| Message ID | 1588922717-63697-1-git-send-email-feng.tang@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | make vm_committed_as_batch aware of vm overcommit policy | expand |
On Tue, May 26, 2020 at 02:14:59PM -0400, Qian Cai wrote: > On Thu, May 21, 2020 at 05:27:26PM -0400, Qian Cai wrote: > > On Fri, May 08, 2020 at 03:25:14PM +0800, Feng Tang wrote: > > > When checking a performance change for will-it-scale scalability > > > mmap test [1], we found very high lock contention for spinlock of > > > percpu counter 'vm_committed_as': > > > > > > 94.14% 0.35% [kernel.kallsyms] [k] _raw_spin_lock_irqsave > > > 48.21% _raw_spin_lock_irqsave;percpu_counter_add_batch;__vm_enough_memory;mmap_region;do_mmap; > > > 45.91% _raw_spin_lock_irqsave;percpu_counter_add_batch;__do_munmap; > > > > > > Actually this heavy lock contention is not always necessary. The > > > 'vm_committed_as' needs to be very precise when the strict > > > OVERCOMMIT_NEVER policy is set, which requires a rather small batch > > > number for the percpu counter. > > > > > > So lift the batch number to 16X for OVERCOMMIT_ALWAYS and > > > OVERCOMMIT_GUESS policies, and add a sysctl handler to adjust it > > > when the policy is reconfigured. > > > > > > Benchmark with the same testcase in [1] shows 53% improvement on a > > > 8C/16T desktop, and 2097%(20X) on a 4S/72C/144T server. And no change > > > for some platforms, due to the test mmap size of the case is bigger > > > than the batch number computed, though the patch will help mmap/munmap > > > generally. > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > There are some style complain from checkpatch for patch 3, as it > > > follows the similar format of sibling functions > > > > > > patch1: a cleanup for /proc/meminfo > > > patch2: a preparation patch which also improve the accuracy of > > > vm_memory_committed > > > patch3: the main change > > > > > > Please help to review, thanks! > > > > Reverted this series fixed a warning under memory pressue. > > Andrew, Stephen, can you drop this series? Hope you're happy now. The warning is just not needed for the always case. The whole point of the patch was to maintain these counters only approximately. We could of course use _sum or _sum_positive in the warning, but that would just be unnecessary overhead in this moderately hot path. So this patch should fix it: (untested) Possibly might need to move it even further down because the patch also affected the _GUESS case. diff --git a/mm/util.c b/mm/util.c index 988d11e6c17c..fa78f90e29a1 100644 --- a/mm/util.c +++ b/mm/util.c @@ -798,10 +798,6 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) { long allowed; - VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < - -(s64)vm_committed_as_batch * num_online_cpus(), - "memory commitment underflow"); - vm_acct_memory(pages); /* @@ -810,6 +806,10 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) if (sysctl_overcommit_memory == OVERCOMMIT_ALWAYS) return 0; + VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < + -(s64)vm_committed_as_batch * num_online_cpus(), + "memory commitment underflow"); + if (sysctl_overcommit_memory == OVERCOMMIT_GUESS) { if (pages > totalram_pages() + total_swap_pages) goto error; > > > > > [ 3319.257898] LTP: starting oom01 > > [ 3319.284417] ------------[ cut here ]------------ > > [ 3319.284439] memory commitment underflow > > [ 3319.284456] WARNING: CPU: 13 PID: 130949 at mm/util.c:835 __vm_enough_memory+0x1dc/0x1f0 > > [ 3319.284485] Modules linked in: brd ext4 crc16 mbcache jbd2 loop kvm_hv kvm ip_tables x_tables xfs sd_mod ahci libahci bnx2x libata mdio tg3 libphy firmware_class dm_mirror dm_region_hash dm_log dm_mod > > [ 3319.284538] CPU: 13 PID: 130949 Comm: oom01 Not tainted 5.7.0-rc6-next-20200521+ #115 > > [ 3319.284551] NIP: c00000000043027c LR: c000000000430278 CTR: c0000000007bd120 > > [ 3319.284572] REGS: c00000064a42f710 TRAP: 0700 Not tainted (5.7.0-rc6-next-20200521+) > > [ 3319.284602] MSR: 900000000282b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28222422 XER: 20040000 > > [ 3319.284638] CFAR: c000000000119098 IRQMASK: 0 > > GPR00: c000000000430278 c00000064a42f9a0 c000000001765500 000000000000001b > > GPR04: c00000000179c3b0 000000000000000b 0000000066726564 c00000000429ac28 > > GPR08: 0000000000000001 0000000000000000 0000000000000000 c000001301facc80 > > GPR12: 0000000000002000 c000001fffff3d80 0000000000000000 0000000000000000 > > GPR16: 0000000000000000 0000000000000000 c0000014f358dd60 fffffffffff7dfff > > GPR20: c0000003725a9128 c0000003725a3928 c000000b0a0c0670 c000000b0a0c0678 > > GPR24: c000000b0a0c0660 c00000064a42fa58 c000000b0a0c0650 c0000003725a9080 > > GPR28: 0000000000000001 0000000000000000 c000000001620178 0000000000000001 > > [ 3319.284825] NIP [c00000000043027c] __vm_enough_memory+0x1dc/0x1f0 > > [ 3319.284853] LR [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 > > [ 3319.284872] Call Trace: > > [ 3319.284889] [c00000064a42f9a0] [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 (unreliable) > > [ 3319.284930] [c00000064a42fa30] [c000000000114258] dup_mm+0x2a8/0x700 > > [ 3319.284960] [c00000064a42fae0] [c000000000115c7c] copy_process+0xeac/0x1a00 > > [ 3319.284981] [c00000064a42fbf0] [c000000000116b3c] _do_fork+0xac/0xce0 > > [ 3319.285002] [c00000064a42fd00] [c000000000117928] __do_sys_clone+0x98/0xe0 > > [ 3319.285032] [c00000064a42fdc0] [c000000000039e78] system_call_exception+0x108/0x1d0 > > [ 3319.285064] [c00000064a42fe20] [c00000000000c9f0] system_call_common+0xf0/0x278 > > [ 3319.285093] Instruction dump: > > [ 3319.285109] 60000000 73a90001 4182000c e8be0002 4bfffeb8 3c62ff51 39200001 3d42fff9 > > [ 3319.285142] 38636690 992a6ad6 4bce8dbd 60000000 <0fe00000> e8be0002 4bfffe90 60000000 > > [ 3319.285166] irq event stamp: 0 > > [ 3319.285183] hardirqs last enabled at (0): [<0000000000000000>] 0x0 > > [ 3319.285195] hardirqs last disabled at (0): [<c0000000001154c8>] copy_process+0x6f8/0x1a00 > > [ 3319.285216] softirqs last enabled at (0): [<c0000000001154c8>] copy_process+0x6f8/0x1a00 > > [ 3319.285238] softirqs last disabled at (0): [<0000000000000000>] 0x0 > > [ 3319.285267] ---[ end trace 115e3f7fbb0653c2 ]--- > > > > > > > > - Feng > > > > > > ---------------------------------------------------------------- > > > Changelog: > > > v2: > > > * add the sysctl handler to cover runtime overcommit policy > > > change, as suggested by Andres Morton > > > * address the accuracy concern of vm_memory_committed() > > > from Andi Kleen > > > > > > Feng Tang (3): > > > proc/meminfo: avoid open coded reading of vm_committed_as > > > mm/util.c: make vm_memory_committed() more accurate > > > mm: adjust vm_committed_as_batch according to vm overcommit policy > > > > > > fs/proc/meminfo.c | 2 +- > > > include/linux/mm.h | 2 ++ > > > include/linux/mman.h | 4 ++++ > > > kernel/sysctl.c | 2 +- > > > mm/mm_init.c | 19 +++++++++++++++---- > > > mm/util.c | 15 ++++++++++++++- > > > 6 files changed, 37 insertions(+), 7 deletions(-) > > > > > > -- > > > 2.7.4 > > > > > >
