firmware: stratix10-svc: Fix probe failure with old ATF

[PATCH 0/2] firmware: stratix10-svc: Fix probe failure with old ATF

Posted by Muhammad Amirul Asyraf Mohamad Jamian 2 months ago

Since commit bcb9f4f07061 ("firmware: stratix10-svc: Add support for
async communication"), the SVC driver fails to probe entirely when
running with ATF versions older than 3.0 (e.g. ATF 2.5) that do not
support SIP SVC v3 asynchronous operations.

stratix10_svc_async_init() returns -EINVAL for old ATF, and the probe
function treats any non-zero return as fatal, causing:

  stratix10-svc firmware:svc: probe with driver stratix10-svc failed \
    with error -22

This prevents all dependent client drivers (hwmon, RSU, FCS) from
probing even though they can operate correctly via the synchronous V1
SMC path.

This series fixes the issue in two steps:
  1. Return -EOPNOTSUPP (instead of -EINVAL) when ATF async is
     unsupported, so callers can distinguish "not supported" from
     "bad argument / programming error".
  2. Treat -EOPNOTSUPP as non-fatal in probe, allowing the SVC driver
     to load in sync-only mode so all client drivers can probe normally.

Both patches fix bcb9f4f07061 and are tagged for stable.

Muhammad Amirul Asyraf Mohamad Jamian (2):
  firmware: stratix10-svc: Return -EOPNOTSUPP when ATF async unsupported
  firmware: stratix10-svc: Don't fail probe when async ops unsupported

 drivers/firmware/stratix10-svc.c | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

-- 
2.43.7

Re: [PATCH 0/2] firmware: stratix10-svc: Fix probe failure with old ATF

Posted by Dinh Nguyen 1 month, 1 week ago


On 4/16/26 02:22, Muhammad Amirul Asyraf Mohamad Jamian wrote:
> Since commit bcb9f4f07061 ("firmware: stratix10-svc: Add support for
> async communication"), the SVC driver fails to probe entirely when
> running with ATF versions older than 3.0 (e.g. ATF 2.5) that do not
> support SIP SVC v3 asynchronous operations.
> 
> stratix10_svc_async_init() returns -EINVAL for old ATF, and the probe
> function treats any non-zero return as fatal, causing:
> 
>    stratix10-svc firmware:svc: probe with driver stratix10-svc failed \
>      with error -22
> 
> This prevents all dependent client drivers (hwmon, RSU, FCS) from
> probing even though they can operate correctly via the synchronous V1
> SMC path.
> 
> This series fixes the issue in two steps:
>    1. Return -EOPNOTSUPP (instead of -EINVAL) when ATF async is
>       unsupported, so callers can distinguish "not supported" from
>       "bad argument / programming error".
>    2. Treat -EOPNOTSUPP as non-fatal in probe, allowing the SVC driver
>       to load in sync-only mode so all client drivers can probe normally.
> 
> Both patches fix bcb9f4f07061 and are tagged for stable.
> 
>
I think it makes more sense to squash these 2 patches together. Patch 1 
adds the -EOPNOTSUPP, but does not make use of it. Patch 2 actually 
makes use of the -EOPNOTSUPP. So I was a bit confused on how the change 
is getting used.

Dinh

Re: [PATCH 0/2] firmware: stratix10-svc: Fix probe failure with old ATF

Posted by MOHAMAD JAMIAN, MUHAMMAD AMIRUL ASYRAF 1 month, 1 week ago

On 5/5/2026 8:14 pm, Dinh Nguyen wrote:
> 
> 
> On 4/16/26 02:22, Muhammad Amirul Asyraf Mohamad Jamian wrote:
>> Since commit bcb9f4f07061 ("firmware: stratix10-svc: Add support for
>> async communication"), the SVC driver fails to probe entirely when
>> running with ATF versions older than 3.0 (e.g. ATF 2.5) that do not
>> support SIP SVC v3 asynchronous operations.
>>
>> stratix10_svc_async_init() returns -EINVAL for old ATF, and the probe
>> function treats any non-zero return as fatal, causing:
>>
>>    stratix10-svc firmware:svc: probe with driver stratix10-svc failed \
>>      with error -22
>>
>> This prevents all dependent client drivers (hwmon, RSU, FCS) from
>> probing even though they can operate correctly via the synchronous V1
>> SMC path.
>>
>> This series fixes the issue in two steps:
>>    1. Return -EOPNOTSUPP (instead of -EINVAL) when ATF async is
>>       unsupported, so callers can distinguish "not supported" from
>>       "bad argument / programming error".
>>    2. Treat -EOPNOTSUPP as non-fatal in probe, allowing the SVC driver
>>       to load in sync-only mode so all client drivers can probe normally.
>>
>> Both patches fix bcb9f4f07061 and are tagged for stable.
>>
>>
> I think it makes more sense to squash these 2 patches together. Patch 1 
> adds the -EOPNOTSUPP, but does not make use of it. Patch 2 actually 
> makes use of the -EOPNOTSUPP. So I was a bit confused on how the change 
> is getting used.
> 
> Dinh
Addressed in v2

https://lore.kernel.org/all/20260508092936.18380-1-muhammad.amirul.asyraf.mohamad.jamian@altera.com/

Thanks,
Amirul