Commit 519129f added support to multi-thread 'zpool import' for
the case where block devices are scanned for under /dev/. This
commit generalizes that logic and applies it to the case where
device names are acquired from libblkid.
The zpool_find_import_scan() and zpool_find_import_blkid()
functions create an AVL tree containing each device name. Each
entry in this tree is dispatched to a taskq where the function
zpool_open_func() validates the device by opening it and reading
the label. This may result in additional entries being added
to the tree and those device paths being verified.
This is largely how the upstream OpenZFS code behaves but due to
significant differences the non-Linux code has been dropped for
readability. Additionally, this code makes use of taskqs and
kmutexs which are normally not available to the command line tools.
Special care has been taken to allow their use in the import
functions.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Olaf Faaland <faaland1@llnl.gov>
Closes#4794
Metadata-intensive workloads can cause the ARC to become permanently
filled with dnode_t objects as they're pinned by the VFS layer.
Subsequent data-intensive workloads may only benefit from about
25% of the potential ARC (arc_c_max - arc_meta_limit).
In order to help track metadata usage more precisely, the other_size
metadata arcstat has replaced with dbuf_size, dnode_size and bonus_size.
The new zfs_arc_dnode_limit tunable, which defaults to 10% of
zfs_arc_meta_limit, defines the minimum number of bytes which is desirable
to be consumed by dnodes. Attempts to evict non-metadata will trigger
async prune tasks if the space used by dnodes exceeds this limit.
The new zfs_arc_dnode_reduce_percent tunable specifies the amount by
which the excess dnode space is attempted to be pruned as a percentage of
the amount by which zfs_arc_dnode_limit is being exceeded. By default,
it tries to unpin 10% of the dnodes.
The problem of dnode metadata pinning was observed with the following
testing procedure (in this example, zfs_arc_max is set to 4GiB):
- Create a large number of small files until arc_meta_used exceeds
arc_meta_limit (3GiB with default tuning) and arc_prune
starts increasing.
- Create a 3GiB file with dd. Observe arc_mata_used. It will still
be around 3GiB.
- Repeatedly read the 3GiB file and observe arc_meta_limit as before.
It will continue to stay around 3GiB.
With this modification, space for the 3GiB file is gradually made
available as subsequent demands on the ARC are made. The previous behavior
can be restored by setting zfs_arc_dnode_limit to the same value as the
zfs_arc_meta_limit.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4345
Issue #4512
Issue #4773Closes#4858
Prior to b39c22b, which was first generally available in the 0.6.5
release as b39c22b, ZoL never actually submitted synchronous read or write
requests to the Linux block layer. This means the vdev_disk_dio_is_sync()
function had always returned false and, therefore, the completion in
dio_request_t.dr_comp was never actually used.
In b39c22b, synchronous ZIO operations were translated to synchronous
BIO requests in vdev_disk_io_start(). The follow-on commits 5592404 and
aa159af fixed several problems introduced by b39c22b. In particular,
5592404 introduced the new flag parameter "wait" to __vdev_disk_physio()
but under ZoL, since vdev_disk_physio() is never actually used, the wait
flag was always zero so the new code had no effect other than to cause
a bug in the use of the dio_request_t.dr_comp which was fixed by aa159af.
The original rationale for introducing synchronous operations in b39c22b
was to hurry certains requests through the BIO layer which would have
otherwise been subject to its unplug timer which would increase the
latency. This behavior of the unplug timer, however, went away during the
transition of the plug/unplug system between kernels 2.6.32 and 2.6.39.
To handle the unplug timer behavior on 2.6.32-2.6.35 kernels the
BIO_RW_UNPLUG flag is used as a hint to suppress the plugging behavior.
For kernels 2.6.36-2.6.38, the REQ_UNPLUG macro will be available and
ise used for the same purpose.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4858
Silence the following warning when compiling with gcc 5.4.0.
