It's my understanding that the zfs_fsyncer_key TSD was added as
a performance omtimization to reduce contention on the zl_lock
from zil_commit(). This issue manifested itself as very long
(100+ms) fsync() system call times for fsync() heavy workloads.
However, under Linux I'm not seeing the same contention that
was originally described. Therefore, I'm removing this code
in order to ween ourselves off any dependence on TSD. If the
original performance issue reappears on Linux we can revisit
fixing it without resorting to TSD.
This just leaves one small ZFS TSD consumer. If it can be
cleanly removed from the code we'll be able to shed the SPL
TSD implementation entirely.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closeszfsonlinux/spl#174
The current state of udev and devicer-mapper devices makes it difficult
to construct a mapping of DM partitions and their underlying DM device.
For example, with a /dev directory with the following contents:
$ ls -d /dev/dm-*
/dev/dm-0
/dev/dm-1
/dev/dm-2
/dev/dm-3
it is not immediately apparent if these are completely separate devices,
or partitions and real devices intermixed. In contrast, SCSI devices
would appear as so:
$ ls -d /dev/sd*
/dev/sda
/dev/sda1
/dev/sdb
/dev/sdb1
Here, one can immediately determine that there are two devices (sda and
sdb), each containing a single partition. The lack of a predictable and
consistent mapping from DM devices to DM device partitions makes it
difficult for user space to process these devices the same way it does
SCSI devices.
As a result, the ZFS utilities do not partition DM devices, and instead
set the "vdev_wholedisk" label to 0 and treat them as partitions. This
has the side effect that, even if ZFS has sole ownership of the device,
the IO scheduler will not be modified because it is treated as a
partition.
This change adds an exception for DM devices in vdev_elevator_switch,
allowing the elevator to be modified even though the "vdev_wholedisk"
property is not set.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1149
During the original ZoL port the vdev_uses_zvols() function was
disabled until it could be properly implemented. This prevented
a zpool from use a zvol for its slog device.
This patch implements that missing functionality by adding a
zvol_is_zvol() function to zvol.c. Given the full path to a
device it will lookup the device and verify its major number
against the registered zvol major number for the system. If
they match we know the device is a zvol.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1131
Fix setting/getting users/groups in quota properties through
numeric identifier. This support was accidentally disabled
in the original port by applying the HAVE_IDMAP wrapper macro
too broadly.
Fix obtained by moving #ifdef HAVE_IDMAP to exclude only
the part of code that really needs IDMAP. Now zfs (get|set)
(user|group)quota@1000 works as expected.
Signed-off-by: Massimo Maggi <massimo@mmmm.it>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1147
A Gentoo user reported an issue where the build system would
attempt to recurse into the kernel source tree if KERNEL_DIR
is set in the environment. KERNEL_DIR is an environment variable
that is used when the kernel sources are in a non-standard
location, so it is necessary to stop relying on it to prevent
this issue.
https://bugs.gentoo.org/show_bug.cgi?id=433946
Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Revert the portion of commit d3aa3ea which always resulted in the
SAs being update when an mmap()'ed file was closed. That change
accidentally resulted in unexpected ctime updates which upset tools
like git. That was always a horrible hack and I'm happy it will
never make it in to a tagged release.
The right fix is something I initially resisted doing because I
was worried about the additional overhead. However, in hindsight
the overhead isn't as bad as I feared.
This patch implemented the sops->dirty_inode() callback which is
unsurprisingly called when an inode is dirtied. We leverage this
callback to keep the znode SAs strictly in sync with the inode.
However, for now we're going to go slowly to avoid introducing
any new unexpected issues by only updating the atime, mtime, and
ctime. This will cover the callpath of most concern to us.
->filemap_page_mkwrite->file_update_time->update_time->
mark_inode_dirty_sync->__mark_inode_dirty->dirty_inode
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#764Closes#1140
Commit 2957f38 renamed 60-vdev.rules to 69-vdev.rules but failed
to update the .gitignore file to reflect this change.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Ensure that the path member pointers are associated with the
newly-mounted snapshot when zpl_snapdir_automount() returns. Otherwise
the follow_automount() function may be called repeatedly, leading to an
incorrect ELOOP error return. This problem was observed as a 'Too many
levels of symbolic links' error from user-space commands accessing an
unmounted snapshot in the .zfs/snapshot directory.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#816
Linux kernel commit d8e794d accidentally broke the delayed work
APIs for non-GPL callers. While the APIs to schedule a delayed
work item are still available to all callers, it is no longer
possible to initialize the delayed work item.
