This patch series corrects our inconsist use of DMA. I'm pretty certain about these patches now, I've tested them myself and got one lock up but it was only after doing some really crazy things to my machine I rather not share in public (and unable to get a trace). The beacon patches should be correct as previous to this we were pci_map_single()'ing more data than we told hardware it can get from us on TX. Regular TX is fine as we always tell it to use the skb->len and also account for the 4 byte padding it always adds. Additionally we now have one user which reports it fixed the issue for him, but this time on a MacBook pro v2 (what is it with those things?!). Unfortunately I cannot find a solution to having the kernel give us 8KB instead of 4KB which is all we need for our buffers. The reason for askign at least for ~3849 bytes (before the encryption overhead) is we _can_ expect a max AMSDU of size 3849. If someone finds a clever solution to this please let me know. Many thanks to Benny and Maciej for helping to review this issue. Luis -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
