core/uboot_image.h hw/sh4/sh7750_regnames.c hw/sh4/sh7750_regs.h include/hw/cris/etraxfs_dma.h linux-user/alpha/termbits.h linux-user/arm/nwfpe/fpopcode.h linux-user/arm/nwfpe/fpsr.h linux-user/arm/syscall_nr.h linux-user/arm/target_signal.h linux-user/cris/target_signal.h linux-user/i386/target_signal.h linux-user/linux_loop.h linux-user/m68k/target_signal.h linux-user/microblaze/target_signal.h linux-user/mips64/target_signal.h linux-user/mips/target_signal.h linux-user/mips/target_syscall.h linux-user/mips/termbits.h linux-user/ppc/target_signal.h linux-user/sh4/target_signal.h linux-user/sh4/termbits.h linux-user/sparc64/target_syscall.h linux-user/sparc/target_signal.h linux-user/x86_64/target_signal.h linux-user/x86_64/termbits.h pc-bios/optionrom/optionrom.h slirp/mbuf.h slirp/misc.h slirp/sbuf.h slirp/tcp.h slirp/tcp_timer.h slirp/tcp_var.h target/i386/svm.h target/sparc/asi.h target/xtensa/core-dc232b/xtensa-modules.inc.c target/xtensa/core-dc233c/xtensa-modules.inc.c target/xtensa/core-de212/core-isa.h target/xtensa/core-de212/xtensa-modules.inc.c target/xtensa/core-fsf/xtensa-modules.inc.c target/xtensa/core-sample_controller/core-isa.h target/xtensa/core-sample_controller/xtensa-modules.inc.c target/xtensa/core-test_kc705_be/core-isa.h target/xtensa/core-test_kc705_be/xtensa-modules.inc.c tests/tcg/cris/check_abs.c tests/tcg/cris/check_addc.c tests/tcg/cris/check_addcm.c tests/tcg/cris/check_addoq.c tests/tcg/cris/check_bound.c tests/tcg/cris/check_ftag.c tests/tcg/cris/check_int64.c tests/tcg/cris/check_lz.c tests/tcg/cris/check_openpf5.c tests/tcg/cris/check_sigalrm.c tests/tcg/cris/crisutils.h tests/tcg/cris/sys.c tests/tcg/i386/test-i386-ssse3.c ui/vgafont.h Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Message-Id: <20181213223737.11793-3-pbonzini@redhat.com> Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> Acked-by: Richard Henderson <richard.henderson@linaro.org> Acked-by: Eric Blake <eblake@redhat.com> Acked-by: David Gibson <david@gibson.dropbear.id.au> Reviewed-by: Stefan Markovic <smarkovic@wavecomp.com> Reviewed-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2018-12-13 23:37:37 +01:00
if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size))
#elif defined(DIOCGPART)
{
struct partinfo pi;
if (ioctl(fd, DIOCGPART, &pi) == 0)
size = pi.media_size;
else
size = 0;
}
if (size == 0)
#endif
#if defined(__APPLE__) && defined(__MACH__)
{
uint64_t sectors = 0;
uint32_t sector_size = 0;
if (ioctl(fd, DKIOCGETBLOCKCOUNT, &sectors) == 0
&& ioctl(fd, DKIOCGETBLOCKSIZE, &sector_size) == 0) {
size = sectors * sector_size;
} else {
size = lseek(fd, 0LL, SEEK_END);
if (size < 0) {
return -errno;
}
}
}
#else
size = lseek(fd, 0LL, SEEK_END);
if (size < 0) {
return -errno;
}
#endif
#if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
switch(s->type) {
case FTYPE_CD:
/* XXX FreeBSD acd returns UINT_MAX sectors for an empty drive */
if (size == 2048LL * (unsigned)-1)
size = 0;
/* XXX no disc? maybe we need to reopen... */
if (size <= 0 && !reopened && cdrom_reopen(bs) >= 0) {
reopened = 1;
goto again;
}
}
#endif
} else {
size = lseek(fd, 0, SEEK_END);
if (size < 0) {
return -errno;
}
}
return size;
}
#else
static int64_t raw_getlength(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
int ret;
int64_t size;
ret = fd_open(bs);
if (ret < 0) {
return ret;
}
size = lseek(s->fd, 0, SEEK_END);
if (size < 0) {
return -errno;
}
return size;
}
#endif
static int64_t raw_get_allocated_file_size(BlockDriverState *bs)
{
struct stat st;
BDRVRawState *s = bs->opaque;
if (fstat(s->fd, &st) < 0) {
return -errno;
}
return (int64_t)st.st_blocks * 512;
}
static int coroutine_fn
raw_co_create(BlockdevCreateOptions *options, Error **errp)
{
BlockdevCreateOptionsFile *file_opts;
Error *local_err = NULL;
int fd;
uint64_t perm, shared;
int result = 0;
/* Validate options and set default values */
assert(options->driver == BLOCKDEV_DRIVER_FILE);
file_opts = &options->u.file;
if (!file_opts->has_nocow) {
file_opts->nocow = false;
}
if (!file_opts->has_preallocation) {
file_opts->preallocation = PREALLOC_MODE_OFF;
}
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
if (!file_opts->has_extent_size_hint) {
file_opts->extent_size_hint = 1 * MiB;
}
if (file_opts->extent_size_hint > UINT32_MAX) {
result = -EINVAL;
error_setg(errp, "Extent size hint is too large");
goto out;
}
/* Create file */
fd = qemu_create(file_opts->filename, O_RDWR | O_BINARY, 0644, errp);
if (fd < 0) {
result = -errno;
goto out;
}
/* Take permissions: We want to discard everything, so we need
* BLK_PERM_WRITE; and truncation to the desired size requires
* BLK_PERM_RESIZE.
* On the other hand, we cannot share the RESIZE permission
* because we promise that after this function, the file has the
* size given in the options. If someone else were to resize it
* concurrently, we could not guarantee that.
* Note that after this function, we can no longer guarantee that
* the file is not touched by a third party, so it may be resized
* then. */
perm = BLK_PERM_WRITE | BLK_PERM_RESIZE;
shared = BLK_PERM_ALL & ~BLK_PERM_RESIZE;
/* Step one: Take locks */
result = raw_apply_lock_bytes(NULL, fd, perm, ~shared, false, errp);
if (result < 0) {
goto out_close;
}
/* Step two: Check that nobody else has taken conflicting locks */
result = raw_check_lock_bytes(fd, perm, shared, errp);
if (result < 0) {
error_append_hint(errp,
"Is another process using the image [%s]?\n",
file_opts->filename);
goto out_unlock;
}
/* Clear the file by truncating it to 0 */
result = raw_regular_truncate(NULL, fd, 0, PREALLOC_MODE_OFF, errp);
if (result < 0) {
goto out_unlock;
}
if (file_opts->nocow) {
#ifdef __linux__
/* Set NOCOW flag to solve performance issue on fs like btrfs.
* This is an optimisation. The FS_IOC_SETFLAGS ioctl return value
* will be ignored since any failure of this operation should not
* block the left work.
*/
int attr;
if (ioctl(fd, FS_IOC_GETFLAGS, &attr) == 0) {
attr |= FS_NOCOW_FL;
ioctl(fd, FS_IOC_SETFLAGS, &attr);
}
#endif
}
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
#ifdef FS_IOC_FSSETXATTR
/*
* Try to set the extent size hint. Failure is not fatal, and a warning is
* only printed if the option was explicitly specified.