I'm cautiously optimistic we could get the delayed_work_timer_fn
exported for all callers in the upstream kernel. But frankly
the compatibility code to use this kernel interface has always
been problematic.
Therefore, this patch abandons direct use the of the Linux
kernel interface in favor of the new delayed taskq interface.
It provides roughly the same functionality as delayed work queues
but it's a stable interface under our control.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1053
When writes to zvols invoke ZIL, zfs_range_new_proxy() is called,
which allocates memory using KM_SLEEP, triggering a warning.
Switch to KM_PUSHPAGE to silence that warning. See commit
b8d06fca08 for additional details.
Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1138
This reverts commit b00131d43c which
is no longer needed due to e89260a1c8.
This change forces all xattr znodes to hold a reference on their
parent which ensures prune_icache() will never attempt to evict
both the parent and child concurrently. This effectively prevents
the deadlock condition from ever occuring.
Therefore we can safely revert back to the upstream synchronous
cleanup code. This is nice because it keeps our code base closer
to upstream and resolves the performance issues introduced by the
original deadlock fix.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#457
When updating a file via mmap()'ed I/O preserve the mtime/ctime
which were updated when the page was made writable by the generic
callback filemap_page_mkwrite().
But more importantly than preserving the exact time add the missing
call to sa_bulk_update(). This ensures that the znode modifications
are written to disk as part of the transaction. Without this the
inode may mistaken rollback to the previous on-disk znode state.
Additionally, for mmap()'ed znodes explicitly set the atime, mtime,
and ctime on close using the up to date values in the inode. This
is critical because writepage() may occur after close and on close
we need to ensure the values are correct.
Original-patch-by: Richard Yao <ryao@cs.stonybrook.edu>
Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#764
Incorrect syntax should never cause a segfault. In this case
listing multiple comma delimited options after '-o' triggered
the problem. For example:
zpool create -o ashift=12,listsnaps=on
This patch resolves the issue by wrapping the calls which use
hdr with a NULL test.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1118
Add a vdev_id feature to map device names based on already defined
udev device links. To increase the odds that vdev_id will run after
the rules it depends on, increase the vdev.rules rule number from 60
to 69. With this change, vdev_id now provides functionality analogous
to zpool_id and zpool_layout, paving the way to retire those tools.
A defined alias takes precedence over a topology-derived name, but the
two naming methods can otherwise coexist. For example, one might name
drives in a JBOD with the sas_direct topology while naming an internal
L2ARC device with an alias.
For example, the following lines in vdev_id.conf will result in the
creation of links /dev/disk/by-vdev/{d1,d2}, each pointing to the same
target as the device link specified in the third field.
# by-vdev
# name fully qualified or base name of device link
alias d1 /dev/disk/by-id/wwn-0x5000c5002de3b9ca
alias d2 wwn-0x5000c5002def789e
Also perform some minor vdev_id cleanup, such as removal of the unused
-s command line option.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#981
Unlike normal file or directory znodes, an xattr znode is
guaranteed to only have a single parent. Therefore, we can
take a refernce on that parent if it is provided at create
time and cache it. Additionally, we take care to cache it
on any subsequent zfs_zaccess() where the parent is provided
as an optimization.
This allows us to avoid needing to do a zfs_zget() when
setting up the SELinux security xattr in the create path.
This is critical because a hash lookup on the directory
will deadlock since it is locked.
The zpl_xattr_security_init() call has also been moved up
to the zpl layer to ensure TXs to create the required
xattrs are performed after the create TX. Otherwise we
run the risk of deadlocking on the open create TX.
Ideally the security xattr should be fully constructed
before the new inode is unlocked. However, doing so would
require far more extensive changes to ZFS.
This change may also have the benefitial side effect of
ensuring xattr directory znodes are evicted from the cache
before normal file or directory znodes due to the extra
reference.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#671
Add the initial support for the 'smbshare' option using the
existing libshare infrastructure. Because this implementation
relies on usershares samba version 3.0.23 is required.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#493
Commit df83110856 missed update to
getopt() call, while delivering all the rest. This commit adds
"o" to getopt().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #566
Add the missing error handling to load_nvlist(). There's no good
reason this needs to be fatal. All callers of load_nvlist() do
correctly handle an error condition and it is preferable that an
error be returned. This will allow 'zpool import -FX' to safely
attempt to rollback through previous txgs looking for a good one.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1120
Commit 57a4edd allows the bootfs property to be set on any pool.
