Page 7 of 55 FirstFirst ... 2345678910111217 ... LastLast
Results 121 to 140 of 1081
Like Tree13Likes
  1. #121  
    for some reason when i run the first line in terminal it says it cant do it, no such file or directory. strange i know i have it written correct. or i think i do. there is a space after cp and then after the first path correct?
  2. #122  
    Quote Originally Posted by loopytee View Post
    a question, If I switch kernels, does it effect all of this? does it overwrite any of these files I have changed? I like to try different kernels as they are updated, do i need to plan on re-implementing these steps if i switch kernels?
    I can answer all questions in about an hour. I'm not near a computer and am in a place I can't really sit and type...

    No. Playing with kernels doesn't touch these files. Set it once and forget.

    Nick, There's a problem with the direction. DO NOT make the backup in event.d. I make a new dir /etc/back and put everything there. upstart will start ANYTHING in event.d.

    The other couple weeks (sorry guys, I don't see people ask) is just a couple quick lines that I need a better screen than my Pre as to refresh my memory. I'll post that when I get back to a computer.


    M.
  3. #123  
    Quote Originally Posted by NickVTPre View Post
    Read this post: http://forums.precentral.net/palm-pr...ml#post2951712.

    Xanadu73's approach is basically to tune the linux parameters to the 256mb of ram that the Pre- has. It doesn't disable, or wipe the swap, it makes sure that the RAM is the preferred storage, and that only things which should be in swap make it there. That's my understanding, anyway.
    I have a Pre Plus and rarely see swap in use. Part of what I understand from reading the UK and other kernel threads is that the Palm kernel is less likely to swap than the other kernels. Does anyone know where these changes are being made since there was a statement by Xanadu73 that his changes would not be likely to make it into the kernels due to their modification of other system files?

    Further - since this has moved on from Clearing Swap Memory to Optimizing Swap Memory - shouldn't this part of the thread be forked into a new thread?

    Lastly - does this approach offer tangible benefits to those running 1.4.5 on Pre Plus?
  4. #124  
    Quote Originally Posted by graffix31 View Post
    for some reason when i run the first line in terminal it says it cant do it, no such file or directory. strange i know i have it written correct. or i think i do. there is a space after cp and then after the first path correct?
    are you observing case? LunaSysMgr is case sensitive.
  5. #125  
    well i guess its good that mine didnt make that copy. so what we would need to do as step one with this new info?
  6. #126  
    Quote Originally Posted by Xanadu73 View Post
    ...
    Nick, There's a problem with the direction. DO NOT make the backup in event.d. I make a new dir /etc/back and put everything there. upstart will start ANYTHING in event.d.
    ...
    I'm confused... The instruction is to "cp /etc/event.d/LunaSysMgr /media/internal/LunaSysMgrOld" isn't backing up anything to /etc/event.d...
  7. #127  
    i know what the problem was. in your original post you had internals not internal
  8. #128  
    Quote Originally Posted by graffix31 View Post
    i know what the problem was. in your original post you had internals not internal
    Nice catch. Thats what typing at midnight will do to me...

    I think Xanadu read my instructions wrong, nothing is being backed up to /etc/event.d. I would go ahead with it.
  9. #129  
    Quote Originally Posted by Unclevanya View Post
    I have a Pre Plus and rarely see swap in use. Part of what I understand from reading the UK and other kernel threads is that the Palm kernel is less likely to swap than the other kernels. Does anyone know where these changes are being made since there was a statement by Xanadu73 that his changes would not be likely to make it into the kernels due to their modification of other system files?

    Further - since this has moved on from Clearing Swap Memory to Optimizing Swap Memory - shouldn't this part of the thread be forked into a new thread?

    Lastly - does this approach offer tangible benefits to those running 1.4.5 on Pre Plus?
    The changes are being made to the two files that Xanadu73 mentions in his post: /etc/event.d/LunaSysMgr and a newly created /etc/sysctl.conf file. This is different than the kernel, so no, this would not be put into uberkernel. I think Govnah could tune these parameters, perhaps, if WOSI decided to include these... Not 100% about that.

    It could go to a new thread, but i'm not sure it really needs to... At some point once this method has been tested to be a known good maybe it should sprout a new thread.

    If you aren't using swap, it probably won't benefit you. I'm on 2.1 with a pre+ and i modified the swappiness to 10 just to see if it had any effect. So far none. I also added the vm.dirty_expire_centisecs = 180000, which might allow more RAM items to be GC'd sooner. I'm doubtful that either is helping me.

    What i have noticed is that when the cache fills my free RAM i get lags and stutters in music players. Thats a topic for another thread though.
  10. #130  
    Quote Originally Posted by NickVTPre View Post
    Nice catch. Thats what typing at midnight will do to me...

    I think Xanadu read my instructions wrong, nothing is being backed up to /etc/event.d. I would go ahead with it.
    I did read wrong. What you typed is fine.


