Windows: afsio reduce library link list

When linking afsio many of the libraries specified in the EXELIB3
list were unnecessary.  The only lib in the list that is required
is libcmd.lib.  Rename EXELIB3 to LIBCMD and prune the rest.

Change-Id: I87bf0583f312d7b3dadc5581c08fedaa6b754c77
Reviewed-on: http://gerrit.openafs.org/9301
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Remove RXAFS* from afsrpc.dll

afsrpc.dll is supposed to be the multi-threaded dynamic loadable
version of the RX package.  It has also included the RXAFS and
RXAFSCB client functions from src/fsint.  Including the RXAFS* is
nice in theory but in practice the afsd_service.exe must link to
afsint.lib anyway in order to obtain access to RXAFSCB_ExecuteRequest()
which is part of the server portion.  Linking to both results in
multiple instances of the same symbols in which case afsint.lib
must be linked to first and the dynamic version of the functions
are ignored.

The only other module that used RXAFS* functions from afsrpc.dll
is the adminutil library from the libadmin package.  adminutil
doesn't implement the server functions but it can just as easily
link against afsint.lib.

Removing RXAFS* from afsrpc.dll results in a library that only
contains RX, RXSTATS, core xdr and rxkad.

This patchset also fixes some minor bugs:

 1. $(UTILOBJS) were not being linked to the library due to
    a typo.

 2. Always use multi-thread safe versions of afsutil.lib

Change-Id: I5cc6d485a763d5a17735abd35d193fb7b0d280e5
Reviewed-on: http://gerrit.openafs.org/9300
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Move grmutex, etc. to afspthread.dll

afsauthent.dll exported:

  pthread_recursive_mutex_lock
  pthread_recursive_mutex_unlock
  grmutex (DATA)

even though a dedicated afspthread.dll existed and all pthreaded
modules must link to afspthread.dll and not all link to
afsauthent.dll.  Move these functions and the global mutex variable
to afspthread.dll to ensure that only one instance of the variable
is present in any binary.

Also remove from src/libafsauthent/NTMakefile the local building
of src/util object files and link to mtafsutil.lib.

Change-Id: I9f8e76165f9c1b1be7ed1a90cdaac03b629d9591
Reviewed-on: http://gerrit.openafs.org/9303
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

ubik: Avoid unlinking garbage

If SDISK_SendFail fails before the pathname to the temporary file
has been constructed, then the failure handler will unlink stack
garbage, with variable results.

Initialise the string, and check to see if it has contents before
calling unlink

Change-Id: I1de898d5f0f0dc608e9c9c62dc66e6b4b9fdc5bf
Reviewed-on: http://gerrit.openafs.org/9299
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Unix CM: Don't free cell, then release lock on it

If afs_NewCell fails, then we can end up releasing a lock on a
section of memory that we have already freed. As this only happens
if the memory we're operating on is newly allocated and not yet
visible to anyone else, it is safe to release the lock before
starting to tidy things up.

Caught by coverity (#986054)

Change-Id: Ie8651c61790d57a9fd7bbbafcaf78e37b8222bae
Reviewed-on: http://gerrit.openafs.org/9298
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

libafscp: Can't unlock something we've freed

When we call _StatCleanup on a stored statent structure, it
deletes the mutex, and frees the structure itself. This means it
can't be called with a locked structure as the mutex deletion
will fail, and then we'll try to reference freed memory when we
later unlock that mutex.

Fix this by unlocking the mutex before calling _StatCleanup. This
is safe because the only reference to the structure visible to other
threads must have been deleted by the time we reach this point.

Caught by coverity (#986058, #986059)

Change-Id: I346d4c8a7cd478db044af919662c1cf1c093e205
Reviewed-on: http://gerrit.openafs.org/9297
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

pt_util: Protect against corrupt input files

If we have an input file which contains a group membership line
(with a leading space) before any group definitions occur, pt_util
would use stack garbage as the group to create these members in.

Avoid this by requiring the presence of a group line before any
membership lines.

Caught by coverity (#92180, #986000, #92248)

Change-Id: I536f089d67041dbf90f35b6dfeed5ba3303318a8
Reviewed-on: http://gerrit.openafs.org/9296
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

pt_util: Catch sscanf failures

If there isn't sufficient data in the input line to satisfy sscanf,
fail with an error, rather than continuing with potentially corrupt
data.

Change-Id: I5f963b5623176ac7631693b093d23cf9f0405edc
Reviewed-on: http://gerrit.openafs.org/9295
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

kauth: Fix overflow when writing ticket file

krb_write_ticket_file uses a fixed length buffer to store the name
of the ticket file, but copies into this from an environment variable.

Remove the fixed length buffer, and use a mixture of the variable
itself, and dynamically allocated strings.

Caught by coverity (#985909)

Change-Id: I997fd1812c45e730db39497a27a5e168f102fee5
Reviewed-on: http://gerrit.openafs.org/9294
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

auth: Refactor afsconf_Open failure handler

Refactor the code which returns a failure from afsconf_Open into
a single section, rather than having multiple copies through out
the routine.

Change-Id: I70b3fec98321bf89e6d59055545d5d4193026d28
Reviewed-on: http://gerrit.openafs.org/9293
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

auth: Fix buffer overflow in afsconf_Open

If we fallback to the .AFSCONF file in the user's homedirectory,
the results of getenv("HOME") are copied into a fixed length string,
without checking for overflows.

Instead of risking this, just use asprintf to dynamically construct
a string, and free it when we are done.

Caught by coverity (#985905)

Change-Id: Id8769ede841165d3ff3104143e55767d550d6f87
Reviewed-on: http://gerrit.openafs.org/9292
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bos_util: Fix buffer overflow

Get rid of a buffer overflow in the bos_util utility, by just
printing the key from the 'tbuffer' string, rather than copying
it into 'x' which is too small for it.

Change-Id: Ia364fb63edb9e40a887e77aad833689a99b4ea7c
Reviewed-on: http://gerrit.openafs.org/9291
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

volser: Fix bad readlink usage

readlink fills the buffer passed to it with a non-terminated string.
It can legitimately fill the whole of this buffer. So, if we require
a string to be NUL terminated, we must give readlink one less than
the string length so that the termination character can be safely
appended.

Caught by coverity (#985611)

Change-Id: Ie6c651a9c0fd94d83bd05395204341d71e353146
Reviewed-on: http://gerrit.openafs.org/9290
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

afsmonitor: Use NULL, rather than (struct foo *)0

Change-Id: Ib011da471465dae056d47f5bbaad5a27b5c85966
Reviewed-on: http://gerrit.openafs.org/9285
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

afsmonitor: Check correct variable is non-NULL

Before freeing curr_fsData, check that it is non-NULL, rather than
checking prev_fsData in error.

Caught by coverity (#985289)

Change-Id: I8fb74ce0c4e447de65924f2afb6075213d839760
Reviewed-on: http://gerrit.openafs.org/9284
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bosserver: Return error if key is bad

If a key is bad, have SBOZO_ListKeys return an error, rather than
silently ignoring the result from ka_KeyCheckSum.

Caught by coverity (#988413)

Change-Id: Ifacc9bac654d2632bd31731995795b786d485327
Reviewed-on: http://gerrit.openafs.org/9270
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

ptserver/testpt.c: remove dead code in ListUsedIds

A part of the current ListUsedIds code is:

    range = abs(startId - maxId);
    if (range < 0)
       range = -range;

The only way abs() could return a negative value would be if its
argument was INT_MIN (-2147483648) to begin with, because -INT_MIN
cannot be represented in two's complement.  However, calling
abs(INT_MIN) is undefined behavior in C [see C99 7.20.6.1], and for
that matter, so would be computing -range (-INT_MIN) in that case,
so we could still be left with a negative range value.

Luckily, (startId - maxId) can never be INT_MIN.  If startId < 0,
then maxId <= startId, so in the worst case, when maxId = INT_MIN
and startId = -1, (startId-maxId)=INT_MAX.  If startId >= 0, then
maxId >= startId, so in the worst case, when maxId = INT_MAX and
startId = 0, (startId-maxId)=-INT_MAX=INT_MIN+1.

This patch removes the useless if statement.

Change-Id: Ia754fcf3e59354afb40dbbbb95623e27285a5f82
Reviewed-on: http://gerrit.openafs.org/9289
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Simon Wilkinson <simonxwilkinson@gmail.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

volser: Catch EOFs when reading dumps

When reading a standard tag, make sure and catch an EOF error
from iod_getc. Otherwise we'll end up trying to parse 0xFF as a
tag length.

Caught by coverity (#985603)

Change-Id: Iae0b1bf71ed495b4f454975a1bfe3aef8fca3e9e
Reviewed-on: http://gerrit.openafs.org/9275
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

ptserver: Zero ubik header before writing to it

When using pt_util's ubik shim, if we're creating a new ubik label
make sure to zero the whole structure before writing it out to disk.
Otherwise we get a shorts worth of stack garbage in the resulting
file.

Caught by coverity (#986010)

Change-Id: I0abb8b7968d4f25c97dc0927d63dd01099b829b7
Reviewed-on: http://gerrit.openafs.org/9274
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

vos: Unrecognized or ambiguous switch '-c'

The documented abbreviation '-c' for '-cell' no longer works since
the -config option was added.  Add a parm alias '-c' to restore
compatibility with previous syntax.

Change-Id: Ie98e470d204ea2186a95668bba2581cf3e88f729
Reviewed-on: http://gerrit.openafs.org/9276
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

ptserver: Rename ubik.c as ptubik.c

Coverity gets confused between ubik/ubik.c and ptserver/ubik.c,
and produces a load of false positives. Rename the ptserver ubik
shim (which is only used by pt_util) in order to reduce this
confusion.

Change-Id: Ida606fccebcb4edfed706405424ff3ee4858020f
Reviewed-on: http://gerrit.openafs.org/9273
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bosserver: Catch failures to create daemon thread

If we can't create the bozo daemon thread, then don't keep going
regardless. Just warn the user and exit.

