| Message ID | 20200130034938.158504-1-gshan@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drivers/firmware/psci: Fix memory leak in alloc_init_cpu_groups() | expand |
On 1/30/20 2:49 PM, Gavin Shan wrote: > The CPU mask (@tmp) should be free'd on failing to allocating the element > of @cpu_groups[]. Otherwise, it leads to memory leakage because the CPU > mask variable is allocated with CONFIG_CPUMASK_OFFSTACK. > > Signed-off-by: Gavin Shan <gshan@redhat.com> > --- > drivers/firmware/psci/psci_checker.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/firmware/psci/psci_checker.c b/drivers/firmware/psci/psci_checker.c > index 6a445397771c..d1d6d1135fed 100644 > --- a/drivers/firmware/psci/psci_checker.c > +++ b/drivers/firmware/psci/psci_checker.c > @@ -167,6 +167,7 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) > topology_core_cpumask(cpumask_any(tmp)); > > if (!alloc_cpumask_var(&cpu_groups[num_groups], GFP_KERNEL)) { > + free_cpumask_var(tmp); > free_cpu_groups(num_groups, &cpu_groups); > return -ENOMEM; > } > I think @tmp has to be free'd either when failing to allocate @cpu_groups. However, I'm holding to post v2 until feedback is received on v1. diff --git a/drivers/firmware/psci/psci_checker.c b/drivers/firmware/psci/psci_checker.c index d1d6d1135fed..03eb798ad3ed 100644 --- a/drivers/firmware/psci/psci_checker.c +++ b/drivers/firmware/psci/psci_checker.c @@ -157,8 +157,10 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) cpu_groups = kcalloc(nb_available_cpus, sizeof(cpu_groups), GFP_KERNEL); - if (!cpu_groups) + if (!cpu_groups) { + free_cpumask_var(tmp); return -ENOMEM; + } cpumask_copy(tmp, cpu_online_mask); Thanks, Gavin
On Thu, Jan 30, 2020 at 02:49:38PM +1100, Gavin Shan wrote: > The CPU mask (@tmp) should be free'd on failing to allocating the element > of @cpu_groups[]. Otherwise, it leads to memory leakage because the CPU > mask variable is allocated with CONFIG_CPUMASK_OFFSTACK. > > Signed-off-by: Gavin Shan <gshan@redhat.com> > --- > drivers/firmware/psci/psci_checker.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/firmware/psci/psci_checker.c b/drivers/firmware/psci/psci_checker.c > index 6a445397771c..d1d6d1135fed 100644 > --- a/drivers/firmware/psci/psci_checker.c > +++ b/drivers/firmware/psci/psci_checker.c > @@ -167,6 +167,7 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) > topology_core_cpumask(cpumask_any(tmp)); > > if (!alloc_cpumask_var(&cpu_groups[num_groups], GFP_KERNEL)) { > + free_cpumask_var(tmp); You may need also below diff, right ? -->8 diff --git i/drivers/firmware/psci/psci_checker.c w/drivers/firmware/psci/psci_checker.c index 6a445397771c..c5a5c76a9fda 100644 --- i/drivers/firmware/psci/psci_checker.c +++ w/drivers/firmware/psci/psci_checker.c @@ -157,8 +157,10 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) cpu_groups = kcalloc(nb_available_cpus, sizeof(cpu_groups), GFP_KERNEL); - if (!cpu_groups) + if (!cpu_groups) { + free_cpumask_var(tmp); return -ENOMEM; + } cpumask_copy(tmp, cpu_online_mask);
