)]}' { "log": [ { "commit": "159ec1fc060ab22b157a62364045f5e98749c4d3", "tree": "1de0edfd782245b271d2898e36ae76c00e1e1b6d", "parents": [ "ccacc7d2cf03114a24ab903f710118e9e5d43273" ], "author": { "name": "Cheng Renquan", "email": "crquan@gmail.com", "time": "Fri Jan 09 08:31:08 2009 +1100" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Fri Jan 09 08:31:08 2009 +1100" }, "message": "md: use list_for_each_entry macro directly\n\nThe rdev_for_each macro defined in \u003clinux/raid/md_k.h\u003e is identical to\nlist_for_each_entry_safe, from \u003clinux/list.h\u003e, it should be defined to\nuse list_for_each_entry_safe, instead of reinventing the wheel.\n\nBut some calls to each_entry_safe don\u0027t really need a safe version,\njust a direct list_for_each_entry is enough, this could save a temp\nvariable (tmp) in every function that used rdev_for_each.\n\nIn this patch, most rdev_for_each loops are replaced by list_for_each_entry,\ntotally save many tmp vars; and only in the other situations that will call\nlist_del to delete an entry, the safe version is used.\n\nSigned-off-by: Cheng Renquan \u003ccrquan@gmail.com\u003e\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "4bbf3771ca40d0aaec8316d0e7476b16010288e5", "tree": "f7cba1f72612c5d2132f549881c46ff572e01933", "parents": [ "d710e13812600037a723a673dc5c96a071de98d3" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Oct 13 11:55:12 2008 +1100" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Oct 13 11:55:12 2008 +1100" }, "message": "md: Relax minimum size restrictions on chunk_size.\n\nCurrently, the \u0027chunk_size\u0027 of an array must be at-least PAGE_SIZE.\n\nThis makes moving an array to a machine with a larger PAGE_SIZE, or\nchanging the kernel to use a larger PAGE_SIZE, can stop an array from\nworking.\n\nFor RAID10 and RAID4/5/6, this is non-trivial to fix as the resync\nprocess works on whole pages at a time, and assumes them to be wholly\nwithin a stripe. For other raid personalities, this restriction is\nnot needed at all and can be dropped.\n\nSo remove the test on chunk_size from common can, and add it in just\nthe places where it is needed: raid10 and raid4/5/6.\n\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "d710e13812600037a723a673dc5c96a071de98d3", "tree": "df8e34fb02cb7909090aa697da7c1db4814ee122", "parents": [ "fb4d8c76e56a887b9eee99fbc55fe82b18625d30" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Oct 13 11:55:12 2008 +1100" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Oct 13 11:55:12 2008 +1100" }, "message": "md: remove space after function name in declaration and call.\n\nHaving\n function (args)\ninstead of\n function(args)\n\nmake is harder to search for calls of particular functions.\nSo remove all those spaces.\n\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "fb4d8c76e56a887b9eee99fbc55fe82b18625d30", "tree": "0ca96887e40db671957b79d493830b9527b04949", "parents": [ "ab5bd5cbc8d4b868378d062eed3d4240930fbb86" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Oct 13 11:55:12 2008 +1100" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Oct 13 11:55:12 2008 +1100" }, "message": "md: Remove unnecessary #includes, #defines, and function declarations.\n\nA lot of cruft has gathered over the years. Time to remove it.\n\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "074a7aca7afa6f230104e8e65eba3420263714a5", "tree": "f418313e45bd55be8156c8a3e8f9a216cf63058d", "parents": [ "eddb2e26b5ee3c5da68ba4bf1921ba20e2097bff" ], "author": { "name": "Tejun Heo", "email": "tj@kernel.org", "time": "Mon Aug 25 19:56:14 2008 +0900" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Thu Oct 09 08:56:08 2008 +0200" }, "message": "block: move stats from disk to part0\n\nMove stats related fields - stamp, in_flight, dkstats - from disk to\npart0 and unify stat handling such that...\n\n* part_stat_*() now updates part0 together if the specified partition\n is not part0. ie. part_stat_*() are now essentially all_stat_*().\n\n* {disk|all}_stat_*() are gone.\n\n* part_round_stats() is updated similary. It handles part0 stats\n automatically and disk_round_stats() is killed.\n\n* part_{inc|dec}_in_fligh() is implemented which automatically updates\n part0 stats for parts other than part0.\n\n* disk_map_sector_rcu() is updated to return part0 if no part matches.\n Combined with the above changes, this makes NULL special case\n handling in callers unnecessary.\n\n* Separate stats show code paths for disk are collapsed into part\n stats show code paths.\n\n* Rename disk_stat_lock/unlock() to part_stat_lock/unlock()\n\nWhile at it, reposition stat handling macros a bit and add missing\nparentheses around macro parameters.\n\nSigned-off-by: Tejun Heo \u003ctj@kernel.org\u003e\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "c9959059161ddd7bf4670cf47367033d6b2f79c4", "tree": "6454db55f8e34361fe472358e10e0c5cfac1e366", "parents": [ "e71bf0d0ee89e51b92776391c5634938236977d5" ], "author": { "name": "Tejun Heo", "email": "tj@kernel.org", "time": "Mon Aug 25 19:47:21 2008 +0900" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Thu Oct 09 08:56:06 2008 +0200" }, "message": "block: fix diskstats access\n\nThere are two variants of stat functions - ones prefixed with double\nunderbars which don\u0027t care about preemption and ones without which\ndisable preemption before manipulating per-cpu counters. It\u0027s unclear\nwhether the underbarred ones assume that preemtion is disabled on\nentry as some callers don\u0027t do that.\n\nThis patch unifies diskstats access by implementing disk_stat_lock()\nand disk_stat_unlock() which take care of both RCU (for partition\naccess) and preemption (for per-cpu counter access). diskstats access\nshould always be enclosed between the two functions. As such, there\u0027s\nno need for the versions which disables preemption. They\u0027re removed\nand double underbars ones are renamed to drop the underbars. As an\nextra argument is added, there\u0027s no danger of using the old version\nunconverted.\n\ndisk_stat_lock() uses get_cpu() and returns the cpu index and all\ndiskstat functions which access per-cpu counters now has @cpu\nargument to help RT.\n\nThis change adds RCU or preemption operations at some places but also\ncollapses several preemption ops into one at others. Overall, the\nperformance difference should be negligible as all involved ops are\nvery lightweight per-cpu ones.\n\nSigned-off-by: Tejun Heo \u003ctj@kernel.org\u003e\nCc: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "5b99c2ffa980528a197f26c7d876cceeccce8dd5", "tree": "65e3bbbccb3f796f0569c47fcfc00e222cd3066e", "parents": [ "960e739d9e9f1c2346d8bdc65299ee2e1ed42218" ], "author": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Fri Aug 15 10:56:11 2008 +0200" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Thu Oct 09 08:56:03 2008 +0200" }, "message": "block: make bi_phys_segments an unsigned int instead of short\n\nraid5 can overflow with more than 255 stripes, and we can increase it\nto an int for free on both 32 and 64-bit archs due to the padding.\n\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "960e739d9e9f1c2346d8bdc65299ee2e1ed42218", "tree": "d3b38cbfbf503338cdf08e9ad6c4d41f7684ce34", "parents": [ "5df97b91b5d7ed426034fcc84cb6e7cf682b8838" ], "author": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Fri Aug 15 10:41:18 2008 +0200" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Thu Oct 09 08:56:03 2008 +0200" }, "message": "block: raid fixups for removal of bi_hw_segments\n\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "ac4090d24c6a26211bc4523d920376e054d4f3f8", "tree": "5aabdbb2448967cbfc23be2fa6714d1c1bbf29dc", "parents": [ "dba034eef2456d2a9f9a76806846c97acf6c3ad1" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Tue Aug 05 15:54:13 2008 +1000" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Tue Aug 05 15:56:32 2008 +1000" }, "message": "Don\u0027t let a blocked_rdev interfere with read request in raid5/6\n\nWhen we have externally managed metadata, we need to mark a failed\ndevice as \u0027Blocked\u0027 and not allow any writes until that device\nhave been marked as faulty in the metadata and the Blocked flag has\nbeen removed.\n\nHowever it is perfectly OK to allow read requests when there is a\nBlocked device, and with a readonly array, there may not be any\nmetadata-handler watching for blocked devices.\n\nSo in raid5/raid6 only allow a Blocked device to interfere with\nWrite request or resync. Read requests go through untouched.\n\nraid1 and raid10 already differentiate between read and write\nproperly.\n\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "dba034eef2456d2a9f9a76806846c97acf6c3ad1", "tree": "3cf51b72fc5be671e986a39c389aa5332994e2a2", "parents": [ "2b25000bf5157c28d8591f03f0575248a8cbd900" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Tue Aug 05 15:54:13 2008 +1000" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Tue Aug 05 15:56:32 2008 +1000" }, "message": "Fail safely when trying to grow an array with a write-intent bitmap.\n\nWe cannot currently change the size of a write-intent bitmap.\nSo if we change the size of an array which has such a bitmap, it\ntries to set bits beyond the end of the bitmap.\n\nFor now, simply reject any request to change the size of an array\nwhich has a bitmap. mdadm can remove the bitmap and add a new one\nafter the array has changed size.\n\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "1e24b15b267293567a8d752721c7ae63f281325a", "tree": "ee63644d2dc59899e42ec0802c201140e3a92e5e", "parents": [ "63a16f90167850010864a9e8ebb71d216983090f", "388667bed591b2359713bb17d5de0cf56e961447" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Fri Aug 01 11:56:07 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Fri Aug 01 11:56:07 2008 -0700" }, "message": "Merge branch \u0027for-linus\u0027 of git://neil.brown.name/md\n\n* \u0027for-linus\u0027 of git://neil.brown.name/md:\n md: raid10: wake up frozen array\n md: do not count blocked devices as spares\n md: do not progress the resync process if the stripe was blocked\n md: delay notification of \u0027active_idle\u0027 to the recovery thread\n md: fix merge error\n md: move async_tx_issue_pending_all outside spin_lock_irq\n" }, { "commit": "df10cfbc4d7ab93260d997df754219d390d62a9d", "tree": "23908d827de85f16f9404875859e51babb966856", "parents": [ "d8e64406a037a64444175730294e449c9e21f5ec" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Jul 28 23:10:39 2008 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Jul 28 17:52:37 2008 -0700" }, "message": "md: do not progress the resync process if the stripe was blocked\n\nhandle_stripe will take no action on a stripe when waiting for userspace\nto unblock the array, so do not report completed sectors.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\n" }, { "commit": "2339788376e2d69a9154130e4dacd5b21ce63094", "tree": "ed8f21a7a726472fd089365b8913f0e950edf12a", "parents": [ "c9f21aaff1d1fb5629325130af469532d19beb93" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Wed Jul 23 20:05:34 2008 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Wed Jul 23 13:09:45 2008 -0700" }, "message": "md: fix merge error\n\nThe original STRIPE_OP_IO removal patch had the following hunk:\n\n- for (i \u003d conf-\u003eraid_disks; i--; ) {\n+ for (i \u003d conf-\u003eraid_disks; i--; )\n set_bit(R5_Wantwrite, \u0026sh-\u003edev[i].flags);\n- if (!test_and_set_bit(STRIPE_OP_IO, \u0026sh-\u003eops.pending))\n- sh-\u003eops.count++;\n- }\n\nHowever it appears the hunk became broken after merging:\n- for (i \u003d conf-\u003eraid_disks; i--; ) {\n+ for (i \u003d conf-\u003eraid_disks; i--; )\n set_bit(R5_Wantwrite, \u0026sh-\u003edev[i].flags);\n set_bit(R5_LOCKED, \u0026dev-\u003eflags);\n s.locked++;\n- if (!test_and_set_bit(STRIPE_OP_IO, \u0026sh-\u003eops.pending))\n- sh-\u003eops.count++;\n- }\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\n" }, { "commit": "c9f21aaff1d1fb5629325130af469532d19beb93", "tree": "41bb8d2e9d0416d76698d7610b2cbda5ea496cda", "parents": [ "4b80991c6cb9efa607bc4fd6f3ecdf5511c31bb0" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Wed Jul 23 12:05:51 2008 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Wed Jul 23 12:05:51 2008 -0700" }, "message": "md: move async_tx_issue_pending_all outside spin_lock_irq\n\nSome dma drivers need to call spin_lock_bh in their device_issue_pending\nroutines. This change avoids:\n\nWARNING: at kernel/softirq.c:136 local_bh_enable_ip+0x3a/0x85()\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\n" }, { "commit": "8a392625b665c676a77c62f8608d10ff430bcb83", "tree": "4000a65d61baed73200e47f91dea5263ed16edd0", "parents": [ "519f0141f1c42e2b8b59c7dea005cbf6095358e8", "4b80991c6cb9efa607bc4fd6f3ecdf5511c31bb0" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Jul 21 10:29:12 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Jul 21 10:29:12 2008 -0700" }, "message": "Merge branch \u0027for-linus\u0027 of git://neil.brown.name/md\n\n* \u0027for-linus\u0027 of git://neil.brown.name/md: (52 commits)\n md: Protect access to mddev-\u003edisks list using RCU\n md: only count actual openers as access which prevent a \u0027stop\u0027\n md: linear: Make array_size sector-based and rename it to array_sectors.\n md: Make mddev-\u003earray_size sector-based.\n md: Make super_type-\u003erdev_size_change() take sector-based sizes.