*/
{
struct fsxattr attr;
result = ioctl(fd, FS_IOC_FSGETXATTR, &attr);
if (result == 0) {
attr.fsx_xflags |= FS_XFLAG_EXTSIZE;
attr.fsx_extsize = file_opts->extent_size_hint;
result = ioctl(fd, FS_IOC_FSSETXATTR, &attr);
}
if (result < 0 && file_opts->has_extent_size_hint &&
file_opts->extent_size_hint)
{
warn_report("Failed to set extent size hint: %s",
strerror(errno));
}
}
#endif
/* Resize and potentially preallocate the file to the desired
* final size */
result = raw_regular_truncate(NULL, fd, file_opts->size,
file_opts->preallocation, errp);
if (result < 0) {
goto out_unlock;
}
out_unlock:
raw_apply_lock_bytes(NULL, fd, 0, 0, true, &local_err);
if (local_err) {
/* The above call should not fail, and if it does, that does
* not mean the whole creation operation has failed. So
* report it the user for their convenience, but do not report
* it to the caller. */
warn_report_err(local_err);
}
out_close:
if (qemu_close(fd) != 0 && result == 0) {
result = -errno;
error_setg_errno(errp, -result, "Could not close the new file");
}
out:
return result;
}
static int coroutine_fn raw_co_create_opts(BlockDriver *drv,
const char *filename,
QemuOpts *opts,
Error **errp)
{
BlockdevCreateOptions options;
int64_t total_size = 0;
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
int64_t extent_size_hint = 0;
bool has_extent_size_hint = false;
bool nocow = false;
PreallocMode prealloc;
char *buf = NULL;
Error *local_err = NULL;
/* Skip file: protocol prefix */
strstart(filename, "file:", &filename);
/* Read out options */
total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
BDRV_SECTOR_SIZE);
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
if (qemu_opt_get(opts, BLOCK_OPT_EXTENT_SIZE_HINT)) {
has_extent_size_hint = true;
extent_size_hint =
qemu_opt_get_size_del(opts, BLOCK_OPT_EXTENT_SIZE_HINT, -1);
}
nocow = qemu_opt_get_bool(opts, BLOCK_OPT_NOCOW, false);
buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC);
prealloc = qapi_enum_parse(&PreallocMode_lookup, buf,
PREALLOC_MODE_OFF, &local_err);
g_free(buf);
if (local_err) {
error_propagate(errp, local_err);
return -EINVAL;
}
options = (BlockdevCreateOptions) {
.driver = BLOCKDEV_DRIVER_FILE,
.u.file = {
.filename = (char *) filename,
.size = total_size,
.has_preallocation = true,
.preallocation = prealloc,
.has_nocow = true,
.nocow = nocow,
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
.has_extent_size_hint = has_extent_size_hint,
.extent_size_hint = extent_size_hint,
},
};
return raw_co_create(&options, errp);
}
static int coroutine_fn raw_co_delete_file(BlockDriverState *bs,
Error **errp)
{
struct stat st;
int ret;
if (!(stat(bs->filename, &st) == 0) || !S_ISREG(st.st_mode)) {
error_setg_errno(errp, ENOENT, "%s is not a regular file",
bs->filename);
return -ENOENT;
}
ret = unlink(bs->filename);
if (ret < 0) {
ret = -errno;
error_setg_errno(errp, -ret, "Error when deleting file %s",
bs->filename);
}
return ret;
}
/*
* Find allocation range in @bs around offset @start.
* May change underlying file descriptor's file offset.
* If @start is not in a hole, store @start in @data, and the
* beginning of the next hole in @hole, and return 0.
* If @start is in a non-trailing hole, store @start in @hole and the
* beginning of the next non-hole in @data, and return 0.
* If @start is in a trailing hole or beyond EOF, return -ENXIO.
* If we can't find out, return a negative errno other than -ENXIO.
*/
static int find_allocation(BlockDriverState *bs, off_t start,
off_t *data, off_t *hole)
{
#if defined SEEK_HOLE && defined SEEK_DATA
BDRVRawState *s = bs->opaque;
off_t offs;
/*
* SEEK_DATA cases:
* D1. offs == start: start is in data
* D2. offs > start: start is in a hole, next data at offs
* D3. offs < 0, errno = ENXIO: either start is in a trailing hole
* or start is beyond EOF
* If the latter happens, the file has been truncated behind
* our back since we opened it. All bets are off then.
* Treating like a trailing hole is simplest.
* D4. offs < 0, errno != ENXIO: we learned nothing
*/
offs = lseek(s->fd, start, SEEK_DATA);
if (offs < 0) {
return -errno; /* D3 or D4 */
}
if (offs < start) {
/* This is not a valid return by lseek(). We are safe to just return
* -EIO in this case, and we'll treat it like D4. */
return -EIO;
}
if (offs > start) {
/* D2: in hole, next data at offs */
*hole = start;
*data = offs;
return 0;
}
/* D1: in data, end not yet known */
/*
* SEEK_HOLE cases:
* H1. offs == start: start is in a hole
* If this happens here, a hole has been dug behind our back
* since the previous lseek().
* H2. offs > start: either start is in data, next hole at offs,
* or start is in trailing hole, EOF at offs
* Linux treats trailing holes like any other hole: offs ==
* start. Solaris seeks to EOF instead: offs > start (blech).
* If that happens here, a hole has been dug behind our back
* since the previous lseek().
* H3. offs < 0, errno = ENXIO: start is beyond EOF
* If this happens, the file has been truncated behind our
* back since we opened it. Treat it like a trailing hole.
* H4. offs < 0, errno != ENXIO: we learned nothing
* Pretend we know nothing at all, i.e. "forget" about D1.
*/
offs = lseek(s->fd, start, SEEK_HOLE);
if (offs < 0) {
return -errno; /* D1 and (H3 or H4) */
}
if (offs < start) {
/* This is not a valid return by lseek(). We are safe to just return
* -EIO in this case, and we'll treat it like H4. */
return -EIO;
}
if (offs > start) {
/*
* D1 and H2: either in data, next hole at offs, or it was in
* data but is now in a trailing hole. In the latter case,
* all bets are off. Treating it as if it there was data all
* the way to EOF is safe, so simply do that.
*/
*data = start;
*hole = offs;
return 0;
}
/* D1 and H1 */
return -EBUSY;
#else
return -ENOTSUP;
#endif
}
/*
* Returns the allocation status of the specified offset.
*
* The block layer guarantees 'offset' and 'bytes' are within bounds.
*
* 'pnum' is set to the number of bytes (including and immediately following
* the specified offset) that are known to be in the same
* allocated/unallocated state.
*
* 'bytes' is the max value 'pnum' should be set to.
*/
static int coroutine_fn raw_co_block_status(BlockDriverState *bs,
bool want_zero,
int64_t offset,
int64_t bytes, int64_t *pnum,
int64_t *map,
BlockDriverState **file)
{
off_t data = 0, hole = 0;
int ret;
block/file-posix: Unaligned O_DIRECT block-status Currently, qemu crashes whenever someone queries the block status of an unaligned image tail of an O_DIRECT image: $ echo > foo $ qemu-img map --image-opts driver=file,filename=foo,cache.direct=on Offset Length Mapped to File qemu-img: block/io.c:2093: bdrv_co_block_status: Assertion `*pnum && QEMU_IS_ALIGNED(*pnum, align) && align > offset - aligned_offset' failed. This is because bdrv_co_block_status() checks that the result returned by the driver's implementation is aligned to the request_alignment, but file-posix can fail to do so, which is actually mentioned in a comment there: "[...] possibly including a partial sector at EOF". Fix this by rounding up those partial sectors. There are two possible alternative fixes: (1) We could refuse to open unaligned image files with O_DIRECT altogether. That sounds reasonable until you realize that qcow2 does necessarily not fill up its metadata clusters, and that nobody runs qemu-img create with O_DIRECT. Therefore, unpreallocated qcow2 files usually have an unaligned image tail. (2) bdrv_co_block_status() could ignore unaligned tails. It actually throws away everything past the EOF already, so that sounds reasonable. Unfortunately, the block layer knows file lengths only with a granularity of BDRV_SECTOR_SIZE, so bdrv_co_block_status() usually would have to guess whether its file length information is inexact or whether the driver is broken. Fixing what raw_co_block_status() returns is the safest thing to do. There seems to be no other block driver that sets request_alignment and does not make sure that it always returns aligned values. Cc: qemu-stable@nongnu.org Signed-off-by: Max Reitz <mreitz@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-05-15 06:15:40 +02:00
assert(QEMU_IS_ALIGNED(offset | bytes, bs->bl.request_alignment));
ret = fd_open(bs);
if (ret < 0) {
return ret;
}
if (!want_zero) {
*pnum = bytes;
*map = offset;
*file = bs;
return BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID;
}
ret = find_allocation(bs, offset, &data, &hole);
if (ret == -ENXIO) {
/* Trailing hole */
*pnum = bytes;
ret = BDRV_BLOCK_ZERO;
} else if (ret < 0) {
/* No info available, so pretend there are no holes */
*pnum = bytes;
ret = BDRV_BLOCK_DATA;
} else if (data == offset) {
/* On a data extent, compute bytes to the end of the extent,
* possibly including a partial sector at EOF. */
*pnum = MIN(bytes, hole - offset);
block/file-posix: Unaligned O_DIRECT block-status Currently, qemu crashes whenever someone queries the block status of an unaligned image tail of an O_DIRECT image: $ echo > foo $ qemu-img map --image-opts driver=file,filename=foo,cache.direct=on Offset Length Mapped to File qemu-img: block/io.c:2093: bdrv_co_block_status: Assertion `*pnum && QEMU_IS_ALIGNED(*pnum, align) && align > offset - aligned_offset' failed. This is because bdrv_co_block_status() checks that the result returned by the driver's implementation is aligned to the request_alignment, but file-posix can fail to do so, which is actually mentioned in a comment there: "[...] possibly including a partial sector at EOF". Fix this by rounding up those partial sectors. There are two possible alternative fixes: (1) We could refuse to open unaligned image files with O_DIRECT altogether. That sounds reasonable until you realize that qcow2 does necessarily not fill up its metadata clusters, and that nobody runs qemu-img create with O_DIRECT. Therefore, unpreallocated qcow2 files usually have an unaligned image tail. (2) bdrv_co_block_status() could ignore unaligned tails. It actually throws away everything past the EOF already, so that sounds reasonable. Unfortunately, the block layer knows file lengths only with a granularity of BDRV_SECTOR_SIZE, so bdrv_co_block_status() usually would have to guess whether its file length information is inexact or whether the driver is broken. Fixing what raw_co_block_status() returns is the safest thing to do. There seems to be no other block driver that sets request_alignment and does not make sure that it always returns aligned values. Cc: qemu-stable@nongnu.org Signed-off-by: Max Reitz <mreitz@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2019-05-15 06:15:40 +02:00
/*
* We are not allowed to return partial sectors, though, so
* round up if necessary.
