nfs: fix UAF in direct writes

[ Upstream commit 17f46b803d4f23c66cacce81db35fef3adb8f2af ]

In production we have been hitting the following warning consistently

------------[ cut here ]-----------

nfs: fix UAF in direct writes

[ Upstream commit 17f46b803d4f23c66cacce81db35fef3adb8f2af ]

In production we have been hitting the following warning consistently

------------[ cut here ]------------
refcount_t: underflow; use-after-free.
WARNING: CPU: 17 PID: 1800359 at lib/refcount.c:28 refcount_warn_saturate+0x9c/0xe0
Workqueue: nfsiod nfs_direct_write_schedule_work [nfs]
RIP: 0010:refcount_warn_saturate+0x9c/0xe0
PKRU: 55555554
Call Trace:
<TASK>
? __warn+0x9f/0x130
? refcount_warn_saturate+0x9c/0xe0
? report_bug+0xcc/0x150
? handle_bug+0x3d/0x70
? exc_invalid_op+0x16/0x40
? asm_exc_invalid_op+0x16/0x20
? refcount_warn_saturate+0x9c/0xe0
nfs_direct_write_schedule_work+0x237/0x250 [nfs]
process_one_work+0x12f/0x4a0
worker_thread+0x14e/0x3b0
? ZSTD_getCParams_internal+0x220/0x220
kthread+0xdc/0x120
? __btf_name_valid+0xa0/0xa0
ret_from_fork+0x1f/0x30

This is because we're completing the nfs_direct_request twice in a row.

The source of this is when we have our commit requests to submit, we
process them and send them off, and then in the completion path for the
commit requests we have

if (nfs_commit_end(cinfo.mds))
nfs_direct_write_complete(dreq);

However since we're submitting asynchronous requests we sometimes have
one that completes before we submit the next one, so we end up calling
complete on the nfs_direct_request twice.

The only other place we use nfs_generic_commit_list() is in
__nfs_commit_inode, which wraps this call in a

nfs_commit_begin();
nfs_commit_end();

Which is a common pattern for this style of completion handling, one
that is also repeated in the direct code with get_dreq()/put_dreq()
calls around where we process events as well as in the completion paths.

Fix this by using the same pattern for the commit requests.

Before with my 200 node rocksdb stress running this warning would pop
every 10ish minutes. With my patch the stress test has been running for
several hours without popping.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Cc: stable@vger.kernel.org
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

show more ...

nfs: fix UAF in direct writes

[ Upstream commit 17f46b803d4f23c66cacce81db35fef3adb8f2af ]

In production we have been hitting the following warning consistently

------------[ cut here ]-----------

nfs: fix UAF in direct writes

[ Upstream commit 17f46b803d4f23c66cacce81db35fef3adb8f2af ]

In production we have been hitting the following warning consistently

------------[ cut here ]------------
refcount_t: underflow; use-after-free.
WARNING: CPU: 17 PID: 1800359 at lib/refcount.c:28 refcount_warn_saturate+0x9c/0xe0
Workqueue: nfsiod nfs_direct_write_schedule_work [nfs]
RIP: 0010:refcount_warn_saturate+0x9c/0xe0
PKRU: 55555554
Call Trace:
<TASK>
? __warn+0x9f/0x130
? refcount_warn_saturate+0x9c/0xe0
? report_bug+0xcc/0x150
? handle_bug+0x3d/0x70
? exc_invalid_op+0x16/0x40
? asm_exc_invalid_op+0x16/0x20
? refcount_warn_saturate+0x9c/0xe0
nfs_direct_write_schedule_work+0x237/0x250 [nfs]
process_one_work+0x12f/0x4a0
worker_thread+0x14e/0x3b0
? ZSTD_getCParams_internal+0x220/0x220
kthread+0xdc/0x120
? __btf_name_valid+0xa0/0xa0
ret_from_fork+0x1f/0x30

This is because we're completing the nfs_direct_request twice in a row.

