Skip to contentSkip to conversation

bpf: Allow bpf_map_lookup_elem() on an xskmap

PATCH Superseded Jonathan Lemon 3 participants 7 messages May 30, 2019 – May 31, 2019

Subsystems (8)

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/2 bpf: Allow bpf_map_lookup_elem() on an xskmap

Currently, the AF_XDP code uses a separate map in order to determine if an xsk is bound to a queue. Instead of doing this, have bpf_map_lookup_elem() return the queue_id, as a way of indicating that there is a valid entry at the map index.

Rearrange some xdp_sock members to eliminate structure holes.

---
 include/net/xdp_sock.h                            |  6 +++---
 kernel/bpf/verifier.c                             |  6 +++++-
 kernel/bpf/xskmap.c                               |  4 +++-
 .../selftests/bpf/verifier/prevent_map_lookup.c   | 15 ---------------
 4 files changed, 11 insertions(+), 20 deletions(-)
include/net/xdp_sock.h +3 -3
index d074b6d60f8a..7d84b1da43d2 100644
@@ -57,12 +57,12 @@ struct xdp_sock {
 	struct net_device *dev;
 	struct xdp_umem *umem;
 	struct list_head flush_node;
-	u16 queue_id;
-	struct xsk_queue *tx ____cacheline_aligned_in_smp;
-	struct list_head list;
+	u32 queue_id;
 	bool zc;
 	/* Protects multiple processes in the control path */
 	struct mutex mutex;
+	struct xsk_queue *tx ____cacheline_aligned_in_smp;
+	struct list_head list;
 	/* Mutual exclusion of NAPI TX thread and sendmsg error paths
 	 * in the SKB destructor callback.
 	 */
kernel/bpf/verifier.c +5 -1
index 2778417e6e0c..91c730f85e92 100644
@@ -2905,10 +2905,14 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env,
 	 * appear.
 	 */
 	case BPF_MAP_TYPE_CPUMAP:
-	case BPF_MAP_TYPE_XSKMAP:
 		if (func_id != BPF_FUNC_redirect_map)
 			goto error;
 		break;
+	case BPF_MAP_TYPE_XSKMAP:
+		if (func_id != BPF_FUNC_redirect_map &&
+		    func_id != BPF_FUNC_map_lookup_elem)
+			goto error;
+		break;
 	case BPF_MAP_TYPE_ARRAY_OF_MAPS:
 	case BPF_MAP_TYPE_HASH_OF_MAPS:
 		if (func_id != BPF_FUNC_map_lookup_elem)
kernel/bpf/xskmap.c +3 -1
index 686d244e798d..249b22089014 100644
@@ -154,7 +154,9 @@ void __xsk_map_flush(struct bpf_map *map)
 
 static void *xsk_map_lookup_elem(struct bpf_map *map, void *key)
 {
-	return ERR_PTR(-EOPNOTSUPP);
+	struct xdp_sock *xs = __xsk_map_lookup_elem(map, *(u32 *)key);
+
+	return xs ? &xs->queue_id : NULL;
 }
 
 static int xsk_map_update_elem(struct bpf_map *map, void *key, void *value,
tools/testing/selftests/bpf/verifier/prevent_map_lookup.c +0 -15
index bbdba990fefb..da7a4b37cb98 100644
@@ -28,21 +28,6 @@
 	.errstr = "cannot pass map_type 18 into func bpf_map_lookup_elem",
 	.prog_type = BPF_PROG_TYPE_SOCK_OPS,
 },
-{
-	"prevent map lookup in xskmap",
-	.insns = {
-	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
-	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
-	BPF_LD_MAP_FD(BPF_REG_1, 0),
-	BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
-	BPF_EXIT_INSN(),
-	},
-	.fixup_map_xskmap = { 3 },
-	.result = REJECT,
-	.errstr = "cannot pass map_type 17 into func bpf_map_lookup_elem",
-	.prog_type = BPF_PROG_TYPE_XDP,
-},
 {
 	"prevent map lookup in stack trace",
 	.insns = {
Signature · 1 line
2.17.1
#1 lore raw

PATCH v2 2/2 libbpf: remove qidconf and better support external bpf programs.

