[RFC/RFT] ssb: Attempt to recover from SPROM CRC error

[Larry Finger]

On 12/02/2010 04:15 PM, Michael Büsch wrote:
> On Thu, 2010-12-02 at 10:36 -0600, Larry Finger wrote: 
>> Current code defaults to SPROM revision 1 if there is a CRC error. In at
>> least one known case, most of the corrupt contents are reasonable and
>> it is possible to extract the correct MAC address and TX power settings
>> from what is read. With this patch, an attempt is made to match the
>> apparent revision number with certain SPROM signatures. For those revisions
>> without such a feature, a reasonable guess is made. If the apparent
>> revision is invalid, or if the signature does not match, the previous
>> behavior is kept.
>>
>> Signed-off-by: Larry Finger <Larry.Finger at lwfinger.net>
>> ---
>>
>> John,
>>
>> Could you please test this patch with your card?
>>
>> Thanks,
>>
>> Larry
>> ---
>>
>> Index: linux-2.6/drivers/ssb/pci.c
>> ===================================================================
>> --- linux-2.6.orig/drivers/ssb/pci.c
>> +++ linux-2.6/drivers/ssb/pci.c
>> @@ -620,6 +620,7 @@ static int ssb_pci_sprom_get(struct ssb_
>>  	const struct ssb_sprom *fallback;
>>  	int err = -ENOMEM;
>>  	u16 *buf;
>> +	u16 revision;
>>  
>>  	if (!ssb_is_sprom_available(bus)) {
>>  		ssb_printk(KERN_ERR PFX "No SPROM available!\n");
>> @@ -671,6 +672,50 @@ static int ssb_pci_sprom_get(struct ssb_
>>  			}
>>  			ssb_printk(KERN_WARNING PFX "WARNING: Invalid"
>>  				   " SPROM CRC (corrupt SPROM)\n");
>> +			/* At this point, we have a faulty SPROM image.
>> +			 * In case only part of it is corrupt, try to
>> +			 * determine what rev we might have */
>> +			revision = buf[SSB_SPROMSIZE_WORDS_R4 - 1] & 0x00FF;
> 
> I think this could possibly overrun the buffer, or did I get something
> wrong?

At this point, we have gotten a CRC error with the small SPROM and expanded the
memory for the 440 byte size, and gotten a second CRC error. This test will be
OK here.

Larry