[PATCH 1/2] pcapng: fix handling of chained mbufs

[PATCH 1/2] pcapng: fix handling of chained mbufs

[Oleksandr Nahnybida]

The pcapng generates corrupted files when dealing with chained mbufs.
This issue arises because in rte_pcapng_copy the length of the EPB block
is incorrectly calculated using the data_len of the first mbuf instead
of the pkt_len, despite that rte_pcapng_write_packets correctly writing
the mbuf chain to disk.

This fix ensures that the block length is calculated based on the pkt_len,
aligning it with the actual data written to disk.

Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemaster.com>
---
 lib/pcapng/rte_pcapng.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/lib/pcapng/rte_pcapng.c b/lib/pcapng/rte_pcapng.c
index 7254defce7..e5326c1d38 100644
--- a/lib/pcapng/rte_pcapng.c
+++ b/lib/pcapng/rte_pcapng.c
@@ -475,7 +475,7 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		const char *comment)
 {
 	struct pcapng_enhance_packet_block *epb;
-	uint32_t orig_len, data_len, padding, flags;
+	uint32_t orig_len, pkt_len, padding, flags;
 	struct pcapng_option *opt;
 	uint64_t timestamp;
 	uint16_t optlen;
@@ -516,8 +516,8 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		    (md->ol_flags & RTE_MBUF_F_RX_RSS_HASH));
 
 	/* pad the packet to 32 bit boundary */
-	data_len = rte_pktmbuf_data_len(mc);
-	padding = RTE_ALIGN(data_len, sizeof(uint32_t)) - data_len;
+	pkt_len = rte_pktmbuf_pkt_len(mc);
+	padding = RTE_ALIGN(pkt_len, sizeof(uint32_t)) - pkt_len;
 	if (padding > 0) {
 		void *tail = rte_pktmbuf_append(mc, padding);
 
@@ -584,7 +584,7 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		goto fail;
 
 	epb->block_type = PCAPNG_ENHANCED_PACKET_BLOCK;
-	epb->block_length = rte_pktmbuf_data_len(mc);
+	epb->block_length = rte_pktmbuf_pkt_len(mc);
 
 	/* Interface index is filled in later during write */
 	mc->port = port_id;
@@ -593,7 +593,7 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 	timestamp = rte_get_tsc_cycles();
 	epb->timestamp_hi = timestamp >> 32;
 	epb->timestamp_lo = (uint32_t)timestamp;
-	epb->capture_length = data_len;
+	epb->capture_length = pkt_len;
 	epb->original_length = orig_len;
 
 	/* set trailer of block length */
@@ -623,7 +623,7 @@ rte_pcapng_write_packets(rte_pcapng_t *self,
 		/* sanity check that is really a pcapng mbuf */
 		epb = rte_pktmbuf_mtod(m, struct pcapng_enhance_packet_block *);
 		if (unlikely(epb->block_type != PCAPNG_ENHANCED_PACKET_BLOCK ||
-			     epb->block_length != rte_pktmbuf_data_len(m))) {
+			     epb->block_length != rte_pktmbuf_pkt_len(m))) {
 			rte_errno = EINVAL;
 			return -1;
 		}
-- 
2.46.0

[PATCH v2 1/2] pcapng: fix handling of chained mbufs

[Oleksandr Nahnybida]

The pcapng generates corrupted files when dealing with chained mbufs.
This issue arises because in rte_pcapng_copy the length of the EPB block
is incorrectly calculated using the data_len of the first mbuf instead
of the pkt_len, despite that rte_pcapng_write_packets correctly writing
the mbuf chain to disk.

This fix ensures that the block length is calculated based on the pkt_len,
aligning it with the actual data written to disk.

Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>
---
v2 - fixed typo in signoff email

 lib/pcapng/rte_pcapng.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/lib/pcapng/rte_pcapng.c b/lib/pcapng/rte_pcapng.c
index 7254defce7..e5326c1d38 100644
--- a/lib/pcapng/rte_pcapng.c
+++ b/lib/pcapng/rte_pcapng.c
@@ -475,7 +475,7 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		const char *comment)
 {
 	struct pcapng_enhance_packet_block *epb;
-	uint32_t orig_len, data_len, padding, flags;
+	uint32_t orig_len, pkt_len, padding, flags;
 	struct pcapng_option *opt;
 	uint64_t timestamp;
 	uint16_t optlen;
@@ -516,8 +516,8 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		    (md->ol_flags & RTE_MBUF_F_RX_RSS_HASH));

 	/* pad the packet to 32 bit boundary */
-	data_len = rte_pktmbuf_data_len(mc);
-	padding = RTE_ALIGN(data_len, sizeof(uint32_t)) - data_len;
+	pkt_len = rte_pktmbuf_pkt_len(mc);
+	padding = RTE_ALIGN(pkt_len, sizeof(uint32_t)) - pkt_len;
 	if (padding > 0) {
 		void *tail = rte_pktmbuf_append(mc, padding);

@@ -584,7 +584,7 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 		goto fail;

 	epb->block_type = PCAPNG_ENHANCED_PACKET_BLOCK;
-	epb->block_length = rte_pktmbuf_data_len(mc);
+	epb->block_length = rte_pktmbuf_pkt_len(mc);

 	/* Interface index is filled in later during write */
 	mc->port = port_id;
@@ -593,7 +593,7 @@ rte_pcapng_copy(uint16_t port_id, uint32_t queue,
 	timestamp = rte_get_tsc_cycles();
 	epb->timestamp_hi = timestamp >> 32;
 	epb->timestamp_lo = (uint32_t)timestamp;
-	epb->capture_length = data_len;
+	epb->capture_length = pkt_len;
 	epb->original_length = orig_len;

 	/* set trailer of block length */
@@ -623,7 +623,7 @@ rte_pcapng_write_packets(rte_pcapng_t *self,
 		/* sanity check that is really a pcapng mbuf */
 		epb = rte_pktmbuf_mtod(m, struct pcapng_enhance_packet_block *);
 		if (unlikely(epb->block_type != PCAPNG_ENHANCED_PACKET_BLOCK ||
-			     epb->block_length != rte_pktmbuf_data_len(m))) {
+			     epb->block_length != rte_pktmbuf_pkt_len(m))) {
 			rte_errno = EINVAL;
 			return -1;
 		}
--
2.46.0

[PATCH v2 2/2] test/pcapng: test chained mbufs

[Oleksandr Nahnybida]

Adjust test to check if pcapng works with chained mbufs

Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>
---
 app/test/test_pcapng.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/app/test/test_pcapng.c b/app/test/test_pcapng.c
index 2665b08c76..b219873c3a 100644
--- a/app/test/test_pcapng.c
+++ b/app/test/test_pcapng.c
@@ -102,6 +102,14 @@ mbuf1_prepare(struct dummy_mbuf *dm, uint32_t plen)
 	pkt.udp.dgram_len = rte_cpu_to_be_16(plen);
 
 	memcpy(rte_pktmbuf_mtod(dm->mb, void *), &pkt, sizeof(pkt));
+
+	/* Idea here is to create mbuf chain big enough that after mbuf deep copy they won't be
+	 * compressed into single mbuf to properly test store of chained mbufs
+	 */
+	dummy_mbuf_prep(&dm->mb[1], dm->buf[1], sizeof(dm->buf[1]), pkt_len);
+	dummy_mbuf_prep(&dm->mb[2], dm->buf[2], sizeof(dm->buf[2]), pkt_len);
+	rte_pktmbuf_chain(&dm->mb[0], &dm->mb[1]);
+	rte_pktmbuf_chain(&dm->mb[0], &dm->mb[2]);
 }
 
 static int
@@ -117,7 +125,7 @@ test_setup(void)
 
 	/* Make a pool for cloned packets */
 	mp = rte_pktmbuf_pool_create_by_ops("pcapng_test_pool",
-					    MAX_BURST, 0, 0,
+					    MAX_BURST * 32, 0, 0,
 					    rte_pcapng_mbuf_size(pkt_len) + 128,
 					    SOCKET_ID_ANY, "ring_mp_sc");
 	if (mp == NULL) {
@@ -155,7 +163,7 @@ fill_pcapng_file(rte_pcapng_t *pcapng, unsigned int num_packets)
 		for (i = 0; i < burst_size; i++) {
 			struct rte_mbuf *mc;
 
