[PATCH v3 1/2] aarch64: align address for BTI protection [BZ #26988]

Adhemerval Zanella adhemerval.zanella@linaro.org
Thu Dec 10 18:49:37 GMT 2020



On 02/12/2020 05:55, Szabolcs Nagy via Libc-alpha wrote:
> Handle unaligned executable load segments (the bfd linker is not
> expected to produce such binaries, but other linkers may).
> 
> Computing the mapping bounds follows _dl_map_object_from_fd more
> closely now.
> 
> Fixes bug 26988.

LGTM, thanks.

Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>

> ---
> v3:
> - split the last patch in two so this bug is fixed separately.
> - pushed to nsz/btifix-v3 branch.
> 
>  sysdeps/aarch64/dl-bti.c | 14 ++++++++------
>  1 file changed, 8 insertions(+), 6 deletions(-)
> 
> diff --git a/sysdeps/aarch64/dl-bti.c b/sysdeps/aarch64/dl-bti.c
> index 8f4728adce..67d63c8a73 100644
> --- a/sysdeps/aarch64/dl-bti.c
> +++ b/sysdeps/aarch64/dl-bti.c
> @@ -20,19 +20,22 @@
>  #include <libintl.h>
>  #include <ldsodefs.h>
>  
> -static int
> +static void
>  enable_bti (struct link_map *map, const char *program)
>  {

Ok.

> +  const size_t pagesz = GLRO(dl_pagesize);
>    const ElfW(Phdr) *phdr;
> -  unsigned prot;
>  
>    for (phdr = map->l_phdr; phdr < &map->l_phdr[map->l_phnum]; ++phdr)
>      if (phdr->p_type == PT_LOAD && (phdr->p_flags & PF_X))
>        {
> -	void *start = (void *) (phdr->p_vaddr + map->l_addr);
> -	size_t len = phdr->p_memsz;
> +	size_t vstart = ALIGN_DOWN (phdr->p_vaddr, pagesz);
> +	size_t vend = ALIGN_UP (phdr->p_vaddr + phdr->p_filesz, pagesz);
> +	off_t off = ALIGN_DOWN (phdr->p_offset, pagesz);
> +	void *start = (void *) (vstart + map->l_addr);
> +	size_t len = vend - vstart;
>  
> -	prot = PROT_EXEC | PROT_BTI;
> +	unsigned prot = PROT_EXEC | PROT_BTI;
>  	if (phdr->p_flags & PF_R)
>  	  prot |= PROT_READ;
>  	if (phdr->p_flags & PF_W)
> @@ -48,7 +51,6 @@ enable_bti (struct link_map *map, const char *program)
>  				N_("mprotect failed to turn on BTI"));
>  	  }
>        }
> -  return 0;
>  }
>  
>  /* Enable BTI for MAP and its dependencies.  */
> 

Ok.


More information about the Libc-alpha mailing list