/*
* Free space in a file.
*
* IN: zp - znode of file to free data in.
* off - start of section to free.
* len - length of section to free.
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_free_range(znode_t *zp, uint64_t off, uint64_t len)
{
zfs_sb_t *zsb = ZTOZSB(zp);
rl_t *rl;
int error;
/*
* Lock the range being freed.
*/
rl = zfs_range_lock(zp, off, len, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (off >= zp->z_size) {
zfs_range_unlock(rl);
return (0);
}
if (off + len > zp->z_size)
len = zp->z_size - off;
error = dmu_free_long_range(zsb->z_os, zp->z_id, off, len);
zfs_range_unlock(rl);
return (error);
}
static int
zvol_read(zvol_state_t *zv, uio_t *uio)
{
uint64_t volsize = zv->zv_volsize;
rl_t *rl;
int error = 0;
ASSERT(zv && zv->zv_open_count > 0);
rl = zfs_range_lock(&zv->zv_range_lock, uio->uio_loffset,
uio->uio_resid, RL_READER);
while (uio->uio_resid > 0 && uio->uio_loffset < volsize) {
uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1);
/* don't read past the end */
if (bytes > volsize - uio->uio_loffset)
bytes = volsize - uio->uio_loffset;
error = dmu_read_uio_dbuf(zv->zv_dbuf, uio, bytes);
if (error) {
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
break;
}
}
zfs_range_unlock(rl);
return (error);
}
static void
zvol_discard(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error;
rl_t *rl;
if (offset + size > zv->zv_volsize) {
blk_end_request(req, -EIO, size);
return;
}
if (size == 0) {
blk_end_request(req, 0, size);
return;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset, size);
/*
* TODO: maybe we should add the operation to the log.
*/
zfs_range_unlock(rl);
blk_end_request(req, -error, size);
}
/*
* Common read path running under the zvol taskq context. This function
* is responsible for copying the requested data out of the DMU and in to
* a linux request structure. It then must signal the request queue with
* an error code describing the result of the copy.
*/
static void
zvol_read(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error;
rl_t *rl;
if (size == 0) {
blk_end_request(req, 0, size);
return;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER);
error = dmu_read_req(zv->zv_objset, ZVOL_OBJ, req);
zfs_range_unlock(rl);
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = EIO;
blk_end_request(req, -error, size);
}
/*
* Common read path running under the zvol taskq context. This function
* is responsible for copying the requested data out of the DMU and in to
* a linux request structure. It then must signal the request queue with
* an error code describing the result of the copy.
*/
static void
zvol_read(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
fstrans_cookie_t cookie = spl_fstrans_mark();
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error;
rl_t *rl;
if (size == 0) {
error = 0;
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER);
error = dmu_read_req(zv->zv_objset, ZVOL_OBJ, req);
zfs_range_unlock(rl);
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
out:
blk_end_request(req, -error, size);
spl_fstrans_unmark(cookie);
}
static int
zvol_read(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t offset = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
int error;
rl_t *rl;
uio_t uio;
if (size == 0)
return (0);
uio.uio_bvec = &bio->bi_io_vec[BIO_BI_IDX(bio)];
uio.uio_skip = BIO_BI_SKIP(bio);
uio.uio_resid = size;
uio.uio_iovcnt = bio->bi_vcnt - BIO_BI_IDX(bio);
uio.uio_loffset = offset;
uio.uio_limit = MAXOFFSET_T;
uio.uio_segflg = UIO_BVEC;
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER);
error = dmu_read_uio(zv->zv_objset, ZVOL_OBJ, &uio, size);
zfs_range_unlock(rl);
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
return (error);
}
static int
zvol_read(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t offset = BIO_BI_SECTOR(bio) << 9;
uint64_t len = BIO_BI_SIZE(bio);
int error;
rl_t *rl;
if (len == 0)
return (0);
rl = zfs_range_lock(&zv->zv_znode, offset, len, RL_READER);
error = dmu_read_bio(zv->zv_objset, ZVOL_OBJ, bio);
zfs_range_unlock(rl);
/* convert checksum errors into IO errors */
if (error == ECKSUM)
error = SET_ERROR(EIO);
return (error);
}
static int
zvol_write(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t offset = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
int error = 0;
dmu_tx_t *tx;
rl_t *rl;
uio_t uio;
if (bio->bi_rw & VDEV_REQ_FLUSH)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
/*
* Some requests are just for flush and nothing else.
