freebsd-dev/module/zfs/zvol.c

1983 lines
46 KiB
C
Raw Normal View History

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright (C) 2008-2010 Lawrence Livermore National Security, LLC.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Rewritten for Linux by Brian Behlendorf <behlendorf1@llnl.gov>.
* LLNL-CODE-403049.
*
* ZFS volume emulation driver.
*
* Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes.
* Volumes are accessed through the symbolic links named:
*
* /dev/<pool_name>/<dataset_name>
*
* Volumes are persistent through reboot and module load. No user command
* needs to be run before opening and using a device.
*
* Copyright 2014 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2016 Actifio, Inc. All rights reserved.
*/
#include <sys/dbuf.h>
#include <sys/dmu_traverse.h>
#include <sys/dsl_dataset.h>
#include <sys/dsl_prop.h>
#include <sys/dsl_dir.h>
#include <sys/zap.h>
#include <sys/zfeature.h>
#include <sys/zil_impl.h>
#include <sys/dmu_tx.h>
#include <sys/zio.h>
#include <sys/zfs_rlock.h>
#include <sys/zfs_znode.h>
#include <sys/spa_impl.h>
#include <sys/zvol.h>
#include <linux/blkdev_compat.h>
unsigned int zvol_inhibit_dev = 0;
unsigned int zvol_major = ZVOL_MAJOR;
unsigned int zvol_prefetch_bytes = (128 * 1024);
unsigned long zvol_max_discard_blocks = 16384;
static kmutex_t zvol_state_lock;
static list_t zvol_state_list;
OpenZFS 2605, 6980, 6902 2605 want to resume interrupted zfs send Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed by: Paul Dagnelie <pcd@delphix.com> Reviewed by: Richard Elling <Richard.Elling@RichardElling.com> Reviewed by: Xin Li <delphij@freebsd.org> Reviewed by: Arne Jansen <sensille@gmx.net> Approved by: Dan McDonald <danmcd@omniti.com> Ported-by: kernelOfTruth <kerneloftruth@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> OpenZFS-issue: https://www.illumos.org/issues/2605 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/9c3fd12 6980 6902 causes zfs send to break due to 32-bit/64-bit struct mismatch Reviewed by: Paul Dagnelie <pcd@delphix.com> Reviewed by: George Wilson <george.wilson@delphix.com> Approved by: Robert Mustacchi <rm@joyent.com> Ported by: Brian Behlendorf <behlendorf1@llnl.gov> OpenZFS-issue: https://www.illumos.org/issues/6980 OpenZFS-commit: https://github.com/openzfs/openzfs/commit/ea4a67f Porting notes: - All rsend and snapshop tests enabled and updated for Linux. - Fix misuse of input argument in traverse_visitbp(). - Fix ISO C90 warnings and errors. - Fix gcc 'missing braces around initializer' in 'struct send_thread_arg to_arg =' warning. - Replace 4 argument fletcher_4_native() with 3 argument version, this change was made in OpenZFS 4185 which has not been ported. - Part of the sections for 'zfs receive' and 'zfs send' was rewritten and reordered to approximate upstream. - Fix mktree xattr creation, 'user.' prefix required. - Minor fixes to newly enabled test cases - Long holds for volumes allowed during receive for minor registration.
2016-01-06 21:22:48 +00:00
void *zvol_tag = "zvol_tag";
/*
* The in-core state of each volume.
*/
typedef struct zvol_state {
char zv_name[MAXNAMELEN]; /* name */
uint64_t zv_volsize; /* advertised space */
uint64_t zv_volblocksize; /* volume block size */
objset_t *zv_objset; /* objset handle */
uint32_t zv_flags; /* ZVOL_* flags */
uint32_t zv_open_count; /* open counts */
uint32_t zv_changed; /* disk changed */
zilog_t *zv_zilog; /* ZIL handle */
zfs_rlock_t zv_range_lock; /* range lock */
dmu_buf_t *zv_dbuf; /* bonus handle */
dev_t zv_dev; /* device id */
struct gendisk *zv_disk; /* generic disk */
struct request_queue *zv_queue; /* request queue */
list_node_t zv_next; /* next zvol_state_t linkage */
} zvol_state_t;
typedef enum {
ZVOL_ASYNC_CREATE_MINORS,
ZVOL_ASYNC_REMOVE_MINORS,
ZVOL_ASYNC_RENAME_MINORS,
ZVOL_ASYNC_SET_SNAPDEV,
ZVOL_ASYNC_MAX
} zvol_async_op_t;
typedef struct {
zvol_async_op_t op;
char pool[MAXNAMELEN];
char name1[MAXNAMELEN];
char name2[MAXNAMELEN];
zprop_source_t source;
uint64_t snapdev;
} zvol_task_t;
#define ZVOL_RDONLY 0x1
/*
* Find the next available range of ZVOL_MINORS minor numbers. The
* zvol_state_list is kept in ascending minor order so we simply need
* to scan the list for the first gap in the sequence. This allows us
* to recycle minor number as devices are created and removed.
*/
static int
zvol_find_minor(unsigned *minor)
{
zvol_state_t *zv;
*minor = 0;
ASSERT(MUTEX_HELD(&zvol_state_lock));
for (zv = list_head(&zvol_state_list); zv != NULL;
zv = list_next(&zvol_state_list, zv), *minor += ZVOL_MINORS) {
if (MINOR(zv->zv_dev) != MINOR(*minor))
break;
}
/* All minors are in use */
if (*minor >= (1 << MINORBITS))
return (SET_ERROR(ENXIO));
return (0);
}
/*
* Find a zvol_state_t given the full major+minor dev_t.
*/
static zvol_state_t *
zvol_find_by_dev(dev_t dev)
{
zvol_state_t *zv;
ASSERT(MUTEX_HELD(&zvol_state_lock));
for (zv = list_head(&zvol_state_list); zv != NULL;
zv = list_next(&zvol_state_list, zv)) {
if (zv->zv_dev == dev)
return (zv);
}
return (NULL);
}
/*
* Find a zvol_state_t given the name provided at zvol_alloc() time.
*/
static zvol_state_t *
zvol_find_by_name(const char *name)
{
zvol_state_t *zv;
ASSERT(MUTEX_HELD(&zvol_state_lock));
for (zv = list_head(&zvol_state_list); zv != NULL;
zv = list_next(&zvol_state_list, zv)) {
if (strncmp(zv->zv_name, name, MAXNAMELEN) == 0)
return (zv);
}
return (NULL);
}
/*
* Given a path, return TRUE if path is a ZVOL.
*/
boolean_t
zvol_is_zvol(const char *device)
{
struct block_device *bdev;
unsigned int major;
bdev = lookup_bdev(device);
if (IS_ERR(bdev))
return (B_FALSE);
major = MAJOR(bdev->bd_dev);
bdput(bdev);
if (major == zvol_major)
return (B_TRUE);
return (B_FALSE);
}
/*
* ZFS_IOC_CREATE callback handles dmu zvol and zap object creation.
*/
void
zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
{
zfs_creat_t *zct = arg;
nvlist_t *nvprops = zct->zct_props;
int error;
uint64_t volblocksize, volsize;
VERIFY(nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0);
if (nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0)
volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
/*
* These properties must be removed from the list so the generic
* property setting step won't apply to them.
*/
VERIFY(nvlist_remove_all(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0);
(void) nvlist_remove_all(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE));
error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx);
ASSERT(error == 0);
}
/*
* ZFS_IOC_OBJSET_STATS entry point.
*/
int
zvol_get_stats(objset_t *os, nvlist_t *nv)
{
int error;
dmu_object_info_t *doi;
uint64_t val;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val);
if (error)
return (SET_ERROR(error));
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val);
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error == 0) {
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE,
doi->doi_data_block_size);
}
kmem_free(doi, sizeof (dmu_object_info_t));
return (SET_ERROR(error));
}
static void
zvol_size_changed(zvol_state_t *zv, uint64_t volsize)
{
struct block_device *bdev;
bdev = bdget_disk(zv->zv_disk, 0);
if (bdev == NULL)
return;
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
check_disk_size_change(zv->zv_disk, bdev);
bdput(bdev);
}
/*
* Sanity check volume size.
