* [PATCH v2] rds: avoid potential stack overflow
@ 2015-03-11 21:46 Arnd Bergmann
2015-03-11 21:50 ` Sowmini Varadhan
2015-03-12 4:28 ` David Miller
0 siblings, 2 replies; 3+ messages in thread
From: Arnd Bergmann @ 2015-03-11 21:46 UTC (permalink / raw)
To: netdev
Cc: roland, rds-devel, linux-kernel, chien.yen, sowmini.varadhan,
linux-arm-kernel, David Miller
The rds_iw_update_cm_id function stores a large 'struct rds_sock' object
on the stack in order to pass a pair of addresses. This happens to just
fit withint the 1024 byte stack size warning limit on x86, but just
exceed that limit on ARM, which gives us this warning:
net/rds/iw_rdma.c:200:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=]
As the use of this large variable is basically bogus, we can rearrange
the code to not do that. Instead of passing an rds socket into
rds_iw_get_device, we now just pass the two addresses that we have
available in rds_iw_update_cm_id, and we change rds_iw_get_mr accordingly,
to create two address structures on the stack there.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
v2: fix incorrect commit message
diff --git a/net/rds/iw_rdma.c b/net/rds/iw_rdma.c
index a817705ce2d0..dba8d0864f18 100644
--- a/net/rds/iw_rdma.c
+++ b/net/rds/iw_rdma.c
@@ -88,7 +88,9 @@ static unsigned int rds_iw_unmap_fastreg_list(struct rds_iw_mr_pool *pool,
int *unpinned);
static void rds_iw_destroy_fastreg(struct rds_iw_mr_pool *pool, struct rds_iw_mr *ibmr);
-static int rds_iw_get_device(struct rds_sock *rs, struct rds_iw_device **rds_iwdev, struct rdma_cm_id **cm_id)
+static int rds_iw_get_device(struct sockaddr_in *src, struct sockaddr_in *dst,
+ struct rds_iw_device **rds_iwdev,
+ struct rdma_cm_id **cm_id)
{
struct rds_iw_device *iwdev;
struct rds_iw_cm_id *i_cm_id;
@@ -112,15 +114,15 @@ static int rds_iw_get_device(struct rds_sock *rs, struct rds_iw_device **rds_iwd
src_addr->sin_port,
dst_addr->sin_addr.s_addr,
dst_addr->sin_port,
- rs->rs_bound_addr,
- rs->rs_bound_port,
- rs->rs_conn_addr,
- rs->rs_conn_port);
+ src->sin_addr.s_addr,
+ src->sin_port,
+ dst->sin_addr.s_addr,
+ dst->sin_port);
#ifdef WORKING_TUPLE_DETECTION
- if (src_addr->sin_addr.s_addr == rs->rs_bound_addr &&
- src_addr->sin_port == rs->rs_bound_port &&
- dst_addr->sin_addr.s_addr == rs->rs_conn_addr &&
- dst_addr->sin_port == rs->rs_conn_port) {
+ if (src_addr->sin_addr.s_addr == src->sin_addr.s_addr &&
+ src_addr->sin_port == src->sin_port &&
+ dst_addr->sin_addr.s_addr == dst->sin_addr.s_addr &&
+ dst_addr->sin_port == dst->sin_port) {
#else
/* FIXME - needs to compare the local and remote
* ipaddr/port tuple, but the ipaddr is the only
@@ -128,7 +130,7 @@ static int rds_iw_get_device(struct rds_sock *rs, struct rds_iw_device **rds_iwd
* zero'ed. It doesn't appear to be properly populated
* during connection setup...
