Michael Buesch wrote: > On Monday 04 February 2008, Michael Buesch wrote: >> On Monday 04 February 2008 18:21:42 Matti Viljanen wrote: >>> Michael Buesch kirjoitti: >>>> On Sunday 03 February 2008 22:15:57 David Cohen wrote: >>>>>> b43-phy0 ERROR: Fatal DMA error: 0x00001000, 0x00000000, 0x00000000, >>>>>> 0x00000000, 0x00000000, 0x00000000 >>>> This is a "Descriptor Protocol Error" on the RX DMA engine. >>>> On which architecture are you running the card and how much RAM >>>> do you have? >>>> It seems that the descriptor ring is allocated in a memory region >>>> not accessible to the DMA engine (above 1G). >>>> Or the wrong DMA address is passed to the DMA engine, but I'd >>>> rule that out, as it works for everybody else. >>> I am running 2.6.24-git13 i386 kernel on Athlon64 with 2GiB of RAM >>> (Gentoo updates tend to kill HD's...) >> Try an x86_64 kernel. I bet it will work. >> > > Can you also try the following patch on an i386 kernel?
I didn't think we would ever need to revisit this code; however, our use of the PIO fall through that hid DMA problems was too clever. Your implementation of an explicit DMA testing should be an improvement on the kernel's routine. I agree with the intent of the patch. Assuming that it fixes the problem found by Matti Viljanen, you have my ACK. BTW, b43legacy will also need a variation of the patch as the earlier BCM4306 models could have the same problem in case they get used with more than 1 GB RAM. Larry _______________________________________________ Bcm43xx-dev mailing list [email protected] https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
