Bug 111892

https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_6989/shard-hsw5/igt@kms_flip@2x-flip-vs-panning.html

<4> [25.732321] ======================================================
<4> [25.732321] WARNING: possible circular locking dependency detected
<4> [25.732323] 5.4.0-rc1-CI-CI_DRM_6989+ #1 Tainted: G     U           
<4> [25.732323] ------------------------------------------------------
<4> [25.732324] kms_flip/1036 is trying to acquire lock:
<4> [25.732325] ffff8883eb92cce8 (
<4> [25.732327] hardirqs last disabled at (1064412): [<ffffffff8123f4b9>] __slab_alloc.isra.84.constprop.89+0x19/0x70
<4> [25.732328] &mapping->i_mmap_rwsem){++++}, at: unmap_mapping_pages+0x48/0x130
<4> [25.732332] 
but task is already holding lock:
<4> [25.732333] ffff8883f99093a0 (&vm->mutex){+.+.}, at: i915_vma_unbind+0xe6/0x4a0 [i915]
<4> [25.732366] softirqs last  enabled at (1063898): [<ffffffff81c00385>] __do_softirq+0x385/0x47f
<4> [25.732367] softirqs last disabled at (1063889): [<ffffffff810b7f4a>] irq_exit+0xba/0xc0
<4> [25.732369] 
which lock already depends on the new lock.

<4> [25.732370] 
the existing dependency chain (in reverse order) is:
<4> [25.732370] 
-> #2 (&vm->mutex){+.+.}:
<4> [25.732373]        __mutex_lock+0x9a/0x9d0
<4> [25.732402]        i915_vma_remove+0x53/0x250 [i915]
<4> [25.732431]        i915_vma_unbind+0x19c/0x4a0 [i915]
<4> [25.732485]        i915_gem_object_unbind+0x153/0x1c0 [i915]
<4> [25.732513]        userptr_mn_invalidate_range_start+0x9f/0x200 [i915]
<4> [25.732515]        __mmu_notifier_invalidate_range_start+0xa3/0x180
<4> [25.732516]        unmap_vmas+0x143/0x150
<4> [25.732517]        unmap_region+0xa3/0x100
<4> [25.732518]        __do_munmap+0x25d/0x490
<4> [25.732519]        __vm_munmap+0x6e/0xc0
<4> [25.732520]        __x64_sys_munmap+0x12/0x20
<4> [25.732521]        do_syscall_64+0x4f/0x210
<4> [25.732523]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
<4> [25.732524] 
-> #1 (mmu_notifier_invalidate_range_start){+.+.}:
<4> [25.732526]        page_mkclean_one+0xda/0x210
<4> [25.732527]        rmap_walk_file+0xff/0x260
<4> [25.732528]        page_mkclean+0x9f/0xb0
<4> [25.732530]        clear_page_dirty_for_io+0xa2/0x300
<4> [25.732532]        mpage_submit_page+0x1a/0x70
<4> [25.732533]        mpage_process_page_bufs+0xe7/0x110
<4> [25.732534]        mpage_prepare_extent_to_map+0x1d2/0x2b0
<4> [25.732536]        ext4_writepages+0x592/0x1230
<4> [25.732536]        do_writepages+0x46/0xe0
<4> [25.732538]        __filemap_fdatawrite_range+0xc6/0x100
<4> [25.732548]        file_write_and_wait_range+0x3c/0x90
<4> [25.732549]        ext4_sync_file+0x154/0x500
<4> [25.732551]        do_fsync+0x33/0x60
<4> [25.732553]        __x64_sys_fsync+0xb/0x10
<4> [25.732553]        do_syscall_64+0x4f/0x210
<4> [25.732555]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
<4> [25.732555] 
-> #0 (&mapping->i_mmap_rwsem){++++}:
<4> [25.732558]        __lock_acquire+0x1328/0x15d0
<4> [25.732559]        lock_acquire+0xa7/0x1c0
<4> [25.732560]        down_write+0x33/0x70
<4> [25.732561]        unmap_mapping_pages+0x48/0x130
<4> [25.732595]        i915_vma_revoke_mmap+0x81/0x1b0 [i915]
<4> [25.732637]        i915_vma_unbind+0xee/0x4a0 [i915]
<4> [25.732674]        i915_gem_object_ggtt_pin+0xee/0x430 [i915]
<4> [25.732699]        i915_gem_object_pin_to_display_plane+0xd1/0x130 [i915]
<4> [25.732728]        intel_pin_and_fence_fb_obj+0xb3/0x230 [i915]
<4> [25.732757]        intel_plane_pin_fb+0x3c/0xd0 [i915]
<4> [25.732786]        intel_prepare_plane_fb+0x144/0x5d0 [i915]
<4> [25.732788]        drm_atomic_helper_prepare_planes+0x85/0x110
<4> [25.732816]        intel_atomic_commit+0xc6/0x2f0 [i915]
<4> [25.732818]        drm_atomic_helper_set_config+0x61/0x90
<4> [25.732818]        drm_mode_setcrtc+0x18e/0x720
<4> [25.732820]        drm_ioctl_kernel+0xa7/0xf0
<4> [25.732821]        drm_ioctl+0x2e1/0x390
<4> [25.732823]        do_vfs_ioctl+0xa0/0x6f0
<4> [25.732824]        ksys_ioctl+0x35/0x60
<4> [25.732825]        __x64_sys_ioctl+0x11/0x20
<4> [25.732826]        do_syscall_64+0x4f/0x210
<4> [25.732827]        entry_SYSCALL_64_after_hwframe+0x49/0xbe
<4> [25.732828] 
other info that might help us debug this:

<4> [25.732828] Chain exists of:
  &mapping->i_mmap_rwsem --> mmu_notifier_invalidate_range_start --> &vm->mutex

<4> [25.732830]  Possible unsafe locking scenario:

<4> [25.732830]        CPU0                    CPU1
<4> [25.732831]        ----                    ----
<4> [25.732831]   lock(&vm->mutex);
<4> [25.732832]                                lock(mmu_notifier_invalidate_range_start);
<4> [25.732833]                                lock(&vm->mutex);
<4> [25.732833]   lock(&mapping->i_mmap_rwsem);
<4> [25.732834] 
 *** DEADLOCK ***