STABLE12-linux-try-more-kernel-paths-20030410

FIXES 1364

try /usr/src/linux-2.4 in addition to /usr/src/linux

(cherry picked from commit 5946081876332bfb53eecf48448b42412ddf9925)

STABLE12-darwin-map-more-versions-20030410

Recognize MacOS X 10.2.5 as ppc_darwin_60.

(cherry picked from commit abc9f6b19029fc76093d8c2d1326ad1b36fa1b00)

STABLE12-rxkad-2b-deal-with-expiry-20030410

FIXES 1369

return the actual error and not RXKADBADTICKET always

(cherry picked from commit 429cc8b6467726795bf46bdf1aa612400aa7afdf)

STABLE12-dux51-basic-build-support-20030409

so servers can be built from 1.2.9-release

STABLE12-h-gethost-r-race-20030401

FIXES 1308

avoid deref of identP when code is -1

(cherry picked from commit f40a6356f7917b164c4b949bc1916dbcb83a2f38)

STABLE12-salvager-aix-update-20030407

make it work with namei

(cherry picked from commit 3ab37f714824d90701caa9123fc7c820aa411bcb)

STABLE12-flushhostcps-20030407

make it so flushhostcps can't try to Release when no host was found,
which was apparently the problem dlc@cs really observed.

(cherry picked from commit dd517125c21e010954989d60caa31bc2da7ca394)

STABLE12-volser-dump-vnode-include-volid-in-error-20030213

include a volume number in the error message

(cherry picked from commit 7c8bd23f40a18027b9229376bf9dacdbc9a2879b)

STABLE12-h-gethost-r-race-20030401

FIXES 1308

Thanks to Chaskiel Grundman for explaining what was happening:
- the connection is old and pre-existing, but has no host structure.
- 2 calls come in
- the first one enters h_GetHost_r, and h_Lookup_r returns null (but identP is
non-null, since rx keeps it around until it gc's the connection)
  The first thread calls WhoAreYou, which succeeds, it then calls
InitCallBackState3 (after H_UNLOCK) note that the host has been inserted into
the hashtable
- the second thread enters h_GetHost_r, and calls rx_GetSpecific. it then
calls h_Lookup_r. h_Lookup_r will block (new host is locked), but eventually
returns the new host
- InitCallBackState3 returns, and the frees the old identP, replaces it, and
unlocks the host.
- the first thread returns from h_Lookup_r. boom.

the changes:
-call rx_GetSpecific after h_Lookup_r returns (and potentially slept)
-removes an if wrapping which always is true (since !interfValid is always
 true)
-don't realloc identP if it exists
-don't free an old one by calling rx_SetSpecific either

(cherry picked from commit 5967b22698c8aeb51131a62c56a2f7fbf1f8e79e)

STABLE12-darwin-map-more-version-20030331

Recognize MacOS X 10.2.3 and 10.2.4.

(cherry picked from commit f14686ea2e4d10381554bd466d236dd9b9479e4a)

STABLE12-ubik-allow-netinfo-netrestrict-to-mask-address-pointed-to-by-hostname-20020417

if you really want to mask the IP address from gethostbyname(gethostname())
there's not really any reason to stand in your way

(cherry picked from commit 7158ef3ae57a97c4b7bb03527cb162ba968a2b35)

STABLE12-solaris-fs-rolled-20030328

for solaris 9 u2

(cherry picked from commit b9eb8cc6695a0dc903b10cd763041f5cfd6a1764)

STABLE12-flushhostcps-init-held-20030328

init held so flushhostcps is more useful.

(cherry picked from commit 665e86f6fc4fbca43fd0d827d25f3de5f074e082)

STABLE12-viced-allow-hinting-of-all-new-clients-for-vbusy-vrestarting-handling-20030215

For sites with only new (>3.4) clients (or who decide to not care
about older clients) and want to force VRESTARTING instead of VBUSY.

(cherry picked from commit 214c2e7b19333380cb7f3aa5c997f35543bd3d15)

STABLE12-shlibafsrpc-export-missing-des-symbols-20030326

symbols libafsauthent wants but we don't have

(cherry picked from commit cd159fddccbf19448e73a3a468268590f26ba59c)

STABLE12-linux-rxk-listener-shutdown-cleanup-20030323

avoid need for sys_kill

====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================

get rid of need for sys_kill

====================

get rid of need for sys_kill

(cherry picked from commit 937ec63685f69f55eb0a6ce39ab94dc475eb5fe6)

STABLE12-rxkad2b-allow-no-kvno-20030325

allow a missing kvno field to work usefully

(cherry picked from commit 45dd8988f28ff573200136edda16d5592a757bff)

