Re: [PATCH v10 00/11] malloc hooks removal

[Stafford Horne <shorne@gmail.com>]

On Tue, Jul 20, 2021 at 3:48 AM Siddhesh Poyarekar via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> This patchset removes the malloc hooks __malloc_hook, __free_hook,
> __realloc_hook and __memalign_hook from the API and leaves compatibility
> symbols so that existing applications can continue to link to them.  The
> reading and execution of the hooks has been moved to a DSO
> libc_malloc_debug.so, which can be preloaded for applications that need
> it.  By default these hooks no longer have any effect in the library.
>
> Further, debugging features such as MALLOC_CHECK_, mcheck() and mtrace
> have been weaned away from these hooks and also moved to
> libc_malloc_debug.so.  With this change, these features are only enabled
> when libc_malloc_debug.so is preloaded using LD_PRELOAD.
>
> Finally, the __morecore, __morecore_after_hook and __default_morecore
> hooks have also been moved to compat symbols and removed from the API.
> Existing applications will continue to link to them but they won't have
> any effect on malloc behaviour.

Hello,

I am working on an OpenRISC port and have just rebased and picked up
this series.  I
also have updated the api version to 2.35.  After this all of the
mtrace related *-mem
tests seem to be failing.

Is this expected to be completely removed after going to 2.35 and the
compatibility APIs
get compiled out?  If so, then the tests should be skipped, if not
then then something is broken
for glibcs with base version 2.35:

For example:

    cat sysdeps/unix/sysv/linux/or1k/shlib-versions
    DEFAULT                 GLIBC_2.35
    ld=ld-linux-or1k.so.1

    FAIL: posix/bug-glob2-mem
    FAIL: posix/bug-regex14-mem
    FAIL: posix/bug-regex2-mem
    FAIL: posix/bug-regex21-mem
    FAIL: posix/bug-regex31-mem
    FAIL: posix/bug-regex36-mem
    FAIL: malloc/tst-mtrace.

When I have looked into mtrace I can see a few strange things:

  In libc.so mtrace is a nop (that seems to be what you wanted for this series).
    000b5bb4 <mtrace>:
       b5bb4:       44 00 48 00     l.jr r9
       b5bb8:       15 00 00 00     l.nop 0x0

    000b5bbc <muntrace>:
       b5bbc:       44 00 48 00     l.jr r9
       b5bc0:       15 00 00 00     l.nop 0x0

Also, I am also seeing basically nothing in libc_malloc_debug.so  either:

    < shorne@antec ~/work/gnu-toolchain/build-glibc > nm
../build-glibc/malloc/libc_malloc_debug.so
    000000f4 r __abi_tag
    00004020 b completed.0
             w __cxa_finalize@GLIBC_2.35
    00000314 t deregister_tm_clones
    00000404 t __do_global_dtors_aux
    00003f0c d __do_global_dtors_aux_fini_array_entry
    00004000 d __dso_handle
    00003f10 a _DYNAMIC
    00000484 t frame_dummy
    00003f08 d __frame_dummy_init_array_entry
    0000048c r __FRAME_END__
    00000000 A GLIBC_2.35
    00004004 a _GLOBAL_OFFSET_TABLE_
             w _ITM_deregisterTMCloneTable
             w _ITM_registerTMCloneTable
    0000037c t register_tm_clones
    00004004 d __TMC_END__

-Stafford