cpufreq: imx6q: fix devres accumulation across driver rebind
Subsystems (2)
PATCH 1/1 cpufreq: imx6q: fix devres accumulation across driver rebind
imx6_soc_volt is allocated with devm_kcalloc(cpu_dev, ...), where cpu_dev is the CPU device from get_cpu_device(0). That device is never unbound, so its devres list is never released, and imx6q_cpufreq_remove() does not free the array either. Every probe therefore adds an allocation that stays for the lifetime of the system.
Allocate against the platform device instead. Its devres is released when the driver is unbound, which is exactly the lifetime the array wants:imx6q_set_target() reads it, and nothing may reach that after cpufreq_unregister_driver().
That makes the array actually go away on unbind, so also clear the file-scope pointer in remove and on the failed-probe path, rather than leave it pointing at memory devres is about to release.
Tested by rebinding the driver on qemu's mcimx6ul-evk.
--- drivers/cpufreq/imx6q-cpufreq.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/cpufreq/imx6q-cpufreq.c +3 -1
@@ -400,7 +400,7 @@ static int imx6q_cpufreq_probe(struct platform_device *pdev) }/* Make imx6_soc_volt array's size same as arm opp number */ - imx6_soc_volt =devm_kcalloc (cpu_dev, num,sizeof (*imx6_soc_volt), + imx6_soc_volt =devm_kcalloc (&pdev->dev, num,sizeof (*imx6_soc_volt),GFP_KERNEL );if (imx6_soc_volt ==NULL ) { ret = -ENOMEM ;
@@ -485,6 +485,7 @@ static int imx6q_cpufreq_probe(struct platform_device *pdev)return 0 ; free_freq_table: + imx6_soc_volt =NULL ;dev_pm_opp_free_cpufreq_table (cpu_dev, &freq_table); out_free_opp:dev_pm_opp_of_remove_table (cpu_dev);
@@ -506,6 +507,7 @@ static int imx6q_cpufreq_probe(struct platform_device *pdev)static void imx6q_cpufreq_remove (struct platform_device *pdev) {cpufreq_unregister_driver (&imx6q_cpufreq_driver); + imx6_soc_volt =NULL ;dev_pm_opp_free_cpufreq_table (cpu_dev, &freq_table);dev_pm_opp_of_remove_table (cpu_dev);regulator_put (arm_reg);
Signature · 1 line
2.53.0
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
Pre-existing issues:
- [High] The file-scope static variable soc_opp_count is never reset to 0 across driver unbind/rebind or deferred probe cycles, leading to a severe out-of-bounds write array overflow. --
--- Patch [1]: [PATCH] cpufreq: imx6q: fix devres accumulation across driver rebind --- commit d8bcb5a2ed1157b0b23fd4c657a9fd392c54309b
Author: Karl Mehltretter <[email protected]>
cpufreq: imx6q: fix devres accumulation across driver rebind
This commit fixes a memory leak caused by devres accumulation during driver unbind and rebind cycles. It changes the memory allocation of imx6_soc_volt to tie its lifetime to the platform device, and clears the file-scope pointer in the remove and probe-failure paths.
Quoted 4 lines
drivers/cpufreq/imx6q-cpufreq.c +0
-0
[ ... ]
drivers/cpufreq/imx6q-cpufreq.c static int imx6q_cpufreq_probe(struct platform_device *pdev) · line 510
Quoted 5 lines
drivers/cpufreq/imx6q-cpufreq.c +1
-0@@ -506,6 +507,7 @@ static int imx6q_cpufreq_probe(struct platform_device *pdev)static void imx6q_cpufreq_remove (struct platform_device *pdev) {cpufreq_unregister_driver (&imx6q_cpufreq_driver); + imx6_soc_volt =NULL ;
[Severity: High]
This is a pre-existing issue, but since this patch explicitly tests and fixes rebind behavior, should soc_opp_count also be reset to 0 here?
The file-scope static variable soc_opp_count is never reset to 0 across driver unbind/rebind or deferred probe cycles.
In imx6q_cpufreq_probe(), the loop increments soc_opp_count:
for (j = 0; j < num; j++) {
val = prop->value;
for (i = 0; i < nr / 2; i++) {
unsigned long freq = be32_to_cpup(val++);
unsigned long volt = be32_to_cpup(val++);
if (freq_table[j].frequency == freq) {
imx6_soc_volt[soc_opp_count++] = volt;
break;
}
}
}On subsequent probes (e.g. after the unbind/rebind cycle you tested), it resumes counting from its previous value instead of 0. Since imx6_soc_volt is newly allocated with size 'num', this will write beyond the bounds of the array, causing memory corruption.
Does the driver need to reset soc_opp_count to 0 during removal and on the error paths in probe?
Quoted 3 lines
drivers/cpufreq/imx6q-cpufreq.c +0
-0dev_pm_opp_free_cpufreq_table (cpu_dev, &freq_table);dev_pm_opp_of_remove_table (cpu_dev);regulator_put (arm_reg);
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
Quoted 15 lines · Karl Mehltretter
imx6_soc_voltis allocated withdevm_kcalloc(cpu_dev, ...), wherecpu_devis the CPU device fromget_cpu_device(0). That device is never unbound, so its devres list is never released, andimx6q_cpufreq_remove()does not free the array either. Every probe therefore adds an allocation that stays for the lifetime of the system.Allocate against the platform device instead. Its devres is released when the driver is unbound, which is exactly the lifetime the array wants:
imx6q_set_target()reads it, and nothing may reach that aftercpufreq_unregister_driver().That makes the array actually go away on unbind, so also clear the file-scope pointer in remove and on the failed-probe path, rather than leave it pointing at memory devres is about to release.
Quoted 8 lines
Tested by rebinding the driver on qemu's mcimx6ul-evk.
Fixes b4573d1d657a ("cpufreq: imx6q: correct VDDSOC/PU voltage scaling when cpufreq is changed")Assisted-by Claude:claude-opus-5Signed-off-by Karl Mehltretter <[email protected]>--- drivers/cpufreq/imx6q-cpufreq.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
Applied. Thanks.
Signature · 1 line
viresh