Hi Trond,
Recently we were observing the behaviour difference between a 2.4.x and
2.6.x kernel with respect to O_EXCL. A comment from 2.4.x era, "For now,
we don't implement O_EXCL." seems inaccurate in TOT.
If so, here's a patch to remove the comment.
This patch is against:
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
Signed-off-by: Harshula Jayasuriya <harshula@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Some releases of Linux rpc.mountd (nfs-utils 1.1.4 and later) return an
empty auth flavor list if no sec= was specified for the export. This is
notably broken server behavior.
The new auth flavor list checking added in a recent commit rejects this
case. The OpenSolaris client does too.
The broken mountd implementation is already widely deployed. To avoid
a behavioral regression, the kernel's mount client skips flavor checking
(ie reverts to the pre-2.6.32 behavior) if mountd returns an empty
flavor list.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
generic_file_direct_write() no longer calls generic_osync_inode() so remove the
comment.
CC: linux-nfs@vger.kernel.org
CC: Neil Brown <neilb@suse.de>
CC: "J. Bruce Fields" <bfields@fieldses.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
In the referral code, use it to look up the new server's ip address if the
fs_locations attribute contains a hostname.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The NFSv4 and NFSv4.1 protocols both allow for the redirection of a client
from one server to another in order to support filesystem migration and
replication. For full protocol support, we need to add the ability to
convert a DNS host name into an IP address that we can feed to the RPC
client.
We'll reuse the sunrpc cache, now that it has been converted to work with
rpc_pipefs.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
do not increment decoding ptr if not needed.
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Part fo the nfs4xdr cleanup. READ_BUF will go away.
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
do not increment encoding ptr if not needed.
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
In order to open code and expose the result pointer assignment.
Alternatively, we can open code the call to xdr_reserve_space
and do the BUG_ON an the error case at the call site.
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This is already done by xdr_reserve_space and since encode_compound_hdr
is adding a byte count to "12" which is already word aligned, the xdr
level rounding will work just as well.
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We can't call nfs_readdata_release()/nfs_writedata_release() without
first initialising and referencing args.context. Doing so inside
nfs_direct_read_schedule_segment()/nfs_direct_write_schedule_segment()
causes an Oops.
We should rather be calling nfs_readdata_free()/nfs_writedata_free() in
those cases.
Looking at the O_DIRECT code, the "struct nfs_direct_req" is already
referencing the nfs_open_context for us. Since the readdata and writedata
structures carry a reference to that, we can simplify things by getting rid
of the extra nfs_open_context references, so that we can replace all
instances of nfs_readdata_release()/nfs_writedata_release().
Reported-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Subject: [PATCH] nfs: remove superfluous BUG_ON()s
Remove duplicated BUG_ON()s from nfs[4]_create_server()
(we make the same checks earlier in both functions).
This takes care of the following entries from Dan's list:
fs/nfs/client.c +1078 nfs_create_server(47) warning: variable derefenced before check 'server->nfs_client'
fs/nfs/client.c +1079 nfs_create_server(48) warning: variable derefenced before check 'server->nfs_client->rpc_ops'
fs/nfs/client.c +1363 nfs4_create_server(43) warning: variable derefenced before check 'server->nfs_client'
fs/nfs/client.c +1364 nfs4_create_server(44) warning: variable derefenced before check 'server->nfs_
Reported-by: Dan Carpenter <error27@gmail.com>
Cc: corbet@lwn.net
Cc: eteo@redhat.com
Cc: Julia Lawall <julia@diku.dk>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Hi.
I have a proposal for possibly resolving this issue.
I believe that this situation occurs due to the way that the
Linux NFS client handles writes which modify partial pages.
The Linux NFS client handles partial page modifications by
allocating a page from the page cache, copying the data from
the user level into the page, and then keeping track of the
offset and length of the modified portions of the page. The
page is not marked as up to date because there are portions
of the page which do not contain valid file contents.
When a read call comes in for a portion of the page, the
contents of the page must be read in the from the server.
However, since the page may already contain some modified
data, that modified data must be written to the server
before the file contents can be read back in the from server.
And, since the writing and reading can not be done atomically,
the data must be written and committed to stable storage on
the server for safety purposes. This means either a
FILE_SYNC WRITE or a UNSTABLE WRITE followed by a COMMIT.
This has been discussed at length previously.
This algorithm could be described as modify-write-read. It
is most efficient when the application only updates pages
and does not read them.
My proposed solution is to add a heuristic to decide whether
to do this modify-write-read algorithm or switch to a read-
modify-write algorithm when initially allocating the page
in the write system call path. The heuristic uses the modes
that the file was opened with, the offset in the page to
read from, and the size of the region to read.
If the file was opened for reading in addition to writing
and the page would not be filled completely with data from
the user level, then read in the old contents of the page
and mark it as Uptodate before copying in the new data. If
the page would be completely filled with data from the user
level, then there would be no reason to read in the old
contents because they would just be copied over.
This would optimize for applications which randomly access
and update portions of files. The linkage editor for the
C compiler is an example of such a thing.
I tested the attached patch by using rpmbuild to build the
current Fedora rawhide kernel. The kernel without the
patch generated about 269,500 WRITE requests. The modified
kernel containing the patch generated about 261,000 WRITE
requests. Thus, about 8,500 fewer WRITE requests were
generated. I suspect that many of these additional
WRITE requests were probably FILE_SYNC requests to WRITE
a single page, but I didn't test this theory.
