Hello Paul Cercueil,
The patch fc1acf317b01: "drm/ingenic: Add support for the IPU" from Jul 16, 2020, leads to the following static checker warning:
drivers/gpu/drm/ingenic/ingenic-drm-drv.c:232 ingenic_drm_crtc_atomic_check() error: potentially dereferencing uninitialized 'ipu_state'.
drivers/gpu/drm/ingenic/ingenic-drm-drv.c 197 static int ingenic_drm_crtc_atomic_check(struct drm_crtc *crtc, 198 struct drm_crtc_state *state) 199 { 200 struct ingenic_drm *priv = drm_crtc_get_priv(crtc); 201 struct drm_plane_state *f1_state, *f0_state, *ipu_state; 202 long rate; 203 204 if (!drm_atomic_crtc_needs_modeset(state)) 205 return 0; 206 207 if (state->mode.hdisplay > priv->soc_info->max_width || 208 state->mode.vdisplay > priv->soc_info->max_height) 209 return -EINVAL; 210 211 rate = clk_round_rate(priv->pix_clk, 212 state->adjusted_mode.clock * 1000); 213 if (rate < 0) 214 return rate; 215 216 if (priv->soc_info->has_osd) { 217 f1_state = drm_atomic_get_plane_state(state->state, &priv->f1); 218 f0_state = drm_atomic_get_plane_state(state->state, &priv->f0); 219 220 if (IS_ENABLED(CONFIG_DRM_INGENIC_IPU) && priv->ipu_plane) {
Do we need to check both the CONFIG and the priv->ipu_plane or would it be sufficient to just check if (priv->ipu_plane) {?
221 ipu_state = drm_atomic_get_plane_state(state->state, priv->ipu_plane); 222 223 /* IPU and F1 planes cannot be enabled at the same time. */ 224 if (f1_state->fb && ipu_state->fb) { 225 dev_dbg(priv->dev, "Cannot enable both F1 and IPU\n"); 226 return -EINVAL; 227 } 228 } 229 230 /* If all the planes are disabled, we won't get a VBLANK IRQ */ 231 priv->no_vblank = !f1_state->fb && !f0_state->fb && 232 !(priv->ipu_plane && ipu_state->fb); ^^^^^^^^^^^^^^^ Because here we're only checking "priv->ipu_plane".
233 } 234 235 return 0; 236 }
regards, dan carpenter
On Mon, Jul 20, 2020 at 10:22:37AM +0300, dan.carpenter@oracle.com wrote:
Hello Paul Cercueil,
The patch fc1acf317b01: "drm/ingenic: Add support for the IPU" from Jul 16, 2020, leads to the following static checker warning:
drivers/gpu/drm/ingenic/ingenic-drm-drv.c:232 ingenic_drm_crtc_atomic_check() error: potentially dereferencing uninitialized 'ipu_state'.
drivers/gpu/drm/ingenic/ingenic-drm-drv.c 197 static int ingenic_drm_crtc_atomic_check(struct drm_crtc *crtc, 198 struct drm_crtc_state *state) 199 { 200 struct ingenic_drm *priv = drm_crtc_get_priv(crtc); 201 struct drm_plane_state *f1_state, *f0_state, *ipu_state; 202 long rate; 203 204 if (!drm_atomic_crtc_needs_modeset(state)) 205 return 0; 206 207 if (state->mode.hdisplay > priv->soc_info->max_width || 208 state->mode.vdisplay > priv->soc_info->max_height) 209 return -EINVAL; 210 211 rate = clk_round_rate(priv->pix_clk, 212 state->adjusted_mode.clock * 1000); 213 if (rate < 0) 214 return rate; 215 216 if (priv->soc_info->has_osd) { 217 f1_state = drm_atomic_get_plane_state(state->state, &priv->f1); 218 f0_state = drm_atomic_get_plane_state(state->state, &priv->f0);
drivers/gpu/drm/ingenic/ingenic-drm-drv.c:224 ingenic_drm_crtc_atomic_check() error: 'f1_state' dereferencing possible ERR_PTR() drivers/gpu/drm/ingenic/ingenic-drm-drv.c:224 ingenic_drm_crtc_atomic_check() error: 'ipu_state' dereferencing possible ERR_PTR() drivers/gpu/drm/ingenic/ingenic-drm-drv.c:231 ingenic_drm_crtc_atomic_check() error: 'f1_state' dereferencing possible ERR_PTR() drivers/gpu/drm/ingenic/ingenic-drm-drv.c:231 ingenic_drm_crtc_atomic_check() error: 'f0_state' dereferencing possible ERR_PTR() drivers/gpu/drm/ingenic/ingenic-drm-drv.c:232 ingenic_drm_crtc_atomic_check() error: 'ipu_state' dereferencing possible ERR_PTR()
Also Smatch complains that these need error checking.
regards, dan carpenter
On Mon, Jul 20, 2020 at 10:22:37AM +0300, dan.carpenter@oracle.com wrote:
Hello Paul Cercueil,
The patch fc1acf317b01: "drm/ingenic: Add support for the IPU" from Jul 16, 2020, leads to the following static checker warning:
drivers/gpu/drm/ingenic/ingenic-drm-drv.c:232 ingenic_drm_crtc_atomic_check() error: potentially dereferencing uninitialized 'ipu_state'.
