Re: [Qemu-devel] FreeBSD status

[Juergen Lock]

On Wed, May 26, 2004 at 03:56:18PM +0000, Fabrice Bellard wrote:
> Juergen Lock wrote:
> > As mentioned on irc I was having some success with FreeBSD:
> > 
> >  4.10RC3 install cd: boots, but kernel doesnt see any disks
> > (error message when trying to partition, -hda with empty image)
> > 

4.10-RELEASE boot.flp can be found using
        http://mirrorlist.freebsd.org/
 btw. (yes 4.10-RELEASE is out, i'm currently ftp'ing it.  4.10 is
off the -stable branch so its also important to get working.)

> >  5.2.1 install cd and boot.flp: boot now with the fixes in cvs,
> > but panic when trying to write to disk (same empty image).
> > to repeat: at the mountroot prompt type:
> >     ufs:md0
> > then in sysinstall select custom install, partition, `c',
> > enter slice (partition) size, hit `w' to write immediately,
> > confirm, select mbr -> panic writing to disk
> 
> I found the problem : it seems that FreeBSD expects the IDE READY status 
> bit just after the IDE WRITE command. It is weird as it is not 
> mentionned in the IDE specs. I would like to be sure it is a standard 
> IDE behaviour.
> 
 Hmm.  I since built a debug kernel and it looks like the panic is in
sys/dev/ata/ata-lowlevel.c line 276:
        if (request->device->channel->dma->flags & ATA_DMA_ACTIVE)
request->device->channel->dma is NULL.  The 5.2.1 ata-lowlevel.c can
be viewed here:
        
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/dev/ata/ata-lowlevel.c?annotate=1.24.2.3

 (to see other 5.2.1 sources look for the RELENG_5_2_1_RELEASE tag.)

 This seems to have something to do with the fact that qemu doesn't
support ata (ide) dma (right?), but it is a bit strange as i do have
seen 5.2.1 running on an old 486 about which i doubt that it has
ide dma hardware, and anyway i'm pretty sure such a bug would have
been known by now...

 Btw can you post or send me your patch so i can test it?

> >  I also get: npx0: WARNING: no FPU! from the kernel (kernel
> > messages can be viewed with scroll-lock and then page-up/down,
> > cursur-up/down), Fabrice says that the FPU error comes from the fact
> > that FreeBSD uses FPU interrupts to test the FPU exception reporting.
> 
> I made a fix (not commited yet).
> 
 Eagerly awaiting that one also... :)

        Juergen