Add a prototype of afs_GetRealCellByIndex to fix Irix build.

(cherry picked from commit fb6d2422f3b0b9ee756def4d91f11983fd767ad0)

Include errno.h so this compiles again.

(cherry picked from commit 49e0c6409a92e707901139789ca79d11a8aa8a67)

STABLE12-xdr-updates-20020731

update xdr so internal variables match what's being passed in

(cherry picked from commit 2d6ebd0cda81cff70a22a4393aaf58c5b7ccbb2e)

STABLE12-linux-start-if-dynroot-even-if-no-network-20020731

if we start with dynroot, dont worry if we have net or not

(cherry picked from commit aea332588be791dd79c467e727621be382d18a21)

STABLE12-deal-with-linux-EAGAIN-returns-when-receiving-20020731

sometimes you get EAGAIN from recvfrom, apparently if a packet comes in with a
bad checksum for instance. we should deal better.

(cherry picked from commit 6c5439b80934cd6e3f0539a07874c3bda18d676d)

STABLE12-linux-alloc-drop-glock-if-we-can-20020731

if we can drop the glock, do so. allocate up to PAGE_SIZE with kmalloc before
switching to vmalloc

(cherry picked from commit 4d75219bedc12833c6265b6e018e6455aa88921c)

STABLE12-afs-getcell-pioctl-should-be-able-to-provide-complete-list-20020731

don't use the LRUq since it can make us miss cells. don't use the existant
index because it includes cell aliases

(cherry picked from commit f2ab9fd440bcccb34e7875c06e00136fbefc10d1)

STABLE12-linux-note-we-use-interruptible-sleeps-in-non-compliant-manner-20020731

based on discussion with David Howells of RedHat

(cherry picked from commit f063c0db2109b3e089c54b4c6dd948bbd72a0ab1)

fix-fakestat-support-for-hardlinks-20020731 made afs_vnop_link.c
not compile under Sun's cc, since it tried to take the address of
a register variable.  Don't declare the arguments as register.

(cherry picked from commit 6c9fd25ae779bc092b9c1eb531d08483b4a96a53)

STABLE12-fix-fakestat-support-for-hardlinks-20020731

Support afs_link() in volume roots when using fakestat.

(cherry picked from commit 59741a821caec4ed83b87dcadbd4771e8d181c38)

STABLE12-make-mutex-tryenter-consistent-with-head-20020730

pullup change from head to MUTEX_TRYENTER

STABLE12-linux-enable-partial-writes-20020730

backport of linux-enable-partial-writes-and-prefetch-20020730
which does not include prefetch and dcache locking changes
(dcache locking isn't in 1.2 but prefetch should be pulled up
for 1.2.7)

STABLE12-return-correct-values-for-fcntl-FGETLK-20020727

Re-encode struct fcntl back into Linux's struct file_lock so
that fcntl(F_GETLK) returns proper values to the user.

(cherry picked from commit b8ebb1a225c55f38aa2a59b1ffd77c2bdc06ff9f)

Remove two unnecessary casts that prevented compilation on Linux 2.2.

(cherry picked from commit e9a6a539449ff3144169aa8ca5fc38bab613ab53)

STABLE12-solaris-nfs-module-updates-20020726

make it compile again

(cherry picked from commit 094adf6b12f3461a7f0ea4ec0a9f268c0c50336b)

STABLE12-verify-getcellbyindex-return-value-20020726

Verify that afs_GetCellByIndex() returns a valid cell -- it might
be NULL if afs_NewCell() aborted partway through, which could be
the case of the CellServDB file has a typo.

