Tom Caputi [Fri, 19 May 2017 00:35:49 +0000 (20:35 -0400)]
Fixed small memory leak in ereport handling
One pre-check in zfs_ereport_start() was being called after
the nvlists were being allocated. This simply corrects that
issue.
Reviewed-by: Tony Hutter <hutter2@llnl.gov> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #6140
Brian Behlendorf [Thu, 18 May 2017 17:02:16 +0000 (10:02 -0700)]
Fix large dnode send stream flag conflict
Bit 21 of the send stream flags was inadvertently used for two
different features under concurrent development. To avoid any
future compatibility problems the large dnode flag is being
switched to bit 23 which is unused.
The large dnode feature has only been present in pre-releases of
ZoL and dnodesize defaults to legacy which is compatible with
existing OpenZFS implementations. Users with dnodesize=auto
needing to use zfs send/recv must update ZoL on both the
source and destination systems.
Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Ned Bass <bass6@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #6139
Boris Protopopov [Wed, 10 May 2017 17:51:29 +0000 (13:51 -0400)]
Introduce zv_state_lock
The lock is designed to protect internal state of zvol_state_t and
to avoid taking spa_namespace_lock (e.g. in dmu_objset_own() code path)
while holding zvol_stat_lock. Refactor the code accordingly.
Signed-off-by: Boris Protopopov <boris.protopopov@actifio.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #3484
Closes #6065
Closes #6134
Fix lock order inversion with zvol_open() as it did not account
for use of zvols as vdevs. The latter use cases resulted in the
lock order inversion deadlocks that involved spa_namespace_lock
and bdev->bd_mutex.
Signed-off-by: Boris Protopopov <boris.protopopov@actifio.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #6065
Issue #6134
Isaac Huang [Sat, 13 May 2017 00:28:03 +0000 (18:28 -0600)]
Skip spurious resilver IO on raidz vdev
On a raidz vdev, a block that does not span all child vdevs, excluding
its skip sectors if any, may not be affected by a child vdev outage or
failure. In such cases, the block does not need to be resilvered.
However, current resilver algorithm simply resilvers all blocks on a
degraded raidz vdev. Such spurious IO is not only wasteful, but also
adds the risk of overwriting good data.
This patch eliminates such spurious IOs.
Reviewed-by: Gvozden Neskovic <neskovic@gmail.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Matthew Ahrens <mahrens@delphix.com> Signed-off-by: Isaac Huang <he.huang@intel.com>
Closes #5316
Brian Behlendorf [Thu, 11 May 2017 21:27:57 +0000 (14:27 -0700)]
Enable additional test cases
Enable additional test cases, in most cases this required a few
minor modifications to the test scripts. In a few cases a real
bug was uncovered and fixed. And in a handful of cases where pools
are layered on pools the test case will be skipped until this is
supported. Details below for each test case.
* zpool_add_004_pos - Skip test on Linux until adding zvols to pools
is fully supported and deadlock free.
* zpool_add_005_pos.ksh - Skip dumpadm portion of the test which isn't
relevant for Linux. The find_vfstab_dev, find_mnttab_dev, and
save_dump_dev functions were updated accordingly for Linux. Add
O_EXCL to the in-use check to prevent the -f (force) option from
working for mounted filesystems and improve the resulting error.
* zpool_add_006_pos - Update test case such that it doesn't depend
on nested pools. Switch to truncate from mkfile to reduce space
requirements and speed up the test case.
* zpool_clear_001_pos - Speed up test case by filling filesystem to
25% capacity.
* zpool_create_002_pos, zpool_create_004_pos - Use sparse files for
file vdevs in order to avoid increasing the partition size.
* zpool_create_006_pos - 6ba1ce9 allows raidz+mirror configs with
similar redundancy. Updating the valid_args and forced_args cases.
* zpool_create_011_neg - Fix to correctly create the extra partition.
Modified zpool_vdev.c to use fstat64_blk() wrapper which includes
the st_size even for block devices.
* zpool_create_012_neg - Updated to properly find swap devices.
* zpool_create_014_neg, zpool_create_015_neg - Updated to use
swap_setup() and swap_cleanup() wrappers which do the right thing
on Linux and Illumos. Removed '-n' option which succeeds under
Linux due to differences in the in-use checks.
* zpool_create_016_pos.ksh - Skipped test case isn't useful.
* zpool_create_020_pos - Added missing / to cleanup() function.
Remove cache file prior to test to ensure a clean environment
and avoid false positives.
* zpool_destroy_001_pos - Removed test case which creates a pool on
a zvol. This is more likely to deadlock under Linux and has never
been completely supported on any platform.
* zpool_destroy_002_pos - 'zpool destroy -f' is unsupported on Linux.
Mount point must not be busy in order to unmount them.
* zfs_destroy_001_pos - Handle EBUSY error which can occur with
volumes when racing with udev.
* zpool_expand_001_pos, zpool_expand_003_neg - Skip test on Linux
until adding zvols to pools is fully supported and deadlock free.
The test could be modified to use loop-back devices but it would
be preferable to use the test case as is for improved coverage.
* zpool_export_004_pos - Updated test case to such that it doesn't
depend on nested pools. Normal file vdev under /var/tmp are fine.
* zpool_import_all_001_pos - Updated to skip partition 1, which is
known as slice 2, on Illumos. This prevents overwriting the
default TESTPOOL which was causing the failure.
* zpool_import_002_pos, zpool_import_012_pos - No changes needed.
* zpool_remove_003_pos - No changes needed
* zpool_upgrade_002_pos, zpool_upgrade_004_pos - Root cause addressed
by upstream OpenZFS commit 3b7f360.
* zpool_upgrade_007_pos - Disabled in test case due to known failure.
Opened issue https://github.com/zfsonlinux/zfs/issues/6112
* zvol_misc_002_pos - Updated to to use ext2.
* zvol_misc_001_neg, zvol_misc_003_neg, zvol_misc_004_pos,
zvol_misc_005_neg, zvol_misc_006_pos - Moved to skip list, these
test case could be updated to use Linux's crash dump facility.
* zvol_swap_* - Updated to use swap_setup/swap_cleanup helpers.
