)]}' { "log": [ { "commit": "03cbc358aab3faf34bfeaa02206fa660127e9b3f", "tree": "8d95237532390a1beeef37158174fd045b0f8819", "parents": [ "632dd2053a1146c826ceb6f26ab689389c05e751" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Fri Sep 29 02:01:46 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Fri Sep 29 09:18:25 2006 -0700" }, "message": "[PATCH] lockdep core: improve the lock-chain-hash\n\nWith CONFIG_DEBUG_LOCK_ALLOC turned off i was getting sporadic failures in\nthe locking self-test:\n\n ------------\u003e\n | Locking API testsuite:\n ----------------------------------------------------------------------------\n | spin |wlock |rlock |mutex | wsem | rsem |\n --------------------------------------------------------------------------\n A-A deadlock: ok | ok | ok | ok | ok | ok |\n A-B-B-A deadlock: ok | ok | ok | ok | ok | ok |\n A-B-B-C-C-A deadlock: ok | ok | ok | ok | ok | ok |\n A-B-C-A-B-C deadlock: ok | ok | ok | ok | ok | ok |\n A-B-B-C-C-D-D-A deadlock: ok |FAILED| ok | ok | ok | ok |\n A-B-C-D-B-D-D-A deadlock: ok | ok | ok | ok | ok | ok |\n A-B-C-D-B-C-D-A deadlock: ok | ok | ok | ok | ok |FAILED|\n\nafter much debugging it turned out to be caused by accidental chain-hash\nkey collisions. The current hash is:\n\n #define iterate_chain_key(key1, key2) \\\n\t(((key1) \u003c\u003c MAX_LOCKDEP_KEYS_BITS/2) ^ \\\n\t((key1) \u003e\u003e (64-MAX_LOCKDEP_KEYS_BITS/2)) ^ \\\n \t(key2))\n\nwhere MAX_LOCKDEP_KEYS_BITS is 11. This hash is pretty good as it will\nshift by 5 bits in every iteration, where every new ID \u0027mixed\u0027 into the\nhash would have up to 11 bits. But because there was a 6 bits overlap\nbetween subsequent IDs and their high bits tended to be similar, there was\na chance for accidental chain-hash collision for a low number of locks\nheld.\n\nthe solution is to shift by 11 bits:\n\n #define iterate_chain_key(key1, key2) \\\n\t(((key1) \u003c\u003c MAX_LOCKDEP_KEYS_BITS) ^ \\\n\t((key1) \u003e\u003e (64-MAX_LOCKDEP_KEYS_BITS)) ^ \\\n \t(key2))\n\nThis keeps the hash perfect up to 5 locks held, but even above that the\nhash is still good because 11 bits is a relative prime to the total 64\nbits, so a complete match will only occur after 64 held locks (which doesnt\nhappen in Linux). Even after 5 locks held, entropy of the 5 IDs mixed into\nthe hash is already good enough so that overlap doesnt generate a colliding\nhash ID.\n\nwith this change the false positives went away.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "99de055ac065e19ed69de961e97c6336a261b34e", "tree": "edc25ea5f39c09e4e273b8824f2306655d6f4726", "parents": [ "2b33b4dcbe5e09e683eef281f72aef951e17061c" ], "author": { "name": "Dave Jones", "email": "davej@redhat.com", "time": "Fri Sep 29 02:00:10 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Fri Sep 29 09:18:13 2006 -0700" }, "message": "[PATCH] lockdep: print kernel version\n\nLets do the same thing we do for oopses - print out the version in the\nreport. It\u0027s an extra line of output though. We could tack it on the end\nof the INFO: lines, but that screws up Ingo\u0027s pretty output.\n\nSigned-off-by: Dave Jones \u003cdavej@redhat.com\u003e\nCc: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: Arjan van de Ven \u003carjan@linux.intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "3fa7c794fe4dc127f7fac3fad4d13628e68f89ce", "tree": "855a6737a2275317bcd3f3a9934c33dcadba6e6f", "parents": [ "5a1b3999d6cb7ab87f1f3b1700bc91839fd6fa29" ], "author": { "name": "Andi Kleen", "email": "ak@suse.de", "time": "Tue Sep 26 10:52:34 2006 +0200" }, "committer": { "name": "Andi Kleen", "email": "andi@basil.nowhere.org", "time": "Tue Sep 26 10:52:34 2006 +0200" }, "message": "[PATCH] Avoid recursion in lockdep when stack tracer takes locks\n\nThe new dwarf2 unwinder needs to take locks to do backtraces\ninside modules. This patch makes sure lockdep which calls\nstacktrace is not reentered.\n\nThanks to Ingo for suggesting this simpler approach.