Specifically gcc (Ubuntu 5.4.0-6ubuntu1~16.04.1) 5.4.0 20160609.
module/avl/avl.c: In function ‘avl_add’:
module/avl/avl.c:647:2: warning: ‘where’ may be used uninitialized
in this function [-Wmaybe-uninitialized]
avl_insert(tree, new_node, where);
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Remove duplicate z_uid/z_gid member which are also held in the
generic vfs inode struct. This is done by first removing the members
from struct znode and then using the KUID_TO_SUID/KGID_TO_SGID
macros to access the respective member from struct inode. In cases
where the uid/gids are being marshalled from/to disk, use the newly
introduced zfs_(uid|gid)_(read|write) functions to properly
save the uids rather than the internal kernel representation.
Signed-off-by: Nikolay Borisov <n.borisov.lkml@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4685
Issue #227
Since the concept of a kuid and the need to translate from it to
ordinary integer type was added in kernel version 3.5 implement necessary
plumbing to be able to detect this condition during compile time. If
the kernel doesn't support the kuid then just fall back to directly
accessing the respective struct inode's members
Signed-off-by: Nikolay Borisov <n.borisov.lkml@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4685
Issue #227
Currently there is an issue where metaslab_fastwrite_unmark() unmarks
fastwrites on vdev_t's that have never had fastwrites marked on them.
The 'fastwrite mark' is essentially a count of outstanding bytes that
will be written to a vdev and is used in syncing context. The problem
stems from the fact that the vdev_pending_fastwrite field is not being
transferred over when replacing a top-level vdev. As a result, the
metaslab is marked for fastwrite on the old vdev and unmarked on the
new one, which brings the fastwrite count below zero. This fix simply
assigns vdev_pending_fastwrite from the old vdev to the new one so
this count is not lost.
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4267
Find the core file by using `/proc/sys/kernel/core_pattern`
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4874
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4329
A port of the Illumos Crypto Framework to a Linux kernel module (found
in module/icp). This is needed to do the actual encryption work. We cannot
use the Linux kernel's built in crypto api because it is only exported to
GPL-licensed modules. Having the ICP also means the crypto code can run on
any of the other kernels under OpenZFS. I ended up porting over most of the
internals of the framework, which means that porting over other API calls (if
we need them) should be fairly easy. Specifically, I have ported over the API
functions related to encryption, digests, macs, and crypto templates. The ICP
is able to use assembly-accelerated encryption on amd64 machines and AES-NI
instructions on Intel chips that support it. There are place-holder
directories for similar assembly optimizations for other architectures
(although they have not been written).
Signed-off-by: Tom Caputi <tcaputi@datto.com>
Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4329
When zfs_domount fails zsb will be freed, and its caller
mount_nodev/get_sb_nodev will do deactivate_locked_super and calls into
zfs_preumount.
In order to make sure we don't touch any nonexistent stuff, we must make sure
s_fs_info is NULL in the fail path so zfs_preumount can easily check that.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4867
Issue #4854
Print table with speed of methods for each implementation.
Last line describes contents of [fastest] selection.
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4860
- Implementation lock replaced with atomic variable
- Trailing whitespace is removed from user specified parameter, to enhance
experience when using commands that add newline, e.g. `echo`
- raidz_test: remove dependency on `getrusage()` and RUSAGE_THREAD, Issue #4813
- silence `cppcheck` in vdev_raidz, partial solution of Issue #1392
- Minor fixes and cleanups
- Enable use of original parity methods in [fastest] configuration.
New opaque original ops structure, representing native methods, is added
to supported raidz methods. Original parity methods are executed if selected
implementation has NULL fn pointer.
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4813
Issue #1392
Wait for iput_async before entering evict_inodes in
generic_shutdown_super. The reason we must finish before
evict_inodes is when lazytime is on, or when zfs_purgedir calls
zfs_zget, iput would bump i_count from 0 to 1. This would race
with the i_count check in evict_inodes. This means it could
destroy the inode while we are still using it.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4854
Commit 7f60329 removed several kstats which arc_summary.py read.
Remove these kstats from arc_summary.py in the same way this was
handled in FreeNAS.
FreeNAS-commit: https://github.com/freenas/freenas/commit/3901f73
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4695
In some cases, the compiler was not respecting the GNU aligned
attribute for stack variables in 35a76a0. This was resulting in
a segfault on CentOS 6.7 hosts using gcc 4.4.7-17. This issue
was fixed in gcc 4.6.
To prevent this from occurring, use unaligned loads and stores
for all stack and global memory references in the SSE optimized
Fletcher-4 code.
