Reply
Thread Tools
Posts: 1,680 | Thanked: 3,685 times | Joined on Jan 2011
#161
Ok, I R TOTAL NOOBER.

I was not using a disksize ^2.

All works great, no babies shall be stabbed this time.
__________________
N900: One of God's own prototypes. A high-powered mutant of some kind never even considered for mass production. Too weird to live, and too rare to die.
 
Gusse's Avatar
Posts: 168 | Thanked: 206 times | Joined on Apr 2010 @ Finland
#162
Originally Posted by maacruz View Post
It never crashes for me, are you using power of 2 sizes for disksize_kb?
EDIT: I should add that I use my own kernel build, but it is currently derived from kp49 and parts of bfs. I can make it available if anyone so wants.
What are your changes compared to KP49? Also if you are willing to make available your version (and maybe patches), that would be nice.

Reason for asking this, is that now I have instant reboots with KP49 when applying insmod-command. Doesn't matter what size I use (tried several values and aways the same), neither reverting other recent changes made any difference (DSP freq, SR).

Earlier with KP48, I managed to use several different disksize's without problems, meaning without problems with insmod part. Of cource large swaps made device eventually very slow and useless.

EDIT: This post seems to be valid also for eMMC (mmcblk0p3).
@ gregoranderson: Thanks for your trials!

Last edited by Gusse; 2011-11-27 at 21:47.
 

The Following 2 Users Say Thank You to Gusse For This Useful Post:
Posts: 1,680 | Thanked: 3,685 times | Joined on Jan 2011
#163
Code:
rzscontrol /dev/ramzswap0 --stats
DiskSize:	  131072 kB
NumReads:	       0
NumWrites:	       0
FailedReads:	       0
FailedWrites:	       0
InvalidIO:	       0
NotifyFree:	       0
ZeroPages:	       0
GoodCompress:	       0 %
NoCompress:	       0 %
PagesStored:	       0
PagesUsed:	       0
OrigDataSize:	       0 kB
ComprDataSize:	       0 kB
MemUsedTotal:	       0 kB
Code:
cat /proc/swaps 
Filename				Type		Size	Used	Priority
/dev/mmcblk0p3                          partition	786424	0	-2
/dev/ramzswap0                          partition	131064	79060	-1
Code:
free
             total         used         free       shared      buffers
Mem:        235260       206888        28372            0           76
-/+ buffers:             206812        28448
Swap:       917488       133856       783632

Can anyone explain why my compcache stats say 0 for everything? Also, why must my disksize be of the order ^2 when before it was not important?
__________________
N900: One of God's own prototypes. A high-powered mutant of some kind never even considered for mass production. Too weird to live, and too rare to die.
 

The Following 5 Users Say Thank You to vi_ For This Useful Post:
Posts: 244 | Thanked: 354 times | Joined on Jul 2010 @ Scotland
#164
It's worth noting that my N900 blew chunks at the idea of attempting to swapoff swap and swapon ramzswap0 except at any point other than post-boot. So, best time to perform this is at startup. Maybe even executed from a nohup'd script:

E.g, what I do is have my wee script in /root

Code:
Nokia-N900:~# cat compcache.sh 
swapoff /dev/mmcblk1p2
insmod /lib/modules/current/ramzswap.ko disksize_kb=65536
swapon /dev/ramzswap0
swapon /dev/mmcblk1p2
Nokia-N900:~#
Then I launch Terminal, hold my sword aloft, cry "By the power of Greyskull!" to become root then run:

Code:
nohup /root/compcache.sh &
This gives it a fighting chance. Once I'm confident enough, then I'll consider adding a runlevel script.

Last edited by gregoranderson; 2011-11-28 at 15:09.
 

The Following 2 Users Say Thank You to gregoranderson For This Useful Post:
Posts: 244 | Thanked: 354 times | Joined on Jul 2010 @ Scotland
#165
Originally Posted by vi_ View Post
Can anyone explain why my compcache stats say 0 for everything? Also, why must my disksize be of the order ^2 when before it was not important?
Now that's interesting: I recall now that free showed no swap usage before I activated secondary swap on my sd-card. For instance, I tried launching phone-ui with just compcache swap running (who wouldn't try this) and I was struck with one of the "not enough memory for that lark, son" error messages.