*/
if (!QEMU_IS_ALIGNED(*pnum, bs->bl.request_alignment)) {
int64_t file_length = raw_getlength(bs);
if (file_length > 0) {
/* Ignore errors, this is just a safeguard */
assert(hole == file_length);
}
*pnum = ROUND_UP(*pnum, bs->bl.request_alignment);
}
ret = BDRV_BLOCK_DATA;
} else {
/* On a hole, compute bytes to the beginning of the next extent. */
assert(hole == offset);
*pnum = MIN(bytes, data - offset);
ret = BDRV_BLOCK_ZERO;
}
*map = offset;
*file = bs;
return ret | BDRV_BLOCK_OFFSET_VALID;
}
#if defined(__linux__)
/* Verify that the file is not in the page cache */
static void check_cache_dropped(BlockDriverState *bs, Error **errp)
{
const size_t window_size = 128 * 1024 * 1024;
BDRVRawState *s = bs->opaque;
void *window = NULL;
size_t length = 0;
unsigned char *vec;
size_t page_size;
off_t offset;
off_t end;
/* mincore(2) page status information requires 1 byte per page */
page_size = sysconf(_SC_PAGESIZE);
vec = g_malloc(DIV_ROUND_UP(window_size, page_size));
end = raw_getlength(bs);
for (offset = 0; offset < end; offset += window_size) {
void *new_window;
size_t new_length;
size_t vec_end;
size_t i;
int ret;
/* Unmap previous window if size has changed */
new_length = MIN(end - offset, window_size);
if (new_length != length) {
munmap(window, length);
window = NULL;
length = 0;
}
new_window = mmap(window, new_length, PROT_NONE, MAP_PRIVATE,
s->fd, offset);
if (new_window == MAP_FAILED) {
error_setg_errno(errp, errno, "mmap failed");
break;
}
window = new_window;
length = new_length;
ret = mincore(window, length, vec);
if (ret < 0) {
error_setg_errno(errp, errno, "mincore failed");
break;
}
vec_end = DIV_ROUND_UP(length, page_size);
for (i = 0; i < vec_end; i++) {
if (vec[i] & 0x1) {
break;
}
}
if (i < vec_end) {
error_setg(errp, "page cache still in use!");
break;
}
}
if (window) {
munmap(window, length);
}
g_free(vec);
}
#endif /* __linux__ */
static void coroutine_fn raw_co_invalidate_cache(BlockDriverState *bs,
Error **errp)
{
BDRVRawState *s = bs->opaque;
int ret;
ret = fd_open(bs);
if (ret < 0) {
error_setg_errno(errp, -ret, "The file descriptor is not open");
return;
}
if (!s->drop_cache) {
return;
}
if (s->open_flags & O_DIRECT) {
return; /* No host kernel page cache */
}
#if defined(__linux__)
/* This sets the scene for the next syscall... */
ret = bdrv_co_flush(bs);
if (ret < 0) {
error_setg_errno(errp, -ret, "flush failed");
return;
}
/* Linux does not invalidate pages that are dirty, locked, or mmapped by a
* process. These limitations are okay because we just fsynced the file,
* we don't use mmap, and the file should not be in use by other processes.
*/
ret = posix_fadvise(s->fd, 0, 0, POSIX_FADV_DONTNEED);
if (ret != 0) { /* the return value is a positive errno */
error_setg_errno(errp, ret, "fadvise failed");
return;
}
if (s->check_cache_dropped) {
check_cache_dropped(bs, errp);
}
#else /* __linux__ */
/* Do nothing. Live migration to a remote host with cache.direct=off is
* unsupported on other host operating systems. Cache consistency issues
* may occur but no error is reported here, partly because that's the
* historical behavior and partly because it's hard to differentiate valid
* configurations that should not cause errors.
*/
#endif /* !__linux__ */
}
static void raw_account_discard(BDRVRawState *s, uint64_t nbytes, int ret)
{
if (ret) {
s->stats.discard_nb_failed++;
} else {
s->stats.discard_nb_ok++;
s->stats.discard_bytes_ok += nbytes;
}
}
static coroutine_fn int
raw_do_pdiscard(BlockDriverState *bs, int64_t offset, int bytes, bool blkdev)
{
BDRVRawState *s = bs->opaque;
RawPosixAIOData acb;
int ret;
acb = (RawPosixAIOData) {
.bs = bs,
.aio_fildes = s->fd,
.aio_type = QEMU_AIO_DISCARD,
.aio_offset = offset,
.aio_nbytes = bytes,
};
if (blkdev) {
acb.aio_type |= QEMU_AIO_BLKDEV;
}
ret = raw_thread_pool_submit(bs, handle_aiocb_discard, &acb);
raw_account_discard(s, bytes, ret);
return ret;
}
static coroutine_fn int
raw_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)
{
return raw_do_pdiscard(bs, offset, bytes, false);
}
static int coroutine_fn
raw_do_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int bytes,
BdrvRequestFlags flags, bool blkdev)
{
BDRVRawState *s = bs->opaque;
RawPosixAIOData acb;
ThreadPoolFunc *handler;
#ifdef CONFIG_FALLOCATE
if (offset + bytes > bs->total_sectors * BDRV_SECTOR_SIZE) {
BdrvTrackedRequest *req;
/*
* This is a workaround for a bug in the Linux XFS driver,
* where writes submitted through the AIO interface will be
* discarded if they happen beyond a concurrently running
* fallocate() that increases the file length (i.e., both the
* write and the fallocate() happen beyond the EOF).
*
* To work around it, we extend the tracked request for this
* zero write until INT64_MAX (effectively infinity), and mark
* it as serializing.
*
* We have to enable this workaround for all filesystems and
* AIO modes (not just XFS with aio=native), because for
* remote filesystems we do not know the host configuration.
