Whamcloud - gitweb
fs/lustre-release.git
21 years ago[liblustre]: merge LiuKai's cygwin patch
ericm [Thu, 21 Aug 2003 08:57:39 +0000 (08:57 +0000)]
[liblustre]: merge LiuKai's cygwin patch

21 years agoif the connection fails for whatever reason and we pass an invalid
phil [Thu, 21 Aug 2003 06:40:50 +0000 (06:40 +0000)]
if the connection fails for whatever reason and we pass an invalid
export handle into oscc_init, don't dereference a NULL

21 years agofix double-unlock in osccd_main
phil [Thu, 21 Aug 2003 06:32:25 +0000 (06:32 +0000)]
fix double-unlock in osccd_main

21 years agoanother spinlock typo
phil [Thu, 21 Aug 2003 06:10:08 +0000 (06:10 +0000)]
another spinlock typo

21 years ago- initialize spinlocks
phil [Thu, 21 Aug 2003 06:08:25 +0000 (06:08 +0000)]
- initialize spinlocks
- fix spinlock name typo
- uninitialized variable warning

21 years agooscc_init should not be calling oscc_precreate, because this happens
phil [Thu, 21 Aug 2003 04:18:18 +0000 (04:18 +0000)]
oscc_init should not be calling oscc_precreate, because this happens
before recovery.

21 years agoone big fix:
phil [Wed, 20 Aug 2003 23:32:38 +0000 (23:32 +0000)]
one big fix:
- in osc_create, set oa->o_id AFTER copying oscc_oa overtop of it, so
  that we don't give every new file object 6

lots of little fixes:
- call mds_lov_set_nextid() from the end of mds_postsetup()
- ...make that safe by calling mds_lov_connect() at the beginning of
  mds_lov_set_nextid()
- in filter_create, stop holding the directory lock for reply acks.
  it's not needed anymore, and causes obvious deadlock during precreation
- clean up the error handling in filter_precreate
- start precreating when there are fewer than "kick_barrier" objects
  remaining, and then create "grow_count" additional objects

moving on to larger tests:
- initial_create_count is now 100
- kick_barrier is now 50
- grow_count is now 100

21 years ago- a bunch of "fixes" on b_llpmd
braam [Wed, 20 Aug 2003 15:38:56 +0000 (15:38 +0000)]
- a bunch of "fixes" on b_llpmd
  - hook initialization and orphan removal functions
    [this needs review and redo, after some config issues are settled.]
    [nasty hackery continues in this area until lconf sees a change.]

    - things that would help:
     1. pass the lov_uuid to the MDS at setup time
     2. call the mds ioctl to do lovinfo setup during setup
        [add yet another parameter to mds_setup?]
        [currently this is a race]
     3. get lov/osc devices set up before the mds is set up.

     All 3 look doable to me.

  - moved code from the class ioctl handler to genops, to begin a
  framework for in-kernel configuration.  For now we have:
    - class_newdev,
    - class_attach
    - class_setup functions.

  - added Alex pre-creation code

NOTE: oscc_precreate doesn't wake up unless has_objects returns
true. That prevents pre-create to be called multiple times. Should be
fixed, but it's unlikely we will see this bug again

21 years agoNFS export of Lustre FS
girishc [Wed, 20 Aug 2003 07:43:26 +0000 (07:43 +0000)]
NFS export of Lustre FS
-Some of the review comments addressed

21 years agoRemove "ERR" trap, which doesn't work in my UML, nor at PNNL. r=robert
adilger [Tue, 19 Aug 2003 23:34:57 +0000 (23:34 +0000)]
Remove "ERR" trap, which doesn't work in my UML, nor at PNNL. r=robert

Use "MOUNT" and "DIR" to be more in line with sanity.sh.  This also
allows us to run these tests in a subdir instead of the root, in case
it makes a difference (it doesn't currently).

Use checkstat instead of "ls" so that we can do some verification of results.

Don't pause for GDB debugging symbols to be loaded.

21 years agoClose and unlink test file if we didn't encounter any errors.
adilger [Tue, 19 Aug 2003 22:22:50 +0000 (22:22 +0000)]
Close and unlink test file if we didn't encounter any errors.
Set default iteration count to 50k instead of 100k.