(cherry picked from commit 9d8ad37ced11afac107a2b7d8d1bd95c7e1aa9f8)

STABLE12-cleanup-pioctl-avoid-inadvertantly-setting-new-pags-due-to-gcc-20020726

fix operator precedence bug

(cherry picked from commit 07a4570310f6bdddbe07ffccb3ece579ff4bdff8)

STABLE12-windows-no-more-files-workaround-20020725

work around erroneous "no more files" error in winnt client

(cherry picked from commit 595f9ae4334f1dcf665944b8e6493a4f82855d13)

STABLE12-tbutc-uses-not-thread-safe-status-20020725

"The multithreaded butc (src/tbutc/butc) includes bucoord/status.o, which is
compiled together with the other bucoord stuff without MT_CFLAGS and hence
without AFS_PTHREAD_ENV defined. Alas, it uses ObtainWriteLock and other
macros out of lock.h which *are* sensitive to AFS_PTHREAD_ENV. Consequently,
butc can hang trying to acquire the statusQueueLock when dumping volumes, in
particular if 'backup status' commands are issued frequently.

A proper fix would probably be to push some of those macros out of lock.h and
make them subroutines in lock.c.

A quick fix is ensuring status.c gets recompiled in tbutc/Makefile.in"

(cherry picked from commit fd2cb884e24b5bf48392db4222ee10125014a3ee)

STABLE12-linux-hold-glock-across-dentry-trace-entries-while-tracing-20020725

this is actually rather scary

(cherry picked from commit 0f4021128e1016ec9a6fea3a52f47e616fcb2022)

STABLE12-linux-cleanup-rx-cv-locking-20020725

cleanup wait queue usage
grab glock before mutex (like all other ports)

(cherry picked from commit fc71eb6e255b6d2bc1d24e9aae799329c675ec49)

Oops, missed a portion of diff.

Fix typo introduced while merging in fakestat support.

Pullup all fakestat support to 1.2 branch.  This includes all of the
following deltas:

- initial-fakestat-support-20020402
- fakestat-fix-eval-args-ordering-20020409
- fix-evalfakestat-argument-ordering-20020409
- minor-fakestat-cleanup-20020720
- update-volume-root-attributes-on-linux-fakestat-20020720

Translate faultcode_t return values from vm routines into errno codes.

(cherry picked from commit 9cd177c636c59a7369155d56582bffffb744aeb7)

Correctly check for 32-bit file size limitations on Solaris;
the MAXOFF_T value changed from 32-bit to 64-bit in Solaris 7.

(cherry picked from commit fbb81cb4ed02a8b9a41657d7ebfdb26b73bdf935)

make it work for 2.2 kernels also
also some debugging

(cherry picked from commit 73cdd10a1b5bda70cb0a1f6299c3cc570ac97b69)

"d_delete: needs to call iput() to decrement i_count, which
you get by returning 1 instead of 0.

looking at the other fs schemes (actually its pretty clear in the
nfs code too) i see that they return 1 for 'is_bad_inode()'."

(cherry picked from commit cabd0bbd4f4129efec2b87b6934c691bf91e883c)

STABLE12-uuid-dont-require-snprintf-in-ukernel-20020711

because obviously not everything has it

(cherry picked from commit c927f48236dd489a33b0bc9fc3959baf1fae0e45)

STABLE12-darwin-vm-bugfix-for-running-apps-out-of-afs-20020710

this "fixes a problem that causes the afs kernel code to spew
into syslog if a long running application is run out of /afs"

(cherry picked from commit 7e0d50a1083e765d12a4978150002cfd7d806a91)

STABLE12-inux-kernel-detect-gfp-mask-in-struct-address-space-20020624

based on report that new redhat kernels need this

STABLE12-linux-sleep-wakeup-cleanup-20020624

"cleanup osi_sleep.c to remove some additional races.  this patches
removes AFS_GLOBAL_SUNLOCK and just uses that case to handle both
smp and non smp cases.  osi_TimedSleep always sleeps atleast 1 hz.
(it also always sleeps interruptibly -- you might get some extra
loops but this is better than the load average running away)
sleep_on's are replaced with the more appropriate add_waitq/schedule
sequence.  getevent is called after addevent since there is no
danger of GLOCK dropping.  getevent/addevent isnt re-merged since
osi_Wakeup doesnt create entries in the event table."