Disable zimport testing against master where this flaw exists:
TEST_ZIMPORT_VERSIONS="installed"
Signed-off-by: Tyler J. Stachecki <stachecki.tyler@gmail.com>
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4862
It is possible that the given DS may have hidden child (%recv)
datasets - "leftovers" resulting from the previously interrupted
'zfs receieve'. Try to remove the hidden child (%recv) and after
that try to remove the target dataset. If the hidden child
(%recv) does not exist the original error (EEXIST) will be returned.
Signed-off-by: Roman Strashkin <roman.strashkin@nexenta.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4818
Builds off of 1eeb4562 (Implementation of AVX2 optimized Fletcher-4)
This commit adds another implementation of the Fletcher-4 algorithm.
It is automatically selected at module load if it benchmarks higher
than all other available implementations.
The module benchmark was also amended to analyze the performance of
the byteswap-ed version of Fletcher-4, as well as the non-byteswaped
version. The average performance of the two is used to select the
the fastest implementation available on the host system.
Adds a pair of fields to an existing zcommon module parameter:
- zfs_fletcher_4_impl (str)
"sse2" - new SSE2 implementation if available
"ssse3" - new SSSE3 implementation if available
Signed-off-by: Tyler J. Stachecki <stachecki.tyler@gmail.com>
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4789
A mostly mechanical change, taking into account i_nlink is 32 bits vs ZFS's
64 bit on-disk link count.
We revert "xattr dir doesn't get purged during iput" (ddae16a) as this is a
more Linux-integrated fix for the same issue.
In addition, setting the initial link count on a new node has been changed
from setting one less than required in zfs_mknode() then incrementing to the
correct count in zfs_link_create() (which was somewhat bizarre in the first
place), to setting the correct count in zfs_mknode() and not incrementing it
in zfs_link_create(). This both means we no longer set the link count in
sa_bulk_update() twice (once for the initial incorrect count then again for
the correct count), as well as adhering to the Linux requirement of not
incrementing a zero link count without I_LINKABLE (see linux commit
f4e0c30c).
Signed-off-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Closes#4838
Issue #227
Dropping DBUF_HASH_MUTEX when walking the hash list is unsafe. The dbuf
can be freed at any time.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4846
In arc_buf_info(), the arc_buf_t may have no header. If not, don't try
to fetch the arc buffer stats and instead just zero them.
The null dereferences were observed while accessing the dbuf kstat with
awk on a system in which millions of small files were being created in
order to overflow the system's metadata limit.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Closes#4837
Creating the pool in a striped rather than mirrored configuration
provides enough space for all upgrade tests to run. Test case
zpool_upgrade_007_pos still fails and must be investigated so
it has been left disabled.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4852
zfs_ioc_recv_impl() is changed to always allocate the 'errors'
nvlist, its callers are responsible for freeing it.
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4829
The following scenario can result in garbage in the dn_spill field.
The db->db_blkptr must be set to NULL when DNODE_FLAG_SPILL_BLKPTR
is clear to ensure the dn_spill field is cleared.
Current txg = A.
* A new spill buffer is created. Its dbuf is initialized with
db_blkptr = NULL and it's dirtied.
Current txg = B.
* The spill buffer is modified. It's marked as dirty in this txg.
* Additional changes make the spill buffer unnecessary because the
xattr fits into the bonus buffer, so it's removed. The dbuf is
undirtied in this txg, but it's still referenced and cannot be
destroyed.
Current txg = C.
* Starts syncing of txg A
* dbuf_sync_leaf() is called for the spill buffer. Since db_blkptr
is NULL, dbuf_check_blkptr() is called.
* The dbuf starts being written and it reaches the ready state
(not done yet).
* A new change makes the spill buffer necessary again.
sa_build_layouts() ends up calling dbuf_find() to locate the
dbuf. It finds the old dbuf because it has not been destroyed yet
(it will be destroyed when the previous write is done and there
are no more references). The old dbuf has db_blkptr != NULL.
* txg A write is complete and the dbuf released. However it's still
referenced, so it's not destroyed.
Current txg = D.
* Starts syncing of txg B
* dbuf_sync_leaf() is called for the bonus buffer. Its contents are
directly copied into the dnode, overwriting the blkptr area because,
in txg B, the bonus buffer was big enough to hold the entire xattr.
