Whamcloud - gitweb
yury [Fri, 14 Apr 2006 15:13:58 +0000 (15:13 +0000)]
- fix in makefile
yury [Fri, 14 Apr 2006 15:05:26 +0000 (15:05 +0000)]
- basic lmv to new fids adaptation.
yury [Fri, 14 Apr 2006 10:51:49 +0000 (10:51 +0000)]
- get rid of ll_fid in mgs
tappro [Fri, 14 Apr 2006 10:33:00 +0000 (10:33 +0000)]
add lmv code from head. Need to be adapted.
yury [Fri, 14 Apr 2006 07:37:24 +0000 (07:37 +0000)]
- update from parent
nikita [Thu, 13 Apr 2006 22:58:28 +0000 (22:58 +0000)]
add lu_context everywhere. Mountability preserved.
yury [Thu, 13 Apr 2006 21:51:21 +0000 (21:51 +0000)]
file vfs_intent-2.6-fc3.patch was added on branch b_new_cmd on 2006-05-19 11:29:25 +0000
wangdi [Thu, 13 Apr 2006 14:09:47 +0000 (14:09 +0000)]
Branch: b_new_cmd
prototype of fld
wangdi [Thu, 13 Apr 2006 13:37:58 +0000 (13:37 +0000)]
Branch: b_new_cmd
add fld prototype
yury [Thu, 13 Apr 2006 11:43:33 +0000 (11:43 +0000)]
- bringing liblustre to use fids like llite does - just to not lost changes and do not do them later;
- fix all legacy res_id.name generation in both llite and liblustre;
- get rid of ino and generation in mdt_body;
- cleanups about using fids in llite and liblustre;
- remove ll_test_inode() - we can live without it;
nikita [Thu, 13 Apr 2006 09:33:44 +0000 (09:33 +0000)]
add prefix to lu_attr fields
ericm [Thu, 13 Apr 2006 03:46:13 +0000 (03:46 +0000)]
file gss_rawobj.c was added on branch b_new_cmd on 2006-09-11 22:05:00 +0000
tappro [Wed, 12 Apr 2006 22:20:37 +0000 (22:20 +0000)]
lu_context changes
tappro [Wed, 12 Apr 2006 22:18:56 +0000 (22:18 +0000)]
fill lu_attr while object initialization
lu_context changes
tappro [Wed, 12 Apr 2006 22:17:26 +0000 (22:17 +0000)]
remove unneeded methods and apply changes due to lu_context
tappro [Wed, 12 Apr 2006 22:16:47 +0000 (22:16 +0000)]
add lu_context to the methods
use lu_attr in mdt_getattr()
tappro [Wed, 12 Apr 2006 22:15:20 +0000 (22:15 +0000)]
introduce the lu_context and lu_attr structures
convert struct context to the struct md_params
add lu_context parameter to the lu_ and md_ methods
tappro [Wed, 12 Apr 2006 22:12:32 +0000 (22:12 +0000)]
dt_data initialization is commented out cause it is broken. Umka, take a look please
calculated inode number is copied to the mdt_body
nikita [Wed, 12 Apr 2006 21:02:45 +0000 (21:02 +0000)]
pointer to super-block lives in osd_device, use it
yury [Wed, 12 Apr 2006 15:07:45 +0000 (15:07 +0000)]
- added lustre_fid.h into Makefile
yury [Wed, 12 Apr 2006 14:40:42 +0000 (14:40 +0000)]
- added fid.ko module. Simple seq management moved to it from mdt. Mdt uses it as well as ost can use it. In principle fid.ko may contain all fid related stuff like client's fid allocation one, etc.;
- renamed ll_fid2ino() to ll_fid_build_ino()
nikita [Wed, 12 Apr 2006 14:20:12 +0000 (14:20 +0000)]
split struct dt_device_operations.
fix layering in cmm.
nikita [Wed, 12 Apr 2006 13:48:42 +0000 (13:48 +0000)]
split struct md_device_operations into struct md_device_operations and struct md_object_operations
tappro [Wed, 12 Apr 2006 13:36:36 +0000 (13:36 +0000)]
wrong assignment for md_data, dt_data
put ll_i2info(root) using after check for root validity
tappro [Wed, 12 Apr 2006 13:33:51 +0000 (13:33 +0000)]
change child getting from macros to inline
add attr_get support
nikita [Wed, 12 Apr 2006 12:41:14 +0000 (12:41 +0000)]
add const qualifiers
nikita [Wed, 12 Apr 2006 12:40:23 +0000 (12:40 +0000)]
fix module initialization
yury [Wed, 12 Apr 2006 11:31:28 +0000 (11:31 +0000)]
- fixes in lli_nfs.c. Seems nfs will not work due to incompatibility with fids, but in any case it will not cause oops.
