Reduce some branches and memory accesses in dcache lookup by adding dentry
flags to indicate common d_ops are set, rather than having to check them.
This saves a pointer memory access (dentry->d_op) in common path lookup
situations, and saves another pointer load and branch in cases where we
have d_op but not the particular operation.
Patched with:
git grep -E '[.>]([[:space:]])*d_op([[:space:]])*=' | xargs sed -e 's/\([^\t ]*\)->d_op = \(.*\);/d_set_d_op(\1, \2);/' -e 's/\([^\t ]*\)\.d_op = \(.*\);/d_set_d_op(\&\1, \2);/' -i
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
RCU free the struct inode. This will allow:
- Subsequent store-free path walking patch. The inode must be consulted for
permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
to take i_lock no longer need to take sb_inode_list_lock to walk the list in
the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
page lock to follow page->mapping.
The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.
In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.
The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
The remaining usages for dcache_lock is to allow atomic, multi-step read-side
operations over the directory tree by excluding modifications to the tree.
Also, to walk in the leaf->root direction in the tree where we don't have
a natural d_lock ordering.
This could be accomplished by taking every d_lock, but this would mean a
huge number of locks and actually gets very tricky.
Solve this instead by using the rename seqlock for multi-step read-side
operations, retry in case of a rename so we don't walk up the wrong parent.
Concurrent dentry insertions are not serialised against. Concurrent deletes
are tricky when walking up the directory: our parent might have been deleted
when dropping locks so also need to check and retry for that.
We can also use the rename lock in cases where livelock is a worry (and it
is introduced in subsequent patch).
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Add a new lock, dcache_inode_lock, to protect the inode's i_dentry list
from concurrent modification. d_alias is also protected by d_lock.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Make d_count non-atomic and protect it with d_lock. This allows us to ensure a
0 refcount dentry remains 0 without dcache_lock. It is also fairly natural when
we start protecting many other dentry members with d_lock.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_delete from a dentry deletion notification to a dentry caching
advise, more like ->drop_inode. Require it to be constant and idempotent,
and not take d_lock. This is how all existing filesystems use the callback
anyway.
This makes fine grained dentry locking of dput and dentry lru scanning
much simpler.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
After a few unsuccessful NFS mount attempts in which the client and
server cannot agree on an authentication flavor both support, the
client panics. nfs_umount() is invoked in the kernel in this case.
Turns out nfs_umount()'s UMNT RPC invocation causes the RPC client to
write off the end of the rpc_clnt's iostat array. This is because the
mount client's nrprocs field is initialized with the count of defined
procedures (two: MNT and UMNT), rather than the size of the client's
proc array (four).
The fix is to use the same initialization technique used by most other
upper layer clients in the kernel.
Introduced by commit 0b524123, which failed to update nrprocs when
support was added for UMNT in the kernel.
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=24302
BugLink: http://bugs.launchpad.net/bugs/683938
Reported-by: Stefan Bader <stefan.bader@canonical.com>
Tested-by: Stefan Bader <stefan.bader@canonical.com>
Cc: stable@kernel.org # >= 2.6.32
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
When a nfs_page is freed, nfs_free_request is called which also calls
nfs_clear_request to clean out the lock and open contexts and free the
pagecache page.
However, a couple of places in the nfs code call nfs_clear_request
themselves. What happens here if the refcount on the request is still high?
We'll be releasing contexts and freeing pointers while the request is
possibly still in use.
Remove those bare calls to nfs_clear_context. That should only be done when
the request is being freed.
Note that when doing this, we need to watch out for tests of req->wb_page.
Previously, nfs_set_page_tag_locked() and nfs_clear_page_tag_locked()
would check the value of req->wb_page to figure out if the page is mapped
into the nfsi->nfs_page_tree. We now indicate the page is mapped using
the new bit PG_MAPPED in req->wb_flags .
Reported-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
When nfs client(kernel) don't support NFSv4, maybe user build
kernel without NFSv4, there is a problem.
