[Trac #839] MMC has timeout problems

[Zarro Boogs per Child]

#839: MMC has timeout problems
--------------------------+-------------------------------------------------
 Reporter:  PierreOssman  |       Owner:  mlj      
     Type:  defect        |      Status:  new      
 Priority:  normal        |   Milestone:  Untriaged
Component:  hardware      |    Keywords:           
--------------------------+-------------------------------------------------
 MMC cards do not work well at all in the XO. In the FPGA model, it worked
 somewhat reliable with MMCv4 cards, but v3:s weren't event detected. With
 the ASIC, all cards are detected but they are so unreliable that it is
 completely unusable.

 Report from the last set of tests I did:

 First of, the write to the GPIO registers that recently got removed are
 for some reason still needed for MMC. Without them, all commands simply
 time out.

 With those writes back in place, things get a bit further but results
 are varied. None of my cards work fully, the just have different odds of
 failing. The problem is also present on the FPGA, just very hard to
 trigger.

 The symptoms are that a few transfers succeed, but then we start getting
 data timeouts. The card still responds to commands, so it doesn't seem
 to have shut down.

 The cards that exhibit this problem are both MMCv4. I also have a MMCv3
 card for testing, but it works even worse. With the FPGA it isn't even
 detected (I assume DAT3 is used for card detection there) and with the
 ASIC I get command timeouts no matter what. I guess someone needs to
 scope a card like this to verify that all signals are correct.

 I'm a bit out of ideas on what could be wrong here. The only way forward
 seems to be to verify with a scope and/or logic analyzer that the
 controller is in fact behaving as it should.

-- 
Ticket URL: <http://dev.laptop.org/ticket/839>
One Laptop Per Child <http://laptop.org/>