linux.git/fs/nfs/write.c, branch v5.17

nfs: Implement cache I/O by accessing the cache directly

2022-01-10T11:53:25+00:00

Move NFS to using fscache DIO API instead of the old upstream I/O API as
that has been removed.  This is a stopgap solution as the intention is that
at sometime in the future, the cache will move to using larger blocks and
won't be able to store individual pages in order to deal with the potential
for data corruption due to the backing filesystem being able insert/remove
bridging blocks of zeros into its extent list[1].

NFS then reads and writes cache pages synchronously and one page at a time.

The preferred change would be to use the netfs lib, but the new I/O API can
be used directly.  It's just that as the cache now needs to track data for
itself, caching blocks may exceed page size...

This code is somewhat borrowed from my "fallback I/O" patchset[2].

Changes
=======
ver #3:
 - Restore lost =n fallback for nfs_fscache_release_page()[2].

Signed-off-by: David Howells 
Tested-by: Dave Wysochanski 
Acked-by: Jeff Layton 
cc: Trond Myklebust 
cc: Anna Schumaker 
cc: linux-nfs@vger.kernel.org
cc: linux-cachefs@redhat.com
Link: https://lore.kernel.org/r/YO17ZNOcq+9PajfQ@mit.edu [1]
Link: https://lore.kernel.org/r/202112100957.2oEDT20W-lkp@intel.com/ [2]
Link: https://lore.kernel.org/r/163189108292.2509237.12615909591150927232.stgit@warthog.procyon.org.uk/ [2]
Link: https://lore.kernel.org/r/163906981318.143852.17220018647843475985.stgit@warthog.procyon.org.uk/ # v2
Link: https://lore.kernel.org/r/163967184451.1823006.6450645559828329590.stgit@warthog.procyon.org.uk/ # v3
Link: https://lore.kernel.org/r/164021577632.640689.11069627070150063812.stgit@warthog.procyon.org.uk/ # v4

nfs: Convert to new fscache volume/cookie API

2022-01-10T11:53:25+00:00

Change the nfs filesystem to support fscache's indexing rewrite and
reenable caching in nfs.

The following changes have been made:

 (1) The fscache_netfs struct is no more, and there's no need to register
     the filesystem as a whole.

 (2) The session cookie is now an fscache_volume cookie, allocated with
     fscache_acquire_volume().  That takes three parameters: a string
     representing the "volume" in the index, a string naming the cache to
     use (or NULL) and a u64 that conveys coherency metadata for the
     volume.

     For nfs, I've made it render the volume name string as:

        "nfs,,,,,,*<,param>[,]"

 (3) The fscache_cookie_def is no more and needed information is passed
     directly to fscache_acquire_cookie().  The cache no longer calls back
     into the filesystem, but rather metadata changes are indicated at
     other times.

     fscache_acquire_cookie() is passed the same keying and coherency
     information as before.

 (4) fscache_enable/disable_cookie() have been removed.

     Call fscache_use_cookie() and fscache_unuse_cookie() when a file is
     opened or closed to prevent a cache file from being culled and to keep
     resources to hand that are needed to do I/O.

     If a file is opened for writing, we invalidate it with
     FSCACHE_INVAL_DIO_WRITE in lieu of doing writeback to the cache,
     thereby making it cease caching until all currently open files are
     closed.  This should give the same behaviour as the uptream code.
     Making the cache store local modifications isn't straightforward for
     NFS, so that's left for future patches.

 (5) fscache_invalidate() now needs to be given uptodate auxiliary data and
     a file size.  It also takes a flag to indicate if this was due to a
     DIO write.

 (6) Call nfs_fscache_invalidate() with FSCACHE_INVAL_DIO_WRITE on a file
     to which a DIO write is made.

 (7) Call fscache_note_page_release() from nfs_release_page().

 (8) Use a killable wait in nfs_vm_page_mkwrite() when waiting for
     PG_fscache to be cleared.

 (9) The functions to read and write data to/from the cache are stubbed out
     pending a conversion to use netfslib.

Changes
=======
ver #3:
 - Added missing =n fallback for nfs_fscache_release_file()[1][2].

ver #2:
 - Use gfpflags_allow_blocking() rather than using flag directly.
 - fscache_acquire_volume() now returns errors.
 - Remove NFS_INO_FSCACHE as it's no longer used.
 - Need to unuse a cookie on file-release, not inode-clear.

