I have been looking at this. I'm testing the per function time limit
fix. I think the problem is that Smatch is using more than 3G of RAM.
The problem is that memcpy() really has a lot of implications and Smatch
is trying to keep track of it. Consider the following one line program:
/* test.c */
#include "check_debug.h"
void memcpy(void *dest, void *src, int size);
struct foo {
int x, y, z;
};
struct foo *p, *q;
static void function(void)
{
memcpy(p, q, sizeof(p));
__smatch_cur_slist();
}
/* end of file */
If we run Smatch on it then it gives the following output.
$ ./smatch test.c
dumping slist at 14
[internal] 'unnull_path' = true (true)
[check_memory] '-p' = assigned (assigned)
[check_memory] '-q' = assigned (assigned)
[check_missing_break] '__smatch_cur_slist' = used (used)
[check_missing_break] 'memcpy' = used (used)
[check_missing_break] 'p' = used (used)
[check_missing_break] 'q' = used (used)
[check_assigned_expr] 'p->x' = q->x (q->x)
[check_assigned_expr] 'p->y' = q->y (q->y)
[check_assigned_expr] 'p->z' = q->z (q->z)
[register_comparison_links] 'p->x' = p->x vs q->x (p->x vs q->x)
[register_comparison_links] 'p->y' = p->y vs q->y (p->y vs q->y)
[register_comparison_links] 'p->z' = p->z vs q->z (p->z vs q->z)
[register_comparison_links] 'q->x' = p->x vs q->x (p->x vs q->x)
[register_comparison_links] 'q->y' = p->y vs q->y (p->y vs q->y)
[register_comparison_links] 'q->z' = p->z vs q->z (p->z vs q->z)
[register_comparison] 'p->x vs q->x' = == (==)
[register_comparison] 'p->y vs q->y' = == (==)
[register_comparison] 'p->z vs q->z' = == (==)
[register_param_set] 'p->x' = s32min-s32max (s32min-s32max)
[register_param_set] 'p->y' = s32min-s32max (s32min-s32max)
[register_param_set] 'p->z' = s32min-s32max (s32min-s32max)
[register_smatch_extra] 'p->x' = s32min-s32max (s32min-s32max)
[register_smatch_extra] 'p->y' = s32min-s32max (s32min-s32max)
[register_smatch_extra] 'p->z' = s32min-s32max (s32min-s32max)
[register_smatch_extra] 'q->x' = s32min-s32max (s32min-s32max)
[register_smatch_extra] 'q->y' = s32min-s32max (s32min-s32max)
[register_smatch_extra] 'q->z' = s32min-s32max (s32min-s32max)
---
It used to be that we didn't track that stuff but it's all useful
information. I don't know which bits I could get rid of.
Perhaps there is a way to handle large switch statements better.
regards,
dan carpenter
--
To unsubscribe from this list: send the line "unsubscribe smatch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
