[PATCH v2] drm/i915: use 'depends on' with visible DEBUG_OBJECTS for DRM_I915_DEBUG and DRM_I915_SW_FENCE_DEBUG_OBJECTS

[PATCH v2] drm/i915: use 'depends on' with visible DEBUG_OBJECTS for DRM_I915_DEBUG and DRM_I915_SW_FENCE_DEBUG_OBJECTS

[Julian Braha]

A previous version of this patch[1] accidentally introduced an unmet
dependency bug, because DRM_I915_DEBUG selects
DRM_I915_SW_FENCE_DEBUG_OBJECTS without also depending on DEBUG_OBJECTS.

DRM_I915_SW_FENCE_DEBUG_OBJECTS currently selects DEBUG_OBJECTS even though
DEBUG_OBJECTS is visible to users. Other config options use 'depends on'
for DEBUG_OBJECTS, so let's do the same here.

This select-visible Kconfig misusage was detected by Kconfirm, a static
analysis tool for Kconfig.

Link 1: https://lore.kernel.org/all/20260502191932.4491-1-julianbraha@gmail.com/

Signed-off-by: Julian Braha <julianbraha@gmail.com>
---
v2:
- select PREEMPT_COUNT, avoid recursive dependency

 drivers/gpu/drm/i915/Kconfig.debug | 3 ++-
 lib/Kconfig.debug                  | 2 +-
 2 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/Kconfig.debug b/drivers/gpu/drm/i915/Kconfig.debug
index 3562a02ef7ad..657cdf49181e 100644
--- a/drivers/gpu/drm/i915/Kconfig.debug
+++ b/drivers/gpu/drm/i915/Kconfig.debug
@@ -35,6 +35,7 @@ config DRM_I915_REPLAY_GPU_HANGS_API
 
 config DRM_I915_DEBUG
 	bool "Enable additional driver debugging"
+	depends on DEBUG_OBJECTS
 	depends on DRM_I915
 	depends on EXPERT # only for developers
 	depends on !COMPILE_TEST # never built by robots
@@ -153,7 +154,7 @@ config DRM_I915_TRACE_GTT
 config DRM_I915_SW_FENCE_DEBUG_OBJECTS
 	bool "Enable additional driver debugging for fence objects"
 	depends on DRM_I915
-	select DEBUG_OBJECTS
+	depends on DEBUG_OBJECTS
 	default n
 	help
 	  Choose this option to turn on extra driver debugging that may affect
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8ff5adcfe1e0..6501b7040706 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -760,7 +760,7 @@ source "mm/Kconfig.debug"
 
 config DEBUG_OBJECTS
 	bool "Debug object operations"
-	depends on PREEMPT_COUNT || !DEFERRED_STRUCT_PAGE_INIT
+	select PREEMPT_COUNT if !DEFERRED_STRUCT_PAGE_INIT
 	depends on DEBUG_KERNEL
 	help
 	  If you say Y here, additional code will be inserted into the
-- 
2.54.0

Claude review: drm/i915: use 'depends on' with visible DEBUG_OBJECTS for DRM_I915_DEBUG and DRM_I915_SW_FENCE_DEBUG_OBJECTS

[Claude Code Review Bot]

Overall Series Review

Subject: drm/i915: use 'depends on' with visible DEBUG_OBJECTS for DRM_I915_DEBUG and DRM_I915_SW_FENCE_DEBUG_OBJECTS
Author: Julian Braha <julianbraha@gmail.com>
Patches: 1
Reviewed: 2026-05-25T17:35:09.900931

---

This is a single-patch series that attempts to fix a Kconfig "select-visible" anti-pattern: `DRM_I915_SW_FENCE_DEBUG_OBJECTS` was using `select DEBUG_OBJECTS` on a user-visible symbol, which is considered Kconfig misuse. The i915-side changes (converting `select` to `depends on`) are correct, but **the patch introduces a bug in `lib/Kconfig.debug`** by inverting the condition on the `select PREEMPT_COUNT` replacement. Additionally, the `lib/Kconfig.debug` change modifies core kernel debug infrastructure and should have broader review beyond i915/DRM maintainers.

**Recommendation: Needs revision.** The `lib/Kconfig.debug` change has an inverted condition that breaks the original constraint, and the change to core kernel config should be split out and reviewed by the debug infrastructure maintainers.

