Skip to contentSkip to conversation

nouveau: explosion in nouveau_fence_signal()

DISCUSSION Mike Galbraith 1 participant 1 message Oct 26, 2013

lore
Share & tools
lore
#0 lore raw

3.12.0-rt1 only exists here, so add a pile of salt. Kernel seems stable though, first sign of troubles was when I decided to see if nouveau rendering speed has improved since last time I tried it (ages ago), so I'll post the crash just in case.

I didn't get to check rendering speed..

[    56.529] (EE) Failed to initialize GLX extension (Compatible NVIDIA X driver not found)

..hohum, back to plain ole nv.

crash> bt
PID: 8580 TASK: ffff88021a3998a0 CPU: 0 COMMAND: "Xorg"
#0 [ffff88003731d7e0] machine_kexec at ffffffff8103a361 #1 [ffff88003731d840] crash_kexec at ffffffff810d1763 #2 [ffff88003731d910] oops_end at ffffffff815e8da8
#3 [ffff88003731d940] no_context at ffffffff815d1ad4 #4 [ffff88003731d990] __bad_area_nosemaphore at ffffffff815d1cb8 #5 [ffff88003731d9f0] bad_area at ffffffff815d1d2a
#6 [ffff88003731da20] __do_page_fault at ffffffff815eb566 #7 [ffff88003731db40] do_page_fault at ffffffff815eb699 #8 [ffff88003731db50] page_fault at ffffffff815e81c8

    [exception RIP: nouveau_fence_signal+102]
    RIP: ffffffffa0439356  RSP: ffff88003731dc08  RFLAGS: 00010246
    RAX: 002e0201003602b2  RBX: ffff88007985a3c0  RCX: dead000000100100
    RDX: 0000000000000000  RSI: dead000000200200  RDI: 0000000000000001
    RBP: ffff88003731dc28   R8: 0000000000000000   R9: 0000000000000000
    R10: 0000000000000001  R11: 0000000000000246  R12: ffff88007985aa30
    R13: ffff88007985aa50  R14: ffff88007985aa40  R15: ffff8802221f7c48
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018

#9 [ffff88003731dc30] nouveau_fence_done at ffffffffa04395e5 [nouveau]
#10 [ffff88003731dc80] nouveau_fence_work at ffffffffa0439c1e [nouveau] #11 [ffff88003731dcc0] nouveau_gem_object_unmap at ffffffffa043e911 [nouveau] #12 [ffff88003731dd00] nouveau_gem_object_close at ffffffffa043fd6b [nouveau] #13 [ffff88003731dd20] drm_gem_handle_delete at ffffffffa0198f7a [drm] #14 [ffff88003731dd70] drm_gem_close_ioctl at ffffffffa01992d3 [drm] #15 [ffff88003731dd80] drm_ioctl at ffffffffa0196eca [drm]
#16 [ffff88003731dea0] nouveau_drm_ioctl at ffffffffa0437f3e [nouveau] #17 [ffff88003731dee0] do_vfs_ioctl at ffffffff811b1955
#18 [ffff88003731df20] sys_ioctl at ffffffff811b1c41
#19 [ffff88003731df80] system_call_fastpath at ffffffff815ef5f9

    RIP: 00007f5cd3b0d837  RSP: 00007ffff73ccd70  RFLAGS: 00003202
    RAX: 0000000000000010  RBX: ffffffff815ef5f9  RCX: 0000000000000001
    RDX: 00007ffff73ccaf0  RSI: 0000000040086409  RDI: 000000000000000a
    RBP: 0000000040086409   R8: 000000000000032b   R9: 00000000414b0000
    R10: 0000000000000001  R11: 0000000000003246  R12: 0000000001463f00
    R13: 0000000000000000  R14: 0000000001463f18  R15: 000000000000000a
    ORIG_RAX: 0000000000000010  CS: 0033  SS: 002b
crash> mod -s nouveau
     MODULE       NAME                     SIZE  OBJECT FILE
ffffffffa04a3f00  nouveau                965202  /lib/modules/3.12.0-rt1/kernel/drivers/gpu/drm/nouveau/nouveau.ko
crash> gdb list *nouveau_fence_signal+102
0xffffffffa0439356 is in nouveau_fence_signal (include/linux/list.h:88).
83       * This is only for internal list manipulation where we know
84       * the prev/next entries already!
85       */

86 static inline void __list_del(struct list_head * prev, struct list_head * next)
87 {
88 next->prev = prev;
89 prev->next = next;
90 }
91
92 /**
crash> gdb list *nouveau_fence_signal+50
0xffffffffa0439322 is in nouveau_fence_signal (drivers/gpu/drm/nouveau/nouveau_fence.c:50).

45      static void
46      nouveau_fence_signal(struct nouveau_fence *fence)
47      {

48 struct fence_work *work, *temp;
49
50 list_for_each_entry_safe(work, temp, &fence->work, head) {
51 schedule_work(&work->base);
52 list_del(&work->head);
53 }
54
crash> gdb list *0xffffffffa04395e5
0xffffffffa04395e5 is in nouveau_fence_done (drivers/gpu/drm/nouveau/nouveau_fence.c:134).
129 list_for_each_entry_safe(fence, fnext, &fctx->pending, head) {
130 if (fctx->read(chan) < fence->sequence)
131 break;
132
133 nouveau_fence_signal(fence);
134 nouveau_fence_unref(&fence);
135 }
136 spin_unlock(&fctx->lock);
137 }
138
crash>