Caught by coverity (#988414)

Change-Id: I46445b1744f7c3c944e917b33a6fc6fb669df002
Reviewed-on: http://gerrit.openafs.org/9269
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

libadmin: Catch VL_GetNewVolumeId failures

If ubik_VL_GetVolumeId fails in the vos ProcessEntries routine,
report the failure and give up, rather than continuing with a
0 maxVolid

Caught by coverity (#988417)

Change-Id: I281e1c518813f9d7fab46e4f87a091fac6b3a754
Reviewed-on: http://gerrit.openafs.org/9268
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

afs: removing trailing semicolons

Remove spurious trailing semicolons from macros.

Change-Id: I26550a4ca8e8fe5232dad4cdb2c42a5f44a28f32
Reviewed-on: http://gerrit.openafs.org/9092
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

ubik: Zero header before writing to disk

The ubik disk header contains a padding field, which is never
zeroed before being written out to disk. This means that there is
a shorts worth of stack garbage in every ubik label!

Just zero the whole structure with memset before we populate it.

Caught by coverity (#986011)

Change-Id: Ib9f8b45111a83178372409c717c30ccb6a1044f4
Reviewed-on: http://gerrit.openafs.org/9272
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

viced: Don't write out garbage when extending file

When we extend a file, we write a single byte of data to it - tlen
is used as a source for this single byte. However, in the current
code, tlen is used uninitialised. Set it to 0, so we don't write a
byte of stack garbage.

Caught by coverity (#986015)

Change-Id: I11480a5ba0a5544437e18a6b314ff98ba6256df0
Reviewed-on: http://gerrit.openafs.org/9271
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

fs: Catch pioctl failure in mkmount

If the VIOC_FILE_CELL_NAME pioctl fails in fs mkmount, return an
error to the caller, instead of ploughing on with potentially bad
data.

Caught by coverity (#988418)

Change-Id: I68df5dd26d955db860dd016c76109f7d262bad17
Reviewed-on: http://gerrit.openafs.org/9267
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

libafscp: NULL return value not pointer

When afscp_FindCallBack sets its return-by-reference value to NULL,
it should do so to the value itself, not the pointer to it.

Caught by coverity (#988419)

Change-Id: I6b227465d7aa945d871b2db1cc4d1357cae493d3
Reviewed-on: http://gerrit.openafs.org/9266
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

Add krb5_enomem for config parser & kernel crypto

The imported code form Heimdal that forms our config file parser
and kernel rfc3961 library now makes use of krb5_enomem. Provide
an implementation so we build again ...

Change-Id: Ie7162d73aa97e6f7594c0e7c68ed6939f57ae884
Reviewed-on: http://gerrit.openafs.org/9265
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

Import of code from heimdal

This commit updates the code imported from heimdal to
66f4c441e9e0de68fbcf81763642779ac5c33631 (git2svn-syncpoint-master-199-g66f4c44)

Upstream changes are:

Kumar Thangavelu (1):
      unning "kinit --fast-armor-cache=xxx" against a Win2K3 domain resulted in a crash with the attached backtrace. FAST is not supported with RC4 keys which are used in Win2K3. The code already handles this but the error code is not propagated up the stack.

Love Hornquist Astrand (1):
      add socket_set_nonblocking

Roland C. Dowdeswell (3):
      Optimise _krb5_n_fold() a bit.
      Use krb5_enomem() more consistently in lib/krb5.
      Further improvements to lib/krb5/n-fold.c:

Change-Id: I3b652e3ea3c84a939c4988bc8c24b3edee06446f
Reviewed-on: http://gerrit.openafs.org/9264
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: Derrick Brashear <shadow@your-file-system.com>

crypto: _afscrypto_strdup must copy trailing NUL

The _afscrypto_strdup helper function was failing to alloc enough
space for the string's trailing NUL, and to copy that NUL over.

Caught by coverity (#985580)

Change-Id: I61c9a96275a79682a9521846ab52615ea5bc061d
Reviewed-on: http://gerrit.openafs.org/9263
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

volser: Make it clear what's copied from partList

XVolListPartitions uses partList to maintain a list of partition
IDs - it does not use the partFlags field. Make it clear when we
populate the partEntries field that we are only copying the
partition list.

Caught by coverity (#986030)

Change-Id: I8c46ee5c220f0cb2b65943ad5faa80877ca46685
Reviewed-on: http://gerrit.openafs.org/9262
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

bucoord: Remove dead assignments in cmd init

Don't assign command syntax objects to 'ts' if we're not going
to add any parameters to them

Change-Id: I69f6e090fc115ab213c1f20dc46da81f2878e0a3
Reviewed-on: http://gerrit.openafs.org/9260
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

butc: Remove dead initialisation

The 'bytesread' variable is set by all code paths, so don't bother
assigning a default value each time we pass through the loop

Caught by clang-analyzer

Change-Id: I8b79bc888ba23e2bcfa9befb5bf034db8f3ca3f8
Reviewed-on: http://gerrit.openafs.org/9259
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

volser: Don't catch errors in cleanup

We can't do anything if cleaning up the clone volume fails, so
don't bother storing the error code.

Caught by clang-analyzer

Change-Id: I5249dbad9a1bb732aba73e7a09f74c83e606b45d
Reviewed-on: http://gerrit.openafs.org/9258
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bucoord: Remove dead initialisation

matchLength is set regardless of what the separator is, so we don't need
to provide a default value.

Caught by clang-analyzer

Change-Id: Iac8fe8db8832cf2b3347ec1bdacc7245bddf542b
Reviewed-on: http://gerrit.openafs.org/9257
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

butc: Remove dead initialisation

code is set to a real value on every code path, so don't bother
initialising it to 0 midway through the function.

Change-Id: I334286f2ae854c57df781af8df78e5a1bf8e9b2b
Reviewed-on: http://gerrit.openafs.org/9256
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

volser: Remove redundant success flag

In UV_RestoreVolume2, there is a success flag whose only use is ...

   success = 0
   [ ... ]
   success = 1
   if (success) {
      [ ... ]
   }

It's pointless, so just remove it, and one clang warning

Caught by clang-analyzer

Change-Id: I7d57531c4dfb8955b813d534ad3b72b182fdbc76
Reviewed-on: http://gerrit.openafs.org/9255
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

budb: Don't double initialise iterator

The blockaddr iterator is initialised as part of the for loop.
Don't also initialise it immediately before.

Caught by clang-analyzer

Change-Id: I97e7ced15fe204b3b2d797abbdce8b3f9a2b0c8d
Reviewed-on: http://gerrit.openafs.org/9253
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

budb: Store dumper information correctly

Change f498c18a6f7300eb89ca3b9d9b85c4d461fa8474 attempted to fix
a problem with using an inappropriate function to copy the
ktc_principal value 'principal' into the d.dumper structure element.

Originally the code read:
    principal_hton(&principal, &d.dumper)

which copies the 'host' principal principal into the 'net' structure
d.dumper. However, this was changed to
    principal = d.dumper

Which populates 'principal' with whatever garbage is in the d.dumper
structure.

Fix this so that the assignment is done the right way around.

Caught by clang-analyzer

Change-Id: Id9865a36b5a9351d013c695ed301b627ddc4a915
Reviewed-on: http://gerrit.openafs.org/9252
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

budb: Simplify scanHashTableBlock

We don't need to maintain two copies (entryAddr and nextEntryAddr)
of the entry address - rework the while loop so we just use one,
and make clang a little happier

Caught by clang-analyzer

Change-Id: I6f5d76e114ad09a4fd3fc36b4d64f63daa9e1e78
Reviewed-on: http://gerrit.openafs.org/9251
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

viced: Unused sentinel in h_stateVerifyUuuidHash

The 'found' sentinel in h_stateVerifyUuidHash is unused, because
as soon as we set it, we jump out to the 'done' label. Simplify the
function's logic, and remove a static analysis warning, by just
removing the variable.

Caught by clang-analyzer

Change-Id: I02e5ccf857f5d84cd24975b7a1b9d516624ecd64
Reviewed-on: http://gerrit.openafs.org/9250
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

bucoord: Remove dead initialisation

offset is simply used as a copy of nextOffset for the GetText
function call. We don't need to initialise it to 0 at the start of
the loop (and arguably, as it is passed to the function by value,
we could just remove the copy entirely)

Caught by clang-analyzer

Change-Id: I137ea3ba1ff44d94b90287d9735db69d6ff8717d
Reviewed-on: http://gerrit.openafs.org/9249
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

butc: Remove dead assignment

We never use the value of mvFlag once the string has been restored,
so don't bother reseting it.

Caught by clang-analyzer

Change-Id: I12db6957c9b8ec3bb9c43f10d9e9754d4afa3ed7
Reviewed-on: http://gerrit.openafs.org/9248
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

budb: Fail if afsconf_GetExtendedCellInfo does

Rather than silently ignoring the failure of
afsconf_GetExtendedCellInfo, and then using garbage cell
configuration, just fail to start if afsconf can't parse the
config directory.

Change-Id: Ib3d26bc67844de10f7eb8c2ef555c5e607e4bf61
Reviewed-on: http://gerrit.openafs.org/9247
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

volser: Remove dead assignment

Don't initialise code immediately before we assign a real value
to it various UV_ functions

Caught by clang-analyzer

Change-Id: I91d9300715d44fafd73b1a61be548589468e3209
Reviewed-on: http://gerrit.openafs.org/9246
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

butc: Catch failures to start keep alive thread

If we can't start the keep alive thread, in either the LWP or
pthread versions of a code, report an error and exit instead of
carrying on regardless.

Caught by clang-analyzer

Change-Id: I0a25a236aadeedf10d3c0c129d716b1397c9a4e3
Reviewed-on: http://gerrit.openafs.org/9245
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

butc: Remove dead loop pointer allocations

All of these *Ptr values are reset upon each iteration of the
while loop, so we don't need to NULL them immediately before
we go around again.

Caught by clang-analyzer

Change-Id: Iaa5ff6c7b154072e23eae272031919b424568eff
Reviewed-on: http://gerrit.openafs.org/9244
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>

Windows: Test NameArrayReferenceCount before deletion

In commit c753d398f2005c886932f20b49af15efc489a829 the
AFSDirectoryCB.DirOpenReferenceCount was split and the
NameArrayReferenceCount field was added.  However, it failed to modify all
of the DirOpenReferenceCount checks prior to calling AFSDeleteDirEntry()
to include a check on the NameArrayReferenceCount.  This patchset corrects
that oversight.

