[PATCH AUTOSEL 6.17-6.6] firmware: arm_scmi: Skip RAW initialization on failure
Sasha Levin
sashal at kernel.org
Tue Oct 28 05:48:07 PDT 2025
From: Cristian Marussi <cristian.marussi at arm.com>
[ Upstream commit 092b9e2ce6dd63d2f36822751a51957412706986 ]
Avoid attempting to initialize RAW mode when the debug subsystem itself
has failed to initialize, since doing so is pointless and emits
misleading error messages.
Signed-off-by: Cristian Marussi <cristian.marussi at arm.com>
Message-Id: <20251014115346.2391418-3-cristian.marussi at arm.com>
Signed-off-by: Sudeep Holla <sudeep.holla at arm.com>
Signed-off-by: Sasha Levin <sashal at kernel.org>
---
LLM Generated explanations, may be completely bogus:
Now I have enough information to make my determination. Let me create a
comprehensive analysis:
## BACKPORT ANALYSIS DECISION
**Backport Status: YES**
---
## COMPREHENSIVE ANALYSIS
### 1. CODE CHANGE ANALYSIS
The commit makes two related changes:
**Change 1** (drivers/firmware/arm_scmi/driver.c:3047-3048):
```c
- if (!info->dbg)
- return -EINVAL;
```
Removes the NULL check from inside `scmi_debugfs_raw_mode_setup()`.
**Change 2** (drivers/firmware/arm_scmi/driver.c:3221):
```c
- if (IS_ENABLED(CONFIG_ARM_SCMI_RAW_MODE_SUPPORT)) {
+ if (info->dbg &&
IS_ENABLED(CONFIG_ARM_SCMI_RAW_MODE_SUPPORT)) {
```
Adds guard condition to prevent calling the function when `info->dbg` is
NULL.
### 2. SEMANTIC ANALYSIS PERFORMED
**Tools Used:**
- `git log` and `git show` for commit history analysis
- `Grep` for code pattern searches across SCMI subsystem
- `Read` for examining driver code and Kconfig
- Task agent (Explore) for understanding SCMI ecosystem context
- Git blame and tag analysis for version tracking
**Key Findings from Analysis:**
**Impact Scope (from code examination):**
- Function `scmi_debugfs_raw_mode_setup()` is called from only ONE
location: `scmi_probe()` at line 3222
- Function accesses `info->dbg->top_dentry` at line 3068, which would be
a **NULL pointer dereference** if the removed check wasn't there
- The calling site now has proper guard, so function is never called
with NULL dbg
**Caller Analysis:**
- `scmi_probe()` is the driver probe function - called during device
initialization
- Prior commits in series (2290ab43b9d8e, 289ce7e9a5e1a) fixed NULL
handling in debug helpers
- This commit completes the debug failure handling by preventing
pointless RAW init attempt
**Dependency Analysis:**
- RAW mode requires debugfs (CONFIG dependency in Kconfig line 35)
- `scmi_raw_mode_init()` expects valid `top_dentry` parameter (line
3068)
- If debug setup fails, `info->dbg` is NULL, making RAW mode init
impossible
### 3. BUG IMPACT ANALYSIS
**Affected Kernel Versions:**
- Introduced: v6.3-rc1 (commit 7063887b53860, January 2023)
- Fixed: v6.18-rc3 (commit 092b9e2ce6dd6, October 2025)
- **Affected range: v6.3 through v6.17**
**Failure Scenario:**
```c
// Before fix - in scmi_probe():
if (scmi_top_dentry) {
info->dbg = scmi_debugfs_common_setup(info); // Returns NULL on
failure
if (!info->dbg)
dev_warn(dev, "Failed to setup SCMI debugfs.\n");
if (IS_ENABLED(CONFIG_ARM_SCMI_RAW_MODE_SUPPORT)) {
ret = scmi_debugfs_raw_mode_setup(info); // Called with NULL
dbg!
if (!coex) { // Default config
if (ret) // ret = -EINVAL
goto clear_dev_req_notifier; // PROBE FAILS!