However, many of the zpool commands still prevent you from using
EFI labeled devices for the root pool. For example:
# zpool attach rpool /dev/sda /dev/sdb
cannot label 'sdb': EFI labeled devices are not supported on
root pools. on root devices.
For non-Solaris builds such as Linux disable this error.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1077
Due to the slightly increased size of the ZFS super block
caused by 30315d2 there are now allocation warnings. The
allocation size is still small (just over 8k) and super
blocks are rarely allocated so we suppress the warning.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1101
Previously this check was only performed when ./configure was
attempting to autodetect your kernel source directory. But we
should also handle the case where --with-linux was provided
and is obviously wrong. This way we catch the error before
invoking make and compiling the source with an incorrect
autoconf results.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closeszfsonlinux/spl#162
While expanding positional parameters shell requires non-single
digits to be enclosed in braces. When the SAS topology is
non-trivial the number of positional parameters generated internally
by vdev_id script (using set -- ...) easily crosses single digit limit
and vdev_id fails to generate links.
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1119
Full bash may not be available in all environments where udev helpers
run, such as in an initial ramdisk. To avoid breakage in this case,
remove use of bash-specific features such as variable arrays and the
`declare' keyword from the vdev_id script.
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#870
If zvol_alloc() fails zv will be set to NULL and dereferenced
in out_dmu_objset_disown. To avoid this entirely the zv->objset
line is moved up in to the success block.
Original-patch-by: Jorgen Lundman <lundman@lundman.net>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1109
Increasing this limit costs us 6144 bytes of memory per mounted
filesystem, but this is small price to pay for accomplishing
the following:
* Allows for up to 256-way concurreny when performing lookups
which helps performance when there are a large number of
processes.
* Minimizes the likelyhood of encountering the deadlock
described in issue #1101. Because vmalloc() won't strictly
honor __GFP_FS there is still a very remote chance of a
deadlock. See the zfsonlinux/spl@043f9b57 commit.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1101
When a zvol with snapshots is renamed the device files under
/dev/zvol/ are not renamed. This patch resolves the problem
by destroying and recreating the minors with the new name so
the links can be recreated bu udev.
Original-patch-by: Suman Chakravartula <schakrava@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#408
Canonicalize the mount point passed to the mount.zfs helper.
This way a clean path is always added to mtab which ensures
the umount can properly locate and remove the entry.
Test case:
$ mkdir /mnt/foo
$ mount -t zfs zpool/foo /mnt/../mnt/foo////
$ umount /mnt/foo
$ cat /etc/mtab | grep zpool/foo
zpool/foo /mnt/../mnt/foo//// zfs rw 0 0
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#573
This branch adds some overdue ashift improvements.
* Add '-o ashift' to 'zpool add' and 'zpool attach'
* Improve AF hard disk detection
* Allow 'zpool import' to handle increases in ashift
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
When adding devices to an existing pool "ashift" property is
auto-detected. However, if this property was overridden at
the pool creation time (i.e. zpool create -o ashift=12 tank ...)
this may not be what the user wants. This commit lets the user
specify the value of "ashift" property to be used with newly
added drives. For example,
zpool add -o ashift=12 tank disk1
zpool attach -o ashift=12 tank disk1 disk2
Signed-off-by: Cyril Plisko <cyril.plisko@mountall.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#566
Use the bdev_physical_block_size() interface to determine the
minimize write size which can be issued without incurring a
read-modify-write operation. This is used to set the ashift
correctly to prevent a performance penalty when using AF hard
disks.
Unfortunately, this interface isn't entirely reliable because
it's not uncommon for disks to misreport this value. For this
reason you may still need to manually set your ashift with:
zpool create -o ashift=12 ...
The solution to this in the upstream Illumos source was to add
a white list of known offending drives. Maintaining such a list
will be a burden, but it still may be worth doing if we can
detect a large number of these drives. This should be considered
as future work.
