Whamcloud - gitweb
adilger [Mon, 16 Sep 2002 22:46:35 +0000 (22:46 +0000)]
Put checks for LOV magic into LOV code, what a good idea.
adilger [Mon, 16 Sep 2002 22:42:09 +0000 (22:42 +0000)]
Don't look at the EA magic in the filesystem, but use the FLEASIZE flag.
adilger [Mon, 16 Sep 2002 22:40:50 +0000 (22:40 +0000)]
TODO message.
adilger [Mon, 16 Sep 2002 22:37:25 +0000 (22:37 +0000)]
Remove setup code from runregression-net.sh and runregression-brw.sh.
You need to set up lustre before running these scripts now.
braam [Mon, 16 Sep 2002 22:33:16 +0000 (22:33 +0000)]
- a fix for Gary's directory read-back bug. You can now create lots
of files and ls shows them. Fixes include updating the directory
size during readdir by passing back the size in the readdir call.
Every directory has at least one page, so this works.
braam [Mon, 16 Sep 2002 21:27:49 +0000 (21:27 +0000)]
- a fix for single node rm `pwd` problems. The distributed case is
more involved and we'll have to see about that.
rread [Mon, 16 Sep 2002 19:35:31 +0000 (19:35 +0000)]
* print a message when a duplicate mds or lov name is used
adilger [Mon, 16 Sep 2002 18:51:18 +0000 (18:51 +0000)]
Fix help message for --threads.
adilger [Mon, 16 Sep 2002 18:50:37 +0000 (18:50 +0000)]
Allow OBD UUID to be specified explicitly for BA OSTs.
pschwan [Mon, 16 Sep 2002 18:19:20 +0000 (18:19 +0000)]
b=609205
Reenable the throwing away of data locks on file close. This may or may not
be enough to fix 609205, depending on whether that bug is purely related to
running out of memory.
pschwan [Mon, 16 Sep 2002 18:13:03 +0000 (18:13 +0000)]
- add resource and lock counters to the namespace
- add them to the namespace proc directories (which are, tragically, in the
top level of the /proc tree until lprocfs gets sorted out)
- fix a dangling resource refcount bug in ldlm_cli_cancel_unused
rread [Mon, 16 Sep 2002 17:59:49 +0000 (17:59 +0000)]
* add --obduuid <uuid> option so the OBD UUID can be specified
when create an ost. For example:
lmc --node localhost --lov lov1 --obduuid 'REAL_UUID' --ost bluearc
braam [Mon, 16 Sep 2002 17:29:01 +0000 (17:29 +0000)]
minor edits to the patch to work with the Red Hat 2.4.18-12.5 kernel.
adilger [Mon, 16 Sep 2002 06:18:21 +0000 (06:18 +0000)]
Take a LOV default stripe count of "0" to mean striping across all OSTs.
braam [Sat, 14 Sep 2002 07:02:20 +0000 (07:02 +0000)]
- don't forget to pass data and data_len on to the local enqueue
functions.
- do directory truncations when the clients don't modify directory
data and change the dcache.
radhikav [Sat, 14 Sep 2002 05:53:57 +0000 (05:53 +0000)]
Made correction to mdc_statfs prototype and a change to mdc_open API.
rread [Fri, 13 Sep 2002 00:20:13 +0000 (00:20 +0000)]
* ensure mds name used for mtpt is really an mds
* standardize lookup error messages
adilger [Thu, 12 Sep 2002 22:33:05 +0000 (22:33 +0000)]
Minor fix for getting the EA size.
pschwan [Thu, 12 Sep 2002 17:47:29 +0000 (17:47 +0000)]
b=597333
Remove unused ELDLM_RESOURCE_FREED
adilger [Thu, 12 Sep 2002 15:39:20 +0000 (15:39 +0000)]
Allos statfs to return values larger than 16TB.
adilger [Thu, 12 Sep 2002 13:33:57 +0000 (13:33 +0000)]
This should fix the LOV problem. All is well.
adilger [Thu, 12 Sep 2002 12:13:59 +0000 (12:13 +0000)]
Fix lov stripe offset calculations.
adilger [Thu, 12 Sep 2002 09:48:07 +0000 (09:48 +0000)]
Set lsm_magic to LOV_MAGIC.
