30 May, 2012
1 commit
-
Pull CIFS updates from Steve French.
* 'for-next' of git://git.samba.org/sfrench/cifs-2.6: (29 commits)
cifs: fix oops while traversing open file list (try #4)
cifs: Fix comment as d_alloc_root() is replaced by d_make_root()
CIFS: Introduce SMB2 mounts as vers=2.1
CIFS: Introduce SMB2 Kconfig option
CIFS: Move add/set_credits and get_credits_field to ops structure
CIFS: Move protocol specific demultiplex thread calls to ops struct
CIFS: Move protocol specific part from cifs_readv_receive to ops struct
CIFS: Move header_size/max_header_size to ops structure
CIFS: Move protocol specific part from SendReceive2 to ops struct
cifs: Include backup intent search flags during searches {try #2)
CIFS: Separate protocol specific part from setlk
CIFS: Separate protocol specific part from getlk
CIFS: Separate protocol specific lock type handling
CIFS: Convert lock type to 32 bit variable
CIFS: Move locks to cifsFileInfo structure
cifs: convert send_nt_cancel into a version specific op
cifs: add a smb_version_operations/values structures and a smb_version enum
cifs: remove the vers= and version= synonyms for ver=
cifs: add warning about change in default cache semantics in 3.7
cifs: display cache= option in /proc/mounts
...
29 May, 2012
1 commit
-
Pull writeback tree from Wu Fengguang:
"Mainly from Jan Kara to avoid iput() in the flusher threads."* tag 'writeback' of git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux:
writeback: Avoid iput() from flusher thread
vfs: Rename end_writeback() to clear_inode()
vfs: Move waiting for inode writeback from end_writeback() to evict_inode()
writeback: Refactor writeback_single_inode()
writeback: Remove wb->list_lock from writeback_single_inode()
writeback: Separate inode requeueing after writeback
writeback: Move I_DIRTY_PAGES handling
writeback: Move requeueing when I_SYNC set to writeback_sb_inodes()
writeback: Move clearing of I_SYNC into inode_sync_complete()
writeback: initialize global_dirty_limit
fs: remove 8 bytes of padding from struct writeback_control on 64 bit builds
mm: page-writeback.c: local functions should not be exposed globally
23 May, 2012
9 commits
-
While traversing the linked list of open file handles, if the identfied
file handle is invalid, a reopen is attempted and if it fails, we
resume traversing where we stopped and cifs can oops while accessing
invalid next element, for list might have changed.So mark the invalid file handle and attempt reopen if no
valid file handle is found in rest of the list.
If reopen fails, move the invalid file handle to the end of the list
and start traversing the list again from the begining.
Repeat this four times before giving up and returning an error if
file reopen keeps failing.Cc:
Signed-off-by: Shirish Pargaonkar
Reviewed-by: Jeff Layton
Signed-off-by: Steve French -
For more details see .
Signed-off-by: Sedat Dilek
Signed-off-by: Steve French -
As with Linux nfs client, which uses "nfsvers=" or "vers=" to
indicate which protocol to use for mount, specifying"vers=2.1"
will force an SMB2 mount. When vers is not specified CIFS is used
"vers=1"
We can eventually autonegotiate down from SMB2 to CIFS
when SMB2 is stable enough to make it the default, but this
is for the future. At that time we could also implement a
"maxprotocol" mount option as smbclient and Samba have today,
but that would be premature until SMB2 is stable.Intially the SMB2 Kconfig option will depend on "BROKEN"
until the merge is complete, and then be "EXPERIMENTAL"
When it is no longer experimental we can consider changing
the default protocol to attempt first.Signed-off-by: Pavel Shilovsky
Signed-off-by: Jeff Layton
Acked-by: Shirish Pargaonkar
Signed-off-by: Steve French -
SMB2 is the followon to the CIFS (and SMB) protocols
and the default for Windows since Windows Vista, and also
now implemented by various non-Windows servers. SMB2
is more secure, has various performance advantages, including
larger i/o sizes, flow control, better caching model and more.