*/
int
zvol_check_volsize(uint64_t volsize, uint64_t blocksize)
{
if (volsize == 0)
return (SET_ERROR(EINVAL));
if (volsize % blocksize != 0)
return (SET_ERROR(EINVAL));
#ifdef _ILP32
if (volsize - 1 > MAXOFFSET_T)
return (SET_ERROR(EOVERFLOW));
#endif
return (0);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(uint64_t volsize, objset_t *os)
{
dmu_tx_t *tx;
int error;
uint64_t txg;
ASSERT(MUTEX_HELD(&zvol_state_lock));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (SET_ERROR(error));
}
txg = dmu_tx_get_txg(tx);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
txg_wait_synced(dmu_objset_pool(os), txg);
if (error == 0)
error = dmu_free_long_range(os,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
return (error);
}
static int
zvol_update_live_volsize(zvol_state_t *zv, uint64_t volsize)
{
zvol_size_changed(zv, volsize);
/*
* We should post a event here describing the expansion. However,
* the zfs_ereport_post() interface doesn't nicely support posting
* events for zvols, it assumes events relate to vdevs or zios.
*/
return (0);
}
/*
* Set ZFS_PROP_VOLSIZE set entry point.
*/
int
zvol_set_volsize(const char *name, uint64_t volsize)
{
zvol_state_t *zv = NULL;
objset_t *os = NULL;
int error;
dmu_object_info_t *doi;
uint64_t readonly;
boolean_t owned = B_FALSE;
error = dsl_prop_get_integer(name,
zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL);
if (error != 0)
return (SET_ERROR(error));
if (readonly)
return (SET_ERROR(EROFS));
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv == NULL || zv->zv_objset == NULL) {
if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE,
FTAG, &os)) != 0) {
mutex_exit(&zvol_state_lock);
return (SET_ERROR(error));
}
owned = B_TRUE;
if (zv != NULL)
zv->zv_objset = os;
} else {
os = zv->zv_objset;
}
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
if ((error = dmu_object_info(os, ZVOL_OBJ, doi)) ||
(error = zvol_check_volsize(volsize, doi->doi_data_block_size)))
goto out;
error = zvol_update_volsize(volsize, os);
kmem_free(doi, sizeof (dmu_object_info_t));
if (error == 0 && zv != NULL)
error = zvol_update_live_volsize(zv, volsize);
out:
if (owned) {
dmu_objset_disown(os, FTAG);
if (zv != NULL)
zv->zv_objset = NULL;
}
mutex_exit(&zvol_state_lock);
return (error);
}
/*
* Sanity check volume block size.
*/
int
zvol_check_volblocksize(const char *name, uint64_t volblocksize)
{
/* Record sizes above 128k need the feature to be enabled */
if (volblocksize > SPA_OLD_MAXBLOCKSIZE) {
spa_t *spa;
int error;
if ((error = spa_open(name, &spa, FTAG)) != 0)
return (error);
if (!spa_feature_is_enabled(spa, SPA_FEATURE_LARGE_BLOCKS)) {
spa_close(spa, FTAG);
return (SET_ERROR(ENOTSUP));
}
/*
* We don't allow setting the property above 1MB,
* unless the tunable has been changed.
*/
if (volblocksize > zfs_max_recordsize)
return (SET_ERROR(EDOM));
spa_close(spa, FTAG);
}
if (volblocksize < SPA_MINBLOCKSIZE ||
volblocksize > SPA_MAXBLOCKSIZE ||
!ISP2(volblocksize))
return (SET_ERROR(EDOM));
return (0);
}
/*
* Set ZFS_PROP_VOLBLOCKSIZE set entry point.
*/
int
zvol_set_volblocksize(const char *name, uint64_t volblocksize)
{
zvol_state_t *zv;
dmu_tx_t *tx;
int error;
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv == NULL) {
error = SET_ERROR(ENXIO);
goto out;
}
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
if (zv->zv_flags & ZVOL_RDONLY) {
error = SET_ERROR(EROFS);
goto out;
}
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_bonus(tx, ZVOL_OBJ);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
error = dmu_object_set_blocksize(zv->zv_objset, ZVOL_OBJ,
volblocksize, 0, tx);
if (error == ENOTSUP)
error = SET_ERROR(EBUSY);
dmu_tx_commit(tx);
if (error == 0)
zv->zv_volblocksize = volblocksize;
}
out:
mutex_exit(&zvol_state_lock);
return (SET_ERROR(error));
}
/*
* Replay a TX_TRUNCATE ZIL transaction if asked. TX_TRUNCATE is how we
* implement DKIOCFREE/free-long-range.
*/
static int
zvol_replay_truncate(zvol_state_t *zv, lr_truncate_t *lr, boolean_t byteswap)
{
uint64_t offset, length;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
offset = lr->lr_offset;
length = lr->lr_length;
return (dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset, length));
}
/*
* Replay a TX_WRITE ZIL transaction that didn't get committed
* after a system failure
*/
static int
zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap)
{
objset_t *os = zv->zv_objset;
char *data = (char *)(lr + 1); /* data follows lr_write_t */
uint64_t off = lr->lr_offset;
uint64_t len = lr->lr_length;
dmu_tx_t *tx;
int error;
if (byteswap)
byteswap_uint64_array(lr, sizeof (*lr));
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, ZVOL_OBJ, off, len);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
} else {
dmu_write(os, ZVOL_OBJ, off, len, data, tx);
dmu_tx_commit(tx);
}
return (SET_ERROR(error));
}
static int
zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap)
{
return (SET_ERROR(ENOTSUP));
}
/*
* Callback vectors for replaying records.
* Only TX_WRITE and TX_TRUNCATE are needed for zvol.
*/
zil_replay_func_t zvol_replay_vector[TX_MAX_TYPE] = {
(zil_replay_func_t)zvol_replay_err, /* no such transaction type */
(zil_replay_func_t)zvol_replay_err, /* TX_CREATE */
(zil_replay_func_t)zvol_replay_err, /* TX_MKDIR */
(zil_replay_func_t)zvol_replay_err, /* TX_MKXATTR */
(zil_replay_func_t)zvol_replay_err, /* TX_SYMLINK */
(zil_replay_func_t)zvol_replay_err, /* TX_REMOVE */
(zil_replay_func_t)zvol_replay_err, /* TX_RMDIR */
(zil_replay_func_t)zvol_replay_err, /* TX_LINK */
(zil_replay_func_t)zvol_replay_err, /* TX_RENAME */
(zil_replay_func_t)zvol_replay_write, /* TX_WRITE */
(zil_replay_func_t)zvol_replay_truncate, /* TX_TRUNCATE */
(zil_replay_func_t)zvol_replay_err, /* TX_SETATTR */
(zil_replay_func_t)zvol_replay_err, /* TX_ACL */
};
/*
* zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions.
*
* We store data in the log buffers if it's small enough.
* Otherwise we will later flush the data out via dmu_sync().
*/
ssize_t zvol_immediate_write_sz = 32768;
static void
zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, uint64_t offset,
uint64_t size, int sync)
{
uint32_t blocksize = zv->zv_volblocksize;
zilog_t *zilog = zv->zv_zilog;
boolean_t slogging;
ssize_t immediate_write_sz;
if (zil_replaying(zilog, tx))
return;
immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT)
? 0 : zvol_immediate_write_sz;
slogging = spa_has_slogs(zilog->zl_spa) &&
(zilog->zl_logbias == ZFS_LOGBIAS_LATENCY);
while (size) {
itx_t *itx;
lr_write_t *lr;
ssize_t len;
itx_wr_state_t write_state;
/*
* Unlike zfs_log_write() we can be called with
* up to DMU_MAX_ACCESS/2 (5MB) writes.
*/
if (blocksize > immediate_write_sz && !slogging &&
size >= blocksize && offset % blocksize == 0) {
write_state = WR_INDIRECT; /* uses dmu_sync */
len = blocksize;
} else if (sync) {
write_state = WR_COPIED;
len = MIN(ZIL_MAX_LOG_DATA, size);
} else {
write_state = WR_NEED_COPY;
len = MIN(ZIL_MAX_LOG_DATA, size);
}
itx = zil_itx_create(TX_WRITE, sizeof (*lr) +
(write_state == WR_COPIED ? len : 0));
lr = (lr_write_t *)&itx->itx_lr;
if (write_state == WR_COPIED && dmu_read(zv->zv_objset,
ZVOL_OBJ, offset, len, lr+1, DMU_READ_NO_PREFETCH) != 0) {
zil_itx_destroy(itx);
itx = zil_itx_create(TX_WRITE, sizeof (*lr));
lr = (lr_write_t *)&itx->itx_lr;
write_state = WR_NEED_COPY;
}
itx->itx_wr_state = write_state;
if (write_state == WR_NEED_COPY)
itx->itx_sod += len;
lr->lr_foid = ZVOL_OBJ;
lr->lr_offset = offset;
lr->lr_length = len;
lr->lr_blkoff = 0;
BP_ZERO(&lr->lr_blkptr);
itx->itx_private = zv;
itx->itx_sync = sync;
(void) zil_itx_assign(zilog, itx, tx);
offset += len;
size -= len;
}
}
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
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);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
return (error);
}
/*
* Log a DKIOCFREE/free-long-range to the ZIL with TX_TRUNCATE.
