ubik: don't rely on timeout value after select()

The value of timeout after a select() call should be considered
undefined; relying on its value is not portable.
Since IOMGR_Select doesn't modify the timeout it is given, the
intention of the code seems to be to wait for gradually increasing
timeout values, starting at 50ms.  At least under Linux, the
timeout gets set to 0 by select() if it waited for the full specified
time, resulting in a much shorter maximum possible wait period.

Initialize the timeout value for each loop according to the existing
logic, to get consistent behaviour between the lwp and pthreaded code.

Reviewed-on: http://gerrit.openafs.org/4441
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 0b510fe30afb34202342364e96bd9030052e1567)

Change-Id: I24eb4d4b1f758f33e3517671cb576ff23e641fb3
Reviewed-on: http://gerrit.openafs.org/6282
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Irix: Supress some more warnings in kernel builds

Suppress the following warnings in kernel builds:
 *) The function "blah" was declared but never referenced (1174)
 *) The indicated trailing comma is non-standard (1201)
 *) The parameter "blah" was never referenced (3201)
 *) implicit conversion of a 64-bit integral type to a smaller
    integral type (3968)

Reviewed-on: http://gerrit.openafs.org/4419
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 353857e7685c9acfcf7474fabe57c928c70c778f)

Change-Id: I4a5b85ffc56a801e19a181f8d2661eba9c975dcc
Reviewed-on: http://gerrit.openafs.org/6281
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

fsprobe: Fix usage message

Fix the usage message to display the correct command name.

Reviewed-on: http://gerrit.openafs.org/4398
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit a0389f5592c25afb19befb5bf62833a0b0bc59cc)

Change-Id: I6105eff9f4cfae56d85df40589bd4b175aa760d0
Reviewed-on: http://gerrit.openafs.org/6280
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

fsprobe: call RXAFS_GetTime directly

Call RXAFS_GetTime directly instead of using ubik_Call which is
not prototyped and has logic that we don't need/want here since
we're targeting a specific server.

Reviewed-on: http://gerrit.openafs.org/4397
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit ef105dcc4932da804750dea0c922d5814fc5bf15)

Change-Id: If4e430a55d36725f2cf8e5bf04a036062480cfac
Reviewed-on: http://gerrit.openafs.org/6279
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Linux: Fix USE_UCONTEXT detection

On Linux, afs/param.h determines whether to use the ucontext()
function based on the glibc version. However, the glibc version
macros aren't available until a C library header has been included.

There have been a couple of attempts to fix this. The first included
afs_sysnames.h before the C library check, but this wasn't sufficient
to pull in a header for all builds. The second included stdio.h before
including afs/param.h in process.c.

Whilst the second approach works, it has the drawback that it breaks
our convention that afsconfig.h and afs/param.h must always be the first
includes in a file. More seriously, it means that the behaviour of
afs/param.h is different depending on where it appears in the include
order - which is especially dangerous for 3rd party user.

So, this patch tries to resolve all of this by explicitly including
features.h before doing the glibc version checks. The only danger I
can see here is that Linux platforms without glibc may lack a
features.h, but I suspect we have trouble building on such platforms
in any case.

Reviewed-on: http://gerrit.openafs.org/4389
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit f5f0ce525a63264098cd0625769d642fe6f106ba)

Change-Id: Ia94487f35ad0fc3143faafd2fa12a6c71d40f116
Reviewed-on: http://gerrit.openafs.org/6278
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

vol: remove warning on use of DFlushVolume

Include the afs/dir.h header on Windows to avoid
warnings caused by the missing prototype.

Reviewed-on: http://gerrit.openafs.org/4148
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit d69da816c6e961ef5e22df289ad5667cd8568b4c)

Change-Id: I1f128f676a6d099b2d9534e28c6302de78b0ff2d
Reviewed-on: http://gerrit.openafs.org/6273
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

vol: Move VOL_CV_TIMEDWAIT to volume_inline.h

Some platforms will give errors for the vol package glock not being
defined in VOL_CV_TIMEDWAIT, since some utilities include volume.h
(with pthreaded ubik enabled) but do not have the vol glock. This
isn't a problem for the other CV_WAIT (et al) macros, since, being
macros, they are not expanded if never referenced. Since
VOL_CV_TIMEDWAIT is a static inline function, however, and at least
some platforms try to compile/link it even if it's not referenced.

So, move it to volume_inline.h, which is only used by the volume
package itself, where we will have the vol glock.

Reviewed-on: http://gerrit.openafs.org/3363
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit e4059c1b812410bd96481cc60c86c67f8733b66f)

Change-Id: Ic1d6286bb148f202b8da010edf48fff93a5a48fb
Reviewed-on: http://gerrit.openafs.org/6268
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

vol: Add timeouts to SYNC server select() calls

Normally *SYNC server processes wait indefinitely for activity to
occur on one of the SYNC sockets. On some Linux kernels, there exists
a race condition where data can come in on a socket, but the select()
call continues to wait. To ensure that we do not hang forever in such
a scenario, add a timeout to the select() call, which will ensure we
notice the new data within 10 seconds. Raise the timeout on non-Linux
to reduce impact elsewhere.

The Linux kernel bug is tracked in
<https://bugzilla.redhat.com/show_bug.cgi?id=494404>, though that bug
report may not represent all affected kernels.

Reviewed-on: http://gerrit.openafs.org/4377
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 920a00e075b679f59e31b9fcbe7f5db15e345a95)

Change-Id: Iae4035bc952b66f88843d51ff341b0576f15163d
Reviewed-on: http://gerrit.openafs.org/6276
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

namei: Log ListViceInodes write failures

If we fail to write to the inodeFile given by ListViceInodes, we
should log the failure. Otherwise, no indication is given as to what
failed, when a salvage gathers the inode list.

Reviewed-on: http://gerrit.openafs.org/4294
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 22149d82d465f5defbe329b3f9b3c5040a6293e9)

Change-Id: I7fb56f0af99bb040e64cf54bf05e01306a609326
Reviewed-on: http://gerrit.openafs.org/6275
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

macos: dont want bind 8 compat on leopard

we removed it for i386, kill it for ppc also.

Reviewed-on: http://gerrit.openafs.org/4229
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 4a72ebfe324874dd405d18561baf5eaacb007950)

Change-Id: I9e625f79c047b61fc3675e6732eaa7d2dd34214f
Reviewed-on: http://gerrit.openafs.org/6274
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

LINUX: afs_linux_put_link is void

The put_link callback doesn't return a value, so declare it as such.

Reviewed-on: http://gerrit.openafs.org/3636
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 336256c7a0e7d54495a070895c9d8c2057168839)

Change-Id: I5985316ea37f572039267ea117dfeae55ea96bc0
Reviewed-on: http://gerrit.openafs.org/6271
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

DAFS: listvol + unsalvagable volumes = intolerable delay

a non- inService volume (or more) plus vos listvol equals pain.
return Salvaging instead of waiting for it in the listvol case.