The source of this is when we have our commit requests to submit, we
process them and send them off, and then in the completion path for the
commit requests we have

if (nfs_commit_end(cinfo.mds))
nfs_direct_write_complete(dreq);

However since we're submitting asynchronous requests we sometimes have
one that completes before we submit the next one, so we end up calling
complete on the nfs_direct_request twice.

The only other place we use nfs_generic_commit_list() is in
__nfs_commit_inode, which wraps this call in a

nfs_commit_begin();
nfs_commit_end();

Which is a common pattern for this style of completion handling, one
that is also repeated in the direct code with get_dreq()/put_dreq()
calls around where we process events as well as in the completion paths.

Fix this by using the same pattern for the commit requests.

Before with my 200 node rocksdb stress running this warning would pop
every 10ish minutes. With my patch the stress test has been running for
several hours without popping.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Cc: stable@vger.kernel.org
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

show more ...

NFSv3: handle out-of-order write replies.

NFSv3 includes pre/post wcc attributes which allow the client to
determine if all changes to the file have been made by the client
itself, or if any might h

NFSv3: handle out-of-order write replies.

NFSv3 includes pre/post wcc attributes which allow the client to
determine if all changes to the file have been made by the client
itself, or if any might have been made by some other client.

If there are gaps in the pre/post ctime sequence it must be assumed that
some other client changed the file in that gap and the local cache must
be suspect. The next time the file is opened the cache should be
invalidated.

Since Commit 1c341b777501 ("NFS: Add deferred cache invalidation for
close-to-open consistency violations") in linux 5.3 the Linux client has
been triggering this invalidation. The chunk in nfs_update_inode() in
particularly triggers.

Unfortunately Linux NFS assumes that all replies will be processed in
the order sent, and will arrive in the order processed. This is not
true in general. Consequently Linux NFS might ignore the wcc info in a
WRITE reply because the reply is in response to a WRITE that was sent
before some other request for which a reply has already been seen. This
is detected by Linux using the gencount tests in nfs_inode_attr_cmp().

Also, when the gencount tests pass it is still possible that the request
were processed on the server in a different order, and a gap seen in
the ctime sequence might be filled in by a subsequent reply, so gaps
should not immediately trigger delayed invalidation.

The net result is that writing to a server and then reading the file
back can result in going to the server for the read rather than serving
it from cache - all because a couple of replies arrived out-of-order.
This is a performance regression over kernels before 5.3, though the
change in 5.3 is a correctness improvement.

This has been seen with Linux writing to a Netapp server which
occasionally re-orders requests. In testing the majority of requests
were in-order, but a few (maybe 2 or three at a time) could be
re-ordered.

This patch addresses the problem by recording any gaps seen in the
pre/post ctime sequence and not triggering invalidation until either
there are too many gaps to fit in the table, or until there are no more
active writes and the remaining gaps cannot be resolved.

We allocate a table of 16 gaps on demand. If the allocation fails we
revert to current behaviour which is of little cost as we are unlikely
to be able to cache the writes anyway.

In the table we store "start->end" pair when iversion is updated and
"end<-start" pairs pre/post pairs reported by the server. Usually these
exactly cancel out and so nothing is stored. When there are
out-of-order replies we do store gaps and these will eventually be
cancelled against later replies when this client is the only writer.

If the final write is out-of-order there may be one gap remaining when
the file is closed. This will be noticed and if there is precisely on
gap and if the iversion can be advanced to match it, then we do so.

This patch makes no attempt to handle directories correctly. The same
problem potentially exists in the out-of-order replies to create/unlink
requests can cause future lookup requires to be sent to the server
unnecessarily. A similar scheme using the same primitives could be used
to notice and handle out-of-order replies.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

show more ...

NFS: Remove fscache specific trace points and NFS_INO_FSCACHE bit

The NFS specific trace points are no longer needed as tracing is well
covered by netfs and fscache.

Signed-off-by: Dave Wysochanski

NFS: Remove fscache specific trace points and NFS_INO_FSCACHE bit

The NFS specific trace points are no longer needed as tracing is well
covered by netfs and fscache.

Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Tested-by: Daire Byrne <daire@dneg.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

show more ...

NFS: Configure support for netfs when NFS fscache is configured

As first steps for support of the netfs library when NFS_FSCACHE is
configured, add NETFS_SUPPORT to Kconfig and add the required netf

NFS: Configure support for netfs when NFS fscache is configured

As first steps for support of the netfs library when NFS_FSCACHE is
configured, add NETFS_SUPPORT to Kconfig and add the required netfs_inode
into struct nfs_inode.

Using netfs requires we move the VFS inode structure to be stored
inside struct netfs_inode, along with the fscache_cookie.
Thus, if NFS_FSCACHE is configured, place netfs_inode inside an
anonymous union so the vfs_inode memory is the same and we do
not need to modify other non-fscache areas of NFS.
In addition, inside the NFS fscache code, use the new helpers,
netfs_inode() and netfs_i_cookie() helpers, and remove our own
helper, nfs_i_fscache().

Later patches will convert NFS fscache to fully use netfs.

Signed-off-by: Dave Wysochanski <dwysocha@redhat.com>
Tested-by: Daire Byrne <daire@dneg.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

show more ...

NFS: Remove unused function nfs_wb_page()

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

NFS: Convert buffered writes to use folios

Mostly mechanical conversion of struct page and functions into struct
folio equivalents.
The lack of support for folios in write_cache_pages(), means we st

NFS: Convert buffered writes to use folios

Mostly mechanical conversion of struct page and functions into struct
folio equivalents.
The lack of support for folios in write_cache_pages(), means we still
only support order 0 folio allocations. However the rest of the
writeback code should now be ready for order n > 0.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

show more ...

NFS: Add a helper nfs_wb_folio()

...and use it in nfs_launder_folio().

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

fs: port ->permission() to pass mnt_idmap

Convert to struct mnt_idmap.

Last cycle we merged the necessary infrastructure in
256c8aed2b42 ("fs: introduce dedicated idmap type for mounts").
This is j

fs: port ->permission() to pass mnt_idmap

Convert to struct mnt_idmap.

Last cycle we merged the necessary infrastructure in
256c8aed2b42 ("fs: introduce dedicated idmap type for mounts").
This is just the conversion to struct mnt_idmap.

Currently we still pass around the plain namespace that was attached to a
mount. This is in general pretty convenient but it makes it easy to
conflate namespaces that are relevant on the filesystem with namespaces
that are relevent on the mount level. Especially for non-vfs developers
without detailed knowledge in this area this can be a potential source for
bugs.

Once the conversion to struct mnt_idmap is done all helpers down to the
really low-level helpers will take a struct mnt_idmap argument instead of
two namespace arguments. This way it becomes impossible to conflate the two
eliminating the possibility of any bugs. All of the vfs and all filesystems
only operate on struct mnt_idmap.

Acked-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>

show more ...

fs: port ->getattr() to pass mnt_idmap

Convert to struct mnt_idmap.

Last cycle we merged the necessary infrastructure in
256c8aed2b42 ("fs: introduce dedicated idmap type for mounts").
This is just

fs: port ->getattr() to pass mnt_idmap

Convert to struct mnt_idmap.

Last cycle we merged the necessary infrastructure in
256c8aed2b42 ("fs: introduce dedicated idmap type for mounts").
This is just the conversion to struct mnt_idmap.

Currently we still pass around the plain namespace that was attached to a
mount. This is in general pretty convenient but it makes it easy to
conflate namespaces that are relevant on the filesystem with namespaces
that are relevent on the mount level. Especially for non-vfs developers
without detailed knowledge in this area this can be a potential source for
bugs.

Once the conversion to struct mnt_idmap is done all helpers down to the
really low-level helpers will take a struct mnt_idmap argument instead of
two namespace arguments. This way it becomes impossible to conflate the two
eliminating the possibility of any bugs. All of the vfs and all filesystems
only operate on struct mnt_idmap.

