Skip to contentSkip to conversation

selftests: xsk: fix bpf_res cleanup test

PATCH Applied Magnus Karlsson 4 participants 4 messages Jan 25, 2022 – Jan 27, 2022

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 selftests: xsk: fix bpf_res cleanup test

From: Magnus Karlsson <[email protected]>

After commit 710ad98c363a ("veth: Do not record rx queue hint in veth_xmit"), veth no longer receives traffic on the same queue as it was sent on. This breaks the bpf_res test for the AF_XDP selftests as the socket tied to queue 1 will not receive traffic anymore. Modify the test so that two sockets are tied to queue id 0 using a shared umem instead. When killing the first socket enter the second socket into the xskmap so that traffic will flow to it. This will still test that the resources are not cleaned up until after the second socket dies, without having to rely on veth supporting rx_queue hints.

---
 tools/testing/selftests/bpf/xdpxceiver.c | 80 +++++++++++++++---------
 tools/testing/selftests/bpf/xdpxceiver.h |  2 +-
 2 files changed, 50 insertions(+), 32 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +49 -31
index ffa5502ad95e..5f8296d29e77 100644
@@ -266,22 +266,24 @@ static int xsk_configure_umem(struct xsk_umem_info *umem, void *buffer, u64 size
 }
 
 static int xsk_configure_socket(struct xsk_socket_info *xsk, struct xsk_umem_info *umem,
-				struct ifobject *ifobject, u32 qid)
+				struct ifobject *ifobject, bool shared)
 {
-	struct xsk_socket_config cfg;
+	struct xsk_socket_config cfg = {};
 	struct xsk_ring_cons *rxr;
 	struct xsk_ring_prod *txr;
 
 	xsk->umem = umem;
 	cfg.rx_size = xsk->rxqsize;
 	cfg.tx_size = XSK_RING_PROD__DEFAULT_NUM_DESCS;
-	cfg.libbpf_flags = 0;
+	cfg.libbpf_flags = XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD;
 	cfg.xdp_flags = ifobject->xdp_flags;
 	cfg.bind_flags = ifobject->bind_flags;
+	if (shared)
+		cfg.bind_flags |= XDP_SHARED_UMEM;
 
 	txr = ifobject->tx_on ? &xsk->tx : NULL;
 	rxr = ifobject->rx_on ? &xsk->rx : NULL;
-	return xsk_socket__create(&xsk->xsk, ifobject->ifname, qid, umem->umem, rxr, txr, &cfg);
+	return xsk_socket__create(&xsk->xsk, ifobject->ifname, 0, umem->umem, rxr, txr, &cfg);
 }
 
 static struct option long_options[] = {
@@ -387,7 +389,6 @@ static void __test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,
 	for (i = 0; i < MAX_INTERFACES; i++) {
 		struct ifobject *ifobj = i ? ifobj_rx : ifobj_tx;
 
-		ifobj->umem = &ifobj->umem_arr[0];
 		ifobj->xsk = &ifobj->xsk_arr[0];
 		ifobj->use_poll = false;
 		ifobj->pacing_on = true;
@@ -401,11 +402,12 @@ static void __test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,
 			ifobj->tx_on = false;
 		}
 
+		memset(ifobj->umem, 0, sizeof(*ifobj->umem));
+		ifobj->umem->num_frames = DEFAULT_UMEM_BUFFERS;
+		ifobj->umem->frame_size = XSK_UMEM__DEFAULT_FRAME_SIZE;
+
 		for (j = 0; j < MAX_SOCKETS; j++) {
-			memset(&ifobj->umem_arr[j], 0, sizeof(ifobj->umem_arr[j]));
 			memset(&ifobj->xsk_arr[j], 0, sizeof(ifobj->xsk_arr[j]));
-			ifobj->umem_arr[j].num_frames = DEFAULT_UMEM_BUFFERS;
-			ifobj->umem_arr[j].frame_size = XSK_UMEM__DEFAULT_FRAME_SIZE;
 			ifobj->xsk_arr[j].rxqsize = XSK_RING_CONS__DEFAULT_NUM_DESCS;
 		}
 	}
@@ -950,7 +952,10 @@ static void tx_stats_validate(struct ifobject *ifobject)
 
 static void thread_common_ops(struct test_spec *test, struct ifobject *ifobject)
 {
+	u64 umem_sz = ifobject->umem->num_frames * ifobject->umem->frame_size;
 	int mmap_flags = MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE;
+	int ret, ifindex;
+	void *bufs;
 	u32 i;
 
 	ifobject->ns_fd = switch_namespace(ifobject->nsname);
@@ -958,23 +963,20 @@ static void thread_common_ops(struct test_spec *test, struct ifobject *ifobject)
 	if (ifobject->umem->unaligned_mode)
 		mmap_flags |= MAP_HUGETLB;
 