*/
req = bdrv_co_get_self_request(bs);
assert(req);
assert(req->type == BDRV_TRACKED_WRITE);
assert(req->offset <= offset);
assert(req->offset + req->bytes >= offset + bytes);
block: introduce BDRV_MAX_LENGTH We are going to modify block layer to work with 64bit requests. And first step is moving to int64_t type for both offset and bytes arguments in all block request related functions. It's mostly safe (when widening signed or unsigned int to int64_t), but switching from uint64_t is questionable. So, let's first establish the set of requests we want to work with. First signed int64_t should be enough, as off_t is signed anyway. Then, obviously offset + bytes should not overflow. And most interesting: (offset + bytes) being aligned up should not overflow as well. Aligned to what alignment? First thing that comes in mind is bs->bl.request_alignment, as we align up request to this alignment. But there is another thing: look at bdrv_mark_request_serialising(). It aligns request up to some given alignment. And this parameter may be bdrv_get_cluster_size(), which is often a lot greater than bs->bl.request_alignment. Note also, that bdrv_mark_request_serialising() uses signed int64_t for calculations. So, actually, we already depend on some restrictions. Happily, bdrv_get_cluster_size() returns int and bs->bl.request_alignment has 32bit unsigned type, but defined to be a power of 2 less than INT_MAX. So, we may establish, that INT_MAX is absolute maximum for any kind of alignment that may occur with the request. Note, that bdrv_get_cluster_size() is not documented to return power of 2, still bdrv_mark_request_serialising() behaves like it is. Also, backup uses bdi.cluster_size and is not prepared to it not being power of 2. So, let's establish that Qemu supports only power-of-2 clusters and alignments. So, alignment can't be greater than 2^30. Finally to be safe with calculations, to not calculate different maximums for different nodes (depending on cluster size and request_alignment), let's simply set QEMU_ALIGN_DOWN(INT64_MAX, 2^30) as absolute maximum bytes length for Qemu. Actually, it's not much less than INT64_MAX. OK, then, let's apply it to block/io. Let's consider all block/io entry points of offset/bytes: 4 bytes/offset interface functions: bdrv_co_preadv_part(), bdrv_co_pwritev_part(), bdrv_co_copy_range_internal() and bdrv_co_pdiscard() and we check them all with bdrv_check_request(). We also have one entry point with only offset: bdrv_co_truncate(). Check the offset. And one public structure: BdrvTrackedRequest. Happily, it has only three external users: file-posix.c: adopted by this patch write-threshold.c: only read fields test-write-threshold.c: sets obviously small constant values Better is to make the structure private and add corresponding interfaces.. Still it's not obvious what kind of interface is needed for file-posix.c. Let's keep it public but add corresponding assertions. After this patch we'll convert functions in block/io.c to int64_t bytes and offset parameters. We can assume that offset/bytes pair always satisfy new restrictions, and make corresponding assertions where needed. If we reach some offset/bytes point in block/io.c missing bdrv_check_request() it is considered a bug. As well, if block/io.c modifies a offset/bytes request, expanding it more then aligning up to request_alignment, it's a bug too. For all io requests except for discard we keep for now old restriction of 32bit request length. iotest 206 output error message changed, as now test disk size is larger than new limit. Add one more test case with new maximum disk size to cover too-big-L1 case. Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> Message-Id: <20201203222713.13507-5-vsementsov@virtuozzo.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-12-03 23:27:13 +01:00
req->bytes = BDRV_MAX_LENGTH - req->offset;
bdrv_check_request(req->offset, req->bytes, &error_abort);
bdrv_make_request_serialising(req, bs->bl.request_alignment);
}
#endif
acb = (RawPosixAIOData) {
.bs = bs,
.aio_fildes = s->fd,
.aio_type = QEMU_AIO_WRITE_ZEROES,
.aio_offset = offset,
.aio_nbytes = bytes,
};
if (blkdev) {
acb.aio_type |= QEMU_AIO_BLKDEV;
}
if (flags & BDRV_REQ_NO_FALLBACK) {
acb.aio_type |= QEMU_AIO_NO_FALLBACK;
}
if (flags & BDRV_REQ_MAY_UNMAP) {
acb.aio_type |= QEMU_AIO_DISCARD;
handler = handle_aiocb_write_zeroes_unmap;
} else {
handler = handle_aiocb_write_zeroes;
}
return raw_thread_pool_submit(bs, handler, &acb);
}
static int coroutine_fn raw_co_pwrite_zeroes(
BlockDriverState *bs, int64_t offset,
int bytes, BdrvRequestFlags flags)
{
return raw_do_pwrite_zeroes(bs, offset, bytes, flags, false);
}
static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
{
return 0;
}
static BlockStatsSpecificFile get_blockstats_specific_file(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
return (BlockStatsSpecificFile) {
.discard_nb_ok = s->stats.discard_nb_ok,
.discard_nb_failed = s->stats.discard_nb_failed,
.discard_bytes_ok = s->stats.discard_bytes_ok,
};
}
static BlockStatsSpecific *raw_get_specific_stats(BlockDriverState *bs)
{
BlockStatsSpecific *stats = g_new(BlockStatsSpecific, 1);
stats->driver = BLOCKDEV_DRIVER_FILE;
stats->u.file = get_blockstats_specific_file(bs);
return stats;
}
static BlockStatsSpecific *hdev_get_specific_stats(BlockDriverState *bs)
{
BlockStatsSpecific *stats = g_new(BlockStatsSpecific, 1);
stats->driver = BLOCKDEV_DRIVER_HOST_DEVICE;
stats->u.host_device = get_blockstats_specific_file(bs);
return stats;
}
static QemuOptsList raw_create_opts = {
.name = "raw-create-opts",
.head = QTAILQ_HEAD_INITIALIZER(raw_create_opts.head),
.desc = {
{
.name = BLOCK_OPT_SIZE,
.type = QEMU_OPT_SIZE,
.help = "Virtual disk size"
},
{
.name = BLOCK_OPT_NOCOW,
.type = QEMU_OPT_BOOL,
.help = "Turn off copy-on-write (valid only on btrfs)"
},
{
.name = BLOCK_OPT_PREALLOC,
.type = QEMU_OPT_STRING,
.help = "Preallocation mode (allowed values: off"
#ifdef CONFIG_POSIX_FALLOCATE
", falloc"
#endif
", full)"
},
file-posix: Mitigate file fragmentation with extent size hints Especially when O_DIRECT is used with image files so that the page cache indirection can't cause a merge of allocating requests, the file will fragment on the file system layer, with a potentially very small fragment size (this depends on the requests the guest sent). On Linux, fragmentation can be reduced by setting an extent size hint when creating the file (at least on XFS, it can't be set any more after the first extent has been allocated), basically giving raw files a "cluster size" for allocation. This adds a create option to set the extent size hint, and changes the default from not setting a hint to setting it to 1 MB. The main reason why qcow2 defaults to smaller cluster sizes is that COW becomes more expensive, which is not an issue with raw files, so we can choose a larger size. The tradeoff here is only potentially wasted disk space. For qcow2 (or other image formats) over file-posix, the advantage should even be greater because they grow sequentially without leaving holes, so there won't be wasted space. Setting even larger extent size hints for such images may make sense. This can be done with the new option, but let's keep the default conservative for now. The effect is very visible with a test that intentionally creates a badly fragmented file with qemu-img bench (the time difference while creating the file is already remarkable) and then looks at the number of extents and the time a simple "qemu-img map" takes. Without an extent size hint: $ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 25.848 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 19.616 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 2000000 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m1,279s user 0m0,043s sys 0m1,226s With the new default extent size hint of 1 MB: $ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576 $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192) Run completed in 11.833 seconds. $ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096 Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192) Run completed in 10.155 seconds. $ filefrag ~/tmp/test.raw /home/kwolf/tmp/test.raw: 178 extents found $ time ./qemu-img map ~/tmp/test.raw Offset Length Mapped to File 0 0x1e8480000 0 /home/kwolf/tmp/test.raw real 0m0,061s user 0m0,040s sys 0m0,014s Signed-off-by: Kevin Wolf <kwolf@redhat.com> Message-Id: <20200707142329.48303-1-kwolf@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2020-07-07 16:23:29 +02:00
{
.name = BLOCK_OPT_EXTENT_SIZE_HINT,
.type = QEMU_OPT_SIZE,
.help = "Extent size hint for the image file, 0 to disable"
},
{ /* end of list */ }
}
};
static int raw_check_perm(BlockDriverState *bs, uint64_t perm, uint64_t shared,
Error **errp)
{
BDRVRawState *s = bs->opaque;
BDRVRawReopenState *rs = NULL;
int open_flags;
int ret;
if (s->perm_change_fd) {
/*
* In the context of reopen, this function may be called several times
* (directly and recursively while change permissions of the parent).
* This is even true for children that don't inherit from the original
* reopen node, so s->reopen_state is not set.
*
* Ignore all but the first call.