File creation switched from /tmp to /var/tmp. Enabled minimal
useful tests for Linux, skip test cases which aren't applicable.
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #3484
Issue #5634
Issue #2437
Issue #5202
Issue #4034
Closes #6095
Matthew Ahrens [Mon, 24 Apr 2017 16:34:36 +0000 (09:34 -0700)]
OpenZFS 8063 - verify that we do not attempt to access inactive txg
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Serapheim Dimitropoulos <serapheim@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: George Melikov <mail@gmelikov.ru>
A standard practice in ZFS is to keep track of "per-txg" state. Any of
the 3 active TXG's (open, quiescing, syncing) can have different values
for this state. We should assert that we do not attempt to modify other
(inactive) TXG's.
Porting Notes:
- ASSERTV added to txg_sync_waiting() for unused variable.
Matthew Ahrens [Wed, 10 May 2017 17:32:40 +0000 (10:32 -0700)]
OpenZFS 8166 - zpool scrub thinks it repaired offline device
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: Matthew Ahrens <mahrens@delphix.com>
If we do a scrub while a leaf device is offline (via "zpool offline"),
we will inadvertently clear the DTL (dirty time log) of the offline
device, even though it is still damaged. When the device comes back
online, we will incompletely resilver it, thinking that the scrub
repaired blocks written before the scrub was started. The incomplete
resilver can lead to data loss if there is a subsequent failure of a
different leaf device.
The fix is to never clear the DTL of offline devices. Note that if a
device is onlined while a scrub is in progress, the scrub will be
restarted.
The problem can be worked around by running "zpool scrub" after
"zpool online".
Tom Caputi [Wed, 10 May 2017 17:25:27 +0000 (13:25 -0400)]
Add missing arc_free_cksum() to arc_release()
The arc layer tracks checksums of its data in the arc header
so that it can ensure that buffers haven't changed when they're
not supposed to. This checksum is only maintained while there
is an uncompressed buffer still attached to the header.
Unfortunately there is a missing call to arc_free_cksum() in
arc_release() that can trigger ASSERTs. This has not been a
common issue because the checksums are only maintained for
debug builds and triggering the bug requires writing a block
(and therefore calling arc_release()) while a compressed buffer
is still being used on a debug build. This simply corrects the
issue.
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tom Caputi <tcaputi@datto.com>
Closes #6105
LOLi [Tue, 9 May 2017 23:21:09 +0000 (01:21 +0200)]
Add property overriding (-o|-x) to 'zfs receive'
This allows users to specify "-o property=value" to override and
"-x property" to exclude properties when receiving a zfs send stream.
Both native and user properties can be specified.
This is useful when using zfs send/receive for periodic
backup/replication because it lets users change properties such as
canmount, mountpoint, or compression without modifying the source.
Document the existence of `createtxg` and `guid` native properties
in man pages and zfs command output.
One of the great features of ZFS is incremental replication of
snapshots, possibly between pools on different machines.
Shell scripts are commonly used to auomate this procedure. They have to
find the most recent common snapshot between both sides and then
perform incremental send & recv.
Currently, scripts rely on the sorting order of `zfs list`, which
defaults to `createtxg`, and the assumption that snapshot names on
either side do not change.
By making `createtxg` and `guid` part of the public ZFS interface,
scripts are enabled to use
a) `createtxg` to determine the logical & temporal order of snapshots
(the creation property is not an equivalent substitute since
multiple snapshots may be created within one second)
b) `guid` to uniquely identify a snapshot, independent of its current
display name
This has the potential of making scripts safer and correct.
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: DHE <git@dehacked.net> Reviewed-by: Richard Laager <rlaager@wiktel.com> Signed-off-by: Christian Schwarz <me@cschwarz.com>
Closes #6102
LOLi [Tue, 9 May 2017 22:22:46 +0000 (00:22 +0200)]
Fix NULL pointer dereference in 'zfs create'
A race condition between 'zpool export' and 'zfs create' can crash the
latter: this is because we never check libzfs`zpool_open() return
value in libzfs`zfs_create().
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
Closes #6096
LOLi [Tue, 9 May 2017 18:51:40 +0000 (20:51 +0200)]
Fix zfs .deb package warning in prerm script
Debian zfs package generated by alien doesn't call the prerm script
(rpm's %preun) with an integer as first parameter, which results in
the following warning:
"zfs.prerm: line 2: [: remove: integer expression expected"
Modify the if-condition to avoid the warning.
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
Closes #6108
Richard Yao [Fri, 11 Jul 2014 18:35:58 +0000 (14:35 -0400)]
Enable Linux read-ahead for a single page on ZVOLs
Linux has read-ahead logic designed to accelerate sequential workloads.
ZFS has its own read-ahead logic called zprefetch that operates on both
ZVOLs and datasets. Having two prefetchers active at the same time can
cause overprefetching, which unnecessarily reduces IOPS performance on
CoW filesystems like ZFS.
Testing shows that entirely disabling the Linux prefetch results in
a significant performance penalty for reads while commensurate benefits
are seen in random writes. It appears that read-ahead benefits are
inversely proportional to random write benefits, and so a single page
of Linux-layer read-ahead appears to offer the middle ground for both
workloads.
Reviewed-by: Chunwei Chen <david.chen@osnexus.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Richard Yao <ryao@gentoo.org>
Issue #5902
RageLtMan [Sat, 18 Mar 2017 04:51:36 +0000 (00:51 -0400)]
Disable write merging on ZVOLs
The current ZVOL implementation does not explicitly set merge
options on ZVOL device queues, which results in the default merge
behavior.
Explicitly set QUEUE_FLAG_NOMERGES on ZVOL queues allowing the
ZIO pipeline to do its work.
Initial benchmarks (tiotest with no O_DIRECT) show random write
performance going up almost 3X on 8K ZVOLs, even after significant
rewrites of the logical space allocation.
Reviewed-by: Richard Yao <ryao@gentoo.org> Reviewed-by: Chunwei Chen <david.chen@osnexus.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: RageLtMan <rageltman@sempervictus>
Issue #5902
The send-c_volume test case has been observed to occasionally
fail on 32-bit systems. Until this issue is fully understood
disable this test case.
The rsend_014_pos test case can occasionally fail due to an
EBUSY during export. This can lead to subsequent test failures.
Resolve the issue by retrying the export on EBUSY. Additionally,
remove the gratuitous use of eval.
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #6088
* zfs_destroy_001_pos - Unable to reproduce the failures locally.
Re-enabled to determine observed buildbot failure rate.
* zfs_destroy_005_neg - Updated for expected Linux behavior.
Busy mount points, even snapshots, are expected to fail.
* zfs_destroy_010_pos - Resolved transient EBUSY with retry.
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #5635
Issue #5893
Closes #6091
LOLi [Wed, 3 May 2017 16:31:05 +0000 (18:31 +0200)]
More ashift improvements
This commit allow higher ashift values (up to 16) in 'zpool create'
The ashift value was previously limited to 13 (8K block) in b41c990
because the limited number of uberblocks we could fit in the
statically sized (128K) vdev label ring buffer could prevent the
ability the safely roll back a pool to recover it.
Since b02fe35 the largest uberblock size we support is 8K: this
allow us to store a minimum number of 16 uberblocks in the vdev
label, even with higher ashift values.
Additionally change 'ashift' pool property behaviour: if set it will
be used as the default hint value in subsequent vdev operations
('zpool add', 'attach' and 'replace'). A custom ashift value can still
be specified from the command line, if desired.
Finally, fix a bug in add-o_ashift.ksh caused by a missing variable.
Olaf Faaland [Tue, 2 May 2017 20:55:24 +0000 (13:55 -0700)]
Write label 2,3 uberblocks when vdev expands
When vdev_psize increases, the location of labels 2 and 3 changes
because their location is relative to the end of the device.
The configs for labels 2 and 3 are written during the next spa_sync()
because the vdev is added to the dirty config list. However, the
uberblock rings are not re-written in their new location, leaving the
device vulnerable to the beginning of the device being overwritten or
damaged.
This patch copies the uberblock ring from label 0 to labels 2 and 3,
in their new locations, at the next sync after vdev_psize increases.
Also, add a test zpool_expand_004_pos.ksh to confirm the uberblocks
are copied.
Reviewed-by: BearBabyLiu <liu.huang@zte.com.cn> Reviewed-by: Andreas Dilger <andreas.dilger@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov>
Closes #5108
LOLi [Tue, 2 May 2017 20:43:53 +0000 (22:43 +0200)]
Add zfs_nicebytes() to print human-readable sizes
* Add zfs_nicebytes() to print human-readable sizes
Some 'zfs', 'zpool' and 'zdb' output strings can be confusing to the
user when no units are specified. This add a new zfs_nicenum_format
"ZFS_NICENUM_BYTES" used to print bytes in their human-readable form.
Additionally, update some test cases to use machine-parsable 'zfs get'.
When multiple filesystems are in use, memory pressure causes arc_cache
to collapse to a minimum. Allow arc_cache to maintain proportional size
even when hit rates are disproportionate. We do this only via evictable
size from the kernel shrinker, thus it's only in effect under memory
pressure.
AKAMAI: zfs: CR 3695072 Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Richard Yao <ryao@gentoo.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Debabrata Banerjee <dbanerje@akamai.com>
Closes #6035
Don't run the reaper if we didn't shrink the cache
Calling it when nothing is evictable will cause extra kswapd cpu. Also
if we didn't shrink it's unlikely to have memory to reap because we
likely just called it microseconds ago. The exception is if we are in
direct reclaim.
You can see how hard this is being hit in kswapd with a light test
workload:
Lock contention, by itself, shouldn't indicate a stop condition to the
kernel's slab shrinker. Doing so can cause stalls when the kernel is
trying to free large parts of the cache such as is done by drop_caches
Also, perhaps arc_reclaim_lock should be a spinlock, and this code
eliminated.
AKAMAI: zfs: CR 3593801 Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Richard Yao <ryao@gentoo.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Debabrata Banerjee <dbanerje@akamai.com>
Issue #6035
Move arcstat_need_free increment from all direct calls to when
arc_reclaim_lock is busy and we exit wihout doing anything. Data will
be reclaimed in reclaim thread. The previous location meant that we
both reclaim the memory in this thread, and also schedule the same
amount of memory for reclaim in arc_reclaim, effectively doubling the
requested reclaim.
AKAMAI: zfs: CR 3695072 Reviewed-by: Tim Chase <tim@chase2k.com> Reviewed-by: Richard Yao <ryao@gentoo.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Debabrata Banerjee <dbanerje@akamai.com>
Issue #6035
jxiong [Tue, 2 May 2017 17:06:18 +0000 (10:06 -0700)]
minor improvement to abd_free_pages()
It doesn't need to have a loop to free page in a single scatterlist
entry because it should be single or compound page. The pages can be
freed in one invocation to __free_pages() for both cases.
Reviewed-by: Gvozden Neskovic <neskovic@gmail.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Jinshan Xiong <jinshan.xiong@gmail.com>
Closes #6057
jxiong [Tue, 2 May 2017 17:04:30 +0000 (10:04 -0700)]
Guarantee PAGESIZE alignment for large zio buffers
In current implementation, only zio buffers in 16KB and bigger are
guaranteed PAGESIZE alignment. This breaks Lustre since it assumes
that 'arc_buf_t::b_data' must be page aligned when zio buffers are
greater than or equal to PAGESIZE.
This patch will make the zio buffers to be PAGESIZE aligned when
the sizes are not less than PAGESIZE.
This change may cause a little bit memory waste but that should be
fine because after ABD is introduced, zio buffers are used to hold
data temporarily and live in memory for a short while.
Reviewed-by: Don Brady <don.brady@intel.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Jinshan Xiong <jinshan.xiong@gmail.com> Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Closes #6084
All filesystems were converted to dynamically allocated BDIs. The
destruction of backing_dev_info structures is handled as part of
super block destruction. Refactor the code to abstract away the
details of creating and destroying a BDI.
