Date: Wed, 18 Jun 2003 19:37:26 +0200 From: Bernd Walter <ticso@cicely12.cicely.de> To: Andrew Gallatin <gallatin@cs.duke.edu> Cc: freebsd-alpha@freebsd.org Subject: Re: 5.1beta2 on Alpha ES40 w/ 32GB pys ram Message-ID: <20030618173725.GU3626@cicely12.cicely.de> In-Reply-To: <16112.39014.751410.929689@grasshopper.cs.duke.edu> References: <16094.16676.273466.121560@grasshopper.cs.duke.edu> <20030613190624.W36168-100000@mail.chesapeake.net> <16110.27799.68056.48679@grasshopper.cs.duke.edu> <20030618005958.GD3626@cicely12.cicely.de> <16111.54084.284151.643826@grasshopper.cs.duke.edu> <20030618123315.GH3626@cicely12.cicely.de> <16112.39014.751410.929689@grasshopper.cs.duke.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Jun 18, 2003 at 12:50:46PM -0400, Andrew Gallatin wrote: > > Bernd Walter writes: > > > > > You can see that the second hose has pci0 as well. > > That should be fine.. they are in separate PCI address spaces. > > > Much worse a bridged card could get a bus number that is also used on > > another hose and it fails obscurly within FreeBSD. > > Having the same bus number appear on multiple bridges should also be > OK. Again, the hose is its own address space. > > Are you worried about rman? I'm not worried - I tryed and failed Its is OK to overlap, but then we need to fix the code to handle this. First point is that programms like pciconf doesn't work, because you tell the bus number, but not the hose. The second one is that the bus behind the bridge on hose 1 had the same number than a bus on hose 0 and probing that one on hose 1 showed only single device from the bus on hose 0. I have some speculation about what happens here, but I still have to check. > > Currently I'm forced to keep bridged cards in the same hose. > > > > This machine also doesn't initialize io space on bridges and some > > drivers have problems with memory space. > > That's bad. E.g. I had no succes to get these controllers working: AHC-3985 (bridged 3 channel AIC-7870) AHC-2940UW AHC-2944W You can tell the driver by a compile flag to try memio. The memio tests fail randomly from time time time as well as usage fails. My current kernel is compiled without ahc because it may hang in probing the scbus behind some of the channels. That's especially bad as there are waiting some HVD disks for the 2944 to run. > > And if we do it ourself than it might be reasonable to think about > > hot-plug PCI some day. > > > > I havn't followed that idea very much because I worried about what SRM > > would do if we tamper with his configuration. > > Your Linux example sounds promissing. > > On miatas, linux shuffles pci memory mappings around so that it can > use a 2GB direct map. This is great, except that linux does not halt > cleanly back to the SRM. At shutdown time, it attempts to restore the > same settings it found, but when halting, all you get is a repeating > register dump. FreeBSD, which doesn't do anything this fancy (and > which is limited to a 1GB direct map, due to the @#@#% stupid place > the SRM puts the PCI memory spaces), halts fine. Perhaps we can make > it work right, but there are drawbacks. OK - my personal priority is low as they should have enough windows and we implement the SG map. There only a performance argument left. If it would just have been reinitializing the PCI bus then thing would have been different. -- B.Walter BWCT http://www.bwct.de ticso@bwct.de info@bwct.de
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030618173725.GU3626>