(cherry picked from commit 048543ade8eab56d47e5094fdb56bb7f49983dcf)

STABLE12-openafs-tools-update-20020630

updates for openafs-tools command suite

(cherry picked from commit c746dbb3cb3fa97c4f8253eef39cd3f01d0b0e12)

STABLE12-linux-gcpags-support-20020625

might even work

(cherry picked from commit b2788cf3abf20f404c16c57314e126a2b1b5800f)

STABLE12-vol-allocate-partition-names-dynamically-to-avoid-problems-with-vx-stuff-20020624

otherwise long device paths can screw you

(cherry picked from commit 78b1f7ec565ec32e114e80035581f95de50a9f04)

STABLE12-linux-kernel-detect-gfp-mask-in-struct-address-space-20020624

based on report that new redhat kernels need this

(cherry picked from commit 19ffb82266bebfbcd529a85b43765b95355195f9)

STABLE12-linux-more-osi-sleep-wakeup-fun-20020624

"the attached patch 'fixes', in my opinion, in osi_alloc.  objects <=
AFS_SMALLOCSIZ are alloced with kmalloc, other vmalloc.  if kmalloc
fails it doesnt fail over to vmalloc.  TASK_RUNNING is set after
schedule.  it doesnt drop the GLOCK.  i moved the allocator init
later so that it doesnt need to toggle the semaphore just before
going to linux_alloc."

(cherry picked from commit fe1648ca02137cac306d5f5002d8f79b580694ba)

STABLE12-ufsread-save-the-oops-for-where-we-expect-it-20020624

based on comment from matt@slackers.net

(cherry picked from commit 60659d2c8fed0c90584bfffb33c2efa6e6270343)

STABLE12-add-uuid-utility-functions-20020622

Add afsUUID_to_string() and afsUUID_from_string() utility functions, from
arla.

(cherry picked from commit 664e4f6f0ddbe30ebc78053a7b16634bac926e37)

STABLE12-define-config_smp-to-some-value-not-empty-string-20020622

Use:

#define CONFIG_SMP 1

instead of:

#define CONFIG_SMP

to fix recent Redhat kernel which used #if CONFIG_SMP instead of
#ifdef CONFIG_SMP in some kernel headers.

(cherry picked from commit 691edd4ea6c117b25aeb124ce7c4a7755694cb37)

STABLE12-linux-newvcache-make-sure-we-dont-try-to-iterate-null-pointer-20020615

based on report from wilder@us.ibm.com
not clear this is sufficient

(cherry picked from commit 307543629ddeb4b4f34ff5947038fd6a224547b5)

STABLE12-afsd-dont-print-garbage-on-bad-cacheinfo-file-20020615

based on report from matt@slackers.net

don't bother printing anything we parsed, since we're guaranteed at least
one item is trash anyhow

(cherry picked from commit d13f701812dd19e26317c602d9a09f0d53fe40b6)

STABLE12-dux-sysent-update-20020614

cast this to deal with changes in sysent in later tru64

(cherry picked from commit da8653c30b7b850269ac20c3c94c37c7d36a59c9)

STABLE12-vos-listaddrs-cleanup-and-allow-printing-of-ip-or-uuid-or-both-20020613

lha@stacken.kth.se complained that we sometimes printed garbage, so i took
the opportunity to both fix it and update ours to support the options
the arla vos listaddrs does

(cherry picked from commit a7500e87f5d3ef97daf158411af4b47fce91aa54)

STABLE12-windows-update-dirpath-header-20020612

add missing definitions

(cherry picked from commit 0f64b6b7528d61199c7c169d06325f5fde2e14f6)

STABLE12-arla-pioctls-now-bogus-20020612

as reported by lha@stacken.kth.se

(cherry picked from commit 544fe2ef8ed929e805bac27faa3c2e62aaaadcba)