Reviewed-by: Chunwei Chen <david.chen@osnexus.com> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #6089
OpenZFS 7786 - zfs`vdev_online() needs better notification about state changes
Authored by: Yuri Pankov <yuri.pankov@nexenta.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Albert Lee <trisk@forkgnu.org> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: bunder2015 <omfgbunder@gmail.com>
OpenZFS-issue: https://www.illumos.org/issues/7786
OpenZFS-commit: http://github.com/openzfs/openzfs/commit/db8498f
Closes #6074
Chunwei Chen [Thu, 23 Feb 2017 00:08:04 +0000 (16:08 -0800)]
Reinstate zvol_taskq to fix aio on zvol
Commit 37f9dac removed the zvol_taskq for processing zvol requests.
This was removed as part of switching to make_request_fn and was
motivated by a concern at the time over dispatch latency.
However, this also made all bio request synchronous, and caused
serious performance issues as the bio submitter would wait for
every bio it submitted, effectively making the IO depth 1.
This patch reinstate zvol_taskq, and to make sure overlapped I/Os
are ordered properly, we take range lock in zvol_request, and pass
it along with bio to the I/O functions zvol_{write,discard,read}.
In order to facilitate benchmarks a zvol_request_sync module
option was added to switch between sync and async request handling.
For the moment, the default behavior is synchronous but this is
likely to change pending additional testing.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Closes #5824
Tim Chase [Tue, 25 Apr 2017 04:01:04 +0000 (23:01 -0500)]
Update documentation for zfs_vdev_queue_depth_pct
It was documented as being related to zfs_vdev_async_max_active
when it is actually related to zfs_vdev_async_write_max_active.
Also, expand the documentation to describe the allocation throttle
which was introduced as part of OpenZFS 7090 in 3dfb57a.
Reviewed-by: Richard Yao <ryao@gentoo.org> Reviewed-by: Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com>
Closes #6064
Dan Kimmel [Tue, 11 Apr 2017 21:56:54 +0000 (21:56 +0000)]
OpenZFS 7252 - compressed zfs send / receive
OpenZFS 7252 - compressed zfs send / receive
OpenZFS 7628 - create long versions of ZFS send / receive options
Authored by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: Sebastien Roy <sebastien.roy@delphix.com>
Reviewed by: David Quigley <dpquigl@davequigley.com>
Reviewed by: Thomas Caputi <tcaputi@datto.com>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: David Quigley <dpquigl@davequigley.com> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Ported-by: bunder2015 <omfgbunder@gmail.com> Ported-by: Don Brady <don.brady@intel.com> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- Most of 7252 was already picked up during ABD work. This
commit represents the gap from the final commit to openzfs.
- Fixed split_large_blocks check in do_dump()
- An alternate version of the write_compressible() function was
implemented for Linux which does not depend on fio. The behavior
of fio differs significantly based on the exact version.
- mkholes was replaced with truncate for Linux.
After run a long time with QAT compression, the variable "inst_num"
is overflow by "atomic_inc_32_nv", which causes its neighbor
variable overwritten. Change its definition from U16 to U32.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Weigang Li <weigang.li@intel.com>
Closes #6051
Matthew Ahrens [Thu, 13 Apr 2017 21:35:00 +0000 (14:35 -0700)]
OpenZFS 8025 - dbuf_read() creates unnecessary zio_root() for bonus buf
Authored by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: Prashanth Sreenivasa <pks@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
dbuf_read() creates a zio_root() to track and wait for all the zio's
that may happen as part of this call. However, if the blkptr_t for
this buffer is NULL or a hole, we will not create any more zio's, so
this zio_root() is unnecessary. This is always the case when calling
dbuf_read() on a bonus buffer, because it has no blkptr (it's part of
the containing dnode). For workloads that read a lot of bonus buffers
(e.g. file creation and removal), creating and destroying these
unnecessary zio's can decrease performance by around 3%.
The fix is to only create/destroy the zio_root() in dbuf_read() if the
blkptr is not NULL and not a hole.
Porting Notes:
- The error handling for when dbuf_read_impl() fails which was
originally added in commit 5f6d0b6f5 has been preserved.
Don Brady [Mon, 24 Apr 2017 17:31:45 +0000 (11:31 -0600)]
Fixed zdb -e regression for active cacheless pools
zdb -e for active cache-less pools fails:
$ sudo zpool create -o cachefile=none basic mirror sdk sdl
$ sudo zdb -e -b basic
zdb: can't open 'basic': No such file or directory
This is a recent regression introduce by commit c30d8de.
Reviewed-by: Richard Yao <ryao@gentoo.org> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Don Brady <don.brady@intel.com>
Closes #6059
Olaf Faaland [Fri, 21 Apr 2017 21:26:43 +0000 (14:26 -0700)]
Correct lock ASSERTs in vdev_label_read/write
The existing assertions in vdev_label_read() and vdev_label_write(),
testing which config locks are held, are incorrect. The assertions
test for locks which exceed what is required for safety.
Both vdev_label_{read,write}() are changed to assert SCL_STATE is held
as RW_READER or RW_WRITER. This is safe because:
Changes to the vdev tree occur under SCL_ALL as RW_WRITER, via
spa_vdev_enter() and spa_vdev_exit().
Changes to vdev state occur under SCL_STATE_ALL as RW_WRITER, via
spa_vdev_state_enter() and spa_vdev_state_exit().
Therefore, the new assertions guarantee that the vdev cannot change
out from under a zio, and I/O to a specified leaf vdev's label is
safe.
Furthermore, this is consistent with the SPA locking discussion in
spa_misc.c, "For any zio operation that takes an explicit vdev_t
argument ... zio_read_phys(), or zio_write_phys() ... SCL_STATE as
reader suffices."
Reviewed-by: Chunwei Chen <david.chen@osnexus.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Olaf Faaland <faaland1@llnl.gov>
Closes #5983
Tony Hutter [Fri, 21 Apr 2017 16:27:04 +0000 (09:27 -0700)]
Prebaked scripts for zpool status/iostat -c
This patch updates the "zpool status/iostat -c" commands to only run
"pre-baked" scripts from the /etc/zfs/zpool.d directory (or wherever
you install to). The scripts can only be run from -c as an unprivileged
user (unless the ZPOOL_SCRIPTS_AS_ROOT environment var is
set by root). This was done to encourage scripts to be written is such
a way that normal users can use them, and to be cautious. If your
script needs to run a privileged command, consider adding the
appropriate line in /etc/sudoers. See zpool(8) for an example of how
to do this.
The patch also allows the scripts to output custom column names. If
the script outputs a line like:
name=value
then "name" is used for the column name, and "value" is its value.
Multiple columns can be specified by outputting multiple lines. Column
names and values can have spaces. If the value is empty, a dash (-) is
printed instead.
After all the "name=value" lines are read (if any), zpool will take the
next the next line of output (if any) and print it without a column
header. After that, no more lines will be processed. This can be
useful for printing errors.
Lastly, this patch also disables the -c option with the latency and
request size histograms, since it produced awkward output and made the
code harder to maintain.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Closes #5852
Ned Bass [Thu, 20 Apr 2017 19:10:55 +0000 (12:10 -0700)]
vdev_id: fix failure due to multipath -l bug
Udev may fail to create the expected symbolic links in
/dev/disk/by-vdev on systems with the
device-mapper-multipath-0.4.9-100.el6 package installed. This affects
RHEL 6.9 and possibly other downstream distributions.
That version of the multipath command may incorrectly list a drive
state as "unkown" instead of "running". The issue was introduced
in the patch for https://bugzilla.redhat.com/show_bug.cgi?id=1401769
The vdev_id udev helper uses the state reported by "multipath -l" to
detect an online component disk of a multipath device in order to
resolve its physical slot and enclosure. Changing the command
invocation to "multipath -ll" works around the above issue by causing
multipath to consult additional sources of information to determine
the drive state.
Reviewed-by: Tony Hutter <hutter2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: Ned Bass <bass6@llnl.gov>
Closes #6039
This lets users create a bookmark from the command line by its name
only, without the need to specify the dataset path which is extacted
from the snapshot parameter.
George Melikov [Thu, 20 Apr 2017 19:05:39 +0000 (23:05 +0400)]
zfs_receive_010_pos: change dd arguments
The `dd` command as written will not create a hole in the file.
Additionally, the `stride` argument isn't understood by `dd` so
it's replaced with `seek` which isn't equivilant but will result in
a single whole which is sufficient for the test case. Finally,
`conv=notrunc` is added to avoid truncating the file.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes #6023
Tim Crawford [Wed, 19 Apr 2017 23:36:32 +0000 (19:36 -0400)]
Fix leak in send_iterate_fs
Fix a leak when generating a replication stream of a cloned dataset.
Reviewed-by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: Tim Crawford <tcrawford@datto.com>
Closes #6034
Richard Yao [Thu, 13 Apr 2017 21:28:46 +0000 (14:28 -0700)]
OpenZFS 6392 - zdb: introduce -V for verbatim import
Authored by: Richard Yao <ryao@gentoo.org>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Yuri Pankov <yuri.pankov@gmail.com>
Reviewed by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com> Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Porting Notes:
This was already implemented in ZFS on Linux. This patch
is to resolved the deltas present in our version.
Ensure `zinject -c` all gets called whenever
zpool_scrub_004_pos exits.
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Issue #5444
Closes #6021
George Melikov [Tue, 18 Apr 2017 16:44:17 +0000 (20:44 +0400)]
zfstest: add dmesg command to $PATH
Error example in `zfs_list_007_pos`:
`sudo: dmesg: command not found`
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes #6024
DHE [Sun, 26 Mar 2017 02:36:28 +0000 (22:36 -0400)]
Increase zfs_vdev_async_write_min_active to 2
Resilver operations frequently cause only a small amount of dirty data
to be written to disk at a time, resulting in the IO scheduler to only
issue 1 write at a time to the resilvering disk. When it is rotational
media the drive will often travel past the next sector to be written
before receiving a write command from ZFS, significantly delaying the
write of the next sector.
Raise zfs_vdev_async_write_min_active so that drives are kept fed
during resilvering.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: DHE <git@dehacked.net>
Issue #4825
Closes #5926
Matthew Ahrens [Thu, 13 Apr 2017 21:38:16 +0000 (14:38 -0700)]
OpenZFS 8061 - sa_find_idx_tab can be declared more type-safely
Authored by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Chris Williamson <chris.williamson@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
sa_find_idx_tab() is declared as taking and returning "void *" parameters.
These can be declared to be the specific types.
Alan Somers [Thu, 13 Apr 2017 21:22:32 +0000 (14:22 -0700)]
OpenZFS 7900 - zdb shouldn't print the path of a znode at verbosity < 5
Authored by: Alan Somers <asomers@gmail.com>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
There are two reasons:
1) Finding a znode's path is slower than printing any other znode
information at verbosity < 5.
2) On a corrupted pool like the one mentioned below, zdb will crash when it
tries to determine the znode's path. But with this patch, zdb can still
extract useful information from such pools.
OpenZFS 6101 - attempt to lzc_create() a filesystem under a volume results in a panic
Authored by: Andriy Gapon <avg@FreeBSD.org>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
When querying ZPL properties verify that the objset is of type
DMU_OST_ZFS.
Force flushing of txg's can be painfully slow when competing for disk
IO, since this is a process meant to execute asynchronously. Optimize
this path via allowing data/hole seeking if the file is clean, but if
dirty fall back to old logic. This is a compromise to disabling the
feature entirely.
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Debabrata Banerjee <dbanerje@akamai.com>
Closes #4306
Closes #5962
Brian Behlendorf [Thu, 13 Apr 2017 16:40:56 +0000 (09:40 -0700)]
OpenZFS 6410 - teach zdb to perform object lookups by path
Authored by: Yuri Pankov <yuri.pankov@nexenta.com>
Reviewed by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Will Andrews <will@freebsd.org>
Approved by: Dan McDonald <danmcd@omniti.com> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- Replaced zdb.8 with upstream mdoc zdb.1m version. Updated to
include Linux specific features: -V verbatium imports and
improved label printing (-u, and -l).
- Minor changes to `zdb -h` output to honor 80 character limit.
Brian Behlendorf [Thu, 13 Apr 2017 16:40:00 +0000 (09:40 -0700)]
OpenZFS 5120 - zfs should allow large block/gzip/raidz boot pool (loader project)
Authored by: Toomas Soome <tsoome@me.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Yuri Pankov <yuri.pankov@nexenta.com>
Reviewed by: Andrew Stormont <andyjstormont@gmail.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Don Brady <don.brady@intel.com> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- grub-2.02-beta2-422-gcad5cc0 includes support for large blocks.