\n md: Fix check for overlapping devices.\n md: Tidy up rdev_size_store a bit:\n md: Remove some unused macros.\n md: Turn rdev-\u003esb_offset into a sector-based quantity.\n md: Make calc_dev_sboffset() return a sector count.\n md: Replace calc_dev_size() by calc_num_sectors().\n md: Make update_size() take the number of sectors.\n md: Better control of when do_md_stop is allowed to stop the array.\n md: get_disk_info(): Don\u0027t convert between signed and unsigned and back.\n md: Simplify restart_array().\n md: alloc_disk_sb(): Return proper error value.\n md: Simplify sb_equal().\n md: Simplify uuid_equal().\n md: sb_equal(): Fix misleading printk.\n md: Fix a typo in the comment to cmd_match().\n ...\n" }, { "commit": "f233ea5c9e0d8b95e4283bf6a3436b88f6fd3586", "tree": "8a36b621de46da23f3957c6e3ea4d653c2937544", "parents": [ "15f4a5fdf3aa07b53f6a7969664741db5882e485" ], "author": { "name": "Andre Noll", "email": "maan@systemlinux.org", "time": "Mon Jul 21 17:05:22 2008 +1000" }, "committer": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Jul 21 17:05:22 2008 +1000" }, "message": "md: Make mddev-\u003earray_size sector-based.\n\nThis patch renames the array_size field of struct mddev_s to array_sectors\nand converts all instances to use units of 512 byte sectors instead of 1k\nblocks.\n\nSigned-off-by: Andre Noll \u003cmaan@systemlinux.org\u003e\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "dddec01eb8e2b56267b37a6f9f0997a64b4e0b2a", "tree": "b6d8bfbce9abd105384b9d116499afbe306b9c22", "parents": [ "7daf705f362e349983e92037a198b8821db198af", "32502b8413a77b54b9e19809404109590c32dfb7" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Jul 14 13:15:14 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Jul 14 13:15:14 2008 -0700" }, "message": "Merge branch \u0027for-linus\u0027 of git://git.kernel.dk/linux-2.6-block\n\n* \u0027for-linus\u0027 of git://git.kernel.dk/linux-2.6-block: (37 commits)\n splice: fix generic_file_splice_read() race with page invalidation\n ramfs: enable splice write\n drivers/block/pktcdvd.c: avoid useless memset\n cdrom: revert commit 22a9189 (cdrom: use kmalloced buffers instead of buffers on stack)\n scsi: sr avoids useless buffer allocation\n block: blk_rq_map_kern uses the bounce buffers for stack buffers\n block: add blk_queue_update_dma_pad\n DAC960: push down BKL\n pktcdvd: push BKL down into driver\n paride: push ioctl down into driver\n block: use get_unaligned_* helpers\n block: extend queue_flag bitops\n block: request_module(): use format string\n Add bvec_merge_data to handle stacked devices and -\u003emerge_bvec()\n block: integrity flags can\u0027t use bit ops on unsigned short\n cmdfilter: extend default read filter\n sg: fix odd style (extra parenthesis) introduced by cmd filter patch\n block: add bounce support to blk_rq_map_user_iov\n cfq-iosched: get rid of enable_idle being unused warning\n allow userspace to modify scsi command filter on per device basis\n ...\n" }, { "commit": "7a1fc53c5adb910751a9b212af90302eb4ffb527", "tree": "9006ecc4323f2a55e6e16f2e19e41ec3c4c3a1ea", "parents": [ "9bbbca3a0ee09293108b67835c6bdf6196d7bcb3" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Thu Jul 10 04:54:57 2008 -0700" }, "committer": { "name": "Neil Brown", "email": "neilb@suse.de", "time": "Thu Jul 10 15:25:18 2008 +1000" }, "message": "md: ensure all blocks are uptodate or locked when syncing\n\nRemove the dubious attempt to prefer \u0027compute\u0027 over \u0027read\u0027. Not only is it\nwrong given commit c337869d (md: do not compute parity unless it is on a failed\ndrive), but it can trigger a BUG_ON in handle_parity_checks5().\n\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "cc371e66e340f35eed8dc4651c7c18e754c7fb26", "tree": "5a2d6727eb07a05999c531a90da43ab1c36b713e", "parents": [ "b24498d477a14680fc3bb3ad884fa9fa76a2d237" ], "author": { "name": "Alasdair G Kergon", "email": "agk@redhat.com", "time": "Thu Jul 03 09:53:43 2008 +0200" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Thu Jul 03 13:21:15 2008 +0200" }, "message": "Add bvec_merge_data to handle stacked devices and -\u003emerge_bvec()\n\nWhen devices are stacked, one device\u0027s merge_bvec_fn may need to perform\nthe mapping and then call one or more functions for its underlying devices.\n\nThe following bio fields are used:\n bio-\u003ebi_sector\n bio-\u003ebi_bdev\n bio-\u003ebi_size\n bio-\u003ebi_rw using bio_data_dir()\n\nThis patch creates a new struct bvec_merge_data holding a copy of those\nfields to avoid having to change them directly in the struct bio when\ngoing down the stack only to have to change them back again on the way\nback up. (And then when the bio gets mapped for real, the whole\nexercise gets repeated, but that\u0027s a problem for another day...)\n\nSigned-off-by: Alasdair G Kergon \u003cagk@redhat.com\u003e\nCc: Neil Brown \u003cneilb@suse.de\u003e\nCc: Milan Broz \u003cmbroz@redhat.com\u003e\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "b5470dc5fc18a8ff6517c3bb538d1479e58ecb02", "tree": "37b0eb3a4691bdbe58dc5c6c73b2dc8d3925b332", "parents": [ "1fe797e67fb07d605b82300934d0de67068a0aca" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jun 27 21:44:04 2008 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Jun 30 17:18:19 2008 -0700" }, "message": "md: resolve external metadata handling deadlock in md_allow_write\n\nmd_allow_write() marks the metadata dirty while holding mddev-\u003elock and then\nwaits for the write to complete. For externally managed metadata this causes a\ndeadlock as userspace needs to take the lock to communicate that the metadata\nupdate has completed.\n\nChange md_allow_write() in the \u0027external\u0027 case to start the \u0027mark active\u0027\noperation and then return -EAGAIN. The expected side effects while waiting for\nuserspace to write \u0027active\u0027 to \u0027array_state\u0027 are holding off reshape (code\ncurrently handles -ENOMEM), cause some \u0027stripe_cache_size\u0027 change requests to\nfail, cause some GET_BITMAP_FILE ioctl requests to fall back to GFP_NOIO, and\ncause updates to \u0027raid_disks\u0027 to fail. Except for \u0027stripe_cache_size\u0027 changes\nthese failures can be mitigated by coordinating with mdmon.\n\nmd_write_start() still prevents writes from occurring until the metadata\nhandler has had a chance to take action as it unconditionally waits for\nMD_CHANGE_CLEAN to be cleared.\n\n[neilb@suse.de: return -EAGAIN, try GFP_NOIO]\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\n" }, { "commit": "1fe797e67fb07d605b82300934d0de67068a0aca", "tree": "b34186b86f446bfaafec0fda3f1f42d2f3a47613", "parents": [ "7b3a871ed995270268a481404454ceafe1a87478" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 09:16:30 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 09:16:30 2008 +1000" }, "message": "md: rationalize raid5 function names\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nCommit a4456856 refactored some of the deep code paths in raid5.c into separate\nfunctions. The names chosen at the time do not consistently indicate what is\ngoing to happen to the stripe. So, update the names, and since a stripe is a\ncache element use cache semantics like fill, dirty, and clean.\n\n(also, fix up the indentation in fetch_block5)\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "7b3a871ed995270268a481404454ceafe1a87478", "tree": "1d86a84bd885f6c24213dd84b5406a38597b457e", "parents": [ "d8ee0728b5b30d7a6f62c399a95e953616d31f23" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:32:09 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:32:09 2008 +1000" }, "message": "md: handle operation chaining in raid5_run_ops\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nNeil said:\n\u003e At the end of ops_run_compute5 you have:\n\u003e /* ack now if postxor is not set to be run */\n\u003e if (tx \u0026\u0026 !test_bit(STRIPE_OP_POSTXOR, \u0026s-\u003eops_run))\n\u003e async_tx_ack(tx);\n\u003e\n\u003e It looks odd having that test there. Would it fit in raid5_run_ops\n\u003e better?\n\nThe intended global interpretation is that raid5_run_ops can build a chain\nof xor and memcpy operations. When MD registers the compute-xor it tells\nasync_tx to keep the operation handle around so that another item in the\ndependency chain can be submitted. If we are just computing a block to\nsatisfy a read then we can terminate the chain immediately. raid5_run_ops\ngives a better context for this test since it cares about the entire chain.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "d8ee0728b5b30d7a6f62c399a95e953616d31f23", "tree": "7f9fcf2b0f6222b2a2b8ee44d69af1fd6990064d", "parents": [ "600aa10993012ff2dd5617720dac081e4f992017" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:32:06 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:32:06 2008 +1000" }, "message": "md: replace R5_WantPrexor with R5_WantDrain, add \u0027prexor\u0027 reconstruct_states\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nCurrently ops_run_biodrain and other locations have extra logic to determine\nwhich blocks are processed in the prexor and non-prexor cases. This can be\neliminated if handle_write_operations5 flags the blocks to be processed in all\ncases via R5_Wantdrain. The presence of the prexor operation is tracked in\nsh-\u003ereconstruct_state.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "600aa10993012ff2dd5617720dac081e4f992017", "tree": "ead3e928aaa45e68bc65ea9aedc513c5b9c3d2d6", "parents": [ "976ea8d475675da6e86bd434328814ccbf5ae641" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:32:05 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:32:05 2008 +1000" }, "message": "md: replace STRIPE_OP_{BIODRAIN,PREXOR,POSTXOR} with \u0027reconstruct_states\u0027\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nTrack the state of reconstruct operations (recalculating the parity block\nusually due to incoming writes, or as part of array expansion) Reduces the\nscope of the STRIPE_OP_{BIODRAIN,PREXOR,POSTXOR} flags to only tracking whether\na reconstruct operation has been requested via the ops_request field of struct\nstripe_head_state.\n\nThis is the final step in the removal of ops.{pending,ack,complete,count}, i.e.\nthe STRIPE_OP_{BIODRAIN,PREXOR,POSTXOR} flags only request an operation and do\nnot track the state of the operation.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "976ea8d475675da6e86bd434328814ccbf5ae641", "tree": "087b3c8a44b6cfce0ff085faab0bc66871c455fb", "parents": [ "83de75cc92be599850e5ef3928e07cd840833499" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:32:03 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:32:03 2008 +1000" }, "message": "md: replace STRIPE_OP_COMPUTE_BLK with STRIPE_COMPUTE_RUN\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nTrack the state of compute operations (recalculating a block from all the other\nblocks in a stripe) with a state flag. Reduces the scope of the\nSTRIPE_OP_COMPUTE_BLK flag to only tracking whether a compute operation has\nbeen requested via the ops_request field of struct stripe_head_state.\n\nNote, the compute operation that is performed in the course of doing a \u0027repair\u0027\noperation (check the parity block, recalculate it and write it back if the\ncheck result is not zero) is tracked separately with the \u0027check_state\u0027\nvariable. Compute operations are held off while a \u0027check\u0027 is in progress, and\nmoving this check out to handle_issuing_new_read_requests5 the helper routine\n__handle_issuing_new_read_requests5 can be simplified.\n\nThis is another step towards the removal of ops.{pending,ack,complete,count},\ni.e. STRIPE_OP_COMPUTE_BLK only requests an operation and does not track the\nstate of the operation.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "83de75cc92be599850e5ef3928e07cd840833499", "tree": "19995f66767debc27bf207ad1ec73280c31ca8fc", "parents": [ "ecc65c9b3f9b9d740a5deade3d85b39be56401b6" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:31:58 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:58 2008 +1000" }, "message": "md: replace STRIPE_OP_BIOFILL with STRIPE_BIOFILL_RUN\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nTrack the state of read operations (copying data from the stripe cache to bio\nbuffers outside the lock) with a state flag. Reduce the scope of the\nSTRIPE_OP_BIOFILL flag to only tracking whether a biofill operation has been\nrequested via the ops_request field of struct stripe_head_state.\n\nThis is another step towards the removal of ops.{pending,ack,complete,count},\ni.e. STRIPE_OP_BIOFILL only requests an operation and does not track the state\nof the operation.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "ecc65c9b3f9b9d740a5deade3d85b39be56401b6", "tree": "e6b4e827befc6849716689f573c89aa0a41e5d26", "parents": [ "f0e43bcdebf709d747a3effb210aff1941e819ab" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:31:57 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:57 2008 +1000" }, "message": "md: replace STRIPE_OP_CHECK with \u0027check_states\u0027\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nThe STRIPE_OP_* flags record the state of stripe operations which are\nperformed outside the stripe lock. Their use in indicating which\noperations need to be run is straightforward; however, interpolating what\nthe next state of the stripe should be based on a given combination of\nthese flags is not straightforward, and has led to bugs. An easier to read\nimplementation with minimal degrees of freedom is needed.