*/
static void
zvol_log_truncate(zvol_state_t *zv, dmu_tx_t *tx, uint64_t off, uint64_t len,
boolean_t sync)
{
itx_t *itx;
lr_truncate_t *lr;
zilog_t *zilog = zv->zv_zilog;
if (zil_replaying(zilog, tx))
return;
itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr));
lr = (lr_truncate_t *)&itx->itx_lr;
lr->lr_foid = ZVOL_OBJ;
lr->lr_offset = off;
lr->lr_length = len;
itx->itx_sync = sync;
zil_itx_assign(zilog, itx, tx);
}
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
static int
zvol_discard(struct bio *bio)
{
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
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);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
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;
}
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
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);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
return (error);
}
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
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);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
return (error);
}
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
static MAKE_REQUEST_FN_RET
zvol_request(struct request_queue *q, struct bio *bio)
{
uio_t uio;
zvol_state_t *zv = q->queuedata;
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
fstrans_cookie_t cookie = spl_fstrans_mark();
int rw = bio_data_dir(bio);
#ifdef HAVE_GENERIC_IO_ACCT
unsigned long start = jiffies;
#endif
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
int error = 0;
uio.uio_bvec = &bio->bi_io_vec[BIO_BI_IDX(bio)];
uio.uio_skip = BIO_BI_SKIP(bio);
uio.uio_resid = BIO_BI_SIZE(bio);
uio.uio_iovcnt = bio->bi_vcnt - BIO_BI_IDX(bio);
uio.uio_loffset = BIO_BI_SECTOR(bio) << 9;
uio.uio_limit = MAXOFFSET_T;
uio.uio_segflg = UIO_BVEC;
if (bio_has_data(bio) && uio.uio_loffset + uio.uio_resid >
zv->zv_volsize) {
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
printk(KERN_INFO
"%s: bad access: offset=%llu, size=%lu\n",
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
zv->zv_disk->disk_name,
(long long unsigned)uio.uio_loffset,
(long unsigned)uio.uio_resid);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
error = SET_ERROR(EIO);
goto out1;
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
}
generic_start_io_acct(rw, bio_sectors(bio), &zv->zv_disk->part0);
if (rw == WRITE) {
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
if (unlikely(zv->zv_flags & ZVOL_RDONLY)) {
error = SET_ERROR(EROFS);
goto out2;
}
if (bio_is_discard(bio) || bio_is_secure_erase(bio)) {
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
error = zvol_discard(bio);
goto out2;
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
}
/*
* Some requests are just for flush and nothing else.
*/
if (uio.uio_resid == 0) {
if (bio_is_flush(bio))
zil_commit(zv->zv_zilog, ZVOL_OBJ);
goto out2;
}
error = zvol_write(zv, &uio,
bio_is_flush(bio) || bio_is_fua(bio) ||
zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
} else
error = zvol_read(zv, &uio);
out2:
generic_end_io_acct(rw, &zv->zv_disk->part0, start);
out1:
BIO_END_IO(bio, -error);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
spl_fstrans_unmark(cookie);
#ifdef HAVE_MAKE_REQUEST_FN_RET_INT
return (0);
#elif defined(HAVE_MAKE_REQUEST_FN_RET_QC)
return (BLK_QC_T_NONE);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
#endif
}
static void
zvol_get_done(zgd_t *zgd, int error)
{
if (zgd->zgd_db)
dmu_buf_rele(zgd->zgd_db, zgd);
zfs_range_unlock(zgd->zgd_rl);
if (error == 0 && zgd->zgd_bp)
zil_add_block(zgd->zgd_zilog, zgd->zgd_bp);
kmem_free(zgd, sizeof (zgd_t));
}
/*
* 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 object = ZVOL_OBJ;
uint64_t offset = lr->lr_offset;
uint64_t size = lr->lr_length;
blkptr_t *bp = &lr->lr_blkptr;
dmu_buf_t *db;
zgd_t *zgd;
int error;
ASSERT(zio != NULL);
ASSERT(size != 0);
zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP);
zgd->zgd_zilog = zv->zv_zilog;
zgd->zgd_rl = zfs_range_lock(&zv->zv_range_lock, 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, object, offset, size, buf,
DMU_READ_NO_PREFETCH);
} else {
size = zv->zv_volblocksize;
offset = P2ALIGN_TYPED(offset, size, uint64_t);
error = dmu_buf_hold(os, object, offset, zgd, &db,
DMU_READ_NO_PREFETCH);
if (error == 0) {
blkptr_t *obp = dmu_buf_get_blkptr(db);
if (obp) {
ASSERT(BP_IS_HOLE(bp));
*bp = *obp;
}
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 (SET_ERROR(error));
}
/*
* The zvol_state_t's are inserted in increasing MINOR(dev_t) order.
*/
static void
zvol_insert(zvol_state_t *zv_insert)
{
zvol_state_t *zv = NULL;
ASSERT(MUTEX_HELD(&zvol_state_lock));
ASSERT3U(MINOR(zv_insert->zv_dev) & ZVOL_MINOR_MASK, ==, 0);
for (zv = list_head(&zvol_state_list); zv != NULL;
zv = list_next(&zvol_state_list, zv)) {
if (MINOR(zv->zv_dev) > MINOR(zv_insert->zv_dev))
break;
}
list_insert_before(&zvol_state_list, zv, zv_insert);
}
/*
* Simply remove the zvol from to list of zvols.
*/
static void
zvol_remove(zvol_state_t *zv_remove)
{
ASSERT(MUTEX_HELD(&zvol_state_lock));
list_remove(&zvol_state_list, zv_remove);
}
static int
zvol_first_open(zvol_state_t *zv)
{
objset_t *os;
uint64_t volsize;
int error;
uint64_t ro;
/* lie and say we're read-only */
error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
if (error)
return (SET_ERROR(-error));
zv->zv_objset = os;
error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL);
if (error)
goto out_owned;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_owned;
error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
if (error)
goto out_owned;
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
zv->zv_zilog = zil_open(os, zvol_get_data);
if (ro || dmu_objset_is_snapshot(os) ||
!spa_writeable(dmu_objset_spa(os))) {
set_disk_ro(zv->zv_disk, 1);
zv->zv_flags |= ZVOL_RDONLY;
} else {
set_disk_ro(zv->zv_disk, 0);
zv->zv_flags &= ~ZVOL_RDONLY;
}
out_owned:
if (error) {
dmu_objset_disown(os, zvol_tag);
zv->zv_objset = NULL;
}
return (SET_ERROR(-error));
}
static void
zvol_last_close(zvol_state_t *zv)
{
zil_close(zv->zv_zilog);
zv->zv_zilog = NULL;
dmu_buf_rele(zv->zv_dbuf, zvol_tag);
zv->zv_dbuf = NULL;
/*
* Evict cached data
*/
if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) &&
!(zv->zv_flags & ZVOL_RDONLY))
txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
(void) dmu_objset_evict_dbufs(zv->zv_objset);
dmu_objset_disown(zv->zv_objset, zvol_tag);
zv->zv_objset = NULL;
}
static int
zvol_open(struct block_device *bdev, fmode_t flag)
{
zvol_state_t *zv;
int error = 0, drop_mutex = 0;
/*
* If the caller is already holding the mutex do not take it
* again, this will happen as part of zvol_create_minor_impl().
* Once add_disk() is called the device is live and the kernel
* will attempt to open it to read the partition information.
*/
if (!mutex_owned(&zvol_state_lock)) {
mutex_enter(&zvol_state_lock);
drop_mutex = 1;
}
/*
* Obtain a copy of private_data under the lock to make sure
* that either the result of zvol_freeg() setting
* bdev->bd_disk->private_data to NULL is observed, or zvol_free()
* is not called on this zv because of the positive zv_open_count.
