On 2026-04-30 12:23, Jens Remus wrote:
On 4/29/2026 10:03 PM, Steven Rostedt wrote:
On Wed, 29 Apr 2026 14:58:15 -0400
Steven Rostedt <[email protected]> wrote:
If the main executable doesn't have a sframe section, I don't see why this
shouldn't be allowed to add one.
The address is added via mtree_insert_range() which states it will return
-EEXISTS if the range is occupied.
I confirmed this. I changed the test program to load its own sframe section
again, and it errors out with -EEXISTS.
Should an executable be allowed to remove its main executable sframe
section? Currently the kernel takes care of registering/unregistering
that.
I wonder if there are programs out there that replace their exec'd
image with mmap MAP_FIXED, or just rewrite their code without exec ?
If so, it would be good to allow those program to unregister the
initial sframe and perhaps re-register a new one. JITs perhaps ?
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