SMB2 also resolves some scalability limits in the CIFS
protocol and adds many new features while being much
simpler (only a few dozen commands instead of hundreds)
and since the protocol is clearer it is also more consistently
implemented across servers and thus easier to optimize.After much discussion with Jeff Layton, Jeremy Allison
and others at Connectathon, we decided to move the SMB2
code from a distinct .ko and fstype into distinct
C files that optionally build in cifs.ko. As a result
the Kconfig gets simpler.To avoid destabilizing CIFS, the SMB2 code is going
to be moved into its own experimental CONFIG_CIFS_SMB2 ifdef
as it is merged and rereviewed. The changes to stable
CIFS (builds with the SMB2 ifdef off) are expected to be
fairly small.Reviewed-by: Jeff Layton
Signed-off-by: Steve French
Signed-off-by: Pavel Shilovsky
Signed-off-by: Steve French -
Acked-by: Shirish Pargaonkar
Reviewed-by: Jeff Layton
Signed-off-by: Pavel Shilovsky
Signed-off-by: Steve French -
Acked-by: Shirish Pargaonkar
Signed-off-by: Pavel Shilovsky
Signed-off-by: Steve French -
Acked-by: Shirish Pargaonkar
Reviewed-by: Jeff Layton
Signed-off-by: Pavel Shilovsky
Signed-off-by: Steve French -
Acked-by: Shirish Pargaonkar
Reviewed-by: Jeff Layton
Signed-off-by: Pavel Shilovsky
Signed-off-by: Steve French -
Acked-by: Shirish Pargaonkar
Reviewed-by: Jeff Layton
Signed-off-by: Pavel Shilovsky
Signed-off-by: Steve French
17 May, 2012
21 commits
-
As observed and suggested by Tushar Gosavi...
---------
readdir calls these function to send TRANS2_FIND_FIRST and
TRANS2_FIND_NEXT command to the server. The current cifs module is
not specifying CIFS_SEARCH_BACKUP_SEARCH flag while sending these
command when backupuid/backupgid is specified. This can be resolved
by specifying CIFS_SEARCH_BACKUP_SEARCH flag.
---------Cc:
Reported-and-Tested-by: Tushar Gosavi
Signed-off-by: Shirish Pargaonkar
Acked-by: Jeff Layton
Signed-off-by: Steve French -
Signed-off-by: Pavel Shilovsky
-
Signed-off-by: Pavel Shilovsky
-
Signed-off-by: Pavel Shilovsky
-
to handle SMB2 lock type field further.
Acked-by: Jeff Layton
Signed-off-by: Pavel Shilovsky -
CIFS brlock cache can be used by several file handles if we have a
write-caching lease on the file that is supported by SMB2 protocol.
Prepate the code to handle this situation correctly by sorting brlocks
by a fid to easily push them in portions when lease break comes.Signed-off-by: Pavel Shilovsky
-
For SMB2, this should be a no-op. Obviously if we wanted to do something
for the SMB2 case, we could also define an operation here for it.Signed-off-by: Jeff Layton
Signed-off-by: Pavel Shilovsky -
We need a way to dispatch different operations for different versions.
Behold the smb_version_operations/values structures. For now, those
structures just hold the version enum value and nothing uses them.
Eventually, we'll expand them to cover other operations/values as we
change the callers to dispatch from here.Signed-off-by: Jeff Layton
Signed-off-by: Pavel Shilovsky -
We want these to mean something different entirely, and the mount.cifs
helper only ever passed in ver= automatically. Also, don't allow
ver=cifs anymore since that was never passed in by the mount helper.Signed-off-by: Jeff Layton
-
Add a warning that will be displayed when there is no cache= option
specified. We want to ensure that users are aware of the change in
defaults coming in 3.7.Reviewed-by: Pavel Shilovsky
Signed-off-by: Jeff Layton -
...and deprecate the display of strictcache, forcedirectio, and fsc
as separate options.Reviewed-by: Pavel Shilovsky
Signed-off-by: Jeff Layton -
Leave them in for 2 releases and remove for 3.7.
Reviewed-by: Pavel Shilovsky
Signed-off-by: Jeff Layton -
Currently, we have several mount options that control cifs' cache
behavior, but those options aren't considered to be mutually exclusive.
The result is poorly-defined when someone specifies more than one of
these options at mount time.Fix this by adding a new cache= mount option that will supercede
"strictcache", and "forcedirectio". That will help make it clear that
these options are mutually exclusive. Also, change the legacy options to
be mutually exclusive too, to ensure that users don't get surprises.Reviewed-by: Pavel Shilovsky
Signed-off-by: Jeff Layton -
This was used by an ancient version of umount.cifs and in nowhere else
that I'm aware of. Let's add a warning now and dump it for 3.7.Signed-off-by: Jeff Layton
-
We've now warned about this for two releases. Remove it for 3.5.