Using command "mount SERVER-IP:/nfsv3 /mnt/" to mount NFSv3
filesystem, mount should should success, but fail and get error:
"mount.nfs: an incorrect mount option was specified"
System call mount "nfs"(not "nfs4") with "vers=4",
if CONFIG_NFS_V4 is not defined, the "vers=4" will be parsed
as invalid argument and kernel return EINVAL to nfs-utils.
About that, we really want get EPROTONOSUPPORT rather than
EINVAL. This path make sure kernel parses argument success,
and return EPROTONOSUPPORT at nfs_validate_mount_data().
Signed-off-by: Mi Jinlong <mijinlong@cn.fujitsu.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The commit 129a84de23 (locks: fix F_GETLK
regression (failure to find conflicts)) fixed the posix_test_lock()
function by itself, however, its usage in NFS changed by the commit
9d6a8c5c21 (locks: give posix_test_lock
same interface as ->lock) remained broken - subsequent NFS-specific
locking code received F_UNLCK instead of the user-specified lock type.
To fix the problem, fl->fl_type needs to be saved before the
posix_test_lock() call and restored if no local conflicts were reported.
Reference: https://bugzilla.kernel.org/show_bug.cgi?id=23892
Tested-by: Alexander Morozov <amorozov@etersoft.ru>
Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
Cc: <stable@kernel.org>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
An update of mode bits can result in ACL value being changed. We need
to mark the acl cache invalid when we update mode. Similarly we need
to update file attribute when we change ACL value
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If we're searching for a specific cookie, and it isn't found in the page
cache, we should try an uncached_readdir(). To do so, we return EBADCOOKIE,
but we don't set desc->eof.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We need to ensure that the entries in the nfs_cache_array get cleared
when the page is removed from the page cache. To do so, we use the
freepage address_space operation.
Change nfs_readdir_clear_array to use kmap_atomic(), so that the
function can be safely called from all contexts.
Finally, modify the cache_page_release helper to call
nfs_readdir_clear_array directly, when dealing with an anonymous
page from 'uncached_readdir'.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We need to use the cookie from the previous array entry, not the
actual cookie that we are searching for (except for the case of
uncached_readdir).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
When comparing filehandles in the helper nfs_same_file(), we should not be
using 'strncmp()': filehandles are not null terminated strings.
Instead, we should just use the existing helper nfs_compare_fh().
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Store the dirent->d_type in the struct nfs_cache_array_entry so that we
can use it in getdents() calls.
This fixes a regression with the new readdir code.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
It looks as if the array size calculation in MAX_READDIR_ARRAY does not
take the alignment of struct nfs_cache_array_entry into account.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We should ignore the errors from the filldir callback, and just interpret
them as meaning we should exit, however we should definitely pass back
ENOMEM errors.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Currently, uncached_readdir() is broken because if fails to handle
the results from nfs_readdir_xdr_to_array() correctly.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
nfs_do_filldir() must always free desc->page when it is done, otherwise
we end up leaking the page.
Also remove unused variable 'dentry'.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Overflowing the buffer in the readdir ->decode_dirent() should not lead to
a fatal error, but rather to an attempt to reread the record in question.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
When an application opens a file with O_DIRECT flag, if the size of
the data that is written is equal to wsize, the client sends a
WRITE RPC with stable flag set to UNSTABLE followed by a single
COMMIT RPC rather than sending a single WRITE RPC with the stable
flag set to FILE_SYNC. This a bug.
Patch to fix this.
Signed-off-by: Arun R Bharadwaj <arun@linux.vnet.ibm.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The big kernel lock has been removed from all these files at some point,
leaving only the #include.
Remove this too as a cleanup.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Strings allocated via kmemdup() in nfs_readdir_make_qstr() are
referenced from the nfs_cache_array which is stored in a page cache
page. Kmemleak does not scan such pages and it reports several false
positives. This patch annotates the string->name pointer so that
kmemleak does not consider it a real leak.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Bryan Schumaker <bjschuma@netapp.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Fix up the issue that array->eof_index needs to be able to be set
even if array->size == 0.