\n\nCc: mingo@elte.hu\nSigned-off-by: Andi Kleen \u003cak@suse.de\u003e\n" }, { "commit": "5a1b3999d6cb7ab87f1f3b1700bc91839fd6fa29", "tree": "ec969406fe68d875d7904c364ea9a5b74c084567", "parents": [ "4ea8a5d8b57cd504b4b2de1212523848e7ab50cf" ], "author": { "name": "Andi Kleen", "email": "ak@suse.de", "time": "Tue Sep 26 10:52:34 2006 +0200" }, "committer": { "name": "Andi Kleen", "email": "andi@basil.nowhere.org", "time": "Tue Sep 26 10:52:34 2006 +0200" }, "message": "[PATCH] x86: Some preparationary cleanup for stack trace\n\n- Remove unused all_contexts parameter\nNo caller used it\n- Move skip argument into the structure (needed for\nfollowon patches)\n\nCc: mingo@elte.hu\n\nSigned-off-by: Andi Kleen \u003cak@suse.de\u003e\n" }, { "commit": "d6d897cec29252b8d0785198cfa6ca16d30c739d", "tree": "01b138b10c7afa0e88e9720be3d3616e4f0bccc1", "parents": [ "55794a412fdf9af1744800e5020a4ec6b21e3cdc" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Jul 10 04:44:04 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Mon Jul 10 13:24:14 2006 -0700" }, "message": "[PATCH] lockdep: core, reduce per-lock class-cache size\n\nlockdep_map is embedded into every lock, which blows up data structure\nsizes all around the kernel. Reduce the class-cache to be for the default\nclass only - that is used in 99.9% of the cases and even if we dont have a\nclass cached, the lookup in the class-hash is lockless.\n\nThis change reduces the per-lock dep_map overhead by 56 bytes on 64-bit\nplatforms and by 28 bytes on 32-bit platforms.\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nCc: Arjan van de Ven \u003carjan@linux.intel.com\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "55794a412fdf9af1744800e5020a4ec6b21e3cdc", "tree": "3429cb26c5dfd484487b15827c655b5d080ce0f8", "parents": [ "5fca80e8b4bf5d69b900115b14342133ce81d79e" ], "author": { "name": "Arjan van de Ven", "email": "arjan@Linux.intel.com", "time": "Mon Jul 10 04:44:03 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Mon Jul 10 13:24:14 2006 -0700" }, "message": "[PATCH] lockdep: improve debug output\n\nMake lockdep print which lock is held, in the \"kfree() of a live lock\"\nscenario.\n\nSigned-off-by: Arjan van de Ven \u003carjan@linux.intel.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "f9829cceb686f3719215fe43c8593e5f3efe1710", "tree": "eb6f721915c879646f81e423b6606d6b4376bed9", "parents": [ "8f72e4028a1ff968000cec4a034f45619fbd7ec4" ], "author": { "name": "Andi Kleen", "email": "ak@suse.de", "time": "Mon Jul 10 04:44:01 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Mon Jul 10 13:24:14 2006 -0700" }, "message": "[PATCH] Minor cleanup to lockdep.c\n\n- Use printk formatting for indentation\n- Don\u0027t leave NTFS in the default event filter\n\nSigned-off-by: Andi Kleen \u003cak@suse.de\u003e\nAcked-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "6c9076ec9cd448f43bbda871352a7067f456ee26", "tree": "8691366aa511f20146eb9cfe942e0d94c4136e1a", "parents": [ "fbb9ce9530fd9b66096d5187fa6a115d16d9746c" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Jul 03 00:24:51 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Mon Jul 03 15:27:04 2006 -0700" }, "message": "[PATCH] lockdep: allow read_lock() recursion of same class\n\nFrom: Ingo Molnar \u003cmingo@elte.hu\u003e\n\nlockdep so far only allowed read-recursion for the same lock instance.\nThis is enough in the overwhelming majority of cases, but a hostap case\ntriggered and reported by Miles Lane relies on same-class\ndifferent-instance recursion. So we relax the restriction on read-lock\nrecursion.\n\n(This change does not allow rwsem read-recursion, which is still\nforbidden.)\n\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" }, { "commit": "fbb9ce9530fd9b66096d5187fa6a115d16d9746c", "tree": "1151a55e5d56045bac17b9766e6a4696cff0a26f", "parents": [ "cae2ed9aa573415c6e5de9a09b7ff0d74af793bc" ], "author": { "name": "Ingo Molnar", "email": "mingo@elte.hu", "time": "Mon Jul 03 00:24:50 2006 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@g5.osdl.org", "time": "Mon Jul 03 15:27:03 2006 -0700" }, "message": "[PATCH] lockdep: core\n\nDo \u0027make oldconfig\u0027 and accept all the defaults for new config options -\nreboot into the kernel and if everything goes well it should boot up fine and\nyou should have /proc/lockdep and /proc/lockdep_stats files.\n\nTypically if the lock validator finds some problem it will print out\nvoluminous debug output that begins with \"BUG: ...\" and which syslog output\ncan be used by kernel developers to figure out the precise locking scenario.