In <20000602164716.D2893@mclinux.com> Josh Huber (huber@missioncriticallinux.com) wrote:
> Is there some sort of precedent/procedure for reserving syscall table
> entries?
Very few precedents. No procedure (not since nobody setuped web server
with database but since syscalls should not be added without deep need).
I can recall just one: streams syscalls. So it's
a) Possible (= was done at least once before)
b) Very hard (as there are VERY few such syscalls)
Plus in this case it's kind of standard. Not implemented in standard linux
by delibirate decision but still standard.
> I'm working on getting a new version of our crash dump system
> ready, and binary compatibility for the user-space part between kernel
> versions would be a big plus for users.
First you should convince Linus that
a) this thing really needed for someone (= is REALLY helpful and not just
needed to add one more point to press-release)
b) can not be added to kernel without adding syscall
Both steps are far from trivial.
> I'm told that Linus is somewhat anti debugging systems in the kernel,
Not anti-debugging. His opinion is simple: kernel debuggers tend to affect
system functionaly and it's NOT what you want when kernel works non-properly
and when kernel works properly why you need kernel debugger in first place ?
Some debugging facilities can be added (was added, will be added) if it does
not affect work of kernel when there are no bugs... On other hand he has VERY
strong opinion AGAINST crash dump systems. EVEN if it does not affect normal
system behaviour. Of course he can change his mind sometime in future but
before it happens probability of getting ANY support for such system
(including syscall numbers) is exactly zero.
> so it seems like the only route to make using our system and painless as
> possible would be to have system call entries marked as unimplemented, which
> would be filled in when you apply our patch.
> Comments? Suggestions?
First step is to convince Linus that availability of you system is helping
linux not hurting it. Good luck is all what can say :-)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:16 EST