kern_sendfile() uses vm_rdwr() to read-ahead blocks of data to populate
page cache. When sendfile stumbles upon a page that is not populated
yet, it sends out all the mbufs that it collected so far. This
resulted in very poor performance with ZFS when file data is not in the
page cache, because ZFS vop_read for UIO_NOCOPY case populated only
those pages that are already in cache, but not valid. Which means that
most of the time it populated only the first requested page in the
described above scenario.
Reported by: Alexander Zagrebin <alexz@visp.ru>
Tested by: Alexander Zagrebin <alexz@visp.ru>,
Artemiev Igor <ai@kliksys.ru>
MFC after: 12 days
and VFS_RELE on a non-existing hold on snapshot parent's z_vfs.
This disables the changes from OpenSolaris onnv-revision 9234:bffdc4fc05c4
(bug IDs: 6792139, 6794830) - not applicable to FreeBSD.
This fixes the process hang if umounting a manually mounted snapshot.
Reported by: Alexander Zagrebin <alexz@visp.ru>
Approved by: delphij (mentor)
MFC after: 1 week
what we have. Without the check the kernel could accessing memory that
does not belong to the request struct.
Note that we do not test if the struct equals in size at this time, which
may faciliate forward compatibility with newer binaries.
Reviewed by: pjd at MeetBSD CA '2010
MFC after: 1 week
OpenSolaris onnv-revision: 10209:91f47f0e7728
6830541 zfs_get_data_trips on a verify
6696242 multiple zfs_fillpage() zfs: accessing past end of object panics
6785914 zfs fails to drop dn_struct_rwlock in recovery code path
Approved by: delphij (mentor)
Obtained from: OpenSolaris (Bug ID 6830541, 6696242, 6785914)
MFC after: 2 weeks
This should make vnode_pager_getpages path a bit shorter and clearer.
Also this should eliminate problems with partially valid pages.
Having this method opens room for future optimizations.
To do: try to satisfy other pages besides the required one taking into
account tradeofs between number of page faults, read throughput and read
latency. Also, eventually vop_putpages should be added too.
Reviewed by: kib, mm, pjd
MFC after: 3 weeks
Since r212650 and before this change sendfile(2) could produce
a partially valid page for a trailing portion of a ZFS vnode.
vm_fault() always wants to see a fully valid page even if it's the last
page that partially extends beyond vnode's end. Otherwise it calls
vop_getpages() to bring in the page. In the case of ZFS this means
that the data is read from the page into the same page and this breaks
checks in ZFS mappedread() - a thread that set VPO_BUSY on the page in
vm_fault() will get blocked forever waiting for it to be cleared.
Many thanks to Kai and Jeremy for reproducing the issue and providing
important debugging information and help.
Reported by: Kai Gallasch <gallasch@free.de>,
Jeremy Chadwick <freebsd@jdc.parodius.com>
Tested by: Kai Gallasch <gallasch@free.de>,
Jeremy Chadwick <freebsd@jdc.parodius.com>
Reviewed by: kib
MFC after: 3 days
To-Do: apply the same treatment to tmpfs + sendfile
Retry IO once with ZIO_FLAG_TRYHARD before declaring a pool faulted
OpenSolaris revision and Bug IDs:
9725:0bf7402e8022
6843014 ZFS B_FAILFAST handling is broken
Approved by: delphij (mentor)
Obtained from: OpenSolaris (Bug ID 6843014)
MFC after: 3 weeks
OpenSolaris revision and Bug IDs:
9701:cc5b64682e64
6803605 should be able to offline log devices
6726045 vdev_deflate_ratio is not set when offlining a log device
6599442 zpool import has faults in the display
Approved by: delphij (mentor)
Obtained from: OpenSolaris (Bug ID 6803605, 6726045, 6599442)
MFC after: 3 weeks
zfs_map_page/zfs_unmap_page are mostly called around potential I/O paths
and it seems to be a not very good idea to do cpu pinning there.
