)]}' { "log": [ { "commit": "1187c968852e3c668f3b9376083851f81f6eee22", "tree": "23a34ab89f724cc015f9e4f6e8bc6ed0fc20ce51", "parents": [ "08413899db89d8d636c2a2d4ba5c356ab587d7ef" ], "author": { "name": "Joel Becker", "email": "joel.becker@oracle.com", "time": "Wed Sep 03 20:03:39 2008 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 16:57:07 2008 -0700" }, "message": "ocfs2: Limit inode allocation to 32bits.\n\nocfs2 inode numbers are block numbers. For any filesystem with less\nthan 2^32 blocks, this is not a problem. However, when ocfs2 starts\nusing JDB2, it will be able to support filesystems with more than 2^32\nblocks. This would result in inode numbers higher than 2^32.\n\nThe problem is that stat(2) can\u0027t handle those numbers on 32bit\nmachines. The simple solution is to have ocfs2 allocate all inodes\nbelow that boundary.\n\nThe suballoc code is changed to honor an optional block limit. Only the\ninode suballocator sets that limit - all other allocations stay unlimited.\n\nThe biggest trick is to grow the inode suballocator beneath that limit.\nThere\u0027s no point in allocating block groups that are above the limit,\nthen rejecting their elements later on. We want to prevent the inode\nallocator from ever having block groups above the limit. This involves\na little gyration with the local alloc code. If the local alloc window\nis above the limit, it signals the caller to try the global bitmap but\ndoes not disable the local alloc file (which can be used for other\nallocations).\n\n[ Minor cleanup - removed an ML_NOTICE comment. --Mark ]\n\nSigned-off-by: Joel Becker \u003cjoel.becker@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "f99b9b7ccf6a691f653cec45f36bfdd1e94769c7", "tree": "1c6ff6ea1fa1bb86b70f1fd78dd725b559c729e4", "parents": [ "1e61ee79e2a96f62c007486677319814ce621c3c" ], "author": { "name": "Joel Becker", "email": "joel.becker@oracle.com", "time": "Wed Aug 20 19:36:33 2008 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 16:57:05 2008 -0700" }, "message": "ocfs2: Make ocfs2_extent_tree the first-class representation of a tree.\n\nWe now have three different kinds of extent trees in ocfs2: inode data\n(dinode), extended attributes (xattr_tree), and extended attribute\nvalues (xattr_value). There is a nice abstraction for them,\nocfs2_extent_tree, but it is hidden in alloc.c. All the calling\nfunctions have to pick amongst a varied API and pass in type bits and\noften extraneous pointers.\n\nA better way is to make ocfs2_extent_tree a first-class object.\nEveryone converts their object to an ocfs2_extent_tree() via the\nocfs2_get_*_extent_tree() calls, then uses the ocfs2_extent_tree for all\ntree calls to alloc.c.\n\nThis simplifies a lot of callers, making for readability. It also\nprovides an easy way to add additional extent tree types, as they only\nneed to be defined in alloc.c with a ocfs2_get_\u003cnew\u003e_extent_tree()\nfunction.\n\nSigned-off-by: Joel Becker \u003cjoel.becker@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "cf1d6c763fbcb115263114302485ad17e7933d87", "tree": "85717dc9d20b7ac1e31b683653933f2d30099f44", "parents": [ "fdd77704a8b4666a32120fcd1e4a9fedaf3263d8" ], "author": { "name": "Tiger Yang", "email": "tiger.yang@oracle.com", "time": "Mon Aug 18 17:11:00 2008 +0800" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 16:57:02 2008 -0700" }, "message": "ocfs2: Add extended attribute support\n\nThis patch implements storing extended attributes both in inode or a single\nexternal block. We only store EA\u0027s in-inode when blocksize \u003e 512 or that\ninode block has free space for it. When an EA\u0027s value is larger than 80\nbytes, we will store the value via b-tree outside inode or block.\n\nSigned-off-by: Tiger Yang \u003ctiger.yang@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "f56654c435c06f2b2bd5751889b1a08a3add7d6c", "tree": "b186d68aedc5dda7afe435f5a68c03937ae382ff", "parents": [ "ac11c827192272eabb68b8f4cf844066461d9690" ], "author": { "name": "Tao Ma", "email": "tao.ma@oracle.com", "time": "Mon Aug 18 17:38:48 2008 +0800" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 16:57:01 2008 -0700" }, "message": "ocfs2: Add extent tree operation for xattr value btrees\n\nAdd some thin wrappers around ocfs2_insert_extent() for each of the 3\ndifferent btree types, ocfs2_inode_insert_extent(),\nocfs2_xattr_value_insert_extent() and ocfs2_xattr_tree_insert_extent(). The\nlast is for the xattr index btree, which will be used in a followup patch.