l4-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: status of l4-hurd


From: Michael Hohmuth
Subject: Re: status of l4-hurd
Date: 04 Sep 2001 02:14:16 +0200
User-agent: Gnus/5.0807 (Gnus v5.8.7) XEmacs/21.1 (Channel Islands)

Farid Hajji <address@hidden> writes:
> So to put it more clearly: glibc currently _needs_ a Linux "VM" (be it
> as a kernel or through an emulation library) to work. Porting glibc to
> L4 (whatever this means) would require writing sysdeps comparable to
> the Mach/Hurd stuff already present in glibc, right? 

Yes, that's true.  

> This seems like a formidable task (no pun intended).

Depending on where you draw the line with your kernel-abstraction
layer (KAL), you may be able to modify or reuse Glibc's Hurd port,
creating a kernel-independent Glibc that runs directly on top of the
KAL.  In other words, maybe you should take for granted that Glibc is
an integral part of the Hurd, and port it to your KAL together with
all the other Hurd components.

> address@hidden wrote:
> > However, there is a port of glibc to L4Linux that replaces the "int
> > 0x80" with a direct call into the emulation library (which is mapped
> > in all user processes and IPCs the L4Linux server) -- effectively
> > emulating the effect of "int 0x80" on L4Linux.  This version saves the
> > cost of a trampoline call (an extra trap into the microkernel).  (In
> > other words, that's the version we use for benchmarks. ;-)

> That is an obvious optimization. It certainly helps to save some
> cycles. If the emulation library caches some results like Lites (e.g.
> getpid() or some such), then it would even save a complete kernel
> call! 

Currently, our emulation library does not do any kind of
system-call--specific processing.  It just calls into the Linux
server.

> Did you generate this modified library manually, or do you use
> some kind of {sed,perl,awk...}script that would work on the most
> actual glibc versions?

I mostly forgot, but I seem to remember that Jean Wolter (who
implemented this hack) just modified a single macro in Glibc.  I'm
Cc'ing him this message and hope that he will elaborate.

Michael
-- 
address@hidden, address@hidden
http://home.pages.de/~hohmuth/




reply via email to

[Prev in Thread] Current Thread [Next in Thread]