We see a stack smashing as a result of defensive code missing. Once the nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to zero after doing a floor align, and we can not exit the following receiving packets loop. And the buffers will be overwrite, then the stack frame was ruined. Fix the problem by adding defensive code, once the nb_pkts is zero, just directly return with no packets. __GI___backtrace (array=0x7fcec7ac3f00, size=256) at ../sysdeps/x86_64/backtrace.c:103 catch_segfault () from /lib64/libSegFault.so <signal handler called> __GI___backtrace (array=array@entry=0x7fcec7ac62e0, size=size@entry=64) at ../sysdeps/x86_64/backtrace.c:103 backtrace_and_maps (do_abort=do_abort@entry=2, written=<optimized out>, fd=fd@entry=2) at ../sysdeps/unix/sysv/linux/libc_fatal.c:47 __libc_message (do_abort=do_abort@entry=2, fmt=fmt@entry=0x7fced6091c60 "*** %s ***: %s terminated\n") at ../sysdeps/posix/libc_fatal.c:172 __GI___fortify_fail (msg=msg@entry=0x7fced6091c48 "stack smashing detected") at fortify_fail.c:31 __stack_chk_fail () at stack_chk_fail.c:28 bnxt_recv_pkts_vec (rx_queue=0x14c571f00, rx_pkts=0x7fcec7ac6f28, nb_pkts=0) rte_eth_rx_burst (port_id=1, queue_id=3, rx_pkts=0x7fcec7ac6f28, nb_pkts=1) Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com> --- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c index d0e7910e7..c4adccdbc 100644 --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts, /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ + /* + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet + */ nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); + if (!nb_pkts) + return 0; /* Handle RX burst request */ while (1) { -- 2.11.0
We see a stack smashing as a result of defensive code missing. Once the nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to zero after doing a floor align, and we can not exit the following receiving packets loop. And the buffers will be overwrite, then the stack frame was ruined. Fix the problem by adding defensive code, once the nb_pkts is zero, just directly return with no packets. __GI___backtrace (array=0x7fcec7ac3f00, size=256) at ../sysdeps/x86_64/backtrace.c:103 catch_segfault () from /lib64/libSegFault.so <signal handler called> __GI___backtrace (array=array@entry=0x7fcec7ac62e0, size=size@entry=64) at ../sysdeps/x86_64/backtrace.c:103 backtrace_and_maps (do_abort=do_abort@entry=2, written=<optimized out>, fd=fd@entry=2) at ../sysdeps/unix/sysv/linux/libc_fatal.c:47 __libc_message (do_abort=do_abort@entry=2, fmt=fmt@entry=0x7fced6091c60 "*** %s ***: %s terminated\n") at ../sysdeps/posix/libc_fatal.c:172 __GI___fortify_fail (msg=msg@entry=0x7fced6091c48 "stack smashing detected") at fortify_fail.c:31 __stack_chk_fail () at stack_chk_fail.c:28 bnxt_recv_pkts_vec (rx_queue=0x14c571f00, rx_pkts=0x7fcec7ac6f28, nb_pkts=0) rte_eth_rx_burst (port_id=1, queue_id=3, rx_pkts=0x7fcec7ac6f28, nb_pkts=1) Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com> --- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c index d0e7910e7..c4adccdbc 100644 --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts, /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ + /* + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet + */ nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); + if (!nb_pkts) + return 0; /* Handle RX burst request */ while (1) { -- 2.11.0
We see a stack smashing as a result of defensive code missing. Once the nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to zero after doing a floor align, and we can not exit the following receiving packets loop. And the buffers will be overwrite, then the stack frame was ruined. Fix the problem by adding defensive code, once the nb_pkts is zero, just directly return with no packets. Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com> --- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c index d0e7910e7..c4adccdbc 100644 --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts, /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ + /* + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet + */ nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); + if (!nb_pkts) + return 0; /* Handle RX burst request */ while (1) { -- 2.11.0
+Lance Richardson
Thanks for the patch, could you please add the 'Fixes' tag as well ?
