Run arc_evict thread at higher priority, nice=0, to give it more CPU
time which can improve performance for workload with high ARC evict
activities.
On mixed read/write and sequential read workloads, I've seen between
10-40% better performance.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Tony Nguyen <tony.nguyen@delphix.com>
Closes#12397
It turns out, there are a lot of possible reasons for fopen to fail.
Let's share which reason we failed for today.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
Closes#12410
The /proc/diskstats accounting needs to be explicitly enabled
for block devices which do not use multi-queue.
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#12440Closes#12066
Describe sequential scrub and add examples of scrub status.
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes#12429
CpaDcGeneratefooter function that obtain the checksum code
does not support the CPA_DC_STATELESS mode. So we get the
adler32 chencksum of the end of the zlib from dc_results.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Chengfei Zhu <chengfeix.zhu@intel.com>
Signed-off-by: hedong.zhang <h_d_zhang@163.com>
Closes#12343
We have a tunable which permits one to disable the use of unmapped I/O
for the buffer cache. Respect it in ZFS as well. This is useful for
KMSAN, which cannot easily maintain shadow state for unmapped pages.
No functional change intended, as unmapped I/O is permitted by default
and there's no real reason to disable it in practice except for
debugging.
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes#12446
- Bail out early if we're running the perf tests and forget to
specify disks.
- Allow perf tests to run with any number of disks.
- Remove weekly vs. nightly settings
- Move variables with common values to perf.shlib
- Use zinject to clear the ARC over export/import
- Fix dbuf cache size calculation
When the meaning of `dbuf_cache_max_bytes` changed, the performance
test that covers the dbuf cache started to fail. The test would try to
write files for the test using the max possible size of the cache,
inevitably filling the pool and failing. This change uses
`dbuf_cache_shift` to correctly calculate the dbuf cache size.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: John Kennedy <john.kennedy@delphix.com>
Closes#12408
`zpool_do_import()` passes `argv[0]`, (optionally) `argv[1]`, and
`pool_specified` to `import_pools()`. If `pool_specified==FALSE`, the
`argv[]` arguments are not used. However, these values may be off the
end of the `argv[]` array, so loading them could dereference unmapped
memory. This error is reported by the asan build:
```
=================================================================
==6003==ERROR: AddressSanitizer: heap-buffer-overflow
READ of size 8 at 0x6030000004a8 thread T0
#0 0x562a078b50eb in zpool_do_import zpool_main.c:3796
#1 0x562a078858c5 in main zpool_main.c:10709
#2 0x7f5115231bf6 in __libc_start_main
#3 0x562a07885eb9 in _start
0x6030000004a8 is located 0 bytes to the right of 24-byte region
allocated by thread T0 here:
#0 0x7f5116ac6b40 in __interceptor_malloc
#1 0x562a07885770 in main zpool_main.c:10699
#2 0x7f5115231bf6 in __libc_start_main
```
This commit passes NULL for these arguments if they are off the end
of the `argv[]` array.
Reviewed-by: George Wilson <gwilson@delphix.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Allan Jude <allan@klarasystems.com>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Closes#12339
In l2arc_add_vdev() first decide whether the device is eligible for
L2ARC rebuild or whole device trim and then add it to the list of cache
devices. Otherwise l2arc_feed_thread() might already start writing on
the device invalidating previous content as l2ad_hand = l2ad_start.
However l2arc_rebuild_vdev() needs the device present in the cache
device list to figure out its l2arc_dev_t. Fix this by moving most of
l2arc_rebuild_vdev() in a new function l2arc_rebuild_dev() which does
not need to search in the cache device list.
In contrast to l2arc_add_vdev() we do not have to worry about
l2arc_feed_thread() invalidating previous content when onlining a
cache device. The device parameters (l2ad*) are not cleared when
offlining the device and writing new buffers will not invalidate
all previous content. In worst case only buffers that have not had
their log block written to the device will be lost.
Retire persist_l2arc_00{4,5,8} tests since they cover code already
covered by the remaining ones. Test persist_l2arc_006 is renamed to
persist_l2arc_004 and persist_l2arc_007 is renamed to persist_l2arc_005.