STABLE12-move-up-cell-initialization-in-cachemgr-20030323

to potentially avoid an oops at startup

(cherry picked from commit 3fa5f389b2b7778cf0df5a506c91b427b147c4c2)

STABLE12-linux-kthreads-reparent-20030323

reparent kernel threads to avoid zombies at shutdown

(cherry picked from commit d14d08e35b95b2482bb074fe174db1e2b55da49d)

STABLE12-xdrmem-getbytes-20030321

See, for example:
http://web.mit.edu/kerberos/www/advisories/MITKRB5-SA-2003-003-xdr.txt
This source is unused but it seems like a good idea to patch it anyway.

(cherry picked from commit 7f4d43ac2f6d7b44a5f07be6e59d45bf9c6736dc)

STABLE12-linux-hold-tasklist-lock-in-get-page-offset-20030321

tasklist_lock must be held to traverse list

(cherry picked from commit 38cec12f2c45d4d2f544acedeeffafec70a89582)

STABLE12-linux-one-more-stupid-configure-test-20030321

pointed out by zacheiss@mit.edu that i missed this. oops

(cherry picked from commit 4731e86a441df09b893a8bf3973a51a404f69d9a)

STABLE12-redhat-81-support-again-20030319

once more. maybe this time they'll release something that looks like this

(cherry picked from commit 284d3edd2375f672057aef638c0e0afd41cc559d)

STABLE12-kaserver-disable-krb4-crossrealm-20030317

to deal with krb4 security advisory

====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================

to deal with krb4 security advisory
type mismatch for krb4_cross pointed out by Jack Neely <slack@quackmaster.net>

(cherry picked from commit 1312d1179d4a97e9c62eb26ef55f60007cd35872)

STABLE12-sunx86-update-20030318

make client useful on solaris x86 again

STABLE12-viced-vbusy-when-getclient-fails-20030317

more paranoia to not crash if GetClient fails, since there's apparently a
race we haven't found yet

(cherry picked from commit 4f35fbb68005579f626d021632e899301f551c66)

STABLE12-winnt-no-softsig-20030314

FIXES 1333

the softsig stuff messed up the windows signal stuff. MSVC doesn't
recognize #elsif. if #elif isn't standard, then it needs to be split into
#else and another #if/#endif block.

STABLE12-winnt-rxkad5-20030314

FIXES 1333

stddef.h needed for NULL

STABLE12-winnt-afsrpc-assert-20030314

FIXES 1333

something in the fileserver started using osi_Assert, instead
of whatever assert comes from util, so we need osi_AssertFailU in addition
to AssertionFailed.

STABLE12-winnt-client-shortcut-20030314

FIXES 1333

leaving the #include <initguid.h> in caused Bad Things to
happen when I did a build. I'm pretty sure initguid.h is only supposed to
be used by things that *impliment* COM objects.

STABLE12-winnt-afsd-initialize-rect-20030314

FIXES 1333

rect is used uninitialized, I'm pretty sure they really meant to
refer to main_rect.top, not rect.top

STABLE12-winnt-fix-duplication-20030314

get rid of duplication

STABLE12-winnt-client-config-submount-fix-20030314

"Atempts to add a new drive, to the root, with a sub mount name, will
sometimes return error #35"

(cherry picked from commit 69a511d5b8a051f4961c3a6503566537893cb96a)

STABLE12-winnt-include-sanity-20030314

FIXES 1332

include headers by their usual name instead of trying to be special

(cherry picked from commit 48acf9c2d71ccd34ab7a95a8593a1689131fc032)

STABLE12-winnt-give-nosuchpath-if-thats-what-we-mean-20030314

FIXES 1329

"The Windows AFS client returns the error
ERROR_NOSUCHFILE, where the same request on a local NTFS volume returns
the error ERROR_NOSUCHPATH."

(cherry picked from commit ff6394b2e40b4c7c58b32a6e901656a12eb1faf1)

STABLE12-winnt-stat-cache-ref-leak-fix-20030314

FIXES 1327

avoid leaking stat cache entries

(cherry picked from commit de4a11b01da1fbf9dd253d16134e200486b76ff3)

STABLE12-rx-init-stats-mutex-20030314

instead of what rx-init-mutex-before-calling-getudpsocket-20030313, just
init the mutex earlier where others are done

(cherry picked from commit 3c3971439731f691dc9faa55ace72596e808b4d9)

STABLE12-winnt-client-config-submount-fix-20030313

"Atempts to add a new drive, to the root, with a sub mount name, will
sometimes return error #35"

(cherry picked from commit 3c22e61e23aae56cf8e40776c4327f7d640efa10)

STABLE12-rx-init-mutex-before-calling-getudpsocket-20030313

pointed out by cg2v@andrew.cmu.edu

(cherry picked from commit 7187d07423c1f3b4db3abc57a44f28c394a9e8ee)

STABLE12-cbd-all-flag-in-usage-20030310

cbd supports -all; put it in usage.

