openafs.git
12 years agoDEVEL15-tweak-regen-20080521
Simon Wilkinson [Wed, 21 May 2008 16:41:07 +0000]
DEVEL15-tweak-regen-20080521

LICENSE IPL10

make regen quieter if you don't want man pages

(cherry picked from commit 1705bda53e4ca198cf78d70ff424fc7af69b1879)

12 years agoDEVEL15-macos-tuning-20080521
Derrick Brashear [Wed, 21 May 2008 14:52:11 +0000]
DEVEL15-macos-tuning-20080521

LICENSE IPL10

be more aggressive to get better performance

(cherry picked from commit 30f65085f7a53ae073e2c4649a462a3a555ebb61)

12 years agoDEVEL15-pmtu-and-prefetch-20080520
Derrick Brashear [Wed, 21 May 2008 05:22:27 +0000]
DEVEL15-pmtu-and-prefetch-20080520

LICENSE IPL10

try hard to optimize using the wire for high latency connections

(cherry picked from commit 1206e7538be86f073b21cd289266286b60a95d0a)

12 years agoDEVEL15-linux-hlist-unhashed-opencoding-20080520
Derrick Brashear [Tue, 20 May 2008 22:03:20 +0000]
DEVEL15-linux-hlist-unhashed-opencoding-20080520

LICENSE IPL10
FIXES 93871

if we can't use hlist_unhashed (which is dumb) or we don't have it (less dumb) w
ork around it.

(cherry picked from commit c761f248397334400f8b4bb71cc4d01d3e4fc6ff)

12 years agoDEVEL15-rx-connection-clones-20080520
Matt Benjamin [Tue, 20 May 2008 21:24:38 +0000]
DEVEL15-rx-connection-clones-20080520

LICENSE IPL10
FIXES 89557

add connection clones to allow more than maxcalls per "connection"

(cherry picked from commit f00df529b79d9df64b77dd7b054eb362a92c279e)

12 years agoDEVEL15-aix-unpin-after-free-20080520
Derrick Brashear [Tue, 20 May 2008 20:50:41 +0000]
DEVEL15-aix-unpin-after-free-20080520

LICENSE IPL10
FIXES 99456

in order that cleanup be safe we need to do it this way

(cherry picked from commit 3edb97360b7d971750e54471037584d7a9a74993)

12 years agoDEVEL15-windows-smb-parse-string-20080513
Jeffrey Altman [Wed, 14 May 2008 05:16:51 +0000]
DEVEL15-windows-smb-parse-string-20080513

LICENSE MIT

if the input length is 0, smb_ParseString should return the NUL string
not NULL.

Do not permit unprotected DebugBreak() calls.

(cherry picked from commit 090b2eebfa9fb48deea94bf7b511e693a251e041)

12 years agoDEVEL15-windows-unicode-normalize-btree-20080511
Jeffrey Altman [Mon, 12 May 2008 04:03:57 +0000]
DEVEL15-windows-unicode-normalize-btree-20080511

LICENSE MIT

Attempt to normalize afs directory strings before inserting them into
the B+tree.

(cherry picked from commit 53cffb32ab7cc5a650d0bbaa66686a44ea206888)

12 years agoDEVEL15-windows-unicode-support-normalize-20080511
Asanka Herath [Mon, 12 May 2008 02:37:56 +0000]
DEVEL15-windows-unicode-support-normalize-20080511

LICENSE MIT

Actually normalize when normaliz.dll is present.

(cherry picked from commit 9e67c4f766892f7c0b79a2d6a1d8b96a7e4b7100)

13 years agoDEVEL15-windows-unicode-support-20080511
Jeffrey Altman [Sun, 11 May 2008 14:17:17 +0000]
DEVEL15-windows-unicode-support-20080511

LICENSE MIT

In NTCreateX a zero length name is permitted.  This can occur when
opening the root directory of a share.

Change cm_NameEntries() to accept a size_t* instead of long* since
that permits better type safety.

(cherry picked from commit 9d28d95151418189e8004734eca0edd498edd46b)

13 years agoDEVEL15-udebug-display-lastyeshost-order-20080510
Jeffrey Altman [Sat, 10 May 2008 16:21:19 +0000]
DEVEL15-udebug-display-lastyeshost-order-20080510

LICENSE MIT

lastYesHost is already in network byte order.  do not reorder it again
when displaying it.

(cherry picked from commit b5dc1903a4bca5bae099beec0b95840df7282250)

13 years agoDEVEL15-server-meltdown-avoidance-20080509
Derrick Brashear [Fri, 9 May 2008 19:50:56 +0000]
DEVEL15-server-meltdown-avoidance-20080509

LICENSE IPL10

if the server is melting down or otherwise unable to reply with data, it will
cut itself off.

(cherry picked from commit c26dc0e6aaefedc55ed5c35a5744b5c01ba39ea1)

13 years agoDEVEL15-windows-unicode-support-20080509
Asanka Herath [Fri, 9 May 2008 16:10:06 +0000]
DEVEL15-windows-unicode-support-20080509

LICENSE MIT

This delta provides a fairly complete implementation of Unicode character
set support for the Windows Cache Manager and supporting tools including
fs.exe, symlink.exe, the pioctl library, and the explorer shell extension.

New Build requirements:

In order to build the Microsoft IDN Mitigation APIs SDK 1.1 is now required.
This SDK provides the normalization.h header and the redistribution
install packages required to install normaliz.dll.   The AFSDEV_INCLUDE
path must be modified to point to the directory containing normalization.h.
There are no lib files for this package.