    M.
    Last edited by Xanadu73; 05/06/2011 at 09:36 AM. Reason: typos
  11. #131  
    Other little tweak.

    I marked this as "dangerous" because I don't know exactly what this does. I do believe, however, this is probably only really helpful to us (cast aside...) Pre- users. Just follow me for a sec here.

    In the file /etc/palm/browser.conf There's these lines:

    Code:
    # These sizes are set default based on physical memory of the device. You can override below
    #YoungSpaceSize=4MB
    #OldSpaceSize=64MB
    HOWEVER

    In /etc/palm/browser-app.conf The "YoungSpaceSize" and "OldSpaceSize" have values:

    Code:
    YoungSpaceSize=2MB
    OldSpaceSize=32MB
    Now keep in mind that the meta-doctor we've all used is based on Pre+ (512M RAM) devices NOT Pre- devices (256M RAM). So I've assumed those values are "hard codded" for devices that have 512M RAM. My little tweak is to comment out those two lines in browser-app.conf and let the system decide what to set those at because of the comment "based on physical memory of the device".

    Again, I don't know what that means, and I don't know a way to check. All I do know is that those values assume that there's 512M RAM in the device which the Pre- clearly does not. I can only assume that Palm's math will lower those values for the Pre-.

    I do also know that 32M of "Old Space Size" suggests to me that they all 32M of known "Old" crap to sit in "memory" (RAM / swap), which is suicide for the Pre-.

    I have seen no adverse effects that I'm aware of, and memory management does seem to be somewhat better. I did have the YouTube app crash my phone 2 or 3 times in a row a couple days ago, but, I've always had problems with YouTube on my Pre- even when running 1.4.5, so I don't know that's a great indication of this being a bad idea.


    M.
  12. #132  
    Quote Originally Posted by Xanadu73 View Post
    Other little tweak.

    I marked this as "dangerous" because I don't know exactly what this does. I do believe, however, this is probably only really helpful to us (cast aside...) Pre- users. Just follow me for a sec here.

    In the file /etc/palm/browser.conf There's these lines:

    Code:
    # These sizes are set default based on physical memory of the device. You can override below
    #YoungSpaceSize=4MB
    #OldSpaceSize=64MB
    HOWEVER

    In /etc/palm/browser-app.conf The "YoungSpaceSize" and "OldSpaceSize" have values:

    Code:
    YoungSpaceSize=2MB
    OldSpaceSize=32MB
    Now keep in mind that the meta-doctor we've all used is based on Pre+ (512M RAM) devices NOT Pre- devices (256M RAM). So I've assumed those values are "hard codded" for devices that have 512M RAM. My little tweak is to comment out those two lines in browser-app.conf and let the system decide what to set those at because of the comment "based on physical memory of the device".
    Does anyone have a standard Pre - with 1.4.5 on it to check the values and give you validation of this theory?

    BTW I have a Pre Plus on 1.4.5 and I would be happy to check if that's of any help at all.
  13. #133  
    Quote Originally Posted by Unclevanya View Post
    WHY is this clearing of swap a good thing? It would seem to me that if I turned swap off on any normal OS the effect would be to shift the memory that is on disk (Flash on the Pre) back to RAM and to discard cached memory contents that were no longer strictly necessary.

    The impact of this would be to either overload RAM and run out of memory entirely OR to slow down performance because previously cached memory was flushed and would have to be reread from "disk" and the proper calculations done to allocate it to RAM etc.

    It's obvious that this is helping some people - but what is the theory behind WHY?

    IN THIS CASE, the complete clearing of swap is a helpful thing because of the crappy memory management that Palm has put into our devices. The tweaks I'm playing with minimize the amount of stuff that gets put into RAM in the first place (especially disabling overcommit) and (help) minimize stuff that gets put into swap when RAM isn't available.

    No, forcibly dumping swap won't over load RAM *under normal conditions*. The kernel will simply decide to throw junk out because it'll trying and keep itself alive. Yes, of course, there's that chance that "swapoff -a" will belly up your device, but, this isn't Windows. Linux is smarter than that. Again, it'll just throw stuff out so it can keep itself alive.

    Yes, there would then be some "slow down" because previously cached things would have to be looked up again, but, the inode cache will minimize the time that takes because it wouldn't have to bother looking up the file/directory locations *and* re-read the files. It'll just have to re-read the files as it needs too.

    I hope this helps some.


    M.
  14. #134  
    Quote Originally Posted by Unclevanya View Post
    Does anyone have a standard Pre - with 1.4.5 on it to check the values and give you validation of this theory?

    BTW I have a Pre Plus on 1.4.5 and I would be happy to check if that's of any help at all.
    Good question.

    The 1.4.5 emulator has the 2M and 32M there, but, that's not the greatest example because the emulator is set for 512M RAM as well.


