On 5/27/2016 4:02 AM, Andrey Ryabinin wrote:

On 05/26/2016 09:43 PM, Yang Shi wrote:

quoted

The upstream commit 1771c6e1a567ea0ba2cccc0a4ffe68a1419fd8ef
("x86/kasan: instrument user memory access API") added KASAN instrument to
x86 user memory access API, so added such instrument to ARM64 too.

Tested by test_kasan module.

Signed-off-by: Yang Shi <redacted>
---
 arch/arm64/include/asm/uaccess.h | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

Please, cover __copy_from_user() and __copy_to_user() too.
Unlike x86, your patch doesn't instrument these two.

I should elaborated this in my review. Yes, I did think about it, but 
unlike x86, __copy_to/from_user are implemented by asm code on ARM64. If 
I add kasan_check_read/write into them, I have to move the registers 
around to prepare the parameters for kasan calls, then restore them 
after the call, for example the below code for __copy_to_user:

         mov     x9, x0
         mov     x10, x1
         mov     x11, x2
         mov     x0, x10
         mov     x1, x11
         bl      kasan_check_read
         mov     x0, x9
         mov     x1, x10


So, I'm wondering if it is worth or not since __copy_to/from_user are 
just called at a couple of places, i.e. sctp, a couple of drivers, etc 
and not used too much. Actually, I think some of them could be replaced 
by __copy_to/from_user_inatomic.

Any idea is appreciated.

Thanks,
Yang

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>