Cache Manager:

The CM SMB server has been modified to negotiate the use of UNICODE.
By default it is on.  The "fs smbunicode" command can be used to disable
UNICODE at runtime.

cm_utils now contains UNICODE normalization and conversion routines built
on top of the IDN Mitigation APIs.

All input strings are normalized with UNICODE Normalization Form C.

The pioctl interface now supports UTF8 strings in addition to ANSI.
UTF8 strings are prefixed with <ESC> % 8 as per the ISO 2022 extension.

Pioctl Library:

New apps should use the pioctl_utf8() function instead of pioctl().
pioctl() is for ANSI strings and provides backward compatibility
with third party apps.

fs.exe and symlink.exe:

Converted to Unicode applications.  All strings are now communicated
using UTF-8.

Explorer Shell Extension:

Converted to a Unicode DLL.  All strings are now communicated using
UTF-8.  Fonts on U.S. Windows distributions are not fully populated.
As a result not all of the characters can be displayed in all of the
dialog boxes.

Still to do:

1. Add Unicode normalization to AFS directory entries and mountpoint
   and symlink target strings.

2. Use the Unicode version of GetCurrentDirectory in fs_utils.c

3. Update the installers to install the IDN Mitigation APIs on XP
   and 2003.  Vista already has them and they are not supported on
   pre XP SP2 releases.

(cherry picked from commit a36ec90cff608c3805f524070eb21aacc118f9e2)

13 years agoDEVEL15-windows-have-snprintf-20080509
Jeffrey Altman [Fri, 9 May 2008 04:57:23 +0000]
DEVEL15-windows-have-snprintf-20080509

LICENSE MIT

Visual Studio provides [v]snprintf.  No need to define and export it.

(cherry picked from commit 048ae7c7e63e03af16925613aeec8243b45ffe70)

13 years agoDEVEL15-windows-version-20080508
Jeffrey Altman [Fri, 9 May 2008 00:25:35 +0000]
DEVEL15-windows-version-20080508

LICENSE MIT

1.5.35

13 years agoDEVEL15-windows-release-notes-20080508
Jeffrey Altman [Fri, 9 May 2008 00:25:00 +0000]
DEVEL15-windows-release-notes-20080508

LICENSE MIT

1.5.35

13 years agoDEVEL15-windows-notes-20080508
Jeffrey Altman [Fri, 9 May 2008 00:01:30 +0000]
DEVEL15-windows-notes-20080508

LICENSE MIT

1.5.36

13 years agoDEVEL15-openafs-1-5-36-20080508 BP-openafs-devel-1_5_3x openafs-devel-1_5_36
Derrick Brashear [Fri, 9 May 2008 02:06:27 +0000]
DEVEL15-openafs-1-5-36-20080508

LICENSE IPL10

make 1.5.36

13 years agoDEVEL15-rx-retain-windowing-per-peer-20080508
Derrick Brashear [Thu, 8 May 2008 22:25:58 +0000]
DEVEL15-rx-retain-windowing-per-peer-20080508

LICENSE IPL10

we learned about the peer in a previous connection... retain the information
and keep using it. widen the available window.

makes rx perform better over high latency wans. needs to be present in both
sides for maximal effect.

(cherry picked from commit 3feee9278bc8d0a22630508f3aca10835bf52866)

13 years agoDEVEL15-viced-avoid-infinite-loop-with-altaddr-unset-20080508
Jeffrey Altman [Thu, 8 May 2008 22:18:56 +0000]
DEVEL15-viced-avoid-infinite-loop-with-altaddr-unset-20080508

LICENSE IPL10

if we return leaving a non HOSTDELETED host while it does not have ALTADDR set,
a loop can happen. avoid it by simply setting ALTADDR again when we're done,
if we believe some addresses are still good this is fine.

(cherry picked from commit e6c818f070022529c362e42dfed340f54b2fd26e)

13 years agoDEVEL15-windows-buf-invalidate-on-vnovnode-20080507
Jeffrey Altman [Thu, 8 May 2008 06:27:14 +0000]
DEVEL15-windows-buf-invalidate-on-vnovnode-20080507

LICENSE MIT

Once we know that an attempt to stablize a buffer has failed due to
VNOVNODE, stop trying to Stablize future buffers and just invalidate
all of them belonging to the same FID.

(cherry picked from commit ef619db804d471fda3bf3f77d55e8cd4e4699b29)

13 years agoDEVEL15-viced-large-more-threads-20080506
Derrick Brashear [Tue, 6 May 2008 16:03:07 +0000]
DEVEL15-viced-large-more-threads-20080506

LICENSE IPL10

fileserver -L should imply -p 128, not -p 12

(cherry picked from commit c847fb4964c0ec24c404cf1fb353b08af81b5eac)

13 years agoDEVEL15-windows-server-retry-20080502
Jeffrey Altman [Fri, 2 May 2008 20:28:18 +0000]
DEVEL15-windows-server-retry-20080502

LICENSE MIT

do not retry if there is no server available for the current attempt

(cherry picked from commit 98e658b7821f5020c67c05503f9e2c61c945554a)

13 years agoDEVEL15-butc-xbsa-lwp-protoize-damage-20080501
Derrick Brashear [Fri, 2 May 2008 01:59:38 +0000]
DEVEL15-butc-xbsa-lwp-protoize-damage-20080501

LICENSE IPL10

make xbsa variant work again

(cherry picked from commit abff169b58764380650e6e162707be68d3aa9bf1)

13 years agoDEVEL15-uuid-corrected-duplicate-check-20080501
Derrick Brashear [Thu, 1 May 2008 20:00:10 +0000]
DEVEL15-uuid-corrected-duplicate-check-20080501

LICENSE IPL10

apply check to chain, not input

(cherry picked from commit 7915302f28091dd0a6547f0f58fa5ceb581cc21a)

13 years agoDEVEL15-client-idledeadtime-support-20080430
Derrick Brashear [Wed, 30 Apr 2008 20:08:04 +0000]
DEVEL15-client-idledeadtime-support-20080430

LICENSE IPL10

in the event a server is "melting down" and not responding with data to RPCs
this adds the ability to time out the calls rather than letting keepalives
effectively hang the client forever.  does not mark the server down since...
it's not down.