* At this point, the db_blkptr of the spill buffer used in txg C
gets corrupted.
Signed-off-by: Peng <peng.hse@xtaotech.com>
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#3937
zp->z_xattr_parent will pin the parent. This will cause huge issue
when unlink a file with xattr. Because the unlinked file is pinned, it
will never get purged immediately. And because of that, the xattr
stuff will never be marked as unlinked. So the whole unlinked stuff
will stay there until shrink cache or umount.
This change partially reverts e89260a. This is safe because only the
zp->z_xattr_parent optimization is removed, zpl_xattr_security_init()
is still called from the zpl outside the inode lock.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlop <chris@onthe.net.au>
Issue #4359
Issue #3508
Issue #4413
Issue #4827
We need to set inode->i_nlink to zero so iput will purge it. Without this, it
will get purged during shrink cache or umount, which would likely result in
deadlock due to zfs_zget waiting forever on its children which are in the
dispose_list of the same thread.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlop <chris@onthe.net.au>
Issue #4359
Issue #3508
Issue #4413
Issue #4827
When generation mismatch, it usually means the file pointed by the file handle
was deleted. We should return ESTALE to indicate this. We return ENOENT in
zfs_vget since zpl_fh_to_dentry will convert it to ESTALE.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4828
Allow accessing XATTR through export handle is a very bad idea. It
would allow user to write whatever they want in fields where they
otherwise could not.
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #4828
Certain ioctl operations will call get_zfs_sb, which will holds an active
count on sb without checking whether it's active or not. This will result
in use-after-free. We fix this by using atomic_inc_not_zero to make sure
we got an active sb.
P1 P2
--- ---
deactivate_locked_super(): s_active = 0
zfs_sb_hold()
->get_zfs_sb(): s_active = 1
->zpl_kill_sb()
-->zpl_put_super()
--->zfs_umount()
---->zfs_sb_free(zsb)
zfs_sb_rele(zsb)
Signed-off-by: Chunwei Chen <david.chen@osnexus.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
If compiled with -O0, gcc doesn't do any stack frame coalescing
and -Wframe-larger-than=1024 is triggered in debug mode.
Starting with gcc 4.8, new opt level -Og is introduced for debugging, which
does not trigger this warning.
Fix bench zio size, using SPA_OLD_MAXBLOCKSHIFT
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4799
Adds support for resuming interrupted zfs send streams and include
all related send/recv bug fixes from upstream OpenZFS.
Unlike the upstream implementation this branch does not change
the existing ioctl interface. Instead a new ZFS_IOC_RECV_NEW ioctl
was added to support resuming zfs send streams. This was done by
applying the original upstream patch and then reverting the ioctl
changes in a follow up patch. For this reason there are a handful
on commits between the relevant patches on this branch which are
not interoperable. This was done to make it easier to extract
the new ZFS_IOC_RECV_NEW and submit it upstream.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#4742
The fletcher_4_native() and fletcher_4_byteswap() functions may only
safely use the vectorized implementations when the buffer is 128-bit
aligned. This is because both the AVX2 and SSE implementations process
four 32-bit words per iterations. Fallback to the scalar implementation
which only processes a single 32-bit word for unaligned buffers.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Gvozden Neskovic <neskovic@gmail.com>
Issue #4330
Reviewed by: Prakash Surya <prakash.surya@delphix.com>
Reviewed by: Dan Kimmel <dan.kimmel@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Yuri Pankov <yuri.pankov@nexenta.com>
Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Calling dsl_dataset_name on a dataset with a 256 byte buffer is asking
for trouble. We should check every dataset on import, using a 1024 byte
buffer and checking each time to see if the dataset's new name is longer
than 256 bytes.
OpenZFS-issue: https://www.illumos.org/issues/6876
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/ca8674e
Adds ZFS_IOC_RECV_NEW for resumable streams and preserves the legacy
ZFS_IOC_RECV user/kernel interface. The new interface supports all
stream options but is currently only used for resumable streams.
This way updated user space utilities will interoperate with older
kernel modules.
