| 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 linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Wed, 15 Jul 2009 22:15:42 +0200 From: Paweł Staszewski <pstaszewski@...are.pl> To: Jarek Poplawski <jarkao2@...il.com> CC: Eric Dumazet <dada1@...mosbay.com>, Eric Dumazet <eric.dumazet@...il.com>, Linux Network Development list <netdev@...r.kernel.org> Subject: Re: weird problem Jarek Poplawski pisze: > On Tue, Jul 14, 2009 at 01:26:46AM +0200, Paweł Staszewski wrote: > >> Jarek Poplawski pisze: >> >>> On Fri, Jul 10, 2009 at 04:47:54PM +0200, Jarek Poplawski wrote: >>> >>> >>>> On Fri, Jul 10, 2009 at 01:59:00AM +0200, Paweł Staszewski wrote: >>>> >>>> >>>>> Today i make other tests with change of >>>>> /proc/sys/net/ipv4/rt_cache_rebuild_count and kernel 2.6.30.1 >>>>> >>>>> And when rt_cache_rebuild_count is set to "-1" i have always load >>>>> on x86_64 machine approx 40-50% of each cpu where network card is >>>>> binded by irq_aff >>>>> >>>>> when rt_cache_rebuild_count is set to more than "-1" i have 15 to >>>>> 20 sec of 1 to 3% cpu and after 40-50% cpu >>>>> >>>>> >>>> ... >>>> >>>> Here is one more patch for testing (with caution!). It adds possibility >>>> to turn off cache disabling (so it should even more resemble 2.6.28) >>>> after setting: rt_cache_rebuild_count = 0 >>>> >>>> I'd like you to try this patch: >>>> 1) together with the previous patch and "rt_cache_rebuild_count = 0" >>>> to check if there is still the difference wrt. 2.6.28; Btw., let >>>> me know which /proc/sys/net/ipv4/route/* settings do you need to >>>> change and why >>>> >>>> 2) alone (without the previous patch) and "rt_cache_rebuild_count = 0" >>>> >>>> 3) if it's possible to try 2.6.30.1 without these patches, but with >>>> default /proc/sys/net/ipv4/route/* settings, and higher >>>> rt_cache_rebuild_count, e.g. 100; I'm interested if/how long it >>>> takes to trigger higher cpu load and the warning "... rebuilds is >>>> over limit, route caching disabled"; (Btw., I wonder why you didn't >>>> mention about these or maybe also other route caching warnings?) >>>> >>>> >>> Here is take 2 to respect setting "rt_cache_rebuild_count = 0" even >>> after cache rebuild counter has been increased earlier. (Btw, don't >>> forget about this setting after going back to vanilla kernel.) >>> >>> >>> >> Applied to 2.6.30.1 >> 1) With >> >> rt_cache_rebuild_count = 0 >> grep . /proc/sys/net/ipv4/route/* >> /proc/sys/net/ipv4/route/error_burst:1250 >> /proc/sys/net/ipv4/route/error_cost:250 >> /proc/sys/net/ipv4/route/gc_elasticity:4 >> /proc/sys/net/ipv4/route/gc_interval:15 >> /proc/sys/net/ipv4/route/gc_min_interval:0 >> /proc/sys/net/ipv4/route/gc_min_interval_ms:0 >> /proc/sys/net/ipv4/route/gc_thresh:190536 >> /proc/sys/net/ipv4/route/gc_timeout:15 >> /proc/sys/net/ipv4/route/max_size:1524288 >> /proc/sys/net/ipv4/route/min_adv_mss:256 >> /proc/sys/net/ipv4/route/min_pmtu:552 >> /proc/sys/net/ipv4/route/mtu_expires:600 >> /proc/sys/net/ipv4/route/redirect_load:5 >> /proc/sys/net/ipv4/route/redirect_number:9 >> /proc/sys/net/ipv4/route/redirect_silence:5120 >> /proc/sys/net/ipv4/route/secret_interval:3600 >> >> I tune this route parameters after looking of traffic/route cache to have not many entries in cache that are not needed anymore >> so gc_timeout = 15 >> limit of max entries = 1524288 >> And make route cahce a little more "faster" for me after tune >> gc_elasticity >> secret_interval >> gc_interval >> gc_thresh >> >> So with this parameters 15 sec of something like this: >> 00:41:23 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:41:24 all 0.00 0.00 0.12 0.00 1.49 10.46 0.00 0.00 87.92 >> 00:41:24 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:24 1 0.00 0.00 0.00 0.00 4.00 36.00 0.00 0.00 60.00 >> 00:41:24 2 0.00 0.00 0.00 0.00 8.91 47.52 0.00 0.00 43.56 >> 00:41:24 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:24 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:24 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:24 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:24 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> and 15 sec of something like this: >> 00:41:44 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:41:45 all 0.00 0.00 0.00 0.00 0.00 0.42 0.00 0.00 99.58 >> 00:41:45 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:45 1 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 99.00 >> 00:41:45 2 0.00 0.00 0.00 0.00 0.00 2.04 0.00 0.00 97.96 >> 00:41:45 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:45 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:45 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:45 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:41:45 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> So i change /proc/sys/net/ipv4/route/gc_timeout to 1 >> with rt_cache_rebuild_count = 0 >> And output is like 20 sec of something like this >> 00:48:52 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:48:53 all 0.00 0.00 0.19 0.00 0.19 0.58 0.00 0.00 99.03 >> 00:48:53 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:53 1 0.00 0.00 0.99 0.00 0.99 0.00 0.00 0.00 98.02 >> 00:48:53 2 0.00 0.00 0.00 0.00 0.00 2.00 0.00 0.00 98.00 >> 00:48:53 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:53 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:53 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:53 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:53 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> and after this two second of something like this: >> 00:48:49 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:48:50 all 0.00 0.00 0.09 0.00 0.27 2.17 0.00 0.00 97.46 >> 00:48:50 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:50 1 0.00 0.00 0.00 0.00 1.96 6.86 0.00 0.00 91.18 >> 00:48:50 2 0.00 0.00 0.00 0.00 0.99 16.83 0.00 0.00 82.18 >> 00:48:50 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:50 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:50 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:50 