openafs.git
14 years agoFix typo in AdminGuide
Asanka Herath [Fri, 15 Jan 2010 00:15:03 +0000]
Fix typo in AdminGuide

</pare> -> </para>

Change-Id: If0f8183f59f038b1b5e033d5ebd2607b5091f6da
Reviewed-on: http://gerrit.openafs.org/1103
Reviewed-by: Russ Allbery <rra@stanford.edu>
Tested-by: Russ Allbery <rra@stanford.edu>

14 years agovolmonitor keep vtrans lock
Dan Hyde [Wed, 13 Jan 2010 19:38:47 +0000]
volmonitor keep vtrans lock

VolMonitor must maintain the VTRANS lock the whole time it is walking
the allTrans list.  Failure to do so can cause core dumps trying to
access memory that has already been free'd.

Two versions of this change were coded and tested.  The other version
used the VTRANS lock only around a THOLD, but needed a TRELE, too.
Timing tests were run counting the number of vos status, vos listvol,
and vos backupsys operations that could be performed during a fixed
number of vos status operations.  The THOLD/TRELE version caused other
vos operations to run about 5% slower.

FIXES 126110

Change-Id: I7e749d30c955867faacafa978d7d643dee648ca3
Reviewed-on: http://gerrit.openafs.org/1098
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoKernel is always defined
Simon Wilkinson [Tue, 17 Nov 2009 20:46:17 +0000]
Kernel is always defined

When we build the kernel module, either in the real or the ukernel
case, KERNEL is defined. So, remove the #ifdefs around our locking
primitives, so that things are a little bit clearer.

Change-Id: I60893cc630d8cd4dd6cb11cdc44e8bf4f2f9f863
Reviewed-on: http://gerrit.openafs.org/838
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoMove GLOCK initialisation to platform directories
Simon Wilkinson [Tue, 17 Nov 2009 19:52:15 +0000]
Move GLOCK initialisation to platform directories

Rework the GLOCK initialisation code so that it's moved out into
platform directories, rather than all being done in osi_Init.

Change-Id: I1aae76ba12cd4e45f54881f5573ed1713159b64b
Reviewed-on: http://gerrit.openafs.org/837
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agorewrite afs_MemWriteBlk() using afs_MemWritevBlk()
Chas Williams [Tue, 5 Jan 2010 00:22:19 +0000]
rewrite afs_MemWriteBlk() using afs_MemWritevBlk()

this avoids the code duplication of extending/shrinking cache entries

Change-Id: Ieeae668caa506e934f92edf193b99dabdba2eabe
Reviewed-on: http://gerrit.openafs.org/1062
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoQuick fix for readpages when using memcache
Chas Williams [Tue, 5 Jan 2010 00:18:36 +0000]
Quick fix for readpages when using memcache

Just don't do readpages if we're using memcache

Change-Id: I6b8b45f291d106ad1303d5bea1caf7f910b7ecf4
Reviewed-on: http://gerrit.openafs.org/1061
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoafsd: decprecate -mem_alloc_sleep flag (and related code)
Chas Williams - CONTRACTOR [Thu, 7 Jan 2010 23:07:43 +0000]
afsd: decprecate -mem_alloc_sleep flag (and related code)

On most platforms, afs_osi_Alloc_NoSleep() is already afs_osi_Alloc().
I have tested this on Solaris (the major exception) and haven't seen
any problems.

Change-Id: I0e8bfd76b2bd25889c1491633a3a6d50aceac510
Reviewed-on: http://gerrit.openafs.org/1079
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoAlways check VInitVolumePackage2 return code
Andrew Deason [Mon, 11 Jan 2010 19:36:32 +0000]
Always check VInitVolumePackage2 return code

Some programs were not checking the return code of VInitVolumePackage2.
Although some programs do not care so much if we fail to properly init
the volume package (SYNC debug tools), at the very least log/print an
error, so it doesn't silently fail.

Other programs are changed to exit when VInitVolumePackage2 fail, so
e.g. salvages don't accidentally cause corruption on 'logging'-mounted
UFS partitions.

In any case, ensure the return code is always checked anywhere it is
called.

Change-Id: I164d5920a5ea2dd7e5ed9ad4ccc578e9bdf0db0b
Reviewed-on: http://gerrit.openafs.org/1090
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove weekly bosserver restarts
Simon Wilkinson [Wed, 13 Jan 2010 17:28:24 +0000]
Remove weekly bosserver restarts

Change the default so new installations of the bosserver have
no weekly restarts. Update the manpage and XML documentation to
reflect this change.

FIXES 126138

Change-Id: Ic22b750a602f6d2a22be881f5e1b04cd4fa132ae
Reviewed-on: http://gerrit.openafs.org/1097
Reviewed-by: Russ Allbery <rra@stanford.edu>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Rework PAG to group conversions
Simon Wilkinson [Sat, 19 Dec 2009 23:58:54 +0000]
Linux: Rework PAG to group conversions

The change to using a single group for PAGs on Linux led to a load
of special cases throughout the code. This patch moves the OS
dependent parts of this into the LINUX/ platform directory, and
rearranges that file so that there is a clear distinction between
the one and two group sections of the code.

Change-Id: Ib6e6f081519b23494dcf731b4be689c006180c68
Reviewed-on: http://gerrit.openafs.org/1007
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoWindows: remove signed vs unsigned error in smb_Init
Jeffrey Altman [Mon, 11 Jan 2010 19:35:12 +0000]
Windows: remove signed vs unsigned error in smb_Init

Use UINT_PTR instead of INT_PTR.

LICENSE MIT

Change-Id: Id10cd7b7a18688c9eee68e3eac9c20348519fd37
Reviewed-on: http://gerrit.openafs.org/1089
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: Protect buffers in smb_WriteData from simultaneous writes
Jeffrey Altman [Mon, 11 Jan 2010 19:21:11 +0000]
Windows: Protect buffers in smb_WriteData from simultaneous writes

smb_WriteData does not properly use CM_SCACHESYNC_WRITE to
protect buffers from simultaneous writes.  Instead of simply
testing CM_SCACHESYNC_WRITE at the top of the while loop,
the flag must remain set until the entire write completes.
cm_SyncOp is now called once and cm_SyncOpDone is only called
upon final success or error.

In addition, as 'count' is unsigned, the test for count < 0
is replaced with count != 0.

LICENSE MIT

Change-Id: I82c8dc20e62079b13bf305e906f4744756aa0ac2
Reviewed-on: http://gerrit.openafs.org/1087
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: fs examine owner and group are signed
Jeffrey Altman [Mon, 21 Dec 2009 17:35:12 +0000]
Windows: fs examine owner and group are signed

Change the owner array to be signed for displaying owner
and group within fs examine output.

