[plug] Laptop hard disk access

Brad Campbell brad at seme.com.au
Tue Nov 28 14:51:36 WST 2000


Matt Kemner wrote:
 
> I remember a thread on linux-kernel about the same issue a long time ago,
> and the problem turned out to be init running sync() at regular intervals.
> 
> If I remember rightly, the author of init (Miquel V. Smoorenburg) wanted
> to keep the sync() in there because it was "safer" in case of power
> failure.  The laptop community thought otherwise.
> 
> I was fairly sure the laptop guys won, and they fixed that, though.  Maybe
> they haven't (or Miquel snuck it back in :)
> 
> "apt-get source init" and see if you can locate the sync() and kill it :)

Yeah, changelog says he took it out in 1997, and I can't locate a sync() in
the source anywhere.. doing further investigation

Nup, it is doing it with init=/bin/sh too..
It must be a kernel thing, but only when the root fs is mounted rw

<Pauses to do more testing..>
remounting / with -o noatime seems to help..
So, it looks like something in my 2.2.17 kernel keeps touching the drive,
except if I have noatime option in my mounts. The reason I did not see it before, was
the partition was mounted ro.
<Pauses for more testing..>

Yup, that fixes it. If I let the system run for a while, then spin the hd down
manually, I can keep working and it stays spun down. So therefore the auto-spindown
parm should work now, and it won't touch the drive unless it has to.

With the old 2.0.x kernels, I could tell update to only run every 10 minutes or so.
Dangerous I know, but I could always manually sync if it was crucial. Is there any
way of telling kupdated to do this? Short of re-compiling the kernel that is..

-- 
Brad....
                   /"\
Save the Forests   \ /     ASCII RIBBON CAMPAIGN
Burn a Greenie.     X      AGAINST HTML MAIL
                   / \



More information about the plug mailing list