(cherry picked from commit 4f33d5b42854623a38d73e2b32d4fc872db356f7)

STABLE12-add-useful-string-functions-20020822

Add two useful functions for string operations: afs_strdup() and
afs_osi_FreeStr().  Will be used by an upcoming patch.

(cherry picked from commit 5d36376c07c665b175440e51b03fd85fc58422b0)

STABLE12-new-and-maybe-improved-cell-code-20020822

This is mostly a rewrite of src/afs/afs_cell.c, and associated changes
to other things.  Features of the new cell code:

 * Persistent name-to-cell-number mapping across reboots, to ensure the
   cell numbers in the dcache are meaningful.  Stored in the new cache
   file, /usr/vice/cache/CellInfo.
 * Allow cmdebug to query the cell number to cell name mapping.
 * Allow clients to have an empty CellServDB as long as AFSDB is enabled
   and ThisCell has AFSDB records.
 * The home cell retains the setuid bit even if looked up via AFSDB.
 * Dynroot gets its own cell, rather than piggybacking on cell 1.
 * Cell 1 is no longer special; cell numbers are now semi-opaque.
 * Convert cell traversal code to using GetCellByIndex rather than
   directly poking at CellLRU / afs_xcell.
 * Separate cells from aliases, which shouldn't have been the same
   in the first place.
 * Cleaner code (IMHO).

====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================

correct a code block so patch is correct

(cherry picked from commit 629d0806549d5b9cff9cd6f8eae50cf11c3e85ac)

STABLE12-misc-build-fix-20030306

fix error logging code

STABLE12-windows-cm-getcellbynum-rpc-20030306

missing stub

(cherry picked from commit 7ee56b700d73fd524bb92c628fe43d62e4365d6c)

STABLE12-rx-thread-id-startup-20030303

FIXES 1304

avoid using rxi_availProcs to allocate thread ids as it may decrement from under us

(cherry picked from commit 45a4d9cb08d92c10332bda071afb35461d5a4cc3)

STABLE12-winnt-cleanup-memory-leak-20030303

FIXES 1322

due to a change in the smb protocol we leaked memory. fix it.

(cherry picked from commit 241c06281036aedbce4c4e8c6a5ce6c981eef27a)

STABLE12-nuke-create-windows-pathnames-only-on-windows-thanks-20030228

don't construct a windows path on unix

(cherry picked from commit 54f056cc7a4a496c82eb6fdb9ba943c079702527)

STABLE12-salvager-create-larger-buffer-for-inodes-tmpfile-and-avoid-overrun-20030224

FIXES 1318

pointed out by Hartmut Reuter

(cherry picked from commit 0385de10a1cde2e520dc38fdb9ce1707e36afe1d)

STABLE12-viced-fix-port-logging-20030218

FIXES 1307

put port in correct byte order when logging

(cherry picked from commit 45a43078f0580c11e24ba34ea82518e5d4445967)

STABLE12-viced-host-logging-update-20030218

FIXES 1308

print info before letting host get gc'd

(cherry picked from commit 6c308978e1195246625babee91b5c49d84e23ffd)

STABLE12-callback-interfaces-check-malloc-result-for-error-20030208

if we run out of memory, maybe we should assert. we should probably log.
this is just so i remember to deal better.

(cherry picked from commit 1e4b22a9f579cd53a4aa17caedf04d06fe965feb)

STABLE12-winnt-find-token-event-missing-braces-20030206

FIXES 1288

add missing {} braces

(cherry picked from commit 922e0cce763f522c9fdf8df7d60d73f07056f45a)

STABLE12-butc-xbsa-cflags-20030203

FIXES 1283

still need to link in libs but this should at least get us further

STABLE12-lwp-tests-20030203

FIXES 1280

make the tests work

(cherry picked from commit f15e1e3f8f6d50ca931f2853bf9a166cddaa4c81)

STABLE12-cmd-print-help-in-cases-without-subcommands-20030203

FIXES 1277

so libcmd supports help for e.g. pt_util

(cherry picked from commit d32c70edbc0ed4e4821deb5542ef11637e8c04e0)

STABLE12-linux-include-thread-h-to-allow-variable-num-of-cpus-20030203

FIXES 1275

allow variable cpu numbers

(cherry picked from commit 5121aa82b5471b1b2088da53e7471bd2dc1cd3c9)

STABLE12-linux-config-defines-need-1-20030203

FIXES 1265

have to define to a value, not just define, local APIC stuff

STABLE12-rx-thread-id-assignment-fixes-20030203

in rx_pthread.c the
++rxi_pthread_hinum;
has to be protected by the rx_stats_mutex in all three cases. In the file I
attached in the openafs-devel article the last one was accidentally
unprotected.

the FSYNC_sync thread (fssync.c) needs needs to be handled as well.