Side effect: adds similar tracking for authentication errors so tokens need
not be discarded if one server's clock has drifted.

(cherry picked from commit 0a2c144307d2471450c0660e141906a1f7b1ea36)

13 years agoDEVEL15-ubik-recovery-swap-in-new-fd-20080428
Derrick Brashear [Mon, 28 Apr 2008 22:48:11 +0000]
DEVEL15-ubik-recovery-swap-in-new-fd-20080428

LICENSE IPL10

when recovery is done, forcibly invalidate the old fd

(cherry picked from commit 888bc28b1886e994d430ddc61657e60a910208a6)

13 years agoDEVEL15-tbutc-wrong-lp-getrwindex-20080428
Hans-Werner Paulsen [Mon, 28 Apr 2008 17:15:06 +0000]
DEVEL15-tbutc-wrong-lp-getrwindex-20080428

LICENSE IPL10
FIXES 96575

The pthread butc crashes with SIGSEGV, when I restore to a volume
which already exists. This does happen with OpenAFS 1.4.7pre4 on
i386_linux26.
Looking into the code I think butc is using the wrong function
Lp_GetRwIndex (at volser/vsprocs.c:4394). It uses the one in the
library lib/libvosadmin.a, but should use the function defined
in volser/lockprocs.c.

(cherry picked from commit 668cd5331e4857a73ba1d849446e0dfcae97eadf)

13 years agoDEVEL15-udebug-64bit-time_t-vs-printf-20080428
Jeffrey Altman [Mon, 28 Apr 2008 16:20:41 +0000]
DEVEL15-udebug-64bit-time_t-vs-printf-20080428

LICENSE MIT

64-bit time_t values cannot be passed to printf() for evaluation
by a %d variable.  To avoid the problem we cast the value to int.

(cherry picked from commit 25d9dbb6be970d54c4711890b2085c153eda15fa)

13 years agoDEVEL15-windows-rx-idle-timeout-20080428
Jeffrey Altman [Mon, 28 Apr 2008 16:06:14 +0000]
DEVEL15-windows-rx-idle-timeout-20080428

LICENSE MIT

Add "IdleDeadTimeout" registry value and call rx_SetConnIdleDeadTime()
with it when creating new rx connections.   This forces the Rx stack
to return RX_CALL_TIMEOUT if the file server has returned VBUSY for
longer than the IdleDeadTimeout.

When RX_CALL_TIMEOUT is received, the cache manager will now attempt to
find another server for the object that might be less busy.

(cherry picked from commit 2f2d5fd8a7e80802fead9ee944e33c1a0c818abd)

13 years agoDEVEL15-kdump-install-fixups-20080427
Chaskiel M Grundman [Mon, 28 Apr 2008 04:45:15 +0000]
DEVEL15-kdump-install-fixups-20080427

LICENSE IPL10
FIXES 96465

The install rule:
- installs the empty 'kdump-build' file instead of the binary on all
platforms except irix
- overwrites the selection script with the binary on linux and hpux
- doesn't have an alpha_linux branch
- installs the wrapper script as kdump32 and the binary as kdump on
solaris (instead of vice-versa)

The dest rule:
- overwrites the selection script with the binary on alpha_linux and hpux

(cherry picked from commit 7a6751c08800da5d49bf3def9175174499f6337e)

13 years agoDEVEL15-redhat-cellservdb-update-20080423
Derrick Brashear [Sun, 27 Apr 2008 05:00:22 +0000]
DEVEL15-redhat-cellservdb-update-20080423

LICENSE IPL10

take update from upstream

(cherry picked from commit c8f4d9f0079b7eeacf831061a0c39310aafaa045)

13 years agoDEVEL15-zero-uio-len-when-trimming-20080426
Derrick Brashear [Sun, 27 Apr 2008 04:54:14 +0000]
DEVEL15-zero-uio-len-when-trimming-20080426

LICENSE IPL10

make it match ufsread case. should change this later.

(cherry picked from commit 2aa615fed2b801a7ed3abf7da5aaa293a4f783bf)

13 years agoDEVEL15-vos-move-avoid-spurious-unlock-20080424
Derrick Brashear [Thu, 24 Apr 2008 23:45:47 +0000]
DEVEL15-vos-move-avoid-spurious-unlock-20080424

LICENSE IPL10

lightly modified from what Bren Mills submitted; the second unlock call after mfail may be entirely useless.

(cherry picked from commit b87935a4c5f26357e29f1db46eff8b1cf9185f3b)

13 years agoDEVEL15-windows-buf-refcount-20080424
Jeffrey Altman [Thu, 24 Apr 2008 17:22:44 +0000]
DEVEL15-windows-buf-refcount-20080424

LICENSE MIT

Implement DEBUG_REFCOUNT refcount tracking code to debug refcount issues
in the cm_buf module.

Fix a refcount leak caused by buf_IncrSyncer() incorrectly removing cm_buf_t
objects from the dirty list.

Fix the dumping of the dirty list to actually dump the dirty list.