- Commit 8aab121 allowed GZIP[1-9].
- Grub allows pools with multiple top-level vdevs.
Be sure to invalidate a vdev's cache before performing
a zpool labelclear. There are cases where the cache is
stale because we did some operation that bypassed it,
and since we are doing an open with only O_RDWR, we
should invalidate it to be safe.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Closes #6009
Brian Behlendorf [Wed, 12 Apr 2017 20:36:48 +0000 (13:36 -0700)]
OpenZFS 7503 - zfs-test should tail ::zfs_dbgmsg on test failure
Authored by: Pavel Zakharov <pavel.zakharov@delphix.com>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Approved by: Gordon Ross <gordon.w.ross@gmail.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- Enable internal log for DEBUG builds and in zfs-tests.sh.
- callbacks/zfs_dbgmsg.ksh - Dump interal log via kstat.
- callbacks/zfs_dmesg.ksh - Dump dmesg log.
- default.cfg - 'Test Suite Specific Commands' dropped.
Richard Yao [Sat, 8 Apr 2017 16:51:04 +0000 (12:51 -0400)]
Fix header inclusions for standards conformance
musl's sys/errno.h is literally:
/#warning redirecting incorrect #include <sys/errno.h> to <errno.h>
/#include <errno.h>
It does the same for sys/{poll,signal}.h. This is rather noisy when
building ZoL against musl. musl is also correct in pointing out that the
correct headers are outside of sys/ according to the single unix
specification:
Lets implement our own sys/* versions of these headers to redirect to
the proper userland ones when building in userspace. That will silence
the warning.
There are also some instances where we include incorrectly from sys/ or
from outside of sys/ in userspace only code. In these instances, lets
just fix the includes directly.
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes #5993
Richard Yao [Sun, 9 Apr 2017 19:00:03 +0000 (15:00 -0400)]
Fix `zpool iostat -T d 1` on musl
When building on Gentoo against musl, GCC complains:
timestamp.c: In function ‘print_timestamp’:
timestamp.c:32:19: warning: passing argument 1 of ‘nl_langinfo’ makes
integer from pointer without a cast
#define _DATE_FMT "%+"
^
timestamp.c:47:21: note: in expansion of macro ‘_DATE_FMT’
fmt = nl_langinfo(_DATE_FMT);
^
The error was wrapped to meet comment style requirements.
This code is used by `zpool iostat -T d 1` to print a date and upon
testing it, I see no date printed. Lets use D_T_FMT so that something
gets printed and if D_T_FMT is not avaliable, then we can fall back to
"%+".
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes #5993
Richard Yao [Sat, 8 Apr 2017 17:14:14 +0000 (13:14 -0400)]
Add missing includes to zed_log.c
GCC 4.9.4 complains about implicit function declarations when building
against musl on Gentoo.
zed_log.c: In function ‘zed_log_pipe_open’:
zed_log.c:69:7: warning: implicit declaration of function ‘getpid’
(int)getpid());
^
zed_log.c:71:2: warning: implicit declaration of function ‘pipe’
if (pipe(_ctx.pipe_fd) < 0)
^
zed_log.c: In function ‘zed_log_pipe_close_reads’:
zed_log.c:90:2: warning: implicit declaration of function ‘close’
if (close(_ctx.pipe_fd[0]) < 0)
^
zed_log.c: In function ‘zed_log_pipe_wait’:
zed_log.c:141:3: warning: implicit declaration of function ‘read’
n = read(_ctx.pipe_fd[0], &c, sizeof (c));
The [-Wimplicit-function-declaration] at the end of each warning has
been removed to meet comment style requirements.
The man pages say to include <sys/types.h> and <unistd.h>. Doing that
silences the warnings.
Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes #5993
Brian Behlendorf [Wed, 12 Apr 2017 15:47:42 +0000 (08:47 -0700)]
OpenZFS 7535 - need test for resumed send of top most filesystem
Authored by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Richard Lowe <richlowe@richlowe.net> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- zfs_share_001_pos.ksh - Older versions of exportfs will match
multiple exports that share a common prefix. Reorder the 'fs'
list so unshares occur from most to least unique.
- zfs_share_005_pos.ksh - Enabled and updated for Linux.
Yuri Pankov [Fri, 13 Jan 2017 17:25:15 +0000 (09:25 -0800)]
OpenZFS 6865 - want zfs-tests cases for zpool labelclear command
Authored by: Yuri Pankov <yuri.pankov@nexenta.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com> Reviewed-by: loli10K <ezomori.nozomu@gmail.com> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- Updated 'zpool labelclear' and 'zdb -l' such that they attempt
to find a vdev given solely its short name. This behavior is
consistent with the upstream OpenZFS code and the test cases
depend on it. The actual implementation differs slightly due
to device naming conventions on Linux.
- auto_online_001_pos, auto_replace_001_pos and add-o_ashift
test cases updated to expect failure when no label exists.
- read_efi_label() and zpool_label_disk_check() are read-only
operations and should use O_RDONLY at open time to enforce this.
- zpool_label_disk() and zpool_relabel_disk() write the partition
information using O_DIRECT an fsync() and page cache invalidation
to ensure a consistent view of the device.
- dump_label() in zdb should invalidate the page cache in order
to get the authoritative label from disk.
When we try assign a new transaction to a TXG we must know beforehand
if there is sufficient free space on disk. This is to decide,
in dmu_tx_assign(), if we should reject the TX with ENOSPC.
We rely on spa_get_worst_case_asize() to inflate the size of our
logical writes by a factor of spa_asize_inflation which is
calculated as:
The problem with the current implementation is that we don't take
into account what happens with very small writes on VDEVs with large
physical block sizes.
Consider the case of writes to a dataset with recordsize=512,
copies=3 on a VDEV with ashift=13 (usually SSD with 8K block size):
every logical IO will end up allocating 3 * 8K = 24K on disk, so 512
bytes multiplied by 48, which is double the size we account for.
