Re: [PATCH v2] platform/chrome: Expose resume result via debugfs

[Enric Balletbo i Serra]

Hi Evan,

On 27/6/19 18:07, Evan Green wrote:
> On Wed, Jun 26, 2019 at 1:55 PM Enric Balletbo i Serra
> <enric.balletbo@xxxxxxxxxxxxx> wrote:
>>
>> Hi Evan,
>>
>> Two few comments and I think I'm fine with it.
>>
>> On 25/6/19 15:05, Lee Jones wrote:
>>> On Mon, 17 Jun 2019, Evan Green wrote:
>>>
>>>> For ECs that support it, the EC returns the number of slp_s0
>>>> transitions and whether or not there was a timeout in the resume
>>>> response. Expose the last resume result to usermode via debugfs so
>>>> that usermode can detect and report S0ix timeouts.
>>>>
>>>> Signed-off-by: Evan Green <evgreen@xxxxxxxxxxxx>
>>>
>>> This still needs a platform/chrome Ack.
>>>
>>>> ---
>>>>
>>>> Changes in v2:
>>>>  - Moved from sysfs to debugfs (Enric)
>>>>  - Added documentation (Enric)
>>>>
>>>>
>>>> ---
>>>>  Documentation/ABI/testing/debugfs-cros-ec | 22 ++++++++++++++++++++++
>>>>  drivers/mfd/cros_ec.c                     |  6 +++++-
>>>>  drivers/platform/chrome/cros_ec_debugfs.c |  7 +++++++
>>>>  include/linux/mfd/cros_ec.h               |  1 +
>>>>  4 files changed, 35 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/Documentation/ABI/testing/debugfs-cros-ec b/Documentation/ABI/testing/debugfs-cros-ec
>>>> index 573a82d23c89..008b31422079 100644
>>>> --- a/Documentation/ABI/testing/debugfs-cros-ec
>>>> +++ b/Documentation/ABI/testing/debugfs-cros-ec
>>>> @@ -32,3 +32,25 @@ Description:
>>>>              is used for synchronizing the AP host time with the EC
>>>>              log. An error is returned if the command is not supported
>>>>              by the EC or there is a communication problem.
>>>> +
>>>> +What:               /sys/kernel/debug/cros_ec/last_resume_result
>>
>> Thinking about it, as other the other interfaces, I'd do
>>
>> s/cros_ec/<cros-ec-device>/
>>
>> I know that for now only cros_ec supports that, but we don't know what will
>> happen in the future, specially now that the number of cros devices is incrementing.
> 
> See my comment below, I suppose the fate of these two comments are
> tied together.
> 
>>
>>>> +Date:               June 2019
>>>> +KernelVersion:      5.3
>>>> +Description:
>>>> +            Some ECs have a feature where they will track transitions to
>>>> +            the (Intel) processor's SLP_S0 line, in order to detect cases
>>>> +            where a system failed to go into S0ix. When the system resumes,
>>>> +            an EC with this feature will return a summary of SLP_S0
>>>> +            transitions that occurred. The last_resume_result file returns
>>>> +            the most recent response from the AP's resume message to the EC.
>>>> +
>>>> +            The bottom 31 bits contain a count of the number of SLP_S0
>>>> +            transitions that occurred since the suspend message was
>>>> +            received. Bit 31 is set if the EC attempted to wake the
>>>> +            system due to a timeout when watching for SLP_S0 transitions.
>>>> +            Callers can use this to detect a wake from the EC due to
>>>> +            S0ix timeouts. The result will be zero if no suspend
>>>> +            transitions have been attempted, or the EC does not support
>>>> +            this feature.
>>>> +
>>>> +            Output will be in the format: "0x%08x\n".
>>>> diff --git a/drivers/mfd/cros_ec.c b/drivers/mfd/cros_ec.c
>>>> index 5d5c41ac3845..2a9ac5213893 100644
>>>> --- a/drivers/mfd/cros_ec.c
>>>> +++ b/drivers/mfd/cros_ec.c
>>>> @@ -102,12 +102,16 @@ static int cros_ec_sleep_event(struct cros_ec_device *ec_dev, u8 sleep_event)
>>>>
>>>>      /* For now, report failure to transition to S0ix with a warning. */
>>>>      if (ret >= 0 && ec_dev->host_sleep_v1 &&
>>>> -        (sleep_event == HOST_SLEEP_EVENT_S0IX_RESUME))
>>>> +        (sleep_event == HOST_SLEEP_EVENT_S0IX_RESUME)) {
>>>> +            ec_dev->last_resume_result =
>>>> +                    buf.u.resp1.resume_response.sleep_transitions;
>>>> +
>>>>              WARN_ONCE(buf.u.resp1.resume_response.sleep_transitions &
>>>>                        EC_HOST_RESUME_SLEEP_TIMEOUT,
>>>>                        "EC detected sleep transition timeout. Total slp_s0 transitions: %d",
>>>>                        buf.u.resp1.resume_response.sleep_transitions &
>>>>                        EC_HOST_RESUME_SLEEP_TRANSITIONS_MASK);
>>>> +    }
>>>>
>>>>      return ret;
>>>>  }
>>>> diff --git a/drivers/platform/chrome/cros_ec_debugfs.c b/drivers/platform/chrome/cros_ec_debugfs.c
>>>> index cd3fb9c22a44..663bebf699bf 100644
>>>> --- a/drivers/platform/chrome/cros_ec_debugfs.c
>>>> +++ b/drivers/platform/chrome/cros_ec_debugfs.c
>>>> @@ -447,6 +447,13 @@ static int cros_ec_debugfs_probe(struct platform_device *pd)
>>>>      debugfs_create_file("uptime", 0444, debug_info->dir, debug_info,
>>>>                          &cros_ec_uptime_fops);
>>>>
>>>> +    if (!strcmp(ec->class_dev.kobj.name, CROS_EC_DEV_NAME)) {
>>
>> For debugfs I don't care having the file exposed even is not supported, anyway
>> there are some CROS_EC_DEV_NAME that won't support it, so just make this simple
>> and create the file always.
> 
> Aw, really? This file is very specific to system suspend/resume. My
> original patch had it everywhere, but I was finding it very ugly that
> this was showing up on things like the fingerprint device. I can
> change it if you think it's better to have it everywhere, but it also
> seems like an easy change to make in the future if this file is for
> some reason needed on other EC types.
> 

