ARM: 7598/1: net: bpf_jit_32: fix sp-relative load/stores offsets.

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

@@ -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))