This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 5/7] elf: Enable relro for static build
- From: Adhemerval Zanella <adhemerval dot zanella at linaro dot org>
- To: Florian Weimer <fw at deneb dot enyo dot de>
- Cc: libc-alpha at sourceware dot org
- Date: Mon, 2 Dec 2019 10:57:50 -0300
- Subject: Re: [PATCH 5/7] elf: Enable relro for static build
- References: <20191129210327.26434-1-adhemerval.zanella@linaro.org> <20191129210327.26434-5-adhemerval.zanella@linaro.org> <874kykfjyy.fsf@mid.deneb.enyo.de>
On 01/12/2019 06:55, Florian Weimer wrote:
> * Adhemerval Zanella:
>
>> The code is similar to the one at rtld.c, where its check for the
>> PT_GNU_RELRO header values from program headers and call
>> _dl_protected_relro with the updated l_relro_{addr,size} values.
>
> This is not the actual code that does RELRO in most cases, it's only
> used with prelink. _dl_relocate_object is what is used.
Ack, I changed the commit message to:
The code is similar to the one at elf/dl-reloc.c, where it checks for
the l_relro_size from the link_map (obtained from PT_GNU_RELRO header
from program headers) and calls_dl_protected_relro.
>
>> diff --git a/elf/dl-support.c b/elf/dl-support.c
>> index 5526d5ee6e..bdb5c2ae91 100644
>> --- a/elf/dl-support.c
>> +++ b/elf/dl-support.c
>> @@ -367,14 +367,24 @@ _dl_non_dynamic_init (void)
>> if (_dl_platform != NULL)
>> _dl_platformlen = strlen (_dl_platform);
>>
>> - /* Scan for a program header telling us the stack is nonexecutable. */
>> if (_dl_phdr != NULL)
>> - for (uint_fast16_t i = 0; i < _dl_phnum; ++i)
>> - if (_dl_phdr[i].p_type == PT_GNU_STACK)
>> + for (const ElfW(Phdr) *ph = _dl_phdr; ph < &_dl_phdr[_dl_phnum]; ++ph)
>> + switch (ph->p_type)
>> {
>> - _dl_stack_flags = _dl_phdr[i].p_flags;
>> + /* Check if the stack is nonexecutable. */
>> + case PT_GNU_STACK:
>> + _dl_stack_flags = ph->p_flags;
>> + break;
>> +
>> + case PT_GNU_RELRO:
>> + _dl_main_map.l_relro_addr = ph->p_vaddr;
>> + _dl_main_map.l_relro_size = ph->p_memsz;
>> break;
>> }
>> +
>> + /* Setup relro on the binary itself. */
>> + if (_dl_main_map.l_relro_size)
>> + _dl_protect_relro (&_dl_main_map);
>
> Please use an explicit comparison with != 0.
Ack.
>
> I have a test case for this which I can post.
Sure, I can attach on the patch itself.
> Somewhat bizarrely,
> full RELRO for statically linked binaries requires linking with -z now.
>
My understanding it is arch-specific and also depends on how bintuils was
build. For instance, with my system ld (GNU ld (GNU Binutils for Ubuntu) 2.30)
seemed to be built with DEFAULT_LD_Z_RELRO (set by --enable-relro) which sets
relro by default. With this binutils I could only disable relro by explicit
add norelro, the -z {lazy,now} did not change the GNU_RELRO header creation.
Also the code in ld/emultempl/elf.em does seem to select different linker
scripts for both link_info.relro and (link_info.flags & DF_BIND_NOW), however
it does no have a special case for
link_info.relro and !(link_info.flags & DF_BIND_NOW). I don't see how -relro
is requiring -z now on ld code.