Use the recent change to XSKMAP bpf_map_lookup_elem() to test if there is a xsk present in the map instead of duplicating the work with qidconf.

Fix things so callers using XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD bypass any internal bpf maps, so xsk_socket__{create|delete} works properly.

---
 tools/lib/bpf/xsk.c | 79 +++++++++------------------------------------
 1 file changed, 16 insertions(+), 63 deletions(-)
tools/lib/bpf/xsk.c +16 -63
index 38667b62f1fe..a150493d51ec 100644
@@ -60,10 +60,8 @@ struct xsk_socket {
 	struct xsk_umem *umem;
 	struct xsk_socket_config config;
 	int fd;
-	int xsks_map;
 	int ifindex;
 	int prog_fd;
-	int qidconf_map_fd;
 	int xsks_map_fd;
 	__u32 queue_id;
 	char ifname[IFNAMSIZ];
@@ -265,15 +263,11 @@ static int xsk_load_xdp_prog(struct xsk_socket *xsk)
 	/* This is the C-program:
 	 * SEC("xdp_sock") int xdp_sock_prog(struct xdp_md *ctx)
 	 * {
-	 *     int *qidconf, index = ctx->rx_queue_index;
+	 *     int index = ctx->rx_queue_index;
 	 *
 	 *     // A set entry here means that the correspnding queue_id
 	 *     // has an active AF_XDP socket bound to it.
-	 *     qidconf = bpf_map_lookup_elem(&qidconf_map, &index);
-	 *     if (!qidconf)
-	 *         return XDP_ABORTED;
-	 *
-	 *     if (*qidconf)
+	 *     if (bpf_map_lookup_elem(&xsks_map, &index))
 	 *         return bpf_redirect_map(&xsks_map, index, 0);
 	 *
 	 *     return XDP_PASS;
@@ -286,15 +280,10 @@ static int xsk_load_xdp_prog(struct xsk_socket *xsk)
 		BPF_STX_MEM(BPF_W, BPF_REG_10, BPF_REG_1, -4),
 		BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
 		BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-		BPF_LD_MAP_FD(BPF_REG_1, xsk->qidconf_map_fd),
+		BPF_LD_MAP_FD(BPF_REG_1, xsk->xsks_map_fd),
 		BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
 		BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-		BPF_MOV32_IMM(BPF_REG_0, 0),
-		/* if r1 == 0 goto +8 */
-		BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 8),
 		BPF_MOV32_IMM(BPF_REG_0, 2),
-		/* r1 = *(u32 *)(r1 + 0) */
-		BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_1, 0),
 		/* if r1 == 0 goto +5 */
 		BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 5),
 		/* r2 = *(u32 *)(r10 - 4) */
@@ -366,18 +355,11 @@ static int xsk_create_bpf_maps(struct xsk_socket *xsk)
 	if (max_queues < 0)
 		return max_queues;
 
-	fd = bpf_create_map_name(BPF_MAP_TYPE_ARRAY, "qidconf_map",
+	fd = bpf_create_map_name(BPF_MAP_TYPE_ARRAY, "xsks_map",
 				 sizeof(int), sizeof(int), max_queues, 0);
 	if (fd < 0)
 		return fd;
-	xsk->qidconf_map_fd = fd;
 
-	fd = bpf_create_map_name(BPF_MAP_TYPE_XSKMAP, "xsks_map",
-				 sizeof(int), sizeof(int), max_queues, 0);
-	if (fd < 0) {
-		close(xsk->qidconf_map_fd);
-		return fd;
-	}
 	xsk->xsks_map_fd = fd;
 
 	return 0;
@@ -385,10 +367,8 @@ static int xsk_create_bpf_maps(struct xsk_socket *xsk)
 
 static void xsk_delete_bpf_maps(struct xsk_socket *xsk)
 {
-	close(xsk->qidconf_map_fd);
+	bpf_map_delete_elem(xsk->xsks_map_fd, &xsk->queue_id);
 	close(xsk->xsks_map_fd);
-	xsk->qidconf_map_fd = -1;
-	xsk->xsks_map_fd = -1;
 }
 
 static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
@@ -417,10 +397,9 @@ static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
 	if (err)
 		goto out_map_ids;
 
-	for (i = 0; i < prog_info.nr_map_ids; i++) {
-		if (xsk->qidconf_map_fd != -1 && xsk->xsks_map_fd != -1)
-			break;
+	xsk->xsks_map_fd = -1;
 
+	for (i = 0; i < prog_info.nr_map_ids; i++) {
 		fd = bpf_map_get_fd_by_id(map_ids[i]);
 		if (fd < 0)
 			continue;
@@ -431,11 +410,6 @@ static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
 			continue;
 		}
 
-		if (!strcmp(map_info.name, "qidconf_map")) {
-			xsk->qidconf_map_fd = fd;
-			continue;
-		}
-
 		if (!strcmp(map_info.name, "xsks_map")) {
 			xsk->xsks_map_fd = fd;
 			continue;
@@ -445,40 +419,18 @@ static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
 	}
 
 	err = 0;
-	if (xsk->qidconf_map_fd < 0 || xsk->xsks_map_fd < 0) {
+	if (xsk->xsks_map_fd == -1)
 		err = -ENOENT;
-		xsk_delete_bpf_maps(xsk);
-	}
 
 out_map_ids:
 	free(map_ids);
 	return err;
 }
 
-static void xsk_clear_bpf_maps(struct xsk_socket *xsk)
-{
-	int qid = false;
-
-	bpf_map_update_elem(xsk->qidconf_map_fd, &xsk->queue_id, &qid, 0);
-	bpf_map_delete_elem(xsk->xsks_map_fd, &xsk->queue_id);
-}
-
 static int xsk_set_bpf_maps(struct xsk_socket *xsk)
 {
-	int qid = true, fd = xsk->fd, err;
-
-	err = bpf_map_update_elem(xsk->qidconf_map_fd, &xsk->queue_id, &qid, 0);
-	if (err)
-		goto out;
-
-	err = bpf_map_update_elem(xsk->xsks_map_fd, &xsk->queue_id, &fd, 0);
-	if (err)
-		goto out;
-
-	return 0;
-out:
-	xsk_clear_bpf_maps(xsk);
-	return err;
+	return bpf_map_update_elem(xsk->xsks_map_fd, &xsk->queue_id,
+				   &xsk->fd, 0);
 }
 
 static int xsk_setup_xdp_prog(struct xsk_socket *xsk)
@@ -514,6 +466,7 @@ static int xsk_setup_xdp_prog(struct xsk_socket *xsk)
 
 out_load:
 	close(xsk->prog_fd);
+	xsk->prog_fd = -1;
 out_maps:
 	xsk_delete_bpf_maps(xsk);
 	return err;
@@ -643,9 +596,7 @@ int xsk_socket__create(struct xsk_socket **xsk_ptr, const char *ifname,
 		goto out_mmap_tx;
 	}
 
-	xsk->qidconf_map_fd = -1;
-	xsk->xsks_map_fd = -1;
-
+	xsk->prog_fd = -1;
 	if (!(xsk->config.libbpf_flags & XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD)) {
 		err = xsk_setup_xdp_prog(xsk);
 		if (err)
@@ -708,8 +659,10 @@ void xsk_socket__delete(struct xsk_socket *xsk)
 	if (!xsk)
 		return;
 
-	xsk_clear_bpf_maps(xsk);
-	xsk_delete_bpf_maps(xsk);
+	if (xsk->prog_fd != -1) {
+		xsk_delete_bpf_maps(xsk);
+		close(xsk->prog_fd);
+	}
 
 	optlen = sizeof(off);
 	err = getsockopt(xsk->fd, SOL_XDP, XDP_MMAP_OFFSETS, &off, &optlen);
Signature · 1 line
2.17.1
#2 lore raw

Re: [PATCH v2 bpf-next 2/2] libbpf: remove qidconf and better support external bpf programs.

Quoted 70 lines · Jonathan Lemon

Use the recent change to XSKMAP bpf_map_lookup_elem() to test if there is a xsk present in the map instead of duplicating the work with qidconf.

Fix things so callers using XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD bypass any internal bpf maps, so xsk_socket__{create|delete} works properly.

