For sites with only new (>3.4) clients (or who decide to not care
about older clients) and want to force VRESTARTING instead of VBUSY.
(cherry picked from commit 214c2e7b19333380cb7f3aa5c997f35543bd3d15)
avoid need for sys_kill
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
get rid of need for sys_kill
====================
get rid of need for sys_kill
(cherry picked from commit 937ec63685f69f55eb0a6ce39ab94dc475eb5fe6)
to deal with krb4 security advisory
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
to deal with krb4 security advisory
type mismatch for krb4_cross pointed out by Jack Neely <slack@quackmaster.net>
(cherry picked from commit 1312d1179d4a97e9c62eb26ef55f60007cd35872)
more paranoia to not crash if GetClient fails, since there's apparently a
race we haven't found yet
(cherry picked from commit 4f35fbb68005579f626d021632e899301f551c66)
FIXES 1333
the softsig stuff messed up the windows signal stuff. MSVC doesn't
recognize #elsif. if #elif isn't standard, then it needs to be split into
#else and another #if/#endif block.
FIXES 1333
something in the fileserver started using osi_Assert, instead
of whatever assert comes from util, so we need osi_AssertFailU in addition
to AssertionFailed.
FIXES 1333
leaving the #include <initguid.h> in caused Bad Things to
happen when I did a build. I'm pretty sure initguid.h is only supposed to
be used by things that *impliment* COM objects.
"Atempts to add a new drive, to the root, with a sub mount name, will
sometimes return error #35"
(cherry picked from commit 69a511d5b8a051f4961c3a6503566537893cb96a)
FIXES 1329
"The Windows AFS client returns the error
ERROR_NOSUCHFILE, where the same request on a local NTFS volume returns
the error ERROR_NOSUCHPATH."
(cherry picked from commit ff6394b2e40b4c7c58b32a6e901656a12eb1faf1)
instead of what rx-init-mutex-before-calling-getudpsocket-20030313, just
init the mutex earlier where others are done
(cherry picked from commit 3c3971439731f691dc9faa55ace72596e808b4d9)
"Atempts to add a new drive, to the root, with a sub mount name, will
sometimes return error #35"
(cherry picked from commit 3c22e61e23aae56cf8e40776c4327f7d640efa10)
Add two useful functions for string operations: afs_strdup() and
afs_osi_FreeStr(). Will be used by an upcoming patch.
(cherry picked from commit 5d36376c07c665b175440e51b03fd85fc58422b0)
This is mostly a rewrite of src/afs/afs_cell.c, and associated changes
to other things. Features of the new cell code:
* Persistent name-to-cell-number mapping across reboots, to ensure the
cell numbers in the dcache are meaningful. Stored in the new cache
file, /usr/vice/cache/CellInfo.
* Allow cmdebug to query the cell number to cell name mapping.
* Allow clients to have an empty CellServDB as long as AFSDB is enabled
and ThisCell has AFSDB records.
* The home cell retains the setuid bit even if looked up via AFSDB.
* Dynroot gets its own cell, rather than piggybacking on cell 1.
* Cell 1 is no longer special; cell numbers are now semi-opaque.
* Convert cell traversal code to using GetCellByIndex rather than
directly poking at CellLRU / afs_xcell.
* Separate cells from aliases, which shouldn't have been the same
in the first place.
* Cleaner code (IMHO).
====================
This delta was composed from multiple commits as part of the CVS->Git migration.
The checkin message with each commit was inconsistent.
The following are the additional commit messages.
====================
correct a code block so patch is correct
(cherry picked from commit 629d0806549d5b9cff9cd6f8eae50cf11c3e85ac)
FIXES 1304
avoid using rxi_availProcs to allocate thread ids as it may decrement from under us
(cherry picked from commit 45a4d9cb08d92c10332bda071afb35461d5a4cc3)
if we run out of memory, maybe we should assert. we should probably log.
this is just so i remember to deal better.
(cherry picked from commit 1e4b22a9f579cd53a4aa17caedf04d06fe965feb)
in rx_pthread.c the
++rxi_pthread_hinum;
has to be protected by the rx_stats_mutex in all three cases. In the file I
attached in the openafs-devel article the last one was accidentally
unprotected.
the FSYNC_sync thread (fssync.c) needs needs to be handled as well.
(cherry picked from commit cff48dcb5210e38e0048a04fc3151d045bdca4ce)
apparently having recalc_sigpending stay the same was just as hard
as continuing to export the same level of functionality, so we get neither.
(cherry picked from commit 9a07ae91c39ded5541302d91aae81e5d137e4d06)
Avoid a subtle cycle in the locking order hierarchy involving
the rxi_ServerThreadSelectingCall flag, and instead rely on
call->lock to make sure we don't attach the same call twice.
Also some miscellaneous cleanup and code unification.
(cherry picked from commit a6d9298d3d299cc2c776b22a2800b5c2044ea497)
Wrap our assert() macro in do{..}while(0) to make it look
just like a real function. Fix instances of use of this
macro that lacked a trailing semicolon of their own.
(cherry picked from commit 3ba961ca00f5d8ec2edd32e0b1e5dac1c9864ef9)
Improved signal-thread support for the pthread fileserver,
which avoids blocking signals in any thread, to appease Linux's
thread signaling semantics.
(cherry picked from commit eff534794e7f49166094fa9279bc4d51520d62a8)