(cherry picked from commit 088eb9158bf553d5b3dcea8f37d3bec439419d44)

13 years agoDEVEL15-windows-cellservdb-20080423
Jeffrey Hutzelman [Thu, 24 Apr 2008 04:08:31 +0000]
DEVEL15-windows-cellservdb-20080423

LICENSE IPL

latest greatest CellServDB

(cherry picked from commit 6143f2eb5c85f5c428578b60b3b8892321f7fbe1)

13 years agoDEVEL15-vol-setaside-one-fd-per-thread-20080423
Dan Hyde [Wed, 23 Apr 2008 19:57:13 +0000]
DEVEL15-vol-setaside-one-fd-per-thread-20080423

LICENSE IPL10
FIXES 87977

kids these days share badly. give each one their own fd otherwise they fight.

(cherry picked from commit 6809ba29fadbd0d84ff6a9bf2ccd909b2f54b186)

13 years agoDEVEL15-linux-fedora9-packaging-updates-20080422
Simon Wilkinson [Tue, 22 Apr 2008 13:50:38 +0000]
DEVEL15-linux-fedora9-packaging-updates-20080422

LICENSE IPL10

make build system able to cope with fedora 9

(cherry picked from commit d18f5104f57152dfb6d1125a2d3a8bf17a266f44)

13 years agoDEVEL15-openafs-1-5-35-20080421 openafs-devel-1_5_35
Derrick Brashear [Mon, 21 Apr 2008 20:48:41 +0000]
DEVEL15-openafs-1-5-35-20080421

LICENSE IPL10

make 1.5.35

13 years agoDEVEL15-libadmin-vos-irix-fixes-20080421
Derrick Brashear [Mon, 21 Apr 2008 20:40:01 +0000]
DEVEL15-libadmin-vos-irix-fixes-20080421

LICENSE IPL10

irix is cranky about missing prototypes

(cherry picked from commit 8bf142ae2a979447b259b0945187379812883bee)

13 years agoDEVEL15-ppc64-linux-26-thread-info-20080418
Derrick Brashear [Mon, 21 Apr 2008 19:59:34 +0000]
DEVEL15-ppc64-linux-26-thread-info-20080418

LICENSE IPL10

thread_info is dead

(cherry picked from commit 2a0f8bfda8b00de429cb15481e3e20d5d140a81c)

13 years agoDEVEL15-dafs-irix-prec99-20080421
Derrick Brashear [Mon, 21 Apr 2008 19:55:48 +0000]
DEVEL15-dafs-irix-prec99-20080421

LICENSE IPL10

unless/until we build c99, irix needs this

(cherry picked from commit 92b5087b42653d5358433c06d60f3adc67a11015)

13 years agoDEVEL15-windows-readme-20080421
Jeffrey Altman [Mon, 21 Apr 2008 17:03:50 +0000]
DEVEL15-windows-readme-20080421

LICENSE MIT

update the build instructions

13 years agoDEVEL15-windows-notes-20080421
Jeffrey Altman [Mon, 21 Apr 2008 16:51:28 +0000]
DEVEL15-windows-notes-20080421

LICENSE MIT

1.5.35

13 years agoDEVEL15-windows-release-notes-20080421
Jeffrey Altman [Mon, 21 Apr 2008 16:44:06 +0000]
DEVEL15-windows-release-notes-20080421

LICENSE MIT

no changes.  just version number increase to 1.5.35

13 years agoDEVEL15-windows-cache-version-20080421
Jeffrey Altman [Mon, 21 Apr 2008 16:36:49 +0000]
DEVEL15-windows-cache-version-20080421

LICENSE MIT

increase the cache version because the cm_volume_t data structure was modified.

(cherry picked from commit a6ef7fed731628a5e5b8af62f7e986d6f7ccf444)

13 years agoDEVEL15-windows-version-20080421
Jeffrey Altman [Mon, 21 Apr 2008 16:27:55 +0000]
DEVEL15-windows-version-20080421

LICENSE MIT

1.5.3500.0 == 1.5.35

13 years agoDEVEL15-linux-fc9-20080419
Simon Wilkinson [Sat, 19 Apr 2008 22:55:44 +0000]
DEVEL15-linux-fc9-20080419

LICENSE IPL10

add in needed changes for fc9 kernels

(cherry picked from commit 0caa578a05150451c706c2ec6c3d095c671a7043)

13 years agoDEVEL15-klog-deal-with-aix-stupidity-20080418
Derrick Brashear [Fri, 18 Apr 2008 21:49:58 +0000]
DEVEL15-klog-deal-with-aix-stupidity-20080418

LICENSE IPL10

aix krb5 is broken. deal.

(cherry picked from commit 194067be89293d2f3ee4e791515c3ac020d3a551)

13 years agoDEVEL15-windows-volume-status-20080418
Jeffrey Altman [Fri, 18 Apr 2008 17:46:07 +0000]
DEVEL15-windows-volume-status-20080418

LICENSE MIT

A problem was discovered with cm_Analyze when serverRef instances are
marked offline.  cm_CheckOfflineVolume() was not resetting the serverRef
state.  cm_Analyze was also waiting in sleep calls when it was it is
clear that waiting would not result in a change of state.

cm_CheckOfflineVolume() was updated to always reset the serverRef state
and indicate that the volume should be considered online for the purpose
of retrying when the serverRef state is reset to srv_not_busy.

The problems identified in the MIT dev.mit.edu stress test environment
are resolved by these changes.

(cherry picked from commit c3eb805fc0309efca059a26b946eb4350417867e)

13 years agoDEVEL15-lwp-fix-protoize-20080418
Derrick Brashear [Fri, 18 Apr 2008 14:57:13 +0000]
DEVEL15-lwp-fix-protoize-20080418

LICENSE IPL10

make this cast correct.

