Bug 68456

Summary: [NV4B] null deref on load, NvI2C=1 makes it work
Product: xorg Reporter: Hans-Peter Deifel <hpdeifel>
Component: Driver/nouveauAssignee: Nouveau Project <nouveau>
Status: RESOLVED FIXED QA Contact: Xorg Project Team <xorg-team>
Severity: normal    
Priority: medium    
Version: git   
Hardware: Other   
OS: All   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
pass i2c functions into create func none

Description Hans-Peter Deifel 2013-08-22 22:57:00 UTC 
Loading the nouveau module leads to an instant crash. I was advised on #nouveau to try the config option NvI2C=1, which worked. But the default config is nevertheless broken. I have a "GeForce 7600 GT" card and use the latest (commit 3b56bba6) kernel from anongit.freedesktop.org master which seems to be rebased onto linux-3.11-rc6.

Here is the generated backtrace:

kernel: BUG: unable to handle kernel NULL pointer dereference at           (null)
kernel: IP: [<ffffffffa028e94b>] nouveau_i2c_pre_xfer+0xf/0x1d [nouveau]
kernel: PGD 0 
kernel: Oops: 0000 [#1] SMP 
kernel: Modules linked in: nouveau(+) video mxm_wmi i2c_algo_bit ttm drm_kms_helper drm hwmon_vid i2c_core usbhid usb_storage kvm_amd kvm pcspkr ohci_pci snd_hda_codec_via ohci_hcd ehci_pci k10temp r8169 snd_hda_intel ehci_hcd wmi snd_hda_codec snd_pcm snd_page_alloc snd_timer snd acpi_cpufreq mperf processor thermal_sys hwmon
kernel: CPU: 0 PID: 439 Comm: kworker/0:1 Not tainted 3.11.0-rc6-g3b56bba #2
kernel: Hardware name: System manufacturer System Product Name/M4A77TD, BIOS 0305    09/10/2009
kernel: Workqueue: events work_for_cpu_fn
kernel: task: ffff88012aff2050 ti: ffff88012ab4a000 task.ti: ffff88012ab4a000
kernel: RIP: 0010:[<ffffffffa028e94b>]  [<ffffffffa028e94b>] nouveau_i2c_pre_xfer+0xf/0x1d [nouveau]
kernel: RSP: 0018:ffff88012ab4b918  EFLAGS: 00010286
kernel: RAX: 0000000000000000 RBX: ffff880128fb7800 RCX: 0000000000000000
kernel: RDX: ffff880128fb7800 RSI: ffffffffa01ebdf5 RDI: ffff88012a2ba800
kernel: RBP: ffff88012a98ecc0 R08: 000000000000000a R09: 00000000fffffffb
kernel: R10: 0000000000000000 R11: ffff880123900001 R12: ffff88012a2ba820
kernel: R13: ffffffffa01ebdf5 R14: ffffffffa02f80c0 R15: ffff88012a98ef00
kernel: FS:  00007f273a9c0780(0000) GS:ffff88012fc00000(0000) knlGS:0000000000000000
kernel: CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
kernel: CR2: 0000000000000000 CR3: 0000000001527000 CR4: 00000000000007f0
kernel: Stack:
kernel: 0000000000000000 ffffffffa026871a ffff880128fb7800 ffff88012a98ecc0
kernel: ffff88012a2ba800 ffff88012a2ba820 ffffffffa02f80c0 ffffffffa028ed5a
kernel: 0000000000000020 ffff88012ab4ba7c ffff88012ab4b9e0 0000000000000000
kernel: Call Trace:
kernel: [<ffffffffa026871a>] ? __i2c_bit_add_bus+0x2a/0x2b3 [i2c_algo_bit]
kernel: [<ffffffffa028ed5a>] ? nouveau_i2c_port_create_+0x136/0x18a [nouveau]
kernel: [<ffffffffa02900b4>] ? nv04_i2c_port_ctor+0x2b/0x5c [nouveau]
kernel: [<ffffffffa027fa84>] ? dcb_i2c_entry+0x24/0x48 [nouveau]
kernel: [<ffffffffa027b00f>] ? nouveau_object_ctor+0x2b/0xb7 [nouveau]
kernel: [<ffffffffa028f030>] ? nouveau_i2c_create_+0xce/0x1de [nouveau]
kernel: [<ffffffffa0279ad3>] ? nouveau_event_create+0x1d/0x5e [nouveau]
kernel: [<ffffffffa029007b>] ? nv04_i2c_ctor+0x1f/0x2d [nouveau]
kernel: [<ffffffffa027b00f>] ? nouveau_object_ctor+0x2b/0xb7 [nouveau]
kernel: [<ffffffffa029b461>] ? nouveau_devobj_ctor+0x52f/0x5a3 [nouveau]
kernel: [<ffffffffa027b00f>] ? nouveau_object_ctor+0x2b/0xb7 [nouveau]
kernel: [<ffffffffa027b81e>] ? nouveau_object_new+0x162/0x20e [nouveau]
kernel: [<ffffffffa02c9c87>] ? nouveau_drm_load+0x154/0x565 [nouveau]
kernel: [<ffffffffa020c455>] ? drm_get_minor+0x196/0x1e8 [drm]
kernel: [<ffffffffa020e08b>] ? drm_get_pci_dev+0x141/0x243 [drm]
kernel: [<ffffffff811a762c>] ? __pci_set_master+0x22/0x6d
kernel: [<ffffffffa02c9656>] ? nouveau_drm_probe+0x1cb/0x1ee [nouveau]
kernel: [<ffffffff811aa8de>] ? local_pci_probe+0x34/0x5b
kernel: [<ffffffff8105350e>] ? work_for_cpu_fn+0xb/0x11
kernel: [<ffffffff8105502d>] ? process_one_work+0x1c1/0x2c8
kernel: [<ffffffff8105514c>] ? process_scheduled_works+0x18/0x25
kernel: [<ffffffff81055875>] ? worker_thread+0x1eb/0x29b
kernel: [<ffffffff8105568a>] ? manage_workers.isra.25+0x1ae/0x1ae
kernel: [<ffffffff81059f28>] ? kthread+0xad/0xb5
kernel: [<ffffffff81059e7b>] ? __kthread_parkme+0x5e/0x5e
kernel: [<ffffffff813b0d6c>] ? ret_from_fork+0x7c/0xb0
kernel: [<ffffffff81059e7b>] ? __kthread_parkme+0x5e/0x5e
kernel: Code: 48 81 c6 80 dc 00 00 e8 69 fd f0 e0 8b 44 24 08 48 83 c4 10 5b c3 e9 10 f2 ff ff 90 51 48 8b 47 18 48 8b 38 48 8b 87 50 03 00 00 <48> 8b 00 48 85 c0 74 02 ff d0 31 c0 5a c3 48 8b 87 50 03 00 00 
kernel: RIP  [<ffffffffa028e94b>] nouveau_i2c_pre_xfer+0xf/0x1d [nouveau]
kernel: RSP <ffff88012ab4b918>
kernel: CR2: 0000000000000000
kernel: ---[ end trace 451adc65612b1d9f ]---
Comment 1 Ilia Mirkin 2013-08-23 01:13:40 UTC 
The code decodes to

