Skip to contentSkip to conversation

drm: use drm_warn() in validate_blend_mode_for_alpha_formats()

PATCH Superseded Leandro Ribeiro 4 participants 9 messages Jul 29 – Jul 30

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover drm: use drm_warn() in validate_blend_mode_for_alpha_formats()

This replaces a WARN() with drm_warn() in
validate_blend_mode_for_alpha_formats().

The condition is expected to be triggered by existing drivers while they adapt to the new blend mode requirement, so it should not use the kernel warning report mechanism. WARN() output can cause fuzzers to stop when they encounter a "WARNING:" message.

drm_warn() keeps the warning visible to driver developers without triggering a WARN().

Leandro Ribeiro (1):
drm: use drm_warn() in validate_blend_mode_for_alpha_formats()

 drivers/gpu/drm/drm_mode_config.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
Signature · 1 line
2.55.0
#1 lore raw

PATCH 1/1 drm: use drm_warn() in validate_blend_mode_for_alpha_formats()

Commit 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed") introduced a WARN() to let driver developers know that a previously valid behavior should now be changed.

But WARN() should not be used for that, as it's a kernel warning report mechanism for conditions that are not expected to happen. It also produces a stack trace. Instead, a simple warning-level log message should have been used, as drivers were expected to trigger the condition.

This is causing problems for fuzzers, as they may stop when encountering a "BUG:" or "WARNING:" in the logs.

Replace WARN() with drm_warn() in this function, avoiding these issues.

---
 drivers/gpu/drm/drm_mode_config.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/drm_mode_config.c +3 -2
index 3bcc7bf0900c..3d9c554e08a8 100644
@@ -869,8 +869,9 @@ static void validate_blend_mode_for_alpha_formats(struct drm_plane *plane)
 	for (i = 0; i < plane->format_count; i++) {
 		fmt = drm_format_info(plane->format_types[i]);
 		if (fmt->has_alpha) {
-			WARN(1, "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
-			     plane->base.id, plane->name);
+			drm_warn(plane->dev,
+				 "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
+				 plane->base.id, plane->name);
 			break;
 		}
 	}
Signature · 1 line
2.55.0
#2 lore raw
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] Unconditional dereference of drm_format_info() return value leads to NULL pointer dereference for unrecognized formats. --

--- Patch [1]: [PATCH 1/1] drm: use drm_warn() in validate_blend_mode_for_alpha_formats() --- commit 8bcc8b572cfd0496af9f1e8dd202524f50670fa0
Author: Leandro Ribeiro <[email protected]>

drm: use drm_warn() in validate_blend_mode_for_alpha_formats()

This commit replaces WARN() with drm_warn() in validate_blend_mode_for_alpha_formats() to prevent fuzzers from tripping over kernel warning reports for expected driver validation issues.

drivers/gpu/drm/drm_mode_config.c static void validate_blend_mode_for_alpha_formats(struct drm_plane *plane) · line 871

Quoted 8 lines
drivers/gpu/drm/drm_mode_config.c +0 -0
index 3bcc7bf0900c4..3d9c554e08a87 100644
@@ -869,8 +869,9 @@ static void validate_blend_mode_for_alpha_formats(struct drm_plane *plane)
 	for (i = 0; i < plane->format_count; i++) {
 		fmt = drm_format_info(plane->format_types[i]);
 		if (fmt->has_alpha) {

[Severity: High]
This is a pre-existing issue, but could this unconditionally dereference fmt if drm_format_info() returns NULL?

If a driver exposes an unrecognized or custom pixel format that the core DRM does not know about, drm_format_info() returns NULL. Since the purpose of this validation function is to catch driver bugs, should we check if fmt is NULL before checking fmt->has_alpha to avoid a kernel panic?

Quoted 8 lines
drivers/gpu/drm/drm_mode_config.c +3 -2
-			WARN(1, "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
-			     plane->base.id, plane->name);
+			drm_warn(plane->dev,
+				 "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
+				 plane->base.id, plane->name);
 			break;
 		}
 	}
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
#3 lore raw
In reply to Leandro Ribeiro
Quoted 10 lines · Leandro Ribeiro

Commit 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed") introduced a WARN() to let driver developers know that a previously valid behavior should now be changed.

But WARN() should not be used for that, as it's a kernel warning report mechanism for conditions that are not expected to happen. It also produces a stack trace. Instead, a simple warning-level log message should have been used, as drivers were expected to trigger the condition.

Quoted 4 lines

This is causing problems for fuzzers, as they may stop when encountering a "BUG:" or "WARNING:" in the logs.

Replace WARN() with drm_warn() in this function, avoiding these issues.

Thank you. But could something like

pr_warn("[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup. Please fix.", plane->base.id, plane->name); dump_stack();

be less prone to get oversighted (because of printing about 10+ lines)?

#4 lore raw
Quoted 14 lines · Tetsuo Handa
Quoted 13 lines · Leandro Ribeiro

Commit 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed") introduced a WARN() to let driver developers know that a previously valid behavior should now be changed.

But WARN() should not be used for that, as it's a kernel warning report mechanism for conditions that are not expected to happen. It also produces a stack trace. Instead, a simple warning-level log message should have been used, as drivers were expected to trigger the condition.

This is causing problems for fuzzers, as they may stop when encountering a "BUG:" or "WARNING:" in the logs.

