[PATCH AUTOSEL 6.16-5.4] scsi: bfa: Double-free fix

[Sasha Levin <sashal@kernel.org>]

From: jackysliu <1972843537@qq.com>

[ Upstream commit add4c4850363d7c1b72e8fce9ccb21fdd2cf5dc9 ]

When the bfad_im_probe() function fails during initialization, the memory
pointed to by bfad->im is freed without setting bfad->im to NULL.

Subsequently, during driver uninstallation, when the state machine enters
the bfad_sm_stopping state and calls the bfad_im_probe_undo() function,
it attempts to free the memory pointed to by bfad->im again, thereby
triggering a double-free vulnerability.

Set bfad->im to NULL if probing fails.

Signed-off-by: jackysliu <1972843537@qq.com>
Link: https://lore.kernel.org/r/tencent_3BB950D6D2D470976F55FC879206DE0B9A09@qq.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---

LLM Generated explanations, may be completely bogus:

**Backport Status: YES**

This commit is a clear candidate for backporting to stable kernel trees.
Here's my extensive analysis:

## Bug Analysis

The commit fixes a **critical double-free vulnerability** in the BFA
(Brocade Fibre Channel Adapter) SCSI driver. The issue occurs in the
following scenario:

1. During driver initialization, `bfad_im_probe()` allocates memory for
   `bfad->im` (line 700: `im = kzalloc(sizeof(struct bfad_im_s),
   GFP_KERNEL)`)
2. The pointer is assigned to `bfad->im` (line 704: `bfad->im = im`)
3. If `bfad_thread_workq()` fails (line 707), the function frees the
   memory (line 708: `kfree(im)`) but **fails to set `bfad->im = NULL`**
4. Later, during driver uninstallation, when the state machine enters
   `bfad_sm_stopping` state, it calls `bfad_im_probe_undo()` (line 714
   in bfad.c)
5. `bfad_im_probe_undo()` checks `if (bfad->im)` (line 720) and since
   the pointer wasn't nulled, it attempts to free the already-freed
   memory again (line 722: `kfree(bfad->im)`)

## Why This Should Be Backported

1. **Security Impact**: Double-free vulnerabilities can lead to memory
   corruption, system crashes, and potentially be exploited for
   privilege escalation. This is a real security issue.

2. **Simple Fix**: The fix is minimal - just a single line addition
   (`bfad->im = NULL;`) after the `kfree(im)` call. This follows the
   exact same pattern already used in `bfad_im_probe_undo()` where after
   freeing, the pointer is set to NULL (line 723).

3. **Clear Bug**: The bug is straightforward - the code inconsistently
   handles the cleanup path. The fix makes it consistent with the
   cleanup function's expectations.

4. **Limited Scope**: The change is confined to error handling in a
   single function (`bfad_im_probe()`). It doesn't modify any normal
   operation paths or introduce new features.

5. **Low Risk**: Setting a pointer to NULL after freeing is a standard
   defensive programming practice with essentially zero risk of
   regression.

6. **Affects Users**: Any system using Brocade Fibre Channel adapters
   could hit this issue if the driver initialization fails (e.g., due to
   memory allocation failure in `bfad_thread_workq()`).

7. **Follows Stable Rules**: The commit:
   - Fixes a real bug (double-free vulnerability)
   - Is small and contained (1 line change)
   - Has clear causality
   - Doesn't introduce new features
   - Has minimal risk of causing regressions

The commit follows the Linux stable kernel rules perfectly - it's a
small, obvious fix for a real bug that could affect users in production
environments.

 drivers/scsi/bfa/bfad_im.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/scsi/bfa/bfad_im.c b/drivers/scsi/bfa/bfad_im.c
index a719a18f0fbc..f56e008ee52b 100644
--- a/drivers/scsi/bfa/bfad_im.c
+++ b/drivers/scsi/bfa/bfad_im.c
@@ -706,6 +706,7 @@ bfad_im_probe(struct bfad_s *bfad)
 
 	if (bfad_thread_workq(bfad) != BFA_STATUS_OK) {
 		kfree(im);
+		bfad->im = NULL;
 		return BFA_STATUS_FAILED;
 	}
 
-- 
2.39.5