Reviewed-on: http://gerrit.openafs.org/3588
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit affc978be7ba3c36c9b7a6a94fb5e907beebed99)

Change-Id: If871ce0cdd3593250f22d74de184f92b046b5f6e
Reviewed-on: http://gerrit.openafs.org/6270
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

LINUX: Reduce stack depth on recursive symlink res

Instead of calling vfs_follow_link inside afs_linux_follow_link
ourselves, we can just resolve the next step of the symlink resolution
and set the result in nd_set_link(), freeing the string in
.put_link().

For kernels without a usable symlink text cache, this reduces call
depth when resolving a path containing many symlinks by two frames per
layer of indirection, allowing for more deeply-nested symlink paths to
be usable.

Reviewed-on: http://gerrit.openafs.org/3433
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 2ce159fbf474b96ecd2a4224751d4655230bba51)

Change-Id: Idb91c2bb1a10b60d125dccbc98731ace63d4add8
Reviewed-on: http://gerrit.openafs.org/6269
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

rxgen: add *TranslateOpcode functions to XDR code

Administrative commands and server binaries can benefit from the
ability to translate RPC OpCode numbers to proper function names.

For this purpose, rxgen now emits the function

char *PKGTranslateOpCode(int)

where PKG is replaced by the respective package prefix.

The function is generated in the exact same way that ExecuteRequest is.
This way, the right optimization (array storage vs. generated case
statement) according to the respective sparseness of OpCodes in the
different RX packages is maintained.

Reviewed-on: http://gerrit.openafs.org/3305
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 5c77a6ca2460c3e79125a614037824db35480905)

Change-Id: I23d7af2dd07e02255783de0523fce58dbe46d31e
Reviewed-on: http://gerrit.openafs.org/6267
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

vol: Add interfaces for registering RX calls

Add VGetVolumeWithCall and VPutVolumeWithCall, to associate RX calls
with volume heavyweight references. Also add the interrupt_rxcall
field to the volume package options structure.

This also adds the VIsGoingOffline function, so a caller can tell when
a volume is going offline.

Reviewed-on: http://gerrit.openafs.org/3215
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 33409b2c849d7fae3b10b3202bb15e338ae5b982)

Change-Id: Ib50a17b3479453bbbaef571711ede7115b17318e
Reviewed-on: http://gerrit.openafs.org/6265
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

vol: Add VGetVolumeTimed

Replace the VGetVolumeNoWait interface with the more general
VGetVolumeTimed interface, which allows for waiting for offlining
volume for arbitrary amounts of time (instead of just "waiting
forever" or "not waiting at all"). Also add VOL_CV_TIMEDWAIT and
VTimedWaitStateChange_r as necessary to implement this.

Reviewed-on: http://gerrit.openafs.org/3214
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit db6ee95864a8fc5f33b7e95c19c8ff5058d37e92)

Change-Id: I6c7b3a0b9fe174ebffeb03153dda1c4705d7dac5
Reviewed-on: http://gerrit.openafs.org/6264
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

rx: Call rxgen_consts.h by its proper name

rxgen_consts.h is actually in the afs/ directory. Include it from
their, rather than requiring that every user have $(TOP_INCDIR)/afs
in their search path.

Reviewed-on: http://gerrit.openafs.org/2750
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 34df5da39871ae5231048a416f191784c5c2d127)

Change-Id: I915cf9d975efbb7c73c1e57750227421362648c8
Reviewed-on: http://gerrit.openafs.org/6263
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

udebug: Always show tidCounter

If the queried site doesn't have an active write transaction,
currently udebug doesn't show the tidCounter. The tidCounter can still
be useful to know (especially since some sites will become buggy when
tidCounter rolls over), so always show it.

This adds the "The last trans I handled was" message to udebug.

Reviewed-on: http://gerrit.openafs.org/2730
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 51806bd21af5bcc70c27cda5787ae853fa3f495e)

Change-Id: I861804c4381a75626c5259a4896642f0e4064b7d
Reviewed-on: http://gerrit.openafs.org/6262
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

doc: limitations of addsite on different partitions

A user on the openafs-info mailing list noted that the Admin Guide is
unclear about creating read-only replicas on different partitions on
the same fileserver. Clarify the rules here.

Reviewed-on: http://gerrit.openafs.org/5745
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit c59c07132ef17589ad3466ded418c1370539b7af)

Change-Id: I68898c21e2c1aa7d6828b475dff4c1cb6d14327c
Reviewed-on: http://gerrit.openafs.org/6317
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

ubik: add interface for reading during write locks

Add ubik_BeginTransReadAnyWrite, which allows for reading from the
database, even while there is a conflicting ubik write lock. Reads are
still blocked while the local database is updating due to a write
transaction commit.

Reviewed-on: http://gerrit.openafs.org/2592
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit fd7f7469bdbde24320c3ac8b4f7015d495b47bff)

Change-Id: I46cf35beeea6dc997976ea2c33e45c1a5f4904df
Reviewed-on: http://gerrit.openafs.org/6261
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

ubik: Fix buffers for reading-during-writes

If we are reading while a write transaction is in progress, we can
encounter a buffer that is dirty if we're on the same site as the
writer. Ignore these buffers for readers, since they contain
uncommitted changes. Then, when the writer commits, invalidate the
resultant duplicate buffer, if one exists.

Reviewed-on: http://gerrit.openafs.org/2231
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 570236dd55e79e8886b45f19cef59ec3fa9d53f6)

Change-Id: If83e077b99a77339377a36f1d1bac9182ef26e4f
Reviewed-on: http://gerrit.openafs.org/6260
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

ubik: Abstract buffer matching and pass trans ptrs

Abstract the code for matching buffers in DRead, and pass around
ubik_trans pointers instead of ubik_dbase pointers. This changes no
behavior, but makes changing ubik buffer code a bit easier.

Reviewed-on: http://gerrit.openafs.org/2230
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 5c7297a6c76777aec6209aa191b067ddc73b1a87)

Change-Id: I6cc673d965b5e06361fff8ea259ba552446e3213
Reviewed-on: http://gerrit.openafs.org/6259
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

macos: no more startupitems

if we're upgrading, a "cruft" startupitems script should just
be nuked.

FIXES 129601

Reviewed-on: http://gerrit.openafs.org/4387
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 841c8022e3978fffe1357f339d948e93c9db9d52)

Change-Id: Ibe2dde067df1fe22da1bb09039d07ebbc178af08
Reviewed-on: http://gerrit.openafs.org/6277
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

fix manpage for udebug -servers

The proper option to udebug is "-server", not "-servers". Fix the manpage to line up with the binary.