*/
- if (src_addr->sin_addr.s_addr == rs->rs_bound_addr) {
+ if (src_addr->sin_addr.s_addr == src->sin_addr.s_addr) {
#endif
spin_unlock_irq(&iwdev->spinlock);
*rds_iwdev = iwdev;
@@ -180,19 +182,13 @@ int rds_iw_update_cm_id(struct rds_iw_device *rds_iwdev, struct rdma_cm_id *cm_i
{
struct sockaddr_in *src_addr, *dst_addr;
struct rds_iw_device *rds_iwdev_old;
- struct rds_sock rs;
struct rdma_cm_id *pcm_id;
int rc;
src_addr = (struct sockaddr_in *)&cm_id->route.addr.src_addr;
dst_addr = (struct sockaddr_in *)&cm_id->route.addr.dst_addr;
- rs.rs_bound_addr = src_addr->sin_addr.s_addr;
- rs.rs_bound_port = src_addr->sin_port;
- rs.rs_conn_addr = dst_addr->sin_addr.s_addr;
- rs.rs_conn_port = dst_addr->sin_port;
-
- rc = rds_iw_get_device(&rs, &rds_iwdev_old, &pcm_id);
+ rc = rds_iw_get_device(src_addr, dst_addr, &rds_iwdev_old, &pcm_id);
if (rc)
rds_iw_remove_cm_id(rds_iwdev, cm_id);
@@ -598,9 +594,17 @@ void *rds_iw_get_mr(struct scatterlist *sg, unsigned long nents,
struct rds_iw_device *rds_iwdev;
struct rds_iw_mr *ibmr = NULL;
struct rdma_cm_id *cm_id;
+ struct sockaddr_in src = {
+ .sin_addr.s_addr = rs->rs_bound_addr,
+ .sin_port = rs->rs_bound_port,
+ };
+ struct sockaddr_in dst = {
+ .sin_addr.s_addr = rs->rs_conn_addr,
+ .sin_port = rs->rs_conn_port,
+ };
int ret;
- ret = rds_iw_get_device(rs, &rds_iwdev, &cm_id);
+ ret = rds_iw_get_device(&src, &dst, &rds_iwdev, &cm_id);
if (ret || !cm_id) {
ret = -ENODEV;
goto out;
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH v2] rds: avoid potential stack overflow
2015-03-11 21:46 [PATCH v2] rds: avoid potential stack overflow Arnd Bergmann
@ 2015-03-11 21:50 ` Sowmini Varadhan
2015-03-12 4:28 ` David Miller
1 sibling, 0 replies; 3+ messages in thread
From: Sowmini Varadhan @ 2015-03-11 21:50 UTC (permalink / raw)
To: Arnd Bergmann
Cc: netdev, roland, rds-devel, linux-kernel, chien.yen,
linux-arm-kernel, David Miller
On (03/11/15 22:46), Arnd Bergmann wrote:
>
> The rds_iw_update_cm_id function stores a large 'struct rds_sock' object
> on the stack in order to pass a pair of addresses. This happens to just
> fit withint the 1024 byte stack size warning limit on x86, but just
> exceed that limit on ARM, which gives us this warning:
:
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
--Sowmini
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH v2] rds: avoid potential stack overflow
2015-03-11 21:46 [PATCH v2] rds: avoid potential stack overflow Arnd Bergmann
2015-03-11 21:50 ` Sowmini Varadhan
@ 2015-03-12 4:28 ` David Miller
1 sibling, 0 replies; 3+ messages in thread
From: David Miller @ 2015-03-12 4:28 UTC (permalink / raw)
To: arnd
Cc: netdev, roland, rds-devel, linux-kernel, chien.yen,
sowmini.varadhan, linux-arm-kernel
From: Arnd Bergmann <arnd@arndb.de>
Date: Wed, 11 Mar 2015 22:46:59 +0100
> The rds_iw_update_cm_id function stores a large 'struct rds_sock' object
> on the stack in order to pass a pair of addresses. This happens to just
> fit withint the 1024 byte stack size warning limit on x86, but just
> exceed that limit on ARM, which gives us this warning:
>
> net/rds/iw_rdma.c:200:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=]
>
> As the use of this large variable is basically bogus, we can rearrange
> the code to not do that. Instead of passing an rds socket into
> rds_iw_get_device, we now just pass the two addresses that we have
> available in rds_iw_update_cm_id, and we change rds_iw_get_mr accordingly,
> to create two address structures on the stack there.
>
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
> v2: fix incorrect commit message
Applied, thanks Arnd.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2015-03-12 4:28 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-03-11 21:46 [PATCH v2] rds: avoid potential stack overflow Arnd Bergmann
2015-03-11 21:50 ` Sowmini Varadhan
2015-03-12 4:28 ` David Miller
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).