ZFS_IOC_RECV_NEW is modeled after the existing ZFS_IOC_SEND_NEW
handler. Non-Linux OpenZFS platforms have opted to change the
legacy interface in an incompatible fashion instead of adding a
new ioctl.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Authored by: Dan McDonald <danmcd@omniti.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Yuri Pankov <yuri.pankov@nexenta.com>
Reviewed by: Toomas Soome <tsoome@me.com>
Approved by: Gordon Ross <gwr@nexenta.com>
Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/6562
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/5f7a8e6
Authored by: Dan McDonald <danmcd@omniti.com>
Reviewed by: John Kennedy <john.kennedy@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Gordon Ross <gordon.ross@nexenta.com>
Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/4986
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/5878fad
2605 want to resume interrupted zfs send
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: Richard Elling <Richard.Elling@RichardElling.com>
Reviewed by: Xin Li <delphij@freebsd.org>
Reviewed by: Arne Jansen <sensille@gmx.net>
Approved by: Dan McDonald <danmcd@omniti.com>
Ported-by: kernelOfTruth <kerneloftruth@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/2605
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/9c3fd12
6980 6902 causes zfs send to break due to 32-bit/64-bit struct mismatch
Reviewed by: Paul Dagnelie <pcd@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Robert Mustacchi <rm@joyent.com>
Ported by: Brian Behlendorf <behlendorf1@llnl.gov>
OpenZFS-issue: https://www.illumos.org/issues/6980
OpenZFS-commit: https://github.com/openzfs/openzfs/commit/ea4a67f
Porting notes:
- All rsend and snapshop tests enabled and updated for Linux.
- Fix misuse of input argument in traverse_visitbp().
- Fix ISO C90 warnings and errors.
- Fix gcc 'missing braces around initializer' in
'struct send_thread_arg to_arg =' warning.
- Replace 4 argument fletcher_4_native() with 3 argument version,
this change was made in OpenZFS 4185 which has not been ported.
- Part of the sections for 'zfs receive' and 'zfs send' was
rewritten and reordered to approximate upstream.
- Fix mktree xattr creation, 'user.' prefix required.
- Minor fixes to newly enabled test cases
- Long holds for volumes allowed during receive for minor registration.
Flag 20 was used in OpenZFS as DMU_BACKUP_FEATURE_RESUMING. The
DMU_BACKUP_FEATURE_LARGE_DNODE flag must be shifted to 21 and
then reserved in the upstream OpenZFS implementation.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Ned Bass <bass6@llnl.gov>
Closes#4795
Justification
-------------
This feature adds support for variable length dnodes. Our motivation is
to eliminate the overhead associated with using spill blocks. Spill
blocks are used to store system attribute data (i.e. file metadata) that
does not fit in the dnode's bonus buffer. By allowing a larger bonus
buffer area the use of a spill block can be avoided. Spill blocks
potentially incur an additional read I/O for every dnode in a dnode
block. As a worst case example, reading 32 dnodes from a 16k dnode block
and all of the spill blocks could issue 33 separate reads. Now suppose
those dnodes have size 1024 and therefore don't need spill blocks. Then
the worst case number of blocks read is reduced to from 33 to two--one
per dnode block. In practice spill blocks may tend to be co-located on
disk with the dnode blocks so the reduction in I/O would not be this
drastic. In a badly fragmented pool, however, the improvement could be
significant.
ZFS-on-Linux systems that make heavy use of extended attributes would
benefit from this feature. In particular, ZFS-on-Linux supports the
xattr=sa dataset property which allows file extended attribute data
to be stored in the dnode bonus buffer as an alternative to the
traditional directory-based format. Workloads such as SELinux and the
Lustre distributed filesystem often store enough xattr data to force
spill bocks when xattr=sa is in effect. Large dnodes may therefore
provide a performance benefit to such systems.
Other use cases that may benefit from this feature include files with
large ACLs and symbolic links with long target names. Furthermore,
this feature may be desirable on other platforms in case future
applications or features are developed that could make use of a
larger bonus buffer area.
Implementation
--------------
The size of a dnode may be a multiple of 512 bytes up to the size of
a dnode block (currently 16384 bytes). A dn_extra_slots field was
added to the current on-disk dnode_phys_t structure to describe the
size of the physical dnode on disk. The 8 bits for this field were
taken from the zero filled dn_pad2 field. The field represents how
many "extra" dnode_phys_t slots a dnode consumes in its dnode block.
