Whamcloud - gitweb
ericm [Mon, 28 Jul 2003 16:34:14 +0000 (16:34 +0000)]
merge b_devel to b_eq:
20030728
(4) mgmt obdclass obdfilter obdecho
ericm [Mon, 28 Jul 2003 16:29:18 +0000 (16:29 +0000)]
merge b_devel to b_eq:
20030728
(3) ldlm liblustre llite lov mdc mds
eeb [Sun, 27 Jul 2003 02:40:01 +0000 (02:40 +0000)]
* Merged b_devel updates
shaver [Sat, 26 Jul 2003 17:58:57 +0000 (17:58 +0000)]
b=1590, (partial 948):
Stop flushing on close, and instead flush all outstanding client data
atomically with the destruction of the final stripe.
And when flushing for destroy, or other cases of truncate-to-zero, send a
hint with the AST to indicate that dirty data should be discarded, instead of
being written back.
r=zab/phil
adilger [Sat, 26 Jul 2003 03:05:16 +0000 (03:05 +0000)]
Remove variable that is only set but not used.
phil [Fri, 25 Jul 2003 22:49:24 +0000 (22:49 +0000)]
merge b_devel into b_llpio
phil [Fri, 25 Jul 2003 21:56:16 +0000 (21:56 +0000)]
merge b_devel into b_filterio
phil [Fri, 25 Jul 2003 21:18:25 +0000 (21:18 +0000)]
land b_unify on b_devel
- patches for 2.6
- linus-approved APIs
- lots of little reorg and fixes for 2.4/2.6 compatibility
- better code to manage MDC lock cancellation and dentry memory pressure
braam [Wed, 23 Jul 2003 04:31:28 +0000 (04:31 +0000)]
- THIS BREAKS b_unify
2.6.0 patches now fixed to Linus specs. 2.6.0 compiles, untested.
beginning of similar changes for 2.4 changes
phil [Tue, 22 Jul 2003 16:29:19 +0000 (16:29 +0000)]
- in ll_umount_begin, protect against NULL obd
- fix debug subsystem in obdclass/obdo.c
- remove zero-length buffer check from lustre_msg_buf; we already have
the minimum-size check, and zero length is common in the "no EA yet"
getattr case
phil [Tue, 22 Jul 2003 13:50:46 +0000 (13:50 +0000)]
partial merge of b_unify into b_devel (20030722_0202)
braam [Tue, 22 Jul 2003 07:20:23 +0000 (07:20 +0000)]
- these should apply, but they don't quite compile (something with netdump)
adilger [Sat, 19 Jul 2003 06:09:32 +0000 (06:09 +0000)]
Fix mtime corruption bug 1570. Was checking wrong valid flags, after
setting them for truncates.
adilger [Fri, 18 Jul 2003 00:05:10 +0000 (00:05 +0000)]
Fixes for remaining POSIX failures, except utime 4.
- set the ATTR_MTIME and ATTR_CTIME flags if truncating, so that we update
the mtime on either the MDS or OST (depending where we are storing the size)
- check the original ia_valid flag, and not the modified attr->ia_valid flag
for utime changes
- add a CERROR for ll_truncate() called without a LOV EA (should never happen
any more)
- do an OST setattr RPC for either MTIME or MTIME_SET flags
- quiet CERRORs for permission problems
- fix another bogus "don't fail if not EPERM" check (we would only have hit
that if we had data corruption on the wire or on disk)
- allow obdo_refresh_inode() to update IO blocksize.
- fix up some error messages in sanity.sh and sanityN.sh
zab [Thu, 17 Jul 2003 17:35:37 +0000 (17:35 +0000)]
- bring recent b_unify changes into b_filterio
(UNIFY_FILTERIO_UPDATE_CHILD_20030717)
eeb [Thu, 17 Jul 2003 13:03:16 +0000 (13:03 +0000)]
* merged current b_devel
adilger [Thu, 17 Jul 2003 07:57:38 +0000 (07:57 +0000)]
Landing statfs cache code.
- preliminary for prealloc (bug 974)
- preliminary for QOS (bug 855)
- speeds up parallel df
- speeds up lprocfs
Lprocfs underwent a bunch of cleanup (mostly from Intel, partly from me)
to remove lots of macros and strangeness w.r.t. per-target statfs values.
Still open for discussion is how old the cached statfs values can be.
I arbitrarily picked 1 second for client statfs, lprocfs, and target
statfs caches. We may want to change that, or allow it to be tunable.
Note that this code is different than that in bug 855, in that the
max_age parameter is an absolute age in jiffies, whereas the patch
in bug 855 used a relative time in seconds. The thing I'm trying
to avoid is if we say "I want a result up to 1 second old", but we
get delayed during an RPC and we don't know what "1 second old" is
related to.
braam [Wed, 16 Jul 2003 22:40:48 +0000 (22:40 +0000)]
1486, 1456, 1546
- Fixes for these bugs included
- Includes substantial cleanup of lock teardown in llite.
- Ignore mountpoints for ll_revalidate_it
- Upgrade to uml-6
Fixes for
adilger [Wed, 16 Jul 2003 12:58:21 +0000 (12:58 +0000)]
Fix flush-on-destroy so that it locks the same resource (i.e. the objid and
not the inode number) as what the clients are locking.
Re-enable flush-on-close, because flush-on-destroy has a deadlock:
- client is writing to object (so has a write extent lock), so
filter_brw_write() (via filter_fid2dentry()) needs a read lock on
parent dir to do the object lookup
- client unlinks object (so gets write lock on parent) and then grabs the
write extent lock on same object in order to flush pages
Solutions are:
- grab extent lock before parent directory lock (a bit messy, but doable)
- send file handle with writes to avoid lookup (needs some way to pass
handle to inode, and doesn't really help us since file is now allowed
to be closed before writing is complete)
shaver [Wed, 16 Jul 2003 11:24:26 +0000 (11:24 +0000)]
include linux/string.h to quell memset warnings and get inline versions
zab [Tue, 15 Jul 2003 19:37:14 +0000 (19:37 +0000)]
- bring recent b_unify fixes into b_filterio
(UNIFY_FILTERIO_UPDATE_CHILD_20030715)
shaver [Tue, 15 Jul 2003 08:54:03 +0000 (08:54 +0000)]
Land b_unify on b_devel.
Highlights include:
- intent rework from Peter and Phil
- unification of VFS intent strategy for 2.4 and 2.5/2.6
- select 2.5 kernel patches from b_ad
- all the whitespace fixes you can imagine.
Complete diff for sanity-checking at http://off.net/~shaver/unify-landing.diff.
braam [Tue, 15 Jul 2003 08:33:07 +0000 (08:33 +0000)]
- missing patches for b_unify from the 2.5 series
braam [Tue, 15 Jul 2003 08:27:59 +0000 (08:27 +0000)]
- add the unhash method to the kernel
braam [Tue, 15 Jul 2003 07:36:41 +0000 (07:36 +0000)]
-make fixes for 2.5, almost there I think
braam [Tue, 15 Jul 2003 03:51:14 +0000 (03:51 +0000)]
- 2.5 kernel patches from ad for unify
braam [Mon, 14 Jul 2003 23:42:10 +0000 (23:42 +0000)]
- merge b_unify into b_ad. This makes b_ad almost ready to be landed
on b_unify and we may just do that next.
braam [Wed, 9 Jul 2003 18:39:59 +0000 (18:39 +0000)]
- another merge of b_devel into b_ad
braam [Tue, 8 Jul 2003 22:40:40 +0000 (22:40 +0000)]
- we could call this "merely a merge from b_devel". Notes:
- lprocfs is not managable
- this oopses on mount, will be fixed soon.