Ensure that we catch all important memory allocation error conditions
and/or kmap() failures.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This reverts commit 80e60639f1.
This change requires further fixes to ensure that the open doesn't
succeed if the lookup later results in a regular file being created.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Trying to mount NFS (root partition in my case) fails if CONFIG_NFS_V3
is not selected. nfs_validate_mount_data() returns EPROTONOSUPPORT,
because of this check:
#ifndef CONFIG_NFS_V3
if (args->version == 3)
goto out_v3_not_compiled;
#endif /* !CONFIG_NFS_V3 */
and args->version was always initialized to 3.
It was working in 2.6.36
Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The caller allocated it, the caller should free it.
The only issue so far is that we could change the flp pointer even on an
error return if the fl_change callback failed. But we can simply move
the flp assignment after the fl_change invocation, as the callers don't
care about the flp return value if the setlease call failed.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We modified setlease to require the caller to allocate the new lease in
the case of creating a new lease, but forgot to fix up the filesystem
methods.
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
On m68k, which is 32-bit:
fs/nfs/nfs4proc.c: In function ‘nfs41_sequence_done’:
fs/nfs/nfs4proc.c:432: warning: format ‘%ld’ expects type ‘long int’, but argument 3 has type ‘int’
fs/nfs/nfs4proc.c: In function ‘nfs4_setup_sequence’:
fs/nfs/nfs4proc.c:576: warning: format ‘%ld’ expects type ‘long int’, but argument 5 has type ‘int’
On 32-bit, ptrdiff_t is int; on 64-bit, ptrdiff_t is long.
Introduced by commit dfb4f30983 ("NFSv4.1: keep
seq_res.sr_slot as pointer rather than an index")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The intent was to test "*desc" for allocation failures, but it tests
"desc" which is always a valid pointer here.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
nfs_get_lock_context can return NULL on an allocation failure.
Regression introduced by commit f11ac8db.
Reported-by: Steve Dickson <steved@redhat.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
A typo, introduced by commit f11ac8db, in the nfs_direct_write()
routine causes writes with O_DIRECT set to fail with a ENOMEM error.
Found-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve Dickson <steved@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
* 'flock' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/bkl:
locks: turn lock_flocks into a spinlock
fasync: re-organize fasync entry insertion to allow it under a spinlock
locks/nfsd: allocate file lock outside of spinlock
lockd: fix nlmsvc_notify_blocked locking
lockd: push lock_flocks down
lockd should use lock_flocks() instead of lock_kernel()
to lock against posix locks accessing the i_flock list.
This is a prerequisite to turning lock_flocks into a
spinlock.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: J. Bruce Fields <bfields@redhat.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6: (52 commits)
split invalidate_inodes()
fs: skip I_FREEING inodes in writeback_sb_inodes
fs: fold invalidate_list into invalidate_inodes
fs: do not drop inode_lock in dispose_list
fs: inode split IO and LRU lists
fs: switch bdev inode bdi's correctly
fs: fix buffer invalidation in invalidate_list
fsnotify: use dget_parent
smbfs: use dget_parent
exportfs: use dget_parent
fs: use RCU read side protection in d_validate
fs: clean up dentry lru modification
fs: split __shrink_dcache_sb
fs: improve DCACHE_REFERENCED usage
fs: use percpu counter for nr_dentry and nr_dentry_unused
fs: simplify __d_free
fs: take dcache_lock inside __d_path
fs: do not assign default i_ino in new_inode
fs: introduce a per-cpu last_ino allocator
new helper: ihold()
...
This removes more dead code that was somehow missed by commit 0d99519efe
(writeback: remove unused nonblocking and congestion checks). There are
no behavior change except for the removal of two entries from one of the
ext4 tracing interface.
The nonblocking checks in ->writepages are no longer used because the
flusher now prefer to block on get_request_wait() than to skip inodes on
IO congestion. The latter will lead to more seeky IO.