Signed-off-by: Jeff Layton
-
Convert cifs_iovec_read to use async I/O. This also raises the limit on
the rsize for uncached reads. We first allocate a set of pages to hold
the replies, then issue the reads in parallel and then collect the
replies and copy the results into the iovec.A possible future optimization would be to kmap and inline the iovec
buffers and read the data directly from the socket into that. That would
require some rather complex conversion of the iovec into a kvec however.Signed-off-by: Jeff Layton
-
We'll need this same bit of code for the uncached case.
Signed-off-by: Jeff Layton
-
This isn't strictly necessary for the async readpages code, but the
uncached version will need to be able to collect the replies after
issuing the calls. Add a kref to cifs_readdata and use change the
code to take and put references appropriately.Signed-off-by: Jeff Layton
-
Cached and uncached reads will need to do different things here to
handle the difference when the pages are in pagecache and not. Abstract
out the function that marshals the page list into a kvec array.Signed-off-by: Jeff Layton
-
We'll need different completion routines for an uncached read. Allow
the caller to set the one he needs at allocation time. Also, move
most of these functions to file.c so we can make more of them static.Signed-off-by: Jeff Layton
-
...and add a "directio" synonym since that's what the manpage has
always advertised.Acked-by: Sachin Prabhu
Signed-off-by: Jeff Layton
Signed-off-by: Steve French
10 May, 2012
1 commit
-
This test is always true so it means we revalidate the length every
time, which generates more network traffic. When it is SEEK_SET or
SEEK_CUR, then we don't need to revalidate.Signed-off-by: Dan Carpenter
Reviewed-by: Jeff Layton
Signed-off-by: Steve French
06 May, 2012
1 commit
-
After we moved inode_sync_wait() from end_writeback() it doesn't make sense
to call the function end_writeback() anymore. Rename it to clear_inode()
which well says what the function really does - set I_CLEAR flag.Signed-off-by: Jan Kara
Signed-off-by: Fengguang Wu
04 May, 2012
4 commits
-
The problem was that the first referral was parsed more than once
and so the caller tried the same referrals multiple times.The problem was introduced partly by commit
066ce6899484d9026acd6ba3a8dbbedb33d7ae1b,
where 'ref += le16_to_cpu(ref->Size);' got lost,
but that was also wrong...Cc:
Signed-off-by: Stefan Metzmacher
Tested-by: Björn Jacke
Reviewed-by: Jeff Layton
Signed-off-by: Steve French -
Older mount.cifs programs passed this on to the kernel after parsing
the file. Make sure the kernel ignores that option.Should fix:
https://bugzilla.kernel.org/show_bug.cgi?id=43195
Cc: Sachin Prabhu
Reported-by: Ronald
Signed-off-by: Jeff Layton
Signed-off-by: Steve French -
Signed-off-by: Steve French
-
When revalidating a dentry, if the inode wasn't known to be a dfs
entry when the dentry was instantiated, such as when created via
->readdir(), the DCACHE_NEED_AUTOMOUNT flag needs to be set on the
dentry in ->d_revalidate().The false return from cifs_d_revalidate(), due to the inode now
being marked with the S_AUTOMOUNT flag, might not invalidate the
dentry if there is a concurrent unlazy path walk. This is because
the dentry reference count will be at least 2 in this case causing
d_invalidate() to return EBUSY. So the asumption that the dentry
will be discarded then correctly instantiated via ->lookup() might
not hold.Signed-off-by: Ian Kent
Reviewed-by: Jeff Layton
Cc: Steve French
Cc: linux-cifs@vger.kernel.org
Signed-off-by: Steve French
02 May, 2012
2 commits
-
Cc: Pavel Shilovsky
Signed-off-by: Jeff Layton
Signed-off-by: Steve French -
While testing, I've found that even when we are able to negotiate a
much larger rsize with the server, on-the-wire reads often end up being
capped at 128k because of ra_pages being capped at that level.Lifting this restriction gave almost a twofold increase in sequential
read performance on my craptactular KVM test rig with a 1M rsize.I think this is safe since the actual ra_pages that the VM requests
is run through max_sane_readahead() prior to submitting the I/O. Under
memory pressure we should end up with large readahead requests being
suppressed anyway.Signed-off-by: Jeff Layton
Signed-off-by: Steve French