*/
zv = bdev->bd_disk->private_data;
if (zv == NULL) {
error = -ENXIO;
goto out_mutex;
}
if (zv->zv_open_count == 0) {
error = zvol_first_open(zv);
if (error)
goto out_mutex;
}
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
if ((flag & FMODE_WRITE) && (zv->zv_flags & ZVOL_RDONLY)) {
error = -EROFS;
goto out_open_count;
}
zv->zv_open_count++;
check_disk_change(bdev);
out_open_count:
if (zv->zv_open_count == 0)
zvol_last_close(zv);
out_mutex:
if (drop_mutex)
mutex_exit(&zvol_state_lock);
return (SET_ERROR(error));
}
#ifdef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID
static void
#else
static int
#endif
zvol_release(struct gendisk *disk, fmode_t mode)
{
zvol_state_t *zv = disk->private_data;
int drop_mutex = 0;
ASSERT(zv && zv->zv_open_count > 0);
if (!mutex_owned(&zvol_state_lock)) {
mutex_enter(&zvol_state_lock);
drop_mutex = 1;
}
zv->zv_open_count--;
if (zv->zv_open_count == 0)
zvol_last_close(zv);
if (drop_mutex)
mutex_exit(&zvol_state_lock);
#ifndef HAVE_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID
return (0);
#endif
}
static int
zvol_ioctl(struct block_device *bdev, fmode_t mode,
unsigned int cmd, unsigned long arg)
{
zvol_state_t *zv = bdev->bd_disk->private_data;
int error = 0;
ASSERT(zv && zv->zv_open_count > 0);
switch (cmd) {
case BLKFLSBUF:
zil_commit(zv->zv_zilog, ZVOL_OBJ);
break;
case BLKZNAME:
error = copy_to_user((void *)arg, zv->zv_name, MAXNAMELEN);
break;
default:
error = -ENOTTY;
break;
}
return (SET_ERROR(error));
}
#ifdef CONFIG_COMPAT
static int
zvol_compat_ioctl(struct block_device *bdev, fmode_t mode,
unsigned cmd, unsigned long arg)
{
return (zvol_ioctl(bdev, mode, cmd, arg));
}
#else
#define zvol_compat_ioctl NULL
#endif
static int zvol_media_changed(struct gendisk *disk)
{
zvol_state_t *zv = disk->private_data;
ASSERT(zv && zv->zv_open_count > 0);
return (zv->zv_changed);
}
static int zvol_revalidate_disk(struct gendisk *disk)
{
zvol_state_t *zv = disk->private_data;
ASSERT(zv && zv->zv_open_count > 0);
zv->zv_changed = 0;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
return (0);
}
/*
* Provide a simple virtual geometry for legacy compatibility. For devices
* smaller than 1 MiB a small head and sector count is used to allow very
* tiny devices. For devices over 1 Mib a standard head and sector count
* is used to keep the cylinders count reasonable.
*/
static int
zvol_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
zvol_state_t *zv = bdev->bd_disk->private_data;
sector_t sectors;
ASSERT(zv && zv->zv_open_count > 0);
sectors = get_capacity(zv->zv_disk);
if (sectors > 2048) {
geo->heads = 16;
geo->sectors = 63;
} else {
geo->heads = 2;
geo->sectors = 4;
}
geo->start = 0;
geo->cylinders = sectors / (geo->heads * geo->sectors);
return (0);
}
static struct kobject *
zvol_probe(dev_t dev, int *part, void *arg)
{
zvol_state_t *zv;
struct kobject *kobj;
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_dev(dev);
kobj = zv ? get_disk(zv->zv_disk) : NULL;
mutex_exit(&zvol_state_lock);
return (kobj);
}
#ifdef HAVE_BDEV_BLOCK_DEVICE_OPERATIONS
static struct block_device_operations zvol_ops = {
.open = zvol_open,
.release = zvol_release,
.ioctl = zvol_ioctl,
.compat_ioctl = zvol_compat_ioctl,
.media_changed = zvol_media_changed,
.revalidate_disk = zvol_revalidate_disk,
.getgeo = zvol_getgeo,
.owner = THIS_MODULE,
};
#else /* HAVE_BDEV_BLOCK_DEVICE_OPERATIONS */
static int
zvol_open_by_inode(struct inode *inode, struct file *file)
{
return (zvol_open(inode->i_bdev, file->f_mode));
}
static int
zvol_release_by_inode(struct inode *inode, struct file *file)
{
return (zvol_release(inode->i_bdev->bd_disk, file->f_mode));
}
static int
zvol_ioctl_by_inode(struct inode *inode, struct file *file,
unsigned int cmd, unsigned long arg)
{
if (file == NULL || inode == NULL)
return (SET_ERROR(-EINVAL));
return (zvol_ioctl(inode->i_bdev, file->f_mode, cmd, arg));
}
#ifdef CONFIG_COMPAT
static long
zvol_compat_ioctl_by_inode(struct file *file,
unsigned int cmd, unsigned long arg)
{
if (file == NULL)
return (SET_ERROR(-EINVAL));
return (zvol_compat_ioctl(file->f_dentry->d_inode->i_bdev,
file->f_mode, cmd, arg));
}
#else
#define zvol_compat_ioctl_by_inode NULL
#endif
static struct block_device_operations zvol_ops = {
.open = zvol_open_by_inode,
.release = zvol_release_by_inode,
.ioctl = zvol_ioctl_by_inode,
.compat_ioctl = zvol_compat_ioctl_by_inode,
.media_changed = zvol_media_changed,
.revalidate_disk = zvol_revalidate_disk,
.getgeo = zvol_getgeo,
.owner = THIS_MODULE,
};
#endif /* HAVE_BDEV_BLOCK_DEVICE_OPERATIONS */
/*
* Allocate memory for a new zvol_state_t and setup the required
* request queue and generic disk structures for the block device.
*/
static zvol_state_t *
zvol_alloc(dev_t dev, const char *name)
{
zvol_state_t *zv;
zv = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP);
Cleanup zvol initialization code The following error will occur on some (possibly all) kernels because blk_init_queue() will try to take the spinlock before we initialize it. BUG: spinlock bad magic on CPU#0, zpool/4054 lock: 0xffff88021a73de60, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0 Pid: 4054, comm: zpool Not tainted 3.9.3 #11 Call Trace: [<ffffffff81478ef8>] spin_dump+0x8c/0x91 [<ffffffff81478f1e>] spin_bug+0x21/0x26 [<ffffffff812da097>] do_raw_spin_lock+0x127/0x130 [<ffffffff8147d851>] _raw_spin_lock_irq+0x21/0x30 [<ffffffff812c2c1e>] cfq_init_queue+0x1fe/0x350 [<ffffffff812aacb8>] elevator_init+0x78/0x140 [<ffffffff812b2677>] blk_init_allocated_queue+0x87/0xb0 [<ffffffff812b26d5>] blk_init_queue_node+0x35/0x70 [<ffffffff812b271e>] blk_init_queue+0xe/0x10 [<ffffffff8125211b>] __zvol_create_minor+0x24b/0x620 [<ffffffff81253264>] zvol_create_minors_cb+0x24/0x30 [<ffffffff811bd9ca>] dmu_objset_find_spa+0xea/0x510 [<ffffffff811bda71>] dmu_objset_find_spa+0x191/0x510 [<ffffffff81253ea2>] zvol_create_minors+0x92/0x180 [<ffffffff811f8d80>] spa_open_common+0x250/0x380 [<ffffffff811f8ece>] spa_open+0xe/0x10 [<ffffffff8122817e>] pool_status_check.part.22+0x1e/0x80 [<ffffffff81228a55>] zfsdev_ioctl+0x155/0x190 [<ffffffff8116a695>] do_vfs_ioctl+0x325/0x5a0 [<ffffffff8116a950>] sys_ioctl+0x40/0x80 [<ffffffff814812c9>] ? do_page_fault+0x9/0x10 [<ffffffff81483929>] system_call_fastpath+0x16/0x1b zd0: unknown partition table We fix this by calling spin_lock_init before blk_init_queue. The manner in which zvol_init() initializes structures is suspectible to a race between initialization and a probe on a zvol. We reorganize zvol_init() to prevent that. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2013-07-02 18:59:10 +00:00
list_link_init(&zv->zv_next);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
zv->zv_queue = blk_alloc_queue(GFP_ATOMIC);
if (zv->zv_queue == NULL)
goto out_kmem;
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
blk_queue_make_request(zv->zv_queue, zvol_request);
blk_queue_set_write_cache(zv->zv_queue, B_TRUE, B_TRUE);
zv->zv_disk = alloc_disk(ZVOL_MINORS);
if (zv->zv_disk == NULL)
goto out_queue;
zv->zv_queue->queuedata = zv;
zv->zv_dev = dev;
zv->zv_open_count = 0;
strlcpy(zv->zv_name, name, MAXNAMELEN);
zfs_rlock_init(&zv->zv_range_lock);
zv->zv_disk->major = zvol_major;
zv->zv_disk->first_minor = (dev & MINORMASK);
zv->zv_disk->fops = &zvol_ops;
zv->zv_disk->private_data = zv;
zv->zv_disk->queue = zv->zv_queue;
snprintf(zv->zv_disk->disk_name, DISK_NAME_LEN, "%s%d",
ZVOL_DEV_NAME, (dev & MINORMASK));
return (zv);
out_queue:
blk_cleanup_queue(zv->zv_queue);
out_kmem:
kmem_free(zv, sizeof (zvol_state_t));
return (NULL);
}
/*
* Cleanup then free a zvol_state_t which was created by zvol_alloc().