---
  tools/lib/bpf/xsk.c | 79 +++++++++------------------------------------
  1 file changed, 16 insertions(+), 63 deletions(-)
tools/lib/bpf/xsk.c +4 -15
index 38667b62f1fe..a150493d51ec 100644
@@ -60,10 +60,8 @@ struct xsk_socket {
  	struct xsk_umem *umem;
  	struct xsk_socket_config config;
  	int fd;
-	int xsks_map;
  	int ifindex;
  	int prog_fd;
-	int qidconf_map_fd;
  	int xsks_map_fd;
  	__u32 queue_id;
  	char ifname[IFNAMSIZ];
@@ -265,15 +263,11 @@ static int xsk_load_xdp_prog(struct xsk_socket *xsk)
  	/* This is the C-program:
  	 * SEC("xdp_sock") int xdp_sock_prog(struct xdp_md *ctx)
  	 * {
-	 *     int *qidconf, index = ctx->rx_queue_index;
+	 *     int index = ctx->rx_queue_index;
  	 *
  	 *     // A set entry here means that the correspnding queue_id
  	 *     // has an active AF_XDP socket bound to it.
-	 *     qidconf = bpf_map_lookup_elem(&qidconf_map, &index);
-	 *     if (!qidconf)
-	 *         return XDP_ABORTED;
-	 *
-	 *     if (*qidconf)
+	 *     if (bpf_map_lookup_elem(&xsks_map, &index))
  	 *         return bpf_redirect_map(&xsks_map, index, 0);
  	 *
  	 *     return XDP_PASS;
@@ -286,15 +280,10 @@ static int xsk_load_xdp_prog(struct xsk_socket *xsk)
  		BPF_STX_MEM(BPF_W, BPF_REG_10, BPF_REG_1, -4),
  		BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
  		BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-		BPF_LD_MAP_FD(BPF_REG_1, xsk->qidconf_map_fd),
+		BPF_LD_MAP_FD(BPF_REG_1, xsk->xsks_map_fd),
  		BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
  		BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-		BPF_MOV32_IMM(BPF_REG_0, 0),
-		/* if r1 == 0 goto +8 */
-		BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 8),
  		BPF_MOV32_IMM(BPF_REG_0, 2),
-		/* r1 = *(u32 *)(r1 + 0) */
-		BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_1, 0),
  		/* if r1 == 0 goto +5 */
  		BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 5),
  		/* r2 = *(u32 *)(r10 - 4) */
@@ -366,18 +355,11 @@ static int xsk_create_bpf_maps(struct xsk_socket *xsk)
  	if (max_queues < 0)
  		return max_queues;
 
-	fd = bpf_create_map_name(BPF_MAP_TYPE_ARRAY, "qidconf_map",
+	fd = bpf_create_map_name(BPF_MAP_TYPE_ARRAY, "xsks_map",
  				 sizeof(int), sizeof(int), max_queues, 0);

tools/lib/bpf/xsk.c static int xsk_create_bpf_maps(struct xsk_socket *xsk) · line 380

Quoted 10 lines
tools/lib/bpf/xsk.c +0 -7
  	if (fd < 0)
  		return fd;
-	xsk->qidconf_map_fd = fd;
 
-	fd = bpf_create_map_name(BPF_MAP_TYPE_XSKMAP, "xsks_map",
-				 sizeof(int), sizeof(int), max_queues, 0);
-	if (fd < 0) {
-		close(xsk->qidconf_map_fd);
-		return fd;
-	}

Uhm, you're removing the XSKMAP here, replacing it with an ARRAY. Have you run this?

