Re: [PATCH 1/2] tmpfs mempolicy: fix /proc/mounts corrupting memory

[KOSAKI Motohiro]

On Wed, Jan 2, 2013 at 5:01 AM, Hugh Dickins <hughd@xxxxxxxxxx> wrote:
> Recently I suggested using "mount -o remount,mpol=local /tmp" in NUMA
> mempolicy testing.  Very nasty.  Reading /proc/mounts, /proc/pid/mounts
> or /proc/pid/mountinfo may then corrupt one bit of kernel memory, often
> in a page table (causing "Bad swap" or "Bad page map" warning or "Bad
> pagetable" oops), sometimes in a vm_area_struct or rbnode or somewhere
> worse.  "mpol=prefer" and "mpol=prefer:Node" are equally toxic.
>
> Recent NUMA enhancements are not to blame: this dates back to 2.6.35,
> when commit e17f74af351c "mempolicy: don't call mpol_set_nodemask()
> when no_context" skipped mpol_parse_str()'s call to mpol_set_nodemask(),
> which used to initialize v.preferred_node, or set MPOL_F_LOCAL in flags.
> With slab poisoning, you can then rely on mpol_to_str() to set the bit
> for node 0x6b6b, probably in the next page above the caller's stack.
>
> mpol_parse_str() is only called from shmem_parse_options(): no_context
> is always true, so call it unused for now, and remove !no_context code.
> Set v.nodes or v.preferred_node or MPOL_F_LOCAL as mpol_to_str() might
> expect.  Then mpol_to_str() can ignore its no_context argument also,
> the mpol being appropriately initialized whether contextualized or not.
> Rename its no_context unused too, and let subsequent patch remove them
> (that's not needed for stable backporting, which would involve rejects).
>
> I don't understand why MPOL_LOCAL is described as a pseudo-policy:
> it's a reasonable policy which suffers from a confusing implementation
> in terms of MPOL_PREFERRED with MPOL_F_LOCAL.  I believe this would be
> much more robust if MPOL_LOCAL were recognized in switch statements
> throughout, MPOL_F_LOCAL deleted, and MPOL_PREFERRED use the (possibly
> empty) nodes mask like everyone else, instead of its preferred_node
> variant (I presume an optimization from the days before MPOL_LOCAL).
> But that would take me too long to get right and fully tested.
>
> Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx
> ---
>
>  mm/mempolicy.c |   64 +++++++++++++++++++----------------------------
>  1 file changed, 26 insertions(+), 38 deletions(-)
>
> --- 3.8-rc1/mm/mempolicy.c      2012-12-22 09:43:27.636015582 -0800
> +++ linux/mm/mempolicy.c        2013-01-01 23:44:10.715017466 -0800
> @@ -2595,8 +2595,7 @@ void numa_default_policy(void)
>   */
>
>  /*
> - * "local" is pseudo-policy:  MPOL_PREFERRED with MPOL_F_LOCAL flag
> - * Used only for mpol_parse_str() and mpol_to_str()
> + * "local" is implemented internally by MPOL_PREFERRED with MPOL_F_LOCAL flag.
>   */
>  static const char * const policy_modes[] =
>  {
> @@ -2610,28 +2609,21 @@ static const char * const policy_modes[]
>
>  #ifdef CONFIG_TMPFS
>  /**
> - * mpol_parse_str - parse string to mempolicy
> + * mpol_parse_str - parse string to mempolicy, for tmpfs mpol mount option.
>   * @str:  string containing mempolicy to parse
>   * @mpol:  pointer to struct mempolicy pointer, returned on success.
> - * @no_context:  flag whether to "contextualize" the mempolicy
> + * @unused:  redundant argument, to be removed later.
>   *
>   * Format of input:
>   *     <mode>[=<flags>][:<nodelist>]
>   *
> - * if @no_context is true, save the input nodemask in w.user_nodemask in
> - * the returned mempolicy.  This will be used to "clone" the mempolicy in
> - * a specific context [cpuset] at a later time.  Used to parse tmpfs mpol
> - * mount option.  Note that if 'static' or 'relative' mode flags were
> - * specified, the input nodemask will already have been saved.  Saving
> - * it again is redundant, but safe.
> - *
>   * On success, returns 0, else 1
>   */
> -int mpol_parse_str(char *str, struct mempolicy **mpol, int no_context)
> +int mpol_parse_str(char *str, struct mempolicy **mpol, int unused)
>  {
>         struct mempolicy *new = NULL;
>         unsigned short mode;
> -       unsigned short uninitialized_var(mode_flags);
> +       unsigned short mode_flags;
>         nodemask_t nodes;
>         char *nodelist = strchr(str, ':');
>         char *flags = strchr(str, '=');
> @@ -2719,24 +2711,23 @@ int mpol_parse_str(char *str, struct mem
>         if (IS_ERR(new))
>                 goto out;
>
> -       if (no_context) {
> -               /* save for contextualization */
> -               new->w.user_nodemask = nodes;
> -       } else {
> -               int ret;
> -               NODEMASK_SCRATCH(scratch);
> -               if (scratch) {
> -                       task_lock(current);
> -                       ret = mpol_set_nodemask(new, &nodes, scratch);
> -                       task_unlock(current);
> -               } else
> -                       ret = -ENOMEM;
> -               NODEMASK_SCRATCH_FREE(scratch);
> -               if (ret) {
> -                       mpol_put(new);
> -                       goto out;
> -               }
> -       }
> +       /*
> +        * Save nodes for mpol_to_str() to show the tmpfs mount options
> +        * for /proc/mounts, /proc/pid/mounts and /proc/pid/mountinfo.
> +        */
> +       if (mode != MPOL_PREFERRED)
> +               new->v.nodes = nodes;
> +       else if (nodelist)
> +               new->v.preferred_node = first_node(nodes);
> +       else
> +               new->flags |= MPOL_F_LOCAL;
> +
> +       /*
> +        * Save nodes for contextualization: this will be used to "clone"
> +        * the mempolicy in a specific context [cpuset] at a later time.
> +        */
> +       new->w.user_nodemask = nodes;
> +
>         err = 0;

Ugh, Indeed.
I should have realized this mistake at my last full review.



>
>  out:
> @@ -2756,13 +2747,13 @@ out:
>   * @buffer:  to contain formatted mempolicy string
>   * @maxlen:  length of @buffer
>   * @pol:  pointer to mempolicy to be formatted
> - * @no_context:  "context free" mempolicy - use nodemask in w.user_nodemask
> + * @unused:  redundant argument, to be removed later.
>   *
>   * Convert a mempolicy into a string.
>   * Returns the number of characters in buffer (if positive)
>   * or an error (negative)
>   */
> -int mpol_to_str(char *buffer, int maxlen, struct mempolicy *pol, int no_context)
> +int mpol_to_str(char *buffer, int maxlen, struct mempolicy *pol, int unused)
>  {
>         char *p = buffer;
>         int l;
> @@ -2788,7 +2779,7 @@ int mpol_to_str(char *buffer, int maxlen
>         case MPOL_PREFERRED:
>                 nodes_clear(nodes);
>                 if (flags & MPOL_F_LOCAL)
> -                       mode = MPOL_LOCAL;      /* pseudo-policy */
> +                       mode = MPOL_LOCAL;
>                 else
>                         node_set(pol->v.preferred_node, nodes);
>                 break;
> @@ -2796,10 +2787,7 @@ int mpol_to_str(char *buffer, int maxlen
>         case MPOL_BIND:
>                 /* Fall through */
>         case MPOL_INTERLEAVE:
> -               if (no_context)
> -                       nodes = pol->w.user_nodemask;
> -               else
> -                       nodes = pol->v.nodes;
> +               nodes = pol->v.nodes;
>                 break;

Hmm. yes, tmpfs mempolicy shoule be out of cpuset contextualization.
then when no_context is true, w.user_nodemask is alywas same v.nodes.

But note, I'd like to change this to make memory hot-plug safe. then I
may resurrect
this test for distinguish before and after hot plugging.

Anyway, I have no seen any problems.

Acked-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/