  1c:   51                      push   %rcx
  1d:   48 8b 47 18             mov    0x18(%rdi),%rax
  21:   48 8b 38                mov    (%rax),%rdi
  24:   48 8b 87 50 03 00 00    mov    0x350(%rdi),%rax
  2b:*  48 8b 00                mov    (%rax),%rax              <-- trapping instruction
  2e:   48 85 c0                test   %rax,%rax
  31:   74 02                   je     0x35
  33:   ff d0                   callq  *%rax
  35:   31 c0                   xor    %eax,%eax
  37:   5a                      pop    %rdx
  38:   c3                      retq   

Which means that port->func is NULL. Still trying to work out exactly how that happens.
Comment 2 Ilia Mirkin 2013-08-23 01:46:45 UTC 
Created attachment 84484 [details] [review]
pass i2c functions into create func

Can you give this patch a shot? It compiles, but I haven't tested it beyond that. Should fix the null deref. But it might still not work. (Obviously without the NvI2C=1 thing.) BTW, how are you hitting this? Did you set something in i2c like bit_test=1?
Comment 3 Hans-Peter Deifel 2013-08-23 16:54:15 UTC 
Your patch did work, thank you very much. I could load the module and use it for a few hours without problems.

I had indeed set bit_test=1 for i2c_algo_bit although I had forgotten about it.
Comment 4 Ilia Mirkin 2013-09-26 23:06:10 UTC 
This patch should now be in 3.12-rc1.

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.