Reported-by: Richard Yao <ryao@cs.stonybrook.edu>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#916
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Eric Schrock <eric.schrock@delphix.com>
Reviewed by: Richard Elling <richard.elling@richardelling.com>
Reviewed by: Gordon Ross <gwr@nexenta.com>
Reviewed by: Garrett D'Amore <garrett@damore.org>
Approved by: Richard Lowe <richlowe@richlowe.net>
Refererces to Illumos issue:
https://www.illumos.org/issues/2671
This patch has been slightly modified from the upstream Illumos
version. In the upstream implementation a warning message is
logged to the console. To prevent pointless console noise this
notification is now posted as a "ereport.fs.zfs.vdev.bad_ashift"
event.
The event indicates a non-optimial (but entirely safe) ashift
value was used to create the pool. Depending on your workload
this may impact pool performance. Unfortunately, the only way
to correct the issue is to recreate the pool with a new ashift.
NOTE: The unrelated fix to the comment in zpool_main.c appears
in the upstream commit and was preserved for consistnecy.
Ported-by: Cyril Plisko <cyril.plisko@mountall.com>
Reworked-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#955
The udev data directory was hard coded in 60-vdev.rules.in. That causes
a problem when a distribution changes the location of the directory.
This was not an issue in the past because virtually all distributions
used the same path, but that is beginning to change following a decision
by the systemd developers to change the directory location to reflect
their take-over of udev maintainership. The testing branch of Gentoo
Linux adopted this change, which enabled the hardcoded directory
location to trigger a regression.
Signed-off-by: Richard Yao <ryao@cs.stonybrook.edu>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1085
Gunnar Beutner did all the hard work on this one by correctly
identifying that this issue is a race between dmu_sync() and
dbuf_dirty().
Now in all cases the caller is responsible for preventing this
race by making sure the zfs_range_lock() is held when dirtying
a buffer which may be referenced in a log record. The mmap
case which relies on zfs_putpage() was not taking the range
lock. This code was accidentally dropped when the function
was rewritten for the Linux VFS.
This patch adds the required range locking to zfs_putpage().
It also adds the missing ZFS_ENTER()/ZFS_EXIT() macros which
aren't strictly required due to the VFS holding a reference.
However, this makes the code more consistent with the upsteam
code and there's no harm in being extra careful here.
Original-patch-by: Gunnar Beutner <gunnar@beutner.name>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#541
When using a zvol to back a btrfs filesystem the btrfs mount
would hang. This was due to the bio completion callback used
in btrfs assuming that lower level drivers would never modify
the bio->bi_io_vecs after they were submitted via bio_submit().
If they are modified btrfs will miscalculate which pages need
to be unlocked resulting in a hang.
It's worth mentioning that other file systems such as ext[234]
and xfs work fine because they do not make the same assumption
in the bio completion callback.
The most straight forward way to fix the issue is to present
the semantics expected by btrfs. This is done by cloning the
bios attached to each request and then using the clones bvecs
to perform the required accounting. The clones are freed after
each read/write and the original unmodified bios are linked back
in to the request.
Signed-off-by: Chris Wedgwood <cw@f00f.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#469
Bring in support for the new KSTAT_TYPE_TXG type. This allows for
additional visibility in to the txg handling.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
There have been reports of ZFS deadlocking due to what appears to
be a lost IO. This patch addes some debugging to determine the
exact state of the IO which neither 1) completed, 2) failed, or
3) timed out after zio_delay_max (30) seconds.
This information will be logged using the ZFS FMA infrastructure
as a 'delay' event and posted to the internal zevent log. By
default the last 64 events will be kept in the log but the limit
is configurable via the zfs_zevent_len_max module option.
To dump the contents of the log use the 'zpool events -v' command
and look for the resource.fs.zfs.delay event. It will include
various information about the pool, vdev, and zio which may shed
some light on the issue.
In the context of this change the 120 second kernel blocked thread
watchdog has been disabled for synchronous IOs.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #930
Create a kstat file which contains useful statistics about the
last N txgs processed. This can be helpful when analyzing pool
performance. The new KSTAT_TYPE_TXG type was added for this
purpose and it tracks the following statistics per-txg.
txg - Unique txg number
state - State (O)pen/(Q)uiescing/(S)yncing/(C)ommitted
birth; - Creation time
nread - Bytes read
nwritten; - Bytes written
reads - IOPs read
writes - IOPs write
open_time; - Length in nanoseconds the txg was open
quiesce_time - Length in nanoseconds the txg was quiescing
sync_time; - Length in nanoseconds the txg was syncing
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
The interface for the ddt_zap_count() function assumes it can
never fail. However, internally ddt_zap_count() is implemented
with zap_count() which can potentially fail. Now because there
was no way to return the error to the caller a VERIFY was used
to ensure this case never happens.
Unfortunately, it has been observed that pools can be damaged in
such a way that zap_count() fails. The result is that the pool can
not be imported without hitting the VERIFY and crashing the system.
This patch reworks ddt_object_count() so the error can be safely
caught and returned to the caller. This allows a pool which has
be damaged in this way to be safely rewound for import.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#910
This reverts commit a5c20e2a0a which
accidentally introduced a regression for real 4k sector devices.
See issue #1065 for details.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1065
The following warning was originally added to provide visibility
in to how often a dio gets heavily fragmented in to over 16 bios.
This can happen due to constraints imposed by the block device
and may have a negitive impact on performance but is otherwise
harmless. To prevent needless confusion and worry the message
has been removed.
kernel: WARNING: Resized bio's/dio to 32
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
During 'zpool import' all ZPOOL_CONFIG_PATH names are supposed
to be updated by fix_paths(). This was not happening for spare
and cache devices because the proper names were getting filtered
out of the pool_list_t->names. Interestingly, the names were
being filtered because the spare and cache devices do not
contain the pool name in their vdev label.
The fix is to exclude the device path from the list only if:
1) has a valid ZPOOL_CONFIG_POOL_NAME key in the label, and
2) that pool name does not match the specified pool name.
Since the label is valid and because it does properly store the
vdev guid it will be correctly assembled without the pool name.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#725
The 'zpool replace' command would fail when given a short name
because unlike on other platforms the short name cannot be
deterministically expanded to a single path. Multiple path
prefixes must be checked and in addition the partition suffix
for whole disks is determined by the prefix.
To handle this complexity a zfs_strcmp_pathname() function was
added which takes either a short or fully qualified device name.
Short names will be expanded using the prefixes in the default
import search path, or the ZPOOL_IMPORT_PATH environment variable
if it's defined. All posible expansions are then compared against
the comparison path. Care is taken to strip redundant slashes to
ensure legitimate matches are not missed.
In the context of this work the existing zfs_resolve_shortname()
function was extended to consider the ZPOOL_IMPORT_PATH when set.
The zfs_append_partition() interface was also simplified to take
only a single buffer.
The vast majority of these changes rework existing Linux specific
code which was originally written to accomidate udev. However,
there is some minimal cleanup which removes Illumos specific code.
This was done to improve readability but the basic flow and intent
of the upstream code was maintained.
These changes are the logical conclusion of the previos work to
adjust the 'zpool import' search behavior, see commit 44867b6a.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#544Closes#976
When automounting a snapshot in the .zfs/snapshot directory
make sure to quote both the dataset name and the mount point.
This ensures that if either component contains spaces, which
are allowed, they get handled correctly.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1027
This brnach brings some ZIL performance optimizations, with
significant increases in synchronous write performance for
some workloads and pool configurations.
See the individual commit messages for details.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1013
In the current code, logbias=throughput implies the following:
1) All synchronous writes are logged in indirect mode.
2) The slog is not used.
(1) makes sense because it avoids writing the data twice, which is
obviously a good thing when the user wants maximum pool throughput.
(2), however, is a surprising decision. Considering all writes are
indirect, the log record doesn't contain the actual data, only pointers
to DMU blocks. As a result, log records written in logbias=throughput
mode are quite small, and as such, it doesn't make any sense to write
them to the main pool since slogs are usually optimized for small
synchronous writes.
In fact, the current behavior is actually harmful for performance,
because log blocks and data blocks from dmu_sync() seldom have the same
allocation size and as a result are usually allocated from different
metaslabs. This means that if a spindle has to write both log blocks and
DMU blocks (which is likely to happen under heavy load), it will have to
seek between the two. Allocating the log blocks from the slog pool
instead of the main pool avoids these unnecessary seeks.
This commit makes ZFS use the slog on datasets with logbias=throughput.
