openafs.git
16 years agowindows-cellservdb-20071103
Jeffrey Hutzelman [Sat, 3 Nov 2007 15:59:12 +0000]
windows-cellservdb-20071103

GCO Public CellServDB 25 Oct 2007

16 years agowindows-install-des-prototypes-20071102
Jeffrey Altman [Sat, 3 Nov 2007 01:05:53 +0000]
windows-install-des-prototypes-20071102

required by auth-getkey-prototyping-20071102

16 years agoauth-getkey-prototyping-20071102
Derrick Brashear [Fri, 2 Nov 2007 18:01:32 +0000]
auth-getkey-prototyping-20071102

make this sane

16 years agoviced-new-host-reup-h-lock-20071101
Ken Raeburn [Thu, 1 Nov 2007 19:12:38 +0000]
viced-new-host-reup-h-lock-20071101

FIXES 69384

don't leave lock dropped

16 years agosolaris-cc-deal-with-new-64-bit-arch-specifier-20071101
Dale Ghent [Thu, 1 Nov 2007 17:05:33 +0000]
solaris-cc-deal-with-new-64-bit-arch-specifier-20071101

FIXES 75908

switch from -xmarch=amd64 to -m64

16 years agosolaris-nevada-update-20071101
Dale Ghent [Thu, 1 Nov 2007 16:57:26 +0000]
solaris-nevada-update-20071101

FIXES 75896

update for current solaris nevada

16 years agoauth-cleanup-20071101
Simon Wilkinson [Thu, 1 Nov 2007 16:09:22 +0000]
auth-cleanup-20071101

kill warnings in auth directory

16 years agorxgen-ansi-again-20071101
Simon Wilkinson [Thu, 1 Nov 2007 15:58:59 +0000]
rxgen-ansi-again-20071101

another look at lettign rxgen do ansification

16 years agodafs-salvage-deal-with-clones-20071101
Tom Keiser [Thu, 1 Nov 2007 14:59:38 +0000]
dafs-salvage-deal-with-clones-20071101

FIXES 75591

cope with trying to salvage a clone correctly

16 years agolinux-afs-unlinked-file-dentry-flagging-20071031
Chaskiel M Grundman [Thu, 1 Nov 2007 03:23:09 +0000]
linux-afs-unlinked-file-dentry-flagging-20071031

fix case:
 perl -e 'open FOO, ">foo"; unlink(foo); @stat = stat(FOO); print "inode for foo(unlinked) is $stat[1]\n"; system("/bin/ls -ali | grep $stat[1]");'

16 years agomore-anonymous-pointer-cleanup-20071031
Simon Wilkinson [Wed, 31 Oct 2007 22:31:38 +0000]
more-anonymous-pointer-cleanup-20071031

more char*->void* conversion

16 years agowindows-btree-20071031
Jeffrey Altman [Wed, 31 Oct 2007 15:33:27 +0000]
windows-btree-20071031

Add additional validation and error handling code after each call to
getSlot().  If an invalid slot is returned, return NONODE.  If the
invalid slot is returned when extracting a data node, invalidate the
tree.

Modify compareKeys() to always perform a case-insensitive comparison
and only perform a case sensistive comparison if the case-insensitive
one matches.  This ensures the ordering is consistently reported.

Add lock assertions to ensure that all calls are being performed with
the correct locks being held.  There have been some crash reports that
provide stack data that does not appear to be possible unless there is
a race.  However, there are no obvious locations where the race is
taking place and the test suite indicates that all of the correct locks
are being held. We shall see what happens in the field.

For consistency replace all calls to findKey in which the range is
(1,numentries) with calls to getSlot().

Optimize the depth search loop by testing the slot value in the for
statement instead of forcing the loop to be broken later.

16 years agowindows-begindirop-20071031
Jeffrey Altman [Wed, 31 Oct 2007 15:23:42 +0000]
windows-begindirop-20071031

Avoid a race condition in cm_BeginDirOp() caused by the failure to
hold the cm_scache_t mutex while copying status data from the scp to
the dirop

16 years agowindows-cm_analyze-ewouldblock-20071031
Jeffrey Altman [Wed, 31 Oct 2007 15:20:14 +0000]
windows-cm_analyze-ewouldblock-20071031

Add support for EWOULDBLOCK to cm_Analyze.  If the file server returns
EWOULDBLOCK, retry the request every two seconds for up to the RDRtimeout.

