blob: 44e16dda876b25241037a083e139dc72f58aff3f [file] [log] [blame] [edit]
/*
* 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 "qapi/error.h"
#include "qemu/cutils.h"
#include "qemu/error-report.h"
#include "block/block-io.h"
#include "block/block_int.h"
#include "qemu/module.h"
#include "qemu/option.h"
#include "qemu/units.h"
#include "qemu/memalign.h"
#include "trace.h"
#include "block/thread-pool.h"
#include "qemu/iov.h"
#include "block/raw-aio.h"
#include "qobject/qdict.h"
#include "qobject/qstring.h"
#include "scsi/pr-manager.h"
#include "scsi/constants.h"
#if defined(__APPLE__) && (__MACH__)
#include <sys/ioctl.h>
#if defined(HAVE_HOST_BLOCK_DEVICE)
#include <paths.h>
#include <sys/param.h>
#include <sys/mount.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 /* defined(HAVE_HOST_BLOCK_DEVICE) */
#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>
#if defined(CONFIG_BLKZONED)
#include <linux/blkzoned.h>
#endif
#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
/* 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;
uint64_t aio_max_batch;
int perm_change_fd;
int perm_change_flags;
BDRVReopenState *reopen_state;
bool has_discard:1;
bool has_write_zeroes:1;
bool use_linux_aio:1;
bool has_laio_fdsync:1;
bool use_linux_io_uring:1;
int page_cache_inconsistent; /* errno from fdatasync failure */
bool has_fallocate;
bool needs_alignment;
bool force_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;
PRManager *pr_mgr;
} BDRVRawState;
typedef struct BDRVRawReopenState {
int open_flags;
bool drop_cache;
bool check_cache_dropped;
} BDRVRawReopenState;
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 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;
struct {
unsigned int *nr_zones;
BlockZoneDescriptor *zones;
} zone_report;
struct {
unsigned long op;
} zone_mgmt;
};
} 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;
}
static bool raw_needs_alignment(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
if ((bs->open_flags & BDRV_O_NOCACHE) != 0 && !dio_byte_aligned(s->fd)) {
return true;
}
return s->force_alignment;
}
/* 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());
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 __linux__
/*
* The XFS ioctl definitions are shipped in extra packages that might
* not always be available. Since we just need the XFS_IOC_DIOINFO ioctl
* here, we simply use our own definition instead:
*/
struct xfs_dioattr {
uint32_t d_mem;
uint32_t d_miniosz;
uint32_t d_maxiosz;
} da;
if (ioctl(fd, _IOR('X', 30, struct xfs_dioattr), &da) >= 0) {
bs->bl.request_alignment = da.d_miniosz;
/* The kernel returns wrong information for d_mem */
/* s->buf_align = da.d_mem; */
}
#endif
/*
* 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;
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);
}
if (!s->buf_align) {
int i;
size_t align;
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. */
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)
{
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 = "aio-max-batch",
.type = QEMU_OPT_NUMBER,
.help = "AIO max batch size (0 = auto handled by AIO backend, default: 0)",
},
{
.name = "locking",
.type = QEMU_OPT_STRING,
.help = "file locking mode (on/off/auto, default: auto)",
},
{
.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;
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
s->aio_max_batch = qemu_opt_get_number(opts, "aio-max-batch", 0);
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();
}
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;
/* 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 && !(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 (s->use_linux_aio) {
s->has_laio_fdsync = laio_has_fdsync(s->fd);
}
#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) */
#ifndef CONFIG_LINUX_IO_URING
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 (fstat(s->fd, &st) < 0) {
ret = -errno;
error_setg_errno(errp, errno, "Could not stat file");
goto fail;
}
if (!device) {
if (!S_ISREG(st.st_mode)) {
error_setg(errp, "'%s' driver requires '%s' to be a regular file",
bs->drv->format_name, bs->filename);
ret = -EINVAL;
goto fail;
} else {
s->has_fallocate = true;
}
} else {
if (!(S_ISCHR(st.st_mode) || S_ISBLK(st.st_mode))) {
error_setg(errp, "'%s' driver requires '%s' to be either "
"a character or block device",
bs->drv->format_name, bs->filename);
ret = -EINVAL;
goto fail;
}
}
#ifdef CONFIG_BLKZONED
/*
* The kernel page cache does not reliably work for writes to SWR zones
* of zoned block device because it can not guarantee the order of writes.
