Skip to contentSkip to conversation

net: filter: cleanup invocation of internal BPF

PATCH Applied Alexei Starovoitov 3 participants 3 messages May 19, 2014 – May 21, 2014

Subsystems (7)
Networking Network drivers Library code BPF +3 more subsystems
lore
Share & tools
lore

Alexei Starovoitov

author
#0 lore raw

PATCH 1/1 net: filter: cleanup invocation of internal BPF

Kernel API for classic BPF socket filters is:

sk_unattached_filter_create() - validate classic BPF, convert, JIT SK_RUN_FILTER() - run it
sk_unattached_filter_destroy() - destroy socket filter

Cleanup internal BPF kernel API as following:

sk_filter_select_runtime() - final step of internal BPF creation.
Try to JIT internal BPF program, if JIT is not available select interpreter
SK_RUN_FILTER() - run it
sk_filter_free() - free internal BPF program

Disallow direct calls to BPF interpreter. Execution of the BPF program should be done with SK_RUN_FILTER() macro.

Example of internal BPF create, run, destroy:

struct sk_filter *fp;

  fp = kzalloc(sk_filter_size(prog_len), GFP_KERNEL);
  memcpy(fp->insni, prog, prog_len * sizeof(fp->insni[0]));
  fp->len = prog_len;

  sk_filter_select_runtime(fp);

  SK_RUN_FILTER(fp, ctx);

  sk_filter_free(fp);

Sockets, seccomp, testsuite, tracing are using different ways to populate sk_filter, so first steps of program creation are not common.

Signed-off-by Alexei Starovoitov <[email protected]>

---

This patch blurs the line between BPF program running in interpeter and JITed program. The 'jited' flag in 'struct sk_filter' indicates to the caller whether program was JITed or not, but that shouldn't matter to the caller. JITed program must be equivalent.
Another alternative is to export bpf_int_jit_compile()/bpf_jit_free() and let the users decide whether to use interpreter or JIT, but that seems ugly. We already have net.core.bpf_jit_enable for that.

 include/linux/filter.h |    6 ++----
 kernel/seccomp.c       |    6 ++----
 lib/test_bpf.c         |    4 ++--
 net/core/filter.c      |   44 ++++++++++++++++++++++++++++----------------
 4 files changed, 34 insertions(+), 26 deletions(-)
include/linux/filter.h +2 -4
index 9d5ae0a2c954..7977b3958e25 100644
@@ -184,10 +184,8 @@ static inline unsigned int sk_filter_size(unsigned int proglen)
 
 int sk_filter(struct sock *sk, struct sk_buff *skb);
 
-u32 sk_run_filter_int_seccomp(const struct seccomp_data *ctx,
-			      const struct sock_filter_int *insni);
-u32 sk_run_filter_int_skb(const struct sk_buff *ctx,
-			  const struct sock_filter_int *insni);
+void sk_filter_select_runtime(struct sk_filter *fp);
+void sk_filter_free(struct sk_filter *fp);
 
 int sk_convert_filter(struct sock_filter *prog, int len,
 		      struct sock_filter_int *new_prog, int *new_len);
kernel/seccomp.c +2 -4
index 7e02d624cc50..1036b6f2fded 100644
@@ -273,10 +273,8 @@ static long seccomp_attach_filter(struct sock_fprog *fprog)
 
 	atomic_set(&filter->usage, 1);
 	filter->prog->len = new_len;
-	filter->prog->bpf_func = (void *)sk_run_filter_int_seccomp;
 
-	/* JIT internal BPF into native HW instructions */
-	bpf_int_jit_compile(filter->prog);
+	sk_filter_select_runtime(filter->prog);
 
 	/*
 	 * If there is an existing filter, make it the prev and don't drop its

@@ -340,7 +338,7 @@ void put_seccomp_filter(struct task_struct *tsk)
 	while (orig && atomic_dec_and_test(&orig->usage)) {
 		struct seccomp_filter *freeme = orig;
 		orig = orig->prev;
-		bpf_jit_free(freeme->prog);
+		sk_filter_free(freeme->prog);
 		kfree(freeme);
 	}
 }
lib/test_bpf.c +2 -2
index 3603ebcd5d65..e160934430eb 100644
@@ -1489,7 +1489,7 @@ static __init int test_bpf(void)
 			memcpy(fp_ext->insns, tests[i].insns_int,
 			       fprog.len * 8);
 			fp->len = fprog.len;
-			fp->bpf_func = sk_run_filter_int_skb;
+			sk_filter_select_runtime(fp);
 		} else {
 			err = sk_unattached_filter_create(&fp, &fprog);
 			if (tests[i].data_type == EXPECTED_FAIL) {

@@ -1516,7 +1516,7 @@ static __init int test_bpf(void)
 		if (tests[i].data_type != SKB_INT)
 			sk_unattached_filter_destroy(fp);
 		else
-			kfree(fp);
+			sk_filter_free(fp);
 
 		if (err) {
 			pr_cont("FAIL %d\n", err);
net/core/filter.c +28 -16

88 lines not on this page · raw message

Signature · 1 line
1.7.9.5
Acked-by on v1 1/1
#1 lore raw
Quoted 30 lines · Alexei Starovoitov

Kernel API for classic BPF socket filters is:

sk_unattached_filter_create() - validate classic BPF, convert, JIT SK_RUN_FILTER() - run it
sk_unattached_filter_destroy() - destroy socket filter

Cleanup internal BPF kernel API as following:

sk_filter_select_runtime() - final step of internal BPF creation.
Try to JIT internal BPF program, if JIT is not available select interpreter
SK_RUN_FILTER() - run it
sk_filter_free() - free internal BPF program

Disallow direct calls to BPF interpreter. Execution of the BPF program should be done with SK_RUN_FILTER() macro.

Example of internal BPF create, run, destroy:

   struct sk_filter *fp;

   fp = kzalloc(sk_filter_size(prog_len), GFP_KERNEL);
   memcpy(fp->insni, prog, prog_len * sizeof(fp->insni[0]));
   fp->len = prog_len;

   sk_filter_select_runtime(fp);

   SK_RUN_FILTER(fp, ctx);

   sk_filter_free(fp);
Quoted 4 lines

Sockets, seccomp, testsuite, tracing are using different ways to populate sk_filter, so first steps of program creation are not common.

Signed-off-by Alexei Starovoitov <[email protected]>

I think it makes sense and we can avoid directly exposing the symbol __sk_run_filter() resp. its aliases.

Acked-by Daniel Borkmann <[email protected]>
maintainer Applied
#2 lore raw
In reply to Alexei Starovoitov

From: Alexei Starovoitov <[email protected]>
Date: Mon, 19 May 2014 14:56:14 -0700

Quoted 30 lines

Kernel API for classic BPF socket filters is:

sk_unattached_filter_create() - validate classic BPF, convert, JIT SK_RUN_FILTER() - run it
sk_unattached_filter_destroy() - destroy socket filter

Cleanup internal BPF kernel API as following:

sk_filter_select_runtime() - final step of internal BPF creation.
Try to JIT internal BPF program, if JIT is not available select interpreter
SK_RUN_FILTER() - run it
sk_filter_free() - free internal BPF program

Disallow direct calls to BPF interpreter. Execution of the BPF program should be done with SK_RUN_FILTER() macro.

Example of internal BPF create, run, destroy:

struct sk_filter *fp;

  fp = kzalloc(sk_filter_size(prog_len), GFP_KERNEL);
  memcpy(fp->insni, prog, prog_len * sizeof(fp->insni[0]));
  fp->len = prog_len;

  sk_filter_select_runtime(fp);

  SK_RUN_FILTER(fp, ctx);

  sk_filter_free(fp);
Quoted 4 lines

Sockets, seccomp, testsuite, tracing are using different ways to populate sk_filter, so first steps of program creation are not common.

Signed-off-by Alexei Starovoitov <[email protected]>

Applied, thank you.