Björn

Quoted 117 lines
tools/lib/bpf/xsk.c +12 -41
  	xsk->xsks_map_fd = fd;
 
  	return 0;
@@ -385,10 +367,8 @@ static int xsk_create_bpf_maps(struct xsk_socket *xsk)
 
  static void xsk_delete_bpf_maps(struct xsk_socket *xsk)
  {
-	close(xsk->qidconf_map_fd);
+	bpf_map_delete_elem(xsk->xsks_map_fd, &xsk->queue_id);
  	close(xsk->xsks_map_fd);
-	xsk->qidconf_map_fd = -1;
-	xsk->xsks_map_fd = -1;
  }
 
  static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
@@ -417,10 +397,9 @@ static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
  	if (err)
  		goto out_map_ids;
 
-	for (i = 0; i < prog_info.nr_map_ids; i++) {
-		if (xsk->qidconf_map_fd != -1 && xsk->xsks_map_fd != -1)
-			break;
+	xsk->xsks_map_fd = -1;
 
+	for (i = 0; i < prog_info.nr_map_ids; i++) {
  		fd = bpf_map_get_fd_by_id(map_ids[i]);
  		if (fd < 0)
  			continue;
@@ -431,11 +410,6 @@ static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
  			continue;
  		}
 
-		if (!strcmp(map_info.name, "qidconf_map")) {
-			xsk->qidconf_map_fd = fd;
-			continue;
-		}
-
  		if (!strcmp(map_info.name, "xsks_map")) {
  			xsk->xsks_map_fd = fd;
  			continue;
@@ -445,40 +419,18 @@ static int xsk_lookup_bpf_maps(struct xsk_socket *xsk)
  	}
 
  	err = 0;
-	if (xsk->qidconf_map_fd < 0 || xsk->xsks_map_fd < 0) {
+	if (xsk->xsks_map_fd == -1)
  		err = -ENOENT;
-		xsk_delete_bpf_maps(xsk);
-	}
 
  out_map_ids:
  	free(map_ids);
  	return err;
  }
 
-static void xsk_clear_bpf_maps(struct xsk_socket *xsk)
-{
-	int qid = false;
-
-	bpf_map_update_elem(xsk->qidconf_map_fd, &xsk->queue_id, &qid, 0);
-	bpf_map_delete_elem(xsk->xsks_map_fd, &xsk->queue_id);
-}
-
  static int xsk_set_bpf_maps(struct xsk_socket *xsk)
  {
-	int qid = true, fd = xsk->fd, err;
-
-	err = bpf_map_update_elem(xsk->qidconf_map_fd, &xsk->queue_id, &qid, 0);
-	if (err)
-		goto out;
-
-	err = bpf_map_update_elem(xsk->xsks_map_fd, &xsk->queue_id, &fd, 0);
-	if (err)
-		goto out;
-
-	return 0;
-out:
-	xsk_clear_bpf_maps(xsk);
-	return err;
+	return bpf_map_update_elem(xsk->xsks_map_fd, &xsk->queue_id,
+				   &xsk->fd, 0);
  }
 
  static int xsk_setup_xdp_prog(struct xsk_socket *xsk)
@@ -514,6 +466,7 @@ static int xsk_setup_xdp_prog(struct xsk_socket *xsk)
 
  out_load:
  	close(xsk->prog_fd);
+	xsk->prog_fd = -1;
  out_maps:
  	xsk_delete_bpf_maps(xsk);
  	return err;
@@ -643,9 +596,7 @@ int xsk_socket__create(struct xsk_socket **xsk_ptr, const char *ifname,
  		goto out_mmap_tx;
  	}
 
-	xsk->qidconf_map_fd = -1;
-	xsk->xsks_map_fd = -1;
-
+	xsk->prog_fd = -1;
  	if (!(xsk->config.libbpf_flags & XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD)) {
  		err = xsk_setup_xdp_prog(xsk);
  		if (err)
@@ -708,8 +659,10 @@ void xsk_socket__delete(struct xsk_socket *xsk)
  	if (!xsk)
  		return;
 
-	xsk_clear_bpf_maps(xsk);
-	xsk_delete_bpf_maps(xsk);
+	if (xsk->prog_fd != -1) {
+		xsk_delete_bpf_maps(xsk);
+		close(xsk->prog_fd);
+	}
 
  	optlen = sizeof(off);
  	err = getsockopt(xsk->fd, SOL_XDP, XDP_MMAP_OFFSETS, &off, &optlen);
#3 lore raw

Re: [PATCH v2 bpf-next 2/2] libbpf: remove qidconf and better support external bpf programs.

Quoted 78 lines · Björn Töpel
Quoted 77 lines · Jonathan Lemon

Use the recent change to XSKMAP bpf_map_lookup_elem() to test if there is a xsk present in the map instead of duplicating the work with qidconf.

Fix things so callers using XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD bypass any internal bpf maps, so xsk_socket__{create|delete} works properly.

---
  tools/lib/bpf/xsk.c | 79 
+++++++++------------------------------------
  1 file changed, 16 insertions(+), 63 deletions(-)
tools/lib/bpf/xsk.c +2 -8
index 38667b62f1fe..a150493d51ec 100644
@@ -60,10 +60,8 @@ struct xsk_socket {
  	struct xsk_umem *umem;
  	struct xsk_socket_config config;
  	int fd;
-	int xsks_map;
  	int ifindex;
  	int prog_fd;
-	int qidconf_map_fd;
  	int xsks_map_fd;
  	__u32 queue_id;
  	char ifname[IFNAMSIZ];
@@ -265,15 +263,11 @@ static int xsk_load_xdp_prog(struct xsk_socket 
 *xsk)
  	/* This is the C-program:
  	 * SEC("xdp_sock") int xdp_sock_prog(struct xdp_md *ctx)
  	 * {
-	 *     int *qidconf, index = ctx->rx_queue_index;
+	 *     int index = ctx->rx_queue_index;
  	 *
  	 *     // A set entry here means that the correspnding queue_id
  	 *     // has an active AF_XDP socket bound to it.
-	 *     qidconf = bpf_map_lookup_elem(&qidconf_map, &index);
-	 *     if (!qidconf)
-	 *         return XDP_ABORTED;
-	 *
-	 *     if (*qidconf)
+	 *     if (bpf_map_lookup_elem(&xsks_map, &index))
  	 *         return bpf_redirect_map(&xsks_map, index, 0);
  	 *

* return XDP_PASS;

tools/lib/bpf/xsk.c +1 -6
@@ -286,15 +280,10 @@ static int xsk_load_xdp_prog(struct xsk_socket 
 *xsk)
  		BPF_STX_MEM(BPF_W, BPF_REG_10, BPF_REG_1, -4),
  		BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
  		BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -4),
-		BPF_LD_MAP_FD(BPF_REG_1, xsk->qidconf_map_fd),
+		BPF_LD_MAP_FD(BPF_REG_1, xsk->xsks_map_fd),
  		BPF_EMIT_CALL(BPF_FUNC_map_lookup_elem),
  		BPF_MOV64_REG(BPF_REG_1, BPF_REG_0),
-		BPF_MOV32_IMM(BPF_REG_0, 0),
-		/* if r1 == 0 goto +8 */
-		BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 8),
  		BPF_MOV32_IMM(BPF_REG_0, 2),
-		/* r1 = *(u32 *)(r1 + 0) */
-		BPF_LDX_MEM(BPF_W, BPF_REG_1, BPF_REG_1, 0),
  		/* if r1 == 0 goto +5 */
  		BPF_JMP_IMM(BPF_JEQ, BPF_REG_1, 0, 5),

/* r2 = *(u32 *)(r10 - 4) */

tools/lib/bpf/xsk.c +1 -1
@@ -366,18 +355,11 @@ static int xsk_create_bpf_maps(struct 
 xsk_socket *xsk)
  	if (max_queues < 0)
  		return max_queues;
 -	fd = bpf_create_map_name(BPF_MAP_TYPE_ARRAY, "qidconf_map",
+	fd = bpf_create_map_name(BPF_MAP_TYPE_ARRAY, "xsks_map",
  				 sizeof(int), sizeof(int), max_queues, 0);
  	if (fd < 0)
  		return fd;
-	xsk->qidconf_map_fd = fd;
 -	fd = bpf_create_map_name(BPF_MAP_TYPE_XSKMAP, "xsks_map",
Quoted 8 lines

tools/lib/bpf/xsk.c static int xsk_create_bpf_maps(struct xsk_socket *xsk) · line 380

Quoted 5 lines
tools/lib/bpf/xsk.c +0 -5
-				 sizeof(int), sizeof(int), max_queues, 0);
-	if (fd < 0) {
-		close(xsk->qidconf_map_fd);
-		return fd;
-	}

Uhm, you're removing the XSKMAP here, replacing it with an ARRAY. Have you run this?

Err... I've been running the code, but this version is wrong. Let me respin this.

Signature · 1 line
Jonathan
Acked-by on v2 1/2
#4 lore raw
In reply to Jonathan Lemon
Quoted 20 lines · Jonathan Lemon

Currently, the AF_XDP code uses a separate map in order to determine if an xsk is bound to a queue. Instead of doing this, have bpf_map_lookup_elem() return the queue_id, as a way of indicating that there is a valid entry at the map index.

Rearrange some xdp_sock members to eliminate structure holes.

---
  include/net/xdp_sock.h                            |  6 +++---
  kernel/bpf/verifier.c                             |  6 +++++-
  kernel/bpf/xskmap.c                               |  4 +++-
  .../selftests/bpf/verifier/prevent_map_lookup.c   | 15 ---------------
  4 files changed, 11 insertions(+), 20 deletions(-)
include/net/xdp_sock.h +0 -0
index d074b6d60f8a..7d84b1da43d2 100644

include/net/xdp_sock.h struct xdp_sock { · line 60

Quoted 7 lines
include/net/xdp_sock.h +1 -3
@@ -57,12 +57,12 @@ struct xdp_sock {
  	struct net_device *dev;
  	struct xdp_umem *umem;
  	struct list_head flush_node;
-	u16 queue_id;
-	struct xsk_queue *tx ____cacheline_aligned_in_smp;
-	struct list_head list;
+	u32 queue_id;

Why the increase of size?

Quoted 33 lines
include/net/xdp_sock.h +2 -0
  	bool zc;
  	/* Protects multiple processes in the control path */
  	struct mutex mutex;
+	struct xsk_queue *tx ____cacheline_aligned_in_smp;
+	struct list_head list;
  	/* Mutual exclusion of NAPI TX thread and sendmsg error paths
  	 * in the SKB destructor callback.
  	 */
kernel/bpf/verifier.c +5 -1
index 2778417e6e0c..91c730f85e92 100644
@@ -2905,10 +2905,14 @@ static int check_map_func_compatibility(struct bpf_verifier_env *env,
  	 * appear.
  	 */
  	case BPF_MAP_TYPE_CPUMAP:
-	case BPF_MAP_TYPE_XSKMAP:
  		if (func_id != BPF_FUNC_redirect_map)
  			goto error;
  		break;
+	case BPF_MAP_TYPE_XSKMAP:
+		if (func_id != BPF_FUNC_redirect_map &&
+		    func_id != BPF_FUNC_map_lookup_elem)
+			goto error;
+		break;
  	case BPF_MAP_TYPE_ARRAY_OF_MAPS:
  	case BPF_MAP_TYPE_HASH_OF_MAPS:
  		if (func_id != BPF_FUNC_map_lookup_elem)
kernel/bpf/xskmap.c +0 -0
index 686d244e798d..249b22089014 100644

kernel/bpf/xskmap.c void __xsk_map_flush(struct bpf_map *map) · line 159

Quoted 7 lines
kernel/bpf/xskmap.c +3 -1
@@ -154,7 +154,9 @@ void __xsk_map_flush(struct bpf_map *map)
 
  static void *xsk_map_lookup_elem(struct bpf_map *map, void *key)
  {
-	return ERR_PTR(-EOPNOTSUPP);
+	struct xdp_sock *xs = __xsk_map_lookup_elem(map, *(u32 *)key);
+
+	return xs ? &xs->queue_id : NULL;

I like this! I haven't taken it for a spin, but still:

Quoted 30 lines
kernel/bpf/xskmap.c +0 -0
  }
 
  static int xsk_map_update_elem(struct bpf_map *map, void *key, void *value,
tools/testing/selftests/bpf/verifier/prevent_map_lookup.c +0 -15
index bbdba990fefb..da7a4b37cb98 100644
@@ -28,21 +28,6 @@
  	.errstr = "cannot pass map_type 18 into func bpf_map_lookup_elem",
  	.prog_type = BPF_PROG_TYPE_SOCK_OPS,
  },
-{
-	"prevent map lookup in xskmap",
-	.insns = {
-	BPF_ST_MEM(BPF_DW, BPF_REG_10, -8, 0),
-	BPF_MOV64_REG(BPF_REG_2, BPF_REG_10),
-	BPF_ALU64_IMM(BPF_ADD, BPF_REG_2, -8),
-	BPF_LD_MAP_FD(BPF_REG_1, 0),
-	BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_map_lookup_elem),
-	BPF_EXIT_INSN(),
-	},
-	.fixup_map_xskmap = { 3 },
-	.result = REJECT,
-	.errstr = "cannot pass map_type 17 into func bpf_map_lookup_elem",
-	.prog_type = BPF_PROG_TYPE_XDP,
-},
  {
  	"prevent map lookup in stack trace",
  	.insns = {
#5 lore raw
Quoted 16 lines · Björn Töpel
Quoted 15 lines · Jonathan Lemon

Currently, the AF_XDP code uses a separate map in order to determine if an xsk is bound to a queue. Instead of doing this, have bpf_map_lookup_elem() return the queue_id, as a way of indicating that there is a valid entry at the map index.

Rearrange some xdp_sock members to eliminate structure holes.

---
  include/net/xdp_sock.h                            |  6 +++---
  kernel/bpf/verifier.c                             |  6 +++++-
  kernel/bpf/xskmap.c                               |  4 +++-
  .../selftests/bpf/verifier/prevent_map_lookup.c   | 15 ---------------
  4 files changed, 11 insertions(+), 20 deletions(-)
Quoted 14 lines

include/net/xdp_sock.h struct xdp_sock { · line 60

Quoted 12 lines
include/net/xdp_sock.h +1 -3
index d074b6d60f8a..7d84b1da43d2 100644
@@ -57,12 +57,12 @@ struct xdp_sock {
  	struct net_device *dev;
  	struct xdp_umem *umem;
  	struct list_head flush_node;
-	u16 queue_id;
-	struct xsk_queue *tx ____cacheline_aligned_in_smp;
-	struct list_head list;
+	u32 queue_id;

Why the increase of size?

Currently xskmaps are defined as size=4, so the returned pointer for bpf must have that many bytes. I was kicking around the idea of using a union and returning { qid, zc }, but Alexei pointed out that doesn't make a good API.

Besides, queue_index in struct xdp_rxq_info is already a u32.

Signature · 1 line
Jonathan
#6 lore raw
Quoted 34 lines · Jonathan Lemon
Quoted 16 lines · Björn Töpel
Quoted 15 lines · Jonathan Lemon

Currently, the AF_XDP code uses a separate map in order to determine if an xsk is bound to a queue. Instead of doing this, have bpf_map_lookup_elem() return the queue_id, as a way of indicating that there is a valid entry at the map index.

Rearrange some xdp_sock members to eliminate structure holes.

---
  include/net/xdp_sock.h                            |  6 +++---
  kernel/bpf/verifier.c                             |  6 +++++-
  kernel/bpf/xskmap.c                               |  4 +++-
  .../selftests/bpf/verifier/prevent_map_lookup.c   | 15 ---------------
  4 files changed, 11 insertions(+), 20 deletions(-)
Quoted 14 lines

include/net/xdp_sock.h struct xdp_sock { · line 60

Quoted 12 lines
include/net/xdp_sock.h +1 -3
index d074b6d60f8a..7d84b1da43d2 100644
@@ -57,12 +57,12 @@ struct xdp_sock {
     struct net_device *dev;
     struct xdp_umem *umem;
     struct list_head flush_node;
-    u16 queue_id;
-    struct xsk_queue *tx ____cacheline_aligned_in_smp;
-    struct list_head list;
+    u32 queue_id;

Why the increase of size?

Quoted 6 lines

Currently xskmaps are defined as size=4, so the returned pointer for bpf must have that many bytes. I was kicking around the idea of using a union and returning { qid, zc }, but Alexei pointed out that doesn't make a good API.

Besides, queue_index in struct xdp_rxq_info is already a u32.

(Resending w/o HTML from mobile client.)

Makes sense. Thanks for clarifying!

Björn

Quoted 3 lines

--
Jonathan