16 years agowindows-uaewouldblock-20071031
Jeffrey Altman [Wed, 31 Oct 2007 15:18:10 +0000]
windows-uaewouldblock-20071031

Reorganize the order of the includes to ensure that EWOULDBLOCK is not
assigned the same value as EIO.

16 years agoutil-cleanup-20071031
Simon Wilkinson [Wed, 31 Oct 2007 12:36:33 +0000]
util-cleanup-20071031

kill the warnings in src/util

16 years agoopenafs-bozo-varargs-20071031
Simon Wilkinson [Wed, 31 Oct 2007 04:20:33 +0000]
openafs-bozo-varargs-20071031

convert bozo to be prototypable

16 years agoopenafs-void-star-pointers-20071031
Simon Wilkinson [Wed, 31 Oct 2007 04:05:27 +0000]
openafs-void-star-pointers-20071031

convert anonymous pointers from char* to void*

16 years agoopenafs-string-header-cleanup-20071030
Simon Wilkinson [Tue, 30 Oct 2007 15:10:37 +0000]
openafs-string-header-cleanup-20071030

include only what we need

16 years agoredhat-packaging-update-20071029
Simon Wilkinson [Tue, 30 Oct 2007 03:53:08 +0000]
redhat-packaging-update-20071029

update rpm packaging stuff for what shipped in 1.4.5

16 years agoupdate-bsd-readme-20071026
Jim Rees [Fri, 26 Oct 2007 17:41:54 +0000]
update-bsd-readme-20071026

Update build info for Open/FreeBSD.

16 years agoredhat-fedora-build-driver-20071026
Simon Wilkinson [Fri, 26 Oct 2007 17:20:22 +0000]
redhat-fedora-build-driver-20071026

driver script for fedora builds

16 years agoredhat-spec-fix-hugemem-modpath-20071026
Derrick Brashear [Fri, 26 Oct 2007 17:09:40 +0000]
redhat-spec-fix-hugemem-modpath-20071026

make this match other kernsls

16 years agorelease-notes-1-4-5-20071025
Derrick Brashear [Thu, 25 Oct 2007 06:19:17 +0000]
release-notes-1-4-5-20071025

release notes for 1.4.5

16 years agowindows-aclent-remove-extraneous-code-20071024
Jeffrey Altman [Thu, 25 Oct 2007 05:47:23 +0000]
windows-aclent-remove-extraneous-code-20071024

Microsoft provided a dump showing cm_data.aclLRUEndp == NULL in
GetFreeACLEnt().  Couldn't find any reason why that would be true.
However, there is extraneous code that manipulates the value of
aclLRUEndp when it should be left to osi_QRemoveHT() to update
the pointer values.

Also add an additional explicit test for aclLRUEndp == NULL in
GetFreeACLEnt().

16 years agomacos-installer-crap-20071025
Derrick Brashear [Thu, 25 Oct 2007 04:42:07 +0000]
macos-installer-crap-20071025

put it back the way it was

16 years agolinux-nfstrans-20071024
Simon Wilkinson [Wed, 24 Oct 2007 16:44:26 +0000]
linux-nfstrans-20071024

FIXES 75233

better track changes so we can build an nfstranslator module for new linux kernels

16 years agocblater-bitmask-bug-20071024
Dan Hyde [Wed, 24 Oct 2007 16:33:31 +0000]
cblater-bitmask-bug-20071024

FIXES 75304

add missing ()

16 years agodarwin-allow-finder-to-unlink-rsrc-files-20071024
Derrick Brashear [Wed, 24 Oct 2007 06:40:12 +0000]
darwin-allow-finder-to-unlink-rsrc-files-20071024

finder tries hard to clean up resource fork files it creates when you drop something in a dropbox that don't then get updates.

let it think it won

16 years agofinder-and-tiger-cp-deal-with-dropboxes-20071024
Chaskiel M Grundman [Wed, 24 Oct 2007 06:25:46 +0000]
finder-and-tiger-cp-deal-with-dropboxes-20071024

FIXES 69363

this way copying finishes

16 years agowindows-wix-nsis-uninstall-20071023
Jeffrey Altman [Wed, 24 Oct 2007 04:49:07 +0000]
windows-wix-nsis-uninstall-20071023

At somepoint windows installer broke the use of the 'file' RegistrySearch
type.  Instead of returning the full file name path we are only getting
the directory.  Switch to using the 'raw' type since we know we aren't
prefixing the file name with a hash.