(cherry picked from commit cff48dcb5210e38e0048a04fc3151d045bdca4ce)

STABLE12-more-redhate-beta-fun-20030131

apparently having recalc_sigpending stay the same was just as hard
as continuing to export the same level of functionality, so we get neither.

(cherry picked from commit 9a07ae91c39ded5541302d91aae81e5d137e4d06)

STABLE12-rx-avoid-deadlock-in-attaching-call-20030122

Avoid a subtle cycle in the locking order hierarchy involving
the rxi_ServerThreadSelectingCall flag, and instead rely on
call->lock to make sure we don't attach the same call twice.
Also some miscellaneous cleanup and code unification.

(cherry picked from commit a6d9298d3d299cc2c776b22a2800b5c2044ea497)

STABLE12-fix-netreceive-memleak-20030130

for darwin and freebsd, simplified for pullup

(cherry picked from commit 530c491c673154c5c935bd339c6d00850d454190)

STABLE12-better-assert-macro-20030113

Wrap our assert() macro in do{..}while(0) to make it look
just like a real function.  Fix instances of use of this
macro that lacked a trailing semicolon of their own.

(cherry picked from commit 3ba961ca00f5d8ec2edd32e0b1e5dac1c9864ef9)

STABLE12-better-signal-thread-support-for-fileserver-20030113

Improved signal-thread support for the pthread fileserver,
which avoids blocking signals in any thread, to appease Linux's
thread signaling semantics.

(cherry picked from commit eff534794e7f49166094fa9279bc4d51520d62a8)

STABLE12-pthread-fileserver-create-signal-thread-20030103

Create a special thread to receive signals in the pthread fileserver, in
order to avoid deadlock (we have signal handlers that unsafely grab locks).

(cherry picked from commit d86a3e307ec8ef4d35d8a9b96f3a5db1fbc5cd86)

STABLE12-darwin-creds-update-20030125

pullup from head

STABLE12-darwin-deal-with-unclean-pages-from-vm-20020107

based on report from lha@stacken.kth.se

(cherry picked from commit 4b690db92f2098f332132996a11ff07322284802)

STABLE12-pam-no-system-include-path-20021210

breaks modern gcc.  also, wrong.

(cherry picked from commit 7e2fcac3c15811b1210d08f86b3d288c2d430a2c)

STABLE12-rx-assign-thread-id-key-to-periodic-helpers-20030117

so each gets own holdbits

(cherry picked from commit ea4f1a80d73f62d91f0375d75dec67d38bd86f24)

STABLE12-viced-getclient-assert-logging-20030117

log some more info before asserting

(cherry picked from commit 2e3678c8829761506b9179d3660448c9ae21942b)

STABLE12-vol-clone-inc-failure-logging-20030117

log more info on asserts

(cherry picked from commit ec111ccc442439f68e9052721571ec8ef1754644)

STABLE12-viced-improved-host-logging-20030117

to deal with condition reported by lha@stacken.kth.se

(cherry picked from commit 12eb511d4516dc5c82dc565bf1ae634e479b09ed)

STABLE12-solaris-deal-with-gmake-stupidity-20030117

make gmake not try to link libafs.o and libafs.nonfs.o into another object

(cherry picked from commit 587304c2ba6f092c963dda8866cd273cf7a68d0e)

STABLE12-viced-alloc-hosts-held-and-locked-20030114

to avoid potential race as we chain in while allocating then try to lock
after returning to h_GetHost_r

(cherry picked from commit cff9ae0f1036782018c818ba7176ad16e9ae202c)

STABLE12-linux-ifdef-stupidity-cleanup-20030117

as pointed out by zacheiss@mit.edu

(cherry picked from commit 82d009d480c6b2a664264329f6940c1a519ea3df)

STABLE12-linux-fast-netdown-20020107

notice the NETUNREACH quickly and mark a host down

(cherry picked from commit b72c14c4cfdae85c469bcc1b88659e5bd6fb620c)

STABLE12-linux-lock64-fix-20030116

fix 64bit lock functions

(cherry picked from commit 3fa43862aa17467795121eb21f191066ab399980)

STABLE12-undo-unintended-change-20020817

An unintentional change accidentally leaked out when the
rx-avoid-using-newproc-on-solaris-20020812 delta was checked
in.  Undo it, since it breaks things.