return 0;
}
}
}
```
**Real-World Impact:**
- **Who is affected**: Development/testing systems with
CONFIG_ARM_SCMI_RAW_MODE_SUPPORT=y
- **When it triggers**: When debugfs initialization fails (no debugfs
support, memory failure, mount issues)
- **What breaks**: **Entire SCMI driver probe fails** → system may not
boot
- **Criticality**: SCMI manages power domains, clocks, DVFS, thermal
sensors - critical for ARM platforms
>From Explore agent findings:
- SCMI is used on STM32MP, ARM Juno, NXP i.MX95 platforms
- Controls essential services: CPU frequency, power states, system
shutdown/reboot
- Kconfig notes it "may be needed early in boot" for power control
**Severity Assessment:**
- **Medium-High for affected configurations**
- Not production issue (RAW mode is debug/testing feature per Kconfig
line 42)
- But complete driver failure is **unacceptable** for development
systems
- Misleading error messages make debugging difficult
### 4. ARCHITECTURAL & SIDE EFFECTS
**Architectural Impact:** None
- No data structure changes
- No API modifications
- No changes to core SCMI protocol handling
**Side Effects:** Positive only
- Prevents probe failure when debugfs fails
- Eliminates misleading error messages
- Improves graceful degradation
- No functional changes to success path
### 5. STABLE TREE COMPLIANCE
**Compliance Check:**
✅ **Bug fix**: Yes - fixes probe failure
✅ **Small and contained**: 2 simple changes in 1 file
✅ **Obviously correct**: Guard check before function call
✅ **No new features**: Pure bug fix
✅ **No architectural changes**: Minimal scope
✅ **Low regression risk**: Only affects error path
**Stable Tree Rules Alignment:**
- Fixes important bug (probe failure)
- Self-contained change
- Easy to review and verify
- Part of coordinated fix series (related commits already backported)
### 6. BACKPORT INDICATORS
**Evidence for backporting:**
- **Related commits already backported**: Commit 2290ab43b9d8e (with
Fixes tag) has been backported to multiple stable trees
(7056e61a28a83, 4c0f9a50f2e98, fae8405cfddb0, 58d3e2680bea9,
30c89140a4ddc, 75446183128d2)
- **Part of fix series**: This is patch 3/3 in a coordinated series
addressing debug initialization failures
- **Fixes merged in mainline**: Present in v6.18-rc3, merged via
arm/fixes tree (71a5970259c3f)
- **No Cc: stable tag**: But related patches were backported, suggesting
subsystem maintainers want the series in stable
**Evidence against backporting:**
- No explicit "Fixes:" tag (but commit message clearly states it fixes a
problem)
- No "Cc: stable at vger.kernel.org" tag
- Only affects debug/testing configuration
### 7. DEPENDENCY CHECK
**Depends on:**
- None - independent fix
- Works better with commits 2290ab43b9d8e and 289ce7e9a5e1a (already
backported)
**Required by:**
- None identified
---
## FINAL RECOMMENDATION
**YES - This commit SHOULD be backported to stable kernel trees.**
**Justification:**
1. **Fixes probe failure**: When debugfs fails + RAW mode enabled
(without coex), entire SCMI driver fails to load, breaking critical
platform services
2. **Completes backported series**: Related fixes (2290ab43b9d8e,
289ce7e9a5e1a) already backported to stable. This commit completes
the debug failure handling
3. **Low risk, high correctness**: Minimal change, obviously correct
logic (guard before function call)
4. **Stable tree compliant**: Small, contained, bug-fix-only change with
no architectural impact
5. **Affects supported feature**: CONFIG_ARM_SCMI_RAW_MODE_SUPPORT
exists in stable kernels since v6.3
6. **Graceful degradation**: Enables proper fallback behavior when debug
subsystem fails
**Target stable trees:** v6.3+ (where RAW mode was introduced)
**Priority:** Medium - affects development/testing systems, not
production, but fixes complete driver failure
drivers/firmware/arm_scmi/driver.c | 5 +----
1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/firmware/arm_scmi/driver.c b/drivers/firmware/arm_scmi/driver.c
index bd56a877fdfc8..85392228d1739 100644
--- a/drivers/firmware/arm_scmi/driver.c
+++ b/drivers/firmware/arm_scmi/driver.c
@@ -3044,9 +3044,6 @@ static int scmi_debugfs_raw_mode_setup(struct scmi_info *info)
u8 channels[SCMI_MAX_CHANNELS] = {};
DECLARE_BITMAP(protos, SCMI_MAX_CHANNELS) = {};
- if (!info->dbg)
- return -EINVAL;
-
/* Enumerate all channels to collect their ids */
idr_for_each_entry(&info->tx_idr, cinfo, id) {
/*
@@ -3218,7 +3215,7 @@ static int scmi_probe(struct platform_device *pdev)
if (!info->dbg)
dev_warn(dev, "Failed to setup SCMI debugfs.\n");
- if (IS_ENABLED(CONFIG_ARM_SCMI_RAW_MODE_SUPPORT)) {
+ if (info->dbg && IS_ENABLED(CONFIG_ARM_SCMI_RAW_MODE_SUPPORT)) {
ret = scmi_debugfs_raw_mode_setup(info);
if (!coex) {
if (ret)
--
2.51.0
More information about the linux-arm-kernel
mailing list