Re: [PATCH 1/3] kcov: support comparison operands collection

[Mark Rutland]

Hi,

On Wed, Aug 30, 2017 at 06:23:29PM +0200, Dmitry Vyukov wrote:
> From: Victor Chibotaru <tchibo@xxxxxxxxxx>
> 
> Enables kcov to collect comparison operands from instrumented code.
> This is done by using Clang's -fsanitize=trace-cmp instrumentation
> (currently not available for GCC).

What's needed to build the kernel with Clang these days?

I was under the impression that it still wasn't possible to build arm64
with clang due to a number of missing features (e.g. the %a assembler
output template).

> The comparison operands help a lot in fuzz testing. E.g. they are
> used in Syzkaller to cover the interiors of conditional statements
> with way less attempts and thus make previously unreachable code
> reachable.
> 
> To allow separate collection of coverage and comparison operands two
> different work modes are implemented. Mode selection is now done via
> a KCOV_ENABLE ioctl call with corresponding argument value.
> 
> Signed-off-by: Victor Chibotaru <tchibo@xxxxxxxxxx>
> Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> Cc: Mark Rutland <mark.rutland@xxxxxxx>
> Cc: Alexander Popov <alex.popov@xxxxxxxxx>
> Cc: Andrey Ryabinin <aryabinin@xxxxxxxxxxxxx>
> Cc: Kees Cook <keescook@xxxxxxxxxxxx>
> Cc: Vegard Nossum <vegard.nossum@xxxxxxxxxx>
> Cc: Quentin Casasnovas <quentin.casasnovas@xxxxxxxxxx>
> Cc: syzkaller@xxxxxxxxxxxxxxxx
> Cc: linux-mm@xxxxxxxxx
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> ---
> Clang instrumentation:
> https://clang.llvm.org/docs/SanitizerCoverage.html#tracing-data-flow

How stable is this?

The comment at the end says "This interface is a subject to change."

[...]

> diff --git a/kernel/kcov.c b/kernel/kcov.c
> index cd771993f96f..2abce5dfa2df 100644
> --- a/kernel/kcov.c
> +++ b/kernel/kcov.c
> @@ -21,13 +21,21 @@
>  #include <linux/kcov.h>
>  #include <asm/setup.h>
>  
> +/* Number of words written per one comparison. */
> +#define KCOV_WORDS_PER_CMP 3

Could you please expand the comment to cover what a "word" is?

Generally, "word" is an ambiguous term, and it's used inconsitently in
this file as of this patch. For comparison coverage, a "word" is assumed
to always be 64-bit, (which makes sxense given 64-bit comparisons), but
for branch coverage a "word" refers to an unsigned long, which would be
32-bit on a 32-bit platform.

[...]

> +static bool check_kcov_mode(enum kcov_mode needed_mode, struct task_struct *t)

Perhaps kcov_mode_is_active()?

That would better describe what is being checked.

> +{
> +	enum kcov_mode mode;
> +
> +	/*
> +	 * We are interested in code coverage as a function of a syscall inputs,
> +	 * so we ignore code executed in interrupts.
> +	 */
> +	if (!t || !in_task())
> +		return false;
> +	mode = READ_ONCE(t->kcov_mode);
> +	/*
> +	 * There is some code that runs in interrupts but for which
> +	 * in_interrupt() returns false (e.g. preempt_schedule_irq()).
> +	 * READ_ONCE()/barrier() effectively provides load-acquire wrt
> +	 * interrupts, there are paired barrier()/WRITE_ONCE() in
> +	 * kcov_ioctl_locked().
> +	 */
> +	barrier();
> +	if (mode != needed_mode)
> +		return false;
> +	return true;

This would be simlper as:

	barrier();
	return mode == needed_mode;

[...]

> +#ifdef CONFIG_KCOV_ENABLE_COMPARISONS
> +static void write_comp_data(u64 type, u64 arg1, u64 arg2)
> +{
> +	struct task_struct *t;
> +	u64 *area;
> +	u64 count, start_index, end_pos, max_pos;
> +
> +	t = current;
> +	if (!check_kcov_mode(KCOV_MODE_TRACE_CMP, t))
> +		return;
> +
> +	/*
> +	 * We write all comparison arguments and types as u64.
> +	 * The buffer was allocated for t->kcov_size unsigned longs.
> +	 */
> +	area = (u64 *)t->kcov_area;
> +	max_pos = t->kcov_size * sizeof(unsigned long);

Perhaps it would make more sense for k->kcov_size to be in bytes, if
different options will have differing record sizes?

> +
> +	count = READ_ONCE(area[0]);
> +
> +	/* Every record is KCOV_WORDS_PER_CMP words. */

As above, please be explicit about what a "word" is, or avoid using
"word" terminology.

Thanks,
Mark.