On Thu, Apr 30, 2020 at 5:35 PM yuanlinsi01 <yuanlinsi01@baidu.com> wrote:
>
> We see a stack smashing as a result of defensive code missing. Once the
> nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> zero after doing a floor align, and we can not exit the following
> receiving packets loop. And the buffers will be overwrite, then the
> stack frame was ruined.
>
> Fix the problem by adding defensive code, once the nb_pkts is zero, just
> directly return with no packets.
>
> Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com>
> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com>
> ---
> drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> index d0e7910e7..c4adccdbc 100644
> --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
> /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */
> nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST);
>
> - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */
> + /*
> + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP
> + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet
> + */
> nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP);
> + if (!nb_pkts)
> + return 0;
>
> /* Handle RX burst request */
> while (1) {
> --
> 2.11.0
>
Sure, I'll add it.
Thanks,
Yuan Linsi
________________________________
发件人: Somnath Kotur <somnath.kotur@broadcom.com>
发送时间: 2020年4月30日 20:55
收件人: Yuan,Linsi; Lance Richardson
抄送: Ajit Kumar Khaparde; dev
主题: Re: [PATCH] net/bnxt: fix a possible stack smashing
+Lance Richardson
Thanks for the patch, could you please add the 'Fixes' tag as well ?
On Thu, Apr 30, 2020 at 5:35 PM yuanlinsi01 <yuanlinsi01@baidu.com> wrote:
>
> We see a stack smashing as a result of defensive code missing. Once the
> nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> zero after doing a floor align, and we can not exit the following
> receiving packets loop. And the buffers will be overwrite, then the
> stack frame was ruined.
>
> Fix the problem by adding defensive code, once the nb_pkts is zero, just
> directly return with no packets.
>
> Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com>
> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com>
> ---
> drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> index d0e7910e7..c4adccdbc 100644
> --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
> /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */
> nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST);
>
> - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */
> + /*
> + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP
> + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet
> + */
> nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP);
> + if (!nb_pkts)
> + return 0;
>
> /* Handle RX burst request */
> while (1) {
> --
> 2.11.0
>
From: yuanlinsi01 <yuanlinsi01@baidu.com> We see a stack smashing as a result of defensive code missing. Once the nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to zero after doing a floor align, and we can not exit the following receiving packets loop. And the buffers will be overwrite, then the stack frame was ruined. Fix the problem by adding defensive code, once the nb_pkts is zero, just directly return with no packets. Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode") Cc: stable@dpdk.org Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com> --- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c index d0e7910e7..8f73add9b 100644 --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts, /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ + /* + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP. + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet + */ nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); + if (!nb_pkts) + return 0; /* Handle RX burst request */ while (1) { -- 2.11.0
On Thu, Apr 30, 2020 at 9:37 AM Yuan Linsi <yuanlinsi01@baidu.com> wrote:
>
> From: yuanlinsi01 <yuanlinsi01@baidu.com>
>
> We see a stack smashing as a result of defensive code missing. Once the
> nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> zero after doing a floor align, and we can not exit the following
> receiving packets loop. And the buffers will be overwrite, then the
> stack frame was ruined.
>
> Fix the problem by adding defensive code, once the nb_pkts is zero, just
> directly return with no packets.
>
> Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode")
> Cc: stable@dpdk.org
>
> Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com>
> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com>
> ---
> drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> index d0e7910e7..8f73add9b 100644
> --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c
> @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
> /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */
> nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST);
>
> - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */
> + /*
> + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP.
> + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet
> + */
> nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP);
> + if (!nb_pkts)
> + return 0;
>
> /* Handle RX burst request */
> while (1) {
> --
> 2.11.0
>
Thanks for the fix!
Acked-by: Lance Richardson <lance.richardson@broadcom.com>
On Thu, Apr 30, 2020 at 6:45 AM Lance Richardson <
lance.richardson@broadcom.com> wrote:
> On Thu, Apr 30, 2020 at 9:37 AM Yuan Linsi <yuanlinsi01@baidu.com> wrote:
> >
> > From: yuanlinsi01 <yuanlinsi01@baidu.com>
> >
> > We see a stack smashing as a result of defensive code missing. Once the
> > nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> > zero after doing a floor align, and we can not exit the following
> > receiving packets loop. And the buffers will be overwrite, then the
> > stack frame was ruined.
