| Message ID | YA6F3oF8mRaNQWjb@mwanda (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915/gvt: fix uninitialized return in intel_gvt_update_reg_whitelist() | expand |
Quoting Dan Carpenter (2021-01-25 08:48:30) > Smatch found an uninitialized variable bug in this code: > > drivers/gpu/drm/i915/gvt/cmd_parser.c:3191 intel_gvt_update_reg_whitelist() > error: uninitialized symbol 'ret'. > > The first thing that Smatch complains about is that "ret" isn't set if > we don't enter the "for_each_engine(engine, &dev_priv->gt, id) {" loop. > Presumably we always have at least one engine so that's a false > positive. > > But it's definitely a bug to not set "ret" if i915_gem_object_pin_map() > fails. True. > Let's fix the bug and silence the false positive. > > Fixes: 493f30cd086e ("drm/i915/gvt: parse init context to update cmd accessible reg whitelist") > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> -Chris
On 2021.01.25 09:44:53 +0000, Chris Wilson wrote: > Quoting Dan Carpenter (2021-01-25 08:48:30) > > Smatch found an uninitialized variable bug in this code: > > > > drivers/gpu/drm/i915/gvt/cmd_parser.c:3191 intel_gvt_update_reg_whitelist() > > error: uninitialized symbol 'ret'. > > > > The first thing that Smatch complains about is that "ret" isn't set if > > we don't enter the "for_each_engine(engine, &dev_priv->gt, id) {" loop. > > Presumably we always have at least one engine so that's a false > > positive. > > > > But it's definitely a bug to not set "ret" if i915_gem_object_pin_map() > > fails. > > True. > > > Let's fix the bug and silence the false positive. > > > > Fixes: 493f30cd086e ("drm/i915/gvt: parse init context to update cmd accessible reg whitelist") > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> > Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Thanks, Dan & Chris! Queued this up.
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c index 7fb91de06557..d54ea0e4681d 100644 --- a/drivers/gpu/drm/i915/gvt/cmd_parser.c +++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c @@ -3103,7 +3103,7 @@ void intel_gvt_update_reg_whitelist(struct intel_vgpu *vgpu) struct intel_vgpu_submission *s = &vgpu->submission; struct i915_request *requests[I915_NUM_ENGINES] = {}; bool is_ctx_pinned[I915_NUM_ENGINES] = {}; - int ret; + int ret = 0; if (gvt->is_reg_whitelist_updated) return; @@ -3157,6 +3157,7 @@ void intel_gvt_update_reg_whitelist(struct intel_vgpu *vgpu) if (IS_ERR(vaddr)) { gvt_err("failed to pin init ctx obj, ring=%d, err=%lx\n", id, PTR_ERR(vaddr)); + ret = PTR_ERR(vaddr); goto out; }
Smatch found an uninitialized variable bug in this code: drivers/gpu/drm/i915/gvt/cmd_parser.c:3191 intel_gvt_update_reg_whitelist() error: uninitialized symbol 'ret'. The first thing that Smatch complains about is that "ret" isn't set if we don't enter the "for_each_engine(engine, &dev_priv->gt, id) {" loop. Presumably we always have at least one engine so that's a false positive. But it's definitely a bug to not set "ret" if i915_gem_object_pin_map() fails. Let's fix the bug and silence the false positive. Fixes: 493f30cd086e ("drm/i915/gvt: parse init context to update cmd accessible reg whitelist") Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> --- drivers/gpu/drm/i915/gvt/cmd_parser.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)