Content-Type: multipart/mixed; boundary="===============7331898122518673369==" MIME-Version: 1.0 From: Alexey Gladkov To: lkp@lists.01.org Subject: Re: 08ed4efad6: stress-ng.sigsegv.ops_per_sec -41.9% regression Date: Thu, 08 Apr 2021 18:44:26 +0200 Message-ID: <20210408164426.o5cfvv3ixowsto62@example.org> In-Reply-To: List-Id: --===============7331898122518673369== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On Thu, Apr 08, 2021 at 09:22:40AM -0700, Linus Torvalds wrote: > On Thu, Apr 8, 2021 at 1:32 AM kernel test robot wrote: > > > > FYI, we noticed a -41.9% regression of stress-ng.sigsegv.ops_per_sec du= e to commit > > 08ed4efad684 ("[PATCH v10 6/9] Reimplement RLIMIT_SIGPENDING on top of = ucounts") > = > Ouch. > = > I *think* this test may be testing "send so many signals that it > triggers the signal queue overflow case". > = > And I *think* that the performance degradation may be due to lots of > unnecessary allocations, because ity looks like that commit changes > __sigqueue_alloc() to do > = > struct sigqueue *q =3D kmem_cache_alloc(sigqueue_cachep, flags); > = > *before* checking the signal limit, and then if the signal limit was > exceeded, it will just be free'd instead. > = > The old code would check the signal count against RLIMIT_SIGPENDING > *first*, and if there were m ore pending signals then it wouldn't do > anything at all (including not incrementing that expensive atomic > count). > = > Also, the old code was very careful to only do the "get_user()" for > the *first* signal it added to the queue, and do the "put_user()" for > when removing the last signal. Exactly because those atomics are very > expensive. > = > The new code just does a lot of these atomics unconditionally. Yes and right now I'm trying to rewrite this patch. > I dunno. The profile data in there is a bit hard to read, but there's > a lot more cachee misses, and a *lot* of node crossers: > = > > 5961544 +190.4% 17314361 perf-stat.i.cache-misses > > 22107466 +119.2% 48457656 perf-stat.i.cache-referen= ces > > 163292 =C4=85 3% +4582.0% 7645410 perf-stat.i.node-loa= d-misses > > 227388 =C4=85 2% +3708.8% 8660824 perf-stat.i.node-loa= ds > = > and (probably as a result) average instruction costs have gone up enormou= sly: > = > > 3.47 +66.8% 5.79 perf-stat.overall.cpi > > 22849 -65.6% 7866 perf-stat.overall.cycles-= between-cache-misses > = > and it does seem to be at least partly about "put_ucounts()": > = > > 0.00 +4.5 4.46 perf-profile.calltrace.cy= cles-pp.put_ucounts.__sigqueue_free.get_signal.arch_do_signal_or_restart.ex= it_to_user_mode_prepare > = > and a lot of "get_ucounts()". > = > But it may also be that the new "get sigpending" is just *so* much > more expensive than it used to be. Thanks for decrypting this! I spent some time to understand this report and still wasn't sure I understood it. -- = Rgrds, legion --===============7331898122518673369==--