Re: [PATCH v2 1/4] kasan, arm64: Add KASAN light mode

[Vincenzo Frascino <vincenzo.frascino@arm.com>]

On 1/13/21 5:16 PM, Catalin Marinas wrote:
> On Thu, Jan 07, 2021 at 05:29:05PM +0000, Vincenzo Frascino wrote:
>> Architectures supported by KASAN HW can provide a light mode of
>> execution. On an MTE enabled arm64 hw for example this can be identified
>> with the asynch mode of execution. If an async exception occurs, the
>> arm64 core updates a register which is asynchronously detected the next
>> time in which the kernel is accessed.
> 
> What do you mean by "the kernel is accessed"? Also, there is no
> "exception" as such, only a bit in a register updated asynchronously. So
> the last sentence could be something like:
> 
>   In this mode, if a tag check fault occurs, the TFSR_EL1 register is
>   updated asynchronously. The kernel checks the corresponding bits
>   periodically.
> 
> (or you can be more precise on when the kernel checks for such faults)
>

Yes, I agree, I will change it accordingly. What I wrote has a similar meaning
but your exposition is more clear.

>> KASAN requires a specific mode of execution to make use of this hw feature.
>>
>> Add KASAN HW light execution mode.
> 
> Shall we call it "fast"? ;)
> 
>> --- /dev/null
>> +++ b/include/linux/kasan_def.h
>> @@ -0,0 +1,25 @@
>> +/* SPDX-License-Identifier: GPL-2.0 */
>> +#ifndef _LINUX_KASAN_DEF_H
>> +#define _LINUX_KASAN_DEF_H
>> +
>> +enum kasan_arg_mode {
>> +	KASAN_ARG_MODE_DEFAULT,
>> +	KASAN_ARG_MODE_OFF,
>> +	KASAN_ARG_MODE_LIGHT,
>> +	KASAN_ARG_MODE_PROD,
>> +	KASAN_ARG_MODE_FULL,
>> +};
>> +
>> +enum kasan_arg_stacktrace {
>> +	KASAN_ARG_STACKTRACE_DEFAULT,
>> +	KASAN_ARG_STACKTRACE_OFF,
>> +	KASAN_ARG_STACKTRACE_ON,
>> +};
>> +
>> +enum kasan_arg_fault {
>> +	KASAN_ARG_FAULT_DEFAULT,
>> +	KASAN_ARG_FAULT_REPORT,
>> +	KASAN_ARG_FAULT_PANIC,
>> +};
>> +
>> +#endif /* _LINUX_KASAN_DEF_H */
> 
> I thought we agreed not to expose the KASAN internal but come up with
> another abstraction. Maybe this was after you posted these patches.
> 

Yes, indeed we agreed and I am going to change it in v3. The agreement
temporally came after I posted v2 hence it is not reflected here.

-- 
Regards,
Vincenzo