Fix stripe OST allocation policy slightly.
adilger [Thu, 12 Sep 2002 09:24:32 +0000 (09:24 +0000)]
Allow the default stripe_count to be specified in the XML config, and also
read this value instead of the "stripe_offset" for which we have no use.
Set an objid for lctl create, so we get one step closer to support of
test_brw for LOV devices.
adilger [Thu, 12 Sep 2002 09:02:10 +0000 (09:02 +0000)]
Make sure we remove extN_fs.h when building new.
adilger [Thu, 12 Sep 2002 09:01:27 +0000 (09:01 +0000)]
Fix minor printf breakage introduced today.
adilger [Thu, 12 Sep 2002 07:32:54 +0000 (07:32 +0000)]
Update the README to describe the usage of the XML configuration files.
Also update runtests to do setup and cleanup with an XML file as a parameter.
Allow the llecho.sh script to build a config that will run on two separate
hosts if desired.
adilger [Thu, 12 Sep 2002 06:54:45 +0000 (06:54 +0000)]
Make runtests take an XML config file as a parameter.
braam [Thu, 12 Sep 2002 06:03:46 +0000 (06:03 +0000)]
Put the HOWTO back in the book.
adilger [Thu, 12 Sep 2002 05:23:02 +0000 (05:23 +0000)]
Add a bit of debugging for LOV sub-object allocation, so we can track it
more easily.
adilger [Thu, 12 Sep 2002 05:08:40 +0000 (05:08 +0000)]
Fix minor braino.
adilger [Thu, 12 Sep 2002 04:24:17 +0000 (04:24 +0000)]
Add definition for O_DIRECT since it is not in the headers yet.
adilger [Thu, 12 Sep 2002 04:05:52 +0000 (04:05 +0000)]
Add O_LARGEFILE and O_DIRECT to openclose test so we can eventually stress
test the kernel for O_DIRECT again.
adilger [Thu, 12 Sep 2002 00:47:45 +0000 (00:47 +0000)]
Large commit which implements more "intelligent" offsets for stripe
creation in the LOV. Most of it is just variable renaming.
Needs some testing.
rread [Thu, 12 Sep 2002 00:42:13 +0000 (00:42 +0000)]
print message if a server is not found, instead of failing with stack trace.
adilger [Wed, 11 Sep 2002 22:47:20 +0000 (22:47 +0000)]
TODOs for proper LOV test_brw operations (not implemented yet).
adilger [Wed, 11 Sep 2002 22:21:14 +0000 (22:21 +0000)]
Allow the start offset for test_brw be based on the thread number. This
allows the threads to be writing to different parts of the file (instead
of all overwriting the same part of the file), to better simulate a real
client load.
Make thread "offsets" start at zero for objids and file offsets (we need
to keep the actual thread numbers > 0, because thread 0 is the master).
Make be_verbose 64-bit clean, and also handle very large counts.
Validate LOV default stripe count <= OST count.
adilger [Wed, 11 Sep 2002 22:08:40 +0000 (22:08 +0000)]
Use OBD_PUNCH_EOF constant instead of lots of f's.
adilger [Wed, 11 Sep 2002 22:06:51 +0000 (22:06 +0000)]
Move definition of LP[DUX]64 to lustre_lib.h so we can use it in obd.c
adilger [Wed, 11 Sep 2002 22:01:27 +0000 (22:01 +0000)]
Rework the code that Phil spent so much time fixing to be much simpler,
much easier to see correctness, and nano-faster also.
pschwan [Wed, 11 Sep 2002 18:22:57 +0000 (18:22 +0000)]
b=600245
Fixes this bug by remember that sizeof() returns the size in bytes, not bits.
adilger [Wed, 11 Sep 2002 17:41:01 +0000 (17:41 +0000)]
Use sizeof(var) instead of sizeof(type).
adilger [Wed, 11 Sep 2002 17:07:58 +0000 (17:07 +0000)]
Add a define for the "punch to EOF = truncate" constant.
