Please pull the NFS client changes...
Subsystems (2)
Trond Myklebust
Acked-by Peter Zijlstra Acked-by Wu Fengguang Acked-by Jeff Layton Acked-by Jeff LaytonHi Linus,
Please pull from the "nfs-for-2.6.34" branch of the repository at
git pull git://git.linux-nfs.org/projects/trondmy/nfs-2.6.git nfs-for-2.6.34
This will update the following files through the appended changesets.
Cheers,
Trond
----
fs/nfs/callback.h | 8 ++ fs/nfs/callback_proc.c | 165 ++++++++++++++++++++++----- fs/nfs/callback_xdr.c | 105 ++++++++++++----- fs/nfs/client.c | 48 ++++---- fs/nfs/dir.c | 2 +- fs/nfs/dns_resolve.c | 18 +++- fs/nfs/file.c | 30 ++++-- fs/nfs/inode.c | 90 ++------------- fs/nfs/nfs3proc.c | 9 +- fs/nfs/nfs4_fs.h | 2 + fs/nfs/nfs4proc.c | 103 +++++++++++------ fs/nfs/nfs4renewd.c | 24 +--- fs/nfs/nfs4state.c | 118 ++++++++++++++++---- fs/nfs/nfs4xdr.c | 10 ++- fs/nfs/proc.c | 41 +++++++ fs/nfs/symlink.c | 2 +- fs/nfs/write.c | 247 ++++++++++++--------------------------- include/linux/nfs_fs.h | 14 +-- include/linux/nfs_fs_sb.h | 2 + include/linux/sunrpc/bc_xprt.h | 15 +++ net/sunrpc/addr.c | 8 +- net/sunrpc/auth_gss/auth_gss.c | 11 ++- net/sunrpc/svc.c | 4 + net/sunrpc/xprtsock.c | 9 +- 24 files changed, 642 insertions(+), 443 deletions(-)
commit 1cda707d52e51a6cafac0aef12d2bd7052d572e6
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:03:30 2010 -0800
NFS: Remove requirement for inode->i_mutex from nfs_invalidate_mapping
commit 5cf95214ccb915591e2214f81de4659302d3e452
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:03:29 2010 -0800
NFS: Clean up nfs_sync_mapping
Remove the redundant call to filemap_write_and_wait().
commit 7f2f12d963e7c33a93bfb0b22f0178eb1e6a4196
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:03:28 2010 -0800
NFS: Simplify nfs_wb_page()
commit acdc53b2146c7ee67feb1f02f7bc3020126514b8
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:03:26 2010 -0800
NFS: Replace __nfs_write_mapping with sync_inode()
Now that we have correct COMMIT semantics in writeback_single_inode, we can reduce and simplify nfs_wb_all(). Also replace nfs_wb_nocommit() with a call to filemap_write_and_wait(), which doesn't need to hold the inode->i_mutex.
With that done, we can eliminate nfs_write_mapping() altogether.
commit c988950eb6dd6f8e6d98503ca094622729e9aa13
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:03:21 2010 -0800
NFS: Simplify nfs_wb_page_cancel()
In all cases we should be able to just remove the request and call cancel_dirty_page().
commit 2928db1ffeacc9717c2d5c230d450bcc377b3ae9
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:03:18 2010 -0800
NFS: Ensure inode is always marked I_DIRTY_DATASYNC, if it has unstable pages
Since nfs_scan_list() doesn't wait for locked pages, we have a race in which it is possible to end up with an inode that needs to send a COMMIT, but which does not have the I_DIRTY_DATASYNC flag set.
commit 5bad5abec4058c5214bfc72cec418348d6747977
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:02:24 2010 -0800
NFS: Run COMMIT as an asynchronous RPC call when wbc->for_background is set
commit 420e3646bb7d93a571734034249fbb1ae1a7a5c7
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 17:00:02 2010 -0800
NFS: Reduce the number of unnecessary COMMIT calls
If the caller is doing a non-blocking flush, and there are still writebacks pending on the wire, we can usually defer the COMMIT call until those writes are done.
Also ensure that we honour the wbc->nonblocking flag.
commit ff778d02bf867e1733a09b34ad6dbb723b024814
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 16:53:39 2010 -0800
NFS: Add a count of the number of unstable writes carried by an inode
In order to know when we should do opportunistic commits of the unstable writes, when the VM is doing a background flush, we add a field to count the number of unstable writes.
commit 8fc795f703c5138e1a8bfb88c69f52632031aa6a
Author: Trond Myklebust <[email protected]> Date: Fri Feb 19 16:46:56 2010 -0800
NFS: Cleanup - move nfs_write_inode() into fs/nfs/write.c
The sole purpose of nfs_write_inode is to commit unstable writes, so move it into fs/nfs/write.c, and make nfs_commit_inode static.
commit 180b62a3d837613fcac3ce89576526423926c3c3
Author: Andy Adamson <[email protected]> Date: Tue Mar 2 13:19:36 2010 -0500
nfs41 fix NFS4ERR_CLID_INUSE for exchange id
commit ebed9203b68a4f333ce5d17e874b26c3afcfeff1
Author: Trond Myklebust <[email protected]> Date: Tue Mar 2 13:06:22 2010 -0500
NFS: Fix an allocation-under-spinlock bug
sunrpc_cache_update() will always call detail->update() from inside the detail->hash_lock, so it cannot allocate memory.
commit 9fcfe0c83c3b04a759cde6b8c5f961237f17808b
Author: Trond Myklebust <[email protected]> Date: Tue Mar 2 13:06:21 2010 -0500
SUNRPC: Handle EINVAL error returns from the TCP connect operation
This can, for instance, happen if the user specifies a link local IPv6 address.
commit 0f79fd6f5c52e05918e44996b0a1b18383d0fbc2
Author: Trond Myklebust <[email protected]> Date: Tue Mar 2 13:06:21 2010 -0500
NFSv4.1: Various fixes to the sequence flag error handling
Ensure that we change the EXCHANGE_ID verifier (i.e. clp->cl_boot_time) when we want to reset all state. This is mainly needed when the server tells us that it is revoking our open or lock stateids.
Handle revoking of recallable state by expiring the delegations.
Handle callback path issues by expiring the delegations and then resetting the session.
commit 0851de06174e9800e76b26e4be0ca94294c09c8c
Author: Alexandros Batsakis <[email protected]> Date: Fri Feb 5 03:45:06 2010 -0800
nfs4: renewd renew operations should take/put a client reference
renewd sends RENEW requests to the NFS server in order to renew state. As the request is asynchronous, renewd should take a reference to the nfs_client to prevent concurrent umounts from freeing the client
commit 7135840fc74699513d50e0c9c64922f2d38aa5e3
Author: Alexandros Batsakis <[email protected]> Date: Fri Feb 5 03:45:05 2010 -0800
nfs41: renewd sequence operations should take/put client reference
renewd sends SEQUENCE requests to the NFS server in order to renew state. As the request is asynchronous, renewd should take a reference to the nfs_client to prevent concurrent umounts from freeing the session/client
commit dc96aef96a75348b4d1b01c4c0429ab52780683e
Author: Alexandros Batsakis <[email protected]> Date: Fri Feb 5 03:45:04 2010 -0800
nfs: prevent backlogging of renewd requests
If the renewd send queue gets backlogged (e.g., if the server goes down), we will keep filling the queue with periodic RENEW/SEQUENCE requests.
This patch schedules a new renewd request if and only if the previous one returns (either success or failure)
[[email protected]: moved nfs4_schedule_state_renewal() into separate nfs4_renew_release() and nfs41_sequence_release() callbacks to ensure correct behaviour on call setup failure]
commit 888ef2e3f8b7b8daeb031bfb4ad1fd4fa817e193
Author: Alexandros Batsakis <[email protected]> Date: Fri Feb 5 03:45:03 2010 -0800
nfs: kill renewd before clearing client minor version
renewd should be synchronously killed before we destroy the session in nfs4_clear_minor_version
[[email protected]: clean up to remove 'unused function warning when !CONFIG_NFS_V4]
commit f895c53f8ace3c3e49ebf9def90e63fc6d46d2bf
Author: Chuck Lever <[email protected]> Date: Mon Feb 1 14:17:50 2010 -0500
NFS: Make close(2) asynchronous when closing NFS O_DIRECT files
For NFSv2 and v3:
O_DIRECT writes are always synchronous, and aren't cached, so nothing should be flushed when closing an NFS O_DIRECT file descriptor. Thus there are no write errors to report on close(2).
In addition, there's no cached data to verify on the next open(2), so we don't need clean GETATTR results at close time to compare with.
Thus, there's no need for the nfs_revalidate_inode() call when closing an NFS O_DIRECT file. This reduces the number of synchronous on-the-wire requests for a simple open-write-close of an NFS O_DIRECT file by roughly 20%.
For NFSv4:
Call nfs4_do_close() with wait set to zero when closing an NFS O_DIRECT file. The CLOSE will go on the wire, but the application won't wait for it to complete.
commit 7e381172cf6e0282a56374e50667515aed55166a
Author: Chuck Lever <[email protected]> Date: Mon Feb 1 14:17:41 2010 -0500
NFS: Improve NFS iostat byte count accuracy for writes
The bytes counted by the performance counters for NFS writes should reflect write and sync errors. If the write(2) system call reports an error, the bytes should not be counted. And, if the write is short, the actual number of bytes that was written should be counted, not the number of bytes that was requested.
commit aa2f1ef10e6ad65c9138ec576f82c08f32e6f32c
Author: Chuck Lever <[email protected]> Date: Mon Feb 1 14:17:32 2010 -0500
NFS: Account for NFS bytes read via the splice API
Bytes read via the splice API should be accounted for in the NFS performance statistics.
commit 4184dcf2dbde481b34d370e1704f2b91a8c9f0d1
Author: Chuck Lever <[email protected]> Date: Mon Feb 1 14:17:23 2010 -0500
NFS: Fix byte accounting for generic NFS reads
Currently, the NFS I/O counters count the number of bytes requested by applications, rather than the number of bytes actually read by the system calls.
The number of bytes requested for reads is actually not that useful, because the value is usually a buffer size for reads. That is, that requested number is usually a maximum, and frequently doesn't reflect the actual number of bytes read.
commit c2459dc46269728e4a080ec8d5a316b2bba2e142
Author: Chuck Lever <[email protected]> Date: Mon Feb 1 14:17:14 2010 -0500
NFS: Proper accounting for NFS VFS calls
Nit: The VFSOPEN and VFSFLUSH counters are function call counters. Count every call to these routines.
commit ba17686f62db88f6a591121e768a0c83a2a2647d
Author: Andy Adamson <[email protected]> Date: Tue Jan 26 21:24:04 2010 -0500
nfs41 do not allocate unused back channel pages
[[email protected]: moved definition of svc_is_backchannel()
into include/linux/sunrpc/bc_xprt.h.]
commit 9733f0d9289cbcac4fa03db0cb5aec1ab01c6bc9
Author: Andy Adamson <[email protected]> Date: Fri Jan 22 12:03:08 2010 -0500
nfs41: cleanup callback code to use __be32 type
commit 41f54a55480c752d9419cac5e647785cb794142e
Author: Andy Adamson <[email protected]> Date: Thu Jan 21 14:54:13 2010 -0500
nfs41: clear NFS4CLNT_RECALL_SLOT bit on session reset
commit bae0ac0ee1839e345a9b26d8c00eb3ef565caad1
Author: Andy Adamson <[email protected]> Date: Thu Jan 21 14:19:16 2010 -0500
nfs41: fix nfs4_callback_recallslot
Return NFS4_OK if target high slotid equals enforced high slotid. Fix nfs_client reference leak.
commit 104aeba484c9291cde2def6d037b836af46d8eb0
Author: Andy Adamson <[email protected]> Date: Thu Jan 14 17:45:10 2010 -0500
nfs41: resize slot table in reset
When session is reset, client can renegotiate slot table size.
commit b9efa1b27e25b1286504973c0a6bf0f24106faa8
Author: Andy Adamson <[email protected]> Date: Wed Jan 20 16:06:27 2010 -0500
nfs41: implement cb_recall_slot
Drain the fore channel and reset the max_slots to the new value.
commit 4911096f1a5df73c12c287a42ece4e7b5d9c19ec
Author: Andy Adamson <[email protected]> Date: Thu Jan 14 17:45:08 2010 -0500
nfs41: back channel drc minimal implementation
For now the back channel ca_maxresponsesize_cached is 0 and there is no backchannel DRC. Return NFS4ERR_REP_TOO_BIG_TO_CACHE when a cb_sequence cachethis is true. When it is false, return NFS4ERR_RETRY_UNCACHED_REP as the next operation error.
Remember the replay error accross compound operation processing.
commit b2f28bd78354b9bbcd178bf6bbf6b2277cd9b761
Author: Andy Adamson <[email protected]> Date: Thu Jan 14 17:45:07 2010 -0500
nfs41: prepare for back channel drc
Make all cb_sequence arguments available to verify_seqid which will make replay decisions.
commit e95e60daee44fade63f32429ddcf1c2012a95632
Author: Andy Adamson <[email protected]> Date: Thu Jan 14 17:45:06 2010 -0500
nfs41: remove uneeded checks in callback processing
All callback operations have arguments to decode and require processing. The preprocess_nfs4X_op functions catch unsupported or illegal ops so decode_args and process_op pointers are always non NULL.
commit b92b30190093377828efcde5fc4cf7598fa1ee46
Author: Andy Adamson <[email protected]> Date: Thu Jan 14 17:45:05 2010 -0500
nfs41: directly encode back channel error
Skip all other processing when error is encountered.
commit 31d2b4356b054537c35f4f8a7533e0b4a494dcc6
Author: Andy Adamson <[email protected]> Date: Thu Jan 14 17:45:04 2010 -0500
nfs41: fix wrong error on callback header xdr overflow
Set NFS4ERR_RESOURCE as CB_COMPOUND status and do not return an op on decode_op_hdr or encode_op_hdr buffer overflow.
