Date: Thu, 14 Mar 2019 00:58:39 +0000 (UTC) From: Alexander Motin <mav@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-stable@freebsd.org, svn-src-stable-12@freebsd.org Subject: svn commit: r345122 - stable/12/sys/cddl/contrib/opensolaris/uts/common/fs/zfs Message-ID: <201903140058.x2E0wdqD082710@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mav Date: Thu Mar 14 00:58:39 2019 New Revision: 345122 URL: https://svnweb.freebsd.org/changeset/base/345122 Log: MFC r344903: Improve entropy for ZFS taskqueue selection. I just found that at least on Skylake CPUs cpu_ticks() never returns odd values, only even, and possibly has even bigger step (176/2?), that makes its lower bits very bad entropy source, leaving half of taskqueues unused. Switch to sbinuptime(), closer to upstreams, mitigates the problem by the rate conversion working as kind of hash function. In case that is somehow not enough (timer rate is too low or too divisible) mix in curcpu. Modified: stable/12/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c Directory Properties: stable/12/ (props changed) Modified: stable/12/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c ============================================================================== --- stable/12/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c Thu Mar 14 00:57:54 2019 (r345121) +++ stable/12/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/spa.c Thu Mar 14 00:58:39 2019 (r345122) @@ -1070,7 +1070,8 @@ spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_t tq = tqs->stqs_taskq[0]; } else { #ifdef _KERNEL - tq = tqs->stqs_taskq[cpu_ticks() % tqs->stqs_count]; + tq = tqs->stqs_taskq[(u_int)(sbinuptime() + curcpu) % + tqs->stqs_count]; #else tq = tqs->stqs_taskq[gethrtime() % tqs->stqs_count]; #endif
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201903140058.x2E0wdqD082710>