This convention results in a value of 0 for 512 byte dnodes which
preserves on-disk format compatibility with older software.
Similarly, the in-memory dnode_t structure has a new dn_num_slots field
to represent the total number of dnode_phys_t slots consumed on disk.
Thus dn->dn_num_slots is 1 greater than the corresponding
dnp->dn_extra_slots. This difference in convention was adopted
because, unlike on-disk structures, backward compatibility is not a
concern for in-memory objects, so we used a more natural way to
represent size for a dnode_t.
The default size for newly created dnodes is determined by the value of
a new "dnodesize" dataset property. By default the property is set to
"legacy" which is compatible with older software. Setting the property
to "auto" will allow the filesystem to choose the most suitable dnode
size. Currently this just sets the default dnode size to 1k, but future
code improvements could dynamically choose a size based on observed
workload patterns. Dnodes of varying sizes can coexist within the same
dataset and even within the same dnode block. For example, to enable
automatically-sized dnodes, run
# zfs set dnodesize=auto tank/fish
The user can also specify literal values for the dnodesize property.
These are currently limited to powers of two from 1k to 16k. The
power-of-2 limitation is only for simplicity of the user interface.
Internally the implementation can handle any multiple of 512 up to 16k,
and consumers of the DMU API can specify any legal dnode value.
The size of a new dnode is determined at object allocation time and
stored as a new field in the znode in-memory structure. New DMU
interfaces are added to allow the consumer to specify the dnode size
that a newly allocated object should use. Existing interfaces are
unchanged to avoid having to update every call site and to preserve
compatibility with external consumers such as Lustre. The new
interfaces names are given below. The versions of these functions that
don't take a dnodesize parameter now just call the _dnsize() versions
with a dnodesize of 0, which means use the legacy dnode size.
New DMU interfaces:
dmu_object_alloc_dnsize()
dmu_object_claim_dnsize()
dmu_object_reclaim_dnsize()
New ZAP interfaces:
zap_create_dnsize()
zap_create_norm_dnsize()
zap_create_flags_dnsize()
zap_create_claim_norm_dnsize()
zap_create_link_dnsize()
The constant DN_MAX_BONUSLEN is renamed to DN_OLD_MAX_BONUSLEN. The
spa_maxdnodesize() function should be used to determine the maximum
bonus length for a pool.
These are a few noteworthy changes to key functions:
* The prototype for dnode_hold_impl() now takes a "slots" parameter.
When the DNODE_MUST_BE_FREE flag is set, this parameter is used to
ensure the hole at the specified object offset is large enough to
hold the dnode being created. The slots parameter is also used
to ensure a dnode does not span multiple dnode blocks. In both of
these cases, if a failure occurs, ENOSPC is returned. Keep in mind,
these failure cases are only possible when using DNODE_MUST_BE_FREE.
If the DNODE_MUST_BE_ALLOCATED flag is set, "slots" must be 0.
dnode_hold_impl() will check if the requested dnode is already
consumed as an extra dnode slot by an large dnode, in which case
it returns ENOENT.
* The function dmu_object_alloc() advances to the next dnode block
if dnode_hold_impl() returns an error for a requested object.
This is because the beginning of the next dnode block is the only
location it can safely assume to either be a hole or a valid
starting point for a dnode.
* dnode_next_offset_level() and other functions that iterate
through dnode blocks may no longer use a simple array indexing
scheme. These now use the current dnode's dn_num_slots field to
advance to the next dnode in the block. This is to ensure we
properly skip the current dnode's bonus area and don't interpret it
as a valid dnode.
zdb
---
The zdb command was updated to display a dnode's size under the
"dnsize" column when the object is dumped.
For ZIL create log records, zdb will now display the slot count for
the object.
ztest
-----
Ztest chooses a random dnodesize for every newly created object. The
random distribution is more heavily weighted toward small dnodes to
better simulate real-world datasets.
Unused bonus buffer space is filled with non-zero values computed from
the object number, dataset id, offset, and generation number. This
helps ensure that the dnode traversal code properly skips the interior
regions of large dnodes, and that these interior regions are not
overwritten by data belonging to other dnodes. A new test visits each
object in a dataset. It verifies that the actual dnode size matches what
was stored in the ztest block tag when it was created. It also verifies
that the unused bonus buffer space is filled with the expected data
patterns.