LICENSE MIT

Change-Id: I807e3f1b3c07eb26c1782d6b76cd49977467ce33
Reviewed-on: http://gerrit.openafs.org/1017
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: do not call cm_SearchCellByDNS if AFS_AFSDB_ENV is not defined
Jeffrey Altman [Mon, 11 Jan 2010 18:59:12 +0000]
Windows: do not call cm_SearchCellByDNS if AFS_AFSDB_ENV is not defined

cm_SearchCellByDNS is only available when AFS_AFSDB_ENV is
defined.  Do not call it when it isn't.

LICENSE MIT

Change-Id: I203ab775513bd75ff6a547c557579c0bc095d555
Reviewed-on: http://gerrit.openafs.org/1086
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Asanka Herath <asanka@secure-endpoints.com>
Tested-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: buf_Recycle should clean up flags and reset version
Jeffrey Altman [Mon, 11 Jan 2010 18:51:42 +0000]
Windows: buf_Recycle should clean up flags and reset version

The CM_BUF_EOF and CM_BUF_ERROR buffer flags and the
buffer dataVersion should be reset within buf_Recycle()
instead of by the caller of buf_Recycle().

LICENSE MIT

Change-Id: I643da102553200b20f5ecc23ec43974581663a8e
Reviewed-on: http://gerrit.openafs.org/1085
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Asanka Herath <asanka@secure-endpoints.com>
Tested-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: do not leak scp->dirlock if cm_BPlusDirBuildTree fails
Jeffrey Altman [Sat, 9 Jan 2010 05:26:37 +0000]
Windows: do not leak scp->dirlock if cm_BPlusDirBuildTree fails

In cm_BeginDirOp, the scp->dirlock would be leaked if
cm_BPlusDirBuildTree() failed.   This would either result in
a panic later on if lock order validation is active; or as
an inability to process subsequent requests on the directory.

LICENSE MIT

Change-Id: I03afd0c9e6296c0f43ae39e5a7b1ff29a1619a43
Reviewed-on: http://gerrit.openafs.org/1083
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Asanka Herath <asanka@secure-endpoints.com>
Tested-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: alter nmtest CreateFile modes
Jeffrey Altman [Mon, 21 Dec 2009 17:29:10 +0000]
Windows: alter nmtest CreateFile modes

Change -r to open the file for FILE_SHARE_READ

Change -w to open the file for FILE_SHARE_WRITE

Change -w to open the file twice.  Once for sequential access
and once for random access.

LICENSE MIT

Change-Id: I295a7e7408e7e7eac06901108a726fe9e94bf18f
Reviewed-on: http://gerrit.openafs.org/1016
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Asanka Herath <asanka@secure-endpoints.com>
Tested-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: return ENOMEM from cm_BPlus functions on malloc failure
Jeffrey Altman [Tue, 8 Dec 2009 14:38:53 +0000]
Windows: return ENOMEM from cm_BPlus functions on malloc failure

There are several cm_BPlus functions that do not return
an error code on malloc failure.

LICENSE MIT

Change-Id: I3a3d78cc6d6844c78e51fa81ebf330f5af88a7a5
Reviewed-on: http://gerrit.openafs.org/900
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Rod Widdowson <rdw@steadingsoftware.com>
Reviewed-by: Asanka Herath <asanka@secure-endpoints.com>
Tested-by: Asanka Herath <asanka@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoUse dget_parent to handle getting inode of parent
Simon Wilkinson [Sun, 22 Nov 2009 23:31:00 +0000]
Use dget_parent to handle getting inode of parent

We can't safely do dp->d_parent->d_inode to access the parent inode
of a given dentry. Instead, use dget_parent() to get safely get a
reference on the parent dcache, and access the inode using that
reference. Dispose of this reference with dput() as appropriate.

Change-Id: Ic4f2159f68f95744bca08ebc99ecd5f45dc46966
Reviewed-on: http://gerrit.openafs.org/1075
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agofreebsd: CM changes targeting RELENG_8
Matt Benjamin [Tue, 5 Jan 2010 02:31:27 +0000]
freebsd:  CM changes targeting RELENG_8

Force all vnodes onto the fs mount queue when allocated.  This fixes
a long-standing vnode recycling problem.  Don't call vgone() on a vnode
whose refcount is 0.  Always destroy vnodes in VOP_RECLAIM.  This is work
in progress towards fixing old reclaim bug mentioned in Rees comment.
Hold vnode returned from gop_lookupname_user in afs_pioctl_syscall,
to avoid it going inactive before we're finished.  Also unlock it if
necessary.  Don't use custom vop_lock impl when AFS_FBSD80_ENV.  Remove
duplicate conditional code in vnode pretty-print (old cruft).  Also don't
format fid members as hex. Revert vn_lock exclusve in osi_VM_StoreAllSegments
(fixes deadlock introduced by me in a 2009 changeset).  Remove unused
variables in osi_VM_StoreAllSegments.

Change-Id: I9fd146d6f405382a20a75523ec2b75c62ac6d17a
Reviewed-on: http://gerrit.openafs.org/1068
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoDAFS: avoid saving fileserver state when panicing
Andrew Deason [Thu, 17 Dec 2009 19:59:57 +0000]
DAFS: avoid saving fileserver state when panicing

When ShutDownAndCore is called with dopanic=1, avoid trying to save
fileserver state. When we are panic'ing it is very possible that the
thread that called ShutDownAndCore is one of the background threads we
are waiting for, or that the calling thread is holding H_LOCK. Since we
are panicing, the fileserver state is probably not consistent anyway, so
just avoid trying to save state altogether, and avoid a possible
deadlock.

Change-Id: If727808bfdfda74a4ec0e65b27ef9c77fbee4aae
Reviewed-on: http://gerrit.openafs.org/997
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoDAFS: Allow non-fileserver to schedule salvages
Andrew Deason [Fri, 6 Nov 2009 20:05:16 +0000]
DAFS: Allow non-fileserver to schedule salvages

Allow non-fileserver programs to schedule salvages through the fileserver
via FSSYNC (VOL_FORCE_ERROR with the FSYNC_SALVAGE reason code). Also
make the volserver schedule salvages this way when it encounters the
appropriate errors.

FIXES 124484

Change-Id: I03ecf6302436c35fec705cd6c84a40b7cdbf6f97
Reviewed-on: http://gerrit.openafs.org/787
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoRevert "Linux: Replace invalidate_inode_pages"
Simon Wilkinson [Fri, 8 Jan 2010 17:29:16 +0000]
Revert "Linux: Replace invalidate_inode_pages"

This reverts commit 40719534d5d906f97a6ae7d28a00289270daf2a6.

RHEL5 kernels publish the invalidate_mapping_pages symbol as
EXPORT_SYMBOL_GPL, so we can't use it with them.

For now, revert this change - we can live with the deprecated
warning at present...

