Whamcloud - gitweb
adilger [Fri, 22 Mar 2002 18:55:44 +0000 (18:55 +0000)]
The loopback device helper will use the next unused loopback device instead
of just starting at minor 0.
pschwan [Fri, 22 Mar 2002 18:54:22 +0000 (18:54 +0000)]
- Take advantage of new disconnect semantics to make llmountcleanup.sh a more
general script
adilger [Fri, 22 Mar 2002 18:52:06 +0000 (18:52 +0000)]
Add ext3_bread to exported symbols. Needed because ext3 does not have
directories in the page cache yet.
pschwan [Fri, 22 Mar 2002 18:31:34 +0000 (18:31 +0000)]
- Removed an overzealous BUG() that keeps us from cleaning up after C-c
adilger [Fri, 22 Mar 2002 05:45:52 +0000 (05:45 +0000)]
Indenting only.
braam [Thu, 21 Mar 2002 20:06:01 +0000 (20:06 +0000)]
- Added UUID description to network document
- fixed leaked refcount in failed osc_request startup
- small debugging message fix
- removed extraneous PtlMDUnlink
- try to remove kqswnal in llmountcleanup
braam [Thu, 21 Mar 2002 05:39:14 +0000 (05:39 +0000)]
add path
braam [Thu, 21 Mar 2002 05:37:09 +0000 (05:37 +0000)]
dd kernel version deps
Cdd \7f\7f\7f\7fVS: ----------------------------------------------------------------------
braam [Thu, 21 Mar 2002 05:31:03 +0000 (05:31 +0000)]
- remove unhelpful CERRORs
braam [Thu, 21 Mar 2002 01:50:34 +0000 (01:50 +0000)]
- do the right thing when we catch SIGTERM
- removed extraneous MDUnlink
- added CERRORs to catch MD leak
braam [Wed, 20 Mar 2002 23:47:14 +0000 (23:47 +0000)]
- updated documentation (introduction)
- race fix in md_active
- new lock stuff
pschwan [Wed, 20 Mar 2002 22:27:53 +0000 (22:27 +0000)]
Yet another test script, this one for the DLM.
pschwan [Wed, 20 Mar 2002 22:27:04 +0000 (22:27 +0000)]
- mds_open/mds_close had an unbalanced mnt put/get. I'm still not clear on why
_every_ caller to mds_fid2dentry doesn't have this problem, but an extra
mntput fixed this one.
- oops. I committed my really large llmount.sh. Fixed.
adilger [Wed, 20 Mar 2002 20:52:11 +0000 (20:52 +0000)]
Update older ll*.sh scripts to use helper functions.
adilger [Wed, 20 Mar 2002 20:47:57 +0000 (20:47 +0000)]
Add "llrmount.sh" script, which does the same setup as llmount.sh, but does
not create new filesystems. It uses the existing filesystems from a previous
invocation of llmount.sh. This is useful for persistency testing or testing
recoverability.
adilger [Wed, 20 Mar 2002 20:46:25 +0000 (20:46 +0000)]
Add "old_fs" helper function (takes same args as new_fs helper) to just
set up an existing filesystem instead of creating a new one.
pschwan [Wed, 20 Mar 2002 19:58:43 +0000 (19:58 +0000)]
- don't free the request when you drop the packet--it's on the stack
- brought mdcreq.sh into the year 2002 (even though the mdc ioctl interface is
still broken)
pschwan [Wed, 20 Mar 2002 19:07:46 +0000 (19:07 +0000)]
- Removed the (nearly) unused obd_print_entry and fixed the code that
depended on it.
- Added code to drop packets based on /proc/sys/obd/drop_packet
- De-tabbified and switched a bunch of code to use RETURN and GOTO
braam [Mon, 18 Mar 2002 15:15:06 +0000 (15:15 +0000)]
- may not need this client lock
braam [Mon, 18 Mar 2002 15:00:07 +0000 (15:00 +0000)]
- add a figure
- fix the problem that the refcount is reduced on the wrong md in the
ring buffer
braam [Sun, 17 Mar 2002 05:39:22 +0000 (05:39 +0000)]
- temporarily change client.c which bums out on killing a client (PHIL?)
