Nvidia driver lockup with 2 PCI express cards

[vflorins]

I am seeing a consistent system freeze (as I run startx) with 2 PCIE cards (a 8800GTS and a Quadro 1400). The system is configured to drive 4 screens (2 TwinView connected with Xinerama). Here is the relevant output from the system log:

Apr 5 16:27:06 pierce-dhcp-201 kernel: NVRM: Xid (0002:00): 6, PE0000 0000 31415930 00000000 ffffffff 00000000
Apr 5 16:27:16 pierce-dhcp-201 kernel: BUG: soft lockup detected on CPU#0!
Apr 5 16:27:16 pierce-dhcp-201 kernel:
Apr 5 16:27:16 pierce-dhcp-201 kernel: Call Trace:
Apr 5 16:27:16 pierce-dhcp-201 kernel: <IRQ> [<ffffffff802b1004>] softlockup_tick+0xdb/0xf6
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff80211fa5>] __do_softirq+0x55/0xc3
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8028f925>] update_process_times+0x42/0x68
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff80272084>] smp_local_timer_interrupt+0x34/0x55
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff80272760>] smp_apic_timer_interrupt+0x51/0x69
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8025ace6>] apic_timer_interrupt+0x66/0x70
Apr 5 16:27:16 pierce-dhcp-201 kernel: <EOI> [<ffffffff80407976>] pci_conf1_read+0x0/0xc0
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881c8967>] :nvidia:_nv003350rm+0xf/0x10
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff883af5af>] :nvidia:_nv007025rm+0x7d/0xb0
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88394155>] :nvidia:_nv008120rm+0x16d/0x550
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88394150>] :nvidia:_nv008120rm+0x168/0x550
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881b5927>] :nvidia:_nv003595rm+0x3f/0x7c
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff883af6c2>] :nvidia:_nv007240rm+0x30/0x54
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881b5ab1>] :nvidia:_nv003589rm+0x27/0x36
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff883af5af>] :nvidia:_nv007025rm+0x7d/0xb0
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88395d78>] :nvidia:_nv009704rm+0x182/0x1be
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8837d48c>] :nvidia:_nv008127rm+0x8c/0x196
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88197c12>] :nvidia:_nv012070rm+0x18/0x44
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8819c554>] :nvidia:_nv012073rm+0x2c/0x14e
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88197038>] :nvidia:_nv012098rm+0x38/0x3e
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881a424d>] :nvidia:_nv002491rm+0x2f7/0x5ae
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88197c12>] :nvidia:_nv012070rm+0x18/0x44
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8819bf61>] :nvidia:_nv012097rm+0x65/0x100
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881aa893>] :nvidia:_nv002607rm+0x3e3/0x410
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff88197c12>] :nvidia:_nv012070rm+0x18/0x44
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881a9b31>] :nvidia:_nv002608rm+0xed/0x27a
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881cfb23>] :nvidia:rm_set_interrupts+0x11f/0x136
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff884d6835>] :nvidias_acquire_sema+0x61/0x79
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881a5602>] :nvidia:_nv004380rm+0x70/0xaa
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881d2701>] :nvidia:_nv002557rm+0x1a9/0x63a
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff881cfded>] :nvidia:rm_ioctl+0x9/0xe
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff884d37fc>] :nvidia:nv_kern_ioctl+0x35c/0x3ed
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff884d38cc>] :nvidia:nv_kern_unlocked_ioctl+0x1c/0x23
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8023f198>] do_ioctl+0x21/0x6b
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8022f347>] vfs_ioctl+0x25c/0x275
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff80249381>] sys_ioctl+0x59/0x78
Apr 5 16:27:16 pierce-dhcp-201 kernel: [<ffffffff8025a11e>] system_call+0x7e/0x83


The bug report is also attached. This was run on a freshly rebooted system, because the /proc/driver/nvidia entries become unreadable after a crash. Cards work fine individually.

[netllama]

Have you verified that you're using the most recent motherboard BIOS?

thanks,
Lonni

[llavalle]

Just FYI, I'm running dual PCIe and it works fine.

Card #1 is a 7900GT
Card #2 is a 7600GS

Both plugged in a Asus P5W DH (Dual PCIe 8x)

[vflorins]

Originally Posted by netllama

Have you verified that you're using the most recent motherboard BIOS?

thanks,
Lonni

Yes, this is running the most recent BIOS.

[vflorins]

An update: tried the same with a GeForce 6800 as the second card - still locks up. The kernel error messages are slightly different though and there is no "soft lockup" message.

Apr 10 16:20:52 pierce-dhcp-201 kernel: NVRM: Xid (0002:00): 6, PE0000 0000 31415930 00000000 ffffffff 00000000
Apr 10 16:21:28 pierce-dhcp-201 kernel: NVRM: Xid (0002:00): 8, Channel 00000020
Apr 10 16:22:12 pierce-dhcp-201 kernel: NVRM: Xid (0002:00): 8, Channel 00000020

Any suggestions?

[netllama]

I'll see if I can locate your motherboard to reproduce this failure.

thanks,
Lonni

[netllama]

I'm afraid that I'm not able to locate your motherboard, and this does not reproduce on any of the systems that I have here.

[vflorins]

I tested the same 2-card configuration on a different motherboard (Nforce 600 series chipset, SLI-capable) and found that it runs properly. So it was indeed the motherboard that was at fault.