Fix a typo in persist_l2arc_004, and remove an assertion that is not
always true from l2arc_arcstats_pos. Also update an assertion in
persist_l2arc_005 and explain why in a comment.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: George Amanakis <gamanakis@gmail.com>
Closes#12365
It seems nothing ensures that this array is zeroed when a dnode is
freshly allocated, so in principle it retains the values from the
previous allocation. In practice it seems to be the case that the
fields should end up zeroed, but we can zero the field anyway for
consistency.
This was found using KMSAN.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes#12383
When logging a TX_WRITE record in the case where file data has to be
copied from the DMU, we pad the log record size to a multiple of 8
bytes. In this case, any padding bytes should be zeroed, otherwise the
contents of uninitialized memory are written to the ZIL.
This was found using KMSAN.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes#12383
When allocating a record, we round up the allocation size to a multiple
of 8. In this case, any padding bytes should be zeroed, otherwise the
contents of uninitialized memory are written to the ZIL.
This was found using KMSAN.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes#12383
When logging TX_SETATTR, we could otherwise fail to initialize part of
the corresponding ZIL record depending on which fields are present in
the xvattr. Initialize the creation time and the AV scan timestamp to
zero so that uninitialized bytes are not written to the ZIL.
This was found using KMSAN.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes#12383
spa_prop_find() may fail to find the specified property, in which case
it suppresses ENOENT from zap_lookup(). In this case, the return value
is left uninitialized, so spa_autoreplace was being initialized using an
uninitialized stack variable.
This was found using KMSAN. It appears to be a regression from commit
9eb7b46ed0, which removed the initialization of "autoreplace" from the
definition.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Mark Johnston <markj@FreeBSD.org>
Closes#12383
Kernel 5.14 introduced a change where set_page_dirty of
struct address_space_operations is no longer implicitly set to
__set_page_dirty_buffers(), which ended up resulting in a NULL
pointer deref in the kernel when it is attempted to be called.
This change sets .set_page_dirty in the structure to
__set_page_dirty_nobuffers(), which was introduced with the
related patch set. The breaking change was introduce in commit
0af573780b0b13fceb7fabd49dc1b073cee9a507 to torvalds/linux.git.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Coleman Kane <ckane@colemankane.org>
Closes#12427
After 1325434b, we can in certain circumstances end up calling
spa_update_dspace with vd->vdev_mg NULL, which ends poorly during
vdev removal.
So let's not do that further space adjustment when we can't.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
Closes#12380Closes#12428
In Linux 5.14, blk_alloc_queue is no longer exported, and its usage
has been superseded by blk_alloc_disk, which returns a gendisk struct
from which we can still retrieve the struct request_queue* that is
needed in the one place where it is used. This also replaces the call
to alloc_disk(minors), and minors is now set via struct member
assignment.
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Reviewed-by: Olaf Faaland <faaland1@llnl.gov>
Reviewed-by: Coleman Kane <ckane@colemankane.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#12362Closes#12409
It is useful to have control over the number of iterations of zloop so
we can easily produce "x core dumps found *in y iterations*" metrics.
Using random values for run/pass times doesn't improve coverage in a
meaningful way.
Randomizing run time could be seen as a compromise between running a
greater variety of shorter tests versus a smaller variety of longer
tests within a fixed time span. However, it is not desirable when
running a fixed number of iterations.
Pass time already incorporates randomness within ztest.
Either parameter can be passed to ztest explicitly if the defaults are
not satisfactory.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Signed-off-by: Ryan Moeller <ryan@iXsystems.com>
Closes#12411
Since errors returned by zvol_create_minor_impl() are ignored by the
common code, it is more convenient to ignore make_dev_s() errors there.
It allows, for example, to get device created for the zvol after later
rename instead of having it further stuck in half-created state.
zvol_rename_minor() already ignores those errors.
While there, switch from MAXPHYS to maxphys in FreeBSD 13+.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12375
Possibly required in the past, but is currently fills no purpose.
Ordinarily such tiny cleanup is not generally worth it, however
on the macOS port, in a future commit, we do unspeakable things to the
"fd" for send/recv, and it would be easier to only have to deal with
one "fd" instead of two.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes#12404
Remove mc_lock use from metaslab_class_throttle_*(). The math there
is based on refcounts and so atomic, so the only race possible there
is between zfs_refcount_count() and zfs_refcount_add(). But in most
cases metaslab_class_throttle_reserve() is called with the allocator
lock held, which covers the race. In cases where the lock is not
held, GANG_ALLOCATION() or METASLAB_MUST_RESERVE are set, and so we
do not use zfs_refcount_count(). And even if we assume some other
non-existing scenario, the worst that may happen from this race is
few more I/Os get to allocation earlier, that is not a problem.
