Hi Linus,
Please pull from the repository at
git pull git://git.linux-nfs.org/pub/linux/nfs-2.6.git
This will update the following files through the appended changesets.
Cheers,
Trond
----
fs/Kconfig | 1=20
fs/lockd/clntlock.c | 112 +---
fs/lockd/clntproc.c | 319 ++++-------
fs/lockd/host.c | 12=20
fs/lockd/mon.c | 11=20
fs/lockd/svc4proc.c | 157 ++---
fs/lockd/svclock.c | 349 +++++++-----
fs/lockd/svcproc.c | 151 ++---
fs/lockd/svcshare.c | 4=20
fs/lockd/svcsubs.c | 7=20
fs/lockd/xdr.c | 17 -
fs/lockd/xdr4.c | 21 -
fs/locks.c | 106 ++--
fs/namespace.c | 38 +
fs/nfs/callback.c | 20 -
fs/nfs/callback_xdr.c | 28 +
fs/nfs/delegation.c | 19 +
fs/nfs/delegation.h | 1=20
fs/nfs/dir.c | 114 +++-
fs/nfs/direct.c | 955 ++++++++++++++++++----------=
----
fs/nfs/file.c | 49 +-
fs/nfs/idmap.c | 47 +-
fs/nfs/inode.c | 229 ++++++--
fs/nfs/iostat.h | 164 +++++
fs/nfs/mount_clnt.c | 17 -
fs/nfs/nfs2xdr.c | 4=20
fs/nfs/nfs3acl.c | 16 -
fs/nfs/nfs3proc.c | 246 ++++----
fs/nfs/nfs3xdr.c | 6=20
fs/nfs/nfs4proc.c | 180 +++---
fs/nfs/nfs4state.c | 1=20
fs/nfs/nfs4xdr.c | 2=20
fs/nfs/pagelist.c | 16 -
fs/nfs/proc.c | 156 +++--
fs/nfs/read.c | 102 +++
fs/nfs/unlink.c | 3=20
fs/nfs/write.c | 288 +++++++---
fs/nfsd/nfs4callback.c | 2=20
fs/nfsd/nfs4state.c | 13=20
fs/proc/base.c | 39 +
include/linux/fs.h | 7=20
include/linux/lockd/lockd.h | 27 +
include/linux/lockd/share.h | 2=20
include/linux/lockd/xdr.h | 1=20
include/linux/nfs_fs.h | 102 +--
include/linux/nfs_fs_i.h | 8=20
include/linux/nfs_fs_sb.h | 6=20
include/linux/nfs_xdr.h | 5=20
include/linux/sunrpc/clnt.h | 20 -
include/linux/sunrpc/gss_krb5.h | 2=20
include/linux/sunrpc/metrics.h | 77 +++
include/linux/sunrpc/rpc_pipe_fs.h | 2=20
include/linux/sunrpc/sched.h | 9=20
include/linux/sunrpc/xprt.h | 13=20
net/sunrpc/auth.c | 16 -
net/sunrpc/auth_gss/auth_gss.c | 2=20
net/sunrpc/auth_gss/gss_krb5_seal.c | 15 -
net/sunrpc/auth_gss/gss_krb5_unseal.c | 4=20
net/sunrpc/auth_gss/gss_krb5_wrap.c | 17 -
net/sunrpc/auth_gss/gss_spkm3_mech.c | 6=20
net/sunrpc/auth_gss/gss_spkm3_seal.c | 5=20
net/sunrpc/auth_gss/gss_spkm3_unseal.c | 4=20
net/sunrpc/clnt.c | 53 +-
net/sunrpc/pmap_clnt.c | 41 +
net/sunrpc/rpc_pipe.c | 31 +
net/sunrpc/sched.c | 12=20
net/sunrpc/stats.c | 115 ++++
net/sunrpc/xprt.c | 29 +
net/sunrpc/xprtsock.c | 49 ++
69 files changed, 2829 insertions(+), 1873 deletions(-)
commit df6db302cb236ac3a683d535a3e2073d9f4b2833
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:25:10 2006 -0500
SUNRPC,RPCSEC_GSS: spkm3--fix config dependencies
=20
Add default selection of CRYPTO_CAST5 when selecting RPCSEC_GSS_SPKM3.
=20
Signed-off-by: Kevin Coffman <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 0e19c1ea2fa57f0612c80595f534b4ddcf69ad8f
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:24:40 2006 -0500
SUNRPC,RPCSEC_GSS: spkm3: import contexts using NID_cast5_cbc
=20
Import the NID_cast5_cbc from the userland context. Not used.
=20
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 5f12191bc000ea31970339a5f54c11087506711c
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:24:25 2006 -0500
LOCKD: Make nlmsvc_traverse_shares return void
=20
The nlmsvc_traverse_shares return value is always zero, hence useless.
=20
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit f3ee439f43381e45b191cf721b4a51d41f33301f
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:24:13 2006 -0500
LOCKD: nlmsvc_traverse_blocks return is unused
=20
Note that we never return non-zero.
=20
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit eaa82edf20d738a7ae31f4b0a5f72f64c14a58df
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:24:04 2006 -0500
SUNRPC,RPCSEC_GSS: fix krb5 sequence numbers.
=20
Use a spinlock to ensure unique sequence numbers when creating krb5 gss=
tokens.
=20
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 096455a22acac06fb6d0d75f276170ab72d55ba6
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:23:42 2006 -0500
NFSv4: Dont list system.nfs4_acl for filesystems that don't support it.
=20
Thanks to Frank Filz for pointing out that we list system.nfs4_acl exte=
nded
attribute even on filesystems where we don't actually support nfs4_acl.
This is inconsistent with the e.g. ext3 POSIX ACL behaviour, and seems =
to
annoy cp.
=20
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 9e57b302cf0f27063184196def620f39ca7a5fc6
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 23:23:11 2006 -0500
SUNRPC,RPCSEC_GSS: remove unnecessary kmalloc of a checksum
=20
Remove unnecessary kmalloc of temporary space to hold the md5 result; i=
t's
small enough to just put on the stack.
=20
This code may be called to process rpc's necessary to perform writes, s=
o
there's a potential deadlock whenever we kmalloc() here. After this a
couple kmalloc()'s still remain, to be removed soon.
=20
This also fixes a rare double-free on error noticed by coverity.
=20
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 7a1218a277c45cba1fb8d7089407a1769c645c43
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 18:11:10 2006 -0500
SUNRPC: Ensure rpc_call_async() always calls tk_ops->rpc_release()
=20
Currently this will not happen if we exit before rpc_new_task() was cal=
led.