Reviewed-on: http://gerrit.openafs.org/4457
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit d5ded9afa1764016977a17ef9346dc979bf70ca0)

Change-Id: I66bea4923717a6b7df5fbefa1dd93d469d96be3c
Reviewed-on: http://gerrit.openafs.org/6287
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

doc: Add aklog_dynamic_auth manpage

Add a manpage for the aklog_dynamic_auth LAM module.

Reviewed-on: http://gerrit.openafs.org/4485
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 24ddd06a0dc19ad7c0b2a2847d41a5ea69665b15)

Change-Id: Ib5f7463b75ee8d58ecc6b725f629fd7600c64792
Reviewed-on: http://gerrit.openafs.org/6297
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

afs: prevent nat pings on destroyed connections

If the connection is no longer in use, reset the nat keep alive
timer to 0 seconds.

Reviewed-on: http://gerrit.openafs.org/6230
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 7a93c827f353e2ad695d8019ab789b02a3596f2e)

Change-Id: I019bdce7a8f9f0123786b46d0563f42f7b26c14b
Reviewed-on: http://gerrit.openafs.org/6232
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Stephan Wiesand <stephan.wiesand@desy.de>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

FBSD: switch afsi_SetServerIPRank implementation

Upstream has removed the ia_net{,mask} elements from
struct in_ifaddr, so we can no longer use them directly.
Switch to passing an rx_ifaddr_t (i.e. struct ifaddr*) in instead,
as that uses a slightly different codepath which still works
for our purposes.

We compile the kernel module with -Werror, so storing a pointer
(memcpy return value) in an int is forbidden, hence the conditional
declaration of 't'.

Reviewed-on: http://gerrit.openafs.org/6203
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 1ef8dc3dfbddcbe5610c276afc627c9fcfe30a65)

Change-Id: Icb338331c879dc4cbd69ae565b115de000e8d3d0
Reviewed-on: http://gerrit.openafs.org/6235
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>

FBSD: cleanup dvp locking for ISDOTDOT

This is a more correct version of
c2ed2577f9c16df3088158fb593d7aab6e8690d0, which was reverted since
it caused build issues on some versions and kernel panics on others.

We do want to always unlock dvp before calling over the network
in the ISDOTDOT case, but be sure to use the proper spelling
for this operation (as the syntax has changed between FreeBSD versions).
This requires not unlocking dvp right after the afs_lookup() call if
it succeeds, letting us just lock the "child" vp (which is actually
the parent starting from '/') first, and then re-lock dvp.

The error case of afs_lookup() was already handled correctly in
this logic, which is to say that it was incorrect before this change,
attempting to recursively lock dvp which causes a panic.

Reviewed-on: http://gerrit.openafs.org/6127
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 8e14168c9c77850ce0603d56f8aa280f73cb3114)

Change-Id: I2d929875a9b008f0aa0b64bce7971932691d9956
Reviewed-on: http://gerrit.openafs.org/6234
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>

FBSD: typo fix

Gerrit/5572 added conditionals on __FreeBSD_version >= 900044, which
is (approximately) when a bunch of kernel API renames happened.
(There has since been a dedicated version bump to 900045 a month
or two post-facto, but 900044 should be fine for now.)
However, 900044 is not 90004.

Reviewed-on: http://gerrit.openafs.org/5657
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit ca8ba7e7a41e61cc943b34196872b72a190432e3)

Change-Id: Ibacd6d037590df0696b8dab8107d36bca470785a
Reviewed-on: http://gerrit.openafs.org/6233
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>

remove CopyOnWrite2 and unused vars

unused static functions make grumpy compilers grumpy.

(cherry picked from commit 7a6cf450babea4e706392cb8ad83586fa7f8e558)

Change-Id: Ib40be2b9458ca799ba2b808ed86a627c40cf6ec4
Reviewed-on: http://gerrit.openafs.org/6161
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

viced: disable accelerated copyonwrite

don't do multistage CoW

FIXES 130295
(cherry picked from commit 97540f78aa333feefe9f567ee71695c1d6451f5c)

Change-Id: I1cd9d54ec9202aa4a11828cd443622413f195687
Reviewed-on: http://gerrit.openafs.org/6160
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

Unix CM: Fix PutVolume in afs_BlackListOnce

It isn't safe to keep accessing a structure once we have Put our
reference to it.

Reviewed-on: http://gerrit.openafs.org/6117
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 2373547a86c12f72261072971b8d47e6b7cdd4e6)

Change-Id: I013b5d6ecf7705745ab16c8736f32ca7750b933b
Reviewed-on: http://gerrit.openafs.org/6124
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

afs: Leave cellnum alone for explicit mtpt cell

When a mountpoint is given an explicit cell, don't alter cellnum.
Cellnum represents the cell for the parent, and is used for
determining whether or not we're crossing a cell boundary.

Previously, this code forced the mount point to always be treated as
foreign (for a mountpoint prefixed with a cell name), or to always be
treated as local (for a mountpoint prefixed with a cell number).

Reviewed-on: http://gerrit.openafs.org/6051
Reviewed-by: Marc Dionne <marc.c.dionne@gmail.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit e14dec55e6600edb60ce5184b4ab1f646c68947b)

Change-Id: I8d2b378a95533aabd65e7121000ce34238d728b7
Reviewed-on: http://gerrit.openafs.org/6123
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

dafs: avoid null deref getting volume header

we don't assign hd before dereferencing. stop dereferencing
and just use the queue cursor.

Reviewed-on: http://gerrit.openafs.org/5814
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 00beae8c4e3b390637f20584e676cf86c817dadd)

Change-Id: Iec2dbbf84e4f08f02998c9e998d2ea899ce48a5d
Reviewed-on: http://gerrit.openafs.org/6122
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Linux: 3: Update specfile to know about 3.* kernels.

Update spec file to be consistent with acinclude.m4 with regards to
sysnames.  We don't bother updating the code inside the legacy kernel
build section, as it doesn't get triggered for 3.* kernels (it should
probably get cleaned up at some point.)

Also, fix a bug in error message printing of unrecognized kernel.

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
Reviewed-on: http://gerrit.openafs.org/6120
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 0f9214164ad56bfe74d0f2cec8775a312f5128dd)

Change-Id: Ie414751cf01c44a73bf9eaef67c593cced2680d9
Reviewed-on: http://gerrit.openafs.org/6121
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Linux: make sure backing_dev_info is zeroed

The afs backing_dev_info structure is allocated dynamically
without zeroing out the contents.  In particular there's no
guarantee that congested_fn is NULL, causing spurious oopses
when bdi_congested in the kernel tries to call it.

(adapted from commit 8e97cf6f215d5575c63d86eaec59031399f4beda)

Change-Id: I83755b6bb5ec2fada7e077c00d3d8edf8af1cae4
Reviewed-on: http://gerrit.openafs.org/6104
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Linux: 3.1: adapt to fsync changes

The fsync file operation gets new arguments to specify a range.
Add a configure test to check for the API change.