Change-Id: I22875683e643c4184bb2c1e06f37d4c6eb9ea0be
Reviewed-on: http://gerrit.openafs.org/1082
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agolinux kernel lacks uintptr
Derrick Brashear [Mon, 4 Jan 2010 22:57:30 +0000]
linux kernel lacks uintptr

sadly only recent kernels include uintptr_t. change the cast to use unsigned
long

Change-Id: Ib69f7aa0f2d316c5ea8bdc960b15ee0f89f96250
Reviewed-on: http://gerrit.openafs.org/1059
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoUNIX CM: remove the last remaining reference to afs_rxglobal_lock
Chas Williams - CONTRACTOR [Thu, 7 Jan 2010 22:48:15 +0000]
UNIX CM: remove the last remaining reference to afs_rxglobal_lock

Somehow this was missed earlier.

Change-Id: Iebebdb90cb42cc9ef716409af435509916e4ab58
Reviewed-on: http://gerrit.openafs.org/1078
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLINUX/: ifdef for GFP_NOFS conditionals
Chas Williams - CONTRACTOR [Thu, 7 Jan 2010 22:35:20 +0000]
LINUX/: ifdef for GFP_NOFS conditionals

Remove the conditionals in the linux 2.6 tree for GFP_NOFS.  This feature
has existed since the beginning of the 2.6 tree.  This makes the code
easier to read.

Change-Id: Ib42af012c57912f72effb836743ab8f1ddc30486
Reviewed-on: http://gerrit.openafs.org/1077
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agodarwin module prototype and cleanup
Derrick Brashear [Mon, 4 Jan 2010 00:01:47 +0000]
darwin module prototype and cleanup

tidy up the osi_module and osi_misc prototyping, and the error checking
and returns at module load time. side effect: also use the afs3_syscall
prototype for the BSDs.

Change-Id: I373f44f3b5999dc05ba23f09c74149aaf237edcc
Reviewed-on: http://gerrit.openafs.org/1056
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agovolmonitor copy link before calling free
Dan Hyde [Wed, 6 Jan 2010 15:49:39 +0000]
volmonitor copy link before calling free

Copy tt->next before TRELE(tt) calls free(tt).

We have a core file from a VTRANS_OBJ_LOCK(tt) assert failure, with tt
pointing into glibc's malloc data structures.

Change-Id: Id52b774520c59c224b58f5d507cc490dafea5ca1
Reviewed-on: http://gerrit.openafs.org/1069
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Thomas L. Kula <tkula@umich.edu>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Dan Hyde <drh@umich.edu>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Mark our super block as not updating access time
Simon Wilkinson [Wed, 6 Jan 2010 15:43:18 +0000]
Linux: Mark our super block as not updating access time

We don't update access times, so mark our mount options as such.

FIXES 126084

Change-Id: Ic1591b943288dae45267676046f78e02c306ec08
Reviewed-on: http://gerrit.openafs.org/1071
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Use the correct ATIME flag
Simon Wilkinson [Wed, 6 Jan 2010 15:12:32 +0000]
Linux: Use the correct ATIME flag

Since Linux kernel 2.4.0, MS_NOATIME has been a super block flag,
and S_NOATIME has been the corresponding inode flag. Use the
correct flag when we're opening cache files.

FIXES 126084

Change-Id: I69cbcd651584cc86568c51793dd3f6d3e8644709
Reviewed-on: http://gerrit.openafs.org/1070
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agocachemanager additional cleanup
Derrick Brashear [Tue, 5 Jan 2010 05:14:26 +0000]
cachemanager additional cleanup

integrate support for additional cachemanager cleanup. free
volume, server, unixuser objects. cleanup messages printed at shutdown.
zero/null/init additional locks and structures.

FIXES 126069

Change-Id: Id00219b679a0b4ad15ca06fc16335b0e4282e0eb
Reviewed-on: http://gerrit.openafs.org/1065
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Marc Dionne <marc.c.dionne@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agomacos code signature for afsd
Derrick Brashear [Mon, 4 Jan 2010 18:44:48 +0000]
macos code signature for afsd

this is a dodge; we should sign with a real certificate and distribute
signed binaries. until we more formally exist, this allows application
firewall to at least cope better with us.

Change-Id: I84d18f72f7b9a42f4eb41f86e2e6dc1ae54f662a
Reviewed-on: http://gerrit.openafs.org/1057
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoRefuse to attach inode partitions with UFS logging
Andrew Deason [Fri, 18 Dec 2009 18:17:41 +0000]
Refuse to attach inode partitions with UFS logging

Partitions with the 'logging' UFS mount option are known to cause
corruption when using the inode fileserver backend. So, if we detect
that we are attempting to attach a 'logging' partition, refuse to attach
it.

Change-Id: I0ee23b9935ac6ffc66e6228fe03d6bbfb5d64574
Reviewed-on: http://gerrit.openafs.org/999
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: don't oops on writeback if vcache has no stashed credentials
Marc Dionne [Tue, 5 Jan 2010 00:37:40 +0000]
Linux: don't oops on writeback if vcache has no stashed credentials

When doing writeback, use current credentials if we can't find
stashed credentials in the vcache entry.
This is normally stashed in afs_open, so it's not clear how we
get to that point, but that's not a reason to trigger an oops.

Change-Id: I60937e3f9e360e68ea84205570c05cabb47ef08e
Reviewed-on: http://gerrit.openafs.org/1063
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoviced: allow alternate addresses on the same hash chain again
Michael Meffie [Tue, 29 Dec 2009 18:40:15 +0000]
viced: allow alternate addresses on the same hash chain again

Revert the logic meant to prevent duplicate entries in the
host address hash table so hosts with multiple addresses can
be stored in the same hash bucket again.

Add a new log message to show when the host cannot be stored
in the hash table because of an address-port pair collision
with a host already in the hash table.

Change-Id: I24bf48b912dcf3062201691b446372d78de9bc14
Reviewed-on: http://gerrit.openafs.org/1046
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoFixes for supergroups on Darwin 64bit
Simon Wilkinson [Tue, 5 Jan 2010 00:14:06 +0000]
Fixes for supergroups on Darwin 64bit

Some fixes needed to make supergroups build on 64bit Mac OS 10.5

Change-Id: I1aea100b138a8212010dd9f511377c993589977f
Reviewed-on: http://gerrit.openafs.org/1060
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoUnix CM: Use xdr_free to free memory allocated by XDR
Jeffrey Altman [Wed, 30 Dec 2009 11:32:58 +0000]
Unix CM: Use xdr_free to free memory allocated by XDR

In src/afs/afs_volume.c VL_GetAddrsU() returns memory allocated by the
xdr package. In 1.4 there was no method of freeing this safely. On the
head there is using the xdr_free() function.

FIXES 124937

Change-Id: I72b8333e02e0e58dd16d25e9ed258a2551cdef9d
Reviewed-on: http://gerrit.openafs.org/1047
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agofreebsd: fix afs_root signature (RELENG_8)
Matt Benjamin [Thu, 31 Dec 2009 00:19:09 +0000]
freebsd: fix afs_root signature (RELENG_8)

Track removal of thread id param, which had become redundant.