Also fix up rpc_run_task() to do the same (for consistency).
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 43ac3f2961b8616da26114ec6dc76ac2a61f76ad
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:51 2006 -0500
SUNRPC: Fix memory barriers for req->rq_received
=20
We need to ensure that all writes to the XDR buffers are done before
req->rq_received is visible to other processors.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit c42de9dd67250fe984e0e31c9b542d721af6454b
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:51 2006 -0500
NFS: Fix a race in nfs_sync_inode()
=20
Kudos to Neil Brown for spotting the problem:
=20
"in nfs_sync_inode, there is effectively the sequence:
=20
nfs_wait_on_requests
nfs_flush_inode
nfs_commit_inode
=20
This seems a bit racy to me as if the only requests are on the
->commit list, and nfs_commit_inode is called separately after
nfs_wait_on_requests completes, and before nfs_commit_inode start
(say: by nfs_write_inode) then none of these function will return
>0, yet there will be some pending request that aren't waited for."
=20
The solution is to search for requests to wait upon, search for dirty
requests, and search for uncommitted requests while holding the
nfsi->req_lock
=20
The patch also cleans up nfs_sync_inode(), getting rid of the redundant
FLUSH_WAIT flag. It turns out that we were always setting it.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 7d46a49f512e8d10b23353781a8ba85edd4fa640
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:50 2006 -0500
NFS: Clean up nfs_flush_list()
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit deb7d638262019cbac5d15ab74ffd1c29242c7cb
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:50 2006 -0500
NFS: Fix a race with PG_private and nfs_release_page()
=20
We don't need to set PG_private for readahead pages, since they never g=
et
unlocked while I/O is in progress. However there is a small race in
nfs_readpage_release() whereby the page may be unlocked, and have
PG_private set.
=20
Fix is to have PG_private set only for the case of writes...
=20
Also fix a bug in nfs_clear_page_writeback(): Don't attempt to clear th=
e
radix_tree tag if we've already deleted the radix tree entry.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 1dd761e9070aa2e543df3db41bd75ed4b8f2fab9
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:49 2006 -0500
NFSv4: Ensure the callback daemon flushes signals
=20
If the callback daemon is signalled, but is unable to exit because it s=
till
has users, then we need to flush signals. If not, then svc_recv() can
never sleep, and so we hang.
If we flush signals, then we also have to be prepared to resend them wh=
en
we want the thread to exit.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 5428154827c2bf7cfdc9dab60db1e0eaa57c027a
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:49 2006 -0500
SUNRPC: Fix a 'Busy inodes' error in rpc_pipefs
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit a9a801787a761616589a6526d7a29c13f4deb3d8
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:48 2006 -0500
NFS, NLM: Allow blocking locks to respect signals
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 03f28e3a2059fc466761d872122f30acb7be61ae
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:48 2006 -0500
NFS: Make nfs_fhget() return appropriate error values
=20
Currently it returns NULL, which usually gets interpreted as ENOMEM. In
fact it can mean a host of issues.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 01d0ae8beaee75d954900109619b700fe68707d9
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:48 2006 -0500
NFSv4: Fix an oops in nfs4_fill_super
=20
The mount statistics patches introduced a call to nfs_free_iostats that=
is
not only redundant, but actually causes an oops.
=20
Also fix a memory leak due to the lack of a call to nfs_free_iostats on
unmount.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit d9f6eb75d4900782a095b98470decfe98971f920
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:47 2006 -0500
lockd: blocks should hold a reference to the nlm_file
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 51581f3bf922512880f52a7777923fd6dcfc792b
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:47 2006 -0500
NFSv4: SETCLIENTID_CONFIRM should handle NFS4ERR_DELAY/NFS4ERR_RESOURCE
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 3e4f6290ca4df7464ee066123f2bca4298c2dab4
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:46 2006 -0500
NFSv4: Send the delegation stateid for SETATTR calls
=20
In the case where we hold a delegation stateid, use that in for inside
SETATTR calls.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit f25bc34967d76610d17bc70769d7c220976eeeb1
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:46 2006 -0500
NFSv4: Ensure nfs_callback_down() calls svc_destroy()
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 6041b79192bdf0e7ab18ea6859effa5d8311391b
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:45 2006 -0500
lockd: Fix a typo in nlmsvc_grant_release()
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit d47166244860eb5dfdb12ee4703968beef8a0db2
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:45 2006 -0500
lockd: Add helper for *_RES callbacks
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 92737230dd3f1478033819d4bc20339f8da852da
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:45 2006 -0500
NLM: Add nlmclnt_release_call
=20
Add a helper function to simplify the freeing of NLM client requests.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit e4cd038a45a46ffbe06a1a72f3f15246e5b041ca
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:44 2006 -0500
NLM: Fix nlmclnt_test to not copy private part of locks
=20
The struct file_lock does not carry a properly initialised lock,
so don't copy it as if it were.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 3a649b884637c4fdff50a6beebc3dc0e6082e048
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:44 2006 -0500
NLM: Simplify client locks
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit d72b7a6b26b9009b7a05117fe2e04b3a73ae4a5c
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:43 2006 -0500
NFS: O_DIRECT needs to use a completion
=20
Now that we have aio writes, it is possible for dreq->outstanding to be
zero, but for the I/O not to have completed. Convert struct nfs_direct_=
req
to use a completion to signal when the I/O is done.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 6b45d858ed6821dd687efd3b68929de2e4954fec
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:43 2006 -0500
NFS: Clean up nfs_get_user_pages
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 606bbba06b11ebcbdf3a4fcd8cce4507c5bd7a4b
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:42 2006 -0500
NFS: fix compiler warnings on 64-bit platforms
=20
Introduced by NFS aio+dio patches.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled on 64-bit hardware.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 5eb53f41d11c04aa4ddb9f168b6bbb27b9790348
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:42 2006 -0500
SUNRPC: fix compile warnings on 64-bit platforms
=20
Introduced by NFS metrics patch.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled on a 64-bit platform.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 35576cba57f1c042b87d6586b3229d13067264c6
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:41 2006 -0500
NLM: nlmclnt_cancel_callback should accept NLM_LCK_DENIED errors
=20
NLM_LCK_DENIED is a valid error return for an NLM_CANCEL call by the
client.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 4c060b531006e0711db32a132d6ac7661594b280
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:41 2006 -0500
lockd: Fix Oopses due to list manipulation errors.
=20
The patch "stop abusing file_lock_list introduces a couple of bugs sinc=
e
the locks may be copied and need to be removed from the lists when they=
are
destroyed.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 26bcbf965f857c710adafd16cf424f043006b5dd
Author: Christoph Hellwig <[email protected]>
Date: Mon Mar 20 13:44:40 2006 -0500
lockd: stop abusing file_lock_list
=20
Currently lockd directly access the file_lock_list from fs/locks.c.
It does so to mark locks granted or reclaimable. This is very
suboptimal, because a) lockd needs to poke into locks.c internals, and
b) it needs to iterate over all locks in the system for marking locks
granted or reclaimable.
=20
This patch adds lists for granted and reclaimable locks to the nlm_host
structure instead, and adds locks to those.
=20
nlmclnt_lock:
now adds the lock to h_granted instead of setting the
NFS_LCK_GRANTED, still O(1)
=20
nlmclnt_mark_reclaim:
goes away completely, replaced by a list_splice_init.
Complexity reduced from O(locks in the system) to O(1)
=20
reclaimer:
iterates over h_reclaim now, complexity reduced from
O(locks in the system) to O(locks per nlm_host)
=20
Signed-off-by: Christoph Hellwig <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 04266473ecf5cdca242201d9f1ed890afe070fb6
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:40 2006 -0500
lockd: Make lockd use rpc_new_client() instead of rpc_create_client
=20
When doing NLM_GRANTED requests, lockd may end up blocking if we use
rpc_create_client() due to the synchronous call to rpc_ping(). Instead,=
use
rpc_new_client().
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 686517f1ad1630c11964d668b556aab79b8c942e
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:39 2006 -0500
lockd: Make nlmsvc_create_block() use nlmsvc_lookup_host()
=20
Currently it uses nlmclnt_lookup_host(), which puts the resulting host
structure on a different list.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 5e1abf8cb713a0b94f5a400c7b9b797990cd9dec
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:39 2006 -0500
lockd: Clean up of the server-side GRANTED code
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 6849c0cab69f5d1a0fc7b05fa5bfb3dec53f86df
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:39 2006 -0500
lockd: Add refcounting to struct nlm_block
=20
Otherwise, the block may disappear from underneath us when in
nlmsvc_retry_blocked.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 09c7938c5640a6f22bef074ca6b803dccfdb93e3
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:38 2006 -0500
lockd: Fix server-side lock blocking code
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 0996905f9301c2ff4c021982c42a15b35e74bf1c
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:38 2006 -0500
lockd: posix_test_lock() should not call locks_copy_lock()
=20
The caller of posix_test_lock() should never need to look at the lock
private data, so do not copy that information. This also means that the=
re
is no need to call the fl_release_private methods.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 3feb2d49394b7874348a6e43c076b780c1d222c5
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:37 2006 -0500
NFS: Uninline nfs_writedata_(alloc|free) and nfs_readdata_(alloc|free)
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 5db3a7b2cabe8f0957683f798c4f8fa8605f9ebb
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:37 2006 -0500
NFS: Debugging code for nfs_direct_(read|write)_schedule()
=20
Make sure that we're doing our list accounting correctly.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit a8881f5a5c723f82da84b786d3ca83a0df9e0c33
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:36 2006 -0500
NFS: O_DIRECT async IO may lose context
=20
The struct nfs_direct_req currently keeps a pointer to the file descrip=
tor
without referencing it. This may cause problems if the parent process i=
s
killed.
=20
The nfs_open_context should normally have all the information that we'r=
e
currently using the filp for, and unlike fput(), is safe to release fro=
m
an rpciod process context.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit fad61490419b3e494f300e9b2579810ef3bcda31
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:36 2006 -0500
nfs: Use UNSTABLE + COMMIT for NFS O_DIRECT writes
=20
Currently NFS O_DIRECT writes use FILE_SYNC so that a COMMIT is not
necessary. This simplifies the internal logic, but this could be a
difficult workload for some servers.
=20
Instead, let's send UNSTABLE writes, and after they all complete, send =
a
COMMIT for the dirty range. After the COMMIT returns successfully, the=
n do
the wake_up or fire off aio_complete().
=20
Test plan:
Async direct I/O tests against Solaris (or any server that requires
committed unstable writes). Reboot server during test.
=20
Based on an earlier patch by Chuck Lever <[email protected]>
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit e17b1fc4b35399935f00a635206e183d9292fe4f
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:35 2006 -0500
NFS: Make nfs_commit_alloc() extern
=20
We need to use nfs_commit_alloc() in fs/nfs/direct.c.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit a37ec012d7fd352648c8455d3396ea24001efcd3
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:35 2006 -0500
NFS: fix data_update accounting in NFS direct I/O path
=20
^C against "iozone -I" is hitting the assertion in nfs_clear_inode().
=20
Test plan:
"iozone -i0 -I -a -c" against a slow server, then control C. This shou=
ld
not cause an oops.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 15ce4a0c1ce0d5e288398cb9e5493fd4e55e2025
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:34 2006 -0500
NFS: Replace atomic_t variables in nfs_direct_req with a single spin lo=
ck
=20
Three atomic_t variables cause a lot of bus locking. Because they are =
all
used in the same places in the code, just use a single spin lock.
=20
Now that the atomic_t variables are gone, we can remove the request siz=
e
limitation since the code no longer depends on the limited width of ato=
mic_t
on some platforms.
=20
Test plan:
Compile with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled. Millions of f=
sx
operations, iozone, OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 88467055f7654302c12df74e5fe4d12516656a39
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:34 2006 -0500
NFS: clean up comments and tab damage in direct.c
=20
Clean up tab damage and comments. Replace "file_offset" with more comm=
only
used "pos".
=20
Test plan:
Compile with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 9eafa8cc521b489f205bf7b0634c99e34e046606
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:33 2006 -0500
NFS: support EIOCBQUEUED return in direct write path
=20
For async iocb's, the NFS direct write path now returns EIOCBQUEUED,
and calls aio_complete when all the requested writes are finished. The
synchronous part of the NFS direct write path behaves exactly as it
was before.
=20
Shared mapped NFS files will have some coherency difficulties when
accessed concurrently with aio+dio. Will need to explore how this
is handled in the local file system case.
=20
Test plan:
aio-stress with "-O". OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit c89f2ee5f9223b864725f7344f24a037dfa76568
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:33 2006 -0500
NFS: make iocb available everywhere in direct write path
=20
Pass the iocb argument all the way down to the direct write request
scheduler, and make it available in nfs_direct_write_result.
=20
Test plan:
Compile the kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
Millions of fsx-odirect ops. OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 47989d7454398827500d0e73766270986a3b488f
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:32 2006 -0500
NFS: remove support for multi-segment iovs in the direct write path
=20
Eliminate the persistent use of automatic storage in all parts of the
NFS client's direct write path to pave the way for introducing support
for aio against files opened with the O_DIRECT flag.
=20
Test plan:
Compile the kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
Millions of fsx-odirect ops. OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 462d5b3296b56289efec426499a83faad4c08d9e
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:32 2006 -0500
NFS: make direct write path generate write requests concurrently
=20
Duplicate infrastructure from direct read path that will allow write
path to generate multiple write requests concurrently. This will
enable us to add support for aio in this path.
=20
Temporarily we will lose the ability to do UNSTABLE writes followed by
a COMMIT in the direct write path. However, all applications I am
aware of that use NFS O_DIRECT currently write in relatively small
chunks, so this should not be inconvenient in any way.
=20
Test plan:
Millions of fsx-odirect ops. OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 63ab46abc70b01cb0711301f5ddb08c1c0bb9b1c
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:31 2006 -0500
NFS: create common routine for handling direct I/O completion
=20
Factor out the common piece of completing an NFS direct I/O request.
=20
Test plan:
Compile kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 93619e5989173614bef0013b0bb8a3fe3dbd5a95
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:31 2006 -0500
NFS: create common routine for allocating nfs_direct_req
=20
Factor out a small common piece of the path that allocate nfs_direct_re=
q
structures.
=20
Test plan:
Compile kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit bc0fb201b34b12e2d16e8cbd5bb078c1db936304
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:31 2006 -0500
NFS: create common routine for waiting for direct I/O to complete
=20
We're about to add asynchrony to the NFS direct write path. Begin by
abstracting out the common pieces in the read path.
=20
The first piece is nfs_direct_read_wait, which works the same whether t=
he
process is waiting for a read or a write.
=20
Test plan:
Compile kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 487b83723ed4d4eaafd5109f36560da4f15c6578
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:30 2006 -0500
NFS: support EIOCBQUEUED return in direct read path
=20
For async iocb's, the NFS direct read path should return EIOCBQUEUED an=
d
call aio_complete when all the requested reads are finished. The
synchronous part of the NFS direct read path behaves exactly as it was
before.
=20
Test plan:
aio-stress with "-O". OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 99514f8fdda2beef1ca922b7f9d89c1a2c57fec0
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:30 2006 -0500
NFS: make iocb available everywhere in direct read path
=20
Pass the iocb argument all the way down to the direct read request
scheduler, and make it available in nfs_direct_read_result.
=20
Test plan:
Compile the kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
Millions of fsx-odirect ops. OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 0cdd80d07fb0f558dfdb30f6e0b9905f5e5475f1
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:29 2006 -0500
NFS: remove support for multi-segment iovs in the direct read path
=20
Eliminate the persistent use of automatic storage in all parts of the N=
FS
client's direct read path to pave the way for introducing support for a=
io
against files opened with the O_DIRECT flag.
=20
Test plan:
Compile the kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled.
Millions of fsx-odirect ops. OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 5dd602f20688e08c85ac91e0451c4e6321ed25d7
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:29 2006 -0500
NFS: use size_t type for holding rsize bytes in NFS O_DIRECT read path
=20
size_t is used for holding byte counts, so use it for variables storing=
rsize.
Note that the write path will be updated as we add support for async
O_DIRECT writes.
=20
Test plan:
Need to verify that existing comparisons against new size_t variables b=
ehave
correctly.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit d4cc948ba97980c55a308eab167a695109796456
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:28 2006 -0500
NFS: update comments and function definitions in fs/nfs/direct.c
=20
Update to latest coding style standards. Remove block comments on
statically defined functions, and place function definitions all on
one line.
=20
Test plan:
Compile kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit b8a32e2b8b7fefff994c89d398b6ac920a195b43
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:28 2006 -0500
NFS: clean up NFS client's a_ops->direct_IO method
=20
The NFS client's a_ops->direct_IO method, nfs_direct_IO, is required to
be present to allow NFS files to be opened with O_DIRECT, but is never
called because the NFS client shunts reads and writes to files opened
with O_DIRECT directly to its own routines.
=20
Gut the nfs_direct_IO function. This eliminates the only part of the
NFS client's direct I/O path that requires support for multi-segment
iovs, allowing further simplification in subsequent patches.
=20
Test plan:
Compile the kernel with CONFIG_NFS and CONFIG_NFS_DIRECTIO enabled. Mi=
llions
of fsx-odirect ops. OraSim.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit ec06c096edec0755534c7126f4caded69de131c2
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:27 2006 -0500
NFS: Cleanup of NFS read code
=20
Same callback hierarchy inversion as for the NFS write calls. This patc=
h is
not strictly speaking needed by the O_DIRECT code, but avoids confusing
differences between the asynchronous read and write code.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 788e7a89a03e364855583c0ab4649b94925efbb9
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:27 2006 -0500
NFS: Cleanup of NFS write code in preparation for asynchronous o_direct
=20
This patch inverts the callback hierarchy for NFS write calls.
=20
Instead of having the NFSv2/v3/v4-specific code set up the RPC callback
ops, we allow the original caller to do so. This allows for more
flexibility w.r.t. how to set up and tear down the nfs_write_data
structure while still allowing the NFSv3/v4 code to perform error
handling.
=20
The greater flexibility is needed by the asynchronous O_DIRECT code, wh=
ich
wants to be able to hold on to the original nfs_write_data structures a=
fter
the WRITE RPC call has completed in order to be able to replay them if =
the
COMMIT call determines that the server has rebooted.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 7117bf3dfb10b534a017260d9fc643bc1d0afd2a
Author: J. Bruce Fields <[email protected]>
Date: Mon Mar 20 13:44:26 2006 -0500
lockd: Remove FL_LOCKD flag
=20
Currently lockd identifies its own locks using the FL_LOCKD flag. This
doesn't scale well to multiple lock managers--if we did this in nfsv4 t=
oo,
for example, we'd be left with only one free flag bit.
=20
Instead, we just check whether the file manager ops (fl_lmops) set on t=
his
lock are our own.
=20
The only use for this is in nlm_traverse_locks, which uses it to find l=
ocks
that need cleaning up when freeing a host or a file.
=20
In the long run it might be nice to do reference counting instead of
traversing all the locks like this....
=20
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 8dc7c3115b611c00006eac3ee5b108296432aab7
Author: Andy Adamson <[email protected]>
Date: Mon Mar 20 13:44:26 2006 -0500
locks,lockd: fix race in nlmsvc_testlock
=20
posix_test_lock() returns a pointer to a struct file_lock which is unpr=
otected
and can be removed while in use by the caller. Move the conflicting lo=
ck from
the return to a parameter, and copy the conflicting lock.
=20
In most cases the caller ends up putting the copy of the conflicting lo=
ck on
the stack. On i386, sizeof(struct file_lock) appears to be about 100 b=
ytes.
We're assuming that's reasonable.
=20
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 2e0af86f618c697b44e2d67dff151256c58201c4
Author: Andy Adamson <[email protected]>
Date: Mon Mar 20 13:44:26 2006 -0500
locks: remove unused posix_block_lock
=20
posix_lock_file() is used to add a blocked lock to Lockd's block, so
posix_block_lock() is no longer needed.
=20
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit a85f193e2fb7d53e48ae6a9d9ea990bfb4cea555
Author: Andy Adamson <[email protected]>
Date: Mon Mar 20 13:44:25 2006 -0500
lockd: make nlmsvc_lock use only posix_lock_file
=20
Reorganize nlmsvc_lock() to make full use of posix_lock_file(), which d=
oes
eveything nlmsvc_lock() needs - no need to call posix_test_lock(),
posix_locks_deadlock(), or posix_block_lock() separately.
=20
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 5de0e5024a4e21251fd80dbfdb83316ce97086bc
Author: Andy Adamson <[email protected]>
Date: Mon Mar 20 13:44:25 2006 -0500
lockd: simplify nlmsvc_grant_blocked
=20
Reorganize nlmsvc_grant_blocked() to make full use of posix_lock_file()=
. Note
that there's no need for separate calls to posix_test_lock(),
posix_locks_deadlock(), or posix_block_lock().
=20
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 15dadef9460ad8d3b1d5ede1c1697dc79af44a72
Author: Andy Adamson <[email protected]>
Date: Mon Mar 20 13:44:24 2006 -0500
lockd: clean up nlmsvc_lock
=20
Slightly more consistent dprintk error reporting, consolidate some up()=
's.
=20
Signed-off-by: Andy Adamson <[email protected]>
Signed-off-by: J. Bruce Fields <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 1e7cb3dc12dbbac690d78c84f9c7cb11132ed121
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:24 2006 -0500
NFS: directory trace messages
=20
Reuse NFSDBG_DIRCACHE and NFSDBG_LOOKUPCACHE to provide additional
diagnostic messages that trace the operation of the NFS client's
directory behavior. A few new messages are now generated when NFSDBG_V=
FS
is active, as well, to trace normal VFS activity. This compromise
provides better trace debugging for those who use pre-built kernels,
without adding a lot of extra noise to the standard debug settings.
=20
Test-plan:
Enable NFS trace debugging with flags 1, 2, or 4. You should be able t=
o
see different types of trace messages with each flag setting.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit e95b85ec9d8c8ad4667f746aa4c9d22c281efc44
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:23 2006 -0500
SUNRPC: minor cleanup
=20
RPC_DEBUG_DATA no longer needed in net/sunrpc/xprt.c.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit dead28da8e3fb32601d38fb32b7021122e0a3d21
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:23 2006 -0500
SUNRPC: eliminate rpc_call()
=20
Clean-up: replace rpc_call() helper with direct call to rpc_call_sync.
=20
This makes NFSv2 and NFSv3 synchronous calls more computationally
efficient, and reduces stack consumption in functions that used to
invoke rpc_call more than once.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled. Connectathon on NFS version 2,
version 3, and version 4 mount points.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit cc0175c1dc1de8f6af0eb0631dcc5b999a6fcc42
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:22 2006 -0500
SUNRPC: display human-readable procedure name in rpc_iostats output
=20
Add fields to the rpc_procinfo struct that allow the display of a
human-readable name for each procedure in the rpc_iostats output.
=20
Also fix it so that the NFSv4 stats are broken up correctly by
sub-procedure number. NFSv4 uses only two real RPC procedures:
NULL, and COMPOUND.
=20
Test plan:
Mount with NFSv2, NFSv3, and NFSv4, and do "cat /proc/self/mountstats".
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 4ece3a2d18fd7fe1d4972284a8c98c569020093f
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:22 2006 -0500
NFS: add RPC I/O statistics to /proc/self/mountstats
=20
NFS client now shows various RPC I/O metrics in /proc/self/mountstats.
=20
Test plan:
Mount/umount while doing "cat /proc/self/mountstats", multiple iteratio=
ns
of connectathon locking suite. Test with NFS version 2, 3, and 4.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 11c556b3d8d481829ab5f9933a25d29b00913b5a
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:22 2006 -0500
SUNRPC: provide a mechanism for collecting stats in the RPC client
=20
Add a simple mechanism for collecting stats in the RPC client. Stats a=
re
tabulated during xprt_release. Note that per_cpu shenanigans are not
required here because the RPC client already serializes on the transpor=
t
write lock.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled. Basic performance regression
testing with high-speed networking and high performance server.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit ef759a2e54ed434b2f72b52a14edecd6d4eadf74
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:17 2006 -0500
SUNRPC: introduce per-task RPC iostats
=20
Account for various things that occur while an RPC task is executed.
Separate timers for RPC round trip and RPC execution time show how
long RPC requests wait in queue before being sent. Eventually these
will be accumulated at xprt_release time in one place where they can
be viewed from userland.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 262ca07de4d7f1bff20361c1353bb14b3607afb2
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:16 2006 -0500
SUNRPC: add a handful of per-xprt counters
=20
Monitor generic transport events. Add a transport switch callout to
format transport counters for export to user-land.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit e19b63dafdf7d615b0d36b90990a07e7792b9d3a
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:15 2006 -0500
SUNRPC: track length of RPC wait queues
=20
RPC wait queue length will eventually be exported to userland via the R=
PC
iostats interface.
=20
Test plan:
Compile kernel with CONFIG_NFS enabled.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 67ec9f46b889bfb1ab0a4e307d53929d5f0692bf
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:15 2006 -0500
NFS: report how long an NFS file system has been mounted
=20
Add a field in nfs_server to record a timestamp when a mount succeeds.
Report the number of seconds the file system has been mounted via
nfs_show_stats().
=20
Test plan:
Mount an NFS file system, watch the mountstats reports and compare with
clock time.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 006ea73e5fa82915d0ac7a3f15ee7c688433236d
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:14 2006 -0500
NFS: add hooks to account for NFSERR_JUKEBOX errors
=20
Make an inode or an nfs_server struct available in the logic that handl=
es
JUKEBOX/DELAY type errors so the NFS client can account for them.
=20
This patch is split out from the main nfs iostat patch to highlight min=
or
architectural changes required to support this statistic.
=20
Test plan:
None.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 91d5b47023b608227d605d1e916b29dd0215bff7
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:14 2006 -0500
NFS: add I/O performance counters
=20
Invoke the byte and event counter macros where we want to count bytes a=
nd
events.
=20
Clean-up: fix a possible NULL dereference in nfs_lock, and simplify
nfs_file_open.
=20
Test-plan:
fsx and iozone on UP and SMP systems, with and without pre-emption. Wa=
tch
for memory overwrite bugs, and performance loss (significantly more CPU
required per op).
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit d9ef5a8c26aab09762afce43df64736720b4860e
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:13 2006 -0500
NFS: introduce mechanism for tracking NFS client metrics
=20
Add a per-superblock performance counter facility to the NFS client. T=
his
facility mimics the counters available for block devices and for
networking. Expose these new counters via the new /proc/self/mountstat=
s
interface.
=20
Thanks to Andrew Morton and Trond Myklebust for their review and commen=
ts.
=20
Test plan:
fsx and iozone on UP and SMP systems, with and without pre-emption. Wa=
tch
for memory overwrite bugs, and performance loss (significantly more CPU
required per op).
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit c8bded96aa8735823e53c95a26177987ebb19a90
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:13 2006 -0500
NFS: clean up some mount options
=20
Get rid of "lock" and "posix", and spell out "vers=3D".
=20
Test plan:
None.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 7a480e250c7ca9187275d8574ae9e48a6b602cb9
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:12 2006 -0500
NFS: show retransmit settings when displaying mount options
=20
Sometimes it's important to know the exact RPC retransmit settings the
kernel is using for an NFS mount point. Add this facility to the NFS
client's show_options method.
=20
Test plan:
Set various retransmit settings via the mount command, and check that t=
he
settings are reflected in /proc/mounts.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit b4629fe2f094b719847f31be1ee5ab38300038b2
Author: Chuck Lever <[email protected]>
Date: Mon Mar 20 13:44:12 2006 -0500
VFS: New /proc file /proc/self/mountstats
=20
Create a new file under /proc/self, called mountstats, where mounted fi=
le
systems can export information (configuration options, performance coun=
ters,
and so on). Use a mechanism similar to /proc/mounts and s_ops->show_op=
tions.
=20
This mechanism does not violate namespace security, and is safe to use =
while
other processes are unmounting file systems.
=20
Thanks to Mike Waychison for his review and comments.
=20
Test-plan:
Test concurrent mount/unmount operations while cat'ing /proc/self/mount=
stats.
=20
Signed-off-by: Chuck Lever <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 1356b8c28d67cafd74f7e7dcfb39bf53681790a5
Author: Levent Serinol <[email protected]>
Date: Mon Mar 20 13:44:11 2006 -0500
SUNRPC: more verbose output for rpc auth weak error
=20
This patch adds server ip address to be printed out when "server
requires stronger authentication" error occured.
=20
Signed-off-by: Levent Serinol <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit 24bd68f46b1ad08d69bf32779f860df867780a7a
Author: Goldwyn Rodrigues <[email protected]>
Date: Mon Mar 20 13:44:11 2006 -0500
NFS: Code comments update in NFS
=20
read_cache_mtime is no longer used in nfs_inode. This patch removes
references of read_cache_mtime in the code comments.
=20
Signed-off-by: Goldwyn Rodrigues <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit c9d5128a10a4974f72674ff3463da4db439e8b04
Author: Ingo Molnar <[email protected]>
Date: Mon Mar 20 13:44:11 2006 -0500
NFS: sem2mutex idmap.c
=20
semaphore to mutex conversion.
=20
the conversion was generated via scripts, and the result was validated
automatically via a script as well.
=20
build and boot tested.
=20
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit bd6475454c774bd9dbe6078d94bbf72b1d3b65f4
Author: Eric Sesterhenn <[email protected]>
Date: Mon Mar 20 13:44:10 2006 -0500
NFS: kzalloc conversion in fs/nfs
=20
this converts fs/nfs to kzalloc() usage.
compile tested with make allyesconfig
=20
Signed-off-by: Eric Sesterhenn <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit a162a6b804b48c605d1fd35e1861a5d32d00ad3f
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:10 2006 -0500
NFSv4: Kill braindead gcc warnings
=20
nfs4_open_revalidate: 'res' may be used uninitialized
nfs4_callback_compound: =E2=80=98hdr_res.nops=E2=80=99 may be used unin=
itialized
'op_nr=E2=80=99 may be used uninitialized
encode_getattr_res: =E2=80=98savep=E2=80=99 may be used uninitialized
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 967b9281361481aecf323563886ef972ee88c681
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:09 2006 -0500
NFSv4: Do not call rpciod_down() before call to destroy_nfsv4_state()
=20
The reason is that the idmapper cleanup may call flush_workqueue() on
rpciod_workqueue.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 12de3b35ea549c5819f287508d7afab0bf3ac44d
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:09 2006 -0500
SUNRPC: Ensure that rpc_mkpipe returns a refcounted dentry
=20
If not, we cannot guarantee that idmap->idmap_dentry, gss_auth->dentry =
and
clnt->cl_dentry are valid dentries.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 24c5d9d7ea5a64fb5f157d17aa2c67a3300f8a08
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:08 2006 -0500
SUNRPC: Run rpci->queue_timeout on the rpciod workqueue instead of gene=
ric
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit f344f6df4b2baa3e5c553c461735dfaf92f44be7
Author: Olaf Kirch <[email protected]>
Date: Mon Mar 20 13:44:08 2006 -0500
SUNRPC: Auto-load RPC authentication kernel modules
=20
This patch adds a request_module call to rpcauth_create which will try
to auto-load the kernel module for the requested authentication flavor.
For kernels with modular sunrpc, this reduces the admin overhead for
the user.
=20
Signed-off-by: Olaf Kirch <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit fb374d24f225f38f13dbffb65dd7ec72daf08dba
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:08 2006 -0500
NFS: reduce the number of false cache invalidations.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit c8d149f3dbd582a101aa7da7bdd6c3316efd11b4
Author: Jesper Juhl <[email protected]>
Date: Mon Mar 20 13:44:07 2006 -0500
NFS: "const static" vs "static const" in nfs4
=20
My previous "const static" vs "static const" cleanup missed a single ca=
se,
patch below takes care of it.
=20
Signed-off-by: Jesper Juhl <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit ca62b9c3f7b8679ada4de94d2ab7098c6860c3d7
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:07 2006 -0500
NFSv4: Don't invalidate cached attributes if change attribute is unchan=
ged
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 755c1e20cd2ad56e5c567fa05769eb98a3eef72b
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:06 2006 -0500
NFS: writes should not clobber utimes() calls
=20
Ensure that we flush out writes in the case when someone calls utimes()=
in
order to set the file times.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 7bab377fcb495ee2e5a1cd69d235f8d84c76e3af
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:06 2006 -0500
lockd: Don't expose the process pid to the NLM server
=20
Instead we use the nlm_lockowner->pid.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 36943fa4b2701b9ef2d60084c85ecbe634aec252
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:05 2006 -0500
NLM: nlm_alloc_call should not immediately fail on signal
=20
Currently, nlm_alloc_call tests for a signal before it even tries to
allocate memory.
Fix it so that it tries at least once.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 47831f35b83e43c804215712dd0c834c92e8a441
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:05 2006 -0500
VFS: Fix __posix_lock_file() copy of private lock area
=20
The struct file_lock->fl_u area must be copied using the fl_copy_lock()
operation.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit 1dd594b21b2d98e56f2b1fe92bb222276b28de41
Author: Neil Brown <[email protected]>
Date: Mon Mar 20 13:44:04 2006 -0500
NFS: Fix buglet in fs/nfs/write.c
=20
I've been reading through fs/nfs/write.c trying to track down a bug
that seems to be related to pages loosing a refcount and getting
freed too early (you interested in detail??) and I spotted a little
bug which the following patch should fix.
=20
Signed-off-by: Neil Brown <[email protected]>
Signed-off-by: Trond Myklebust <[email protected]>
commit cd52ed35535ef443f08bf5cd3331d350272885b8
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:04 2006 -0500
NFS: Avoid races between writebacks and truncation
=20
Currently, there is no serialisation between NFS asynchronous writeback=
s
and truncation at the page level due to the fact that nfs_sync_inode()
cannot lock the pages that it is about to write out.
=20
This means that it is possible to be flushing out data (and calling som=
ething
like set_page_writeback()) while the page cache is busy evicting the pa=
ge.
Oops...
=20
Use the hooks provided in try_to_release_page() to ensure that dirty pa=
ges
are always written back to storage before we evict them.
=20
Signed-off-by: Trond Myklebust <[email protected]>
commit b92dccf65bab3b6b7deb79ff3321dc256eb0f53b
Author: Trond Myklebust <[email protected]>
Date: Mon Mar 20 13:44:03 2006 -0500
NFS: Fix a busy inodes issue...
=20
The nfs_open_context may live longer than the file descriptor that spaw=
ned
it, so it needs to carry a reference to the vfsmount. If not, then
generic_shutdown_super() may end up being called before reads and write=
s
have been flushed out.
=20
Make a couple of functions static while we're at it...
=20
Signed-off-by: Trond Myklebust <[email protected]>
> commit 47989d7454398827500d0e73766270986a3b488f
> Author: Chuck Lever <[email protected]>
> Date: Mon Mar 20 13:44:32 2006 -0500
>
> NFS: remove support for multi-segment iovs in the direct write path
>
> Eliminate the persistent use of automatic storage in all parts of the
> NFS client's direct write path to pave the way for introducing support
> for aio against files opened with the O_DIRECT flag.
NACK. We have patches pending that consolidate ->aio_read/write and
->read/writev into one operation. this change is completely counterproductive
towards that goal which has been discussed on -fsdevel for a while.
Trond Myklebust wrote:
> Hi Linus,
>
> Please pull from the repository at
>
> git pull git://git.linux-nfs.org/pub/linux/nfs-2.6.git
>
> This will update the following files through the appended changesets.
It would IMO be nice if you would pipe the changelog through 'git
shortlog', to better summarize the submission. And if the cumulative
patch isn't over 100k or so, include that in the email for any last
minute review.
I've attached the simple script I use to generate pull emails for
Andrew/Linus...
Jeff
_______________________________________________
NFSv4 mailing list
[email protected]
http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4
On Tue, 2006-03-21 at 17:46 +0000, Christoph Hellwig wrote:
> > commit 47989d7454398827500d0e73766270986a3b488f
> > Author: Chuck Lever <[email protected]>
> > Date: Mon Mar 20 13:44:32 2006 -0500
> >
> > NFS: remove support for multi-segment iovs in the direct write path
> >
> > Eliminate the persistent use of automatic storage in all parts of the
> > NFS client's direct write path to pave the way for introducing support
> > for aio against files opened with the O_DIRECT flag.
>
> NACK. We have patches pending that consolidate ->aio_read/write and
> ->read/writev into one operation. this change is completely counterproductive
> towards that goal which has been discussed on -fsdevel for a while.
How so? The interface is _exactly_ the same as before.
We never had support for multiple iovecs in O_DIRECT, but were passing
around a single iovec entry deep into code that couldn't care less.
Cheers,
Trond
On Tue, 2006-03-21 at 12:50 -0500, Jeff Garzik wrote:
> Trond Myklebust wrote:
> > Hi Linus,
> >
> > Please pull from the repository at
> >
> > git pull git://git.linux-nfs.org/pub/linux/nfs-2.6.git
> >
> > This will update the following files through the appended changesets.
>
> It would IMO be nice if you would pipe the changelog through 'git
> shortlog', to better summarize the submission. And if the cumulative
> patch isn't over 100k or so, include that in the email for any last
> minute review.
The cumulative patch in this case is 291k (smaller than usual, but not
insignificant).
In any case, the individual patches and the cumulative patch are
available, as always, from linux-nfs.org. Please see
http://client.linux-nfs.org/Linux-2.6.x/2.6.16/
> I've attached the simple script I use to generate pull emails for
> Andrew/Linus...
Thanks.
Cheers,
Trond
On Tue, Mar 21, 2006 at 01:08:52PM -0500, Trond Myklebust wrote:
> We never had support for multiple iovecs in O_DIRECT, but were passing
> around a single iovec entry deep into code that couldn't care less.
anthing that moves from iovecs back to plain buffers is counterproductive.
The plan is that every fullblown fs will only deal with iovecs, onlt drivers
and synthetic filesystems will implement the plain buffers.
On Tue, Mar 21, 2006 at 01:03:47PM -0500, Chuck Lever wrote:
> and when do you expect these changes to be integrated?
I'd expect in 2.6.18
> we've had this
> implementation for almost a year, and zach brown has reviewed it.
would be nice to post such things to -fsdevel and -aio..
On Tue, 2006-03-21 at 17:46 +0000, Christoph Hellwig wrote:
> > commit 47989d7454398827500d0e73766270986a3b488f
> > Author: Chuck Lever <[email protected]>
> > Date: Mon Mar 20 13:44:32 2006 -0500
> >
> > NFS: remove support for multi-segment iovs in the direct write path
> >
> > Eliminate the persistent use of automatic storage in all parts of the
> > NFS client's direct write path to pave the way for introducing support
> > for aio against files opened with the O_DIRECT flag.
>
> NACK. We have patches pending that consolidate ->aio_read/write and
> ->read/writev into one operation. this change is completely counterproductive
> towards that goal which has been discussed on -fsdevel for a while.
Could you, BTW, please post these pending patches? They are clearly not
in -mm, since these patches have been there for some time without
conflicting with anything else. The only references I could find were to
the following thread:
http://thread.gmane.org/gmane.linux.file-systems/7748
which basically just adds a BUG() if you call NFS O_DIRECT with a
multi-segment iovec.
I can't see from those patches how Chuck's O_DIRECT changes can be
construed as being "counterproductive". AFAICS those changes are
completely orthogonal to your proposal.
Cheers,
Trond
On Tue, 2006-03-21 at 18:57 +0000, Christoph Hellwig wrote:
> On Tue, Mar 21, 2006 at 01:08:52PM -0500, Trond Myklebust wrote:
> > We never had support for multiple iovecs in O_DIRECT, but were passing
> > around a single iovec entry deep into code that couldn't care less.
>
> anthing that moves from iovecs back to plain buffers is counterproductive.
> The plan is that every fullblown fs will only deal with iovecs, onlt drivers
> and synthetic filesystems will implement the plain buffers.
You need to do more than just add an iovec argument to
nfs_file_direct_read()/nfs_file_direct_write() if you want to achieve
this. The new call interface actually just clarifies something that was
implicit in the old one.
As I said in my other posting, I believe Chuck's changes are relatively
orthogonal to what you want to do: they neither make the low-level
plumbing better or worse for readv()/writev().
We'd be happy to work with you in the run-up to 2.6.18 to add
multi-segment support for the existing patchsets. It makes more sense to
me to append that functionality to the existing patchsets rather than
trigger a complete rewrite (and thus have a sh_tload more code to
retest).
Cheers,
Trond
_______________________________________________
NFSv4 mailing list
[email protected]
http://linux-nfs.org/cgi-bin/mailman/listinfo/nfsv4
On Tue, Mar 21, 2006 at 01:58:01PM -0500, Trond Myklebust wrote:
> Could you, BTW, please post these pending patches? They are clearly not
> in -mm, since these patches have been there for some time without
> conflicting with anything else. The only references I could find were to
> the following thread:
>
> http://thread.gmane.org/gmane.linux.file-systems/7748
>
> which basically just adds a BUG() if you call NFS O_DIRECT with a
> multi-segment iovec.
Yes, that's the latest public version. and nfs is one of the last missing
items.
On Tue, Mar 21, 2006 at 02:12:03PM -0500, Chuck Lever wrote:
> i have been watching the multi-segment iovec work since then, and fully
> intended to add the support for readv/writev aio in the NFS direct path
> when the generic support becomes available.
we agreed to not add another set of methods but rather consolidate the
existing two sets of aio and vectored methods into one. So to merge the
core support all users including nfs need to be updated. the last wip
patchset is posted in the at:
http://marc.theaimsgroup.com/?l=linux-kernel&m=114177713027505&w=2
http://marc.theaimsgroup.com/?l=linux-kernel&m=114177739515518&w=2
http://marc.theaimsgroup.com/?l=linux-kernel&m=114177739313588&w=2
http://marc.theaimsgroup.com/?l=linux-kernel&m=114177739407636&w=2
any nfs work should happen ontop of that.
I'm sill the opinion removing the iovec arguments isn't helpfull, but
it's your code and if you think it helps you move forwarg go ahead with
this.
-------------------------------------------------------
This SF.Net email is sponsored by xPML, a groundbreaking scripting language
that extends applications into web and mobile media. Attend the live webcast
and join the prime developer group breaking into this new coding territory!
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642
_______________________________________________
NFS maillist - [email protected]
https://lists.sourceforge.net/lists/listinfo/nfs
On Wed, 2006-03-22 at 10:30 +0000, Christoph Hellwig wrote:
> On Tue, Mar 21, 2006 at 02:12:03PM -0500, Chuck Lever wrote:
> > i have been watching the multi-segment iovec work since then, and fully
> > intended to add the support for readv/writev aio in the NFS direct path
> > when the generic support becomes available.
>
>
> we agreed to not add another set of methods but rather consolidate the
> existing two sets of aio and vectored methods into one. So to merge the
> core support all users including nfs need to be updated. the last wip
> patchset is posted in the at:
>
> http://marc.theaimsgroup.com/?l=linux-kernel&m=114177713027505&w=2
> http://marc.theaimsgroup.com/?l=linux-kernel&m=114177739515518&w=2
> http://marc.theaimsgroup.com/?l=linux-kernel&m=114177739313588&w=2
> http://marc.theaimsgroup.com/?l=linux-kernel&m=114177739407636&w=2
>
> any nfs work should happen ontop of that.
>
> I'm sill the opinion removing the iovec arguments isn't helpfull, but
> it's your code and if you think it helps you move forwarg go ahead with
> this.
Adding multi-segment support on top of this is not going to be a huge
job since the new code already does a form of multi-segment tracking in
the lower layers. By that I mean that it has to convert an arbitrarily
sized user buffer into a series of asynchronous NFS requests to be put
on the wire and then track them until they have completed. That is
pretty much all you have to do for the multi-iovec case too.
Cheers,
Trond