2002-08-09 12:51:33

by jbradford

[permalink] [raw]
Subject: No reset of IDE disk

Running 2.4.19, when I use:

hdparm -Y /dev/hda

to put my Maxtor IDE HD in to 'Sleep' mode, it doesn't get automatically woken up again when it's accessed. I presume that this is a kernel issue, (an interrupt not being issued), because resetting the device with:

hdparm -w /dev/hda

restarts it successfully.

This is using a PIIX3 IDE interface, incidently.

John.


2002-08-09 17:41:24

by jbradford

[permalink] [raw]
Subject: Re: No reset of IDE disk

> This is because you *have* to use -w after -Y. The "sleep" method you're
> searching is -y, not -Y (dont ask me why.... ;)

Are you absolutely sure? My laptop, (same software versions), wakes up the hard disk successfully after a -Y sleep, (although on 2.2.13 it used to generate a spurious interupt message first).

Also, from the hdparm MAN page, it says "A hard or soft reset is required before the drive can be accessed again (the Linux IDE driver will automatically handle issuing a reset if/when needed).

A -y standby achieves the expected behaviour, (I.E. automatic spin up), with no errors.

2002-08-09 19:34:31

by jbradford

[permalink] [raw]
Subject: Re: No reset of IDE disk

> I posted the same message as you some weeks ago.

Sorry for the wasted bandwidth then - I am usually subscribed to the list, but temporarily de-subscribed for the last week.

> People said me that
> after a -Y you must always do a -w. Yes, the manpage can say that the
> kernel should do a -w of needed...but perhaps the man page is out of
> date...people here should know more...

Hmmm, the man page may well be out of date.

However, I would have thought that anything that stalls the IDE queue was to be considered a bug. I've E-Mailed Mark Lord, (the hdparm maintainer), to get some more input on this.

The interesting thing is that it does automatically recover from a -Y on the laptop running 2.4.19, (as it did with 2.2.13).

> the -w is the hard soft reset i think ;)

Yeah, it's a device reset.

> I was told to use -y. And it works well. I dont know the differences
> between -y and -Y (apparently, it should do the same ;)

Well, as far as I know, -y is intended to be a fast recovery, whereas -Y is more or less a full shut down, somewhat analogous to the various levels of power saving implemented in DPMS-compliant "green" monitors.

Somewhat related to all this, I am interested in having a variable delay in the write-back caching, (yeah, I know, keeping data cached in RAM on a battery powered laptop is asking for it to get lost, but I'd like the option), that way I could spin the HD down and enjoy a much extended battery life.