Acked-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>

show more ...

fs: port ->setattr() to pass mnt_idmap

Convert to struct mnt_idmap.

Last cycle we merged the necessary infrastructure in
256c8aed2b42 ("fs: introduce dedicated idmap type for mounts").
This is just

fs: port ->setattr() to pass mnt_idmap

Convert to struct mnt_idmap.

Last cycle we merged the necessary infrastructure in
256c8aed2b42 ("fs: introduce dedicated idmap type for mounts").
This is just the conversion to struct mnt_idmap.

Currently we still pass around the plain namespace that was attached to a
mount. This is in general pretty convenient but it makes it easy to
conflate namespaces that are relevant on the filesystem with namespaces
that are relevent on the mount level. Especially for non-vfs developers
without detailed knowledge in this area this can be a potential source for
bugs.

Once the conversion to struct mnt_idmap is done all helpers down to the
really low-level helpers will take a struct mnt_idmap argument instead of
two namespace arguments. This way it becomes impossible to conflate the two
eliminating the possibility of any bugs. All of the vfs and all filesystems
only operate on struct mnt_idmap.

Acked-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner (Microsoft) <brauner@kernel.org>

show more ...

NFS: Clear the file access cache upon login

POSIX typically only refreshes the user's supplementary group
information upon login. Since NFS servers may often refresh their
concept of the user supple

NFS: Clear the file access cache upon login

POSIX typically only refreshes the user's supplementary group
information upon login. Since NFS servers may often refresh their
concept of the user supplementary group membership at their own cadence,
it is possible for the NFS client's access cache to become stale due to
the user's group membership changing on the server after the user has
already logged in on the client.
While it is reasonable to expect that such group membership changes are
rare, and that we do not want to optimise the cache to accommodate them,
it is also not unreasonable for the user to expect that if they log out
and log back in again, that the staleness would clear up.

Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Tested-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: Cleanup to remove unused flag NFS_CONTEXT_RESEND_WRITES

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

NFS: Fix another fsync() issue after a server reboot

Currently, when the writeback code detects a server reboot, it redirties
any pages that were not committed to disk, and it sets the flag
NFS_CONT

NFS: Fix another fsync() issue after a server reboot

Currently, when the writeback code detects a server reboot, it redirties
any pages that were not committed to disk, and it sets the flag
NFS_CONTEXT_RESEND_WRITES in the nfs_open_context of the file descriptor
that dirtied the file. While this allows the file descriptor in question
to redrive its own writes, it violates the fsync() requirement that we
should be synchronising all writes to disk.
While the problem is infrequent, we do see corner cases where an
untimely server reboot causes the fsync() call to abandon its attempt to
sync data to disk and causing data corruption issues due to missed error
conditions or similar.

In order to tighted up the client's ability to deal with this situation
without introducing livelocks, add a counter that records the number of
times pages are redirtied due to a server reboot-like condition, and use
that in fsync() to redrive the sync to disk.

Fixes: 2197e9b06c22 ("NFS: Fix up fsync() when the server rebooted")
Cc: stable@vger.kernel.org
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: don't unhash dentry during unlink/rename

NFS unlink() (and rename over existing target) must determine if the
file is open, and must perform a "silly rename" instead of an unlink (or
before ren

NFS: don't unhash dentry during unlink/rename

NFS unlink() (and rename over existing target) must determine if the
file is open, and must perform a "silly rename" instead of an unlink (or
before rename) if it is. Otherwise the client might hold a file open
which has been removed on the server.

Consequently if it determines that the file isn't open, it must block
any subsequent opens until the unlink/rename has been completed on the
server.

This is currently achieved by unhashing the dentry. This forces any
open attempt to the slow-path for lookup which will block on i_rwsem on
the directory until the unlink/rename completes. A future patch will
change the VFS to only get a shared lock on i_rwsem for unlink, so this
will no longer work.

Instead we introduce an explicit interlock. A special value is stored
in dentry->d_fsdata while the unlink/rename is running and
->d_revalidate blocks while that value is present. When ->d_revalidate
unblocks, the dentry will be invalid. This closes the race
without requiring exclusion on i_rwsem.

d_fsdata is already used in two different ways.
1/ an IS_ROOT directory dentry might have a "devname" stored in
d_fsdata. Such a dentry doesn't have a name and so cannot be the
target of unlink or rename. For safety we check if an old devname
is still stored, and remove it if it is.
2/ a dentry with DCACHE_NFSFS_RENAMED set will have a 'struct
nfs_unlinkdata' stored in d_fsdata. While this is set maydelete()
will fail, so an unlink or rename will never proceed on such
a dentry.

Neither of these can be in effect when a dentry is the target of unlink
or rename. So we can expect d_fsdata to be NULL, and store a special
value ((void*)1) which is given the name NFS_FSDATA_BLOCKED to indicate
that any lookup will be blocked.

The d_count() is incremented under d_lock() when a lookup finds the
dentry, so we check d_count() is low, and set NFS_FSDATA_BLOCKED under
the same lock to avoid any races.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

nfs: rename nfs_direct_IO and use as ->swap_rw

The nfs_direct_IO() exists to support SWAP IO, but hasn't worked for a
while. We now need a ->swap_rw function which behaves slightly
differently, ret

nfs: rename nfs_direct_IO and use as ->swap_rw

The nfs_direct_IO() exists to support SWAP IO, but hasn't worked for a
while. We now need a ->swap_rw function which behaves slightly
differently, returning zero for success rather than a byte count.

So modify nfs_direct_IO accordingly, rename it, and use it as the
->swap_rw function.

Link: https://lkml.kernel.org/r/165119301493.15698.7491285551903597618.stgit@noble.brown
Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Geert Uytterhoeven <geert+renesas@glider.be> (on Renesas RSK+RZA1 with 32 MiB of SDRAM)
Cc: David Howells <dhowells@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

show more ...

nfs: Convert nfs to read_folio

This is a "weak" conversion which converts straight back to using pages.
A full conversion should be performed at some point, hopefully by
someone familiar with the fi

nfs: Convert nfs to read_folio

This is a "weak" conversion which converts straight back to using pages.
A full conversion should be performed at some point, hopefully by
someone familiar with the filesystem.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>

show more ...

NFS: nfsiod should not block forever in mempool_alloc()

The concern is that since nfsiod is sometimes required to kick off a
commit, it can get locked up waiting forever in mempool_alloc() instead
o

NFS: nfsiod should not block forever in mempool_alloc()

The concern is that since nfsiod is sometimes required to kick off a
commit, it can get locked up waiting forever in mempool_alloc() instead
of failing gracefully and leaving the commit until later.

Try to allocate from the slab first, with GFP_KERNEL | __GFP_NORETRY,
then fall back to a non-blocking attempt to allocate from the memory
pool.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

nfs: Convert from invalidatepage to invalidate_folio

Print the folio index instead of the pointer, since this is more
useful. We also don't need to use page_file_mapping() as we do not
invalidate s

nfs: Convert from invalidatepage to invalidate_folio

Print the folio index instead of the pointer, since this is more
useful. We also don't need to use page_file_mapping() as we do not
invalidate swapcache pages. Since this is the only caller of
nfs_wb_page_cancel(), convert it to nfs_wb_folio_cancel().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Tested-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Acked-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
Tested-by: Mike Marshall <hubcap@omnibond.com> # orangefs
Tested-by: David Howells <dhowells@redhat.com> # afs

show more ...

NFS: swap IO handling is slightly different for O_DIRECT IO

1/ Taking the i_rwsem for swap IO triggers lockdep warnings regarding
possible deadlocks with "fs_reclaim". These deadlocks could, I b

NFS: swap IO handling is slightly different for O_DIRECT IO

1/ Taking the i_rwsem for swap IO triggers lockdep warnings regarding
possible deadlocks with "fs_reclaim". These deadlocks could, I believe,
eventuate if a buffered read on the swapfile was attempted.