\n\nTowards this goal, this patch introduces explicit states to replace what was\npreviously interpolated from the STRIPE_OP_* flags. For now this only converts\nthe handle_parity_checks5 path, removing a user of the\nops.{pending,ack,complete,count} fields of struct stripe_operations.\n\nThis conversion also found a remaining issue with the current code. There is\na small window for a drive to fail between when we schedule a repair and when\nthe parity calculation for that repair completes. When this happens we will\nwriteback to \u0027failed_num\u0027 when we really want to write back to \u0027pd_idx\u0027.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "f0e43bcdebf709d747a3effb210aff1941e819ab", "tree": "3d95952224a1e31a49f9e635470aad3cdf6a707a", "parents": [ "c4e5ac0a22e664eecf29249553cf16c2433f5f25" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:31:55 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:55 2008 +1000" }, "message": "md: unify raid5/6 i/o submission\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nLet the raid6 path call ops_run_io to get pending i/o submitted.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "c4e5ac0a22e664eecf29249553cf16c2433f5f25", "tree": "898a3b7f4b80d54d15d5b415397b1f4dd588e116", "parents": [ "2b7497f0e0a0b9cf21d822e427d5399b2056501a" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:31:53 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:53 2008 +1000" }, "message": "md: use stripe_head_state in ops_run_io()\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nIn handle_stripe after taking sh-\u003elock we sample some bits into \u0027s\u0027 (struct\nstripe_head_state):\n\n\ts.syncing \u003d test_bit(STRIPE_SYNCING, \u0026sh-\u003estate);\n\ts.expanding \u003d test_bit(STRIPE_EXPAND_SOURCE, \u0026sh-\u003estate);\n\ts.expanded \u003d test_bit(STRIPE_EXPAND_READY, \u0026sh-\u003estate);\n\nUse these values from \u0027s\u0027 in ops_run_io() rather than re-sampling the bits.\nThis ensures a consistent snapshot (as seen under sh-\u003elock) is used.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "2b7497f0e0a0b9cf21d822e427d5399b2056501a", "tree": "8cf4d8e056ddafe48d49af0d8afe600868d2d21b", "parents": [ "b203886edbcaac3ca427cf4dbcb50b18bdb346fd" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:31:52 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:52 2008 +1000" }, "message": "md: kill STRIPE_OP_IO flag\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nThe R5_Want{Read,Write} flags already gate i/o. So, this flag is\nsuperfluous and we can unconditionally call ops_run_io().\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "b203886edbcaac3ca427cf4dbcb50b18bdb346fd", "tree": "d96cf939fd3a7ab454f61110b91b3a928eb5fe7c", "parents": [ "0cd17fec983b6bca505eecee1af33138687220b6" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Sat Jun 28 08:31:50 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:50 2008 +1000" }, "message": "md: kill STRIPE_OP_MOD_DMA in raid5 offload\n\nFrom: Dan Williams \u003cdan.j.williams@intel.com\u003e\n\nThis micro-optimization allowed the raid code to skip a re-read of the\nparity block after checking parity. It took advantage of the fact that\nxor-offload-engines have their own internal result buffer and can check\nparity without writing to memory. Remove it for the following reasons:\n\n1/ It is a layering violation for MD to need to manage the DMA and\n non-DMA paths within async_xor_zero_sum\n2/ Bad precedent to toggle the \u0027ops\u0027 flags outside the lock\n3/ Hard to realize a performance gain as reads will not need an updated\n parity block and writes will dirty it anyways.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "199050ea1ff2270174ee525b73bc4c3323098897", "tree": "b2a7851511b2ddb943b9ff7ce7c8a96c3ac2a8bb", "parents": [ "6c2fce2ef6b4821c21b5c42c7207cb9cf8c87eda" ], "author": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:33 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:33 2008 +1000" }, "message": "rationalise return value for -\u003ehot_add_disk method.\n\nFor all array types but linear, -\u003ehot_add_disk returns 1 on\nsuccess, 0 on failure.\nFor linear, it returns 0 on success and -errno on failure.\n\nThis doesn\u0027t cause a functional problem because the -\u003ehot_add_disk\nfunction of linear is used quite differently to the others.\nHowever it is confusing.\n\nSo convert all to return 0 for success or -errno on failure\nand fix call sites to match.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "6c2fce2ef6b4821c21b5c42c7207cb9cf8c87eda", "tree": "726b16f46c039df387f7cdfe0d195821d8955532", "parents": [ "8ed0a5216a0238f53b482ec88ce4aeed4b9f0da1" ], "author": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:31 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:31 2008 +1000" }, "message": "Support adding a spare to a live md array with external metadata.\n\ni.e. extend the \u0027md/dev-XXX/slot\u0027 attribute so that you can\ntell a device to fill an vacant slot in an and md array.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "0e13fe23a00ad88c737d91d94a050707c6139ce4", "tree": "7b52d698f200751cd332948a6ad982d8e839ee2d", "parents": [ "13864515f7bf6cabd60e63c62e09d311386ae1f1" ], "author": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:20 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:20 2008 +1000" }, "message": "use bio_endio instead of a call to bi_end_io\n\nTurn calls to bi-\u003ebi_end_io() into bio_endio(). Apparently bio_endio does\nexactly the same error processing as is hardcoded at these places.\n\nbio_endio() avoids recursion (or will soon), so it should be used.\n\nSigned-off-by: Mikulas Patocka \u003cmpatocka@redhat.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "efe311431869b40d67911820a309f9a1a41306f3", "tree": "3496fb7eca85cd629c19cd23afb5341007e9fa21", "parents": [ "8c2e870a625bd336b2e7a65a97c1836acef07322" ], "author": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:14 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:31:14 2008 +1000" }, "message": "Don\u0027t acknowlege that stripe-expand is complete until it really is.\n\nWe shouldn\u0027t acknowledge that a stripe has been expanded (When\nreshaping a raid5 by adding a device) until the moved data has\nactually been written out. However we are currently\nacknowledging (by calling md_done_sync) when the POST_XOR\nis complete and before the write.\n\nSo track in s.locked whether there are pending writes, and don\u0027t\ncall md_done_sync yet if there are.\n\nNote: we all set R5_LOCKED on devices which are are about to\nread from. This probably isn\u0027t technically necessary, but is\nusually done when writing a block, and justifies the use of\ns.locked here.\n\nThis bug can lead to a crash if an array is stopped while an reshape\nis in progress.\n\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "8c2e870a625bd336b2e7a65a97c1836acef07322", "tree": "1d0650ce9affed0bfae754e17bd5a86a563f8037", "parents": [ "543cf4cb3fe6f6cae3651ba918b9c56200b257d0" ], "author": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:30:52 2008 +1000" }, "committer": { "name": "Neil Brown", "email": "neilb@notabene.brown", "time": "Sat Jun 28 08:30:52 2008 +1000" }, "message": "Ensure interrupted recovery completed properly (v1 metadata plus bitmap)\n\nIf, while assembling an array, we find a device which is not fully\nin-sync with the array, it is important to set the \"fullsync\" flags.\nThis is an exact analog to the setting of this flag in hot_add_disk\nmethods.\n\nCurrently, only v1.x metadata supports having devices in an array\nwhich are not fully in-sync (it keep track of how in sync they are).\nThe \u0027fullsync\u0027 flag only makes a difference when a write-intent bitmap\nis being used. In this case it tells recovery to ignore the bitmap\nand recovery all blocks.\n\nThis fix is already in place for raid1, but not raid5/6 or raid10.\n\nSo without this fix, a raid1 ir raid4/5/6 array with version 1.x\nmetadata and a write intent bitmaps, that is stopped in the middle\nof a recovery, will appear to complete the recovery instantly\nafter it is reassembled, but the recovery will not be correct.\n\nIf you might have an array like that, issueing\n echo repair \u003e /sys/block/mdXX/md/sync_action\n\nwill make sure recovery completes properly.\n\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\n" }, { "commit": "c337869d95011495fa181536786e74aa2d7ff031", "tree": "786cdf119ce33641438cd841deb0c480dd92f900", "parents": [ "a6d8113a986c66aeb379a26b6e0062488b3e59e1" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Thu Jun 05 22:45:54 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Fri Jun 06 11:29:08 2008 -0700" }, "message": "md: do not compute parity unless it is on a failed drive\n\nIf a block is computed (rather than read) then a check/repair operation\nmay be lead to believe that the data on disk is correct, when infact it\nisn\u0027t. So only compute blocks for failed devices.\n\nThis issue has been around since at least 2.6.12, but has become harder to\nhit in recent kernels since most reads bypass the cache.\n\necho repair \u003e /sys/block/mdN/md/sync_action will set the parity blocks to the\ncorrect state.\n\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e0a115e5aa554b93150a8dc1c3fe15467708abb2", "tree": "c8d869cba362f3728c528d696e3985f1c30b0a7b", "parents": [ "b2c8daddcbe03a22402ecf943bb88302601c6835" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Thu Jun 05 22:45:52 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Fri Jun 06 11:29:08 2008 -0700" }, "message": "md: fix prexor vs sync_request race\n\nDuring the initial array synchronization process there is a window between\nwhen a prexor operation is scheduled to a specific stripe and when it\ncompletes for a sync_request to be scheduled to the same stripe. When\nthis happens the prexor completes and the stripe is unconditionally marked\n\"insync\", effectively canceling the sync_request for the stripe. Prior to\n2.6.23 this was not a problem because the prexor operation was done under\nsh-\u003elock. The effect in older kernels being that the prexor would still\nerroneously mark the stripe \"insync\", but sync_request would be held off\nand re-mark the stripe as \"!in_sync\".\n\nChange the write completion logic to not mark the stripe \"in_sync\" if a\nprexor was performed. The effect of the change is to sometimes not set\nSTRIPE_INSYNC. The worst this can do is cause the resync to stall waiting\nfor STRIPE_INSYNC to be set. If this were happening, then STRIPE_SYNCING\nwould be set and handle_issuing_new_read_requests would cause all\navailable blocks to eventually be read, at which point prexor would never\nbe used on that stripe any more and STRIPE_INSYNC would eventually be set.\n\necho repair \u003e /sys/block/mdN/md/sync_action will correct arrays that may\nhave lost this race.\n\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "dfc7064500061677720fa26352963c772d3ebe6b", "tree": "a8ca495bccf98837c6762ffba54a8009c9772259", "parents": [ "90b08710e41a07d4ff0fb8940dcce3a552991a56" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Fri May 23 13:04:39 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Sat May 24 09:56:10 2008 -0700" }, "message": "md: restart recovery cleanly after device failure.\n\nWhen we get any IO error during a recovery (rebuilding a spare), we abort\nthe recovery and restart it.\n\nFor RAID6 (and multi-drive RAID1) it may not be best to restart at the\nbeginning: when multiple failures can be tolerated, the recovery may be\nable to continue and re-doing all that has already been done doesn\u0027t make\nsense.\n\nWe already have the infrastructure to record where a recovery is up to\nand restart from there, but it is not being used properly.\nThis is because:\n - We sometimes abort with MD_RECOVERY_ERR rather than just MD_RECOVERY_INTR,\n which causes the recovery not be be checkpointed.\n - We remove spares and then re-added them which loses important state\n information.\n\nThe distinction between MD_RECOVERY_ERR and MD_RECOVERY_INTR really isn\u0027t\nneeded. If there is an error, the relevant drive will be marked as\nFaulty, and that is enough to ensure correct handling of the error. So we\nfirst remove MD_RECOVERY_ERR, changing some of the uses of it to\nMD_RECOVERY_INTR.\n\nThen we cause the attempt to remove a non-faulty device from an array to\nfail (unless recovery is impossible as the array is too degraded). Then\nwhen remove_and_add_spares attempts to remove the devices on which\nrecovery can continue, it will fail, they will remain in place, and\nrecovery will continue on them as desired.\n\nIssue: If we are halfway through rebuilding a spare and another drive\nfails, and a new spare is immediately available, do we want to:\n 1/ complete the current rebuild, then go back and rebuild the new spare or\n 2/ restart the rebuild from the start and rebuild both devices in\n parallel.\n\nBoth options can be argued for. The code currently takes option 2 as\n a/ this requires least code change\n b/ this results in a minimally-degraded array in minimal time.\n\nCc: \"Eivind Sarto\" \u003civan@kasenna.