yury [Wed, 12 Apr 2006 10:45:26 +0000 (10:45 +0000)]
- pass ino into ILOOKUP() in search_inode_for_lustre()
yury [Wed, 12 Apr 2006 10:30:28 +0000 (10:30 +0000)]
- convert llite_nfs.c to using fids
yury [Wed, 12 Apr 2006 09:44:53 +0000 (09:44 +0000)]
file iallocsem_consistency.patch was added on branch b_new_cmd on 2006-05-19 11:29:25 +0000
huanghua [Wed, 12 Apr 2006 08:25:25 +0000 (08:25 +0000)]
module_setup.sh: copy newly added modules into kernel module directory
loadmod_all.sh: load modules needed for Lustre. Users may only load modules they need.
also setup some debug features.
huanghua [Wed, 12 Apr 2006 07:12:57 +0000 (07:12 +0000)]
add "#if 0" to remove "defined but not used" warnings.
huanghua [Wed, 12 Apr 2006 06:30:40 +0000 (06:30 +0000)]
Fixed a bug: when stealing config log for add mdc & osc from clients', I should
change the name of the mdc&osc, instead of uuid.
nikita [Tue, 11 Apr 2006 20:47:57 +0000 (20:47 +0000)]
rename functions to follow prefix_object_verb() style; export mdt_seq_alloc
nikita [Tue, 11 Apr 2006 18:55:16 +0000 (18:55 +0000)]
osd: implement ->dt_get_root() method
nikita [Tue, 11 Apr 2006 18:54:34 +0000 (18:54 +0000)]
add const here and there
nikita [Tue, 11 Apr 2006 18:34:48 +0000 (18:34 +0000)]
osd: add ->dt_root_get() and ->object_release() methods
nikita [Tue, 11 Apr 2006 18:34:02 +0000 (18:34 +0000)]
mdt_init0(): cleanup
nikita [Tue, 11 Apr 2006 18:33:01 +0000 (18:33 +0000)]
move device_{init,fini}() into lu_device_type_operations, and add "const" to second argument of ->ldto_device_init()
nikita [Tue, 11 Apr 2006 17:45:12 +0000 (17:45 +0000)]
fix object allocation sequence
yury [Tue, 11 Apr 2006 16:36:40 +0000 (16:36 +0000)]
- for proto better use simple seq. allocation without using meta-sequence which is going to improve later;
- implement simple ll_fid_alloc() which allocates new fids on client;
- implemented simple ll_fid2ino() which allocates client inode numbers from passed fid;
- root object is allocated in separate sequence with number 1. Sequences and fid numbers with value 0 are not allowed by ldlm - fixing that. First seq is 1, first fid in seq is 1;
- implemented simple mdt_alloc_seq() which allocates new sequences to clients;
- fixes in mdt_init_seq();
- using sema instead of spinlock in mdt_alloc_seq(), as bumping the seq and saving it to back store should be atomic operation and sleeping on possible IO with lcoked spin lock is not allowed.
yury [Tue, 11 Apr 2006 13:27:30 +0000 (13:27 +0000)]
- added ->md_config() and ->dt_config() method for getting/setting named configuration data from/to back store device. For instance, MDT needs to get saved meta-sequence info from lower layers (osd);
- added mdt_metaseq_init() method called from mdt_init0();
- added basic and primitive stuff about meta-sequence management in mdt;
- added some comments, fixes in mdt_init0() about correct handling error cases (memort freeing, etc.);
- fixed memory leak in mdt_fini(). There d->ld_site was not freed;
tappro [Tue, 11 Apr 2006 06:48:17 +0000 (06:48 +0000)]
new init
get lustre_mount_info and attach the mounted device to the osd
working statfs()
tappro [Tue, 11 Apr 2006 06:46:48 +0000 (06:46 +0000)]
new init schema
statfs, root_get support
root fid is pre-defined
tappro [Tue, 11 Apr 2006 06:45:31 +0000 (06:45 +0000)]
new init schema
statfs, root_get support
tappro [Tue, 11 Apr 2006 06:44:31 +0000 (06:44 +0000)]
add support for new init schema
add .mdo_statfs(), .mdo_root_get() with calling the child device
add mdt_connect() which accept client connection
add check for req->rq_export in mdt_handle()
add initial proto for getattr()
tappro [Tue, 11 Apr 2006 06:41:18 +0000 (06:41 +0000)]
- add two methods for lu_device - init/fini. Aftef all lu_device are allocateed,
the top device call init() through all stack to start them. It is needed
because there are insufficient info for startup until all layers are setup
- add statfs() support in dt_device_ops/md_device_ops
tappro [Tue, 11 Apr 2006 06:37:57 +0000 (06:37 +0000)]
- make MDT the 'server' device. It is needed for client connection
- comment out lmv stuff temporary because there is no lmv code in b_new_cmd yet
- add old-way mdc setup in mds logs
huanghua [Tue, 11 Apr 2006 04:26:50 +0000 (04:26 +0000)]
msg_llog.c: modify a debug message.