NFS4ERR_RESOURCE is correct for v4.0. Will fix the return for v4.1 along with all the other NFS4ERR_RESOURCE errors in a later patch.
commit 72ce2b3c064471fc511a9ca2fb6c38d90d2ab826
Author: Mike Sager <[email protected]> Date: Tue Jan 19 12:54:41 2010 -0500
nfs41: Process callback's referring call list
If a CB_SEQUENCE referring call triple matches a slot table entry, the client is still waiting for a response to the original request. In this case, return NFS4ERR_DELAY as the response to the callback.
commit a7989c3e4702203baa5ddb3614f92bfc49a6e491
Author: Mike Sager <[email protected]> Date: Tue Jan 19 12:54:40 2010 -0500
nfs41: Check slot table for referring calls
Traverse a list of referring calls and look for a session/slot/seq number match.
commit 8e0d46e13833b06832395e7eacccae8af8743461
Author: Mike Sager <[email protected]> Date: Thu Dec 17 12:06:26 2009 -0500
nfs41: Adjust max cache response size value
For the CREATE_SESSION attribute ca_maxresponsesize_cached, calculate the value based on the rpc reply header size plus the maximum nfs compound reply size.
commit 5a51f13adf7909caec2f8182767485c30e21364e
Author: H Hartley Sweeten <[email protected]> Date: Thu Jan 14 15:38:31 2010 -0700
xprtsock.c: make bc_{malloc/free} static
xprtsock.c: make bc_{malloc/free} static
The server backchannel buf_alloc and buf_free methods should be static since they are not used outside this file.
commit 7a88efe9760de3bc75bde61b0a4c56dbb6cf2494
Author: Chuck Lever <[email protected]> Date: Tue Jan 12 17:41:20 2010 -0500
SUNRPC: Don't display zero scope IDs
A zero scope ID means that it wasn't set, so we don't need to append it to presentation format addresses.
commit f1a89a118299df9e5afbaaedf83e5709a28632b6
Author: Chuck Lever <[email protected]> Date: Tue Jan 12 17:41:10 2010 -0500
SUNRPC: Deprecate support for site-local addresses
RFC 3879 "formally deprecates" site-local IPv6 addresses. We interpret that to mean that the scope ID is ignored for all but link-local addresses.
commit 97cefcc6d0aa6b4fc9ba67eb1ef4cc9e25f826f2
Author: Jeff Layton <[email protected]> Date: Fri Jan 8 12:17:21 2010 -0500
nfs: handle NFSv2 -EKEYEXPIRED returns from RPC layer appropriately
Add a wrapper around rpc_call_sync that handles -EKEYEXPIRED errors from the RPC layer as it would an -EJUKEBOX error if NFSv2 had such a thing. Also, add a handler for that error for async calls that makes it resubmit the RPC on -EKEYEXPIRED.
commit b68d69b8c6d19f4c2174f26fe8b750a0e82eb732
Author: Jeff Layton <[email protected]> Date: Thu Jan 7 09:42:04 2010 -0500
nfs: handle NFSv3 -EKEYEXPIRED errors as we would -EJUKEBOX
We're using -EKEYEXPIRED to indicate that a krb5 credcache contains an expired ticket and that we should have the NFS layer retry the RPC call instead of returning an error back to the caller. Handle this as we would an -EJUKEBOX error return.
commit 2c6434888cef9e5f450d6c5b7df6d8c625ed27c1
Author: Jeff Layton <[email protected]> Date: Thu Jan 7 09:42:03 2010 -0500
nfs4: handle -EKEYEXPIRED errors from RPC layer
If a KRB5 TGT ticket expires, we don't want to return an error immediatel. If someone has a long running job and just forgets to run "kinit" in time then this will make it fail.
Instead, we want to treat this situation as we would NFS4ERR_DELAY and retry the upcall after delaying a bit with an exponential backoff.
This patch just makes any place that would handle NFS4ERR_DELAY also handle -EKEYEXPIRED the same way. In the future, we may want to be more sophisticated however and handle hard vs. soft mounts differently, or specify some upper limit on how long we'll wait for a new TGT to be acquired.
commit dc5ddce956660247e004a4b20a26b7d137ab1644
Author: Jeff Layton <[email protected]> Date: Thu Jan 7 09:42:02 2010 -0500
sunrpc: parse and return errors reported by gssd
The kernel currently ignores any error code sent by gssd and always considers it to be -EACCES. In order to better handle the situation of an expired KRB5 TGT, the kernel needs to be able to parse and deal with the errors that gssd sends. Aside from -EACCES the only error we care about is -EKEYEXPIRED, which we're using to indicate that the upper layers should retry the call a little later.
To maintain backward compatibility with older gssd's, any error other than -EKEYEXPIRED is interpreted as -EACCES.