Farewell report size can be bigger than one byte, increase array
size to accomodate maximum 32 bytes of farewell report.
Signed-off-by: Rupesh Gujare <[email protected]>
---
drivers/staging/ozwpan/ozpd.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/staging/ozwpan/ozpd.h b/drivers/staging/ozwpan/ozpd.h
index a281753..57e98c8 100644
--- a/drivers/staging/ozwpan/ozpd.h
+++ b/drivers/staging/ozwpan/ozpd.h
@@ -48,7 +48,7 @@ struct oz_farewell {
struct list_head link;
u8 ep_num;
u8 index;
- u8 report[1];
+ u8 report[32];
u8 len;
};
--
1.7.9.5
Fixes a bug where we were not setting length field causing wrong
report size to be copied.
Signed-off-by: Rupesh Gujare <[email protected]>
---
drivers/staging/ozwpan/ozproto.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/drivers/staging/ozwpan/ozproto.c b/drivers/staging/ozwpan/ozproto.c
index ec60286..084307a 100644
--- a/drivers/staging/ozwpan/ozproto.c
+++ b/drivers/staging/ozwpan/ozproto.c
@@ -296,6 +296,7 @@ static void oz_add_farewell(struct oz_pd *pd, u8 ep_num, u8 index,
return;
f->ep_num = ep_num;
f->index = index;
+ f->len = len;
memcpy(f->report, report, len);
oz_dbg(ON, "RX: Adding farewell report\n");
spin_lock(&g_polling_lock);
--
1.7.9.5
On Thu, Aug 01, 2013 at 06:45:01PM +0100, Rupesh Gujare wrote:
> Farewell report size can be bigger than one byte, increase array
> size to accomodate maximum 32 bytes of farewell report.
>
Gar... No. This is not right.
1) There is no check limiting the size to 32 and it could be up to
253 bytes.
2) Use defines instead of magic numbers.
3) The oz_farewell struct is supposed to be a variable length struct
but the variable part is put in the middle. It doesn't make any
sense to put the length of the variable size array after then end
of the array because we can never find it again! Put the
variable size array at the end. Make it a zero length array.
u8 len;
u8 report[0];
4) In oz_add_farewell() we do this:
f = kmalloc(sizeof(struct oz_farewell) + len - 1, GFP_ATOMIC);
The "- 1" refers to sizeof(f->report) but because it was a magic
number then it was missed when the sizeof(f->report) changed.
5) In [patch 6/6] we set the ->len member. But because it is at the
end of a variable length array with no limit check the remote
attacker can just rewrite it using the memcpy() on the next line.
regards,
dan carpenter
On 02/08/13 11:27, Dan Carpenter wrote:
> On Thu, Aug 01, 2013 at 06:45:01PM +0100, Rupesh Gujare wrote:
>> Farewell report size can be bigger than one byte, increase array
>> size to accomodate maximum 32 bytes of farewell report.
>>
> Gar... No. This is not right.
>
> 1) There is no check limiting the size to 32 and it could be up to
> 253 bytes.
>
> 2) Use defines instead of magic numbers.
>
> 3) The oz_farewell struct is supposed to be a variable length struct
> but the variable part is put in the middle. It doesn't make any
> sense to put the length of the variable size array after then end
> of the array because we can never find it again! Put the
> variable size array at the end. Make it a zero length array.
> u8 len;
> u8 report[0];
>
> 4) In oz_add_farewell() we do this:
>
> f = kmalloc(sizeof(struct oz_farewell) + len - 1, GFP_ATOMIC);
>
> The "- 1" refers to sizeof(f->report) but because it was a magic
> number then it was missed when the sizeof(f->report) changed.
>
> 5) In [patch 6/6] we set the ->len member. But because it is at the
> end of a variable length array with no limit check the remote
> attacker can just rewrite it using the memcpy() on the next line.
>
>
Thanks Dan.
A patch follows to fix above issues.