Change-Id: I8db1a68e5240088bcfeb226bd1eafac0c83a0b69
Reviewed-on: http://gerrit.openafs.org/9254
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: Jeffrey Altman <jaltman@your-file-system.com>

kauth: realm can't be NULL when calling CellToRealm

Despite what the static analyzer thinks, ka_CellToRealm can't be
called with a NULL realm. Just add an assert to close down that
code path, and stop it from producing an error.

Change-Id: Ifc8cf7e5dca5ae29eaaa91c5b238a97197d89ee3
Reviewed-on: http://gerrit.openafs.org/9174
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: no side effects from AFSDbgLogMsg calls

Do not perform assignments or call side-effect resulting functions
as parameters to AFSDbgLogMsg.

Change-Id: I76eca3a82377543f4148d03af51a07886127b385
Reviewed-on: http://gerrit.openafs.org/9242
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>
Tested-by: Jeffrey Altman <jaltman@your-file-system.com>

viced: Allow GetCapabilities to return an error

At the moment, the GetCapabilities RPC handler ignors errors from
CallPreamble or CallPostamble. Instead of silently swallowing these,
and potentially reutnring a bogus capabilities structure, make
it possible for the RPC to return failure to the caller.

Caught by clang-analyzer

Change-Id: Ic36e59e8cce8be4920f041a86d5636d54668d887
Reviewed-on: http://gerrit.openafs.org/9241
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

viced: Catch GetCellInfo failure

If GetCellInfo returns a non-zero result, catch the error, rather
than just carrying on regardless.

Caught by clang-analyzer

Change-Id: I6b9ca91c8cea2ff0d68873913abdc0daadcc0e27
Reviewed-on: http://gerrit.openafs.org/9240
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

viced: Don't store rc from IH_DEC on error path

If CopyOnWrite has already failed, and is taking the volume offline
for salvage we can't (and don't) do anything with further errors from
IH_DEC. Instead of assigning these to 'rc', just ignore them.

Caught by clang-analyzer

Change-Id: I7bcd82f124740e3ca3d11cb2b079aa7ea5751cf9
Reviewed-on: http://gerrit.openafs.org/9239
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

scout: Fix a lot of dead assignments

In lots of places scout takes the return value of a graphics
function, then decides it doesn't actually care about the result.

Get rid of the assignments, as they provide a false sense of
security.

Caught by clang-analyzer

Change-Id: I1ea8514b969777f1276b4e8f8ab946a4d36a870c
Reviewed-on: http://gerrit.openafs.org/9238
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

gtx: Return failure if gtx_Init can't start thread

If LWP_CreateProcess fails in gtx_Init return NULL, rather than
silently swallowing the failure.

Caught by clang-analyzer

Change-Id: I5db5667f9dae3f405e88cf9c57e6b8a069ea28a3
Reviewed-on: http://gerrit.openafs.org/9236
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

afsmonitor: Allow CBSTATS collection to work

The switch which selects the collection number was missing a
'break', so selecting the CBSTATS collection would always fall
through to the default, error, case.

Caught by clang-analyzer

Change-Id: I354873d711c5f49ddad51f1b9fb332af91a6d13a
Reviewed-on: http://gerrit.openafs.org/9235
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bosserver: Don't store result when renaming core file

The bosserver doesn't do anything if the core file rename fails, so
don't bother storing the result code.

Caught by clang-analyzer

Change-Id: I9b1b7d77404b0fd560c7735646e5afe1fbcc0752
Reviewed-on: http://gerrit.openafs.org/9232
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bosserver: Remove dead assignment

In the Restart RPC handler, don't assign a value to code
immediately before we use it to hold the return value of
bozo_WaitStatus.

Caught by clang-analyzer

Change-Id: I0781c57c87088cb3ba6170ca1e5c8d5053cf9001
Reviewed-on: http://gerrit.openafs.org/9231
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

bosserver: Remove dead assignment in SaveOldFiles

We don't care about whether the renames succeed in SaveOldFiles,
so don't bother assigning their results to a variable

Caught by clang-analyzer

Change-Id: I0605fc71d243ef0f587706f9f3856aa4d722fb82
Reviewed-on: http://gerrit.openafs.org/9207
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Unix CM: Remove dead assignment in readdir

len is zeroed at the start of the next block, so don't bother assigning
tlen to it if we're not going to do anything with it.

Caught by clang-analyzer

Change-Id: Ie1cd5a4d9dd96920018f054b6edfc18c569c9807
Reviewed-on: http://gerrit.openafs.org/9206
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

libadmin: Fix a lot of dead assignments in vsprocs

Tidy up a lot of places where we initialise a variable, then
immediately assign a proper value to it, or store a return value
that we don't actually care about.

Caught by clang-analyzer

Change-Id: Icf0ec713dac6f2c4f0a60e8ead52a8d3e19381f0
Reviewed-on: http://gerrit.openafs.org/9205
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

gtx: Remove pointless assignments from gtxtest

Remove assignments to 'code' when we're not going to pay attention
to them.