On Tue, May 26, 2020 at 06:14:13PM -0700, Andi Kleen wrote: > On Tue, May 26, 2020 at 02:14:59PM -0400, Qian Cai wrote: > > On Thu, May 21, 2020 at 05:27:26PM -0400, Qian Cai wrote: > > > On Fri, May 08, 2020 at 03:25:14PM +0800, Feng Tang wrote: > > > > When checking a performance change for will-it-scale scalability > > > > mmap test [1], we found very high lock contention for spinlock of > > > > percpu counter 'vm_committed_as': > > > > > > > > 94.14% 0.35% [kernel.kallsyms] [k] _raw_spin_lock_irqsave > > > > 48.21% _raw_spin_lock_irqsave;percpu_counter_add_batch;__vm_enough_memory;mmap_region;do_mmap; > > > > 45.91% _raw_spin_lock_irqsave;percpu_counter_add_batch;__do_munmap; > > > > > > > > Actually this heavy lock contention is not always necessary. The > > > > 'vm_committed_as' needs to be very precise when the strict > > > > OVERCOMMIT_NEVER policy is set, which requires a rather small batch > > > > number for the percpu counter. > > > > > > > > So lift the batch number to 16X for OVERCOMMIT_ALWAYS and > > > > OVERCOMMIT_GUESS policies, and add a sysctl handler to adjust it > > > > when the policy is reconfigured. > > > > > > > > Benchmark with the same testcase in [1] shows 53% improvement on a > > > > 8C/16T desktop, and 2097%(20X) on a 4S/72C/144T server. And no change > > > > for some platforms, due to the test mmap size of the case is bigger > > > > than the batch number computed, though the patch will help mmap/munmap > > > > generally. > > > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > > > There are some style complain from checkpatch for patch 3, as it > > > > follows the similar format of sibling functions > > > > > > > > patch1: a cleanup for /proc/meminfo > > > > patch2: a preparation patch which also improve the accuracy of > > > > vm_memory_committed > > > > patch3: the main change > > > > > > > > Please help to review, thanks! > > > > > > Reverted this series fixed a warning under memory pressue. > > > > Andrew, Stephen, can you drop this series? > > Hope you're happy now. > > The warning is just not needed for the always case. The whole point > of the patch was to maintain these counters only approximately. > We could of course use _sum or _sum_positive in the warning, but that would > just be unnecessary overhead in this moderately hot path. > > So this patch should fix it: (untested) > > Possibly might need to move it even further down because the patch > also affected the _GUESS case. Thanks Andi for the patch. And I agree that it could be after _GUESS case, as existig check for _GUESS is loose: if (pages > totalram_pages() + total_swap_pages) goto error; Thanks, Feng > > > diff --git a/mm/util.c b/mm/util.c > index 988d11e6c17c..fa78f90e29a1 100644 > --- a/mm/util.c > +++ b/mm/util.c > @@ -798,10 +798,6 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) > { > long allowed; > > - VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < > - -(s64)vm_committed_as_batch * num_online_cpus(), > - "memory commitment underflow"); > - > vm_acct_memory(pages); > > /* > @@ -810,6 +806,10 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) > if (sysctl_overcommit_memory == OVERCOMMIT_ALWAYS) > return 0; > > + VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < > + -(s64)vm_committed_as_batch * num_online_cpus(), > + "memory commitment underflow"); > + > if (sysctl_overcommit_memory == OVERCOMMIT_GUESS) { > if (pages > totalram_pages() + total_swap_pages) > goto error;
Hi Qian, On Tue, May 26, 2020 at 02:14:59PM -0400, Qian Cai wrote: > On Thu, May 21, 2020 at 05:27:26PM -0400, Qian Cai wrote: > > On Fri, May 08, 2020 at 03:25:14PM +0800, Feng Tang wrote: > > > When checking a performance change for will-it-scale scalability > > > mmap test [1], we found very high lock contention for spinlock of > > > percpu counter 'vm_committed_as': > > > > > > 94.14% 0.35% [kernel.kallsyms] [k] _raw_spin_lock_irqsave > > > 48.21% _raw_spin_lock_irqsave;percpu_counter_add_batch;__vm_enough_memory;mmap_region;do_mmap; > > > 45.91% _raw_spin_lock_irqsave;percpu_counter_add_batch;__do_munmap; > > > > > > Actually this heavy lock contention is not always necessary. The > > > 'vm_committed_as' needs to be very precise when the strict > > > OVERCOMMIT_NEVER policy is set, which requires a rather small batch > > > number for the percpu counter. > > > > > > So lift the batch number to 16X for OVERCOMMIT_ALWAYS and > > > OVERCOMMIT_GUESS policies, and add a sysctl handler to adjust it > > > when the policy is reconfigured. > > > > > > Benchmark with the same testcase in [1] shows 53% improvement on a > > > 8C/16T desktop, and 2097%(20X) on a 4S/72C/144T server. And no change > > > for some platforms, due to the test mmap size of the case is bigger > > > than the batch number computed, though the patch will help mmap/munmap > > > generally. > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > Reverted this series fixed a warning under memory pressue. > > Andrew, Stephen, can you drop this series? > > > > > [ 3319.257898] LTP: starting oom01 > > [ 3319.284417] ------------[ cut here ]------------ > > [ 3319.284439] memory commitment underflow Thanks for the catch! Could you share the info about the platform, like the CPU numbers and RAM size, and what's the mmap test size of your test program. It would be great if you can point me the link to the test program. Thanks, Feng > > [ 3319.284456] WARNING: CPU: 13 PID: 130949 at mm/util.c:835 __vm_enough_memory+0x1dc/0x1f0 > > [ 3319.284485] Modules linked in: brd ext4 crc16 mbcache jbd2 loop kvm_hv kvm ip_tables x_tables xfs sd_mod ahci libahci bnx2x libata mdio tg3 libphy firmware_class dm_mirror dm_region_hash dm_log dm_mod > > [ 3319.284538] CPU: 13 PID: 130949 Comm: oom01 Not tainted 5.7.0-rc6-next-20200521+ #115 > > [ 3319.284551] NIP: c00000000043027c LR: c000000000430278 CTR: c0000000007bd120 > > [ 3319.284572] REGS: c00000064a42f710 TRAP: 0700 Not tainted (5.7.0-rc6-next-20200521+) > > [ 3319.284602] MSR: 900000000282b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28222422 XER: 20040000 > > [ 3319.284638] CFAR: c000000000119098 IRQMASK: 0 > > GPR00: c000000000430278 c00000064a42f9a0 c000000001765500 000000000000001b > > GPR04: c00000000179c3b0 000000000000000b 0000000066726564 c00000000429ac28 > > GPR08: 0000000000000001 0000000000000000 0000000000000000 c000001301facc80 > > GPR12: 0000000000002000 c000001fffff3d80 0000000000000000 0000000000000000 > > GPR16: 0000000000000000 0000000000000000 c0000014f358dd60 fffffffffff7dfff > > GPR20: c0000003725a9128 c0000003725a3928 c000000b0a0c0670 c000000b0a0c0678 > > GPR24: c000000b0a0c0660 c00000064a42fa58 c000000b0a0c0650 c0000003725a9080 > > GPR28: 0000000000000001 0000000000000000 c000000001620178 0000000000000001 > > [ 3319.284825] NIP [c00000000043027c] __vm_enough_memory+0x1dc/0x1f0 > > [ 3319.284853] LR [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 > > [ 3319.284872] Call Trace: > > [ 3319.284889] [c00000064a42f9a0] [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 (unreliable) > > [ 3319.284930] [c00000064a42fa30] [c000000000114258] dup_mm+0x2a8/0x700 > > [ 3319.284960] [c00000064a42fae0] [c000000000115c7c] copy_process+0xeac/0x1a00 > > [ 3319.284981] [c00000064a42fbf0] [c000000000116b3c] _do_fork+0xac/0xce0 > > [ 3319.285002] [c00000064a42fd00] [c000000000117928] __do_sys_clone+0x98/0xe0 > > [ 3319.285032] [c00000064a42fdc0] [c000000000039e78] system_call_exception+0x108/0x1d0 > > [ 3319.285064] [c00000064a42fe20] [c00000000000c9f0] system_call_common+0xf0/0x278 > > [ 3319.285093] Instruction dump: > > [ 3319.285109] 60000000 73a90001 4182000c e8be0002 4bfffeb8 3c62ff51 39200001 3d42fff9 > > [ 3319.285142] 38636690 992a6ad6 4bce8dbd 60000000 <0fe00000> e8be0002 4bfffe90 60000000