Move locks and data of different allocators into different cache
lines to avoid false sharing. Group spa_alloc_* arrays together
into single array of aligned struct spa_alloc spa_allocs. Align
struct metaslab_class_allocator.
Reviewed-by: Paul Dagnelie <pcd@delphix.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Reviewed-by: Don Brady <don.brady@delphix.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12314
So commit author can just download them as
artifacts and commit.
Reviewed-by: Ryan Moeller <ryan@ixsystems.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes#12379
Remove unneeded global, practically constant, state pointer variables
(arc_anon, arc_mru, etc.), replacing them with macros of real state
variables addresses (&ARC_anon, &ARC_mru, etc.).
Change ARC_EVICT_ALL from -1ULL to UINT64_MAX, not requiring special
handling in inner loop of ARC reclamation. Respectively change bytes
argument of arc_evict_state() from int64_t to uint64_t.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Mark Maybee <mark.maybee@delphix.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12348
Ensure all calls to bqueue_init() has a corresponding call to bqueue_destroy()
Reviewed-by: Paul Dagnelie <pcd@delphix.com>
Co-authored-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes#12118
* zio: avoid callback typecasting
* zil: avoid zil_itxg_clean() callback typecasting
* zpl: decouple zpl_readpage() into two separate callbacks
* nvpair: explicitly declare callbacks for xdr_array()
* linux/zfs_nvops: don't use external iput() as a callback
* zcp_synctask: don't use fnvlist_free() as a callback
* zvol: don't use ops->zv_free() as a callback for taskq_dispatch()
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Mark Maybee <mark.maybee@delphix.com>
Signed-off-by: Alexander Lobakin <alobakin@pm.me>
Closes#12260
We don't use or need the pool name or value source in the zvol tasks.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Ryan Moeller <ryan@iXsystems.com>
Closes#12361
Reviewed-by: Allan Jude <allan@klarasystems.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12378
Use strlcpy instead of problematic strncpy
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: George Melikov <mail@gmelikov.ru>
Closes#12344
zfs-send(8) claimed in the flags list you could use -pR when sending
a readonly filesystem or volume. You cannot.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
Closes#12336
Most of dsl_dir_diduse_space() and dsl_dir_transfer_space() CPU time
is a dd_lock overhead and time spent in dmu_buf_will_dirty(). Calling
them one after another is a waste of time and even more contention.
Doing that twice for each rewritten block within dbuf_write_done()
via dsl_dataset_block_kill() and dsl_dataset_block_born() created one
of the biggest CPU overheads in case of small blocks rewrite.
dsl_dir_diduse_transfer_space() combines functionality of these two
functions for cases where it is needed, but without double overhead,
practically for the cost of dsl_dir_diduse_space() or even cheaper.
While there, optimize dsl_dir_phys() calls in dsl_dir_diduse_space()
and dsl_dir_transfer_space(). It seems Clang detects some aliasing
there, repeating dd->dd_dbuf->db_data dereference multiple times,
increasing dd_lock scope and contention.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Author: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12300
In absence of LTO, and dynamic libatomic, la.so ends up in the needs
section of every toolchain executable; some consider this an issue.
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Closes#12345Closes#12359
* Tinker with slop space accounting with dedup
Do not include the deduplicated space usage in the slop space
reservation, it leads to surprising outcomes.
* Update spa_dedup_dspace sometimes
Sometimes, we get into spa_get_slop_space() with
spa_dedup_dspace=~0ULL, AKA "unset", while spa_dspace is correctly set.
So call the code to update it before we use it if we hit that case.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Mark Maybee <mark.maybee@delphix.com>
Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
Closes#12271
arc_evict_hdr() returns number of evicted bytes in scope of specific
state. For ghost states it does not mean the amount of really freed
memory, but the logical buffer size. It is correct for the eviction
process, but not for waking up threads waiting for ARC size reduction,
as added in "Revise ARC shrinker algorithm" commit, causing premature
wakeups while ARC is still overflowed, allowing even bigger overflow,
plus processing overhead when next allocation will also get blocked,
probably also for too short time.
To fix that make arc_evict_hdr() also return the amount of really
freed memory, which for the ghost states is only the header, and use
it to update arc_evict_count instead. Originally I was thinking to
not return it at all, since arc_get_data_impl() does not account for
the headers, but decided that some slow allocation progress is better
than long waits, reaching on my tests up to 100ms.
To reduce negative latency effects of long time periods when reclaim
thread can free little real memory, start reclamation process earlier,
before we actually reached the overflow threshold, when we have to
throttle new allocations. We can also do it without taking global
arc_evict_lock, reducing the contention.
Reviewed-by: George Wilson <gwilson@delphix.com>
Reviewed-by: Allan Jude <allan@klarasystems.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12279
- Remove the "SPL Version" line, the repositories have been merged
since the 0.8 release and we no longer need to ask about this.
- Simply ask for the kernel version / patch level and add a hint
about how to get this information on Linux and FreeBSD.
- Remove "Status: Triage Needed" from the template, in practice
we really haven't been using this label so let's step setting it.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Reviewed-by: John Kennedy <john.kennedy@delphix.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes: #12340
Callers of zfs_file_get and zfs_file_put can corrupt the reference
counts for the file structure resulting in a panic or a soft lockup.
When zfs send/recv runs, it will add a reference count to the
open file, and begin to send or recv the stream. If the file descriptor
is closed, then when dmu_recv_stream() or dmu_send() return we will
call zfs_file_put to remove the reference we placed on the file
structure. Unfortunately, because zfs_file_put() uses the file
descriptor to lookup the file structure, it may end up finding that
the file descriptor table no longer contains the file struct, thus
leaking the file structure. Or it might end up finding a file
descriptor for a different file and blindly updating its reference
counts. Other failure modes probably exists.
This change reworks the zfs_file_[get|put] interface to not rely
on the file descriptor but instead pass the zfs_file_t pointer around.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Mark Maybee <mark.maybee@delphix.com>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Co-authored-by: Allan Jude <allan@klarasystems.com>
Signed-off-by: George Wilson <gwilson@delphix.com>
External-issue: DLPX-76119
Closes#12299
Missed a couple of strcpy() in earlier commit, this is only used with
--enable-debug.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Tony Nguyen <tony.nguyen@delphix.com>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes#12311
Could have gone either way with this one, either adding it to
macOS/Windows SPL, or returning it to "classic" usage with strrchr().
Since the new special way isn't really used, and only used once,
we have this commit.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes #12312
Many FreeBSD disk drivers support "unmapped" I/O mode, when data
buffer represented not with a virtually contiguous KVA-mapped address
range, but with a list of physical memory pages. Originally it was
designed to do I/O from buffers without KVA mapping (unmapped). But
moving virtual addresses out of equation allows us to operate even
non-contiguous data buffers with one condition: all buffer discon-
tinuities must be aligned to memory page borders.
Doing I/O to capable GEOM device this patch traverses through non-
linear ABD buffers, validating the chunks borders. If the condition
is met, it supplies GEOM with the list of original physical memory
pages instead of copying the data into temporary contiguous buffer.
On capable hardware on pools with ashift=12 and default ABD chunk of
4KB it should handle all the I/O without additional memory copying.
Reviewed-by: Brian Atkinson <batkinson@lanl.gov>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12320
It makes no sense to set it below PAGE_SIZE, since it increases all
overheads and makes returning memory to OS problematic. It makes no
sense to set it above PAGE_SIZE, since such allocations and especially
frees are too expensive and cause KVA fragmentation to benefit from
fewer chunks. After that it makes no sense to keep more complicated
math here.
What may have sense though is just a tunable border between linear and
scatter ABDs, previously also controlled by this tunable. Retain that
functionality by taking abd_scatter_min_size tunable from Linux, just
with different default value.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Brian Atkinson <batkinson@lanl.gov>
Reviewed-by: Ryan Moeller <ryan@iXsystems.com>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12328
This dramatically reduces the lock contention on systems with slower
(non-TSC) timecounters. With TSC the difference is minimal, but since
this lock is pretty congested, any improvement counts. Plus I don't
see any reason to do it under the lock other than the latency of the
lock itself, which this change actually reduces.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12281
Currently, there are several places in zvol_id where the program logic
returns particular errno values, or even particular ioctl return values,
as the program exit status, rather than a straightforward system of
explicit zero on success and explicit nonzero value(s) on failure.
