On Sat, 23 Feb 2013, Philipp Kern wrote:
> On Sat, Feb 23, 2013 at 09:19:53AM -0300, Henrique de Moraes Holschuh wrote:
> > I recommend that you just ignore the messages, that way you will still get
> > an updated microcode if Intel decides to start shipping one for your
> > processor.
> 
> well, I thought you had the toolchain for that already, but your argument 
> makes
> sense to me. I wonder if it'd make sense to document that in README.Debian,

I do, but I run it when _updating_ the initramfs, not during boot... and
only for the Intel case (AMD has so little microcode, I didn't bother
writing a tool to parse its container format).

When iucode-tool is installed and the config allows it (by default it does),
it will be used to probe the online processors for their signatures, and
only install microcode for those to the initramfs.  I do ignore processor
feature masks for this, so it can cause false positives (microcode for sig A
added to the initramfs, but it won't be loaded to processor B that shares
the same signature, but has incompatible processor features mask).

If no microcode ends up inside the initramfs, intel-microcode attempts to
don't do anything.  But if the microcode module is around, it will be
autoloaded by the kernel, and end up requesting microcode anyway.

> but maybe this bug report is already enough if somebody's wondering about
> those lines. (I found somebody being slightly concerned about them, thinking
> that some action would be required. I can't blame him for that.)

Indeed.  But fixing this upstream is not that easy, we'd need to teach the
firmware subsystem about "probe" firmware loading attempts, and to be quiet
when it results in a "ok, done but file not found" error (instead of any
other error)...

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh


-- 
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to