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6be9d4940134b36f9ed020aead36f831f19b49f1", "tree": "af74a03897cc8ee2ce4ae68a82a18f1d86d34f31", "parents": [ "6bcfd601861cce45ca73ac1d714f1286b6b3f0d4" ], "author": { "name": "Bernd Schubert", "email": "bernd-schubert@gmx.de", "time": "Fri May 23 13:04:34 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Sat May 24 09:56:10 2008 -0700" }, "message": "md: md: raid5 rate limit error printk\n\nLast night we had scsi problems and a hardware raid unit was offlined\nduring heavy i/o. While this happened we got for about 3 minutes a huge\nnumber messages like these\n\nApr 12 03:36:07 pfs1n14 kernel: [197510.696595] raid5:md7: read error not correctable (sector 2993096568 on sdj2).\n\nI guess the high error rate is responsible for not scheduling other events\n- during this time the system was not pingable and in the end also other\ndevices run into scsi command timeouts causing problems on these unrelated\ndevices as well.\n\nSigned-off-by: Bernd Schubert \u003cbernd-schubert@gmx.de\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e7e72bf641b1fc7b9df6f40bd2c36dfccd8d647c", "tree": "81b1db5434c9635bf23fb40415056e10390cd692", "parents": [ "4920916f728fe3c51f54c25ab7b3d271254aab5a" ], "author": { "name": "Neil Brown", "email": "neilb@suse.de", "time": "Wed May 14 16:05:54 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Wed May 14 19:11:15 2008 -0700" }, "message": "Remove blkdev warning triggered by using md\n\nAs setting and clearing queue flags now requires that we hold a spinlock\non the queue, and as blk_queue_stack_limits is called without that lock,\nget the lock inside blk_queue_stack_limits.\n\nFor blk_queue_stack_limits to be able to find the right lock, each md\npersonality needs to set q-\u003equeue_lock to point to the appropriate lock.\nThose personalities which didn\u0027t previously use a spin_lock, us\nq-\u003e__queue_lock. So always initialise that lock when allocated.\n\nWith this in place, setting/clearing of the QUEUE_FLAG_PLUGGED bit will no\nlonger cause warnings as it will be clear that the proper lock is held.\n\nThanks to Dan Williams for review and fixing the silly bugs.\n\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\nCc: Dan Williams \u003cdan.j.williams@intel.com\u003e\nCc: Jens Axboe \u003cjens.axboe@oracle.com\u003e\nCc: Alistair John Strachan \u003calistair@devzero.co.uk\u003e\nCc: Nick Piggin \u003cnpiggin@suse.de\u003e\nCc: \"Rafael J. Wysocki\" \u003crjw@sisk.pl\u003e\nCc: Jacek Luczak \u003cdifrost.kernel@gmail.com\u003e\nCc: Prakash Punnoor \u003cprakash@punnoor.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "c8894419acf5e56851de9741c5047bebd78acd1f", "tree": "6294b87ebf31d577cd48880a201ff4b1a6a50ffc", "parents": [ "3f275ea3086054205795972b8e87f2046fd3de98" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon May 12 14:02:12 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Tue May 13 08:02:24 2008 -0700" }, "message": "md: fix raid5 \u0027repair\u0027 operations\n\ncommit bd2ab67030e9116f1e4aae1289220255412b37fd \"md: close a livelock window\nin handle_parity_checks5\" introduced a bug in handling \u0027repair\u0027 operations.\nAfter a repair operation completes we clear the state bits tracking this\noperation. However, they are cleared too early and this results in the code\ndeciding to re-run the parity check operation. Since we have done the repair\nin memory the second check does not find a mismatch and thus does not do a\nwriteback.\n\nTest results:\n$ echo repair \u003e /sys/block/md0/md/sync_action\n$ cat /sys/block/md0/md/mismatch_cnt\n51072\n$ echo repair \u003e /sys/block/md0/md/sync_action\n$ cat /sys/block/md0/md/mismatch_cnt\n0\n\n(also fix incorrect indentation)\n\nCc: \u003cstable@kernel.org\u003e\nTested-by: George Spelvin \u003clinux@horizon.com\u003e\nAcked-by: NeilBrown \u003cneilb@suse.de\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6bfe0b499082fd3950429017cd8ebf2a6c458aa5", "tree": "81476cf7f7ddbea135bdb93729e0bffae0e7c163", "parents": [ "11e2ede0228ee0f81ccacd15894908c3bf241f73" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Wed Apr 30 00:52:32 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Wed Apr 30 08:29:33 2008 -0700" }, "message": "md: support blocking writes to an array on device failure\n\nAllows a userspace metadata handler to take action upon detecting a device\nfailure.\n\nBased on an original patch by Neil Brown.\n\nChanges:\n-added blocked_wait waitqueue to rdev\n-don\u0027t qualify Blocked with Faulty always let userspace block writes\n-added md_wait_for_blocked_rdev to wait for the block device to be clear, if\n userspace misses the notification another one is sent every 5 seconds\n-set MD_RECOVERY_NEEDED after clearing \"blocked\"\n-kill DoBlock flag, just test mddev-\u003eexternal\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "d7a420c9472a95c46600a0345434b7b166e0b9c7", "tree": "b95005ce8ab92efdb5cd04b3d85c3b01a0455ebc", "parents": [ "4ef197d87ad7d4bb326de3e9b8ecbb26f9e86253" ], "author": { "name": "Nick Andrew", "email": "nick@nick-andrew.net", "time": "Mon Apr 28 02:15:55 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Apr 28 08:58:42 2008 -0700" }, "message": "raid: remove leading TAB on printk messages\n\nMD drivers use one printk() call to print 2 log messages and the second line\nmay be prefixed by a TAB character. It may also output a trailing space\nbefore newline. klogd (I think) turns the TAB character into the 2 characters\n\u0027^I\u0027 when logging to a file. This looks ugly.\n\nInstead of a leading TAB to indicate continuation, prefix both output lines\nwith \u0027raid:\u0027 or similar. Also remove any trailing space in the vicinity of\nthe affected code and consistently end the sentences with a period.\n\nSigned-off-by: Nick Andrew \u003cnick@nick-andrew.net\u003e\nCc: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4ef197d87ad7d4bb326de3e9b8ecbb26f9e86253", "tree": "f6bc4fde8c1955d9b8d6669fcc08e15e3cd6fa7e", "parents": [ "8b3e6cdc53b7f29f7026955d6cb6902a49322a15" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Apr 28 02:15:54 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Apr 28 08:58:42 2008 -0700" }, "message": "md: raid5.c convert simple_strtoul to strict_strtoul\n\nstrict_strtoul handles the open-coded sanity checks in\nraid5_store_stripe_cache_size and raid5_store_preread_threshold\n\nAcked-by: NeilBrown \u003cneilb@suse.de\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "8b3e6cdc53b7f29f7026955d6cb6902a49322a15", "tree": "69d9f22a526e687fe2892d327caa3fa6ddd65cf6", "parents": [ "e46b272b6608783ed7aa7b0594871550ce20b849" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Apr 28 02:15:53 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Apr 28 08:58:42 2008 -0700" }, "message": "md: introduce get_priority_stripe() to improve raid456 write performance\n\nImprove write performance by preventing the delayed_list from dumping all its\nstripes onto the handle_list in one shot. Delayed stripes are now further\ndelayed by being held on the \u0027hold_list\u0027. The \u0027hold_list\u0027 is bypassed when:\n\n * a STRIPE_IO_STARTED stripe is found at the head of \u0027handle_list\u0027\n * \u0027handle_list\u0027 is empty and i/o is being done to satisfy full stripe-width\n write requests\n * \u0027bypass_count\u0027 is less than \u0027bypass_threshold\u0027. By default the threshold\n is 1, i.e. every other stripe handled is a preread stripe provided the\n top two conditions are false.\n\nBenchmark data:\nSystem: 2x Xeon 5150, 4x SATA, mem\u003d1GB\nBaseline: 2.6.24-rc7\nConfiguration: mdadm --create /dev/md0 /dev/sd[b-e] -n 4 -l 5 --assume-clean\nTest1: dd if\u003d/dev/zero of\u003d/dev/md0 bs\u003d1024k count\u003d2048\n * patched: +33% (stripe_cache_size \u003d 256), +25% (stripe_cache_size \u003d 512)\n\nTest2: tiobench --size 2048 --numruns 5 --block 4096 --block 131072 (XFS)\n * patched: +13%\n * patched + preread_bypass_threshold \u003d 0: +37%\n\nChanges since v1:\n* reduce bypass_threshold from (chunk_size / sectors_per_chunk) to (1) and\n make it configurable. This defaults to fairness and modest performance\n gains out of the box.\nChanges since v2:\n* [neilb@suse.de]: kill STRIPE_PRIO_HI and preread_needed as they are not\n necessary, the important change was clearing STRIPE_DELAYED in\n add_stripe_bio and this has been moved out to make_request for the hang\n fix.\n* [neilb@suse.de]: simplify get_priority_stripe\n* [dan.j.williams@intel.com]: reset the bypass_count when -\u003ehold_list is\n sampled empty (+11%)\n* [dan.j.williams@intel.com]: decrement the bypass_count at the detection\n of stripes being naturally promoted off of hold_list +2%. Note, resetting\n bypass_count instead of decrementing on these events yields +4% but that is\n probably too aggressive.\nChanges since v3:\n* cosmetic fixups\n\nTested-by: James W. Laferriere \u003cbabydr@baby-dragons.com\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e46b272b6608783ed7aa7b0594871550ce20b849", "tree": "a65834f60bee7e77f9669a7ff7e95a1ec3cd74ee", "parents": [ "9a7b2b0f36c40beeb252cb5ceff36bb295e88d97" ], "author": { "name": "Harvey Harrison", "email": "harvey.harrison@gmail.com", "time": "Mon Apr 28 02:15:50 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Apr 28 08:58:42 2008 -0700" }, "message": "md: replace remaining __FUNCTION__ occurrences\n\n__FUNCTION__ is gcc-specific, use __func__\n\nSigned-off-by: Harvey Harrison \u003charvey.harrison@gmail.com\u003e\nCc: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "bd2ab67030e9116f1e4aae1289220255412b37fd", "tree": "da52cf8952a660f17ce45c8c510ded9a3487180a", "parents": [ "231bc2a222411f43bfb0fbb6d64c0f34c7b1039f" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Thu Apr 10 21:29:27 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Fri Apr 11 08:06:44 2008 -0700" }, "message": "md: close a livelock window in handle_parity_checks5\n\nIf a failure is detected after a parity check operation has been initiated,\nbut before it completes handle_parity_checks5 will never quiesce operations on\nthe stripe.\n\nExplicitly handle this case by \"canceling\" the parity check, i.e. clear the\nSTRIPE_OP_CHECK flags and queue the stripe on the handle list again to refresh\nany non-uptodate blocks.\n\nKernel versions \u003e\u003d 2.6.23 are susceptible.\n\nCc: \u003cstable@kernel.org\u003e\nCc: NeilBrown \u003cneilb@suse.de\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "9ea85ebae1e05100cdb4807db4f265b0ede7aad8", "tree": "c12e7a0a7339291f0dcf3deff6dcabf40c00e29f", "parents": [ "8195096bf2e3756264561d7a47d547bafd4e9100" ], "author": { "name": "Andrew Morton", "email": "akpm@linux-foundation.org", "time": "Wed Mar 19 17:01:05 2008 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Wed Mar 19 18:53:37 2008 -0700" }, "message": "drivers/md/raid5.c: fix printk warnings\n\ngcc-3.4.5 on sparc64:\n\ndrivers/md/raid5.c: In function `raid5_end_read_request\u0027:\ndrivers/md/raid5.c:1147: warning: long long unsigned int format, long unsigned int arg (arg 4)\ndrivers/md/raid5.c:1164: warning: long long unsigned int format, long unsigned int arg (arg 3)\ndrivers/md/raid5.c:1170: warning: long long unsigned int format, long unsigned int arg (arg 3)\n\nsector_t is u64, and we don\u0027t know what type the architecture uses to\nimplement u64 (on some it is unsigned long).\n\nCc: Neil Brown \u003cneilb@suse.de\u003e\nCc: \"J. Bruce Fields\" \u003cbfields@fieldses.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6ed3003c19a96fe18edf8179c4be6fe14abbebbc", "tree": "dba37b52eb4c68fedc9e842a69e2c0c3b4a5aa31", "parents": [ "73c34431c7119d0bc7d3436abfad75fe47b2c51f" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Feb 06 01:40:00 2008 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed Feb 06 10:41:19 2008 -0800" }, "message": "md: fix an occasional deadlock in raid5\n\nraid5\u0027s \u0027make_request\u0027 function calls generic_make_request on underlying\ndevices and if we run out of stripe heads, it could end up waiting for one of\nthose requests to complete. This is bad as recursive calls to\ngeneric_make_request go on a queue and are not even attempted until\nmake_request completes.\n\nSo: don\u0027t make any generic_make_request calls in raid5 make_request until all\nwaiting has been done. We do this by simply setting STRIPE_HANDLE instead of\ncalling handle_stripe().\n\nIf we need more stripe_heads, raid5d will get called to process the pending\nstripe_heads which will call generic_make_request from a\n\nThis change by itself causes a performance hit. So add a change so that\nraid5_activate_delayed is only called at unplug time, never in raid5. This\nseems to bring back the performance numbers. Calling it in raid5d was\nsometimes too soon...\n\nNeil said:\n\n How about we queue it for 2.6.25-rc1 and then about when -rc2 comes out,\n we queue it for 2.6.24.y?\n\nAcked-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nTested-by: dean gaudet \u003cdean@arctic.org\u003e\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "d089c6af10c2be5988f03667d6d22fe6085fbe5e", "tree": "815eb0596bd7c908f30c268da6b5c5eff0e0bc21", "parents": [ "29ac4aa3fc68a86279aca50f20df4d614de2e204" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Feb 06 01:39:59 2008 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed Feb 06 10:41:19 2008 -0800" }, "message": "md: change ITERATE_RDEV to rdev_for_each\n\nAs this is more in line with common practice in the kernel. Also swap the\nargs around to be more like list_for_each.