[arch-general] stability from pm-suspend ?
I've been using pm-suspend for temporarily shutting down the computer for later use, but now I raised the question whether it is safe or stable to do so at a constant basis. That is, seldom real reboots and often just suspend. For me the ratio of reboot and suspend is like 1:5. As you know that suspend don't really unmount the drives to read-only before it goes into suspension, when resumption had failed, you usually need to repair it and check for errors. This is at least the case for me when I use ext4. Even if it did resume successfully, I started to wonder if it also would harm the filesystems. I really like to hear some experiences whether that you have been using ACPI S3 kernel suspension for quite some time now, and feels it is really stable and safe to use, or that, you had ran into troubles using them. -- 李彥學 (Ian-Xue Li) http://b4283.ath.cx A student.
On Tue, Apr 20, 2010 at 09:37, Ian-Xue Li <da.mi.spirit@gmail.com> wrote:
I really like to hear some experiences whether that you have been using ACPI S3 kernel suspension for quite some time now, and feels it is really stable and safe to use, or that, you had ran into troubles using them.
I put my laptop in S3 sleep multiple times a day and pretty much only reboot for kernel upgrades now. Been doing this for months with no problems so far (I also use ext4). The file system corruption you seem worried should only occur if you don't resume properly, so as long as your resume is reliable I'd say it's perfectly safe to do this. -Aren Olson -- "Whoever said sunshine brings happiness has never danced in the rain." - K. Jackson
On Tue, Apr 20, 2010 at 10:37 AM, Ian-Xue Li <da.mi.spirit@gmail.com> wrote:
I really like to hear some experiences whether that you have been using ACPI S3 kernel suspension for quite some time now, and feels it is really stable and safe to use, or that, you had ran into troubles using them.
One thing you must avoid is to boot on one kernel version, install a kernel upgrade, then suspend and resume on the newer kernel. That will cause problems, so if you upgrade the kernel, you need to do a real reboot next time. Other than that, it seems to work well.
On 04/20/10 10:51, Ray Kohler wrote:
One thing you must avoid is to boot on one kernel version, install a kernel upgrade, then suspend and resume on the newer kernel. That will cause problems, so if you upgrade the kernel, you need to do a real reboot next time. Other than that, it seems to work well.
Really?? I've never had that problem IIRC... Is it to do with suspend-fixing-hacks that unload a module before suspend and reload it upon resume?(I think none of those are needed for my machine anymore).
On 04/20/10 10:37, Ian-Xue Li wrote:
I've been using pm-suspend for temporarily shutting down the computer for later use, but now I raised the question whether it is safe or stable to do so at a constant basis. That is, seldom real reboots and often just suspend.
me too, sometimes
As you know that suspend don't really unmount the drives to read-only before it goes into suspension, when resumption had failed, you usually need to repair it and check for errors. This is at least the case for me when I use ext4.
It's no worse than power failure, or system crash where you kill it with the power button! Actually I think it's better, because Linux does sync filesystems before suspending, so at least you won't have loss of recent data. Besides, I get more crashes when it's running normally than when it's suspended. However, restarting frequently tends to increase stability. It makes you restart your applications such as Firefox (which definitely benefits from a restart now and then). If you've upgraded (-Syu), it makes sure the running version of everything is the current version, including system libs. If there was corrupted memory due either to a bug or a hardware glitch, it cleans that up. Suspend/resume doesn't do any of these -- but then, leaving your computer on constantly is at least as bad as suspend/resume, probably worse. On the other hand, sometimes suspend is broken for some hardware. You should be able to observe this pretty easily for yourself. If, on the other hand, it looks like it's working, and if the system post-resume seems to be working just as well as the system pre-resume, then you should feel comfortable using it. After all, it's a great convenience! (modulo security considerations, if you keep an encrypted disk and don't want people who get their hands on your computer to be able to read your data! Or just login-password-wise if you're dealing with tech-ignorant friends.) -Isaac
On Tue, Apr 20, 2010 at 12:57:35PM -0400, Isaac Dupree wrote: Thanks for the detailed reply in advance.
On the other hand, sometimes suspend is broken for some hardware. You should be able to observe this pretty easily for yourself. If, on the other hand, it looks like it's working, and if the system post-resume seems to be working just as well as the system pre-resume, then you should feel comfortable using it. After all, it's a great convenience!
I do have a USB mouse that won't work, usually reloading ehci_hcd does the trick.
(modulo security considerations, if you keep an encrypted disk and don't want people who get their hands on your computer to be able to read your data! Or just login-password-wise if you're dealing with tech-ignorant friends.)
