drm/i915/gvt: Fix use-after-free in intel_vgpu_create_workload

Submitted by Gustavo A. R. Silva on Aug. 7, 2019, 2:20 a.m.

Details

Message ID 20190807022033.GA22623@embeddedor
State New
Headers show
Series "drm/i915/gvt: Fix use-after-free in intel_vgpu_create_workload" ( rev: 1 ) in Intel GVT devel

Not browsing as part of any series.

Commit Message

Gustavo A. R. Silva Aug. 7, 2019, 2:20 a.m.
kmem_cache_free() frees *workload*, hence there is a use-after-free bug
when calling function gvt_vgpu_err().

Fix this by storing the value of workload->wa_ctx.indirect_ctx.guest_gma
and workload->wa_ctx.per_ctx.guest_gma into automatic variable
guest_gma before freeing *workload*, for its further use.

Addresses-Coverity-ID: 1452235 ("Read from pointer after free")
Fixes: 2089a76ade90 ("drm/i915/gvt: Checking workload's gma earlier")
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
---
 drivers/gpu/drm/i915/gvt/scheduler.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

Patch hide | download patch | download mbox

diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 32ae6b5b7e16..c8cdb4a309f6 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -1525,9 +1525,11 @@  intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id,
 			if (!intel_gvt_ggtt_validate_range(vgpu,
 				workload->wa_ctx.indirect_ctx.guest_gma,
 				workload->wa_ctx.indirect_ctx.size)) {
+				unsigned long guest_gma =
+				       workload->wa_ctx.indirect_ctx.guest_gma;
 				kmem_cache_free(s->workloads, workload);
 				gvt_vgpu_err("invalid wa_ctx at: 0x%lx\n",
-				    workload->wa_ctx.indirect_ctx.guest_gma);
+					     guest_gma);
 				return ERR_PTR(-EINVAL);
 			}
 		}
@@ -1539,9 +1541,11 @@  intel_vgpu_create_workload(struct intel_vgpu *vgpu, int ring_id,
 			if (!intel_gvt_ggtt_validate_range(vgpu,
 				workload->wa_ctx.per_ctx.guest_gma,
 				CACHELINE_BYTES)) {
+				unsigned long guest_gma =
+					workload->wa_ctx.per_ctx.guest_gma;
 				kmem_cache_free(s->workloads, workload);
 				gvt_vgpu_err("invalid per_ctx at: 0x%lx\n",
-					workload->wa_ctx.per_ctx.guest_gma);
+					     guest_gma);
 				return ERR_PTR(-EINVAL);
 			}
 		}