Suggested by: kib
MFC after: 2 weeks
Those checks are not present in upstream code and they are enforced in
actual calculations of delta by which ARC size can be grown or should be
reduced.
MFC after: 3 weeks
vm_paging_target() is not a trigger of any kind for pageademon, but
rather a "soft" target for it when it's already triggered.
Thus, trying to keep 2048 pages above that level at the expense of ARC
was simply driving ARC size into the ground even with normal memory
loads.
Instead, use a threshold at which a pagedaemon scan is triggered, so
that ARC reclaiming helps with pagedaemon's task, but the latter still
recycles active and inactive pages.
PR: kern/146410, kern/138790
MFC after: 3 weeks
Fix possible loss of correct error return code in ZFS mount
OpenSolaris revisions and Bug IDs:
11824:53128e5db7cf
6863610 ZFS mount can lose correct error return
12079:13822b941977
6939941 problem with moving files in zfs (142901-12)
Approved by: delphij (mentor)
Obtained from: OpenSolaris (Bug ID 6863610, 6939941)
MFC after: 3 days
This mirrors code in tmpfs.
This changge shouldn't affect much read path, it may cause unnecessary
vm_page_lookup calls in the case where v_object has no active or inactive
pages but has some cache pages. I believe this situation to be non-essential.
In write path this change should allow us to properly detect the above
case and free a cache page when we write to a range that corresponds to it.
If this situation is undetected then we could have a discrepancy between
data in page cache and in ARC or on disk.
This change allows us to re-enable vn_has_cached_data() check in zfs_write.
NOTE: strictly speaking resident_page_count and cache fields of v_object
should be exmined under VM_OBJECT_LOCK, but for this particular usage
we may get away with it.
Discussed with: alc, kib
Approved by: pjd
Tested with: tools/regression/fsx
MFC after: 3 weeks
Otherwise, adding insult to injury, in addition to double-caching of data
we would always copy the data into a vnode's vm object page from backend.
This is specific to sendfile case only (VOP_READ with UIO_NOCOPY).
PR: kern/141305
Reported by: Wiktor Niesiobedzki <bsd@vink.pl>
Reviewed by: alc
Tested by: tools/regression/sockets/sendfile
MFC after: 2 weeks
* processes now can't go away while we are inserting probes (fixes a panic)
* if a trap happens, we won't be holding the process lock (fixes a hang)
* fix a LOR between the process lock and the fasttrap bucket list lock
Thanks to kib for pointing some problems.
Sponsored by: The FreeBSD Foundation
code associated with overflow or with the drain function. While this
function is not expected to be used often, it produces more information
in the form of an errno that sbuf_overflowed() did.
Add the BIO_ORDERED flag for struct bio and update bio clients to use it.
The barrier semantics of bioq_insert_tail() were broken in two ways:
o In bioq_disksort(), an added bio could be inserted at the head of
the queue, even when a barrier was present, if the sort key for
the new entry was less than that of the last queued barrier bio.
o The last_offset used to generate the sort key for newly queued bios
did not stay at the position of the barrier until either the
barrier was de-queued, or a new barrier (which updates last_offset)
was queued. When a barrier is in effect, we know that the disk
will pass through the barrier position just before the
"blocked bios" are released, so using the barrier's offset for
last_offset is the optimal choice.
sys/geom/sched/subr_disk.c:
sys/kern/subr_disk.c:
o Update last_offset in bioq_insert_tail().
o Only update last_offset in bioq_remove() if the removed bio is
at the head of the queue (typically due to a call via
bioq_takefirst()) and no barrier is active.
o In bioq_disksort(), if we have a barrier (insert_point is non-NULL),
set prev to the barrier and cur to it's next element. Now that
last_offset is kept at the barrier position, this change isn't
strictly necessary, but since we have to take a decision branch
anyway, it does avoid one, no-op, loop iteration in the while
loop that immediately follows.
o In bioq_disksort(), bypass the normal sort for bios with the
BIO_ORDERED attribute and instead insert them into the queue
with bioq_insert_tail(). bioq_insert_tail() not only gives
the desired command order during insertion, but also provides
barrier semantics so that commands disksorted in the future
cannot pass the just enqueued transaction.
sys/sys/bio.h:
Add BIO_ORDERED as bit 4 of the bio_flags field in struct bio.
sys/cam/ata/ata_da.c:
sys/cam/scsi/scsi_da.c
Use an ordered command for SCSI/ATA-NCQ commands issued in
response to bios with the BIO_ORDERED flag set.
sys/cam/scsi/scsi_da.c
Use an ordered tag when issuing a synchronize cache command.
Wrap some lines to 80 columns.
sys/cddl/contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c
sys/geom/geom_io.c
Mark bios with the BIO_FLUSH command as BIO_ORDERED.
Sponsored by: Spectra Logic Corporation
MFC after: 1 month
least one execute bit set, otherwise execve(2) will return EACCES even
for an user with PRIV_VFS_EXEC privilege.
Add the check also to vaccess(9), vaccess_acl_nfs4(9) and
vaccess_acl_posix1e(9). This makes access(2) to better agree with
execve(2). Because ZFS doesn't use vaccess(9) for VEXEC, add the check
to zfs_freebsd_access() too. There may be other file systems which are
not using vaccess*() functions and need to be handled separately.
PR: kern/125009
Reviewed by: bde, trasz
Approved by: pjd (ZFS part)
- better ACL caching and speedup of ACL permission checks
- faster handling of stat()
- lowered mutex contention in the read/writer lock (rrwlock)
- several related bugfixes
Detailed information (OpenSolaris onnv changesets and Bug IDs):
9749:105f407a2680
6802734 Support for Access Based Enumeration (not used on FreeBSD)
6844861 inconsistent xattr readdir behavior with too-small buffer
9866:ddc5f1d8eb4e
6848431 zfs with rstchown=0 or file_chown_self privilege allows user to "take" ownership
9981:b4907297e740
6775100 stat() performance on files on zfs should be improved
6827779 rrwlock is overly protective of its counters
10143:d2d432dfe597
6857433 memory leaks found at: zfs_acl_alloc/zfs_acl_node_alloc
6860318 truncate() on zfsroot succeeds when file has a component of its path set without access permission
10232:f37b85f7e03e
6865875 zfs sometimes incorrectly giving search access to a dir
10250:b179ceb34b62
6867395 zpool_upgrade_007_pos testcase panic'd with BAD TRAP: type=e (#pf Page fault)
10269:2788675568fd
6868276 zfs_rezget() can be hazardous when znode has a cached ACL
10295:f7a18a1e9610
6870564 panic in zfs_getsecattr
Approved by: delphij (mentor)
Obtained from: OpenSolaris (multiple Bug IDs)
MFC after: 2 weeks
This provides a noticeable write speedup, especially on pools with
less than 30% of free space.
Detailed information (OpenSolaris onnv changesets and Bug IDs):
11146:7e58f40bcb1c
6826241 Sync write IOPS drops dramatically during TXG sync
6869229 zfs should switch to shiny new metaslabs more frequently
11728:59fdb3b856f6
6918420 zdb -m has issues printing metaslab statistics
12047:7c1fcc8419ca
6917066 zfs block picking can be improved
Approved by: delphij (mentor)
Obtained from: OpenSolaris (Bug ID 6826241, 6869229, 6918420, 6917066)
MFC after: 2 weeks
resetting needfree
needfree is checked at the very start of arc_reclaim_needed.
This change makes code easier to follow and maintain in face of
potential changed in arc_reclaim_needed.
Also, put the whole sub-block under _KERNEL because needfree can be set
only in kernel code.
To do: rename needfree to something else to aovid confusion with
OpenSolaris global variable of the same name which is used in the same
code, but has different meaning (page deficit).
Note: I have an impression that locking around accesses to this variable
as well as mutual notifications between arc_reclaim_thread and
arc_lowmem are not proper.
MFC after: 1 week
