Re: [PATCH RFC 0/5] uprobes: kill xol vma

[Oleg Nesterov]

On 11/28, Oleg Nesterov wrote:
>
> On top of this series, not for inclusion yet, just to explain what
> I mean. May be someone can test it ;)
>
> This series kills xol_vma. Instead we use the per_cpu-like xol slots.
>
> This is much more simple and efficient. And this of course solves
> many problems we currently have with xol_vma.
>
> For example, we simply can not trust it. We do not know what actually
> we are going to execute in UTASK_SSTEP mode. An application can unmap
> this area and then do mmap(PROT_EXEC|PROT_WRITE, MAP_FIXED) to fool
> uprobes.
>
> The only disadvantage is that this adds a bit more arch-dependant
> code.
>
> The main question, can this work?

OK, it almost works.

But, this way we can't probe the compat tasks. A __USER32_CS task can't
access the fix_to_virt() area, so it can't use uprobe_xol_slots[].

Many thanks to Josh who noticed this.

I'll try to think more, but so far I do not see any simple solution.

Oleg.

--
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/