---

---
Generated by Claude Code Patch Reviewer

Claude review: drm/i915: use 'depends on' with visible DEBUG_OBJECTS for DRM_I915_DEBUG and DRM_I915_SW_FENCE_DEBUG_OBJECTS

[Claude Code Review Bot]

Patch Review

**i915 Kconfig changes — Correct**

The two changes in `drivers/gpu/drm/i915/Kconfig.debug` are logically sound:

1. Changing `DRM_I915_SW_FENCE_DEBUG_OBJECTS` from `select DEBUG_OBJECTS` to `depends on DEBUG_OBJECTS` (line 134→157) is the right fix for the select-visible anti-pattern.

2. Adding `depends on DEBUG_OBJECTS` to `DRM_I915_DEBUG` (line 125→38) is necessary because `DRM_I915_DEBUG` selects `DRM_I915_SW_FENCE_DEBUG_OBJECTS` (line 63), which now depends on `DEBUG_OBJECTS`. Without this, Kconfig would select a symbol whose dependency isn't met.

These two changes are consistent: when `DRM_I915_DEBUG=y`, `DEBUG_OBJECTS` must be on (new depends), so the selected `DRM_I915_SW_FENCE_DEBUG_OBJECTS` has its `DEBUG_OBJECTS` dependency satisfied.

**lib/Kconfig.debug change — Bug: inverted condition**

The change to `DEBUG_OBJECTS` in `lib/Kconfig.debug` has the condition backwards:

```
-	depends on PREEMPT_COUNT || !DEFERRED_STRUCT_PAGE_INIT
+	select PREEMPT_COUNT if !DEFERRED_STRUCT_PAGE_INIT
```

The original `depends on` expression means: `DEBUG_OBJECTS` requires `PREEMPT_COUNT` **when `DEFERRED_STRUCT_PAGE_INIT=y`**. When `DEFERRED_STRUCT_PAGE_INIT=n`, `PREEMPT_COUNT` is not needed (the `!DEFERRED_STRUCT_PAGE_INIT` clause satisfies the dependency).

The replacement `select PREEMPT_COUNT if !DEFERRED_STRUCT_PAGE_INIT` does the opposite:
- When `DEFERRED_STRUCT_PAGE_INIT=n`: forces `PREEMPT_COUNT=y` — **unnecessary**, the original constraint was already satisfied without it.
- When `DEFERRED_STRUCT_PAGE_INIT=y`: does nothing — **the dangerous case is left unguarded**. `DEBUG_OBJECTS` can now be enabled with `PREEMPT_COUNT=n` and `DEFERRED_STRUCT_PAGE_INIT=y`, a configuration the original dependency explicitly prohibited.

The correct replacement would be:
```
select PREEMPT_COUNT if DEFERRED_STRUCT_PAGE_INIT
```

This ensures that when `DEFERRED_STRUCT_PAGE_INIT=y`, `PREEMPT_COUNT` is force-enabled, which is the constraint the original `depends on` was enforcing.

**Scope concern: modifying core kernel Kconfig from a driver patch**

The `lib/Kconfig.debug` change modifies `DEBUG_OBJECTS`, a core kernel debug option that affects far more than i915. The patch is CC'd only to i915/DRM maintainers and mailing lists. This change should be reviewed by the maintainers of `lib/Kconfig.debug` (likely Andrew Morton / the debug infrastructure maintainers). At minimum it needs a CC to `linux-kernel@vger.kernel.org` with relevant subsystem maintainers, and ideally should be a separate patch with its own justification rather than bundled into an i915 fix.

**Why the `select` approach is needed at all**

The commit message says "select PREEMPT_COUNT, avoid recursive dependency." The underlying issue is real: `PREEMPT_COUNT` is a prompt-less bool (in `kernel/Kconfig.preempt`), so it can only be enabled via `select`. With the original `depends on PREEMPT_COUNT` in `DEBUG_OBJECTS`, if nothing else in the config happens to select `PREEMPT_COUNT`, then `DEBUG_OBJECTS` becomes unavailable. Converting the dependency to `select` is a reasonable approach — but the condition must be correct.

**Minor: `default n` is unnecessary**

```
	depends on DEBUG_OBJECTS
	default n
```

In Kconfig, `bool` options default to `n` implicitly. The `default n` on `DRM_I915_SW_FENCE_DEBUG_OBJECTS` (line 158) is redundant. This is pre-existing and not introduced by this patch, but could be cleaned up opportunistically.

**Summary of required changes:**
1. Fix the inverted condition: `select PREEMPT_COUNT if DEFERRED_STRUCT_PAGE_INIT` (not `if !DEFERRED_STRUCT_PAGE_INIT`)
2. Consider splitting the `lib/Kconfig.debug` change into a separate patch with appropriate CC to debug infrastructure maintainers
3. Test the resulting config to verify `DEBUG_OBJECTS` remains selectable in all relevant configurations

---
Generated by Claude Code Patch Reviewer