When verifying a directory entry we must update the metadata
for a file therefore we must also purge the data if the version
number changed. This is not optional.
Change-Id: Icb673c7a6c52b3778978248fcb243d110685aa8d
Reviewed-on: http://gerrit.openafs.org/7208
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
The redirector must flush the data to the file server
and then drop the file locks. Doing it the other way
creates races.
Change-Id: Ic3a48546ee9e0066df9d228fc4e57f52b43587c2
Reviewed-on: http://gerrit.openafs.org/7207
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Check for the gencat program and give a missing warning if
not present.
Change-Id: I71cdaa0840215b53071be4481a5f23b2f8667bda
Reviewed-on: http://gerrit.openafs.org/7203
Reviewed-by: Simon Wilkinson <simonxwilkinson@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Fail configure if the solariscc is not found and give the builder
a message to cope with non-standard installation paths for
sunstudio.
Change-Id: I64c1e752833f2aa824f4a15cc144e54bb03a1ec8
Reviewed-on: http://gerrit.openafs.org/7204
Reviewed-by: Simon Wilkinson <simonxwilkinson@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Both rxi_ReadProc and rxi_FillReadVector contained copies of the
same code to pull a packet out of the receive queue, and turn it
into the call's currentPacket. Abstract this out into a single common
function, so we're not maintaining the same code in two different
places.
Change-Id: I20af6b4ff19f05e21ffde1a80609be12ad6cfeee
Reviewed-on: http://gerrit.openafs.org/7195
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Instead of making a copy of the call->currentPacket variable in our
read/write routines, reference it directly. Make it clear that
currentPacket is used solely by the application thread, and remove
a number of mistaken comments that suggest otherwise.
Change-Id: I7ad799acbf110422df7c1e18ab552caf26b7766a
Reviewed-on: http://gerrit.openafs.org/7194
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Remove includes of system headers where roken.h already takes care of
including them. This simplifies the source tree, reduces the amount of
work done by the compiler, and ensures that all of our headers are
included with the correct guards
The list of files to edit was generated with the following script:
list=`grep include external/heimdal/roken/roken.h.in \
| sed -e's/#include//g' | sort | uniq`; \
for A in `find . -name *.c | xargs grep -l roken.h \
| grep -v external/ | grep -v WINNT/`; do \
found=0; \
for B in $list; do \
if grep "$B" $A > /dev/null; then \
echo "$A : $B"; \
found=1; \
fi; \
done; \
if [ $found == 1 ] ; then mvim -f $A; fi; \
done
Change-Id: I2edbda550a129709b1dc6860b17d6a8a7509af58
Reviewed-on: http://gerrit.openafs.org/5815
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Alistair Ferguson <alistair.ferguson@mac.com>
Reviewed-by: Michael Meffie <mmeffie@sinenomine.net>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Add index offset as a possible volscan output column.
Change-Id: I1bf6e8dc3c506a2ee013536e19141ddb214879a6
Reviewed-on: http://gerrit.openafs.org/6980
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Option to ignore directory vnode magic when looking
up the path names. Print which vnode failed when
doing the checks.
Change-Id: Id4b74ee9f139305025b849a007a4534b8ed04dbf
Reviewed-on: http://gerrit.openafs.org/6979
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Currently the salvaging code looks for special inodes, and infers the
volume id and inode type from the OGM data in each special inode file.
However, we can already derive this information from the inode number
itself for the special inode, so if they disagree, use the values
based off of the inode number and correct the OGM data.
The inode number should be more likely to be correct, since that is
how we look up the special inode from the header when attaching the
volume. It is also impossible to get special inode files with the same
name, so this ensures we don't get duplicates. And for people that go
snooping around /vicepX/AFSIDat even though we tell them not to, it
seems more likely that they go around 'chmod'ing or 'chown'ing rather
than 'mv'ing.
This change avoids an abort in the salvaging code when the OGM data is
wrong. If we trust the OGM data when it is incorrect, we assume the
special inode file is for a different volume. So when we go to
recreate one of the special files for the volume we're actually
working with, the IH_CREATE fails (from EEXIST) and so we abort.