-			mc = rte_pcapng_copy(port_id, 0, orig, mp, pkt_len,
+			mc = rte_pcapng_copy(port_id, 0, orig, mp, rte_pktmbuf_pkt_len(orig),
 					     RTE_PCAPNG_DIRECTION_IN, NULL);
 			if (mc == NULL) {
 				fprintf(stderr, "Cannot copy packet\n");
-- 
2.46.0

[PATCH v2 2/2] test/pcapng: test chained mbufs

[Oleksandr Nahnybida]

Adjust test to check if pcapng works with chained mbufs

Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>
---
Resending PATCH v2 2/2 as the previous submission was incorrectly
threaded to the v1. Please ignore the earlier email

 app/test/test_pcapng.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/app/test/test_pcapng.c b/app/test/test_pcapng.c
index 2665b08c76..b219873c3a 100644
--- a/app/test/test_pcapng.c
+++ b/app/test/test_pcapng.c
@@ -102,6 +102,14 @@ mbuf1_prepare(struct dummy_mbuf *dm, uint32_t plen)
 	pkt.udp.dgram_len = rte_cpu_to_be_16(plen);
 
 	memcpy(rte_pktmbuf_mtod(dm->mb, void *), &pkt, sizeof(pkt));
+
+	/* Idea here is to create mbuf chain big enough that after mbuf deep copy they won't be
+	 * compressed into single mbuf to properly test store of chained mbufs
+	 */
+	dummy_mbuf_prep(&dm->mb[1], dm->buf[1], sizeof(dm->buf[1]), pkt_len);
+	dummy_mbuf_prep(&dm->mb[2], dm->buf[2], sizeof(dm->buf[2]), pkt_len);
+	rte_pktmbuf_chain(&dm->mb[0], &dm->mb[1]);
+	rte_pktmbuf_chain(&dm->mb[0], &dm->mb[2]);
 }
 
 static int
@@ -117,7 +125,7 @@ test_setup(void)
 
 	/* Make a pool for cloned packets */
 	mp = rte_pktmbuf_pool_create_by_ops("pcapng_test_pool",
-					    MAX_BURST, 0, 0,
+					    MAX_BURST * 32, 0, 0,
 					    rte_pcapng_mbuf_size(pkt_len) + 128,
 					    SOCKET_ID_ANY, "ring_mp_sc");
 	if (mp == NULL) {
@@ -155,7 +163,7 @@ fill_pcapng_file(rte_pcapng_t *pcapng, unsigned int num_packets)
 		for (i = 0; i < burst_size; i++) {
 			struct rte_mbuf *mc;
 
-			mc = rte_pcapng_copy(port_id, 0, orig, mp, pkt_len,
+			mc = rte_pcapng_copy(port_id, 0, orig, mp, rte_pktmbuf_pkt_len(orig),
 					     RTE_PCAPNG_DIRECTION_IN, NULL);
 			if (mc == NULL) {
 				fprintf(stderr, "Cannot copy packet\n");
-- 
2.46.0

[PATCH v2 2/2] test/pcapng: test chained mbufs

[Oleksandr Nahnybida]

Adjust test to check if pcapng works with chained mbufs

Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>
---
Sorry for the mess.
Resending PATCH v2 2/2 as the previous submission was incorrectly
threaded to the v1. Please ignore the earlier email

 app/test/test_pcapng.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/app/test/test_pcapng.c b/app/test/test_pcapng.c
index 2665b08c76..b219873c3a 100644
--- a/app/test/test_pcapng.c
+++ b/app/test/test_pcapng.c
@@ -102,6 +102,14 @@ mbuf1_prepare(struct dummy_mbuf *dm, uint32_t plen)
 	pkt.udp.dgram_len = rte_cpu_to_be_16(plen);

 	memcpy(rte_pktmbuf_mtod(dm->mb, void *), &pkt, sizeof(pkt));
+
+	/* Idea here is to create mbuf chain big enough that after mbuf deep copy they won't be
+	 * compressed into single mbuf to properly test store of chained mbufs
+	 */
+	dummy_mbuf_prep(&dm->mb[1], dm->buf[1], sizeof(dm->buf[1]), pkt_len);
+	dummy_mbuf_prep(&dm->mb[2], dm->buf[2], sizeof(dm->buf[2]), pkt_len);
+	rte_pktmbuf_chain(&dm->mb[0], &dm->mb[1]);
+	rte_pktmbuf_chain(&dm->mb[0], &dm->mb[2]);
 }

 static int
@@ -117,7 +125,7 @@ test_setup(void)