The inode lock is also pushed down into the operation, so we need
to take it ourselves to keep the original behaviour.

Reviewed-on: http://gerrit.openafs.org/5332
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit cbaefa266d433af3b9a082a360e23a42f161d80f)

Change-Id: Idb6770204b014c62a8611548509240f8b5f950bc
Reviewed-on: http://gerrit.openafs.org/6098
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

Linux: 3.2: Use set_nlink to update i_nlink

As of v3.2-rc1~84^2~1, struct inode.i_nlink is now const to prevent
direct modification.

Signed-off-by: Anders Kaseorg <andersk@mit.edu>
Reviewed-on: http://gerrit.openafs.org/6096
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit f1dd2d696fb9ab71b4192b156042e0c63019c58a)

Change-Id: I685aa6e8638e8fe864f1a6a7e428dfb6839ebcea
Reviewed-on: http://gerrit.openafs.org/6099
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Linux: d_delete now takes a const argument

The d_delete dentry operation now takes a const argument.  Test for
this and define our function accordingly to avoid warnings.

Reviewed-on: http://gerrit.openafs.org/5335
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit def00ae8ba774f4d7290c29208484dda756dc0e0)

Change-Id: I128a721587f4963849652cb389472a550d81d267
Reviewed-on: http://gerrit.openafs.org/6097
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

redhat: make rpms build on pre-f15 again

we don't have systemd everywhere. don't assume it

Reviewed-on: http://gerrit.openafs.org/5813
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Tested-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 47b010d1e1d0e8bdbbcad73d5aedc19719d5d96a)

Change-Id: I200ca2306b0e6a22cdcb0515641977b2ea6d21a0
Reviewed-on: http://gerrit.openafs.org/6090
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

linux: Update Packaging to build OpenAFS services for Fedora's systemd

Fedora 15 now uses systemd (see http://fedoraproject.org/wiki/Systemd)
for the OS init system. While it currently has backwards
compatibility with older SysV-style init scripts, future versions of
Fedora may no longer support it, and OS startup tends to be faster
with the systemd service units. Also, systemd runs all the service's
processes within a linux kernel cgroup.
(see http://www.kernel.org/doc/Documentation/cgroups/cgroups.txt)

This change includes an openafs-client.service and
openafs-server.service unit files for the client and server packages
respectively.

Client
- Loading the openafs module was moved into
/etc/sysconfig/modules/openafs-client.modules. This causes the OS to
load the module on boot. This is the preferred way for modules to be
loaded with Fedora. (See
http://docs.fedoraproject.org/en-US/Fedora/15/html/Deployment_Guide/sec-Persistent_Module_Loading.html
for more details)
- The CellServDB file is generated with sed rather than cat.
This change was made because Systemd doesn't execute as a shell
script, but rather executes processes directly. Rather than invoking
a shell to concatenate the CellServDB.* files, they're written to the
CellServDB file using a sed oneliner.
- Do all of the proper kernel module loading and unloading.

Server
- Since systemd uses cgroups, when the service is shut down, all
processes in the openafs-server.service cgroup will be terminated.

The other changes are standard as per:
    http://fedoraproject.org/wiki/Packaging:ScriptletSnippets#Systemd

Original version by Jonathan Billings <jsbillin@umich.edu>.

Signed-off-by: Edward Z. Yang <ezyang@mit.edu>
Reviewed-on: http://gerrit.openafs.org/5637
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Alex Chernyakhovsky <achernya@mit.edu>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 286ffa0d7c4d594ff107b70f9e930271c027a79e)

Change-Id: I653b211ef24fb9010f0a8ff0ad70c11703a64e82
Reviewed-on: http://gerrit.openafs.org/6089
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Linux: 3.1: update RCU path walking detection in permission i_op

The permission() inode operation changed again with kernel 3.1,
back to the form it had before 2.6.38.  This compiles fine,
but is missing the new way of detecting when we get called in
RCU path walking mode, resulting in system hangs.

Reviewed-on: http://gerrit.openafs.org/5740
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 4952df3f0359531e4a660c99c94c51eb0b169f59)

Change-Id: Ibd497309e6699fb585cf70e618373e800b73cbb8
Reviewed-on: http://gerrit.openafs.org/6088
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

klog.krb5: enforce DES for rxkad

0. Always request a TGT regardless of the state of
   writeTicketFile.

1. request des-cbc-crc when requesting a ticket for an
   rxkad service principal

2. check the returned key length to ensure that it matches
   the permitted length of an rxkad key.  If not, generate
   an error instead of overwriting memory and continuing.

FIXES 130278

Reviewed-on: http://gerrit.openafs.org/5619
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 3a9a5783cd1fd73902655f0876e2069b42688c94)

Change-Id: Ia162cb9fc1d67a38cde26e52b9077f5634a62c84
Reviewed-on: http://gerrit.openafs.org/6087
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Solaris: Specify ARCHFLAGS in CFLAGS

Various autoconf tests which use the C compiler may yield different
results depending on whether or not we are compiling for x86 or amd64
on Solaris (different libraries are available, structures may be
different, et al). So, set CFLAGS depending on which arch we are
targeting, so the autoconf results are more consistent with the actual
compilation during the build.

Reviewed-on: http://gerrit.openafs.org/5786
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 5eba97a8960c49aeea957dc94408368e1324eff0)

Change-Id: Idd0f7d2220ae9762af60e96d087e7eb7ee8c0e63
Reviewed-on: http://gerrit.openafs.org/6086
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

DAFS: Deal with exclusive-state volume headers

GetVolumeHeader assumes that headers on the LRU are not associated
with a volume in an exclusive state. This is known to not be true for
some cases when salvage requests are received over FSSYNC, and may be
true in other scenarios. It's easy to just skip such headers, so skip
them.

Reviewed-on: http://gerrit.openafs.org/5808
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit c2ee7b00ddfe8b1b8c296a99f0244eb33c726e8e)

Change-Id: I0eda511071114b01cd8984d30858b0aa9157727a
Reviewed-on: http://gerrit.openafs.org/6085
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

ukernel: handle pioctl errors properly in ktc

the straight return from call_syscall isn't what we want. munge so
e.g. EDOM handling for ktc_GetToken does the right thing.

Reviewed-on: http://gerrit.openafs.org/5807
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit b06c700ee40ca70ef08259af819e2a6084137457)

Change-Id: I8929615940cb4f025bdd04139399f0ceb82955c9
Reviewed-on: http://gerrit.openafs.org/6084
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

salvager: Implement AskDAFS via SYNC flags

Instead of probing the DAFS-ness of the fileserver by probing which
FSSYNC opcodes it supports, detect DAFS-ness by looking at the SYNC
response header flags, which explicitly state whether or not the
endpoint is DAFS. This avoids unnecessary "protocol mismatch" log
messages when the endpoint is not DAFS.

Reviewed-on: http://gerrit.openafs.org/5800
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 7fee28bae30c5a65898a2e3fd7149f07e8ca0df6)

Change-Id: I0643b0e61795f39b63783c937e88c4a74c3c8312
Reviewed-on: http://gerrit.openafs.org/6083
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

regen: Fail on failures

If something fails, don't ignore the error.

Reviewed-on: http://gerrit.openafs.org/5817
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 6d519cf1234d54630dcfe48f6ecd393e2dfe71dd)

Change-Id: I5e5ac4d843b9dafce7b7f3acadfcdb0901aed214
Reviewed-on: http://gerrit.openafs.org/6082
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

viced: avoid bogus handle in rename

because fileptr's handle can be reset during CopyOnWrite
we need to SetDirHandle after that call

FIXES 130215

Reviewed-on: http://gerrit.openafs.org/5818
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit a1669f3e37299c11bf29a458212e03d1bdb22334)

Change-Id: I57e084102048916b04e45aa32792dc91193e43d8
Reviewed-on: http://gerrit.openafs.org/6081
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

volser: Preserve needsSalvaged during restore

Some of the routines during a volume restore may set needsSalvaged, if
an inconsistency is detected while writing the given volume data.
However, after the data is read, we set the volume header information
to what was found in the dump stream, ignoring any needsSalvaged that
may have been set.

To ensure that inconsistent volumes in this situation actually get
demand-salvaged (for DAFS) or offlined (non-DAFS), keep the value of
needsSalvaged in the header, if it was set.

Reviewed-on: http://gerrit.openafs.org/5822
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit 9ba891676a15512fa58915662cf599ea73eac430)

Change-Id: Iece9f5b03f0454b7f833e4f3e86da4478b4441fe
Reviewed-on: http://gerrit.openafs.org/6080
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

namei: Remove extraneous rmdir

We just unlinked the file, so we know we won't be able to rmdir() the
same thing. Give a path one level higher to
namei_RemoveDataDirectories, so we start rmdir()ing at the parent dir.

Reviewed-on: http://gerrit.openafs.org/5833
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
(cherry picked from commit b41e30a55c8aeb6d7f42d74f9eb090232032057f)

Change-Id: Ibd14490e2d85bd78522472203ad57120237bff22
Reviewed-on: http://gerrit.openafs.org/6079
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

namei: tweak getlinkbyte error handling

in the event of a pread error, add the same error exit path
trigger that we have elsewhere

Reviewed-on: http://gerrit.openafs.org/5843
Reviewed-by: Andrew Deason <adeason@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 51f71e533b20cd875eec8adda71edea044a606da)

Change-Id: I05623d392e7e3446b7b41dcc1778c6aca986bca1
Reviewed-on: http://gerrit.openafs.org/6078
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

Windows: Track active RPCs per scache_t

It has been noticed that multiple RPCs can be active on
a cm_scache_t object at the same time.  This is especially
true of directory objects with the redirector.  Track the
number of active RPCs and use that number in cm_MergeStatus
when deciding whether or not to discard the cached data for
the object.

Reviewed-on: http://gerrit.openafs.org/6001
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 090f6279c5496f648893606d298c698f376c7ae0)

Change-Id: Ic40b73ae44c47ad6077fcbbe41bf0d783ab776fc
Reviewed-on: http://gerrit.openafs.org/6045
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: Add Vista/Win7 Firewall Configuration

Make use of the INetFwPolicy2 COM interface to configure
the Vista/Win7 firewall for all network profile types.

Reviewed-on: http://gerrit.openafs.org/5812
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit b15b4b15822bee87720bb0d78c2ce54051180339)

Change-Id: I3022b2480c7c7273e819d0a10c092f22f4af3800
Reviewed-on: http://gerrit.openafs.org/6044
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: cm_Rename break in 9cf7a628c2ff178b2fa236d2c0a41ca9be315036

9cf7a628c2ff178b2fa236d2c0a41ca9be315036 broke renaming of
a file to a name that does not exist.  Put back the code that
makes that possible with a comment.

Reviewed-on: http://gerrit.openafs.org/5735
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 7a426424b7fb45659f6000a564e8c71a9108a33c)

Change-Id: Ia82ddc46b303cb80380c342d44db437c490b8508
Reviewed-on: http://gerrit.openafs.org/6043
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: out of date version not in current chunk

In buf_GetNewLocked(), the comparision to decide whether a
cm_buf_t is a member of the current chunk must take the data
version into account.  If the data version is out of date, it
is not part of the current chunk and is an object that can be
safely recycled.

Reviewed-on: http://gerrit.openafs.org/5734
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 1dad690fc34fdd8d9df8a3be6d580441b7cae8bb)

Change-Id: I626562d2189182b434f8da0f4f0a6213c5b695fe
Reviewed-on: http://gerrit.openafs.org/6042
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: only flush buffers on shutdown if running

If a service shutdown message is received prior to the
service entering the running state, do not attempt to
buf_CleanAndReset() because the required data structures
and locks are not initialized.

Reviewed-on: http://gerrit.openafs.org/5733
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 9743f5729d1a55513e861bb1334599a8e597de6b)

Change-Id: I2bda1c4a538c83595f2d60ad4661c7e6865c5fc3
Reviewed-on: http://gerrit.openafs.org/6041
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: Do not EEXIST exact match during rename

AFS Rename operations on the file server will delete a
target file if it exists.  Do not prevent renames because
an exact match of the target name exists in the target
directory.

Reviewed-on: http://gerrit.openafs.org/5731
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
(cherry picked from commit 9cf7a628c2ff178b2fa236d2c0a41ca9be315036)

Change-Id: I02c32fab18053e1a37811089812f402d351a6666
Reviewed-on: http://gerrit.openafs.org/6040
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: validate buffer hash tables in cm_MergeStatus

Reviewed-on: http://gerrit.openafs.org/5425
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit cf57eb294f28084cbb138b1a06bc813e14e6d0ec)

Change-Id: Ia685aebdc443113b4ddb968155f306cc53680af9
Reviewed-on: http://gerrit.openafs.org/6039
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: record history of read lock threads

Maintance an array of up to 32 reader threads that have
acquired a rwlock.  Use it for debugging when things go bad.

Reviewed-on: http://gerrit.openafs.org/5411
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit c9d655b9a6edf02fd92455961445f239f4667821)

Change-Id: Icab5ffc36e4c6245b7d0699596eca66a66bc9538
Reviewed-on: http://gerrit.openafs.org/6038
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: ignore SYNCHRONIZE privilege in cm_CheckNTOpen

SYNCHRONIZE has not meaning for AFS so just ignore it.  Do not
deny an open request because it is set.

Reviewed-on: http://gerrit.openafs.org/5410
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 793fdd3360982c48bf60eefbd9c56b4a3c1e09f0)

Change-Id: Ice8758dcbd0b355c9fdfb5af293b6d670ec8c1e4
Reviewed-on: http://gerrit.openafs.org/6037
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: permit offline volume check to be disabled

Setting the registry value to 0 can now be used as a
method of disabling the offline volume check.

