On Wed, 1 Jun 2005, Con Kolivas wrote:
On Wed, 1 Jun 2005 07:20 am, Pekka Enberg wrote:
On Tue, 2005-05-31 at 20:41 +0200, Ingo Molnar wrote:
Now, assuming you can confirm that doing:
echo 5 > /proc/sys/kernel/INTERACTIVE_DELTA
The hang goes away with a magic number of 6 (although it does not seem
as smooth as with turning off interactivity completely). With 5, I still
get the hang but it is noticeable shorter than before. Number 4 gives me
the same old hang.
Ingo, are there other patches you wanted me to try out?
Generally when I was playing with it I found that if something was critically
dependant on the _value_ of a tunable rather than the algorithm design, it
would simply take a different piece of hardware or test case to induce the
problem, and the algorithm would have to be changed instead of tweaking the
parameters. This will require a some thought as to the design and an
algorithm change for the long term rather than tweaking a value :-|
None of this behaviour existed when the interactivity code went in during 2.5
and I could never have anticipated this pipe design change coming and
affecting it in this way.
IIRC back when the interactivity code was introduces someone raised the
question about a set of malicious cooperating processes takeing advantage
of this feature to hog system resources. This seems to match what is
happening accidently with these pipe processes
at the time an answer was given that satisfied everyone (but I don't
remember what it was and don't have the time to do the digging for it), I
seem to remember something about an option to detect the dependancies
between processes and boost/penalize them as a group in addition to the
individual bonuses (with this deemed unessasary overhead at the time as
there were no example cases to show that this was more then a theoretical
case)
possibly someone with a better memory (or more time to research) can find
more.
David Lang
--
There are two ways of constructing a software design. One way is to make it so simple that there are obviously no deficiencies. And the other way is to make it so complicated that there are no obvious deficiencies.
-- C.A.R. Hoare
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
[Index of Archives]
[Kernel Newbies]
[Netfilter]
[Bugtraq]
[Photo]
[Stuff]
[Gimp]
[Yosemite News]
[MIPS Linux]
[ARM Linux]
[Linux Security]
[Linux RAID]
[Video 4 Linux]
[Linux for the blind]
[Linux Resources]