*/
if ((bs->bl.zoned != BLK_Z_NONE) &&
(!(s->open_flags & O_DIRECT))) {
error_setg(errp, "The driver supports zoned devices, and it requires "
"cache.direct=on, which was not specified.");
return -EINVAL; /* No host kernel page cache */
}
#endif
if (S_ISBLK(st.st_mode)) {
#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->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->force_alignment = true;
}
#endif
s->needs_alignment = raw_needs_alignment(bs);
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;
}
/* Sets a specific flag */
static int fcntl_setfl(int fd, int flag)
{
int flags;
flags = fcntl(fd, F_GETFL);
if (flags == -1) {
return -errno;
}
if (fcntl(fd, F_SETFL, flags | flag) == -1) {
return -errno;
}
return 0;
}
static int raw_reconfigure_getfd(BlockDriverState *bs, int flags,
int *open_flags, uint64_t perm, 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 (*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;
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);
/*
* As part of reopen prepare we also want to create new fd by
* raw_reconfigure_getfd(). But it wants updated "perm", when in
* bdrv_reopen_multiple() .bdrv_reopen_prepare() callback called prior to
* permission update. Happily, permission update is always a part
* (a separate stage) of bdrv_reopen_multiple() so we can rely on this
* fact and reconfigure fd in raw_check_perm().
*/
s->reopen_state = state;
ret = 0;
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;
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;
}
g_free(state->opaque);
state->opaque = NULL;
assert(s->reopen_state == state);
s->reopen_state = NULL;
}
static int hdev_get_max_hw_transfer(int fd, struct stat *st)
{
#ifdef BLKSECTGET
if (S_ISBLK(st->st_mode)) {
unsigned short max_sectors = 0;
if (ioctl(fd, BLKSECTGET, &max_sectors) == 0) {
return max_sectors * 512;
}
} else {
int max_bytes = 0;
if (ioctl(fd, BLKSECTGET, &max_bytes) == 0) {
return max_bytes;
}
}
return -errno;
#else
return -ENOSYS;
#endif
}
/*
* Get a sysfs attribute value as character string.
*/
#ifdef CONFIG_LINUX
static int get_sysfs_str_val(struct stat *st, const char *attribute,
char **val) {
g_autofree char *sysfspath = NULL;
size_t len;
if (!S_ISBLK(st->st_mode)) {
return -ENOTSUP;
}
sysfspath = g_strdup_printf("/sys/dev/block/%u:%u/queue/%s",
major(st->st_rdev), minor(st->st_rdev),
attribute);
if (!g_file_get_contents(sysfspath, val, &len, NULL)) {
return -ENOENT;
}
/* The file is ended with '\n' */
char *p;
p = *val;
if (*(p + len - 1) == '\n') {
*(p + len - 1) = '\0';
}
return 0;
}
#endif
#if defined(CONFIG_BLKZONED)
static int get_sysfs_zoned_model(struct stat *st, BlockZoneModel *zoned)
{
g_autofree char *val = NULL;
int ret;
ret = get_sysfs_str_val(st, "zoned", &val);
if (ret < 0) {
return ret;
}
if (strcmp(val, "host-managed") == 0) {
*zoned = BLK_Z_HM;
} else if (strcmp(val, "host-aware") == 0) {
*zoned = BLK_Z_HA;
} else if (strcmp(val, "none") == 0) {
*zoned = BLK_Z_NONE;
} else {
return -ENOTSUP;
}
return 0;
}
#endif /* defined(CONFIG_BLKZONED) */
/*
* Get a sysfs attribute value as a long integer.