*/
return 0;
}
if (s->reopen_state) {
/* We already have a new file descriptor to set permissions for */
assert(s->reopen_state->perm == perm);
assert(s->reopen_state->shared_perm == shared);
rs = s->reopen_state->opaque;
s->perm_change_fd = rs->fd;
s->perm_change_flags = rs->open_flags;
} else {
/* We may need a new fd if auto-read-only switches the mode */
ret = raw_reconfigure_getfd(bs, bs->open_flags, &open_flags, perm,
false, errp);
if (ret < 0) {
return ret;
} else if (ret != s->fd) {
s->perm_change_fd = ret;
s->perm_change_flags = open_flags;
}
}
/* Prepare permissions on old fd to avoid conflicts between old and new,
* but keep everything locked that new will need. */
ret = raw_handle_perm_lock(bs, RAW_PL_PREPARE, perm, shared, errp);
if (ret < 0) {
goto fail;
}
/* Copy locks to the new fd */
if (s->perm_change_fd && s->use_lock) {
ret = raw_apply_lock_bytes(NULL, s->perm_change_fd, perm, ~shared,
false, errp);
if (ret < 0) {
raw_handle_perm_lock(bs, RAW_PL_ABORT, 0, 0, NULL);
goto fail;
}
}
return 0;
fail:
if (s->perm_change_fd && !s->reopen_state) {
qemu_close(s->perm_change_fd);
}
s->perm_change_fd = 0;
return ret;
}
static void raw_set_perm(BlockDriverState *bs, uint64_t perm, uint64_t shared)
{
BDRVRawState *s = bs->opaque;
/* For reopen, we have already switched to the new fd (.bdrv_set_perm is
* called after .bdrv_reopen_commit) */
if (s->perm_change_fd && s->fd != s->perm_change_fd) {
qemu_close(s->fd);
s->fd = s->perm_change_fd;
s->open_flags = s->perm_change_flags;
}
s->perm_change_fd = 0;
raw_handle_perm_lock(bs, RAW_PL_COMMIT, perm, shared, NULL);
s->perm = perm;
s->shared_perm = shared;
}
static void raw_abort_perm_update(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
/* For reopen, .bdrv_reopen_abort is called afterwards and will close
* the file descriptor. */
if (s->perm_change_fd && !s->reopen_state) {
qemu_close(s->perm_change_fd);
}
s->perm_change_fd = 0;
raw_handle_perm_lock(bs, RAW_PL_ABORT, 0, 0, NULL);
}
static int coroutine_fn raw_co_copy_range_from(
BlockDriverState *bs, BdrvChild *src, uint64_t src_offset,
BdrvChild *dst, uint64_t dst_offset, uint64_t bytes,
BdrvRequestFlags read_flags, BdrvRequestFlags write_flags)
{
return bdrv_co_copy_range_to(src, src_offset, dst, dst_offset, bytes,
read_flags, write_flags);
}
static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs,
BdrvChild *src,
uint64_t src_offset,
BdrvChild *dst,
uint64_t dst_offset,
uint64_t bytes,
BdrvRequestFlags read_flags,
BdrvRequestFlags write_flags)
{
RawPosixAIOData acb;
BDRVRawState *s = bs->opaque;
BDRVRawState *src_s;
assert(dst->bs == bs);
if (src->bs->drv->bdrv_co_copy_range_to != raw_co_copy_range_to) {
return -ENOTSUP;
}
src_s = src->bs->opaque;
if (fd_open(src->bs) < 0 || fd_open(dst->bs) < 0) {
return -EIO;
}
acb = (RawPosixAIOData) {
.bs = bs,
.aio_type = QEMU_AIO_COPY_RANGE,
.aio_fildes = src_s->fd,
.aio_offset = src_offset,
.aio_nbytes = bytes,
.copy_range = {
.aio_fd2 = s->fd,
.aio_offset2 = dst_offset,
},
};
return raw_thread_pool_submit(bs, handle_aiocb_copy_range, &acb);
}
BlockDriver bdrv_file = {
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
.format_name = "file",
.protocol_name = "file",
.instance_size = sizeof(BDRVRawState),
.bdrv_needs_filename = true,
.bdrv_probe = NULL, /* no probe for protocols */
.bdrv_parse_filename = raw_parse_filename,
.bdrv_file_open = raw_open,
.bdrv_reopen_prepare = raw_reopen_prepare,
.bdrv_reopen_commit = raw_reopen_commit,
.bdrv_reopen_abort = raw_reopen_abort,
.bdrv_close = raw_close,
.bdrv_co_create = raw_co_create,
.bdrv_co_create_opts = raw_co_create_opts,
.bdrv_has_zero_init = bdrv_has_zero_init_1,
.bdrv_co_block_status = raw_co_block_status,
.bdrv_co_invalidate_cache = raw_co_invalidate_cache,
.bdrv_co_pwrite_zeroes = raw_co_pwrite_zeroes,
.bdrv_co_delete_file = raw_co_delete_file,
.bdrv_co_preadv = raw_co_preadv,
.bdrv_co_pwritev = raw_co_pwritev,
.bdrv_co_flush_to_disk = raw_co_flush_to_disk,
.bdrv_co_pdiscard = raw_co_pdiscard,
.bdrv_co_copy_range_from = raw_co_copy_range_from,
.bdrv_co_copy_range_to = raw_co_copy_range_to,
.bdrv_refresh_limits = raw_refresh_limits,
.bdrv_io_plug = raw_aio_plug,
.bdrv_io_unplug = raw_aio_unplug,
.bdrv_attach_aio_context = raw_aio_attach_aio_context,
.bdrv_co_truncate = raw_co_truncate,
.bdrv_getlength = raw_getlength,
.bdrv_get_info = raw_get_info,
.bdrv_get_allocated_file_size
= raw_get_allocated_file_size,
.bdrv_get_specific_stats = raw_get_specific_stats,
.bdrv_check_perm = raw_check_perm,
.bdrv_set_perm = raw_set_perm,
.bdrv_abort_perm_update = raw_abort_perm_update,
.create_opts = &raw_create_opts,
.mutable_opts = mutable_opts,
};
/***********************************************/
/* host device */
#if defined(__APPLE__) && defined(__MACH__)
static kern_return_t GetBSDPath(io_iterator_t mediaIterator, char *bsdPath,
CFIndex maxPathSize, int flags);
static char *FindEjectableOpticalMedia(io_iterator_t *mediaIterator)
{
kern_return_t kernResult = KERN_FAILURE;
mach_port_t masterPort;
CFMutableDictionaryRef classesToMatch;
const char *matching_array[] = {kIODVDMediaClass, kIOCDMediaClass};
char *mediaType = NULL;
kernResult = IOMasterPort( MACH_PORT_NULL, &masterPort );
if ( KERN_SUCCESS != kernResult ) {
printf( "IOMasterPort returned %d\n", kernResult );
}
int index;
for (index = 0; index < ARRAY_SIZE(matching_array); index++) {
classesToMatch = IOServiceMatching(matching_array[index]);
if (classesToMatch == NULL) {
error_report("IOServiceMatching returned NULL for %s",
matching_array[index]);
continue;
}
CFDictionarySetValue(classesToMatch, CFSTR(kIOMediaEjectableKey),
kCFBooleanTrue);
kernResult = IOServiceGetMatchingServices(masterPort, classesToMatch,
mediaIterator);
if (kernResult != KERN_SUCCESS) {
error_report("Note: IOServiceGetMatchingServices returned %d",
kernResult);
continue;
}
/* If a match was found, leave the loop */
if (*mediaIterator != 0) {
trace_file_FindEjectableOpticalMedia(matching_array[index]);
mediaType = g_strdup(matching_array[index]);
break;
}
}
return mediaType;
}
kern_return_t GetBSDPath(io_iterator_t mediaIterator, char *bsdPath,
CFIndex maxPathSize, int flags)
{
io_object_t nextMedia;
kern_return_t kernResult = KERN_FAILURE;
*bsdPath = '\0';
nextMedia = IOIteratorNext( mediaIterator );
if ( nextMedia )
{
CFTypeRef bsdPathAsCFString;
bsdPathAsCFString = IORegistryEntryCreateCFProperty( nextMedia, CFSTR( kIOBSDNameKey ), kCFAllocatorDefault, 0 );
if ( bsdPathAsCFString ) {
size_t devPathLength;
strcpy( bsdPath, _PATH_DEV );
if (flags & BDRV_O_NOCACHE) {
strcat(bsdPath, "r");
}
devPathLength = strlen( bsdPath );
if ( CFStringGetCString( bsdPathAsCFString, bsdPath + devPathLength, maxPathSize - devPathLength, kCFStringEncodingASCII ) ) {
kernResult = KERN_SUCCESS;
}
CFRelease( bsdPathAsCFString );
}
IOObjectRelease( nextMedia );
}
return kernResult;
}
/* Sets up a real cdrom for use in QEMU */
static bool setup_cdrom(char *bsd_path, Error **errp)
{
int index, num_of_test_partitions = 2, fd;
char test_partition[MAXPATHLEN];
bool partition_found = false;
/* look for a working partition */
for (index = 0; index < num_of_test_partitions; index++) {
snprintf(test_partition, sizeof(test_partition), "%ss%d", bsd_path,
index);
fd = qemu_open(test_partition, O_RDONLY | O_BINARY | O_LARGEFILE, NULL);
if (fd >= 0) {
partition_found = true;
qemu_close(fd);
break;
}
}
/* if a working partition on the device was not found */
if (partition_found == false) {
error_setg(errp, "Failed to find a working partition on disc");
} else {
trace_file_setup_cdrom(test_partition);
pstrcpy(bsd_path, MAXPATHLEN, test_partition);
}
return partition_found;
}
/* Prints directions on mounting and unmounting a device */
static void print_unmounting_directions(const char *file_name)
{
error_report("If device %s is mounted on the desktop, unmount"
" it first before using it in QEMU", file_name);
