Re: exynos4: cpu hotplug + cpuidle/AFTR broken

[Daniel Lezcano <daniel.lezcano@linaro.org>]

On 04/01/2014 01:04 PM, Krzysztof Kozlowski wrote:
>
> On Tue, 2014-04-01 at 12:43 +0200, Daniel Lezcano wrote:
>> Hi all,
>>
>> I tried the v3.14 kernel and unplugged the cpu1. That leads to a kernel
>> hang without any trace.
>>
>> If I disable the cpuidle AFTR state through sysfs before unplugging
>> cpu1, it works well.
>
> Hi,
>
> Can you describe the hardware (at least SoC)?

Yes, Exynos 4210.

>
> I encountered stalls after CPU hotplugging. In my case affected were
> SOC-s s where MCT local timer interrupt is shared (SPI). It could be
> reproduced easily by continuous hotplugging CPU1.
>
> Fix for one issue (CPU1 stall) in my case:
> http://thread.gmane.org/gmane.linux.kernel.stable/85047
> However I still encounter CPU0 stall in the same test case (hotplug
> CPU1).
>
> Maybe these issues are similar?

I don't think so. IMO, it is directly related to the AFTR state because 
disabling this state (via sysfs) before unplugging the cpu1, does *not* 
lead to a kernel hang and everything is working fine.