*/
#ifdef CONFIG_LINUX
static long get_sysfs_long_val(struct stat *st, const char *attribute)
{
g_autofree char *str = NULL;
const char *end;
long val;
int ret;
ret = get_sysfs_str_val(st, attribute, &str);
if (ret < 0) {
return ret;
}
/* The file is ended with '\n', pass 'end' to accept that. */
ret = qemu_strtol(str, &end, 10, &val);
if (ret == 0 && end && *end == '\0') {
ret = val;
}
return ret;
}
#endif
static int hdev_get_max_segments(int fd, struct stat *st)
{
#ifdef CONFIG_LINUX
int ret;
if (S_ISCHR(st->st_mode)) {
if (ioctl(fd, SG_GET_SG_TABLESIZE, &ret) == 0) {
return ret;
}
return -ENOTSUP;
}
return get_sysfs_long_val(st, "max_segments");
#else
return -ENOTSUP;
#endif
}
#if defined(CONFIG_BLKZONED)
/*
* If the reset_all flag is true, then the wps of zone whose state is
* not readonly or offline should be all reset to the start sector.
* Else, take the real wp of the device.
*/
static int get_zones_wp(BlockDriverState *bs, int fd, int64_t offset,
unsigned int nrz, bool reset_all)
{
struct blk_zone *blkz;
size_t rep_size;
uint64_t sector = offset >> BDRV_SECTOR_BITS;
BlockZoneWps *wps = bs->wps;
unsigned int j = offset / bs->bl.zone_size;
unsigned int n = 0, i = 0;
int ret;
rep_size = sizeof(struct blk_zone_report) + nrz * sizeof(struct blk_zone);
g_autofree struct blk_zone_report *rep = NULL;
rep = g_malloc(rep_size);
blkz = (struct blk_zone *)(rep + 1);
while (n < nrz) {
memset(rep, 0, rep_size);
rep->sector = sector;
rep->nr_zones = nrz - n;
do {
ret = ioctl(fd, BLKREPORTZONE, rep);
} while (ret != 0 && errno == EINTR);
if (ret != 0) {
error_report("%d: ioctl BLKREPORTZONE at %" PRId64 " failed %d",
fd, offset, errno);
return -errno;
}
if (!rep->nr_zones) {
break;
}
for (i = 0; i < rep->nr_zones; ++i, ++n, ++j) {
/*
* The wp tracking cares only about sequential writes required and
* sequential write preferred zones so that the wp can advance to
* the right location.
* Use the most significant bit of the wp location to indicate the
* zone type: 0 for SWR/SWP zones and 1 for conventional zones.
*/
if (blkz[i].type == BLK_ZONE_TYPE_CONVENTIONAL) {
wps->wp[j] |= 1ULL << 63;
} else {
switch(blkz[i].cond) {
case BLK_ZONE_COND_FULL:
case BLK_ZONE_COND_READONLY:
/* Zone not writable */
wps->wp[j] = (blkz[i].start + blkz[i].len) << BDRV_SECTOR_BITS;
break;
case BLK_ZONE_COND_OFFLINE:
/* Zone not writable nor readable */
wps->wp[j] = (blkz[i].start) << BDRV_SECTOR_BITS;
break;
default:
if (reset_all) {
wps->wp[j] = blkz[i].start << BDRV_SECTOR_BITS;
} else {
wps->wp[j] = blkz[i].wp << BDRV_SECTOR_BITS;
}
break;
}
}
}
sector = blkz[i - 1].start + blkz[i - 1].len;
}
return 0;
}
static void update_zones_wp(BlockDriverState *bs, int fd, int64_t offset,
unsigned int nrz)
{
if (get_zones_wp(bs, fd, offset, nrz, 0) < 0) {
error_report("update zone wp failed");
}
}
static void raw_refresh_zoned_limits(BlockDriverState *bs, struct stat *st,
Error **errp)
{
BDRVRawState *s = bs->opaque;
BlockZoneModel zoned = BLK_Z_NONE;
int ret;
ret = get_sysfs_zoned_model(st, &zoned);
if (ret < 0 || zoned == BLK_Z_NONE) {
goto no_zoned;
}
bs->bl.zoned = zoned;
ret = get_sysfs_long_val(st, "max_open_zones");
if (ret >= 0) {
bs->bl.max_open_zones = ret;
}
ret = get_sysfs_long_val(st, "max_active_zones");
if (ret >= 0) {
bs->bl.max_active_zones = ret;
}
/*
* The zoned device must at least have zone size and nr_zones fields.