*/
if (size == 0)
goto out;
uio.uio_bvec = &bio->bi_io_vec[BIO_BI_IDX(bio)];
uio.uio_skip = BIO_BI_SKIP(bio);
uio.uio_resid = size;
uio.uio_iovcnt = bio->bi_vcnt - BIO_BI_IDX(bio);
uio.uio_loffset = offset;
uio.uio_limit = MAXOFFSET_T;
uio.uio_segflg = UIO_BVEC;
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
goto out;
}
error = dmu_write_uio(zv->zv_objset, ZVOL_OBJ, &uio, size, tx);
if (error == 0)
zvol_log_write(zv, tx, offset, size,
!!(bio->bi_rw & VDEV_REQ_FUA));
dmu_tx_commit(tx);
zfs_range_unlock(rl);
if ((bio->bi_rw & VDEV_REQ_FUA) ||
zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
out:
return (error);
}
/*
* Get data to generate a TX_WRITE intent log record.
*/
static int
zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio)
{
zvol_state_t *zv = arg;
objset_t *os = zv->zv_objset;
uint64_t offset = lr->lr_offset;
uint64_t size = lr->lr_length;
dmu_buf_t *db;
zgd_t *zgd;
int error;
ASSERT(zio != NULL);
ASSERT(size != 0);
zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_PUSHPAGE);
zgd->zgd_zilog = zv->zv_zilog;
zgd->zgd_rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER);
/*
* Write records come in two flavors: immediate and indirect.
* For small writes it's cheaper to store the data with the
* log record (immediate); for large writes it's cheaper to
* sync the data and get a pointer to it (indirect) so that
* we don't have to write the data twice.
*/
if (buf != NULL) { /* immediate write */
error = dmu_read(os, ZVOL_OBJ, offset, size, buf,
DMU_READ_NO_PREFETCH);
} else {
size = zv->zv_volblocksize;
offset = P2ALIGN_TYPED(offset, size, uint64_t);
error = dmu_buf_hold(os, ZVOL_OBJ, offset, zgd, &db,
DMU_READ_NO_PREFETCH);
if (error == 0) {
zgd->zgd_db = db;
zgd->zgd_bp = &lr->lr_blkptr;
ASSERT(db != NULL);
ASSERT(db->db_offset == offset);
ASSERT(db->db_size == size);
error = dmu_sync(zio, lr->lr_common.lrc_txg,
zvol_get_done, zgd);
if (error == 0)
return (0);
}
}
zvol_get_done(zgd, error);
return (error);
}
/*
* Common write path running under the zvol taskq context. This function
* is responsible for copying the request structure data in to the DMU and
* signaling the request queue with the result of the copy.
*/
static void
zvol_write(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
fstrans_cookie_t cookie = spl_fstrans_mark();
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error = 0;
dmu_tx_t *tx;
rl_t *rl;
if (req->cmd_flags & VDEV_REQ_FLUSH)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
/*
* Some requests are just for flush and nothing else.
*/
if (size == 0) {
error = 0;
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
goto out;
}
error = dmu_write_req(zv->zv_objset, ZVOL_OBJ, req, tx);
if (error == 0)
zvol_log_write(zv, tx, offset, size,
req->cmd_flags & VDEV_REQ_FUA);
dmu_tx_commit(tx);
zfs_range_unlock(rl);
if ((req->cmd_flags & VDEV_REQ_FUA) ||
zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
out:
blk_end_request(req, -error, size);
spl_fstrans_unmark(cookie);
}
static void
zvol_discard(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
uint64_t start = blk_rq_pos(req) << 9;
uint64_t end = start + blk_rq_bytes(req);
int error;
rl_t *rl;
/*
* Annotate this call path with a flag that indicates that it is
* unsafe to use KM_SLEEP during memory allocations due to the
* potential for a deadlock. KM_PUSHPAGE should be used instead.