--
Regards,
Rupesh Gujare
This patch fixes issues reported by Dan here:-
http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2013-August/040052.html
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Rupesh Gujare <[email protected]>
---
drivers/staging/ozwpan/ozpd.h | 2 +-
drivers/staging/ozwpan/ozproto.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/staging/ozwpan/ozpd.h b/drivers/staging/ozwpan/ozpd.h
index 57e98c8..996ef65 100644
--- a/drivers/staging/ozwpan/ozpd.h
+++ b/drivers/staging/ozwpan/ozpd.h
@@ -48,8 +48,8 @@ struct oz_farewell {
struct list_head link;
u8 ep_num;
u8 index;
- u8 report[32];
u8 len;
+ u8 report[0];
};
/* Data structure that holds information on a specific peripheral device (PD).
diff --git a/drivers/staging/ozwpan/ozproto.c b/drivers/staging/ozwpan/ozproto.c
index 084307a..3d1a89f 100644
--- a/drivers/staging/ozwpan/ozproto.c
+++ b/drivers/staging/ozwpan/ozproto.c
@@ -291,7 +291,7 @@ static void oz_add_farewell(struct oz_pd *pd, u8 ep_num, u8 index,
struct oz_farewell *f;
struct oz_farewell *f2;
int found = 0;
- f = kmalloc(sizeof(struct oz_farewell) + len - 1, GFP_ATOMIC);
+ f = kmalloc(sizeof(struct oz_farewell) + len, GFP_ATOMIC);
if (!f)
return;
f->ep_num = ep_num;
--
1.7.9.5
On Fri, Aug 02, 2013 at 12:04:16PM +0100, Rupesh Gujare wrote:
> This patch fixes issues reported by Dan here:-
> http://driverdev.linuxdriverproject.org/pipermail/driverdev-devel/2013-August/040052.html
Please list the issues in the patch changelog itself, as sometimes you
don't have web access, and I sure wouldn't rely on that email archive to
always be around (hint, it's been down a bunch recently...)
Can you please resend this with a better changelog description?
thanks,
greg k-h
This patch fix following issues reported by Dan:-
1) There is no check limiting the size to 32 and it could be up to
253 bytes.
2) Use defines instead of magic numbers.
3) The oz_farewell struct is supposed to be a variable length struct
but the variable part is put in the middle. It doesn't make any
sense to put the length of the variable size array after then end
of the array because we can never find it again! Put the
variable size array at the end. Make it a zero length array.
u8 len;
u8 report[0];
4) In oz_add_farewell() we do this:
f = kmalloc(sizeof(struct oz_farewell) + len - 1, GFP_ATOMIC);
The "- 1" refers to sizeof(f->report) but because it was a magic
number then it was missed when the sizeof(f->report) changed.
5) In [patch 6/6] we set the ->len member. But because it is at the
end of a variable length array with no limit check the remote
attacker can just rewrite it using the memcpy() on the next line.
Reported-by: Dan Carpenter <[email protected]>
Signed-off-by: Rupesh Gujare <[email protected]>
---
drivers/staging/ozwpan/ozpd.h | 2 +-
drivers/staging/ozwpan/ozproto.c | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/staging/ozwpan/ozpd.h b/drivers/staging/ozwpan/ozpd.h
index 57e98c8..996ef65 100644
--- a/drivers/staging/ozwpan/ozpd.h
+++ b/drivers/staging/ozwpan/ozpd.h
@@ -48,8 +48,8 @@ struct oz_farewell {
struct list_head link;
u8 ep_num;
u8 index;
- u8 report[32];
u8 len;
+ u8 report[0];
};
/* Data structure that holds information on a specific peripheral device (PD).
diff --git a/drivers/staging/ozwpan/ozproto.c b/drivers/staging/ozwpan/ozproto.c
index 084307a..3d1a89f 100644
--- a/drivers/staging/ozwpan/ozproto.c
+++ b/drivers/staging/ozwpan/ozproto.c
@@ -291,7 +291,7 @@ static void oz_add_farewell(struct oz_pd *pd, u8 ep_num, u8 index,
struct oz_farewell *f;
struct oz_farewell *f2;
int found = 0;
- f = kmalloc(sizeof(struct oz_farewell) + len - 1, GFP_ATOMIC);
+ f = kmalloc(sizeof(struct oz_farewell) + len, GFP_ATOMIC);
if (!f)
return;
f->ep_num = ep_num;
--
1.7.9.5