On Wednesday 27 August 2008 23:22, Harald Küthe wrote:
> Hello list,
>
> I'm getting coredumps in bb-1.12.0 in obj_create_alloced_section at
> xrealloc_vector at an insmod call.
> It looks like xrealloc_vector() is assuming that the memory of
> f->sections is quantisized
> in steps as stated in the 2nd parameter of xrealloc_vector().
> In insmod it seems that the inital malloc of f->sections is not.
>
> static struct obj_section *obj_create_alloced_section_first(struct
> obj_file *f,...)
> ...
> f->sections = xrealloc(f->sections, (newidx + 1) * sizeof(sec));
>
> At the 1st call of xrealloc_vector() no xrealloc is done.
What do you mean?
> Then the 2nd
> call is causing a memory error/coredump. (*** glibc detected ***
> realloc(): invalid next size: 0x10068710Aborted (core dumped))
> When I remove "if (!(idx & (mask - 1))) in xrealloc_vector_helper() so
> that xrealloc is called each time things are running.
Can you instrument it like this:
void* FAST_FUNC xrealloc_vector_helper(void *vector, unsigned sizeof_and_shift,
int idx)
{
int mask = 1 << (uint8_t)sizeof_and_shift;
bb_error_msg("xrealloc_vector_helper(v:%p, mask:%x, sizeof:%d, idx:%d",
vector, mask, sizeof_and_shift >> 8, idx);
if (!(idx & (mask - 1))) {
rebuild it, run insmod and post the result to ml?
--
vda
_______________________________________________
busybox mailing list
[email protected]
http://busybox.net/cgi-bin/mailman/listinfo/busybox
