Re: [PATCH] compatibility syscall layer (lets try again)

[Jim Houston <jim.houston@ccur.com>]

Linus Torvalds wrote:
> 
> On Wed, 4 Dec 2002, george anzinger wrote:
> >
> > As a suggestion for a solution for this, is it true that
> > regs, on a system call, will ALWAYS be at the end of the
> > stack?
> 
> No. Some architectures do not save enough state on the stack by default,
> and need to do more to use do_signal(). Look at alpha, for example - the
> default kernel stack doesn't contain all tbe registers needed, and
> the alpha do_signal() calling convention is different.
> 
> If you want to handle do_signal(), then you need to do _all_ of this in
> architecture-specific files. You simply cannot do what you want to do in a
> generic way.
> 
>                 Linus

Hi Linus,

Agreed!  In my alternative version of the Posix timers patch, I avoid
calling do_signal() from clock_nanosleep by using a variant of the 
existing ERESTARTNOHAND mechanism.  The problem I ran into was that I
could not tell on entry to clock_nanosleep if it was a new call or
an old one being restarted.  I solved this by adding a new 
ERESTARTNANOSLP error code and making a small change in do_signal().
The handling of ERESTARTNANOSLP is the same as ERESTARTNOHAND but also
sets a new flag in the task_struct before restarting the system call.

This is still an architecture-specific change but atleast it is simple.

Jim Houston - Concurrent Computer Corp.