memory: fixes: driver fixes for v7.4
Hi Arnd and SoC folks,
A few less and more important fixes. Sashiko as usually exaggerates and calls some of these "Critical" but exploitation likely requires already root privileges, e.g. through device unbind. Cannot rule out chaining vulnerabilities these days, thus sending these as fixes.
Best regards,
Krzysztof
The following changes since commit 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4:
Linux 7.4-rc1 (2026-11-01 13:26:31 -0800)
are available in the Git repository at:
https://git.kernel.org/pub/scm/linux/kernel/git/krzk/linux-mem-ctrl.git tags/memory-controller-drv-fixes-7.4
for you to fetch changes up to 1a509eca07176bc3363031c030d038ca9be9dfc0:
memory: fsl_ifc: Do not dereference the unset global pointer in probe (2026-11-11 16:22:36 +0100)
----------------------------------------------------------------
Memory controller drivers - fixes for v7.4
A set of fixes for more serious looking issues reported by Sashiko, although in practice these target only a very specific condition, like probe/removal racing:
1. Tegra210: During probe error paths or device removal, disable timers before teardown to avoid theoreticall possible use-after-free.
2. Tegra186: Fix OF node reference leak.
3. Freescale IFC:
- Fix possible probe error path use-after-free on unrequested NAND interrupt,
- Fix possible use-after-free of a globally exported device control structure after device unbind, if used by out of tree drivers (in-tree drivers do not use that export).
4. Atmel EBI: Fix leaked SMC clock reference with missing enable/disable balance on probe failure (e.g. deferred probe).
----------------------------------------------------------------
Krzysztof Kozlowski (2):
memory: atmel-ebi: Disable SMC clock on probe failure
memory: fsl_ifc: Clear fsl_ifc_ctrl_dev on probe failure and removal
Myeonghun Pak (1):
memory: tegra: Shut down Tegra210 EMC timers before teardown
Runyu Xiao (1):
memory: fsl_ifc: guard optional NAND IRQ cleanup
Sasha Levin (1):
memory: fsl_ifc: Do not dereference the unset global pointer in probe
Selby Mashiki (1):
memory: tegra: fix device_node leak in tegra186_mc_probe_device()
drivers/memory/atmel-ebi.c | 27 ++++++-- drivers/memory/fsl_ifc.c | 107 +++++++++++++++++++------------ drivers/memory/tegra/tegra186.c | 1 + drivers/memory/tegra/tegra210-emc-core.c | 16 +++-- 4 files changed, 97 insertions(+), 54 deletions(-)
Quoted 10 lines · Krzysztof Kozlowski
Hi Arnd and SoC folks,
A few less and more important fixes. Sashiko as usually exaggerates and calls some of these "Critical" but exploitation likely requires already root privileges, e.g. through device unbind. Cannot rule out chaining vulnerabilities these days, thus sending these as fixes.
Best regards,
Krzysztof
I forgot, there will be a conflict here with Linus' tree. Resolution from linux-next should be corrected a bit:
- fsl_ifc_ctrl_dev = devm_kzalloc_obj(&dev->dev, *fsl_ifc_ctrl_dev);
- if (!fsl_ifc_ctrl_dev)
- ctrl = devm_kzalloc(&dev->dev, sizeof(*ctrl), GFP_KERNEL);
++ ctrl = devm_kzalloc_obj(&dev->dev, *ctrl);
+ if (!ctrl)
return -ENOMEM;
Best regards,
Krzysztof