On 8/16/2014 2:16 PM, Kim Phillips wrote:
> On Thu, 14 Aug 2014 15:54:22 +0300
> Horia Geanta <horia.geanta@xxxxxxxxxxxxx> wrote:
>
>> This patch set adds Run Time Assembler (RTA) SEC descriptor library.
>> RTA is a replacement for incumbent "inline append".
>>
>> The library is intended to be a single code base for SEC descriptors creation
>> for all Freescale products. This comes with a series of advantages, such as
>> library being maintained / kept up-to-date with latest platforms, i.e. SEC
>> functionalities (for e.g. SEC incarnations present in Layerscape LS1 and LS2).
>>
>> RTA detects options in SEC descriptors that are not supported
>> by a SEC HW revision ("Era") and reports this back.
>> Say a descriptor uses Sequence Out Pointer (SOP) option for the SEQINPTR
>> command, which is supported starting from SEC Era 5. If the descriptor would
>> be built on a P4080R3 platform (which has SEC Era 4), RTA would report
>> "SEQ IN PTR: Flag(s) not supported by SEC Era 4".
>> This is extremely useful and saves a lot of time wasted on debugging.
>> SEC HW detects only *some* of these problems, leaving user wonder what causes
>> a "DECO Watchdog Timeout". And when it prints something more useful, sometimes
>> it does not point to the exact opcode.
>
> again, RTA just adds bloat to the kernel driver - the kernel driver
> is supposed to generate the appropriate descriptor for its target
> running SEC version no matter what, not "report back" what is/is not
> supported. This is a flaw at the RTA design level, as far as the
> kernel driver is concerned.
What is your understanding of developing a descriptor?
First it needs to be written, then tested - within the kernel driver.
Having no error checking in the code that generates descriptors
increases testing / debugging time significantly. Again, SEC HW provides
some error reporting, but in many cases this is a clueless Watchdog Timeout.
SEC descriptors development is complex enough to deserve a few
indications along the way.
Regards,
Horia
--
To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
