Commit Graph

113 Commits

Author SHA1 Message Date
Konstantin Belousov
223073fd1a Do not synchronously start the nfsiod threads at all. The r212506
fixed the issues with file descriptor locks, but the same problems are
present for vnode lock/user map lock.

If the nfs_asyncio() cannot find the free nfsiod, schedule task to
create new nfsiod and return error. This causes fall back to the
synchronous i/o for nfs_strategy(), or does not start read at all in
the case of readahead. The caller that holds vnode and potentially
user map lock does not wait for kproc_create() to finish, preventing
the LORs.

The change effectively reverts r203072, because we never hand off the
request to newly created nfsiod thread anymore.

Reviewed by:	jhb
Tested by:	jhb, pluknet
MFC after:	3 weeks
2010-10-18 19:06:46 +00:00
Konstantin Belousov
57bfe0a9f8 Do not fork nfsiod directly from the vop methods. This causes LORs between
vnode lock and several locks needed during fork, like fd lock.

Instead, schedule the task to be executed in the taskqueue context. We
still waiting for the fork to finish, but the context of the thread
executing the task does not make real LORs with our vnode lock.

Submitted by:	pluknet at gmail com
Reviewed by:	jhb
Tested by:	pho
MFC after:	3 weeks
2010-09-12 19:06:08 +00:00
Rick Macklem
651ff543f8 Fix a race that can occur when nfs nfsiod threads are being created.
Without this patch it was possible for a different thread that calls
nfs_asyncio() to snitch a newly created nfsiod thread that was
intended for another caller of nfs_asyncio(), because the nfs_iod_mtx
mutex was unlocked while the new nfsiod thread was created. This patch
labels the newly created nfsiod, so that it is not taken by another
caller of nfs_asyncio(). This is believed to fix the problem reported
on the freebsd-stable email list under the subject:
FreeBSD NFS client/Linux NFS server issue.

Tested by:	to DOT my DOT trociny AT gmail DOT com
Reviewed by:	jhb
MFC after:	2 weeks
2010-01-27 15:22:20 +00:00
Doug Rabson
98c497255b Adjust the internal NFS KPI to avoid the last traces of NFS_LEGACYRPC.
Approved by: re
2009-06-30 19:10:17 +00:00
Doug Rabson
b49a2b39fd Remove the old kernel RPC implementation and the NFS_LEGACYRPC option.
Approved by: re
2009-06-30 19:03:27 +00:00
Robert Watson
86ce6a83d1 Remove the unmaintained University of Michigan NFSv4 client from 8.x
prior to 8.0-RELEASE.  Rick Macklem's new and more feature-rich NFSv234
client and server are replacing it.

Discussed with:	rmacklem
2009-05-22 12:35:12 +00:00
Attilio Rao
dfd233edd5 Remove the thread argument from the FSD (File-System Dependent) parts of
the VFS.  Now all the VFS_* functions and relating parts don't want the
context as long as it always refers to curthread.

In some points, in particular when dealing with VOPs and functions living
in the same namespace (eg. vflush) which still need to be converted,
pass curthread explicitly in order to retain the old behaviour.
Such loose ends will be fixed ASAP.

While here fix a bug: now, UFS_EXTATTR can be compiled alone without the
UFS_EXTATTR_AUTOSTART option.

VFS KPI is heavilly changed by this commit so thirdy parts modules needs
to be recompiled.  Bump __FreeBSD_version in order to signal such
situation.
2009-05-11 15:33:26 +00:00
John Baldwin
3429118095 When a stale file handle is encountered, purge all cached information about
an NFS node including the access and attribute caches.  Previously the NFS
client only purged any name cache entries associated with the file.

PR:		kern/123755
Submitted by:	Jaakko Heinonen  jh of saunalahti fi
Reported by:	Timo Sirainen  tss of iki fi
Reviewed by:	rwatson, rmacklem
MFC after:	1 month
2009-04-06 21:11:08 +00:00
John Baldwin
a58158ef08 Change the default timeout for caching attributes of a directory in the NFS
client from 30 seconds to 3 seconds.  After the recent changes to add
caching of negative name cache lookups, a negative name cache hit will
persist until the client notices the parent directory has changed.  The
higher the attribute cache timeout on directories, the longer that can take,
so lower the default timeout for directories to match that of regular files.

Suggested by:	bde, mohans
MFC after:	1 month
2009-04-06 19:12:47 +00:00
John Baldwin
2a3f3a09c6 Expand the per-node access cache to cache permissions for multiple users.
The number of entries in the cache defaults to 8 but is easily changed in
nfsclient/nfs.h.  When the cache is filled, the oldest cache entry is
evicted when space is needed.

