lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening PHC | |
Open Source and information security mailing list archives
| ||
|
Date: Fri, 15 Nov 2019 17:09:13 +0800 From: Shaokun Zhang <zhangshaokun@...ilicon.com> To: Michal Hocko <mhocko@...nel.org> CC: <linux-kernel@...r.kernel.org>, yuqi jin <jinyuqi@...wei.com>, "Andrew Morton" <akpm@...ux-foundation.org>, Mike Rapoport <rppt@...ux.ibm.com>, "Paul Burton" <paul.burton@...s.com>, Michael Ellerman <mpe@...erman.id.au>, Anshuman Khandual <anshuman.khandual@....com>, <netdev@...r.kernel.org> Subject: Re: [PATCH v3] lib: optimize cpumask_local_spread() Hi Michal, On 2019/11/14 22:43, Michal Hocko wrote: > On Wed 13-11-19 10:46:05, Shaokun Zhang wrote: > [...] >>>> available: 4 nodes (0-3) >>>> 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 >>>> node 0 size: 63379 MB >>>> node 0 free: 61899 MB >>>> node 1 cpus: 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 >>>> node 1 size: 64509 MB >>>> node 1 free: 63942 MB >>>> node 2 cpus: 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 >>>> node 2 size: 64509 MB >>>> node 2 free: 63056 MB >>>> node 3 cpus: 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 >>>> node 3 size: 63997 MB >>>> node 3 free: 63420 MB >>>> node distances: >>>> node 0 1 2 3 >>>> 0: 10 16 32 33 >>>> 1: 16 10 25 32 >>>> 2: 32 25 10 16 >>>> 3: 33 32 16 10 > [...] >> before patch >> Euler:/sys/bus/pci/devices/0000:7d:00.2 # cat numa_node >> 2 >> Euler:/sys/bus/pci # cat /proc/irq/345/smp_affinity_list >> 48 > > node 2 > >> Euler:/sys/bus/pci # cat /proc/irq/369/smp_affinity_list >> 0 > > node 0 > >> Euler:/sys/bus/pci # cat /proc/irq/393/smp_affinity_list >> 24 > > node 1 > >> Euler:/sys/bus/pci # >> >> after patch >> Euler:/sys/bus/pci/devices/0000:7d:00.2 # cat numa_node >> 2 >> Euler:/sys/bus/pci # cat /proc/irq/345/smp_affinity_list >> 48 > > node 2 > >> Euler:/sys/bus/pci # cat /proc/irq/369/smp_affinity_list >> 72 > > node 3 > >> Euler:/sys/bus/pci # cat /proc/irq/393/smp_affinity_list >> 24 > > node 1 > > So few more questions. The only difference seems to be IRQ369 > moving from 0 to 3 and having the device affinity to node 2 > makes some sense because node 3 is closer. So far so good. Right, it is what we want. > I still have a large gap to get the whole picture. Namely why those > other IRQs are not using any of the existing CPUs on the node 2. > Could you explain that please? > Oh, my mistake, for the previous instance, I don't list all IRQs and just choose one IRQ from one NUMA node. You can see that the IRQ number is not consistent :-). IRQ from 345 to 368 will be bound to CPU cores which are in NUMA node2 and each IRQ is corresponding to one core. Euler:/sys/bus/pci # cat /proc/irq/346/smp_affinity_list 49 Others are the similar. > Btw. this all should be in the changelog. Ok, I will follow it in future. Thanks, Shaokun >
Powered by blists - more mailing lists