| Summary: | X.org doesn't come up w/ an ancient i815 most of the time | ||
|---|---|---|---|
| Product: | xorg | Reporter: | Bernhard Rosenkraenzer <bero> |
| Component: | Driver/intel | Assignee: | Xorg Project Team <xorg-team> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | high | ||
| Version: | git | ||
| Hardware: | Other | ||
| OS: | other | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
This problem has apparently been around forever and on every OS, try searching google for "AGPIOC_INFO failed". There are hundreds of mailing list archive entries about exactly this issue [though usually without the "it works sometimes" part], dated between August 2000 and now, reported by users of Linux, NetBSD and OpenBSD. running "lsof|grep agp" before starting Xorg confirms that nothing else has opened /dev/agpgart. This seems to be related to 4k stacks in the kernel... Switching to 8k stacks makes it behave much better. So is this problem fixed with 8k stacks ? It seems that 4K stacks is an option and 8K stacks is the default. Bernhard, I'm going to close this now. If it's still a problem with X.Org 6.8.2, can you post a full log and reopen the bug. Thanks. |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.
X.org fails to start on this device: 00:02.0 Class 0300: 8086:1132 (rev 04) Subsystem: 0e11:001a Flags: bus master, 66Mhz, medium devsel, latency 0, IRQ 11 Memory at 44000000 (32-bit, prefetchable) [size=64M] Memory at 40400000 (32-bit, non-prefetchable) [size=512K] Expansion ROM at <unassigned> [disabled] Capabilities: [dc] Power Management version 2 The error is: (--) I810(0): Chipset: "i815" (--) I810(0): Linear framebuffer at 0x44000000 (--) I810(0): IO registers at addr 0x40400000 (WW) xf86AcquireGART: AGPIOC_ACQUIRE failed (Device or resource busy) (EE) GARTInit: AGPIOC_INFO failed (Invalid argument) (EE) I810(0): AGP GART support is not available. Make sure your kernel has agpgart support or that the agpgart kernel module is loaded. The AGP bridge is integrated into the i815 chip, so nothing unusual here either. AGP support is definitely there, and (without doing a lot of debugging yet) I believe the AGPIOC_ACQUIRE Device or resource busy must be caused by X.org opening the device twice; there's nothing else that might want to use AGP (no framebuffer drivers, no other X servers, nothing). An even weirder thing is, sometimes X.org does come up (approximately once in 15 tries), and once it is up, it works perfectly, even with DRI. I suspect there's some race condition (/dev/agpgart opened again before its first usage is closed under certain circumstances?) in the driver. Verified with current X.org and Linux kernels 2.6.10, 2.6.10-mm1 and 2.6.10-ac12.