On Wed, May 27, 2020 at 09:46:47AM +0800, Feng Tang wrote: > Hi Qian, > > On Tue, May 26, 2020 at 02:14:59PM -0400, Qian Cai wrote: > > On Thu, May 21, 2020 at 05:27:26PM -0400, Qian Cai wrote: > > > On Fri, May 08, 2020 at 03:25:14PM +0800, Feng Tang wrote: > > > > When checking a performance change for will-it-scale scalability > > > > mmap test [1], we found very high lock contention for spinlock of > > > > percpu counter 'vm_committed_as': > > > > > > > > 94.14% 0.35% [kernel.kallsyms] [k] _raw_spin_lock_irqsave > > > > 48.21% _raw_spin_lock_irqsave;percpu_counter_add_batch;__vm_enough_memory;mmap_region;do_mmap; > > > > 45.91% _raw_spin_lock_irqsave;percpu_counter_add_batch;__do_munmap; > > > > > > > > Actually this heavy lock contention is not always necessary. The > > > > 'vm_committed_as' needs to be very precise when the strict > > > > OVERCOMMIT_NEVER policy is set, which requires a rather small batch > > > > number for the percpu counter. > > > > > > > > So lift the batch number to 16X for OVERCOMMIT_ALWAYS and > > > > OVERCOMMIT_GUESS policies, and add a sysctl handler to adjust it > > > > when the policy is reconfigured. > > > > > > > > Benchmark with the same testcase in [1] shows 53% improvement on a > > > > 8C/16T desktop, and 2097%(20X) on a 4S/72C/144T server. And no change > > > > for some platforms, due to the test mmap size of the case is bigger > > > > than the batch number computed, though the patch will help mmap/munmap > > > > generally. > > > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > Reverted this series fixed a warning under memory pressue. > > > > Andrew, Stephen, can you drop this series? > > > > > > > > [ 3319.257898] LTP: starting oom01 > > > [ 3319.284417] ------------[ cut here ]------------ > > > [ 3319.284439] memory commitment underflow > > Thanks for the catch! > > Could you share the info about the platform, like the CPU numbers > and RAM size, and what's the mmap test size of your test program. > It would be great if you can point me the link to the test program. I have been reproduced this on both AMD and Intel. The test just allocating memory and swapping. https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c It might be better to run the whole LTP mm tests if none of the above triggers it for you which has quite a few memory pressurers. /opt/ltp/runltp -f mm # lscpu Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian CPU(s): 48 On-line CPU(s) list: 0-47 Thread(s) per core: 1 Core(s) per socket: 12 Socket(s): 4 NUMA node(s): 4 Vendor ID: GenuineIntel CPU family: 6 Model: 63 Model name: Intel(R) Xeon(R) CPU E5-4650 v3 @ 2.10GHz Stepping: 2 CPU MHz: 1198.498 BogoMIPS: 4195.14 Virtualization: VT-x L1d cache: 32K L1i cache: 32K L2 cache: 256K L3 cache: 30720K NUMA node0 CPU(s): 0-5,24-29 NUMA node1 CPU(s): 6-11,30-35 NUMA node2 CPU(s): 12-17,36-41 NUMA node3 CPU(s): 18-23,42-47 # numactl -H available: 4 nodes (0-3) node 0 cpus: 0 1 2 3 4 5 24 25 26 27 28 29 node 0 size: 27768 MB node 0 free: 27539 MB node 1 cpus: 6 7 8 9 10 11 30 31 32 33 34 35 node 1 size: 32225 MB node 1 free: 32009 MB node 2 cpus: 12 13 14 15 16 17 36 37 38 39 40 41 node 2 size: 32253 MB node 2 free: 32174 MB node 3 cpus: 18 19 20 21 22 23 42 43 44 45 46 47 node 3 size: 32252 MB node 3 free: 31330 MB node distances: node 0 1 2 3 0: 10 21 21 21 1: 21 10 21 21 2: 21 21 10 21 3: 21 21 21 10 > > Thanks, > Feng > > > > [ 3319.284456] WARNING: CPU: 13 PID: 130949 at mm/util.c:835 __vm_enough_memory+0x1dc/0x1f0 > > > [ 3319.284485] Modules linked in: brd ext4 crc16 mbcache jbd2 loop kvm_hv kvm ip_tables x_tables xfs sd_mod ahci libahci bnx2x libata mdio tg3 libphy firmware_class dm_mirror dm_region_hash dm_log dm_mod > > > [ 3319.284538] CPU: 13 PID: 130949 Comm: oom01 Not tainted 5.7.0-rc6-next-20200521+ #115 > > > [ 3319.284551] NIP: c00000000043027c LR: c000000000430278 CTR: c0000000007bd120 > > > [ 3319.284572] REGS: c00000064a42f710 TRAP: 0700 Not tainted (5.7.0-rc6-next-20200521+) > > > [ 3319.284602] MSR: 900000000282b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28222422 XER: 20040000 > > > [ 3319.284638] CFAR: c000000000119098 IRQMASK: 0 > > > GPR00: c000000000430278 c00000064a42f9a0 c000000001765500 000000000000001b > > > GPR04: c00000000179c3b0 000000000000000b 0000000066726564 c00000000429ac28 > > > GPR08: 0000000000000001 0000000000000000 0000000000000000 c000001301facc80 > > > GPR12: 0000000000002000 c000001fffff3d80 0000000000000000 0000000000000000 > > > GPR16: 0000000000000000 0000000000000000 c0000014f358dd60 fffffffffff7dfff > > > GPR20: c0000003725a9128 c0000003725a3928 c000000b0a0c0670 c000000b0a0c0678 > > > GPR24: c000000b0a0c0660 c00000064a42fa58 c000000b0a0c0650 c0000003725a9080 > > > GPR28: 0000000000000001 0000000000000000 c000000001620178 0000000000000001 > > > [ 3319.284825] NIP [c00000000043027c] __vm_enough_memory+0x1dc/0x1f0 > > > [ 3319.284853] LR [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 > > > [ 3319.284872] Call Trace: > > > [ 3319.284889] [c00000064a42f9a0] [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 (unreliable) > > > [ 3319.284930] [c00000064a42fa30] [c000000000114258] dup_mm+0x2a8/0x700 > > > [ 3319.284960] [c00000064a42fae0] [c000000000115c7c] copy_process+0xeac/0x1a00 > > > [ 3319.284981] [c00000064a42fbf0] [c000000000116b3c] _do_fork+0xac/0xce0 > > > [ 3319.285002] [c00000064a42fd00] [c000000000117928] __do_sys_clone+0x98/0xe0 > > > [ 3319.285032] [c00000064a42fdc0] [c000000000039e78] system_call_exception+0x108/0x1d0 > > > [ 3319.285064] [c00000064a42fe20] [c00000000000c9f0] system_call_common+0xf0/0x278 > > > [ 3319.285093] Instruction dump: > > > [ 3319.285109] 60000000 73a90001 4182000c e8be0002 4bfffeb8 3c62ff51 39200001 3d42fff9 > > > [ 3319.285142] 38636690 992a6ad6 4bce8dbd 60000000 <0fe00000> e8be0002 4bfffe90 60000000
