Re: [PATCH v7 09/12] crypto: qce - remove the BROKEN label

[Demi Marie Obenour <demiobenour@gmail.com>]

[-- Attachment #1.1: Type: text/plain, Size: 1656 bytes --]

On 9/11/26 07:14, Bartosz Golaszewski wrote:
> On Thu, 10 Sep 2026 16:31:40 +0200, Eric Biggers <ebiggers@kernel.org> said:
>> On Thu, Sep 10, 2026 at 03:00:50PM +0200, Bartosz Golaszewski wrote:
>>> All self-tests now pass.
>>
>> Have the other bugs that were reported been fixed?  For example:
>> https://lore.kernel.org/linux-crypto/20260723202704.GB110634@quark/
>> https://lore.kernel.org/linux-crypto/20260723183438.GB93534@quark/
>> https://lore.kernel.org/linux-crypto/20260723205342.GC110634@quark/
>>
> 
> I will get to these too. For now, the series allows the self-tests to pass,
> I assume it it's enough to not keep the BROKEN label.

Personally, I think the driver should stay BROKEN until all the
known bugs have been fixed.  The self-tests aren't enough for HW
crypto drivers.

>>>  	help
>>>  	  This driver supports Qualcomm crypto engine accelerator
>>
>> As this patch would make the kconfig option visible again and would
>> result in everyone running "make oldconfig" being asked about it, it
> 
> I don't mind changing the symbol if you prefer it to avoid any accidental
> `make oldconfig` carry-overs.
> 
>> needs to be fixed to accurately document why anyone would want to enable
>> this option.  As you've stated, it's actually just for "testing".  The
>> false claim that it's an "accelerator" also needs to go.
>>
> 
> "Offloader" for lack of a better word?
Could it be called a "decelerator"?  That would give users an idea
of what to expect from the driver.  Maybe explicitly mention that
you will take at least a 50x performance hit?
-- 
Sincerely,
Demi Marie Obenour (she/her/hers)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]