[Top] [All Lists]

Re: sys_syscall patch.

To: Ralf Baechle <>
Subject: Re: sys_syscall patch.
From: "Maciej W. Rozycki" <>
Date: Mon, 24 Jun 2002 16:49:00 +0200 (MET DST)
Cc: Carsten Langgaard <>,
In-reply-to: <>
Organization: Technical University of Gdansk
On Mon, 24 Jun 2002, Ralf Baechle wrote:

> It's in the kernel for no better reason than Risc/OS and IRIX having this
> syscall.  Also the glibc syscall implementation was historically broken
> wrt. syscall restarting and a few other subtilities.

 Well, userland implementations for other archs seem quite
straightforward.  So should be ours -- we only have to shuffle arguments
appropriately.  Restarting is easy -- we just have to make sure to reload
v0 just before "syscall" reliably (we can use a static register or an
automatic variable to preserve it).  What are the few other subtleties?

 Also I can't see an implementation of syscall() for MIPS/Linux anywhere
in glibc.  What implementation do you refer to?  The Mach one?

 The win is we don't have to mess with user accesses specially -- the
final syscall will handle them. 

+  Maciej W. Rozycki, Technical University of Gdansk, Poland   +
+        e-mail:, PGP key available        +

<Prev in Thread] Current Thread [Next in Thread>