    M.
  15. #135  
    Quote Originally Posted by Xanadu73 View Post
    No, forcibly dumping swap won't over load RAM *under normal conditions*. The kernel will simply decide to throw junk out because it'll trying and keep itself alive. Yes, of course, there's that chance that "swapoff -a" will belly up your device, but, this isn't Windows. Linux is smarter than that. Again, it'll just throw stuff out so it can keep itself alive.
    You won't get more TMC errors with this toss away model?

    Quote Originally Posted by Xanadu73 View Post
    Yes, there would then be some "slow down" because previously cached things would have to be looked up again, but, the inode cache will minimize the time that takes because it wouldn't have to bother looking up the file/directory locations *and* re-read the files. It'll just have to re-read the files as it needs too.

    I hope this helps some.
    It does.

    However I'm still confused. Do you need BOTH the tweaks - the clearing script and the tuning? Also for devices like mine with 512 is this needed?

    Lastly - I have seen numerous things in the Kernel threads that claim that the alternate kernels swap more aggressively than the stock one and this is why swap/compcache is more important. What are they changing if that's true and how does this interact with your changes?
  16. #136  
    when trying to check to see if i am in write only it says is: not found. any ideas? thanks

    scratch that. all set
    Last edited by graffix31; 05/06/2011 at 10:09 AM.
  17. #137  
    Quote Originally Posted by Unclevanya View Post
    You won't get more TMC errors with this toss away model?

    However I'm still confused. Do you need BOTH the tweaks - the clearing script and the tuning? Also for devices like mine with 512 is this needed?

    Lastly - I have seen numerous things in the Kernel threads that claim that the alternate kernels swap more aggressively than the stock one and this is why swap/compcache is more important. What are they changing if that's true and how does this interact with your changes?

    The TMC is a Palm thing. 'Psycho has disabled that in his kernels (UK simply has the limit raised and not disabled) which perhaps may lead to "excessive swapping" (though I haven't seen posts that you're referring to) because without the TMC forcing us to reboot our phones, Linux will keep right on going trying to keep itself alive and needing swap to do so (especially on the Pre-). I don't have an Plam device with more than 256M RAM (outside of the emulators - and that hardly counts) so I can't really say either way.

    I do have to say that any ov these kernel paramentrs that I'm playing with are Linux defaults, not Harrier/SR-71/etc. defaults. I've checked the values here on my main machine (Gentoo!) and they are the same on webOS. The only one that Palm screwed with is the overcommit value. They changed it to 1 when the default is 0 (zero).

    Quote Originally Posted by /usr/src/linux/Documentation/sysctl/vm.txt
    overcommit_memory:

    This value contains a flag that enables memory overcommitment.

    When this flag is 0, the kernel attempts to estimate the amount
    of free memory left when userspace requests more memory.

    When this flag is 1, the kernel pretends there is always enough
    memory until it actually runs out.

    When this flag is 2, the kernel uses a "never overcommit"
    policy that attempts to prevent any overcommit of memory.

    This feature can be very useful because there are a lot of
    programs that malloc() huge amounts of memory "just-in-case"
    and don't use much of it.

    The default value is 0.
    That goes on to reference another document in the kernel source tree that further explains it. It's far too long to paste here. Here's a HTML version I just found:

    Linux Kernel Documentation :: vm : overcommit-accounting


    2 sounds really bad for our low RAM devices. I haven't even tried that. 1 is what Palm did. Zero is the default action (hence commenting out that line in LunaSysMgr).


    M.
  18. #138  
    Quote Originally Posted by graffix31 View Post
    when trying to check to see if i am in write only it says is: not found. any ideas? thanks
    Nick, there's a minor "typo". That should probably read:

    Code:
    mount -oremount,rw /
    Not:

    Code:
    mount -wo remount rootfs /
    M.
  19. #139  
    Just wondering but does browser-app.conf have any affect on Luna or is it just the web browser? Changing the browser-app-config seems like it would only affect those who browse the web or use form tools that work off the browser a lot.
    Sprint pre -> Motorola Photon 4G
  20. #140  
    Quote Originally Posted by Unclevanya View Post
    Do you need BOTH the tweaks - the clearing script and the tuning? Also for devices like mine with 512 is this needed?

    Sorry, I didn't answer this part...

    BOTH? I'd say probably not, especially since the swap cleaning thing I couldn't get to run in the background at all anyway...

    For 512M, I'd venture to say "probably not", BUT, "proper" memory management never hurts, so I guess that would really be up to you. It's us cast-aside Sprint Guinea Pigs that can really benefit from some Linux TLC and that's what I am and use so that's where my focus is. If any of you "Supported Users" find that this helps, cool! If you find no real change, in some ways I'm not surprised. Using 512M RAM in the emulators is a real dream, I think Ruby REALLY screwed us Sprint Guinea Pigs and it seems things aren't changing any for us under Leo's reign...


    M.
    Last edited by Xanadu73; 05/06/2011 at 10:19 AM. Reason: typos
Page 7 of 55 FirstFirst ... 2345678910111217 ... LastLast

Posting Permissions