-	for (i = 0; i < test->nb_sockets; i++) {
-		u64 umem_sz = ifobject->umem->num_frames * ifobject->umem->frame_size;
-		u32 ctr = 0;
-		void *bufs;
-		int ret;
+	bufs = mmap(NULL, umem_sz, PROT_READ | PROT_WRITE, mmap_flags, -1, 0);
+	if (bufs == MAP_FAILED)
+		exit_with_error(errno);
 
-		bufs = mmap(NULL, umem_sz, PROT_READ | PROT_WRITE, mmap_flags, -1, 0);
-		if (bufs == MAP_FAILED)
-			exit_with_error(errno);
+	ret = xsk_configure_umem(ifobject->umem, bufs, umem_sz);
+	if (ret)
+		exit_with_error(-ret);
 
-		ret = xsk_configure_umem(&ifobject->umem_arr[i], bufs, umem_sz);
-		if (ret)
-			exit_with_error(-ret);
+	for (i = 0; i < test->nb_sockets; i++) {
+		u32 ctr = 0;
 
 		while (ctr++ < SOCK_RECONF_CTR) {
-			ret = xsk_configure_socket(&ifobject->xsk_arr[i], &ifobject->umem_arr[i],
-						   ifobject, i);
+			ret = xsk_configure_socket(&ifobject->xsk_arr[i], ifobject->umem,
+						   ifobject, !!i);
 			if (!ret)
 				break;
 
@@ -985,8 +987,22 @@ static void thread_common_ops(struct test_spec *test, struct ifobject *ifobject)
 		}
 	}
 
-	ifobject->umem = &ifobject->umem_arr[0];
 	ifobject->xsk = &ifobject->xsk_arr[0];
+
+	if (!ifobject->rx_on)
+		return;
+
+	ifindex = if_nametoindex(ifobject->ifname);
+	if (!ifindex)
+		exit_with_error(errno);
+
+	ret = xsk_setup_xdp_prog(ifindex, &ifobject->xsk_map_fd);
+	if (ret)
+		exit_with_error(-ret);
+
+	ret = xsk_socket__update_xskmap(ifobject->xsk->xsk, ifobject->xsk_map_fd);
+	if (ret)
+		exit_with_error(-ret);
 }
 
 static void testapp_cleanup_xsk_res(struct ifobject *ifobj)
@@ -1142,14 +1158,16 @@ static void testapp_bidi(struct test_spec *test)
 
 static void swap_xsk_resources(struct ifobject *ifobj_tx, struct ifobject *ifobj_rx)
 {
+	int ret;
+
 	xsk_socket__delete(ifobj_tx->xsk->xsk);
-	xsk_umem__delete(ifobj_tx->umem->umem);
 	xsk_socket__delete(ifobj_rx->xsk->xsk);
-	xsk_umem__delete(ifobj_rx->umem->umem);
-	ifobj_tx->umem = &ifobj_tx->umem_arr[1];
 	ifobj_tx->xsk = &ifobj_tx->xsk_arr[1];
-	ifobj_rx->umem = &ifobj_rx->umem_arr[1];
 	ifobj_rx->xsk = &ifobj_rx->xsk_arr[1];
+
+	ret = xsk_socket__update_xskmap(ifobj_rx->xsk->xsk, ifobj_rx->xsk_map_fd);
+	if (ret)
+		exit_with_error(-ret);
 }
 
 static void testapp_bpf_res(struct test_spec *test)
@@ -1408,13 +1426,13 @@ static struct ifobject *ifobject_create(void)
 	if (!ifobj->xsk_arr)
 		goto out_xsk_arr;
 
-	ifobj->umem_arr = calloc(MAX_SOCKETS, sizeof(*ifobj->umem_arr));
-	if (!ifobj->umem_arr)
-		goto out_umem_arr;
+	ifobj->umem = calloc(1, sizeof(*ifobj->umem));
+	if (!ifobj->umem)
+		goto out_umem;
 
 	return ifobj;
 