Reviewed-on: http://gerrit.openafs.org/5370
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 19a5b5e67c65f0a9109b0fb5410b5ba74ad85b41)

Change-Id: I50dfa3c5387ba8a1cbdd3ca80f9e8dc03e92c404
Reviewed-on: http://gerrit.openafs.org/6036
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: add clean rule to src/afs/NTMakefile

Reviewed-on: http://gerrit.openafs.org/5368
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 4ec6fd003084b02c299466bcaeac3850be10f79c)

Change-Id: I0cadd04598e78f9b8a5c0774e0636af153a5030c
Reviewed-on: http://gerrit.openafs.org/6035
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: correct comments in cm_BkgStore

Reviewed-on: http://gerrit.openafs.org/5369
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit b6e5493437b942d5550247761dac3311c6d73a86)

Change-Id: I73c2fcab9ccff73e4419d3edf3ffc7395230e6e5
Reviewed-on: http://gerrit.openafs.org/6034
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: remove CM_BUF_CMBKGFETCH flag

The processing of the CM_BUF_CMBKGFETCH flag adds more
overhead than it prevents.  Get rid of it.

LICENSE MIT

Reviewed-on: http://gerrit.openafs.org/5359
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit b6f23864ffa380923f50d424063ad0804e407f4a)

Change-Id: I423aa105bad99c70942d6b967a1dde70216fb096
Reviewed-on: http://gerrit.openafs.org/6033
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: use lock conversion in cm_EndDirOp

Instead of dropping the lock for read and reacquiring for write
use lock_ConvertRToW() which will make the change atomicly if
it is possible or place the thread into the wait list if not.

LICENSE MIT

Reviewed-on: http://gerrit.openafs.org/5360
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 33af0f92e9b8a50c5a6dd17a9065c0ef58597c28)

Change-Id: Id92298fa848becb4131139579fd31b0c5f98a4ad
Reviewed-on: http://gerrit.openafs.org/6032
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: cm_SymLink export created cm_scache_t

Permit the caller to request the cm_scache_t that represents
the created symlink or mount point object.

Reviewed-on: http://gerrit.openafs.org/5355
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 9b0cb614387403ff433cf39f84c13b626b1b9bdc)

Change-Id: I170059255e3bbc7e3ccbd59b66d7d8b11005a94f
Reviewed-on: http://gerrit.openafs.org/6031
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: correct cm_buf use of Head/Tail queues

The buffer free list least recently used queue has both
head and tail points.  Use the proper versions of the queue
mgmt functions and do not handle edge cases as special cases.

Reviewed-on: http://gerrit.openafs.org/5354
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 537592f6bbda0c659c25ae7068f50504e2a5e116)

Change-Id: I0d5afaf59895f34ced739f4626a351fe41186008
Reviewed-on: http://gerrit.openafs.org/6030
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: remove dead code from cm_scache.c

Reviewed-on: http://gerrit.openafs.org/5353
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 1e23761b7fadc260b32bebb0343a9b5b1de37cf4)

Change-Id: Ifd8cf9ec7b517da80b22853e0f90034586560e0d
Reviewed-on: http://gerrit.openafs.org/6029
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: add logging to smb_IoctlRead

Reviewed-on: http://gerrit.openafs.org/5350
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 69463ac0cb367202485aac9977f8b4c47fb067f2)

Change-Id: I1f72db5148d2f596f4b0b58cb6788a244fd4fa44
Reviewed-on: http://gerrit.openafs.org/6028
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: remove potential data loss warning with cast

Reviewed-on: http://gerrit.openafs.org/5349
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 2588dd12478afb96a72f0cd788233ecf2145830f)

Change-Id: I02a87c4980c6ebe6ef671f90e2c6000ab2eeb60e
Reviewed-on: http://gerrit.openafs.org/6027
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

windows: pass cm_req_t through background ops

add cm_req_t object to background daemon operations in order to
permit request source to be propagated to the background daemon.

pass a cm_req_t into buf_SetDirty().

Reviewed-on: http://gerrit.openafs.org/5348
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 722843e2df533e8fcf245f2a3b10ef686992deb2)

Change-Id: I2fbc78cf765c534fc6fc673695634fcf946dca0b
Reviewed-on: http://gerrit.openafs.org/6026
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: correct log message in buf_Sync()

Reviewed-on: http://gerrit.openafs.org/5347
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 20772897654b58da6f7d4cd0bd2caf169abddd50)

Change-Id: I676a919701c4edfbc130a04c94f4092676a1d53c
Reviewed-on: http://gerrit.openafs.org/6025
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: buf_SetDirty do nothing if length is zero

Reviewed-on: http://gerrit.openafs.org/5358
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 7dcc4766377b99c60db60328c21a3990db132268)

Change-Id: I30dc0f218df56e6e38fe6ce3736a5450b7036e51
Reviewed-on: http://gerrit.openafs.org/6024
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: remove MULTIHOMED ifdef

All builds include the MULTIHOMED functionality.
Get rid of the #ifdef MULTIHOMED.

Reviewed-on: http://gerrit.openafs.org/5333
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 2271d481d33b2b2ad56dcd4fc9fbdd726fc1967b)

Change-Id: Ia11e4bfdc83737720e5274bad3d01232f0f2dbbe
Reviewed-on: http://gerrit.openafs.org/6023
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: missing ! in update volume location

The check to see if the volume name is numeric or not was
missing a ! in order to perform the correct test.  Add it.

Reviewed-on: http://gerrit.openafs.org/5331
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 8127e9332be8cadda001f05c6036d4c87e04cb4a)

Change-Id: I9760845b597c1653a78ab3e496166e5df03b7575
Reviewed-on: http://gerrit.openafs.org/6022
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: improve handling of volume rename

The windows cache manager tracks volumes by volume group.
Up to this point all volume location updates have been performed
by the volume name.  What if the volume name was altered?  In this
case the volume location information for the in use volume ids will
fail until a mount point to the new name is queried.  Before
marking the volume group as non-existent attempt to perform a
lookup using either the volume id for the readwrite or readonly
volume.

Reviewed-on: http://gerrit.openafs.org/5330
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit f37b6dfaf050ec0270979c8310fb471c16b66cd3)

Change-Id: I33771780a4ace4aa5b4613bef4590d5efaf15922
Reviewed-on: http://gerrit.openafs.org/6021
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: refactor volume location updates

Break out the VL_GetEntryByName RPC calls into support
functions so we can reduce the amount of duplicated code.

Reviewed-on: http://gerrit.openafs.org/5329
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
(cherry picked from commit 9563dd68fc1fb4b86016106aaacd33732d4cf6ed)

Change-Id: I7fc8b9ac2902c0909b39c446468307fba30b2a51
Reviewed-on: http://gerrit.openafs.org/6020
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

Windows: netidmgr krb5_cc_get_principal can fail

Do not dereference a NULL pointer if krb5_cc_get_principal fails.

On master this bug is fixed by e55d1774b1b5b27a3617467b5e2a24ee2be3a38c
but that change is after the conversion to the Kerberos Compatibility
SDK and cannot be applied to openafs-stable-1_6_x.

Change-Id: I646011dd3171fd1aa6f40f349329f8261546b809
Reviewed-on: http://gerrit.openafs.org/6019
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>

volser: Remove debugging log messages

While the -log option to volserver is supposed to print additional log
information, it shouldn't spam the log with useless data. Remove some
of the log lines that are really more "debug" information, so we log
the same amount of information as in the 1.4 series.

Reviewed-on: http://gerrit.openafs.org/5810
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 2a1251c5d0c1b0fd4024d7b3282187a838b551fd)

Change-Id: I1af47fa1910a98a6175f1f000d0432af1e1126e8
Reviewed-on: http://gerrit.openafs.org/5831
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

rx: Don't clear the receive queue when out of packets

We can end up discarding a receive queue that's been soft acked,
effectively taking back soft acks we sent. Whilst the RX
documentation says that a client can drop soft acked packets at
will, our RX implementation assumes that if the final packet in
a call has been soft acked, we won't clear the queue. If a client
clears the queue in this situation, the call will hang.

What *should* happen is that we should take necessary locks,
confirm that we have not soft-acked all of the packets in a flow,
and then discard, or, if we're just going to discard, error the
call.

Reviewed-on: http://gerrit.openafs.org/5603
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>
(cherry-picked from f722a75fe21963d64217375f1f7bbb0eb14befb4)

Change-Id: Ie2f14a0ed7b6d258a7f456387df44ac95fd1748c
Reviewed-on: http://gerrit.openafs.org/5820
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

afs: Only use actual connections for GetTime calls

The for() loop that makes an RXAFS_GetTime call in afs_CheckServers
was iterating over conns and rxconns from 0 to j. However, 'j' here is
just the size of the allocated array, whereas 'nconns' is the number
of structures in the array actually initialized. So, just go up to
nconns to avoid using unitialized connections and Rx connections.

This is a 1.6-only change. On master, the -settime code has been
completely removed in change
Id291f5f88b1ad84594706f2a1a02a933dddd0cb9.

Issue reported by Sebastian Hanigk.

Change-Id: Iec782a4464899b5a7e49e193b04f79e31c6b37f2
Reviewed-on: http://gerrit.openafs.org/5788
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>

make bozo honor -rxbind correctly

Bozo needs to call rxInitHost() rather than rxInit() when -rxbind is
present. This patch causes it to read NetInfo/NetRestrict earlier in
the startup process so it can make that decision.

FIXES 57286

Reviewed-on: http://gerrit.openafs.org/4729
Tested-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 6260c4ae51e7a2807596ddf1bd49620f98eed94c)

Change-Id: If873686f7a7742f82fd5fadf43c6f44379448c29
Reviewed-on: http://gerrit.openafs.org/5771
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

Suppress cmp component version error messages

When we use cmp to determine whether to replace
AFS_component_version_number.c, suppress stderr in addition to stdout,
to slightly reduce output during the build.

Reviewed-on: http://gerrit.openafs.org/4471
Reviewed-by: Simon Wilkinson <sxw@inf.ed.ac.uk>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 21578144e08d46eeec9a2944e92e8d0d7a6dba57)

Change-Id: Ie06f293e3750abb7514ee3795c1af41a26f46642
Reviewed-on: http://gerrit.openafs.org/5770
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

viced: Check vnode length on dir ops

The commit aadf69eabb1962496fa93745ab560a5b48cacd61 added checks on
vnode length whenever we read or write from a vnode. Add the same
check on directory vnodes when we modify the directory (whenever
entries are added or deleted).

Reviewed-on: http://gerrit.openafs.org/4233
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 2578555d7e08131bf2fe4cdd0aa4b32567a76eb2)

Change-Id: I105f0dc7b480f7328205ae02929f6ca187aa962a
Reviewed-on: http://gerrit.openafs.org/5769
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

viced: Check vnode length on read and write

When reading or writing a file vnode, check that the length of the
vnode in the vnode index matches the size of the on-disk file
containing the data for the file. If it does not match, take the
volume offline (and for DAFS, demand-salvage it).

Reviewed-on: http://gerrit.openafs.org/4121
Tested-by: Andrew Deason <adeason@sinenomine.net>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit aadf69eabb1962496fa93745ab560a5b48cacd61)

Change-Id: I324dc79a24fc8c199825ffefcc1dd2b0097942a1
Reviewed-on: http://gerrit.openafs.org/5768
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

DAFS: DFlushVolume outside of vol glock

DFlushVolume may traverse a long list of directory objects, and can
even hit the disk, so we should drop the glock for it. This should be
safe in DAFS, since we already transition the volume to an exclusive
state before doing this, and DFlushVolume only deals with structures
internal to the directory package and maintains its own locking.

Reviewed-on: http://gerrit.openafs.org/4242
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit a18175f479437f3a4fe68460538927630070cd13)

Change-Id: I355c53c6f224c5471da70827f30511986abeac27
Reviewed-on: http://gerrit.openafs.org/5767
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

afs: Retry unlock after afs_StoreAllSegments

HandleFlock calls afs_StoreAllSegments when unlocking an exclusive
flock lock. This can drop the write lock on avc, so we must
effectively retry the entire lock operation again, since the world may
have changed while we were waiting to reacquire the lock on avc. So,
retry once all of the lock checks up to that point, to ensure that a
lock on the file actually still exists.

FIXES 125446

Reviewed-on: http://gerrit.openafs.org/4393
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 601fcf1d7f7c88cfc0ffd877c5458340b3e59098)

Change-Id: Ibd215d586485d8f18f77665f5f85e6bee579e35e
Reviewed-on: http://gerrit.openafs.org/5766
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

afs: Avoid memory leak on recursive write flock

When a process requests an exclusive lock on a file on which it
already holds an exclusive lock, we basically form a no-op. However,
HandleFlock was allocating a new SimpleLocks and attaching it to
avc->slocks, without freeing the old SimpleLocks structure.

Since we don't need to do anything if we already hold an exclusive
lock, just break out of the loop right away when we detect that
scenario. Thus we avoid adding a new structure to avc->slocks, and we
avoid a memory leak.

