Re: [PATCH v8 06/11] drm/xe/guc: Use a two stage dump for GuC logs and add more info

[John Harrison <john.c.harrison@intel.com>]

On 10/2/2024 13:53, Julia Filipchuk wrote:
> On 9/19/2024 8:21 PM, John.C.Harrison@Intel.com wrote:
>> +#define GUC_PMTIMESTAMP				XE_REG(0xc3e8)
> This register seems to be the lower-32 bits of a pair. Does it make
> sense to read the full 64-bit counter for the log?
Hmm. Probably. Eventually.

The problem is that the GuC log only has 32bit timestamps. The extra 
four bits might be available but they are not used (in most 
circumstances) within the GuC firmware. And the whole point of printing 
this out is to allow correlation between the GuC log entries and dmesg 
entries. So printing a larger value than is used in the log is not 
currently useful.

And note that it is still not a 64bit counter. There might be 64bits of 
register space but the counter is only 36 bits total. So it still wraps 
in under an hour.

John.

>
>
>> +	err = xe_force_wake_get(gt_to_fw(gt), XE_FW_GT);
>> +	if (err) {
>> +		snapshot->stamp = ~0;
>> +	} else {
>> +		snapshot->stamp = xe_mmio_read32(&gt->mmio, GUC_PMTIMESTAMP);
>> +		xe_force_wake_put(gt_to_fw(gt), XE_FW_GT);
>> +	}