Date: Fri, 18 Apr 2003 14:58:38 -0400 From: Jake Burkholder <jake@locore.ca> To: "Andrew R. Reiter" <arr@watson.org> Cc: freebsd-sparc@freebsd.org Subject: Re: devices on my sparc64 machine Message-ID: <20030418185838.GB33257@locore.ca> In-Reply-To: <20030417161244.B90240@fledge.watson.org> References: <20030417161244.B90240@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Apparently, On Thu, Apr 17, 2003 at 04:17:19PM -0400, Andrew R. Reiter said words to the effect of; > > I think there was a previous email regarding this, but I was not able to > easily find it. > > Are there any basic things that I should look out for when trying various > PCI devices on my sparc64 machine? It seems a number of devices have > already been tested, but I am more curious about things such as ATM > devices ... I have hfa and en cards and would like to try them on the u60. As far as I know Harti Brandt uses ngATM on sparc64, but I think that's with separate drivers. I looked briefly at the atm drivers for PAE safeness and ran away screaming, many of the issues are the same. > > The issues Im thinking of looking out for are obvious things .. 32 v. 64 > bit issues etc. The compiler will catch many 32/64 bit issues (casting pointer to integer of different size). The biggest things to look for are endian problems and proper use of busdma. The bus space accessors will do byte swapping in hardware for pci devices, but if the device directly accesses data in host memory it needs to have the right endianness and may need to be byte swapped by hand. Some devices have big and little endian modes. Anything that does dma and uses vtophys will not work, it needs to use busdma. Also, busdma consumers must not assume that bus addresses are actually physical addresses, because they're actually dvma addresses on sparc64. Jake
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030418185838.GB33257>