(cherry picked from commit 9a871520c951db8f68b7b08acc26de5ffe8b5f57)

STABLE12-solaris-x86-58-support-20030113

needed for patch

====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================

make 1.2.x work on solaris 8 x86

STABLE12-linux-redhat-81-beta-support-20020107

thanks to Sean McDermott for the errno.h fixes.
the rest is my fault, the configure tests as well as the kernel changes

(cherry picked from commit 4624fd2206027a83ca3a27dd701f0ef4d3f01cef)

STABLE12-rx-avoid-using-newproc-on-solaris-20020812

Second attempt at avoiding the use of newproc() on Solaris.
This time, use thread_create to spawn a kernel thread, and
use a dummy packet to wake up the listener during shutdown.

(cherry picked from commit 5d3545610b83905e220d18c948d0fc621db851a5)

STABLE12-rxkad-2b-principal-translation-20021223

Do v5/v4 principal name translation.

(cherry picked from commit b8dbb6e0c9aaa6091080be01ac5426c52356196b)

STABLE12-vol-package-avoid-race-on-invalidating-hashed-vnode-20030111

thanks to whichever of jhutz@cmu.edu or cg2v@andrew.cmu.edu (or both)
noticed the race; also deal with wrap problem in VolumeCacheCheck that i
noticed

STABLE12-include-errno-dont-declare-it-20030111

include errno.h instead of extern int errno ourselves

(cherry picked from commit 92e76f7eb20ea548dfc2d688d9d6eeb6a58e2c18)

STABLE12-rxkad-be-openssl-keyschedule-safe-20030111

openssl has larger keyschedules than we do

crypto api is all so unfortunate

(cherry picked from commit 2be60b7ee43cb3baa1709506d685c86ea643815e)

STABLE12-linux-placate-whiny-gcc-20020110

debian unstable gcc gets bitter and cries to mom over newlines in strings.

(cherry picked from commit e1c86d76917ee5960de4a55f529f58e8c57dbcfb)

STABLE12-nfstrans-solaris-pre-7-support-20021210

pullup

Fix symlink support on Windows XP.

(cherry picked from commit d057f179506a94db0cc178e5df60977ee753f1f7)

STABLE12-nfstrans-solaris-pre-7-support-20021210

make it work for 2.6 and earlier

STABLE12-windows-wide-char-fix-20021210

include afxpriv.h so we get macros
call GetSize() and not GetCount()

(cherry picked from commit 7c5798f3902f7329511f127d7f6c8922885da383)

STABLE12-use-afs-strchr-instead-of-strchr-20021210

make dux happy

(cherry picked from commit 889da5053cb1b89ce25caec1bf9fa5a7dbb7830b)

STABLE12-windows-loopback-adapter-support-20021126

try to bind to loopback on windows

(cherry picked from commit fd856437c88b3a4f3954db1bf06f20ad67e77b44)

STABLE12-linux-ia64-updates-20021209

deal with *LK64 not defined
deal with ia64 sys_call_table

(cherry picked from commit f7f1c50256b4073f3f8b762468bda450ad288267)

Add 1.2.8 release notes.

Update NEWS file with user visible changes and other noteworthy things
about OpenAFS 1.2.8.

(cherry picked from commit 7a449be5fc0edc1ba4b86927856a44b865587c8b)

Begin release cycle.

STABLE12-windows-versioning-updates-20021205

conditionalize parts of windows build process on what platform the build is on

STABLE12-deal-with-non-modversioned-non-exported-sys-call-table-20021204

fix broken test

(cherry picked from commit 1bd76f6fe56f231d5fbc271cc7768c519058890e)

STABLE12-windows-fake-cml-info-20021126

fake up cml info for windows

(cherry picked from commit 0dd158bb6f09089928cfa80dbc2e1e13f12fda98)

STABLE12-windows-rename-warning-20021126

Currently AFS doesn't handle Explorer operations 'rename' and 'move'
correctly.  Normal operation of 'rename' and 'move' is to warn the operator
when the destination file name already exist.

(cherry picked from commit f5ff298ec5dcc3607621321e5785374bbed591fc)

STABLE12-windows-rename-20021126

fix rename on windows client

(cherry picked from commit fffba45737f37a0944491928352369e91547ff07)