Time for another fiddle.
 

The Following 3 Users Say Thank You to gregoranderson For This Useful Post:
Posts: 1,808 | Thanked: 4,272 times | Joined on Feb 2011 @ Germany
#166
@gregoranderson,

Why the "nohup ... &". Your compcache.sh script does its thing and exits, no need to continue running or anything.

The only reason I can imagine for the nohup & is if you want to close your terminal really quickly before the script ends.. which is IMHO not a good idea if you're debugging

Last edited by reinob; 2011-11-28 at 15:16. Reason: s/exists/exits/
 
Posts: 244 | Thanked: 354 times | Joined on Jul 2010 @ Scotland
#167
Originally Posted by reinob View Post
@gregoranderson,

Why the "nohup ... &". Your compcache.sh script does its thing and exits, no need to continue running or anything.
Decent question. No reason other than a veteran's badge in the black art of shoving square pegs into round holes. I was guessing the extra beef required to run Terminal may have been the straw that broke the camel's back (straw being memory, camel's back being the ... ach you get the idea). Less beef used to retain the shell itself.

If I wanted debug (which I did at one point) then I redirect STDERR to a handle.

It may all be to pot anyway because, as per vi_'s post, I'm not actually seeing any I/O activity (this is KP49 ; KP48 definitely did "stuff").

If I can find my E90, or the spare ZTE Blade so I've still got comms, I'll chuck my SIM in there so I can play with the N900 later and see WTF is going on.
 

The Following 4 Users Say Thank You to gregoranderson For This Useful Post:
Posts: 1,680 | Thanked: 3,685 times | Joined on Jan 2011
#168
I await your results with baited breath fellow scotch man. Also there is a script just appeared in dev-repos called 'swap-set'. It is for detecting where your swap is, unounting, mounting ramzswap, mounting old swap.

It is a good start however it needs to be either:

1. To be encased in an event.d 'wrapper'
2. Called from Xsession.d
3. Support the few of us that use a swap file.

Anyway, the point is there is someone already making a nice compcache loading script.
__________________
N900: One of God's own prototypes. A high-powered mutant of some kind never even considered for mass production. Too weird to live, and too rare to die.
 

The Following 3 Users Say Thank You to vi_ For This Useful Post:
Posts: 1,101 | Thanked: 1,184 times | Joined on Aug 2008 @ Spain
#169
Originally Posted by Gusse View Post
What are your changes compared to KP49? Also if you are willing to make available your version (and maybe patches), that would be nice.

Reason for asking this, is that now I have instant reboots with KP49 when applying insmod-command. Doesn't matter what size I use (tried several values and aways the same), neither reverting other recent changes made any difference (DSP freq, SR).

Earlier with KP48, I managed to use several different disksize's without problems, meaning without problems with insmod part. Of cource large swaps made device eventually very slow and useless.

EDIT: This post seems to be valid also for eMMC (mmcblk0p3).
@ gregoranderson: Thanks for your trials!
I follow KP49 conservatively (disabling patches when I think they are still unstable), plus I have added some patches from Con Kolivas (extracted from BFS kernel patchset), and one of my own.
Features in KP49 not present: no 720p recording (dspbrige), no bq27x00 module
Features from BFS not present in KP49: more responsive (ck patches), BFQ disk scheduler, ubifs module
Other features not present in KP49 or BFS: enabled comcache statistics (rzscontrol will show data, not 0's), ability to disable the stupid cam's red light
I've been testing my last build for a week now, completely stable, but since I load test kernels from usb at boot, I'll have to rebuild a new set of debs. As soon as it's done I'll upload it to my garage page and I'll post here the link.
Of course I'll make the patchset available with the binaries.
 

The Following 5 Users Say Thank You to maacruz For This Useful Post:
Posts: 1,101 | Thanked: 1,184 times | Joined on Aug 2008 @ Spain
#170
Originally Posted by vi_ View Post
Can anyone explain why my compcache stats say 0 for everything?
Because KP49/BFS kernels have compcache statistics disabled.
Also, why must my disksize be of the order ^2 when before it was not important?
Oh, I was just guessing for a possible cause of reboots.
 

The Following User Says Thank You to maacruz For This Useful Post:
Reply

Thread Tools

 
Forum Jump


All times are GMT. The time now is 07:39.