[PATCH] setenv(): fix memory leak when setting large, duplicate string
H.J. Lu
hjl.tools@gmail.com
Tue Jan 6 17:09:00 GMT 2015
On Mon, Dec 1, 2014 at 2:23 AM, Siddhesh Poyarekar <siddhesh@redhat.com> wrote:
> On Sat, Nov 22, 2014 at 02:02:40PM -0600, Eric Biggers wrote:
>> glibc maintains a binary tree of environment strings it malloc()ed
>> itself. However, it's possible for it to malloc() a string, then find
>> that an identical string is already in the tree. In this case, the
>> memory is leaked and is not freed if the application later calls
>> __libc_freeres(). Fix this by freeing 'new_value' when it's unneeded.
>>
>> Test case:
>> #include <stdlib.h>
>> #include <string.h>
>>
>> int main()
>> {
>> char *p = calloc(100000, 1);
>> memset(p, 'A', 99999);
>> setenv("TESTVAR", p, 1);
>> setenv("TESTVAR", p, 1);
>> free(p);
>> }
>>
>> Leak that was reported by valgrind:
>> 100,008 bytes in 1 blocks are definitely lost in loss record 1 of 1
>> at 0x4C29F90: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
>> by 0x4E6B3D4: __add_to_environ (setenv.c:176)
>> by 0x4C31B8F: setenv (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
>> by 0x400642: main (in /mnt/tmpfs/a.out)
>
> Thanks, the patch looks good but before I apply it, please provide a
> ChangeLog for the patch and also file a bug report and give me the bug
> number.
>
> Siddhesh
I opened:
https://sourceware.org/bugzilla/show_bug.cgi?id=17805
--
H.J.
More information about the Libc-alpha
mailing list