\n\nWhat does the lock validator do? It \"observes\" and maps all locking rules as\nthey occur dynamically (as triggered by the kernel\u0027s natural use of spinlocks,\nrwlocks, mutexes and rwsems). Whenever the lock validator subsystem detects a\nnew locking scenario, it validates this new rule against the existing set of\nrules. If this new rule is consistent with the existing set of rules then the\nnew rule is added transparently and the kernel continues as normal. If the\nnew rule could create a deadlock scenario then this condition is printed out.\n\nWhen determining validity of locking, all possible \"deadlock scenarios\" are\nconsidered: assuming arbitrary number of CPUs, arbitrary irq context and task\ncontext constellations, running arbitrary combinations of all the existing\nlocking scenarios. In a typical system this means millions of separate\nscenarios. This is why we call it a \"locking correctness\" validator - for all\nrules that are observed the lock validator proves it with mathematical\ncertainty that a deadlock could not occur (assuming that the lock validator\nimplementation itself is correct and its internal data structures are not\ncorrupted by some other kernel subsystem). [see more details and conditionals\nof this statement in include/linux/lockdep.h and\nDocumentation/lockdep-design.txt]\n\nFurthermore, this \"all possible scenarios\" property of the validator also\nenables the finding of complex, highly unlikely multi-CPU multi-context races\nvia single single-context rules, increasing the likelyhood of finding bugs\ndrastically. In practical terms: the lock validator already found a bug in\nthe upstream kernel that could only occur on systems with 3 or more CPUs, and\nwhich needed 3 very unlikely code sequences to occur at once on the 3 CPUs.\nThat bug was found and reported on a single-CPU system (!). So in essence a\nrace will be found \"piecemail-wise\", triggering all the necessary components\nfor the race, without having to reproduce the race scenario itself! In its\nshort existence the lock validator found and reported many bugs before they\nactually caused a real deadlock.\n\nTo further increase the efficiency of the validator, the mapping is not per\n\"lock instance\", but per \"lock-class\". For example, all struct inode objects\nin the kernel have inode-\u003einotify_mutex. If there are 10,000 inodes cached,\nthen there are 10,000 lock objects. But -\u003einotify_mutex is a single \"lock\ntype\", and all locking activities that occur against -\u003einotify_mutex are\n\"unified\" into this single lock-class. The advantage of the lock-class\napproach is that all historical -\u003einotify_mutex uses are mapped into a single\n(and as narrow as possible) set of locking rules - regardless of how many\ndifferent tasks or inode structures it took to build this set of rules. The\nset of rules persist during the lifetime of the kernel.\n\nTo see the rough magnitude of checking that the lock validator does, here\u0027s a\nportion of /proc/lockdep_stats, fresh after bootup:\n\n lock-classes: 694 [max: 2048]\n direct dependencies: 1598 [max: 8192]\n indirect dependencies: 17896\n all direct dependencies: 16206\n dependency chains: 1910 [max: 8192]\n in-hardirq chains: 17\n in-softirq chains: 105\n in-process chains: 1065\n stack-trace entries: 38761 [max: 131072]\n combined max dependencies: 2033928\n hardirq-safe locks: 24\n hardirq-unsafe locks: 176\n softirq-safe locks: 53\n softirq-unsafe locks: 137\n irq-safe locks: 59\n irq-unsafe locks: 176\n\nThe lock validator has observed 1598 actual single-thread locking patterns,\nand has validated all possible 2033928 distinct locking scenarios.\n\nMore details about the design of the lock validator can be found in\nDocumentation/lockdep-design.txt, which can also found at:\n\n http://redhat.com/~mingo/lockdep-patches/lockdep-design.txt\n\n[bunk@stusta.de: cleanups]\nSigned-off-by: Ingo Molnar \u003cmingo@elte.hu\u003e\nSigned-off-by: Arjan van de Ven \u003carjan@linux.intel.com\u003e\nSigned-off-by: Adrian Bunk \u003cbunk@stusta.de\u003e\nSigned-off-by: Andrew Morton \u003cakpm@osdl.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@osdl.org\u003e\n" } ] }