Whamcloud - gitweb
grev [Tue, 28 Oct 2008 16:05:54 +0000 (16:05 +0000)]
b=17477
i=Adilger
i=Tappro
init facet vars for mounted lustre
huanghua [Tue, 28 Oct 2008 14:06:30 +0000 (14:06 +0000)]
Branch HEAD
b=17192
i=robert.read
i=adilger
clear the proc stats counter. This value might be used later again.
bobijam [Tue, 28 Oct 2008 05:34:14 +0000 (05:34 +0000)]
Branch HEAD
b=17093
o=Btian Behlendorf (behlendorf1@llnl.gov) for mgs part
o=E. Gryaznova (grev) for test framework
i=nathan.rutman
i=adilger
Do writeconf only explicitly required.
nikita [Mon, 27 Oct 2008 12:15:58 +0000 (12:15 +0000)]
lockdep: add lockdep_{on,off}() functions.
johann [Mon, 27 Oct 2008 12:08:33 +0000 (12:08 +0000)]
Branch HEAD
b=17385
i=green
i=shadow
grab lock reference when the lock is added to the waiting or expired list.
zhanghc [Mon, 27 Oct 2008 06:04:19 +0000 (06:04 +0000)]
Branch HEAD
handle 'Unexpected: can't find mdc_open_data,
but the close succeeded with replayabel imp.
Please tell http://bugzilla.lustre.org/"
printed by mdc_close in mdc_request.c
b=17089
i=johann
bobijam [Mon, 27 Oct 2008 02:15:02 +0000 (02:15 +0000)]
Branch HEAD
b=11089
i=nathan.rutman
i=adilger
per-nid client statistics for MDT.
fanyong [Sat, 25 Oct 2008 04:37:55 +0000 (04:37 +0000)]
Branch HEAD
b=17193
i=eric.mei
i=minh.diep
1) set root permission mode as 0755 before test.
2) scripts cleanup.
adilger [Fri, 24 Oct 2008 20:52:32 +0000 (20:52 +0000)]
Branch HEAD
Return a protocol error from ost_punch() like all other handler methods
if there is a request format error, instead of -EINVAL.
adilger [Fri, 24 Oct 2008 20:47:38 +0000 (20:47 +0000)]
Branch HEAD
Rename "pages" pointer to "npages", because it otherwise looks like an
array of pointers to pages instead of a pass-by-reference number of pages.
b=12682
adilger [Fri, 24 Oct 2008 20:40:56 +0000 (20:40 +0000)]
Branch HEAD
Fix whitespace, line wrapping.
grev [Fri, 24 Oct 2008 20:10:57 +0000 (20:10 +0000)]
b=13297
add test_27u to ALWAYS_EXCEPT list
grev [Fri, 24 Oct 2008 19:46:25 +0000 (19:46 +0000)]
b=16551 (att 19784)
o=Adilger
i=grev
skip replay/recovery tests if remote MDS/OSS nodsh
b=16551 (att 19835)
i=Nathan
do skip_rem[mds|ost] check only if test suit is run; fail acc-sm if
tests suits were skipped due to nodsh
b=17326
i=Alexey.Lyashkov
remove now-useless remount/recofig for liblustre;
always add accept=all to lnet module
yury [Fri, 24 Oct 2008 05:05:11 +0000 (05:05 +0000)]
b=17323
- fixed using "mds" in test 59b
shadow [Fri, 24 Oct 2008 05:03:41 +0000 (05:03 +0000)]
Kill extra argument for llog_connect, and don't access to ld_tgt_count
without protection.
Branch HEAD
b=16693
i=umka
i=tappro
stinson1947 [Thu, 23 Oct 2008 23:53:36 +0000 (23:53 +0000)]
updated for next bi-weekly build of 2.0.0
grev [Thu, 23 Oct 2008 19:33:30 +0000 (19:33 +0000)]
file large-scale.sh was initially added on branch b1_8_gate.
yury [Thu, 23 Oct 2008 18:10:26 +0000 (18:10 +0000)]
b=17323
r=shadow
- small to test from Mike
girish [Thu, 23 Oct 2008 16:59:43 +0000 (16:59 +0000)]
Initialize i_file_acl for EXT3_BAD_INO (Johann)
i=adilger
i=kalpak
b=17414
yury [Thu, 23 Oct 2008 14:33:01 +0000 (14:33 +0000)]
b=17447
r=deen,adilger
- fixes memory leaks;
- cleanups and more accurate error handling.