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "c620727779f7cc8ea96efb71f0651a26349e59c1", "tree": "777abdad9c9ef10cb4df5c0efc736e6c64851ed8", "parents": [ "c303da6d713b87b7b3f999f5acce8ecc76ff1adb" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Feb 06 01:39:52 2008 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed Feb 06 10:41:18 2008 -0800" }, "message": "md: allow a maximum extent to be set for resyncing\n\nThis allows userspace to control resync/reshape progress and synchronise it\nwith other activities, such as shared access in a SAN, or backing up critical\nsections during a tricky reshape.\n\nWriting a number of sectors (which must be a multiple of the chunk size if\nsuch is meaningful) causes a resync to pause when it gets to that point.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b47490c9bc73d0b34e4c194db40de183e592e446", "tree": "8334581aff7aa0b199df04c6f9d707c1902ad14f", "parents": [ "66c811e99322767bad5db4368de93aac604f02a2" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Feb 06 01:39:50 2008 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed Feb 06 10:41:18 2008 -0800" }, "message": "md: Update md bitmap during resync.\n\nCurrently an md array with a write-intent bitmap does not updated that bitmap\nto reflect successful partial resync. Rather the entire bitmap is updated\nwhen the resync completes.\n\nThis is because there is no guarentee that resync requests will complete in\norder, and tracking each request individually is unnecessarily burdensome.\n\nHowever there is value in regularly updating the bitmap, so add code to\nperiodically pause while all pending sync requests complete, then update the\nbitmap. Doing this only every few seconds (the same as the bitmap update\ntime) does not notciably affect resync performance.\n\n[snitzer@gmail.com: export bitmap_cond_end_sync]\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nCc: \"Mike Snitzer\" \u003csnitzer@gmail.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "0f94e87cdeaaac9f0f9a28a5dd2a5070b87cd3e8", "tree": "6c35efe23f565d27bec046d8eed3c5901e1d0090", "parents": [ "5b7741b3325d0d94c88b2ad46657a536890aaa2f" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 08 15:32:53 2008 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Jan 08 16:10:35 2008 -0800" }, "message": "md: fix data corruption when a degraded raid5 array is reshaped\n\nWe currently do not wait for the block from the missing device to be\ncomputed from parity before copying data to the new stripe layout.\n\nThe change in the raid6 code is not techincally needed as we don\u0027t delay\ndata block recovery in the same way for raid6 yet. But making the change\nnow is safer long-term.\n\nThis bug exists in 2.6.23 and 2.6.24-rc\n\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6c55be8b962f1bdc592d579e81fc27b11ea53dfc", "tree": "3413cc2c8e26afdbc498cbdbd96bc2fa9733b2a6", "parents": [ "5b23dbe8173c212d6a326e35347b038705603d39" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Wed Nov 14 16:59:35 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed Nov 14 18:45:39 2007 -0800" }, "message": "raid5: fix unending write sequence\n\n\u003cdebug output from Joel\u0027s system\u003e\nhandling stripe 7629696, state\u003d0x14 cnt\u003d1, pd_idx\u003d2 ops\u003d0:0:0\ncheck 5: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800ffcffcc0 written 0000000000000000\ncheck 4: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800fdd4e360 written 0000000000000000\ncheck 3: state 0x1 toread 0000000000000000 read 0000000000000000 write 0000000000000000 written 0000000000000000\ncheck 2: state 0x1 toread 0000000000000000 read 0000000000000000 write 0000000000000000 written 0000000000000000\ncheck 1: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800ff517e40 written 0000000000000000\ncheck 0: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800fd4cae60 written 0000000000000000\nlocked\u003d4 uptodate\u003d2 to_read\u003d0 to_write\u003d4 failed\u003d0 failed_num\u003d0\nfor sector 7629696, rmw\u003d0 rcw\u003d0\n\u003c/debug\u003e\n\nThese blocks were prepared to be written out, but were never handled in\nops_run_biodrain(), so they remain locked forever. The operations flags\nare all clear which means handle_stripe() thinks nothing else needs to be\ndone.\n\nThis state suggests that the STRIPE_OP_PREXOR bit was sampled \u0027set\u0027 when it\nshould not have been. This patch cleans up cases where the code looks at\nsh-\u003eops.pending when it should be looking at the consistent stack-based\nsnapshot of the operations flags.\n\nReport from Joel:\n\tResync done. Patch fix this bug.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nTested-by: Joel Bertrand \u003cjoel.bertrand@systella.fr\u003e\nCc: \u003cstable@kernel.org\u003e\nCc: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "2ad8b1ef11c98c5603580878aebf9f1bc74129e4", "tree": "f7bdc2484513f6ffd174b1385bb216dcf97d2c78", "parents": [ "d85532ed284e63b5c56eaf2418f262822af60be4" ], "author": { "name": "Alan D. Brunelle", "email": "Alan.Brunelle@hp.com", "time": "Wed Nov 07 14:26:56 2007 -0500" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Fri Nov 09 13:41:32 2007 +0100" }, "message": "Add UNPLUG traces to all appropriate places\n\nAdded blk_unplug interface, allowing all invocations of unplugs to result\nin a generated blktrace UNPLUG.\n\nSigned-off-by: Alan D. Brunelle \u003cAlan.Brunelle@hp.com\u003e\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "def6ae26a9e69c3e6d0f0054524c76fd32420ecd", "tree": "3688a3d5d8507835e0f55fb3e3d83b52cfa049a7", "parents": [ "139b82984af5a98e4b03fd01616d79fc4970128a" ], "author": { "name": "Neil Brown", "email": "neilb@suse.de", "time": "Mon Nov 05 14:51:02 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Mon Nov 05 15:12:32 2007 -0800" }, "message": "md: fix misapplied patch in raid5.c\n\ncommit 4ae3f847e49e3787eca91bced31f8fd328d50496 (\"md: raid5: fix\nclearing of biofill operations\") did not get applied correctly,\npresumably due to substantial similarities between handle_stripe5 and\nhandle_stripe6.\n\nThis patch moves the chunk of new code from handle_stripe6 (where it isn\u0027t\nneeded (yet)) to handle_stripe5.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nCc: \"Dan Williams\" \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "4ae3f847e49e3787eca91bced31f8fd328d50496", "tree": "9fab60c4898e49cf1cf04e360dc63ae1d0ef5bd2", "parents": [ "85bfb4da8cad483a4e550ec89060d05a4daf895b" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Oct 22 20:45:11 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Oct 23 08:32:06 2007 -0700" }, "message": "md: raid5: fix clearing of biofill operations\n\nops_complete_biofill() runs outside of spin_lock(\u0026sh-\u003elock) and clears the\n\u0027pending\u0027 and \u0027ack\u0027 bits. Since the test_and_ack_op() macro only checks\nagainst \u0027complete\u0027 it can get an inconsistent snapshot of pending work.\n\nMove the clearing of these bits to handle_stripe5(), under the lock.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nTested-by: Joel Bertrand \u003cjoel.bertrand@systella.fr\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nCc: Stable \u003cstable@kernel.org\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "fd5d806266935179deda1502101624832eacd01f", "tree": "b0d01c2ed7d754f2a6daef6c77d108a81d8a470e", "parents": [ "bf2de6f5a4faf0197268f18d08969b003b87b6e8" ], "author": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Tue Oct 16 11:05:02 2007 +0200" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Tue Oct 16 11:05:02 2007 +0200" }, "message": "block: convert blkdev_issue_flush() to use empty barriers\n\nThen we can get rid of -\u003eissue_flush_fn() and all the driver private\nimplementations of that.\n\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "6712ecf8f648118c3363c142196418f89a510b90", "tree": "347d39a7d5a7ed96d3b1afecd28de2a0f98b98c9", "parents": [ "5bb23a688b2de23d7765a1dd439d89c038378978" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Thu Sep 27 12:47:43 2007 +0200" }, "committer": { "name": "Jens Axboe", "email": "axboe@carl.home.kernel.dk", "time": "Wed Oct 10 09:25:57 2007 +0200" }, "message": "Drop \u0027size\u0027 argument from bio_endio and bi_end_io\n\nAs bi_end_io is only called once when the reqeust is complete,\nthe \u0027size\u0027 argument is now redundant. Remove it.\n\nNow there is no need for bio_endio to subtract the size completed\nfrom bi_size. So don\u0027t do that either.\n\nWhile we are at it, change bi_end_io to return void.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "e4d84909dd48b5e5806a5d18b881e1ca1610ba9b", "tree": "85aa712ebc034a31b9a2a5db2f0e7ce4951cb440", "parents": [ "6247cdc2cd334dad0ea5428245a7d8f4b075f21e" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Sep 24 10:06:13 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Sep 24 13:23:35 2007 -0700" }, "message": "raid5: fix 2 bugs in ops_complete_biofill\n\n1/ ops_complete_biofill tried to avoid calling handle_stripe since all the\nstate necessary to return read completions is available. However the\nprocess of determining whether more read requests are pending requires\nlocking the stripe (to block add_stripe_bio from updating dev-\u003etoead).\nops_complete_biofill can run in tasklet context, so rather than upgrading\nall the stripe locks from spin_lock to spin_lock_bh this patch just\nunconditionally reschedules handle_stripe after completing the read\nrequest.\n\n2/ ops_complete_biofill needlessly qualified processing R5_Wantfill with\ndev-\u003etoread. The result being that the \u0027biofill\u0027 pending bit is cleared\nbefore handling the pending read-completions on dev-\u003eread. R5_Wantfill can\nbe unconditionally handled because the \u0027biofill\u0027 pending bit prevents new\nR5_Wantfill requests from being seen by ops_run_biofill and\nops_complete_biofill.\n\nFound-by: Yuri Tikhonov \u003cyur@emcraft.com\u003e\n[neilb@suse.de: simpler fix for bug 1 than moving code]\nSigned-off-by: NeilBrown \u003cneilb@suse.de\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\n" }, { "commit": "a2e0855182e2be26b252745b2bb7558705cb0dd2", "tree": "d7a1457b1785cff5c018b37587ce2dbef159e83d", "parents": [ "b70ae1d9f69ba52767af89f90fd79587669bc7ff" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Tue Sep 11 15:23:36 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Sep 11 17:21:19 2007 -0700" }, "message": "md: fix some bugs with growing raid5/raid6 arrays.\n\nThe recent changed to raid5 to allow offload of parity calculation etc\nintroduced some bugs in the code for growing (i.e. adding a disk to) raid5\nand raid6. This fixes them\n\nAcked-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "165125e1e480f9510a5ffcfbfee4e3ee38c05f23", "tree": "8009c8a5ff09e26dc2418d42f66ecafb055c52a2", "parents": [ "f695baf2df9e0413d3521661070103711545207a" ], "author": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Tue Jul 24 09:28:11 2007 +0200" }, "committer": { "name": "Jens Axboe", "email": "jens.axboe@oracle.com", "time": "Tue Jul 24 09:28:11 2007 +0200" }, "message": "[BLOCK] Get rid of request_queue_t typedef\n\nSome of the code has been gradually transitioned to using the proper\nstruct request_queue, but there\u0027s lots left. So do a full sweet of\nthe kernel and get rid of this typedef and replace its uses with\nthe proper type.\n\nSigned-off-by: Jens Axboe \u003cjens.axboe@oracle.com\u003e\n" }, { "commit": "eb0645a8b1f14da300f40bb9f424640cd1181fbf", "tree": "462789626fcd1775bec80d74d19bcd68797589c8", "parents": [ "7c6129c68fe90a61166800b40217a850b8faee98" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 20 00:31:46 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Fri Jul 20 08:44:19 2007 -0700" }, "message": "async_tx: fix kmap_atomic usage in async_memcpy\n\nAndrew Morton:\n\t[async_memcpy] is very wrong if both ASYNC_TX_KMAP_DST and\n\tASYNC_TX_KMAP_SRC can ever be set. We\u0027ll end up using the same kmap\n\tslot for both src add dest and we get either corrupted data or a BUG.\n\nEvgeniy Polyakov:\n\tBtw, shouldn\u0027t it always be kmap_atomic() even if flag is not set.\n\tThat pages are usual one returned by alloc_page().\n\nSo fix the usage of kmap_atomic and kill the ASYNC_TX_KMAP_DST and\nASYNC_TX_KMAP_SRC flags.\n\nCc: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nCc: Evgeniy Polyakov \u003cjohnpol@2ka.mipt.ru\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "20c2df83d25c6a95affe6157a4c9cac4cf5ffaac", "tree": "415c4453d2b17a50abe7a3e515177e1fa337bd67", "parents": [ "64fb98fc40738ae1a98bcea9ca3145b89fb71524" ], "author": { "name": "Paul Mundt", "email": "lethal@linux-sh.org", "time": "Fri Jul 20 10:11:58 2007 +0900" }, "committer": { "name": "Paul Mundt", "email": "lethal@linux-sh.org", "time": "Fri Jul 20 10:11:58 2007 +0900" }, "message": "mm: Remove slab destructors from kmem_cache_create().\n\nSlab destructors were no longer supported after Christoph\u0027s\nc59def9f222d44bb7e2f0a559f2906191a0862d7 change. They\u0027ve been\nBUGs for both slab and slub, and slob never supported them\neither.\n\nThis rips out support for the dtor pointer from kmem_cache_create()\ncompletely and fixes up every single callsite in the kernel (there were\nabout 224, not including the slab allocator definitions themselves,\nor the documentation references).