We don't need coherence with the page cache for a swap file, and
buffered writes are forbidden anyway. There is no other need for
i_rwsem during direct IO. So never take it for swap_rw()

2/ generic_write_checks() explicitly forbids writes to swap, and
performs checks that are not needed for swap. So bypass it
for swap_rw().

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: discard NFS_RPC_SWAPFLAGS and RPC_TASK_ROOTCREDS

NFS_RPC_SWAPFLAGS is only used for READ requests.
It sets RPC_TASK_SWAPPER which gives some memory-allocation priority to
requests. This is not

NFS: discard NFS_RPC_SWAPFLAGS and RPC_TASK_ROOTCREDS

NFS_RPC_SWAPFLAGS is only used for READ requests.
It sets RPC_TASK_SWAPPER which gives some memory-allocation priority to
requests. This is not needed for swap READ - though it is for writes
where it is set via a different mechanism.

RPC_TASK_ROOTCREDS causes the 'machine' credential to be used.
This is not needed as the root credential is saved when the swap file is
opened, and this is used for all IO.

So NFS_RPC_SWAPFLAGS isn't needed, and as it is the only user of
RPC_TASK_ROOTCREDS, that isn't needed either.

Remove both.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: Fix up forced readdirplus

Avoid clearing the entire readdir page cache if we're just doing forced
readdirplus for the 'ls -l' heuristic.

Signed-off-by: Trond Myklebust <trond.myklebust@hammers

NFS: Fix up forced readdirplus

Avoid clearing the entire readdir page cache if we're just doing forced
readdirplus for the 'ls -l' heuristic.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: Convert readdir page cache to use a cookie based index

Instead of using a linear index to address the pages, use the cookie of
the first entry, since that is what we use to match the page anywa

NFS: Convert readdir page cache to use a cookie based index

Instead of using a linear index to address the pages, use the cookie of
the first entry, since that is what we use to match the page anyway.

This allows us to avoid re-reading the entire cache on a seekdir() type
of operation. The latter is very common when re-exporting NFS, and is a
major performance drain.

The change does affect our duplicate cookie detection, since we can no
longer rely on the page index as a linear offset for detecting whether
we looped backwards. However since we no longer do a linear search
through all the pages on each call to nfs_readdir(), this is less of a
concern than it was previously.
The other downside is that invalidate_mapping_pages() no longer can use
the page index to avoid clearing pages that have been read. A subsequent
patch will restore the functionality this provides to the 'ls -l'
heuristic.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: Improve heuristic for readdirplus

The heuristic for readdirplus is designed to try to detect 'ls -l' and
similar patterns. It does so by looking for cache hit/miss patterns in
both the attribut

NFS: Improve heuristic for readdirplus

The heuristic for readdirplus is designed to try to detect 'ls -l' and
similar patterns. It does so by looking for cache hit/miss patterns in
both the attribute cache and in the dcache of the files in a given
directory, and then sets a flag for the readdirplus code to interpret.

The problem with this approach is that a single attribute or dcache miss
can cause the NFS code to force a refresh of the attributes for the
entire set of files contained in the directory.

To be able to make a more nuanced decision, let's sample the number of
hits and misses in the set of open directory descriptors. That allows us
to set thresholds at which we start preferring READDIRPLUS over regular
READDIR, or at which we start to force a re-read of the remaining
readdir cache using READDIRPLUS.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...

NFS: Adjust the amount of readahead performed by NFS readdir

The current NFS readdir code will always try to maximise the amount of
readahead it performs on the assumption that we can cache anything

NFS: Adjust the amount of readahead performed by NFS readdir

The current NFS readdir code will always try to maximise the amount of
readahead it performs on the assumption that we can cache anything that
isn't immediately read by the process.
There are several cases where this assumption breaks down, including
when the 'ls -l' heuristic kicks in to try to force use of readdirplus
as a batch replacement for lookup/getattr.

This patch therefore tries to tone down the amount of readahead we
perform, and adjust it to try to match the amount of data being
requested by user space.

Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>

show more ...