16 years agoafs-shutdown-free-resources-20071023
Mike Meffie [Tue, 23 Oct 2007 23:36:06 +0000]
afs-shutdown-free-resources-20071023

try hard to free resources

16 years agodarwin-uio-64-20071023
Derrick Brashear [Tue, 23 Oct 2007 18:37:58 +0000]
darwin-uio-64-20071023

use the 64 bit awareness in the KPI

16 years agopod-man-fs-uuid-20071023
Jason Edgecombe [Tue, 23 Oct 2007 17:59:16 +0000]
pod-man-fs-uuid-20071023

FIXES 75175

add man page for fs uuid

16 years agodarwin-afsd-cleanup-power-notification-nicely-20071023
Derrick Brashear [Tue, 23 Oct 2007 15:28:22 +0000]
darwin-afsd-cleanup-power-notification-nicely-20071023

don't drop a core on shutdown

16 years agovos-restore-check-doesnt-apply-to-pipes-20071023
Derrick Brashear [Tue, 23 Oct 2007 15:19:07 +0000]
vos-restore-check-doesnt-apply-to-pipes-20071023

as pointed out by hung pham

16 years agodarwin-modify-cp-workaround-20071022
Derrick Brashear [Tue, 23 Oct 2007 01:30:03 +0000]
darwin-modify-cp-workaround-20071022

more work will be needed to fool finder into just copying the d**n file, since
it's checking too hard.

16 years agochange-default-probeint-for-rfc4787-20071022
Derrick Brashear [Tue, 23 Oct 2007 01:02:39 +0000]
change-default-probeint-for-rfc4787-20071022

an RFC compliant port mapping device guarantees less time than we previously probed.

16 years agofssync-poll-interface-20071022
Simon Wilkinson [Mon, 22 Oct 2007 18:29:28 +0000]
fssync-poll-interface-20071022

FIXES 74708

support a poll interface

otherwise, the fd setsize can screw with us

16 years agocpp-avoid-cc-E-20071022
Simon Wilkinson [Mon, 22 Oct 2007 14:21:26 +0000]
cpp-avoid-cc-E-20071022

finding the right cpp is good. having it not be cc -E is better

16 years agowindows-ntstatus-warnings-20071021
Jeffrey Altman [Mon, 22 Oct 2007 04:32:09 +0000]
windows-ntstatus-warnings-20071021

suppress warnings caused by multiple definitions of NTSTATUS values

16 years agowindows-del-symlink-20071021
Jeffrey Altman [Mon, 22 Oct 2007 04:27:14 +0000]
windows-del-symlink-20071021

The cmd.exe "del" command operates by opening a file and then setting
the file disposition to delete on close followed by closing the file.

When the filename is a symlink, the smb_fid_t scp refers to the final
destination object and not the symlink.  In smb_CloseFid() the correct
object would be removed from the directory by name, but the wrong cm_scache_t
would be marked deleted.  This would result in subsequent references to
the target file being considered invalid.

Fix it by looking up the cm_scache_t of the symlink prior to performing
the deletion.

16 years agowindows-registry-OfflineReadOnlyIsValid-20071019
Jeffrey Altman [Sat, 20 Oct 2007 04:58:40 +0000]
windows-registry-OfflineReadOnlyIsValid-20071019

Add registry value "OfflineReadOnlyIsValid" option.  When set to non-zero
value, an offline or down read only volume will be treated as if it has
a valid callback even if it doesn't.

16 years agocpp-autoconf-located-20071019
Arno Schuring [Sat, 20 Oct 2007 01:20:51 +0000]
cpp-autoconf-located-20071019

let autoconf find cpp

16 years agodarwin80-dnlc-dont-return-deadvnodes-20071019
Derrick Brashear [Fri, 19 Oct 2007 21:37:15 +0000]
darwin80-dnlc-dont-return-deadvnodes-20071019

FIXES 41550

obvious. if we are going to be like GetVCache, we should

16 years agowindows-pioctl-parse-path-20071018
Jeffrey Altman [Fri, 19 Oct 2007 05:25:39 +0000]
windows-pioctl-parse-path-20071018

cm_ParseIoctlPath() moves the ioctlp->inDatap past the ioctl path
to the next data value if there is one.  Make sure we move the pointer
before we alter the contents of the buffer.  Otherwise, the caller
uses the wrong string for its next data blob.

16 years agowindows-pioctl-error-codes-20071018
Jeffrey Altman [Thu, 18 Oct 2007 23:06:35 +0000]
windows-pioctl-error-codes-20071018

add knowledge of new cm error codes

16 years agowindows-freelance-symlink-20071018
Jeffrey Altman [Thu, 18 Oct 2007 06:27:40 +0000]
windows-freelance-symlink-20071018

More tweaking of the IoctlPathParsing code.

Add special handling for many more pioctl operations performed on
the Freelance root.afs volume.

Add informative error values for symlink freelance operations.

16 years agowindows-smb-20071018
Jeffrey Altman [Thu, 18 Oct 2007 06:24:39 +0000]
windows-smb-20071018

refactoring, suppression of duplicate status value definition warnings,
and removal of old pattern matching code

16 years agowindows-afskfw-20071018
Jeffrey Altman [Thu, 18 Oct 2007 06:20:08 +0000]
windows-afskfw-20071018

Add SOFTWARE\OpenAFS\Client  AcceptDottedPrincipalNames registry
value to control whether the use of Kerberos v5 principal names
with dots in the first component are permitted.

16 years agoredhat-spec-require-bison-flex-20071018
Derrick Brashear [Thu, 18 Oct 2007 05:11:16 +0000]
redhat-spec-require-bison-flex-20071018

require everything we need to build

16 years agoredhat-packaging-updates-20071018
Derrick Brashear [Thu, 18 Oct 2007 05:02:42 +0000]
redhat-packaging-updates-20071018

make the buildall script know centos and new fedora; make the spec file include it

16 years agodarwin-fsref-other-races-20071017
Derrick Brashear [Wed, 17 Oct 2007 21:32:38 +0000]
darwin-fsref-other-races-20071017

FIXES 41550

not clear this is exactly correct but testing seems to vet it.

16 years agodarwin-finalizevnode-get-vnode-inside-lock-20071017
Chaskiel M Grundman [Wed, 17 Oct 2007 18:16:19 +0000]
darwin-finalizevnode-get-vnode-inside-lock-20071017

FIXES 41550

the ovp initialization should happen while we have the vnode locked, regardless.

16 years agoafs-shutdown-free-resources-20071017
Derrick Brashear [Wed, 17 Oct 2007 15:32:50 +0000]
afs-shutdown-free-resources-20071017

try hard to free resources

16 years agodarwin90-disk-cache-20071016
Derrick Brashear [Wed, 17 Oct 2007 04:46:19 +0000]
darwin90-disk-cache-20071016

FIXES 43303

don't break memcache, also

16 years agotranslator-sysnamelist-bugfix-20071016
Douglas Engert [Tue, 16 Oct 2007 22:55:38 +0000]
translator-sysnamelist-bugfix-20071016

FIXES 74478

fix sysname list support bug when the list is empty

16 years agowindows-fs-uuid-20071016
Jeffrey Altman [Tue, 16 Oct 2007 18:24:31 +0000]
windows-fs-uuid-20071016

do not require afs client administrator access to display the current
Uuid value.

16 years agofs-uuid-20071016
Jason Edgecombe [Tue, 16 Oct 2007 18:18:28 +0000]
fs-uuid-20071016

FIXES 74473

Change "fs uuid" to "fs uuid -generate" where "-generate" is a
required parameter.

This command requires root access to execute it.

