Forwarded from http://lists.debian.org/debian-boot/2007/10/msg00592.html
----- Forwarded message from Gordon Farquharson <[EMAIL PROTECTED]> ----- From: Gordon Farquharson <[EMAIL PROTECTED]> Subject: Re: Memory savings Date: Mon, 22 Oct 2007 06:22:24 -0600 To: Martin Michlmayr <[EMAIL PROTECTED]> Cc: debian-boot@lists.debian.org Hi Martin On 9/28/07, Martin Michlmayr <[EMAIL PROTECTED]> wrote: > * Joey Hess <[EMAIL PROTECTED]> [2007-06-28 15:05]: > > I have high hopes for the nslu2, maybe it will be able to work w/o > > lowmem level 2 now. Haven't been able to check yet. > > I just booted nslu2 in lowmem level 2 and didn't see any big > difference to etch at all (maybe 400 KB). From what I understood, > Colin's memory saving should apply to all lowmem levels, so I'm note > quite sure why the difference is not bigger here. Here are the results from some tests I did over the weekend. I changed the value of level2 from 33 to 24 in S15lowmem, built an installer, and confirmed that /var/lib/lowmem contained 1 after the system had booted. I used guided partitioning with all directories on a single partition. After a couple of attempts which failed with errors that appeared to be due to repository related problems (unable to download package x), the installation worked (sort of). The system boots, but gets stuck at INIT: Entering runlevel: 2 Starting system log daemon.... Starting kernel log daemon.... Starting OpenBSD Secure Shell server: sshd. Starting periodic command scheduler: crond. Below are the last few messages in syslog. I ran out of time before I was able to fully investigate this problem, so I don't know if it is a symptom of using lowmem level 1 for the installation. Oct 22 12:08:44 LKG7102D7 /usr/sbin/cron[1790]: (CRON) INFO (pidfile fd = 3) Oct 22 12:08:44 LKG7102D7 /usr/sbin/cron[1791]: (CRON) STARTUP (fork ok) Oct 22 12:08:44 LKG7102D7 /usr/sbin/cron[1791]: (CRON) INFO (Running @reboot jobs) Oct 22 12:08:51 LKG7102D7 kernel: eth1: no IPv6 routers present Oct 22 12:17:01 LKG7102D7 /USR/SBIN/CRON[1891]: (root) CMD ( cd / && run-parts --report /etc/cron.hourly) This test did show that the installer doesn't run out of memory in lowmem level 1 using guided partitioning, but I should try manual partitioning, and also should look at the available memory at various stages of the installation to see how close the installer is to running out of memory, assuming there is enough memory left to run shell :-) Also, this test was done without the NPE-B microcode, so I'll test the installer with the NPE-B microcode in the installer image, if it seems like a worthwhile thing to do (see below). The installer was much slower than the lowmem level 2 installation I did yesterday. The installation yesterday (lowmem level 2) started at 04:25:27 and reached localechooser at 05:36:20 (elapsed time = 01:10:53). The installation I did today using lowmem level 1 started at 03:29:55 and reached localechooser at 06:00:34 (elapsed time = 02:30:39). I'm not sure whether this is due to the lowmem level or another factor. I'll have to repeat the cases to confirm this difference but, the following log excepts show what tasks are causing such a difference in elapsed time. --- lowmem level 2 install: Oct 21 05:25:11 main-menu[2480]: INFO: Menu item 'flash-kernel-installer' selected Oct 21 05:25:13 apt-install: Reading package lists... Oct 21 05:25:13 apt-install: Oct 21 05:25:13 apt-install: Building dependency tree... Oct 21 05:25:36 apt-install: Oct 21 05:25:36 apt-install: Reading state information... Oct 21 05:25:37 apt-install: Oct 21 05:25:55 apt-install: initramfs-tools is already the newest version. Oct 21 05:25:55 apt-install: 0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded. elapsed time = 44 seconds --- lowmem level 1 install Oct 22 05:26:51 main-menu[2475]: INFO: Menu item 'flash-kernel-installer' selected Oct 22 05:26:54 apt-install: Reading package lists... Oct 22 05:26:54 apt-install: Oct 22 05:26:54 apt-install: Building dependency tree... Oct 22 05:29:33 apt-install: Oct 22 05:29:33 apt-install: Reading state information... Oct 22 05:29:36 apt-install: Oct 22 05:30:22 apt-install: initramfs-tools is already the newest version. Oct 22 05:30:22 apt-install: 0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded. elapsed time = 211 seconds --- The difference is probably due to swapping when apt is running because there is less memory (RAM) available. This hypothesis is supported by non-swap intensive tasks such as creating the initramfs. --- lowmem level 2 install Oct 21 05:27:49 apt-install: update-initramfs: Generating /boot/initrd.img-2.6.22-2-ixp4xx Oct 21 05:31:04 apt-install: Warning: ixp4xx_npe ethernet driver firmware file /lib/firmware/NPE-B not found Oct 21 05:31:04 apt-install: Warning: This system has an ethernet module loaded; it's not safe to Oct 21 05:31:04 apt-install: create an initramfs image for the new kernel without the firmware file. Oct 21 05:31:04 apt-install: Oct 21 05:31:04 apt-install: Unable to abort; system will probably be broken! Oct 21 05:33:05 apt-install: Reading package lists... elapsed time = 316 seconds --- lowmem level 1 install Oct 22 05:44:23 apt-install: update-initramfs: Generating /boot/initrd.img-2.6.22-2-ixp4xx Oct 22 05:47:15 apt-install: Warning: ixp4xx_npe ethernet driver firmware file /lib/firmware/NPE-B not found Oct 22 05:47:15 apt-install: Warning: This system has an ethernet module loaded; it's not safe to Oct 22 05:47:15 apt-install: create an initramfs image for the new kernel without the firmware file. Oct 22 05:47:15 apt-install: Oct 22 05:47:15 apt-install: Unable to abort; system will probably be broken! Oct 22 05:49:23 apt-install: Reading package lists... elapsed time = 300 seconds -- Gordon Farquharson GnuPG Key ID: 32D6D676 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED] ----- End forwarded message ----- -- Martin Michlmayr http://www.cyrius.com/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
