Re: [PATCH] tst-p_alignmod3.so: Disable GNU_RELRO segment

["H.J. Lu" <hjl.tools@gmail.com>]

On Sun, Jan 30, 2022 at 6:36 PM Michael Hudson-Doyle
<michael.hudson@canonical.com> wrote:
>
>
>
> On Thu, 27 Jan 2022 at 10:41, H.J. Lu via Libc-alpha <libc-alpha@sourceware.org> wrote:
>>
>> tst-p_alignmod3.so has invalid p_align on LOAD segments which can't work
>> with GNU_RELRO.  Pass -z norelro to linker to disable GNU_RELRO segment
>> to trigger
>
>
> This helps for me on most Ubuntu architectures (s390x, arm64, ppc64el, probably armhf although that build hasn't finished yet) but I still see a failure on amd64 which still seems to hit the "cannot change memory protections" case (full log here: https://launchpad.net/~mwhudson/+archive/ubuntu/devirt/+build/23110381) and i386 (full log here: https://launchpad.net/~mwhudson/+archive/ubuntu/devirt/+build/23110384) where the loader seems to be segfaulting:
>
> (gdb) r
> Starting program: /build/glibc-xBQSrs/glibc-2.34.9000-596-g4556b6edae/build-tree/i386-libc/elf/ld-linux.so.2 --library-path ../build-tree/i386-libc:../build-tree/i386-libc/math:../build-tree/i386-libc/elf:../build-tree/i386-libc/dlfcn:../build-tree/i386-libc/nss:../build-tree/i386-libc/nis:../build-tree/i386-libc/rt:../build-tree/i386-libc/resolv:../build-tree/i386-libc/mathvec:../build-tree/i386-libc/support:../build-tree/i386-libc/nptl ../build-tree/i386-libc/elf/tst-p_align3
>
> Program received signal SIGSEGV, Segmentation fault.
> 0xf7fe9f08 in mprotect () at ../sysdeps/unix/syscall-template.S:120
> 120 T_PSEUDO (SYSCALL_SYMBOL, SYSCALL_NAME, SYSCALL_NARGS)
> (gdb) bt
> #0  0xf7fe9f08 in mprotect () at ../sysdeps/unix/syscall-template.S:120
> #1  0xf7fcd299 in _dl_map_segments (loader=<optimized out>, has_holes=true, maplength=2044, nloadcmds=<optimized out>, loadcmds=0xffffc650, type=3,
>     header=<optimized out>, fd=3, l=0xf7fb7a70) at ./dl-map-segments.h:116
> #2  _dl_map_object_from_fd (name=name@entry=0xf7fb9b03 "/build/glibc-xBQSrs/glibc-2.34.9000-596-g4556b6edae/build-tree/i386-libc/elf/tst-p_alignmod3.so",
>     origname=origname@entry=0x0, fd=<optimized out>, fbp=<optimized out>, realname=<optimized out>, loader=<optimized out>, l_type=<optimized out>, mode=<optimized out>,
>     stack_endp=<optimized out>, nsid=<optimized out>) at dl-load.c:1258
> #3  0xf7fcebcd in _dl_map_object (loader=0xf7ffda70, name=0xf7fb9b03 "/build/glibc-xBQSrs/glibc-2.34.9000-596-g4556b6edae/build-tree/i386-libc/elf/tst-p_alignmod3.so",
>     type=1, trace_mode=0, mode=0, nsid=<optimized out>) at dl-load.c:2327
> #4  0xf7fc8378 in openaux (a=0xffffcc98) at dl-deps.c:64
> #5  0xf7fdf8c4 in _dl_catch_exception (exception=0xffffcc8c, operate=0xf7fc8340 <openaux>, args=0xffffcc98) at dl-error-skeleton.c:208
> #6  0xf7fc87f0 in _dl_map_object_deps (map=<optimized out>, preloads=<optimized out>, npreloads=<optimized out>, trace_mode=<optimized out>, open_mode=<optimized out>)
>     at dl-deps.c:248
> #7  0xf7fe5bcf in dl_main (phdr=<optimized out>, phnum=<optimized out>, user_entry=<optimized out>, auxv=<optimized out>) at rtld.c:1969
> #8  0xf7fe1c66 in _dl_sysdep_start (start_argptr=0xffffd440, dl_main=0xf7fe3ba0 <dl_main>) at ../elf/dl-sysdep.c:256
> #9  0xf7fe393f in _dl_start_final (arg=0xffffd440) at rtld.c:506
> #10 _dl_start (arg=<optimized out>) at rtld.c:595
> #11 0xf7fe273b in _start () from /build/glibc-xBQSrs/glibc-2.34.9000-596-g4556b6edae/build-tree/i386-libc/elf/ld-linux.so.2
>
> I think but am not sure that "loadcmds[nloadcmds - 1].mapstart" is 0 and "c->mapend" is 4096 so mprotect is getting called with an insane len in this code:
>
>           if (__glibc_unlikely
>               (__mprotect ((caddr_t) (l->l_addr + c->mapend),
>                            loadcmds[nloadcmds - 1].mapstart - c->mapend,
>                            PROT_NONE) < 0))
>             return DL_MAP_SEGMENTS_ERROR_MPROTECT;
>         }
>
> but it's all pretty new to me. It's also possible that this is something about how Ubuntu's binutils is configured, I suppose.

tst-p_alignmod3.so is invalid and is very sensitive to how it is built.
But the loader shouldn't crash in any case.  Please provide i386 and
x86-64 tst-p_alignmod3.so so that I can fix it.

> Cheers,
> mwh
>
>>
>> .../elf/tst-p_alignmod3.so: ELF load command address/offset not page-aligned
>>
>> instead of
>>
>> .../elf/tst-p_alignmod3.so: cannot change memory protections
>> ---
>>  elf/Makefile | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/elf/Makefile b/elf/Makefile
>> index daafb5cf12..6229add1fc 100644
>> --- a/elf/Makefile
>> +++ b/elf/Makefile
>> @@ -2619,7 +2619,7 @@ $(objpfx)tst-p_alignmod2.so: $(objpfx)tst-p_alignmod-base.so
>>         cp $(objpfx)tst-p_alignmod-base.so $@
>>         $(PYTHON) $(..)scripts/tst-elf-edit.py -a 1 $@
>>
>> -LDFLAGS-tst-p_alignmod3.so += -Wl,-z,max-page-size=0x100,-z,common-page-size=0x100
>> +LDFLAGS-tst-p_alignmod3.so += -Wl,-z,max-page-size=0x100,-z,common-page-size=0x100,-z,norelro
>>
>>  $(objpfx)tst-p_align3: $(objpfx)tst-p_alignmod3.so
>>  $(objpfx)tst-p_align3.out: tst-p_align3.sh $(objpfx)tst-p_align3
>> --
>> 2.34.1
>>


-- 
H.J.