Unlike the Windows variant, the UNIX version does not yet support displaying
the current UUID to the user.  When this functionality is added, "-generate"
should become an optional parameter.

16 years agorestore-20071016
Jeffrey Altman [Tue, 16 Oct 2007 17:55:10 +0000]
restore-20071016

FIXES 74531

afs_com_err() was called without specifying the error code.

16 years agodarwin-access-avoid-confusing-cp-20071015
Derrick Brashear [Mon, 15 Oct 2007 20:19:03 +0000]
darwin-access-avoid-confusing-cp-20071015

FIXES 69363

it wants EPERM; EACCES, and it just retries forever.

"no means no. now go away"

16 years agodarwin-install-fstrace-catalog-20071015
Derrick Brashear [Mon, 15 Oct 2007 18:23:54 +0000]
darwin-install-fstrace-catalog-20071015

fstrace needs the message catalog to be useful. install it.

16 years agoredhat-spec-no-kdump-really-20071015
Derrick Brashear [Mon, 15 Oct 2007 18:00:58 +0000]
redhat-spec-no-kdump-really-20071015

make really, really sure it's not there

16 years agovolser-return-extended-stats-to-callers-20071015
Rainer Toebbicke [Mon, 15 Oct 2007 15:13:42 +0000]
volser-return-extended-stats-to-callers-20071015

FIXES 74398

make sure when we roll over stats we return the new values

16 years agowindows-clean-eventlog-20071014
Jeffrey Altman [Mon, 15 Oct 2007 04:18:30 +0000]
windows-clean-eventlog-20071014

delete AFS_component_version_number.h in the lang subdirectory

16 years agowindows-pioctl-two-20071014
Jeffrey Altman [Mon, 15 Oct 2007 03:32:12 +0000]
windows-pioctl-two-20071014

Two more locations where an intermediary path of length zero should
be ignored.

16 years agowindows-pioctl-20071014
Jeffrey Altman [Sun, 14 Oct 2007 06:54:06 +0000]
windows-pioctl-20071014

do not parse an intermediary path if the path length is 0

16 years agorpm-kmodule-patch-20071012
Derek Atkins [Fri, 12 Oct 2007 05:28:55 +0000]
rpm-kmodule-patch-20071012

install the openafs module in the kernel module area, in rpms

16 years agogcc42-integer-overflow-fix-20071007
Marc Dionne [Thu, 11 Oct 2007 18:44:53 +0000]
gcc42-integer-overflow-fix-20071007

fix issue where gcc 4.2 compiled kernel modules were breaking on linux

====================
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.
====================

with fix from Hans-Werner Pulsen

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

with fix from Hans-Werner Pulsen

16 years agomacos-version-generation-20071011
Derrick Brashear [Thu, 11 Oct 2007 17:51:56 +0000]
macos-version-generation-20071011

make release time simpler

16 years agorpm-files-update-20071011
Simon Wilkinson [Thu, 11 Oct 2007 17:38:56 +0000]
rpm-files-update-20071011

updates for the specfile and rpm generation

16 years agowindows-freelance-pioctl-two-20071010
Jeffrey Altman [Wed, 10 Oct 2007 19:50:25 +0000]
windows-freelance-pioctl-two-20071010

explicitly process the last component of the path separately and do not
follow symlinks or chase mountpoints on the last component.

16 years agodarwin90-disk-cache-20071010
Derrick Brashear [Wed, 10 Oct 2007 18:38:45 +0000]
darwin90-disk-cache-20071010

FIXES 43303

support disk cache in darwin90

16 years agowindows-build-xstat-20071010
Jeffrey Altman [Wed, 10 Oct 2007 17:05:47 +0000]
windows-build-xstat-20071010

build xstat directory

16 years agowindows-clean-eventlog-20071010
Jeffrey Altman [Wed, 10 Oct 2007 16:43:30 +0000]
windows-clean-eventlog-20071010

clean eventlog/lang/AFS_component_version_number.h

16 years agowindows-installer-wix-20071010
Jeffrey Altman [Wed, 10 Oct 2007 16:37:25 +0000]
windows-installer-wix-20071010

Later versions of the wix toolset require that all variables be defined
with a value.

