RepoMirrors/ceph
1
0
Fork 0
mirror of https://github.com/ceph/ceph synced 2025-01-11 05:29:51 +00:00
Code Issues Packages Projects Releases Wiki Activity
2,615 Commits 214 Branches 394 Tags 1.3 GiB
a98d66aa63
Commit Graph

2615 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Sage Weil
a98d66aa63 mds: ignore atime updates until code can handle non-wrlockable updates 2008-03-05 15:00:21 -08:00
Sage Weil
d586b8a5fc client: error handling, open cleanups 2008-03-05 14:59:56 -08:00
Sage Weil
401ff1b482 Merge branch 'unstable' into always_insert_trace 2008-03-05 14:47:18 -08:00
Sage Weil
831c4af927 client: disable ceph_lookup_open intent stuff for now 2008-03-05 13:46:41 -08:00
Sage Weil
4c6431bb0c client: refactor mds requests to put most processing in reply handler 
The goal is to fix race issues between open reply processing (previously
in calling processes context) and subsequent file_cap messages.  This just
pushes generic mds reply handling into the reply handler, including
the ceph_insert_trace (somewhat heavyweight, unfortunately) and the cap
creation/update.

This effectively pushes a lot more code into the single thread thats
doing mds replies, but simplifies a lot of code.  Until scaling issues
arise, I think it's fine.
 2008-03-05 13:29:03 -08:00
Sage Weil
9acc42b9b9 mds: fix open flags check on directories 2008-03-05 13:06:02 -08:00
Sage Weil
371f4bc32b client: inode.i_size locking, cap TRUNC handler 2008-03-04 14:39:08 -08:00
Sage Weil
8bcbeb5f56 FileLayout rename 2008-03-04 14:36:11 -08:00
Sage Weil
8e064794c9 client: addr.c tabs 2008-03-04 12:08:36 -08:00
Sage Weil
da06f086fc client: fix dout/derr macros 2008-03-04 12:05:08 -08:00
Sage Weil
466fdb4c9e client: require rd cap before checking size in Client::_read() 2008-03-04 11:06:50 -08:00
Sage Weil
83e4a66f6a journal: some fixes, still some issues left 2008-03-03 21:46:23 -08:00
Sage Weil
9e29169a4b Merge branch 'mdscapthing' into unstable 2008-03-03 15:34:35 -08:00
Sage Weil
4c9987a0bb mds, client: fixed races with cap release, open, size/mtime update journaling, and file size 
- there's now an explicit TRUNC cap message when file size decreases
- mds holds wrlock during duration of size|mtime journaling
 2008-03-03 15:26:34 -08:00
Sage Weil
317bb0f5f6 todo, client verbosity 2008-03-02 22:41:06 -08:00
Sage Weil
0a135059de cap fix unfinished 2008-03-02 22:40:30 -08:00
Sage Weil
2db74860f0 mon: osd commands to mark osds in/out/down 2008-03-02 14:39:00 -08:00
Sage Weil
f3f3bf1630 manual merge of volker and dongsu's write stuff 
this is based on a diff of
    7469f26a33
and 933226e5cdea16d311f15d47d5db852cf2107e57

i left out the dcache locking bits in dir.c, for now.
 2008-02-29 22:13:51 -08:00
Sage Weil
ab5be39486 cas: pg_pool and file_layout bits to pass object pool down through stack 2008-02-29 22:12:36 -08:00
Sage Weil
ece3fef191 cleanups 2008-02-29 15:41:10 -08:00
Sage Weil
c8d2a8ff00 byteorder fixup 2008-02-29 15:35:34 -08:00
Sage Weil
91582d1fb5 todos 2008-02-29 15:35:34 -08:00
Sage Weil
d99d887ba8 mds: fix pre_dirty + projected_parent to bump pv properly; fix unlink vs cap release race 2008-02-29 10:57:02 -08:00
Sage Weil
1f2db92f78 fakestore: various fixes (esp on 32bit) 2008-02-28 14:45:58 -08:00
Sage Weil
fe91c06cd3 mon: oops 2008-02-28 11:32:32 -08:00
Sage Weil
22d102ad5e mon: get/set maxosd 2008-02-28 11:11:59 -08:00
Sage Weil
3cfd18b553 mon: reworked command interface to use preprocess_ and prepare_command 2008-02-28 10:21:48 -08:00
Sage Weil
5dd9f1a17c mds: projected_parent on inode to allow concurrent rename and inode updates (due to cap changes) 
The problem is that a cap update would give a new mtime mid-rename.  Locker would pre-dirty the inode and _old_ parent dentry, and when it went to apply the projected update, it would have a pv from the old dentry and use it to dirty the new parent dentry.  Now, rename sets projected_parent in _rename_prepare and clears it via CDir::link_inode, and pre_dirty follows that pointer when present.
 2008-02-28 09:50:13 -08:00
Sage Weil
63187b6192 mon: setcrushmap hack; basic fwd to leader strategy 2008-02-27 17:01:38 -08:00
Sage Weil
dd34c81691 cmonctl: send data payload 2008-02-27 16:52:13 -08:00
Sage Weil
42ea68fdb8 mon: added getmap and getcrushmap to extract osdmap or just crush map from monitor 2008-02-27 16:43:22 -08:00
Sage Weil
e52743fe25 client: ipquad printf cleanups 2008-02-27 16:12:55 -08:00
Sage Weil
8cf5aea9e4 mds: truncate fixes 2008-02-27 15:44:18 -08:00
Sage Weil
57e0de067a fakestore: fix collection_setattrs 2008-02-27 14:57:36 -08:00
Sage Weil
0aee328593 fakestore: cleanup 2008-02-27 14:45:50 -08:00
Sage Weil
be972e85e5 mds: untested fix for opent/openc inode versioning bug 2008-02-27 14:43:36 -08:00
Sage Weil
fc0ae5f35f fakestore: journal replay 2008-02-27 14:17:36 -08:00
Sage Weil
7d34187b26 fakestore: more journaling bits, fixes 2008-02-27 13:51:05 -08:00
Sage Weil
99f3e71be4 client: ignore long crush blob 2008-02-27 11:58:26 -08:00
Sage Weil
d42d98fa7f fakestore: use a journal, with periodic sync 2008-02-27 11:58:26 -08:00
Sage Weil
30adf9f693 osd: factor journaling out of ebofs 2008-02-27 11:58:26 -08:00
Sage Weil
c288b4f2e5 todos 2008-02-27 11:58:25 -08:00
Sage Weil
c7292dbca3 lower autoconf version req to 2.59 (works, with warnings) 2008-02-25 10:04:01 -08:00
Sage Weil
7afb7e2a66 crush: git bucket id calc for additional buckets 2008-02-24 20:53:52 -08:00
Sage Weil
138d8bda1f mds: add wrlock bits to FileLock 2008-02-13 03:14:53 -08:00
Sage Weil
05d573097d mds: temp workaround mtime update race 2008-02-13 03:14:37 -08:00
Yehuda Sadeh-Weinraub
8603b428d5 temporary fix of a crash due to missing decode 2008-02-10 17:01:36 +02:00
Yehuda Sadeh-Weinraub
224eca6190 start.sh can execute binaries from alternative locations 2008-02-07 15:18:27 +02:00
Yehuda Sadeh-Weinraub
0e2de00545 added missing file 2008-02-06 15:17:01 +02:00
Yehuda Sadeh-Weinraub
4be7eca084 added proc entry for setting debug level 2008-02-06 15:13:15 +02:00