21 years agoAdd truncate-to-zero to multiop commands.
adilger [Tue, 19 Aug 2003 22:19:25 +0000 (22:19 +0000)]
Add truncate-to-zero to multiop commands.

21 years agoRemove unused variable.
adilger [Tue, 19 Aug 2003 22:15:19 +0000 (22:15 +0000)]
Remove unused variable.

21 years ago- #1642
behlendo [Tue, 19 Aug 2003 19:13:26 +0000 (19:13 +0000)]
- #1642
  - backport some small cleanups to the dentry refcounts and cleanup
  - Zach's fix from llpio which fixes a race condition in the
    partial read
- #1592 fixes trying to read a single page that was entirely past
  EOF, we would later oops in commitrw on the OST trying to dput
  a NULL dentry.  This fixes the 'make' issue Richard discovered.
- #1505 patch to quiet four messages
  - "lustre_commitd quitting"
  - "processing error -107" after server is restarted
  - no need to print statfs() errors to the console
  - print one reconnect message instead of two
- Adjusted version_tag.pl to use portals version string if no
  /CVS/Tag exists.  This resolves rpms using only the HEAD tag.

21 years agob=1777
rread [Tue, 19 Aug 2003 19:04:17 +0000 (19:04 +0000)]
b=1777

* Add test to replay open after chmod

21 years ago[liblustre]: merge LiuKai's cygwin patch
ericm [Tue, 19 Aug 2003 12:08:52 +0000 (12:08 +0000)]
[liblustre]: merge LiuKai's cygwin patch
 - add portals/include/cygwin-ioctl.h
 - don't compile accepter.c in liblustre
 - add missing le-cpu convert macro
 - use windows API directly at some places

21 years ago- remainder of the pre-create code (more or less)
braam [Tue, 19 Aug 2003 08:28:36 +0000 (08:28 +0000)]
- remainder of the pre-create code (more or less)
 - routines to set next_id
 - routines to clear orphans
 - routines to set growth count
- minor cleanups

21 years ago- more small fixes for pre-creation, to help bzzz make progress.
braam [Tue, 19 Aug 2003 07:08:47 +0000 (07:08 +0000)]
- more small fixes for pre-creation, to help bzzz make progress.
 - set o_id as a hint upon creation.

21 years ago- new series uml_2.6.0_test3
alex [Mon, 18 Aug 2003 09:40:35 +0000 (09:40 +0000)]
- new series uml_2.6.0_test3
- latest uml patch against 2.6.0-test3

21 years ago- oscc_has_objects needs to do a signed comparison, or osc_create will
phil [Sun, 17 Aug 2003 20:52:58 +0000 (20:52 +0000)]
- oscc_has_objects needs to do a signed comparison, or osc_create will
  spin forever
- print some lines to the log about which objects are being returned,
  and which were preallocated
- default to starting at 2 instead of 0, because that's the first
  object that the OST will actually create.  This will be replaced
  shortly with an actual understanding of which objects are available
  at startup
- exclude tests 34c-34e until I fix "create with existing size"; the
  rest of sanity passes if you start with a newly formatted OST (see
  last point)

21 years agoFix some tab damage and missing modeline
phil [Sun, 17 Aug 2003 17:42:36 +0000 (17:42 +0000)]
Fix some tab damage and missing modeline

21 years agoseveral fixes to the pre-create daemon. I think it survives a few creates
braam [Sun, 17 Aug 2003 15:25:32 +0000 (15:25 +0000)]
several fixes to the pre-create daemon.  I think it survives a few creates
now.

21 years ago[liblustre]: liblustre sanity script fix.
ericm [Sun, 17 Aug 2003 07:20:00 +0000 (07:20 +0000)]
[liblustre]: liblustre sanity script fix.

21 years ago[liblustre]: set intent pointer to NULL after intent released.
ericm [Sun, 17 Aug 2003 07:19:04 +0000 (07:19 +0000)]
[liblustre]: set intent pointer to NULL after intent released.