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:50 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> 00:48:50 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:48:51 all 0.00 0.00 0.00 0.00 1.86 10.41 0.00 0.00 87.73 >> 00:48:51 0 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 99.00 >> 00:48:51 1 0.00 0.00 0.00 0.00 4.85 26.21 0.00 0.00 68.93 >> 00:48:51 2 0.00 0.00 1.00 0.00 5.00 29.00 0.00 0.00 65.00 >> 00:48:51 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:51 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:51 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:51 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:48:51 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> > > Could you remind us how it differs from 2.6.28 with the same settings? > With the same settings and 2.6.28 there was always cpu load from 1% to 3% with gc_timeout = 15 > >> Another test: >> >> gc_timeout = 1 >> rt_cache_rebuild_count = 100 >> 10 to 14 sec of something like this: >> 00:51:36 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:51:37 all 0.00 0.00 0.00 0.00 0.00 0.27 0.00 0.00 99.73 >> 00:51:37 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:51:37 1 0.00 0.00 0.00 0.00 0.00 2.00 0.00 0.00 98.00 >> 00:51:37 2 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 99.00 >> 00:51:37 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:51:37 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:51:37 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:51:37 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:51:37 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> and two seconds of 10 to 30% cpu load more >> >> >> 2). >> Only last patch and almost all the time output like this >> 00:59:49 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 00:59:50 all 0.00 0.00 0.13 0.00 1.73 8.00 0.00 0.00 90.13 >> 00:59:50 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:59:50 1 0.00 0.00 0.00 0.00 4.00 24.00 0.00 0.00 72.00 >> 00:59:50 2 0.00 0.00 0.00 0.00 8.91 34.65 0.00 0.00 56.44 >> 00:59:50 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:59:50 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:59:50 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:59:50 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 00:59:50 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> sometimes after 15 to 30 sec i have 1 to 2% cpu load >> > > And how long do you have this 1 to 2% load? Is it with: > rt_cache_rebuild_count = 0 > gc_timeout = 1? > Maybe you could describe the main difference with or without the first > patch? > > >> 3). >> >> with default settings and without this patch i have almost all the time output like this: >> > > You mean without these two patches, right? So, there is no breaks with > less load like above? > > Yes. >> 01:21:40 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 01:21:41 all 0.00 0.00 0.00 0.00 2.14 10.97 0.00 0.00 86.89 >> 01:21:41 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:21:41 1 0.00 0.00 0.00 0.00 6.93 34.65 0.00 0.00 58.42 >> 01:21:41 2 0.00 0.00 0.00 0.00 7.07 42.42 0.00 0.00 50.51 >> 01:21:41 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:21:41 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:21:41 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:21:41 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:21:41 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> >> >> with my settings: >> /proc/sys/net/ipv4/route/error_burst:1250 >> /proc/sys/net/ipv4/route/error_cost:250 >> /proc/sys/net/ipv4/route/gc_elasticity:4 >> /proc/sys/net/ipv4/route/gc_interval:15 >> /proc/sys/net/ipv4/route/gc_min_interval:0 >> /proc/sys/net/ipv4/route/gc_min_interval_ms:0 >> /proc/sys/net/ipv4/route/gc_thresh:190536 >> /proc/sys/net/ipv4/route/gc_timeout:15 >> /proc/sys/net/ipv4/route/max_size:1524288 >> /proc/sys/net/ipv4/route/min_adv_mss:256 >> /proc/sys/net/ipv4/route/min_pmtu:552 >> /proc/sys/net/ipv4/route/mtu_expires:600 >> /proc/sys/net/ipv4/route/redirect_load:5 >> /proc/sys/net/ipv4/route/redirect_number:9 >> /proc/sys/net/ipv4/route/redirect_silence:5120 >> /proc/sys/net/ipv4/route/secret_interval:3600 >> >> >> 15 sec of 30 to 50 % cpu and 15 sec 1 to 2 % cpu >> >> with /proc/sys/net/ipv4/route/gc_interval:1 >> almost all the time like this >> 01:23:45 CPU %usr %nice %sys %iowait %irq %soft %steal %guest %idle >> 01:23:46 all 0.00 0.00 0.00 0.00 0.00 0.12 0.00 0.00 99.88 >> 01:23:46 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:23:46 1 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 99.00 >> 01:23:46 2 0.00 0.00 0.00 0.00 0.00 1.02 0.00 0.00 98.98 >> 01:23:46 3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:23:46 4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:23:46 5 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:23:46 6 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> 01:23:46 7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 100.00 >> >> with max two outputs of 20 to 30% cpu in different times from 12 to 15sec >> > > Didn't you see any: "... rebuilds is over limit, route caching > disabled" warning? > > No i don't any info. >> And i dont know but i think patch for turning off route cache is not >> working because with this patches and rt_cache_rebuild_count = 0 >> > > If you mean the patch #2, it does something opposite: with > rt_cache_rebuild_count = 0 it turns off automatic "cache disabling" > after rt_cache_rebuild_count events signaled with the above-mentionned > warning, which was introduced in 2.6.29. Sorry for not describing this > enough. > > Thanks, > Jarek P. > > > So is there some patch or there will be patch that turn off definitely route cache ? For now i use gc_timeout = 1 in my routers and all is working fine - there is only 1 second of 20% of cpu load after every 20 sec. Regards Pawel Staszewski -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@...r.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists