W dniu 14 sierpnia 2011 11:24 użytkownik Rafał Miłecki <zaj...@gmail.com> napisał: > W dniu 14 sierpnia 2011 11:07 użytkownik Rafał Miłecki > <zaj...@gmail.com> napisał: >> I guess we should just increase RX ring size instead hacking >> (stripping) 0x1000 bit. We just need to check on which hardware wl >> uses 256 ring size. > > Ignore that. I just got 8 KiB aligned ring when testing 256 ring size. > > I agree with David, that 0x1000 comes from ring address. Depending on > address alignment it's 0 or 1.
The question I now have is: 1) Should be keep using 4 KiB aligned rings (sometimes, depending on the luck, 8 KiB) with ignoring 0x1000 bit 2) Switch to 8 KiB aligned rings Broadcom seems to use second method, it never uses just 4 KiB aligned ring. From dmb's log: write32 0xb0600228 <- 0x00098000 write32 0xb060022c <- 0x80000000 0x0000000000098000 From dwmw2's log: write32 0xb0600228 <- 0x1f88a000 write32 0xb060022c <- 0x80000000 0x000000001f88a000 From kitelau's log: write32 0xa0600228 <- 0x00098000 write32 0xa060022c <- 0x80000000 0x0000000000098000 All are 8 KiB aligned. David also dropped following comment from brcm80211: > Each descriptor ring must be 8kB aligned, and fit within a contiguous 8kB > physical address. -- Rafał _______________________________________________ b43-dev mailing list b43-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/b43-dev