If we allow this kind of writes to be assigned a TX it is possible,
when the pool is almost full, to trigger an allocation failure
(ENOSPC) in the ZIO pipeline, which will in turn result in the whole
pool being suspended.
The bug is fixed by using, in spa_get_worst_case_asize(), the MAX()
value chosen between the logical io size from zfs_write() and the
maximum physical block size used among our VDEVs.
Reviewed by: Matthew Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: loli10K <ezomori.nozomu@gmail.com>
Closes #5941
Matthew Ahrens [Mon, 10 Apr 2017 22:21:45 +0000 (15:21 -0700)]
OpenZFS 8005 - poor performance of 1MB writes on certain RAID-Z configurations
Authored by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: Saso Kiselkov <saso.kiselkov@nexenta.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Don Brady <don.brady@intel.com> Ported-by: Matt Ahrens <mahrens@delphix.com>
RAID-Z requires that space be allocated in multiples of P+1 sectors,
because this is the minimum size block that can have the required amount
of parity. Thus blocks on RAIDZ1 must be allocated in a multiple of 2
sectors; on RAIDZ2 multiple of 3; and on RAIDZ3 multiple of 4. A sector
is a unit of 2^ashift bytes, typically 512B or 4KB.
To satisfy this constraint, the allocation size is rounded up to the
proper multiple, resulting in up to 3 "pad sectors" at the end of some
blocks. The contents of these pad sectors are not used, so we do not
need to read or write these sectors. However, some storage hardware
performs much worse (around 1/2 as fast) on mostly-contiguous writes
when there are small gaps of non-overwritten data between the writes.
Therefore, ZFS creates "optional" zio's when writing RAID-Z blocks that
include pad sectors. If writing a pad sector will fill the gap between
two (required) writes, we will issue the optional zio, thus doubling
performance. The gap-filling performance improvement was introduced in
July 2009.
Writing the optional zio is done by the io aggregation code in
vdev_queue.c. The problem is that it is also subject to the limit on
the size of aggregate writes, zfs_vdev_aggregation_limit, which is by
default 128KB. For a given block, if the amount of data plus padding
written to a leaf device exceeds zfs_vdev_aggregation_limit, the
optional zio will not be written, resulting in a ~2x performance
degradation.
The problem occurs only for certain values of ashift, compressed block
size, and RAID-Z configuration (number of parity and data disks). It
cannot occur with the default recordsize=128KB. If compression is
enabled, all configurations with recordsize=1MB or larger will be
impacted to some degree.
The problem notably occurs with recordsize=1MB, compression=off, with 10
disks in a RAIDZ2 or RAIDZ3 group (with 512B or 4KB sectors). Therefore
this problem has been known as "the 1MB 10-wide RAIDZ2 (or 3) problem".
The problem also occurs with the following configurations:
With recordsize=512KB or 256KB, compression=off, the problem occurs only
in rarely-used configurations:
* 4-wide RAIDZ1 with recordsize=512KB and ashift=12 (4KB sectors)
* 4-wide RAIDZ2 (either recordsize, either ashift)
* 5-wide RAIDZ2 with recordsize=512KB (either ashift)
* 6-wide RAIDZ2 with recordsize=512KB (either ashift)
With recordsize=1MB, compression=off, ashift=9 (512B sectors)
* RAIDZ1 with 4 or 8 disks
* RAIDZ2 with 4, 8, or 10 disks
* RAIDZ3 with 6, 8, 9, or 10 disks
With recordsize=1MB, compression=off, ashift=12 (4KB sectors)
* RAIDZ1 with 7 or 8 disks
* RAIDZ2 with 4, 5, or 10 disks
* RAIDZ3 with 6, 9, or 10 disks
With recordsize=2MB and larger (which can only be selected by changing
kernel tunables), many configurations are affected, including with
higher numbers of disks (up to 18 disks with recordsize=2MB).
Increase zfs_vdev_aggregation_limit to allow the optional zio to be
aggregated, thus eliminating the problem. Setting it to 256KB fixes all
commonly-used configurations.
The solution is to aggregate optional zio's regardless of the
aggregation size limit.
George Melikov [Sun, 9 Apr 2017 23:17:55 +0000 (03:17 +0400)]
zfstest - replace dircmp with diff
`dircmp` doesn't exist in Linux while `diff` is already used
by zfstests on all platforms.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: John Wren Kennedy <john.kennedy@delphix.com> Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes #5996
George Melikov [Sun, 9 Apr 2017 23:15:44 +0000 (03:15 +0400)]
zfstest reservation_009_pos.sh missed backslash
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: John Wren Kennedy <john.kennedy@delphix.com> Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes #5997
George Wilson [Fri, 7 Apr 2017 20:50:18 +0000 (13:50 -0700)]
OpenZFS 8023 - Panic destroying a metaslab deferred range tree
Authored by: George Wilson <george.wilson@delphix.com>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Brad Lewis <brad.lewis@delphix.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: Saso Kiselkov <saso.kiselkov@nexenta.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
We don't want to dirty any data when we're in the final txgs of the pool
export logic. This change introduces checks to make sure that no data is
dirtied after a certain point. It also addresses the culprit of this
specific bug – the space map cannot be upgraded when we're in final
stages of pool export. If we encounter a space map that wants to be
upgraded in this phase, then we simply ignore the request as it will get
retried the next time we set the fragmentation metric on that metaslab.
OpenZFS 5380 - receive of a send -p stream doesn't need to try renaming snapshots
Authored by: Andriy Gapon <avg@FreeBSD.org>
Approved by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
recv_incremental_replication() takes care of things like removing
datasets that have been removed on the sending side, detecting renamed
datasets, ensuring that all datasets in the affected hierarchy have the
same properties as their counterparts on the sending side.
All of the above are not necessary if we are receiving a stream for a
single dataset that has been generated with zfs send -p, that is, a
stream that includes properties. zfs_receive_one() already takes care
of applying the properties to the received datasets.
