)]}' { "log": [ { "commit": "f9f9ffc237dd924f048204e8799da74f9ecf40cf", "tree": "81ed0c3435dfe54781d0f120d3a5938d571bacd1", "parents": [ "0ac9b1c21874d2490331233b3242085f8151e166" ], "author": { "name": "Ben Segall", "email": "bsegall@google.com", "time": "Wed Oct 16 11:16:32 2013 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 29 12:02:32 2013 +0100" }, "message": "sched: Avoid throttle_cfs_rq() racing with period_timer stopping\n\nthrottle_cfs_rq() doesn\u0027t check to make sure that period_timer is running,\nand while update_curr/assign_cfs_runtime does, a concurrently running\nperiod_timer on another cpu could cancel itself between this cpu\u0027s\nupdate_curr and throttle_cfs_rq(). If there are no other cfs_rqs running\nin the tg to restart the timer, this causes the cfs_rq to be stranded\nforever.\n\nFix this by calling __start_cfs_bandwidth() in throttle if the timer is\ninactive.\n\n(Also add some sched_debug lines for cfs_bandwidth.)\n\nTested: make a run/sleep task in a cgroup, loop switching the cgroup\nbetween 1ms/100ms quota and unlimited, checking for timer_active\u003d0 and\nthrottled\u003d1 as a failure. With the throttle_cfs_rq() change commented out\nthis fails, with the full patch it passes.\n\nSigned-off-by: Ben Segall \u003cbsegall@google.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: pjt@google.com\nLink: http://lkml.kernel.org/r/20131016181632.22647.84174.stgit@sword-of-the-dawn.mtv.corp.google.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "0ac9b1c21874d2490331233b3242085f8151e166", "tree": "76199cdef57f1bacd3b0d81aba280e062d81bd73", "parents": [ "927b54fccbf04207ec92f669dce6806848cbec7d" ], "author": { "name": "Paul Turner", "email": "pjt@google.com", "time": "Wed Oct 16 11:16:27 2013 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 29 12:02:23 2013 +0100" }, "message": "sched: Guarantee new group-entities always have weight\n\nCurrently, group entity load-weights are initialized to zero. This\nadmits some races with respect to the first time they are re-weighted in\nearlty use. ( Let g[x] denote the se for \"g\" on cpu \"x\". )\n\nSuppose that we have root-\u003ea and that a enters a throttled state,\nimmediately followed by a[0]-\u003et1 (the only task running on cpu[0])\nblocking:\n\n put_prev_task(group_cfs_rq(a[0]), t1)\n put_prev_entity(..., t1)\n check_cfs_rq_runtime(group_cfs_rq(a[0]))\n throttle_cfs_rq(group_cfs_rq(a[0]))\n\nThen, before unthrottling occurs, let a[0]-\u003eb[0]-\u003et2 wake for the first\ntime:\n\n enqueue_task_fair(rq[0], t2)\n enqueue_entity(group_cfs_rq(b[0]), t2)\n enqueue_entity_load_avg(group_cfs_rq(b[0]), t2)\n account_entity_enqueue(group_cfs_ra(b[0]), t2)\n update_cfs_shares(group_cfs_rq(b[0]))\n \u003c skipped because b is part of a throttled hierarchy \u003e\n enqueue_entity(group_cfs_rq(a[0]), b[0])\n ...\n\nWe now have b[0] enqueued, yet group_cfs_rq(a[0])-\u003eload.weight \u003d\u003d 0\nwhich violates invariants in several code-paths. Eliminate the\npossibility of this by initializing group entity weight.\n\nSigned-off-by: Paul Turner \u003cpjt@google.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/20131016181627.22647.47543.stgit@sword-of-the-dawn.mtv.corp.google.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "927b54fccbf04207ec92f669dce6806848cbec7d", "tree": "4ee5a37f4796df9883c6beb9b7a91ca8c8534049", "parents": [ "db06e78cc13d70f10877e0557becc88ab3ad2be8" ], "author": { "name": "Ben Segall", "email": "bsegall@google.com", "time": "Wed Oct 16 11:16:22 2013 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 29 12:02:21 2013 +0100" }, "message": "sched: Fix hrtimer_cancel()/rq-\u003elock deadlock\n\n__start_cfs_bandwidth calls hrtimer_cancel while holding rq-\u003elock,\nwaiting for the hrtimer to finish. However, if sched_cfs_period_timer\nruns for another loop iteration, the hrtimer can attempt to take\nrq-\u003elock, resulting in deadlock.\n\nFix this by ensuring that cfs_b-\u003etimer_active is cleared only if the\n_latest_ call to do_sched_cfs_period_timer is returning as idle. Then\n__start_cfs_bandwidth can just call hrtimer_try_to_cancel and wait for\nthat to succeed or timer_active \u003d\u003d 1.\n\nSigned-off-by: Ben Segall \u003cbsegall@google.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: pjt@google.com\nLink: http://lkml.kernel.org/r/20131016181622.22647.16643.stgit@sword-of-the-dawn.mtv.corp.google.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "db06e78cc13d70f10877e0557becc88ab3ad2be8", "tree": "867b99d17a5df49fc4c413940a99e6209e81e7a7", "parents": [ "1ee14e6c8cddeeb8a490d7b54cd9016e4bb900b4" ], "author": { "name": "Ben Segall", "email": "bsegall@google.com", "time": "Wed Oct 16 11:16:17 2013 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 29 12:02:20 2013 +0100" }, "message": "sched: Fix cfs_bandwidth misuse of hrtimer_expires_remaining\n\nhrtimer_expires_remaining does not take internal hrtimer locks and thus\nmust be guarded against concurrent __hrtimer_start_range_ns (but\nreturning HRTIMER_RESTART is safe). Use cfs_b-\u003elock to make it safe.\n\nSigned-off-by: Ben Segall \u003cbsegall@google.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: pjt@google.com\nLink: http://lkml.kernel.org/r/20131016181617.22647.73829.stgit@sword-of-the-dawn.mtv.corp.google.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "1ee14e6c8cddeeb8a490d7b54cd9016e4bb900b4", "tree": "1fe48f0122fdda3bf75bb637a9e103e541598e8a", "parents": [ "ac9ff7997b6f2b31949dcd2495ac671fd9ddc990" ], "author": { "name": "Ben Segall", "email": "bsegall@google.com", "time": "Wed Oct 16 11:16:12 2013 -0700" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 29 12:02:19 2013 +0100" }, "message": "sched: Fix race on toggling cfs_bandwidth_used\n\nWhen we transition cfs_bandwidth_used to false, any currently\nthrottled groups will incorrectly return false from cfs_rq_throttled.\nWhile tg_set_cfs_bandwidth will unthrottle them eventually, currently\nrunning code (including at least dequeue_task_fair and\ndistribute_cfs_runtime) will cause errors.\n\nFix this by turning off cfs_bandwidth_used only after unthrottling all\ncfs_rqs.\n\nTested: toggle bandwidth back and forth on a loaded cgroup. Caused\ncrashes in minutes without the patch, hasn\u0027t crashed with it.\n\nSigned-off-by: Ben Segall \u003cbsegall@google.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: pjt@google.com\nLink: http://lkml.kernel.org/r/20131016181611.22647.80365.stgit@sword-of-the-dawn.mtv.corp.google.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ac9ff7997b6f2b31949dcd2495ac671fd9ddc990", "tree": "e4f028e899cf804d60a20b4aed3a4473a996fb6c", "parents": [ "e9aa39bb7c4415ca26484239cc3a6686d549bf4f" ], "author": { "name": "Michael wang", "email": "wangyun@linux.vnet.ibm.com", "time": "Mon Oct 28 10:50:22 2013 +0800" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Mon Oct 28 11:36:50 2013 +0100" }, "message": "sched: Remove extra put_online_cpus() inside sched_setaffinity()\n\nCommit 6acce3ef8:\n\n\tsched: Remove get_online_cpus() usage\n\nhas left one extra put_online_cpus() inside sched_setaffinity(),\nremove it to fix the WARN:\n\n ------------[ cut here ]------------\n WARNING: CPU: 0 PID: 3166 at kernel/cpu.c:84 put_online_cpus+0x43/0x70()\n ...\n [\u003cffffffff810c3fef\u003e] put_online_cpus+0x43/0x70 [\n [\u003cffffffff810efd59\u003e] sched_setaffinity+0x7d/0x1f9 [\n ...\n\nReported-by: Fengguang Wu \u003cfengguang.wu@intel.com\u003e\nTested-by: Fengguang Wu \u003cfengguang.wu@intel.com\u003e\nSigned-off-by: Michael Wang \u003cwangyun@linux.vnet.ibm.com\u003e\nCc: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/526DD0EE.1090309@linux.vnet.ibm.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "e9aa39bb7c4415ca26484239cc3a6686d549bf4f", "tree": "04185e0cfa4b097f95ce532506b1895a13281e0f", "parents": [ "92ec11809565cf6429c75204e99e0f583b5c9d7c" ], "author": { "name": "Li Bin", "email": "huawei.libin@huawei.com", "time": "Mon Oct 21 20:15:43 2013 +0800" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Sat Oct 26 12:25:21 2013 +0200" }, "message": "sched/rt: Fix task_tick_rt() comment\n\nThis issue was introduced by 454c79999f7e (\"sched/rt: Fix SCHED_RR\nacross cgroups\") that missed the word \u0027not\u0027. Fix it.\n\nSigned-off-by: Li Bin \u003chuawei.libin@huawei.com\u003e\nCc: \u003cguohanjun@huawei.com\u003e\nCc: \u003cxiexiuqi@huawei.com\u003e\nCc: \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1382357743-54136-1-git-send-email-huawei.libin@huawei.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "92ec11809565cf6429c75204e99e0f583b5c9d7c", "tree": "cc0f9678b17b8b02ded386c3ff605423dd593fa9", "parents": [ "c2d816443ef305aba8eaf0bf368f4d3d87494f06" ], "author": { "name": "Thierry Reding", "email": "thierry.reding@gmail.com", "time": "Wed Oct 23 13:40:55 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 23 14:44:10 2013 +0200" }, "message": "sched/wait: Fix build breakage\n\nThe wait_event_interruptible_lock_irq() macro is missing a\nsemi-colon which causes a build failure in the i915 DRM driver.\n\nSigned-off-by: Thierry Reding \u003ctreding@nvidia.com\u003e\nCc: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1382528455-29911-1-git-send-email-treding@nvidia.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "c2d816443ef305aba8eaf0bf368f4d3d87494f06", "tree": "0331463c4ea621c1467e83894a9cebf3a91cb136", "parents": [ "8922915b38cd8b72f8e5af614b95be71d1d299d4" ], "author": { "name": "Oleg Nesterov", "email": "oleg@redhat.com", "time": "Mon Oct 07 18:18:24 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 16 14:22:18 2013 +0200" }, "message": "sched/wait: Introduce prepare_to_wait_event()\n\nAdd the new helper, prepare_to_wait_event() which should only be used\nby ___wait_event().\n\nprepare_to_wait_event() returns -ERESTARTSYS if signal_pending_state()\nis true, otherwise it does prepare_to_wait/exclusive. This allows to\nuninline the signal-pending checks in wait_event*() macros.\n\nAlso, it can initialize wait-\u003eprivate/func. We do not care if they were\nalready initialized, the values are the same. This also shaves a couple\nof insns from the inlined code.\n\nThis obviously makes prepare_*() path a little bit slower, but we are\nlikely going to sleep anyway, so I think it makes sense to shrink .text:\n\n text data bss dec hex filename\n \u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\n before: 5126092 2959248 10117120 18202460 115bf5c vmlinux\n after: 5124618 2955152 10117120 18196890 115a99a vmlinux\n\non my build.\n\nSigned-off-by: Oleg Nesterov \u003coleg@redhat.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/20131007161824.GA29757@redhat.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "8922915b38cd8b72f8e5af614b95be71d1d299d4", "tree": "d0770b40ff4d01aaaf050a651c717337b237294e", "parents": [ "6acce3ef84520537f8a09a12c9ddbe814a584dd2" ], "author": { "name": "Oleg Nesterov", "email": "oleg@redhat.com", "time": "Mon Oct 07 20:31:06 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 16 14:22:17 2013 +0200" }, "message": "sched/wait: Add ___wait_cond_timeout() to wait_event*_timeout() too\n\nCommit 4c663cfc (\"wait: fix false timeouts when using\nwait_event_timeout()\") introduced the additional condition checks\nafter a timeout but only in the \"slow\" __wait*() paths.\n\nwait_event_timeout(wq, CONDITION, 0) still returns 0 if CONDITION\nis already true and we do not call __wait*().\n\nNow that we have ___wait_cond_timeout() we can use it instead to\nensure that __ret will be properly updated.\n\nSigned-off-by: Oleg Nesterov \u003coleg@redhat.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/20131007183106.GA10973@redhat.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "6acce3ef84520537f8a09a12c9ddbe814a584dd2", "tree": "b4e117df4a57be6a040529c148480227c3d100cc", "parents": [ "746023159c40c523b08a3bc3d213dac212385895" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Fri Oct 11 14:38:20 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 16 14:22:16 2013 +0200" }, "message": "sched: Remove get_online_cpus() usage\n\nRemove get_online_cpus() usage from the scheduler; there\u0027s 4 sites that\nuse it:\n\n - sched_init_smp(); where its completely superfluous since we\u0027re in\n \u0027early\u0027 boot and there simply cannot be any hotplugging.\n\n - sched_getaffinity(); we already take a raw spinlock to protect the\n task cpus_allowed mask, this disables preemption and therefore\n also stabilizes cpu_online_mask as that\u0027s modified using\n stop_machine. However switch to active mask for symmetry with\n sched_setaffinity()/set_cpus_allowed_ptr(). We guarantee active\n mask stability by inserting sync_rcu/sched() into _cpu_down.\n\n - sched_setaffinity(); we don\u0027t appear to need get_online_cpus()\n either, there\u0027s two sites where hotplug appears relevant:\n * cpuset_cpus_allowed(); for the !cpuset case we use possible_mask,\n for the cpuset case we hold task_lock, which is a spinlock and\n thus for mainline disables preemption (might cause pain on RT).\n * set_cpus_allowed_ptr(); Holds all scheduler locks and thus has\n preemption properly disabled; also it already deals with hotplug\n races explicitly where it releases them.