pschwan [Wed, 11 Sep 2002 16:11:09 +0000 (16:11 +0000)]
b=600245
This probably won't fix 600245, but it is related, and would have bitten us
once we grew to more than 32 clients with an ia32 MDS.
adilger [Wed, 11 Sep 2002 05:17:04 +0000 (05:17 +0000)]
Add 0x prefix to hex numbers.
rread [Wed, 11 Sep 2002 00:50:40 +0000 (00:50 +0000)]
* support 'ethX:*' style wildcard for tcp clients
* grab output of acceptor so errors can be displayed
rread [Tue, 10 Sep 2002 17:36:36 +0000 (17:36 +0000)]
disconnect first, then del_uuid
rread [Tue, 10 Sep 2002 07:31:41 +0000 (07:31 +0000)]
* also del_uuid when deleting routes
rread [Tue, 10 Sep 2002 06:34:55 +0000 (06:34 +0000)]
new --batch <file> option
File contains a series of lmc command lines. All commands are run and
then xml file is saved.
adilger [Mon, 9 Sep 2002 19:09:07 +0000 (19:09 +0000)]
Update changelog for release 0_5_7
pschwan [Mon, 9 Sep 2002 18:00:50 +0000 (18:00 +0000)]
- osc_create was setting lmd_stripe_count to 1 instead of 0
- ll_read_inode2 was requiring lmd_stripe_count to be non-zero; instead, we now
check the magic number in ll_lookup2 before calling iget
braam [Mon, 9 Sep 2002 16:01:17 +0000 (16:01 +0000)]
- change field names for clarity.
adilger [Mon, 9 Sep 2002 15:19:11 +0000 (15:19 +0000)]
Clarify that ll_brw() gets lustre flags and not the kernel rw flags.
pschwan [Mon, 9 Sep 2002 01:47:05 +0000 (01:47 +0000)]
If the 'nolocks' mount option is set, avoid taking or releasing OST locks.
This is hopefully a suitable workaround for 606116
braam [Mon, 9 Sep 2002 01:29:24 +0000 (01:29 +0000)]
- defiition of the flag and value
braam [Mon, 9 Sep 2002 01:28:40 +0000 (01:28 +0000)]
- add nolocks mount option osc=foo,mdc=bar,nolocks
pschwan [Mon, 9 Sep 2002 00:52:32 +0000 (00:52 +0000)]
b=606116
Add a check for req->rq_export in ldlm_callback_handler, to immediately return
an error instead of later triggering the 'conn' assertion in ptl_send_buf
adilger [Sun, 8 Sep 2002 22:28:43 +0000 (22:28 +0000)]
Drop the spinlock before returning if the export already has a connection.
I'm still not sure I agree that this is an error, but be that as it may,
I leave it for Mike & Peter to work out.
pschwan [Sun, 8 Sep 2002 20:05:16 +0000 (20:05 +0000)]
- We don't use pga[i].pg->index, so don't overwrite it, since that seems pretty
dangerous
- obd_dev's definition is already in a common header file, it doesn't need to be
in each module
- add an assertion that will perhaps catch the callback bug
- fix PTL_RPC_MSG_ERR reporting, which I don't think has ever worked properly
braam [Sun, 8 Sep 2002 20:01:41 +0000 (20:01 +0000)]
don't crash an MDS if a client send a reconnect.
adilger [Sun, 8 Sep 2002 19:39:30 +0000 (19:39 +0000)]
Don't oops if there is no recovd for a connection. Mike needs to confirm if
that is wrong (to not have a recovd), or if this fix is correct.
adilger [Sun, 8 Sep 2002 19:36:50 +0000 (19:36 +0000)]
Allow an objid to be specified on the CLI (via OID=<foo>) to allow multiple
hosts to run on the same object and/or only do rewrite or read tests on an
existing object.
adilger [Sun, 8 Sep 2002 19:32:24 +0000 (19:32 +0000)]
Add O_DIRECT by default again.
shaver [Sun, 8 Sep 2002 17:21:08 +0000 (17:21 +0000)]
Avoid OOPS on timeout from unconnected client. (Temporary.)