> >
> > Fix the problem by adding defensive code, once the nb_pkts is zero, just
> > directly return with no packets.
> >
> > Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com>
> > Signed-off-by: rongdongsheng <rongdongsheng@baidu.com>
>
> Thanks for the fix!
>
> Acked-by: Lance Richardson <lance.richardson@broadcom.com>
>
Patch applied to dpdk-next-net-brcm. Thanks
On 4/30/2020 2:37 PM, Yuan Linsi wrote:
> From: yuanlinsi01 <yuanlinsi01@baidu.com>
>
> We see a stack smashing as a result of defensive code missing. Once the
> nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> zero after doing a floor align, and we can not exit the following
> receiving packets loop. And the buffers will be overwrite, then the
> stack frame was ruined.
>
> Fix the problem by adding defensive code, once the nb_pkts is zero, just
> directly return with no packets.
>
> Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode")
> Cc: stable@dpdk.org
>
> Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com>
> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com>
Hi Yuan,
For the sign-off tag, we need "Name Surname <email@adress.com>" syntax,
for you I can see from mail thread that it is:
"Signed-off-by: Linsi Yuan <yuanlinsi01@baidu.com>"
Can you please share the same for the other sign-off, 'rongdongsheng'?
On Thu, Apr 30, 2020 at 4:55 PM Ferruh Yigit <ferruh.yigit@intel.com> wrote:
> On 4/30/2020 2:37 PM, Yuan Linsi wrote:
> > From: yuanlinsi01 <yuanlinsi01@baidu.com>
> >
> > We see a stack smashing as a result of defensive code missing. Once the
> > nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> > zero after doing a floor align, and we can not exit the following
> > receiving packets loop. And the buffers will be overwrite, then the
> > stack frame was ruined.
> >
> > Fix the problem by adding defensive code, once the nb_pkts is zero, just
> > directly return with no packets.
> >
> > Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com>
> > Signed-off-by: rongdongsheng <rongdongsheng@baidu.com>
>
> Hi Yuan,
>
> For the sign-off tag, we need "Name Surname <email@adress.com>" syntax,
> for you I can see from mail thread that it is:
> "Signed-off-by: Linsi Yuan <yuanlinsi01@baidu.com>"
>
> Can you please share the same for the other sign-off, 'rongdongsheng'?
>
Yuan, Can you please send the proper first-name, last-name
for rongdongsheng? Thanks
Ok, I'll modify it per you suggestion
Thanks,
Yuan Linsi
________________________________
发件人: Ajit Khaparde <ajit.khaparde@broadcom.com>
发送时间: 2020年5月5日 11:42:20
收件人: Ferruh Yigit
抄送: Yuan,Linsi; Somnath Kotur; Lance Richardson; dpdk-dev
主题: Re: [dpdk-dev] [PATCH] net/bnxt: fix a possible stack smashing
On Thu, Apr 30, 2020 at 4:55 PM Ferruh Yigit <ferruh.yigit@intel.com<mailto:ferruh.yigit@intel.com>> wrote:
On 4/30/2020 2:37 PM, Yuan Linsi wrote:
> From: yuanlinsi01 <yuanlinsi01@baidu.com<mailto:yuanlinsi01@baidu.com>>
>
> We see a stack smashing as a result of defensive code missing. Once the
> nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to
> zero after doing a floor align, and we can not exit the following
> receiving packets loop. And the buffers will be overwrite, then the
> stack frame was ruined.
>
> Fix the problem by adding defensive code, once the nb_pkts is zero, just
> directly return with no packets.
>
> Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode")
> Cc: stable@dpdk.org<mailto:stable@dpdk.org>
>
> Signed-off-by: yuanlinsi01 <yuanlinsi01@baidu.com<mailto:yuanlinsi01@baidu.com>>
> Signed-off-by: rongdongsheng <rongdongsheng@baidu.com<mailto:rongdongsheng@baidu.com>>
Hi Yuan,
For the sign-off tag, we need "Name Surname <email@adress.com<mailto:email@adress.com>>" syntax,
for you I can see from mail thread that it is:
"Signed-off-by: Linsi Yuan <yuanlinsi01@baidu.com<mailto:yuanlinsi01@baidu.com>>"
Can you please share the same for the other sign-off, 'rongdongsheng'?
