|
Message
From: Marko Mlinar <markom@o...>
Date: Fri, 4 Jul 2003 08:13:03 +0200
Subject: Re: [openrisc] l.sys on delay slot
On Thursday 03 July 2003 19:46, Carlos Sánchez de La Lama wrote:
> Hi everybody,
>
> According to the manual, when a l.sys instruction in a delay slot is
> executed EPCR is loaded with address of corresponding jump instruction.
> However, this will result on an infinite loop, as the l.sys will be
> executed again. On a l.trap this is OK as traps are usually software
> breakpoints and the trap instruction will be removed by the service
> routine.
>
> Althought the routine servicing system calls could detect whether the
> l.sys was on a delay slot (by means of SR[DSX]), I think it will be
> better if EPCR was loaded with jump destination address in such cases.
>
> Please post your comments.
as far as I can remember this is correct. The reason is that l.sys instruction
holds immediate, which is almost always parsed in system call.
Is this feature used in uclinux?
Marko
|
 |
Browse