This is problematic for multiple reasons. One particularly interesting
problem that can arise, is that if any of these values happens to have
all 8 least significant bits unset (i.e., it is a positive or negative
multiple of 256), then although the C program sees a nonzero int value
(presumed to be a failure exit status), the actual exit status as seen
by the system is only the bottom 8 bits of that integer: zero.
This can happen in practice, and I have encountered it myself. In a
particularly weird situation, the zvol_open code in the zfs kernel
module was behaving in such a manner that it caused the open() syscall
to fail and for errno to be set to a kernel-private value (ERESTARTSYS,
which happens to be defined as 512). It turns out that 512 is evenly
divisible by 256; or, in other words, its least significant 8 bits are
all-zero. So even though zvol_id believed it was returning a nonzero
(failure) exit status of 512, the system modulo'd that value by 256,
resulting in the actual exit status visible by other programs being 0!
This actually-zero (non-failure) exit status caused problems: udev
believed that the program was operating successfully, when in fact it
was attempting to indicate failure via a nonzero exit status integer.
Combined with another problem, this led to the creation of nonsense
symlinks for zvol dev nodes by udev.
Let's get rid of all this problematic logic, and simply return
EXIT_SUCCESS (0) is everything went fine, and EXIT_FAILURE (1) if
anything went wrong.
Additionally, let's clarify some of the variable names (error is similar
to errno, etc) and clean up the overall program flow a bit.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com>
Signed-off-by: Justin Gottula <justin@jgottula.com>
Closes#12302
The zvol_id program is invoked by udev, via a PROGRAM key in the
60-zvol.rules.in rule file, to determine the "pretty" /dev/zvol/*
symlink paths paths that should be generated for each opaquely named
/dev/zd* dev node.
The udev rule uses the PROGRAM key, followed by a SYMLINK+= assignment
containing the %c substitution, to collect the program's stdout and then
"paste" it directly into the name of the symlink(s) to be created.
Unfortunately, as currently written, zvol_id outputs both its intended
output (a single string representing the symlink path that should be
created to refer to the name of the dataset whose /dev/zd* path is
given) AND its error messages (if any) to stdout.
When processing PROGRAM keys (and others, such as IMPORT{program}), udev
uses only the data written to stdout for functional purposes. Any data
written to stderr is used solely for the purposes of logging (if udev's
log_level is set to debug).
The unintended consequence of this is as follows: if zvol_id encounters
an error condition; and then udev fails to halt processing of the
current rule (either because zvol_id didn't return a nonzero exit
status, or because the PROGRAM key in the rule wasn't written properly
to result in a "non-match" condition that would stop the current rule on
a nonzero exit); then udev will create a space-delimited list of symlink
names derived directly from the words of the error message string!
I've observed this exact behavior on my own system, in a situation where
the open() syscall on /dev/zd* dev nodes was failing sporadically (for
reasons that aren't especially relevant here). Because the open() call
failed, zvol_id printed "Unable to open device file: /dev/zd736\n" to
stdout and then exited.
The udev rule finished with SYMLINK+="zvol/%c %c". Assuming a volume
name like pool/foo/bar, this would ordinarily expand to
SYMLINK+="zvol/pool/foo/bar pool/foo/bar"
and would cause symlinks to be created like this:
/dev/zvol/pool/foo/bar -> /dev/zd736
/dev/pool/foo/bar -> /dev/zd736
But because of the combination of error messages being printed to
stdout, and the udev syntax freely accepting a space-delimited sequence
of names in this context, the error message string
"Unable to open device file: /dev/zd736\n"
in reality expanded to
SYMLINK+="zvol/Unable to open device file: /dev/zd736"
which caused the following symlinks to actually be created:
/dev/zvol/Unable -> /dev/zd736
/dev/to -> /dev/zd736
/dev/open -> /dev/zd736
/dev/device -> /dev/zd736
/dev/file: -> /dev/zd736
/dev//dev/zd736 -> /dev/zd736
(And, because multiple zvols had open() syscall errors, multiple zvols
attempted to claim several of those symlink names, resulting in numerous
udev errors and timeouts and general chaos.)
This commit rectifies all this silliness by simply printing error
messages to stderr, as Dennis Ritchie originally intended.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com>
Signed-off-by: Justin Gottula <justin@jgottula.com>
Closes#12302
Assignment syntax (=) can be used for the PROGRAM key. But the PROGRAM
key is really a match key, not an assign key. The internal logic used by
udev to decide whether a PROGRAM key "matched" or not (which determines
whether the remainder of the rule is evaluated) depends on whether the
operator was OP_MATCH (==) or OP_NOMATCH (!=). [1]
The man page claims that '"=", ":=", and "+=" have the same effect as
"=="' for PROGRAM keys. And, after a brief perusal, the udev source code
does seem to confirm that operators other than OP_MATCH (==) or
OP_NOMATCH (!=) are implicitly converted to OP_MATCH (==). [2] But it's
not entirely clear that this is definitely the case: anecdotal testing
seems to indicate that when OP_ASSIGN (=) is used, the program's exit
status is disregarded and the remainder of the rule is processed
regardless of whether it was, in fact, a successful exit.
The bottom line here is that, if zvol_id hits some snag and returns a
nonzero exit status, then we almost certainly do NOT want to continue on
with the rule and use whatever the stdout contents may have been to
mindlessly create /dev/zvol/* symlinks. Therefore, let's be extra-sure
and use the match (==) operator explicitly, to eliminate any possibility
that udev might do the wrong thing, and ensure that a nonzero exit
status will definitely short-circuit the rest of the rule, bypassing the
SYMLINK+= assignments.
[1]
udev,
file src/udev/udev-rules.c,
func udev_rule_apply_token_to_event,
switch case TK_M_PROGRAM if r != 0 (nonzero exit status):
return token->op == OP_NOMATCH;
switch case TK_M_PROGRAM if r == 0 (zero exit status):
return token->op == OP_MATCH;
func retval 0 => key is considered to have matched
func retval 1 => key is considered to have NOT matched
[2]
udev,
file src/udev/udev-rules.c,
func parse_token,
at func start:
bool is_match = IN_SET(op, OP_MATCH, OP_NOMATCH);
in else-if case streq(key, "PROGRAM"):
if (!is_match) op = OP_MATCH;
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com>
Signed-off-by: Justin Gottula <justin@jgottula.com>
Closes#12302
The $tempnode substitution is so old that it's not even mentioned in the
man page anymore. It is still technically supported by udev, but with
plenty of "deprecated" comments surrounding it.
The preferred modern equivalent of $tempnode is $devnode (or
alternatively, %N).
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com>
Signed-off-by: Justin Gottula <justin@jgottula.com>
Closes#12302
This file is old as dirt. It's entirely possible that commas were
optional in udev back at that time. But they're definitely supposed to
be there nowadays.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Pavel Zakharov <pavel.zakharov@delphix.com>
Signed-off-by: Justin Gottula <justin@jgottula.com>
Closes#12302
With default dbuf cache size of 1/32 of ARC, it makes no sense to have
hash table of the same size (or even bigger on Linux). Reduce it to
1/8 of ARC's one, still leaving some slack, assuming higher I/O rate
via dbuf cache than via ARC.
Remove padding from ARC hash locks array. The idea behind padding
is to avoid false sharing between locks. It would have sense if
there would be a limited number of very busy locks. But since we
have no limit on the number, using the same memory for more locks we
can achieve even lower lock contention with the same false sharing,
or we can use less memory for the same contention level.
Reduce number of hash locks from 8192 to 2048. The number is still
big enough to not cause contention, but reduced memory size improves
cache hit rate for mutex_tryenter() in ARC eviction thread, saving
about 1% of the thread time.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Closes#12289
Fix a leak of abd_t that manifested mostly when using
raidzN with at least as many columns as N (e.g. a
four-disk raidz2 but not a three-disk raidz2).
Sufficiently heavy raidz use would eventually run a system
out of memory.
Additionally:
* Switch abd_cache arena to FIRSTFIT, which empirically
improves perofrmance.
* Make abd_chunk_cache more performant and debuggable.
* Allocate the abd_zero_buf from abd_chunk_cache rather
than the heap.
* Don't try to reap non-existent qcaches in abd_cache arena.
* KM_PUSHPAGE->KM_SLEEP when allocating chunks from their
own arena
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Co-authored-by: Sean Doran <smd@use.net>
Closes#12295
dmu_zfetch_stream_fini() is missing calls to destroy the refcounts,
leaking them and the mutex inside.
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Signed-off-by: Jorgen Lundman <lundman@lundman.net>
Closes#12294
