Umm, guys?
This proposal disturbs me. I find the following exchange particularly
bothersome:
> -----Original Message-----
> From: Tigran Aivazian [mailto:tigran@aivazian.fsnet.co.uk]
>
> Jan Niehusmann wrote:
> > I don't know much about intel microcode, but perhaps a
> microcode update
> > _may_ change cache handling. So it might work with every microcode
> > update available now (because Intel didn't make any incompatible
> > changes), but may fail with the next one.
>
> the approach I take is pragmatic - something works until
> proven broken.
> If Intel give you microcode you just say thank you and use it. If it
> breaks something you complain or work the OS around it yourself. Of
> course it can change anything and break anything. But hopefully it
> doesn't :)
Now maybe I'm just biased by my validation experience, but the attitude I
always take is "it's broken unless demonstrated otherwise". I've had people
describe LILO as a loaded gun. Ucode patch is a nuclear reactor. If a
vendor says, "Don't do a ucode patch after doing X", then we had better not
try a ucode patch after doing X. I know that we like to do uptime
competitions, but a ucode patch can be a bigger deal than a minor stepping
of a cpu--and no one complains about having to reboot their system when they
change processors. :-))
I don't feel that avoiding a reboot is worth risking the damage caused by a
failed hot-patch, particularly if the user MUST do a hot-patch.
Nathan Zook
> -----Original Message-----
> From: Linus Torvalds [mailto:torvalds@transmeta.com]
> Sent: Wednesday, February 16, 2000 11:02 AM
> To: Tigran Aivazian
> Cc: linux-kernel@vger.rutgers.edu; hpa@transmeta.com
> Subject: Re: [patch-2.3.46-p2] P6 microcode update support
>
>
>
> Ok, the final patch I've seen looks ok,
>
> ... BUT ...
>
> I think we shoul dstart getting rid of /dev files for random
> new devices.
-
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/
This archive was generated by hypermail 2b29 : Wed Feb 23 2000 - 21:00:15 EST