Change-Id: Ifa8ddce411373a1aee01b577bf7e5a268e644dd7
Reviewed-on: http://gerrit.openafs.org/1055
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoBuild fix: restore centry in ptutils.c
Marc Dionne [Thu, 31 Dec 2009 16:08:32 +0000]
Build fix: restore centry in ptutils.c

The declaration of centry is needed when supergroups are not
enabled.  Restore it within the ifdef so that we don't get
an unused variable warning if supergroups are enabled.

Change-Id: I0e7916b7898cdd5c47accb7f185764bbb1f2e94d
Reviewed-on: http://gerrit.openafs.org/1054
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove warnings from supergroups code on x86
Russ Allbery [Thu, 31 Dec 2009 00:22:37 +0000]
Remove warnings from supergroups code on x86

Prototype functions where necessary, remove unused variables, fix
mismatches of data types (char * != void *), initialize variables the
compiler can't tell are initialized, compare integers against 0 and
not NULL, and wrap assignments used as conditions in an explicit
comparison.

This removes all warnings that fail --enable-checking on x86 with
--enable-supergroups.  64-bit will still have warnings.

Change-Id: I5ab6e317a280bea29cb44b97f2805a00e387577f
Reviewed-on: http://gerrit.openafs.org/1050
Tested-by: Russ Allbery <rra@stanford.edu>
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoCorrect spelling errors in man pages
Russ Allbery [Thu, 31 Dec 2009 04:30:27 +0000]
Correct spelling errors in man pages

Correct several spelling errors in the man pages detected by Debian's
Lintian package checking tool.

Change-Id: I3834a12776ec5e27b04d80fbfcc43ffe71cf1790
Reviewed-on: http://gerrit.openafs.org/1051
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agofreebsd: track RELENG_8
Matt Benjamin [Wed, 23 Dec 2009 00:48:02 +0000]
freebsd: track RELENG_8

Fixes to re-enable build and load of kernel module on FreeBSD's
RELENG_8.  Includes the following changes:

Remove cred pointer argument to vop_getattr and vop_setattr.
Restore VSUID and VGUID defines, but move to osi_machdep.h.
Fixup calls to vinvalbuf, create a macro to avoid too many ifdefs.
Revert incorrect definition of gop_lookupname_user.  Avoid
uninitialized use of 'code' variable in vop_access, and track
new a_accmode member name.  Remove cred pointer in VFS_STATFS.
Avoid including sys/ioctl.h in kernel mode (formerly had no
effect, it's now an error).  Avoid using an empty file as an
object file in shlibafsrpc link.  Replace suser() calls with calls
to priv_check, using defined OpenAFS privileges (thanks due to
Ben Kaduk).

Change-Id: I96fc29599a797b975630c8fa5b8b84798b3740a6
Reviewed-on: http://gerrit.openafs.org/1023
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Check for multiple silly renames
Simon Wilkinson [Sun, 27 Dec 2009 11:28:43 +0000]
Linux: Check for multiple silly renames

We don't want to do multiple silly renames of the same dcache entry,
so add a check for that occuring, and just return EBUSY if we're
trying to do so.

Change-Id: Ic1cb4061d89bf87926995162f1ac410375bddcb5
Reviewed-on: http://gerrit.openafs.org/1035
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoAbstract out Linux sillyrename function
Simon Wilkinson [Sun, 27 Dec 2009 11:06:01 +0000]
Abstract out Linux sillyrename function

In order to keep the dcache happy, the Linux client has its own
sillyrename function. Abstract this out from afs_linux_unlink into a
function of its own (afs_linux_sillyrename) so we can make use of it
from other vnodeops.

Change-Id: I298251c400dfc22efb3bacaa72612b28a5409112
Reviewed-on: http://gerrit.openafs.org/1034
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove unused configuration tests
Simon Wilkinson [Sat, 26 Dec 2009 18:22:32 +0000]
Remove unused configuration tests

Remove configure tests that are no longer used by the build. Testing
for unused features just slows down the configuration process.

Change-Id: I18e161e8e608a69801c53980e545c3077a7578d7
Reviewed-on: http://gerrit.openafs.org/1033
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove HAVE_STRUCT_BUF test
Simon Wilkinson [Thu, 24 Dec 2009 18:19:01 +0000]
Remove HAVE_STRUCT_BUF test

Ever since b1edf891d717ee8e08c0d93738a204cff09cf68f was committed,
we've not needed to test to see if the OS provides struct buf - we
can just declare it as a structure without providing a definition
in all circumstances.

Change-Id: I81ebad0dcda91229883f44b657e8113ec832668d
Reviewed-on: http://gerrit.openafs.org/1032
Tested-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove --disable-full-vos-listvol-switch option
Simon Wilkinson [Thu, 24 Dec 2009 15:35:47 +0000]
Remove --disable-full-vos-listvol-switch option

Remove the --disable-full-vos-listvol-switch configuration option,
and the corresponding #ifdefs. Full output from vos listvol is now
always enabled.

Change-Id: Ib4900cc0a8532b3db31d13d34fe59946b34e86ed
Reviewed-on: http://gerrit.openafs.org/1031
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove --disable-afsdb
Simon Wilkinson [Thu, 24 Dec 2009 15:14:46 +0000]
Remove --disable-afsdb

Remove the --disable-afsdb option from configure, and the
corresponding AFS_AFSDB_ENV #ifdefs from the code. This means that
the AFSDB code will always be built, but whether it used or not
can still be controlled by runtime options in the cache manager.

Change-Id: I1378c4626568e02345c7c400804747c446078c7e
Reviewed-on: http://gerrit.openafs.org/1030
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoclean up axscache at shutdown
Derrick Brashear [Tue, 29 Dec 2009 16:29:09 +0000]
clean up axscache at shutdown

modify allocations in axscache code to be able to be tracked. at shutdown,
clean up allocations.

Change-Id: I1fbde0fe49cf0cdeea451dac58718a90f1c87e66
Reviewed-on: http://gerrit.openafs.org/1037
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoMake DAFS 'bos salvage' work in restricted mode
Andrew Deason [Wed, 30 Dec 2009 16:44:39 +0000]
Make DAFS 'bos salvage' work in restricted mode

bosserver currently has code to specifically allow 'bos
salvage'-initiated salvages to run when in restricted mode. This only
specifically tests for running salvager, though, and so fails when 'bos
salvage' is run against a DAFS fileserver and tries to run
'salvageserver -client' instead.

Test for 'salvageserver -client' as well, so we can run 'bos salvage'
for DAFS fileservers in restricted mode.

Change-Id: I27052a440ea8d6347a9d86e4a1f1a7131a277b0b
Reviewed-on: http://gerrit.openafs.org/1048
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove --disable-largefile-fileserver
Simon Wilkinson [Thu, 24 Dec 2009 14:34:42 +0000]
Remove --disable-largefile-fileserver

Make largefile fileservers the only option. Remove all of the
AFS_LARGEFILE_ENV ifdefs, and tidy up some code as a result of
this change.