error_report("Command to unmount device: diskutil unmountDisk %s",
file_name);
error_report("Command to mount device: diskutil mountDisk %s", file_name);
}
#endif /* defined(__APPLE__) && defined(__MACH__) */
static int hdev_probe_device(const char *filename)
{
struct stat st;
/* allow a dedicated CD-ROM driver to match with a higher priority */
if (strstart(filename, "/dev/cdrom", NULL))
return 50;
if (stat(filename, &st) >= 0 &&
(S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) {
return 100;
}
return 0;
}
static void hdev_parse_filename(const char *filename, QDict *options,
Error **errp)
{
2017-05-22 21:52:16 +02:00
bdrv_parse_filename_strip_prefix(filename, "host_device:", options);
}
static bool hdev_is_sg(BlockDriverState *bs)
{
#if defined(__linux__)
BDRVRawState *s = bs->opaque;
struct stat st;
struct sg_scsi_id scsiid;
int sg_version;
int ret;
if (stat(bs->filename, &st) < 0 || !S_ISCHR(st.st_mode)) {
return false;
}
ret = ioctl(s->fd, SG_GET_VERSION_NUM, &sg_version);
if (ret < 0) {
return false;
}
ret = ioctl(s->fd, SG_GET_SCSI_ID, &scsiid);
if (ret >= 0) {
trace_file_hdev_is_sg(scsiid.scsi_type, sg_version);
return true;
}
#endif
return false;
}
static int hdev_open(BlockDriverState *bs, QDict *options, int flags,
Error **errp)
{
BDRVRawState *s = bs->opaque;
int ret;
#if defined(__APPLE__) && defined(__MACH__)
block: Document -drive problematic code and bugs -blockdev and blockdev_add convert their arguments via QObject to BlockdevOptions for qmp_blockdev_add(), which converts them back to QObject, then to a flattened QDict. The QDict's members are typed according to the QAPI schema. -drive converts its argument via QemuOpts to a (flat) QDict. This QDict's members are all QString. Thus, the QType of a flat QDict member depends on whether it comes from -drive or -blockdev/blockdev_add, except when the QAPI type maps to QString, which is the case for 'str' and enumeration types. The block layer core extracts generic configuration from the flat QDict, and the block driver extracts driver-specific configuration. Both commonly do so by converting (parts of) the flat QDict to QemuOpts, which turns all values into strings. Not exactly elegant, but correct. However, A few places access the flat QDict directly: * Most of them access members that are always QString. Correct. * bdrv_open_inherit() accesses a boolean, carefully. Correct. * nfs_config() uses a QObject input visitor. Correct only because the visited type contains nothing but QStrings. * nbd_config() and ssh_config() use a QObject input visitor, and the visited types contain non-QStrings: InetSocketAddress members @numeric, @to, @ipv4, @ipv6. -drive works as long as you don't try to use them (they're all optional). @to is ignored anyway. Reproducer: -drive driver=ssh,server.host=h,server.port=22,server.ipv4,path=p -drive driver=nbd,server.type=inet,server.data.host=h,server.data.port=22,server.data.ipv4 both fail with "Invalid parameter type for 'data.ipv4', expected: boolean" Add suitable comments to all these places. Mark the buggy ones FIXME. "Fortunately", -drive's driver-specific options are entirely undocumented. Signed-off-by: Markus Armbruster <armbru@redhat.com> Message-id: 1490895797-29094-5-git-send-email-armbru@redhat.com [mreitz: Fixed two typos] Reviewed-by: Eric Blake <eblake@redhat.com> Signed-off-by: Max Reitz <mreitz@redhat.com>
2017-03-30 19:43:12 +02:00
/*
* Caution: while qdict_get_str() is fine, getting non-string types
* would require more care. When @options come from -blockdev or
* blockdev_add, its members are typed according to the QAPI
* schema, but when they come from -drive, they're all QString.
*/
const char *filename = qdict_get_str(options, "filename");
char bsd_path[MAXPATHLEN] = "";
bool error_occurred = false;
/* If using a real cdrom */
if (strcmp(filename, "/dev/cdrom") == 0) {
char *mediaType = NULL;
kern_return_t ret_val;
io_iterator_t mediaIterator = 0;
mediaType = FindEjectableOpticalMedia(&mediaIterator);
if (mediaType == NULL) {
error_setg(errp, "Please make sure your CD/DVD is in the optical"
" drive");
error_occurred = true;
goto hdev_open_Mac_error;
}
ret_val = GetBSDPath(mediaIterator, bsd_path, sizeof(bsd_path), flags);
if (ret_val != KERN_SUCCESS) {
error_setg(errp, "Could not get BSD path for optical drive");
error_occurred = true;
goto hdev_open_Mac_error;
}
/* If a real optical drive was not found */
if (bsd_path[0] == '\0') {
error_setg(errp, "Failed to obtain bsd path for optical drive");
error_occurred = true;
goto hdev_open_Mac_error;
}
/* If using a cdrom disc and finding a partition on the disc failed */
if (strncmp(mediaType, kIOCDMediaClass, 9) == 0 &&
setup_cdrom(bsd_path, errp) == false) {
print_unmounting_directions(bsd_path);
error_occurred = true;
goto hdev_open_Mac_error;
}
qdict_put_str(options, "filename", bsd_path);
hdev_open_Mac_error:
g_free(mediaType);
if (mediaIterator) {
IOObjectRelease(mediaIterator);
}
if (error_occurred) {
return -ENOENT;
}
}
#endif /* defined(__APPLE__) && defined(__MACH__) */
s->type = FTYPE_FILE;
error: Eliminate error_propagate() with Coccinelle, part 1 When all we do with an Error we receive into a local variable is propagating to somewhere else, we can just as well receive it there right away. Convert if (!foo(..., &err)) { ... error_propagate(errp, err); ... return ... } to if (!foo(..., errp)) { ... ... return ... } where nothing else needs @err. Coccinelle script: @rule1 forall@ identifier fun, err, errp, lbl; expression list args, args2; binary operator op; constant c1, c2; symbol false; @@ if ( ( - fun(args, &err, args2) + fun(args, errp, args2) | - !fun(args, &err, args2) + !fun(args, errp, args2) | - fun(args, &err, args2) op c1 + fun(args, errp, args2) op c1 ) ) { ... when != err when != lbl: when strict - error_propagate(errp, err); ... when != err ( return; | return c2; | return false; ) } @rule2 forall@ identifier fun, err, errp, lbl; expression list args, args2; expression var; binary operator op; constant c1, c2; symbol false; @@ - var = fun(args, &err, args2); + var = fun(args, errp, args2); ... when != err if ( ( var | !var | var op c1 ) ) { ... when != err when != lbl: when strict - error_propagate(errp, err); ... when != err ( return; | return c2; | return false; | return var; ) } @depends on rule1 || rule2@ identifier err; @@ - Error *err = NULL; ... when != err Not exactly elegant, I'm afraid. The "when != lbl:" is necessary to avoid transforming if (fun(args, &err)) { goto out } ... out: error_propagate(errp, err); even though other paths to label out still need the error_propagate(). For an actual example, see sclp_realize(). Without the "when strict", Coccinelle transforms vfio_msix_setup(), incorrectly. I don't know what exactly "when strict" does, only that it helps here. The match of return is narrower than what I want, but I can't figure out how to express "return where the operand doesn't use @err". For an example where it's too narrow, see vfio_intx_enable(). Silently fails to convert hw/arm/armsse.c, because Coccinelle gets confused by ARMSSE being used both as typedef and function-like macro there. Converted manually. Line breaks tidied up manually. One nested declaration of @local_err deleted manually. Preexisting unwanted blank line dropped in hw/riscv/sifive_e.c. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20200707160613.848843-35-armbru@redhat.com>
2020-07-07 18:06:02 +02:00
ret = raw_open_common(bs, options, flags, 0, true, errp);
if (ret < 0) {
#if defined(__APPLE__) && defined(__MACH__)
if (*bsd_path) {
filename = bsd_path;
}
/* if a physical device experienced an error while being opened */
if (strncmp(filename, "/dev/", 5) == 0) {
print_unmounting_directions(filename);
}
#endif /* defined(__APPLE__) && defined(__MACH__) */
return ret;
}