- lock the service structure whenever it mucks with state
- add full debugging to llecho.sh
adilger [Thu, 14 Mar 2002 21:54:43 +0000 (21:54 +0000)]
Check for 1 or 2 obdos being passed from test_brw ioctl.
adilger [Thu, 14 Mar 2002 21:49:44 +0000 (21:49 +0000)]
Change default obdctl test_brw behaviour to have only 1 obdo and 1 page.
Make BRW class ioctl check for one or two obdos.
braam [Thu, 14 Mar 2002 18:37:39 +0000 (18:37 +0000)]
- Small changes for packaging
pschwan [Thu, 14 Mar 2002 06:43:31 +0000 (06:43 +0000)]
- be more strict about what we accept from PtlEQGet in ptlrpc_check_event()
pschwan [Thu, 14 Mar 2002 06:33:35 +0000 (06:33 +0000)]
- fixed some NTOH/HTON mixups
- added some BUG()s in niobuf pack/unpack to catch bogus offsets
- added a couple of BUGs() in dropped-packet cases
braam [Thu, 14 Mar 2002 06:29:41 +0000 (06:29 +0000)]
- add correct file
braam [Thu, 14 Mar 2002 05:26:35 +0000 (05:26 +0000)]
- many small makefile changes to build package conveniently.
- to be done: retain CVS with anonymous root
- fix up the PORTALS path
- build a portals RPM
- include a better tags target that includes vim tags and portals.
adilger [Thu, 14 Mar 2002 00:16:31 +0000 (00:16 +0000)]
Use req->page instead of req->addr inside echo_commitrw, as the latter is
unset by the callback function on a write (it worked fine for reads ;-).
adilger [Wed, 13 Mar 2002 19:07:20 +0000 (19:07 +0000)]
Script to set up MDS using ext3 (basically llmount.sh with ext3 instead of
ext2 for MDS). Will not work unless ext3 has been patched with a_ops=&ext3_ops
in fs/ext3/inode.c and fs/ext3/namei.c. Not sure if that is the correct
solution yet or not.
pschwan [Wed, 13 Mar 2002 17:59:13 +0000 (17:59 +0000)]
- Makefile.am fixes to allow building in objdirs
adilger [Tue, 12 Mar 2002 23:08:32 +0000 (23:08 +0000)]
Add comment about direct access to fs-internal data struct - it works for
ext2 and ext3 (they happen to have the same layout) but it is dangerous.
Since it falls under the "hack - to be fixed with EA support" I didn't fix.
adilger [Tue, 12 Mar 2002 23:06:35 +0000 (23:06 +0000)]
A couple of cleanups while looking for ext3 bug - don't leak buf or bulk
allocations on error.
pschwan [Tue, 12 Mar 2002 20:34:24 +0000 (20:34 +0000)]
- brought lustre-debugging.lyx up to date
- removed unused obd_print_entry code
- fixed the llmount.sh debug level default
- minor debugging cleanups
- moved mdc_reint to where it's actually used
adilger [Tue, 12 Mar 2002 00:19:51 +0000 (00:19 +0000)]
Make llmountcleanup.sh use new helper methods.
adilger [Tue, 12 Mar 2002 00:19:15 +0000 (00:19 +0000)]
Clean up llmount.sh to use new helper methods
adilger [Tue, 12 Mar 2002 00:18:21 +0000 (00:18 +0000)]
Clean up llecho.sh script to use new helper methods
adilger [Tue, 12 Mar 2002 00:17:41 +0000 (00:17 +0000)]
Add helper routines to tests/common.sh:
- centralize location for debugctl, obdctl, ptlctl
- automate test filesystem creation
- check if devices exist before trying to create them (avoid error msg)
adilger [Tue, 12 Mar 2002 00:02:48 +0000 (00:02 +0000)]
Add openunlink binary to .cvsignore.
adilger [Tue, 12 Mar 2002 00:01:26 +0000 (00:01 +0000)]
Add echo_connect/echo_disconnect to do proper module refcounting (avoid
oops if we unload module while we have an open connection to it).
The echo_preprw and echo_commitrw methods now clean up properly on error.
adilger [Mon, 11 Mar 2002 23:56:39 +0000 (23:56 +0000)]
Fix BRW ioctl test interface. I had set up offset/count/flag per obdo
when it is really per-page.
adilger [Mon, 11 Mar 2002 19:45:31 +0000 (19:45 +0000)]
Unlock page in lustre_prepare_page() if there is an error.
pschwan [Mon, 11 Mar 2002 14:58:49 +0000 (14:58 +0000)]
- a few more header cleanups
- removed some unused code
- inserted many extra BUG()s to catch problems early
- fixed a use of the incorrect inode, and fixed the directory bug
- enabled mds_open/mds_close
- fixed a small leak in an error case
- removed an extra set_page_dirty() in the write path
- added an open/write/close/open/unlink/read/truncate/write/read/close test
pschwan [Sun, 10 Mar 2002 19:08:09 +0000 (19:08 +0000)]
We touched everything, but it's not as scary as it looks.
- dramatically reorganized and simplified the headers. Most of the nasty
recursive definitions are gone, and obd_class.h no longer drags in the
entirety of Lustre.
- .cvsignores updated and cleaned
- added another chunk of DLM code and a test, which passes (note that it doesn't
free the locks yet, so don't expect to cleanup afterwards)
- fixed OBD_ALLOC/OBD_FREE in the same ways as the Portals macros
- removed a lot of dead obd_class code
braam [Sat, 9 Mar 2002 18:31:38 +0000 (18:31 +0000)]
- add open and close calls: initial purpose is to support I/O to open,
unlinked files
- fix faulty connection handling in osc/ost methods: unlink, punch (this
should explain and fix the iozone behavior).
adilger [Fri, 8 Mar 2002 21:06:24 +0000 (21:06 +0000)]
Update preprw and commitrw methods on echo OST.
adilger [Fri, 8 Mar 2002 20:57:52 +0000 (20:57 +0000)]
Add BRW_READ and BRW_WRITE ioctls.
They are for BRW bulk transport testing, using a vector of 2 obdos with a
variable number of pages each. No data is currently passed to/from userspace.
In fact, I _think_ that no data is going over the transport either for some
reason, because the operation is too fast, even though I'm not getting any
errors.
adilger [Fri, 8 Mar 2002 20:47:51 +0000 (20:47 +0000)]
Add test_brw command. While it doesn't cause any errors, I don't think it
actually does the right thing yet - it runs much too quickly (faster than
getattr, in fact). Committing it to reduce the size of outstanding changes.
Add some performance stats to test_getattr.
adilger [Fri, 8 Mar 2002 20:41:33 +0000 (20:41 +0000)]
Adding ctags and other built files to .cvsignore to reduce clutter and make
changes more visible.
adilger [Fri, 8 Mar 2002 20:38:37 +0000 (20:38 +0000)]
Tiny cleanup in preprw to match commitrw.
adilger [Fri, 8 Mar 2002 20:34:31 +0000 (20:34 +0000)]
Add BRW_READ and BRW_WRITE ioctls.
adilger [Fri, 8 Mar 2002 20:30:34 +0000 (20:30 +0000)]
Don't try to close file pointer if there was an error opening it.
adilger [Fri, 8 Mar 2002 16:59:29 +0000 (16:59 +0000)]
Added cvsignore files to directories which didn't have them.
braam [Thu, 7 Mar 2002 23:41:09 +0000 (23:41 +0000)]
- first parts of lock management: module & resource hash +
header file with data structures.
pschwan [Thu, 7 Mar 2002 22:36:24 +0000 (22:36 +0000)]
- typing 'quit' was not cleaning up obdctl connections. fixed.
- switching devices also closes connections now.
pschwan [Thu, 7 Mar 2002 20:16:05 +0000 (20:16 +0000)]
- module insert code moved to tests/common.sh
- refactored request freeing code to handle the use of the reply buffer after
the mdc_getattr, etc. returns.
pschwan [Thu, 7 Mar 2002 17:23:02 +0000 (17:23 +0000)]
- made many CERRORs (the informative, rather than bad ones) into CDEBUGs. It
should run silently now.
pschwan [Wed, 6 Mar 2002 20:53:13 +0000 (20:53 +0000)]
- Report any leaked memory at module remove time
- Free the reply buffer when we destroy the request, thus avoiding the
double-free. If we ever move to an event model where we don't put the
ptlrpc_main() request on the stack, be careful that we don't free an outgoing
reply buffer before it's sent!
adilger [Tue, 5 Mar 2002 23:55:38 +0000 (23:55 +0000)]
Add echo_preprw() and echo_commitrw() methods to the echo OST. They have
not been tested yet, need to add test_brw to obdctl first.
adilger [Tue, 5 Mar 2002 23:51:16 +0000 (23:51 +0000)]
Don't dereference conn if we don't have a connection.
pschwan [Tue, 5 Mar 2002 22:02:59 +0000 (22:02 +0000)]
rpc/service.c: set max_offset when creating service MDs
added lustre debugging draft
braam [Tue, 5 Mar 2002 18:44:34 +0000 (18:44 +0000)]
- track offset for debugging
pschwan [Tue, 5 Mar 2002 18:29:37 +0000 (18:29 +0000)]
use debugctl for 'modules'
unload 'obdecho'
braam [Tue, 5 Mar 2002 18:13:18 +0000 (18:13 +0000)]
max offset patch for request headers,
missing obd_echo
pschwan [Tue, 5 Mar 2002 17:15:22 +0000 (17:15 +0000)]
Removed 'modules' functionality, which is now in debugctl
braam [Tue, 5 Mar 2002 01:00:36 +0000 (01:00 +0000)]
- add obdecho directory
- numerous small fixes to obdclass -- yippeah Andreas is back
- test_getattr (jt_multi_getattr) tells obdctl to fire many getattr requests.
- build this code, run obd/tests/llecho.sh and play with it
pschwan [Sun, 3 Mar 2002 08:03:03 +0000 (08:03 +0000)]
Fixed stupid bug, calling kunmap() on an address
pschwan [Sun, 3 Mar 2002 07:56:15 +0000 (07:56 +0000)]
build fix, remove 3rd buffer
braam [Sun, 3 Mar 2002 07:46:16 +0000 (07:46 +0000)]
- small fixes; things work.
braam [Sun, 3 Mar 2002 07:13:51 +0000 (07:13 +0000)]
- fix for reads
pschwan [Sun, 3 Mar 2002 06:50:17 +0000 (06:50 +0000)]
- fixed obd_brw_read bug
pschwan [Sun, 3 Mar 2002 06:19:25 +0000 (06:19 +0000)]
*** empty log message ***
braam [Sun, 3 Mar 2002 06:11:18 +0000 (06:11 +0000)]
- bulk handling from callbacks
pschwan [Sun, 3 Mar 2002 05:46:50 +0000 (05:46 +0000)]
*** empty log message ***
pschwan [Sun, 3 Mar 2002 04:34:22 +0000 (04:34 +0000)]
- Added an 'xid' field to struct niobuf
- Fixed bug in OST/OSC bulk handling by using this new field
- fixed double free in OST
braam [Sun, 3 Mar 2002 04:25:40 +0000 (04:25 +0000)]
- tiny fix to ost_request.c
braam [Sun, 3 Mar 2002 03:17:46 +0000 (03:17 +0000)]
- minor fixes to get mdc_readpage working over again.
pschwan [Sun, 3 Mar 2002 02:29:47 +0000 (02:29 +0000)]
Merged branch 'peter' with the tip. Pre-merge tag is 't_20020302_networking'.
Avert your eyes.
braam [Thu, 28 Feb 2002 08:49:03 +0000 (08:49 +0000)]
new files - restructure RPC further
pschwan [Thu, 28 Feb 2002 03:05:25 +0000 (03:05 +0000)]
- added rq_type field to ptlrpc_request
- this field must be one of PTLRPC_REQUEST, PTLRPC_REPLY, PTLRPC_BULK
- mdc_request now waits for directory pages to be sent via the bulk portal
- change MDS bulk waitq to the new style
- ptl_send_buf no longer takes the 'is_request' argument; this is encoded in
the rq_type mentioned above
- added ptl_abort_rpc() to tear down reply structures for aborted requests.
- refactored ptl_send_buf code to rid ourselves of some of the more gross
bulklen hacks
braam [Wed, 27 Feb 2002 22:02:48 +0000 (22:02 +0000)]
-small is better for now
braam [Wed, 27 Feb 2002 20:57:00 +0000 (20:57 +0000)]
- barrier for potential cache problems
pschwan [Wed, 27 Feb 2002 20:06:14 +0000 (20:06 +0000)]
mds/handler.c: init the waitqueue before we call ptl_send_buf()
braam [Wed, 27 Feb 2002 18:23:31 +0000 (18:23 +0000)]
yet more tracking of lost packets
braam [Wed, 27 Feb 2002 17:49:34 +0000 (17:49 +0000)]
- add open and close to MDC interface (not used yet)
- very defensive logging to track our lost packet
braam [Wed, 27 Feb 2002 10:16:11 +0000 (10:16 +0000)]
- fix connection find on OST
braam [Wed, 27 Feb 2002 09:56:13 +0000 (09:56 +0000)]
- bug fix to unload modules with safe wait functions
- rewrite of ptlrpc client system
- pivot around ptlrpc_client
- mdc client & lustre_light uses generic routines
- osc can do the same (tomorrow)
- different event queues in rpc.c freeing packets & getting replies
- every unexpected event or PtlError is now a BUG in rpc.c
- change gdb script output to allow reloading everything
pschwan [Wed, 27 Feb 2002 05:41:04 +0000 (05:41 +0000)]
- New style waitqueues in ost and mds main
pschwan [Wed, 27 Feb 2002 05:07:30 +0000 (05:07 +0000)]
New style waitqueues for OSC
pschwan [Wed, 27 Feb 2002 05:02:20 +0000 (05:02 +0000)]
- Don't set request->rq_repbuf until the reply has actually come in
pschwan [Wed, 27 Feb 2002 04:58:37 +0000 (04:58 +0000)]
Fix the waitqueue race in the mdc, I hope.
pschwan [Wed, 27 Feb 2002 04:26:22 +0000 (04:26 +0000)]
- llmount.sh creates for you a /tmp/ogdb in the hostfs, and waits for you to
press enter before it continues. This lets you set breakpoints and debug
the mount/startup process.
- Run 'mynid localhost' instead of 'mynid'
pschwan [Wed, 27 Feb 2002 04:16:27 +0000 (04:16 +0000)]
Add many debug messages to the RPC path
pschwan [Tue, 26 Feb 2002 22:32:29 +0000 (22:32 +0000)]
Added paths to obdctl's 'modules' command, making it immediately useful:
/r/usr/src/obd/utils/obdctl modules > /r/root/ogdb
braam [Mon, 25 Feb 2002 23:34:54 +0000 (23:34 +0000)]
- wrong offset in commit page
braam [Mon, 25 Feb 2002 20:45:32 +0000 (20:45 +0000)]
Wrong object id sent to obd - fixed
braam [Mon, 25 Feb 2002 18:57:03 +0000 (18:57 +0000)]
- uninitialized error code
- uninitialized mode of data objects
both fixed
braam [Mon, 25 Feb 2002 18:13:41 +0000 (18:13 +0000)]
minor changes to debug prepw and prepare page
remove some debug crud from the OST handler
braam [Mon, 25 Feb 2002 15:49:21 +0000 (15:49 +0000)]
- add the truncate obd call
pschwan [Mon, 25 Feb 2002 08:00:48 +0000 (08:00 +0000)]
Getting rid of tabs makes me happy, just humour me.