 	/* Make a pool for cloned packets */
 	mp = rte_pktmbuf_pool_create_by_ops("pcapng_test_pool",
-					    MAX_BURST, 0, 0,
+					    MAX_BURST * 32, 0, 0,
 					    rte_pcapng_mbuf_size(pkt_len) + 128,
 					    SOCKET_ID_ANY, "ring_mp_sc");
 	if (mp == NULL) {
@@ -155,7 +163,7 @@ fill_pcapng_file(rte_pcapng_t *pcapng, unsigned int num_packets)
 		for (i = 0; i < burst_size; i++) {
 			struct rte_mbuf *mc;

-			mc = rte_pcapng_copy(port_id, 0, orig, mp, pkt_len,
+			mc = rte_pcapng_copy(port_id, 0, orig, mp, rte_pktmbuf_pkt_len(orig),
 					     RTE_PCAPNG_DIRECTION_IN, NULL);
 			if (mc == NULL) {
 				fprintf(stderr, "Cannot copy packet\n");
--
2.46.0

Re: [PATCH 1/2] pcapng: fix handling of chained mbufs

[Stephen Hemminger]

On Fri, 13 Sep 2024 15:19:18 +0300
Oleksandr Nahnybida <oleksandrn@interfacemasters.com> wrote:

> The pcapng generates corrupted files when dealing with chained mbufs.
> This issue arises because in rte_pcapng_copy the length of the EPB block
> is incorrectly calculated using the data_len of the first mbuf instead
> of the pkt_len, despite that rte_pcapng_write_packets correctly writing
> the mbuf chain to disk.
> 
> This fix ensures that the block length is calculated based on the pkt_len,
> aligning it with the actual data written to disk.
> 
> Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemaster.com>


Fixes: 8d23ce8f5ee9 ("pcapng: add new library for writing pcapng files")

Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH v2 2/2] test/pcapng: test chained mbufs

[Stephen Hemminger]

On Fri, 13 Sep 2024 15:40:47 +0300
Oleksandr Nahnybida <oleksandrn@interfacemasters.com> wrote:

> Adjust test to check if pcapng works with chained mbufs
> 
> Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>

Acked-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH 1/2] pcapng: fix handling of chained mbufs

[Stephen Hemminger]

On Fri, 13 Sep 2024 15:19:18 +0300
Oleksandr Nahnybida <oleksandrn@interfacemasters.com> wrote:

> The pcapng generates corrupted files when dealing with chained mbufs.
> This issue arises because in rte_pcapng_copy the length of the EPB block
> is incorrectly calculated using the data_len of the first mbuf instead
> of the pkt_len, despite that rte_pcapng_write_packets correctly writing
> the mbuf chain to disk.
> 
> This fix ensures that the block length is calculated based on the pkt_len,
> aligning it with the actual data written to disk.
> 
> Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemaster.com>

Fixes: 8d23ce8f5ee9 ("pcapng: add new library for writing pcapng files")
Reviewed-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH v2 2/2] test/pcapng: test chained mbufs

[Stephen Hemminger]

On Fri, 13 Sep 2024 16:06:00 +0300
Oleksandr Nahnybida <oleksandrn@interfacemasters.com> wrote:

> Adjust test to check if pcapng works with chained mbufs
> 
> Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>
> ---

Next time best to use a cover-letter and in-reply-to as described
in DPDK contributors doc.

The test looks good.

Acked-by: Stephen Hemminger <stephen@networkplumber.org>

Re: [PATCH v2 2/2] test/pcapng: test chained mbufs

[David Marchand]

On Sat, Oct 5, 2024 at 6:14 PM Stephen Hemminger
<stephen@networkplumber.org> wrote:
> On Fri, 13 Sep 2024 16:06:00 +0300
> Oleksandr Nahnybida <oleksandrn@interfacemasters.com> wrote:
>
> > Adjust test to check if pcapng works with chained mbufs
> >
> > Signed-off-by: Oleksandr Nahnybida <oleksandrn@interfacemasters.com>
> Acked-by: Stephen Hemminger <stephen@networkplumber.org>

Series applied, thanks Oleksandr.


-- 
David Marchand