Re: [PATCH 3/3] elf: Don't write past end of notes for regset gap

[Kees Cook <keescook@chromium.org>]

On Tue, Mar 15, 2022 at 01:17:06PM -0700, Rick Edgecombe wrote:
> In fill_thread_core_info() the ptrace accessible registers are collected
> to be written out as notes in a core file. The note array is allocated
> from a size calculated by iterating the user regset view, and counting the
> regsets that have a non-zero core_note_type. However, this only allows for
> there to be non-zero core_note_type at the end of the regset view. If
> there are any gaps in the middle, fill_thread_core_info() will overflow the
> note allocation, as it iterates over the size of the view and the
> allocation would be smaller than that.
> 
> There doesn't appear to be any arch that has gaps such that they exceed
> the notes allocation, but the code is brittle and tries to support
> something it doesn't. It could be fixed by increasing the allocation size,
> but instead just have the note collecting code utilize the array better.
> This way the allocation can stay smaller.
> 
> Even in the case of no arch's that have gaps in their regset views, this
> introduces a change in the resulting indicies of t->notes. It does not
> introduce any changes to the core file itself, because any blank notes are
> skipped in write_note_info().

Hm, yes, fill_note_info() does an initial count & allocate. Then
fill_thread_core_info() writes them.

> 
> This fix is derrived from an earlier one[0] by Yu-cheng Yu.
> 
> [0] https://lore.kernel.org/lkml/20180717162502.32274-1-yu-cheng.yu@intel.com/
> 
> Co-developed-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
> Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
> Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
> ---
>  fs/binfmt_elf.c | 15 ++++++++-------
>  1 file changed, 8 insertions(+), 7 deletions(-)
> 
> diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c
> index d61543fbd652..a48f85e3c017 100644
> --- a/fs/binfmt_elf.c
> +++ b/fs/binfmt_elf.c
> @@ -1757,7 +1757,7 @@ static int fill_thread_core_info(struct elf_thread_core_info *t,
>  				 const struct user_regset_view *view,
>  				 long signr, size_t *total)
>  {
> -	unsigned int i;
> +	unsigned int note_iter, view_iter;
>  
>  	/*
>  	 * NT_PRSTATUS is the one special case, because the regset data
> @@ -1777,11 +1777,11 @@ static int fill_thread_core_info(struct elf_thread_core_info *t,
>  
>  	/*
>  	 * Each other regset might generate a note too.  For each regset
> -	 * that has no core_note_type or is inactive, we leave t->notes[i]
> -	 * all zero and we'll know to skip writing it later.
> +	 * that has no core_note_type or is inactive, skip it.
>  	 */
> -	for (i = 1; i < view->n; ++i) {
> -		const struct user_regset *regset = &view->regsets[i];
> +	note_iter = 1;
> +	for (view_iter = 1; view_iter < view->n; ++view_iter) {
> +		const struct user_regset *regset = &view->regsets[view_iter];
>  		int note_type = regset->core_note_type;
>  		bool is_fpreg = note_type == NT_PRFPREG;
>  		void *data;
> @@ -1800,10 +1800,11 @@ static int fill_thread_core_info(struct elf_thread_core_info *t,
>  		if (is_fpreg)
>  			SET_PR_FPVALID(&t->prstatus);
>  

info->thread_notes contains the count. Since fill_thread_core_info()
passes a info member by reference, it might make sense to just pass info
itself, then the size can be written and a bounds-check can be added
here:

		if (WARN_ON_ONCE(i >= info->thread_notes))
			continue;

> -		fill_note(&t->notes[i], is_fpreg ? "CORE" : "LINUX",
> +		fill_note(&t->notes[note_iter], is_fpreg ? "CORE" : "LINUX",
>  			  note_type, ret, data);
>  
> -		*total += notesize(&t->notes[i]);
> +		*total += notesize(&t->notes[note_iter]);
> +		note_iter++;
>  	}
>  
>  	return 1;
> -- 
> 2.17.1
> 

If that can get adjusted, I'd be happy to carry this patch separately in
for-next/execve (or I can Ack it and it can go with the others here in
the series).

(And in a perfect world, I'd *love* a KUnit test to exercise this logic,
but I don't think we're there yet with function mocking, etc.)

-- 
Kees Cook