I mirrored the changes to the NFSv[23] client in the NFSv4 client to fix
compile breakage.  However, the NFSv4 client doesn't actually use the
access cache currently.

Submitted by:	rmacklem
2009-03-20 21:12:38 +00:00
Doug Rabson
335317d291 Switch the default rpc implementation for NFS back to the new code. I believe
I have fixed the reported problems - if you still have trouble with it, please
contact me with as much detail as possible so that I can track down any other
issues as quickly as possible.
2008-11-14 11:27:53 +00:00
Doug Rabson
0eec5c87bf Temporarily switch NFS back to the old RPC code while I try to diagnose and
fix the problems a few people have noticed with the new code. People who want
to continue testing the new code or who need RPCSEC_GSS support should use
the new option NFS_NEWRPC to select it.
2008-11-13 11:35:18 +00:00
Doug Rabson
a9148abd9d Implement support for RPCSEC_GSS authentication to both the NFS client
and server. This replaces the RPC implementation of the NFS client and
server with the newer RPC implementation originally developed
(actually ported from the userland sunrpc code) to support the NFS
Lock Manager.  I have tested this code extensively and I believe it is
stable and that performance is at least equal to the legacy RPC
implementation.

The NFS code currently contains support for both the new RPC
implementation and the older legacy implementation inherited from the
original NFS codebase. The default is to use the new implementation -
add the NFS_LEGACYRPC option to fall back to the old code. When I
merge this support back to RELENG_7, I will probably change this so
that users have to 'opt in' to get the new code.

To use RPCSEC_GSS on either client or server, you must build a kernel
which includes the KGSSAPI option and the crypto device. On the
userland side, you must build at least a new libc, mountd, mount_nfs
and gssd. You must install new versions of /etc/rc.d/gssd and
/etc/rc.d/nfsd and add 'gssd_enable=YES' to /etc/rc.conf.

As long as gssd is running, you should be able to mount an NFS
filesystem from a server that requires RPCSEC_GSS authentication. The
mount itself can happen without any kerberos credentials but all
access to the filesystem will be denied unless the accessing user has
a valid ticket file in the standard place (/tmp/krb5cc_<uid>). There
is currently no support for situations where the ticket file is in a
different place, such as when the user logged in via SSH and has
delegated credentials from that login. This restriction is also
present in Solaris and Linux. In theory, we could improve this in
future, possibly using Brooks Davis' implementation of variant
symlinks.

Supporting RPCSEC_GSS on a server is nearly as simple. You must create
service creds for the server in the form 'nfs/<fqdn>@<REALM>' and
install them in /etc/krb5.keytab. The standard heimdal utility ktutil
makes this fairly easy. After the service creds have been created, you
can add a '-sec=krb5' option to /etc/exports and restart both mountd
and nfsd.

The only other difference an administrator should notice is that nfsd
doesn't fork to create service threads any more. In normal operation,
there will be two nfsd processes, one in userland waiting for TCP
connections and one in the kernel handling requests. The latter
process will create as many kthreads as required - these should be
visible via 'top -H'. The code has some support for varying the number
of service threads according to load but initially at least, nfsd uses
a fixed number of threads according to the value supplied to its '-n'
option.

Sponsored by:	Isilon Systems
MFC after:	1 month
2008-11-03 10:38:00 +00:00
Doug Rabson
c675522fc4 Re-implement the client side of rpc.lockd in the kernel. This implementation
provides the correct semantics for flock(2) style locks which are used by the
lockf(1) command line tool and the pidfile(3) library. It also implements
recovery from server restarts and ensures that dirty cache blocks are written
to the server before obtaining locks (allowing multiple clients to use file
locking to safely share data).

Sponsored by:	Isilon Systems
PR:		94256
MFC after:	2 weeks
2008-06-26 10:21:54 +00:00
Mohan Srinivasan
faf529dce5 NFS MP scaling changes.
- Eliminate the hideous nfs_sndlock that serialized NFS/TCP request senders
  thru the sndlock.
- Institute a new nfs_connectlock that serializes NFS/TCP reconnects. Add
  logic to wait for pending request senders to finish sending before
  reconnecting. Dial down the sb_timeo for NFS/TCP sockets to 1 sec.
- Break out the nfs xid manipulation under a new nfs xid lock, rather than
  over loading the nfs request lock for this purpose.
- Fix some of the locking in nfs_request.
Many thanks to Kris Kennaway for his help with this and for initiating the
MP scaling analysis and work. Kris also tested this patch thorougly.
Approved by: re@ (Ken Smith)
2007-10-12 19:12:21 +00:00
Bruce Evans
e43982a801 Unstaticize nfs_iosize() in nfsclient and use it in nfs4client instead
of duplicating it except for larger style bugs in the copy.

Fix some nearby style bugs (including a harmless type mismatch)
in and near the remaining copy.

This is part of fixing collisions of the 2 nfs*client's names.  Even
static names should have a unique prefixes so that they can be debugged
easily.
2007-01-25 13:07:25 +00:00
Mohan Srinivasan
a18c4dc336 1) Fix up locking in nfs_up() and nfs_down.
2) Reduce the acquisitions of the Giant lock in the nfs_socket.c paths significantly.
- We don't need to acquire Giant before tsleeping on lbolt anymore,
  since jhb specialcased lbolt handling in msleep.
- nfs_up() needs to acquire Giant only if printing the "server up"
  message.
- nfs_timer() held Giant for the duration of the NFS timer processing,
  just because the printing of the message in nfs_down() needed it
  (and we acquire other locks in nfs_timer()). The acquisition of
  Giant is moved down into nfs_down() now, reducing the time Giant is
  held in that path.

Reported by: Kris Kennaway
2006-11-20 04:14:23 +00:00
Bruce Evans
35259c2c89 Fixed some style bugs (especially ones involving long lines and use
of __P(())).  There are many more.
2006-10-17 22:07:07 +00:00
Mohan Srinivasan
5bbfbd1422 Since NFSv4 is not SMP safe, nfsiod needs to acquire Giant for NFSv4 mounts
before doing the read/write.

Reported by:	Chuck Lever.
2006-05-24 23:06:50 +00:00
Chuck Lever
6d0699a5ba NFS over TCP retransmit behavior should default to a 60 second time out,
mimicing the NFS reference implementation.

NFS over TCP does not need fast retransmit timeouts, since network loss
and congestion are managed by the transport (TCP), unlike with NFS over
UDP.  A long timeout prevents the unnecessary retransmission of non-
idempotent NFS requests.

Reviewed by:	mohans, silby, rees?
Sponsored by:	Network Appliance, Incorporated
2006-05-23 18:48:07 +00:00
Chuck Lever
94163ea283 Refactor the NFS over UDP retransmit timeout estimation logic to allow
the estimator to be more easily tuned and maintained.

There should be no functional change except there is now a lower limit
on the retransmit timeout to prevent the client from retransmitting
faster than the server's disks can fill requests, and an upper limit
to prevent the estimator from taking to long to retransmit during a
server outage.

Reviewed by:	mohan, kris, silby
Sponsored by:	Network Appliance, Incorporated
2006-05-23 18:33:58 +00:00
Mohan Srinivasan
f1cdf89911 Changes to make the NFS client MP safe.
Thanks to Kris Kennaway for testing and sending lots of bugs my way.
2006-05-19 00:04:24 +00:00
Paul Saab
4321eae6b7 Make nfs_timer() MPSAFE. With this change, the bottom half of the NFS
client (the interface with the protocol stack and callouts) is
Giant-free.

Submitted by:	Mohan Srinivasan.
2005-07-19 21:27:25 +00:00
Poul-Henning Kamp
56dd36b1a6 Remove unused cred arg from nfs_vinvalbuf() and many bogus arguments
passed for it.
2005-01-24 12:31:06 +00:00
Warner Losh
c398230b64 /* -> /*- for license, minor formatting changes 2005-01-07 01:45:51 +00:00
Paul Saab
a7500bceb0 First cut of NFS direct IO support.
- NFS direct IO completely bypasses the buffer and page caches.
  If a file is open for direct IO all caching is disabled.
- Direct IO for Directories will be addressed later.
- 2 new NFS directio related sysctls are added. One is a knob to
  disable NFS direct IO completely (direct IO is enabled by default).
  The other is to disallow mmaped IO on a file that has at least one
  O_DIRECT open (see the comment in nfs_vnops.c for more details).
  The default is to allow mmaps on a file that has O_DIRECT opens.

Submitted by:	Mohan Srinivasan mohans at yahoo-inc dot com
Obtained from:	Yahoo!
2004-12-15 22:20:22 +00:00
Paul Saab
35ec46b7f2 Rewrite of the NFS client's reply handling. We now have NFS socket
upcalls which do RPC header parsing and match up the reply with the
request. NFS calls now sleep on the nfsreq structure. This enables
us to eliminate the NFS recvlock.

Submitted by:	Mohan Srinivasan mohans at yahoo-inc dot com
2004-12-06 21:11:15 +00:00
Robert Watson
0fe6462ad5 Correct a typo in a comment. 2004-12-06 16:11:25 +00:00
Poul-Henning Kamp
8b431c9576 For reasons unknown, the nfs locking code used a fifo to send requests to
userland and a dedicated system call to get replies.