Signed-off-by: Dave Wysochanski 
Co-developed-by: David Howells 
Signed-off-by: David Howells 
Tested-by: Dave Wysochanski 
Acked-by: Jeff Layton 
cc: Trond Myklebust 
cc: Anna Schumaker 
cc: linux-nfs@vger.kernel.org
cc: linux-cachefs@redhat.com
Link: https://lore.kernel.org/r/202112100804.nksO8K4u-lkp@intel.com/ [1]
Link: https://lore.kernel.org/r/202112100957.2oEDT20W-lkp@intel.com/ [2]
Link: https://lore.kernel.org/r/163819668938.215744.14448852181937731615.stgit@warthog.procyon.org.uk/ # v1
Link: https://lore.kernel.org/r/163906979003.143852.2601189243864854724.stgit@warthog.procyon.org.uk/ # v2
Link: https://lore.kernel.org/r/163967182112.1823006.7791504655391213379.stgit@warthog.procyon.org.uk/ # v3
Link: https://lore.kernel.org/r/164021575950.640689.12069642327533368467.stgit@warthog.procyon.org.uk/ # v4

NFS: Remove redundant call to __set_page_dirty_nobuffers

2021-10-21T21:11:41+00:00

Remove a redundant call in nfs_updatepage(). nfs_writepage_setup() will
have already called nfs_mark_request_dirty() on success.

Signed-off-by: Trond Myklebust

SUNRPC: Trace calls to .rpc_call_done

2021-10-20T22:09:54+00:00

Introduce a single tracepoint that can replace simple dprintk call
sites in upper layer "rpc_call_done" callbacks. Example:

   kworker/u24:2-1254  [001]   771.026677: rpc_stats_latency:    task:00000001@00000002 xid=0x16a6f3c0 rpcbindv2 GETPORT backlog=446 rtt=101 execute=555
   kworker/u24:2-1254  [001]   771.026677: rpc_task_call_done:   task:00000001@00000002 flags=ASYNC|DYNAMIC|SOFT|SOFTCONN|SENT runstate=RUNNING|ACTIVE status=0 action=rpcb_getport_done
   kworker/u24:2-1254  [001]   771.026678: rpcb_setport:         task:00000001@00000002 status=0 port=20048

Signed-off-by: Chuck Lever 
Signed-off-by: Trond Myklebust

NFS: Fix up commit deadlocks

2021-10-20T22:09:45+00:00

If O_DIRECT bumps the commit_info rpcs_out field, then that could lead
to fsync() hangs. The fix is to ensure that O_DIRECT calls
nfs_commit_end().

Fixes: 723c921e7dfc ("sched/wait, fs/nfs: Convert wait_on_atomic_t() usage to the new wait_var_event() API")
Signed-off-by: Trond Myklebust

NFS: Fix deadlocks in nfs_scan_commit_list()

2021-10-10T09:05:54+00:00

Partially revert commit 2ce209c42c01 ("NFS: Wait for requests that are
locked on the commit list"), since it can lead to deadlocks between
commit requests and nfs_join_page_group().
For now we should assume that any locked requests on the commit list are
either about to be removed and committed by another task, or the writes
they describe are about to be retransmitted. In either case, we should
not need to worry.

Fixes: 2ce209c42c01 ("NFS: Wait for requests that are locked on the commit list")
Signed-off-by: Trond Myklebust

NFS: Instrument i_size_write()

2021-10-10T09:05:54+00:00

Generate a trace event whenever the NFS client modifies the size of
a file. These new events aid troubleshooting workloads that trigger
races around size updates.

There are four new trace points, all named nfs_size_something so
they are easy to grep for or enable as a group with a single glob.

Size updated on the server:

  kworker/u24:10-194   [010]   369.939174: nfs_size_update:      fileid=00:28:2 fhandle=0x36fbbe51 version=1752899344277980615 cursize=250471 newsize=172083

Server-side size update reported via NFSv3 WCC attributes:

             fsx-1387  [006]   380.760686: nfs_size_wcc:         fileid=00:28:2 fhandle=0x36fbbe51 version=1752899355909932456 cursize=146792 newsize=171216

File has been truncated locally:

             fsx-1387  [007]   369.437421: nfs_size_truncate:    fileid=00:28:2 fhandle=0x36fbbe51 version=1752899231200117272 cursize=215244 newsize=0

File has been extended locally:

             fsx-1387  [007]   369.439213: nfs_size_grow:        fileid=00:28:2 fhandle=0x36fbbe51 version=1752899343704248410 cursize=258048 newsize=262144

Signed-off-by: Chuck Lever 
Signed-off-by: Trond Myklebust

NFS: Fix up nfs_ctx_key_to_expire()

2021-10-04T00:49:05+00:00

If the cached credential exists but doesn't have any expiration callback
then exit early.
Fix up atomicity issues when replacing the credential with a new one
since the existing code could lead to refcount leaks.

Signed-off-by: Trond Myklebust

NFSv4.1 identify and mark RPC tasks that can move between transports

2021-07-08T18:03:24+00:00

In preparation for when we can re-try a task on a different transport,
identify and mark such RPC tasks as moveable. Only 4.1+ operarations can
be re-tried on a different transport.

Signed-off-by: Olga Kornievskaia 
Signed-off-by: Trond Myklebust

NFSv4: Don't modify the change attribute cached in the inode

2021-04-13T14:04:05+00:00

When the client is caching data and a write delegation is held, then the
server may send a CB_GETATTR to query the attributes. When this happens,
the client is supposed to bump the change attribute value that it
returns if it holds cached data.
However that process uses a value that is stored in the delegation. We
do not want to bump the change attribute held in the inode.

Signed-off-by: Trond Myklebust