ZFS Test Suite
--------------
Added six new large dnode-specific tests, and integrated the dnodesize
property into existing tests for zfs allow and send/recv.
Send/Receive
------------
ZFS send streams for datasets containing large dnodes cannot be received
on pools that don't support the large_dnode feature. A send stream with
large dnodes sets a DMU_BACKUP_FEATURE_LARGE_DNODE flag which will be
unrecognized by an incompatible receiving pool so that the zfs receive
will fail gracefully.
While not implemented here, it may be possible to generate a
backward-compatible send stream from a dataset containing large
dnodes. The implementation may be tricky, however, because the send
object record for a large dnode would need to be resized to a 512
byte dnode, possibly kicking in a spill block in the process. This
means we would need to construct a new SA layout and possibly
register it in the SA layout object. The SA layout is normally just
sent as an ordinary object record. But if we are constructing new
layouts while generating the send stream we'd have to build the SA
layout object dynamically and send it at the end of the stream.
For sending and receiving between pools that do support large dnodes,
the drr_object send record type is extended with a new field to store
the dnode slot count. This field was repurposed from unused padding
in the structure.
ZIL Replay
----------
The dnode slot count is stored in the uppermost 8 bits of the lr_foid
field. The bits were unused as the object id is currently capped at
48 bits.
Resizing Dnodes
---------------
It should be possible to resize a dnode when it is dirtied if the
current dnodesize dataset property differs from the dnode's size, but
this functionality is not currently implemented. Clearly a dnode can
only grow if there are sufficient contiguous unused slots in the
dnode block, but it should always be possible to shrink a dnode.
Growing dnodes may be useful to reduce fragmentation in a pool with
many spill blocks in use. Shrinking dnodes may be useful to allow
sending a dataset to a pool that doesn't support the large_dnode
feature.
Feature Reference Counting
--------------------------
The reference count for the large_dnode pool feature tracks the
number of datasets that have ever contained a dnode of size larger
than 512 bytes. The first time a large dnode is created in a dataset
the dataset is converted to an extensible dataset. This is a one-way
operation and the only way to decrement the feature count is to
destroy the dataset, even if the dataset no longer contains any large
dnodes. The complexity of reference counting on a per-dnode basis was
too high, so we chose to track it on a per-dataset basis similarly to
the large_block feature.
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#3542
Only attempt to backfill lower metadnode object numbers if at least
4096 objects have been freed since the last rescan, and at most once
per transaction group. This avoids a pathology in dmu_object_alloc()
that caused O(N^2) behavior for create-heavy workloads and
substantially improves object creation rates. As summarized by
@mahrens in #4636:
"Normally, the object allocator simply checks to see if the next
object is available. The slow calls happened when dmu_object_alloc()
checks to see if it can backfill lower object numbers. This happens
every time we move on to a new L1 indirect block (i.e. every 32 *
128 = 4096 objects). When re-checking lower object numbers, we use
the on-disk fill count (blkptr_t:blk_fill) to quickly skip over
indirect blocks that don’t have enough free dnodes (defined as an L2
with at least 393,216 of 524,288 dnodes free). Therefore, we may
find that a block of dnodes has a low (or zero) fill count, and yet
we can’t allocate any of its dnodes, because they've been allocated
in memory but not yet written to disk. In this case we have to hold
each of the dnodes and then notice that it has been allocated in
memory.
The end result is that allocating N objects in the same TXG can
require CPU usage proportional to N^2."
Add a tunable dmu_rescan_dnode_threshold to define the number of
objects that must be freed before a rescan is performed. Don't bother
to export this as a module option because testing doesn't show a
compelling reason to change it. The vast majority of the performance
gain comes from limit the rescan to at most once per TXG.
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Brian Behlendorf
Gvozden Neskovic
Tim Chase
Nikolay Borisov
Tom Caputi
Chunwei Chen
Tyler J. Stachecki
Roman Strashkin
Chris Dunlop
Peng
Paul Dagnelie
Igor Kozhukhov
Dan McDonald
Eli Rosenthal
Andrew Stormont
Matthew Ahrens
Ned Bass