Re: therm_adt746x and load average
On 21 May 2004 at 09h05, Albert Cahalan wrote:
> > Yes, that's due to the TASK_UNINTERRUPTIBLE flag. It adds 1 to the load
> > average, however it uses very little CPU. Other macintosh's fan drivers
> > show the same behaviour.
> You're getting a hint that TASK_UNINTERRUPTABLE is not
> playing nice. Don't ignore the hint.
I'm ignoring this "hint" because I spoke with Benjamin H about that and the
results of our discussion was that it was not important. In fact, I wrote
TASK_INTERRUPTIBLE in version 1.1 of the driver, and he fixed that to
TASK_UNINTERRUPTIBLE. (See attached mail)
The hint, in my opinion, is that the load average calculation algorithm is
not perfect, nothing more.
> People get pissed
> off when "kill -9 12345" doesn't work.
Do you often kill -9 modules ? ;-)
--- Begin Message ---
- To: Colin Leroy <firstname.lastname@example.org>
- Subject: Re: TASK_UNINTERRUPTIBLE question
- From: Benjamin Herrenschmidt <email@example.com>
- Date: Sat, 21 Feb 2004 08:51:27 +1100
- Message-id: <1077313887.20789.1427.camel@gaston>
- In-reply-to: <firstname.lastname@example.org>
- References: <email@example.com>
On Sat, 2004-02-21 at 01:22, Colin Leroy wrote:
> Hi Ben,
> You once modified the therm_adt7467 driver (the CSet label was "fixes" or
> something like that), to change the TASK_INTERRUPTIBLE to
> What did motivate this change ? I'm not really competent when it comes to
> this area, but it seems to have two drawbacks:
> 1) Makes your pmdisk patch fail ("cannot stop tasks")
> 2) Load average is always at 1.00
Oh, that is what is causing the load avg issue ? That shouldn't...
I'll investigate... For pmdisk, we need to add some shit for
The TASK_UNINTERRUPTIBLE mean we don't get woken up by signals.
--- End Message ---