The difference between this patch and the previous one was
to remove the unneeded PageDirty() test. I then retested to
ensure that the resulting system continued to behave as
desired.
Thanx...
ps
Signed-off-by: Peter Staubach <staubach@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Clean up: Use the common routine now provided in sunrpc.ko for parsing mount
addresses.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Introduce a set of functions in the kernel's RPC implementation for
converting between a socket address and either a standard
presentation address string or an RPC universal address.
The universal address functions will be used to encode and decode
RPCB_FOO and NFSv4 SETCLIENTID arguments. The other functions are
part of a previous promise to deliver shared functions that can be
used by upper-layer protocols to display and manipulate IP
addresses.
The kernel's current address printf formatters were designed
specifically for kernel to user-space APIs that require a particular
string format for socket addresses, thus are somewhat limited for the
purposes of sunrpc.ko. The formatter for IPv6 addresses, %pI6, does
not support short-handing or scope IDs. Also, these printf formatters
are unique per address family, so a separate formatter string is
required for printing AF_INET and AF_INET6 addresses.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Commit a14017db added support in the kernel's NFS mount client to
decode the authentication flavor list returned by mountd.
The NFS client can now use this list to determine whether the
authentication flavor requested by the user is actually supported
by the server.
Note we don't actually negotiate the security flavor if none was
specified by the user. Instead, we try to use AUTH_SYS, and fail if
the server does not support it. This prevents us from negotiating
an inappropriate security flavor (some servers list AUTH_NULL first).
If the server does not support AUTH_SYS, the user must provide an
appropriate security flavor by specifying the "sec=" mount option.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Previous logic in the NFS mount parsing code path assumed
auth_flavor_len was set to zero for simple authentication flavors
(like AUTH_UNIX), and 1 for compound flavors (like AUTH_GSS).
At some earlier point (maybe even before the option parsers were
merged?) specific checks for auth_flavor_len being zero were removed
from the functions that validate the mount option that sets the mount
point's authentication flavor.
Since we are populating an array for authentication flavors, the
auth_flavor_len should always be set to the number of flavors. Let's
eliminate some cleverness here, and prepare for new logic that needs
to know the number of flavors in the auth_flavors[] array.
(auth_flavors[] is an array because at some point we want to allow a
list of acceptable authentication flavors to be specified via the sec=
mount option. For now it remains a single element array).
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
After certain failure modes of an NFS mount, an NFS client should send
a MOUNTPROC_UMNT request to remove the just-added mount entry from the
server's mount table. While no-one should rely on the accuracy of the
server's mount table, sending a UMNT is simply being a good internet
neighbor.
Since NFS mount processing is handled in the kernel now, we will need
a function in the kernel's mountd client that can post a MOUNTRPC_UMNT
request, in order to handle these failure modes.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The new minorversion= mount option (commit 3fd5be9e) was merged at
the same time as the recent sloppy parser fixes (commit a5a16bae),
so minorversion= still uses the old value parsing logic.
If the minorversion= option specifies a bogus value, it should fail
with "bad value" not "bad option."
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tighten up the validity checking in param_set_port: check for NULL pointers.
Ensure that the option shows up on 'modinfo' output.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If the NFSv4 server doesn't support a POSIX attribute, the generic NFS code
needs to know that, so that it don't keep trying to poll for it.
However, by the same count, if the NFSv4 server does support that
attribute, then we should ensure that the inode metadata is appropriately
labelled as being untrusted. For instance, if we don't know the correct
value of the file's uid, we should certainly not be caching ACLs or ACCESS
results.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If the server is broken, then retrying forever won't fix it. We
should just give up after a while, and return an error to the user.
We set the number of retries to 10 for now...
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Ensure that index i remains within array mnt_errtbl[] and mnt3_errtbl[].
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We just had a case in which a buggy server occasionally returns the wrong
attributes during an OPEN call. While the client does catch this sort of
condition in nfs4_open_done(), and causes the nfs4_atomic_open() to return
-EISDIR, the logic in nfs_atomic_lookup() is broken, since it causes a
fallback to an ordinary lookup instead of just returning the error.
When the buggy server then returns a regular file for the fallback lookup,
the VFS allows the open, and bad things start to happen, since the open
file doesn't have any associated NFSv4 state.
The fix is firstly to return the EISDIR/ENOTDIR errors immediately, and
secondly to ensure that we are always careful when dereferencing the
nfs_open_context state pointer.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Commit 008f55d0e0 (nfs41: recover lease in
_nfs4_lookup_root) forces the state manager to always run on mount. This is
a bug in the case of NFSv4.0, which doesn't require us to send a
setclientid until we want to grab file state.
In any case, this is completely the wrong place to be doing state
management. Moving that code into nfs4_init_session...
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The oops http://www.kerneloops.org/raw.php?rawid=537858&msgid= appears to
be due to the nfs4_lock_state->ls_state field being uninitialised. This
happens if the call to nfs4_free_lock_state() is triggered at the end of
nfs4_get_lock_state().
The fix is to move the initialisation of ls_state into the allocator.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
* Remove smp_lock.h from files which don't need it (including some headers!)
* Add smp_lock.h to files which do need it
* Make smp_lock.h include conditional in hardirq.h
It's needed only for one kernel_locked() usage which is under CONFIG_PREEMPT
This will make hardirq.h inclusion cheaper for every PREEMPT=n config
(which includes allmodconfig/allyesconfig, BTW)
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This reverts commit 097041e576.
Trond had a better fix, which is the parent of this one ("Fix compile
error due to congestion_wait() changes")
Requested-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
