[PATCH] bpf, arm64: start flushing icache range from header

[PATCH] bpf, arm64: start flushing icache range from header

[Daniel Borkmann]

While recently going over ARM64's BPF code, I noticed that the icache
range we're flushing should start at header already and not at ctx.image.

Reason is that after b569c1c622c5 ("net: bpf: arm64: address randomize
and write protect JIT code"), we also want to make sure to flush the
random-sized trap in front of the start of the actual program (analogous
to x86). No operational differences from user side.

Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Zi Shen Lim <zlim.lnx@gmail.com>
Cc: Alexei Starovoitov <ast@kernel.org>
---
 ( As arm64 jit fixes seem to go via arm64 tree, sending them here. )

 arch/arm64/net/bpf_jit_comp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
index a44e529..ee06570 100644
--- a/arch/arm64/net/bpf_jit_comp.c
+++ b/arch/arm64/net/bpf_jit_comp.c
@@ -740,7 +740,7 @@ void bpf_int_jit_compile(struct bpf_prog *prog)
 	if (bpf_jit_enable > 1)
 		bpf_jit_dump(prog->len, image_size, 2, ctx.image);
 
-	bpf_flush_icache(ctx.image, ctx.image + ctx.idx);
+	bpf_flush_icache(header, ctx.image + ctx.idx);
 
 	set_memory_ro((unsigned long)header, header->pages);
 	prog->bpf_func = (void *)ctx.image;
-- 
1.9.3

[PATCH] bpf, arm64: start flushing icache range from header

[Mark Rutland]

On Sat, Nov 14, 2015 at 01:16:18AM +0100, Daniel Borkmann wrote:
> While recently going over ARM64's BPF code, I noticed that the icache
> range we're flushing should start at header already and not at ctx.image.
> 
> Reason is that after b569c1c622c5 ("net: bpf: arm64: address randomize
> and write protect JIT code"), we also want to make sure to flush the
> random-sized trap in front of the start of the actual program (analogous
> to x86). No operational differences from user side.
> 
> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
> Acked-by: Zi Shen Lim <zlim.lnx@gmail.com>
> Cc: Alexei Starovoitov <ast@kernel.org>
> ---
>  ( As arm64 jit fixes seem to go via arm64 tree, sending them here. )
> 
>  arch/arm64/net/bpf_jit_comp.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
> index a44e529..ee06570 100644
> --- a/arch/arm64/net/bpf_jit_comp.c
> +++ b/arch/arm64/net/bpf_jit_comp.c
> @@ -740,7 +740,7 @@ void bpf_int_jit_compile(struct bpf_prog *prog)
>  	if (bpf_jit_enable > 1)
>  		bpf_jit_dump(prog->len, image_size, 2, ctx.image);
>  
> -	bpf_flush_icache(ctx.image, ctx.image + ctx.idx);
> +	bpf_flush_icache(header, ctx.image + ctx.idx);

As far as I can see, ctx.idx doesn't take into account the size of the
header, given we zero it after bpf_jit_binary_alloc, and increment it
for each instruction.

So won't this prevent us from flushing the end of the image? Or did I
miss something?

Mark.

[PATCH] bpf, arm64: start flushing icache range from header

[Daniel Borkmann]

On 11/16/2015 12:39 PM, Mark Rutland wrote:
> On Sat, Nov 14, 2015 at 01:16:18AM +0100, Daniel Borkmann wrote:
>> While recently going over ARM64's BPF code, I noticed that the icache
>> range we're flushing should start at header already and not at ctx.image.
>>
>> Reason is that after b569c1c622c5 ("net: bpf: arm64: address randomize
>> and write protect JIT code"), we also want to make sure to flush the
>> random-sized trap in front of the start of the actual program (analogous
>> to x86). No operational differences from user side.
>>
>> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
>> Acked-by: Zi Shen Lim <zlim.lnx@gmail.com>
>> Cc: Alexei Starovoitov <ast@kernel.org>
>> ---
>>   ( As arm64 jit fixes seem to go via arm64 tree, sending them here. )
>>
>>   arch/arm64/net/bpf_jit_comp.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
>> index a44e529..ee06570 100644
>> --- a/arch/arm64/net/bpf_jit_comp.c
>> +++ b/arch/arm64/net/bpf_jit_comp.c
>> @@ -740,7 +740,7 @@ void bpf_int_jit_compile(struct bpf_prog *prog)
>>   	if (bpf_jit_enable > 1)
>>   		bpf_jit_dump(prog->len, image_size, 2, ctx.image);
>>
>> -	bpf_flush_icache(ctx.image, ctx.image + ctx.idx);
>> +	bpf_flush_icache(header, ctx.image + ctx.idx);
>
> As far as I can see, ctx.idx doesn't take into account the size of the
> header, given we zero it after bpf_jit_binary_alloc, and increment it
> for each instruction.
>
> So won't this prevent us from flushing the end of the image? Or did I
> miss something?

