Trying to reduce booting time with JFFS2 summary.
Dhinakaran K
Dhinakaran.K at Lntemsys.com
Mon Dec 21 10:57:54 EST 2009
Hi ALL,
I'm trying to reduce linux booting time, using JFFS2 block summary
option, but unfortunately unable to do so. Below is description of my
setup and test procedure. Please comment / suggest.
SETUP:
Processor : MPC8313e
OS : linux-2.6.20 (Freescale-LTIB
ltib-mpc8313erdb-20070824)
Flash : Nand Flash with 128KB block size.
JFFS2 : JFFS2 version 2.2. (NAND) (SUMMARY) (C)
2001-2006 Red Hat, Inc.
CONFIGURATION:
Kernel : Enabled
Filesystems -> Miscellaneous fs ->
JFFS2 -> Summary support
RFS image : Converted using
sumtool -b -irootfs.jffs2
-oout-sum.jffs2 -e128KiB
Observation:
With the above mentioned setup and configuration, it almost takes
40secs to prompt for login. Surprisingly even without JFFS2 block summary
support, it takes almost same time.
To have a closer look I just enabled kernel debug printk timestamps(shown
below) for both with and without summary. Looks like there is some
improvement in kernel tick level, but no improvement in boot time in terms
of secs.
Without summary support
[ 0.000000] IPIC (128 IRQ sources) at fdefa700
[ 0.000000] PID hash table entries: 512 (order: 9, 2048 bytes)
[ 31.015118] Dentry cache hash table entries: 16384 (order: 4, 65536
bytes)
[ 31.015892] Inode-cache hash table entries: 8192 (order: 3, 32768
bytes)
:
:
:
:
[ 35.856826] Freeing unused kernel memory: 144k init
:
: < takes quite a lot of time here >
:
[ 60.255748] JFFS2 notice: (769) check_node_data: wrong data CRC in data
node at 0x0037c000: read 0x7885983e, calculated 0x
With summary support
[ 0.000000] IPIC (128 IRQ sources) at fdefa700
[ 0.000000] PID hash table entries: 512 (order: 9, 2048 bytes)
[ 14.757860] Dentry cache hash table entries: 16384 (order: 4, 65536
bytes)
[ 14.758634] Inode-cache hash table entries: 8192 (order: 3, 32768
bytes)
:
:
:
[ 19.363823] Freeing unused kernel memory: 144k init
:
: < takes quite a lot of time here >
:
[ 43.785920] JFFS2 notice: (776) check_node_data: wrong data CRC in data
node at 0x003a1800: read 0x64066ebe, calculated 0x
Though there is significant improvement in timing(kernel tick) during hash
manipulation(see 3rd lines of respective logs) finally when frees unused
memory it takes long nap, and finally it looks like there is not
improvement in boot time. Even after googling I'm left with no clues as of
whats happening. Has someone got to suggest / comment please do reply to
this post.
P.S: For detailed log please see the attachment.
Warm Regards
Dhinakaran K
Software Engineer, EmSyS ( : +91 821 240
5638
Larsen and Toubro Limited, Mysore Campus ( : +91 821 240 5489
KIADB Industrial Area, Hebbal - Hootagalli Ê : +91 821
2403752
Mysore - 570 018 š :
Dhinakaran.K at Lntemsys.com
India ý :
http://www.Lntemsys.com
This mail is classified as
L&T EmSyS Proprietary L&T EmSyS Internal Use
L&T EmSyS Confidential L&T EmSyS General Business
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Jffs2Summary.zip
Type: application/zip
Size: 7429 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-mtd/attachments/20091221/dc316f86/attachment.zip>
More information about the linux-mtd
mailing list