Real-life performance testing shows a 50% synchronous write performance
increase with some large commit sizes, and no negative effect in other
cases.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1013
Currently, ZIL blocks are spread over vdevs using hint block pointers
managed by the ZIL commit code and passed to metaslab_alloc(). Spreading
log blocks accross vdevs is important for performance: indeed, using
mutliple disks in parallel decreases the ZIL commit latency, which is
the main performance metric for synchronous writes. However, the current
implementation suffers from the following issues:
1) It would be best if the ZIL module was not aware of such low-level
details. They should be handled by the ZIO and metaslab modules;
2) Because the hint block pointer is managed per log, simultaneous
commits from multiple logs might use the same vdevs at the same time,
which is inefficient;
3) Because dmu_write() does not honor the block pointer hint, indirect
writes are not spread.
The naive solution of rotating the metaslab rotor each time a block is
allocated for the ZIL or dmu_sync() doesn't work in practice because the
first ZIL block to be written is actually allocated during the previous
commit. Consequently, when metaslab_alloc() decides the vdev for this
block, it will do so while a bunch of other allocations are happening at
the same time (from dmu_sync() and other ZILs). This means the vdev for
this block is chosen more or less at random. When the next commit
happens, there is a high chance (especially when the number of blocks
per commit is slightly less than the number of the disks) that one disk
will have to write two blocks (with a potential seek) while other disks
are sitting idle, which defeats spreading and increases the commit
latency.
This commit introduces a new concept in the metaslab allocator:
fastwrites. Basically, each top-level vdev maintains a counter
indicating the number of synchronous writes (from dmu_sync() and the
ZIL) which have been allocated but not yet completed. When the metaslab
is called with the FASTWRITE flag, it will choose the vdev with the
least amount of pending synchronous writes. If there are multiple vdevs
with the same value, the first matching vdev (starting from the rotor)
is used. Once metaslab_alloc() has decided which vdev the block is
allocated to, it updates the fastwrite counter for this vdev.
The rationale goes like this: when an allocation is done with
FASTWRITE, it "reserves" the vdev until the data is written. Until then,
all future allocations will naturally avoid this vdev, even after a full
rotation of the rotor. As a result, pending synchronous writes at a
given point in time will be nicely spread over all vdevs. This contrasts
with the previous algorithm, which is based on the implicit assumption
that blocks are written instantaneously after they're allocated.
metaslab_fastwrite_mark() and metaslab_fastwrite_unmark() are used to
manually increase or decrease fastwrite counters, respectively. They
should be used with caution, as there is no per-BP tracking of fastwrite
information, so leaks and "double-unmarks" are possible. There is,
however, an assert in the vdev teardown code which will fire if the
fastwrite counters are not zero when the pool is exported or the vdev
removed. Note that as stated above, marking is also done implictly by
metaslab_alloc().
ZIO also got a new FASTWRITE flag; when it is used, ZIO will pass it to
the metaslab when allocating (assuming ZIO does the allocation, which is
only true in the case of dmu_sync). This flag will also trigger an
unmark when zio_done() fires.
A side-effect of the new algorithm is that when a ZIL stops being used,
its last block can stay in the pending state (allocated but not yet
written) for a long time, polluting the fastwrite counters. To avoid
that, I've implemented a somewhat crude but working solution which
unmarks these pending blocks in zil_sync(), thus guaranteeing that
linguering fastwrites will get pruned at each sync event.
The best performance improvements are observed with pools using a large
number of top-level vdevs and heavy synchronous write workflows
(especially indirect writes and concurrent writes from multiple ZILs).
Real-life testing shows a 200% to 300% performance increase with
indirect writes and various commit sizes.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1013
Both the SPL and the ZFS libspl export most of the atomic_* functions,
except atomic_sub_* functions which are only exported by the SPL, not by
libspl. This patch remedies that by implementing atomic_sub_* functions
in libspl.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1013
Auditing the code to verify that all instances of cv_signal() and
cv_broadcast() are called under the proper associated mutex turned
up several races. None of these have been conclusively seen in the
wild but the following patch set resolves them.
For reference, from the cv_signal(9F) man page:
cv_signal() signals the condition and wakes one blocked thread.
All blocked threads can be unblocked by calling cv_broadcast().
You must acquire the mutex passed into cv_wait() before calling
cv_signal() or cv_broadcast()
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1048
The following incorrect usage of cv_broadcast() was caught by
code inspection. The cv_broadcast() function must be called
under the associated mutex to preventing racing with cv_wait().
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf
Prakash Surya
Jorgen Lundman
Massimo Maggi
Richard Yao
Ned Bass
Turbo Fredriksson
Cyril Plisko
nordaux
George Wilson
Etienne Dechamps