\n\n - migrate_swap(); we can make stop_two_cpus() do the heavy lifting for\n us with a little trickery. By adding a sync_sched/rcu() after the\n CPU_DOWN_PREPARE notifier we can provide preempt/rcu guarantees for\n cpu_active_mask. Use these to validate that both our cpus are active\n when queueing the stop work before we queue the stop_machine works\n for take_cpu_down().\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: \"Srivatsa S. Bhat\" \u003csrivatsa.bhat@linux.vnet.ibm.com\u003e\nCc: Paul McKenney \u003cpaulmck@linux.vnet.ibm.com\u003e\nCc: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Rik van Riel \u003criel@redhat.com\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\nCc: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nCc: Steven Rostedt \u003crostedt@goodmis.org\u003e\nCc: Oleg Nesterov \u003coleg@redhat.com\u003e\nLink: http://lkml.kernel.org/r/20131011123820.GV3081@twins.programming.kicks-ass.net\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "746023159c40c523b08a3bc3d213dac212385895", "tree": "5ec30b81b256a202bf44940360dd1849325c4af8", "parents": [ "7c3f2ab7b844f1a859afbc3d41925e8a0faba5fa" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Thu Oct 10 20:17:22 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 16 14:22:14 2013 +0200" }, "message": "sched: Fix race in migrate_swap_stop()\n\nThere is a subtle race in migrate_swap, when task P, on CPU A, decides to swap\nplaces with task T, on CPU B.\n\nTask P:\n - call migrate_swap\nTask T:\n - go to sleep, removing itself from the runqueue\nTask P:\n - double lock the runqueues on CPU A \u0026 B\nTask T:\n - get woken up, place itself on the runqueue of CPU C\nTask P:\n - see that task T is on a runqueue, and pretend to remove it\n from the runqueue on CPU B\n\nNow CPUs B \u0026 C both have corrupted scheduler data structures.\n\nThis patch fixes it, by holding the pi_lock for both of the tasks\ninvolved in the migrate swap. This prevents task T from waking up,\nand placing itself onto another runqueue, until after migrate_swap\nhas released all locks.\n\nThis means that, when migrate_swap checks, task T will be either\non the runqueue where it was originally seen, or not on any\nrunqueue at all. Migrate_swap deals correctly with of those cases.\n\nTested-by: Joe Mario \u003cjmario@redhat.com\u003e\nAcked-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: hannes@cmpxchg.org\nCc: aarcange@redhat.com\nCc: srikar@linux.vnet.ibm.com\nCc: tglx@linutronix.de\nCc: hpa@zytor.com\nLink: http://lkml.kernel.org/r/20131010181722.GO13848@laptop.programming.kicks-ass.net\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "7c3f2ab7b844f1a859afbc3d41925e8a0faba5fa", "tree": "134aee5c31160aeafef83b211eeed5db2b2ce61c", "parents": [ "ed1b7732868035990f07aeb532b1d86272ea909e" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Tue Oct 15 12:35:07 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 16 14:22:13 2013 +0200" }, "message": "sched/rt: Add missing rmb()\n\nWhile discussing the proposed SCHED_DEADLINE patches which in parts\nmimic the existing FIFO code it was noticed that the wmb in\nrt_set_overloaded() didn\u0027t have a matching barrier.\n\nThe only site using rt_overloaded() to test the rto_count is\npull_rt_task() and we should issue a matching rmb before then assuming\nthere\u0027s an rto_mask bit set.\n\nWithout that smp_rmb() in there we could actually miss seeing the\nrto_mask bit.\n\nAlso, change to using smp_[wr]mb(), even though this is SMP only code;\nmemory barriers without smp_ always make me think they\u0027re against\nhardware of some sort.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nCc: vincent.guittot@linaro.org\nCc: luca.abeni@unitn.it\nCc: bruce.ashfield@windriver.com\nCc: dhaval.giani@gmail.com\nCc: rostedt@goodmis.org\nCc: hgu1972@gmail.com\nCc: oleg@redhat.com\nCc: fweisbec@gmail.com\nCc: darren@dvhart.com\nCc: johan.eker@ericsson.com\nCc: p.faure@akatech.ch\nCc: paulmck@linux.vnet.ibm.com\nCc: raistlin@linux.it\nCc: claudio@evidence.eu.com\nCc: insop.song@gmail.com\nCc: michael@amarulasolutions.com\nCc: liming.wang@windriver.com\nCc: fchecconi@gmail.com\nCc: jkacur@redhat.com\nCc: tommaso.cucinotta@sssup.it\nCc: Juri Lelli \u003cjuri.lelli@gmail.com\u003e\nCc: harald.gustafsson@ericsson.com\nCc: nicola.manica@disi.unitn.it\nCc: tglx@linutronix.de\nLink: http://lkml.kernel.org/r/20131015103507.GF10651@twins.programming.kicks-ass.net\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ed1b7732868035990f07aeb532b1d86272ea909e", "tree": "981975df505488fbebbcd1c45cea00eaa29bd806", "parents": [ "62e947cb0cd27c392aabe732c64f5023e272cf0e" ], "author": { "name": "Kamalesh Babulal", "email": "kamalesh@linux.vnet.ibm.com", "time": "Sun Oct 13 23:06:15 2013 +0530" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Mon Oct 14 09:22:55 2013 +0200" }, "message": "sched/fair: Fix trivial typos in comments\n\n - \u0027load_icx\u0027 \u003d\u003e \u0027load_idx\u0027\n - \u0027calculcate_imbalance\u0027 \u003d\u003e \u0027calculate_imbalance\u0027\n\nSigned-off-by: Kamalesh Babulal \u003ckamalesh@linux.vnet.ibm.com\u003e\nCc: peterz@infradead.org\nLink: http://lkml.kernel.org/r/1381685775-3544-1-git-send-email-kamalesh@linux.vnet.ibm.com\n[ Also, don\u0027t capitalize \u0027idle\u0027 unnecessarily. ]\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "62e947cb0cd27c392aabe732c64f5023e272cf0e", "tree": "cbe389e622400c9f610acbe8341aced7c5f46e86", "parents": [ "88f182dd779b9d350b4774c12d16633a5b60f50c" ], "author": { "name": "Ramkumar Ramachandra", "email": "artagnon@gmail.com", "time": "Thu Oct 10 15:50:33 2013 +0530" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Sat Oct 12 19:01:24 2013 +0200" }, "message": "sched: Remove bogus parameter in structured comment\n\nThe balance parameter was removed by 23f0d20 (\"sched: Factor out\ncode to should_we_balance()\", 2013-08-06).\n\nSigned-off-by: Ramkumar Ramachandra \u003cartagnon@gmail.com\u003e\nCc: Joonsoo Kim \u003ciamjoonsoo.kim@lge.com\u003e\nCc: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381400433-2030-1-git-send-email-artagnon@gmail.com\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "88f182dd779b9d350b4774c12d16633a5b60f50c", "tree": "b5f1fa7e2ab82091203e07e5dff41ae257e67d4f", "parents": [ "ec0ad3d01f99d5e5b56a99a58f7003b99250dc65" ], "author": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Thu Oct 10 10:16:30 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Fri Oct 11 07:41:43 2013 +0200" }, "message": "x86: Apply the asm_volatile_goto() compiler quirk\n\nApply the asm_volatile_goto() compiler quirk to the new rmwcc.h\nfile as well, introduced in:\n\n c2daa3bed53a sched, x86: Provide a per-cpu preempt_count implementation\n\nReported-and-tested-by: Fengguang Wu \u003cfengguang.wu@intel.com\u003e\nReported-by: Oleg Nesterov \u003coleg@redhat.com\u003e\nReported-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSuggested-by: Jakub Jelinek \u003cjakub@redhat.com\u003e\nReviewed-by: Richard Henderson \u003crth@twiddle.net\u003e\nCc: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\nCc: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ec0ad3d01f99d5e5b56a99a58f7003b99250dc65", "tree": "85066cbea1cf8da6d099019adefaca90aae39234", "parents": [ "3354781a2184380046c8dd19144628d3c33991e6", "3f0116c3238a96bc18ad4b4acefe4e7be32fa861" ], "author": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Fri Oct 11 07:39:37 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Fri Oct 11 07:39:37 2013 +0200" }, "message": "Merge branch \u0027core/urgent\u0027 into sched/core\n\nMerge in asm goto fix, to be able to apply the asm/rmwcc.h fix.\n\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "3f0116c3238a96bc18ad4b4acefe4e7be32fa861", "tree": "b30e1fd03f2ab222051d8d1eb776bd1895416497", "parents": [ "2fe80d3bbf1c8bd9efc5b8154207c8dd104e7306" ], "author": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Thu Oct 10 10:16:30 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Fri Oct 11 07:39:14 2013 +0200" }, "message": "compiler/gcc4: Add quirk for \u0027asm goto\u0027 miscompilation bug\n\nFengguang Wu, Oleg Nesterov and Peter Zijlstra tracked down\na kernel crash to a GCC bug: GCC miscompiles certain \u0027asm goto\u0027\nconstructs, as outlined here:\n\n http://gcc.gnu.org/bugzilla/show_bug.cgi?id\u003d58670\n\nImplement a workaround suggested by Jakub Jelinek.\n\nReported-and-tested-by: Fengguang Wu \u003cfengguang.wu@intel.com\u003e\nReported-by: Oleg Nesterov \u003coleg@redhat.com\u003e\nReported-by: Peter Zijlstra \u003ca.p.zijlstra@chello.nl\u003e\nSuggested-by: Jakub Jelinek \u003cjakub@redhat.com\u003e\nReviewed-by: Richard Henderson \u003crth@twiddle.net\u003e\nCc: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\nCc: Andrew Morton \u003cakpm@linux-foundation.org\u003e\nCc: \u003cstable@kernel.org\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "2fe80d3bbf1c8bd9efc5b8154207c8dd104e7306", "tree": "729b2f1de3b8bf8eb96e1640c415b8d0c3b5fe1c", "parents": [ "e3e8ded06f0960d642a81dd676fff0367af68ba0" ], "author": { "name": "Kent Overstreet", "email": "kmo@daterainc.com", "time": "Thu Oct 10 17:31:15 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 18:17:39 2013 -0700" }, "message": "bcache: Fix a null ptr deref regression\n\nCommit c0f04d88e46d (\"bcache: Fix flushes in writeback mode\") was fixing\na reported data corruption bug, but it seems some last minute\nrefactoring or rebasing introduced a null pointer deref.\n\nSigned-off-by: Kent Overstreet \u003ckmo@daterainc.com\u003e\nCc: linux-stable \u003cstable@vger.kernel.org\u003e # \u003e\u003d v3.10\nReported-by: Gabriel de Perthuis \u003cg2p.code@gmail.com\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "e3e8ded06f0960d642a81dd676fff0367af68ba0", "tree": "a47f85ecd7e344342c6c8602169f760c929269d5", "parents": [ "2d9d0282834dfdd7a444ca5001935187a38eb598", "25f2bd7f5add608c1d1405938f39c96927b275ca" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 18:16:02 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 18:16:02 2013 -0700" }, "message": "Merge tag \u0027hwmon-for-linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging\n\nPull hwmon fix from Guenter Roeck:\n \"Fix root cause of crash/error seen in applesmc driver\"\n\n* tag \u0027hwmon-for-linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging:\n hwmon: (applesmc) Always read until end of data\n" }, { "commit": "2d9d0282834dfdd7a444ca5001935187a38eb598", "tree": "df6bc18a22c931b4d170acc129a63542f1b1cd85", "parents": [ "f112bf8585b9ce4e3cce2eee6b8319eb9fee3ec1", "19514fc665ffbce624785f76ee7ad0ea6378a527" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 18:15:23 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 18:15:23 2013 -0700" }, "message": "Merge branch \u0027rc-fixes\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild\n\nPull kbuild fix from Michal Marek:\n \"Here is an ARM Makefile fix that you even acked. After nobody wanted\n to take it, it ended up in the kbuild tree\"\n\n* \u0027rc-fixes\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:\n arm, kbuild: make \"make install\" not depend on vmlinux\n" }, { "commit": "f112bf8585b9ce4e3cce2eee6b8319eb9fee3ec1", "tree": "8792159f0212c3b296c5894f85f68b9d443672b0", "parents": [ "0ff18e3734c248fa516b4be942b1196832ee936c", "0821f20d49d19f49e53b05ef6727e010c4b1f11d" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 13:57:10 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 13:57:10 2013 -0700" }, "message": "Merge git://www.linux-watchdog.org/linux-watchdog\n\nPull watchdog fix from Wim Van Sebroeck:\n \"Make sure that the hpwdt driver will not load auxilary iLO devices\"\n\n* git://www.linux-watchdog.org/linux-watchdog:\n watchdog: hpwdt: Patch to ignore auxilary iLO devices\n" }, { "commit": "0ff18e3734c248fa516b4be942b1196832ee936c", "tree": "a978c59f375936c7e37c67a61dc08821a5e674bb", "parents": [ "f715729ee4cb666f51749f6cd86d06fff1e6e17b" ], "author": { "name": "Fengguang Wu", "email": "fengguang.wu@intel.com", "time": "Wed Oct 09 09:26:21 2013 +0800" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 13:56:52 2013 -0700" }, "message": "kobject: show debug info on delayed kobject release\n\nUseful for locating buggy drivers on kernel oops.\n\nIt may add dozens of new lines to boot dmesg. DEBUG_KOBJECT_RELEASE is\nhopefully only enabled in debug kernels (like maybe the Fedora rawhide\none, or at developers), so being a bit more verbose is likely ok.\n\nAcked-by: Russell King \u003crmk+kernel@arm.linux.org.uk\u003e\nCc: Greg Kroah-Hartman \u003cgregkh@linuxfoundation.org\u003e\nSigned-off-by: Fengguang Wu \u003cfengguang.wu@intel.com\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" }, { "commit": "0821f20d49d19f49e53b05ef6727e010c4b1f11d", "tree": "b4190aa3d36d29720aa8b615548cf3859336862e", "parents": [ "0e7a3ed04f0cd4311096d691888f88569310ee6c" ], "author": { "name": "Mingarelli, Thomas", "email": "Thomas.Mingarelli@hp.com", "time": "Fri Aug 09 16:31:09 2013 +0000" }, "committer": { "name": "Wim Van Sebroeck", "email": "wim@iguana.be", "time": "Thu Oct 10 22:23:10 2013 +0200" }, "message": "watchdog: hpwdt: Patch to ignore auxilary iLO devices\n\nThis patch is to prevent hpwdt from loading on any auxilary iLO devices defined\nafter the initial (or main) iLO device. All auxilary iLO devices will have a\nsubsystem device ID set to 0x1979 in order for hpwdt to differentiate between\nthe two types.