(cherry picked from commit 8ea0eef2d6807d2ad51d29853335f115439e616c)

13 years agoDEVEL15-klog-aix-missing-symbols-20080418
Derrick Brashear [Fri, 18 Apr 2008 14:56:21 +0000]
DEVEL15-klog-aix-missing-symbols-20080418

LICENSE IPL10

aix's krb5 isn't heimdal or mit. it has dangling symbols. yay!

(cherry picked from commit f3056feadbd6866bddc20e040b5005baeb4c8cad)

13 years agoDEVEL15-linux-fakestat-avoid-mtpt-fillin-issue-20080415
Rainer Toebbicke [Tue, 15 Apr 2008 13:29:37 +0000]
DEVEL15-linux-fakestat-avoid-mtpt-fillin-issue-20080415

LICENSE IPL10
FIXES 93898

"It is actually the detection of a "cell" in the mount point string
which triggers the (loosely consistent) fakestat handling - it's
treated like a foreign cell and the mount point is never "completely
evaluated", i.e. logically replaced by the root directory of the
volume in question. As a result, callbacks are ignored, as they go
against the directory and not the mount point."

(cherry picked from commit 5f6b59f4601e9892e082a145a75e5872818d2e06)

13 years agoDEVEL15-windows-volume-20080414
Jeffrey Altman [Mon, 14 Apr 2008 23:44:03 +0000]
DEVEL15-windows-volume-20080414

LICENSE MIT

1. Convert cm_volume_t mutex to a rwlock (rw).

2. Don't hold cm_volume_t lock across RPCs

3. Add CM_VOLUMEFLAG_UPDATING_VL flag to prevent multiple threads
   from performing volume update calls at the same time on the same
   volume set.

4. Convert the .rw, .ro, and .bk cm_vol_state_t members of cm_volume_t
   to an array of cm_vol_state_t members indexed using the pre-existing
   RWVOL, ROVOL and BACKVOL preprocessor symbols.  This will permit
   consolidation of code and make it easier to add new volume types
   to the volume set in the future.

(cherry picked from commit 4440b9a61247a0da020042b4262695f0ef7d35de)

13 years agoDEVEL15-volser-prototype-include-fix-20080414
Derrick Brashear [Mon, 14 Apr 2008 21:25:52 +0000]
DEVEL15-volser-prototype-include-fix-20080414

LICENSE IPL10

so tvolser can build again

(cherry picked from commit 7936739f44372458fffb86058b3279d53995bd8b)

13 years agoDEVEL15-windows-freelance-20080411
Jeffrey Altman [Fri, 11 Apr 2008 18:46:24 +0000]
DEVEL15-windows-freelance-20080411

LICENSE MIT

Do not set the DV on the rootSCachep because doing so disables
the use of MergeStatus().

Do not remove the rootSCachep from the hash table because doing
so makes causes a second cm_scache_t object for that fid to be created
that is not the root.

(cherry picked from commit 7bfe889eb2e900719c70cd9fd29f82569ef4c04c)

13 years agoDEVEL15-vos-noresolve-option-20080410
Jeffrey Altman [Thu, 10 Apr 2008 18:52:26 +0000]
DEVEL15-vos-noresolve-option-20080410

LICENSE MIT

Add a -noresolve option to all vos commands.  This permits the display
of servers by IP address instead of DNS name.  Very useful when 127.0.0.1
is registered as the server address or when trying to eliminate
multi-homed addresses.

(cherry picked from commit 40d41c191bdb64256d4ccf6a5f1bab8124249e63)

13 years agoDEVEL15-void-return-casts-20080409
Simon Wilkinson [Wed, 9 Apr 2008 17:39:57 +0000]
DEVEL15-void-return-casts-20080409

LICENSE IPL10

do apporpriate casts for void returns to placate irix cc

(cherry picked from commit fab6653cf714ee6121cd7a22857f22f86a969559)

13 years agoDEVEL15-linux-nfstrans-avoid-access-deadlock-20080409
Jeffrey Hutzelman [Wed, 9 Apr 2008 14:39:19 +0000]
DEVEL15-linux-nfstrans-avoid-access-deadlock-20080409

LICENSE IPL10

avoid deadlocking in access while we're already holding locks as filldir is doing its work

(cherry picked from commit 7faf596b439cce022cfa4b3174cd17a1f0a0901e)

13 years agoDEVEL15-windows-nsis-vs2005-20080409
Asanka Herath [Wed, 9 Apr 2008 05:36:53 +0000]
DEVEL15-windows-nsis-vs2005-20080409

LICENSE MIT

Add support for building with VS2005 Runtime Libraries.
For the release build use the vcredist_<arch>.exe package.
For the debug build, build an msi from the merge modules
and then install that package.

(cherry picked from commit 494568ea929af96bf2b641e31b970cf695f13240)

13 years agoDEVEL15-redhat-spec-updates-20080408
Simon Wilkinson [Tue, 8 Apr 2008 19:36:10 +0000]
DEVEL15-redhat-spec-updates-20080408

LICENSE IPL10
FIXES 93616

update to correctly fill in specfile from source tree

(cherry picked from commit 1495088e4168a2ddf35135db872777b9e51823b3)

13 years agoDEVEL15-windows-strlwr-cellname-20080407
Jeffrey Altman [Mon, 7 Apr 2008 15:20:19 +0000]
DEVEL15-windows-strlwr-cellname-20080407

LICENSE MIT

cellnames are case insensitive and by convention they are lowercased.

(cherry picked from commit 0d2e36cac80d4c49db8e6b12988344107b7b1cd4)

13 years agoDEVEL15-windows-freelance-three-20080407
Jeffrey Altman [Mon, 7 Apr 2008 15:17:47 +0000]
DEVEL15-windows-freelance-three-20080407

LICENSE MIT

deliver notification events to watchers when adding mount points or
symlinks to the freelance root.afs volume.

(cherry picked from commit c108d606f59b6f5edcb863de3f3d9d7ecc44f04a)

13 years agoDEVEL15-windows-freelance-two-20080407
Jeffrey Altman [Mon, 7 Apr 2008 07:00:46 +0000]
DEVEL15-windows-freelance-two-20080407

LICENSE MIT

do not add the scp into the hash table twice.

(cherry picked from commit 8f57819a99273812c1d9ba7ec504689b05b83b14)

13 years agoDEVEL15-windows-freelance-20080407
Jeffrey Altman [Mon, 7 Apr 2008 06:48:19 +0000]
DEVEL15-windows-freelance-20080407

LICENSE MIT

When freelance symlinks and mountpoints were modified the target string
value was not being updated and the old data was not being invalidated.

(cherry picked from commit a642541b960f601d7396c28ca08ee7fd32975ce6)

13 years agoDEVEL15-linux-no-syscall-probe-with-no-init-mm-20080406
Marc Dionne [Mon, 7 Apr 2008 04:49:00 +0000]
DEVEL15-linux-no-syscall-probe-with-no-init-mm-20080406

LICENSE IPL10
FIXES 88000

no better idea for this, got to do something... push this for now

(cherry picked from commit c7ee4828e0d7c631fcd827d1a4c9af8bebb2fc2d)

13 years agoDEVEL15-windows-fid-hash-20080406
Jeffrey Altman [Sun, 6 Apr 2008 14:37:22 +0000]
DEVEL15-windows-fid-hash-20080406

LICENSE MIT

1. During the conversion to hash based FID comparisons the cm_GetSCache()
   lookups from the cm_ioctl.c (CM_IOCTL_QOPTS_HAVE_FID) and from FIDs
   returned by cm_LookupSearchProc() were broken.

2. In cm_LookupInternal(), ensure that an error is returned if rock.found
   is FALSE.  A failure to return an error was resulting in cm_GetSCache()
   producing an object for FIDs with vnode=0,unique=0

3. cm_GetSCache(), remove the warning check for vnode=0,unique=0

4. reapply the synchronization logic to async store requests in cm_SyncOp().
   The lack of synchronization is causing requests to complete out of order
   which is confusing the SMB redirector when under heavy load.

(cherry picked from commit ef19cea44c433ec7c1d214e8229b5fb4ea4b6ef6)

13 years agoDEVEL15-revert-ubik-changes-20080405
Derrick Brashear [Sat, 5 Apr 2008 19:03:24 +0000]
DEVEL15-revert-ubik-changes-20080405

LICENSE IPL10

these changes are unneeded; remove them

(cherry picked from commit 5dd7e65c6fd233e70d98b271cbf48ac2356efe0f)

13 years agoDEVEL15-volser-preclude-alternate-partition-clones-20080404
Kevin McBride [Fri, 4 Apr 2008 20:58:45 +0000]
DEVEL15-volser-preclude-alternate-partition-clones-20080404

LICENSE IPL10

patch based on work from shadow@dementia.org

the idea is to preclude multiple copies of a volume replica on a server,
as there's nothing to indicate which copy is correct and you can end up serving
stale data when you think you've just released and are serving something good

(cherry picked from commit 5640c4a1b5bea4e61ff2e6c2da02265c3e1ba612)

13 years agoDEVEL15-tubik-more-makefile-fun-20080404
Derrick Brashear [Fri, 4 Apr 2008 16:40:04 +0000]
DEVEL15-tubik-more-makefile-fun-20080404

LICENSE IPL10

kill off library name reuse

(cherry picked from commit e62727e328c8fad2a5b7b305745a248d9aabe19a)

13 years agoDEVEL15-tubik-vos-20080403
Kevin McBride [Thu, 3 Apr 2008 17:41:29 +0000]
DEVEL15-tubik-vos-20080403

LICENSE IPL10

build pthreaded vos

(cherry picked from commit 65e701fee4968b17066bb81e25b7adaa4024d4f3)

13 years agoDEVEL15-tubik-20080403
Derrick Brashear [Thu, 3 Apr 2008 15:20:53 +0000]
DEVEL15-tubik-20080403

LICENSE IPL10

avoid bogus dependency here too

(cherry picked from commit 5f67c569455f0597d42b5a2cccf89e470ed74c53)

13 years agoDEVEL15-tubik-one-more-20080402
Derrick Brashear [Thu, 3 Apr 2008 04:54:53 +0000]
DEVEL15-tubik-one-more-20080402

LICENSE IPL10

handle the rest of the issues

(cherry picked from commit bc1a7f1ee8fb12eeb7b54f89987e6f240627a579)

13 years agoDEVEL15-tubik-makefiles-fun-20080402
Derrick Brashear [Thu, 3 Apr 2008 02:20:44 +0000]
DEVEL15-tubik-makefiles-fun-20080402

LICENSE IPL10

these fixes brought to you by delicious beer.