Quoted 9 lines
Quoted 1 line

Replace WARN() with drm_warn() in this function, avoiding these issues.

Thank you. But could something like

pr_warn("[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup. Please fix.", plane->base.id, plane->name); dump_stack();

be less prone to get oversighted (because of printing about 10+ lines)?

I see the point about making the message harder to overlook, but honestly I feel like the stack trace would be overkill here.

The error message is more important than the stack trace itself in this case, and I think the warning should be enough to draw attention to the issue.

Signature · 1 line
Leandro Ribeiro
#5 lore raw
Quoted 6 lines · Leandro Ribeiro

I see the point about making the message harder to overlook, but honestly I feel like the stack trace would be overkill here.

The error message is more important than the stack trace itself in this case, and I think the warning should be enough to draw attention to the issue.

OK. Then, please consider

drm_warn(plane->dev,
- "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
+ "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup. Please fix.,
plane->base.id, plane->name);

to indicate that developers have something to do.

#6 lore raw
Quoted 15 lines · Tetsuo Handa
Quoted 6 lines · Leandro Ribeiro

I see the point about making the message harder to overlook, but honestly I feel like the stack trace would be overkill here.

The error message is more important than the stack trace itself in this case, and I think the warning should be enough to draw attention to the issue.

OK. Then, please consider

(quoted diff) +1 -1
 			drm_warn(plane->dev,
-				 "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
+				 "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup. Please fix.,
 				 plane->base.id, plane->name);
Quoted 2 lines

to indicate that developers have something to do.

Sure, sounds good! I'll change this, thank you.

Signature · 1 line
Leandro Ribeiro
#7 lore raw
In reply to Leandro Ribeiro
Quoted 3 lines · Leandro Ribeiro

Commit 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed") introduced a WARN() to let driver developers know that a previously valid behavior should now be changed.

I should've chimed in way back when, but in retrospect it's a bit presumptuous to add a commit that puts the burden of the change on the driver developers, without so much as identifying which drivers or getting an ack from them.

This is really not unlike the __deprecated attribute, see commit 771c035372a0 ("deprecate the '__deprecated' attribute warnings entirely and for good"), but this one's runtime not build.

Quoted 10 lines

But WARN() should not be used for that, as it's a kernel warning report mechanism for conditions that are not expected to happen. It also produces a stack trace. Instead, a simple warning-level log message should have been used, as drivers were expected to trigger the condition.

This is causing problems for fuzzers, as they may stop when encountering a "BUG:" or "WARNING:" in the logs.

Replace WARN() with drm_warn() in this function, avoiding these issues.

Fixes 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed")
Quoted 21 lines
---
 drivers/gpu/drm/drm_mode_config.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/drm_mode_config.c +3 -2
index 3bcc7bf0900c..3d9c554e08a8 100644
@@ -869,8 +869,9 @@ static void validate_blend_mode_for_alpha_formats(struct drm_plane *plane)
 	for (i = 0; i < plane->format_count; i++) {
 		fmt = drm_format_info(plane->format_types[i]);
 		if (fmt->has_alpha) {
-			WARN(1, "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
-			     plane->base.id, plane->name);
+			drm_warn(plane->dev,
+				 "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
+				 plane->base.id, plane->name);
 			break;
 		}
 	}
Signature · 1 line
Jani Nikula, Intel
#8 lore raw
Quoted 5 lines · Jani Nikula
Quoted 3 lines · Leandro Ribeiro

Commit 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed") introduced a WARN() to let driver developers know that a previously valid behavior should now be changed.

Quoted 9 lines

I should've chimed in way back when, but in retrospect it's a bit presumptuous to add a commit that puts the burden of the change on the driver developers, without so much as identifying which drivers or getting an ack from them.

This is really not unlike the __deprecated attribute, see commit 771c035372a0 ("deprecate the '__deprecated' attribute warnings entirely and for good"), but this one's runtime not build.

I agree. I hadn't considered the driver coordination aspect when proposing that change.

Quoted 35 lines
Quoted 10 lines

But WARN() should not be used for that, as it's a kernel warning report mechanism for conditions that are not expected to happen. It also produces a stack trace. Instead, a simple warning-level log message should have been used, as drivers were expected to trigger the condition.

This is causing problems for fuzzers, as they may stop when encountering a "BUG:" or "WARNING:" in the logs.

Replace WARN() with drm_warn() in this function, avoiding these issues.

Fixes 860e748bddcc ("drm: ensure blend mode supported if pixel format with alpha exposed")
Quoted 21 lines
---
 drivers/gpu/drm/drm_mode_config.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/gpu/drm/drm_mode_config.c +3 -2
index 3bcc7bf0900c..3d9c554e08a8 100644
@@ -869,8 +869,9 @@ static void validate_blend_mode_for_alpha_formats(struct drm_plane *plane)
 	for (i = 0; i < plane->format_count; i++) {
 		fmt = drm_format_info(plane->format_types[i]);
 		if (fmt->has_alpha) {
-			WARN(1, "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
-			     plane->base.id, plane->name);
+			drm_warn(plane->dev,
+				 "[PLANE:%d:%s] pixel format with alpha exposed but blend mode not setup",
+				 plane->base.id, plane->name);
 			break;
 		}
 	}
Signature · 1 line
Leandro Ribeiro