*/
static void
zvol_free(zvol_state_t *zv)
{
ASSERT(MUTEX_HELD(&zvol_state_lock));
ASSERT(zv->zv_open_count == 0);
zfs_rlock_destroy(&zv->zv_range_lock);
zv->zv_disk->private_data = NULL;
del_gendisk(zv->zv_disk);
blk_cleanup_queue(zv->zv_queue);
put_disk(zv->zv_disk);
kmem_free(zv, sizeof (zvol_state_t));
}
/*
* Create a block device minor node and setup the linkage between it
* and the specified volume. Once this function returns the block
* device is live and ready for use.
*/
static int
zvol_create_minor_impl(const char *name)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
uint64_t len;
unsigned minor = 0;
int error = 0;
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv) {
error = SET_ERROR(EEXIST);
goto out;
}
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
if (error)
goto out_doi;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
error = zvol_find_minor(&minor);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = SET_ERROR(EAGAIN);
goto out_dmu_objset_disown;
}
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
blk_queue_max_hw_sectors(zv->zv_queue, (DMU_MAX_ACCESS / 4) >> 9);
blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
blk_queue_max_discard_sectors(zv->zv_queue,
(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
#ifdef QUEUE_FLAG_NONROT
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif
#ifdef QUEUE_FLAG_ADD_RANDOM
queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zv->zv_queue);
#endif
if (spa_writeable(dmu_objset_spa(os))) {
if (zil_replay_disable)
zil_destroy(dmu_objset_zil(os), B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
}
/*
* When udev detects the addition of the device it will immediately
* invoke blkid(8) to determine the type of content on the device.
* Prefetching the blocks commonly scanned by blkid(8) will speed
* up this process.
*/
len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE);
if (len > 0) {
dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
ZIO_PRIORITY_SYNC_READ);
}
zv->zv_objset = NULL;
out_dmu_objset_disown:
dmu_objset_disown(os, zvol_tag);
out_doi:
kmem_free(doi, sizeof (dmu_object_info_t));
out:
if (error == 0) {
zvol_insert(zv);
/*
* Drop the lock to prevent deadlock with sys_open() ->
* zvol_open(), which first takes bd_disk->bd_mutex and then
* takes zvol_state_lock, whereas this code path first takes
* zvol_state_lock, and then takes bd_disk->bd_mutex.
*/
mutex_exit(&zvol_state_lock);
add_disk(zv->zv_disk);
} else {
mutex_exit(&zvol_state_lock);
}
return (SET_ERROR(error));
}
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
/*
* Rename a block device minor mode for the specified volume.
*/
static void
zvol_rename_minor(zvol_state_t *zv, const char *newname)
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
{
int readonly = get_disk_ro(zv->zv_disk);
ASSERT(MUTEX_HELD(&zvol_state_lock));
strlcpy(zv->zv_name, newname, sizeof (zv->zv_name));
/*
* The block device's read-only state is briefly changed causing
* a KOBJ_CHANGE uevent to be issued. This ensures udev detects
* the name change and fixes the symlinks. This does not change
* ZVOL_RDONLY in zv->zv_flags so the actual read-only state never
* changes. This would normally be done using kobject_uevent() but
* that is a GPL-only symbol which is why we need this workaround.
*/
set_disk_ro(zv->zv_disk, !readonly);
set_disk_ro(zv->zv_disk, readonly);
}
/*
* Mask errors to continue dmu_objset_find() traversal
*/
static int
zvol_create_snap_minor_cb(const char *dsname, void *arg)
{
const char *name = (const char *)arg;
ASSERT0(MUTEX_HELD(&spa_namespace_lock));
/* skip the designated dataset */
if (name && strcmp(dsname, name) == 0)
return (0);
/* at this point, the dsname should name a snapshot */
if (strchr(dsname, '@') == 0) {
dprintf("zvol_create_snap_minor_cb(): "
"%s is not a shapshot name\n", dsname);
} else {
(void) zvol_create_minor_impl(dsname);
}
return (0);
}
/*
* Mask errors to continue dmu_objset_find() traversal
*/
static int
zvol_create_minors_cb(const char *dsname, void *arg)
{
uint64_t snapdev;
int error;
ASSERT0(MUTEX_HELD(&spa_namespace_lock));
error = dsl_prop_get_integer(dsname, "snapdev", &snapdev, NULL);
if (error)
return (0);
/*
* Given the name and the 'snapdev' property, create device minor nodes
* with the linkages to zvols/snapshots as needed.
* If the name represents a zvol, create a minor node for the zvol, then
* check if its snapshots are 'visible', and if so, iterate over the
* snapshots and create device minor nodes for those.
*/
if (strchr(dsname, '@') == 0) {
/* create minor for the 'dsname' explicitly */
error = zvol_create_minor_impl(dsname);
if ((error == 0 || error == EEXIST) &&
(snapdev == ZFS_SNAPDEV_VISIBLE)) {
fstrans_cookie_t cookie = spl_fstrans_mark();
/*
* traverse snapshots only, do not traverse children,
* and skip the 'dsname'
*/
error = dmu_objset_find((char *)dsname,
zvol_create_snap_minor_cb, (void *)dsname,
DS_FIND_SNAPSHOTS);
spl_fstrans_unmark(cookie);
}
} else {
dprintf("zvol_create_minors_cb(): %s is not a zvol name\n",
dsname);
}
return (0);
}
/*
* Create minors for the specified dataset, including children and snapshots.
* Pay attention to the 'snapdev' property and iterate over the snapshots
* only if they are 'visible'. This approach allows one to assure that the
* snapshot metadata is read from disk only if it is needed.
*
* The name can represent a dataset to be recursively scanned for zvols and
* their snapshots, or a single zvol snapshot. If the name represents a
* dataset, the scan is performed in two nested stages:
* - scan the dataset for zvols, and
* - for each zvol, create a minor node, then check if the zvol's snapshots
* are 'visible', and only then iterate over the snapshots if needed
*
* If the name represents a snapshot, a check is perfromed if the snapshot is
* 'visible' (which also verifies that the parent is a zvol), and if so,
* a minor node for that snapshot is created.
*/
static int
zvol_create_minors_impl(const char *name)
{
int error = 0;
fstrans_cookie_t cookie;
char *atp, *parent;
if (zvol_inhibit_dev)
return (0);
parent = kmem_alloc(MAXPATHLEN, KM_SLEEP);
(void) strlcpy(parent, name, MAXPATHLEN);
if ((atp = strrchr(parent, '@')) != NULL) {
uint64_t snapdev;
*atp = '\0';
error = dsl_prop_get_integer(parent, "snapdev",
&snapdev, NULL);
if (error == 0 && snapdev == ZFS_SNAPDEV_VISIBLE)
error = zvol_create_minor_impl(name);
} else {
cookie = spl_fstrans_mark();
error = dmu_objset_find(parent, zvol_create_minors_cb,
NULL, DS_FIND_CHILDREN);
spl_fstrans_unmark(cookie);
}
kmem_free(parent, MAXPATHLEN);
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
return (SET_ERROR(error));
}
/*
* Remove minors for specified dataset including children and snapshots.