(cherry picked from commit 33161360381d5ec67f884c087da26bd8009a0511)

13 years agoDEVEL15-tubik-makefile-fixes-20080402
Derrick Brashear [Wed, 2 Apr 2008 22:44:04 +0000]
DEVEL15-tubik-makefile-fixes-20080402

LICENSE IPL10

don't use gmake macros. ick

(cherry picked from commit ec02ff4342844c0d83ed3e9e5e27aef7c3af7204)

13 years agoDEVEL15-pthreaded-ubik-20080402
Steven Jenkins [Wed, 2 Apr 2008 21:06:31 +0000]
DEVEL15-pthreaded-ubik-20080402

LICENSE IPL10

add pthreaded ubik and various binaries

(cherry picked from commit 9fbbc3f50cc6f04796a8a444bc06acc66a9c55a6)

13 years agoDEVEL15-linux-flush-compare-20080402
Marc Dionne [Wed, 2 Apr 2008 19:21:49 +0000]
DEVEL15-linux-flush-compare-20080402

LICENSE IPL10

because i let this slip out like this. sigh.

(cherry picked from commit 93c326ef08cacc730470a7af75041d943825260b)

13 years agoDEVEL15-linux-flush-unlock-20080402
Harald Barth [Wed, 2 Apr 2008 14:10:00 +0000]
DEVEL15-linux-flush-unlock-20080402

LICENSE IPL10

because i let this slip out like this. sigh.

(cherry picked from commit 31d06c78b2f8ea14877bebf05097b33cc7e233f2)

13 years agoDEVEL15-salvageserver-init-handler-lock-20080401
Tom Keiser [Tue, 1 Apr 2008 21:29:12 +0000]
DEVEL15-salvageserver-init-handler-lock-20080401

LICENSE IPL10

init lock

(cherry picked from commit 5fe4de574dba287b8197263028ac49014955a483)

13 years agoDEVEL15-tsalvaged-clean-20080401
Derrick Brashear [Tue, 1 Apr 2008 21:17:27 +0000]
DEVEL15-tsalvaged-clean-20080401

LICENSE IPL10

make clean target work here

(cherry picked from commit 0faa1e0412c4cb22e6820473fa60cdea5d516550)

13 years agoDEVEL15-vol-osi-assert-20080401
Derrick Brashear [Tue, 1 Apr 2008 21:05:36 +0000]
DEVEL15-vol-osi-assert-20080401

LICENSE IPL10

undo the old windows afs assert undoing

(cherry picked from commit 48838bab91b50a5cf8cd3fb7d1902e6b7525bb86)

13 years agoDEVEL15-aklog-debug-text-dont-deref-null-ptr-20080401
Derrick Brashear [Tue, 1 Apr 2008 19:15:40 +0000]
DEVEL15-aklog-debug-text-dont-deref-null-ptr-20080401

LICENSE IPL10

yeah, this is sort of dumb

AUTHOR
LICENSE IPL10

====================
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.
====================
LICENSE IPL10

yeah, this is sort of dumb

(cherry picked from commit 695d1aa896cf38cfe9856ba1c45b6b3cca7565a2)

13 years agoDEVEL15-aklog-tidyup-20080401
Simon Wilkinson [Tue, 1 Apr 2008 18:44:40 +0000]
DEVEL15-aklog-tidyup-20080401

LICENSE IPL10
FIXES 92643

remove windows code (osolete) from unix branch
do other minor cleanup

(cherry picked from commit 3f54c934b9c933d5f34644a096c821375db17d97)

13 years agoDEVEL15-vlserver-checksignal-returns-voidstar-20080401
Derrick Brashear [Tue, 1 Apr 2008 17:51:31 +0000]
DEVEL15-vlserver-checksignal-returns-voidstar-20080401

LICENSE IPL10
FIXES 92662

CheckSignal returns void*, don't try to return an int directly.

(cherry picked from commit 94aa281af8c9cb74561a32a34cb7cd5b5da97cdf)

13 years agoDEVEL15-synvageserver-no-rx-server-refs-20080401
Tom Keiser [Tue, 1 Apr 2008 15:12:06 +0000]
DEVEL15-synvageserver-no-rx-server-refs-20080401

LICENSE IPL10
FIXES 92648

remove reference to rx service mutex and thread tracking. we need to ensure there's nothing else either but this is true now

(cherry picked from commit 9bdb8599c094821d0dfb29505cc63aa2a8460ae2)

13 years agoDEVEL15-man-page-vos-shadow-20080331
Jason Edgecombe [Tue, 1 Apr 2008 07:57:18 +0000]
DEVEL15-man-page-vos-shadow-20080331

LICENSE BSD

Man page for vos shadow.

(cherry picked from commit 4bcb7d01b2a8aa6ce53cf1c0b093058b4e5072f1)

13 years agoDEVEL15-test-updates-20080331
Mike Meffie [Mon, 31 Mar 2008 16:54:02 +0000]
DEVEL15-test-updates-20080331

LICENSE IPL10

update tests directory

(cherry picked from commit a26d7f1403d7d41d51836442a7a4759f4fc2bfe4)

13 years agoDEVEL15-windows-aklog-cross-cell-20080330
Jeffrey Altman [Sun, 30 Mar 2008 17:21:07 +0000]
DEVEL15-windows-aklog-cross-cell-20080330

LICENSE MIT

report the new ID after a successful cross cell registration

(cherry picked from commit 99a29903c8b77e2e3260bf06d913367fadbc4fed)

13 years agoDEVEL15-windows-aklog-robustness-20080330
Jeffrey Altman [Sun, 30 Mar 2008 15:13:17 +0000]
DEVEL15-windows-aklog-robustness-20080330

LICENSE MIT

Fix the order of parameters to strncpy

Add error checking to all krb5 functions.

Initialize the global ccache variable to NULL.