I'd think different if it was a sysfs property but it's a debugfs.

Right now we have:

#define CROS_EC_DEV_NAME        "cros_ec"
#define CROS_EC_DEV_FP_NAME     "cros_fp"
#define CROS_EC_DEV_ISH_NAME    "cros_ish"
#define CROS_EC_DEV_PD_NAME     "cros_pd"
#define CROS_EC_DEV_SCP_NAME    "cros_scp"
#define CROS_EC_DEV_TP_NAME     "cros_tp"

Is really the named cros_ec the only that has this feature? What about cros_scp
(is not supposed to run the same base code as cros_ec)? And cros_ish ?

And all the named cros_ec devices have this feature? Maybe is supported by
Nocturne but not Veyron? Wouldn't be exposed also on those cros_ec that doesn't
support it?

And don't have the answer to all these questions, hence my concerns.

As per your documentation

 +                ...           The result will be zero if no suspend
 +            transitions have been attempted, or the EC does not support
 +            this feature.

I can accept if _all_ the CROS_EC_DEV_NAMEs and _only_ the CROS_EC_DEV_NAME
supports it, is that the case?

But if you are anyway exposing this on CROS_EC_DEV_NAMEs that doesn't support it
why not just expose to all and skip a future bunch of code to filter, it is
clear what a 0 means and at the end it's just a debugfs file.

~ Enric

> -Evan
>