Caught by clang-analyzer

Change-Id: Ib2f8c20c0bc741d0bb1b8ed4fdb46ea61511fc95
Reviewed-on: http://gerrit.openafs.org/9204
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Unix CM: Catch ICL initialisation failures

If afs_icl_CreateLog fails, don't still call afS_icl_CreateSet.
If afs_icl_CreateSetWithFlags fails, don't ignore its error code.

Caught by clang-analyzer

Change-Id: I245f4391bbbf183492e8c13fb8f091b339cea80a
Reviewed-on: http://gerrit.openafs.org/9203
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

rx: Fix wrong loop iterator in rx_clearProcessRPCStats

Fix the use of an incorrect variable as loop iterator which would have
resulted in a crash.

Caught by clang analyzer.

Change-Id: I785d06b9be0915bc540c37b43d1001a562a7cde4
Reviewed-on: http://gerrit.openafs.org/9202
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

volser: Fix a couple of assigned but not read errors

Fix a couple more errors where we assign to a variable, but never
read the result back out.

Caught by clang-analyzer

Change-Id: I78f8f1403d166c4231b0abdb607de77e9ed62a27
Reviewed-on: http://gerrit.openafs.org/9201
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

volser: Don't assign code when we don't care

When we're cleaning up temporary volumes, we don't care whether
it succeeds or not. Don't assign code to the results of these
volume deletions, only to then discard it.

Caught by clang-analyzer

Change-Id: I81583cc82e78db38de8aad92c7c7d0c2f44eb33d
Reviewed-on: http://gerrit.openafs.org/9200
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

volser: Don't NULL conn right before assigning it

Don't do
   conn = NULL;
   conn = rx_newConnection();

and similar, equally pointless things.

Caught by clang-analyzer

Change-Id: I54d9ccd9f3910624b3289ce067a27622e29e4a34
Reviewed-on: http://gerrit.openafs.org/9199
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

fs: Remove redundant variable assignments

Remove a number of redundant variable assignments that clang warns
about.

Caught by clang-analyzer

Change-Id: I248e0e3a9f9241084436604532424d6e4b900e31
Reviewed-on: http://gerrit.openafs.org/9194
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

fs: Free parent_dir later in lsmount and flushmount

If lsmount or flushmount encounter an error, then they may include
the contents of parent_dir in their error message. However, in both
cases, this was freed a couple of lines earlier.

Just move the free() later, so that the contents of this variable
are still available.

Caught by clang-analyzer

Change-Id: Ideba52b7ba98f034baf923194007239a6cf0221f
Reviewed-on: http://gerrit.openafs.org/9193
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

up: Notice errors on both file handles

Report errors on both filehandles in up's Copy routine, rather
than just on the second one that we close.

Caught by clang-analyzer

Change-Id: Ia3aa9be91ccfca2993a4f27c291332e0b0d350f5
Reviewed-on: http://gerrit.openafs.org/9192
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

fstrace: Remove redundant variable assignment

Don't reset the value of 'code' when nothing else references it.

Caught by clang-analyzer

Change-Id: Ifd361ed6eb8c6b8740466f6c1a337669160874f4
Reviewed-on: http://gerrit.openafs.org/9191
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

fstrace: Avoid accessing icl log after zapping it

The for loop in icl_EnumerateLogs looks up the next pointer in the
current entry after zapping it. Depending on reference counts, this
may result in us looking up freed memory.

Take a copy of the next point before zapping the current entry, just
in case.

Caught by clang-analyzer

Change-Id: If38f0af2b01c5b8ea00e68e4432c6ad5517578c8
Reviewed-on: http://gerrit.openafs.org/9190
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

xstat: Ignore return value from xstat_*_Cleanup

We're about to exit anyway, so we don't care whether Cleanup
succeeds or not

Caught by clang-analyzer

Change-Id: I6350d60e6a5d29b8aaa272a54cf9ead47fdb5fad
Reviewed-on: http://gerrit.openafs.org/9186
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

afsmonitor: Ignore return codes as appropriate

When we're not interested in the return value of a function, don't
assign it to a variable.

Caught by clang-analyzer

Change-Id: I5b576f312376780c98c2dbe4f9b85e00d529ddd8
Reviewed-on: http://gerrit.openafs.org/9185
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

sys: Flag rmt_Quit as AFS_NORETURN

Help out the static analyzer by telling it that rmt_Quit never returns

Change-Id: I5ee0d2ea444706e68af5a1298fd638e7d0d7f41d
Reviewed-on: http://gerrit.openafs.org/9184
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

rxkad: Remove redundant checkCksum initialisation

checkCksum is assigned to by both sides of the if() expression, so
there's no need to initialise it to 0.

Caught by clang-analyzer

Change-Id: Id0e286943e11b6948231577085cdb0ce96bb0b97
Reviewed-on: http://gerrit.openafs.org/9183
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

ubik: Don't leak memory if mutex init fails

If we can't initialise the ubik_client mutex, don't leak the
structure when returning an error.

Caught by clang-analyzer

Change-Id: Ib36d2ddf585a31b628e56dce1f33ee859fc835bc
Reviewed-on: http://gerrit.openafs.org/9182
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

ubik: Don't record result of select()

We never do anything with the return value from select(), so don't
bother storing it into code.