STABLE12-solaris-x86-20020611

solaris 7 and 8 for x86

(cherry picked from commit a271a55041086f062ca888e5cdde02fe234f7281)

STABLE12-solaris-9-support-20020611

update hrestime interface

(cherry picked from commit 9c29a6ead1730649ca501c85482288c73e99f693)

STABLE12-linux-dont-install-osi-vfs-h-when-no-kernel-module-20020611

the rest of the patch

(cherry picked from commit de06f45787658ff47d959d782a57e498598c1a4f)

STABLE12-windows-callback-interface-updates-20020611

make callback interface like unix

(cherry picked from commit eb8c33d2993ef279218e970a300d5bac2bb63f9b)

STABLE12-pam-update-for-correct-krb-aware-module-20011102

this still missed being pulled up

(cherry picked from commit 68196d85dbd09220f91421101f651d8e8d7322c3)

STABLE12-revert-dont-spawn-kernel-thread-for-rxk-listener-20020426

in some cases this causes shutdowns to hang forever. notably shutdown -i0 -g6 -y works while reboot does not

(cherry picked from commit fe1beb4f3e69f4ae2dbec60902b60cd601674d15)

STABLE12-warn-of-but-allow-bogus-shutdown-20020608

despite the danger of it, allow shutdown without umount, to prevent hanging

Add OpenAFS 1.2.5 Release Notes.

STABLE12-getcellservdb-cleanup-20020608

based on comments and suggestion from Srikanth Vishwanathan <vsrikanth@in.ibm.com>
avoid deadlock. send back variable length list. allocate what we need to, and
be care to clean up after ourselves if there's an error. whether this
should be variable length is arguable but it's pretty clear no one was using
this before.

(cherry picked from commit 08bb209eef379534268ec6433a131bc52711d111)

STABLE12-use-private-xdr-in-kernel-to-avoid-conflicts-over-memory-ownership-20020608

allocating memory from a pool and then having it be freed out from under you is unwise

(cherry picked from commit b80ae7d78cd944420c54567030e57c4f68a3d482)

Update NEWS file with important changes in 1.2.5 release.

(cherry picked from commit 3651150489cc5e98dc88643d0a24d0f3941f86b8)

STABLE12-create-openafs-1_2_5-20020608

new release cycle

fix afs to enforce that /afs must be unmounted, and afsd to report it

(cherry picked from commit 155c13192339239dd1d794210b5bcedfb3feea15)

Recognize MacOS X 10.1.5 as ppc_darwin_14.  It's compatible.

(cherry picked from commit 3789698e0ab94c4ff0ebb875d336eb34ff2aa5aa)

STABLE12-avoid-deadlock-while-upgrading-vcache-lock-20020605

Delay calling afs_ProcessFS() and upgrading the shared vcache lock
until after the dcache locks have been released, to obey lock ordering
and avoid deadlock.

(cherry picked from commit 9ab8961c3106402de12f4d0dd80c65e249293fdb)

STABLE12-des-fix-linux-64bit-problems-more-usefully-20020607

with a minor fix from shadow@dementia.org

long isn't 32bit on ia64 or alpha

(cherry picked from commit ddfc17dfd19dd4e5bf83f42b92c3a25ebd101c59)

STABLE12-NT-dirpath-copy-in-useful-file-20020516

basically until NT port has been properly set up to deal with dirpath, make
it work as before

(cherry picked from commit 63ea4ee399b9fc2a52d81ace70e9902424195c2e)

STABLE12-afs-attrs-cast-date-to-int-to-avoid-negative-dates-20020516

cheap hack to avoid negative date problem which occasionally plagues solaris

(cherry picked from commit c28c22ceadfb1db8151ae74dacee2a49c42624db)

STABLE12-pioctl-avoid-inadvertantly-setting-new-pags-due-to-gcc-20020516

some versions of gcc compile the if used to decide whether to allocate a pag
badly. make life easier.

