* [PATCH] acpi: Fix potential call to a freed memory section.
@ 2007-05-16 16:09 Aaron Durbin
2007-05-16 16:33 ` Andrew Morton
0 siblings, 1 reply; 3+ messages in thread
From: Aaron Durbin @ 2007-05-16 16:09 UTC (permalink / raw)
To: linux-kernel; +Cc: akpm, len.brown
Strip __cpuinit[data] from Node <-> PXM routines and supporting data structures.
Also make pxm_to_node_map and node_to_pxm_map local to the numa acpi module.
This fixes a bug triggered by the following conditions:
- boot on a machine with a SLIT table defined
- kernel is configured w/ CONFIG_HOTPLUG_CPU=n
- cat /sys/devices/system/node/node*/distance
This will cause an oops by calling into a freed memory section.
In particular, on x86_64, __node_distance calls node_to_pxm().
Signed-off-by: Aaron Durbin <adurbin@google.com>
---
diff --git a/drivers/acpi/numa.c b/drivers/acpi/numa.c
index 8fcd6a1..a2efae8 100644
--- a/drivers/acpi/numa.c
+++ b/drivers/acpi/numa.c
@@ -40,19 +40,19 @@ #define PXM_INVAL -1
#define NID_INVAL -1
/* maps to convert between proximity domain and logical node ID */
-int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS]
+static int pxm_to_node_map[MAX_PXM_DOMAINS]
= { [0 ... MAX_PXM_DOMAINS - 1] = NID_INVAL };
-int __cpuinitdata node_to_pxm_map[MAX_NUMNODES]
+static int node_to_pxm_map[MAX_NUMNODES]
= { [0 ... MAX_NUMNODES - 1] = PXM_INVAL };
-int __cpuinit pxm_to_node(int pxm)
+int pxm_to_node(int pxm)
{
if (pxm < 0)
return NID_INVAL;
return pxm_to_node_map[pxm];
}
-int __cpuinit node_to_pxm(int node)
+int node_to_pxm(int node)
{
if (node < 0)
return PXM_INVAL;
diff --git a/include/acpi/acpi_numa.h b/include/acpi/acpi_numa.h
index f9d2bde..b62cd36 100644
--- a/include/acpi/acpi_numa.h
+++ b/include/acpi/acpi_numa.h
@@ -11,11 +11,8 @@ #else
#define MAX_PXM_DOMAINS (256) /* Old pxm spec is defined 8 bit */
#endif
-extern int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS];
-extern int __cpuinitdata node_to_pxm_map[MAX_NUMNODES];
-
-extern int __cpuinit pxm_to_node(int);
-extern int __cpuinit node_to_pxm(int);
+extern int pxm_to_node(int);
+extern int node_to_pxm(int);
extern int __cpuinit acpi_map_pxm_to_node(int);
extern void __cpuinit acpi_unmap_pxm_to_node(int);
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH] acpi: Fix potential call to a freed memory section.
2007-05-16 16:09 [PATCH] acpi: Fix potential call to a freed memory section Aaron Durbin
@ 2007-05-16 16:33 ` Andrew Morton
2007-05-16 16:38 ` Aaron Durbin
0 siblings, 1 reply; 3+ messages in thread
From: Andrew Morton @ 2007-05-16 16:33 UTC (permalink / raw)
To: Aaron Durbin; +Cc: linux-kernel, len.brown
On Wed, 16 May 2007 09:09:11 -0700 Aaron Durbin <adurbin@google.com> wrote:
> Strip __cpuinit[data] from Node <-> PXM routines and supporting data structures.
> Also make pxm_to_node_map and node_to_pxm_map local to the numa acpi module.
>
> This fixes a bug triggered by the following conditions:
> - boot on a machine with a SLIT table defined
> - kernel is configured w/ CONFIG_HOTPLUG_CPU=n
> - cat /sys/devices/system/node/node*/distance
> This will cause an oops by calling into a freed memory section.
>
> In particular, on x86_64, __node_distance calls node_to_pxm().
>
Yeah, those section warnings we get aren't just noise.
I think this patch is also needed in 2.6.21.x, yes?
