Pinpointing source of corruption?

[curaga]

Ever since updating from Xorg 7.4 to 7.5 there's been corruption, in the background pixmap and in app widgets. It is capturable with xwd:



The radeon driver is the latest from the 6.12 branch, X 1.7.5, Mesa 7.7, Radeon X1300 Pro pcie, UMS.

Since the first suspect, radeon driver, only got an update from 6.12.4 to 6.12 git, what else to look for?


Oh, any 3d also gives a black screen and Xorg taking 100% cpu, unkillable. SSH and Sysrq works, and nothing in logs. Ain't it fun with the new stacks

[whizse]

Downgrading package by package until you can pinpoint which component the problem is in, and then (from a known working version) use git bisect?

[curaga]

The issue with that would be that downgrading one package likely requires a rebuild of everything atop it, and the fact I don't have a fast trigger for the corruption.
It takes some tens of minutes from a clean boot for it to show.

[eslrahc]

Are you running a kms enabled kernel?

If you are, load the radeon module Before you start X.
You can do so manually or by adding it to modprobe.preload.

[curaga]

No KMS, I thought I mentioned UMS in the post.
I wonder, does anyone else see something like this?

[whizse]

Unless you're using gl compositing or seeing the corruption after using a gl app I would guess the problem is somewhere in radeon driver. (6.12.4 does seem to be quite old. I guess a lot of development have happened since...)

I tried building 6.12.4 against 1.7.5 (not running, no hardware to test on yet, sorry) and it works, so it's probably easy to downgrade and bisect if that's where the problem lies.

[curaga]

Well crap, I'm afraid my card is dying. Glxgears brings the system down even on the old stack now, as well as on some live cd's.

I haven't seen the corruption yet on other stacks than this, but I haven't used them long either.

Got one log piece after a crash, dunno if it says anything:

Mar 4 11:42:50 Valinor kernel: irq 16: nobody cared (try booting with the "irqpoll" option)
Mar 4 11:42:50 Valinor kernel: Pid: 0, comm: swapper Not tainted 2.6.32.7-laxy #333
Mar 4 11:42:50 Valinor kernel: Call Trace:
Mar 4 11:42:50 Valinor kernel: <IRQ> [<ffffffff81059b3a>] ? __report_bad_irq+0x30/0x7d
Mar 4 11:42:50 Valinor kernel: [<ffffffff81059c8c>] ? note_interrupt+0x105/0x151
Mar 4 11:42:50 Valinor kernel: [<ffffffff8105a4e3>] ? handle_fasteoi_irq+0x8a/0xaa
Mar 4 11:42:50 Valinor kernel: [<ffffffff8100d75b>] ? handle_irq+0x17/0x20
Mar 4 11:42:50 Valinor kernel: [<ffffffff8100ce25>] ? do_IRQ+0x57/0xb6
Mar 4 11:42:50 Valinor kernel: [<ffffffff8100b653>] ? ret_from_intr+0x0/0xa
Mar 4 11:42:50 Valinor kernel: <EOI> [<ffffffff810119a2>] ? mwait_idle+0x55/0x58
Mar 4 11:42:50 Valinor kernel: [<ffffffff81009eda>] ? cpu_idle+0x53/0x6d
Mar 4 11:42:50 Valinor kernel: [<ffffffff81590c20>] ? start_kernel+0x325/0x331
Mar 4 11:42:50 Valinor kernel: [<ffffffff81590381>] ? x86_64_start_kernel+0xea/0xee
Mar 4 11:42:50 Valinor kernel: handlers:
Mar 4 11:42:50 Valinor kernel: [<ffffffff8125aa67>] (usb_hcd_irq+0x0/0x5d)
Mar 4 11:42:50 Valinor kernel: [<ffffffffa0028377>] (radeon_driver_irq_handler+0x0/0x145 [radeon])
Mar 4 11:42:50 Valinor kernel: Disabling IRQ #16

Anyways, the 2d crash, 3d crash, and corruption might all be because of the card, or not :P
Will buy a new r500 soon. Might have gone 'green, except that it's not that well supported yet.