cmm_object.c:cmm_object_alloc: call lu_object_init to initialize it.
yury [Mon, 10 Apr 2006 14:21:44 +0000 (14:21 +0000)]
- allocation of new fids before sending open or create RPC, using allocated fid in ll_prep_inode()
tappro [Sun, 9 Apr 2006 11:01:52 +0000 (11:01 +0000)]
-remove unneeded stuff from cmm
-update mds layers names and creation
-osd gets mountpoint
-fix issue in ldlm_lockd - wrong comparision
tappro [Sat, 8 Apr 2006 08:17:25 +0000 (08:17 +0000)]
add layers setup and initialization.
nikita [Fri, 7 Apr 2006 21:03:56 +0000 (21:03 +0000)]
mdd_device: adjust to dt_object, and some cleanups
nikita [Fri, 7 Apr 2006 21:02:55 +0000 (21:02 +0000)]
osd_device: adjust for dt_object
nikita [Fri, 7 Apr 2006 21:02:13 +0000 (21:02 +0000)]
lu_object: 0. add ->ldo_object_delete() method; 1. add dt_object interface
tappro [Fri, 7 Apr 2006 14:46:15 +0000 (14:46 +0000)]
add osd
tappro [Fri, 7 Apr 2006 14:40:36 +0000 (14:40 +0000)]
add prototype method for mdt log which setup whole MDS stack
tappro [Fri, 7 Apr 2006 14:38:59 +0000 (14:38 +0000)]
use new simple MDS
tappro [Fri, 7 Apr 2006 14:38:35 +0000 (14:38 +0000)]
use mdt0 - new mdt device
tappro [Fri, 7 Apr 2006 14:36:12 +0000 (14:36 +0000)]
make MDS very simple. It is needed only for setup and maybe for some global
things later.
tappro [Fri, 7 Apr 2006 14:33:55 +0000 (14:33 +0000)]
add name for MDD
tappro [Fri, 7 Apr 2006 14:33:05 +0000 (14:33 +0000)]
add lu2md device convertor
tappro [Fri, 7 Apr 2006 14:30:31 +0000 (14:30 +0000)]
export methods for MDT
tappro [Fri, 7 Apr 2006 14:28:07 +0000 (14:28 +0000)]
update CMM device
tappro [Fri, 7 Apr 2006 14:25:56 +0000 (14:25 +0000)]
init procfs correctly
tappro [Fri, 7 Apr 2006 14:25:08 +0000 (14:25 +0000)]
update alloc() method to link CMM
tappro [Fri, 7 Apr 2006 14:23:53 +0000 (14:23 +0000)]
move md_ stuff from here and use md_object.h
yury [Fri, 7 Apr 2006 13:52:44 +0000 (13:52 +0000)]
- fixes in llite with using mds_body vs. mdt_body
tappro [Fri, 7 Apr 2006 12:06:11 +0000 (12:06 +0000)]
-wrong name 'mdt' should be 'osd'
huanghua [Fri, 7 Apr 2006 10:34:16 +0000 (10:34 +0000)]
add some code to parse newly added command: lmv,cmm.
yury [Fri, 7 Apr 2006 10:29:35 +0000 (10:29 +0000)]
- many changes about fids:
(1) client, mdc and liblustre converted to use new fids;
(2) some md related structures renamed into mdt ones to use new fids with no changing anything about MDS code;
(3) md methods used obdo switched to use mdc_op_data like others;
(4) f_num is now replaced to f_ver and f_oid.
huanghua [Fri, 7 Apr 2006 09:09:33 +0000 (09:09 +0000)]
Fixex two bugs: 1. add lov & lmv for clients which have already had lov&lmv in mgs_write_log_mdt;
2. skip adding mdt to cmm which resides on this mdt itself.
basically, it passed simple tests. It seems that MGS generates correct config logs. :)
huanghua [Fri, 7 Apr 2006 04:41:33 +0000 (04:41 +0000)]
finished the last function in MGS: mgs_steal_llog_for_mdt_from_client.
compilation passed.