\n\nSigned-off-by: Thomas Mingarelli \u003cthomas.mingarelli@hp.com\u003e\nTested-by: Lisa Mitchell \u003clisa.mitchell@hp.com\u003e\nSigned-off-by: Wim Van Sebroeck \u003cwim@iguana.be\u003e\n" }, { "commit": "f715729ee4cb666f51749f6cd86d06fff1e6e17b", "tree": "cf8f2c6c2d90441d6a978915b5d7eb0b19981f44", "parents": [ "8273548c5455e3ae27e905a77bad277535837329", "61875f30daf60305712e25b209ef41ced2635bad" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 12:31:43 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 12:31:43 2013 -0700" }, "message": "Merge tag \u0027random_for_linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/random\n\nPull /dev/random changes from Ted Ts\u0027o:\n \"These patches are designed to enable improvements to /dev/random for\n non-x86 platforms, in particular MIPS and ARM\"\n\n* tag \u0027random_for_linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/random:\n random: allow architectures to optionally define random_get_entropy()\n random: run random_int_secret_init() run after all late_initcalls\n" }, { "commit": "8273548c5455e3ae27e905a77bad277535837329", "tree": "5b539100070e5f908df39a7de955084cf2990c75", "parents": [ "bd7df5ad056995e083e682b8e4901031a7b38836", "40fde70d0df993d1a652d6cc69f8b4e967656170" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 11:33:48 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 11:33:48 2013 -0700" }, "message": "Merge tag \u0027for-linus\u0027 of git://git.kernel.org/pub/scm/virt/kvm/kvm\n\nPull kvm fixes from Paolo Bonzini:\n \"Fixes for 3.12-rc5: two old PPC bugs and one new (3.12-rc2) x86 bug\"\n\n* tag \u0027for-linus\u0027 of git://git.kernel.org/pub/scm/virt/kvm/kvm:\n kvm: ppc: booke: check range page invalidation progress on page setup\n KVM: PPC: Book3S HV: Fix typo in saving DSCR\n KVM: nVMX: fix shadow on EPT\n" }, { "commit": "bd7df5ad056995e083e682b8e4901031a7b38836", "tree": "4a0da3c9c4b72e38d35686d1e37c70ffb67214b8", "parents": [ "b291a22952e74bb2fcf9c57c6d8bae52a4f197e2", "ac9fdc88641cf68b1077afb0b5cd7ed87fe4a5c2" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 11:33:02 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 11:33:02 2013 -0700" }, "message": "Merge tag \u0027spi-v3.12-rc4\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi\n\nPull spi fixes from Mark Brown:\n \"This is all driver updates, mostly fixes for error handling paths\n except for the s3c64xx and hspi fixes for trying to use runtime PM\n before it is enabled and the pxa2xx fix for interactions between power\n management and interrupt handling\"\n\n* tag \u0027spi-v3.12-rc4\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:\n spi: atmel: Fix incorrect error path\n spi/hspi: fixup Runtime PM enable timing\n spi/s3c64xx: Ensure runtime PM is enabled prior to registration\n spi/clps711x: drop clk_put for devm_clk_get in spi_clps711x_probe()\n spi: fix return value check in dspi_probe()\n spi: mpc512x: fix error return code in mpc512x_psc_spi_do_probe()\n spi: clps711x: Don\u0027t call kfree() after spi_master_put/spi_unregister_master\n spi/pxa2xx: check status register as well to determine if the device is off\n" }, { "commit": "61875f30daf60305712e25b209ef41ced2635bad", "tree": "176cda3ae729c520bff1b0f6a19806c18c0a7523", "parents": [ "47d06e532e95b71c0db3839ebdef3fe8812fca2c" ], "author": { "name": "Theodore Ts\u0027o", "email": "tytso@mit.edu", "time": "Sat Sep 21 13:58:22 2013 -0400" }, "committer": { "name": "Theodore Ts\u0027o", "email": "tytso@mit.edu", "time": "Thu Oct 10 14:30:53 2013 -0400" }, "message": "random: allow architectures to optionally define random_get_entropy()\n\nAllow architectures which have a disabled get_cycles() function to\nprovide a random_get_entropy() function which provides a fine-grained,\nrapidly changing counter that can be used by the /dev/random driver.\n\nFor example, an architecture might have a rapidly changing register\nused to control random TLB cache eviction, or DRAM refresh that\ndoesn\u0027t meet the requirements of get_cycles(), but which is good\nenough for the needs of the random driver.\n\nSigned-off-by: \"Theodore Ts\u0027o\" \u003ctytso@mit.edu\u003e\nCc: stable@vger.kernel.org\n" }, { "commit": "b291a22952e74bb2fcf9c57c6d8bae52a4f197e2", "tree": "d216a0e6e29c16e4dc18d0c637799b22ad9f2c61", "parents": [ "0e7a3ed04f0cd4311096d691888f88569310ee6c", "2b468ef0e7959b703626b64c4d264ef822c9267a" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 11:30:33 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Thu Oct 10 11:30:33 2013 -0700" }, "message": "Merge tag \u0027for-linus-20131008\u0027 of git://git.infradead.org/linux-mtd\n\nPull MTD fixes from Brian Norris:\n - fix a small memory leak in some new ONFI code\n - account for additional odd variations of Micron SPI flash\n\nAcked by David Woodhouse.\n\n* tag \u0027for-linus-20131008\u0027 of git://git.infradead.org/linux-mtd:\n mtd: m25p80: Fix 4 byte addressing mode for Micron devices.\n mtd: nand: fix memory leak in ONFI extended parameter page\n" }, { "commit": "40fde70d0df993d1a652d6cc69f8b4e967656170", "tree": "9d9ec53946c35f6335c87a6ec6ba504ed7690fbc", "parents": [ "cfc860253abd73e1681696c08ea268d33285a2c4" ], "author": { "name": "Bharat Bhushan", "email": "r65777@freescale.com", "time": "Wed Aug 07 15:33:46 2013 +0530" }, "committer": { "name": "Paolo Bonzini", "email": "pbonzini@redhat.com", "time": "Thu Oct 10 11:40:08 2013 +0200" }, "message": "kvm: ppc: booke: check range page invalidation progress on page setup\n\nWhen the MM code is invalidating a range of pages, it calls the KVM\nkvm_mmu_notifier_invalidate_range_start() notifier function, which calls\nkvm_unmap_hva_range(), which arranges to flush all the TLBs for guest pages.\nHowever, the Linux PTEs for the range being flushed are still valid at\nthat point. We are not supposed to establish any new references to pages\nin the range until the ...range_end() notifier gets called.\nThe PPC-specific KVM code doesn\u0027t get any explicit notification of that;\ninstead, we are supposed to use mmu_notifier_retry() to test whether we\nare or have been inside a range flush notifier pair while we have been\nreferencing a page.\n\nThis patch calls the mmu_notifier_retry() while mapping the guest\npage to ensure we are not referencing a page when in range invalidation.\n\nThis call is inside a region locked with kvm-\u003emmu_lock, which is the\nsame lock that is called by the KVM MMU notifier functions, thus\nensuring that no new notification can proceed while we are in the\nlocked region.\n\nSigned-off-by: Bharat Bhushan \u003cbharat.bhushan@freescale.com\u003e\nAcked-by: Alexander Graf \u003cagraf@suse.de\u003e\n[Backported to 3.12 - Paolo]\nReviewed-by: Bharat Bhushan \u003cbharat.bhushan@freescale.com\u003e\nSigned-off-by: Paolo Bonzini \u003cpbonzini@redhat.com\u003e\n" }, { "commit": "cfc860253abd73e1681696c08ea268d33285a2c4", "tree": "b0a307c8ef521952c2cc7157807a935424247e85", "parents": [ "d0d538b9d1239ed6ed318a2304ba5a763f835089" ], "author": { "name": "Paul Mackerras", "email": "paulus@samba.org", "time": "Sat Sep 21 09:53:28 2013 +1000" }, "committer": { "name": "Paolo Bonzini", "email": "pbonzini@redhat.com", "time": "Thu Oct 10 11:40:07 2013 +0200" }, "message": "KVM: PPC: Book3S HV: Fix typo in saving DSCR\n\nThis fixes a typo in the code that saves the guest DSCR (Data Stream\nControl Register) into the kvm_vcpu_arch struct on guest exit. The\neffect of the typo was that the DSCR value was saved in the wrong place,\nso changes to the DSCR by the guest didn\u0027t persist across guest exit\nand entry, and some host kernel memory got corrupted.\n\nCc: stable@vger.kernel.org [v3.1+]\nSigned-off-by: Paul Mackerras \u003cpaulus@samba.org\u003e\nAcked-by: Alexander Graf \u003cagraf@suse.de\u003e\nSigned-off-by: Paolo Bonzini \u003cpbonzini@redhat.com\u003e\n" }, { "commit": "d0d538b9d1239ed6ed318a2304ba5a763f835089", "tree": "31475e605ac4068d647cb2be76d9851910302751", "parents": [ "6d15ee492809d38bd62237b6d0f6a81d4dd12d15" ], "author": { "name": "Gleb Natapov", "email": "gleb@redhat.com", "time": "Wed Oct 09 19:13:19 2013 +0300" }, "committer": { "name": "Paolo Bonzini", "email": "pbonzini@redhat.com", "time": "Thu Oct 10 11:39:57 2013 +0200" }, "message": "KVM: nVMX: fix shadow on EPT\n\n72f857950f6f19 broke shadow on EPT. This patch reverts it and fixes PAE\non nEPT (which reverted commit fixed) in other way.\n\nShadow on EPT is now broken because while L1 builds shadow page table\nfor L2 (which is PAE while L2 is in real mode) it never loads L2\u0027s\nGUEST_PDPTR[0-3]. They do not need to be loaded because without nested\nvirtualization HW does this during guest entry if EPT is disabled,\nbut in our case L0 emulates L2\u0027s vmentry while EPT is enables, so we\ncannot rely on vmcs12-\u003eguest_pdptr[0-3] to contain up-to-date values\nand need to re-read PDPTEs from L2 memory. This is what kvm_set_cr3()\nis doing, but by clearing cache bits during L2 vmentry we drop values\nthat kvm_set_cr3() read from memory.\n\nSo why the same code does not work for PAE on nEPT? kvm_set_cr3()\nreads pdptes into vcpu-\u003earch.walk_mmu-\u003epdptrs[]. walk_mmu points to\nvcpu-\u003earch.nested_mmu while nested guest is running, but ept_load_pdptrs()\nuses vcpu-\u003earch.mmu which contain incorrect values. Fix that by using\nwalk_mmu in ept_(load|save)_pdptrs.\n\nSigned-off-by: Gleb Natapov \u003cgleb@redhat.com\u003e\nReviewed-by: Paolo Bonzini \u003cpbonzini@redhat.com\u003e\nTested-by: Paolo Bonzini \u003cpbonzini@redhat.com\u003e\nSigned-off-by: Paolo Bonzini \u003cpbonzini@redhat.com\u003e\n" }, { "commit": "25f2bd7f5add608c1d1405938f39c96927b275ca", "tree": "7eaef0e6f3618ace1096905bd4ceb2d5597517f6", "parents": [ "15c03dd4859ab16f9212238f29dd315654aa94f6" ], "author": { "name": "Henrik Rydberg", "email": "rydberg@euromail.se", "time": "Wed Oct 02 19:15:03 2013 +0200" }, "committer": { "name": "Guenter Roeck", "email": "linux@roeck-us.net", "time": "Wed Oct 09 09:48:55 2013 -0700" }, "message": "hwmon: (applesmc) Always read until end of data\n\nThe crash reported and investigated in commit 5f4513 turned out to be\ncaused by a change to the read interface on newer (2012) SMCs.\n\nTests by Chris show that simply reading the data valid line is enough\nfor the problem to go away. Additional tests show that the newer SMCs\nno longer wait for the number of requested bytes, but start sending\ndata right away. Apparently the number of bytes to read is no longer\nspecified as before, but instead found out by reading until end of\ndata. Failure to read until end of data confuses the state machine,\nwhich eventually causes the crash.\n\nAs a remedy, assuming bit0 is the read valid line, make sure there is\nnothing more to read before leaving the read function.\n\nTested to resolve the original problem, and runtested on MBA3,1,\nMBP4,1, MBP8,2, MBP10,1, MBP10,2. The patch seems to have no effect on\nmachines before 2012.\n\nTested-by: Chris Murphy \u003cchris@cmurf.com\u003e\nCc: stable@vger.kernel.org\nSigned-off-by: Henrik Rydberg \u003crydberg@euromail.se\u003e\nSigned-off-by: Guenter Roeck \u003clinux@roeck-us.net\u003e\n" }, { "commit": "3354781a2184380046c8dd19144628d3c33991e6", "tree": "013a9f5584e5e4518374db44a20385ca062fb985", "parents": [ "2739d3eef3a93a92c366a3a0bb85a0afe09e8b8c" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Wed Oct 09 10:24:48 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:27 2013 +0200" }, "message": "sched/numa: Reflow task_numa_group() to avoid a compiler warning\n\nReflow the function a bit because GCC gets confused:\n\n kernel/sched/fair.c: In function ‘task_numa_fault’:\n kernel/sched/fair.c:1448:3: warning: ‘my_grp’ may be used uninitialized in this function [-Wmaybe-uninitialized]\n kernel/sched/fair.c:1463:27: note: ‘my_grp’ was declared here\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/n/tip-6ebt6x7u64pbbonq1khqu2z9@git.kernel.org\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "2739d3eef3a93a92c366a3a0bb85a0afe09e8b8c", "tree": "74f70c1af9469a659758d8caaacaafafe17df28f", "parents": [ "989348b5fc2367d6880d23a1c779a90bbb6f9baf" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:41 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:25 2013 +0200" }, "message": "sched/numa: Retry task_numa_migrate() periodically\n\nShort spikes of CPU load can lead to a task being migrated\naway from its preferred node for temporary reasons.\n\nIt is important that the task is migrated back to where it\nbelongs, in order to avoid migrating too much memory to its\nnew location, and generally disturbing a task\u0027s NUMA location.\n\nThis patch fixes NUMA placement for 4 specjbb instances on\na 4 node system. Without this patch, things take longer to\nconverge, and processes are not always completely on their\nown node.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-64-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "989348b5fc2367d6880d23a1c779a90bbb6f9baf", "tree": "cc5b883e5e2534b6a29c01bee153efd8daf445d5", "parents": [ "de1c9ce6f07fec0381a39a9d0b379ea35aa1167f" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:40 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:23 2013 +0200" }, "message": "sched/numa: Use unsigned longs for numa group fault stats\n\nAs Peter says \"If you\u0027re going to hold locks you can also do away with all\nthat atomic_long_*() nonsense\". Lock aquisition moved slightly to protect\nthe updates.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-63-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "de1c9ce6f07fec0381a39a9d0b379ea35aa1167f", "tree": "d96bf1a2b25dfa84d3fe5f6fe00fb780800e3ef3", "parents": [ "1e3646ffc64b232cb14a5ef01d7b98997c1b73f9" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:39 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:21 2013 +0200" }, "message": "sched/numa: Skip some page migrations after a shared fault\n\nShared faults can lead to lots of unnecessary page migrations,\nslowing down the system, and causing private faults to hit the\nper-pgdat migration ratelimit.\n\nThis patch adds sysctl numa_balancing_migrate_deferred, which specifies\nhow many shared page migrations to skip unconditionally, after each page\nmigration that is skipped because it is a shared fault.\n\nThis reduces the number of page migrations back and forth in\nshared fault situations. It also gives a strong preference to\nthe tasks that are already running where most of the memory is,\nand to moving the other tasks to near the memory.\n\nTesting this with a much higher scan rate than the default\nstill seems to result in fewer page migrations than before.\n\nMemory seems to be somewhat better consolidated than previously,\nwith multi-instance specjbb runs on a 4 node system.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-62-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "1e3646ffc64b232cb14a5ef01d7b98997c1b73f9", "tree": "6bb7610078609d8d84e1b7d535d958de41c8d588", "parents": [ "930aa174fcc8b0efaad102fd80f677b92f35eaa2" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:38 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:20 2013 +0200" }, "message": "mm: numa: Revert temporarily disabling of NUMA migration\n\nWith the scan rate code working (at least for multi-instance specjbb),\nthe large hammer that is \"sched: Do not migrate memory immediately after\nswitching node\" can be replaced with something smarter. Revert temporarily\nmigration disabling and all traces of numa_migrate_seq.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-61-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "930aa174fcc8b0efaad102fd80f677b92f35eaa2", "tree": "7746bbcf350f3ce305d9e55435f7a5e3c41b9c8e", "parents": [ "04bb2f9475054298f0c67a89ca92cade42d3fe5e" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:37 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:18 2013 +0200" }, "message": "sched/numa: Remove the numa_balancing_scan_period_reset sysctl\n\nWith scan rate adaptions based on whether the workload has properly\nconverged or not there should be no need for the scan period reset\nhammer. Get rid of it.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-60-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "04bb2f9475054298f0c67a89ca92cade42d3fe5e", "tree": "ab48887e23b7f820380a3f415cbe0a6f64f7fecc", "parents": [ "3e6a9418cf05638b103e34f5d13be0321872e623" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:36 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:16 2013 +0200" }, "message": "sched/numa: Adjust scan rate in task_numa_placement\n\nAdjust numa_scan_period in task_numa_placement, depending on how much\nuseful work the numa code can do. The more local faults there are in a\ngiven scan window the longer the period (and hence the slower the scan rate)\nduring the next window. If there are excessive shared faults then the scan\nperiod will decrease with the amount of scaling depending on whether the\nratio of shared/private faults. If the preferred node changes then the\nscan rate is reset to recheck if the task is properly placed.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-59-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "3e6a9418cf05638b103e34f5d13be0321872e623", "tree": "5ef3c692c7cd728f53ebabcfebdc94a21db7a8ea", "parents": [ "dabe1d992414a6456e60e41f1d1ad8affc6d444d" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:35 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:14 2013 +0200" }, "message": "sched/numa: Take false sharing into account when adapting scan rate\n\nScan rate is altered based on whether shared/private faults dominated.\ntask_numa_group() may detect false sharing but that information is not\ntaken into account when adapting the scan rate. Take it into account.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-58-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "dabe1d992414a6456e60e41f1d1ad8affc6d444d", "tree": "213b65fbc7864687d0ea9db96d7e768b6ff587cf", "parents": [ "0ec8aa00f2b4dc457836ef4e2662b02483e94fb7" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:34 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:12 2013 +0200" }, "message": "sched/numa: Be more careful about joining numa groups\n\nDue to the way the pid is truncated, and tasks are moved between\nCPUs by the scheduler, it is possible for the current task_numa_fault\nto group together tasks that do not actually share memory together.\n\nThis patch adds a few easy sanity checks to task_numa_fault, joining\ntasks together if they share the same tsk-\u003emm, or if the fault was on\na page with an elevated mapcount, in a shared VMA.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-57-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "0ec8aa00f2b4dc457836ef4e2662b02483e94fb7", "tree": "ffd621a5f639a10b1a7213892649e512ea0ee1ba", "parents": [ "ca28aa53dd95868c9e38917b9881c09dacfacf1a" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:29:33 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:10 2013 +0200" }, "message": "sched/numa: Avoid migrating tasks that are placed on their preferred node\n\nThis patch classifies scheduler domains and runqueues into types depending\nthe number of tasks that are about their NUMA placement and the number\nthat are currently running on their preferred node. The types are\n\nregular: There are tasks running that do not care about their NUMA\n\tplacement.\n\nremote: There are tasks running that care about their placement but are\n\tcurrently running on a node remote to their ideal placement\n\nall: No distinction\n\nTo implement this the patch tracks the number of tasks that are optimally\nNUMA placed (rq-\u003enr_preferred_running) and the number of tasks running\nthat care about their placement (nr_numa_running). The load balancer\nuses this information to avoid migrating idea placed NUMA tasks as long\nas better options for load balancing exists. For example, it will not\nconsider balancing between a group whose tasks are all perfectly placed\nand a group with remote tasks.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-56-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ca28aa53dd95868c9e38917b9881c09dacfacf1a", "tree": "82d068c0c711dcba8a66d3760d6679586ae07638", "parents": [ "887c290e82e8950d854730c084904c115fc367ac" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:32 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:08 2013 +0200" }, "message": "sched/numa: Fix task or group comparison\n\nThis patch separately considers task and group affinities when\nsearching for swap candidates during NUMA placement. If tasks\nare part of the same group, or no group at all, the task weights\nare considered.\n\nSome hysteresis is added to prevent tasks within one group from\ngetting bounced between NUMA nodes due to tiny differences.\n\nIf tasks are part of different groups, the code compares group\nweights, in order to favor grouping task groups together.\n\nThe patch also changes the group weight multiplier to be the\nsame as the task weight multiplier, since the two are no longer\nadded up like before.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-55-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "887c290e82e8950d854730c084904c115fc367ac", "tree": "9d7064eeb89e08c72b33ae973399ddac59468876", "parents": [ "b32e86b4301e345611f0446265f782a229faadf6" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:31 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:06 2013 +0200" }, "message": "sched/numa: Decide whether to favour task or group weights based on swap candidate relationships\n\nThis patch separately considers task and group affinities when searching\nfor swap candidates during task NUMA placement. If tasks are not part of\na group or the same group then the task weights are considered.\nOtherwise the group weights are compared.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-54-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "b32e86b4301e345611f0446265f782a229faadf6", "tree": "8cc0a8b731e63f2b3450f89330d49009374c99fb", "parents": [ "7dbd13ed06513b047216a7ffc718bad9df0660f1" ], "author": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Mon Oct 07 11:29:30 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:04 2013 +0200" }, "message": "sched/numa: Add debugging\n\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-53-git-send-email-mgorman@suse.de\n" }, { "commit": "7dbd13ed06513b047216a7ffc718bad9df0660f1", "tree": "6f752254b51954b129fa3a35f44968342060afdb", "parents": [ "82727018b0d33d188e9916bcf76f18387484cb04" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:29 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:02 2013 +0200" }, "message": "sched/numa: Prevent parallel updates to group stats during placement\n\nHaving multiple tasks in a group go through task_numa_placement\nsimultaneously can lead to a task picking a wrong node to run on, because\nthe group stats may be in the middle of an update. This patch avoids\nparallel updates by holding the numa_group lock during placement\ndecisions.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-52-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "82727018b0d33d188e9916bcf76f18387484cb04", "tree": "429e0db98c4512f714a668ae9ef9d3bb6e85ce0f", "parents": [ "83e1d2cd9eabec5164afea295ff06b941ae8e4a9" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:28 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:48:00 2013 +0200" }, "message": "sched/numa: Call task_numa_free() from do_execve()\n\nIt is possible for a task in a numa group to call exec, and\nhave the new (unrelated) executable inherit the numa group\nassociation from its former self.\n\nThis has the potential to break numa grouping, and is trivial\nto fix.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-51-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "83e1d2cd9eabec5164afea295ff06b941ae8e4a9", "tree": "f1f23d5483b00be3ce851c941de72ea52d6f7a4b", "parents": [ "5e1576ed0e54d419286a8096133029062b6ad456" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:27 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:58 2013 +0200" }, "message": "sched/numa: Use group fault statistics in numa placement\n\nThis patch uses the fraction of faults on a particular node for both task\nand group, to figure out the best node to place a task. If the task and\ngroup statistics disagree on what the preferred node should be then a full\nrescan will select the node with the best combined weight.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-50-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "5e1576ed0e54d419286a8096133029062b6ad456", "tree": "2394e5c0ed3783e36f23961c0330276cde820e03", "parents": [ "0f19c17929c952c6f0966d93ab05558e7bf814cc" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:26 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:57 2013 +0200" }, "message": "sched/numa: Stay on the same node if CLONE_VM\n\nA newly spawned thread inside a process should stay on the same\nNUMA node as its parent. This prevents processes from being \"torn\"\nacross multiple NUMA nodes every time they spawn a new thread.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-49-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "0f19c17929c952c6f0966d93ab05558e7bf814cc", "tree": "a881a5c520d8d0791dd73859f51c87285d3a06be", "parents": [ "6688cc05473b36a0a3d3971e1adf1712919b32eb" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:25 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:55 2013 +0200" }, "message": "mm: numa: Do not batch handle PMD pages\n\nWith the THP migration races closed it is still possible to occasionally\nsee corruption. The problem is related to handling PMD pages in batch.