mattwu [Thu, 23 Oct 2008 13:33:28 +0000 (13:33 +0000)]
Remove files imported by Cariddi (osr) carelessly
kalpak [Thu, 23 Oct 2008 10:02:19 +0000 (10:02 +0000)]
b=12800
o=alex.zhuravlev
i=kalpak
i=adilger
Add support for tunable preallocation window and new tunables for large/small requests
adilger [Thu, 23 Oct 2008 04:13:42 +0000 (04:13 +0000)]
Branch HEAD
Remove trailing whitespace.
adilger [Wed, 22 Oct 2008 23:39:19 +0000 (23:39 +0000)]
Branch HEAD
Make releasepage check failsafe for newer kernels, per Robert's suggestion.
adilger [Wed, 22 Oct 2008 21:33:28 +0000 (21:33 +0000)]
Branch HEAD
Improve error messages.
adilger [Wed, 22 Oct 2008 21:32:09 +0000 (21:32 +0000)]
Branch HEAD
Start modifying test 32a, 32b (currently disabled) to handle 1.8->2.0
filesystem upgrade test.
adilger [Wed, 22 Oct 2008 21:26:57 +0000 (21:26 +0000)]
Branch HEAD
Fix whitespace.
adilger [Wed, 22 Oct 2008 21:25:23 +0000 (21:25 +0000)]
Branch HEAD
Move HAVE_PAGE_CONSTANT checks out-of-line via compat macros.
adilger [Wed, 22 Oct 2008 21:19:51 +0000 (21:19 +0000)]
Branch HEAD
Remove unnecessary cast of void pointers.
adilger [Wed, 22 Oct 2008 21:17:28 +0000 (21:17 +0000)]
Branch HEAD
copy_{to,from}_user() do not return an error code, but rather the number
of uncopied bytes (0 if all bytes were copied). Storing this return value
into "rc" is misleading because it then seems this should be returned to
the caller of the function.
Code was audited to remote storing of the return value into "rc" and
several incorrect uses were found.
adilger [Wed, 22 Oct 2008 21:06:44 +0000 (21:06 +0000)]
Branch HEAD
Don't print two console messages where one would do.
adilger [Wed, 22 Oct 2008 21:04:39 +0000 (21:04 +0000)]
Branch HEAD
The infiniband code includes its own lockdep compatibility code and this
#defines lockdep_set_class but doesn't include mutex_lock_nested() or
down_write_nested(), causing a compiler error in obdclass/llog_cat.c.
mutex_{un,}lock_nested() and down_{read,write}_nested() are functions
under CONFIG_DEBUG_LOCK_ALLOC, but #defined to regular locks without
CONFIG_DEBUG_LOCK_ALLOC so just checking #ifndef isn't sufficient.
b=16450
adilger [Wed, 22 Oct 2008 20:56:18 +0000 (20:56 +0000)]
Branch HEAD
Fix compile warning for 2.6.9 kernels that define gfp_t but don't use
it in the ->releasepage() prototype.
Make the AC_MSG_RESULT() output consistent.
adilger [Wed, 22 Oct 2008 20:22:58 +0000 (20:22 +0000)]
Branch HEAD
Limit bio_alloc() size to the maximum possible value, or it will always fail.
b=9945
i=johann
grev [Wed, 22 Oct 2008 19:26:12 +0000 (19:26 +0000)]
b=15711
i=Johann
recovery-small exception for FAILURE_MODE=HARD and mixed ost devices
adilger [Wed, 22 Oct 2008 18:58:20 +0000 (18:58 +0000)]
Branch HEAD
Fix debug message.
grev [Wed, 22 Oct 2008 18:57:06 +0000 (18:57 +0000)]
b=17442
i=Nikita
do_nodes fix for case when list contains a single node $HOSTNAME
adilger [Wed, 22 Oct 2008 18:55:40 +0000 (18:55 +0000)]
Branch HEAD
Fix ChangeLog comments
wangdi [Wed, 22 Oct 2008 18:32:57 +0000 (18:32 +0000)]
Branch: HEAD
rq_xid should be "LPU64", instead of "LPD64"
cariddi [Wed, 22 Oct 2008 16:46:40 +0000 (16:46 +0000)]
Setup changed
grev [Wed, 22 Oct 2008 16:39:53 +0000 (16:39 +0000)]
b=17443
i=Nikita
fix calc_osc_kbytes
yury [Wed, 22 Oct 2008 14:25:38 +0000 (14:25 +0000)]
b=17323
r=johann,adilger
- fixes resent rpcs for those sent by set and also fixes log_cancel resent handling;
- cleanups.
