My AGP X1300 often (maybe 80% of boots) fails to boot with radeon.modeset=1. No error messages are reported, but when the GPU is being initialized, my CRT screen goes blank (with signal on, the monitor reports 720x400@70Hz). After a few seconds, the machine reboots. Sometimes, instead of getting a blank screen, irregular vertical white&black stripes are displayed (at the aforementioned resolution). I've used netconsole to get the dmesg output, the log ends with: [drm] GART: num cpu pages 131072, num gpu pages 131072 I've tried to pinpoint the exact location of the problem by adding printk's between the relevant function calls in rv515.c (tested this on 2.6.32.3-17.fc12.x86_64, but the problem is here since the birth of the Universe). However, this made the bug disappear, as it apparently is some kind of a race condition. Adding printk's one by one I've found that the bug is somewhere between rv515_mc_program() and r100_rb2d_reset() in rv515_startup/rv515_gpu_init. Inserting anything between those two function calls makes the bug disappear. I've tried to fiddle with various options, notably acpi=off, pci=nomsi, radeon.agpmode=-1 and various AGP aperture sizes, with no effect.
Actually, there's a bug in my bugreport :-) There should be r100_hdp_reset() instead of r100_rb2d_reset() (i.e. inserting a printk between those two doesn't make the bug disappear).
My Radeon X1600 doesn't work with KMS too. However, sometimes the boot is sucessful (about 10% of tries). The kernel version is 2.6.34.
I have the same problem. X1600, Кernel 2.6.32-5
i am also having described boot failures. it started with 2.6.31 for me. the successful boot rate is ~ 1 of out 3 tries for me. i sometimes get lockups with black screen of mentioned vertical stripes. the problem is reproductible with gentoo 32bit kernel, arch linux 64bit kernel, and occasionally with fedora 12/13 kernel.
i forgot to add - X1300pro card (rv515) here. various kernels.
AGP ?
pci-e on my end. gigabyte k8nf9u mainboard (rev 1.x).
Closing as obsolete, please update the bug if it is seen with a modern kernel
KMS works just fine for me for a long time now (can't say exactly since when, but it's more than a year), I just forgot about this bug.