On 04/28/2014 07:34 PM, Oleg Nesterov wrote:
> Thanks...
>
> Again, the change in riprel_analyze() needs the review from someone
> who understands the instruction decoding/encoding.
>
> On 04/28, Denys Vlasenko wrote:
>>
>> Otherwise, instructions such as cmpxchg and div will be mishandled.
>
> It seems that you are right. But it would be really great if you also
> provide the test-case which proves the fix ;)
Working on a testcase for this. So far covered div (test1)
and cmpxchg (test2).
Reproduced failure on a fairly old 3.10.11 kernel:
# gcc -Os -Wall test_riprel.c -o test_riprel
# ./test_riprel
test1: pass
test2: pass
# perf probe -x ./test_riprel probe1
# perf record -e probe_test:probe1 ./test_riprel
test1: FAIL
test2: pass
# perf probe -x ./test_riprel probe2
# perf record -e probe_test:probe2 ./test_riprel
test1: pass
test2: FAIL
Source:
test_riprel.c
==================
#include <stdio.h>
static const char *const fail_pass[] = { "FAIL", "pass" };
long two = 2;
long test1()
{
long ax=0, dx=0;
asm volatile("\n"
" xor %%edx,%%edx\n"
" lea 2(%%edx),%%eax\n"
// We divide 2 by 2. Result (in eax) should be 1:
" probe1: .globl probe1\n"
" divl two(%%rip)\n"
// If we have a bug (eax mangled on entry) the result will be 2,
// because eax gets restored by probe machinery.
: "=a" (ax), "=d" (dx) /*out*/
: "0" (ax), "1" (dx) /*in*/
: "memory" /*clobber*/
);
dprintf(2, "%s: %s\n", __func__, fail_pass[ax == 1]);
return ax;
}
long val2 = 0;
long test2()
{
long old_val2 = val2;
long ax=0, dx=0;
asm volatile("\n"
" mov val2,%%eax\n" // eax := val2
" lea 1(%%eax),%%edx\n" // edx := eax+1
// eax is equal to val2. cmpxchg should store edx to val2:
" probe2: .globl probe2\n"
" cmpxchg %%edx,val2(%%rip)\n"
// If we have a bug (eax mangled on entry), val2 will stay unchanged
: "=a" (ax), "=d" (dx) /*out*/
: "0" (ax), "1" (dx) /*in*/
: "memory" /*clobber*/
);
dprintf(2, "%s: %s\n", __func__, fail_pass[val2 == old_val2 + 1]);
return ax == dx;
}
int main()
{
test1();
test2();
return 0;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
