linux-kernel - Re: REGRESSION: kernel BUG at arch/arm64/kernel/alternative.c:157!

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
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <878s4cv20v.wl-maz@kernel.org>
Date:   Tue, 18 May 2021 10:12:32 +0100
From:   Marc Zyngier <maz@...nel.org>
To:     John Stultz <john.stultz@...aro.org>
Cc:     Catalin Marinas <catalin.marinas@....com>,
        Kees Cook <keescook@...omium.org>,
        Will Deacon <will@...nel.org>,
        Sami Tolvanen <samitolvanen@...gle.com>,
        linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Bjorn Andersson <bjorn.andersson@...aro.org>,
        YongQin Liu <yongqin.liu@...aro.org>,
        Amit Pundir <amit.pundir@...aro.org>,
        Michael Walle <michael@...le.cc>
Subject: Re: REGRESSION: kernel BUG at arch/arm64/kernel/alternative.c:157!

+ Michael

On Mon, 17 May 2021 22:52:59 +0100,
John Stultz <john.stultz@...aro.org> wrote:
> 
> With v5.13-rc2, I've been seeing an odd boot regression with the
> DragonBoard 845c:
> 
> Unfortunately, trying to bisect it down (v5.13-rc1 works ok) is giving
> me inconsistent results so far. It feels a bit like maybe some config
> option gets enabled moving forward, and then sticks around when we go
> back.  I'll take another swing at bisecting it later today, but I have
> to move on to some other work right now, so I figured I'd share (with
> folks who better know the recent __apply_alternatives changes) in case
> folks have a better idea:
> 
> [    0.254384] CPU features: detected: RAS Extension Support
> [    0.259928] CPU: All CPU(s) started at EL1
> [    0.264127] alternatives: patching kernel code
> [    0.268635] ------------[ cut here ]------------
> [    0.273303] kernel BUG at arch/arm64/kernel/alternative.c:157!
> [    0.279192] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
> [    0.284736] Modules linked in:
> [    0.287833] CPU: 0 PID: 14 Comm: migration/0 Not tainted
> 5.13.0-rc2-mainline #4501
> [    0.295472] Hardware name: Thundercomm Dragonboard 845c (DT)
> [    0.301182] Stopper: multi_cpu_stop+0x0/0x1a0 <-
> stop_machine_cpuslocked+0x128/0x160
> [    0.309020] pstate: 204000c5 (nzCv daIF +PAN -UAO -TCO BTYPE=--)
> [    0.315086] pc : __apply_alternatives+0x1f0/0x270
> [    0.319847] lr : __apply_alternatives+0xf4/0x270
> [    0.324515] sp : ffffffc01020bca0
> [    0.327874] x29: ffffffc01020bca0 x28: 00000000000000a0 x27: ffffffd7f5c11124
> [    0.335086] x26: ffffffd7f5c11128 x25: 00000000001b0020 x24: ffffffd7f700ab90
> [    0.342297] x23: 0000000000000000 x22: ffffffc01020bd20 x21: ffffffd7f7bea374
> [    0.349508] x20: ffffffc01020bd30 x19: ffffffd7f72194fc x18: ffffffffffffffff
> [    0.356718] x17: ffffffd7f7bdce40 x16: 000000005c8e1b43 x15: ffffffd7f76d9d10
> [    0.363929] x14: ffffffc09020b967 x13: ffffffc01020b975 x12: ffffffd7f76d9e30
> [    0.371140] x11: 0000000005f5e0ff x10: ffffffc01020b8c0 x9 : 00000000ffffffd0
> [    0.378350] x8 : 6b20676e69686374 x7 : ffffffd7f79b9238 x6 : c0000000ffff7fff
> [    0.385560] x5 : 0000000000000000 x4 : ffffffd7f5c22898 x3 : 0000000000000010
> [    0.392771] x2 : 0000000000000004 x1 : 0000000000000000 x0 : 000000000000003f
> [    0.399982] Call trace:
> [    0.402461]  __apply_alternatives+0x1f0/0x270
> [    0.406873]  __apply_alternatives_multi_stop+0xc0/0xe0
> [    0.412062]  multi_cpu_stop+0xb8/0x1a0
> [    0.415851]  cpu_stopper_thread+0xac/0x120
> [    0.419997]  smpboot_thread_fn+0x200/0x238
> [    0.424146]  kthread+0x14c/0x158
> [    0.427423]  ret_from_fork+0x10/0x1c
> [    0.431045] Code: 39402e61 39402a62 6b01005f 54fff500 (d4210000)
> [    0.437199] ---[ end trace 523e13d9d60a992d ]---
> [    0.441868] note: migration/0[14] exited with preempt_count 2
> [    0.447739] migration/0 (14) used greatest stack depth: 12448 bytes left

