Comment 310 for bug 59695

Matthieu:

Your case seems inconsistent with itself:

On Sat, 2007-12-22 at 22:34 +0000, Mathieu HAVEL wrote:
> [...]they replace it by another one (a Seagate ST91208220AS) on which I can apply partly the bug fix :
> * with -B 255 it completely stops the parking/unparking cycle
> * with -B 254 it does nothing.
[...]
> Last but not least, hdparm is always giving me the BIOS value for
> PowerManagement (set to 254) but this value is NOT applied by default on
> Ubuntu, and if I change it to 255, hdparm still indicates 254... Here I
> do not understand : either I have a bad configuration or either the
> problem is not "that simple" as the ugly fix let us suggest !

..so, you are saying that '255 completely stops the parking/unparking
cycle', yet you are also saying that the drive never gets set to '255'.
If you have a particular setting that completely stops the load cycle
count from increasing -- that would be the setting to use, and your
problem is solved. If you don't, then you stated as fact that 'with -B
255 it completely stops the parking/unparking cycle', yet you had not
verified that at all. Perhaps it is merely a difficulty of
communication.

> My conclusion is that, as I said, the ugly fix works (fortunatly) for
> about 95% of cases, but it is not fixing directly the "real" problem : I
> think it is not yet a direct solution. We have to find (and define) the
> true cause of this "too much" cycling.

It has already been established that the ugly fix is:
A) only temporary
B) does not work for everyone, although it works for most
C) that in most of the cases where 255 does not work, 254 does, and vice
versa
D) That it is only *part* of a problem that must have three factors
present --
    1- a maximum load cycle count on the drive (most/all drives)
    2- Aggressive APM (Default of some BIOS software, default of some
       drives)
    3- Frequent disk activity that is not frequent enough to prevent the
       drive from parking, nor infrequent enough to let it stay parked.

Factors 2 and 3 should both be dealt with. However, removing any one of
the above contributing factors to the problem prevents the problem, and
removing factor 3 is definitely non-trivial. Therefore, we remove
factor 2 -- by increasing the delay of or turning off the head parking.

No, the 'ugly fix' is not ideal. Thus, it has a name like "ugly fix."
However, it is the most likely to be useful in a reasonable amount of
time.