* Re: [PATCH] Fix a memory leak reported by kmemleak.
2011-06-21 11:08 [PATCH] Fix a memory leak reported by kmemleak Bernd Schubert
@ 2011-06-21 12:46 ` Catalin Marinas
0 siblings, 0 replies; 2+ messages in thread
From: Catalin Marinas @ 2011-06-21 12:46 UTC (permalink / raw)
To: Bernd Schubert; +Cc: x86, linux-kernel, Bjorn Helgaas
On Tue, Jun 21, 2011 at 12:08:53PM +0100, Bernd Schubert wrote:
> Memory allocated by get_current_resources() will be never free'ed
> if 'pci_use_crs' is not set.
>
> Signed-off-by: Bernd Schubert <bernd.schubert@itwm.fraunhofer.de>
> ---
> arch/x86/pci/acpi.c | 5 ++++-
> 1 files changed, 4 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/pci/acpi.c b/arch/x86/pci/acpi.c
> index 0972315..3ad040a 100644
> --- a/arch/x86/pci/acpi.c
> +++ b/arch/x86/pci/acpi.c
> @@ -228,8 +228,11 @@ static void add_resources(struct pci_root_info *info)
> int i;
> struct resource *res, *root, *conflict;
>
> - if (!pci_use_crs)
> + if (!pci_use_crs) {
> + kfree(info->res);
> + kfree(info->name);
> return;
> + }
>
> coalesce_windows(info, IORESOURCE_MEM);
> coalesce_windows(info, IORESOURCE_IO);
Thanks for looking into this. I cc'ed Bjorn as he seems to be the last
one changing this file. I'm not familiar with part of the kernel, so not
sure whether info->res or info->name could be used somewhere else under
certain conditions.
An alternative patch (just to keep allocation/freeing in the same
function):
diff --git a/arch/x86/pci/acpi.c b/arch/x86/pci/acpi.c
index 0972315..684cf82 100644
--- a/arch/x86/pci/acpi.c
+++ b/arch/x86/pci/acpi.c
@@ -228,9 +228,6 @@ static void add_resources(struct pci_root_info *info)
int i;
struct resource *res, *root, *conflict;
- if (!pci_use_crs)
- return;
-
coalesce_windows(info, IORESOURCE_MEM);
coalesce_windows(info, IORESOURCE_IO);
@@ -286,8 +283,11 @@ get_current_resources(struct acpi_device *device, int busnum,
acpi_walk_resources(device->handle, METHOD_NAME__CRS, setup_resource,
&info);
- add_resources(&info);
- return;
+ if (pci_use_crs) {
+ add_resources(&info);
+ return;
+ }
+ kfree(info.name);
name_alloc_fail:
kfree(info.res);
--
Catalin
^ permalink raw reply [flat|nested] 2+ messages in thread