RE: unaligned accesses
>> It's a random D-Link card using via-rhine.
>This is my prime suspect.
>> I have - the messages are from various places in the ipv4 network
>> code, tcp_ack, udp_recvmsg, tcp_recvmsg are the more recent places.
>Well, that's what happens if the packet is not "aligned", and the
>components of the "headers" are accessed off-alignment. The driver may
>be fixable, if it's one of Don Becker's bunch, in that you can force
>the message to be copied out of the receive buffer into an aligned
>data area; please look at one of his other drivers (the original TULIP
>was one, perhaps it still has the remnants), looking for code that is
>conditionally compiled (or not) when ALPHA is the architecture.
Yep - seems to be the case. The via-rhine.c didn't have an architecture
condition to force copy the entire packet to aligned memory.
Added that code, built and reloaded the module and so far, no more alignment
To UNSUBSCRIBE, email to email@example.com
with a subject of "unsubscribe". Trouble? Contact firstname.lastname@example.org