Change-Id: Ibfaa1bfb27d190c5fec6fc673d88df366268e302
Reviewed-on: http://gerrit.openafs.org/6946
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Add GetWinOGM and SetWinOGM for getting and setting the
Windows-equivalent of the Unix OGM data. Make those and CheckOGM use
GetFileTime/SetFileTime so we can operate just via an FD_t, without
needing the full pathname. Modify the NT namei_icreate to use
SetWinOGM.
Change-Id: I5077911d8473d44ad9ba1d17d3b6deec3a12ba39
Reviewed-on: http://gerrit.openafs.org/6945
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>
libaudit depends on util symbols, so link it first to avoid
unresolved symbols. Include libaudit in the list of libs
instead of making it a special case.
Change-Id: Ide6d5f7e3ce8fd76b504088dcae3110872a87092
Reviewed-on: http://gerrit.openafs.org/6847
Tested-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Do not log a code of zero if the name plus instance length exceeds
the max pr name buffer.
Change-Id: I8b8f970213c0beb35a728bcc2158796f84120a3b
Reviewed-on: http://gerrit.openafs.org/6861
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Alistair Ferguson <alistair.ferguson@mac.com>
Tested-by: Derrick Brashear <shadow@dementix.org>
Add the dest path component in the quick start guide instructions for
copying files from a binary OpenAFS distribution file.
Change-Id: I5c39ac0426ef2eab89ae5d3df260628e949dc21e
Reviewed-on: http://gerrit.openafs.org/7187
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Move afsd over to using the newer cmd functions, and avoid using
direct array access, and fixed offsets.
Change-Id: I6e09823d764cbe6109c9a72700239d814b91db0c
Reviewed-on: http://gerrit.openafs.org/7137
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
If an error was detected when processing RDR_CleanupFileEntry()
do not perform the file deletion. Just return the error as
attempting the deletion could result in an exception.
Change-Id: I57b17a15d1bdda4d0890dbe59f9e2432d465250f
Reviewed-on: http://gerrit.openafs.org/7186
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Log the error code when marking the server down due
to server/networks issues.
Change-Id: Ib3d57562642ca564dcfeaa775de66f54d08f23af
Reviewed-on: http://gerrit.openafs.org/7138
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Remodel the ugen_* ubik initialisation interface so that more complex
sets of security flags can be passed through it. A number of new
functions are provided, all of which take the AFSCONF_* security flags
defined in libauth:
*) ugen_ClientInitCell is used to initialise a ubik client structure
when the caller already has afsconf_dir and afsconf_cell
structures for the cell being contacted
*) ugen_ClientInitServer is used to initialise a client which will
contact a specific server within a given cell, rather than
selecting one from the pool
*) ugen_ClientInitFlags is used to initialise a client given a
cellname, and a path to a configuration directory. It is is
closest to the legacy ugen_ClientInit function.
All of the in tree callers are updated to use this new interface
Change-Id: I6a438aa70d94114c8b6626fa1396b7580ab07c07
Reviewed-on: http://gerrit.openafs.org/7131
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Add a remote flag, skip rights related code for the remote case.
This will need to be revisited later; in the case that what ends
up on disk is slightly different than the InStatus passed in, the
changes should be preserved to pass to the replicas.
Change-Id: I596698ba00174efbdc6a170a24f86ea88554f9af
Reviewed-on: http://gerrit.openafs.org/6675
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Add a remote flag to GetVolumePackageWithCall to make it usable for
remote replication RPCs. For the remote case the checks related
to access rights are skipped.
Change-Id: I81db0bb1ae90c5f98cf4e59d9ba28466b312c763
Reviewed-on: http://gerrit.openafs.org/6673
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Add parameters to VAllocVnode to allow the caller to specifiy the
vnode and unique numbers to use. This will be used by the RW
replication code to keep vnode numbers in sync between the master
volume and the replicas.
Adapted from code by Vishal Powar and Derrick Brashear.
Change-Id: Ibaf79aad2b3e7a52802f5e01f7e4c7730c3f5090
Reviewed-on: http://gerrit.openafs.org/6672
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
In AFSValidateProcessEntry(), ullProcessId is 64-bit and must be
printed as an %I64X in AFSDbgLogMsg() formats.
Change-Id: I0e1cd45383c4512eb271c294674bbfe3f0f39bef
Reviewed-on: http://gerrit.openafs.org/7172
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Add the rk_getprogname symbol to the Windows mapfile so it can be
used by roken consumers.
Change-Id: I77c7db3fcd9136d36e2b17b6250bfcfa6327b0e5
Reviewed-on: http://gerrit.openafs.org/7139
Tested-by: Simon Wilkinson <simonxwilkinson@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Remove the hcrypto/des.h include directory from the rxkad_prototypes
header file. We ship rxkad_prototypes.h, but we don't ship the hcrypto
headers. Having an installed header depend upon an uninstalled one is
a pain when building out-of-tree code like perl-AFS.
Change-Id: Idac93ba834c84f6f9e2e49361f88d0e357a328d6
Reviewed-on: http://gerrit.openafs.org/7068
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Move the various key conversion functions into thir own header file,
rather than rxkad_prototypes.h. This means we can isolate the dependency
on the hcrypto DES headers (which we don't install)
Change-Id: I8efcbf97b6794d9573f35eec9d4a32e9ff4cf352
Reviewed-on: http://gerrit.openafs.org/7067
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
The perl-AFS module needs to link against a pthreaded, PIC, library
that contains all of the routines that it provides perl shims for.
At the moment, libafsrpc_pic and libafsauthent_pic are the only such
libraries that OpenAFS provides.
Add all of the necessary objects to libafsauthent_pic such that the
2.6.2 version of perl-AFS can build with a minimal set of patches.
Minimise the damage by only making these available through the static
version of the libafsauthent_pic library.
Change-Id: I5a2aa5f5fbcfbe15aa7a23c3ab57b34b979ab2f3
Reviewed-on: http://gerrit.openafs.org/7070
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
The opr queue definition is required by a structure in the installed
cellconfig.h header. So that header can still be used by out-of-tree
applications, install the queue.h header too.
Change-Id: Ifaf49d59df24275ff50c215de6a0f40fc2bd2230
Reviewed-on: http://gerrit.openafs.org/7069
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
If afsd_notifier() is called the service is going to exit(1).
Report to the service manager that the service exited with an
error as well.
Change-Id: I88bb2f315016ab17a526d1eab162ccde8beb0dd0
Reviewed-on: http://gerrit.openafs.org/7148
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Refactor AFSValidateEntry to avoid obtaining the
ObjectInformation->Fcb->Resource when it isn't necessary.
This will avoid contention and improve performance.
The only time that the Fcb->Resource is required is when
the object requires verification, the object is a FILE,
and the object was successfully evaluated.
Even with this reorganization there is a small window
of opportunity for a deadlock to occur if a CcPurgeCacheSection()
which is called with the Fcb->Resource held triggers a filter
driver to issue a CreateFile and in between the two operations
an invalidate object is received.
Change-Id: Iddf64f030c6a608ac29a10b7a9a7e130ae6c4249
Reviewed-on: http://gerrit.openafs.org/7143
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Move the new background daemon code out from under the AFS_DARWIN80_ENV
ifdef into its own AFS_NEW_BKG define. This lets us use it for things
other than the Darwin mv helper, on other platforms.
Based on code originally written by Derrick Brashear
Change-Id: Iba9431f5dc4bbbe10222b25a595cdd1ea2f592f4
Reviewed-on: http://gerrit.openafs.org/7136
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Currently we always try to delete a BK volume if we're deleting the
RW. If the BK volume id is 0, this is never going to do anything, so
don't try to delete it.
Change-Id: I0c003040cface9999944d61637077fce1736445e
Reviewed-on: http://gerrit.openafs.org/7140
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Derrick Brashear <shadow@dementix.org>
Import expand_path.c from Heimdal so that it can be used for path
parsing with the Windows version of the INI file parser.
Change-Id: I764ea10edb8bb63bd45216c912c0f67597c4467a
Reviewed-on: http://gerrit.openafs.org/7144
Tested-by: Simon Wilkinson <simonxwilkinson@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
This commit updates the code imported from heimdal to
ee7340860a22f81fb869cc431efc1fd4e5c77d34 (switch-from-svn-to-git-2542-gee73408)
New files are:
krb5/expand_path.c
Change-Id: Ib4093b72257f9a0d1a49edb0fe6779908a78cd83
Reviewed-on: http://gerrit.openafs.org/7145
Tested-by: Simon Wilkinson <simonxwilkinson@gmail.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
The peer structure contains a number of elements left over from an
earlier, burst based, congestion control strategy. As OpenAFS has
never implemented this strategy, just remove these elements from
the structure, and store 0 values for these elements in the debug
packets.