(cherry picked from commit 10486b5652f0fb5a1d8c1fe5ddbbf5daa2477974)

STABLE12-viced-malloc-dont-make-assumptions-about-the-compiler-20020524

based on report from lha@stacken.kth.se
"identP = (struct Identity *)malloc(1);

This can't be right, there should not be an assumption how the
c-compiler will lay out code. I find it acceptable to do something like this

           /* sizeof valid + offset of valid. */
           i.. = malloc(sizeof(identP->valid) + (int)&(((struct Identity *)0)->valid));

But then, why not just allocate a whole identP and be happy. A afsUUID
is not that bloted (16 bytes) so I can't see why way this is
chosen. 20 bytes will probably be allocated on 32 bit platforms for
the whole struct Identity."

(cherry picked from commit bbab90943637385742923c8bebebdd750a283e5c)

STABLE12-dir-hold-afs-bufferlock-across-increment-of-buffer-lockers-to-prevent-newslot-from-walking-on-stuff-20020514

liberal debugging code and hammering produced a failure where newslot was
actually walking on another call to newslot because the bufferlock was
being dropped before newslot incremented lockers on the buffer it
was allocating, allowing someone else to come along and also think they
could allocate it.

by holding the bufferlock just a little longer we avoid this

(cherry picked from commit 0eb68f307aac84472a13523a0ce8b7a865f01ac7)

STABLE12-viced-cleanup-dirhandle-references-after-ourselves-20020509

doesn't fix any problem that i can find, but results in ihandles not
having thousands of bogus references (and may fix problems if refcnt were to wrap)

(cherry picked from commit ee798e49f23d7894c92275fc3d69b8b5c3c516a2)

STABLE12-vol-ihandle-cleanup-20020429

this does fix some potential problems, even if none of them are the
CopyOnWrite problem. basically, ih_reallyclose() could reinsert a now-unref'd
fd handle into the wrong list when cleaning up, and this cleans up the code
considerably.

(cherry picked from commit 5cbc233c648e968381a473a6bc1d6d599d1738d0)

STABLE12-fix-pointer-argument-type-in-hLookup-20020422

The heldp argument to h_Lookup should be an int* rather than int,
as pointed out by amu@alum.mit.edu.

(cherry picked from commit 27595ffcd995dc31c565acdb630d5b60a275fe27)

STABLE12-prevent-host-deletion-race-20020415

Make h_Lookup_r return the host held, and ensure that it's not
deleted at that point; otherwise, the host may be deleted by
another thread between a call to h_Lookup_r and the time when
we actually get around to h_Hold'ing it.

(cherry picked from commit d4493526a764f5d4d1ef6a437ab8cfa467077be6)

STABLE12-fix-__afs-file-creation-20020411

Pass &treq to afsrename() when preserving a currently-open file.

(cherry picked from commit 301d83202c8d969197ea9515e02eb93c1b2bd72f)

osi_linux_unmask() needs to explicitly flush_signals() before calling
recalc_sigpending(), otherwise recent kernels don't appear to deliver
new instances of any signals already received.

Revert back to using SIGKILL at shutdown time, but use SIGKILL instead of
explicitly using signal 9, because abstraction is your friend.

(cherry picked from commit 8251fc6871a65326150be69b929917b37d83e215)

Seems to fix a hang at shutdown time seen on Redhat 7.3 kernels; doesn't
appear to affect older kernels.

(cherry picked from commit 097df1049a6316f087e1c79b30a332358fa761c9)

Allows us to build modules for the bigmem kernels redhat started
shipping.

(cherry picked from commit fe6b4fd293f789c4d9f8f62401e31a78b466b09b)

Update NEWS file for user visible changes in 1.2.4.

(cherry picked from commit 50473f885f07857d03b963bf43c5b24771b4913b)

Add 1.2.4 release notes.

begin release cycle.

Avoid kernel panic when dynroot and memcache are used simultaneously.