*/
static void
zvol_remove_minors_impl(const char *name)
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
{
zvol_state_t *zv, *zv_next;
int namelen = ((name) ? strlen(name) : 0);
if (zvol_inhibit_dev)
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
return;
mutex_enter(&zvol_state_lock);
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
if (name == NULL || strcmp(zv->zv_name, name) == 0 ||
(strncmp(zv->zv_name, name, namelen) == 0 &&
(zv->zv_name[namelen] == '/' ||
zv->zv_name[namelen] == '@'))) {
/* If in use, leave alone */
if (zv->zv_open_count > 0)
continue;
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
zvol_remove(zv);
zvol_free(zv);
}
}
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
mutex_exit(&zvol_state_lock);
}
/* Remove minor for this specific snapshot only */
static void
zvol_remove_minor_impl(const char *name)
{
zvol_state_t *zv, *zv_next;
if (zvol_inhibit_dev)
return;
if (strchr(name, '@') == NULL)
return;
mutex_enter(&zvol_state_lock);
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
if (strcmp(zv->zv_name, name) == 0) {
/* If in use, leave alone */
if (zv->zv_open_count > 0)
continue;
zvol_remove(zv);
zvol_free(zv);
break;
}
}
mutex_exit(&zvol_state_lock);
}
/*
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
* Rename minors for specified dataset including children and snapshots.
*/
static void
zvol_rename_minors_impl(const char *oldname, const char *newname)
{
zvol_state_t *zv, *zv_next;
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
int oldnamelen, newnamelen;
char *name;
if (zvol_inhibit_dev)
return;
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
oldnamelen = strlen(oldname);
newnamelen = strlen(newname);
name = kmem_alloc(MAXNAMELEN, KM_SLEEP);
mutex_enter(&zvol_state_lock);
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
for (zv = list_head(&zvol_state_list); zv != NULL; zv = zv_next) {
zv_next = list_next(&zvol_state_list, zv);
/* If in use, leave alone */
if (zv->zv_open_count > 0)
continue;
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
if (strcmp(zv->zv_name, oldname) == 0) {
zvol_rename_minor(zv, newname);
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
} else if (strncmp(zv->zv_name, oldname, oldnamelen) == 0 &&
(zv->zv_name[oldnamelen] == '/' ||
zv->zv_name[oldnamelen] == '@')) {
snprintf(name, MAXNAMELEN, "%s%c%s", newname,
zv->zv_name[oldnamelen],
zv->zv_name + oldnamelen + 1);
zvol_rename_minor(zv, name);
}
}
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
mutex_exit(&zvol_state_lock);
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
kmem_free(name, MAXNAMELEN);
}
typedef struct zvol_snapdev_cb_arg {
uint64_t snapdev;
} zvol_snapdev_cb_arg_t;
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
static int
zvol_set_snapdev_cb(const char *dsname, void *param) {
zvol_snapdev_cb_arg_t *arg = param;
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
if (strchr(dsname, '@') == NULL)
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
return (0);
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
switch (arg->snapdev) {
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
case ZFS_SNAPDEV_VISIBLE:
(void) zvol_create_minor_impl(dsname);
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
break;
case ZFS_SNAPDEV_HIDDEN:
(void) zvol_remove_minor_impl(dsname);
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
break;
}
Remove ZFC_IOC_*_MINOR ioctl()s Early versions of ZFS coordinated the creation and destruction of device minors from userspace. This was inherently racy and in late 2009 these ioctl()s were removed leaving everything up to the kernel. This significantly simplified the code. However, we never picked up these changes in ZoL since we'd already significantly adjusted this code for Linux. This patch aims to rectify that by finally removing ZFC_IOC_*_MINOR ioctl()s and moving all the functionality down in to the kernel. Since this cleanup will change the kernel/user ABI it's being done in the same tag as the previous libzfs_core ABI changes. This will minimize, but not eliminate, the disruption to end users. Once merged ZoL, Illumos, and FreeBSD will basically be back in sync in regards to handling ZVOLs in the common code. While each platform must have its own custom zvol.c implemenation the interfaces provided are consistent. NOTES: 1) This patch introduces one subtle change in behavior which could not be easily avoided. Prior to this change callers of 'zfs create -V ...' were guaranteed that upon exit the /dev/zvol/ block device link would be created or an error returned. That's no longer the case. The utilities will no longer block waiting for the symlink to be created. Callers are now responsible for blocking, this is why a 'udev_wait' call was added to the 'label' function in scripts/common.sh. 2) The read-only behavior of a ZVOL now solely depends on if the ZVOL_RDONLY bit is set in zv->zv_flags. The redundant policy setting in the gendisk structure was removed. This both simplifies the code and allows us to safely leverage set_disk_ro() to issue a KOBJ_CHANGE uevent. See the comment in the code for futher details on this. 3) Because __zvol_create_minor() and zvol_alloc() may now be called in a sync task they must use KM_PUSHPAGE. References: illumos/illumos-gate@681d9761e8516a7dc5ab6589e2dfe717777e1123 Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Closes #1969
2013-12-06 22:20:22 +00:00
return (0);
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
}
static void
zvol_set_snapdev_impl(char *name, uint64_t snapdev)
{
zvol_snapdev_cb_arg_t arg = {snapdev};
fstrans_cookie_t cookie = spl_fstrans_mark();
/*
* The zvol_set_snapdev_sync() sets snapdev appropriately
* in the dataset hierarchy. Here, we only scan snapshots.
*/
dmu_objset_find(name, zvol_set_snapdev_cb, &arg, DS_FIND_SNAPSHOTS);
spl_fstrans_unmark(cookie);
}
static zvol_task_t *
zvol_task_alloc(zvol_async_op_t op, const char *name1, const char *name2,
uint64_t snapdev)
{
zvol_task_t *task;
char *delim;
/* Never allow tasks on hidden names. */
if (name1[0] == '$')
return (NULL);
task = kmem_zalloc(sizeof (zvol_task_t), KM_SLEEP);
task->op = op;
task->snapdev = snapdev;
delim = strchr(name1, '/');
strlcpy(task->pool, name1, delim ? (delim - name1 + 1) : MAXNAMELEN);
strlcpy(task->name1, name1, MAXNAMELEN);
if (name2 != NULL)
strlcpy(task->name2, name2, MAXNAMELEN);
return (task);
}
static void
zvol_task_free(zvol_task_t *task)
{
kmem_free(task, sizeof (zvol_task_t));
}
/*
* The worker thread function performed asynchronously.
*/
static void
zvol_task_cb(void *param)
{
zvol_task_t *task = (zvol_task_t *)param;
switch (task->op) {
case ZVOL_ASYNC_CREATE_MINORS:
(void) zvol_create_minors_impl(task->name1);
break;
case ZVOL_ASYNC_REMOVE_MINORS:
zvol_remove_minors_impl(task->name1);
break;
case ZVOL_ASYNC_RENAME_MINORS:
zvol_rename_minors_impl(task->name1, task->name2);
break;
case ZVOL_ASYNC_SET_SNAPDEV:
zvol_set_snapdev_impl(task->name1, task->snapdev);
break;
default:
VERIFY(0);
break;
}
zvol_task_free(task);
}
typedef struct zvol_set_snapdev_arg {
const char *zsda_name;
uint64_t zsda_value;
zprop_source_t zsda_source;
dmu_tx_t *zsda_tx;
} zvol_set_snapdev_arg_t;
/*
* Sanity check the dataset for safe use by the sync task. No additional
* conditions are imposed.
*/
static int
zvol_set_snapdev_check(void *arg, dmu_tx_t *tx)
{
zvol_set_snapdev_arg_t *zsda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *dd;
int error;
error = dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL);
if (error != 0)
return (error);
dsl_dir_rele(dd, FTAG);
return (error);
}
static int
zvol_set_snapdev_sync_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
{
zvol_set_snapdev_arg_t *zsda = arg;
char dsname[MAXNAMELEN];
zvol_task_t *task;
dsl_dataset_name(ds, dsname);
dsl_prop_set_sync_impl(ds, zfs_prop_to_name(ZFS_PROP_SNAPDEV),
zsda->zsda_source, sizeof (zsda->zsda_value), 1,
&zsda->zsda_value, zsda->zsda_tx);
task = zvol_task_alloc(ZVOL_ASYNC_SET_SNAPDEV, dsname,
NULL, zsda->zsda_value);
if (task == NULL)
return (0);
(void) taskq_dispatch(dp->dp_spa->spa_zvol_taskq, zvol_task_cb,
task, TQ_SLEEP);
return (0);
}
/*
* Traverse all child snapshot datasets and apply snapdev appropriately.