Security is indeed one problem I'm looking to resolve, because with suspend, it goes back directly into the system, without having to enter login/password... I wonder if there's some way to achieve something like winxp, which resumes back into the login manager, not the usable system itself. But that would probably require a session manager as well, I don't know. Nonetheless, another bizarre "problem" (quote because it doesn't really affect ordinary use) is that after resume, there're duplicate "basic system process" (I don't know how to put these) possessed by root. E.g. watch `ps -eF', there would be two instances of such processes as below. [events/0] [events/1] [events/2] [events/3] [cpuset] [khelper] [netns] [async/mgr] [pm] . . . I wonder if this is normal... -- 李彥學 (Ian-Xue Li) http://b4283.ath.cx A student.
On Wed, Apr 21, 2010 at 5:27 AM, Ian-Xue Li <da.mi.spirit@gmail.com> wrote:
Security is indeed one problem I'm looking to resolve, because with suspend, it goes back directly into the system, without having to enter login/password...
If you are only using X, maybe you could use xlock just before suspending.
On Tue, Apr 20, 2010 at 9:27 PM, Ian-Xue Li <da.mi.spirit@gmail.com> wrote:
Security is indeed one problem I'm looking to resolve, because with suspend, it goes back directly into the system, without having to enter login/password...
I wonder if there's some way to achieve something like winxp, which resumes back into the login manager, not the usable system itself. But that would probably require a session manager as well, I don't know.
Well, this is what I could dig up: should work if you're using xscreensaver (might have to modify the instructions a little bit to work with arch's pm-utils, but shouldn't be too tough:) http://www.linuxquestions.org/questions/linux-laptop-and-netbook-25/lock-scr...
On Wed, Apr 21, 2010 at 10:50:23AM -0600, Gary Wright wrote:
On Tue, Apr 20, 2010 at 9:27 PM, Ian-Xue Li <da.mi.spirit@gmail.com> wrote:
Security is indeed one problem I'm looking to resolve, because with suspend, it goes back directly into the system, without having to enter login/password...
I wonder if there's some way to achieve something like winxp, which resumes back into the login manager, not the usable system itself. But that would probably require a session manager as well, I don't know.
Well, this is what I could dig up: should work if you're using xscreensaver (might have to modify the instructions a little bit to work with arch's pm-utils, but shouldn't be too tough:)
http://www.linuxquestions.org/questions/linux-laptop-and-netbook-25/lock-scr...
And here's one that works well with pm-utils: http://aur.archlinux.org/packages.php?ID=36219 -- Byron Clark
On 04/20/2010 11:27 PM, Ian-Xue Li wrote:
Security is indeed one problem I'm looking to resolve, because with suspend, it goes back directly into the system, without having to enter login/password...
I don't think you mentioned which WM/DE you're using. If you're using XFCE (or would consider switching to it) the xfce4-power-manager-settings app has a setting for this: http://www.darose.net/xfce4-power-manager-settings_extended-tab.png HTH, DR
I've been using pm-suspend for temporarily shutting down the computer for later use, but now I raised the question whether it is safe or stable to do so at a constant basis. That is, seldom real reboots and often just suspend.
For me the ratio of reboot and suspend is like 1:5. As others have replied, I suspend regularly, getting days or maybe a week or so of 'uptime' on my laptop before bothering to reboot (normally kernel or X related, since i use nvidia's driver restarting X doesn't ALWAYS work).
As you know that suspend don't really unmount the drives to read-only before it goes into suspension, when resumption had failed, you usually need to repair it and check for errors. This is at least the case for me when I use ext4.
Even if it did resume successfully, I started to wonder if it also would harm the filesystems. No, no harm noticed. What I DON'T do is suspend with many or even any
On Tue, 2010-04-20 at 22:37 +0800, Ian-Xue Li wrote: programs open (besides background daemons and stuff like conky/rainlendar). No reason to, I'm anal that way to Alt-F4 everything first. This has the side benefit of protecting me against a possible non-resume (which hasn't happened in months).
I really like to hear some experiences whether that you have been using ACPI S3 kernel suspension for quite some time now, and feels it is really stable and safe to use, or that, you had ran into troubles using them.
Stable and safe for me. Now, talking about HIBERNATE, on the other hand....
participants (9)
-
Aren Olson
-
Byron Clark
-
David Rosenstrauch
-
Gary Wright
-
Ian-Xue Li
-
Isaac Dupree
-
Ng Oon-Ee
-
Ray Kohler
-
Sergi Pons Freixes