Pedro Giffuni [Fri, 7 Apr 2017 20:36:06 +0000 (13:36 -0700)]
OpenZFS 8046 - Let calloc() do the multiplication in libzfs_fru_refresh
Authored by: Pedro Giffuni <pfg@freebsd.org>
Approved by: Robert Mustacchi <rm@joyent.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: George Melikov <mail@gmelikov.ru> Ported-by: Giuseppe Di Natale <dinatale2@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/8046
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/3a3c0d5
Closes #5989
George Melikov [Sun, 9 Apr 2017 23:01:54 +0000 (03:01 +0400)]
zfs_receive_010_pos.ksh local => typeset change
Ksh uses `typeset`, `local` is a Bash analog.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: John Wren Kennedy <john.kennedy@delphix.com> Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes #5995
George Melikov [Sun, 9 Apr 2017 23:00:43 +0000 (03:00 +0400)]
zfstests cli_user/misc/setup.ksh space missed
Ksh syntax requires a space after `!` in if statement.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: John Wren Kennedy <john.kennedy@delphix.com> Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes #5994
Toomas Soome [Sat, 3 Dec 2016 07:13:44 +0000 (23:13 -0800)]
OpenZFS 7404 - rootpool_007_neg, bootfs_006_pos and bootfs_008_neg tests fail with the loader project bits
Authored by: Toomas Soome <tsoome@me.com>
Reviewed by: Igor Kozhukhov <igor@dilos.org>
Reviewed by: Marcel Telka <marcel@telka.sk>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Richard Lowe <richlowe@richlowe.net> Reviewed-by: George Melikov <mail@gmelikov.ru> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- Removed gzip and zle compression restriction on bootfs
datasets. Grub added support for these long ago. Ay
version of grub which understands lz4 also supports this.
- Enabled rootpool tests in runfile but skipped by default
in setup on Linux since they modify the rootpool.
- bootfs_006_pos.ksh, striped pools are allowed as bootfs.
OpenZFS 7629 - Fix for 7290 neglected to remove some escape sequences
Authored by: John Wren Kennedy <john.kennedy@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com> Reviewed-by: George Melikov <mail@gmelikov.ru> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Porting Notes:
- Multiple changes in this commit were applied in c1d9abf.
Consolidated the shellcheck call in the
make recipe down to a single call of
shellcheck. Corrected script errors that
have been skipped. Corrected script errors
that have been introduced because make
wasn't reporting any errors from shellcheck.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Closes #5976
The ZFS enabled versions of xfstests fails to build cleanly on
Amazon Linux. This issue should be resolved by rebasing the ZFS
patches against the latest xfstests and pushing those patches
upstream. This would allow us to use an unmodified xfstests.
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #5481
Closes #5977
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Closes #5974
OpenZFS 7290 - ZFS test suite needs to control what utilities it can run
Authored by: John Wren Kennedy <john.kennedy@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Dan McDonald <danmcd@omniti.com>
Approved by: Gordon Ross <gordon.w.ross@gmail.com> Ported-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: George Melikov <mail@gmelikov.ru>
Porting Notes:
- Utilities which aren't available under Linux have been removed.
- Because of sudo's default secure path behavior PATH must be
explicitly reset at the top of libtest.shlib. This avoids the
need for all users to customize secure path on their system.
- Updated ZoL infrastructure to manage constrained path
- Updated all test cases
- Check permissions for usergroup tests
- When testing in-tree create links under bin/
- Update fault cleanup such that missing files during
cleanup aren't fatal.
- Configure su environment with constrained path
Sydney Vanda [Thu, 2 Mar 2017 16:47:26 +0000 (09:47 -0700)]
Added auto-replace FMA test for the ZFS Test Suite
Also included are updates to auto-online test
Automated auto-replace test to go along with ZED FMA integration
(PR 4673) auto-replace_001.pos works using a scsi_debug device
(the only usable virtual device currently due to whole_disk var
needing to be set)
Functionality for automated FMA auto-replace test to work with
scsi_debug devs: Some functionality/exceptions needed to be
added for automation of auto-replace to work correctly.
In the test an alias vdev_id rule is added for any scsi_debug
device which sets the phys_path="scsidebug" after a udevadm
trigger command.
A symlink is created for the vdev_id.conf file (in /etc/zfs/ by
default) to be used in-tree for the test suite
(/var/tmp/zfs/vdev_id.conf). "./scripts/zfs-helpers.sh -i" needs
to be run before fault tests in the ZTS (to use udev rules in-tree)
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Don Brady <don.brady@intel.com> Reviewed-by: David Quigley <david.quigley@intel.com> Signed-off-by: Sydney Vanda <sydney.m.vanda@intel.com>
Closes #5944
Don Brady [Wed, 5 Apr 2017 21:24:26 +0000 (15:24 -0600)]
Fix regression in zfs_ereport_start()
On 32-bit platforms spa_state is 32 bits without cast, and thus
caused a NULL pointer dereference when treated as 64bit in
var arg. Accidentally introduced by bcdb96a.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Nathaniel Clark <nathaniel.l.clark@intel.com> Signed-off-by: Don Brady <don.brady@intel.com>
Closes #5966
Closes #5965
In _zed_event_add_nvpair, when handling DATA_TYPE_UINT64,
we should be using i64 throughout the entire case.
Reviewed-by: George Melikov <mail@gmelikov.ru> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Don Brady <don.brady@intel.com> Signed-off-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Closes #5964
Steven Hartland [Mon, 3 Apr 2017 23:38:51 +0000 (16:38 -0700)]
OpenZFS 7885 - zpool list can report 16.0e for expandsz
Authored by: Steven Hartland <steven.hartland@multiplay.co.uk>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Gordon Ross <gordon.w.ross@gmail.com>
When a member of a RAIDZ has been replaced with a device smaller than
the original, then the top level vdev can report its expand size as
16.0E.
The reduced child asize causes the RAIDZ to have a vdev_asize lower than
its vdev_max_asize which then results in an underflow during the
calculation of the parents expand size.
Fix this by updating the vdev_asize if it shrinks, which is already
protected by a check against vdev_min_asize so should always be safe.
Also for RAIDZ vdevs, ensure that the sum of their child vdev_min_asize
is always greater than the parents vdev_min_size.
Reviewed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Ported-by: George Melikov <mail@gmelikov.ru>
OpenZFS-issue: https://www.illumos.org/issues/7885
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/bb0dbaa
Closes #5963
projects / zfs / log