b=13930
i=johann,adilger

ll_inode_fill_obdo() expects OBD_BRW_WRITE or OBD_BRW_READ as rw argument
rather than READ/WRITE.

i=adilger, johann

cleanup in *_get_info methods to use the KEY_IS() macro and to abstract the datatypes for key and reply on the OST.

i=adilger
i=eeb
Track actual lock callback times, instead of the rpc times, for setting ldlm
timeouts.  Add a module parameter for the minimum lock wait time.
Also add a module parameter for a minimum AT time in general.

i=Adilger
i=Walter
test_54e fix

b=13769
i=Adilger
i=Nathan

t-f helper check_runas_id()

i=grev
Remove code accidentally landed with 3055 while I was doing some investigation
into the occasional test 50/51/52 failures.

i=yangsheng,mjmac

More correctly set the EXTRAVERSION in the sles10 kernel-lustre-source.

r=Andreas, Nathan

Patch to organize the stats on per-nid basis instead of the existing per-export basis.

Moving bug 12652 EXCEPT to the correct test.

b=13888
i=nikita
i=alex

b=13622
i=tappro
i=green

1) do not zero cookie in class_handle_unhash_nolock() to let it be rehashed back later;
2) dlm_lock_match() returns the matched mode instead of 1; pass bitwise sets of modes into;

b=12858
i=green
i=adilger

r=Nathan, Adilger

Patch to handle symlinks while setting tunables.

b=13564
i=johann
i=nathan

error handling cleanup.

b=13828
i=adilger
i=shadow

Land 2nd attachment (#13342):
- use a higher value for ATTR_CTIME_SET
- define MDS_ATTR_* in hex
- improve attr_unpack/attr_pack/mds_pack_open_flags

b=12584
i=johann
i=nathan

add sub-routine declaration before use it.

b=13497
i=deen
i=johann

Description: LASSERT_{REQ,REP}SWAB macros are buggy
Details    : If SWAB_PARANOIA is disabled, the LASSERT_REQSWAB and
             LASSERT_REPSWAB macros become no-ops, which is incorrect. Drop
             these macros and replace them with their difinitions instead.

i=Adilger

force runtests to use t-f log()

i=Nathan
i=Tianzy

remove run test_0 from test_14 to pre_test_14

add test_68(using lustre file as swap space) into slow test set because it always takes a long time to finish.

i=adilger,shadow
fix the direct I/O path in the case of read hitting hole.
improve the write direct I/O path: submit the entire write in one RPC set.
add generic completion callback set support in ptlrpc_request_set.

i=Adilger

t-f wait_delete_complete() function; force test_79 to wait;
quiet test_65l output.

b=13556
i=scjody

Update version

Update release dates

Update version

b=13730
i=johann
i=green

b=13521
i=scjody
i=nathan

Update which_patch & target file for SLES10 2.6.16.53-0.8 kernel.

r=adilger,nathan
- fixes in LVF calculation in sanity test 124a.

i=Adilger

skip test_125 if mds acl disabled

i=adilger
add error_ignore method to run a test but ignore failures

Revert Qian's recent changes related to bug 13099 since they caused build
issues and differ significantly from the patch that was inspected.

b=13828
i=adilger,shadow

Don't use kernel's ATTR_* attributes over the network since the client and MDS
may run different kernels. Instead, use MDS_ATTR_* attributes (defined in
lustre_idl.h) for sa_valid.

i=adilger,shadow
remove the useless 'nob'.

i=adilger,shadow
fix the directio path: submit entire write in one RPC set.
change the test script name.

b=13626
i=johann, adilger

use k(un)map_atomic instead of k(un)map for short/tight code paths.

lost test_7 fix

b=13626

reverse the patch since been reported broken test on box with 2G ram.

 - revert patch landed by mistake

i=adilger
i=eeb

 - improved sd iostats:
  * time requests spend in elevator's queue
  * time requests are served by disk
  * seek stats
  * queue depth stats