2018-05-16 15:35:21

by Alexander Potapenko

[permalink] [raw]
Subject: [PATCH] lib/stackdepot.c: use a non-instrumented version of memcpy()

stackdepot used to call memcpy(), which compiler tools normally
instrument, therefore every lookup used to unnecessarily call instrumented
code. This is somewhat ok in the case of KASAN, but under KMSAN a lot of
time was spent in the instrumentation.

(A similar change has been previously committed for memcmp())

Signed-off-by: Alexander Potapenko <[email protected]>
Cc: Andrey Ryabinin <[email protected]>
Cc: Dmitry Vyukov <[email protected]>
---
lib/stackdepot.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib/stackdepot.c b/lib/stackdepot.c
index e513459a5601..d48c744fa750 100644
--- a/lib/stackdepot.c
+++ b/lib/stackdepot.c
@@ -140,7 +140,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
stack->handle.slabindex = depot_index;
stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
stack->handle.valid = 1;
- memcpy(stack->entries, entries, size * sizeof(unsigned long));
+ __memcpy(stack->entries, entries, size * sizeof(unsigned long));
depot_offset += required_size;

return stack;
--
2.17.0.441.gb46fe60e1d-goog



2018-05-16 16:46:53

by Andrey Ryabinin

[permalink] [raw]
Subject: Re: [PATCH] lib/stackdepot.c: use a non-instrumented version of memcpy()

On 05/16/2018 06:34 PM, Alexander Potapenko wrote:
> stackdepot used to call memcpy(), which compiler tools normally
> instrument, therefore every lookup used to unnecessarily call instrumented
> code. This is somewhat ok in the case of KASAN, but under KMSAN a lot of
> time was spent in the instrumentation.
>
> (A similar change has been previously committed for memcmp())
>
> Signed-off-by: Alexander Potapenko <[email protected]>
> Cc: Andrey Ryabinin <[email protected]>
> Cc: Dmitry Vyukov <[email protected]>
> ---
> lib/stackdepot.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/lib/stackdepot.c b/lib/stackdepot.c
> index e513459a5601..d48c744fa750 100644
> --- a/lib/stackdepot.c
> +++ b/lib/stackdepot.c
> @@ -140,7 +140,7 @@ static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
> stack->handle.slabindex = depot_index;
> stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
> stack->handle.valid = 1;
> - memcpy(stack->entries, entries, size * sizeof(unsigned long));
> + __memcpy(stack->entries, entries, size * sizeof(unsigned long));

This has no effect. Since the whole file is not instrumented memcpy automagically replaced with __memcpy.


2018-05-17 08:55:18

by Alexander Potapenko

[permalink] [raw]
Subject: Re: [PATCH] lib/stackdepot.c: use a non-instrumented version of memcpy()

On Wed, May 16, 2018 at 6:45 PM Andrey Ryabinin <[email protected]>
wrote:

> On 05/16/2018 06:34 PM, Alexander Potapenko wrote:
> > stackdepot used to call memcpy(), which compiler tools normally
> > instrument, therefore every lookup used to unnecessarily call
instrumented
> > code. This is somewhat ok in the case of KASAN, but under KMSAN a lot
of
> > time was spent in the instrumentation.
> >
> > (A similar change has been previously committed for memcmp())
> >
> > Signed-off-by: Alexander Potapenko <[email protected]>
> > Cc: Andrey Ryabinin <[email protected]>
> > Cc: Dmitry Vyukov <[email protected]>
> > ---
> > lib/stackdepot.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/lib/stackdepot.c b/lib/stackdepot.c
> > index e513459a5601..d48c744fa750 100644
> > --- a/lib/stackdepot.c
> > +++ b/lib/stackdepot.c
> > @@ -140,7 +140,7 @@ static struct stack_record
*depot_alloc_stack(unsigned long *entries, int size,
> > stack->handle.slabindex = depot_index;
> > stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
> > stack->handle.valid = 1;
> > - memcpy(stack->entries, entries, size * sizeof(unsigned long));
> > + __memcpy(stack->entries, entries, size * sizeof(unsigned long));

> This has no effect. Since the whole file is not instrumented memcpy
automagically replaced with __memcpy.
You're right, we just didn't have the code defining memcpy() to __memcpy()
in KMSAN. I'll fix that instead.


--
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Straße, 33
80636 München

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg

2018-05-18 01:05:45

by kernel test robot

[permalink] [raw]
Subject: Re: [PATCH] lib/stackdepot.c: use a non-instrumented version of memcpy()

Hi Alexander,

I love your patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on v4.17-rc5 next-20180517]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]

url: https://github.com/0day-ci/linux/commits/Alexander-Potapenko/lib-stackdepot-c-use-a-non-instrumented-version-of-memcpy/20180518-015328
config: sh-allmodconfig (attached as .config)
compiler: sh4-linux-gnu-gcc (Debian 7.2.0-11) 7.2.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
make.cross ARCH=sh

All errors (new ones prefixed by >>):

lib/stackdepot.c: In function 'depot_alloc_stack':
>> lib/stackdepot.c:143:2: error: implicit declaration of function '__memcpy'; did you mean 'memcpy'? [-Werror=implicit-function-declaration]
__memcpy(stack->entries, entries, size * sizeof(unsigned long));
^~~~~~~~
memcpy
cc1: some warnings being treated as errors

vim +143 lib/stackdepot.c

106
107 /* Allocation of a new stack in raw storage */
108 static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
109 u32 hash, void **prealloc, gfp_t alloc_flags)
110 {
111 int required_size = offsetof(struct stack_record, entries) +
112 sizeof(unsigned long) * size;
113 struct stack_record *stack;
114
115 required_size = ALIGN(required_size, 1 << STACK_ALLOC_ALIGN);
116
117 if (unlikely(depot_offset + required_size > STACK_ALLOC_SIZE)) {
118 if (unlikely(depot_index + 1 >= STACK_ALLOC_MAX_SLABS)) {
119 WARN_ONCE(1, "Stack depot reached limit capacity");
120 return NULL;
121 }
122 depot_index++;
123 depot_offset = 0;
124 /*
125 * smp_store_release() here pairs with smp_load_acquire() from
126 * |next_slab_inited| in depot_save_stack() and
127 * init_stack_slab().
128 */
129 if (depot_index + 1 < STACK_ALLOC_MAX_SLABS)
130 smp_store_release(&next_slab_inited, 0);
131 }
132 init_stack_slab(prealloc);
133 if (stack_slabs[depot_index] == NULL)
134 return NULL;
135
136 stack = stack_slabs[depot_index] + depot_offset;
137
138 stack->hash = hash;
139 stack->size = size;
140 stack->handle.slabindex = depot_index;
141 stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
142 stack->handle.valid = 1;
> 143 __memcpy(stack->entries, entries, size * sizeof(unsigned long));
144 depot_offset += required_size;
145
146 return stack;
147 }
148

---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation


Attachments:
(No filename) (2.87 kB)
.config.gz (46.64 kB)
Download all attachments