21 years agob=1541: fix LASSERT on mismatched transnos during open-unlink testing
shaver [Sat, 16 Aug 2003 19:38:35 +0000 (19:38 +0000)]
b=1541: fix LASSERT on mismatched transnos during open-unlink testing
r=phik

- the open request now keeps a pair of pointers in its replay data:
  - the och that has always been there for fixing up open filehandles
  - the close request that has been created to balance it, which may or may
    not have made it over the wire yet, for fixing up the filehandle cookie it
    sends to the MDS

- the explicit updating of the close req replaces the previous walking of the
  sending list, which was both insufficient _and_ excessive

- suppress the "asked for EA, got none" message in the case of a 0-nlink file

- when processing an INTENT_ONLY enqueue request, return a successful 0 rather
  than ELDLM_LOCK_ABORTED, which only confused the clients

- when opening by FID, be sure to set the intent disposition flags
  appropriately

- triple the timeout, instead of doubling it, for the replay-completed PING, to
  better suit the numerology of the pinger and timeout values

21 years agoUse compat. macros to fix building w/ rh-2.4.20 series.
jacob [Sat, 16 Aug 2003 19:05:29 +0000 (19:05 +0000)]
Use compat. macros to fix building w/ rh-2.4.20 series.

r=phil

21 years ago- pre-creation is beginning to work with these small changes to the
braam [Sat, 16 Aug 2003 14:49:46 +0000 (14:49 +0000)]
- pre-creation is beginning to work with these small changes to the
state machine.

21 years ago- initial check in of pre-creation code including:
braam [Sat, 16 Aug 2003 05:08:19 +0000 (05:08 +0000)]
- initial check in of pre-creation code including:
  - alex server side code (still needs to update last allocated id on disk)
  - peter's client code (still needs debugging and some finishing touches)
  - the whole thing needs some thought viz a viz:
    - recovery
    - error notificaiton from the daemon to the caller.
- removed unused commit_cbd, maybe one day we will use it and re-surrect,
  but I think we need this at the osc/mdc level, not at llite level.

21 years ago- #1763 fixes timestamps on files getting advanced incorrectly.
behlendo [Fri, 15 Aug 2003 17:59:35 +0000 (17:59 +0000)]
- #1763 fixes timestamps on files getting advanced incorrectly.
- #1639 client VFS truncate/write lock inversion.  This caused
  the bug in Joe Koning music code.

21 years ago- give llite moderately more reasonable default watermarks so that it can
zab [Thu, 14 Aug 2003 21:00:37 +0000 (21:00 +0000)]
- give llite moderately more reasonable default watermarks so that it can
  fill a fair sized lov in a burst.
- be more careful about unlocking writepage's page as it returns an error
- allocate writeback page state with NOFS so we don't descend into filesystems
  when memory is short

21 years agoUpdate header to give compilation and running instructions.
adilger [Thu, 14 Aug 2003 17:41:21 +0000 (17:41 +0000)]
Update header to give compilation and running instructions.

21 years agoNFS Export of Lustre File System
girishc [Thu, 14 Aug 2003 17:20:41 +0000 (17:20 +0000)]
NFS Export of Lustre File System
-Contains kernel patch nfs_export_kernel-2.4.20.patch
-Lustre Modifications
    fh_to_dentry/dentry_to_fh implementation
    IT_CREAT Implementation (As zero sized file creation on NFS client never calles nfsd_open)

21 years agoCheck for MAP_FAILED return from mmap instead of NULL.
adilger [Thu, 14 Aug 2003 16:26:58 +0000 (16:26 +0000)]
Check for MAP_FAILED return from mmap instead of NULL.
#define _GNU_SOURCE to get O_DIRECT/O_DIRECTORY
In test_brw define a default pg_vec if unset.

21 years agoutility to get GM global nid from node name
mdoyle [Thu, 14 Aug 2003 14:07:53 +0000 (14:07 +0000)]
utility to get GM global nid from node name

21 years agoPaged IO support for scimacnal.
nikke [Thu, 14 Aug 2003 09:13:33 +0000 (09:13 +0000)]
Paged IO support for scimacnal.
Closes bug#1347.

21 years agoZach and I both agreed that we should keep short IO sanity tests in sanity.sh
adilger [Thu, 14 Aug 2003 06:30:35 +0000 (06:30 +0000)]
Zach and I both agreed that we should keep short IO sanity tests in sanity.sh
or sanityN.sh so that they are always run during tests.

Remove tests from ALWAYS_EXCEPT now that their bugs have been fixed and the
tests pass.  Please keep ALWAYS_EXCEPT and the comment above it in sync.

21 years agomerge b_filterio into b_llpio; b_filterio soon to be deleted, given
phil [Thu, 14 Aug 2003 05:54:47 +0000 (05:54 +0000)]
merge b_filterio into b_llpio; b_filterio soon to be deleted, given
that b_llpio is a superset

21 years agob=1513
rread [Wed, 13 Aug 2003 20:16:21 +0000 (20:16 +0000)]
b=1513
r=shaver
* add conn_cnt to import, export, and lustre_msg.  The conn_cnt is
increased by the client whenever it connects or reconnects.  The
server ignores failed BRWs with an old conn_cnt.

Also, when a bulk is resent, the xid is changed, so the previous one
will definitely fail.

21 years agomerge HEAD changes into b_devel
phil [Wed, 13 Aug 2003 19:53:17 +0000 (19:53 +0000)]
merge HEAD changes into b_devel

21 years agoUpdate write_append_truncate program:
adilger [Wed, 13 Aug 2003 19:00:52 +0000 (19:00 +0000)]
Update write_append_truncate program:
- check all return codes
- add understandable error messages
- can select the test file without recompiling.
- can select iteration count (default 100000 loops)
- have some progress output during run (ala fsx)

21 years ago- #1592 fixes trying to read a single page that was entirely past
behlendo [Wed, 13 Aug 2003 18:37:23 +0000 (18:37 +0000)]
- #1592 fixes trying to read a single page that was entirely past
  EOF, we would later oops in commitrw on the OST trying to dput
  a NULL dentry.  This fixes the 'make' issue Richard discovered.
- #1765 fixes MDS cleanup bug where we have an outstanding reply
  at disconnect time.  Remove all exp_outstanding_reply
  manipulation in mds_disconnect.
- #1732 fixes the lprocfs unresolved symbol error.

21 years ago- trivial sanity asserts in llite page accounting
zab [Wed, 13 Aug 2003 17:20:32 +0000 (17:20 +0000)]
- trivial sanity asserts in llite page accounting
- don't leak ocp's in some failure paths
- make sure read-ahead doesn't orphan a locked page
- add some mmap goo to multiop (this is going to conflict, I bet)

21 years agoFirst test for io sanity - force a zero-length read from a sparse stripe.
adilger [Wed, 13 Aug 2003 17:00:05 +0000 (17:00 +0000)]
First test for io sanity - force a zero-length read from a sparse stripe.

21 years agoMPI distributed write/append/truncate coherency stress test - original version.
adilger [Tue, 12 Aug 2003 23:18:11 +0000 (23:18 +0000)]
MPI distributed write/append/truncate coherency stress test - original version.

21 years ago- backport some small cleanups to the dentry refcounts and cleanup
behlendo [Tue, 12 Aug 2003 22:45:14 +0000 (22:45 +0000)]
- backport some small cleanups to the dentry refcounts and cleanup
- Zach's fix from llpio which fixes a race condition in the partial read path

21 years ago- Bug #1749 patch
behlendo [Tue, 12 Aug 2003 22:33:09 +0000 (22:33 +0000)]
- Bug #1749 patch

21 years agob=1720
rread [Tue, 12 Aug 2003 20:27:41 +0000 (20:27 +0000)]
b=1720
r=adilger

* free reply buffer whether we have ack locks or not.

21 years ago- Patch for bug #1600, MDS server data isn't written at setup. This is
behlendo [Tue, 12 Aug 2003 19:46:08 +0000 (19:46 +0000)]
- Patch for bug #1600, MDS server data isn't written at setup.  This is
  only an issue when mounting the MDS for the first time.

21 years agoRevert changes committed to b_devel for MDS-creates-objects.
adilger [Tue, 12 Aug 2003 17:45:39 +0000 (17:45 +0000)]
Revert changes committed to b_devel for MDS-creates-objects.
Changes will be committed into b_llpmd instead.

We could, however, remove the OST open/close RPCs from b_devel as a
starting point.

21 years ago- version_tag.pl now uses the portals version string if no CVS/Tag exists.
behlendo [Tue, 12 Aug 2003 17:03:09 +0000 (17:03 +0000)]
- version_tag.pl now uses the portals version string if no CVS/Tag exists.
  This resolves building rpms which all claim to have the HEAD tag.
- LLNL_ChangeLog updates
- Tagged version llnl_4devel

21 years agoAdd sleeping write/verification test to CVS.
adilger [Tue, 12 Aug 2003 16:59:47 +0000 (16:59 +0000)]
Add sleeping write/verification test to CVS.
usage: sleeptest [basename [sleeptime]]

Fixes a couple of bugs in the current test:
- assumes that there is 100 bytes after offset & ~4095 (fails at iter 233)
- assumes that "Line" appears in the output (fails if < 60 bytes read)

21 years ago- #1751 fix LBUG/bad LOV EA if lov_create fails because of inactive OSCs.
behlendo [Tue, 12 Aug 2003 16:02:30 +0000 (16:02 +0000)]
- #1751 fix LBUG/bad LOV EA if lov_create fails because of inactive OSCs.
  We hit this bug on ALC Monday Afternoon (8/11/03), the fix has been
  tested in b_devel since 07/28.

21 years ago- this contains much of the pre-create code
braam [Tue, 12 Aug 2003 10:22:33 +0000 (10:22 +0000)]
- this contains much of the pre-create code
  - extensive changes to mds_open to call obd_create
  - important changes to the client to avoid obd_create
  - changes to intent locking to avoid giving a client a lock
- deficiencies:
  - there is an open unlink problem (sanity 31).  I think this was lurking and is now exposed
  - lstripe needs some thought; doesn't work right now
  - further refactoring of the object creation in mds_open desirable and forthcoming.

21 years ago- make sure the sync commit write gets the rpc's error code
zab [Mon, 11 Aug 2003 22:04:51 +0000 (22:04 +0000)]
- make sure the sync commit write gets the rpc's error code
- be sure to set the page up to date if commit_write succeeds
- get a page ref when we start async ocp io and drop the ref as the io
  completes
- drop some asserts left over from early refactoring passes
- add a simple io data consistency check to sanity.sh

21 years agor=adilger
rread [Mon, 11 Aug 2003 20:06:27 +0000 (20:06 +0000)]
r=adilger

* Add a flag to mds_mfd_close to prevent unlinking files when we are
  cleaning up with the --failover flag.

21 years ago* test_6 fails if any files are found at the end.
rread [Mon, 11 Aug 2003 19:58:33 +0000 (19:58 +0000)]
* test_6 fails if any files are found at the end.
* test_8 does fchmod after unlink

21 years ago* yank back stat help
rread [Mon, 11 Aug 2003 18:15:24 +0000 (18:15 +0000)]
* yank back stat help

21 years ago* new: t fchmod (set mode to 0)
rread [Mon, 11 Aug 2003 18:01:20 +0000 (18:01 +0000)]
* new: t fchmod (set mode to 0)
* add help for new options

21 years agoupdate 2.4.18 patches for bug 1705
wangdi [Mon, 11 Aug 2003 05:53:22 +0000 (05:53 +0000)]
update 2.4.18 patches for bug 1705

21 years ago* 2 tests for open-unlink, replay close testing.
rread [Fri, 8 Aug 2003 23:49:11 +0000 (23:49 +0000)]
* 2 tests for open-unlink, replay close testing.

21 years agor=shaver
jacob [Fri, 8 Aug 2003 22:53:42 +0000 (22:53 +0000)]
r=shaver
 - sort switch statement for sanity
 - add read/write (r,w)
 - add seek-to-zero (z)

21 years ago- #1158 fix for MDS ldlm_intent_policy
behlendo [Fri, 8 Aug 2003 21:07:20 +0000 (21:07 +0000)]
- #1158 fix for MDS ldlm_intent_policy
- LLNL ChangeLog updates

21 years ago* use multiop and kill keepopen
rread [Fri, 8 Aug 2003 17:02:17 +0000 (17:02 +0000)]
* use multiop and kill keepopen

21 years agoliblustre: fix timeout during mknod tests.
ericm [Fri, 8 Aug 2003 10:14:23 +0000 (10:14 +0000)]
liblustre: fix timeout during mknod tests.

21 years ago* add keepopen.py, which holds a file open until it recieves a
rread [Fri, 8 Aug 2003 10:04:38 +0000 (10:04 +0000)]
* add keepopen.py, which holds a file open until it recieves a
  SIGUSR1, then it reads the file and closes it, returning an error
  if anything fails.

* replay_single.sh/test_8 uses keepopen to test close replay after
  unlink. So far, the test actually passes, but from reading the log I
  can see the MDS_CLOSE is not replayed, because it doesn't get a
  transno. Not sure how to actually test for this.

21 years agob=1614
braam [Fri, 8 Aug 2003 09:51:03 +0000 (09:51 +0000)]
b=1614
r=phil

This bug fix was complicated by the fact that many code paths that
never had run now execute.  We changed intent error returns (already
committed), and error returns inside mds_open.

This includes a kernel fix but no version change, as I made it
backward compatible.

Fixes for open:
 - kernel passes FMODE_EXEC to lustre, which sends it to MDS
 - move all open/close code to mds_open.c, make almost everything static
 - arrange for allow/deny/get/put write access in mds
 - fix several problems with error handling in mds_dentry_open:
   - don't do a wrong mntput
   - do a dget for readability
 - add test to sanityN and sanity
 - simplify confusing permission checking in mds_open with helper

21 years ago* each test is independent and cleans up after itself
rread [Fri, 8 Aug 2003 06:16:53 +0000 (06:16 +0000)]
* each test is independent and cleans up after itself
* test_4: test an error replay

21 years agob=1505
phil [Thu, 7 Aug 2003 18:14:02 +0000 (18:14 +0000)]
b=1505
r=shaver
- Remove confusing and uninteresting "lustre_commitd quitting" message
- downgrade the -ENOTCONN "processing error" message to D_HA from D_ERROR

21 years agoadd a simple liblustre sanity test script
ericm [Thu, 7 Aug 2003 10:27:58 +0000 (10:27 +0000)]
add a simple liblustre sanity test script

21 years agoFix minor typo in error message.
adilger [Thu, 7 Aug 2003 08:00:54 +0000 (08:00 +0000)]
Fix minor typo in error message.

21 years ago- export some simple histograms that let us watch rpc concurrency and the
zab [Wed, 6 Aug 2003 23:00:42 +0000 (23:00 +0000)]
- export some simple histograms that let us watch rpc concurrency and the
  number of pages in each rpc.  writing to the file clears them.

21 years ago- trivially export some rpc/page stats from the osc so we can get some
zab [Wed, 6 Aug 2003 21:30:32 +0000 (21:30 +0000)]
- trivially export some rpc/page stats from the osc so we can get some
  insight into how llite is driving it

21 years ago- Landed b_multinet on b_llnl_devel
behlendo [Tue, 5 Aug 2003 22:13:14 +0000 (22:13 +0000)]
- Landed b_multinet on b_llnl_devel

21 years ago- Several minor changes so llnl can build rpms of the form:
behlendo [Tue, 5 Aug 2003 18:43:23 +0000 (18:43 +0000)]
- Several minor changes so llnl can build rpms of the form:

       lustre-*-VERSION-RELEASE.i386.rpm; where VERSION is the CVS tag
       and RELEASE is the kernel is was built against.

  e.g. lustre-modules-llnl_1devel-2.4.18_p4smp_68chaos.i386.rpm

- I also started a LLNL_ChangeLog for the branch to track what patches we apply.