Change-Id: Ib50a8ac587890efb483d3c2640ead615d170db34
Reviewed-on: http://gerrit.openafs.org/7006
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
We don't (and haven't, since the release of OpenAFS) use the packet
skew calculations for anything. However, maintaining them requires
taking locks in some critical parts of the transmission code path.
For both speed, and maintainability, reasons remove the skew code
from the tree - it's in git if we ever want to go back to it.
Change-Id: Idc6dc80fb06ca959ad02a85f2c536a94ca3d5c02
Reviewed-on: http://gerrit.openafs.org/7004
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Hide functions in rx_packet.c which are only used by that file by
making them static.
Change-Id: I22d4be2130cc4a46183254d3ba36c61644891852
Reviewed-on: http://gerrit.openafs.org/7003
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
The peer structure is an internal RX structure, so it shouldn't
be contained in public header files, or used by non-rx modules.
Hide the structure definition in rx_peer.h, and convert the
rx_HostOf and rx_PortOf accessors from macros into functions. This
will also help with adding support for non-IPv4 addresses to RX.
Change-Id: I790fa3193bfa629a2872fc5395d0146e76a2de9e
Reviewed-on: http://gerrit.openafs.org/7002
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
errors is set but not used - remove it.
Change-Id: I00c99ac8b9c61ab2667aecbdf0fd04401d018bf8
Reviewed-on: http://gerrit.openafs.org/7142
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Rework the API for directly parsing configuration files and move it
out of util/ and into libcmd, where we can actually make use of it.
This is a raw API - it provides direct access to the parser. The
eventual intent is that this will predominantly be contained within
libcmd itself, which will then wrap this with the more general purpose
Option functions
Change-Id: I9f4a9c373c4123023120f69e595068ead6160507
Reviewed-on: http://gerrit.openafs.org/7132
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
The roken symbols for getopt and getprogname are now prefixed with
'rk_'. Fix the map file so it lists the correct symbols.
Change-Id: I61adba7d5bec3e86b8e29d1eefda23df4ef91988
Reviewed-on: http://gerrit.openafs.org/7010
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Just running 'make' in a subdirectory will fire the first rule in that
directory's Makefile. When the buildtools rules were added to roken and
crypto, they were put at the start of the file, meaning that they fire
in perference to 'all'
Swap the ordering back so that 'make' does what would be expected
Change-Id: I598a183f42c150a5da724b7403e91f89f74a0109
Reviewed-on: http://gerrit.openafs.org/7009
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
The ka_IsZero and ka_PrintBytes function just use char strings. Cast
directly, rather than going via a helper functions to simplify this
code.
The helper functions add complexity, and additional dependencies which
cause problems with some uses of this code.
Change-Id: I831ff0f8fbecec602d8c2b9112675938facfb06b
Reviewed-on: http://gerrit.openafs.org/7066
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Export the afs_xdr_pointer and afs_xdr_u_char functions so that
shlibafsauthent can make use of them.
Change-Id: I16cf3f69904bffb3a034fd4052117845c1067a2c
Reviewed-on: http://gerrit.openafs.org/7065
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Derrick Brashear <shadow@dementix.org>
Subsequent patchsets will turn rx_peer into an opaque object
and remove the burst elements entirely from the tree. For
now remove them from rx/test
Change-Id: I728f7b74497701c0ef965009dff5550e1b153696
Reviewed-on: http://gerrit.openafs.org/7130
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
rxi_AckAll is coded as if it is an event handler, yet the only
call site is a direct call from rx.c. So, rework the function to
remove all of the event handling code, change its arguments to match,
and make it static.
Change-Id: I947c02bf91bf5a3f42d2a389f9393e68f3ccee4b
Reviewed-on: http://gerrit.openafs.org/6999
Tested-by: BuildBot <buildbot@rampaginggeek.com>
Reviewed-by: Jeffrey Altman <jaltman@secure-endpoints.com>
Tested-by: Jeffrey Altman <jaltman@secure-endpoints.com>
