It looks like eina_mempool_from() is accessing invalid memory regions in some cases. It looks like that this problem happens when lots of eina_mempool_from() + eina_mempool_free() calls are happening. I've created a sample test program which valgrind will complain about the invalid memory access.
This problems happens using chained mempools and big_one.
here's the valgrind output
==2436== Invalid read of size 8 ==2436== at 0x4EB3045: eina_chained_mempool_from (eina_chained_mempool.c:423) ==2436== by 0x108A8A: eina_mempool_from (in /home/iscaro/mp) ==2436== by 0x108BBA: main (in /home/iscaro/mp) ==2436== Address 0x700d0a0 is 80 bytes inside a recently re-allocated block of size 12,240 alloc'd ==2436== at 0x4C2BE7F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so) ==2436== by 0x4EB287E: _eina_chained_mp_pool_new (eina_chained_mempool.c:134) ==2436== by 0x4EB2DAB: eina_chained_mempool_malloc (eina_chained_mempool.c:312) ==2436== by 0x1089F1: eina_mempool_calloc (in /home/iscaro/mp) ==2436== by 0x108B1B: main (in /home/iscaro/mp)