\n\nSigned-off-by: Paul Mundt \u003clethal@linux-sh.org\u003e\n" }, { "commit": "f6dff381af01006ffae3c23cd2e07e30584de0ec", "tree": "257502c7a1a1392ffa91687d10331b08e21d2012", "parents": [ "830ea01673a397798d1281d2022615559f5001bb" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:31 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:18 2007 -0700" }, "message": "md: remove raid5 compute_block and compute_parity5\n\nreplaced by raid5_run_ops\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "830ea01673a397798d1281d2022615559f5001bb", "tree": "33413032374605648a1e47f059b93ffc39e33717", "parents": [ "f0a50d3754c7f1b7f05f45b1c0b35d20445316b5" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:31 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:17 2007 -0700" }, "message": "md: handle_stripe5 - request io processing in raid5_run_ops\n\nI/O submission requests were already handled outside of the stripe lock in\nhandle_stripe. Now that handle_stripe is only tasked with finding work,\nthis logic belongs in raid5_run_ops.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "f0a50d3754c7f1b7f05f45b1c0b35d20445316b5", "tree": "c54b572ad4c4c9b48f887a9ecc28ec7b6166d552", "parents": [ "b5e98d65d34a1c11a2135ea8a9b2619dbc7216c8" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:31 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:17 2007 -0700" }, "message": "md: handle_stripe5 - add request/completion logic for async expand ops\n\nWhen a stripe is being expanded bulk copying takes place to move the data\nfrom the old stripe to the new. Since raid5_run_ops only operates on one\nstripe at a time these bulk copies are handled in-line under the stripe\nlock. In the dma offload case we poll for the completion of the operation.\n\nAfter the data has been copied into the new stripe the parity needs to be\nrecalculated across the new disks. We reuse the existing postxor\nfunctionality to carry out this calculation. By setting STRIPE_OP_POSTXOR\nwithout setting STRIPE_OP_BIODRAIN the completion path in handle stripe\ncan differentiate expand operations from normal write operations.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "b5e98d65d34a1c11a2135ea8a9b2619dbc7216c8", "tree": "bf88aad375bb0cbf6c346ec912c06c2607850b21", "parents": [ "e89f89629b5de76e504d1be75c82c4a6b2419583" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:31 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:17 2007 -0700" }, "message": "md: handle_stripe5 - add request/completion logic for async read ops\n\nWhen a read bio is attached to the stripe and the corresponding block is\nmarked R5_UPTODATE, then a read (biofill) operation is scheduled to copy\nthe data from the stripe cache to the bio buffer. handle_stripe flags the\nblocks to be operated on with the R5_Wantfill flag. If new read requests\narrive while raid5_run_ops is running they will not be handled until\nhandle_stripe is scheduled to run again.\n\nChangelog:\n* cleanup to_read and to_fill accounting\n* do not fail reads that have reached the cache\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "e89f89629b5de76e504d1be75c82c4a6b2419583", "tree": "96f54c576f51207e4f4a75f4770e9cf2707fc948", "parents": [ "f38e12199a94ca458e4f03c5a2c984fb80adadc5" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:31 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:17 2007 -0700" }, "message": "md: handle_stripe5 - add request/completion logic for async check ops\n\nCheck operations are scheduled when the array is being resynced or an\nexplicit \u0027check/repair\u0027 command was sent to the array. Previously check\noperations would destroy the parity block in the cache such that even if\nparity turned out to be correct the parity block would be marked\n!R5_UPTODATE at the completion of the check. When the operation can be\ncarried out by a dma engine the assumption is that it can check parity as a\nread-only operation. If raid5_run_ops notices that the check was handled\nby hardware it will preserve the R5_UPTODATE status of the parity disk.\n\nWhen a check operation determines that the parity needs to be repaired we\nreuse the existing compute block infrastructure to carry out the operation.\nRepair operations imply an immediate write back of the data, so to\ndifferentiate a repair from a normal compute operation the\nSTRIPE_OP_MOD_REPAIR_PD flag is added.\n\nChangelog:\n* remove test_and_set/test_and_clear BUG_ONs, Neil Brown\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "f38e12199a94ca458e4f03c5a2c984fb80adadc5", "tree": "706f54c46d4a4c839dd43c1403854dde860c6be5", "parents": [ "e33129d84130459dbb764a1a52a4bfceab3da978" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:30 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:17 2007 -0700" }, "message": "md: handle_stripe5 - add request/completion logic for async compute ops\n\nhandle_stripe will compute a block when a backing disk has failed, or when\nit determines it can save a disk read by computing the block from all the\nother up-to-date blocks.\n\nPreviously a block would be computed under the lock and subsequent logic in\nhandle_stripe could use the newly up-to-date block. With the raid5_run_ops\nimplementation the compute operation is carried out a later time outside\nthe lock. To preserve the old functionality we take advantage of the\ndependency chain feature of async_tx to flag the block as R5_Wantcompute\nand then let other parts of handle_stripe operate on the block as if it\nwere up-to-date. raid5_run_ops guarantees that the block will be ready\nbefore it is used in another operation.\n\nHowever, this only works in cases where the compute and the dependent\noperation are scheduled at the same time. If a previous call to\nhandle_stripe sets the R5_Wantcompute flag there is no facility to pass the\nasync_tx dependency chain across successive calls to raid5_run_ops. The\nreq_compute variable protects against this case.\n\nChangelog:\n* remove the req_compute BUG_ON\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "e33129d84130459dbb764a1a52a4bfceab3da978", "tree": "c3f2742dac468a1c62e14ec1f2ec0cb5a37ee966", "parents": [ "d84e0f10d38393f617227f0c831a99c69294651f" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:30 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:16 2007 -0700" }, "message": "md: handle_stripe5 - add request/completion logic for async write ops\n\nAfter handle_stripe5 decides whether it wants to perform a\nread-modify-write, or a reconstruct write it calls\nhandle_write_operations5. A read-modify-write operation will perform an\nxor subtraction of the blocks marked with the R5_Wantprexor flag, copy the\nnew data into the stripe (biodrain) and perform a postxor operation across\nall up-to-date blocks to generate the new parity. A reconstruct write is run\nwhen all blocks are already up-to-date in the cache so all that is needed\nis a biodrain and postxor.\n\nOn the completion path STRIPE_OP_PREXOR will be set if the operation was a\nread-modify-write. The STRIPE_OP_BIODRAIN flag is used in the completion\npath to differentiate write-initiated postxor operations versus\nexpansion-initiated postxor operations. Completion of a write triggers i/o\nto the drives.\n\nChangelog:\n* make the \u0027rcw\u0027 parameter to handle_write_operations5 a simple flag, Neil Brown\n* remove test_and_set/test_and_clear BUG_ONs, Neil Brown\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "d84e0f10d38393f617227f0c831a99c69294651f", "tree": "5d0836f024f1f13ac6f1b2b2d3004244cb0fa649", "parents": [ "91c00924846a0034020451c280c76baa4299f9dc" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:30 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:16 2007 -0700" }, "message": "md: common infrastructure for running operations with raid5_run_ops\n\nAll the handle_stripe operations that are to be transitioned to use\nraid5_run_ops need a method to coherently gather work under the stripe-lock\nand hand that work off to raid5_run_ops. The \u0027get_stripe_work\u0027 routine\nruns under the lock to read all the bits in sh-\u003eops.pending that do not\nhave the corresponding bit set in sh-\u003eops.ack. This modified \u0027pending\u0027\nbitmap is then passed to raid5_run_ops for processing.\n\nThe transition from \u0027ack\u0027 to \u0027completion\u0027 does not need similar protection\nas the existing release_stripe infrastructure will guarantee that\nhandle_stripe will run again after a completion bit is set, and\nhandle_stripe can tolerate a sh-\u003eops.completed bit being set while the lock\nis held.\n\nA call to async_tx_issue_pending_all() is added to raid5d to kick the\noffload engines once all pending stripe operations work has been submitted.\nThis enables batching of the submission and completion of operations.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "91c00924846a0034020451c280c76baa4299f9dc", "tree": "7124ed6706937b793a10c37a861c5fc0f2e5b348", "parents": [ "45b4233caac05da0118b608a9fc2a40a9fc580cd" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 13:52:30 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:15 2007 -0700" }, "message": "md: raid5_run_ops - run stripe operations outside sh-\u003elock\n\nWhen the raid acceleration work was proposed, Neil laid out the following\nattack plan:\n\n1/ move the xor and copy operations outside spin_lock(\u0026sh-\u003elock)\n2/ find/implement an asynchronous offload api\n\nThe raid5_run_ops routine uses the asynchronous offload api (async_tx) and\nthe stripe_operations member of a stripe_head to carry out xor+copy\noperations asynchronously, outside the lock.\n\nTo perform operations outside the lock a new set of state flags is needed\nto track new requests, in-flight requests, and completed requests. In this\nnew model handle_stripe is tasked with scanning the stripe_head for work,\nupdating the stripe_operations structure, and finally dropping the lock and\ncalling raid5_run_ops for processing. The following flags outline the\nrequests that handle_stripe can make of raid5_run_ops:\n\nSTRIPE_OP_BIOFILL\n - copy data into request buffers to satisfy a read request\nSTRIPE_OP_COMPUTE_BLK\n - generate a missing block in the cache from the other blocks\nSTRIPE_OP_PREXOR\n - subtract existing data as part of the read-modify-write process\nSTRIPE_OP_BIODRAIN\n - copy data out of request buffers to satisfy a write request\nSTRIPE_OP_POSTXOR\n - recalculate parity for new data that has entered the cache\nSTRIPE_OP_CHECK\n - verify that the parity is correct\nSTRIPE_OP_IO\n - submit i/o to the member disks (note this was already performed outside\n the stripe lock, but it made sense to add it as an operation type\n\nThe flow is:\n1/ handle_stripe sets STRIPE_OP_* in sh-\u003eops.pending\n2/ raid5_run_ops reads sh-\u003eops.pending, sets sh-\u003eops.ack, and submits the\n operation to the async_tx api\n3/ async_tx triggers the completion callback routine to set\n sh-\u003eops.complete and release the stripe\n4/ handle_stripe runs again to finish the operation and optionally submit\n new operations that were previously blocked\n\nNote this patch just defines raid5_run_ops, subsequent commits (one per\nmajor operation type) modify handle_stripe to take advantage of this\nroutine.\n\nChangelog:\n* removed ops_complete_biodrain in favor of ops_complete_postxor and\n ops_complete_write.\n* removed the raid5_run_ops workqueue\n* call bi_end_io for reads in ops_complete_biofill, saves a call to\n handle_stripe\n* explicitly handle the 2-disk raid5 case (xor becomes memcpy), Neil Brown\n* fix race between async engines and bi_end_io call for reads, Neil Brown\n* remove unnecessary spin_lock from ops_complete_biofill\n* remove test_and_set/test_and_clear BUG_ONs, Neil Brown\n* remove explicit interrupt handling for channel switching, this feature\n was absorbed (i.e. it is now implicit) by the async_tx api\n* use return_io in ops_complete_biofill\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "45b4233caac05da0118b608a9fc2a40a9fc580cd", "tree": "ebc65304c088b6f2162c5ba4e04382bd49736040", "parents": [ "a445685647e825c713175d180ffc8dd54d90589b" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Jul 09 11:56:43 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:15 2007 -0700" }, "message": "raid5: replace custom debug PRINTKs with standard pr_debug\n\nReplaces PRINTK with pr_debug, and kills the RAID5_DEBUG definition in\nfavor of the global DEBUG definition. To get local debug messages just add\n\u0027#define DEBUG\u0027 to the top of the file.\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "a445685647e825c713175d180ffc8dd54d90589b", "tree": "d2db5674e51d33162e1e5993b6e6680ec534e2df", "parents": [ "9bc89cd82d6f88fb0ca39b30445c329a430fd66b" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Jul 09 11:56:43 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:15 2007 -0700" }, "message": "raid5: refactor handle_stripe5 and handle_stripe6 (v3)\n\nhandle_stripe5 and handle_stripe6 have very deep logic paths handling the\nvarious states of a stripe_head. By introducing the \u0027stripe_head_state\u0027\nand \u0027r6_state\u0027 objects, large portions of the logic can be moved to\nsub-routines.\n\n\u0027struct stripe_head_state\u0027 consumes all of the automatic variables that previously\nstood alone in handle_stripe5,6. \u0027struct r6_state\u0027 contains the handle_stripe6\nspecific variables like p_failed and q_failed.