* [PATCH] net/rds: Fix a use after free in rds_message_map_pages
@ 2021-03-30 10:16 Lv Yunlong
2021-03-31 0:02 ` David Miller
0 siblings, 1 reply; 5+ messages in thread
From: Lv Yunlong @ 2021-03-30 10:16 UTC (permalink / raw)
To: santosh.shilimkar, davem, kuba
Cc: netdev, linux-rdma, rds-devel, linux-kernel, Lv Yunlong
In rds_message_map_pages, the rm is freed by rds_message_put(rm).
But rm is still used by rm->data.op_sg in return value.
My patch replaces ERR_CAST(rm->data.op_sg) to ERR_PTR(-ENOMEM) to avoid
the uaf.
Fixes: 7dba92037baf3 ("net/rds: Use ERR_PTR for rds_message_alloc_sgs()")
Signed-off-by: Lv Yunlong <lyl2019@mail.ustc.edu.cn>
---
net/rds/message.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/rds/message.c b/net/rds/message.c
index 071a261fdaab..cecd968c9b25 100644
--- a/net/rds/message.c
+++ b/net/rds/message.c
@@ -348,7 +348,7 @@ struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned in
rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
if (IS_ERR(rm->data.op_sg)) {
rds_message_put(rm);
- return ERR_CAST(rm->data.op_sg);
+ return ERR_PTR(-ENOMEM);
}
for (i = 0; i < rm->data.op_nents; ++i) {
--
2.25.1
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH] net/rds: Fix a use after free in rds_message_map_pages
2021-03-30 10:16 [PATCH] net/rds: Fix a use after free in rds_message_map_pages Lv Yunlong
@ 2021-03-31 0:02 ` David Miller
2021-03-31 2:02 ` lyl2019
0 siblings, 1 reply; 5+ messages in thread
From: David Miller @ 2021-03-31 0:02 UTC (permalink / raw)
To: lyl2019
Cc: santosh.shilimkar, kuba, netdev, linux-rdma, rds-devel, linux-kernel
From: Lv Yunlong <lyl2019@mail.ustc.edu.cn>
Date: Tue, 30 Mar 2021 03:16:02 -0700
> @@ -348,7 +348,7 @@ struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned in
> rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
> if (IS_ERR(rm->data.op_sg)) {
> rds_message_put(rm);
> - return ERR_CAST(rm->data.op_sg);
> + return ERR_PTR(-ENOMEM);
> }
>
> for (i = 0; i < rm->data.op_nents; ++i) {
Maybe instead do:
int err = ERR_CAST(rm->data.op_sg);
rds_message_put(rm);
return err;
Then if rds_message_alloc_sgs() starts to return other errors, they will propagate.
Thank you.
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: Re: [PATCH] net/rds: Fix a use after free in rds_message_map_pages
2021-03-31 0:02 ` David Miller
@ 2021-03-31 2:02 ` lyl2019
0 siblings, 0 replies; 5+ messages in thread
From: lyl2019 @ 2021-03-31 2:02 UTC (permalink / raw)
To: David Miller
Cc: santosh.shilimkar, kuba, netdev, linux-rdma, rds-devel, linux-kernel
> -----原始邮件-----
> 发件人: "David Miller" <davem@davemloft.net>
> 发送时间: 2021-03-31 08:02:28 (星期三)
> 收件人: lyl2019@mail.ustc.edu.cn
> 抄送: santosh.shilimkar@oracle.com, kuba@kernel.org, netdev@vger.kernel.org, linux-rdma@vger.kernel.org, rds-devel@oss.oracle.com, linux-kernel@vger.kernel.org
> 主题: Re: [PATCH] net/rds: Fix a use after free in rds_message_map_pages
>
> From: Lv Yunlong <lyl2019@mail.ustc.edu.cn>
> Date: Tue, 30 Mar 2021 03:16:02 -0700
>
> > @@ -348,7 +348,7 @@ struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned in
> > rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
> > if (IS_ERR(rm->data.op_sg)) {
> > rds_message_put(rm);
> > - return ERR_CAST(rm->data.op_sg);
> > + return ERR_PTR(-ENOMEM);
> > }
> >
> > for (i = 0; i < rm->data.op_nents; ++i) {
>
> Maybe instead do:
>
> int err = ERR_CAST(rm->data.op_sg);
> rds_message_put(rm);
> return err;
>
> Then if rds_message_alloc_sgs() starts to return other errors, they will propagate.
>
> Thank you.
