Segment Violation in Painter::StraightLine()

[scottmc]

While debugging #1479 on my thinkpad 570E I was able to get it to boot up with the crazy color video reported in #1169 and then set the video mode to 1024x768, 16bit color and reboot. It them boots up to a blue screen with a working hand cursor but after about 10 seconds drops to GDB with this error:

[tcsetpgrp failed in terminal_inferior: Invalid Argument] Thread 110 caused an exception: Segment violation Reading symbols yada yada.... ... ... ... Reading symbols from /boot/beos/system/add-ons/accelerants/neomagic.accelerant [tcsetpgrp failed in terminal_inferior: Invalid Argument]

[Switching to team /boot/beos/system/servers/app_server (42) thread w:68:Terminal 1 (110)] 0x002922ca in Painter::StraightLine () (gdb)

Note that I typed this in by hand so might be typos. This was with r22697. I will attach serial debug as well.

[meanwhile]

I've a similar situation with a similar Thinkpad (570), maybe the serial debug from it can help as there are similarities in them as well.

[scottmc]

Sometimes when booting up it hangs with a black screen and the Haiku logo. Serial debug shows it hanging at: via: init_hardware() failed: General system error

never getting to the next line, unless I try rebooting: ps2_hid: init_hardware

So it seems there might also be an issue with ps2_hid: init_hardware as well.

meanwhile do you see this same thing happening on your 570?

[meanwhile]

In the setup here (two BeOS partitons and a Haiku partition, managed by Bootman) each Haiku revision has always hung at its bootscreen after a reboot from one of the BeOS installs.

Restarting from Haiku also doesn't work: the cursor freezes at the stage in the restart process where the message 'Asking other processes to quit' is displayed. When choosing shutdown instead and then clicking the 'Restart System' button in the 'System is Shut Down' window, the cursor also freezes.

Just found that a reboot from BeOS into Haiku gives

"via: init_hardware() failed: General System error"

as the last line of the debug output here too.

At this point I can't reboot though; if you could let me know how you reboot to get to the next line ?

[meanwhile]

BTW, this was all done with r22509. I'll get the latest revision in the next days to see if #1479 changed things here too.

[meanwhile]

As for restarting from Haiku, the reboot gets further in r22724: it now hangs at Haiku's bootscreen. However, except for the successfull very first boot of this revision, a next boot of it gives a GDB screen (see attachment).

[scottmc]

Replying to meanwhile:

As for restarting from Haiku, the reboot gets further in r22724: it now hangs at Haiku's bootscreen. However, except for the successfull very first boot of this revision, a next boot of it gives a GDB screen (see attachment).

OK so it sounds as if both of our systems are doing the exact same thing. First boot hangs, and then a second boot attempy boots up but then drops to GDB with the Painter::StraightLine() seg fault. I'm not sure but this might be 2 or 3 bugs. Anyways we're here to test. Bug one would be why does it hang at "via: init_hardware() failed: General System error" or is it hanging on the ps2 command which is next? Bug two is this ticket, the Painter::StraightLine(). Anyone else seeing this or is it just with Neomagic chipset or just Thinkpad 570 series? A listdev of a 570E can be seen in ticket #1169.

[stippi]

Could you try if you still get the crash in Painter::StraightLine() with r22736?

My suspicion is that the graphics driver and/or accelerant are not being initialized properly, and the Painters are not attached to the frame buffer correctly. I don't really know though. Would be great to have such a machine to test... but maybe not me but someone able to maintain the Via driver...

[meanwhile]

Tried it with r22736, but it still behaves the same. Maybe the debug info tells more... (see r22736debug.txt). Since last week I have a second TP 570, so it wouldn't hurt so much to see one of them leave for a testing period.

[meanwhile]

Clarification: Now only at the third boot there's a GDB screen showing. The r22736.txt is the serial debug output; what shows at this 3rd boot is seen in 'screen.png'.

[meanwhile]

scottmc: maybe you want to have another try with r23281 (or later), because here this bug seems solved.

[scottmc]

I just put r23287 on the Thinkpad 570E and this bug seems to be gone now, so OK to close.

[stippi]

Hm. I am not sure if I understood the original bug, but if you can't reproduce it anymore... :-)

[meanwhile]

Thanks nevertheless, stippi. I'm tempted to put an easy "typical IBM quirk" label on it...