-out_umem_arr:
+out_umem:
 	free(ifobj->xsk_arr);
 out_xsk_arr:
 	free(ifobj);
@@ -1423,7 +1441,7 @@ static struct ifobject *ifobject_create(void)
 
 static void ifobject_delete(struct ifobject *ifobj)
 {
-	free(ifobj->umem_arr);
+	free(ifobj->umem);
 	free(ifobj->xsk_arr);
 	free(ifobj);
 }
tools/testing/selftests/bpf/xdpxceiver.h +1 -1
index 2f705f44b748..62a3e6388632 100644
@@ -125,10 +125,10 @@ struct ifobject {
 	struct xsk_socket_info *xsk;
 	struct xsk_socket_info *xsk_arr;
 	struct xsk_umem_info *umem;
-	struct xsk_umem_info *umem_arr;
 	thread_func_t func_ptr;
 	struct pkt_stream *pkt_stream;
 	int ns_fd;
+	int xsk_map_fd;
 	u32 dst_ip;
 	u32 src_ip;
 	u32 xdp_flags;

base-commit: 74bb0f0c299cdc9c68cb3bc8f452e5812aa9eab0

Signature · 1 line
2.34.1
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 12 lines · Magnus Karlsson

From: Magnus Karlsson <[email protected]>

After commit 710ad98c363a ("veth: Do not record rx queue hint in veth_xmit"), veth no longer receives traffic on the same queue as it was sent on. This breaks the bpf_res test for the AF_XDP selftests as the socket tied to queue 1 will not receive traffic anymore. Modify the test so that two sockets are tied to queue id 0 using a shared umem instead. When killing the first socket enter the second socket into the xskmap so that traffic will flow to it. This will still test that the resources are not cleaned up until after the second socket dies, without having to rely on veth supporting rx_queue hints.

Quoted 3 lines

---

Reviewed-by on v1 1/1
#2 lore raw
In reply to Magnus Karlsson
Quoted 12 lines · Magnus Karlsson

From: Magnus Karlsson <[email protected]>

After commit 710ad98c363a ("veth: Do not record rx queue hint in veth_xmit"), veth no longer receives traffic on the same queue as it was sent on. This breaks the bpf_res test for the AF_XDP selftests as the socket tied to queue 1 will not receive traffic anymore. Modify the test so that two sockets are tied to queue id 0 using a shared umem instead. When killing the first socket enter the second socket into the xskmap so that traffic will flow to it. This will still test that the resources are not cleaned up until after the second socket dies, without having to rely on veth supporting rx_queue hints.

Quoted 2 lines

Thanks, makes sense as we discussed this internally. I didn't have a chance to apologize to the mess I introduced with the stuff that 710ad98c363a fixed, so sorry everyone, I didn't mean that :<

FWIW I think now we can drop the numtxqueues numrxqueues args from test_xsk.sh that are used when creating veth pair.

Besides that:

Quoted 200 lines
---
 tools/testing/selftests/bpf/xdpxceiver.c | 80 +++++++++++++++---------
 tools/testing/selftests/bpf/xdpxceiver.h |  2 +-
 2 files changed, 50 insertions(+), 32 deletions(-)
tools/testing/selftests/bpf/xdpxceiver.c +49 -31
index ffa5502ad95e..5f8296d29e77 100644
@@ -266,22 +266,24 @@ static int xsk_configure_umem(struct xsk_umem_info *umem, void *buffer, u64 size
 }
 
 static int xsk_configure_socket(struct xsk_socket_info *xsk, struct xsk_umem_info *umem,
-				struct ifobject *ifobject, u32 qid)
+				struct ifobject *ifobject, bool shared)
 {
-	struct xsk_socket_config cfg;
+	struct xsk_socket_config cfg = {};
 	struct xsk_ring_cons *rxr;
 	struct xsk_ring_prod *txr;
 
 	xsk->umem = umem;
 	cfg.rx_size = xsk->rxqsize;
 	cfg.tx_size = XSK_RING_PROD__DEFAULT_NUM_DESCS;
-	cfg.libbpf_flags = 0;
+	cfg.libbpf_flags = XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD;
 	cfg.xdp_flags = ifobject->xdp_flags;
 	cfg.bind_flags = ifobject->bind_flags;
+	if (shared)
+		cfg.bind_flags |= XDP_SHARED_UMEM;
 
 	txr = ifobject->tx_on ? &xsk->tx : NULL;
 	rxr = ifobject->rx_on ? &xsk->rx : NULL;
-	return xsk_socket__create(&xsk->xsk, ifobject->ifname, qid, umem->umem, rxr, txr, &cfg);
+	return xsk_socket__create(&xsk->xsk, ifobject->ifname, 0, umem->umem, rxr, txr, &cfg);
 }
 
 static struct option long_options[] = {
@@ -387,7 +389,6 @@ static void __test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,
 	for (i = 0; i < MAX_INTERFACES; i++) {
 		struct ifobject *ifobj = i ? ifobj_rx : ifobj_tx;
 
-		ifobj->umem = &ifobj->umem_arr[0];
 		ifobj->xsk = &ifobj->xsk_arr[0];
 		ifobj->use_poll = false;
 		ifobj->pacing_on = true;
@@ -401,11 +402,12 @@ static void __test_spec_init(struct test_spec *test, struct ifobject *ifobj_tx,
 			ifobj->tx_on = false;
 		}
 
+		memset(ifobj->umem, 0, sizeof(*ifobj->umem));
+		ifobj->umem->num_frames = DEFAULT_UMEM_BUFFERS;
+		ifobj->umem->frame_size = XSK_UMEM__DEFAULT_FRAME_SIZE;
+
 		for (j = 0; j < MAX_SOCKETS; j++) {
-			memset(&ifobj->umem_arr[j], 0, sizeof(ifobj->umem_arr[j]));
 			memset(&ifobj->xsk_arr[j], 0, sizeof(ifobj->xsk_arr[j]));
-			ifobj->umem_arr[j].num_frames = DEFAULT_UMEM_BUFFERS;
-			ifobj->umem_arr[j].frame_size = XSK_UMEM__DEFAULT_FRAME_SIZE;
 			ifobj->xsk_arr[j].rxqsize = XSK_RING_CONS__DEFAULT_NUM_DESCS;
 		}
 	}
@@ -950,7 +952,10 @@ static void tx_stats_validate(struct ifobject *ifobject)
 
 static void thread_common_ops(struct test_spec *test, struct ifobject *ifobject)
 {
+	u64 umem_sz = ifobject->umem->num_frames * ifobject->umem->frame_size;
 	int mmap_flags = MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE;
+	int ret, ifindex;
+	void *bufs;
 	u32 i;
 
 	ifobject->ns_fd = switch_namespace(ifobject->nsname);
@@ -958,23 +963,20 @@ static void thread_common_ops(struct test_spec *test, struct ifobject *ifobject)
 	if (ifobject->umem->unaligned_mode)
 		mmap_flags |= MAP_HUGETLB;
 
-	for (i = 0; i < test->nb_sockets; i++) {
-		u64 umem_sz = ifobject->umem->num_frames * ifobject->umem->frame_size;
-		u32 ctr = 0;
-		void *bufs;
-		int ret;
+	bufs = mmap(NULL, umem_sz, PROT_READ | PROT_WRITE, mmap_flags, -1, 0);
+	if (bufs == MAP_FAILED)
+		exit_with_error(errno);
 
-		bufs = mmap(NULL, umem_sz, PROT_READ | PROT_WRITE, mmap_flags, -1, 0);
-		if (bufs == MAP_FAILED)
-			exit_with_error(errno);
+	ret = xsk_configure_umem(ifobject->umem, bufs, umem_sz);
+	if (ret)
+		exit_with_error(-ret);
 
-		ret = xsk_configure_umem(&ifobject->umem_arr[i], bufs, umem_sz);
-		if (ret)
-			exit_with_error(-ret);
+	for (i = 0; i < test->nb_sockets; i++) {
+		u32 ctr = 0;
 
 		while (ctr++ < SOCK_RECONF_CTR) {
-			ret = xsk_configure_socket(&ifobject->xsk_arr[i], &ifobject->umem_arr[i],
-						   ifobject, i);
+			ret = xsk_configure_socket(&ifobject->xsk_arr[i], ifobject->umem,
+						   ifobject, !!i);
 			if (!ret)
 				break;
 
@@ -985,8 +987,22 @@ static void thread_common_ops(struct test_spec *test, struct ifobject *ifobject)
 		}
 	}
 
-	ifobject->umem = &ifobject->umem_arr[0];
 	ifobject->xsk = &ifobject->xsk_arr[0];
+
+	if (!ifobject->rx_on)
+		return;
+
+	ifindex = if_nametoindex(ifobject->ifname);
+	if (!ifindex)
+		exit_with_error(errno);
+
+	ret = xsk_setup_xdp_prog(ifindex, &ifobject->xsk_map_fd);
+	if (ret)
+		exit_with_error(-ret);
+
+	ret = xsk_socket__update_xskmap(ifobject->xsk->xsk, ifobject->xsk_map_fd);
+	if (ret)
+		exit_with_error(-ret);
 }
 
 static void testapp_cleanup_xsk_res(struct ifobject *ifobj)
@@ -1142,14 +1158,16 @@ static void testapp_bidi(struct test_spec *test)
 
 static void swap_xsk_resources(struct ifobject *ifobj_tx, struct ifobject *ifobj_rx)
 {
+	int ret;
+
 	xsk_socket__delete(ifobj_tx->xsk->xsk);
-	xsk_umem__delete(ifobj_tx->umem->umem);
 	xsk_socket__delete(ifobj_rx->xsk->xsk);
-	xsk_umem__delete(ifobj_rx->umem->umem);
-	ifobj_tx->umem = &ifobj_tx->umem_arr[1];
 	ifobj_tx->xsk = &ifobj_tx->xsk_arr[1];
-	ifobj_rx->umem = &ifobj_rx->umem_arr[1];
 	ifobj_rx->xsk = &ifobj_rx->xsk_arr[1];
+
+	ret = xsk_socket__update_xskmap(ifobj_rx->xsk->xsk, ifobj_rx->xsk_map_fd);
+	if (ret)
+		exit_with_error(-ret);
 }
 
 static void testapp_bpf_res(struct test_spec *test)
@@ -1408,13 +1426,13 @@ static struct ifobject *ifobject_create(void)
 	if (!ifobj->xsk_arr)
 		goto out_xsk_arr;
 
-	ifobj->umem_arr = calloc(MAX_SOCKETS, sizeof(*ifobj->umem_arr));
-	if (!ifobj->umem_arr)
-		goto out_umem_arr;
+	ifobj->umem = calloc(1, sizeof(*ifobj->umem));
+	if (!ifobj->umem)
+		goto out_umem;
 
 	return ifobj;
 
-out_umem_arr:
+out_umem:
 	free(ifobj->xsk_arr);
 out_xsk_arr:
 	free(ifobj);
@@ -1423,7 +1441,7 @@ static struct ifobject *ifobject_create(void)
 
 static void ifobject_delete(struct ifobject *ifobj)
 {
-	free(ifobj->umem_arr);
+	free(ifobj->umem);
 	free(ifobj->xsk_arr);
 	free(ifobj);
 }
tools/testing/selftests/bpf/xdpxceiver.h +1 -1
index 2f705f44b748..62a3e6388632 100644
@@ -125,10 +125,10 @@ struct ifobject {
 	struct xsk_socket_info *xsk;
 	struct xsk_socket_info *xsk_arr;
 	struct xsk_umem_info *umem;
-	struct xsk_umem_info *umem_arr;
 	thread_func_t func_ptr;
 	struct pkt_stream *pkt_stream;
 	int ns_fd;
+	int xsk_map_fd;
 	u32 dst_ip;
 	u32 src_ip;
 	u32 xdp_flags;

base-commit: 74bb0f0c299cdc9c68cb3bc8f452e5812aa9eab0

Signature · 1 line
2.34.1
#3 lore raw
In reply to Magnus Karlsson
This patch was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:

Hello:

This patch was applied to bpf/bpf-next.git (master) by Daniel Borkmann <[email protected]>:

Quoted 12 lines · On Tue, 25 Jan 2022 09:29:45 +0100 you wrote:

From: Magnus Karlsson <[email protected]>

After commit 710ad98c363a ("veth: Do not record rx queue hint in veth_xmit"), veth no longer receives traffic on the same queue as it was sent on. This breaks the bpf_res test for the AF_XDP selftests as the socket tied to queue 1 will not receive traffic anymore. Modify the test so that two sockets are tied to queue id 0 using a shared umem instead. When killing the first socket enter the second socket into the xskmap so that traffic will flow to it. This will still test that the resources are not cleaned up until after the second socket dies, without having to rely on veth supporting rx_queue hints.

Quoted 1 line

[...]

Here is the summary with links:
- [bpf-next] selftests: xsk: fix bpf_res cleanup test

    https://git.kernel.org/bpf/bpf-next/c/3b22523bca02

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html