16 years agowindows-freelance-pioctl-20071010
Jeffrey Altman [Wed, 10 Oct 2007 16:20:29 +0000]
windows-freelance-pioctl-20071010

pioctl commands were failing with CM_ERROR_NOSUCHCELL because the
Freelance cell object was not being added to the Cell Name and ID
hash tables.

ParseIoctlPath was not consistently returning the FID of the specified
symlink or mount point object.  The was still one more case in which
CM_FLAG_FOLLOW was being specified left over from
windows-parse-ioctl-path-20070629.  Changed the use of CM_FLAG_FOLLOW
into a 'follow' variable to ease testing and permit easy migration to
adding a 'follow' parameter to the function.

RXAFS_GetVolumeStatus is always going to fail for Freelance volumes.
Fake the response data.

UpdateVolume() was performing the wrong test for Freelance volumes.
Freelance volumes should not be marked offline even though they have
no servers.

16 years agodarwin90-debug-harder-20071009
Derrick Brashear [Tue, 9 Oct 2007 22:28:18 +0000]
darwin90-debug-harder-20071009

stab the stabs

16 years agotaking-rx-back-into-the-20th-century-20061228
Derrick Brashear [Tue, 9 Oct 2007 22:26:24 +0000]
taking-rx-back-into-the-20th-century-20061228

ken's patch is now on the ipv6 branch

16 years agodarwin90-dsym-and-ppc-cputype-support-20071009
Derrick Brashear [Tue, 9 Oct 2007 16:58:47 +0000]
darwin90-dsym-and-ppc-cputype-support-20071009

build for all ppc cputypes. also, support dsym debug info instead of stabs, when appropriate

16 years agosol10-abstraction-20071005
Derrick Brashear [Fri, 5 Oct 2007 23:16:36 +0000]
sol10-abstraction-20071005

abstract things as needed

16 years agodafs-listen-after-restoring-state-20071004
Tom Keiser [Fri, 5 Oct 2007 04:30:19 +0000]
dafs-listen-after-restoring-state-20071004

don't start rx server until the previous state has been loaded

16 years agovol-salvager-avoid-logging-to-nothing-20071004
Derrick Brashear [Fri, 5 Oct 2007 04:22:27 +0000]
vol-salvager-avoid-logging-to-nothing-20071004

check for logFile before using

16 years agolinux-2623-support-20071004
Marc Dionne [Fri, 5 Oct 2007 04:07:27 +0000]
linux-2623-support-20071004

FIXES 70773

update for support for 2.6.23

16 years agosolaris10-network-updates-20071004
Dale Ghent [Fri, 5 Oct 2007 03:22:11 +0000]
solaris10-network-updates-20071004

FIXES 72759

This patch removes use of the non-Public ILL structures by OpenAFS to
gather network interface information in order to make RX packet size
and server locality decisions.

16 years agowindows-btree-rename-case-only-20071004
Jeffrey Altman [Thu, 4 Oct 2007 15:11:07 +0000]
windows-btree-rename-case-only-20071004

Attempts to rename by case only would fail when there are more than
one ambiguous matches.  The list of data nodes would be corrupted
and future tree manipulations could result in a crash.

16 years agolibadmin-minimize-dns-lookups-20071003
Jeffrey Altman [Wed, 3 Oct 2007 16:27:53 +0000]
libadmin-minimize-dns-lookups-20071003

Some cells have a large number of servers listed in the vldb that
no longer exist and/or do not have PTR records in DNS.  The algorithm
used to obtain the "server names" from IP addresses involves opening
a cell handle, creating an iterator of all servers, and then iterating
over them until a match is found for the IP address.  This is done for
every server reference.  Each time the iterator was constructed the
vldb would be queried and gethostbyaddr() would be called on every listed
ip address.

This patch stores a copy of the server list in the cell handle along with
an expiration time generated from a 10 minute TTL.  This significant
reduces but does not eliminate the number of times that the caller is
blocked on reverse dns lookups.