The type of ERR_CAST() is void *, not int.
I think the correct patch is:
void *err = ERR_CAST(rm->data.op_sg);
rds_message_put(rm);
return err;
I have submitted the PATCH v2 for you to review.
Thanks.
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] net/rds: Fix a use after free in rds_message_map_pages
@ 2021-03-11 8:46 Lv Yunlong
2021-03-19 2:25 ` lyl2019
0 siblings, 1 reply; 5+ messages in thread
From: Lv Yunlong @ 2021-03-11 8:46 UTC (permalink / raw)
To: santosh.shilimkar, davem, kuba
Cc: netdev, linux-rdma, rds-devel, linux-kernel, Lv Yunlong
In rds_message_map_pages, rds_message_put() will free rm.
Maybe store the value of rm->data.op_sg ahead of rds_message_put()
is better. Otherwise other threads could allocate the freed chunk
and may change the value of rm->data.op_sg.
Signed-off-by: Lv Yunlong <lyl2019@mail.ustc.edu.cn>
---
net/rds/message.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/net/rds/message.c b/net/rds/message.c
index 071a261fdaab..392e3a2f41a0 100644
--- a/net/rds/message.c
+++ b/net/rds/message.c
@@ -347,8 +347,9 @@ struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned in
rm->data.op_nents = DIV_ROUND_UP(total_len, PAGE_SIZE);
rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
if (IS_ERR(rm->data.op_sg)) {
+ struct scatterlist *tmp = rm->data.op_sg;
rds_message_put(rm);
- return ERR_CAST(rm->data.op_sg);
+ return ERR_CAST(tmp);
}
for (i = 0; i < rm->data.op_nents; ++i) {
--
2.25.1
^ permalink raw reply related [flat|nested] 5+ messages in thread* Re: [PATCH] net/rds: Fix a use after free in rds_message_map_pages
2021-03-11 8:46 Lv Yunlong
@ 2021-03-19 2:25 ` lyl2019
0 siblings, 0 replies; 5+ messages in thread
From: lyl2019 @ 2021-03-19 2:25 UTC (permalink / raw)
To: Lv Yunlong
Cc: santosh.shilimkar, davem, kuba, netdev, linux-rdma, rds-devel,
linux-kernel
Just as a reminder,
there has been no reply to this message for more than a week.
Could someone help to fix this issue?
> -----原始邮件-----
> 发件人: "Lv Yunlong" <lyl2019@mail.ustc.edu.cn>
> 发送时间: 2021-03-11 16:46:16 (星期四)
> 收件人: santosh.shilimkar@oracle.com, davem@davemloft.net, kuba@kernel.org
> 抄送: netdev@vger.kernel.org, linux-rdma@vger.kernel.org, rds-devel@oss.oracle.com, linux-kernel@vger.kernel.org, "Lv Yunlong" <lyl2019@mail.ustc.edu.cn>
> 主题: [PATCH] net/rds: Fix a use after free in rds_message_map_pages
>
> In rds_message_map_pages, rds_message_put() will free rm.
> Maybe store the value of rm->data.op_sg ahead of rds_message_put()
> is better. Otherwise other threads could allocate the freed chunk
> and may change the value of rm->data.op_sg.
>
> Signed-off-by: Lv Yunlong <lyl2019@mail.ustc.edu.cn>
> ---
> net/rds/message.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/net/rds/message.c b/net/rds/message.c
> index 071a261fdaab..392e3a2f41a0 100644
> --- a/net/rds/message.c
> +++ b/net/rds/message.c
> @@ -347,8 +347,9 @@ struct rds_message *rds_message_map_pages(unsigned long *page_addrs, unsigned in
> rm->data.op_nents = DIV_ROUND_UP(total_len, PAGE_SIZE);
> rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs);
> if (IS_ERR(rm->data.op_sg)) {
> + struct scatterlist *tmp = rm->data.op_sg;
> rds_message_put(rm);
> - return ERR_CAST(rm->data.op_sg);
> + return ERR_CAST(tmp);
> }
>
> for (i = 0; i < rm->data.op_nents; ++i) {
> --
> 2.25.1
>
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2021-03-31 2:03 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-30 10:16 [PATCH] net/rds: Fix a use after free in rds_message_map_pages Lv Yunlong
2021-03-31 0:02 ` David Miller
2021-03-31 2:02 ` lyl2019
-- strict thread matches above, loose matches on Subject: below --
2021-03-11 8:46 Lv Yunlong
2021-03-19 2:25 ` lyl2019
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).