yangsheng [Wed, 22 Oct 2008 06:12:46 +0000 (06:12 +0000)]
Branch HEAD
b=16437
i=cliff.white, brian
Enable the '-m64' flag for powerpc64 patchless build.
robert.read [Tue, 21 Oct 2008 22:30:53 +0000 (22:30 +0000)]
Branch HEAD
b=1819
i=adilger
i=nathan
Add test for new proc file.
robert.read [Tue, 21 Oct 2008 21:40:41 +0000 (21:40 +0000)]
Branch HEAD
b=1819
i=adilger
i=nathan
Add an import file to the osc, mdc, and mgc proc dir.
yury [Tue, 21 Oct 2008 13:53:47 +0000 (13:53 +0000)]
b=17353
r=shadow,wangdi
- fixes wrong work with llog in catalog that led to removing alive object on ost in recovery time;
- cleanups.
nikita [Tue, 21 Oct 2008 10:58:13 +0000 (10:58 +0000)]
fix lu_device_is_md() assertions.
adilger [Tue, 21 Oct 2008 07:06:23 +0000 (07:06 +0000)]
Branch HEAD
Temporary build fix.
b=17444
i=ericm (original fix)
fanyong [Tue, 21 Oct 2008 02:18:04 +0000 (02:18 +0000)]
Branch HEAD
b=16621
i=robert.read
scripts fix.
nikita [Mon, 20 Oct 2008 20:46:41 +0000 (20:46 +0000)]
mdt_init0(): fix memory leak in error handling path.
nikita [Mon, 20 Oct 2008 10:19:30 +0000 (10:19 +0000)]
liblustre build fixes for 2.6.18-53 (found by Shadow).
nikita [Mon, 20 Oct 2008 10:19:09 +0000 (10:19 +0000)]
lockdep build fixes for 2.6.18-53 (found by Shadow).
liuy [Mon, 20 Oct 2008 06:47:01 +0000 (06:47 +0000)]
Branch HEAD
b=12521
remove unused hint "user_cb_nodes"
huanghua [Mon, 20 Oct 2008 05:47:14 +0000 (05:47 +0000)]
Branch HEAD
b=17444
quick fix for HEAD build problem.
nikita may have more fix.
ericm [Mon, 20 Oct 2008 00:06:26 +0000 (00:06 +0000)]
*** empty log message ***
nikita [Sun, 19 Oct 2008 10:10:49 +0000 (10:10 +0000)]
merge with HEAD.
nikita [Sat, 18 Oct 2008 17:43:11 +0000 (17:43 +0000)]
libcfs: add cfs_{need,cond}_resched() definition and implementations for
Linux, NT, and liblustre.
b=16450
nikita [Sat, 18 Oct 2008 17:41:47 +0000 (17:41 +0000)]
Liblustre build fixes.
nikita [Sat, 18 Oct 2008 17:40:51 +0000 (17:40 +0000)]
Use cdebug_show() in CDEBUG-style macros defined outside of libcfs.
b=16450
nikita [Sat, 18 Oct 2008 17:39:45 +0000 (17:39 +0000)]
Ratelimit a memory allocation failure message that can be too chatty.
b=16450
nikita [Sat, 18 Oct 2008 17:38:41 +0000 (17:38 +0000)]
New testing program exercising sendfile(2) (Jay).
b=16450
nikita [Sat, 18 Oct 2008 17:37:57 +0000 (17:37 +0000)]
New testing program exercising readv(2) and writev(2) (Qian).
b=16450
nikita [Sat, 18 Oct 2008 17:36:17 +0000 (17:36 +0000)]
lu_ref is a debugging module allowing to track references to a given
object. It is quite cpu expensive, and has to be explicitly enabled with
--enable-lu_ref. See usage description within the patch.
b=16450
nikita [Sat, 18 Oct 2008 17:34:22 +0000 (17:34 +0000)]
Replace RW_LOCK_UNLOCKED() with rwlock_init() as the former doesn't work with
lockdep.
