[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
Re: Random floating point exceptions and interrupt handler re-entry
- From: Andreas Johansson <ajo wopr campus luth se>
- To: Red Hat Alpha mailing list <axp-list redhat com>
- Subject: Re: Random floating point exceptions and interrupt handler re-entry
- Date: Sat, 16 May 1998 12:30:03 +0200 (CEST)
On Fri, 15 May 1998, Christian Worley wrote:
> 1) Programs with lots of floating point code die randomly with
> floating point exceptions when the network is at peak load (running
> netperfs and such).
I instantly recognize this problem from my own driver hacking. Have you
compiled your driver with the -mno-fp-regs flag? Otherwise gcc will
sometimes use fp registers for move purposes to save integer registers.
The kernel does NOT save the floting point registers when enterred (e.g.
for an interrupt), so it it is reqired that all objects/modules in the
kernel are NOT using fp-regs at all.
> 2) The interrupt handlers, in general, seem to be getting re-entered,
> even though interrupts are disabled.
>
> 3) The yellowfin drivers stops responding, and needs to be
> rmmod/insmod to get it going again.
Sorry, I can't help you with the other problems.
/Andreas
[Date Prev][Date Next] [Thread Prev][Thread Next]
[Thread Index]
[Date Index]
[Author Index]
[]