Hi Qian, On Tue, May 26, 2020 at 10:25:39PM -0400, Qian Cai wrote: > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > > > Reverted this series fixed a warning under memory pressue. > > > > > > Andrew, Stephen, can you drop this series? > > > > > > > > > > > [ 3319.257898] LTP: starting oom01 > > > > [ 3319.284417] ------------[ cut here ]------------ > > > > [ 3319.284439] memory commitment underflow > > > > Thanks for the catch! > > > > Could you share the info about the platform, like the CPU numbers > > and RAM size, and what's the mmap test size of your test program. > > It would be great if you can point me the link to the test program. > > I have been reproduced this on both AMD and Intel. The test just > allocating memory and swapping. > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c > > It might be better to run the whole LTP mm tests if none of the above > triggers it for you which has quite a few memory pressurers. > > /opt/ltp/runltp -f mm Thanks for sharing. I tried to reproduce this on 2 server plaforms, but can't reproduce it, and they are still under testing. Meanwhile, could you help to try the below patch, which is based on Andi's suggestion and have some debug info. The warning is a little strange, as the condition is (percpu_counter_read(&vm_committed_as) < -(s64)vm_committed_as_batch * num_online_cpus()) while for your platform (48 CPU + 128 GB RAM), the '-(s64)vm_committed_as_batch * num_online_cpus()' is a s64 value: '-32G', which makes the condition hard to be true, and when it is, it could be triggered by some magic for s32/s64 operations around the percpu-counter. Thanks, Feng --- mm/util.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/mm/util.c b/mm/util.c index 988d11e6c17c..2f3a2e0f12ae 100644 --- a/mm/util.c +++ b/mm/util.c @@ -797,10 +797,15 @@ EXPORT_SYMBOL_GPL(vm_memory_committed); int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) { long allowed; - - VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < + if (percpu_counter_read(&vm_committed_as) < + -(s64)vm_committed_as_batch * num_online_cpus()) { + printk("as:%lld as_sum:%lld check:%lld batch:%d\n", + percpu_counter_read(&vm_committed_as), + percpu_counter_sum(&vm_committed_as), -(s64)vm_committed_as_batch * num_online_cpus(), - "memory commitment underflow"); + vm_committed_as_batch); + + } vm_acct_memory(pages); @@ -816,6 +821,10 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) return 0; } + VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < + -(s64)vm_committed_as_batch * num_online_cpus(), + "memory commitment underflow"); + allowed = vm_commit_limit(); /* * Reserve some for root
On Wed, May 27, 2020 at 06:46:06PM +0800, Feng Tang wrote: > Hi Qian, > > On Tue, May 26, 2020 at 10:25:39PM -0400, Qian Cai wrote: > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > > > > > Reverted this series fixed a warning under memory pressue. > > > > > > > > Andrew, Stephen, can you drop this series? > > > > > > > > > > > > > > [ 3319.257898] LTP: starting oom01 > > > > > [ 3319.284417] ------------[ cut here ]------------ > > > > > [ 3319.284439] memory commitment underflow > > > > > > Thanks for the catch! > > > > > > Could you share the info about the platform, like the CPU numbers > > > and RAM size, and what's the mmap test size of your test program. > > > It would be great if you can point me the link to the test program. > > > > I have been reproduced this on both AMD and Intel. The test just > > allocating memory and swapping. > > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c > > > > It might be better to run the whole LTP mm tests if none of the above > > triggers it for you which has quite a few memory pressurers. > > > > /opt/ltp/runltp -f mm > > Thanks for sharing. I tried to reproduce this on 2 server plaforms, > but can't reproduce it, and they are still under testing. > > Meanwhile, could you help to try the below patch, which is based on > Andi's suggestion and have some debug info. The warning is a little > strange, as the condition is > > (percpu_counter_read(&vm_committed_as) < > -(s64)vm_committed_as_batch * num_online_cpus()) > > while for your platform (48 CPU + 128 GB RAM), the > '-(s64)vm_committed_as_batch * num_online_cpus()' > is a s64 value: '-32G', which makes the condition hard to be true, > and when it is, it could be triggered by some magic for s32/s64 > operations around the percpu-counter. Here is the information on AMD and powerpc below affected by this. It could need a bit patient to reproduce, but our usual daily CI would trigger it eventually after a few tries. # git clone https://github.com/cailca/linux-mm.git # cd linux-mm # ./compile.sh # systemctl reboot # ./test.sh == AMD == # lscpu Architecture: x86_64 CPU op-mode(s): 32-bit, 64-bit Byte Order: Little Endian CPU(s): 128 On-line CPU(s) list: 0-127 Thread(s) per core: 2 Core(s) per socket: 32 Socket(s): 2 NUMA node(s): 8 Vendor ID: AuthenticAMD CPU family: 23 Model: 1 Model name: AMD EPYC 7601 32-Core Processor Stepping: 2 CPU MHz: 3184.974 CPU max MHz: 2200.0000 CPU min MHz: 1200.0000 BogoMIPS: 4391.87 Virtualization: AMD-V L1d cache: 32K L1i cache: 64K L2 cache: 512K L3 cache: 8192K NUMA node0 CPU(s): 0-7,64-71 NUMA node1 CPU(s): 8-15,72-79 NUMA node2 CPU(s): 16-23,80-87 NUMA node3 CPU(s): 24-31,88-95 NUMA node4 CPU(s): 32-39,96-103 NUMA node5 CPU(s): 40-47,104-111 NUMA node6 CPU(s): 48-55,112-119 NUMA node7 CPU(s): 56-63,120-127 # numactl -H available: 8 nodes (0-7) node 0 cpus: 0 1 2 3 4 5 6 7 64 65 66 67 68 69 70 71 node 0 size: 20028 MB node 0 free: 3837 MB node 1 cpus: 8 9 10 11 12 13 14 15 72 73 74 75 76 77 78 79 node 1 size: 0 MB node 1 free: 0 MB node 2 cpus: 16 17 18 19 20 21 22 23 80 81 82 83 84 85 86 87 node 2 size: 0 MB node 2 free: 0 MB node 3 cpus: 24 25 26 27 28 29 30 31 88 89 90 91 92 93 94 95 node 3 size: 0 MB node 3 free: 0 MB node 4 cpus: 32 33 34 35 36 37 38 39 96 97 98 99 100 101 102 103 node 4 size: 31391 MB node 4 free: 21309 MB node 5 cpus: 40 41 42 43 44 45 46 47 104 105 106 107 108 109 110 111 node 5 size: 0 MB node 5 free: 0 MB node 6 cpus: 48 49 50 51 52 53 54 55 112 113 114 115 116 117 118 119 node 6 size: 0 MB node 6 free: 0 MB node 7 cpus: 56 57 58 59 60 61 62 63 120 121 122 123 124 125 126 127 node 7 size: 0 MB node 7 free: 0 MB node distances: node 0 1 2 3 4 5 6 7 0: 10 16 16 16 32 32 32 32 1: 16 10 16 16 32 32 32 32 2: 16 16 10 16 32 32 32 32 3: 16 16 16 10 32 32 32 32 4: 32 32 32 32 10 16 16 16 5: 32 32 32 32 16 10 16 16 6: 32 32 32 32 16 16 10 16 7: 32 32 32 32 16 16 16 10 == powerpc == # lscpu Architecture: ppc64le Byte Order: Little Endian CPU(s): 128 On-line CPU(s) list: 0-127 Thread(s) per core: 4 Core(s) per socket: 16 Socket(s): 2 NUMA node(s): 2 Model: 2.2 (pvr 004e 1202) Model name: POWER9 (raw), altivec supported CPU max MHz: 3800.0000 CPU min MHz: 2300.0000 L1d cache: 32K L1i cache: 32K L2 cache: 512K L3 cache: 10240K NUMA node0 CPU(s): 0-63 NUMA node8 CPU(s): 64-127 # numactl -H available: 2 nodes (0,8) node 0 cpus: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 node 0 size: 130138 MB node 0 free: 125654 MB node 8 cpus: 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 node 8 size: 130781 MB node 8 free: 124756 MB node distances: node 0 8 0: 10 40 8: 40 10
Hi Qian, On Wed, May 27, 2020 at 08:05:49AM -0400, Qian Cai wrote: > On Wed, May 27, 2020 at 06:46:06PM +0800, Feng Tang wrote: > > Hi Qian, > > > > On Tue, May 26, 2020 at 10:25:39PM -0400, Qian Cai wrote: > > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > > > > > > > Reverted this series fixed a warning under memory pressue. > > > > > > > > > > Andrew, Stephen, can you drop this series? > > > > > > > > > > > > > > > > > [ 3319.257898] LTP: starting oom01 > > > > > > [ 3319.284417] ------------[ cut here ]------------ > > > > > > [ 3319.284439] memory commitment underflow > > > > > > > > Thanks for the catch! > > > > > > > > Could you share the info about the platform, like the CPU numbers > > > > and RAM size, and what's the mmap test size of your test program. > > > > It would be great if you can point me the link to the test program. > > > > > > I have been reproduced this on both AMD and Intel. The test just > > > allocating memory and swapping. > > > > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c > > > > > > It might be better to run the whole LTP mm tests if none of the above > > > triggers it for you which has quite a few memory pressurers. > > > > > > /opt/ltp/runltp -f mm > > > > Thanks for sharing. I tried to reproduce this on 2 server plaforms, > > but can't reproduce it, and they are still under testing. > > > > Meanwhile, could you help to try the below patch, which is based on > > Andi's suggestion and have some debug info. The warning is a little > > strange, as the condition is > > > > (percpu_counter_read(&vm_committed_as) < > > -(s64)vm_committed_as_batch * num_online_cpus()) > > > > while for your platform (48 CPU + 128 GB RAM), the > > '-(s64)vm_committed_as_batch * num_online_cpus()' > > is a s64 value: '-32G', which makes the condition hard to be true, > > and when it is, it could be triggered by some magic for s32/s64 > > operations around the percpu-counter. > > Here is the information on AMD and powerpc below affected by this. It > could need a bit patient to reproduce, but our usual daily CI would > trigger it eventually after a few tries. > > # git clone https://github.com/cailca/linux-mm.git > # cd linux-mm > # ./compile.sh > # systemctl reboot > # ./test.sh I just downloaded it, and it failed on my desktop machine as it failed in 'yum' and 'grub2' setup. The difficulty for me to reproduce is the test platforms are behind the 0day framework, and I can hardly setup external test suits, though I have been trying for all day today :) So if possible, please help to try the patch in my last email. thanks! - Feng
On Wed, May 27, 2020 at 09:33:32PM +0800, Feng Tang wrote: > Hi Qian, > > On Wed, May 27, 2020 at 08:05:49AM -0400, Qian Cai wrote: > > On Wed, May 27, 2020 at 06:46:06PM +0800, Feng Tang wrote: > > > Hi Qian, > > > > > > On Tue, May 26, 2020 at 10:25:39PM -0400, Qian Cai wrote: > > > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > > > > > > > > > Reverted this series fixed a warning under memory pressue. > > > > > > > > > > > > Andrew, Stephen, can you drop this series? > > > > > > > > > > > > > > > > > > > > [ 3319.257898] LTP: starting oom01 > > > > > > > [ 3319.284417] ------------[ cut here ]------------ > > > > > > > [ 3319.284439] memory commitment underflow > > > > > > > > > > Thanks for the catch! > > > > > > > > > > Could you share the info about the platform, like the CPU numbers > > > > > and RAM size, and what's the mmap test size of your test program. > > > > > It would be great if you can point me the link to the test program. > > > > > > > > I have been reproduced this on both AMD and Intel. The test just > > > > allocating memory and swapping. > > > > > > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c > > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c > > > > > > > > It might be better to run the whole LTP mm tests if none of the above > > > > triggers it for you which has quite a few memory pressurers. > > > > > > > > /opt/ltp/runltp -f mm > > > > > > Thanks for sharing. I tried to reproduce this on 2 server plaforms, > > > but can't reproduce it, and they are still under testing. > > > > > > Meanwhile, could you help to try the below patch, which is based on > > > Andi's suggestion and have some debug info. The warning is a little > > > strange, as the condition is > > > > > > (percpu_counter_read(&vm_committed_as) < > > > -(s64)vm_committed_as_batch * num_online_cpus()) > > > > > > while for your platform (48 CPU + 128 GB RAM), the > > > '-(s64)vm_committed_as_batch * num_online_cpus()' > > > is a s64 value: '-32G', which makes the condition hard to be true, > > > and when it is, it could be triggered by some magic for s32/s64 > > > operations around the percpu-counter. > > > > Here is the information on AMD and powerpc below affected by this. It > > could need a bit patient to reproduce, but our usual daily CI would > > trigger it eventually after a few tries. > > > > # git clone https://github.com/cailca/linux-mm.git > > # cd linux-mm > > # ./compile.sh > > # systemctl reboot > > # ./test.sh > > I just downloaded it, and it failed on my desktop machine as it failed > in 'yum' and 'grub2' setup. The difficulty for me to reproduce is the > test platforms are behind the 0day framework, and I can hardly setup > external test suits, though I have been trying for all day today :) I tried your debug patch and it did not even compile on linux-next (where the issue was happened) and I am running out of time today. It probably need to reproduce on large systems as it did not happen on one of our small s390 system here.
On Wed, May 27, 2020 at 06:46:06PM +0800, Feng Tang wrote: > Hi Qian, > > On Tue, May 26, 2020 at 10:25:39PM -0400, Qian Cai wrote: > > > > > > [1] https://lkml.org/lkml/2020/3/5/57 > > > > > > > > > > Reverted this series fixed a warning under memory pressue. > > > > > > > > Andrew, Stephen, can you drop this series? > > > > > > > > > > > > > > [ 3319.257898] LTP: starting oom01 > > > > > [ 3319.284417] ------------[ cut here ]------------ > > > > > [ 3319.284439] memory commitment underflow > > > > > > Thanks for the catch! > > > > > > Could you share the info about the platform, like the CPU numbers > > > and RAM size, and what's the mmap test size of your test program. > > > It would be great if you can point me the link to the test program. > > > > I have been reproduced this on both AMD and Intel. The test just > > allocating memory and swapping. > > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c > > > > It might be better to run the whole LTP mm tests if none of the above > > triggers it for you which has quite a few memory pressurers. > > > > /opt/ltp/runltp -f mm > > Thanks for sharing. I tried to reproduce this on 2 server plaforms, > but can't reproduce it, and they are still under testing. > > Meanwhile, could you help to try the below patch, which is based on > Andi's suggestion and have some debug info. The warning is a little > strange, as the condition is > > (percpu_counter_read(&vm_committed_as) < > -(s64)vm_committed_as_batch * num_online_cpus()) > > while for your platform (48 CPU + 128 GB RAM), the > '-(s64)vm_committed_as_batch * num_online_cpus()' > is a s64 value: '-32G', which makes the condition hard to be true, > and when it is, it could be triggered by some magic for s32/s64 > operations around the percpu-counter. The patch below does not fix anything. [ 3160.275230][ T7955] LTP: starting oom01 [ 3160.306977][T106365] KK as:-59683 as_sum:6896 check:-32768 batch:256 [ 3160.307161][T106365] ------------[ cut here ]------------ [ 3160.307184][T106365] memory commitment underflow [ 3160.307216][T106365] WARNING: CPU: 103 PID: 106365 at mm/util.c:858 __vm_enough_memory+0x204/0x250 [ 3160.307275][T106365] Modules linked in: brd ext4 crc16 mbcache jbd2 loop kvm_hv kvm ip_tables x_tables xfs sd_mod bnx2x ahci libahci mdio tg3 libata libphy firmware_class dm_mirror dm_region_hash dm_log dm_mod [ 3160.307341][T106365] CPU: 103 PID: 106365 Comm: oom01 Not tainted 5.7.0-rc7-next-20200528+ #3 [ 3160.307368][T106365] NIP: c0000000003ee654 LR: c0000000003ee650 CTR: c000000000745b40 [ 3160.307382][T106365] REGS: c0002017f940f730 TRAP: 0700 Not tainted (5.7.0-rc7-next-20200528+) [ 3160.307409][T106365] MSR: 900000000282b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28222482 XER: 00000000 [ 3160.307442][T106365] CFAR: c00000000010e428 IRQMASK: 0 [ 3160.307442][T106365] GPR00: c0000000003ee650 c0002017f940f9c0 c00000000130a300 000000000000001b [ 3160.307442][T106365] GPR04: c0000000017b8d68 0000000000000006 0000000078154885 ffffffff6ca32090 [ 3160.307442][T106365] GPR08: 0000201cc61a0000 0000000000000000 0000000000000000 0000000000000003 [ 3160.307442][T106365] GPR12: 0000000000002000 c000201fff675100 0000000000000000 0000000000000000 [ 3160.307442][T106365] GPR16: 0000000000000000 0000000000000000 c000200350fe3b60 fffffffffff7dfff [ 3160.307442][T106365] GPR20: c000201a36824928 c000201a3682c128 c000200ee33ad3a0 c000200ee33ad3a8 [ 3160.307442][T106365] GPR24: c000200ee33ad390 ffffffffffff16dd 0000000000000000 0000000000000001 [ 3160.307442][T106365] GPR28: c000201a36824880 0000000000000001 c000000003fa2a80 c0000000011fb0a8 [ 3160.307723][T106365] NIP [c0000000003ee654] __vm_enough_memory+0x204/0x250 [ 3160.307759][T106365] LR [c0000000003ee650] __vm_enough_memory+0x200/0x250 [ 3160.307793][T106365] Call Trace: [ 3160.307814][T106365] [c0002017f940f9c0] [c0000000003ee650] __vm_enough_memory+0x200/0x250 (unreliable) [ 3160.307875][T106365] [c0002017f940fa60] [c000000000109ab8] dup_mm+0x2a8/0x700 [ 3160.307901][T106365] [c0002017f940fb10] [c00000000010b3f0] copy_process+0xeb0/0x1920 [ 3160.307927][T106365] [c0002017f940fc20] [c00000000010c1cc] _do_fork+0xac/0xb20 [ 3160.307941][T106365] [c0002017f940fd00] [c00000000010cdf8] __do_sys_clone+0x98/0xe0 [ 3160.307979][T106365] [c0002017f940fdc0] [c000000000038a88] system_call_exception+0x108/0x1d0 [ 3160.308018][T106365] [c0002017f940fe20] [c00000000000c9f0] system_call_common+0xf0/0x278 [ 3160.308054][T106365] Instruction dump: [ 3160.308076][T106365] 60000000 60000000 60000000 73490001 4082fee4 3c62ff8c 39200001 3d42fff9 [ 3160.308128][T106365] 38632788 992a6b4a 4bd1fd75 60000000 <0fe00000> 4bfffec0 7fc3f378 483218f9 [ 3160.308158][T106365] irq event stamp: 760 [ 3160.308181][T106365] hardirqs last enabled at (759): [<c0000000001e9838>] vprintk_emit+0x118/0x430 [ 3160.308231][T106365] hardirqs last disabled at (760): [<c00000000000963c>] program_check_common_virt+0x2bc/0x310 [ 3160.308281][T106365] softirqs last enabled at (0): [<c00000000010ac38>] copy_process+0x6f8/0x1920 [ 3160.308297][T106365] softirqs last disabled at (0): [<0000000000000000>] 0x0 [ 3160.308310][T106365] ---[ end trace e2152aa44c190593 ]--- [ 3160.308478][T106365] KK as:-59683 as_sum:6897 check:-32768 batch:256 [ 3160.308614][T106365] KK as:-59683 as_sum:6898 check:-32768 batch:256 [ 3160.308714][T106365] KK as:-59683 as_sum:6901 check:-32768 batch:256 [ 3160.308806][T106365] KK as:-59683 as_sum:6902 check:-32768 batch:256 [ 3160.308900][T106365] KK as:-59683 as_sum:6903 check:-32768 batch:256 [ 3160.308979][T106365] KK as:-59683 as_sum:6904 check:-32768 batch:256 [ 3160.309064][T106365] KK as:-59683 as_sum:6905 check:-32768 batch:256 [ 3160.309160][T106365] KK as:-59683 as_sum:6906 check:-32768 batch:256 [ 3160.309275][T106365] KK as:-59683 as_sum:6907 check:-32768 batch:256 [ 3160.309356][T106365] KK as:-59683 as_sum:6908 check:-32768 batch:256 [ 3160.309437][T106365] KK as:-59683 as_sum:6909 check:-32768 batch:256 [ 3160.310939][T106366] KK as:-59683 as_sum:6912 check:-32768 batch:256 [ 3160.311134][T106366] KK as:-59270 as_sum:7040 check:-32768 batch:256 [ 3160.311240][T106367] KK as:-59270 as_sum:7168 check:-32768 batch:256 > > Thanks, > Feng > > --- > mm/util.c | 15 ++++++++++++--- > 1 file changed, 12 insertions(+), 3 deletions(-) > > diff --git a/mm/util.c b/mm/util.c > index 988d11e6c17c..2f3a2e0f12ae 100644 > --- a/mm/util.c > +++ b/mm/util.c > @@ -797,10 +797,15 @@ EXPORT_SYMBOL_GPL(vm_memory_committed); > int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) > { > long allowed; > - > - VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < > + if (percpu_counter_read(&vm_committed_as) < > + -(s64)vm_committed_as_batch * num_online_cpus()) { > + printk("as:%lld as_sum:%lld check:%lld batch:%d\n", > + percpu_counter_read(&vm_committed_as), > + percpu_counter_sum(&vm_committed_as), > -(s64)vm_committed_as_batch * num_online_cpus(), > - "memory commitment underflow"); > + vm_committed_as_batch); > + > + } > > vm_acct_memory(pages); > > @@ -816,6 +821,10 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) > return 0; > } > > + VM_WARN_ONCE(percpu_counter_read(&vm_committed_as) < > + -(s64)vm_committed_as_batch * num_online_cpus(), > + "memory commitment underflow"); > + > allowed = vm_commit_limit(); > /* > * Reserve some for root > -- > 2.14.1 > > > > > # lscpu > > Architecture: x86_64 > > CPU op-mode(s): 32-bit, 64-bit > > Byte Order: Little Endian > > CPU(s): 48 > > On-line CPU(s) list: 0-47 > > Thread(s) per core: 1 > > Core(s) per socket: 12 > > Socket(s): 4 > > NUMA node(s): 4 > > Vendor ID: GenuineIntel > > CPU family: 6 > > Model: 63 > > Model name: Intel(R) Xeon(R) CPU E5-4650 v3 @ 2.10GHz > > Stepping: 2 > > CPU MHz: 1198.498 > > BogoMIPS: 4195.14 > > Virtualization: VT-x > > L1d cache: 32K > > L1i cache: 32K > > L2 cache: 256K > > L3 cache: 30720K > > NUMA node0 CPU(s): 0-5,24-29 > > NUMA node1 CPU(s): 6-11,30-35 > > NUMA node2 CPU(s): 12-17,36-41 > > NUMA node3 CPU(s): 18-23,42-47 > > > > # numactl -H > > available: 4 nodes (0-3) > > node 0 cpus: 0 1 2 3 4 5 24 25 26 27 28 29 > > node 0 size: 27768 MB > > node 0 free: 27539 MB > > node 1 cpus: 6 7 8 9 10 11 30 31 32 33 34 35 > > node 1 size: 32225 MB > > node 1 free: 32009 MB > > node 2 cpus: 12 13 14 15 16 17 36 37 38 39 40 41 > > node 2 size: 32253 MB > > node 2 free: 32174 MB > > node 3 cpus: 18 19 20 21 22 23 42 43 44 45 46 47 > > node 3 size: 32252 MB > > node 3 free: 31330 MB > > node distances: > > node 0 1 2 3 > > 0: 10 21 21 21 > > 1: 21 10 21 21 > > 2: 21 21 10 21 > > 3: 21 21 21 10 > > > > > > > > Thanks, > > > Feng > > > > > > > > [ 3319.284456] WARNING: CPU: 13 PID: 130949 at mm/util.c:835 __vm_enough_memory+0x1dc/0x1f0 > > > > > [ 3319.284485] Modules linked in: brd ext4 crc16 mbcache jbd2 loop kvm_hv kvm ip_tables x_tables xfs sd_mod ahci libahci bnx2x libata mdio tg3 libphy firmware_class dm_mirror dm_region_hash dm_log dm_mod > > > > > [ 3319.284538] CPU: 13 PID: 130949 Comm: oom01 Not tainted 5.7.0-rc6-next-20200521+ #115 > > > > > [ 3319.284551] NIP: c00000000043027c LR: c000000000430278 CTR: c0000000007bd120 > > > > > [ 3319.284572] REGS: c00000064a42f710 TRAP: 0700 Not tainted (5.7.0-rc6-next-20200521+) > > > > > [ 3319.284602] MSR: 900000000282b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28222422 XER: 20040000 > > > > > [ 3319.284638] CFAR: c000000000119098 IRQMASK: 0 > > > > > GPR00: c000000000430278 c00000064a42f9a0 c000000001765500 000000000000001b > > > > > GPR04: c00000000179c3b0 000000000000000b 0000000066726564 c00000000429ac28 > > > > > GPR08: 0000000000000001 0000000000000000 0000000000000000 c000001301facc80 > > > > > GPR12: 0000000000002000 c000001fffff3d80 0000000000000000 0000000000000000 > > > > > GPR16: 0000000000000000 0000000000000000 c0000014f358dd60 fffffffffff7dfff > > > > > GPR20: c0000003725a9128 c0000003725a3928 c000000b0a0c0670 c000000b0a0c0678 > > > > > GPR24: c000000b0a0c0660 c00000064a42fa58 c000000b0a0c0650 c0000003725a9080 > > > > > GPR28: 0000000000000001 0000000000000000 c000000001620178 0000000000000001 > > > > > [ 3319.284825] NIP [c00000000043027c] __vm_enough_memory+0x1dc/0x1f0 > > > > > [ 3319.284853] LR [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 > > > > > [ 3319.284872] Call Trace: > > > > > [ 3319.284889] [c00000064a42f9a0] [c000000000430278] __vm_enough_memory+0x1d8/0x1f0 (unreliable) > > > > > [ 3319.284930] [c00000064a42fa30] [c000000000114258] dup_mm+0x2a8/0x700 > > > > > [ 3319.284960] [c00000064a42fae0] [c000000000115c7c] copy_process+0xeac/0x1a00 > > > > > [ 3319.284981] [c00000064a42fbf0] [c000000000116b3c] _do_fork+0xac/0xce0 > > > > > [ 3319.285002] [c00000064a42fd00] [c000000000117928] __do_sys_clone+0x98/0xe0 > > > > > [ 3319.285032] [c00000064a42fdc0] [c000000000039e78] system_call_exception+0x108/0x1d0 > > > > > [ 3319.285064] [c00000064a42fe20] [c00000000000c9f0] system_call_common+0xf0/0x278 > > > > > [ 3319.285093] Instruction dump: > > > > > [ 3319.285109] 60000000 73a90001 4182000c e8be0002 4bfffeb8 3c62ff51 39200001 3d42fff9 > > > > > [ 3319.285142] 38636690 992a6ad6 4bce8dbd 60000000 <0fe00000> e8be0002 4bfffe90 60000000
On Thu, May 28, 2020 at 10:18:02AM -0400, Qian Cai wrote: > > > I have been reproduced this on both AMD and Intel. The test just > > > allocating memory and swapping. > > > > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/oom/oom01.c > > > https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/mem/tunable/overcommit_memory.c > > > > > > It might be better to run the whole LTP mm tests if none of the above > > > triggers it for you which has quite a few memory pressurers. > > > > > > /opt/ltp/runltp -f mm > > > > Thanks for sharing. I tried to reproduce this on 2 server plaforms, > > but can't reproduce it, and they are still under testing. > > > > Meanwhile, could you help to try the below patch, which is based on > > Andi's suggestion and have some debug info. The warning is a little > > strange, as the condition is > > > > (percpu_counter_read(&vm_committed_as) < > > -(s64)vm_committed_as_batch * num_online_cpus()) > > > > while for your platform (48 CPU + 128 GB RAM), the > > '-(s64)vm_committed_as_batch * num_online_cpus()' > > is a s64 value: '-32G', which makes the condition hard to be true, > > and when it is, it could be triggered by some magic for s32/s64 > > operations around the percpu-counter. > > The patch below does not fix anything. Thanks for the info. > > [ 3160.275230][ T7955] LTP: starting oom01 > [ 3160.306977][T106365] KK as:-59683 as_sum:6896 This does show that percpu_counter_read() is not accurate comparing to percpu_counter_sum(). And if we use percpu_counter_sum() then the warning won't happen, as 6896 >> -32768 > check:-32768 batch:256 -32768 means the RAM of the test platform is either 512M or 32GB (more likely) depending on overcommit policy, and it has 128 CPUs. So my guess of the root cause is the overcommit policy is changing while doing the memory stress test, so the vm_commited_as.count could have bigger deviation when the overcommit policy is !OVERCOMMIT_NEVER, and when the policy is changed to OVERCOMMIT_NEVER, the check value is hugely reduced, which trigger the warning. If it's true, then there could be 2 solutions, one is to skip the WARN_ONCE as it has no practical value, as the real check is the following code, the other is to rectify the percpu counter when the policy is changing to OVERCOMMIT_NEVER. Thanks, Feng > [ 3160.307161][T106365] ------------[ cut here ]------------ > [ 3160.307184][T106365] memory commitment underflow > [ 3160.307216][T106365] WARNING: CPU: 103 PID: 106365 at mm/util.c:858 __vm_enough_memory+0x204/0x250 > [ 3160.307275][T106365] Modules linked in: brd ext4 crc16 mbcache jbd2 loop kvm_hv kvm ip_tables x_tables xfs sd_mod bnx2x ahci libahci mdio tg3 libata libphy firmware_class dm_mirror dm_region_hash dm_log dm_mod > [ 3160.307341][T106365] CPU: 103 PID: 106365 Comm: oom01 Not tainted 5.7.0-rc7-next-20200528+ #3 > [ 3160.307368][T106365] NIP: c0000000003ee654 LR: c0000000003ee650 CTR: c000000000745b40 > [ 3160.307382][T106365] REGS: c0002017f940f730 TRAP: 0700 Not tainted (5.7.0-rc7-next-20200528+) > [ 3160.307409][T106365] MSR: 900000000282b033 <SF,HV,VEC,VSX,EE,FP,ME,IR,DR,RI,LE> CR: 28222482 XER: 00000000 > [ 3160.307442][T106365] CFAR: c00000000010e428 IRQMASK: 0 > [ 3160.307442][T106365] GPR00: c0000000003ee650 c0002017f940f9c0 c00000000130a300 000000000000001b > [ 3160.307442][T106365] GPR04: c0000000017b8d68 0000000000000006 0000000078154885 ffffffff6ca32090 > [ 3160.307442][T106365] GPR08: 0000201cc61a0000 0000000000000000 0000000000000000 0000000000000003 > [ 3160.307442][T106365] GPR12: 0000000000002000 c000201fff675100 0000000000000000 0000000000000000 > [ 3160.307442][T106365] GPR16: 0000000000000000 0000000000000000 c000200350fe3b60 fffffffffff7dfff > [ 3160.307442][T106365] GPR20: c000201a36824928 c000201a3682c128 c000200ee33ad3a0 c000200ee33ad3a8 > [ 3160.307442][T106365] GPR24: c000200ee33ad390 ffffffffffff16dd 0000000000000000 0000000000000001 > [ 3160.307442][T106365] GPR28: c000201a36824880 0000000000000001 c000000003fa2a80 c0000000011fb0a8 > [ 3160.307723][T106365] NIP [c0000000003ee654] __vm_enough_memory+0x204/0x250 > [ 3160.307759][T106365] LR [c0000000003ee650] __vm_enough_memory+0x200/0x250 > [ 3160.308310][T106365] ---[ end trace e2152aa44c190593 ]--- > [ 3160.308478][T106365] KK as:-59683 as_sum:6897 check:-32768 batch:256 > [ 3160.308614][T106365] KK as:-59683 as_sum:6898 check:-32768 batch:256 > [ 3160.308714][T106365] KK as:-59683 as_sum:6901 check:-32768 batch:256 > [ 3160.308806][T106365] KK as:-59683 as_sum:6902 check:-32768 batch:256 > [ 3160.308900][T106365] KK as:-59683 as_sum:6903 check:-32768 batch:256 > [ 3160.308979][T106365] KK as:-59683 as_sum:6904 check:-32768 batch:256 > [ 3160.309064][T106365] KK as:-59683 as_sum:6905 check:-32768 batch:256 > [ 3160.309160][T106365] KK as:-59683 as_sum:6906 check:-32768 batch:256 > [ 3160.309275][T106365] KK as:-59683 as_sum:6907 check:-32768 batch:256 > [ 3160.309356][T106365] KK as:-59683 as_sum:6908 check:-32768 batch:256 > [ 3160.309437][T106365] KK as:-59683 as_sum:6909 check:-32768 batch:256 > [ 3160.310939][T106366] KK as:-59683 as_sum:6912 check:-32768 batch:256 > [ 3160.311134][T106366] KK as:-59270 as_sum:7040 check:-32768 batch:256 > [ 3160.311240][T106367] KK as:-59270 as_sum:7168 check:-32768 batch:256 >
>If it's true, then there could be 2 solutions, one is to skip the WARN_ONCE as it has no practical value, as the real >check is the following code, the other is to rectify the percpu counter when the policy is changing to >OVERCOMMIT_NEVER.
