[BISECTED] Regression v4.6-rc1 : nanosleep doesn't return on AT91SAM9

[Richard Genoud]

2016-05-10 16:47 GMT+02:00 Boris Brezillon <boris.brezillon at free-electrons.com>:
> Hi Richard,
>
> On Tue, 10 May 2016 16:25:25 +0200
> Richard Genoud <richard.genoud at gmail.com> wrote:
>
>> 2016-05-10 15:58 GMT+02:00 Richard Genoud <richard.genoud at gmail.com>:
>> > Hi,
>> >
>> > Starting with 4.6-rc1, a sleep 1 doesn't return :
>> >
>> > execve("/bin/sleep", ["sleep", "1"], ["USER=root", "SHLVL=1",
>> > "OLDPWD=/var", "HOME=/", "TERM=vt102",
>> > "PATH=/sbin:/usr/sbin:/bin:/usr/bin", "SHELL=/bin/sh", "PWD=/"]) = 0
>> > uname({sysname="Linux", nodename="LNS",
>> > release="4.5.0-09406-g46e595a17dcf", version="#72 Tue May 10 15:41:34
>> > CEST 2016", machine="armv5tejl", domainname="(none)"}) = 0
>> > brk(0)                                  = 0x153000
>> > brk(0x153d20)                           = 0x153d20
>> > set_tls(0x1534c0, 0x14d1e4, 0, 0x153d20, 0x14e368) = 0
>> > readlink("/proc/self/exe", "/bin/busybox", 4096) = 12
>> > brk(0x174d20)                           = 0x174d20
>> > brk(0x175000)                           = 0x175000
>> > getuid32()                              = 0
>> > rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0
>> > rt_sigaction(SIGCHLD, NULL, {SIG_DFL, [], 0}, 8) = 0
>> > rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
>> > nanosleep({1, 0},
>> >
>> > And the heartbeat led doesn't blink anymore (full on), as if the
>> > timers never fire.
>> >
>> > The platform is an AT91SAM9G35-CM.
>> >
>> > git bisect gives a merge as the 1st bad commit :(
>> > first bad commit: [46e595a17dcf11404f713845ecb5b06b92a94e43] Merge tag
>> > 'armsoc-drivers' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>>
>>
>> I went a little further in tracking that down, starting from the merge
>> (1st bad commit) and reverting some commits.
>> It seems that the culprit is :
>> commit 863a81c3be1d931bdae6426e231add9334311f13
>> Author: Boris Brezillon <boris.brezillon at free-electrons.com>
>> Date:   Fri Sep 5 09:54:13 2014 +0200
>>     clk: at91: make use of syscon to share PMC registers in several drivers
>>
>> Here are the commits I've reverted :
>>
>> git log --oneline --decorate -20
>> 849cdaa0cf04 (HEAD, tag: good) Revert "clk: at91: make use of syscon
>> to share PMC registers in several drivers"
>> 4bd6dde95321 (tag: last_bad) Revert "clk: at91: make use of
>> syscon/regmap internally"
>> a56dacaa24c4 (tag: always_bad) Revert "clk: at91: remove IRQ handling
>> and use polling"
>> 33edfe7a2de6 Revert "clk: at91: pmc: merge at91_pmc_init in atmel_pmc_probe"
>> 0b22bb649eae Revert "clk: at91: pmc: move pmc structures to C file"
>> 14994fdff366 Revert "ARM: at91: pm: simply call at91_pm_init"
>> 6f0821175327 (tag: still_bad) Revert "ARM: at91: pm: find and remap the pmc"
>> b52bbf7fad1b Revert "ARM: at91: pm: move idle functions to pm.c"
>> dffb1a5f68b9 Revert "ARM: at91: remove useless includes and function prototypes"
>> db352691dcc1 Revert "usb: gadget: atmel: access the PMC using regmap"
>> f7b5433dbbc8 Revert "clk: at91: pmc: drop at91_pmc_base"
>> 6e6dfc0d91c0 Revert "clk: at91: pmc: remove useless capacities handling"
>> ceddd3f88ada Revert "clk: at91: remove useless includes"
>> 46e595a17dcf (tag: bad) Merge tag 'armsoc-drivers' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>> e3b1f64e04f5 Merge tag 'armsoc-defconfig' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>> 915c56bc01d6 Merge tag 'armsoc-dt64' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>> 5a6b7e53d035 Merge tag 'armsoc-dt' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>> 8c6d4082fc6a Merge tag 'armsoc-arm64' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>> 33b3d2e88c9e Merge tag 'armsoc-soc' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>> e88fa1b8b00a Merge tag 'armsoc-cleanup' of
>> git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
>
> Thanks for the detailed report.
>
> Can you try again after applying this patch [1]?
>
> Regards,
>
> Boris
>
> [1]http://code.bulix.org/dx1c0t-98377

Yes, it works.
Big thumbs up, that was fast !

Less than 1 hour between the report and the patch ! :)

This is clearly 4.6 material (as it's a regression from 4.5), but it's
maybe a little late.
Anrd, Nicolas, what do you think ?


Thanks,
Richard.