>> I reproduced the issue on v3.14, v3.13, v3.12, v3.11
>>
>> v3.10 works fine.
>>
>> I tried to git bisect but I finally gave up due to too much kernel
>> hanging at boot.
>>
>> The git bisect log below.
>>
>> Thanks
>>     -- Daniel
>>
>>
>> git bisect start
>> # bad: [6e4664525b1db28f8c4e1130957f70a94c19213e] Linux 3.11
>> git bisect bad 6e4664525b1db28f8c4e1130957f70a94c19213e
>> # good: [8bb495e3f02401ee6f76d1b1d77f3ac9f079e376] Linux 3.10
>> git bisect good 8bb495e3f02401ee6f76d1b1d77f3ac9f079e376
>> # good: [8b70a90cabafb6a6e1a0d3f838b38355fe48337e] Merge branch
>> 'for-v3.11' of git://git.linaro.org/people/mszyprowski/linux-dma-mapping
>> git bisect good 8b70a90cabafb6a6e1a0d3f838b38355fe48337e
>> # bad: [b41e6a51d57e231d2ed237f17f002cc536c0987c] sh_eth: SH_ETH should
>> depend on HAS_DMA
>> git bisect bad b41e6a51d57e231d2ed237f17f002cc536c0987c
>> # bad: [2e17c5a97e231f3cb426f4b7895eab5be5c5442e] Merge branch
>> 'drm-next' of git://people.freedesktop.org/~airlied/linux
>> git bisect bad 2e17c5a97e231f3cb426f4b7895eab5be5c5442e
>> # good: [e631227f698f39969eb476d297f3ac65b43b51a5] drm/radeon: fix
>> endian bug in radeon_atom_get_mclk_range_table()
>> git bisect good e631227f698f39969eb476d297f3ac65b43b51a5
>> # bad: [9a5889ae1ce41f376e6a5b56e17e0c5a755fda80] Merge branch
>> 'for-linus' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client
>> git bisect bad 9a5889ae1ce41f376e6a5b56e17e0c5a755fda80
>> # bad: [be0c5d8c0bb0023e11f5c6d38e90f7b0f24edb64] Merge tag
>> 'nfs-for-3.11-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
>> git bisect bad be0c5d8c0bb0023e11f5c6d38e90f7b0f24edb64
>> # bad: [d2b4a646717153a1a180b64d4a8464054dbd700e] Merge branch
>> 'for-linus' of git://git.infradead.org/users/vkoul/slave-dma
>> git bisect bad d2b4a646717153a1a180b64d4a8464054dbd700e
>> # skip: [07bd1172902e782f288e4d44b1fde7dec0f08b6f] tick: Sanitize
>> broadcast control logic
>> git bisect skip 07bd1172902e782f288e4d44b1fde7dec0f08b6f
>> # good: [757f4e51b74f7fb19251020e742a78111fdea194] MIPS: jz4740: Correct
>> clock gate bit for DMA controller
>> git bisect good 757f4e51b74f7fb19251020e742a78111fdea194
>> # skip: [064706514ec3fea740c2656e03c4f01f6a551ac4] clocksource: Add
>> generic dummy timer driver
>> git bisect skip 064706514ec3fea740c2656e03c4f01f6a551ac4
>> # skip: [55a68c23e0a675b2b8ac2656fd6edbf98b78e4c6] dw_apb_timer_of.c:
>> Remove parts that were picoxcell-specific
>> git bisect skip 55a68c23e0a675b2b8ac2656fd6edbf98b78e4c6
>> # skip: [fc1f7d5606487ae28d6c84e95401952927d7379e] clocksource:
>> apb_timer: Remove unsused function
>> git bisect skip fc1f7d5606487ae28d6c84e95401952927d7379e
>> # skip: [6cffe00f7d4e24679eae6b7aae4caaf915288256] alarmtimer: Add
>> functions for timerfd support
>> git bisect skip 6cffe00f7d4e24679eae6b7aae4caaf915288256
>> # good: [e24f6628811e2d4531b443684b598f7050932012] modpost: remove all
>> traces of cpuinit/cpuexit sections
>> git bisect good e24f6628811e2d4531b443684b598f7050932012
>> # skip: [629a6a2b7762a474177d08043bea094dd27e0a54] sched_clock: Add
>> temporary asm/sched_clock.h
>> git bisect skip 629a6a2b7762a474177d08043bea094dd27e0a54
>> # skip: [5c83545f24ab3dd67e0ae0e2b795fea750f08c34] power: Add option to
>> log time spent in suspend
>> git bisect skip 5c83545f24ab3dd67e0ae0e2b795fea750f08c34
>> # good: [2473f3e7a97ce8bc0fe7596cdb361b21221418eb] posix_cpu_timers:
>> consolidate expired timers check
>> git bisect good 2473f3e7a97ce8bc0fe7596cdb361b21221418eb
>> # skip: [336ae1180df5f69b9e0fb6561bec01c5f64361cf] ARM: sched_clock:
>> Load cycle count after epoch stabilizes
>> git bisect skip 336ae1180df5f69b9e0fb6561bec01c5f64361cf
>> # skip: [f5a2e34375a5e2b711aea488ac3ae50eeba6d57c] clocksource: Allow
>> clocksource select to skip current clocksource
>> git bisect skip f5a2e34375a5e2b711aea488ac3ae50eeba6d57c
>> # skip: [ce0b098981544d9f4e910ea48f3af3e726ca6d6b] x86: Fix
>> vrtc_get_time/set_mmss to use new timespec interface
>> git bisect skip ce0b098981544d9f4e910ea48f3af3e726ca6d6b
>> # skip: [d0951a23383d09276f7976ed34d8f1cede629b48] DMA: shdma:
>> shdma_chan_filter() has to be in shdma-base.h
>> git bisect skip d0951a23383d09276f7976ed34d8f1cede629b48
>> # skip: [7c4c3a0f18ba57ea2a2985034532303d2929902a] hrtimers: Support
>> resuming with two or more CPUs online (but stopped)
>> git bisect skip 7c4c3a0f18ba57ea2a2985034532303d2929902a
>> # skip: [a89c7edbe7d7aa80f507915f3dd801211b116b79] clocksource: Let
>> clocksource_unregister() return success/error
>> git bisect skip a89c7edbe7d7aa80f507915f3dd801211b116b79
>> # skip: [2699339361a9bacb3fa663e6b8981a040cfca4ee] clocksource:
>> vf_pit_timer: Use linux/sched_clock.h
>> git bisect skip 2699339361a9bacb3fa663e6b8981a040cfca4ee
>> # skip: [7172a286ced0c1f4f239a0fa09db54ed37d3ead2] clockevents: Get rid
>> of the notifier chain
>> git bisect skip 7172a286ced0c1f4f239a0fa09db54ed37d3ead2
>> # skip: [762cf9695d714d312ef7369bed1b9f9467c9e64e] Merge branch
>> 'timers/clockevents' of git://git.linaro.org/people/dlezcano/clockevents
>> into timers/core
>> git bisect skip 762cf9695d714d312ef7369bed1b9f9467c9e64e
>> # bad: [21884a83b2192a00885d7244a1dda32debd2fbc7] Merge branch
>> 'timers-core-for-linus' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
>> git bisect bad 21884a83b2192a00885d7244a1dda32debd2fbc7
>> # skip: [5a9b5855c248d0298eac4d5490c7bc11c2f1983b] ARM: sched_clock:
>> Remove unused needs_suspend member
>> git bisect skip 5a9b5855c248d0298eac4d5490c7bc11c2f1983b
>> # skip: [45cb8e01b2ecef1c2afb18333e95793fa1a90281] clockevents: Split
>> out selection logic
>> git bisect skip 45cb8e01b2ecef1c2afb18333e95793fa1a90281
>> # skip: [ffbfb5e316f0db486798ccf1db36a577ffe79637] ARM: sched_clock:
>> Return suspended count earlier
>> git bisect skip ffbfb5e316f0db486798ccf1db36a577ffe79637
>> # skip: [04397fe94ad65289884b9862b6a0c722ececaadf] timekeeping: Pass
>> flags instead of multiple bools to timekeeping_update()
>> git bisect skip 04397fe94ad65289884b9862b6a0c722ececaadf
>> # skip: [11682a41618f8094cb7a9330b4b6a12ffaef5774] alarmtimer: Export
>> symbols of functions declared in linux/alarmtimer.h
>> git bisect skip 11682a41618f8094cb7a9330b4b6a12ffaef5774
>> # skip: [fc1f7d5606487ae28d6c84e95401952927d7379e] clocksource:
>> apb_timer: Remove unsused function
>> git bisect skip fc1f7d5606487ae28d6c84e95401952927d7379e
>>
>>
>


-- 
  <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog