[bpf-next,5/8] libbpf, xsk: select AF_XDP BPF program based on kernel version

Commit Message

Björn Töpel Jan. 19, 2021, 3:36 p.m. UTC

From: Björn Töpel <[e-mail]>

Add detection for kernel version, and adapt the BPF program based on
kernel support. This way, users will get the best possible performance
from the BPF program.

Reviewed-by: Maciej Fijalkowski <[e-mail]>
Acked-by: Maciej Fijalkowski <[e-mail]>
Signed-off-by: Björn Töpel <[e-mail]>
Signed-off-by: Marek Majtyka  <[e-mail]>
---
 tools/lib/bpf/libbpf.c          |  2 +-
 tools/lib/bpf/libbpf_internal.h |  2 ++
 tools/lib/bpf/libbpf_probes.c   | 16 -------------
 tools/lib/bpf/xsk.c             | 41 ++++++++++++++++++++++++++++++---
 4 files changed, 41 insertions(+), 20 deletions(-)

Patch

diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
index 2abbc3800568..6a53adf14a9c 100644
--- a/tools/lib/bpf/libbpf.c
+++ b/tools/lib/bpf/libbpf.c
@@ -693,7 +693,7 @@  bpf_object__add_programs(struct bpf_object *obj, Elf_Data *sec_data,
 	return 0;
 }
 
-static __u32 get_kernel_version(void)
+__u32 get_kernel_version(void)
 {
 	__u32 major, minor, patch;
 	struct utsname info;
diff --git a/tools/lib/bpf/libbpf_internal.h b/tools/lib/bpf/libbpf_internal.h
index 969d0ac592ba..dafb780e2dd2 100644
--- a/tools/lib/bpf/libbpf_internal.h
+++ b/tools/lib/bpf/libbpf_internal.h
@@ -349,4 +349,6 @@  struct bpf_core_relo {
 	enum bpf_core_relo_kind kind;
 };
 
+__u32 get_kernel_version(void);
+
 #endif /* __LIBBPF_LIBBPF_INTERNAL_H */
diff --git a/tools/lib/bpf/libbpf_probes.c b/tools/lib/bpf/libbpf_probes.c
index ecaae2927ab8..aae0231371d0 100644
--- a/tools/lib/bpf/libbpf_probes.c
+++ b/tools/lib/bpf/libbpf_probes.c
@@ -48,22 +48,6 @@  static int get_vendor_id(int ifindex)
 	return strtol(buf, NULL, 0);
 }
 
-static int get_kernel_version(void)
-{
-	int version, subversion, patchlevel;
-	struct utsname utsn;
-
-	/* Return 0 on failure, and attempt to probe with empty kversion */
-	if (uname(&utsn))
-		return 0;
-
-	if (sscanf(utsn.release, "%d.%d.%d",
-		   &version, &subversion, &patchlevel) != 3)
-		return 0;
-
-	return (version << 16) + (subversion << 8) + patchlevel;
-}
-
 static void
 probe_load(enum bpf_prog_type prog_type, const struct bpf_insn *insns,
 	   size_t insns_cnt, char *buf, size_t buf_len, __u32 ifindex)
diff --git a/tools/lib/bpf/xsk.c b/tools/lib/bpf/xsk.c
index e3e41ceeb1bc..c8642c6cb5d6 100644
--- a/tools/lib/bpf/xsk.c
+++ b/tools/lib/bpf/xsk.c
@@ -23,6 +23,7 @@ 
 #include <linux/kernel.h>
 #include <linux/list.h>
 #include <linux/sockios.h>
+#include <linux/version.h>
 #include <net/if.h>
 #include <sys/ioctl.h>
 #include <sys/mman.h>
@@ -46,6 +47,11 @@ 
  #define PF_XDP AF_XDP
 #endif
 
+enum xsk_prog {
+	XSK_PROG_FALLBACK,
+	XSK_PROG_REDIRECT_FLAGS,
+};
+
 struct xsk_umem {
 	struct xsk_ring_prod *fill_save;
 	struct xsk_ring_cons *comp_save;
@@ -351,6 +357,13 @@  int xsk_umem__create_v0_0_2(struct xsk_umem **umem_ptr, void *umem_area,
 COMPAT_VERSION(xsk_umem__create_v0_0_2, xsk_umem__create, LIBBPF_0.0.2)
 DEFAULT_VERSION(xsk_umem__create_v0_0_4, xsk_umem__create, LIBBPF_0.0.4)
 
+static enum xsk_prog get_xsk_prog(void)
+{
+	__u32 kver = get_kernel_version();
+
+	return kver < KERNEL_VERSION(5, 3, 0) ? XSK_PROG_FALLBACK : XSK_PROG_REDIRECT_FLAGS;
+}
+
 static int xsk_load_xdp_prog(struct xsk_socket *xsk)
 {
 	static const int log_buf_size = 16 * 1024;
@@ -358,7 +371,7 @@  static int xsk_load_xdp_prog(struct xsk_socket *xsk)
 	char log_buf[log_buf_size];
 	int err, prog_fd;
 
-	/* This is the C-program:
+	/* This is the fallback C-program:
 	 * SEC("xdp_sock") int xdp_sock_prog(struct xdp_md *ctx)
 	 * {
 	 *     int ret, index = ctx->rx_queue_index;
@@ -414,9 +427,31 @@  static int xsk_load_xdp_prog(struct xsk_socket *xsk)
 		/* The jumps are to this instruction */
 		BPF_EXIT_INSN(),
 	};
-	size_t insns_cnt = sizeof(prog) / sizeof(struct bpf_insn);
 
-	prog_fd = bpf_load_program(BPF_PROG_TYPE_XDP, prog, insns_cnt,
+	/* This is the post-5.3 kernel C-program:
+	 * SEC("xdp_sock") int xdp_sock_prog(struct xdp_md *ctx)
+	 * {
+	 *     return bpf_redirect_map(&xsks_map, ctx->rx_queue_index, XDP_PASS);
+	 * }
+	 */
+	struct bpf_insn prog_redirect_flags[] = {
+		/* r2 = *(u32 *)(r1 + 16) */
+		BPF_LDX_MEM(BPF_W, BPF_REG_2, BPF_REG_1, 16),
+		/* r1 = xskmap[] */
+		BPF_LD_MAP_FD(BPF_REG_1, ctx->xsks_map_fd),
+		/* r3 = XDP_PASS */
+		BPF_MOV64_IMM(BPF_REG_3, 2),
+		/* call bpf_redirect_map */
+		BPF_EMIT_CALL(BPF_FUNC_redirect_map),
+		BPF_EXIT_INSN(),
+	};
+	size_t insns_cnt[] = {sizeof(prog) / sizeof(struct bpf_insn),
+			      sizeof(prog_redirect_flags) / sizeof(struct bpf_insn),
+	};
+	struct bpf_insn *progs[] = {prog, prog_redirect_flags};
+	enum xsk_prog option = get_xsk_prog();
+
+	prog_fd = bpf_load_program(BPF_PROG_TYPE_XDP, progs[option], insns_cnt[option],
 				   "LGPL-2.1 or BSD-2-Clause", 0, log_buf,
 				   log_buf_size);
 	if (prog_fd < 0) {