I think it's better to fix it up when the policy changes. That's the right place. The WARN_ON might be useful to catch other bugs.
-Andi
On Thu 28-05-20 23:10:20, Feng Tang wrote: [...] > If it's true, then there could be 2 solutions, one is to > skip the WARN_ONCE as it has no practical value, as the real > check is the following code, the other is to rectify the > percpu counter when the policy is changing to OVERCOMMIT_NEVER. I would simply drop the WARN_ONCE. Looking at the history this has been added by 82f71ae4a2b8 ("mm: catch memory commitment underflow") to have a safety check for issues which have been fixed. There doesn't seem to be any bug reports mentioning this splat since then so it is likely just spending cycles for a hot path (yes many people run with DEBUG_VM) without a strong reason.
On Thu, May 28, 2020 at 11:21:36PM +0800, Kleen, Andi wrote: > > > >If it's true, then there could be 2 solutions, one is to skip the WARN_ONCE as it has no practical value, as the real >check is the following code, the other is to rectify the percpu counter when the policy is changing to >OVERCOMMIT_NEVER. > > I think it's better to fix it up when the policy changes. That's the right place. The WARN_ON might be useful to catch other bugs. If we keep the WARN_ON, then the draft fix patch I can think of looks like: diff --git a/lib/percpu_counter.c b/lib/percpu_counter.c index a66595b..02d87fc 100644 --- a/lib/percpu_counter.c +++ b/lib/percpu_counter.c @@ -98,6 +98,20 @@ void percpu_counter_add_batch(struct percpu_counter *fbc, s64 amount, s32 batch) } EXPORT_SYMBOL(percpu_counter_add_batch); +void percpu_counter_sync(struct percpu_counter *fbc) +{ + unsigned long flags; + s64 count; + + raw_spin_lock_irqsave(&fbc->lock, flags); + count = __this_cpu_read(*fbc->counters); + fbc->count += count; + __this_cpu_sub(*fbc->counters, count); + raw_spin_unlock_irqrestore(&fbc->lock, flags); +} +EXPORT_SYMBOL(percpu_counter_sync); + + /* * Add up all the per-cpu counts, return the result. This is a more accurate * but much slower version of percpu_counter_read_positive() diff --git a/mm/util.c b/mm/util.c index 580d268..24322da 100644 --- a/mm/util.c +++ b/mm/util.c @@ -746,14 +746,24 @@ int overcommit_ratio_handler(struct ctl_table *table, int write, void *buffer, return ret; } +static void sync_overcommit_as(struct work_struct *dummy) +{ + percpu_counter_sync(&vm_committed_as); +} + int overcommit_policy_handler(struct ctl_table *table, int write, void *buffer, size_t *lenp, loff_t *ppos) { int ret; ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); - if (ret == 0 && write) + if (ret == 0 && write) { + if (sysctl_overcommit_memory == OVERCOMMIT_NEVER) + schedule_on_each_cpu(sync_overcommit_as); + mm_compute_batch(); + } return ret; } Any comments? Thanks, Feng > -Andi >
> > ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); > - if (ret == 0 && write) > + if (ret == 0 && write) { > + if (sysctl_overcommit_memory == OVERCOMMIT_NEVER) > + schedule_on_each_cpu(sync_overcommit_as); The schedule is not atomic. There's still a race window here over all the CPUs where the WARN_ON could happen because you change the global first. Probably you would need another global that says "i'm currently changing the mode" and then skip the WARN_ON in that window. Maybe a sequence lock. Seems all overkill to me. Better to kill the warning. -Andi
On Fri, May 29, 2020 at 08:50:25AM -0700, Andi Kleen wrote: > > > > ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos); > > - if (ret == 0 && write) > > + if (ret == 0 && write) { > > + if (sysctl_overcommit_memory == OVERCOMMIT_NEVER) > > + schedule_on_each_cpu(sync_overcommit_as); > > The schedule is not atomic. > There's still a race window here over all the CPUs where the WARN_ON could > happen because you change the global first. The re-computing of batch number comes after this sync, so at this point the batch is still the bigger one, and won't trigger the warning. > Probably you would need another global that says "i'm currently changing > the mode" and then skip the WARN_ON in that window. Maybe a sequence lock. > > Seems all overkill to me. Better to kill the warning. Yes, the cost is high, schedule_on_each_cpu is labeled as "very slow" in the code comments. Thanks, Feng > > -Andi
When checking a performance change for will-it-scale scalability mmap test [1], we found very high lock contention for spinlock of percpu counter 'vm_committed_as': 94.14% 0.35% [kernel.kallsyms] [k] _raw_spin_lock_irqsave 48.21% _raw_spin_lock_irqsave;percpu_counter_add_batch;__vm_enough_memory;mmap_region;do_mmap; 45.91% _raw_spin_lock_irqsave;percpu_counter_add_batch;__do_munmap; Actually this heavy lock contention is not always necessary. The 'vm_committed_as' needs to be very precise when the strict OVERCOMMIT_NEVER policy is set, which requires a rather small batch number for the percpu counter. So lift the batch number to 16X for OVERCOMMIT_ALWAYS and OVERCOMMIT_GUESS policies, and add a sysctl handler to adjust it when the policy is reconfigured. Benchmark with the same testcase in [1] shows 53% improvement on a 8C/16T desktop, and 2097%(20X) on a 4S/72C/144T server. And no change for some platforms, due to the test mmap size of the case is bigger than the batch number computed, though the patch will help mmap/munmap generally. [1] https://lkml.org/lkml/2020/3/5/57 There are some style complain from checkpatch for patch 3, as it follows the similar format of sibling functions patch1: a cleanup for /proc/meminfo patch2: a preparation patch which also improve the accuracy of vm_memory_committed patch3: the main change Please help to review, thanks! - Feng ---------------------------------------------------------------- Changelog: v2: * add the sysctl handler to cover runtime overcommit policy change, as suggested by Andres Morton * address the accuracy concern of vm_memory_committed() from Andi Kleen Feng Tang (3): proc/meminfo: avoid open coded reading of vm_committed_as mm/util.c: make vm_memory_committed() more accurate mm: adjust vm_committed_as_batch according to vm overcommit policy fs/proc/meminfo.c | 2 +- include/linux/mm.h | 2 ++ include/linux/mman.h | 4 ++++ kernel/sysctl.c | 2 +- mm/mm_init.c | 19 +++++++++++++++---- mm/util.c | 15 ++++++++++++++- 6 files changed, 37 insertions(+), 7 deletions(-)