pschwan [Thu, 3 Jul 2003 19:54:47 +0000 (19:54 +0000)]
Merge b_devel into b_orphan (
20030703 update tag)
pschwan [Wed, 25 Jun 2003 22:38:35 +0000 (22:38 +0000)]
- fixes the maximum OST packet size, a lot of list management, and a
couple other bugs
- starts threads at OST and MDS startup
- able to complete a round of create/setattr/cancel if SENDNOW is set
- some corruption happens when I ask it to build up and cancel whole
pages; I need to run on multiple systems to start narrowing down where.
eeb [Wed, 11 Jun 2003 22:28:16 +0000 (22:28 +0000)]
* cp /etc/termcap /mnt/lustre runs OK
eeb [Wed, 11 Jun 2003 18:05:38 +0000 (18:05 +0000)]
* Merged latest b_devel
* **** NOT TESTED ****
pschwan [Tue, 10 Jun 2003 10:48:23 +0000 (10:48 +0000)]
b_devel -> b_orphan merge for
20030610
adilger [Wed, 4 Jun 2003 21:34:50 +0000 (21:34 +0000)]
Merge from b_devel -> b_orphan
DEVEL_ORPHAN_UPDATE_PARENT_20030528 -> DEVEL_ORPHAN_UPDATE_PARENT_20030604
This is a non-functional commit, with osc/osc_request.c not compiling because
it is missing the "llog_commit_master" data struct, and does not have any
commit threads started yet. I'm beginning to think the right thing to do
here is to create a single global llog_commit_master struct for the OSC, and
start a single thread at module setup time, and clean it all up at module
unload time, but I could be convinced otherwise.
adilger [Tue, 3 Jun 2003 07:38:10 +0000 (07:38 +0000)]
Add THREAD_NAME compat macro, to remove duplication of UML thread-name stuff.
Make list_move() available if not defined (based on presence of
list_for_each_entry(), which is also defined in 2.5).
adilger [Wed, 28 May 2003 08:50:46 +0000 (08:50 +0000)]
Merge b_devel->b_orphan:
DEVEL_ORPHAN_UPDATE_PARENT_20030522->DEVEL_ORPHAN_UPDATE_PARENT_20030528
zab [Mon, 26 May 2003 22:50:12 +0000 (22:50 +0000)]
- bring recent b_devel changes into b_ad, only minor corrections needed. the
only significant of which was making sure changes in super.c made it to
b_ad's llite_lib.c
adilger [Fri, 23 May 2003 07:02:54 +0000 (07:02 +0000)]
Merge b_devel -> b_orphan.
DEVEL_ORPHAN_UPDATE_PARENT_20030522 -> DEVEL_ORPHAN_UPDATE_PARENT_20030522
girish_c [Tue, 20 May 2003 12:04:49 +0000 (12:04 +0000)]
*** empty log message ***
freym [Mon, 19 May 2003 17:55:32 +0000 (17:55 +0000)]
merged b_ad into b_qos
meijia [Mon, 19 May 2003 17:05:21 +0000 (17:05 +0000)]
merge b_devel into b_eq:
DEVEL_EQ_UPDATE_PARENT_20030519 -> DEVEL_EQ_UPDATE_CHILD_20030519
only resolve confliction.
pschwan [Sun, 18 May 2003 17:05:46 +0000 (17:05 +0000)]
merged b_devel into b_proto,
20030518
adilger [Thu, 15 May 2003 17:02:03 +0000 (17:02 +0000)]
Merge b_devel (ORPHAN_BASE6 -> ORPHAN_BASE7) to
b_orphan (DEVEL_ORPHAN_UPDATE_CHILD_20030514).
rread [Tue, 13 May 2003 00:10:51 +0000 (00:10 +0000)]
b_devel -> b_proto
from DEVEL_PROTO_UPDATE_PARENT_20030512
to DEVEL_PROTO_UPDATE_CHILD_20030512
meijia [Mon, 12 May 2003 09:45:15 +0000 (09:45 +0000)]
merge b_devel to b_eq
eeb [Wed, 7 May 2003 17:03:16 +0000 (17:03 +0000)]
* Updated ptlrpc_abort_reply() to not simply LBUG(). The current
implementation has to poll for the SENT event at 1 second intervals, since
the SENT callback cannot touch any state in the request.
