[Prism54-devel] [Bug 112] SMC2835W fails with "...your card may be faulty..." in latest cvs

[Jean-Baptiste Note]

Hello,

I don't know if this mail is okay and/or if i should go to bugzilla. Anyways.

>> If you want to jump over the error, edit islpci_dev.c       
>> and comment out lines 526 to 530 -> This :       
>>       
>>            if (result) {       
>>                  printk(KERN_ERR "%s: interface reset failure\n",     
>>                  priv->ndev->name);       
>>                  up_write(&priv->mib_sem);       
>>                  return result;       
>>            }      

Well, unless i'm mistaken, there's a *big race* in the "waiting for the
irq" place. We should use a completion variable to do the wait, with a
calibrated timeout. The race is that if the interrupt is coming in too
soon (and/or the machine is too slow), we'll miss it. With a completion
variable there is no such risk.

Does anyone agree on this, and can i do a quick patch correcting this ?

> So I tried it, and closed the bug. I have been using my card ever since to my    
> great satisfaction and, frankly, it is stable as a rock. It is NOT faulty.     
>
> So the reason I am reopening this bug is, that I believe that this check is    
> not right. Even though the developers think it is necessarry, in my case it is    
> not. The card works when the check is not there.    

Thanks for your patience, and taking the time to share your knowledge.

I know this is not in you're immediate interrest, but i'd really like to
understand what's happening here. Maybe we simply miss the
interrupt. Maybe also we're not waiting on the right interrupt bit (which we
wouldn't have enabled); this case i'd really like to catch if this is
the case.

JB

-- 
Jean-Baptiste Note
+33 (0)6 83 03 42 38
jean-baptiste.note at wanadoo.fr