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

Re: [MiNT] [PATCH] ethernat.xif hangs the system if card isn't present

On 02/16/13 10:48, David Gálvez wrote:

2013/2/15 Alan Hourihane <alanh@fairlite.co.uk>:

On 02/15/13 11:28, David Gálvez wrote:

2013/2/15 Alan Hourihane <alanh@fairlite.co.uk>:

On 02/15/13 09:34, Vincent Rivičre wrote:

Is this the right way to do with FreeMiNT?
I thought there was high level functions like signal catching, etc...


Mmm, yes, I think it should be possible to do....

sigbus()
{
/* failed */
}

SIGACTION(get_curproc(), SIGBUS).sa_handler = (long) sigbus;


Before I made the patch I was unsure if it was a correct practice to
replace the bus error exception handler from a driver,
so I did some investigation, this is what I found out. I think
information below is correct but I'm not 100% sure.

Code inside xdd, xfs and xif modules is run like it was kernel code. Then
PID=0.

Take a look at the file sys/signal.c, specially the code below:

/*
   * check_sigs: see if we have any signals pending. if so,
   * handle them.
   */
void
check_sigs (void)
{
         struct proc *p = get_curproc();
         ulong sigs, sigm;
         int i;
         short deliversig;

         if (p->pid == 0)
                 return;

Returns if PID is 0 without doing anything, before calling
handle_sig() function where signals are handled.
So I think there is no high level way for signal catching for these
drivers.


Good spot. p->pid == 0 would be correct for the handling of kernel mode
drivers.

It would be nice though that drivers didn't have to go replacing the bus
error handler behind the kernels back.


When I was making the patch I also looked at Linux sources to see how
they were probing hardware, they have the function hwreg_present():

http://lxr.free-electrons.com/source/arch/m68k/mm/hwtest.c?a=m68k#L17

(I don't think that the comment about that the function is only used
before the VBR is set up is still right)

We could have a similar function in the kernel and export it to kentry
and kinfo structures to be used by drivers.



Sounds good to me.

Alan.