The nonblocking checks in ->writepage are no longer used because it's
redundant with the WB_SYNC_NONE check.
We no long set ->nonblocking in VM page out and page migration, because
a) it's effectively redundant with WB_SYNC_NONE in current code
b) it's old semantic of "Don't get stuck on request queues" is mis-behavior:
that would skip some dirty inodes on congestion and page out others, which
is unfair in terms of LRU age.
Inspired by Christoph Hellwig. Thanks!
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: David Howells <dhowells@redhat.com>
Cc: Sage Weil <sage@newdream.net>
Cc: Steve French <sfrench@samba.org>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Stephen Rothwell reports:
> /home/test/linux-2.6/fs/nfs/nfsroot.c: In function 'nfs_root_debug':
> /home/test/linux-2.6/fs/nfs/nfsroot.c:110:2: error: 'nfs_debug'
> undeclared (first use in this function)
> /home/test/linux-2.6/fs/nfs/nfsroot.c:110:2: note: each undeclared
> identifier is reported only once for each function it appears in
> make[3]: *** [fs/nfs/nfsroot.o] Error 1
> make[2]: *** [fs/nfs] Error 2
> make[1]: *** [fs] Error 2
> make: *** [sub-make] Error 2
Which is caused by commit 306a075362
(NFS: Allow NFSROOT debugging messages to be enabled dynamically)
Fix is to disable this code when RPC_DEBUG is disabled.
Reported-by: Zimny Lech <napohybelskurwysynom2010@gmail.com>
Tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
* 'for-2.6.37' of git://linux-nfs.org/~bfields/linux: (99 commits)
svcrpc: svc_tcp_sendto XPT_DEAD check is redundant
svcrpc: no need for XPT_DEAD check in svc_xprt_enqueue
svcrpc: assume svc_delete_xprt() called only once
svcrpc: never clear XPT_BUSY on dead xprt
nfsd4: fix connection allocation in sequence()
nfsd4: only require krb5 principal for NFSv4.0 callbacks
nfsd4: move minorversion to client
nfsd4: delay session removal till free_client
nfsd4: separate callback change and callback probe
nfsd4: callback program number is per-session
nfsd4: track backchannel connections
nfsd4: confirm only on succesful create_session
nfsd4: make backchannel sequence number per-session
nfsd4: use client pointer to backchannel session
nfsd4: move callback setup into session init code
nfsd4: don't cache seq_misordered replies
SUNRPC: Properly initialize sock_xprt.srcaddr in all cases
SUNRPC: Use conventional switch statement when reclassifying sockets
sunrpc/xprtrdma: clean up workqueue usage
sunrpc: Turn list_for_each-s into the ..._entry-s
...
Fix up trivial conflicts (two different deprecation notices added in
separate branches) in Documentation/feature-removal-schedule.txt
* 'nfs-for-2.6.37' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6:
net/sunrpc: Use static const char arrays
nfs4: fix channel attribute sanity-checks
NFSv4.1: Use more sensible names for 'initialize_mountpoint'
NFSv4.1: pnfs: filelayout: add driver's LAYOUTGET and GETDEVICEINFO infrastructure
NFSv4.1: pnfs: add LAYOUTGET and GETDEVICEINFO infrastructure
NFS: client needs to maintain list of inodes with active layouts
NFS: create and destroy inode's layout cache
NFSv4.1: pnfs: filelayout: introduce minimal file layout driver
NFSv4.1: pnfs: full mount/umount infrastructure
NFS: set layout driver
NFS: ask for layouttypes during v4 fsinfo call
NFS: change stateid to be a union
NFSv4.1: pnfsd, pnfs: protocol level pnfs constants
SUNRPC: define xdr_decode_opaque_fixed
NFSD: remove duplicate NFS4_STATEID_SIZE
The sanity checks here are incorrect; in the worst case they allow
values that crash the client.
They're also over-reliant on the preprocessor.
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