> ---
>
> diff --git a/drivers/acpi/numa.c b/drivers/acpi/numa.c
> index 8fcd6a1..a2efae8 100644
> --- a/drivers/acpi/numa.c
> +++ b/drivers/acpi/numa.c
> @@ -40,19 +40,19 @@ #define PXM_INVAL -1
> #define NID_INVAL -1
>
> /* maps to convert between proximity domain and logical node ID */
> -int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS]
> +static int pxm_to_node_map[MAX_PXM_DOMAINS]
> = { [0 ... MAX_PXM_DOMAINS - 1] = NID_INVAL };
> -int __cpuinitdata node_to_pxm_map[MAX_NUMNODES]
> +static int node_to_pxm_map[MAX_NUMNODES]
> = { [0 ... MAX_NUMNODES - 1] = PXM_INVAL };
>
> -int __cpuinit pxm_to_node(int pxm)
> +int pxm_to_node(int pxm)
> {
> if (pxm < 0)
> return NID_INVAL;
> return pxm_to_node_map[pxm];
> }
>
> -int __cpuinit node_to_pxm(int node)
> +int node_to_pxm(int node)
> {
> if (node < 0)
> return PXM_INVAL;
> diff --git a/include/acpi/acpi_numa.h b/include/acpi/acpi_numa.h
> index f9d2bde..b62cd36 100644
> --- a/include/acpi/acpi_numa.h
> +++ b/include/acpi/acpi_numa.h
> @@ -11,11 +11,8 @@ #else
> #define MAX_PXM_DOMAINS (256) /* Old pxm spec is defined 8 bit */
> #endif
>
> -extern int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS];
> -extern int __cpuinitdata node_to_pxm_map[MAX_NUMNODES];
> -
> -extern int __cpuinit pxm_to_node(int);
> -extern int __cpuinit node_to_pxm(int);
> +extern int pxm_to_node(int);
> +extern int node_to_pxm(int);
> extern int __cpuinit acpi_map_pxm_to_node(int);
> extern void __cpuinit acpi_unmap_pxm_to_node(int);
>
Moving from __fooinit into .text is always safe. Len, do you want me to
merge this up?
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] acpi: Fix potential call to a freed memory section.
2007-05-16 16:33 ` Andrew Morton
@ 2007-05-16 16:38 ` Aaron Durbin
0 siblings, 0 replies; 3+ messages in thread
From: Aaron Durbin @ 2007-05-16 16:38 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-kernel, len.brown
On 5/16/07, Andrew Morton <akpm@linux-foundation.org> wrote:
> On Wed, 16 May 2007 09:09:11 -0700 Aaron Durbin <adurbin@google.com> wrote:
>
> > Strip __cpuinit[data] from Node <-> PXM routines and supporting data structures.
> > Also make pxm_to_node_map and node_to_pxm_map local to the numa acpi module.
> >
> > This fixes a bug triggered by the following conditions:
> > - boot on a machine with a SLIT table defined
> > - kernel is configured w/ CONFIG_HOTPLUG_CPU=n
> > - cat /sys/devices/system/node/node*/distance
> > This will cause an oops by calling into a freed memory section.
> >
> > In particular, on x86_64, __node_distance calls node_to_pxm().
> >
>
> Yeah, those section warnings we get aren't just noise.
>
> I think this patch is also needed in 2.6.21.x, yes?
Yes, this bug exists in 2.6.21 as well.
>
> > ---
> >
> > diff --git a/drivers/acpi/numa.c b/drivers/acpi/numa.c
> > index 8fcd6a1..a2efae8 100644
> > --- a/drivers/acpi/numa.c
> > +++ b/drivers/acpi/numa.c
> > @@ -40,19 +40,19 @@ #define PXM_INVAL -1
> > #define NID_INVAL -1
> >
> > /* maps to convert between proximity domain and logical node ID */
> > -int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS]
> > +static int pxm_to_node_map[MAX_PXM_DOMAINS]
> > = { [0 ... MAX_PXM_DOMAINS - 1] = NID_INVAL };
> > -int __cpuinitdata node_to_pxm_map[MAX_NUMNODES]
> > +static int node_to_pxm_map[MAX_NUMNODES]
> > = { [0 ... MAX_NUMNODES - 1] = PXM_INVAL };
> >
> > -int __cpuinit pxm_to_node(int pxm)
> > +int pxm_to_node(int pxm)
> > {
> > if (pxm < 0)
> > return NID_INVAL;
> > return pxm_to_node_map[pxm];
> > }
> >
> > -int __cpuinit node_to_pxm(int node)
> > +int node_to_pxm(int node)
> > {
> > if (node < 0)
> > return PXM_INVAL;
> > diff --git a/include/acpi/acpi_numa.h b/include/acpi/acpi_numa.h
> > index f9d2bde..b62cd36 100644
> > --- a/include/acpi/acpi_numa.h
> > +++ b/include/acpi/acpi_numa.h
> > @@ -11,11 +11,8 @@ #else
> > #define MAX_PXM_DOMAINS (256) /* Old pxm spec is defined 8 bit */
> > #endif
> >
> > -extern int __cpuinitdata pxm_to_node_map[MAX_PXM_DOMAINS];
> > -extern int __cpuinitdata node_to_pxm_map[MAX_NUMNODES];
> > -
> > -extern int __cpuinit pxm_to_node(int);
> > -extern int __cpuinit node_to_pxm(int);
> > +extern int pxm_to_node(int);
> > +extern int node_to_pxm(int);
> > extern int __cpuinit acpi_map_pxm_to_node(int);
> > extern void __cpuinit acpi_unmap_pxm_to_node(int);
> >
>
> Moving from __fooinit into .text is always safe. Len, do you want me to
> merge this up?
>
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2007-05-16 16:38 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-05-16 16:09 [PATCH] acpi: Fix potential call to a freed memory section Aaron Durbin
2007-05-16 16:33 ` Andrew Morton
2007-05-16 16:38 ` Aaron Durbin
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.