drm/i915/gt: Check cacheline is valid before acquiring
authorChris Wilson <chris@chris-wilson.co.uk>
Mon, 27 Apr 2020 09:30:38 +0000 (10:30 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Mon, 27 Apr 2020 10:39:23 +0000 (11:39 +0100)
commit2759e395358b2b909577928894f856ab75bea41a
treec576fec4c6dbdbfc68c97530a1a2741108fec835
parent68ace460c5b2a96a82ee49ab0b589ceed8abd000
drm/i915/gt: Check cacheline is valid before acquiring

The hwsp_cacheline pointer from i915_request is very, very flimsy. The
i915_request.timeline (and the hwsp_cacheline) are lost upon retiring
(after an RCU grace). Therefore we need to confirm that once we have the
right pointer for the cacheline, it is not in the process of being
retired and disposed of before we attempt to acquire a reference to the
cacheline.

<3>[  547.208237] BUG: KASAN: use-after-free in active_debug_hint+0x6a/0x70 [i915]
<3>[  547.208366] Read of size 8 at addr ffff88822a0d2710 by task gem_exec_parall/2536

<4>[  547.208547] CPU: 3 PID: 2536 Comm: gem_exec_parall Tainted: G     U            5.7.0-rc2-ged7a286b5d02d-kasan_117+ #1
<4>[  547.208556] Hardware name: Dell Inc. XPS 13 9350/, BIOS 1.4.12 11/30/2016
<4>[  547.208564] Call Trace:
<4>[  547.208579]  dump_stack+0x96/0xdb
<4>[  547.208707]  ? active_debug_hint+0x6a/0x70 [i915]
<4>[  547.208719]  print_address_description.constprop.6+0x16/0x310
<4>[  547.208841]  ? active_debug_hint+0x6a/0x70 [i915]
<4>[  547.208963]  ? active_debug_hint+0x6a/0x70 [i915]
<4>[  547.208975]  __kasan_report+0x137/0x190
<4>[  547.209106]  ? active_debug_hint+0x6a/0x70 [i915]
<4>[  547.209127]  kasan_report+0x32/0x50
<4>[  547.209257]  ? i915_gemfs_fini+0x40/0x40 [i915]
<4>[  547.209376]  active_debug_hint+0x6a/0x70 [i915]
<4>[  547.209389]  debug_print_object+0xa7/0x220
<4>[  547.209405]  ? lockdep_hardirqs_on+0x348/0x5f0
<4>[  547.209426]  debug_object_assert_init+0x297/0x430
<4>[  547.209449]  ? debug_object_free+0x360/0x360
<4>[  547.209472]  ? lock_acquire+0x1ac/0x8a0
<4>[  547.209592]  ? intel_timeline_read_hwsp+0x4f/0x840 [i915]
<4>[  547.209737]  ? i915_active_acquire_if_busy+0x66/0x120 [i915]
<4>[  547.209861]  i915_active_acquire_if_busy+0x66/0x120 [i915]
<4>[  547.209990]  ? __live_alloc.isra.15+0xc0/0xc0 [i915]
<4>[  547.210005]  ? rcu_read_lock_sched_held+0xd0/0xd0
<4>[  547.210017]  ? print_usage_bug+0x580/0x580
<4>[  547.210153]  intel_timeline_read_hwsp+0xbc/0x840 [i915]
<4>[  547.210284]  __emit_semaphore_wait+0xd5/0x480 [i915]
<4>[  547.210415]  ? i915_fence_get_timeline_name+0x110/0x110 [i915]
<4>[  547.210428]  ? lockdep_hardirqs_on+0x348/0x5f0
<4>[  547.210442]  ? _raw_spin_unlock_irq+0x2a/0x40
<4>[  547.210567]  ? __await_execution.constprop.51+0x2e0/0x570 [i915]
<4>[  547.210706]  i915_request_await_dma_fence+0x8f7/0xc70 [i915]

Fixes: 85bedbf191e8 ("drm/i915/gt: Eliminate the trylock for reading a timeline's hwsp")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: <stable@vger.kernel.org> # v5.6+
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200427093038.29219-1-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/gt/intel_timeline.c