/* Since this does ioctl the device must be already opened */
bs->sg = hdev_is_sg(bs);
return ret;
}
#if defined(__linux__)
static int coroutine_fn
hdev_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)
{
BDRVRawState *s = bs->opaque;
RawPosixAIOData acb;
int ret;
ret = fd_open(bs);
if (ret < 0) {
return ret;
}
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
if (req == SG_IO && s->pr_mgr) {
struct sg_io_hdr *io_hdr = buf;
if (io_hdr->cmdp[0] == PERSISTENT_RESERVE_OUT ||
io_hdr->cmdp[0] == PERSISTENT_RESERVE_IN) {
return pr_manager_execute(s->pr_mgr, bdrv_get_aio_context(bs),
s->fd, io_hdr);
scsi, file-posix: add support for persistent reservation management It is a common requirement for virtual machine to send persistent reservations, but this currently requires either running QEMU with CAP_SYS_RAWIO, or using out-of-tree patches that let an unprivileged QEMU bypass Linux's filter on SG_IO commands. As an alternative mechanism, the next patches will introduce a privileged helper to run persistent reservation commands without expanding QEMU's attack surface unnecessarily. The helper is invoked through a "pr-manager" QOM object, to which file-posix.c passes SG_IO requests for PERSISTENT RESERVE OUT and PERSISTENT RESERVE IN commands. For example: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -drive if=none,id=hd,driver=raw,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd or: $ qemu-system-x86_64 -device virtio-scsi \ -object pr-manager-helper,id=helper0,path=/var/run/qemu-pr-helper.sock -blockdev node-name=hd,driver=raw,file.driver=host_device,file.filename=/dev/sdb,file.pr-manager=helper0 -device scsi-block,drive=hd Multiple pr-manager implementations are conceivable and possible, though only one is implemented right now. For example, a pr-manager could: - talk directly to the multipath daemon from a privileged QEMU (i.e. QEMU links to libmpathpersist); this makes reservation work properly with multipath, but still requires CAP_SYS_RAWIO - use the Linux IOC_PR_* ioctls (they require CAP_SYS_ADMIN though) - more interestingly, implement reservations directly in QEMU through file system locks or a shared database (e.g. sqlite) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2017-08-21 18:58:56 +02:00
}
}
acb = (RawPosixAIOData) {
.bs = bs,
.aio_type = QEMU_AIO_IOCTL,
.aio_fildes = s->fd,
.aio_offset = 0,
.ioctl = {
.buf = buf,
.cmd = req,
},
};
return raw_thread_pool_submit(bs, handle_aiocb_ioctl, &acb);
}
#endif /* linux */
static int fd_open(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
/* this is just to ensure s->fd is sane (its called by io ops) */
if (s->fd >= 0)
return 0;
return -EIO;
}
static coroutine_fn int
hdev_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)
{
BDRVRawState *s = bs->opaque;
int ret;
ret = fd_open(bs);
if (ret < 0) {
raw_account_discard(s, bytes, ret);
return ret;
}
return raw_do_pdiscard(bs, offset, bytes, true);
}
static coroutine_fn int hdev_co_pwrite_zeroes(BlockDriverState *bs,
int64_t offset, int bytes, BdrvRequestFlags flags)
{
int rc;
rc = fd_open(bs);
if (rc < 0) {
return rc;
}
return raw_do_pwrite_zeroes(bs, offset, bytes, flags, true);
}
static BlockDriver bdrv_host_device = {
.format_name = "host_device",
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
.protocol_name = "host_device",
.instance_size = sizeof(BDRVRawState),
.bdrv_needs_filename = true,
.bdrv_probe_device = hdev_probe_device,
.bdrv_parse_filename = hdev_parse_filename,
.bdrv_file_open = hdev_open,
.bdrv_close = raw_close,
.bdrv_reopen_prepare = raw_reopen_prepare,
.bdrv_reopen_commit = raw_reopen_commit,
.bdrv_reopen_abort = raw_reopen_abort,
.bdrv_co_create_opts = bdrv_co_create_opts_simple,
.create_opts = &bdrv_create_opts_simple,
.mutable_opts = mutable_opts,
.bdrv_co_invalidate_cache = raw_co_invalidate_cache,
.bdrv_co_pwrite_zeroes = hdev_co_pwrite_zeroes,
.bdrv_co_preadv = raw_co_preadv,
.bdrv_co_pwritev = raw_co_pwritev,
.bdrv_co_flush_to_disk = raw_co_flush_to_disk,
.bdrv_co_pdiscard = hdev_co_pdiscard,
.bdrv_co_copy_range_from = raw_co_copy_range_from,
.bdrv_co_copy_range_to = raw_co_copy_range_to,
.bdrv_refresh_limits = raw_refresh_limits,
.bdrv_io_plug = raw_aio_plug,
.bdrv_io_unplug = raw_aio_unplug,
.bdrv_attach_aio_context = raw_aio_attach_aio_context,
.bdrv_co_truncate = raw_co_truncate,
.bdrv_getlength = raw_getlength,
.bdrv_get_info = raw_get_info,
.bdrv_get_allocated_file_size
= raw_get_allocated_file_size,
.bdrv_get_specific_stats = hdev_get_specific_stats,
.bdrv_check_perm = raw_check_perm,
.bdrv_set_perm = raw_set_perm,
.bdrv_abort_perm_update = raw_abort_perm_update,
.bdrv_probe_blocksizes = hdev_probe_blocksizes,
.bdrv_probe_geometry = hdev_probe_geometry,
/* generic scsi device */
#ifdef __linux__
.bdrv_co_ioctl = hdev_co_ioctl,
#endif
};
#if defined(__linux__) || defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
static void cdrom_parse_filename(const char *filename, QDict *options,
Error **errp)
{
2017-05-22 21:52:16 +02:00
bdrv_parse_filename_strip_prefix(filename, "host_cdrom:", options);
}
#endif
#ifdef __linux__
static int cdrom_open(BlockDriverState *bs, QDict *options, int flags,
Error **errp)
{
BDRVRawState *s = bs->opaque;
s->type = FTYPE_CD;
/* open will not fail even if no CD is inserted, so add O_NONBLOCK */
return raw_open_common(bs, options, flags, O_NONBLOCK, true, errp);
}
static int cdrom_probe_device(const char *filename)
{
int fd, ret;
int prio = 0;
struct stat st;
fd = qemu_open(filename, O_RDONLY | O_NONBLOCK, NULL);
if (fd < 0) {
goto out;
}
ret = fstat(fd, &st);
if (ret == -1 || !S_ISBLK(st.st_mode)) {
goto outc;
}
/* Attempt to detect via a CDROM specific ioctl */
ret = ioctl(fd, CDROM_DRIVE_STATUS, CDSL_CURRENT);
if (ret >= 0)
prio = 100;
outc:
qemu_close(fd);
out:
return prio;
}
static bool cdrom_is_inserted(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
int ret;
ret = ioctl(s->fd, CDROM_DRIVE_STATUS, CDSL_CURRENT);
return ret == CDS_DISC_OK;
}
static void cdrom_eject(BlockDriverState *bs, bool eject_flag)
{
BDRVRawState *s = bs->opaque;
if (eject_flag) {
if (ioctl(s->fd, CDROMEJECT, NULL) < 0)
perror("CDROMEJECT");
} else {
if (ioctl(s->fd, CDROMCLOSETRAY, NULL) < 0)
perror("CDROMEJECT");
}
}
static void cdrom_lock_medium(BlockDriverState *bs, bool locked)
{
BDRVRawState *s = bs->opaque;
if (ioctl(s->fd, CDROM_LOCKDOOR, locked) < 0) {
/*
* Note: an error can happen if the distribution automatically
* mounts the CD-ROM
*/
/* perror("CDROM_LOCKDOOR"); */
}
}
static BlockDriver bdrv_host_cdrom = {
.format_name = "host_cdrom",
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
.protocol_name = "host_cdrom",
.instance_size = sizeof(BDRVRawState),
.bdrv_needs_filename = true,
.bdrv_probe_device = cdrom_probe_device,
.bdrv_parse_filename = cdrom_parse_filename,
.bdrv_file_open = cdrom_open,
.bdrv_close = raw_close,
.bdrv_reopen_prepare = raw_reopen_prepare,
.bdrv_reopen_commit = raw_reopen_commit,
.bdrv_reopen_abort = raw_reopen_abort,
.bdrv_co_create_opts = bdrv_co_create_opts_simple,
.create_opts = &bdrv_create_opts_simple,
.mutable_opts = mutable_opts,
.bdrv_co_invalidate_cache = raw_co_invalidate_cache,
.bdrv_co_preadv = raw_co_preadv,
.bdrv_co_pwritev = raw_co_pwritev,
.bdrv_co_flush_to_disk = raw_co_flush_to_disk,
.bdrv_refresh_limits = raw_refresh_limits,
.bdrv_io_plug = raw_aio_plug,
.bdrv_io_unplug = raw_aio_unplug,
.bdrv_attach_aio_context = raw_aio_attach_aio_context,
.bdrv_co_truncate = raw_co_truncate,
2013-10-29 12:18:58 +01:00
.bdrv_getlength = raw_getlength,
.has_variable_length = true,
.bdrv_get_allocated_file_size
= raw_get_allocated_file_size,
/* removable device support */
.bdrv_is_inserted = cdrom_is_inserted,
.bdrv_eject = cdrom_eject,
.bdrv_lock_medium = cdrom_lock_medium,
/* generic scsi device */
.bdrv_co_ioctl = hdev_co_ioctl,
};
#endif /* __linux__ */
#if defined (__FreeBSD__) || defined(__FreeBSD_kernel__)
static int cdrom_open(BlockDriverState *bs, QDict *options, int flags,
Error **errp)
{
BDRVRawState *s = bs->opaque;
int ret;
s->type = FTYPE_CD;
error: Eliminate error_propagate() with Coccinelle, part 1 When all we do with an Error we receive into a local variable is propagating to somewhere else, we can just as well receive it there right away. Convert if (!foo(..., &err)) { ... error_propagate(errp, err); ... return ... } to if (!foo(..., errp)) { ... ... return ... } where nothing else needs @err. Coccinelle script: @rule1 forall@ identifier fun, err, errp, lbl; expression list args, args2; binary operator op; constant c1, c2; symbol false; @@ if ( ( - fun(args, &err, args2) + fun(args, errp, args2) | - !fun(args, &err, args2) + !fun(args, errp, args2) | - fun(args, &err, args2) op c1 + fun(args, errp, args2) op c1 ) ) { ... when != err when != lbl: when strict - error_propagate(errp, err); ... when != err ( return; | return c2; | return false; ) } @rule2 forall@ identifier fun, err, errp, lbl; expression list args, args2; expression var; binary operator op; constant c1, c2; symbol false; @@ - var = fun(args, &err, args2); + var = fun(args, errp, args2); ... when != err if ( ( var | !var | var op c1 ) ) { ... when != err when != lbl: when strict - error_propagate(errp, err); ... when != err ( return; | return c2; | return false; | return var; ) } @depends on rule1 || rule2@ identifier err; @@ - Error *err = NULL; ... when != err Not exactly elegant, I'm afraid. The "when != lbl:" is necessary to avoid transforming if (fun(args, &err)) { goto out } ... out: error_propagate(errp, err); even though other paths to label out still need the error_propagate(). For an actual example, see sclp_realize(). Without the "when strict", Coccinelle transforms vfio_msix_setup(), incorrectly. I don't know what exactly "when strict" does, only that it helps here. The match of return is narrower than what I want, but I can't figure out how to express "return where the operand doesn't use @err". For an example where it's too narrow, see vfio_intx_enable(). Silently fails to convert hw/arm/armsse.c, because Coccinelle gets confused by ARMSSE being used both as typedef and function-like macro there. Converted manually. Line breaks tidied up manually. One nested declaration of @local_err deleted manually. Preexisting unwanted blank line dropped in hw/riscv/sifive_e.c. Signed-off-by: Markus Armbruster <armbru@redhat.com> Reviewed-by: Eric Blake <eblake@redhat.com> Message-Id: <20200707160613.848843-35-armbru@redhat.com>
2020-07-07 18:06:02 +02:00
ret = raw_open_common(bs, options, flags, 0, true, errp);
if (ret) {
return ret;
}
/* make sure the door isn't locked at this time */
ioctl(s->fd, CDIOCALLOW);
return 0;
}
static int cdrom_probe_device(const char *filename)
{
if (strstart(filename, "/dev/cd", NULL) ||
strstart(filename, "/dev/acd", NULL))
return 100;
return 0;
}
static int cdrom_reopen(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
int fd;
/*
* Force reread of possibly changed/newly loaded disc,
* FreeBSD seems to not notice sometimes...
*/
if (s->fd >= 0)
qemu_close(s->fd);
fd = qemu_open(bs->filename, s->open_flags, NULL);
if (fd < 0) {
s->fd = -1;
return -EIO;
}
s->fd = fd;
/* make sure the door isn't locked at this time */
ioctl(s->fd, CDIOCALLOW);
return 0;
}
static bool cdrom_is_inserted(BlockDriverState *bs)
{
return raw_getlength(bs) > 0;
}
static void cdrom_eject(BlockDriverState *bs, bool eject_flag)
{
BDRVRawState *s = bs->opaque;
if (s->fd < 0)
return;
(void) ioctl(s->fd, CDIOCALLOW);
if (eject_flag) {
if (ioctl(s->fd, CDIOCEJECT) < 0)
perror("CDIOCEJECT");
} else {
if (ioctl(s->fd, CDIOCCLOSE) < 0)
perror("CDIOCCLOSE");
}
cdrom_reopen(bs);
}
static void cdrom_lock_medium(BlockDriverState *bs, bool locked)
{
BDRVRawState *s = bs->opaque;
if (s->fd < 0)
return;
if (ioctl(s->fd, (locked ? CDIOCPREVENT : CDIOCALLOW)) < 0) {
/*
* Note: an error can happen if the distribution automatically
* mounts the CD-ROM
*/
/* perror("CDROM_LOCKDOOR"); */
}
}
static BlockDriver bdrv_host_cdrom = {
.format_name = "host_cdrom",
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
.protocol_name = "host_cdrom",
.instance_size = sizeof(BDRVRawState),
.bdrv_needs_filename = true,
.bdrv_probe_device = cdrom_probe_device,
.bdrv_parse_filename = cdrom_parse_filename,
.bdrv_file_open = cdrom_open,
.bdrv_close = raw_close,
.bdrv_reopen_prepare = raw_reopen_prepare,
.bdrv_reopen_commit = raw_reopen_commit,
.bdrv_reopen_abort = raw_reopen_abort,
.bdrv_co_create_opts = bdrv_co_create_opts_simple,
.create_opts = &bdrv_create_opts_simple,
.mutable_opts = mutable_opts,
.bdrv_co_preadv = raw_co_preadv,
.bdrv_co_pwritev = raw_co_pwritev,
.bdrv_co_flush_to_disk = raw_co_flush_to_disk,
.bdrv_refresh_limits = raw_refresh_limits,
.bdrv_io_plug = raw_aio_plug,
.bdrv_io_unplug = raw_aio_unplug,
.bdrv_attach_aio_context = raw_aio_attach_aio_context,
.bdrv_co_truncate = raw_co_truncate,
2013-10-29 12:18:58 +01:00
.bdrv_getlength = raw_getlength,
.has_variable_length = true,
.bdrv_get_allocated_file_size
= raw_get_allocated_file_size,
/* removable device support */
.bdrv_is_inserted = cdrom_is_inserted,
.bdrv_eject = cdrom_eject,
.bdrv_lock_medium = cdrom_lock_medium,
};
#endif /* __FreeBSD__ */
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
static void bdrv_file_init(void)
{
/*
* Register all the drivers. Note that order is important, the driver
* registered last will get probed first.
*/
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
bdrv_register(&bdrv_file);
bdrv_register(&bdrv_host_device);
#ifdef __linux__
bdrv_register(&bdrv_host_cdrom);
#endif
#if defined(__FreeBSD__) || defined(__FreeBSD_kernel__)
bdrv_register(&bdrv_host_cdrom);
#endif
}
block: separate raw images from the file protocol We're running into various problems because the "raw" file access, which is used internally by the various image formats is entangled with the "raw" image format, which maps the VM view 1:1 to a file system. This patch renames the raw file backends to the file protocol which is treated like other protocols (e.g. nbd and http) and adds a new "raw" image format which is just a wrapper around calls to the underlying protocol. The patch is surprisingly simple, besides changing the probing logical in block.c to only look for image formats when using bdrv_open and renaming of the old raw protocols to file there's almost nothing in there. For creating images, a new bdrv_create_file is introduced which guesses the protocol to use. This allows using qemu-img create -f raw (or just using the default) for both files and host devices. Converting the other format drivers to use this function to create their images is left for later patches. The only issues still open are in the handling of the host devices. Firstly in current qemu we can specifiy the host* format names on various command line acceping images, but the new code can't do that without adding some translation. Second the layering breaks the no_zero_init flag in the BlockDriver used by qemu-img. I'm not happy how this is done per-driver instead of per-state so I'll prepare a separate patch to clean this up. There's some more cleanup opportunity after this patch, e.g. using separate lists and registration functions for image formats vs protocols and maybe even host drivers, but this can be done at a later stage. Also there's a check for protocol in bdrv_open for the BDRV_O_SNAPSHOT case that I don't quite understand, but which I fear won't work as expected - possibly even before this patch. Note that this patch requires various recent block patches from Kevin and me, which should all be in his block queue. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
2010-04-07 22:30:24 +02:00
block_init(bdrv_file_init);