(cherry picked from commit 2169423f6821672be5506ab388d6faa0ed37f0d0)

13 years agoDEVEL15-windows-vlserver-makefile-20080329
Jeffrey Altman [Sun, 30 Mar 2008 05:17:15 +0000]
DEVEL15-windows-vlserver-makefile-20080329

LICENSE MIT

make references to INCFILEDIR consistent

(cherry picked from commit aff582765a8af5ce8743e28ee4d33bf8e4396d63)

13 years agoDEVEL15-windows-use-client-realm-for-tokens-20080329
Jeffrey Altman [Sun, 30 Mar 2008 04:36:55 +0000]
DEVEL15-windows-use-client-realm-for-tokens-20080329

LICENSE MIT

Two recent changes to the AFS/Kerberos landscape have been causing
problems for aklog and related modules.  First, the support for multiple
local realms for the cell has broken the pts auto-registration code
when the realm used for the token acquisition does not match the
realm belonging to the selected vldb server.  Second, Kerberos referrals
prevents detection of the realm of the vldb server.

This commit adds a new method of searching for the afs service principal.
The first attempt is for afs/<cell>@<CLIENT-REALM>.  If found, the
<CLIENT-REALM> is used as the realm of the cell.

The patch adds error handling for KRB5_ERR_HOST_REALM_UNKNOWN which is
returned when krb5_get_host_realm() can't determine the realm.

Duplicate queries are also avoided and copy_realm_of_ticket() is
properly employed.

(cherry picked from commit c6897cee01063d6d6ad302e6527794130f007d74)

13 years agoDEVEL15-nbsd-update-20080328
Tracy Di Marco White [Fri, 28 Mar 2008 17:12:40 +0000]
DEVEL15-nbsd-update-20080328

LICENSE IPL10
FIXES 92073

fix nbsd30 param file, add nbsd40 param file

(cherry picked from commit 07d806845e1bf790c8cbe8e6a711e99e0958235b)

13 years agoDEVEL15-windows-freelance-20080328
Jeffrey Altman [Fri, 28 Mar 2008 14:02:54 +0000]
DEVEL15-windows-freelance-20080328

LICENSE MIT

1. return CM_ERROR_NOSUCHFILE when attempting to remove a symlink or
   mount point that does not exist

2. When adding a new symlink or mountpoint, be sure to increment the
   number of local mount points

3. When re-initializing the fake root.cell the unique value is always
   one, not two.  Use cm_FidCmp() to compare fids.  The use of the
   value two prevented the invalidation of existing values because all
   of the directory entries are created with the unique set to one.

(cherry picked from commit b2e3e6427522c92ce4132abf2ca08ce18599b16b)

13 years agoDEVEL15-krb5-klog-solaris-krb-20080327
Douglas Engert [Thu, 27 Mar 2008 18:48:46 +0000]
DEVEL15-krb5-klog-solaris-krb-20080327

LICENSE IPL10
FIXES 91573

no krb5_mcc_ops on solaris krb5. cope

(cherry picked from commit e0021180665fa75af88c3ebf0d7f58f93918ceab)

13 years agoDEVEL15-ptutil-swap-epoch-version-as-needed-20080327
Derrick Brashear [Thu, 27 Mar 2008 18:40:12 +0000]
DEVEL15-ptutil-swap-epoch-version-as-needed-20080327

LICENSE IPL10
FIXES 90869

do byteswapping of epoch and version so we print and compare sensibly

(cherry picked from commit 87e959e87df52e026ffcb2de8ecfcbf4889f7bd8)

13 years agoDEVEL15-volser-higher-thread-limit-20080327
Derrick Brashear [Thu, 27 Mar 2008 16:20:39 +0000]
DEVEL15-volser-higher-thread-limit-20080327

LICENSE IPL10

with tvolser we can increase this and actually *see benefit*, so...

(cherry picked from commit 7c05046f1f097a6ded89a5928ea48aca224f520a)

13 years agoDEVEL15-dafs-vshutdown-match-id-to-params-20080327
Alistair Ferguson [Thu, 27 Mar 2008 16:16:12 +0000]
DEVEL15-dafs-vshutdown-match-id-to-params-20080327

LICENSE IPL10

we track shutdown thread parms as index rather than device; deal correctly

(cherry picked from commit 6c9647cc614cfdf9b2c9cc8adb7b3c6b741e1f4e)

13 years agoDEVEL15-linux-dont-flush-writes-on-exec-20080326
Chas Williams [Wed, 26 Mar 2008 04:17:18 +0000]
DEVEL15-linux-dont-flush-writes-on-exec-20080326

LICENSE IPL10
FIXES 17509

because of when fds are NULL'd in file_table this should work except when one process has the same file open for read and for write in different FDs; otherwise,
 the last write fd to close collects the error.

(cherry picked from commit 1cf6c12a075cb0705cd8680291d78d215d4fe996)

13 years agoDEVEL15-linux-restore-iget-cachemgr-20080326
Marc Dionne [Wed, 26 Mar 2008 04:10:39 +0000]
DEVEL15-linux-restore-iget-cachemgr-20080326

LICENSE IPL10
FIXES 88000

restore the d_alloc_anon we need

(cherry picked from commit 3f8be40525ae84980f1519e60f4b9f73bd4cc676)

13 years agoDEVEL15-windows-vlserver-include-path-20080323
Jeffrey Altman [Sun, 23 Mar 2008 14:11:57 +0000]
DEVEL15-windows-vlserver-include-path-20080323

LICENSE MIT

Fix the INCFILEDIR definition so that the INCFILES actually exist

(cherry picked from commit f3776341151ad2c9b439567411e908836b16d415)