\n\nAll the old callers in file.c etc will call ocfs2_dinode_insert_extent(),\nwhile the other two handle the xattr issue. And the init of extent tree are\nhandled by these functions.\n\nWhen storing xattr value which is too large, we will allocate some clusters\nfor it and here ocfs2_extent_list and ocfs2_extent_rec will also be used. In\norder to re-use the b-tree operation code, a new parameter named \"private\"\nis added into ocfs2_extent_tree and it is used to indicate the root of\nocfs2_exent_list. The reason is that we can\u0027t deduce the root from the\nbuffer_head now. It may be in an inode, an ocfs2_xattr_block or even worse,\nin any place in an ocfs2_xattr_bucket.\n\nSigned-off-by: Tao Ma \u003ctao.ma@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "e7d4cb6bc19658646357eeff134645cd9bc3479f", "tree": "4e24321e5c28fb90dffa1f396972fddb3c458e58", "parents": [ "811f933df1e55615fd0bb4818f31e3868a8e6e23" ], "author": { "name": "Tao Ma", "email": "tao.ma@oracle.com", "time": "Mon Aug 18 17:38:44 2008 +0800" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 13:57:58 2008 -0700" }, "message": "ocfs2: Abstract ocfs2_extent_tree in b-tree operations.\n\nIn the old extent tree operation, we take the hypothesis that we\nare using the ocfs2_extent_list in ocfs2_dinode as the tree root.\nAs xattr will also use ocfs2_extent_list to store large value\nfor a xattr entry, we refactor the tree operation so that xattr\ncan use it directly.\n\nThe refactoring includes 4 steps:\n1. Abstract set/get of last_eb_blk and update_clusters since they may\n be stored in different location for dinode and xattr.\n2. Add a new structure named ocfs2_extent_tree to indicate the\n extent tree the operation will work on.\n3. Remove all the use of fe_bh and di, use root_bh and root_el in\n extent tree instead. So now all the fe_bh is replaced with\n et-\u003eroot_bh, el with root_el accordingly.\n4. Make ocfs2_lock_allocators generic. Now it is limited to be only used\n in file extend allocation. But the whole function is useful when we want\n to store large EAs.\n\nNote: This patch doesn\u0027t touch ocfs2_commit_truncate() since it is not used\nfor anything other than truncate inode data btrees.\n\nSigned-off-by: Tao Ma \u003ctao.ma@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "811f933df1e55615fd0bb4818f31e3868a8e6e23", "tree": "53335e59bdff02d7ddb3a2ec15d36239fe50003d", "parents": [ "231b87d10920e024efaf0f9e86e1bab7bced1620" ], "author": { "name": "Tao Ma", "email": "tao.ma@oracle.com", "time": "Mon Aug 18 17:38:43 2008 +0800" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 13:57:58 2008 -0700" }, "message": "ocfs2: Use ocfs2_extent_list instead of ocfs2_dinode.\n\nocfs2_extend_meta_needed(), ocfs2_calc_extend_credits() and\nocfs2_reserve_new_metadata() are all useful for extent tree operations. But\nthey are all limited to an inode btree because they use a struct\nocfs2_dinode parameter. Change their parameter to struct ocfs2_extent_list\n(the part of an ocfs2_dinode they actually use) so that the xattr btree code\ncan use these functions.\n\nSigned-off-by: Tao Ma \u003ctao.ma@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "9c7af40b210e87f8fddd97b0badc0a352862234a", "tree": "deb248e5295439b96ad9bd1789e8512f93803142", "parents": [ "ebcee4b5c9136096f64ee6f691a013d7c0a4bc34" ], "author": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Jul 28 18:02:53 2008 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Mon Oct 13 13:57:57 2008 -0700" }, "message": "ocfs2: throttle back local alloc when low on disk space\n\nOcfs2\u0027s local allocator disables itself for the duration of a mount point\nwhen it has trouble allocating a large enough area from the primary bitmap.\nThat can cause performance problems, especially for disks which were only\ntemporarily full or fragmented. This patch allows for the allocator to\nshrink it\u0027s window first, before being disabled. Later, it can also be\nre-enabled so that any performance drop is minimized.\n\nTo do this, we allow the value of osb-\u003elocal_alloc_bits to be shrunk when\nneeded. The default value is recorded in a mostly read-only variable so that\nwe can re-initialize when required.\n\nLocking had to be updated so that we could protect changes to\nlocal_alloc_bits. Mostly this involves protecting various local alloc values\nwith the osb spinlock. A new state is also added, OCFS2_LA_THROTTLED, which\nis used when the local allocator is has shrunk, but is not disabled. If the\navailable space dips below 1 megabyte, the local alloc file is disabled. In\neither case, local alloc is re-enabled 30 seconds after the event, or when\nan appropriate amount of bits is seen in the primary bitmap.\n\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "a4a4891164d4f6f383cc17e7c90828a7ca6a1146", "tree": "a0a881ae3f1910d5f4e4aa9824f93781616ef99f", "parents": [ "ffda89a3bf3b968bdc268584c6bc1da5c173cf12" ], "author": { "name": "Tao Ma", "email": "tao.ma@oracle.com", "time": "Mon Mar 03 17:12:30 2008 +0800" }, "committer": { "name": "Mark Fasheh", "email": "mfasheh@suse.com", "time": "Fri Apr 18 08:56:10 2008 -0700" }, "message": "ocfs2: Add ac_alloc_slot in ocfs2_alloc_context\n\nIn inode stealing, we no longer restrict the allocation to\nhappen in the local node. So it is neccessary for us to add\na new member in ocfs2_alloc_context to indicate which slot\nwe are using for allocation. We also modify the process of\nlocal alloc so that this member can be used there also.\n\nSigned-off-by: Tao Ma \u003ctao.ma@oracle.com\u003e\nSigned-off-by: Sunil Mushran \u003csunil.mushran@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmfasheh@suse.com\u003e\n" }, { "commit": "d659072f736837e56b6433d58e5315ad1d4d5ccf", "tree": "56882b7b36c6b60a8208f6ed5bee9904adc7b649", "parents": [ "7f68fc28219be3b44ef4132f95c6506ff3e806b5" ], "author": { "name": "Tao Ma", "email": "tao.ma@oracle.com", "time": "Tue Dec 18 15:47:03 2007 +0800" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Fri Jan 25 14:53:35 2008 -0800" }, "message": "[PATCH 1/2] ocfs2: Add group extend for online resize\n\nThis patch adds the ability for a userspace program to request an extend of\nlast cluster group on an Ocfs2 file system. The request is made via ioctl,\nOCFS2_IOC_GROUP_EXTEND. This is derived from EXT3_IOC_GROUP_EXTEND, but is\nobviously Ocfs2 specific.\n\ntunefs.ocfs2 would call this for an online-resize operation if the last\ncluster group isn\u0027t full.\n\nSigned-off-by: Tao Ma \u003ctao.ma@oracle.com\u003e\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "415cb800375cc4e89fb5a6a454e484bd4adbffb4", "tree": "816021a5279047702d5fa8180783b1c710f31746", "parents": [ "6d0b842d3bf0cc027dcff57a89fb8a6b1fd610e1" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Sun Sep 16 20:10:16 2007 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Thu Sep 20 15:06:09 2007 -0700" }, "message": "ocfs2: Allow smaller allocations during large writes\n\nThe ocfs2 write code loops through a page much like the block code, except\nthat ocfs2 allocation units can be any size, including larger than page\nsize. Typically it\u0027s equal to or larger than page size - most kernels run 4k\npages, the minimum ocfs2 allocation (cluster) size.\n\nSome changes introduced during 2.6.23 changed the way writes to pages are\nhandled, and inadvertantly broke support for \u003e 4k page size. Instead of just\nwriting one cluster at a time, we now handle the whole page in one pass.\n\nThis means that multiple (small) seperate allocations might happen in the\nsame pass. The allocation code howver typically optimizes by getting the\nmaximum which was reserved. This triggered a BUG_ON in the extend code where\nit\u0027d ask for a single bit (for one part of a \u003e 4k page) and get back more\nthan it asked for.\n\nFix this by providing a variant of the high level allocation function which\nallows the caller to specify a maximum. The traditional function remains and\njust calls the new one with a maximum determined from the initial\nreservation.\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "59a5e416d1ab543a5248a2b34d83202c4d55d132", "tree": "3e97ce48774d26e8a927edd839dd54a7d68f8c7d", "parents": [ "2b604351bc99b4e4504758cbac369b660b71de0b" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Fri Jun 22 15:52:36 2007 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Tue Jul 10 17:31:55 2007 -0700" }, "message": "ocfs2: plug truncate into cached dealloc routines\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "2b604351bc99b4e4504758cbac369b660b71de0b", "tree": "293fa51f1ae9d19db0d09c721cc8433303cc8974", "parents": [ "bce997682fe3121516f5a20cf7bad2e6029ba018" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Fri Jun 22 15:45:27 2007 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Tue Jul 10 17:31:54 2007 -0700" }, "message": "ocfs2: simplify deallocation locking\n\nDeallocation of suballocator blocks, most notably extent blocks, might\ninvolve multiple suballocator inodes.\n\nThe locking for this can get extremely complicated, especially when the\nsuballocator inodes to delete from aren\u0027t known until deep within an\nunrelated codepath.\n\nImplement a simple scheme for recording the blocks to be unlinked so that\nthe actual deallocation can be done in a context which won\u0027t deadlock.\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "1fabe1481fac9e01bf8bffa60a2307ef379aa5de", "tree": "17092c1be837ed95c8f26646003e9e49cfdb9663", "parents": [ "65eff9ccf86d63eb5c3e9071450a36e4e4fa9564" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Mon Oct 09 18:11:45 2006 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Fri Dec 01 18:28:28 2006 -0800" }, "message": "ocfs2: Remove struct ocfs2_journal_handle in favor of handle_t\n\nThis is mostly a search and replace as ocfs2_journal_handle is now no more\nthan a container for a handle_t pointer.\n\nocfs2_commit_trans() becomes very straight forward, and we remove some out\nof date comments / code.\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "da5cbf2f9df922cfdafa39351691fa83517f1e25", "tree": "845dc288b72f0408870f50605bb3c96eec978dd8", "parents": [ "8d5596c687c49c1d8812c3456946dec15d069139" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Fri Oct 06 18:34:35 2006 -0700" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Fri Dec 01 18:27:49 2006 -0800" }, "message": "ocfs2: don\u0027t use handle for locking in allocation functions\n\nInstead we record our state on the allocation context structure which all\ncallers already know about and lifetime correctly. This means the\nreservation functions don\u0027t need a handle passed in any more, and we can\nalso take it off the alloc context.\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "883d4cae4a2b01a05193cf2665c77b7489a8b6a0", "tree": "1d4f06fa7bdadf025159ff5408f14080552a0cb0", "parents": [ "7bf72edee614e10b8d470c40a326f47bfdd69992" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Mon Jun 05 16:41:00 2006 -0400" }, "committer": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Mon Aug 07 11:07:01 2006 -0700" }, "message": "ocfs2: allocation hints\n\nRecord the most recently used allocation group on the allocation context, so\nthat subsequent allocations can attempt to optimize for contiguousness.\nLocal alloc especially should benefit from this as the current chain search\ntends to let it spew across the disk.\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\n" }, { "commit": "ccd979bdbce9fba8412beb3f1de68a9d0171b12c", "tree": "c50ed941849ce06ccadd4ce27599b3ef9fdbe2ae", "parents": [ "8df08c89c668e1bd922a053fdb5ba1fadbecbb38" ], "author": { "name": "Mark Fasheh", "email": "mark.fasheh@oracle.com", "time": "Thu Dec 15 14:31:24 2005 -0800" }, "committer": { "name": "Joel Becker", "email": "joel.becker@oracle.com", "time": "Tue Jan 03 11:45:47 2006 -0800" }, "message": "[PATCH] OCFS2: The Second Oracle Cluster Filesystem\n\nThe OCFS2 file system module.\n\nSigned-off-by: Mark Fasheh \u003cmark.fasheh@oracle.com\u003e\nSigned-off-by: Kurt Hackel \u003ckurt.hackel@oracle.com\u003e\n" } ] }