b=16450
nikita [Sat, 18 Oct 2008 17:33:11 +0000 (17:33 +0000)]
Add lu_env parameter to ->rq_interpreter call-back. NULL is passed
there. Actual usage will be in CLIO.
b=16450
nikita [Sat, 18 Oct 2008 17:31:46 +0000 (17:31 +0000)]
Remove unused code.
b=16450
nikita [Sat, 18 Oct 2008 17:30:57 +0000 (17:30 +0000)]
Add lu_ref tracking to obd_device.
b=16450
nikita [Sat, 18 Oct 2008 17:29:50 +0000 (17:29 +0000)]
Add st_block checking to multistat.c.
b=16450
nikita [Sat, 18 Oct 2008 17:28:44 +0000 (17:28 +0000)]
Remove unused mdd code.
b=16450
nikita [Sat, 18 Oct 2008 17:27:58 +0000 (17:27 +0000)]
Move md-specific fields out of struct lu_site into special struct md_site, so
that lu_site can be used on a client.
b=16450
nikita [Sat, 18 Oct 2008 17:26:37 +0000 (17:26 +0000)]
Constify instances of struct {lu,dt,md}_device_operations.
b=16450
nikita [Sat, 18 Oct 2008 17:25:19 +0000 (17:25 +0000)]
Separate lu_site stats printing code into a separate function that can be
reused on a client.
b=16450
nikita [Sat, 18 Oct 2008 17:24:14 +0000 (17:24 +0000)]
lu_ref is a debugging module allowing to track references to a given
object. It is quite cpu expensive, and has to be explicitly enabled with
--enable-lu_ref. See usage description within the patch.
b=16450
nikita [Sat, 18 Oct 2008 17:22:46 +0000 (17:22 +0000)]
Fix a race between lu_object_find() finding an object and its concurrent
finalization. This race is (most likely) not possible on the server, but might
happen on the client.
b=16450
nikita [Sat, 18 Oct 2008 17:20:41 +0000 (17:20 +0000)]
lu_kmem_descr and its companion interface allow to create and destroy a number
of kmem caches at once.
b=16450
nikita [Sat, 18 Oct 2008 17:19:11 +0000 (17:19 +0000)]
Add lu_ref support to lu_object and lu_device. lu_ref is used to track leaked
references.
b=16450
nikita [Sat, 18 Oct 2008 17:17:44 +0000 (17:17 +0000)]
Introduce two new methods in lu_device_type_operations, that are invoked when
first instance of a given type is created and last one is destroyed
respectively. This is need by CLIO.
b=16450
nikita [Sat, 18 Oct 2008 17:16:32 +0000 (17:16 +0000)]
Introduce new lu_context functions that are needed on the client side, where
some system threads (ptlrpcd) are shared by multiple modules, and so cannot be
stopped during module shutdown.
b=16450
nikita [Sat, 18 Oct 2008 17:14:45 +0000 (17:14 +0000)]
On a server, a file system object is uniquely identified by a fid, which is
sufficient to locate and load all object state (inode). On a client, on the
other hand, more data are necessary instantiate an object. Change
lu_object_find() and friends to take additional `lu_conf' argument describing
object. Typically this includes layout information.
b=16450
nikita [Sat, 18 Oct 2008 17:12:24 +0000 (17:12 +0000)]
Constify instances of struct lsm_operations.
b=16450
nikita [Sat, 18 Oct 2008 17:11:31 +0000 (17:11 +0000)]
Wrap kms updates into a helper function.
b=16450
nikita [Sat, 18 Oct 2008 17:09:55 +0000 (17:09 +0000)]
Use appropriately tagged _nested() locking calls in the places where llog
takes more than one ->lgh_lock lock.
b=16450
nikita [Sat, 18 Oct 2008 17:07:21 +0000 (17:07 +0000)]
Add new ->l_weigh_ast() call-back to ldlm_lock. It is called by
ldlm_cancel_shrink_policy() to estimate lock "value", instead of hard-coded
`number of pages' logic.
b=16450
nikita [Sat, 18 Oct 2008 17:05:50 +0000 (17:05 +0000)]
Introduce ldlm_lock_addref_try() function (used by CLIO) that attempts to
addref a lock that might be being canceled concurrently.
b=16450
nikita [Sat, 18 Oct 2008 17:04:25 +0000 (17:04 +0000)]
lu_ref support for ldlm_lock and ldlm_resource. See lu_ref patch.
lu_ref fields ->l_reference and ->lr_reference are added to ldlm_lock
and ldlm_resource. LDLM interface has to be changed, because code that
releases a reference on a lock, has to "know" what reference this is.