*/
static void
zvol_set_snapdev_sync(void *arg, dmu_tx_t *tx)
{
zvol_set_snapdev_arg_t *zsda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *dd;
VERIFY0(dsl_dir_hold(dp, zsda->zsda_name, FTAG, &dd, NULL));
zsda->zsda_tx = tx;
dmu_objset_find_dp(dp, dd->dd_object, zvol_set_snapdev_sync_cb,
zsda, DS_FIND_CHILDREN);
dsl_dir_rele(dd, FTAG);
}
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
int
zvol_set_snapdev(const char *ddname, zprop_source_t source, uint64_t snapdev)
{
zvol_set_snapdev_arg_t zsda;
zsda.zsda_name = ddname;
zsda.zsda_source = source;
zsda.zsda_value = snapdev;
return (dsl_sync_task(ddname, zvol_set_snapdev_check,
zvol_set_snapdev_sync, &zsda, 0, ZFS_SPACE_CHECK_NONE));
}
void
zvol_create_minors(spa_t *spa, const char *name, boolean_t async)
{
zvol_task_t *task;
taskqid_t id;
task = zvol_task_alloc(ZVOL_ASYNC_CREATE_MINORS, name, NULL, ~0ULL);
if (task == NULL)
return;
id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
if ((async == B_FALSE) && (id != 0))
taskq_wait_id(spa->spa_zvol_taskq, id);
}
void
zvol_remove_minors(spa_t *spa, const char *name, boolean_t async)
{
zvol_task_t *task;
taskqid_t id;
task = zvol_task_alloc(ZVOL_ASYNC_REMOVE_MINORS, name, NULL, ~0ULL);
if (task == NULL)
return;
id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
if ((async == B_FALSE) && (id != 0))
taskq_wait_id(spa->spa_zvol_taskq, id);
}
void
zvol_rename_minors(spa_t *spa, const char *name1, const char *name2,
boolean_t async)
{
zvol_task_t *task;
taskqid_t id;
task = zvol_task_alloc(ZVOL_ASYNC_RENAME_MINORS, name1, name2, ~0ULL);
if (task == NULL)
return;
id = taskq_dispatch(spa->spa_zvol_taskq, zvol_task_cb, task, TQ_SLEEP);
if ((async == B_FALSE) && (id != 0))
taskq_wait_id(spa->spa_zvol_taskq, id);
Add snapdev=[hidden|visible] dataset property The new snapdev dataset property may be set to control the visibility of zvol snapshot devices. By default this value is set to 'hidden' which will prevent zvol snapshots from appearing under /dev/zvol/ and /dev/<dataset>/. When set to 'visible' all zvol snapshots for the dataset will be visible. This functionality was largely added because when automatic snapshoting is enabled large numbers of read-only zvol snapshots will be created. When creating these devices the kernel will attempt to read their partition tables, and blkid will attempt to identify any filesystems on those partitions. This leads to a variety of issues: 1) The zvol partition tables will be read in the context of the `modprobe zfs` for automatically imported pools. This is undesirable and should be done asynchronously, but for now reducing the number of visible devices helps. 2) Udev expects to be able to complete its work for a new block devices fairly quickly. When many zvol devices are added at the same time this is no longer be true. It can lead to udev timeouts and missing /dev/zvol links. 3) Simply having lots of devices in /dev/ can be aukward from a management standpoint. Hidding the devices your unlikely to ever use helps with this. Any snapshot device which is needed can be made visible by changing the snapdev property. NOTE: This patch changes the default behavior for zvols which was effectively 'snapdev=visible'. Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #1235 Closes #945 Issue #956 Issue #756
2013-02-13 23:11:59 +00:00
}
int
zvol_init(void)
{
int error;
Cleanup zvol initialization code The following error will occur on some (possibly all) kernels because blk_init_queue() will try to take the spinlock before we initialize it. BUG: spinlock bad magic on CPU#0, zpool/4054 lock: 0xffff88021a73de60, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0 Pid: 4054, comm: zpool Not tainted 3.9.3 #11 Call Trace: [<ffffffff81478ef8>] spin_dump+0x8c/0x91 [<ffffffff81478f1e>] spin_bug+0x21/0x26 [<ffffffff812da097>] do_raw_spin_lock+0x127/0x130 [<ffffffff8147d851>] _raw_spin_lock_irq+0x21/0x30 [<ffffffff812c2c1e>] cfq_init_queue+0x1fe/0x350 [<ffffffff812aacb8>] elevator_init+0x78/0x140 [<ffffffff812b2677>] blk_init_allocated_queue+0x87/0xb0 [<ffffffff812b26d5>] blk_init_queue_node+0x35/0x70 [<ffffffff812b271e>] blk_init_queue+0xe/0x10 [<ffffffff8125211b>] __zvol_create_minor+0x24b/0x620 [<ffffffff81253264>] zvol_create_minors_cb+0x24/0x30 [<ffffffff811bd9ca>] dmu_objset_find_spa+0xea/0x510 [<ffffffff811bda71>] dmu_objset_find_spa+0x191/0x510 [<ffffffff81253ea2>] zvol_create_minors+0x92/0x180 [<ffffffff811f8d80>] spa_open_common+0x250/0x380 [<ffffffff811f8ece>] spa_open+0xe/0x10 [<ffffffff8122817e>] pool_status_check.part.22+0x1e/0x80 [<ffffffff81228a55>] zfsdev_ioctl+0x155/0x190 [<ffffffff8116a695>] do_vfs_ioctl+0x325/0x5a0 [<ffffffff8116a950>] sys_ioctl+0x40/0x80 [<ffffffff814812c9>] ? do_page_fault+0x9/0x10 [<ffffffff81483929>] system_call_fastpath+0x16/0x1b zd0: unknown partition table We fix this by calling spin_lock_init before blk_init_queue. The manner in which zvol_init() initializes structures is suspectible to a race between initialization and a probe on a zvol. We reorganize zvol_init() to prevent that. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2013-07-02 18:59:10 +00:00
list_create(&zvol_state_list, sizeof (zvol_state_t),
offsetof(zvol_state_t, zv_next));
Cleanup zvol initialization code The following error will occur on some (possibly all) kernels because blk_init_queue() will try to take the spinlock before we initialize it. BUG: spinlock bad magic on CPU#0, zpool/4054 lock: 0xffff88021a73de60, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0 Pid: 4054, comm: zpool Not tainted 3.9.3 #11 Call Trace: [<ffffffff81478ef8>] spin_dump+0x8c/0x91 [<ffffffff81478f1e>] spin_bug+0x21/0x26 [<ffffffff812da097>] do_raw_spin_lock+0x127/0x130 [<ffffffff8147d851>] _raw_spin_lock_irq+0x21/0x30 [<ffffffff812c2c1e>] cfq_init_queue+0x1fe/0x350 [<ffffffff812aacb8>] elevator_init+0x78/0x140 [<ffffffff812b2677>] blk_init_allocated_queue+0x87/0xb0 [<ffffffff812b26d5>] blk_init_queue_node+0x35/0x70 [<ffffffff812b271e>] blk_init_queue+0xe/0x10 [<ffffffff8125211b>] __zvol_create_minor+0x24b/0x620 [<ffffffff81253264>] zvol_create_minors_cb+0x24/0x30 [<ffffffff811bd9ca>] dmu_objset_find_spa+0xea/0x510 [<ffffffff811bda71>] dmu_objset_find_spa+0x191/0x510 [<ffffffff81253ea2>] zvol_create_minors+0x92/0x180 [<ffffffff811f8d80>] spa_open_common+0x250/0x380 [<ffffffff811f8ece>] spa_open+0xe/0x10 [<ffffffff8122817e>] pool_status_check.part.22+0x1e/0x80 [<ffffffff81228a55>] zfsdev_ioctl+0x155/0x190 [<ffffffff8116a695>] do_vfs_ioctl+0x325/0x5a0 [<ffffffff8116a950>] sys_ioctl+0x40/0x80 [<ffffffff814812c9>] ? do_page_fault+0x9/0x10 [<ffffffff81483929>] system_call_fastpath+0x16/0x1b zd0: unknown partition table We fix this by calling spin_lock_init before blk_init_queue. The manner in which zvol_init() initializes structures is suspectible to a race between initialization and a probe on a zvol. We reorganize zvol_init() to prevent that. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2013-07-02 18:59:10 +00:00
mutex_init(&zvol_state_lock, NULL, MUTEX_DEFAULT, NULL);
error = register_blkdev(zvol_major, ZVOL_DRIVER);
if (error) {
printk(KERN_INFO "ZFS: register_blkdev() failed %d\n", error);
