Re: [PATCH v3 19/23] arm64: mte: Add PTRACE_{PEEK,POKE}MTETAGS support

[Catalin Marinas <catalin.marinas@arm.com>]

On Wed, May 13, 2020 at 01:45:27PM -0300, Luis Machado wrote:
> On 5/13/20 12:09 PM, Luis Machado wrote:
> > Let me think about this for a bit. I'm trying to factor in the
> > /proc/<pid>/maps contents. If debuggers know which pages have PROT_MTE
> > set, then we can teach the tools not to PEEK/POKE tags from/to those
> > memory ranges, which simplifies the error handling a bit.
> 
> I was checking the output of /proc/<pid>/maps and it doesn't seem to contain
> flags against which i can match PROT_MTE. It seems /proc/<pid>/smaps is the
> one that contains the flags (mt) for MTE. Am i missing something?

You are right, the smaps is the one with the MTE information.

> Is this the only place debuggers can check for PROT_MTE? If so, that's
> unfortunate. /proc/<pid>/smaps doesn't seem to be convenient for parsing.

We can't change 'maps' as it's a pretty standard format with rwxp
properties only.

If you don't want to check any /proc file, just attempt to read the tags
and check the ptrace return code. The downside is that you can't easily
probe if a process is using MTE or not. But is this piece of information
relevant? The gdb user should know what to look for (well, it's been a
while since I used a debugger ;)).

-- 
Catalin