\nWhen a page fault is handled it can be assumed that the page being\nfaulted will also be flushed from the TLB. The same flushing does not\nhappen when handling PMD pages in batch. Fixing is straight forward but\nthere are a number of reasons not to\n\n1. Multiple TLB flushes may have to be sent depending on what pages get\n migrated\n2. The handling of PMDs in batch means that faults get accounted to\n the task that is handling the fault. While care is taken to only\n mark PMDs where the last CPU and PID match it can still have problems\n due to PID truncation when matching PIDs.\n3. Batching on the PMD level may reduce faults but setting pmd_numa\n requires taking a heavy lock that can contend with THP migration\n and handling the fault requires the release/acquisition of the PTL\n for every page migrated. It\u0027s still pretty heavy.\n\nPMD batch handling is not something that people ever have been happy\nwith. This patch removes it and later patches will deal with the\nadditional fault overhead using more installigent migrate rate adaption.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-48-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "6688cc05473b36a0a3d3971e1adf1712919b32eb", "tree": "2305dca14e2df669b48138088bd93cf51db37721", "parents": [ "7851a45cd3f6198bf542c30e27b330e8eeb3736c" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:29:24 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:53 2013 +0200" }, "message": "mm: numa: Do not group on RO pages\n\nAnd here\u0027s a little something to make sure not the whole world ends up\nin a single group.\n\nAs while we don\u0027t migrate shared executable pages, we do scan/fault on\nthem. And since everybody links to libc, everybody ends up in the same\ngroup.\n\nSuggested-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-47-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "7851a45cd3f6198bf542c30e27b330e8eeb3736c", "tree": "6e81020606c34c95aeded72620738e3a22032e13", "parents": [ "e29cf08b05dc0b8151d65704d96d525a9e179a6b" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:23 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:51 2013 +0200" }, "message": "mm: numa: Copy cpupid on page migration\n\nAfter page migration, the new page has the nidpid unset. This makes\nevery fault on a recently migrated page look like a first numa fault,\nleading to another page migration.\n\nCopying over the nidpid at page migration time should prevent erroneous\nmigrations of recently migrated pages.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-46-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "e29cf08b05dc0b8151d65704d96d525a9e179a6b", "tree": "07eb90e16973acca57ed9bc3866f2bff8ab0750a", "parents": [ "8c8a743c5087bac9caac8155b8f3b367e75cdd0b" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:22 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:49 2013 +0200" }, "message": "sched/numa: Report a NUMA task group ID\n\nIt is desirable to model from userspace how the scheduler groups tasks\nover time. This patch adds an ID to the numa_group and reports it via\n/proc/PID/status.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-45-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "8c8a743c5087bac9caac8155b8f3b367e75cdd0b", "tree": "d5cc5c5f4368cf7d3deb627388c869dbea2e83f2", "parents": [ "90572890d202527c366aa9489b32404e88a7c020" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:29:21 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:47 2013 +0200" }, "message": "sched/numa: Use {cpu, pid} to create task groups for shared faults\n\nWhile parallel applications tend to align their data on the cache\nboundary, they tend not to align on the page or THP boundary.\nConsequently tasks that partition their data can still \"false-share\"\npages presenting a problem for optimal NUMA placement.\n\nThis patch uses NUMA hinting faults to chain tasks together into\nnuma_groups. As well as storing the NID a task was running on when\naccessing a page a truncated representation of the faulting PID is\nstored. If subsequent faults are from different PIDs it is reasonable\nto assume that those two tasks share a page and are candidates for\nbeing grouped together. Note that this patch makes no scheduling\ndecisions based on the grouping information.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-44-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "90572890d202527c366aa9489b32404e88a7c020", "tree": "0577f3b043e312f6d53e50105b236514f7df2455", "parents": [ "e1dda8a797b59d7ec4b17e393152ec3273a552d5" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:29:20 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:45 2013 +0200" }, "message": "mm: numa: Change page last {nid,pid} into {cpu,pid}\n\nChange the per page last fault tracking to use cpu,pid instead of\nnid,pid. This will allow us to try and lookup the alternate task more\neasily. Note that even though it is the cpu that is store in the page\nflags that the mpol_misplaced decision is still based on the node.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-43-git-send-email-mgorman@suse.de\n[ Fixed build failure on 32-bit systems. ]\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "e1dda8a797b59d7ec4b17e393152ec3273a552d5", "tree": "256769c0da413cb1d5fcaeabd06316d24804259c", "parents": [ "2c8a50aa873a7e1d6cc0913362051ff9912dc6ca" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:19 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:43 2013 +0200" }, "message": "sched/numa: Fix placement of workloads spread across multiple nodes\n\nThe load balancer will spread workloads across multiple NUMA nodes,\nin order to balance the load on the system. This means that sometimes\na task\u0027s preferred node has available capacity, but moving the task\nthere will not succeed, because that would create too large an imbalance.\n\nIn that case, other NUMA nodes need to be considered.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-42-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "2c8a50aa873a7e1d6cc0913362051ff9912dc6ca", "tree": "ca32297c046a3789b44e1255b94ef8d0b760d1f9", "parents": [ "fb13c7ee0ed387bd6bec4b4024a4d49b1bd504f1" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:18 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:41 2013 +0200" }, "message": "sched/numa: Favor placing a task on the preferred node\n\nA tasks preferred node is selected based on the number of faults\nrecorded for a node but the actual task_numa_migate() conducts a global\nsearch regardless of the preferred nid. This patch checks if the\npreferred nid has capacity and if so, searches for a CPU within that\nnode. This avoids a global search when the preferred node is not\noverloaded.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-41-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "fb13c7ee0ed387bd6bec4b4024a4d49b1bd504f1", "tree": "b5892db95bf0b47375cc43005291006aeb115772", "parents": [ "ac66f5477239ebd3c4e2cbf2f591ef387aa09884" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:17 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 14:47:25 2013 +0200" }, "message": "sched/numa: Use a system-wide search to find swap/migration candidates\n\nThis patch implements a system-wide search for swap/migration candidates\nbased on total NUMA hinting faults. It has a balance limit, however it\ndoesn\u0027t properly consider total node balance.\n\nIn the old scheme a task selected a preferred node based on the highest\nnumber of private faults recorded on the node. In this scheme, the preferred\nnode is based on the total number of faults. If the preferred node for a\ntask changes then task_numa_migrate will search the whole system looking\nfor tasks to swap with that would improve both the overall compute\nbalance and minimise the expected number of remote NUMA hinting faults.\n\nNot there is no guarantee that the node the source task is placed\non by task_numa_migrate() has any relationship to the newly selected\ntask-\u003enuma_preferred_nid due to compute overloading.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\n[ Do not swap with tasks that cannot run on source cpu]\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\n[ Fixed compiler warning on UP. ]\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-40-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ac66f5477239ebd3c4e2cbf2f591ef387aa09884", "tree": "d340545fa9235f4369c2bd50852fa3a2b7e807f1", "parents": [ "1be0bd77c5dd7c903f46abf52f9a3650face3c1d" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:29:16 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:46 2013 +0200" }, "message": "sched/numa: Introduce migrate_swap()\n\nUse the new stop_two_cpus() to implement migrate_swap(), a function that\nflips two tasks between their respective cpus.\n\nI\u0027m fairly sure there\u0027s a less crude way than employing the stop_two_cpus()\nmethod, but everything I tried either got horribly fragile and/or complex. So\nkeep it simple for now.\n\nThe notable detail is how we \u0027migrate\u0027 tasks that aren\u0027t runnable\nanymore. We\u0027ll make it appear like we migrated them before they went to\nsleep. The sole difference is the previous cpu in the wakeup path, so we\noverride this.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-39-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "1be0bd77c5dd7c903f46abf52f9a3650face3c1d", "tree": "e8b2a6c742951e8a9f98a53a285ab85b79437e4d", "parents": [ "25cbbef1924299249756bc4030fcb2436c019813" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:29:15 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:45 2013 +0200" }, "message": "stop_machine: Introduce stop_two_cpus()\n\nIntroduce stop_two_cpus() in order to allow controlled swapping of two\ntasks. It repurposes the stop_machine() state machine but only stops\nthe two cpus which we can do with on-stack structures and avoid\nmachine wide synchronization issues.\n\nThe ordering of CPUs is important to avoid deadlocks. If unordered then\ntwo cpus calling stop_two_cpus on each other simultaneously would attempt\nto queue in the opposite order on each CPU causing an AB-BA style deadlock.\nBy always having the lowest number CPU doing the queueing of works, we can\nguarantee that works are always queued in the same order, and deadlocks\nare avoided.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\n[ Implemented deadlock avoidance. ]\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-38-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "25cbbef1924299249756bc4030fcb2436c019813", "tree": "9369e1194bf3ddc91a5298b04fe590845851565b", "parents": [ "4591ce4f2d22dc9de7a6719161ce409b5fd1caac" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:14 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:44 2013 +0200" }, "message": "mm: numa: Trap pmd hinting faults only if we would otherwise trap PTE faults\n\nBase page PMD faulting is meant to batch handle NUMA hinting faults from\nPTEs. However, even is no PTE faults would ever be handled within a\nrange the kernel still traps PMD hinting faults. This patch avoids the\noverhead.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-37-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "4591ce4f2d22dc9de7a6719161ce409b5fd1caac", "tree": "1ba60c69568eaa476960bd62e0f1ea2dc8b96a65", "parents": [ "06ea5e035b4e66cc77790457a89fc7e368060c4b" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:13 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:42 2013 +0200" }, "message": "sched/numa: Do not trap hinting faults for shared libraries\n\nNUMA hinting faults will not migrate a shared executable page mapped by\nmultiple processes on the grounds that the data is probably in the CPU\ncache already and the page may just bounce between tasks running on multipl\nnodes. Even if the migration is avoided, there is still the overhead of\ntrapping the fault, updating the statistics, making scheduler placement\ndecisions based on the information etc. If we are never going to migrate\nthe page, it is overhead for no gain and worse a process may be placed on\na sub-optimal node for shared executable pages. This patch avoids trapping\nfaults for shared libraries entirely.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-36-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "06ea5e035b4e66cc77790457a89fc7e368060c4b", "tree": "225dd9d1d180dcca19cce9ce8c3762b9c1cee7dc", "parents": [ "6b9a7460b6baf6c77fc3d23d927ddfc3f3f05bf3" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:12 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:41 2013 +0200" }, "message": "sched/numa: Increment numa_migrate_seq when task runs in correct location\n\nWhen a task is already running on its preferred node, increment\nnuma_migrate_seq to indicate that the task is settled if migration is\ntemporarily disabled, and memory should migrate towards it.