Bob Heise wrote: > > I added pci_dma_supported(ring->bcm->pci_dev, 0x3FFFFFFF) and > pci_set_dma_mask(ring->bcm-pci_dev, 0x3FFFFFFF) after your 64 and 32 bit DMA > attempts in the bcm43xx_dma.c's map_descbuffer from your patch, both > indicated that 30-bit DMA was supported, yet when after that I did: > dmaaddr = pci_map_single(ring->bcm->pci_dev, buf, len, direction) > pci_dma_mapping_error returned true for that address. Very frustrating > indeed. > > Since that experiment I have upgraded my wireless-dev tree to the latest. I > thought, naïvely I suppose, that maybe forcing on the iommu would allow the > correct address mapping since obviously it isn't occurring normally. The > resultant kernel turned up what seems to be a locking problem. I'm not sure > if it's a real problem for normal operation, but it did crash my box so... > hopefully the pursuent /var/log/messages snippet is helpful. > > Of course if anyone has any ideas on how to solve this 30-bit DMA problem I > would be happy to try it out.
My memory should arrive early next week, which will give me 1.5 GB, and I think I can force my system to generate dma addresses in the first 1G. Then I should be able to generate the necessary fixes. I'll let you know when I have a patch to test. Mine will initially be for wireless-2.6, but it should be simple to add to wireless-dev. Larry _______________________________________________ Bcm43xx-dev mailing list [email protected] https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