Change-Id: I126f7dc5505bbdb28c9337dcd2e81403045707f4
Reviewed-on: http://gerrit.openafs.org/1029
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Andrew Deason <adeason@sinenomine.net>

14 years agoTurn on bos restricted code
Simon Wilkinson [Thu, 24 Dec 2009 13:00:53 +0000]
Turn on bos restricted code

Remove the #ifdef's around the bos restricted mode code. This makes
restricted mode available as part of the standard build, but a server
will not go into restricted mode unless the relevant command line
options are specified, or bos setrestricted is run.

Document bos_setrestricted and bos_getrestricted, and the new
'-restricted' command line option. Add a note to the man pages of
all of the commands whose behaviour is affected by restricted mode.

Add 'setr' and 'getr' aliases for setrestart and getrestart so that
these documented shortcuts continue to work (otherwise they'd be
ambiguous against setrestricted and getrestricted). Note that
setre, setres, and setrest will not work once this patch is applied.

Change-Id: Ie69d21493ea5f78757f0a3d478de43fdaabd3c31
Reviewed-on: http://gerrit.openafs.org/1028
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoUpdate .gitignore for copied rxkad files
Russ Allbery [Thu, 31 Dec 2009 04:33:14 +0000]
Update .gitignore for copied rxkad files

src/rxkad/Makefile copies fcrypt.h and sboxes.h from the domestic
subdirectory, so ignore the copies in the src/rxkad directory.

Change-Id: I263c8f5ea45c3c6b32bc377f6451424477d255a2
Reviewed-on: http://gerrit.openafs.org/1052
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: #if BLAH should be #if defined(BLAH)
Simon Wilkinson [Tue, 29 Dec 2009 22:48:39 +0000]
Linux: #if BLAH should be #if defined(BLAH)

As the subject says, using #if BLAH gives warnings. Fix them
in osi_ioctl.c

Change-Id: Iee3da68cd29e84bd201a1ee657d3f80ca9b2f5d5
Reviewed-on: http://gerrit.openafs.org/1041
Reviewed-by: Russ Allbery <rra@stanford.edu>
Tested-by: Russ Allbery <rra@stanford.edu>

14 years agomissing brace in afs_get_sb
Michael Meffie [Wed, 30 Dec 2009 18:02:31 +0000]
missing brace in afs_get_sb

Fix a syntax error when GET_SB_HAS_STRUCT_VFSMOUNT is
not true so we can build on older versions of linux.

Change-Id: I85aac803f68c3f3729abd32dfcecbc20ccac8ab9
Reviewed-on: http://gerrit.openafs.org/1049
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Russ Allbery <rra@stanford.edu>
Tested-by: Russ Allbery <rra@stanford.edu>

14 years agoUnix: Rename aklog_main.c as aklog.c
Simon Wilkinson [Tue, 29 Dec 2009 17:06:42 +0000]
Unix: Rename aklog_main.c as aklog.c

Now the reasons for having aklog_main.c separate from aklog.c have gone,
rename the file so we've just got aklog.c

Change-Id: I32e34be246a5357be25ab93328d434c396483042
Reviewed-on: http://gerrit.openafs.org/1038
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoUnix: Merge aklog with aklog_main
Simon Wilkinson [Tue, 29 Dec 2009 17:03:50 +0000]
Unix: Merge aklog with aklog_main

Historically, aklog was split into aklog.c (which just contained a
wrapper) and aklog_main.c, which did all of the hard work. Now
that we've purged all of the Windows code from aklog_main.c, we
can merge them both into the same file, so do so.

FIXES 20884

Change-Id: Ifdcb33f4eb2b26703c833056882d95f7f0f654b1
Reviewed-on: http://gerrit.openafs.org/1036
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoavoid leaking stat info
Derrick Brashear [Thu, 17 Dec 2009 20:33:42 +0000]
avoid leaking stat info

if a file is already CStatd we just copy out the information without doing
access checks. add an access check.

Change-Id: I94d1dc118189b19759d8e9635988c40f38026cbb
Reviewed-on: http://gerrit.openafs.org/995
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoReturn a cell name from afsconf_LookupServer
Andrew Deason [Tue, 29 Dec 2009 20:45:51 +0000]
Return a cell name from afsconf_LookupServer

Previously, afsconf_LookupServer could not meaningfully return a cell
name, since a char* was passed in for realCellName, and
afsconf_GetAfsdbInfo was expecting the value of the pointer to change.
Change afsconf_LookupServer to take a char**, and pass realCellName by
reference, so we can get a cell name back.

Change-Id: Ib945fcb4be482ee155cdd1dcd62fe577c002082d
Reviewed-on: http://gerrit.openafs.org/1039
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoFix ioctl32 inclusion on Sparc 64
Simon Wilkinson [Tue, 29 Dec 2009 21:56:36 +0000]
Fix ioctl32 inclusion on Sparc 64

ioctl32.h only became available in Linux 2.5.44, so reduce the scope
of the include. It went away again, in Linux 2.6.22. In its short
life, it only ever provided prototypes for registering translation
handlers, and for sys_ioctl.

Only include it where we call those handlers, and when we call them,
to remove compilation errors for Sparc64.

Change-Id: I7816bd8b1e18ead68f7e5078e684d60bf59ac470
Reviewed-on: http://gerrit.openafs.org/1040
Reviewed-by: Russ Allbery <rra@stanford.edu>
Tested-by: Russ Allbery <rra@stanford.edu>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoDocumentation: Fix asetkey syntax
Simon Wilkinson [Wed, 30 Dec 2009 01:39:26 +0000]
Documentation: Fix asetkey syntax

To set a key, use "asetkey add", not just "asetkey"

FIXES 125430

Change-Id: Ifa381ec95f9253bcc5c7a1d374fbf88408f82f67
Reviewed-on: http://gerrit.openafs.org/1045
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Replace invalidate_inode_pages
Simon Wilkinson [Tue, 29 Dec 2009 22:58:56 +0000]
Linux: Replace invalidate_inode_pages

invalidate_inode_pages is now marked as deprecated in the kernel
headers. Use the equivalent call to invalidate_mapping_pages() which
has been available since Linux 2.5

Change-Id: I58de20e16874bb06991320b26bc6ce63685b8bb4
Reviewed-on: http://gerrit.openafs.org/1042
Reviewed-by: Russ Allbery <rra@stanford.edu>
Tested-by: Russ Allbery <rra@stanford.edu>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoDocumentation: vos dump -verbose outputs to stderr
Simon Wilkinson [Wed, 30 Dec 2009 01:04:23 +0000]
Documentation: vos dump -verbose outputs to stderr

vos dump -verbose sends its logging to stderr, not to stdout (where
it would potentially collide with the dump data itself).

FIXES 124911

Change-Id: I515c50df59d2f376787969df59b6e01e244ecbc7
Reviewed-on: http://gerrit.openafs.org/1044
Reviewed-by: Russ Allbery <rra@stanford.edu>
Tested-by: Russ Allbery <rra@stanford.edu>

14 years agorxgen: Remove stray debugging statement
Simon Wilkinson [Tue, 29 Dec 2009 23:20:36 +0000]
rxgen: Remove stray debugging statement

A stray "Printing ex_req" slipped into the last set of rxgen changes.
Remove the unnecessary debugging statement.

