Re: Perform minimal RAM test at boot
H. Peter Anvin (hpa@transmeta.com)
1 Nov 1999 20:03:59 GMT
Followup to: <199911011425.PAA02084@p.steiner.dialin.t-online.de>
By author: Peter Steiner <p.steiner@t-online.de>
In newsgroup: linux.dev.kernel
>
>
> > > > What about the attached code? It's for 2.3.20 but it should be easy to
> > > > apply to other versions.
> > >
> > > Nice. But it my eyes it should panic, not try to mark pages reserved
> > > and continue. If there's one error, there may be undetected errors
> > > nearby. I'd suggest panic() and forcing user deal with that.
>
> I'm not sure if this is really necessary.
>
> > You're very right.
> >
> > GOOD memory tests find all errors that can be modelled.
>
> This is *not* a memory test. It's a memory map veryfier. I wrote a good
> memory test that uses special access patterns to stress test the whole
> memory subsystem. It takes half an hour on my 64MB machine to complete.
> That's to slow for a standard bootup. And I'm still not satisfied with it.
>
> The memory map veryfier is designed to detect wrong "mem=xxx" parameters,
> buggy BIOSses and memory holes (eg. from 15M to 16M). And it seems to work
> reliably.
>
I hope you don't check *outside* the predefined memory map, or you'll
die horribly on some systems. This kind of test certainly can be
useful to *remove* some stuff, but never add.
Also, don't expect it to detect anything reliable (think of a video
card in the 15-16M memory hole.) Rather, think of it as a "last
ditch" sanity check. It probably should print a major warning if it
ever needs to make any adjustments.
-hpa
--
<hpa@transmeta.com> at work, <hpa@zytor.com> in private!
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/