Re: [PATCH v7 3/4] drm/xe/hwmon: Update xe_hwmon_process_reg

["Poosa, Karthik" <karthik.poosa@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1316 bytes --]


On 10-04-2024 00:22, Rodrigo Vivi wrote:
> On Tue, Apr 09, 2024 at 12:52:34PM -0500, Lucas De Marchi wrote:
>> On Fri, Apr 05, 2024 at 06:31:26PM +0530, Karthik Poosa wrote:
>>> Return u64 from xe_hwmon_process_reg, instead of a void return.
>>> u64* input pointer not needed with this change.
>>>
>>> With this caller can directly assign return value to a variable without
>>> need of explicit initialization and pass by reference.
>>>
>>> v2:
>>> - Fix checkpatch warnings.
>>>
>>> v3:
>>> - Rebase
>>> - Removed unncessary break statements.
>>>
>>> Signed-off-by: Karthik Poosa<karthik.poosa@intel.com>
>>> Suggested-by: Lucas De Marchi<lucas.demarchi@intel.com>
>>> Cc: Badal Nilawar<badal.nilawar@intel.com>
>> Applied the other patches.  This one I'm putting on hold to think about.
>>
>> I'm not sure the approach in that hwmon in general is good with the
>> xe_hwmon_get_reg() + xe_hwmon_process_reg(). It seems it's even taking
>> some pm refs when it doesn't need (to decide if attribute is visible).
> I believe this approach is fine.
> We do need to earlier get pm refs if we believe that there will be mmio
> operations underneath. Better more then less in this case.

Rodrigo,

We are already taking pm refs in parent calls of xe_hwmon_get/process_reg.

Anything else needed here ?

>> Lucas De Marchi

[-- Attachment #2: Type: text/html, Size: 2499 bytes --]