\n\nOne of the nice side effects of the \u0027stripe_head_state\u0027 change is that it\nallows for further reductions in code duplication between raid5 and raid6.\nThe following new routines are shared between raid5 and raid6:\n\n\thandle_completed_write_requests\n\thandle_requests_to_failed_array\n\thandle_stripe_expansion\n\nChanges:\n* v2: fixed \u0027conf-\u003eraid_disk-1\u0027 for the raid6 \u0027handle_stripe_expansion\u0027 path\n* v3: removed the unused \u0027dirty\u0027 field from struct stripe_head_state\n* v3: coalesced open coded bi_end_io routines into return_io()\n\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "9bc89cd82d6f88fb0ca39b30445c329a430fd66b", "tree": "7bd0e856abd359f84edea1bacfd1dd32edd93fbb", "parents": [ "685784aaf3cd0e3ff5e36c7ecf6f441cdbf57f73" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Tue Jan 02 11:10:44 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:14 2007 -0700" }, "message": "async_tx: add the async_tx api\n\nThe async_tx api provides methods for describing a chain of asynchronous\nbulk memory transfers/transforms with support for inter-transactional\ndependencies. It is implemented as a dmaengine client that smooths over\nthe details of different hardware offload engine implementations. Code\nthat is written to the api can optimize for asynchronous operation and the\napi will fit the chain of operations to the available offload resources. \n \n\tI imagine that any piece of ADMA hardware would register with the\n\t\u0027async_*\u0027 subsystem, and a call to async_X would be routed as\n\tappropriate, or be run in-line. - Neil Brown\n\nasync_tx exploits the capabilities of struct dma_async_tx_descriptor to\nprovide an api of the following general format:\n\nstruct dma_async_tx_descriptor *\nasync_\u003coperation\u003e(..., struct dma_async_tx_descriptor *depend_tx,\n\t\t\tdma_async_tx_callback cb_fn, void *cb_param)\n{\n\tstruct dma_chan *chan \u003d async_tx_find_channel(depend_tx, \u003coperation\u003e);\n\tstruct dma_device *device \u003d chan ? chan-\u003edevice : NULL;\n\tint int_en \u003d cb_fn ? 1 : 0;\n\tstruct dma_async_tx_descriptor *tx \u003d device ?\n\t\tdevice-\u003edevice_prep_dma_\u003coperation\u003e(chan, len, int_en) : NULL;\n\n\tif (tx) { /* run \u003coperation\u003e asynchronously */\n\t\t...\n\t\ttx-\u003etx_set_dest(addr, tx, index);\n\t\t...\n\t\ttx-\u003etx_set_src(addr, tx, index);\n\t\t...\n\t\tasync_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);\n\t} else { /* run \u003coperation\u003e synchronously */\n\t\t...\n\t\t\u003coperation\u003e\n\t\t...\n\t\tasync_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);\n\t}\n\n\treturn tx;\n}\n\nasync_tx_find_channel() returns a capable channel from its pool. The\nchannel pool is organized as a per-cpu array of channel pointers. The\nasync_tx_rebalance() routine is tasked with managing these arrays. In the\nuniprocessor case async_tx_rebalance() tries to spread responsibility\nevenly over channels of similar capabilities. For example if there are two\ncopy+xor channels, one will handle copy operations and the other will\nhandle xor. In the SMP case async_tx_rebalance() attempts to spread the\noperations evenly over the cpus, e.g. cpu0 gets copy channel0 and xor\nchannel0 while cpu1 gets copy channel 1 and xor channel 1. When a\ndependency is specified async_tx_find_channel defaults to keeping the\noperation on the same channel. A xor-\u003ecopy-\u003exor chain will stay on one\nchannel if it supports both operation types, otherwise the transaction will\ntransition between a copy and a xor resource.\n\nCurrently the raid5 implementation in the MD raid456 driver has been\nconverted to the async_tx api. A driver for the offload engines on the\nIntel Xscale series of I/O processors, iop-adma, is provided in a later\ncommit. With the iop-adma driver and async_tx, raid456 is able to offload\ncopy, xor, and xor-zero-sum operations to hardware engines.\n \nOn iop342 tiobench showed higher throughput for sequential writes (20 - 30%\nimprovement) and sequential reads to a degraded array (40 - 55%\nimprovement). For the other cases performance was roughly equal, +/- a few\npercentage points. On a x86-smp platform the performance of the async_tx\nimplementation (in synchronous mode) was also +/- a few percentage points\nof the original implementation. According to \u0027top\u0027 on iop342 CPU\nutilization drops from ~50% to ~15% during a \u0027resync\u0027 while the speed\naccording to /proc/mdstat doubles from ~25 MB/s to ~50 MB/s.\n \nThe tiobench command line used for testing was: tiobench --size 2048\n--block 4096 --block 131072 --dir /mnt/raid --numruns 5\n* iop342 had 1GB of memory available\n\nDetails:\n* if CONFIG_DMA_ENGINE\u003dn the asynchronous path is compiled away by making\n async_tx_find_channel a static inline routine that always returns NULL\n* when a callback is specified for a given transaction an interrupt will\n fire at operation completion time and the callback will occur in a\n tasklet. if the the channel does not support interrupts then a live\n polling wait will be performed\n* the api is written as a dmaengine client that requests all available\n channels\n* In support of dependencies the api implicitly schedules channel-switch\n interrupts. The interrupt triggers the cleanup tasklet which causes\n pending operations to be scheduled on the next channel\n* Xor engines treat an xor destination address differently than a software\n xor routine. To the software routine the destination address is an implied\n source, whereas engines treat it as a write-only destination. This patch\n modifies the xor_blocks routine to take a an explicit destination address\n to mirror the hardware.\n\nChangelog:\n* fixed a leftover debug print\n* don\u0027t allow callbacks in async_interrupt_cond\n* fixed xor_block changes\n* fixed usage of ASYNC_TX_XOR_DROP_DEST\n* drop dma mapping methods, suggested by Chris Leech\n* printk warning fixups from Andrew Morton\n* don\u0027t use inline in C files, Adrian Bunk\n* select the API when MD is enabled\n* BUG_ON xor source counts \u003c\u003d 1\n* implicitly handle hardware concerns like channel switching and\n interrupts, Neil Brown\n* remove the per operation type list, and distribute operation capabilities\n evenly amongst the available channels\n* simplify async_tx_find_channel to optimize the fast path\n* introduce the channel_table_initialized flag to prevent early calls to\n the api\n* reorganize the code to mimic crypto\n* include mm.h as not all archs include it in dma-mapping.h\n* make the Kconfig options non-user visible, Adrian Bunk\n* move async_tx under crypto since it is meant as \u0027core\u0027 functionality, and\n the two may share algorithms in the future\n* move large inline functions into c files\n* checkpatch.pl fixes\n* gpl v2 only correction\n\nCc: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\nAcked-By: NeilBrown \u003cneilb@suse.de\u003e\n" }, { "commit": "685784aaf3cd0e3ff5e36c7ecf6f441cdbf57f73", "tree": "10f99829f7d877b87614fe69be77e363c026a8d7", "parents": [ "d379b01e9087a582d58f4b678208a4f8d8376fe7" ], "author": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Mon Jul 09 11:56:42 2007 -0700" }, "committer": { "name": "Dan Williams", "email": "dan.j.williams@intel.com", "time": "Fri Jul 13 08:06:14 2007 -0700" }, "message": "xor: make \u0027xor_blocks\u0027 a library routine for use with async_tx\n\nThe async_tx api tries to use a dma engine for an operation, but will fall\nback to an optimized software routine otherwise. Xor support is\nimplemented using the raid5 xor routines. For organizational purposes this\nroutine is moved to a common area.\n\nThe following fixes are also made:\n* rename xor_block \u003d\u003e xor_blocks, suggested by Adrian Bunk\n* ensure that xor.o initializes before md.o in the built-in case\n* checkpatch.pl fixes\n* mark calibrate_xor_blocks __init, Adrian Bunk\n\nCc: Adrian Bunk \u003cbunk@stusta.de\u003e\nCc: NeilBrown \u003cneilb@suse.de\u003e\nCc: Herbert Xu \u003cherbert@gondor.apana.org.au\u003e\nSigned-off-by: Dan Williams \u003cdan.j.williams@intel.com\u003e\n" }, { "commit": "44ce6294d07555c3d313757105fd44b78208407f", "tree": "29659bee16a4d7111124c6f41289bc9c11669e70", "parents": [ "497f050c42e46a4b1f6a9bcd8827fa5d97fe1feb" ], "author": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed May 09 18:51:36 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed May 09 18:51:36 2007 -0700" }, "message": "Revert \"md: improve partition detection in md array\"\n\nThis reverts commit 5b479c91da90eef605f851508744bfe8269591a0.\n\nQuoth Neil Brown:\n\n \"It causes an oops when auto-detecting raid arrays, and it doesn\u0027t\n seem easy to fix.\n\n The array may not be \u0027open\u0027 when do_md_run is called, so\n bdev-\u003ebd_disk might be NULL, so bd_set_size can oops.\n\n This whole approach of opening an md device before it has been\n assembled just seems to get more and more painful. I think I\u0027m going\n to have to come up with something clever to provide both backward\n comparability with usage expectation, and sane integration into the\n rest of the kernel.\"\n\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "5b479c91da90eef605f851508744bfe8269591a0", "tree": "d7c2844926d96d794d66b974b300eba2d047465e", "parents": [ "08a02ecd28bad35a47357e611044dcbeab06e3d7" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed May 09 02:35:39 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed May 09 12:30:57 2007 -0700" }, "message": "md: improve partition detection in md array\n\nmd currently uses -\u003emedia_changed to make sure rescan_partitions\nis call on md array after they are assembled.\n\nHowever that doesn\u0027t happen until the array is opened, which is later\nthan some people would like.\n\nSo use blkdev_ioctl to do the rescan immediately that the\narray has been assembled.\n\nThis means we can remove all the -\u003echange infrastructure as it was only used\nto trigger a partition rescan.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "42b9bebe3fea3d3ce381bc6735a3fb50e6613f06", "tree": "e99aec5ccb154e2f09a89a9b16f9e174b40b2863", "parents": [ "4d167f09375bd6c18447d3fcc18baaf3acd15fbc" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed May 09 02:35:37 2007 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Wed May 09 12:30:57 2007 -0700" }, "message": "md: remove the slash from the name of a kmem_cache used by raid5\n\nSLUB doesn\u0027t like slashes as it wants to use the cache name as the name of a\ndirectory (or symlink) in sysfs.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nAcked-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "5e55e2f5fc95b355d8aa649f346cff69904c8ade", "tree": "6722b3f116bbeacb500ed6900219cfe54eaefa11", "parents": [ "041ae52e265fc432ea5525b1c66720385c2d11f0" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Mar 26 21:32:14 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Mar 27 09:05:15 2007 -0700" }, "message": "[PATCH] md: convert compile time warnings into runtime warnings\n\n... still not sure why we need this ....\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "041ae52e265fc432ea5525b1c66720385c2d11f0", "tree": "b7bab39615cb591c7df82d70e3f5ad5dba200e3d", "parents": [ "3d37890baa4ca962f8a6b77525b8f3d0698eee09" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Mar 26 21:32:14 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Mar 27 09:05:14 2007 -0700" }, "message": "[PATCH] md: clear the congested_fn when stopping a raid5\n\nIf this mddev and queue got reused for another array that doesn\u0027t register a\ncongested_fn, this function would get called incorretly.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "3d37890baa4ca962f8a6b77525b8f3d0698eee09", "tree": "12e58ec8d5369ba270affe1eb8406e3603b33575", "parents": [ "79f6523a16b2010969418f8df25fe61498dec66b" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Mar 26 21:32:13 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Tue Mar 27 09:05:14 2007 -0700" }, "message": "[PATCH] md: allow raid4 arrays to be reshaped\n\nAll that is missing the the function pointers in raid4_pers.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "6d3baf2eb8bd680b2d4f509bc3dbf4dcd6e27a40", "tree": "bb76dd97b16812efa78f546e08d11bd7ed6eefc7", "parents": [ "c6b36e9a3c57b73c7a6bdf787baa55f21195bba9" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Mon Mar 05 00:30:44 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Mon Mar 05 07:57:53 2007 -0800" }, "message": "[PATCH] md: fix for raid6 reshape\n\nRecent patch for raid6 reshape had a change missing that showed up in\nsubsequent review.\n\nMany places in the raid5 code used \"conf-\u003eraid_disks-1\" to mean \"number of\ndata disks\". With raid6 that had to be changed to \"conf-\u003eraid_disk -\nconf-\u003emax_degraded\" or similar. One place was missed.\n\nThis bug means that if a raid6 reshape were aborted in the middle the\nrecorded position would be wrong. On restart it would either fail (as the\nposition wasn\u0027t on an appropriate boundary) or would leave a section of the\narray unreshaped, causing data corruption.