21 years agorebase b_qos on b_devel instead of b_ad
mfrey [Mon, 4 Aug 2003 15:12:55 +0000 (15:12 +0000)]
rebase b_qos on b_devel instead of b_ad

21 years agob=1598
phil [Sun, 3 Aug 2003 23:58:40 +0000 (23:58 +0000)]
b=1598
Merge final read performance fix into b_devel

21 years agomerge HEAD into b_devel:
phil [Sun, 3 Aug 2003 20:57:17 +0000 (20:57 +0000)]
merge HEAD into b_devel:
 - fix for bug 1598
 - temporary flag-clearing until bug 974 is fixed
 - new GM nal

21 years agoImproved replay-single.sh and replay-dual.sh
pengzhao [Sun, 3 Aug 2003 16:52:29 +0000 (16:52 +0000)]
Improved replay-single.sh and replay-dual.sh

21 years agoAs best I could test, none of our tests were using stat.c, it was just
phil [Sun, 3 Aug 2003 16:39:50 +0000 (16:39 +0000)]
As best I could test, none of our tests were using stat.c, it was just
a convenient helper.  Because most of our installations have a broken
version of the 'stat' utility installed (it reports an incorrect block
size), I replaced stat.c with v3.3 of the stat utility.

21 years agovarious fixes liblustre: following the new llite
ericm [Sun, 3 Aug 2003 07:12:11 +0000 (07:12 +0000)]
various fixes liblustre: following the new llite
at this moment, many metadata can't work, and file i/o is tatally broken

21 years agoadd implementation of rbtree from linux.
ericm [Sun, 3 Aug 2003 07:06:37 +0000 (07:06 +0000)]
add implementation of rbtree from linux.

21 years agoThis commit was manufactured by cvs2svn to create branch 'unlabeled-1.1.2'.
cvs2svn [Sun, 3 Aug 2003 07:06:36 +0000 (07:06 +0000)]
This commit was manufactured by cvs2svn to create branch 'unlabeled-1.1.2'.

21 years ago - Fix build with rh 2.4.20 kernel series
jacob [Fri, 1 Aug 2003 18:37:16 +0000 (18:37 +0000)]
 - Fix build with rh 2.4.20 kernel series
r=phil
b=1663

21 years agoPortals NAL for Myrinet GM2 for Lustre (lgmnal)
mdoyle [Fri, 1 Aug 2003 09:31:53 +0000 (09:31 +0000)]
Portals NAL for Myrinet GM2 for Lustre (lgmnal)

21 years agoFix for bug 1609 on CVS HEAD.
adilger [Fri, 1 Aug 2003 09:04:28 +0000 (09:04 +0000)]
Fix for bug 1609 on CVS HEAD.

21 years agoFix ia64 O_DIRECT crash (bug 1609).
adilger [Fri, 1 Aug 2003 09:01:35 +0000 (09:01 +0000)]
Fix ia64 O_DIRECT crash (bug 1609).
- don't send less than PAGE_SIZE O_DIRECT writes...  It appears that lustre
  or the VFS does not handle this properly.  This is the only time we send
  non-page-aligned writes.
- prepare_write() takes parameters "from" and "to", and not "offset" and "len"
  (this is real source of crash - we were not mapping a buffer in this case).

21 years agoupdate changes of lookup_intent for hp kernel
wangdi [Fri, 1 Aug 2003 03:36:44 +0000 (03:36 +0000)]
update changes of lookup_intent for hp kernel

21 years ago* doh! (thanks, mike)
rread [Fri, 1 Aug 2003 00:19:57 +0000 (00:19 +0000)]
* doh!  (thanks, mike)

21 years ago* simple cleanups to get replay-single.sh to run again, also added
rread [Thu, 31 Jul 2003 23:55:56 +0000 (23:55 +0000)]
* simple cleanups to get replay-single.sh to run again, also added
  a trap ERR.

21 years agomake sanity and iozone run from llite over the lov in the new ocp
zab [Thu, 31 Jul 2003 20:29:35 +0000 (20:29 +0000)]
make sanity and iozone run from llite over the lov in the new ocp
io model.
- add ocp_off to the ocp so the lov can remap it as it passes through
- fix up the page accounting a bit 'cause truncate doesn't really know
  right know to subtract pages from the dirty or io counts
- pass our completion struct, _not_ a pointer to the complete function
- make sure to always free the ocp in commit_write_sync
- disable the watermark io until it actually does something
- make sure to alter the filehandle/o_id as we pass through the lov
- initialize our loi bits in lov_unpackmd

21 years ago- introduce tunables for the aggregate amount of cached data that llite will
zab [Thu, 31 Jul 2003 17:15:41 +0000 (17:15 +0000)]
- introduce tunables for the aggregate amount of cached data that llite will
  allow on a machine.  When we have more than 'dirty_high_water_MB' meg of
  data dirty we start async ocp io on the amount of data between the high
  water mark and 'dirty_low_water_MB'.  If we ever have more than
  'max_dirty_MB' we write synchronously in commit_write, batching with other
  dirty data in the inode.  these watermarks are checked every time we pass
  through commit_write and we account seperately for dirty pages and pages
  that are under io.

- add ocp_count to the ocp so that we can write partial pages at the end
  of the file.  make the callback func a proper type.

- remove iod.c, the few code paths we cared about in there are moved to
  rw{,24,26}.c

- turn off generic file read-ahead by setting filp->f_ramax to 0.

- have writepage and commit_write when writing sync use the plugging to batch
  their io with io started from other dirty pages in the inode's dirty list

- woops, missed a hardcoded '16'.  move to max_pags_per_rpc

There's still much to do.  The 2.6 side needs work on its many-page variants,
kiovec->bio, etc.  We need to actually start watermark io from the dirty inode
list.

21 years agoEnsure OST objids are unique even after a crash by using the mount count as
adilger [Wed, 30 Jul 2003 20:35:18 +0000 (20:35 +0000)]
Ensure OST objids are unique even after a crash by using the mount count as
a generation number.  Remove old "superhack" code.

Add a "force_sync" parameter to filter_update_server_data() and
fsfilt_write_record() so that we ensure the mount_count is safe on disk
when the OST has started.

b=294
r=phil

21 years ago* Optimized qswnal GETs use EKC rpcs to "teleport" payload (as far as
eeb [Wed, 30 Jul 2003 04:13:55 +0000 (04:13 +0000)]
*  Optimized qswnal GETs use EKC rpcs to "teleport" payload (as far as
   portals is concerned).

*  allowed qswnal to have knowledge of lib_msg_t.

*  lib_fake_reply_msg() written to allow qswnal to create replies and
   deliver them in the normal fashion (with lib_finalize()).

21 years ago- put in initial lov obd plug/async/unplug APIs. for now its giving
zab [Mon, 28 Jul 2003 20:49:56 +0000 (20:49 +0000)]
- put in initial lov obd plug/async/unplug APIs.  for now its giving
  the stripe to the osc via an explicit loi argument.  this will be
  refactored when the refactoring described in 314 arrives.

21 years ago- simple optimization for EXT3:
alex [Mon, 28 Jul 2003 12:18:00 +0000 (12:18 +0000)]
- simple optimization for EXT3:
  ext3_write_super() doesn't need to start commit as kupdated wants. this
  improves things significantly in some cases

21 years agofile ext3-no-write-super.pc was initially added on branch b_devel.
alex [Mon, 28 Jul 2003 12:17:55 +0000 (12:17 +0000)]
file ext3-no-write-super.pc was initially added on branch b_devel.

21 years agofile ext3-no-write-super.patch was initially added on branch b_devel.
alex [Mon, 28 Jul 2003 12:17:54 +0000 (12:17 +0000)]
file ext3-no-write-super.patch was initially added on branch b_devel.

21 years ago- minor changes in Andreas'es ext3 direct-io patch to be applicable for
alex [Mon, 28 Jul 2003 12:02:04 +0000 (12:02 +0000)]
- minor changes in Andreas'es ext3 direct-io patch to be applicable for
  2.4.18-chaos
- ext3-o_direct-1-2.4.18-chaos.patch have been added to chaos-2.4.18 series

21 years agofile ext3-o_direct-1-2.4.18-chaos.pc was initially added on branch b_devel.
alex [Mon, 28 Jul 2003 12:02:02 +0000 (12:02 +0000)]
file ext3-o_direct-1-2.4.18-chaos.pc was initially added on branch b_devel.

21 years agofile ext3-o_direct-1-2.4.18-chaos.patch was initially added on branch b_devel.
alex [Mon, 28 Jul 2003 12:02:01 +0000 (12:02 +0000)]
file ext3-o_direct-1-2.4.18-chaos.patch was initially added on branch b_devel.