shaver [Sun, 8 Sep 2002 17:19:00 +0000 (17:19 +0000)]
Sample recovery reconnection script for MDS failover.
adilger [Sun, 8 Sep 2002 16:48:22 +0000 (16:48 +0000)]
Show page vector size (useful for very long running tests).
adilger [Sun, 8 Sep 2002 09:56:57 +0000 (09:56 +0000)]
Initialize spinlock for obdecho creates.
adilger [Sun, 8 Sep 2002 09:10:03 +0000 (09:10 +0000)]
Disable the lprocfs deregistration, now that I've disabled the registration.
adilger [Sun, 8 Sep 2002 09:00:53 +0000 (09:00 +0000)]
Disable the procfs setup stuff, because it is having grief with obdecho.
adilger [Sun, 8 Sep 2002 08:55:20 +0000 (08:55 +0000)]
Add obdecho so I don't get a complaint from the proc stuff.
adilger [Sun, 8 Sep 2002 08:51:34 +0000 (08:51 +0000)]
Update scripts to not do vector reads on BA OSTs.
Do some quickie tests in -brw so we know things are at least close to good.
adilger [Sun, 8 Sep 2002 08:24:57 +0000 (08:24 +0000)]
Like runregression-net.sh, but only bulk I/O, and lots of it. It does
reads immediately following the write tests, to catch errors early.
Uses 64GB of space on a real OST.
adilger [Sun, 8 Sep 2002 08:07:16 +0000 (08:07 +0000)]
Fix case where only the OSTNODE is set (test_brw).
adilger [Sun, 8 Sep 2002 07:59:39 +0000 (07:59 +0000)]
Add echo_open() and echo_close() to keep test_brw happy.
adilger [Sun, 8 Sep 2002 01:28:52 +0000 (01:28 +0000)]
Fix 32-bit truncation of test_brw offset calculation.
adilger [Sun, 8 Sep 2002 01:18:08 +0000 (01:18 +0000)]
Updated test_brw user program. Checks args better, uses a long long for
the iteration count, handles default args properly.
adilger [Sun, 8 Sep 2002 00:00:33 +0000 (00:00 +0000)]
Allow test_brw to run in either read-only, write-only, or read-write mode,
with or without O_DIRECT. The read-only mode allows us to offline verify
a file that was created from lctl test_brw write independent of Lustre.
shaver [Sat, 7 Sep 2002 21:19:48 +0000 (21:19 +0000)]
- Add hooks for post-replay fixup on ptlrpc_request.
- Use those hooks to update fh data in client filehandles when replaying
open.
Recovery can now failover to a standby/rebooted MDS and complete a
file write, attr-update and close successfully. HAND.
shaver [Sat, 7 Sep 2002 21:02:39 +0000 (21:02 +0000)]
- Handle setting of existing EA by retrying without XATTR_CREATE.
- Check RESEND instead of RECOVERY in ptlrpc_check_reply, because that's
what ptlrpc_resend_req sets.
- Don't "commit" in-flight requests.
- Update imp_handle when sending delayed requests after recovery.
Recovery now successfully replays enough to resume the request stream
for the client process. (That stream may, in the close case, refer to
stale data, but that's the next bit.)
pschwan [Sat, 7 Sep 2002 20:47:17 +0000 (20:47 +0000)]
- lprocfs_dereg_dev is failing, which prevented cleanup. This is VERY BAD, so
I made it continue with cleanup instead of aborting
rread [Sat, 7 Sep 2002 20:44:25 +0000 (20:44 +0000)]
* always recreat loopback backing file when --reformat is set
adilger [Sat, 7 Sep 2002 08:58:54 +0000 (08:58 +0000)]
Clean up compiler warnings for 64-bit systems where __u64 is a long and
not a long long like on 32-bit systems. From Evan Felix, PNL.
adilger [Sat, 7 Sep 2002 07:58:52 +0000 (07:58 +0000)]
Add a new type, obd_uuid_t (char[37]) to replace uuid_t, which means
something different for the rest of the kernel and user-space too.
This should be used in all places where we specify char[37] directly.
