On Mon, 31 Mar 2025 13:48:37 Tetsuya Isaki wrote: > At Sun, 30 Mar 2025 11:29:38 +1100, > > Nat Sloss wrote: > > I've done some more investigating. It seems it is only unstable > > when using an > > audio kernel. I believe there is a possibility some of the pages > > at the top > > on memory are being used by the ASC/EASC. > > > > The crashes are easily reproducible as simple as listing a directory > > while > > triggering a console bell. > > At Sun, 30 Mar 2025 13:39:47 +1100, > > Nat Sloss wrote: > > Is something like this OK? > > > > --- a/sys/arch/mac68k/mac68k/pmap_bootstrap.c Sat Mar 29 01:43:38 > > 2025 +0000 > > > +++ b/sys/arch/mac68k/mac68k/pmap_bootstrap.c Sun Mar 30 13:36:45 > > 2025 +1100 > > > maxaddr = high[numranges - 1] - m68k_ptob(1); > > > > - high[numranges - 1] -= (m68k_round_page(MSGBUFSIZE) + > > m68k_ptob(1)); > > > + > > + /* > > + * Reduce high by an extra 7 pages which seem to be used by > > the EASC > > > + * ascaudio(4) on some machines. > > The above your messsage says ascaudio has issues. > The next your message says you're not sure what's going on. > It's better to look into it more before modifying pmap. > With typical audio MD drivers, it's unlikely to need pmap > modification (though I'm not sure about mac specific). > --- > Tetsuya Isaki <is...@pastel-flower.jp / is...@netbsd.org>
Sorry I'm mistaken - ascaudio(4) does NOT have issues on the PowerBook 160 or 520/540. Which I have tested it throughly on. - The audio MD code does not modify the top of memory. What I'm observing on the PowerBook 520 is that a portion of the top of memory is being used by the hardare when audio fifos are started.....This can only be observed when using the audio hardware on the PowerBook 520/540. It is not an issue with all macs but as I only have a small number of machines to test with I cannot be certain as to which models are affected. And for the sake of an addiotional 7 pages (56kB) it ensures that the system functions with an AUDIO kernel or GENERIC for all macs - should include all macs I cannot test on too. Unless people with differnent hardware were to test I'm unable to refine this patch further :( Best regards, Nat