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "f416885ef4950501dd3858d1afa1137a0c2905c5", "tree": "77003957b071e5728fefedd905bf6f8c348fb4b3", "parents": [ "b4c4c7b8095298ff4ce20b40bf180ada070812d0" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Feb 28 20:11:53 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Thu Mar 01 14:53:36 2007 -0800" }, "message": "[PATCH] md: add support for reshape of a raid6\n\ni.e. one or more drives can be added and the array will re-stripe\nwhile on-line.\n\nMost of the interesting work was already done for raid5. This just extends it\nto raid6.\n\nmdadm newer than 2.6 is needed for complete safety, however any version of\nmdadm which support raid5 reshape will do a good enough job in almost all\ncases (an \u0027echo repair \u003e /sys/block/mdX/md/sync_action\u0027 is recommended after a\nreshape that was aborted and had to be restarted with an such a version of\nmdadm).\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "b4c4c7b8095298ff4ce20b40bf180ada070812d0", "tree": "2c51fe8ef220be17c791f8aca1218cadcae22b7f", "parents": [ "d1b5380c7f794da16e815c34e54ee7641db8a288" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Feb 28 20:11:48 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Thu Mar 01 14:53:36 2007 -0800" }, "message": "[PATCH] md: restart a (raid5) reshape that has been aborted due to a read/write error\n\nAn error always aborts any resync/recovery/reshape on the understanding that\nit will immediately be restarted if that still makes sense. However a reshape\ncurrently doesn\u0027t get restarted. With this patch it does.\n\nTo avoid restarting when it is not possible to do work, we call into the\npersonality to check that a reshape is ok, and strengthen raid5_check_reshape\nto fail if there are too many failed devices.\n\nWe also break some code out into a separate function: remove_and_add_spares as\nthe indent level for that code was getting crazy.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "387bb17374c5fa057462d00d4ba941d49f45de4d", "tree": "85c68bbd2f077724563cfb26a3357d6679f12104", "parents": [ "6649a3863232eb2e2f15ea6c622bd8ceacf96d76" ], "author": { "name": "Neil Brown", "email": "neilb@suse.de", "time": "Thu Feb 08 14:20:29 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Fri Feb 09 09:25:46 2007 -0800" }, "message": "[PATCH] md: fix various bugs with aligned reads in RAID5\n\nIt is possible for raid5 to be sent a bio that is too big for an underlying\ndevice. So if it is a READ that we pass stright down to a device, it will\nfail and confuse RAID5.\n\nSo in \u0027chunk_aligned_read\u0027 we check that the bio fits within the parameters\nfor the target device and if it doesn\u0027t fit, fall back on reading through\nthe stripe cache and making lots of one-page requests.\n\nNote that this is the earliest time we can check against the device because\nearlier we don\u0027t have a lock on the device, so it could change underneath\nus.\n\nAlso, the code for handling a retry through the cache when a read fails has\nnot been tested and was badly broken. This patch fixes that code.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nCc: \"Kai\" \u003cepimetreus@fastmail.fm\u003e\nCc: \u003cstable@suse.de\u003e\nCc: \u003corg@suse.de\u003e\nCc: Jens Axboe \u003cjens.axboe@oracle.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "c20086de9319ac406f1e96ad459763c9f9965b18", "tree": "281e7863e9702b16520b5cb2b109e234355254f8", "parents": [ "fe33f6f1525113ab0bae416966209438e7a31987" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Fri Jan 26 00:57:14 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Fri Jan 26 13:51:00 2007 -0800" }, "message": "[PATCH] md: remove unnecessary printk when raid5 gets an unaligned read.\n\nraid5_mergeable_bvec tries to ensure that raid5 never sees a read request\nthat does not fit within just one chunk. However as we must always accept\na single-page read, that is not always possible.\n\nSo when \"in_chunk_boundary\" fails, it might be unusual, but it is not a\nproblem and printing a message every time is a bad idea.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "2a2275d630b982e5f90206f9bc497f6695a3ec5d", "tree": "88cddee709b2107b74e5424810d4ffb6e3772382", "parents": [ "a0ad13ef643a5829d63c456ab6143bbda60b44a9" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Fri Jan 26 00:57:11 2007 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.linux-foundation.org", "time": "Fri Jan 26 13:51:00 2007 -0800" }, "message": "[PATCH] md: fix potential memalloc deadlock in md\n\nIf a GFP_KERNEL allocation is attempted in md while the mddev_lock is held,\nit is possible for a deadlock to eventuate.\n\nThis happens if the array was marked \u0027clean\u0027, and the memalloc triggers a\nwrite-out to the md device.\n\nFor the writeout to succeed, the array must be marked \u0027dirty\u0027, and that\nrequires getting the mddev_lock.\n\nSo, before attempting a GFP_KERNEL allocation while holding the lock, make\nsure the array is marked \u0027dirty\u0027 (unless it is currently read-only).\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "802ba064c49f655d20fed563f2a4924c8256ea10", "tree": "f705b597b2da6583239e9f028fa0d6f6ec319682", "parents": [ "011e3a9ad4891ba1b27e281e085228189c56b137" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Dec 13 00:34:13 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Wed Dec 13 09:05:48 2006 -0800" }, "message": "[PATCH] md: Don\u0027t assume that READ\u003d\u003d0 and WRITE\u003d\u003d1 - use the names explicitly\n\nThanks Jens for alerting me to this.\n\nCc: Jens Axboe \u003cjens.axboe@oracle.com\u003e\nCc: \u003craziebe@gmail.com\u003e\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "c2b00852fbae4f8c45c2651530ded3bd01bde814", "tree": "697a2087ffb1e5d3d81af628e818f9d943c4267f", "parents": [ "b8c6b645563d641df91fdcfd84a9c73c91d75b61" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Sun Dec 10 02:20:51 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:21 2006 -0800" }, "message": "[PATCH] md: return a non-zero error to bi_end_io as appropriate in raid5\n\nCurrently raid5 depends on clearing the BIO_UPTODATE flag to signal an error\nto higher levels. While this should be sufficient, it is safer to explicitly\nset the error code as well - less room for confusion.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "b8c6b645563d641df91fdcfd84a9c73c91d75b61", "tree": "434f057ef1523fa84d8e669661c06db569a2f030", "parents": [ "fdee8ae4498c48b44c0eac592f9c6ed24c4517c1" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Sun Dec 10 02:20:50 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:21 2006 -0800" }, "message": "[PATCH] md: remove some old ifdefed-out code from raid5.c\n\nThere are some vestiges of old code that was used for bypassing the stripe\ncache on reads in raid5.c. This was never updated after the change from\nbuffer_heads to bios, but was left as a reminder.\n\nThat functionality has nowe been implemented in a completely different way, so\nthe old code can go.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "b875e531fc82db592d6093594593d5cafde0a1cd", "tree": "4bf23fcee1a1d1675339ea7e5d671dde6dd8641f", "parents": [ "5248861511d6aae4997a5aa7152824d87587b0b6" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Sun Dec 10 02:20:49 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:21 2006 -0800" }, "message": "[PATCH] md: fix innocuous bug in raid6 stripe_to_pdidx\n\nstripe_to_pdidx finds the index of the parity disk for a given stripe. It\nassumes raid5 in that it uses \"disks-1\" to determine the number of data disks.\n\nThis is incorrect for raid6 but fortunately the two usages cancel each other\nout. The only way that \u0027data_disks\u0027 affects the calculation of pd_idx in\nraid5_compute_sector is when it is divided into the sector number. But as\nthat sector number is calculated by multiplying in the wrong value of\n\u0027data_disks\u0027 the division produces the right value.\n\nSo it is innocuous but needs to be fixed.\n\nAlso change the calculation of raid_disks in compute_blocknr to make it\nmore obviously correct (it seems at first to always use disks-1 too).\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "5248861511d6aae4997a5aa7152824d87587b0b6", "tree": "a853392a6ca8fdb10a308e46b90e4ed63e66bf04", "parents": [ "46031f9a38a9773021f1872abc713d62467ac22e" ], "author": { "name": "Raz Ben-Jehuda(caro)", "email": "raziebe@gmail.com", "time": "Sun Dec 10 02:20:48 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:20 2006 -0800" }, "message": "[PATCH] md: enable bypassing cache for reads\n\nCall the chunk_aligned_read where appropriate.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "46031f9a38a9773021f1872abc713d62467ac22e", "tree": "fe91f661fe0aad5f149447797c5d31544453ca38", "parents": [ "f679623f50545bc0577caf2d0f8675b61162f059" ], "author": { "name": "Raz Ben-Jehuda(caro)", "email": "raziebe@gmail.com", "time": "Sun Dec 10 02:20:47 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:20 2006 -0800" }, "message": "[PATCH] md: allow reads that have bypassed the cache to be retried on failure\n\nIf a bypass-the-cache read fails, we simply try again through the cache. If\nit fails again it will trigger normal recovery precedures.\n\nupdate 1:\n\nFrom: NeilBrown \u003cneilb@suse.de\u003e\n\n1/\n chunk_aligned_read and retry_aligned_read assume that\n data_disks \u003d\u003d raid_disks - 1\n which is not true for raid6.\n So when an aligned read request bypasses the cache, we can get the wrong data.\n\n2/ The cloned bio is being used-after-free in raid5_align_endio\n (to test BIO_UPTODATE).\n\n3/ We forgot to add rdev-\u003edata_offset when submitting\n a bio for aligned-read\n\n4/ clone_bio calls blk_recount_segments and then we change bi_bdev,\n so we need to invalidate the segment counts.\n\n5/ We don\u0027t de-reference the rdev when the read completes.\n This means we need to record the rdev to so it is still\n available in the end_io routine. Fortunately\n bi_next in the original bio is unused at this point so\n we can stuff it in there.\n\n6/ We leak a cloned bio if the target rdev is not usable.\n\nFrom: NeilBrown \u003cneilb@suse.de\u003e\n\nupdate 2:\n\n1/ When aligned requests fail (read error) they need to be retried\n via the normal method (stripe cache). As we cannot be sure that\n we can process a single read in one go (we may not be able to\n allocate all the stripes needed) we store a bio-being-retried\n and a list of bioes-that-still-need-to-be-retried.\n When find a bio that needs to be retried, we should add it to\n the list, not to single-bio...\n\n2/ We were never incrementing \u0027scnt\u0027 when resubmitting failed\n aligned requests.\n\n[akpm@osdl.org: build fix]\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "f679623f50545bc0577caf2d0f8675b61162f059", "tree": "7253c48db142ec63e6f22df12dfa2babb3e6129c", "parents": [ "23032a0eb97c8eaae8ac9d17373b53b19d0f5413" ], "author": { "name": "Raz Ben-Jehuda(caro)", "email": "raziebe@gmail.com", "time": "Sun Dec 10 02:20:46 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:20 2006 -0800" }, "message": "[PATCH] md: handle bypassing the read cache (assuming nothing fails)\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "23032a0eb97c8eaae8ac9d17373b53b19d0f5413", "tree": "7021309ed7a591c6ae93f7347fb23482cea85163", "parents": [ "0d4ca600fcf5c5f3a0c195ccf37e989b83451dd4" ], "author": { "name": "Raz Ben-Jehuda(caro)", "email": "raziebe@gmail.com", "time": "Sun Dec 10 02:20:45 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Sun Dec 10 09:57:20 2006 -0800" }, "message": "[PATCH] md: define raid5_mergeable_bvec\n\nThis will encourage read request to be on only one device, so we will often be\nable to bypass the cache for read requests.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nCc: Jens Axboe \u003cjens.axboe@oracle.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "e18b890bb0881bbab6f4f1a6cd20d9c60d66b003", "tree": "4828be07e1c24781c264b42c5a75bcd968223c3f", "parents": [ "441e143e95f5aa1e04026cb0aa71c801ba53982f" ], "author": { "name": "Christoph Lameter", "email": "clameter@sgi.com", "time": "Wed Dec 06 20:33:20 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@woody.osdl.org", "time": "Thu Dec 07 08:39:25 2006 -0800" }, "message": "[PATCH] slab: remove kmem_cache_t\n\nReplace all uses of kmem_cache_t with struct kmem_cache.\n\nThe patch was generated using the following script:\n\n\t#!/bin/sh\n\t#\n\t# Replace one string by another in all the kernel sources.\n\t#\n\n\tset -e\n\n\tfor file in `find * -name \"*.c\" -o -name \"*.h\"|xargs grep -l $1`; do\n\t\tquilt add $file\n\t\tsed -e \"1,\\$s/$1/$2/g\" $file \u003e/tmp/$$\n\t\tmv /tmp/$$ $file\n\t\tquilt refresh\n\tdone\n\nThe script was run like this\n\n\tsh replace kmem_cache_t \"struct kmem_cache\"\n\nSigned-off-by: Christoph Lameter \u003cclameter@sgi.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "0692c6b1cf5537b190f90fb5903f1af89a41b0a8", "tree": "4686d65a3020e41501feda573350f5e2d43629d5", "parents": [ "2f4713036114dd13d1f4fe433b7f236250b65f5a" ], "author": { "name": "NeilBrown", "email": "neilb@suse.de", "time": "Wed Nov 08 17:44:48 2006 -0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Wed Nov 08 18:29:24 2006 -0800" }, "message": "[PATCH] md: fix sizing problem with raid5-reshape and CONFIG_LBD\u003dn\n\nI forgot to has the size-in-blocks to (loff_t) before shifting up to a\nsize-in-bytes.\n\nSigned-off-by: Neil Brown \u003cneilb@suse.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" } ], "next": "52e5f9d1cf0b10b24317037dcd1c9be38ca7011c" }