drivers/gpu/drm/ingenic/ingenic-drm-drv.c 197 static int ingenic_drm_crtc_atomic_check(struct drm_crtc *crtc, 198 struct drm_crtc_state *state) 199 { 200 struct ingenic_drm *priv = drm_crtc_get_priv(crtc); 201 struct drm_plane_state *f1_state, *f0_state, *ipu_state; 202 long rate; 203 204 if (!drm_atomic_crtc_needs_modeset(state)) 205 return 0; 206 207 if (state->mode.hdisplay > priv->soc_info->max_width || 208 state->mode.vdisplay > priv->soc_info->max_height) 209 return -EINVAL;
Random entirely unrelated drive-through comment: These mode checks should be in crtc_helper_funcs->mode_valid so that they're shared between the atomic_check and output probe paths.
But preexisting issues, just something that would be nice to rectify.
Cheers, Daniel
210 211 rate = clk_round_rate(priv->pix_clk, 212 state->adjusted_mode.clock * 1000); 213 if (rate < 0) 214 return rate; 215 216 if (priv->soc_info->has_osd) { 217 f1_state = drm_atomic_get_plane_state(state->state, &priv->f1); 218 f0_state = drm_atomic_get_plane_state(state->state, &priv->f0); 219 220 if (IS_ENABLED(CONFIG_DRM_INGENIC_IPU) && priv->ipu_plane) {
Do we need to check both the CONFIG and the priv->ipu_plane or would it be sufficient to just check if (priv->ipu_plane) {?
221 ipu_state = drm_atomic_get_plane_state(state->state, priv->ipu_plane); 222 223 /* IPU and F1 planes cannot be enabled at the same time. */ 224 if (f1_state->fb && ipu_state->fb) { 225 dev_dbg(priv->dev, "Cannot enable both F1 and IPU\n"); 226 return -EINVAL; 227 } 228 } 229 230 /* If all the planes are disabled, we won't get a VBLANK IRQ */ 231 priv->no_vblank = !f1_state->fb && !f0_state->fb && 232 !(priv->ipu_plane && ipu_state->fb); ^^^^^^^^^^^^^^^ Because here we're only checking "priv->ipu_plane".
233 } 234 235 return 0; 236 }
regards, dan carpenter _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Hi Dan,
Le lun. 20 juil. 2020 à 10:22, dan.carpenter@oracle.com a écrit :
Hello Paul Cercueil,
The patch fc1acf317b01: "drm/ingenic: Add support for the IPU" from Jul 16, 2020, leads to the following static checker warning:
drivers/gpu/drm/ingenic/ingenic-drm-drv.c:232 ingenic_drm_crtc_atomic_check() error: potentially dereferencing uninitialized 'ipu_state'.
I fixed it yesterday, commit 40a55dc13e9d in drm-misc-next.
drivers/gpu/drm/ingenic/ingenic-drm-drv.c 197 static int ingenic_drm_crtc_atomic_check(struct drm_crtc *crtc, 198 struct drm_crtc_state *state) 199 { 200 struct ingenic_drm *priv = drm_crtc_get_priv(crtc); 201 struct drm_plane_state *f1_state, *f0_state, *ipu_state; 202 long rate; 203 204 if (!drm_atomic_crtc_needs_modeset(state)) 205 return 0; 206 207 if (state->mode.hdisplay > priv->soc_info->max_width || 208 state->mode.vdisplay > priv->soc_info->max_height) 209 return -EINVAL; 210 211 rate = clk_round_rate(priv->pix_clk, 212 state->adjusted_mode.clock * 1000); 213 if (rate < 0) 214 return rate; 215 216 if (priv->soc_info->has_osd) { 217 f1_state = drm_atomic_get_plane_state(state->state, &priv->f1); 218 f0_state = drm_atomic_get_plane_state(state->state, &priv->f0); 219 220 if (IS_ENABLED(CONFIG_DRM_INGENIC_IPU) && priv->ipu_plane) {
Do we need to check both the CONFIG and the priv->ipu_plane or would it be sufficient to just check if (priv->ipu_plane) {?
It would be enough, yes.
Cheers, -Paul
221 ipu_state = drm_atomic_get_plane_state(state->state, priv->ipu_plane); 222 223 /* IPU and F1 planes cannot be enabled at the same time. */ 224 if (f1_state->fb && ipu_state->fb) { 225 dev_dbg(priv->dev, "Cannot enable both F1 and IPU\n"); 226 return -EINVAL; 227 } 228 } 229 230 /* If all the planes are disabled, we won't get a VBLANK IRQ */ 231 priv->no_vblank = !f1_state->fb && !f0_state->fb && 232 !(priv->ipu_plane && ipu_state->fb); ^^^^^^^^^^^^^^^ Because here we're only checking "priv->ipu_plane".
233 } 234 235 return 0; 236 }
regards, dan carpenter
dri-devel@lists.freedesktop.org
-
Dan Carpenter -
dan.carpenter＠oracle.com
-
Daniel Vetter -
Paul Cercueil