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
goto out;
}
blk_register_region(MKDEV(zvol_major, 0), 1UL << MINORBITS,
THIS_MODULE, zvol_probe, NULL, NULL);
return (0);
Cleanup zvol initialization code The following error will occur on some (possibly all) kernels because blk_init_queue() will try to take the spinlock before we initialize it. BUG: spinlock bad magic on CPU#0, zpool/4054 lock: 0xffff88021a73de60, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0 Pid: 4054, comm: zpool Not tainted 3.9.3 #11 Call Trace: [<ffffffff81478ef8>] spin_dump+0x8c/0x91 [<ffffffff81478f1e>] spin_bug+0x21/0x26 [<ffffffff812da097>] do_raw_spin_lock+0x127/0x130 [<ffffffff8147d851>] _raw_spin_lock_irq+0x21/0x30 [<ffffffff812c2c1e>] cfq_init_queue+0x1fe/0x350 [<ffffffff812aacb8>] elevator_init+0x78/0x140 [<ffffffff812b2677>] blk_init_allocated_queue+0x87/0xb0 [<ffffffff812b26d5>] blk_init_queue_node+0x35/0x70 [<ffffffff812b271e>] blk_init_queue+0xe/0x10 [<ffffffff8125211b>] __zvol_create_minor+0x24b/0x620 [<ffffffff81253264>] zvol_create_minors_cb+0x24/0x30 [<ffffffff811bd9ca>] dmu_objset_find_spa+0xea/0x510 [<ffffffff811bda71>] dmu_objset_find_spa+0x191/0x510 [<ffffffff81253ea2>] zvol_create_minors+0x92/0x180 [<ffffffff811f8d80>] spa_open_common+0x250/0x380 [<ffffffff811f8ece>] spa_open+0xe/0x10 [<ffffffff8122817e>] pool_status_check.part.22+0x1e/0x80 [<ffffffff81228a55>] zfsdev_ioctl+0x155/0x190 [<ffffffff8116a695>] do_vfs_ioctl+0x325/0x5a0 [<ffffffff8116a950>] sys_ioctl+0x40/0x80 [<ffffffff814812c9>] ? do_page_fault+0x9/0x10 [<ffffffff81483929>] system_call_fastpath+0x16/0x1b zd0: unknown partition table We fix this by calling spin_lock_init before blk_init_queue. The manner in which zvol_init() initializes structures is suspectible to a race between initialization and a probe on a zvol. We reorganize zvol_init() to prevent that. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2013-07-02 18:59:10 +00:00
zvol processing should use struct bio Internally, zvols are files exposed through the block device API. This is intended to reduce overhead when things require block devices. However, the ZoL zvol code emulates a traditional block device in that it has a top half and a bottom half. This is an unnecessary source of overhead that does not exist on any other OpenZFS platform does this. This patch removes it. Early users of this patch reported double digit performance gains in IOPS on zvols in the range of 50% to 80%. Comments in the code suggest that the current implementation was done to obtain IO merging from Linux's IO elevator. However, the DMU already does write merging while arc_read() should implicitly merge read IOs because only 1 thread is permitted to fetch the buffer into ARC. In addition, commercial ZFSOnLinux distributions report that regular files are more performant than zvols under the current implementation, and the main consumers of zvols are VMs and iSCSI targets, which have their own elevators to merge IOs. Some minor refactoring allows us to register zfs_request() as our ->make_request() handler in place of the generic_make_request() function. This eliminates the layer of code that broke IO requests on zvols into a top half and a bottom half. This has several benefits: 1. No per zvol spinlocks. 2. No redundant IO elevator processing. 3. Interrupts are disabled only when actually necessary. 4. No redispatching of IOs when all taskq threads are busy. 5. Linux's page out routines will properly block. 6. Many autotools checks become obsolete. An unfortunate consequence of eliminating the layer that generic_make_request() is that we no longer calls the instrumentation hooks for block IO accounting. Those hooks are GPL-exported, so we cannot call them ourselves and consequently, we lose the ability to do IO monitoring via iostat. Since zvols are internally files mapped as block devices, this should be okay. Anyone who is willing to accept the performance penalty for the block IO layer's accounting could use the loop device in between the zvol and its consumer. Alternatively, perf and ftrace likely could be used. Also, tools like latencytop will still work. Tools such as latencytop sometimes provide a better view of performance bottlenecks than the traditional block IO accounting tools do. Lastly, if direct reclaim occurs during spacemap loading and swap is on a zvol, this code will deadlock. That deadlock could already occur with sync=always on zvols. Given that swap on zvols is not yet production ready, this is not a blocker. Signed-off-by: Richard Yao <ryao@gentoo.org>
2014-07-04 22:43:47 +00:00
out:
Cleanup zvol initialization code The following error will occur on some (possibly all) kernels because blk_init_queue() will try to take the spinlock before we initialize it. BUG: spinlock bad magic on CPU#0, zpool/4054 lock: 0xffff88021a73de60, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0 Pid: 4054, comm: zpool Not tainted 3.9.3 #11 Call Trace: [<ffffffff81478ef8>] spin_dump+0x8c/0x91 [<ffffffff81478f1e>] spin_bug+0x21/0x26 [<ffffffff812da097>] do_raw_spin_lock+0x127/0x130 [<ffffffff8147d851>] _raw_spin_lock_irq+0x21/0x30 [<ffffffff812c2c1e>] cfq_init_queue+0x1fe/0x350 [<ffffffff812aacb8>] elevator_init+0x78/0x140 [<ffffffff812b2677>] blk_init_allocated_queue+0x87/0xb0 [<ffffffff812b26d5>] blk_init_queue_node+0x35/0x70 [<ffffffff812b271e>] blk_init_queue+0xe/0x10 [<ffffffff8125211b>] __zvol_create_minor+0x24b/0x620 [<ffffffff81253264>] zvol_create_minors_cb+0x24/0x30 [<ffffffff811bd9ca>] dmu_objset_find_spa+0xea/0x510 [<ffffffff811bda71>] dmu_objset_find_spa+0x191/0x510 [<ffffffff81253ea2>] zvol_create_minors+0x92/0x180 [<ffffffff811f8d80>] spa_open_common+0x250/0x380 [<ffffffff811f8ece>] spa_open+0xe/0x10 [<ffffffff8122817e>] pool_status_check.part.22+0x1e/0x80 [<ffffffff81228a55>] zfsdev_ioctl+0x155/0x190 [<ffffffff8116a695>] do_vfs_ioctl+0x325/0x5a0 [<ffffffff8116a950>] sys_ioctl+0x40/0x80 [<ffffffff814812c9>] ? do_page_fault+0x9/0x10 [<ffffffff81483929>] system_call_fastpath+0x16/0x1b zd0: unknown partition table We fix this by calling spin_lock_init before blk_init_queue. The manner in which zvol_init() initializes structures is suspectible to a race between initialization and a probe on a zvol. We reorganize zvol_init() to prevent that. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
2013-07-02 18:59:10 +00:00
mutex_destroy(&zvol_state_lock);
list_destroy(&zvol_state_list);
return (SET_ERROR(error));
}
void
zvol_fini(void)
{
zvol_remove_minors_impl(NULL);
blk_unregister_region(MKDEV(zvol_major, 0), 1UL << MINORBITS);
unregister_blkdev(zvol_major, ZVOL_DRIVER);
list_destroy(&zvol_state_list);
mutex_destroy(&zvol_state_lock);
}
module_param(zvol_inhibit_dev, uint, 0644);
MODULE_PARM_DESC(zvol_inhibit_dev, "Do not create zvol device nodes");
module_param(zvol_major, uint, 0444);
MODULE_PARM_DESC(zvol_major, "Major number for zvol device");
module_param(zvol_max_discard_blocks, ulong, 0444);
MODULE_PARM_DESC(zvol_max_discard_blocks, "Max number of blocks to discard");
module_param(zvol_prefetch_bytes, uint, 0644);
MODULE_PARM_DESC(zvol_prefetch_bytes, "Prefetch N bytes at zvol start+end");