Nope, bpf_flush_icache() takes start and end pointer ... header starts
before ctx.image on the linear buffer. Why should this prevent us from
flushing the end of the image?

[PATCH] bpf, arm64: start flushing icache range from header

[Mark Rutland]

On Mon, Nov 16, 2015 at 12:48:08PM +0100, Daniel Borkmann wrote:
> On 11/16/2015 12:39 PM, Mark Rutland wrote:
> >On Sat, Nov 14, 2015 at 01:16:18AM +0100, Daniel Borkmann wrote:
> >>While recently going over ARM64's BPF code, I noticed that the icache
> >>range we're flushing should start at header already and not at ctx.image.
> >>
> >>Reason is that after b569c1c622c5 ("net: bpf: arm64: address randomize
> >>and write protect JIT code"), we also want to make sure to flush the
> >>random-sized trap in front of the start of the actual program (analogous
> >>to x86). No operational differences from user side.
> >>
> >>Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
> >>Acked-by: Zi Shen Lim <zlim.lnx@gmail.com>
> >>Cc: Alexei Starovoitov <ast@kernel.org>
> >>---
> >>  ( As arm64 jit fixes seem to go via arm64 tree, sending them here. )
> >>
> >>  arch/arm64/net/bpf_jit_comp.c | 2 +-
> >>  1 file changed, 1 insertion(+), 1 deletion(-)
> >>
> >>diff --git a/arch/arm64/net/bpf_jit_comp.c b/arch/arm64/net/bpf_jit_comp.c
> >>index a44e529..ee06570 100644
> >>--- a/arch/arm64/net/bpf_jit_comp.c
> >>+++ b/arch/arm64/net/bpf_jit_comp.c
> >>@@ -740,7 +740,7 @@ void bpf_int_jit_compile(struct bpf_prog *prog)
> >>  	if (bpf_jit_enable > 1)
> >>  		bpf_jit_dump(prog->len, image_size, 2, ctx.image);
> >>
> >>-	bpf_flush_icache(ctx.image, ctx.image + ctx.idx);
> >>+	bpf_flush_icache(header, ctx.image + ctx.idx);
> >
> >As far as I can see, ctx.idx doesn't take into account the size of the
> >header, given we zero it after bpf_jit_binary_alloc, and increment it
> >for each instruction.
> >
> >So won't this prevent us from flushing the end of the image? Or did I
> >miss something?
> 
> Nope, bpf_flush_icache() takes start and end pointer ... header starts
> before ctx.image on the linear buffer. Why should this prevent us from
> flushing the end of the image?

I erroneously thought the second parameter was a size (which it clearly
isn't given it's bsed on the ctx.image pointer).

My bad, apologies for the noise.

Mark.

[PATCH] bpf, arm64: start flushing icache range from header

[David Miller]

From: Daniel Borkmann <daniel@iogearbox.net>
Date: Sat, 14 Nov 2015 01:16:18 +0100

> While recently going over ARM64's BPF code, I noticed that the icache
> range we're flushing should start at header already and not at ctx.image.
> 
> Reason is that after b569c1c622c5 ("net: bpf: arm64: address randomize
> and write protect JIT code"), we also want to make sure to flush the
> random-sized trap in front of the start of the actual program (analogous
> to x86). No operational differences from user side.
> 
> Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
> Acked-by: Zi Shen Lim <zlim.lnx@gmail.com>
> Cc: Alexei Starovoitov <ast@kernel.org>
> ---
>  ( As arm64 jit fixes seem to go via arm64 tree, sending them here. )

I'm applying this directly, please CC: netdev on future patches
to the BPF JIT's regardless of where you think the patch will
be applied as it should be _REVIEWED_ on netdev regardless.

[PATCH] bpf, arm64: start flushing icache range from header

[Daniel Borkmann]

On 11/16/2015 08:42 PM, David Miller wrote:
...
> I'm applying this directly, please CC: netdev on future patches
> to the BPF JIT's regardless of where you think the patch will
> be applied as it should be _REVIEWED_ on netdev regardless.

Okay, sorry about that, will do in future.

Thanks, Dave!