net/af_packet: remove timestamp from packet status
Checks
Commit Message
We should eliminate the timestamp status from the packet
status. This should only matter if timestamping is enabled
on the socket, but we might hit a kernel bug, which is fixed
in newer releases.
For interfaces of type 'veth', the sent skb is forwarded
to the peer and back into the network stack which timestamps
it on the RX path if timestamping is enabled globally
(which happens if any socket enables timestamping).
When the skb is destructed, tpacket_destruct_skb() is called
and it calls __packet_set_timestamp() which doesn't check
the flags on the socket and returns the timestamp if it is
set in the skb (and for veth it is, as mentioned above).
See the following kernel commit for reference [1]:
net: packetmmap: fix only tx timestamp on request
The packetmmap tx ring should only return timestamps if requested
via setsockopt PACKET_TIMESTAMP, as documented. This allows
compatibility with non-timestamp aware user-space code which checks
tp_status == TP_STATUS_AVAILABLE; not expecting additional timestamp
flags to be set in tp_status.
[1] https://www.spinics.net/lists/kernel/msg3959391.html
Signed-off-by: Mihai Pogonaru <pogonarumihai@gmail.com>
Signed-off-by: Tudor Cornea <tudor.cornea@gmail.com>
---
drivers/net/af_packet/rte_eth_af_packet.c | 22 ++++++++++++++++++++--
1 file changed, 20 insertions(+), 2 deletions(-)
Comments
On Mon, 13 Sep 2021 17:09:11 +0300
Tudor Cornea <tudor.cornea@gmail.com> wrote:
> +static inline bool tx_ring_status_unavailable(uint32_t tp_status)
> +{
> +#if KERNEL_VERSION(5, 10, 0) > LINUX_VERSION_CODE
No, having kernel dependent userspace in DPDK is not good practice.
Distribution vendors don't number their kernels the same as upstream.
RHEL for example, keeps same version over life or release but backports
many fixes.
Also, the system DPDK runs on is often not the system DPDK is built
on.
Thanks for the observation.
I have removed the compile-time kernel version check in v2 of the patch
On Mon, 13 Sept 2021 at 18:09, Stephen Hemminger <stephen@networkplumber.org>
wrote:
> On Mon, 13 Sep 2021 17:09:11 +0300
> Tudor Cornea <tudor.cornea@gmail.com> wrote:
>
> > +static inline bool tx_ring_status_unavailable(uint32_t tp_status)
> > +{
> > +#if KERNEL_VERSION(5, 10, 0) > LINUX_VERSION_CODE
>
> No, having kernel dependent userspace in DPDK is not good practice.
>
> Distribution vendors don't number their kernels the same as upstream.
> RHEL for example, keeps same version over life or release but backports
> many fixes.
>
> Also, the system DPDK runs on is often not the system DPDK is built
> on.
>
@@ -17,6 +17,7 @@
#include <errno.h>
#include <linux/if_ether.h>
#include <linux/if_packet.h>
+#include <linux/version.h>
#include <arpa/inet.h>
#include <net/if.h>
#include <net/if_arp.h>
@@ -167,6 +168,23 @@ eth_af_packet_rx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
return num_rx;
}
+static inline bool tx_ring_status_unavailable(uint32_t tp_status)
+{
+#if KERNEL_VERSION(5, 10, 0) > LINUX_VERSION_CODE
+ /*
+ * We eliminate the timestamp status from the packet status.
+ * This should only matter if timestamping is enabled on the socket,
+ * but there is a bug in the kernel which is fixed in newer releases.
+ *
+ * See the following kernel commit for reference:
+ * commit 171c3b151118a2fe0fc1e2a9d1b5a1570cfe82d2
+ * net: packetmmap: fix only tx timestamp on request
+ */
+ tp_status &= ~(TP_STATUS_TS_SOFTWARE | TP_STATUS_TS_RAW_HARDWARE);
+#endif
+ return tp_status != TP_STATUS_AVAILABLE;
+}
+
/*
* Callback to handle sending packets through a real NIC.
*/
@@ -212,8 +230,8 @@ eth_af_packet_tx(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
}
/* point at the next incoming frame */
- if ((ppd->tp_status != TP_STATUS_AVAILABLE) &&
- (poll(&pfd, 1, -1) < 0))
+ if (tx_ring_status_unavailable(ppd->tp_status) &&
+ poll(&pfd, 1, -1) < 0)
break;
/* copy the tx frame data */