Date: Tue, 28 Sep 2021 21:41:03 +0900 From: Tomoaki AOKI <junchoon@dec.sakura.ne.jp> To: freebsd-current@freebsd.org Cc: Konstantin Belousov <kostikbel@gmail.com>, Alexander Motin <mav@FreeBSD.org>, joh.hendriks@gmail.com Subject: Re: latest current fails to boot. Message-ID: <20210928214103.32b36a3b627b51700389970a@dec.sakura.ne.jp> In-Reply-To: <9a9a43f9-0474-0c32-afb9-13c2f7c33e51@FreeBSD.org> References: <fa4768fb-4524-551d-1c8d-f144d6e5995b@gmail.com> <YUsl8sSAY0ybMWUI@albert.catwhisker.org> <20210922230905.79a494a4ddbecd7d6b8179ea@dec.sakura.ne.jp> <wnn5-l0uf-wny@FreeBSD.org> <fa7cbdff-fdf0-3994-bd34-1afb108ce964@gmail.com> <36221848-2b85-4a60-fd27-d4bf83d4571e@FreeBSD.org> <20210926102347.c538169dddec5f2ff50d8e4c@dec.sakura.ne.jp> <YU/RHltUD5cHm1%2BK@kib.kiev.ua> <9a9a43f9-0474-0c32-afb9-13c2f7c33e51@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, 26 Sep 2021 12:16:55 -0400 Alexander Motin <mav@FreeBSD.org> wrote: > Thank you for the notification. 08063e9f98a should fix the hang. Thanks for the fix! Rebuilt, Several reboots with varying 0, 1, commented out whole line for kern.sched.steal_thresh and all went fine. > I just want to add that lowering kern.sched.steal_thresh below 2 should > not be a proper fix for any problem, but only a workaround. I guess > either some CPU can't wake up from sleep for too long, or the wakeup > interrupt is not properly sent to it when the load is assigned. In such > case stealing makes other CPU to do the work instead. It would be good > to find and fix the real problem. For me (with Core i7-8750H), lowering kern.sched.steal_thresh didn't made significant improvement but had no reason to go back to default. *As sched_ule got modified now, I'm trying default kern.sched.steal_thresh value now. In the other hand, at least some Ryzen users seem to have much more severe problem than me and the workaround make significant imrovement. > On 25.09.2021 21:47, Konstantin Belousov wrote: > > On Sun, Sep 26, 2021 at 10:23:47AM +0900, Tomoaki AOKI wrote: > >> On Sat, 25 Sep 2021 23:46:48 +0300 > >> Andriy Gapon <avg@FreeBSD.org> wrote: > >> > >>> On 25/09/2021 19:10, Johan Hendriks wrote: > >>>> For me i had kern.sched.steal_thresh=1 in my sysctl as i use this machine mainly > >>>> for tests and so on. > >>>> By removing this sysctl the system boots again. I already used the latest > >>>> snapshot and that booted fine. > >>> > >>> Might have something to do with > >>> https://cgit.FreeBSD.org/src/commit/?id=bd84094a51c4648a7c97ececdaccfb30bc832096 > >>> > >>> -- > >>> Andriy Gapon > >> > >> Commenting out kern.sched.steal_thresh=0 line in /etc/sysctl.conf let > >> me boot fine. No other setting of kern.sched.* affected. > >> I've introduced the setting by reading posts [1] and [2] on > >> freebsd-current ML. Thanks for the hint, Jan! > >> > >> Andriy, I took time to bi-sect and determined the commit triggered > >> this issue was e745d729be60. [3] > >> Worked OK even with kern.sched.steal_thresh=0 at a342ecd326ee. [4] > >> > >> Tested commits are as below (tested order, not using git bisect): > >> 0b79a76f8487: [Known to be OK] > >> 8db1669959ce: [Problematic rev I first encountered] > >> 0f6829488ef3: OK > >> df8dd6025af8: OK > >> 4f917847c903: OK > >> e745d729be60: NG! > >> bd84094a51c4: OK > >> a342ecd326ee: OK > >> > >> Konstantin, no more chance to get into ddb on hang up until my previous > >> post. ^T never worked on hang up situation. Sory. But does info above > >> help? > > Let the author of the commit look. > > > >> > >> > >> [1] > >> https://lists.freebsd.org/pipermail/freebsd-current/2021-March/079237.html > >> > >> [2] > >> https://lists.freebsd.org/pipermail/freebsd-current/2021-March/079240.html > >> > >> [3] > >> https://lists.freebsd.org/pipermail/dev-commits-src-main/2021-September/007513.html > >> > >> [4] > >> https://lists.freebsd.org/pipermail/dev-commits-src-main/2021-September/007512.html > >> > >> -- > >> Tomoaki AOKI <junchoon@dec.sakura.ne.jp> > > -- > Alexander Motin > -- Tomoaki AOKI <junchoon@dec.sakura.ne.jp>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20210928214103.32b36a3b627b51700389970a>