Re: [PATCH] coresight: tmc: fix failure to disable/enable ETF after reading

[Jie Gan <jie.gan@oss.qualcomm.com>]

On 4/23/2025 9:56 PM, Mike Leach wrote:
> Hi,
> 
> On Wed, 23 Apr 2025 at 11:37, Jie Gan <jie.gan@oss.qualcomm.com> wrote:
>>
>> From: Mao Jinlong <quic_jinlmao@quicinc.com>
>>
>> From: Mao Jinlong <quic_jinlmao@quicinc.com>
>>
> 
> What are these extra email addresses for?

Sorry for the mistake. I did not aware the commit author is changed from 
me to the original author.

> 
>> ETF may fail to re-enable after reading, and driver->reading will
>> not be set to false, this will cause failure to enable/disable to ETF.
>> This change set driver->reading to false even if re-enabling fail.
>>
>> Fixes: 669c4614236a7 ("coresight: tmc: Don't enable TMC when it's not ready.")
> 
> This SHA and message appear not be present in any upstream / coresight branch.
> 

detail for the commit from linux-next git:

commit 669c4614236a7f78a2b693d0024cbdfa8536eb5a
Author: Yabin Cui <yabinc@google.com>
Date:   Fri Jan 27 23:10:01 2023 +0000

     coresight: tmc: Don't enable TMC when it's not ready.

     If TMC ETR is enabled without being ready, in later use we may
     see AXI bus errors caused by accessing invalid addresses.

     Signed-off-by: Yabin Cui <yabinc@google.com>
     [ Tweak error message ]
     Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
     Link: 
https://lore.kernel.org/r/20230127231001.1920947-1-yabinc@google.com


>> Co-developed-by: Yuanfang Zhang <quic_yuanfang@quicinc.com>
>> Signed-off-by: Yuanfang Zhang <quic_yuanfang@quicinc.com>
>> Signed-off-by: Mao Jinlong <quic_jinlmao@quicinc.com>
>> Signed-off-by: Jie Gan <jie.gan@oss.qualcomm.com>
>> ---
>>   drivers/hwtracing/coresight/coresight-tmc-etf.c | 13 +++++++------
>>   1 file changed, 7 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/hwtracing/coresight/coresight-tmc-etf.c b/drivers/hwtracing/coresight/coresight-tmc-etf.c
>> index d858740001c2..8c9f14e36bc2 100644
>> --- a/drivers/hwtracing/coresight/coresight-tmc-etf.c
>> +++ b/drivers/hwtracing/coresight/coresight-tmc-etf.c
>> @@ -87,6 +87,12 @@ static void __tmc_etb_disable_hw(struct tmc_drvdata *drvdata)
>>   {
>>          CS_UNLOCK(drvdata->base);
>>
>> +       /* Check if the etf already disabled*/
>> +       if (!(readl_relaxed(drvdata->base + TMC_CTL) & TMC_CTL_CAPT_EN)) {
>> +               CS_LOCK(drvdata->base);
>> +               return;
>> +       }
>> +
> 
> What does this have to do with the stated function of the patch - this
> is unnecessary.

I agree with you. I think it's ok to disable a disabled device again.

> Under what scenario can this function be called with the ETB
> previously disabled?

Combined with the fix code below. It may appear a scenario that the ETB 
is disabled but the mode is CS_MODE_SYSFS.

> 
>>          tmc_flush_and_stop(drvdata);
>>          /*
>>           * When operating in sysFS mode the content of the buffer needs to be
>> @@ -747,7 +753,6 @@ int tmc_read_unprepare_etb(struct tmc_drvdata *drvdata)
>>          char *buf = NULL;
>>          enum tmc_mode mode;
>>          unsigned long flags;
>> -       int rc = 0;
>>
>>          /* config types are set a boot time and never change */
>>          if (WARN_ON_ONCE(drvdata->config_type != TMC_CONFIG_TYPE_ETB &&
>> @@ -773,11 +778,7 @@ int tmc_read_unprepare_etb(struct tmc_drvdata *drvdata)
>>                   * can't be NULL.
>>                   */
>>                  memset(drvdata->buf, 0, drvdata->size);
>> -               rc = __tmc_etb_enable_hw(drvdata);
>> -               if (rc) {
>> -                       raw_spin_unlock_irqrestore(&drvdata->spinlock, flags);
>> -                       return rc;
>> -               }
>> +               __tmc_etb_enable_hw(drvdata);
> 
> Dropping a valid error check is not acceptable. If a TMC cannot be
> re-enabled, then that is a hardware error that needs noting and
> dealing with.
> 

That's the issue about the error check. __tmc_etb_enable_hw may fail due 
to hardware issue or just a timeout for tmcready(may restore). The error 
check captures the error and returns directly without setting 
drvdata->reading = false.

So the current mode for ETB is CS_MODE_SYSFS, which implies the device 
is enabled but actually not. This prevents us from processing another 
enable/disable session for the ETF if drvdata->reading remains true.

Thanks,
Jie

> Regards
> 
> Mike
> 
>>          } else {
>>                  /*
>>                   * The ETB/ETF is not tracing and the buffer was just read.
>> --
>> 2.34.1
>>
> 
>