Change-Id: I2dda4138d70bf3429122f58bbbae996b84baefac
Reviewed-on: http://gerrit.openafs.org/1043
Reviewed-by: Russ Allbery <rra@stanford.edu>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoMove TOP_INCDIR/des dependency to prelude
Andrew Deason [Thu, 24 Dec 2009 18:32:14 +0000]
Move TOP_INCDIR/des dependency to prelude

Make TOP_INCDIR/des a dependency of the 'prelude' target, not all of the
'dest'-ish targets. Without this, things like 'make tviced' fail since
we never create the TOP_INCDIR/des directory, but 'des' needs to copy
headers into there.

Also, remove the other redundant INCDIR/LIBDIR deps in the 'dest'-y
targets, and put them in 'prelude'.

Change-Id: I1beb1a0c8b93a7bf17af71792ab38590d03c9976
Reviewed-on: http://gerrit.openafs.org/1027
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoCheck viced FetchData length for cache bypass
Andrew Deason [Thu, 17 Dec 2009 21:16:50 +0000]
Check viced FetchData length for cache bypass

Same fix as change I413393a7bacbf207332d7f904cf396c79b77b6b5, but for
the cache bypass code.

Change-Id: Ic181e257f7d0e1892bd10bf14d8d5571b4804d63
Reviewed-on: http://gerrit.openafs.org/1000
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoFix warnings in fs.c with --enable-cache-bypass
Andrew Deason [Wed, 23 Dec 2009 20:23:30 +0000]
Fix warnings in fs.c with --enable-cache-bypass

Fix some warnings in src/venus/fs.c so we can compile with
--enable-cache-bypass and --enable-checking:

 -- Include ctype.h so isdigit gets a prototype

 -- Make BypassThresholdCmd have the proper signature

 -- Remove the unused 'size' variable

Change-Id: I09afc6c04c990476226d1c2a751e23d3d39085b6
Reviewed-on: http://gerrit.openafs.org/1025
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoSolaris: Don't access lbolt directly
Simon Wilkinson [Tue, 22 Dec 2009 21:46:43 +0000]
Solaris: Don't access lbolt directly

Recent changes in OpenSolaris have removed the 'lbolt' variable.
However, the ddi_get_lbolt accessor (which has been present since
Solaris 10) can be used to get access to the same value. So, use it.

Change-Id: I7a72bab14b0226fb25adb642f3621512e02ca216
Reviewed-on: http://gerrit.openafs.org/1022
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoSolaris: Don't directly fiddle with the groups structure
Simon Wilkinson [Tue, 22 Dec 2009 21:42:19 +0000]
Solaris: Don't directly fiddle with the groups structure

Solaris doesn't let us directly play with the groups structure
anymore. Instead, there is a crsetgroups accessor which can be
used instead. So, use it.

Change-Id: I4d5b99110318630c8a7744812476e1941c117381
Reviewed-on: http://gerrit.openafs.org/1021
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoSolaris: Be more flexible about compilers
Simon Wilkinson [Tue, 22 Dec 2009 21:38:00 +0000]
Solaris: Be more flexible about compilers

OpenSolaris doesn't have its C compiler in /opt/SUNWspro/bin/cc,
but we still have to be careful to avoid gcc, as that can't
build the kernel module.

As a short term stop gap, add a PATH_PROG test which looks in
both /opt/SUNWspro/bin and /opt/SunStudioExpress/bin for cc.
We probably should look in more places, and in the long run,
we should probably be taking whatever autoconf gives us
(for userspace, at least)

Change-Id: Ie31cf1bf474650a081fc101a2aa40cfa6bd39423
Reviewed-on: http://gerrit.openafs.org/1020
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLook for aclocal in more places
Simon Wilkinson [Tue, 22 Dec 2009 21:34:23 +0000]
Look for aclocal in more places

Some operating systems (cough, OpenSolaris, cough) have multiple
versions of aclocal installed, and don't populate the 'aclocal' name.
If 'aclocal' isn't present, then try using 'aclocal-1.10' before we give
up in disgust.

Change-Id: Iad6daf1038942aeee13f38cb0c00c58da621cfd1
Reviewed-on: http://gerrit.openafs.org/1019
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoFix typo in afs_linux_cred_is_current
Andrew Deason [Tue, 22 Dec 2009 14:49:21 +0000]
Fix typo in afs_linux_cred_is_current

987816dfa852750caca8c64c44174a5333ae0e3b introduced a stray 'cred' in
one of the afs_linux_cred_is_current definitions. Remove it so we can
build without STRUCT_TASK_HAS_CRED.

Change-Id: I7ef94c1c3d15a2a744755d8a20ec2b0c92cd5349
Reviewed-on: http://gerrit.openafs.org/1018
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoPrefix global defines
Michael Meffie [Thu, 10 Dec 2009 21:13:45 +0000]
Prefix global defines

Prefix constants relocated to the common afs_consts.h
header to void naming collisions.

Change-Id: Ib657e00385002c957b40b36058f260c6cabd108c
Reviewed-on: http://gerrit.openafs.org/905
Reviewed-by: Mickey Lane <mlane@sinenomine.net>
Tested-by: Mickey Lane <mlane@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoConsolidate duplicate definitions
Michael Meffie [Thu, 10 Dec 2009 16:37:34 +0000]
Consolidate duplicate definitions

Housekeeping change to consolidate some common definitions.
Create a new common header called afs_consts.h.  This allows
us to remove the afscp.c dependency on afs.h (so src/tests
will build again.) Rename the max filename definition in the
update server package to MAXFNSIZE to avoid a name conflict
with MAXSIZE.  The global defines will be prefixed with
AFS_ in a second patch.

Change-Id: I2b8d555a244cc92d889618de4eec4a99550d7c7f
Reviewed-on: http://gerrit.openafs.org/855
Reviewed-by: Mickey Lane <mlane@sinenomine.net>
Tested-by: Mickey Lane <mlane@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: fix sysctl for 2.6.33
Marc Dionne [Sun, 20 Dec 2009 13:30:07 +0000]
Linux: fix sysctl for 2.6.33

The sysctl patch for 2.6.33 was a bit overzealous and ifdef'ed
a few lines that it shouldn't have.

Change-Id: I7033259e567a06ce05b968366e810c5b2688c271
Reviewed-on: http://gerrit.openafs.org/1011
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: utsrelease.h is moving
Marc Dionne [Sun, 20 Dec 2009 03:22:05 +0000]
Linux: utsrelease.h is moving

In kernel 2.6.33, utsrelease.h has moved to include/generated.
Adapt the configure code to consider that location, and clean
up that section's indentation.

Change-Id: I5061043ff7f46875a39953b11c472693650c7485
Reviewed-on: http://gerrit.openafs.org/1009
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoFix PAGs for all platforms without Linux keyrings
Simon Wilkinson [Sat, 19 Dec 2009 21:53:54 +0000]
Fix PAGs for all platforms without Linux keyrings

The changes in 7b272177de4c6f78db7f2315f4e30e85ab7660bb sadly break
all platforms which aren't Linux. This is because the logic in PagInCred
doesn't handle the non-Linux case at all. Fix this so that we call
afs_get_group_pag whenever we're not running on Linux.

The code also doesn't match the comment - Linux kernels which don't
have keyrings should use afs_get_group_pag(), regardless of whether
STRUCT_TASK_HAS_CRED. It's unlikely this bit would ever bite anyone
though, kernels with cred support almost certainly require keyrings.

Change-Id: I5373da51151229dc13b6dc60c3cf10700f0f881a
Reviewed-on: http://gerrit.openafs.org/1006
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Simplify keyring compatibility code
Simon Wilkinson [Sat, 19 Dec 2009 20:21:36 +0000]
Linux: Simplify keyring compatibility code

This introduces a few inline functions in osi_compat.h, to reduce the
number of #ifdefs in the main chunk of code. In particular, we gain
  * afs_linux_key_alloc : to handle all the different key_alloc
                          signatures
  * afs_linux_search_keyring : to handle our two different mechanisms
                               for searching a keyring
  * afs_linux_cred_is_current : will return true if we're using native
credentials, and the passed creds are also
those of the current task

Change-Id: I138f3533a7e8e88e04e4b5508158e003882d63ee
Reviewed-on: http://gerrit.openafs.org/1005
Reviewed-by: Marc Dionne <marc.c.dionne@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Don't panic when keys aren't found
Simon Wilkinson [Sat, 19 Dec 2009 15:40:49 +0000]
Linux: Don't panic when keys aren't found

This fixes two potential problems in our session keyring lookup code,
which can lead to panics in situations where we're using the new struct
cred based code.

The first is that if there is no session kerying installed for the
current task, we'll attempt to do a lookup on a NULL kerying and oops.

The second is that if the keyring_search returns EPERM, then we can end
up unmasking that error code, and return NULL, rather than an error.

Change-Id: If0e2804408ec17b00f352980cee6a8e56704d93f
Reviewed-on: http://gerrit.openafs.org/1004
Reviewed-by: Marc Dionne <marc.c.dionne@gmail.com>
Tested-by: Marc Dionne <marc.c.dionne@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Flush vcaches when a mount fails
Simon Wilkinson [Sat, 19 Dec 2009 14:48:32 +0000]
Linux: Flush vcaches when a mount fails

GetVCache (well, really NewVCache) creates a new vcache, with an attached
inode, and links it into the VLRU queue, regardless of whether it is
successful in populating that vcache or not. The attached inode, on Linux,
contains a reference to the super block of the filesystem. If the created
vcache is for the root, however, and populating that vcache failes, then mount
fails, and the super block is disposed of. This leaves us with a vcache in
the VLRU queue which contains a reference to a non-existent inode.

When ShakeLooseVCaches comes along a few minutes later, it attempts to discard
this inode. However, doing so fails because the inode no longer has a valid
super block.

Avoid this trauma by ensuring all vcaches are disposed of before the super
block goes away, in the event of a mount failure.

Change-Id: I68864f1ea401d24adba76164905a17de6ab3e6ce
Reviewed-on: http://gerrit.openafs.org/1003
Reviewed-by: Marc Dionne <marc.c.dionne@gmail.com>
Tested-by: Marc Dionne <marc.c.dionne@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoDo not trust FetchData length from fileservers
Andrew Deason [Thu, 17 Dec 2009 20:59:44 +0000]
Do not trust FetchData length from fileservers

Currently the Unix CM implicitly trusts that the length from a FetchData
request from a fileserver will always be less than the requested length.
If the fileserver sends more data than requested, we can use up more
cache space than we intended, possibly exceeding the cacheinfo cache
limits.

Add a check for this, and return EIO to the caller if the fileserver
responds with too much data.

Change-Id: I413393a7bacbf207332d7f904cf396c79b77b6b5
Reviewed-on: http://gerrit.openafs.org/996
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoplatform target depends on cmd and vol
Andrew Deason [Fri, 18 Dec 2009 18:21:35 +0000]
platform target depends on cmd and vol

The 'platform' target depends on cmd and vol, since on solaris
non-namei, we build fs_conv_sol26.c, which depends on some volume
structures, and uses libcmd. So, have Makefile.in accurately reflect
that dependency so we can build.

Change-Id: Ic7038f252dd069522ebfc8e72b9743c01c97d99c
Reviewed-on: http://gerrit.openafs.org/998
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agokernel ioctl32 conversion typecasting
Derrick Brashear [Thu, 17 Dec 2009 14:48:06 +0000]
kernel ioctl32 conversion typecasting

cast types to avoid a pointer from integer warning when using pointer
types for sizing

Change-Id: Idac76608e5352c65a13d07bfadb66d78f9db0ffd
Reviewed-on: http://gerrit.openafs.org/988
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoAdd the 'vos endtrans' command
Andrew Deason [Fri, 16 Oct 2009 22:12:35 +0000]
Add the 'vos endtrans' command

Add a command to vos to explicitly end volume transactions. These can
happen if (for example) we segfault or the user ctrl-C's in 'vos'
before/after a volume operation, but before we AFSVolEndTrans.

Change-Id: Ie34ee1fdff917b56900f456c7cf8b1329533a7da
Reviewed-on: http://gerrit.openafs.org/870
Reviewed-by: Dan Hyde <drh@umich.edu>
Reviewed-by: Alistair Ferguson <alistair.ferguson@mac.com>
Tested-by: Alistair Ferguson <alistair.ferguson@mac.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoOpenAFS Preference Pane
Claudio Bisegni [Fri, 11 Dec 2009 21:03:39 +0000]
OpenAFS Preference Pane
Now the preference pane is fat compiled.

Change-Id: I5c54555d00b16d85f01286719bbc41ea5f1fb67d
Reviewed-on: http://gerrit.openafs.org/906
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoExpand ProgramType enumeration
Andrew Deason [Fri, 6 Nov 2009 20:03:52 +0000]
Expand ProgramType enumeration

The ProgramType values volumeUtility and salvager are overloaded. Expand
the ProgramType enum to include more specific program types, and adjust
conditionals to match.

Also, instead of determining all behavior by checking programType, add
some flags to be passed in to VInitVolumePackage to determine e.g.
whether or not we can use the FSSYNC channel. This makes it easier to
see the intent of some conditionals, and reduces the number of times a
caller must lie about what program it is.

Change-Id: Ic9852a35bb16a6b1f4b0aa9766de63178ecea56f
Reviewed-on: http://gerrit.openafs.org/786
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: Use splice to speed up cache storeback
Simon Wilkinson [Fri, 20 Nov 2009 15:08:25 +0000]
Linux: Use splice to speed up cache storeback

This patch adds a new cache store function for Linux, which uses
splice() and direct access to the page cache, rather than doing
data copies to a temporary buffer between rx and afs. It removes
one copy, and some context switching, from the write codepath.

One side-effect here is that it will delay storebehinds from
returning control to the user. Instead of returning once the first
4k has been successfully transfered, we will wait until a cache
chunk has been transmitted. This is currently unavoidable, as we
can't take the GLOCK within a splice actor.

Change-Id: I5b0284d67febccf099710589908fad18b808332c
Reviewed-on: http://gerrit.openafs.org/903
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoRework cache store back
Simon Wilkinson [Fri, 20 Nov 2009 15:03:37 +0000]
Rework cache store back

This patch reworks the code to store data from the cache to the server,
such that the entire store loop can be replaced. The idea here is that a
platform which wishes to provide a different store loop, in particular
one which doesn't rely upon multiple data copies, may do so simply by
plugging in a new function.

Change-Id: I6573c6a5c3e93a1259266e7e1a2b7bd17357203f
Reviewed-on: http://gerrit.openafs.org/902
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>

14 years agoLinux: deal with ctl_name removal
Marc Dionne [Thu, 10 Dec 2009 00:06:18 +0000]
Linux: deal with ctl_name removal

The binary sysctl interface will be removed in kernel 2.6.33 and
ctl_name will be dropped from the ctl_table structure.
Make the code that uses ctl_name conditional on a configure test.

Change-Id: Iba0f107f299c6515e4e560d7596e6187bd68e399
Reviewed-on: http://gerrit.openafs.org/904
Tested-by: Marc Dionne <marc.c.dionne@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agopt_util runs on database servers
Simon Wilkinson [Tue, 8 Dec 2009 15:21:54 +0000]
pt_util runs on database servers

Fix the pt_util manpage to reflect the fact that it is run on database
servers, not fileservers.

Change-Id: If2a8e5b65ef925c50eb9bfebea4e0d30c20f0970
Reviewed-on: http://gerrit.openafs.org/901
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: cm_BPlusEnumAlloc should not fail for zero entries
Jeffrey Altman [Sat, 5 Dec 2009 15:51:27 +0000]
Windows: cm_BPlusEnumAlloc should not fail for zero entries

If cm_BPlusEnumAlloc returns NULL, the caller assumes a
memory allocation error.  If the enumeration consists of
zero entries, allocate a structure that stores zero entries.

LICENSE MIT

Change-Id: I8ed3811a1b3a0e4262749e110027c5d5812338b6
Reviewed-on: http://gerrit.openafs.org/892
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agoWindows: cm_BPlusDirBuildTree can fail
Jeffrey Altman [Sat, 5 Dec 2009 15:53:03 +0000]
Windows: cm_BPlusDirBuildTree can fail

It is possible that cm_BPlusDirBuildTree can fail.  For example,
the server could be marked down after a callback is obtained
but before all of the directory data buffers have been fetched.
cm_BeginDirOp must check for the failure, destroy the tree,
and return the failure code to the caller.  Otherwise, a tree
with no entries may be created and marked with the current
data version.

LICENSE MIT

Change-Id: I26fbfceaf68389a1906797b12721c49172b027ec
Reviewed-on: http://gerrit.openafs.org/893
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>

14 years agomacos avoid dotunderbar lookups in fakestat mode
Derrick Brashear [Mon, 7 Dec 2009 04:44:53 +0000]
macos avoid dotunderbar lookups in fakestat mode

cocoa tries to lookup ._ files for every object, possibly generating
AFSDB queries and other fun. avoid them

Change-Id: Id70b6335d4dddcb6d5628a9883b2c81193bb39d7
Reviewed-on: http://gerrit.openafs.org/896
Reviewed-by: Rod Widdowson <rdw@steadingsoftware.com>
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoadd newline to lockprocs_prototypes.h
Derrick Brashear [Mon, 7 Dec 2009 20:43:33 +0000]
add newline to lockprocs_prototypes.h

gcc was unhappy about the lack of trailing newline. add one.

Change-Id: Ie9c93bd9ed8c923ea8069fe49e7f5ddb20c28f80
Reviewed-on: http://gerrit.openafs.org/898
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoDump all hostFlags in hosts.dump
Andrew Deason [Tue, 24 Nov 2009 21:09:42 +0000]
Dump all hostFlags in hosts.dump

Currently only certain flags are dumped in hosts.dump. Also dump the
value of hostFlags itself, so we get more information from a host dump.

Change-Id: I2d265b85e057c2d44938e1d92f4c102048c4bf1f
Reviewed-on: http://gerrit.openafs.org/869
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Dan Hyde <drh@umich.edu>
Reviewed-by: Alistair Ferguson <alistair.ferguson@mac.com>
Tested-by: Alistair Ferguson <alistair.ferguson@mac.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoMake HandleClientContext take an afs_ucred_t
Andrew Deason [Sun, 6 Dec 2009 20:56:20 +0000]
Make HandleClientContext take an afs_ucred_t

5d5d18071608d9782a5d936f79215048af641410 missed one occurrence of
AFS_UCRED. Make HandleClientContext take an afs_ucred_t instead of
AFS_UCRED, so afs_pioctl.c builds again.

Change-Id: I04c831ad262cade5da0f90fb8860174657dae4bf
Reviewed-on: http://gerrit.openafs.org/895
Tested-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoRemove AFS_USEBUFFERS
Simon Wilkinson [Sun, 6 Dec 2009 14:25:08 +0000]
Remove AFS_USEBUFFERS

The AFS_USEBUFFERS code has been disabled for all supported platforms
since the initial OpenAFS code drop. Simplify the buffers code by
removing it entirely.

Change-Id: Iab6b2d54361eca1a1446b4804b3d6533d75f2831
Reviewed-on: http://gerrit.openafs.org/894
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agoWindows: Version update for 1.5.68 openafs-devel-1_5_68
Derrick Brashear [Sat, 5 Dec 2009 00:53:21 +0000]
Windows: Version update for 1.5.68

1.5.6800 = 1.5.68

Change-Id: Id3bdc21e5f72ebe2047cb70c2496df2ba7300b4b
Reviewed-on: http://gerrit.openafs.org/891
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agounix 1.5.68
Derrick Brashear [Fri, 4 Dec 2009 16:52:29 +0000]
unix 1.5.68

update version numbers for unix

Change-Id: Ic8a42a9ca1cfc74ab2d8f8d2ac0442efa716a84b
Reviewed-on: http://gerrit.openafs.org/888
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>

14 years agorx packet dumper not in kernel
Derrick Brashear [Fri, 4 Dec 2009 20:31:10 +0000]
rx packet dumper not in kernel

don't compile the rx packet dumper when in kernel

Change-Id: I8484f3ed80d3b97d733091901f6e252a455cefa1
Reviewed-on: http://gerrit.openafs.org/889
Tested-by: Derrick Brashear <shadow@dementia.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>