I have the full change for this already implemented, but I'm going
slowly to avoid causing too many conflicts for Mike.
adilger [Sat, 7 Sep 2002 07:52:46 +0000 (07:52 +0000)]
Change internal statfs calls to pass struct obd_statfs as a parameter
instead of the Linux struct statfs. This allows us to have 64-bit
clean statfs data all the way to the VFS, and avoids problems like
block count overflows in intermediate math in LOV and such.
We still need to resolve the VFS interface to statfs, as it limits the
output to 16TB on 32-bit platforms (4G x 4kb blocks). The likely
answer is to shift the blocksize up, and the blocks counts down until
we fit in under 4G total blocks. For MCR this will be around a blocksize
of 64kB which is nice to specify as the block size anyways.
adilger [Sat, 7 Sep 2002 07:43:02 +0000 (07:43 +0000)]
Move knowledge of the LOV metadata/message structures out of mdc_getlovinfo()
and put it into lov_connect. This will be needed when we have different
kinds of LOVs that store their data differently. We still need to do
something about giving the LOV a "name" so that it can distinguish its data
on the MDS from other LOV's that there.
adilger [Sat, 7 Sep 2002 07:41:03 +0000 (07:41 +0000)]
Quiet a few printk's and turn them into CDEBUGs so they don't go to console.
Remove check for kernel code older than 2.3 - lustre will never go there.
* Add warning for lprocfs_bfs_srch(): having recursive code in the kernel is
* extremely frowned upon, because there is a very limited kernel stack size
* (only 8kB). That has to be enough, not only for your functions, but all
* other functions in that call trace, as well as interrupt handlers and such.
Remove extraneous line continuation characters.
Whitespace reformatting to match the rest of the lustre code.
adilger [Sat, 7 Sep 2002 06:47:07 +0000 (06:47 +0000)]
Invoke shell scripts via "sh" because CVS doesn't keep file perms well.
Add some extra basic checks after the cleanup.
adilger [Sat, 7 Sep 2002 06:32:45 +0000 (06:32 +0000)]
Add objid to test_brw output.
Allow objid for test_getattr to have a "t" (thread) prefix for symmetry
with test_brw, so that it will be possible to do a bunch of creates in
runregression net, and have the BRWs run I/O on different objects in
each thread.
adilger [Sat, 7 Sep 2002 06:01:49 +0000 (06:01 +0000)]
Cosmetic.
adilger [Sat, 7 Sep 2002 05:28:14 +0000 (05:28 +0000)]
Oops, missing part of change.
adilger [Sat, 7 Sep 2002 05:25:50 +0000 (05:25 +0000)]
Disable vector reads until the BA OSTs are updated (tomorrow, hopefully).
Don't do setups and cleanups for each loop because of problems therein...
rread [Sat, 7 Sep 2002 00:23:30 +0000 (00:23 +0000)]
* MDC element is no longer created in config. The MDC UUIDS are generated
by lconf when a client is started. This required some Big Ugly Hacks that
I hope to clean up in the near future.
* LOVs and OSC are treated uniformly. --osc lov1 will now work as expected.
These changes require config files to be recreated to work properly with lconf. No
changes were made to the lmc interface, so the config scripts are unchanged.
rread [Sat, 7 Sep 2002 00:20:25 +0000 (00:20 +0000)]
use a larger buffer for device_list
adilger [Fri, 6 Sep 2002 23:48:37 +0000 (23:48 +0000)]
Call open/close on an object before trying to read/write it.
adilger [Fri, 6 Sep 2002 23:47:26 +0000 (23:47 +0000)]
Add open/close ioctls to obdclass.
thantry [Fri, 6 Sep 2002 23:04:34 +0000 (23:04 +0000)]
Updated makefile to include first cut for LProcFS
thantry [Fri, 6 Sep 2002 23:03:05 +0000 (23:03 +0000)]
Commented proc code
thantry [Fri, 6 Sep 2002 23:00:45 +0000 (23:00 +0000)]
Updated obd to include profiling support
thantry [Fri, 6 Sep 2002 22:59:52 +0000 (22:59 +0000)]
Header file for LprocFS
thantry [Fri, 6 Sep 2002 22:58:30 +0000 (22:58 +0000)]
Updated lprocfs.c. First cut