[PATCH v2 00/12] EDAC/versalnet: Fix error handling, teardown, and robustness
From: Shubhrajyoti Datta
Date: Wed Nov 18 2026 - 09:23:30 EST
This series addresses several correctness and robustness issues in
the VersalNet EDAC driver, including NULL pointer dereferences,
resource lifetime problems, shutdown handling, and probe/remove
path cleanup.
The main fixes are:
* Robustness fixes
- Add bounds validation in rpmsg_cb().
* Probe and initialization cleanup
- Move platform_set_drvdata() from init_one_mc() to mc_probe().
- Reorder probe flow to initialize MCDI before RPMsg registration.
- Remove unused pdev parameters from internal initialization
functions.
* Resource management fixes
- Fix error handling in init_one_mc() to avoid a potential
double-free by reordering device_register() and edac_mc_alloc().
- Replace sprintf() with init_name by dev_set_name() to avoid
use-after-free issues caused by deferred kobject cleanup.
* Code cleanup
- Use a designated initializer for rpmsg_channel_info.
Changes in v2:
Add the codeveloped tag and remove the non standard signature
fix the name stack warning
Use a macro for channel name
Add reviewed by tag
Sashiko comment
"This is a pre-existing issue, but are offset and length validated before
being used as indices?"
New patch to fix shashiko review comment "this is a pre-existing
issue, but is the per-controller width overwriting the shared
state for all controllers?".
Patch addition for the rpmsg endpoint race
Update the labels
Address the sashiko comment
"register_rpmsg_driver() exposes the callback endpoint before setup_mcdi()
initializes the mcdi pointer. Could this lead to a NULL pointer dereference
if a response arrives early?"
Patch addition for the rpmsg drvdata race
Add reviewed-by tag
Add amd,num-ctrl property to device tree
The property was requested in the discussion below
https://lore.kernel.org/all/34ffba5a-8e9a-49b9-9e36-c31d0e4369f8@xxxxxxx/
Adding the device tree parameter was suggested in the review in v1
https://lore.kernel.org/all/34ffba5a-8e9a-49b9-9e36-c31d0e4369f8@xxxxxxx/
New patch added to bailout if the cdx_mcdi_rpc fails.
Hayden Rinn (1):
EDAC/versalnet: Wait for RPMsg endpoint initialization
Prasanna Kumar T S M (1):
EDAC/versalnet: Fix device_register() error handling in init_one_mc()
Shubhrajyoti Datta (10):
EDAC/versalnet: Use designated initializer for rpmsg_channel_info
EDAC/versalnet: Fix use-after-free in remove_one_mc()
EDAC/versalnet: Add bounds validation in rpmsg_cb()
EDAC/versalnet: Make dwidth per-controller
EDAC/versalnet: Initialize MCDI before RPMsg registration
EDAC/versalnet: Guard against NULL drvdata in rpmsg_cb()
EDAC/versalnet: Move platform_set_drvdata() to mc_probe()
dt-bindings: memory: xlnx,versal-net-ddrmc5: Add amd,num-ctrl property
EDAC/versalnet: use device tree property for number of controllers
EDAC/versalnet: Make get_ddr_config() return errors
.../xlnx,versal-net-ddrmc5.yaml | 9 +
drivers/edac/versalnet_edac.c | 283 +++++++++++++-----
2 files changed, 220 insertions(+), 72 deletions(-)
--
2.44.4