16 years agowindows-tviced-resfile-20071003
Jeffrey Altman [Wed, 3 Oct 2007 16:17:43 +0000]
windows-tviced-resfile-20071003

Add explicit build rules for TVICEDRES since we do not
build the non-threaded file server on Windows

16 years agodafs-cbstate-tracking-fix-20071001
Tom Keiser [Tue, 2 Oct 2007 01:14:08 +0000]
dafs-cbstate-tracking-fix-20071001

fix this the rest of the places it happens

16 years agowindows-lanahelper-20071001
Jeffrey Altman [Mon, 1 Oct 2007 21:02:03 +0000]
windows-lanahelper-20071001

uninitialized variable in lana_IsLoopback

16 years agovollistpart-init-error-20070927
Dan Hyde [Fri, 28 Sep 2007 00:15:30 +0000]
vollistpart-init-error-20070927

FIXES 72144

correct an initialization failure in VolListPartitions

16 years agodafs-avoid-volhdr-leak-20070927
Tom Keiser [Thu, 27 Sep 2007 17:30:32 +0000]
dafs-avoid-volhdr-leak-20070927

when there's an error don't leak the loaded header; release it

16 years agoreadme-nt-20070920
Jeffrey Altman [Thu, 20 Sep 2007 12:56:54 +0000]
readme-nt-20070920

update wix and nsis instructions

16 years agowindows-version-info-20070920
Jeffrey Altman [Thu, 20 Sep 2007 12:44:23 +0000]
windows-version-info-20070920

add version info to the threaded version of butc.exe that we ship

16 years agokrb5-aklog-kimpersonate-conditionalize-20070920
Derrick Brashear [Thu, 20 Sep 2007 05:52:23 +0000]
krb5-aklog-kimpersonate-conditionalize-20070920

make kimpersonate conitional

16 years agodafs-loadvolumeheader-before-we-need-it-20070920
Derrick Brashear [Thu, 20 Sep 2007 05:01:44 +0000]
dafs-loadvolumeheader-before-we-need-it-20070920

make sure we have a header instead of potentially dereffing null

16 years agowindows-fs-memdump-20070919
Jeffrey Altman [Thu, 20 Sep 2007 04:54:14 +0000]
windows-fs-memdump-20070919

add directory lookup stats to "fs memdump" output

16 years agowindows-hex-dump-20070919
Jeffrey Altman [Wed, 19 Sep 2007 22:27:53 +0000]
windows-hex-dump-20070919

remove the broken percent quoting from OutputDebugHexDump as it is
not required.  We never use the resulting string as a print format string.

16 years agowindows-version-info-20070919
Jeffrey Altman [Wed, 19 Sep 2007 18:55:55 +0000]
windows-version-info-20070919

FIXES 71397

Many of the OpenAFS binaries that are installed on end user machines
did not have version information.  This is problematic as it is impossible
to determine which version of a binary the crash report was filed against
when a report is received by Microsoft Windows Error Reporting service.

16 years agowindows-vnovnode-while-file-in-use-20070918
Jeffrey Altman [Tue, 18 Sep 2007 17:57:31 +0000]
windows-vnovnode-while-file-in-use-20070918

Discovered another case where VNOVNODE errors were not being handled.
If there are dirty buffers and a VNOVNODE error is received while
writing the buffer, the buffer would be left in the dirty buffers queue.
This caused a couple problems:
(1) any attempt to flush the file, volume, or cache would fail because
    there were unflushed dirty buffers that could not be flushed.
(2) shutdown of the service would hang because the buffers could not
    be flushed.

In addition, while a VNOVNODE error would result in the cm_scache_t
being marked CM_SCACHEFLAG_DELETED, this state was not being checked
at the SMB layer.  As a result, if a smb_fid_t was allocated and it
referenced the deleted cm_scache_t, the SMB operations would continue
to be processed and report success even if the actual file or directory
no longer existed.

We now clear the dirty state on buffers which cannot be written due to
VNOVNODE errors.  We also check the cm_scache_t for deletion prior to
use whenever a smb_fid_t is looked up.  If the cm_scache_t is deleted,
the smb_fid_t is closed and the error CM_ERROR_NOSUCHFILE is returned
for files or CM_ERROR_NOSUCHPATH for directories.