]> rtime.felk.cvut.cz Git - can-eth-gw-linux.git/commitdiff
ARM: 7598/1: net: bpf_jit_32: fix sp-relative load/stores offsets.
authorSchichan Nicolas <nschichan@freebox.fr>
Mon, 10 Dec 2012 13:49:40 +0000 (14:49 +0100)
committerRussell King <rmk+kernel@arm.linux.org.uk>
Tue, 11 Dec 2012 00:19:29 +0000 (00:19 +0000)
The offset must be multiplied by 4 to be sure to access the correct
32bit word in the stack scratch space.

For instance, a store at scratch memory cell #1 was generating the
following:

st r4, [sp, #1]

While the correct code for this is:

st r4, [sp, #4]

To reproduce the bug (assuming your system has a NIC with the mac
address 52:54:00:12:34:56):

echo 0 > /proc/sys/net/core/bpf_jit_enable
tcpdump -ni eth0 "ether[1] + ether[2] - ether[3] * ether[4] - ether[5] \
== -0x3AA" # this will capture packets as expected

echo 1 > /proc/sys/net/core/bpf_jit_enable
tcpdump -ni eth0 "ether[1] + ether[2] - ether[3] * ether[4] - ether[5] \
== -0x3AA" # this will not.

This bug was present since the original inclusion of bpf_jit for ARM
(ddecdfce: ARM: 7259/3: net: JIT compiler for packet filters).

Signed-off-by: Nicolas Schichan <nschichan@freebox.fr>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
arch/arm/net/bpf_jit_32.c

index a64d34968305dac0c76c5ad7f8e8455a990df906..b6f305e3b9085fe932ab056d2a7a68aa3db0f738 100644 (file)
@@ -42,7 +42,7 @@
 #define r_skb_hl       ARM_R8
 
 #define SCRATCH_SP_OFFSET      0
-#define SCRATCH_OFF(k)         (SCRATCH_SP_OFFSET + (k))
+#define SCRATCH_OFF(k)         (SCRATCH_SP_OFFSET + 4 * (k))
 
 #define SEEN_MEM               ((1 << BPF_MEMWORDS) - 1)
 #define SEEN_MEM_WORD(k)       (1 << (k))