(cherry picked from commit 81a3f0068f3488b7fb20fb9d6b7441f64f5b61e0)

STABLE12-dir-buffer-newslot-cant-return-zero-20020508

since newslot will call Die() instead of returning 0 (causing assert(0))
this code can never be called. remove it.

STABLE12-need-prototype-for-osi-dnlc-lookup-20020503

Include afs_osidnlc.h for osi_dnlc_lookup() prototype.  Apparently
fixes a panic on ia64.

STABLE12-linux-include-osidnlc-header-for-ia64-20020506

this is in the head and not here; ia64 needs the prototype

Include afs_osidnlc.h for osi_dnlc_lookup() prototype.  Apparently
fixes a panic on ia64.

keep small globals from being gprel

(cherry picked from commit 1e90eb264895b5bbe974c7c8c32dbf89e48d6081)

rewrite syscall stubs not in machine code

(cherry picked from commit 25de65bc0bba9bc2df382dc34c1c5b0c1460aca5)

avoid ESTALE in 2.4.19-pre7

(cherry picked from commit 7147cc946bead75eee909ea1134bc542c0b09c4f)

Don't spawn a separate kernel thread for rx kernel listener under
Solaris, but instead just use the afsd fork for the listener. Use
of newproc() is deprecated, according to Frank.Batschulat@Sun.COM.

(cherry picked from commit be1458b07817900d15f5c069ea66b72ef50b91e9)

If the inode is not allocated, don't bother updating timestamps.
Pointed out by Frank.Batschulat@Sun.COM.

(cherry picked from commit a31928815796bcbd1acf3005a0f628a21bbb0d8a)

STABLE12-darwin-use-namei-fileserver-20020429

build namei fileserver for darwin

(cherry picked from commit c2e075fba4cec152e772ec5514d52e8ec5b29990)

STABLE12-cvsignore-updates-20020429

stuff missed in the stable branch

(cherry picked from commit 291e8352be1ee0c6e3da79328e752b82a7a13190)

STABLE12-darwin-map-more-versions-20020425

5.3 and 5.4 are "Same" enough

(cherry picked from commit 1ccedea1a0d978f213a9051e15b5e5efb7ccffe8)

STABLE12-darwin-fix-build-20020401

add missing )

(cherry picked from commit fd318a1fc9551a156cdb43916143d3b959b6720f)

STABLE12-rxkad-stats-initialize-to-placate-darwin-compiler-20020425

because it deals badly with common symbols

(cherry picked from commit 0c0eb03e9a806dc84b7c3251478de4766a8fd853)

even though right now a vnode and an inode are basically but not exactly
the same, part of the usefulness of this change is the ability to
convert to inodes linked from vcaches later. make the macros fit with
this view of the fture

(cherry picked from commit 39ee79e7c328d86738356aa5a736e24339f8b126)

convert vcache casting in cache manager to macros

should ease further changes later

(cherry picked from commit 3bf8818d956337fcf70a2734f200275f403e8814)

provide sysctl interface for toggling module state. provide hooks for
all currently known tweakable items

(complete earlier incomplete pullup)

STABLE12-linux-ia64-cleanup-casting-20020423

so we don't need the SYSCALL2POINTER change

(cherry picked from commit 8268bf34acd715d28137a3d601252fd235d1a9fb)

update setgroups32 for 2.4 and make sparc64 linux2.4 work again

(cherry picked from commit 5d57da192ea989d9ab4e83cb5f721a7975def9fb)

STABLE12-linux-ia64-updates-20020422

updates to make it work with later 2.4

SYSCALL2POINTER change may be unnecessary (sparc64 way should work)
but without any way to test...

(cherry picked from commit e12859bed90a0bd896150b7cfc137d0ef52e48c3)

STABLE12-darwin-vnops-make-unlink-fix-fake-references-20020422

so unlink doesn't panic us

(cherry picked from commit 6d43ab450a907bb1eccf9561c37c108202cd8e52)