[SGVLUG] Slow-to-Boot: Why does it try "resume"

[David Lawyer]

I run Debain on a Pentium I PC desktop.  The time it takes to boot has
gone up by a factor of 4 over the past few years and I want that
fixed.  I decided to work on one delay at a time and filed a bug
report with Debian but they stopped trying to send me email since
whoever responded from Debian has an ISP that condones spam so email
to me was blocked by my ISP.  I recently submited another bug report
but Debian claimed that it was an "unknown package" although it was the
linux-image-2.6.17-2-486 package.
			David Lawyer
Here's what I reported (edited).  Any ideas on what "manual resume" is
doing and why.  Do you get this same message id dmesg?
------------------------------------------------------------------------------

>From dave at lafn.org Mon Jan 29 16:16:24 2007
From: David Lawyer <dave at lafn.org>
Subject: When booting: 38 sec. delay at "Attempting manual resume"
Cc: submit at bugs.debian.org

Package: linux-image-2.6.17-2-486
Version: 2.6.17-9_i386

Here's from dmesg at the point of delay.  Note that I've put "debug" and
"time" on the kernel command line (using lilo).

[   46.324557] hdd: cache flushes not supported
[   46.325131]  hdd: hdd1 hdd2 hdd3
[   84.007425] Attempting manual resume
[   84.353163] kjournald starting.  Commit interval 5 seconds

What is going on?  "resume" must mean resume from a suspend, but there
is no suspend in dmesg.  The "Attempting manual resume" message is
found in /linux-2.6.x/kernel/power/disk.c in the function
resume_store().  I don't have source code on my PC but found it on the
Internet at:
http://www.gelato.unsw.edu.au/lxr/source/kernel/power/disk.c#L12

According to this site, the function name that issues this message
(resume_store) only appears in the definition of this function,  In
other words the function is defined but never used (if this site
indexing is correct).

cat /sys/power/resume results in 22:67 which is hdd3, the swap
partition on my "backup" drive.  Apparently the image being restored
comes from there and it happens just after hdd3 prints in the dmesg
output.  Yet I don't see S1SUSPEND on hdd3 which is the SWSUSP_SIG
(signature).

There is a document, swsusp, in the power directory of the kernel
documentation and it says that it's not good to use suspend if the ide
driver is a module (as it is with this package).  But is suspend being
used at all? 
-----------------------------------------------------------------------
Here's the output of dmesg for my Pentium I PC.  There are several
other places where there is excessive delay but those will be the
subject of future bug reports.
[snip]