Caught by clang-analyzer

Change-Id: I6e0352b74f0e90b62ad32a3ef1a9d0173bf307ee
Reviewed-on: http://gerrit.openafs.org/9181
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

kauth: Don't ignore GetExtendedCellInfo failures

Don't ignore a failure from afsconf_GetExtendedCellInfo, as it will
leave buffers that we later read uninitialised.

Caught by clang-analyzer

Change-Id: Ia0c2675a3083385d011fd340f36bf13373f89feb
Reviewed-on: http://gerrit.openafs.org/9180
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

kauth: Catch ka_KeyCheckSum failures

If ka_KeyCheckSum fails, typically because the key being used is bad,
catch the failure rather than just silently ignoring it.

Caught by clang-analyzer

Change-Id: Id51174fdf94356db10961690808670da372bc60e
Reviewed-on: http://gerrit.openafs.org/9179
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

kauth: Fix some unnecessary variable assignments

Remove various places where a variable is assigned to, and then the
result never read.

Caught by clang-analyzer

Change-Id: Id8b7aeb6c54b7c93093ad44df5a9fe01575f2fbe
Reviewed-on: http://gerrit.openafs.org/9172
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Unix CM: Fix byte accounting for storebehind

In the current version of CacheStoreDCaches, the stored variable is
maintained within the for loop that iterates over the chunk list. This
means that it is reset to 0 each time we handle a new chunk.

However, this means that our progress is no longer accurately tracked,
as (bytes - stored) no longer gives the number of bytes which remain to
be transfered. In fact, as stored is zeroed with each loop iteration,
(bytes - stored) == bytes. This means that store behind is no longer
activated according to the users settings.

Prior to commit 334114ac58b0039ae90d7e29fa2f019fe068bd79, the
stored variable was maintained within the outer, function, scope.
Just move it back there to restore the previous behaviour.

Caught by clang-analyzer

Change-Id: I248ed81f1331f4d3d4ed2e1206acebb0a8200762
Reviewed-on: http://gerrit.openafs.org/9237
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

Unix CM: Don't zero args on dcache failure

Even if allocating the dcache fails, there's no point zeroing the
parameters to the allocation function, as those changes aren't
visible outside of that function.

Caught by clang-analyzer

Change-Id: Ieb91b525e818b73e38b8d78a55d170d36273177c
Reviewed-on: http://gerrit.openafs.org/9234
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

Unix CM: Fix bogus character in license string

A stray $ had crept into the license string in afs_dcache.c. Fix it.

Change-Id: I1ab096aadb947cb08318a56b9ae21ed0786d3339
Reviewed-on: http://gerrit.openafs.org/9233
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

bos: Don't set up the conenction twice in AddKey

AddKey calls GetConn twice - once at the start of the function,
and then again immediately before calling BOZO_AddKey. This leaks
a connection...

Caught by clang-analyzer

Change-Id: I5c593de56e7f61fb21b973a34bea1a5b4f4a3177
Reviewed-on: http://gerrit.openafs.org/9230
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

bosserver: Don't check if daemon thread starts

We don't do anything with the return code from the LWP_CreateProcess
that starts the daemon thread, so don't bother recording it at all.

Caught by clang-analyzer

Change-Id: I9310bdd90236a90d71dc202c73d7e06ea58a31f8
Reviewed-on: http://gerrit.openafs.org/9229
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

vlserver: Remove dead assignment

Don't assign the serverNumber to j, when we're just about to try
and get it from a multi homed extent record

Caught by clang-analyzer.

Change-Id: I79349ff211b5df2d16e6ee9122dbb28b04ad7823
Reviewed-on: http://gerrit.openafs.org/9228
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Chas Williams - CONTRACTOR <chas@cmf.nrl.navy.mil>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

bosserver: Remove a couple of dead assignments

Remove a couple of places where we initialise a variable, only to
assign a new value to it a few lines later.

Caught by clang-analyzer

Change-Id: I096b80cce851f3494475e6f9683715b081ff8375
Reviewed-on: http://gerrit.openafs.org/9208
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

afsio: Remove redundant assignment

tbuf is assigned to during the initialisation phase of all of the
for loops that follow, so just remove this assignment

Caught by clang-analyzer

Change-Id: I3f2ffc8cee93768bd03e0abf3b391e1f6c45c70f
Reviewed-on: http://gerrit.openafs.org/9189
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@your-file-system.com>

Windows: ReleaseNotes reformatting

Oxygen 14.2 automatically reformatted the text.

Change-Id: I387627948f449c96ed3ab112371752f4673b24d9
Reviewed-on: http://gerrit.openafs.org/9212
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Direct IO Support for Service

This patchset implements and enables by default the new
Direct IO pathway between the AFS redirector and the afsd_service.exe.
When Direct IO is enabled all reads and writes are performed by the
AFS redirector locking memory allocated by the kernel and mapping it into
the service's memory address space.

The service supports cache bypass in this mode when the
AFS_REQUEST_FLAG_CACHE_BYPASS flag is set in the request from the
redirector.  When cache bypass is active, the AFSCache file is ignored and
data is either directly fetched from or stored to the file server.  Cache
bypass is enabled by IIS and other applications that request no
intermediate buffering when opening file handles.   This is often done
because the application implements its own data caching.  All cache bypass
store operations are synchronous.

When cache bypass is not enabled, the memory region provided by the AFS
redirector is either used to populate the cm_buf_t objects or is populated
by them.  When cache bypass is not enabled, one outstanding store
operation can be in flight asynchronously to improve performance.

Direct IO is enabled by default and can be disabled by creating the
registry value.

  ..\Services\TransarcAFSDaemon\Parameters
  "DirectIO"  DWORD  0x0

Change-Id: I3cac3660c8b8eded58226ba4a819692c454704a8
Reviewed-on: http://gerrit.openafs.org/9211
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Direct IO for AFS Redirector

Implement a new IO processing model in which extents are not passed between
afsredirlib.sys and afsd_service.exe.  Instead the AFSCache file is
maintained exclusively by the service and the redirector locks kernel
memory, maps it into the service's address space, and permits the service
to manage all IO directly.

This interface adds an AFS Cache Bypass option to the AFS Redirector which
is activated when the file handle has been opened with the no intermediate
buffering option.

This patchset implements the kernel interface.  A subsequent
patchset will implement the service component.

Assisted by Jeffrey Altman <jaltman@your-file-system.com>

Change-Id: I25a4764db060b3b3f2b0de4006479dd3a220c6eb
Reviewed-on: http://gerrit.openafs.org/9210
Reviewed-by: Peter Scott <pscott@kerneldrivers.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Add flag BUF_GET_FLAG_BUFCREATE_LOCKED

Modify the buf_Get() and buf_GetNewLocked() interfaces to
permit the cm_scache.bufCreateLocked lock to be held prior to
calling to buf_Get().  Holding the cm_scache.bufCreateLocked lock
before the buf_Get() call prevents a race with another thread
that attempts to set the file size.

Change-Id: I54b13c83801a4d9a2078552d5be7ada6f4607850
Reviewed-on: http://gerrit.openafs.org/9209
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Set AFS_OBJECT_FLAGS_DIRECTORY_ENUMERATED

Set the ObjectInformationCB AFS_OBJECT_FLAGS_DIRECTORY_ENUMERATED flag
from within the AFSEnumerateDirectory() function and not from its callers.

Change-Id: I2bddedba1165557679ddd9c637dbeb85cd74e1de
Reviewed-on: http://gerrit.openafs.org/9119
Reviewed-by: Peter Scott <pscott@kerneldrivers.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Permit direct to service non-wildcard lookups

The AFS redirector has required that directories be fully enumerated
when the directory object is opened.  This is a very expensive
operation involving large numbers of file server RPC round trips for
directories with tens of thousands of objects and those containing
symlinks.

This patchset delays directory enumeration for the last component
in a path until such time as dirctory data is requested by the
application.  If the request is for a non-wildcard pattern, the
service will be asked to provide the details for just the one required
object.

Delaying the directory enumeration improves performance for
GetFileAttributes[Ex], GetDiskFreeSpace[Ex], GetVolumeInformation,
and GetNamedSecurityInfo Win32 API calls.   In those cases it is
no longer necessary to enumerate the target directory at all.

Change-Id: I7ef2fbafff925697d8b40e56837ef53bfcc78542
Reviewed-on: http://gerrit.openafs.org/9118
Reviewed-by: Peter Scott <pscott@kerneldrivers.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: EvalByName pass LastComponent flag

Add AFS_REQUEST_FLAG_LAST_COMPONENT flag for use with
AFS_REQUEST_TYPE_EVAL_TARGET_BY_NAME requests to the service.
When set the service will perform cm_Lookup calls without the
CM_FLAG_CHECKPATH flag set.

Change-Id: I47ec2fb8b1e2699f2d87a6625b1db549ecb4e03d
Reviewed-on: http://gerrit.openafs.org/9133
Reviewed-by: Peter Scott <pscott@kerneldrivers.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: EvaluateByName support case-insensitive lookups

Directory lookups in AFS should favor case-sensivite matches
but permit case-insensitive matches otherwise.  The service
should not follow mount points.  The redirector exposes mount
points as junctions.

Change-Id: I34c9244e2756a926f29bbf184dc250e7a29d264f
Reviewed-on: http://gerrit.openafs.org/9132
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: IO Subsystem Trace modifications

Display file offsets consistently as %0I64X.

Change-Id: Ib1484ef99fbf055ec4bc00e17469e0daa9661e96
Reviewed-on: http://gerrit.openafs.org/9129
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Rod Widdowson <rdw@steadingsoftware.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>

Windows: Move work item queues over to the Control Device

Currently, when the library is unloaded it stops all worker
threands and then evaporates the work item queues.  Thus
any work items which are pending will disappear.

Whilst it is OK that the threads going away, any work items need to
remain queued so that when the library is restarted the work
can continue.  This checkin does this by moving the work item
queues and their synchronization primitives into the FS maintained
Control Device Object Extension.  The list of worker threads
remains in the Library Device Object Extension.

Change-Id: If5c7cd3bdfea1a368c8df69649e627bac3a9585f
Reviewed-on: http://gerrit.openafs.org/9139
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Peter Scott <pscott@kerneldrivers.com>
Reviewed-by: Jeffrey Altman <jaltman@your-file-system.com>