Yuan, Can you please send the proper first-name, last-name for rongdongsheng? Thanks
From: yuanlinsi01 <yuanlinsi01@baidu.com> We see a stack smashing as a result of defensive code missing. Once the nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to zero after doing a floor align, and we can not exit the following receiving packets loop. And the buffers will be overwrite, then the stack frame was ruined. Fix the problem by adding defensive code, once the nb_pkts is zero, just directly return with no packets. Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode") Cc: stable@dpdk.org Signed-off-by: Linsi Yuan <yuanlinsi01@baidu.com> Signed-off-by: Dongsheng Rong <rongdongsheng@baidu.com> --- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c index d0e7910e7..8f73add9b 100644 --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts, /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ + /* + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP. + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet + */ nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); + if (!nb_pkts) + return 0; /* Handle RX burst request */ while (1) { -- 2.11.0
From: Linsi Yuan <yuanlinsi01@baidu.com> We see a stack smashing as a result of defensive code missing. Once the nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to zero after doing a floor align, and we can not exit the following receiving packets loop. And the buffers will be overwrite, then the stack frame was ruined. Fix the problem by adding defensive code, once the nb_pkts is zero, just directly return with no packets. Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode") Cc: stable@dpdk.org Signed-off-by: Linsi Yuan <yuanlinsi01@baidu.com> Signed-off-by: Dongsheng Rong <rongdongsheng@baidu.com> --- drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c index d0e7910e7..8f73add9b 100644 --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts, /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ + /* + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP. + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet + */ nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); + if (!nb_pkts) + return 0; /* Handle RX burst request */ while (1) { -- 2.11.0
On Tue, May 5, 2020 at 8:29 PM Yuan Linsi <yuanlinsi01@baidu.com> wrote: > From: Linsi Yuan <yuanlinsi01@baidu.com> > > We see a stack smashing as a result of defensive code missing. Once the > nb_pkts is less than RTE_BNXT_DESCS_PER_LOOP, it will be modified to > zero after doing a floor align, and we can not exit the following > receiving packets loop. And the buffers will be overwrite, then the > stack frame was ruined. > > Fix the problem by adding defensive code, once the nb_pkts is zero, just > directly return with no packets. > > Fixes: bc4a000f2 ("net/bnxt: implement SSE vector mode") > Cc: stable@dpdk.org > > Signed-off-by: Linsi Yuan <yuanlinsi01@baidu.com> > Signed-off-by: Dongsheng Rong <rongdongsheng@baidu.com> > Thanks. I updated the earlier version [1] with this Signed-off. [1] https://patchwork.dpdk.org/patch/69604/ > --- > drivers/net/bnxt/bnxt_rxtx_vec_sse.c | 7 ++++++- > 1 file changed, 6 insertions(+), 1 deletion(-) > > diff --git a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c > b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c > index d0e7910e7..8f73add9b 100644 > --- a/drivers/net/bnxt/bnxt_rxtx_vec_sse.c > +++ b/drivers/net/bnxt/bnxt_rxtx_vec_sse.c > @@ -233,8 +233,13 @@ bnxt_recv_pkts_vec(void *rx_queue, struct rte_mbuf > **rx_pkts, > /* Return no more than RTE_BNXT_MAX_RX_BURST per call. */ > nb_pkts = RTE_MIN(nb_pkts, RTE_BNXT_MAX_RX_BURST); > > - /* Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP */ > + /* > + * Make nb_pkts an integer multiple of RTE_BNXT_DESCS_PER_LOOP. > + * nb_pkts < RTE_BNXT_DESCS_PER_LOOP, just return no packet > + */ > nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_BNXT_DESCS_PER_LOOP); > + if (!nb_pkts) > + return 0; > > /* Handle RX burst request */ > while (1) { > -- > 2.11.0 > >