Re: [PATCH V5 0/7] cpufreq: suspend early/resume late: dpm_{suspend|resume}()

From: Stephen Warren <swarren@wwwdotorg.org>
To: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Viresh Kumar <viresh.kumar@linaro.org>,
	linaro-kernel@lists.linaro.org, cpufreq@vger.kernel.org,
	linux-pm@vger.kernel.org, linux-kernel@vger.kernel.org,
	nm@ti.com, kgene.kim@samsung.com, jinchoi@broadcom.com,
	tianyu.lan@intel.com, sebastian.capella@linaro.org,
	jhbird.choi@samsung.com
Subject: Re: [PATCH V5 0/7] cpufreq: suspend early/resume late: dpm_{suspend|resume}()
Date: Fri, 14 Feb 2014 17:03:22 -0700	[thread overview]
Message-ID: <52FEAECA.5080905@wwwdotorg.org> (raw)
In-Reply-To: <2869027.iYHxWOSS2C@vostro.rjw.lan>

On 02/14/2014 03:23 PM, Rafael J. Wysocki wrote:
> On Friday, February 14, 2014 12:42:53 PM Stephen Warren wrote:
>> On 02/12/2014 11:50 PM, Viresh Kumar wrote:
>>> This patchset creates/calls cpufreq suspend/resume callbacks from dpm_{suspend|resume}()
>>> for handling suspend/resume of cpufreq governors and core.
>>
>> Are these patches for 3.14 or 3.15?
> 
> I don't think they are suitable for 3.14, honestly.
> 
>> I ask because I just tested Linus's master from a few days back (some
>> point after v3.14-rc2; commit 9398a10cd964 Merge tag
>> 'regulator-v3.14-rc2'), and I see a lot of the following messages during
>> suspend and/or resume (about 2-7 times, perhaps more of them from the
>> resume path, but it's hard to tell):
>>
>> cpufreq: __cpufreq_driver_target: Failed to change cpu frequency: -16
>>
>> This series does appear to solve those, so I think at least part of it
>> needs to be applied for 3.14.
> 
> Well, it would be good to verify which part, then.

Patch 2/7 appears to stop that message from being printed during
suspend, and perhaps reduce the number of times it's printed during
resume. Patch 7/7 stops the message being printed at all.

Looking at patch 7, I wonder if it's simply because tegra_target() was
modified never to return -EBUSY, so the bug is still there, but it's
just been hidden.

>> Also, I sometimes see the following during resume. I saw it twice with
>> Linus's tree, but then I did 10 more reboot+suspend+resume cycles and
>> couldn't repro it, and I saw it once with Linus's tree plus this series
>> applied, then couldn't reproduce it in 5 more tries.

Oops. I screwed up my re-testing (tested on the wrong board, without
cpufreq active:-/). The message below is reproducible 100% of the time
with or without this series.

>>> [ 48.500410] ------------[ cut here ]------------
>>> [ 48.505252] WARNING: CPU: 0 PID: 877 at fs/sysfs/dir.c:52 sysfs_warn_dup+0x68/0x84()
>>> [ 48.513005] sysfs: cannot create duplicate filename '/devices/system/cpu/cpu1/cpufreq'
>>> [ 48.520995] Modules linked in: brcmfmac brcmutil
>>> [ 48.525740] CPU: 0 PID: 877 Comm: test-rtc-resume Not tainted 3.14.0-rc2-00259-g9398a10cd964 #12
...
>>> [ 48.575024] [<c012776c>] (sysfs_warn_dup) from [<c0127a54>] (sysfs_do_create_link_sd+0xb0/0xb8)
>>> [ 48.583772] [<c0127a54>] (sysfs_do_create_link_sd) from [<c038ef64>] (__cpufreq_add_dev.isra.27+0x2a8/0x814)
>>> [ 48.593701] [<c038ef64>] (__cpufreq_add_dev.isra.27) from [<c038f548>] (cpufreq_cpu_callback+0x70/0x8c)
>>> [ 48.603198] [<c038f548>] (cpufreq_cpu_callback) from [<c0043864>] (notifier_call_chain+0x44/0x84)
>>> [ 48.612166] [<c0043864>] (notifier_call_chain) from [<c0025f60>] (__cpu_notify+0x28/0x44)
>>> [ 48.620424] [<c0025f60>] (__cpu_notify) from [<c00261e8>] (_cpu_up+0xf0/0x140)
>>> [ 48.627748] [<c00261e8>] (_cpu_up) from [<c0569eb8>] (enable_nonboot_cpus+0x68/0xb0)
>>> [ 48.635579] [<c0569eb8>] (enable_nonboot_cpus) from [<c006339c>] (suspend_devices_and_enter+0x198/0x2dc)