*/
ret = get_sysfs_long_val(st, "chunk_sectors");
if (ret < 0) {
error_setg_errno(errp, -ret, "Unable to read chunk_sectors "
"sysfs attribute");
goto no_zoned;
} else if (!ret) {
error_setg(errp, "Read 0 from chunk_sectors sysfs attribute");
goto no_zoned;
}
bs->bl.zone_size = ret << BDRV_SECTOR_BITS;
ret = get_sysfs_long_val(st, "nr_zones");
if (ret < 0) {
error_setg_errno(errp, -ret, "Unable to read nr_zones "
"sysfs attribute");
goto no_zoned;
} else if (!ret) {
error_setg(errp, "Read 0 from nr_zones sysfs attribute");
goto no_zoned;
}
bs->bl.nr_zones = ret;
ret = get_sysfs_long_val(st, "zone_append_max_bytes");
if (ret > 0) {
bs->bl.max_append_sectors = ret >> BDRV_SECTOR_BITS;
}
ret = get_sysfs_long_val(st, "physical_block_size");
if (ret >= 0) {
bs->bl.write_granularity = ret;
}
/* The refresh_limits() function can be called multiple times. */
g_free(bs->wps);
bs->wps = g_malloc(sizeof(BlockZoneWps) +
sizeof(int64_t) * bs->bl.nr_zones);
ret = get_zones_wp(bs, s->fd, 0, bs->bl.nr_zones, 0);
if (ret < 0) {
error_setg_errno(errp, -ret, "report wps failed");
goto no_zoned;
}
qemu_co_mutex_init(&bs->wps->colock);
return;
no_zoned:
bs->bl.zoned = BLK_Z_NONE;
g_free(bs->wps);
bs->wps = NULL;
}
#else /* !defined(CONFIG_BLKZONED) */
static void raw_refresh_zoned_limits(BlockDriverState *bs, struct stat *st,
Error **errp)
{
bs->bl.zoned = BLK_Z_NONE;
}
#endif /* !defined(CONFIG_BLKZONED) */
static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
{
BDRVRawState *s = bs->opaque;
struct stat st;
s->needs_alignment = raw_needs_alignment(bs);
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());
/*
* Maximum transfers are best effort, so it is okay to ignore any
* errors. That said, based on the man page errors in fstat would be
* very much unexpected; the only possible case seems to be ENOMEM.
*/
if (fstat(s->fd, &st)) {
return;
}
#if defined(__APPLE__) && (__MACH__)
struct statfs buf;
if (!fstatfs(s->fd, &buf)) {
bs->bl.opt_transfer = buf.f_iosize;
bs->bl.pdiscard_alignment = buf.f_bsize;
}
#endif
if (bdrv_is_sg(bs) || S_ISBLK(st.st_mode)) {
int ret = hdev_get_max_hw_transfer(s->fd, &st);
if (ret > 0 && ret <= BDRV_REQUEST_MAX_BYTES) {
bs->bl.max_hw_transfer = ret;
}
ret = hdev_get_max_segments(s->fd, &st);
if (ret > 0) {
bs->bl.max_hw_iov = ret;
}
}
raw_refresh_zoned_limits(bs, &st, errp);
}
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 or zoned devices, get blocksizes */
if (check_for_dasd(s->fd) < 0) {
/* zoned devices are not DASD */
if (bs->bl.zoned == BLK_Z_NONE) {
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 = RETRY_ON_EINTR(
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 -s->page_cache_inconsistent;
}
ret = qemu_fdatasync(aiocb->aio_fildes);
if (ret == -1) {
trace_file_flush_fdatasync_failed(errno);
/* 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 = errno;
}
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;
len = RETRY_ON_EINTR(
(aiocb->aio_type & (QEMU_AIO_WRITE | QEMU_AIO_ZONE_APPEND)) ?
qemu_pwritev(aiocb->aio_fildes,
aiocb->io.iov,
aiocb->io.niov,
aiocb->aio_offset) :
qemu_preadv(aiocb->aio_fildes,
aiocb->io.iov,
aiocb->io.niov,
aiocb->aio_offset)
);
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 | QEMU_AIO_ZONE_APPEND)) {
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 | QEMU_AIO_ZONE_APPEND))) {
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;
}
}
#if defined(CONFIG_FALLOCATE) || defined(BLKZEROOUT) || defined(BLKDISCARD)
static int translate_err(int err)
{
if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP ||
err == -ENOTTY) {
err = -ENOTSUP;
}
return err;
}
#endif
#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 == -ENOTSUP) {
s->has_write_zeroes = false;
} else if (ret == 0 || ret != -EINVAL) {
return ret;
}
/*
* Note: Some file systems do not like unaligned byte ranges, and
* return EINVAL in such a case, though they should not do it according
* to the man-page of fallocate(). Thus we simply ignore this return
* value and try the other fallbacks instead.
*/
}
#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 == -EINVAL) {
/*
* Some file systems like older versions of GPFS do not like un-
* aligned byte ranges, and return EINVAL in such a case, though
* they should not do it according to the man-page of fallocate().
* Warn about the bad filesystem and try the final fallback instead.
*/
warn_report_once("Your file system is misbehaving: "
"fallocate(FALLOC_FL_PUNCH_HOLE) returned EINVAL. "
"Please report this bug to your file system "
"vendor.");
} 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 = raw_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
/*
* parse_zone - Fill a zone descriptor
*/
#if defined(CONFIG_BLKZONED)
static inline int parse_zone(struct BlockZoneDescriptor *zone,
const struct blk_zone *blkz) {
zone->start = blkz->start << BDRV_SECTOR_BITS;
zone->length = blkz->len << BDRV_SECTOR_BITS;
zone->wp = blkz->wp << BDRV_SECTOR_BITS;
#ifdef HAVE_BLK_ZONE_REP_CAPACITY
zone->cap = blkz->capacity << BDRV_SECTOR_BITS;
#else
zone->cap = blkz->len << BDRV_SECTOR_BITS;
#endif
switch (blkz->type) {
case BLK_ZONE_TYPE_SEQWRITE_REQ:
zone->type = BLK_ZT_SWR;
break;
case BLK_ZONE_TYPE_SEQWRITE_PREF:
zone->type = BLK_ZT_SWP;
break;
case BLK_ZONE_TYPE_CONVENTIONAL:
zone->type = BLK_ZT_CONV;
break;
default:
error_report("Unsupported zone type: 0x%x", blkz->type);
return -ENOTSUP;
}
switch (blkz->cond) {
case BLK_ZONE_COND_NOT_WP:
zone->state = BLK_ZS_NOT_WP;
break;
case BLK_ZONE_COND_EMPTY:
zone->state = BLK_ZS_EMPTY;
break;
case BLK_ZONE_COND_IMP_OPEN:
zone->state = BLK_ZS_IOPEN;
break;
case BLK_ZONE_COND_EXP_OPEN:
zone->state = BLK_ZS_EOPEN;
break;
case BLK_ZONE_COND_CLOSED:
zone->state = BLK_ZS_CLOSED;
break;
case BLK_ZONE_COND_READONLY:
zone->state = BLK_ZS_RDONLY;
break;
case BLK_ZONE_COND_FULL:
zone->state = BLK_ZS_FULL;
break;
case BLK_ZONE_COND_OFFLINE:
zone->state = BLK_ZS_OFFLINE;
break;
default:
error_report("Unsupported zone state: 0x%x", blkz->cond);
return -ENOTSUP;
}
return 0;
}
#endif
#if defined(CONFIG_BLKZONED)
static int handle_aiocb_zone_report(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
int fd = aiocb->aio_fildes;
unsigned int *nr_zones = aiocb->zone_report.nr_zones;
BlockZoneDescriptor *zones = aiocb->zone_report.zones;
/* zoned block devices use 512-byte sectors */
uint64_t sector = aiocb->aio_offset / 512;
struct blk_zone *blkz;
size_t rep_size;
unsigned int nrz;
int ret;
unsigned int n = 0, i = 0;
nrz = *nr_zones;
rep_size = sizeof(struct blk_zone_report) + nrz * sizeof(struct blk_zone);
g_autofree struct blk_zone_report *rep = NULL;
rep = g_malloc(rep_size);
blkz = (struct blk_zone *)(rep + 1);
while (n < nrz) {
memset(rep, 0, rep_size);
rep->sector = sector;
rep->nr_zones = nrz - n;
do {
ret = ioctl(fd, BLKREPORTZONE, rep);
} while (ret != 0 && errno == EINTR);
if (ret != 0) {
error_report("%d: ioctl BLKREPORTZONE at %" PRId64 " failed %d",
fd, sector, errno);
return -errno;
}
if (!rep->nr_zones) {
break;
}
for (i = 0; i < rep->nr_zones; i++, n++) {
ret = parse_zone(&zones[n], &blkz[i]);
if (ret != 0) {
return ret;
}
/* The next report should start after the last zone reported */
sector = blkz[i].start + blkz[i].len;
}
}
*nr_zones = n;
return 0;
}
#endif
#if defined(CONFIG_BLKZONED)
static int handle_aiocb_zone_mgmt(void *opaque)
{
RawPosixAIOData *aiocb = opaque;
int fd = aiocb->aio_fildes;
uint64_t sector = aiocb->aio_offset / 512;
int64_t nr_sectors = aiocb->aio_nbytes / 512;
struct blk_zone_range range;
int ret;
/* Execute the operation */
range.sector = sector;
range.nr_sectors = nr_sectors;
do {
ret = ioctl(fd, aiocb->zone_mgmt.op, &range);
} while (ret != 0 && errno == EINTR);
return ret < 0 ? -errno : ret;
}
#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 = -ENOTSUP;
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 = translate_err(-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);
ret = translate_err(ret);
#elif defined(__APPLE__) && (__MACH__)
fpunchhole_t fpunchhole;
fpunchhole.fp_flags = 0;
fpunchhole.reserved = 0;
fpunchhole.fp_offset = aiocb->aio_offset;
fpunchhole.fp_length = aiocb->aio_nbytes;
if (fcntl(s->fd, F_PUNCHHOLE, &fpunchhole) == -1) {
ret = errno == ENODEV ? -ENOTSUP : -errno;
} else {
ret = 0;
}
#endif
}
if (ret == -ENOTSUP) {
s->has_discard = false;
}
return ret;
}
/*
* 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());
void *buf;
ssize_t n;
int ret;
buf = qemu_memalign(max_align, write_size);
memset(buf, 0, write_size);
n = RETRY_ON_EINTR(pwrite(fd, buf, write_size, 0));
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");
} 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");
} 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(ThreadPoolFunc func, void *arg)
{
return thread_pool_submit_co(func, arg);
}
/*
* Check if all memory in this vector is sector aligned.
*/
static bool bdrv_qiov_is_aligned(BlockDriverState *bs, QEMUIOVector *qiov)
{
int i;
size_t alignment = bdrv_min_mem_align(bs);
size_t len = bs->bl.request_alignment;
IO_CODE();
for (i = 0; i < qiov->niov; i++) {
if ((uintptr_t) qiov->iov[i].iov_base % alignment) {
return false;
}
if (qiov->iov[i].iov_len % len) {
return false;
}
}
return true;
}
#ifdef CONFIG_LINUX_IO_URING
static inline bool raw_check_linux_io_uring(BDRVRawState *s)
{
Error *local_err = NULL;
AioContext *ctx;
if (!s->use_linux_io_uring) {
return false;
}
ctx = qemu_get_current_aio_context();
if (unlikely(!aio_setup_linux_io_uring(ctx, &local_err))) {
error_reportf_err(local_err, "Unable to use linux io_uring, "
"falling back to thread pool: ");
s->use_linux_io_uring = false;
return false;
}
return true;
}
#endif
#ifdef CONFIG_LINUX_AIO
static inline bool raw_check_linux_aio(BDRVRawState *s)
{
Error *local_err = NULL;
AioContext *ctx;
if (!s->use_linux_aio) {
return false;
}
ctx = qemu_get_current_aio_context();
if (unlikely(!aio_setup_linux_aio(ctx, &local_err))) {
error_reportf_err(local_err, "Unable to use Linux AIO, "
"falling back to thread pool: ");
s->use_linux_aio = false;
return false;
}
return true;
}
#endif
static int coroutine_fn raw_co_prw(BlockDriverState *bs, int64_t *offset_ptr,
uint64_t bytes, QEMUIOVector *qiov, int type)
{
BDRVRawState *s = bs->opaque;
RawPosixAIOData acb;
int ret;
uint64_t offset = *offset_ptr;
if (fd_open(bs) < 0)
return -EIO;
#if defined(CONFIG_BLKZONED)
if ((type & (QEMU_AIO_WRITE | QEMU_AIO_ZONE_APPEND)) &&
bs->bl.zoned != BLK_Z_NONE) {
qemu_co_mutex_lock(&bs->wps->colock);
if (type & QEMU_AIO_ZONE_APPEND) {
int index = offset / bs->bl.zone_size;
offset = bs->wps->wp[index];
}
}
#endif
/*
* 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 (raw_check_linux_io_uring(s)) {
assert(qiov->size == bytes);
ret = luring_co_submit(bs, s->fd, offset, qiov, type);
goto out;
#endif
#ifdef CONFIG_LINUX_AIO
} else if (raw_check_linux_aio(s)) {
assert(qiov->size == bytes);
ret = laio_co_submit(s->fd, offset, qiov, type,
s->aio_max_batch);
goto out;
#endif
}
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);
ret = raw_thread_pool_submit(handle_aiocb_rw, &acb);
goto out; /* Avoid the compiler err of unused label */
out:
#if defined(CONFIG_BLKZONED)
if ((type & (QEMU_AIO_WRITE | QEMU_AIO_ZONE_APPEND)) &&
bs->bl.zoned != BLK_Z_NONE) {
BlockZoneWps *wps = bs->wps;
if (ret == 0) {
uint64_t *wp = &wps->wp[offset / bs->bl.zone_size];
if (!BDRV_ZT_IS_CONV(*wp)) {
if (type & QEMU_AIO_ZONE_APPEND) {
*offset_ptr = *wp;
trace_zbd_zone_append_complete(bs, *offset_ptr
>> BDRV_SECTOR_BITS);
}
/* Advance the wp if needed */
if (offset + bytes > *wp) {
*wp = offset + bytes;
}
}
} else {
/*
* write and append write are not allowed to cross zone boundaries
*/
update_zones_wp(bs, s->fd, offset, 1);
}
qemu_co_mutex_unlock(&wps->colock);
}
#endif
return ret;
}
static int coroutine_fn raw_co_preadv(BlockDriverState *bs, int64_t offset,
int64_t bytes, QEMUIOVector *qiov,
BdrvRequestFlags flags)
{
return raw_co_prw(bs, &offset, bytes, qiov, QEMU_AIO_READ);
}
static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, int64_t offset,
int64_t bytes, QEMUIOVector *qiov,
BdrvRequestFlags flags)
{
return raw_co_prw(bs, &offset, bytes, qiov, QEMU_AIO_WRITE);
}
static int coroutine_fn 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 (raw_check_linux_io_uring(s)) {
return luring_co_submit(bs, s->fd, 0, NULL, QEMU_AIO_FLUSH);
}
#endif
#ifdef CONFIG_LINUX_AIO
if (s->has_laio_fdsync && raw_check_linux_aio(s)) {
return laio_co_submit(s->fd, 0, NULL, QEMU_AIO_FLUSH, 0);
}
#endif
return raw_thread_pool_submit(handle_aiocb_flush, &acb);
}
static void raw_close(BlockDriverState *bs)
{
BDRVRawState *s = bs->opaque;
if (s->fd >= 0) {
#if defined(CONFIG_BLKZONED)
g_free(bs->wps);
#endif
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(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)) {
size = 0;
#ifdef DIOCGMEDIASIZE
if (ioctl(fd, DIOCGMEDIASIZE, (off_t *)&size)) {
size = 0;
}
#endif
#ifdef DIOCGPART
if (size == 0) {
struct partinfo pi;
if (ioctl(fd, DIOCGPART, &pi) == 0) {
size = pi.media_size;
}
}
#endif
#if defined(DKIOCGETBLOCKCOUNT) && defined(DKIOCGETBLOCKSIZE)
if (size == 0) {
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;
}
}
#endif
if (size == 0) {
size = lseek(fd, 0LL, SEEK_END);
}
if (size < 0) {
return -errno;
}
#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 coroutine_fn raw_co_getlength(BlockDriverState *bs)
{
return raw_getlength(bs);
}
static int64_t coroutine_fn raw_co_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;
}
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
}
#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 GRAPH_RDLOCK
raw_co_create_opts(BlockDriver *drv, const char *filename,
QemuOpts *opts, Error **errp)
{
BlockdevCreateOptions options;
int64_t total_size = 0;
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