Re: [iproute2 net-next 1/8] lib bpf: Add support for BPF_PROG_ATTACH and BPF_PROG_DETACH

[Daniel Borkmann]

On 12/10/2016 10:16 PM, Daniel Borkmann wrote:

On 12/10/2016 09:32 PM, David Ahern wrote:

For consistency with other bpf commands, the functions are named
bpf_prog_attach and bpf_prog_detach. The existing bpf_prog_attach is
renamed to bpf_prog_load_and_report since it calls bpf_prog_load and
bpf_prog_report.

Signed-off-by: David Ahern <d...@cumulusnetworks.com>
---
  include/bpf_util.h |  3 +++
  lib/bpf.c          | 31 ++++++++++++++++++++++++++-----
  2 files changed, 29 insertions(+), 5 deletions(-)

diff --git a/include/bpf_util.h b/include/bpf_util.h
index 05baeecda57f..49b96bbc208f 100644
--- a/include/bpf_util.h
+++ b/include/bpf_util.h
@@ -75,6 +75,9 @@ int bpf_trace_pipe(void);

  void bpf_print_ops(FILE *f, struct rtattr *bpf_ops, __u16 len);

+int bpf_prog_attach(int prog_fd, int target_fd, enum bpf_attach_type type);
+int bpf_prog_detach(int target_fd, enum bpf_attach_type type);
+
  #ifdef HAVE_ELF
  int bpf_send_map_fds(const char *path, const char *obj);
  int bpf_recv_map_fds(const char *path, int *fds, struct bpf_map_aux *aux,
diff --git a/lib/bpf.c b/lib/bpf.c
index 2a8cd51d4dae..103fc1ef0593 100644
--- a/lib/bpf.c
+++ b/lib/bpf.c
@@ -850,6 +850,27 @@ int bpf_graft_map(const char *map_path, uint32_t *key, int 
argc, char **argv)
      return ret;
  }

+int bpf_prog_attach(int prog_fd, int target_fd, enum bpf_attach_type type)
+{
+    union bpf_attr attr = {
+        .target_fd = target_fd,
+        .attach_bpf_fd = prog_fd,
+        .attach_type = type,
+    };

Please make this consistent with the other bpf(2) cmds we
have in the current lib code. There were some gcc issues in
the past, see:

https://git.kernel.org/cgit/linux/kernel/git/shemminger/iproute2.git/commit/?id=67584e3ab289a22eb9a2e51f90d23e2ced2e76b0

F.e. bpf_map_create() currently looks like:

     union bpf_attr attr = {};

     attr.map_type = type;
     attr.key_size = size_key;
     attr.value_size = size_value;
     attr.max_entries = max_elem;
     attr.map_flags = flags;

+    return bpf(BPF_PROG_ATTACH, &attr, sizeof(attr));
+}
+
+int bpf_prog_detach(int target_fd, enum bpf_attach_type type)
+{
+    union bpf_attr attr = {
+        .target_fd = target_fd,
+        .attach_type = type,
+    };

Ditto.

+    return bpf(BPF_PROG_DETACH, &attr, sizeof(attr));
+}
+
  #ifdef HAVE_ELF
  struct bpf_elf_prog {
      enum bpf_prog_type    type;
@@ -1262,9 +1283,9 @@ static void bpf_prog_report(int fd, const char *section,
      bpf_dump_error(ctx, "Verifier analysis:\n\n");
  }

-static int bpf_prog_attach(const char *section,
-               const struct bpf_elf_prog *prog,
-               struct bpf_elf_ctx *ctx)
+static int bpf_prog_load_and_report(const char *section,
+                    const struct bpf_elf_prog *prog,
+                    struct bpf_elf_ctx *ctx)
  {

Please name it bpf_prog_create() then, it would be consistent to
bpf_map_create() and shorter as well.

Sorry, lack of coffee, scratch that.

Can't the current bpf_prog_attach() stay as is, and you name the above new
functions bpf_prog_attach_fd() and bpf_prog_detach_fd()? I think that would
be better.

      int tries = 0, fd;
  retry:
@@ -1656,7 +1677,7 @@ static int bpf_fetch_prog(struct bpf_elf_ctx *ctx, const 
char *section,
          prog.size    = data.sec_data->d_size;
          prog.license = ctx->license;

-        fd = bpf_prog_attach(section, &prog, ctx);
+        fd = bpf_prog_load_and_report(section, &prog, ctx);
          if (fd < 0)
              return fd;

@@ -1755,7 +1776,7 @@ static int bpf_fetch_prog_relo(struct bpf_elf_ctx *ctx, 
const char *section,
          prog.size    = data_insn.sec_data->d_size;
          prog.license = ctx->license;

-        fd = bpf_prog_attach(section, &prog, ctx);
+        fd = bpf_prog_load_and_report(section, &prog, ctx);
          if (fd < 0) {
              *lderr = true;
              return fd;