On Thu, Jan 30, 2020 at 03:00:13PM +1100, Gavin Shan wrote: > On 1/30/20 2:49 PM, Gavin Shan wrote: > > The CPU mask (@tmp) should be free'd on failing to allocating the element > > of @cpu_groups[]. Otherwise, it leads to memory leakage because the CPU > > mask variable is allocated with CONFIG_CPUMASK_OFFSTACK. > > > > Signed-off-by: Gavin Shan <gshan@redhat.com> > > --- > > drivers/firmware/psci/psci_checker.c | 1 + > > 1 file changed, 1 insertion(+) > > > > diff --git a/drivers/firmware/psci/psci_checker.c b/drivers/firmware/psci/psci_checker.c > > index 6a445397771c..d1d6d1135fed 100644 > > --- a/drivers/firmware/psci/psci_checker.c > > +++ b/drivers/firmware/psci/psci_checker.c > > @@ -167,6 +167,7 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) > > topology_core_cpumask(cpumask_any(tmp)); > > if (!alloc_cpumask_var(&cpu_groups[num_groups], GFP_KERNEL)) { > > + free_cpumask_var(tmp); > > free_cpu_groups(num_groups, &cpu_groups); > > return -ENOMEM; > > } > > > > I think @tmp has to be free'd either when failing to allocate @cpu_groups. > However, I'm holding to post v2 until feedback is received on v1. > Ah sorry, problem with sequential mail reading, ignore my reply as you have already figured this out. -- Regards, Sudeep
On 1/30/20 10:05 PM, Sudeep Holla wrote: > On Thu, Jan 30, 2020 at 03:00:13PM +1100, Gavin Shan wrote: >> On 1/30/20 2:49 PM, Gavin Shan wrote: >>> The CPU mask (@tmp) should be free'd on failing to allocating the element >>> of @cpu_groups[]. Otherwise, it leads to memory leakage because the CPU >>> mask variable is allocated with CONFIG_CPUMASK_OFFSTACK. >>> >>> Signed-off-by: Gavin Shan <gshan@redhat.com> >>> --- >>> drivers/firmware/psci/psci_checker.c | 1 + >>> 1 file changed, 1 insertion(+) >>> >>> diff --git a/drivers/firmware/psci/psci_checker.c b/drivers/firmware/psci/psci_checker.c >>> index 6a445397771c..d1d6d1135fed 100644 >>> --- a/drivers/firmware/psci/psci_checker.c >>> +++ b/drivers/firmware/psci/psci_checker.c >>> @@ -167,6 +167,7 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) >>> topology_core_cpumask(cpumask_any(tmp)); >>> if (!alloc_cpumask_var(&cpu_groups[num_groups], GFP_KERNEL)) { >>> + free_cpumask_var(tmp); >>> free_cpu_groups(num_groups, &cpu_groups); >>> return -ENOMEM; >>> } >>> >> >> I think @tmp has to be free'd either when failing to allocate @cpu_groups. >> However, I'm holding to post v2 until feedback is received on v1. >> > > Ah sorry, problem with sequential mail reading, ignore my reply as you have > already figured this out. > Sudeep, no problem. I'll fold the additional changes into v2. Thanks for your time on this. Thanks, Gavin
diff --git a/drivers/firmware/psci/psci_checker.c b/drivers/firmware/psci/psci_checker.c index 6a445397771c..d1d6d1135fed 100644 --- a/drivers/firmware/psci/psci_checker.c +++ b/drivers/firmware/psci/psci_checker.c @@ -167,6 +167,7 @@ static int alloc_init_cpu_groups(cpumask_var_t **pcpu_groups) topology_core_cpumask(cpumask_any(tmp)); if (!alloc_cpumask_var(&cpu_groups[num_groups], GFP_KERNEL)) { + free_cpumask_var(tmp); free_cpu_groups(num_groups, &cpu_groups); return -ENOMEM; }
The CPU mask (@tmp) should be free'd on failing to allocating the element of @cpu_groups[]. Otherwise, it leads to memory leakage because the CPU mask variable is allocated with CONFIG_CPUMASK_OFFSTACK. Signed-off-by: Gavin Shan <gshan@redhat.com> --- drivers/firmware/psci/psci_checker.c | 1 + 1 file changed, 1 insertion(+)