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\n[ Only increment migrate_seq if migration temporarily disabled. ]\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-35-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "6b9a7460b6baf6c77fc3d23d927ddfc3f3f05bf3", "tree": "b35642ac95837435740ee6f584ca6d169795891d", "parents": [ "58d081b5082dd85e02ac9a1fb151d97395340a09" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:11 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:40 2013 +0200" }, "message": "sched/numa: Retry migration of tasks to CPU on a preferred node\n\nWhen a preferred node is selected for a tasks there is an attempt to migrate\nthe task to a CPU there. This may fail in which case the task will only\nmigrate if the active load balancer takes action. This may never happen if\nthe conditions are not right. This patch will check at NUMA hinting fault\ntime if another attempt should be made to migrate the task. It will only\nmake an attempt once every five seconds.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-34-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "58d081b5082dd85e02ac9a1fb151d97395340a09", "tree": "5bfc7ac630ba62d898dfb860e63f118079cba57d", "parents": [ "fc3147245d193bd0f57307859c698fa28a20b0fe" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:10 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:39 2013 +0200" }, "message": "sched/numa: Avoid overloading CPUs on a preferred NUMA node\n\nThis patch replaces find_idlest_cpu_node with task_numa_find_cpu.\nfind_idlest_cpu_node has two critical limitations. It does not take the\nscheduling class into account when calculating the load and it is unsuitable\nfor using when comparing loads between NUMA nodes.\n\ntask_numa_find_cpu uses similar load calculations to wake_affine() when\nselecting the least loaded CPU within a scheduling domain common to the\nsource and destimation nodes. It avoids causing CPU load imbalances in\nthe machine by refusing to migrate if the relative load on the target\nCPU is higher than the source CPU.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-33-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "fc3147245d193bd0f57307859c698fa28a20b0fe", "tree": "d1c6339ab156fc1f1419fe62164dfd1947cec14d", "parents": [ "6fe6b2d6dabf392aceb3ad3a5e859b46a04465c6" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:09 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:38 2013 +0200" }, "message": "mm: numa: Limit NUMA scanning to migrate-on-fault VMAs\n\nThere is a 90% regression observed with a large Oracle performance test\non a 4 node system. Profiles indicated that the overhead was due to\ncontention on sp_lock when looking up shared memory policies. These\npolicies do not have the appropriate flags to allow them to be\nautomatically balanced so trapping faults on them is pointless. This\npatch skips VMAs that do not have MPOL_F_MOF set.\n\n[riel@redhat.com: Initial patch]\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReported-and-tested-by: Joe Mario \u003cjmario@redhat.com\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-32-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "6fe6b2d6dabf392aceb3ad3a5e859b46a04465c6", "tree": "db4493950d94c418edcce093bd698e79ec1dca1a", "parents": [ "b795854b1fa70f6aee923ae5df74ff7afeaddcaa" ], "author": { "name": "Rik van Riel", "email": "riel@redhat.com", "time": "Mon Oct 07 11:29:08 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:36 2013 +0200" }, "message": "sched/numa: Do not migrate memory immediately after switching node\n\nThe load balancer can move tasks between nodes and does not take NUMA\nlocality into account. With automatic NUMA balancing this may result in the\ntasks working set being migrated to the new node. However, as the fault\nbuffer will still store faults from the old node the schduler may decide to\nreset the preferred node and migrate the task back resulting in more\nmigrations.\n\nThe ideal would be that the scheduler did not migrate tasks with a heavy\nmemory footprint but this may result nodes being overloaded. We could\nalso discard the fault information on task migration but this would still\ncause all the tasks working set to be migrated. This patch simply avoids\nmigrating the memory for a short time after a task is migrated.\n\nSigned-off-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-31-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "b795854b1fa70f6aee923ae5df74ff7afeaddcaa", "tree": "fd109d9f3778c7bc934fedb3cda2b5bfb1293375", "parents": [ "073b5beea735c7e1970686c94ff1f3aaac790a2a" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:07 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:35 2013 +0200" }, "message": "sched/numa: Set preferred NUMA node based on number of private faults\n\nIdeally it would be possible to distinguish between NUMA hinting faults that\nare private to a task and those that are shared. If treated identically\nthere is a risk that shared pages bounce between nodes depending on\nthe order they are referenced by tasks. Ultimately what is desirable is\nthat task private pages remain local to the task while shared pages are\ninterleaved between sharing tasks running on different nodes to give good\naverage performance. This is further complicated by THP as even\napplications that partition their data may not be partitioning on a huge\npage boundary.\n\nTo start with, this patch assumes that multi-threaded or multi-process\napplications partition their data and that in general the private accesses\nare more important for cpu-\u003ememory locality in the general case. Also,\nno new infrastructure is required to treat private pages properly but\ninterleaving for shared pages requires additional infrastructure.\n\nTo detect private accesses the pid of the last accessing task is required\nbut the storage requirements are a high. This patch borrows heavily from\nIngo Molnar\u0027s patch \"numa, mm, sched: Implement last-CPU+PID hash tracking\"\nto encode some bits from the last accessing task in the page flags as\nwell as the node information. Collisions will occur but it is better than\njust depending on the node information. Node information is then used to\ndetermine if a page needs to migrate. The PID information is used to detect\nprivate/shared accesses. The preferred NUMA node is selected based on where\nthe maximum number of approximately private faults were measured. Shared\nfaults are not taken into consideration for a few reasons.\n\nFirst, if there are many tasks sharing the page then they\u0027ll all move\ntowards the same node. The node will be compute overloaded and then\nscheduled away later only to bounce back again. Alternatively the shared\ntasks would just bounce around nodes because the fault information is\neffectively noise. Either way accounting for shared faults the same as\nprivate faults can result in lower performance overall.\n\nThe second reason is based on a hypothetical workload that has a small\nnumber of very important, heavily accessed private pages but a large shared\narray. The shared array would dominate the number of faults and be selected\nas a preferred node even though it\u0027s the wrong decision.\n\nThe third reason is that multiple threads in a process will race each\nother to fault the shared page making the fault information unreliable.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\n[ Fix complication error when !NUMA_BALANCING. ]\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-30-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "073b5beea735c7e1970686c94ff1f3aaac790a2a", "tree": "ba44188b0a4b52fddd1d0a7b9a6a214ea6212def", "parents": [ "1bc115d87dffd1c43bdc3c9c9d1e3a51c195d18e" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:06 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:33 2013 +0200" }, "message": "sched/numa: Remove check that skips small VMAs\n\ntask_numa_work skips small VMAs. At the time the logic was to reduce the\nscanning overhead which was considerable. It is a dubious hack at best.\nIt would make much more sense to cache where faults have been observed\nand only rescan those regions during subsequent PTE scans. Remove this\nhack as motivation to do it properly in the future.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-29-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "1bc115d87dffd1c43bdc3c9c9d1e3a51c195d18e", "tree": "56a26b4f4fe089e3dd1df5a26877d1e4c0114d35", "parents": [ "9ff1d9ff3c2c8ab3feaeb2e8056a07ca293f7bde" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:05 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:32 2013 +0200" }, "message": "mm: numa: Scan pages with elevated page_mapcount\n\nCurrently automatic NUMA balancing is unable to distinguish between false\nshared versus private pages except by ignoring pages with an elevated\npage_mapcount entirely. This avoids shared pages bouncing between the\nnodes whose task is using them but that is ignored quite a lot of data.\n\nThis patch kicks away the training wheels in preparation for adding support\nfor identifying shared/private pages is now in place. The ordering is so\nthat the impact of the shared/private detection can be easily measured. Note\nthat the patch does not migrate shared, file-backed within vmas marked\nVM_EXEC as these are generally shared library pages. Migrating such pages\nis not beneficial as there is an expectation they are read-shared between\ncaches and iTLB and iCache pressure is generally low.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-28-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "9ff1d9ff3c2c8ab3feaeb2e8056a07ca293f7bde", "tree": "678236662d8e0eb734dc7aeb011e50f81b7e9a06", "parents": [ "ac8e895bd260cb8bb19ade6a3abd44e7abe9a01d" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:04 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:31 2013 +0200" }, "message": "sched/numa: Check current-\u003emm before allocating NUMA faults\n\ntask_numa_placement checks current-\u003emm but after buffers for faults\nhave already been uselessly allocated. Move the check earlier.\n\n[peterz@infradead.org: Identified the problem]\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-27-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ac8e895bd260cb8bb19ade6a3abd44e7abe9a01d", "tree": "fe0d50baf0dad412fd7d5ba0286ce95e08a363ac", "parents": [ "e6628d5b0a2979f3e0ee6f7783ede5df50cb9ede" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:03 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:30 2013 +0200" }, "message": "sched/numa: Add infrastructure for split shared/private accounting of NUMA hinting faults\n\nIdeally it would be possible to distinguish between NUMA hinting faults\nthat are private to a task and those that are shared. This patch prepares\ninfrastructure for separately accounting shared and private faults by\nallocating the necessary buffers and passing in relevant information. For\nnow, all faults are treated as private and detection will be introduced\nlater.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-26-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "e6628d5b0a2979f3e0ee6f7783ede5df50cb9ede", "tree": "824c34aa911095ab568e12797db9150949068b8e", "parents": [ "7a0f308337d11fd5caa9f845c6d08cc5d6067988" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:02 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:28 2013 +0200" }, "message": "sched/numa: Reschedule task on preferred NUMA node once selected\n\nA preferred node is selected based on the node the most NUMA hinting\nfaults was incurred on. There is no guarantee that the task is running\non that node at the time so this patch rescheules the task to run on\nthe most idle CPU of the selected node when selected. This avoids\nwaiting for the balancer to make a decision.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-25-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "7a0f308337d11fd5caa9f845c6d08cc5d6067988", "tree": "91203c3f68fcc5d6a63dc7403fe43b1cc0fc3955", "parents": [ "3a7053b3224f4a8b0e8184166190076593621617" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:01 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:27 2013 +0200" }, "message": "sched/numa: Resist moving tasks towards nodes with fewer hinting faults\n\nJust as \"sched: Favour moving tasks towards the preferred node\" favours\nmoving tasks towards nodes with a higher number of recorded NUMA hinting\nfaults, this patch resists moving tasks towards nodes with lower faults.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-24-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "3a7053b3224f4a8b0e8184166190076593621617", "tree": "dfe404bfbc1306fccbc00f2177becf1482504e45", "parents": [ "745d61476ddb737aad3495fa6d9a8f8c2ee59f86" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:29:00 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:26 2013 +0200" }, "message": "sched/numa: Favour moving tasks towards the preferred node\n\nThis patch favours moving tasks towards NUMA node that recorded a higher\nnumber of NUMA faults during active load balancing. Ideally this is\nself-reinforcing as the longer the task runs on that node, the more faults\nit should incur causing task_numa_placement to keep the task running on that\nnode. In reality a big weakness is that the nodes CPUs can be overloaded\nand it would be more efficient to queue tasks on an idle node and migrate\nto the new node. This would require additional smarts in the balancer so\nfor now the balancer will simply prefer to place the task on the preferred\nnode for a PTE scans which is controlled by the numa_balancing_settle_count\nsysctl. Once the settle_count number of scans has complete the schedule\nis free to place the task on an alternative node if the load is imbalanced.