[/me digs in my IRC logs]

This looks a lot like an issue that was reported my Michael Walle a
few days ago on IRC, leading to a crash that looked like this:

[    0.325238] alternatives: patching kernel code
[    0.329735] ------------[ cut here ]------------
[    0.334394] kernel BUG at arch/arm64/kernel/alternative.c:157!
[    0.340300] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[    0.345836] Modules linked in:
[    0.348916] CPU: 0 PID: 14 Comm: migration/0 Not tainted 5.13.0-rc1-next-20210511+ #536
[    0.356998] Hardware name: Kontron SMARC-sAL28 (Single PHY) on SMARC Eval 2.0 carrier (DT)
[    0.365339] Stopper: multi_cpu_stop+0x0/0x1a8 <- stop_cpus.constprop.9+0x78/0xc8
[    0.372820] pstate: 200000c5 (nzCv daIF -PAN -UAO -TCO BTYPE=--)
[    0.378882] pc : __apply_alternatives.isra.1+0x1c4/0x270
[    0.384246] lr : __apply_alternatives.isra.1+0x110/0x270
[    0.389606] sp : ffff800012db3ca0
[    0.392946] x29: ffff800012db3ca0 x28: 0000000000000000 x27: ffff800010011924
[    0.400155] x26: ffff800010011928 x25: 00000000001b0020 x24: ffff8000115ad350
[    0.407364] x23: ffff800012db3d28 x22: 0000000000000000 x21: ffff800011fb24cd
[    0.414571] x20: ffff800012db3d30 x19: ffff800011840b38 x18: 0000000000000010
[    0.421779] x17: 0000000044a56c23 x16: 0000000000000002 x15: ffffffffffffffff
[    0.428986] x14: ffff800011d50a48 x13: ffff800092db3987 x12: ffff800011de6a70
[    0.436193] x11: 0000000000000003 x10: ffff800011dcea30 x9 : ffff8000105d8928
[    0.443401] x8 : 0000000000017fe8 x7 : c0000000ffffefff x6 : 0000000000000001
[    0.450608] x5 : 0000000000000000 x4 : ffff800010024398 x3 : 0000000000000010
[    0.457815] x2 : 0000000000000004 x1 : 0000000000000000 x0 : 000000000000003f
[    0.465022] Call trace:
[    0.467483]  __apply_alternatives.isra.1+0x1c4/0x270
[    0.472493]  __apply_alternatives_multi_stop+0xcc/0xe0
[    0.477679]  multi_cpu_stop+0xac/0x1a8
[    0.481460]  cpu_stopper_thread+0xa4/0x138
[    0.485592]  smpboot_thread_fn+0x12c/0x268
[    0.489725]  kthread+0x164/0x168
[    0.492980]  ret_from_fork+0x10/0x30
[    0.496588] Code: 39402e61 39402a62 6b01005f 54fff6a0 (d4210000) 
[    0.502742] ---[ end trace 24ef7d65759ab825 ]---
[    0.507398] note: migration/0[14] exited with preempt_count 2
[    0.513290] ------------[ cut here ]------------

Michael subsequently reported that:

<quote>
 mhh nevermind, I can't reproduce it anymore. Maybe I should have
 recompiled with a clean build dir at first
</quote>

My gut feeling is that we can end-up with some build leftovers when
going between -rc1 and -rc2, hence the screw-up when the capabilities
get reordered. Dependency issues?

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.