Reviewed-on: http://gerrit.openafs.org/4395
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit ba9ae1ed7c269d7c080b5ce99b3b4bb2fe0a2a6f)

Change-Id: I9e45a51e02774c555c6d5cb5ba9d0407b8442215
Reviewed-on: http://gerrit.openafs.org/5765
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

vol: avoid MAXINT redefinition warning in vnode.c

Reviewed-on: http://gerrit.openafs.org/4179
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit b1d51d2a6030fb8aabdae1fedb6846dc7ef9fbca)

Change-Id: I8ea833cca442c67468982da31ed1b6fe04508a34
Reviewed-on: http://gerrit.openafs.org/5764
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

DAFS: Correct FSYNC_VOL_QUERY_VOP checks

Check that the given partition matches the vp partition, and ensure
the vp is not in an exclusive state when we check the state.
Otherwise, we may return pending vol ops for a volume on a different
partition, or we may incorrectly return that there is no pending vol
op when in fact the volume does not exist at all.

Reviewed-on: http://gerrit.openafs.org/4308
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 8234cc254406173a7ada9fb1b4a63ca8aa626bca)

Change-Id: I912c1d0b34fbbaae9ce4bd8edf6e3dd7f718be00
Reviewed-on: http://gerrit.openafs.org/5763
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

salvager: Give back volumes when exiting early

Sometimes the salvager exits a bit earlier than normal. For instance,
when no applicable inodes are found for a volume group, or if the
-inodes command line option was given. In these cases, we have already
checked out singleVolumeNumber from the fileserver (if we're salvaging
a single VG), so we need to give it back. So, give it back in those
instances.

Reviewed-on: http://gerrit.openafs.org/4297
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 637f5b642aaee3a6f4642fd0ba15e2a9eec6c496)

Change-Id: Ib40ffb84fa6c40cfd692c561b913decb97268403
Reviewed-on: http://gerrit.openafs.org/5762
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

DAFS: Do not record vol ops for DELETED vols

When a volume is VOL_STATE_DELETED, it effectively does not exist, so
there is little point in recording a vp->pending_vol_op structure for
it. Just let callers checkout the volume as they would a nonexistent
volume: without recording anything about the operation.

This just reduces some edge cases and confusing debugging info, so we
don't have to worry about cleaning up pending_vol_op structures for
nonexistent volumes.

Reviewed-on: http://gerrit.openafs.org/4296
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit dba991728ba5a90af316928348053189dac80398)

Change-Id: Ica0c53e62ed885710210c7e1344e12b382302572
Reviewed-on: http://gerrit.openafs.org/5761
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

salvager: Do not AskOnline nonexistent volumes

If singleVolumeNumber is not in our volume summary list, then the
singleVolumeNumber volume does not exist. So, don't try to bring it
back online. Still do try to make sure we don't have the volume
checked out, though, so issue an AskDelete, so ensure that it's not
checked out and that the fileserver does not think it exists.

Change AskDelete so we don't care if we tried to delete a volume that
the fileserver thinks already doesn't exist. Change the FSYNC_VOL_DONE
handler so it does not complain about already-deleted volumes.

Reviewed-on: http://gerrit.openafs.org/4295
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit f2c1f2ee48663d00701e7cee52b99151e36b9534)

Change-Id: I708c173ae6cbd3f40cc09f6fc3229a6f038baf3f
Reviewed-on: http://gerrit.openafs.org/5760
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

Fix ihandle.c indents

ihandle.c had some blocks that were not indented. Indent them.

Reviewed-on: http://gerrit.openafs.org/4273
Reviewed-by: Jeffrey Altman <jaltman@openafs.org>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
(cherry picked from commit 9ac85bb2fd91baa4c3077d091f395f57e933018a)

Change-Id: I83604d19759a183fa10989596734a405515ac19c
Reviewed-on: http://gerrit.openafs.org/5759
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

vol: Handle large volume IDs in VLockFile

VLockVolumeByIdNB currently cannot handle volume IDs larger than
2^31-1. Fix this by using struct flock64, F_SETLKW64, and F_SETLK64 in
the VLockFile functions where possible.

Thanks to Simon Wilkinson for pointing out F_SETLK64.

Reviewed-on: http://gerrit.openafs.org/4198
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 85edec7105dd653ed1d8046b69b8262281e97bba)

Change-Id: Iee09a8bc4a0aabbd12fe156940fa61d6947642e8
Reviewed-on: http://gerrit.openafs.org/5758
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

vol-salvage: Fix AskOnline error message

Reviewed-on: http://gerrit.openafs.org/4116
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 22365ab37ac823b6734f4fc0ad79f10b6ee14390)

Change-Id: I7cf852d563b72c40b183b728143b59e8e4b8da7b
Reviewed-on: http://gerrit.openafs.org/5757
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

Give a default reason in *sync-debug

If no -reason is given for fssync-debug calls, we currently just
transmit garbage to the fileserver or salvageserver. Instead, give a
default (the *_WHATEVER constant), so we do something consistent.

Reviewed-on: http://gerrit.openafs.org/4115
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 7f31a11d110085ae528d4555d34744c0832008b0)

Change-Id: I27aefea0e891444e006e5d4b5ed43ed9cd492242
Reviewed-on: http://gerrit.openafs.org/5756
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

ConvertROtoRW: Use old copyDate for creationDate

When we convert an RO volume to an RW, currently we just copy the
copyDate and creationDate from the RO metadata into the RW. But the
copyDate and creationDate fields have different meanings for RW and RO
volumes: for ROs, the creationDate is merely the last time the data
was updated from the RW during a release operation.

So, if the copyDate is older than the creationDate, use the copyDate
as the new RW creationDate instead. This will probably not match the
creationDate of the original RW, but it will be closer to it, and it
will more accurately represent the conceptual "created time" of the
new RW.

Doing this can avoid forcing an unnecessary full dump on a subsequent
release of the resultant RW volume, since the creationDate is more
accurate.

Reviewed-on: http://gerrit.openafs.org/3891
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit f47487db051ba6b51ae12daf7edc09ecfd3c38cd)

Change-Id: I54516a20ebb87d9c482cbc4840e0d236e7c54f8a
Reviewed-on: http://gerrit.openafs.org/5755
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>

libafs: Set tvcp->callback before BulkStatus

When we call InlineBulkStatus or BulkStatus, we currently do not touch
tvcp->callback for any of the vcaches before making the call. This can
cause us to not notice an InitCallBackState issued by the fileserver
before the BulkStatus call returns, since the InitCallBackState
handler looks at tvcp->callback to determine what vcaches to clear
callbacks for. In turn, this can cause us to think we have a callback
agreement with the fileserver on one of the BulkStatus'd files, when
the fileserver does not actually have such a callback agreement.

So, set tvcp->callback to the server we are contacting, so if we get
an InitCallBackState call from that fileserver, the CBulkFetching
state will be cleared, and we will correctly discard the callback
information for that vcache.

Reviewed-on: http://gerrit.openafs.org/2548
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
(cherry picked from commit 37817796c4890683a7e41ed0f3a2fa6a53e1edc7)

Change-Id: I85c89eff061af799a7d8f612bee9b2f182312e6f
Reviewed-on: http://gerrit.openafs.org/5754
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>