* Blindly wrapped all request flag updates in an IRQ spinlock (or just a
spinlock if already at IRQ priority). My head hurt too much trying to
determine whether it was safe to update unlocked, and we know this is
safe. Maybe we should split out "owner" and "notifier" flag usage to
separate words.
adilger [Sun, 4 May 2003 00:33:13 +0000 (00:33 +0000)]
Merge from b_devel to b_orphan (ORPHAN_BASE5 -> ORPHAN_BASE6).
pschwan [Sat, 3 May 2003 06:01:00 +0000 (06:01 +0000)]
merge b_devel into b_proto
from DEVEL_PROTO_UPDATE_PARENT_20030502 tag on b_devel
to DEVEL_PROTO_UPDATE_CHILD_20030502 tag on b_proto
meijia [Sat, 3 May 2003 03:18:17 +0000 (03:18 +0000)]
merge b_devel -> b_eq
mmex [Mon, 28 Apr 2003 18:48:47 +0000 (18:48 +0000)]
Merge ... b_devel into b_kernel. This is now branched off of BASE_KERNAL2 (yes there is a typo in the tag name), which is b_devel as of Apr 27.
eeb [Fri, 25 Apr 2003 10:54:28 +0000 (10:54 +0000)]
* removed struct obd_brw_set from non-async obd_brw()
* struct ptlrpc_request::rq_lock is a new IRQ spinlock to serialise request
flag updates by the I/O thread and network callbacks.
* Added some #warnings: I/O thread v. recovery thread concurrency may need
to be reviewed.
* The rq_replied flag is now set by the network when the reply is
received. ptlrpc_check_reply() is now a pure predicate (i.e. no side
effects).
* struct ptlrpc_request has a new 'phase' member to show what state the
request is in {new, doing the rpc, interpreting the reply, complete}.
* Client-side RPC aborts should use ptlrpc_unregister_reply(). This
should only be called in thread context. When it return, it's safe to
retry.
* Server-side RPC aborts should use ptlrpc_abort_reply().
* All of the bulk per-desc and per-page callbacks have been removed.
* Client-side bulk descriptors are owned by the struct pltrpc_request.
When the descriptor has been initialised (including setting either its
bd_get_source or its bd_put_sink flags), point the request at it.
ptlrpc_queue_wait() / ptlrpc_set_wait() take care of registering and
unregistering the memory described by the descriptor, and it is freed
when the request is freed.
* Server side bulk descriptors are owned by the handler that creates them.
Completion is flagged by ptlrpc_bulk_complete(), but if you get bored,
ptlrpc_abort_bulk() synchronously prizes it away from the network so
the caller can ptlrpc_free_bulk().
* Removed struct pltrpc_request_set refcounts. A spinlock
guarantees network callback completion before the waiting thread can
nuke the struct.
* Ensured asynch replies are swabbed, just like synch ones.
* tagged eeb_eeb_backoff_tag just prior to checkin
adilger [Thu, 24 Apr 2003 08:28:50 +0000 (08:28 +0000)]
Merge from b_devel -> b_orphan (up to ORPHAN_BASE5 tag on b_devel)
wangdi [Wed, 16 Apr 2003 12:51:46 +0000 (12:51 +0000)]
add patch dev_2.4.20
pschwan [Tue, 15 Apr 2003 00:03:40 +0000 (00:03 +0000)]
merge b_devel into b_proto
pschwan [Sat, 12 Apr 2003 20:06:58 +0000 (20:06 +0000)]
- merge b_devel into b_proto (incl b_swab changes)
- fix leak-all-bulk-pages bug
- Mike fixed a bug where a failed ptlrpc_queue_wait send buf would
fail, but the user would still have to wait for the timeout to pass
before being able to interrupt
adilger [Thu, 10 Apr 2003 09:22:53 +0000 (09:22 +0000)]
Merge from b_devel.
mmex [Wed, 9 Apr 2003 18:27:56 +0000 (18:27 +0000)]
These update the patch files so that they apply cleanly in the
PNNL world.
mmex [Wed, 9 Apr 2003 17:55:44 +0000 (17:55 +0000)]
Added a hp-pnnl-2.4.20 series to the kernel patchs and fixed up the patch
files to work correctly.
nfshp [Tue, 8 Apr 2003 09:35:23 +0000 (09:35 +0000)]
the last merge b_devel to b_cray
eeb [Sun, 6 Apr 2003 14:44:11 +0000 (14:44 +0000)]
* Up to date b_devel (SWAB_LAND)
nfshp [Fri, 4 Apr 2003 16:16:40 +0000 (16:16 +0000)]
merge b_devel into b_cray
nfshp [Tue, 1 Apr 2003 16:23:16 +0000 (16:23 +0000)]
merge b_devel into b_cray
pschwan [Sun, 2 Mar 2003 05:19:20 +0000 (05:19 +0000)]
merge 0.5.20.3 b_devel into b_io (ready to land)
zab [Thu, 27 Feb 2003 21:27:57 +0000 (21:27 +0000)]
- bring b_io up to b_devel
zab [Tue, 25 Feb 2003 01:09:40 +0000 (01:09 +0000)]
- bring b_devel changes into b_io in preparation for file size fixes
zab [Sat, 15 Feb 2003 20:21:38 +0000 (20:21 +0000)]
- rebase b_io against HEAD in preparation for the latest write cache code
adilger [Fri, 7 Feb 2003 09:48:28 +0000 (09:48 +0000)]
Remove CONFIG_DEV_RDONLY so Mike doesn't lose his sanity.
adilger [Mon, 3 Feb 2003 19:21:16 +0000 (19:21 +0000)]
Fix typo for patch series.
adilger [Mon, 3 Feb 2003 18:31:07 +0000 (18:31 +0000)]
Update kernel patches for hp kernel.
adilger [Sat, 1 Feb 2003 07:42:38 +0000 (07:42 +0000)]
Add invalidate-show patch to kernel.
Move docs to a more noticable place.
braam [Mon, 20 Jan 2003 04:20:56 +0000 (04:20 +0000)]
- kernel patches for vanilla 2.4.19, 2.4.19-hp2-2 (not tested much)
- tiny fix to dcache.c
adilger [Thu, 9 Jan 2003 20:56:07 +0000 (20:56 +0000)]
Merge from HEAD to b_intel in preparation for merge of b_intel to b_md.
rread [Thu, 26 Dec 2002 18:19:46 +0000 (18:19 +0000)]
Move the llecho.sh config into seperate script so it now works just like
llmount.sh
rread [Wed, 18 Dec 2002 23:58:10 +0000 (23:58 +0000)]
* update HEAD to b_config
nfshp [Tue, 17 Dec 2002 13:19:09 +0000 (13:19 +0000)]
merge HEAD to b_io branch. (seems quite a large changes from
b_md and b_recover)
cvs2svn [Tue, 17 Dec 2002 06:19:44 +0000 (06:19 +0000)]
This commit was manufactured by cvs2svn to create branch 'unlabeled-1.1.8'.
pschwan [Tue, 17 Dec 2002 06:19:37 +0000 (06:19 +0000)]
move patch-2.4.18-chaos22 (aka EXTN_FIXES) to extN/
pschwan [Tue, 17 Dec 2002 05:50:04 +0000 (05:50 +0000)]
land b_md onto HEAD. almost entirely small cleanups and miscellaneous fixes,
plus:
- fix lock order deadlock on OST (O/R i_sem before journal ops, 478)
- fix follow_symlink bugs (437)
- preparation for some ldlm race fixes (419)
braam [Tue, 17 Dec 2002 05:44:32 +0000 (05:44 +0000)]
- series file for Chaos kernel.
pschwan [Mon, 16 Dec 2002 20:57:38 +0000 (20:57 +0000)]
remove obsolete 'patches' directory from HEAD
braam [Mon, 16 Dec 2002 17:01:10 +0000 (17:01 +0000)]
- Phil's intent fix for middle of symlink lookups.
- Update version
shaver [Mon, 16 Dec 2002 02:04:25 +0000 (02:04 +0000)]
Land b_recovery (other than WIP lock-replay work).
- b=512,others: fully reactivate OST imports after reconnection.
- b=513(partial): make sure client sees out -ENOTCONN from mds_handle
- b=515: more graceful error handling for truncating on dead OST
- b=474: don't error out just because a file is striped across a dead OST; only
if we actually try to access it
- b=410: fix garbage sizes when stat(2)ing a file that includes a stripe on a
dead OST
- print console diagnostic for completion-timeout client recovery
- MSG_RESENT for requests that were retransmitted
braam [Mon, 16 Dec 2002 01:17:16 +0000 (01:17 +0000)]
- add more patch series files. We have now obsoleted most of the
patches directory with the new regime.
hitao [Sun, 15 Dec 2002 07:40:12 +0000 (07:40 +0000)]
file lin-2.5.44.txt was initially added on branch b_lin25.
hitao [Sun, 15 Dec 2002 07:40:11 +0000 (07:40 +0000)]
file lin-2.5.44 was initially added on branch b_lin25.
hitao [Sun, 15 Dec 2002 07:40:10 +0000 (07:40 +0000)]
file lin-2.5.44.pc was initially added on branch b_lin25.
hitao [Sun, 15 Dec 2002 07:40:09 +0000 (07:40 +0000)]
file lin-2.5.44.patch was initially added on branch b_lin25.
pschwan [Sat, 14 Dec 2002 19:40:17 +0000 (19:40 +0000)]
land b_md onto HEAD. the highlights:
- fstat didn't correctly synchronize attributes (399)
- server must handle lock cancellation during blocking AST prep (487)
- bulk descriptors were free()d too soon (511)
- fix paths in lconf, which would load incorrect modules (451, 507)
- fix confusing lconf 'host not found' error message (386)
braam [Sat, 14 Dec 2002 13:56:10 +0000 (13:56 +0000)]
- this time added with -ko
braam [Sat, 14 Dec 2002 13:55:13 +0000 (13:55 +0000)]
- redo the addition of the patch, forgot -ko
braam [Sat, 14 Dec 2002 13:43:37 +0000 (13:43 +0000)]
add patch fixups for Red Hat 2.4.18-18.8.0 kernel
thantry [Sat, 14 Dec 2002 02:49:44 +0000 (02:49 +0000)]
file llparser.pm was initially added on branch b_intel.
shorthair [Thu, 12 Dec 2002 14:11:26 +0000 (14:11 +0000)]
file lustre-2.5-from811-812.patch was initially added on branch b_lin25.
shorthair [Thu, 12 Dec 2002 14:11:25 +0000 (14:11 +0000)]
file lustre-2.5-from809-812.patch was initially added on branch b_lin25.
nfshp [Thu, 12 Dec 2002 12:53:09 +0000 (12:53 +0000)]
file iod.c was initially added on branch b_io.
thantry [Wed, 11 Dec 2002 18:58:13 +0000 (18:58 +0000)]
Merged from HEAD on 12/11/2002.
pschwan [Wed, 11 Dec 2002 16:21:47 +0000 (16:21 +0000)]
backout #2
narasimd [Wed, 11 Dec 2002 06:53:25 +0000 (06:53 +0000)]
Merging to head, BUG#405
yanhao [Wed, 11 Dec 2002 01:07:56 +0000 (01:07 +0000)]
yanhao: check in again for bug 461 wrong path in lconf