"GOT", but the "O" is a cute, smiling pufferfish. Index | Thread | Search

From:
Omar Polo <op@omarpolo.com>
Subject:
Re: [portable] support for landlock ABI v2 and v3
To:
Bryan Steele <brynet@gmail.com>
Cc:
gameoftrees@openbsd.org
Date:
Sun, 16 Jul 2023 00:18:45 +0200

Download raw body.

Thread
On 2023/07/15 15:38:31 -0400, Bryan Steele <brynet@gmail.com> wrote:
> On Sat, Jul 15, 2023 at 03:28:42PM -0400, Bryan Steele wrote:
> > On Sat, Jul 15, 2023 at 04:19:17PM +0200, Omar Polo wrote:
> > > commit 50e298d16389ca1854169f765a1ea4362521c056 (portable)
> > > from: Omar Polo <op@omarpolo.com>
> > > date: Sat Jul 15 13:57:18 2023 UTC
> > >  
> > >  support landlock ABI v2 and v3
> > >  
> > >  Add the right #ifdef for backward-compatibility and block REFER and
> > >  TRUNCATE as well, otherwise they're silently and implicitly allowed.
> > 
> > Definitely saw this coming.. :\
> > 
> > If this is for backwards-compat, is there any reason to not just #define
> > these to 0? Do older Linux kernels handle having "future" bits set?
> 
> Nevermind.. sigh..

yeah, landlock doesn't handle having bits sets from a "future" version
ABI (not actually tested, only gathered from the documentation.)

But since we're talking about Linux there's no required match between
the kernel version at runtime, the one used to compile the binary and
what the source code thinks the kernel should have.  One could build
this on a older kernel (landlock ABI v1 or v2) and then run the binary
on an updated kernel... or build on the new version and run on the
older.

:/