Bug 29299
| Summary: | FreeBSD 4.3 Alpha + Tekram SCSI adapter panics | ||
|---|---|---|---|
| Product: | Base System | Reporter: | thomas.pornin <Thomas.Pornin> |
| Component: | alpha | Assignee: | freebsd-alpha (Nobody) <alpha> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | ||
| Priority: | Normal | ||
| Version: | 4.3-RELEASE | ||
| Hardware: | Any | ||
| OS: | Any | ||
State Changed From-To: open->feedback Is this still a problem for you with 5.1-CURRENT? FYI. ---------- Forwarded message ---------- Date: Fri, 22 Aug 2003 19:55:29 +0200 From: Thomas Pornin <pornin@bolet.org> To: Nate Lawson <njl@freebsd.org> Subject: Re: alpha/29299: FreeBSD 4.3 Alpha + Tekram SCSI adapter panics On Thu, Aug 21, 2003 at 08:20:01PM -0700, Nate Lawson wrote: > Synopsis: FreeBSD 4.3 Alpha + Tekram SCSI adapter panics > > State-Changed-From-To: open->feedback > State-Changed-By: njl > State-Changed-When: Thu Aug 21 20:19:41 PDT 2003 > State-Changed-Why: > Is this still a problem for you with 5.1-CURRENT? > > http://www.freebsd.org/cgi/query-pr.cgi?pr=29299 I am sorry, I cannot test anymore. The machine was may office machine when I was doing my PhD thesis. I now work elsewhere and I unfortunately could not take with me that PWS 500a (even though I would have done it gladly). I have another Alpha machine but it is small (AXPpci33, 32MB ram) and is my primary DNS / mail server, so I cannot afford to convert it to 5.x while it is not perfectly stable, nor to take three days to compile a whole new 5.1 system and possibly witness it crash. --Thomas Pornin FreeBSD 5.4 is quite stable on alpha, a test would be appreciated. Otherwise please sent me the scsi card, I would be willing to test it. I can even provide a sym895 in return ( bootable on modern alpha's ). If that is no option either I don't see much value in keeping this pr open, the hardware and software revisions are pretty old/ hard to find these days. If it was symbios/isp/ata it make sense to bother. -- Sten Spans "There is a crack in everything, that's how the light gets in." Leonard Cohen - Anthem State Changed From-To: feedback->closed Submitter no longer has access to original machine; a latter correspondant notes that he believes that this combination should now be stable. |
The GENERIC kernel does not include support for the Tekram SCSI adapter. I recompiled a kernel with the 'device amd' line. At the end of the SCSI delay (set to 8 seconds in my setup), the kernel enters a fatal trap related to memory management. I tried with a debug kernel, the trap is triggered by the line 455 in alpha/alpha/busdma_machdep.c, function bus_dmamap_load(): if (map->pagesneeded != 0) { apparently, at that point, "map" is equal to NULL, and the memory acces triggers the trap. I tried to replace that line with the following: if (map && map->pagesneeded != 0) { (and something similar for another access to map->pagesneeded line 491), but then the machine crashes later with an "unexpected machine check" (mces = 0x1, vector = 0x660, param = 0xfffffc0000006068). I believe it is not standard behaviour that bus_dmamap_load() is called with a NULL map. The very same kernel, without the 'amd' support, boots and runs fine. The Tekram card and Zip drive are functionnal (they work on the same machine under Linux). Fix: None included. I guess the amd driver requires some tweekings. I have some spare time for the next three weeks, and am willing to help; but I do not know where to begin. How-To-Repeat: Just try to boot an Alpha machine with a Tekram adapter and support compiled in the kernel. Maybe the fact that there is no disk in the Zip drive is important.