svn commit: r223648 - head/sys/dev/gem

[Nathan Whitehorn]

On 07/07/11 11:36, YongHyeon PYUN wrote:
> On Thu, Jul 07, 2011 at 11:26:08AM -0500, Nathan Whitehorn wrote:
>> On 07/07/11 11:07, Marius Strobl wrote:
>>> On Thu, Jul 07, 2011 at 09:34:39AM -0500, Nathan Whitehorn wrote:
>>>> This breaks one of my gem devices (chip=0x0021106b) when the controller
>>>> is initialized from FreeBSD (netbooting works fine, and the controller
>>>> stays working after that), with "cannot disable RX MAC" or "cannot
>>>> disable RX MAC or hash filter" messages.
>>> Are you positive that it's exactly this revision which breaks things?
>>> Actually there was a report from Justin Hibbits that r222135 causing
>>> a GMAC to no longer work with similar symptoms, which still is unsolved
>>> AFAICT. If it's really r223648 which is causing problems in your case
>>> please try to figure out which part is causing that as none of the
>>> changes exactly stands out as disruptive, probably best by reverting
>>> gem_reset_rx() and gem_setladrf() one by one.
>>>
>> You're right. r223648 just added the printf(), r222135 is where it began
>> not working.
> Yes, I know that. I still have no clue and can't explain how
> r222135 can break gem(4). Also note, the "cannot disable RX MAC"
> was there before r222135.
> One vague guess I have at this moment is that firmware may have put
> controller into some deep power save state so the first time
> gem_init() is called it may not have fully initialized the
> controller. I sent a patch to Justin to check that and waiting for
> his reply.
>

That patch doesn't help my case.
-Nathan