Kernel: repeatable hang in 2.2.15pre7, atyfb issues
Greetings,
All I have to do to send 2.2.15pre7 into a hang (well, to about 1% of
the machine's normal performance, judging from the speed of E ripples)
is type "minicom /dev/ttyS1".
In 2.2.13, it says "Device /dev/ttyS1 is locked." I did /etc/init.d/lpd
stop (I have printtool configured to try to use a serial printer), but
that didn't unlock ttyS1 in 2.2.13.
The modem is on S0, and I have a printer on S1 with PowerPrint (I'm
trying to figure out how to talk to it; also in the process of getting
help from InfoWave but it's taking them a while).
About a week ago, I had my first hang since August, on 2.2.15pre5.
Didn't wait long enough to see if it was actually slow and not hung as
described above. This was during heavy activity over PPP on S0, and
while doing something disk-intensive, and I think building a kernel- a
pretty severe stress test... maybe related to this pre7 problem with
ttyS1?
All this is on a Motorola StarMax 3000/160 (pmac 4400 clone, Tanzania
mobo, 160 MHz 603e).
Any ideas? I'd be glad to test patches. Should I send this to
linux-kernel?
Also, I'm seeing several problems on atyfb (Mach64 with 1MB, 2.2.12 thru
.15pre7):
1. The default video/color mode uses more vram than I have, so I have
to specify vmode in the boot line. Fine for me now, but was
confusing as a newbie- I could only get a console on the built-in
Mach64 video if I *disabled* atyfb in the kernel config, otherwise
it would bounce to the unsupported offb on MacPicasso in VGA mode.
Shouldn't the default fb mode be small enough to work on most
machines, then you go to a higher mode in Xfb or using fbset?
2. I boot into atyfb:vmode:18, but Xfb fails miserably in this mode
(when I use Modes "default" in XF86Config)- I get a white screen
with cursor which only redraws if I move the mouse fast. Should it
work in all requested modes, or does it only work in a few? Now I
use an 1152x864 modeline, which is very nice.
3. 16 bit 640x480 is bad in console and X- every pair of columns seems
to be swapped, e.g. 10325476 instead of 01234567 (or maybe it's in
groups of four, 23016745, I haven't looked too closely).
4. Reported this already, but in the above modeline, updates in a
partially obscured window don't seem to happen. For example, if
I scroll an emacs window which is partially under a Netscape
window, and then raise the emacs window, the part that was obscured
has not been scrolled. This doesn't happen in XSVGA (I've tried
Matrox Mill II, Voodoo3- which has other issues, NeoMagic), nor in
Xfb/cyber2000, so I think it is an atyfb issue.
5. I frequently get little random horizontal white lines about 8
pixels long, and often vertically aligned with other little lines
of the same type. Could something be accidentally writing bytes
into one of the bitplanes? (Does it even use bitplanes? I haven't
followed details of video rendering since the Amiga.)
Thanks,
--
Adam Powell http://lyre.mit.edu/~powell/
Thomas B. King Assistant Professor of Materials Engineering
77 Massachusetts Ave. Rm. 4-117 Phone (617) 452-2086
Cambridge, MA 02139 USA Fax (617) 253-5418
Reply to: