On Mon, Jun 01, 2020 at 03:45:45PM +0100, Catalin Marinas wrote:
> On Mon, Jun 01, 2020 at 09:55:38AM +0100, Dave P Martin wrote:
> > On Thu, May 28, 2020 at 05:34:13PM +0100, Catalin Marinas wrote:
> > > On Thu, May 28, 2020 at 12:05:09PM +0100, Szabolcs Nagy wrote:
> > > > The 05/28/2020 10:14, Catalin Marinas wrote:
> > > > > On Wed, May 27, 2020 at 11:57:39AM -0700, Peter Collingbourne wrote:
> > > > > > Should the userspace stack always be mapped as if with PROT_MTE if the
> > > > > > hardware supports it? Such a change would be invisible to non-MTE
> > > > > > aware userspace since it would already need to opt in to tag checking
> > > > > > via prctl. This would let userspace avoid a complex stack
> > > > > > initialization sequence when running with stack tagging enabled on the
> > > > > > main thread.
> > > > >
> > > > > I don't think the stack initialisation is that difficult. On program
> > > > > startup (can be the dynamic loader). Something like (untested):
> > > > >
> > > > > register unsigned long stack asm ("sp");
> > > > > unsigned long page_sz = sysconf(_SC_PAGESIZE);
> > > > >
> > > > > mprotect((void *)(stack & ~(page_sz - 1)), page_sz,
> > > > > PROT_READ | PROT_WRITE | PROT_MTE | PROT_GROWSDOWN);
> > > > >
> > > > > (the essential part it PROT_GROWSDOWN so that you don't have to specify
> > > > > a stack lower limit)
> > > >
> > > > does this work even if the currently mapped stack is more than page_sz?
> > > > determining the mapped main stack area is i think non-trivial to do in
> > > > userspace (requires parsing /proc/self/maps or similar).
> > >
> > > Because of PROT_GROWSDOWN, the kernel adjusts the start of the range
> > > down automatically. It is potentially problematic if the top of the
> > > stack is more than a page away and you want the whole stack coloured. I
> > > haven't run a test but my reading of the kernel code is that the stack
> > > vma would be split in this scenario, so the range beyond sp+page_sz
> > > won't have PROT_MTE set.
> > >
> > > My assumption is that if you do this during program start, the stack is
> > > smaller than a page. Alternatively, could we use argv or envp to
> > > determine the top of the user stack (the bottom is taken care of by the
> > > kernel)?
> >
> > I don't think you can easily know when the stack ends, but perhaps it
> > doesn't matter.
> >
> > From memory, the initial stack looks like:
> >
> > argv/env strings
> > AT_NULL
> > auxv
> > NULL
> > env
> > NULL
> > argv
> > argc <--- sp
> >
> > If we don't care about tagging the strings correctly, we could step to
> > the end of auxv and tag down from there.
> >
> > If we do care about tagging the strings, there's probably no good way
> > to find the end of the string area, other than looking up sp in
> > /proc/self/maps. I'm not sure we should trust all past and future
> > kernels to spit out the strings in a predictable order.
>
> I don't think we care about tagging whatever the kernel places on the
> stack since the argv/envp pointers are untagged. An mprotect(PROT_MTE)
> may or may not cover the environment but it shouldn't matter as the
> kernel clears the tags on the corresponding pages anyway.
We have no match-all tag, right? So we do rely on the tags being
cleared for the initial stack contents so that using untagged pointers
to access it works.
> AFAIK stack tagging works by colouring a stack frame on function entry
> and clearing the tags on return. We would only hit a problem if the
> function issuing mprotect(sp, PROT_MTE) on and its callers already
> assumed a PROT_MTE stack. Without PROT_MTE, an STG would be
> write-ignore, so subsequently turning it on would lead to a mismatch
> between the pointer and the allocation tags.
>
> So PROT_MTE turning on should happen very early in the user process
> startup code before any code with stack tagging enabled. Whether you
> reach the top of the stack with such mprotect() doesn't really matter
> since up to that point there should not be any use of stack tagging. If
> that's not possible, for example the glibc code setting up the stack was
> compiled to stack tagging itself, the kernel would have to enable it
> when the user process starts. However, I'd only do this based on some
> ELF note.
Sounds fair.
This early on, the process shouldn't be exposed to arbitrary, untrusted
data. So it's probably not a problem that tagging isn't turned on right
from the start.
Cheers
---Dave
