On 7/30/21 12:02 PM, Corentin Jabot wrote:

On Fri, Jul 30, 2021 at 5:41 PM Tom Honermann <tom@honermann.net> wrote:
On 7/30/21 10:56 AM, Thiago Macieira via SG16 wrote:
> On Thursday, 29 July 2021 19:11:03 PDT Charlie Barto wrote:
>> If such functionality were to be added it would be neat if the linux
>> implementations worked even if procfs were not mounted (such as when the
>> process is init) unless I'm mistaken fetching the command line through proc
>> for your own process is making like two system calls to de-reference a
>> pointer into your own address space.
> Going off-topic here, but correcting the above.
> Procfs is not required for one's own command-line and environment in any Unix
> I know of. They are duplicated in Linux's /proc, but they are added to the
> process' virtual address space just above the stack (and below the auxv).

True today, but I suspect their location is subject to change (though
probably unlikely).

Is there a straight forward way to determine their location in memory?

Do we need to?

It seems like a robust (yet painful) solution would be to ask the language below (c folks) to make this happen
(When glibc calls main it can put argv in a global)

I only asked because I needed this functionality in the past (10+ years ago) and never found a good solution for Linux.  Most of the other OSs I needed a solution for did expose these in global variables (z/OS was another notable exception; I never was able to get a properly delimited argv there other than from main()).



> Trivia: if you modify that area, the /proc files also change, allowing you to
> change the name presented in ps and similar tools.
> $ ps -o pid,command -Cfile.so
>     4873 file.so [kdeinit5] file local:/run/user/1000/klauncherdAEtky.1.slave-
> socket local:/run/user/1000/kded5hvkirU.1.slave-socket
>     4874 file.so [kdeinit5] file local:/run/user/1000/klauncherdAEtky.1.slave-
> socket local:/run/user/1000/kded5FwKgfB.2.slave-socket