The vnode-bypass of fifos broke this into a panic.

Ditch all the magic and create a device /dev/nfslock instead, and
use that for both directions apart from the shorter path, this is
also faster because the device driver runs Giant free using the
vnode bypass.

Noticed by:	marcel
2004-12-06 08:31:32 +00:00
Poul-Henning Kamp
9f2b7bc4a8 style consistency. 2004-09-27 19:44:39 +00:00
Poul-Henning Kamp
35f134080f Explicitly pass vnode to nfs_doio() and mountpoint to nfs_asyncio(). 2004-09-07 08:56:43 +00:00
Alfred Perlstein
c713aaaeca NFS mobility PHASE I, II & III (phase VI, and V pending):
Rebind the client socket when we experience a timeout.  This fixes
the case where our IP changes for some reason.

Signal a VFS event when NFS transitions from up to down and vice
versa.

Add a placeholder vfs_sysctl where we will put status reporting
shortly.

Also:
Make down NFS mounts return EIO instead of EINTR when there is a
soft timeout or force unmount in progress.
2004-07-06 09:12:03 +00:00
Peter Edwards
1630ff08a3 Clean up properly when unloading NFS client module.
This includes a modified form of some code from Thomas Moestl (tmm@)
to properly clean up the UMA zone and the "nfsnodehashtbl" hash
table.

Reviewed By:	iedowse
PR:		16299
2004-04-11 13:30:20 +00:00
Warner Losh
2fcbca0d85 Remove advertising clause from University of California Regent's
license, per letter dated July 22, 1999 and email from Peter Wemm,
Alan Cox and Robert Watson.

Approved by: core, peter, alc, rwatson
2004-04-07 05:00:01 +00:00
Jim Rees
f9955a5f53 only do nfs rpc callouts if there is work to do.
Submitted by:	kan
Approved by:	alfred
2004-03-25 21:48:09 +00:00
Peter Wemm
6df0617286 Calculate NFS timeouts in units of 10ms, not 5ms. This matches the default
clock precision on i386.  This is a NOP change on i386.  But this stops
the mount_nfs units from suddenly changing to units of 1/20 of a second
(vs the normal 1/10 of a second) if HZ is increased.
2004-03-14 06:21:56 +00:00
Jim Rees
73c02c410e NFSv4 fixes from Connectathon 2004:
remove unused pid field of file context struct
map nfs4 error codes to errnos
eliminate redundant code from nfs4_request
use zero stateid on setattr that doesn't set file size
use same clientid on all mounts until reboot
invalidate dirty bufs in nfs4_close, to play it safe
open file for writing if truncating and it's not already open

Approved by:	alfred
2004-02-27 19:37:43 +00:00
Alfred Perlstein
90abe7f294 Use function pointers to remove the depenancy cross dependancy on nfs4
and the nfs3 client.  Also fix some bugs that happen to be causing crashes
in both v3 and v4 introduced by the v4 import.

Submitted by: Jim Rees <rees@umich.edu>
Approved by: re
2003-11-22 02:21:49 +00:00
Alfred Perlstein
42233ecdc1 Move the declaration for "struct nfs4_fctx" out from under #ifdef KERNEL
for fstat(1).
2003-11-15 05:03:15 +00:00
Alfred Perlstein
1bf8720450 University of Michigan's Citi NFSv4 kernel client code.
Submitted by: Jim Rees <rees@umich.edu>
2003-11-14 20:54:10 +00:00
Poul-Henning Kamp
0c183c5a56 Remove extern declarations of stuff which is static in nfs_node.c
Move related macro to nfs_node.c

Spotted by:	FlexeLint
2002-10-20 21:40:55 +00:00
Bruce Evans
6af7f1e511 Use `struct uma_zone *' instead of uma_zone_t, so that <sys/uma.h> isn't
a prerequisite.
2002-09-05 14:04:34 +00:00
Poul-Henning Kamp
9bf1a75697 Introduce typedefs for the member functions of struct vfsops and employ
these in the main filesystems.  This does not change the resulting code
but makes the source a little bit more grepable.

Sponsored by:	DARPA and NAI Labs.
2002-08-13 10:05:50 +00:00
John Baldwin
56e9ce41a5 In namei(), we use a NULL thread for uio_td when doing a VOP_READLINK().
nfs_readlink() calls nfs_bioread() which passes in uio_td as the thread
argument to nfs_getcacheblk().  In nfs_getcacheblk() we dereference the
thread pointer to get a process pointer to pass to nfs_sigintr().  This
obviously results in a panic. :)

Rather than change nfs_getcacheblk() to check if the thread pointer is
NULL when calling nfs_sigintr() like other callers do, change
nfs_sigintr() to take a thread as the last argument instead of a
process so none of the callers have to care if the thread is NULL or not.
2002-06-28 21:53:08 +00:00
Jeff Roberson
ab426dc822 Remove references to vm_zone.h and switch over to the new uma API. 2002-03-20 10:07:52 +00:00
Peter Wemm
1bde568682 Revise the nfsiod auto tuning code. Now both the upper and lower limits
are specifyable by sysctl and are respected.

Submitted by:	Maxime Henrion <mux@sneakerz.org>
2002-01-15 20:57:21 +00:00
Peter Wemm
117f61374c Implement vfs.nfs.iodmin (minimum number of nfsiod's) and
vfs.nfs.iodmaxidle (idle time before nfsiod's exit).  Make it adaptive
so that we create nfsiod's on demand and they go away after not being
used for a while.  The upper limit is NFS_MAXASYNCDAEMON (currently 20).
More will be done here, but this is a useful checkpoint.

Submitted by:	Maxime Henrion <mux@qualys.com>
2002-01-14 02:13:46 +00:00
Ian Dowse
1278d57acd Permit NFS filesystems to be forcibly unmounted when the server is
down, even if there are hung processes and the mount is non-
interruptible.

This works by having nfs_unmount call a new function nfs_nmcancelreqs()
in the FORCECLOSE case. It scans the list of outstanding requests
and marks as interrupted any requests belonging to the specified
mount. Then it waits up to 30 seconds for all requests to terminate.
A few other changes are necessary to support this:
- Unconditionally set a socket timeout so that even hard mounts
  are guaranteed to occasionally check the R_SOFTTERM flag on
  requests. For hard mounts this flag can only be set by
  nfs_nmcancelreqs().
- Reject requests on a mount that is currently being unmounted.
- Never grant the receive lock to a request that has been cancelled.

This should also avoid an old problem where a forced NFS unmount
could cause a crash; it occurred when a VOP on an unlocked vnode
(usually VOP_GETATTR) was in progress at the time of the forced
unmount.
2002-01-02 00:41:26 +00:00
Robert Watson
147839396c o Pass td into nfs_mountroot() to eliminate an XXX'd curthread use.
Since it's in the parent function anyway, might as well pass it
  another layer down.

Obtained from:	TrustedBSD Project
2001-12-31 21:00:00 +00:00
Matthew Dillon
3ebeaf5984 This fixes a large number of bugs in our NFS client side code. A recent
commit by Kirk also fixed a softupdates bug that could easily be triggered
by server side NFS.

	* An edge case with shared R+W mmap()'s and truncate whereby
	  the system would inappropriately clear the dirty bits on
	  still-dirty data.  (applicable to all filesystems)

	  THIS FIX TEMPORARILY DISABLED PENDING FURTHER TESTING.
	  see vm/vm_page.c line 1641

	* The straddle case for VM pages and buffer cache buffers when
	  truncating.  (applicable to NFS client side)

	* Possible SMP database corruption due to vm_pager_unmap_page()
	  not clearing the TLB for the other cpu's.  (applicable to NFS
	  client side but could effect all filesystems).  Note: not
	  considered serious since the corruption occurs beyond the file
	  EOF.

	* When flusing a dirty buffer due to B_CACHE getting cleared,
	  we were accidently setting B_CACHE again (that is, bwrite() sets
	  B_CACHE), when we really want it to stay clear after the write
	  is complete.  This resulted in a corrupt buffer.  (applicable
	  to all filesystems but probably only triggered by NFS)

	* We have to call vtruncbuf() when ftruncate()ing to remove
	  any buffer cache buffers.  This is still tentitive, I may
	  be able to remove it due to the second bug fix.  (applicable
	  to NFS client side)

	* vnode_pager_setsize() race against nfs_vinvalbuf()... we have
	  to set n_size before calling nfs_vinvalbuf or the NFS code
	  may recursively vnode_pager_setsize() to the original value
	  before the truncate.  This is what was causing the user mmap
	  bus faults in the nfs tester program.  (applicable to NFS
	  client side)

	* Fix to softupdates (see ufs/ffs/ffs_inode.c 1.73, commit made
	  by Kirk).

Testing program written by: Avadis Tevanian, Jr.
Testing program supplied by: jkh / Apple (see Dec2001 posting to freebsd-hackers with Subject 'NFS: How to make FreeBS fall on its face in one easy step')
MFC after:	1 week
2001-12-14 01:16:57 +00:00