[PATCH] MIPS: BPF: Disable JIT on R3000 (MIPS-I)

[PATCH] MIPS: BPF: Disable JIT on R3000 (MIPS-I)

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 739 bytes --]

The JIT does not include the load delay slots required by MIPS-I
processors.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
---
 arch/mips/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 199a835..7ff9cba 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -12,7 +12,7 @@ config MIPS
 	select HAVE_ARCH_KGDB
 	select HAVE_ARCH_SECCOMP_FILTER
 	select HAVE_ARCH_TRACEHOOK
-	select HAVE_BPF_JIT if !CPU_MICROMIPS
+	select HAVE_BPF_JIT if !CPU_MICROMIPS && !CPU_R3000
 	select ARCH_HAVE_CUSTOM_GPIO_H
 	select HAVE_FUNCTION_TRACER
 	select HAVE_DYNAMIC_FTRACE
-- 
Ben Hutchings
friends: People who know you well, but like you anyway.


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]

Re: [PATCH] MIPS: BPF: Disable JIT on R3000 (MIPS-I)

[Ralf Baechle]

On Sat, Sep 05, 2015 at 08:29:28PM +0100, Ben Hutchings wrote:

> The JIT does not include the load delay slots required by MIPS-I
> processors.

See 0c5d187828588dd1b36cb93b4481a8db467ef3e8 (MIPS: BPF: Fix load delay
slots.) for a proper fix.

I'm wondering, how did you find this issue, are you scanning the code
for broken instruction sequences?

  Ralf

Re: [PATCH] MIPS: BPF: Disable JIT on R3000 (MIPS-I)

[Markos Chandras]

On 10/05/2015 10:24 AM, Ralf Baechle wrote:
> On Sat, Sep 05, 2015 at 08:29:28PM +0100, Ben Hutchings wrote:
> 
>> The JIT does not include the load delay slots required by MIPS-I
>> processors.
> 
> See 0c5d187828588dd1b36cb93b4481a8db467ef3e8 (MIPS: BPF: Fix load delay
> slots.) for a proper fix.

Since this patch didn't make it to the LMO and never discussed, i must
say that it's probably not complete. The bpf_jit.c code also emits load
instructions (for example see emit_load_imm). So I guess you need to
extend your patch to cover the C code as well.

-- 
markos

Re: [PATCH] MIPS: BPF: Disable JIT on R3000 (MIPS-I)

[Markos Chandras]

On 10/05/2015 10:24 AM, Ralf Baechle wrote:
> On Sat, Sep 05, 2015 at 08:29:28PM +0100, Ben Hutchings wrote:
> 
>> The JIT does not include the load delay slots required by MIPS-I
>> processors.
> 
> See 0c5d187828588dd1b36cb93b4481a8db467ef3e8 (MIPS: BPF: Fix load delay
> slots.) for a proper fix.

Since this patch didn't make it to the LMO and never discussed, i must
say that it's probably not complete. The bpf_jit.c code also emits load
instructions (for example see emit_load_imm). So I guess you need to
extend your patch to cover the C code as well.

-- 
markos

Re: [PATCH] MIPS: BPF: Disable JIT on R3000 (MIPS-I)

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 688 bytes --]

On Mon, 2015-10-05 at 11:24 +0200, Ralf Baechle wrote:
> On Sat, Sep 05, 2015 at 08:29:28PM +0100, Ben Hutchings wrote:
> 
> > The JIT does not include the load delay slots required by MIPS-I
> > processors.
> 
> See 0c5d187828588dd1b36cb93b4481a8db467ef3e8 (MIPS: BPF: Fix load delay
> slots.) for a proper fix.

How about all the inlined loads (emit_load, emit_load_byte,
emit_half_load)?

> I'm wondering, how did you find this issue, are you scanning the code
> for broken instruction sequences?

No, I was already looking at the BPF JIT implementation for some other
reason.

Ben.

-- 
Ben Hutchings
compatible: Gracefully accepts erroneous data from any source

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]