Re: [PATCH v2 4/4] malloc: Add Huge Page support for sysmalloc

[Matheus Castanho <msc@linux.ibm.com>]

Adhemerval Zanella <adhemerval.zanella@linaro.org> writes:

> On 19/08/2021 14:58, Matheus Castanho wrote:
>> Hi Adhemerval,
>>
>> I tested this patchset on a POWER9, and I'm seeing the following test
>> failures when running make check with glibc.malloc.mmap_hugetlb=1:
>
> Thanks for checking on this.
>
>>
>> malloc/tst-free-errno
>> malloc/tst-free-errno-malloc-check
>> malloc/tst-free-errno-mcheck
>
> These one I couldn't really reproduce it on gcc farm power machines,
> a power9 with 2M huge page default and power8 with 16M default. Both
> didn't have any page allocated in the poll. I don't have admin access
> so I can change the pool size to check what is happening.
>
> I also tested on my x86_64 environment without any pages in the poll,
> with 4 pages in the pool and with 10 pages.
>

I confirm that without pages in the pool the tests pass correctly. Only
when I add them to the pool things start failing. In this case I'm
reserving 500 16 MB pages:

$ grep -i hugepages /proc/meminfo
AnonHugePages:         0 kB
ShmemHugePages:        0 kB
FileHugePages:         0 kB
HugePages_Total:     500
HugePages_Free:      500
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:      16384 kB

> If you could the stacktrace from where we get the
> "Didn't expect signal from child: got `Aborted'" it would be useful.
>

This is what GDB is showing me when the abort happens:

#0  0x00007ffff7dccf00 in __pthread_kill_internal (threadid=<optimized out>, signo=<optimized out>) at pthread_kill.c:44
#1  0x00007ffff7d6e26c in __GI_raise (sig=<optimized out>) at ../sysdeps/posix/raise.c:26
#2  0x00007ffff7d50490 in __GI_abort () at abort.c:79
#3  0x00007ffff7dba770 in __libc_message (action=<optimized out>, fmt=<optimized out>) at ../sysdeps/posix/libc_fatal.c:155
#4  0x00007ffff7ddc4e8 in malloc_printerr (str=<optimized out>, str@entry=0x7ffff7efdc90 "double free or corruption (out)") at malloc.c:5654
#5  0x00007ffff7ddefe8 in _int_free (av=0x7ffff7f60e30 <main_arena>, p=0x7ffff80203d0, have_lock=<optimized out>, have_lock@entry=0) at malloc.c:4555
#6  0x00007ffff7de2160 in __GI___libc_free (mem=<optimized out>) at malloc.c:3358
#7  0x0000000010001ee4 in do_test () at tst-free-errno.c:123
#8  0x0000000010002730 in run_test_function (argc=argc@entry=1, argv=argv@entry=0x7fffffffede0, config=config@entry=0x7fffffffe950) at support_test_main.c:232
#9  0x00000000100032fc in support_test_main (argc=1, argv=0x7fffffffede0, config=0x7fffffffe950) at support_test_main.c:431
#10 0x00000000100019d0 in main (argc=<optimized out>, argv=<optimized out>) at ../support/test-driver.c:168
#11 0x00007ffff7d50818 in __libc_start_call_main (main=main@entry=0x10001980 <main>, argc=argc@entry=1, argv=argv@entry=0x7fffffffede0, auxvec=auxvec@entry=0x7fffffffef68) at ../sysdeps/nptl/libc_start_call_main.h:58
#12 0x00007ffff7d50a00 in generic_start_main (fini=<optimized out>, stack_end=<optimized out>, rtld_fini=<optimized out>, init=<optimized out>, auxvec=<optimized out>, argv=<optimized out>, argc=<optimized out>, main=<optimized out>) at ../csu/libc-start.c:409
#13 __libc_start_main_impl (argc=1, argv=0x7fffffffede0, ev=<optimized out>, auxvec=0x7fffffffef68, rtld_fini=<optimized out>, stinfo=<optimized out>, stack_on_entry=<optimized out>) at ../sysdeps/unix/sysv/linux/powerpc/libc-start.c:98
#14 0x0000000000000000 in ?? ()

> It could be also something related to /proc/sys/vm/max_map_count
> value, since it mmap seems to be failing for some reason.
>

This is what the machine I'm using now has:

$ cat /proc/sys/vm/max_map_count
65530

>> posix/tst-exec
>> posix/tst-exec-static
>> posix/tst-spawn
>> posix/tst-spawn-static
>> posix/tst-spawn5
>
> These are an overlook at 'malloc_default_hugepage_size()' where it
> does not close the file descriptor on success.  I have fixed it.

Ok, thanks!

--
Matheus Castanho