mgs_steal_llog_for_mdt_from_client has the following steps:
1. find a "marker" to identify the start of adding osc/mdc for clients;
2. duplicate config log from clients' to mdt's:
(a) add start marker;
(b) duplicate add_uuid;
(c) duplicate attach, modify uuid for the new obd;
(d) duplicate setup;
(e) duplicate lov_add to lov_add (or lmv_add to cmm_add);
(f) add end marker;
3. got a "marker" to end. goto 1. to find another.
nikita [Thu, 6 Apr 2006 21:25:32 +0000 (21:25 +0000)]
small fixes to mdt: add few statics, and remove osd declaration
nikita [Thu, 6 Apr 2006 21:23:14 +0000 (21:23 +0000)]
add osd skeleton
ericm [Thu, 6 Apr 2006 19:20:21 +0000 (19:20 +0000)]
file sec_gss.c was added on branch b_new_cmd on 2006-09-11 22:05:00 +0000
ericm [Thu, 6 Apr 2006 19:20:19 +0000 (19:20 +0000)]
file gss_cli_upcall.c was added on branch b_new_cmd on 2006-09-11 22:05:00 +0000
nikita [Thu, 6 Apr 2006 13:47:53 +0000 (13:47 +0000)]
lustre_msg_check_version(): straighten a bit
nikita [Thu, 6 Apr 2006 13:47:08 +0000 (13:47 +0000)]
mdt prototype: 0. add handlers for dlm requests; 1. add fid_build_res_name() and fid_res_name_eq() (should go into separate file, ultimately); 2. add some stubs for interoperability code
nikita [Thu, 6 Apr 2006 12:47:08 +0000 (12:47 +0000)]
fix boilerplate
wangdi [Thu, 6 Apr 2006 09:33:03 +0000 (09:33 +0000)]
Branch b_new_cmd
1) add mdd_put_object in release method to destory the object when nlink of the
object is zero.
2) some minor fixes
yury [Thu, 6 Apr 2006 07:41:54 +0000 (07:41 +0000)]
- fixes in lustre_swab_lu_fid()
yury [Thu, 6 Apr 2006 07:40:46 +0000 (07:40 +0000)]
- changes about fids:
- f_seq is 64bit field to allow to create not less than 2 ^ 64 objects;
- f_num is compound 64bit field containing oid (object id) and object version;
- for issuing dlm locks we use f_num.
huanghua [Thu, 6 Apr 2006 05:33:38 +0000 (05:33 +0000)]
partially of adding MDT. passed compile.
nikita [Wed, 5 Apr 2006 19:24:43 +0000 (19:24 +0000)]
generalize fid_{,un}lock() interface
nikita [Wed, 5 Apr 2006 19:05:12 +0000 (19:05 +0000)]
add "inline" to make compiler happy
yury [Wed, 5 Apr 2006 18:09:24 +0000 (18:09 +0000)]
- preparation before converting client to new fids
wangdi [Wed, 5 Apr 2006 06:20:31 +0000 (06:20 +0000)]
Branch: b_new_cmd
update mdd prototype
yury [Tue, 4 Apr 2006 13:24:23 +0000 (13:24 +0000)]
- added fid_lock()/fid_unlock()
- many changes about using fid. Idea is that we do not want to change whole mds stuff to use new fids but we want to use new fids for new modules.
tappro [Tue, 4 Apr 2006 09:08:52 +0000 (09:08 +0000)]
anoter occurance of ll_fid
tappro [Tue, 4 Apr 2006 08:58:37 +0000 (08:58 +0000)]
use lu_fid instead of ll_fid
wangdi [Tue, 4 Apr 2006 08:27:04 +0000 (08:27 +0000)]
Branch: b_new_cmd
1.move share structure to mdd
2.some minor fixes
tappro [Tue, 4 Apr 2006 07:44:01 +0000 (07:44 +0000)]
move cmm name to the obd.h
tappro [Tue, 4 Apr 2006 07:41:12 +0000 (07:41 +0000)]
add missed makefile stuff
yury [Tue, 4 Apr 2006 06:41:15 +0000 (06:41 +0000)]
- renames of ll_fid (what is expanded to llite_fid) into lu_fid (lustre_fid)
tappro [Mon, 3 Apr 2006 18:29:26 +0000 (18:29 +0000)]
add CMM to makefiles
tappro [Mon, 3 Apr 2006 18:28:25 +0000 (18:28 +0000)]
CMM support
tappro [Mon, 3 Apr 2006 18:27:35 +0000 (18:27 +0000)]
generic cmm device support
tappro [Mon, 3 Apr 2006 18:23:41 +0000 (18:23 +0000)]
move generic md_ stuff from mdt.h to md_object.h. It is extention for
lu_object.h