lkml.org 
[lkml]   [1998]   [Oct]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: "per-process" limits (was: Showstopper list)
From
Date
pbd@Op.Net (Paul Barton-Davis) writes:

> >Someone wrote:
> >> I'd like to see per user as well as per process limits for kernel
> >> commitment, address space commitment etc. Those are really needed
> >> for "big iron" machines and Linux is moving more and more into that
> >> area.
>
> "Per process" limits for anything under Linux are difficult to
> envisage in a world of multi-threaded processes. Currently, Linux has
> no kernel abstraction to distinguish the kernel's notion of a task
> (aka "thread") and the unix/user-level notion of a process. Limits
> applied to a given task in a process (conventionally adequate for all
> single-threaded processes) would not apply to other kernel threads
> created using clone(). Presto, any "per process" limits are
> circumvented.
>
> Adding a kernel abstraction, such as a "group of tasks", would allow
> such limits to be implemented in a sane fashion, and is flexible
> enough to even allow the limits applied to groups of tasks not related
> via fork()/clone().

Such a concept is needed for POSIX compliant signal handling too. Also
it is needed to make wait4() Single Unix compliant (I know of at least one
important commercial program - Franz Allegro Common Lisp - that was forced
to go with userspace threads instead of kernel threads because of this)

Also there is another "bug"/"undocumented feature" that needs them to get fixed:
currently credentials like gids/uid/euid etc. are thread local, not process
local in Linux. This differs from all other Unixes (Solaris, UnixWare etc.)
and is probably not Single Unix compliant. For some special applications it
is nice to have (e.g. for a multi threaded non anonymous ftp server), but for
most threaded applications it is a portability problem and likely to introduce
security holes.

-Andi

-
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/

\
 
 \ /
  Last update: 2005-03-22 13:45    [W:0.062 / U:0.108 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site