[Cocci] double_lock.cocci extremely slow for multiple mutex locks check

[Ning Jiang]

Hi,

I met a case when checking my driver by coccicheck. It can be illustrated
by the simplified test-cocci.c.

#include <linux/mutex.h>

static DEFINE_MUTEX(test_lock);

int test_cond;
int cmd;
int type;
int ret;

int test_cocci_func(void)
{
if (ret == 1)
 return -1;

if (!test_cond) {
switch (cmd) {
 case 0:
mutex_lock(&test_lock);
mutex_unlock(&test_lock);
 break;
}
}

if (ret < 0)
 return ret;

if (test_cond) {
switch (cmd) {
 case 0:
if (type == 1) {
mutex_lock(&test_lock);
 mutex_unlock(&test_lock);
} else if (type == 2) {
mutex_lock(&test_lock);
 mutex_unlock(&test_lock);
}
break;
 case 1:
if (type == 1) {
mutex_lock(&test_lock);
 mutex_unlock(&test_lock);
} else {
mutex_lock(&test_lock);
 mutex_unlock(&test_lock);
}
break;
 }
}

return ret;
}

I run the coccicheck like this:
time make coccicheck V=1 MODE=report M=test-cocci
COCCI=scripts/coccinelle/locks/double_lock.cocci 1>test-cocci-result 2>&1

It gives the following output:
real    2m1.676s
user    2m0.904s
sys     0m0.160s

The strange thing is that after I removed the code at the beginning of the
function:
if (ret == 1)
 return -1;

The test result will be much better:
real    0m0.896s
user    0m0.500s
sys     0m0.148s

Could you please check the issue?

Thanks,
Ning




 
test-cocci.c<https://docs.google.com/file/d/0B7BiiwPkh3BwWi1jdVNDWjdaT00/edit?usp=drive_web>


_______________________________________________
Cocci mailing list
Cocci@systeme.lip6.fr
https://systeme.lip6.fr/mailman/listinfo/cocci