\n\n[srikar@linux.vnet.ibm.com: Fixed statistics]\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\n[ Tunable and use higher faults instead of preferred. ]\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-23-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "745d61476ddb737aad3495fa6d9a8f8c2ee59f86", "tree": "ae6ebeefdcee8b5615f21e2f8a960f92fbdd2f71", "parents": [ "688b7585d16ab57a17aa4422a3b290b3a55fa679" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:59 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:25 2013 +0200" }, "message": "sched/numa: Update NUMA hinting faults once per scan\n\nNUMA hinting fault counts and placement decisions are both recorded in the\nsame array which distorts the samples in an unpredictable fashion. The values\nlinearly accumulate during the scan and then decay creating a sawtooth-like\npattern in the per-node counts. It also means that placement decisions are\ntime sensitive. At best it means that it is very difficult to state that\nthe buffer holds a decaying average of past faulting behaviour. At worst,\nit can confuse the load balancer if it sees one node with an artifically high\ncount due to very recent faulting activity and may create a bouncing effect.\n\nThis patch adds a second array. numa_faults stores the historical data\nwhich is used for placement decisions. numa_faults_buffer holds the\nfault activity during the current scan window. When the scan completes,\nnuma_faults decays and the values from numa_faults_buffer are copied\nacross.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-22-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "688b7585d16ab57a17aa4422a3b290b3a55fa679", "tree": "2a016719cf05336bf49db699a9268dae6b99310f", "parents": [ "f809ca9a554dda49fb264c79e31c722e0b063ff8" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:58 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:23 2013 +0200" }, "message": "sched/numa: Select a preferred node with the most numa hinting faults\n\nThis patch selects a preferred node for a task to run on based on the\nNUMA hinting faults. This information is later used to migrate tasks\ntowards the node during balancing.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-21-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "f809ca9a554dda49fb264c79e31c722e0b063ff8", "tree": "dbb0307d8d16c26421c1a08fc11e4bbff1d48f85", "parents": [ "f307cd1a32fab53012b01749a1f5ba10b0a7243f" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:57 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:22 2013 +0200" }, "message": "sched/numa: Track NUMA hinting faults on per-node basis\n\nThis patch tracks what nodes numa hinting faults were incurred on.\nThis information is later used to schedule a task on the node storing\nthe pages most frequently faulted by the task.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-20-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "f307cd1a32fab53012b01749a1f5ba10b0a7243f", "tree": "a3a1f1eb0b3550e730f533bc7c4deb77549bd370", "parents": [ "598f0ec0bc996e90a806ee9564af919ea5aad401" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:56 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:21 2013 +0200" }, "message": "sched/numa: Slow scan rate if no NUMA hinting faults are being recorded\n\nNUMA PTE scanning slows if a NUMA hinting fault was trapped and no page\nwas migrated. For long-lived but idle processes there may be no faults\nbut the scan rate will be high and just waste CPU. This patch will slow\nthe scan rate for processes that are not trapping faults.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-19-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "598f0ec0bc996e90a806ee9564af919ea5aad401", "tree": "9df97675a01340285b792be1909a41a02dbe905f", "parents": [ "7e8d16b6cbccb2f5da579f5085479fb82ba851b8" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:55 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:20 2013 +0200" }, "message": "sched/numa: Set the scan rate proportional to the memory usage of the task being scanned\n\nThe NUMA PTE scan rate is controlled with a combination of the\nnuma_balancing_scan_period_min, numa_balancing_scan_period_max and\nnuma_balancing_scan_size. This scan rate is independent of the size\nof the task and as an aside it is further complicated by the fact that\nnuma_balancing_scan_size controls how many pages are marked pte_numa and\nnot how much virtual memory is scanned.\n\nIn combination, it is almost impossible to meaningfully tune the min and\nmax scan periods and reasoning about performance is complex when the time\nto complete a full scan is is partially a function of the tasks memory\nsize. This patch alters the semantic of the min and max tunables to be\nabout tuning the length time it takes to complete a scan of a tasks occupied\nvirtual address space. Conceptually this is a lot easier to understand. There\nis a \"sanity\" check to ensure the scan rate is never extremely fast based on\nthe amount of virtual memory that should be scanned in a second. The default\nof 2.5G seems arbitrary but it is to have the maximum scan rate after the\npatch roughly match the maximum scan rate before the patch was applied.\n\nOn a similar note, numa_scan_period is in milliseconds and not\njiffies. Properly placed pages slow the scanning rate but adding 10 jiffies\nto numa_scan_period means that the rate scanning slows depends on HZ which\nis confusing. Get rid of the jiffies_to_msec conversion and treat it as ms.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-18-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "7e8d16b6cbccb2f5da579f5085479fb82ba851b8", "tree": "3f44d9f893531bed393017824b41fd3814e6e890", "parents": [ "b726b7dfb400c937546fa91cf8523dcb1aa2fc6e" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:54 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:19 2013 +0200" }, "message": "sched/numa: Initialise numa_next_scan properly\n\nScan delay logic and resets are currently initialised to start scanning\nimmediately instead of delaying properly. Initialise them properly at\nfork time and catch when a new mm has been allocated.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-17-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "b726b7dfb400c937546fa91cf8523dcb1aa2fc6e", "tree": "b4a6f741b630bb22568536860f3f46e94c4e8904", "parents": [ "9e645ab6d089f5822479a833c6977c785bcfffe3" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:53 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:17 2013 +0200" }, "message": "Revert \"mm: sched: numa: Delay PTE scanning until a task is scheduled on a new node\"\n\nPTE scanning and NUMA hinting fault handling is expensive so commit\n5bca2303 (\"mm: sched: numa: Delay PTE scanning until a task is scheduled\non a new node\") deferred the PTE scan until a task had been scheduled on\nanother node. The problem is that in the purely shared memory case that\nthis may never happen and no NUMA hinting fault information will be\ncaptured. We are not ruling out the possibility that something better\ncan be done here but for now, this patch needs to be reverted and depend\nentirely on the scan_delay to avoid punishing short-lived processes.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-16-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "9e645ab6d089f5822479a833c6977c785bcfffe3", "tree": "5f3de0a077588419e0a62e2eccd81c68a51c7de4", "parents": [ "19a78d110d7a8045aeb90d38ee8fe9743ce88c2d" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:28:52 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:40:09 2013 +0200" }, "message": "sched/numa: Continue PTE scanning even if migrate rate limited\n\nAvoiding marking PTEs pte_numa because a particular NUMA node is migrate rate\nlimited sees like a bad idea. Even if this node can\u0027t migrate anymore other\nnodes might and we want up-to-date information to do balance decisions.\nWe already rate limit the actual migrations, this should leave enough\nbandwidth to allow the non-migrating scanning. I think its important we\nkeep up-to-date information if we\u0027re going to do placement based on it.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-15-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "19a78d110d7a8045aeb90d38ee8fe9743ce88c2d", "tree": "722c68fda81970196b84e23126739d9fe00ca0b6", "parents": [ "a1a46184e34cfd0764f06a54870defa052b0a094" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:28:51 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:56 2013 +0200" }, "message": "sched/numa: Mitigate chance that same task always updates PTEs\n\nWith a trace_printk(\"working\\n\"); right after the cmpxchg in\ntask_numa_work() we can see that of a 4 thread process, its always the\nsame task winning the race and doing the protection change.\n\nThis is a problem since the task doing the protection change has a\npenalty for taking faults -- it is busy when marking the PTEs. If its\nalways the same task the -\u003enuma_faults[] get severely skewed.\n\nAvoid this by delaying the task doing the protection change such that\nit is unlikely to win the privilege again.\n\nBefore:\n\nroot@interlagos:~# grep \"thread 0/.*working\" /debug/tracing/trace | tail -15\n thread 0/0-3232 [022] .... 212.787402: task_numa_work: working\n thread 0/0-3232 [022] .... 212.888473: task_numa_work: working\n thread 0/0-3232 [022] .... 212.989538: task_numa_work: working\n thread 0/0-3232 [022] .... 213.090602: task_numa_work: working\n thread 0/0-3232 [022] .... 213.191667: task_numa_work: working\n thread 0/0-3232 [022] .... 213.292734: task_numa_work: working\n thread 0/0-3232 [022] .... 213.393804: task_numa_work: working\n thread 0/0-3232 [022] .... 213.494869: task_numa_work: working\n thread 0/0-3232 [022] .... 213.596937: task_numa_work: working\n thread 0/0-3232 [022] .... 213.699000: task_numa_work: working\n thread 0/0-3232 [022] .... 213.801067: task_numa_work: working\n thread 0/0-3232 [022] .... 213.903155: task_numa_work: working\n thread 0/0-3232 [022] .... 214.005201: task_numa_work: working\n thread 0/0-3232 [022] .... 214.107266: task_numa_work: working\n thread 0/0-3232 [022] .... 214.209342: task_numa_work: working\n\nAfter:\n\nroot@interlagos:~# grep \"thread 0/.*working\" /debug/tracing/trace | tail -15\n thread 0/0-3253 [005] .... 136.865051: task_numa_work: working\n thread 0/2-3255 [026] .... 136.965134: task_numa_work: working\n thread 0/3-3256 [024] .... 137.065217: task_numa_work: working\n thread 0/3-3256 [024] .... 137.165302: task_numa_work: working\n thread 0/3-3256 [024] .... 137.265382: task_numa_work: working\n thread 0/0-3253 [004] .... 137.366465: task_numa_work: working\n thread 0/2-3255 [026] .... 137.466549: task_numa_work: working\n thread 0/0-3253 [004] .... 137.566629: task_numa_work: working\n thread 0/0-3253 [004] .... 137.666711: task_numa_work: working\n thread 0/1-3254 [028] .... 137.766799: task_numa_work: working\n thread 0/0-3253 [004] .... 137.866876: task_numa_work: working\n thread 0/2-3255 [026] .... 137.966960: task_numa_work: working\n thread 0/1-3254 [028] .... 138.067041: task_numa_work: working\n thread 0/2-3255 [026] .... 138.167123: task_numa_work: working\n thread 0/3-3256 [024] .... 138.267207: task_numa_work: working\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-14-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "a1a46184e34cfd0764f06a54870defa052b0a094", "tree": "b4d12f1d23963fb3de14cf8d81f7979c4f58d95a", "parents": [ "f123d74abf91574837d14e5ea58f6a779a387bf5" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:50 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:50 2013 +0200" }, "message": "mm: numa: Do not migrate or account for hinting faults on the zero page\n\nThe zero page is not replicated between nodes and is often shared between\nprocesses. The data is read-only and likely to be cached in local CPUs\nif heavily accessed meaning that the remote memory access cost is less\nof a concern. This patch prevents trapping faults on the zero pages. For\ntasks using the zero page this will reduce the number of PTE updates,\nTLB flushes and hinting faults.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\n[ Correct use of is_huge_zero_page]\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-13-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "f123d74abf91574837d14e5ea58f6a779a387bf5", "tree": "4347ff38dea6a74a53d811e63c21e805a5d25cfd", "parents": [ "e920e14ca29b0b2a981cfc90e4e20edd6f078d19" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:49 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:49 2013 +0200" }, "message": "mm: Only flush TLBs if a transhuge PMD is modified for NUMA pte scanning\n\nNUMA PTE scanning is expensive both in terms of the scanning itself and\nthe TLB flush if there are any updates. The TLB flush is avoided if no\nPTEs are updated but there is a bug where transhuge PMDs are considered\nto be updated even if they were already pmd_numa. This patch addresses\nthe problem and TLB flushes should be reduced.\n\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-12-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "e920e14ca29b0b2a981cfc90e4e20edd6f078d19", "tree": "e8f8530aaddb664bc811289aa0e096890a40dd2e", "parents": [ "afcae2655b0ab67e65f161b1bb214efcfa1db415" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:48 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:48 2013 +0200" }, "message": "mm: Do not flush TLB during protection change if !pte_present \u0026\u0026 !migration_entry\n\nNUMA PTE scanning is expensive both in terms of the scanning itself and\nthe TLB flush if there are any updates. Currently non-present PTEs are\naccounted for as an update and incurring a TLB flush where it is only\nnecessary for anonymous migration entries. This patch addresses the\nproblem and should reduce TLB flushes.\n\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-11-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "afcae2655b0ab67e65f161b1bb214efcfa1db415", "tree": "795ef14f1193117c0b0fd126e12089f0fc4994d0", "parents": [ "a54a407fbf7735fd8f7841375574f5d9b0375f93" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:47 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:46 2013 +0200" }, "message": "mm: Account for a THP NUMA hinting update as one PTE update\n\nA THP PMD update is accounted for as 512 pages updated in vmstat. This is\nlarge difference when estimating the cost of automatic NUMA balancing and\ncan be misleading when comparing results that had collapsed versus split\nTHP. This patch addresses the accounting issue.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-10-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "a54a407fbf7735fd8f7841375574f5d9b0375f93", "tree": "29a4af92708dfc99f1693e9a313b53474d50c496", "parents": [ "8191acbd30c73e45c24ad16c372e0b42cc7ac8f8" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:46 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:45 2013 +0200" }, "message": "mm: Close races between THP migration and PMD numa clearing\n\nTHP migration uses the page lock to guard against parallel allocations\nbut there are cases like this still open\n\n Task A\t\t\t\t\tTask B\n ---------------------\t\t\t\t---------------------\n do_huge_pmd_numa_page\t\t\t\tdo_huge_pmd_numa_page\n lock_page\n mpol_misplaced \u003d\u003d -1\n unlock_page\n goto clear_pmdnuma\n\t\t\t\t\t\tlock_page\n\t\t\t\t\t\tmpol_misplaced \u003d\u003d 2\n\t\t\t\t\t\tmigrate_misplaced_transhuge\n pmd \u003d pmd_mknonnuma\n set_pmd_at\n\nDuring hours of testing, one crashed with weird errors and while I have\nno direct evidence, I suspect something like the race above happened.\nThis patch extends the page lock to being held until the pmd_numa is\ncleared to prevent migration starting in parallel while the pmd_numa is\nbeing cleared. It also flushes the old pmd entry and orders pagetable\ninsertion before rmap insertion.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-9-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "8191acbd30c73e45c24ad16c372e0b42cc7ac8f8", "tree": "1c558203ad1f7e141ef3e587d5d3d846ac82e0ec", "parents": [ "b8916634b77bffb233d8f2f45703c80343457cc1" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:45 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:44 2013 +0200" }, "message": "mm: numa: Sanitize task_numa_fault() callsites\n\nThere are three callers of task_numa_fault():\n\n - do_huge_pmd_numa_page():\n Accounts against the current node, not the node where the\n page resides, unless we migrated, in which case it accounts\n against the node we migrated to.\n\n - do_numa_page():\n Accounts against the current node, not the node where the\n page resides, unless we migrated, in which case it accounts\n against the node we migrated to.\n\n - do_pmd_numa_page():\n Accounts not at all when the page isn\u0027t migrated, otherwise\n accounts against the node we migrated towards.\n\nThis seems wrong to me; all three sites should have the same\nsementaics, furthermore we should accounts against where the page\nreally is, we already know where the task is.\n\nSo modify all three sites to always account; we did after all receive\nthe fault; and always account to where the page is after migration,\nregardless of success.\n\nThey all still differ on when they clear the PTE/PMD; ideally that\nwould get sorted too.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-8-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "b8916634b77bffb233d8f2f45703c80343457cc1", "tree": "e2452c3d74d14ef387b9cce460f3b5597eca3676", "parents": [ "ff9042b11a71c81238c70af168cd36b98a6d5a3c" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:44 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:43 2013 +0200" }, "message": "mm: Prevent parallel splits during THP migration\n\nTHP migrations are serialised by the page lock but on its own that does\nnot prevent THP splits. If the page is split during THP migration then\nthe pmd_same checks will prevent page table corruption but the unlock page\nand other fix-ups potentially will cause corruption. This patch takes the\nanon_vma lock to prevent parallel splits during migration.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-7-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "ff9042b11a71c81238c70af168cd36b98a6d5a3c", "tree": "26a09b0c1e8bf5cd091bed61475c6e0433571924", "parents": [ "0c3a775e1e0b069bf765f8355b723ce0d18dcc6c" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:43 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:41 2013 +0200" }, "message": "mm: Wait for THP migrations to complete during NUMA hinting faults\n\nThe locking for migrating THP is unusual. While normal page migration\nprevents parallel accesses using a migration PTE, THP migration relies on\na combination of the page_table_lock, the page lock and the existance of\nthe NUMA hinting PTE to guarantee safety but there is a bug in the scheme.\n\nIf a THP page is currently being migrated and another thread traps a\nfault on the same page it checks if the page is misplaced. If it is not,\nthen pmd_numa is cleared. The problem is that it checks if the page is\nmisplaced without holding the page lock meaning that the racing thread\ncan be migrating the THP when the second thread clears the NUMA bit\nand faults a stale page.\n\nThis patch checks if the page is potentially being migrated and stalls\nusing the lock_page if it is potentially being migrated before checking\nif the page is misplaced or not.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-6-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "0c3a775e1e0b069bf765f8355b723ce0d18dcc6c", "tree": "c71bab366ceb81cc198f06b3d492aa8c62ea10cb", "parents": [ "c69307d533d7aa7cc8894dbbb8a274599f8630d7" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:42 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:40 2013 +0200" }, "message": "mm: numa: Do not account for a hinting fault if we raced\n\nIf another task handled a hinting fault in parallel then do not double\naccount for it.\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-5-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "c69307d533d7aa7cc8894dbbb8a274599f8630d7", "tree": "874de2fc649ede70e52cf80e4e2fbb22ab5e9f84", "parents": [ "10fc05d0e551146ad6feb0ab8902d28a2d3c5624" ], "author": { "name": "Peter Zijlstra", "email": "peterz@infradead.org", "time": "Mon Oct 07 11:28:41 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:30 2013 +0200" }, "message": "sched/numa: Fix comments\n\nFix a 80 column violation and a PTE vs PMD reference.\n\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-4-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "10fc05d0e551146ad6feb0ab8902d28a2d3c5624", "tree": "3beae144163fae0d9aea50ffb5ad8aaa7a57fc39", "parents": [ "37bf06375c90a42fe07b9bebdb07bc316ae5a0ce" ], "author": { "name": "Mel Gorman", "email": "mgorman@suse.de", "time": "Mon Oct 07 11:28:40 2013 +0100" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:39:20 2013 +0200" }, "message": "mm: numa: Document automatic NUMA balancing sysctls\n\nSigned-off-by: Mel Gorman \u003cmgorman@suse.de\u003e\nReviewed-by: Rik van Riel \u003criel@redhat.com\u003e\nCc: Andrea Arcangeli \u003caarcange@redhat.com\u003e\nCc: Johannes Weiner \u003channes@cmpxchg.org\u003e\nCc: Srikar Dronamraju \u003csrikar@linux.vnet.ibm.com\u003e\nSigned-off-by: Peter Zijlstra \u003cpeterz@infradead.org\u003e\nLink: http://lkml.kernel.org/r/1381141781-10992-3-git-send-email-mgorman@suse.de\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "37bf06375c90a42fe07b9bebdb07bc316ae5a0ce", "tree": "de572dd6d3955b0725001776a7b03796f99e1e8e", "parents": [ "6bfa687c19b7ab8adee03f0d43c197c2945dd869", "d0e639c9e06d44e713170031fe05fb60ebe680af" ], "author": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:36:13 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Wed Oct 09 12:36:13 2013 +0200" }, "message": "Merge tag \u0027v3.12-rc4\u0027 into sched/core\n\nMerge Linux v3.12-rc4 to fix a conflict and also to refresh the tree\nbefore applying more scheduler patches.\n\nConflicts:\n\tarch/avr32/include/asm/Kbuild\n\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "0e7a3ed04f0cd4311096d691888f88569310ee6c", "tree": "f7d246dd2a1f27cc64dda7758a1306ccb6e70129", "parents": [ "8b5ede69d24db939f52b47effff2f6fe1e83e08b", "1651d120baddf68846188bcdf0ef5350068436cf" ], "author": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Tue Oct 08 09:23:12 2013 -0700" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Tue Oct 08 09:23:12 2013 -0700" }, "message": "Merge branch \u0027perf-urgent-for-linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip\n\nPull perf fixes from Ingo Molnar:\n \"Various fixlets:\n\n On the kernel side:\n\n - fix a race\n - fix a bug in the handling of the perf ring-buffer data page\n\n On the tooling side:\n\n - fix the handling of certain corrupted perf.data files\n - fix a bug in \u0027perf probe\u0027\n - fix a bug in \u0027perf record + perf sched\u0027\n - fix a bug in \u0027make install\u0027\n - fix a bug in libaudit feature-detection on certain distros\"\n\n* \u0027perf-urgent-for-linus\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:\n perf session: Fix infinite loop on invalid perf.data file\n perf tools: Fix installation of libexec components\n perf probe: Fix to find line information for probe list\n perf tools: Fix libaudit test\n perf stat: Set child_pid after perf_evlist__prepare_workload()\n perf tools: Add default handler for mmap2 events\n perf/x86: Clean up cap_user_time* setting\n perf: Fix perf_pmu_migrate_context\n" }, { "commit": "1651d120baddf68846188bcdf0ef5350068436cf", "tree": "9316fed404c26d55ae87ae6b68b4842e78c820b9", "parents": [ "d8b11a0cbd1c66ce283eb9dabe0498dfa6483f32", "b314e5cfd11fd78545ce6c2be42646254390c1aa" ], "author": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 08 07:29:09 2013 +0200" }, "committer": { "name": "Ingo Molnar", "email": "mingo@kernel.org", "time": "Tue Oct 08 07:29:09 2013 +0200" }, "message": "Merge tag \u0027perf-urgent-for-mingo\u0027 of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent\n\nPull perf/urgent fixes from Arnaldo Carvalho de Melo:\n\n * The libaudit test was failing in some systems due to a unescaped newline, fix\n it so that the \u0027trace\u0027 tool can be built in such systems.\n\n * Fix installation of libexec components.\n\n * Add default handler for mmap2 events so that tools that don\u0027t explicitely\n define an MMAP2 handler don\u0027t crash, fix from David Ahern.\n\n * Fix to find line information for probe list, from Masami Hiramatsu.\n\n * Set child_pid after perf_evlist__prepare_workload(), fix from Namhyung Kim.\n\n * Fix infinite loop on invalid perf.data file, from Namhyung Kim.\n\nSigned-off-by: Arnaldo Carvalho de Melo \u003cacme@redhat.com\u003e\nSigned-off-by: Ingo Molnar \u003cmingo@kernel.org\u003e\n" }, { "commit": "8b5ede69d24db939f52b47effff2f6fe1e83e08b", "tree": "c492c191c688296cdd708e1d9e5fc6cc63197edf", "parents": [ "fd848319e751739a93aa9fc8182e57e87c5a0df1" ], "author": { "name": "Benjamin Herrenschmidt", "email": "benh@kernel.crashing.org", "time": "Tue Oct 08 08:08:24 2013 +1100" }, "committer": { "name": "Linus Torvalds", "email": "torvalds@linux-foundation.org", "time": "Mon Oct 07 14:19:39 2013 -0700" }, "message": "powerpc/irq: Don\u0027t switch to irq stack from softirq stack\n\nirq_exit() is now called on the irq stack, which can trigger a switch to\nthe softirq stack from the irq stack. If an interrupt happens at that\npoint, we will not properly detect the re-entrancy and clobber the\noriginal return context on the irq stack.\n\nThis fixes it. The side effect is to prevent all nesting from softirq\nstack to irq stack even in the \"safe\" case but it\u0027s simpler that way and\nmatches what x86_64 does.\n\nReported-by: Cédric Le Goater \u003cclg@fr.ibm.com\u003e\nTested-by: Cédric Le Goater \u003cclg@fr.ibm.com\u003e\nSigned-off-by: Benjamin Herrenschmidt \u003cbenh@kernel.crashing.org\u003e\nSigned-off-by: Linus Torvalds \u003ctorvalds@linux-foundation.org\u003e\n" } ], "next": "fd848319e751739a93aa9fc8182e57e87c5a0df1" }