*/
ASSERT(!(current->flags & PF_NOFS));
current->flags |= PF_NOFS;
if (end > zv->zv_volsize) {
blk_end_request(req, -EIO, blk_rq_bytes(req));
goto out;
}
/*
* Align the request to volume block boundaries. If we don't,
* then this will force dnode_free_range() to zero out the
* unaligned parts, which is slow (read-modify-write) and
* useless since we are not freeing any space by doing so.
*/
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
if (start >= end) {
blk_end_request(req, 0, blk_rq_bytes(req));
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, start, end - start, RL_WRITER);
error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, start, end - start);
/*
* TODO: maybe we should add the operation to the log.
*/
zfs_range_unlock(rl);
blk_end_request(req, -error, blk_rq_bytes(req));
out:
current->flags &= ~PF_NOFS;
}
static int
zvol_discard(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t start = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
uint64_t end = start + size;
int error;
rl_t *rl;
dmu_tx_t *tx;
ASSERT(zv && zv->zv_open_count > 0);
if (end > zv->zv_volsize)
return (SET_ERROR(EIO));
/*
* Align the request to volume block boundaries when REQ_SECURE is
* available, but not requested. If we don't, then this will force
* dnode_free_range() to zero out the unaligned parts, which is slow
* (read-modify-write) and useless since we are not freeing any space
* by doing so. Kernels that do not support REQ_SECURE (2.6.32 through
* 2.6.35) will not receive this optimization.
*/
#ifdef REQ_SECURE
if (!(bio->bi_rw & REQ_SECURE)) {
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
size = end - start;
}
#endif
if (start >= end)
return (0);
rl = zfs_range_lock(&zv->zv_znode, start, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0) {
dmu_tx_abort(tx);
} else {
zvol_log_truncate(zv, tx, start, size, B_TRUE);
dmu_tx_commit(tx);
error = dmu_free_long_range(zv->zv_objset,
ZVOL_OBJ, start, size);
}
zfs_range_unlock(rl);
return (error);
}
static int
zvol_discard(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t start = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
uint64_t end = start + size;
int error;
rl_t *rl;
dmu_tx_t *tx;
ASSERT(zv && zv->zv_open_count > 0);
if (end > zv->zv_volsize)
return (SET_ERROR(EIO));
/*
* Align the request to volume block boundaries when a secure erase is
* not required. This will prevent dnode_free_range() from zeroing out
* the unaligned parts which is slow (read-modify-write) and useless
* since we are not freeing any space by doing so.
*/
if (!bio_is_secure_erase(bio)) {
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
size = end - start;
}
if (start >= end)
return (0);
rl = zfs_range_lock(&zv->zv_range_lock, start, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error != 0) {
dmu_tx_abort(tx);
} else {
zvol_log_truncate(zv, tx, start, size, B_TRUE);
dmu_tx_commit(tx);
error = dmu_free_long_range(zv->zv_objset,
ZVOL_OBJ, start, size);
}
zfs_range_unlock(rl);
return (error);
}
static void
zvol_discard(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
fstrans_cookie_t cookie = spl_fstrans_mark();
uint64_t start = blk_rq_pos(req) << 9;
uint64_t end = start + blk_rq_bytes(req);
int error;
rl_t *rl;
if (end > zv->zv_volsize) {
error = EIO;
goto out;
}
/*
* Align the request to volume block boundaries. If we don't,
* then this will force dnode_free_range() to zero out the
* unaligned parts, which is slow (read-modify-write) and
* useless since we are not freeing any space by doing so.
*/
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
if (start >= end) {
error = 0;
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, start, end - start, RL_WRITER);
error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, start, end-start);
/*
* TODO: maybe we should add the operation to the log.
*/
zfs_range_unlock(rl);
out:
blk_end_request(req, -error, blk_rq_bytes(req));
spl_fstrans_unmark(cookie);
}
static int
zvol_discard(struct bio *bio)
{
zvol_state_t *zv = bio->bi_bdev->bd_disk->private_data;
uint64_t start = BIO_BI_SECTOR(bio) << 9;
uint64_t size = BIO_BI_SIZE(bio);
uint64_t end = start + size;
int error;
rl_t *rl;
if (end > zv->zv_volsize)
return (SET_ERROR(EIO));
/*
* Align the request to volume block boundaries when REQ_SECURE is
* available, but not requested. If we don't, then this will force
* dnode_free_range() to zero out the unaligned parts, which is slow
* (read-modify-write) and useless since we are not freeing any space
* by doing so. Kernels that do not support REQ_SECURE (2.6.32 through
* 2.6.35) will not receive this optimization.
*/
#ifdef REQ_SECURE
if (!(bio->bi_rw & REQ_SECURE)) {
start = P2ROUNDUP(start, zv->zv_volblocksize);
end = P2ALIGN(end, zv->zv_volblocksize);
size = end - start;
}
#endif
if (start >= end)
return (0);
rl = zfs_range_lock(&zv->zv_znode, start, size, RL_WRITER);
error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, start, size);
/*
* TODO: maybe we should add the operation to the log.
*/
zfs_range_unlock(rl);
return (error);
}
static void
zvol_discard(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error;
rl_t *rl;
/*
* Annotate this call path with a flag that indicates that it is
* unsafe to use KM_SLEEP during memory allocations due to the
* potential for a deadlock. KM_PUSHPAGE should be used instead.
*/
ASSERT(!(current->flags & PF_NOFS));
current->flags |= PF_NOFS;
if (offset + size > zv->zv_volsize) {
blk_end_request(req, -EIO, size);
goto out;
}
if (size == 0) {
blk_end_request(req, 0, size);
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset, size);
/*
* TODO: maybe we should add the operation to the log.
*/
zfs_range_unlock(rl);
blk_end_request(req, -error, size);
out:
current->flags &= ~PF_NOFS;
}
static int
zvol_write(zvol_state_t *zv, uio_t *uio, boolean_t sync)
{
uint64_t volsize = zv->zv_volsize;
rl_t *rl;
int error = 0;
ASSERT(zv && zv->zv_open_count > 0);
rl = zfs_range_lock(&zv->zv_range_lock, uio->uio_loffset,
uio->uio_resid, RL_WRITER);
while (uio->uio_resid > 0 && uio->uio_loffset < volsize) {
uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1);
uint64_t off = uio->uio_loffset;
dmu_tx_t *tx = dmu_tx_create(zv->zv_objset);
if (bytes > volsize - off) /* don't write past the end */
bytes = volsize - off;
dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
break;
}
error = dmu_write_uio_dbuf(zv->zv_dbuf, uio, bytes, tx);
if (error == 0)
zvol_log_write(zv, tx, off, bytes, sync);
dmu_tx_commit(tx);
if (error)
break;
}
zfs_range_unlock(rl);
if (sync)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
return (error);
}
/*
* Common write path running under the zvol taskq context. This function
* is responsible for copying the request structure data in to the DMU and
* signaling the request queue with the result of the copy.
*/
static void
zvol_write(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error = 0;
dmu_tx_t *tx;
rl_t *rl;
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
blk_end_request(req, -error, size);
return;
}
error = dmu_write_req(zv->zv_objset, ZVOL_OBJ, req, tx);
if (error == 0)
zvol_log_write(zv, tx, offset, size, rq_is_sync(req));
dmu_tx_commit(tx);
zfs_range_unlock(rl);
if (rq_is_sync(req))
zil_commit(zv->zv_zilog, ZVOL_OBJ);
blk_end_request(req, -error, size);
}
/*
* Truncate a file
*
* IN: zp - znode of file to free data in.
* end - new end-of-file.
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_trunc(znode_t *zp, uint64_t end)
{
zfs_sb_t *zsb = ZTOZSB(zp);
dmu_tx_t *tx;
rl_t *rl;
int error;
sa_bulk_attr_t bulk[2];
int count = 0;
/*
* We will change zp_size, lock the whole file.
*/
rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (end >= zp->z_size) {
zfs_range_unlock(rl);
return (0);
}
error = dmu_free_long_range(zsb->z_os, zp->z_id, end, -1);
if (error) {
zfs_range_unlock(rl);
return (error);
}
top:
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
error = dmu_tx_assign(tx, TXG_NOWAIT);
if (error) {
if (error == ERESTART) {
dmu_tx_wait(tx);
dmu_tx_abort(tx);
goto top;
}
dmu_tx_abort(tx);
zfs_range_unlock(rl);
return (error);
}
zp->z_size = end;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zsb),
NULL, &zp->z_size, sizeof (zp->z_size));
if (end == 0) {
zp->z_pflags &= ~ZFS_SPARSE;
SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zsb),
NULL, &zp->z_pflags, 8);
}
VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0);
dmu_tx_commit(tx);
zfs_range_unlock(rl);
return (0);
}
/*
* Common write path running under the zvol taskq context. This function
* is responsible for copying the request structure data in to the DMU and
* signaling the request queue with the result of the copy.
*/
static void
zvol_write(void *arg)
{
struct request *req = (struct request *)arg;
struct request_queue *q = req->q;
zvol_state_t *zv = q->queuedata;
uint64_t offset = blk_rq_pos(req) << 9;
uint64_t size = blk_rq_bytes(req);
int error = 0;
dmu_tx_t *tx;
rl_t *rl;
/*
* Annotate this call path with a flag that indicates that it is
* unsafe to use KM_SLEEP during memory allocations due to the
* potential for a deadlock. KM_PUSHPAGE should be used instead.
*/
ASSERT(!(current->flags & PF_NOFS));
current->flags |= PF_NOFS;
if (req->cmd_flags & VDEV_REQ_FLUSH)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
/*
* Some requests are just for flush and nothing else.
*/
if (size == 0) {
blk_end_request(req, 0, size);
goto out;
}
rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_write(tx, ZVOL_OBJ, offset, size);
/* This will only fail for ENOSPC */
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
blk_end_request(req, -error, size);
goto out;
}
error = dmu_write_req(zv->zv_objset, ZVOL_OBJ, req, tx);
if (error == 0)
zvol_log_write(zv, tx, offset, size,
req->cmd_flags & VDEV_REQ_FUA);
dmu_tx_commit(tx);
zfs_range_unlock(rl);
if ((req->cmd_flags & VDEV_REQ_FUA) ||
zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)
zil_commit(zv->zv_zilog, ZVOL_OBJ);
blk_end_request(req, -error, size);
out:
current->flags &= ~PF_NOFS;
}
/*
* Increase the file length
*
* IN: zp - znode of file to free data in.
* end - new end-of-file
*
* RETURN: 0 on success, error code on failure
*/
static int
zfs_extend(znode_t *zp, uint64_t end)
{
zfs_sb_t *zsb = ZTOZSB(zp);
dmu_tx_t *tx;
rl_t *rl;
uint64_t newblksz;
int error;
/*
* We will change zp_size, lock the whole file.
*/
rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER);
/*
* Nothing to do if file already at desired length.
*/
if (end <= zp->z_size) {
zfs_range_unlock(rl);
return (0);
}
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE);
zfs_sa_upgrade_txholds(tx, zp);
if (end > zp->z_blksz &&
(!ISP2(zp->z_blksz) || zp->z_blksz < zsb->z_max_blksz)) {
/*
* We are growing the file past the current block size.
*/
if (zp->z_blksz > ZTOZSB(zp)->z_max_blksz) {
ASSERT(!ISP2(zp->z_blksz));
newblksz = MIN(end, SPA_MAXBLOCKSIZE);
} else {
newblksz = MIN(end, ZTOZSB(zp)->z_max_blksz);
}
dmu_tx_hold_write(tx, zp->z_id, 0, newblksz);
} else {
newblksz = 0;
}
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
zfs_range_unlock(rl);
return (error);
}
if (newblksz)
zfs_grow_blocksize(zp, newblksz, tx);
zp->z_size = end;
VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(ZTOZSB(zp)),
&zp->z_size, sizeof (zp->z_size), tx));
zfs_range_unlock(rl);
dmu_tx_commit(tx);
return (0);
}