In the most frequent case
lock = ldlm_handle2lock(handle);
...
LDLM_LOCK_PUT(lock);
no changes are required. When any other reference (received _not_ from
ldlm_handle2lock()) is released, LDLM_LOCK_RELEASE() has to be called
instead of LDLM_LOCK_PUT().
Arguably, changes are pervasive, and interface requires some discipline
for proper use. On the other hand, it was very instrumental in finding
a few leaked lock references.
b=16450
nikita [Sat, 18 Oct 2008 17:02:22 +0000 (17:02 +0000)]
Kill unused ldlm_handle2lock_ns() function.
b=16450
nikita [Sat, 18 Oct 2008 17:01:23 +0000 (17:01 +0000)]
Add functions to map (rather arbitrary) between LDLM error codes and standard
errno values. CLIO needs this to prevent LDLM specific constants from escaping
ldlm and osc.
b=16450
nikita [Sat, 18 Oct 2008 15:49:01 +0000 (15:49 +0000)]
CLIO posts ENQUEUE requests asynchronously through ptlrpcd---a case that stock
ldlm_completion_ast() cannot handle as it waits until lock is
granted. Introduce new ldlm_completion_ast_async() for this. Also comment
ldlm_completion_ast().
b=16450
nikita [Sat, 18 Oct 2008 15:45:07 +0000 (15:45 +0000)]
CLIO uses lock weighting policy to keep locks over mmapped regions in
memory---a requirement implemented through ->o_join_lru() obd method in
HEAD. Remove this method and its users.
b=16450
nikita [Sat, 18 Oct 2008 15:42:22 +0000 (15:42 +0000)]
Instead of specifying each ldlm_lock call-back through separate parameter,
wrap them into struct ldlm_callback_suite.
b=16450
nikita [Sat, 18 Oct 2008 15:38:15 +0000 (15:38 +0000)]
Use spin_lock_nested() in (the only) situation where more than one ldlm_lock
is locked simultaneously. Also, fix possible dead-lock in
ldlm_lock_change_resource() by enforcing particular lock ordering.
b=16450
nikita [Sat, 18 Oct 2008 15:35:00 +0000 (15:35 +0000)]
Use LCK_MINMODE rather than 0 to reset lock->l_granted_mode to its initial
state.
b=16450
nikita [Sat, 18 Oct 2008 15:32:59 +0000 (15:32 +0000)]
Add new LINVRNT() macro, optional on new --enable-invariants configure
switch. This macro is to be used for consistency and sanity checks that are
too expensive to be left in `production' mode.
b=16450
nikita [Sat, 18 Oct 2008 15:30:01 +0000 (15:30 +0000)]
Introduce and use new LOCK_INODE_MUTEX_PARENT() macro to be used in the
situations where Lustre has to lock more than one inode mutex at a time.
b=16450
nikita [Sat, 18 Oct 2008 15:27:57 +0000 (15:27 +0000)]
Add two new failloc constants to test lockless IO. Only one of them in
implemented---another is checked in yet to be landed core CLIO code.
b=16450
nikita [Sat, 18 Oct 2008 15:25:48 +0000 (15:25 +0000)]
Augment ->do_{read,write}_lock() prototypes with a `role' parameter indicating
lock ordering. Update mdd code to use new locking interface.
b=16450
nikita [Sat, 18 Oct 2008 15:22:51 +0000 (15:22 +0000)]
Call cmm_device_free() in the failure path of cmm_device_alloc().
b=16450
nikita [Sat, 18 Oct 2008 15:20:34 +0000 (15:20 +0000)]
Make previously optional ->oti_{w,r}_locks sanity checks mandatory to simplify
the code.
b=16450
nikita [Sat, 18 Oct 2008 15:15:47 +0000 (15:15 +0000)]
A couple of trivial spelling fixes.
b=16450
nikita [Sat, 18 Oct 2008 15:13:21 +0000 (15:13 +0000)]
Update documenting comments to match doxygen conventions.
b=16450