I've been asked to forward this bug report to the kernel team. Apparently the latest X (7.5) in Debian uses this new interface. I see exactly the same issue whether I use nvidia binary drivers or nv driver. Downgrading X to 7.4 completely resolves this issue. dmesg reports, [ 840.676052] INFO: task Xorg:3075 blocked for more than 120 seconds. [ 840.676060] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 840.676067] Xorg D 0000000000000000 0 3075 3060 0x00400000 [ 840.676078] ffff880100861e28 0000000000000082 0000000000000000 0000000000000008 [ 840.676088] 0000000000000400 ffffffff81408094 000000000000f8a0 ffff880100861fd8 [ 840.676097] 00000000000155c0 00000000000155c0 ffff88013bcbebd0 ffff88013fab3980 [ 840.676106] Call Trace: [ 840.676123] [<ffffffff812339b7>] vga_get+0x115/0x141 [ 840.676133] [<ffffffff8104f777>] ? default_wake_function+0x0/0x14 [ 840.676141] [<ffffffff81233a14>] ? vga_arb_write+0x31/0x449 [ 840.676148] [<ffffffff81233aa6>] vga_arb_write+0xc3/0x449 [ 840.676157] [<ffffffff8116bbe4>] ? security_file_permission+0x16/0x18 [ 840.676168] [<ffffffff81102538>] vfs_write+0xb0/0x10a [ 840.676176] [<ffffffff81102660>] sys_write+0x4c/0x74 [ 840.676184] [<ffffffff81010db2>] system_call_fastpath+0x16/0x1b Attached is a full Xorg.0.log
Created attachment 24910 [details] Xorg.0.log
Reassigned to DRI. Do you think this is a regression? Is any earlier kernel version OK with that X version? Thanks.
I don't think this is a regression. I've just tested X 7.5 with kernel 2.6.26, 2.6.32-rc8 and 2.6.32 with the same results. I will test if removing one of the GPUs makes a difference.