Hello,
I'm currently working on huge page support for ARM and have also been
working with libhugetlbfs.
I think I ran into the same problem with the malloc_manysmall unit test that
appears to have prompted the commit:
cdd2fa336d9980eb268a775b7cf9c49e3b698cd7 - "exhaust malloc arenas before
malloc tests"
For info, placing a breakpoint at "__GI___default_morecore" (the default
value of __morecore on my ARM Fedora 17 system with glibc 2.15) in gdb
before running the code, turned up an internal malloc (_int_malloc) in the
sscanf function (called in kernel-features.c and hugeutils.c). This malloc
was called before __morecore was re-directed, thus led to some non-huge
memory being returned from malloc when called in the unit test. Replacing
the sscanfs from kernel-features.c and hugeutils.c with some test code
allowed the malloc_manysmall unit test to pass with the consume_heap()
function calls removed.
Best,
--
Steve Capper
------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_nov
_______________________________________________
Libhugetlbfs-devel mailing list
Libhugetlbfs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
