lkml.org 
[lkml]   [2006]   [Feb]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
Subject100% cpu usage (kjournald, pdflush) with encrypted disks (dm-crypt)
hi,

i've also posted this to dm-crypt mailing list, but here is some
additional info.

i've encrypted my hard disks (whole root, and other disks)
with "cryptsetup luksFormat --cipher aes-cbc-essiv:sha256 --key-size
256" and formatted them with ext3 (-j -m0 -O dir_index).

i'm using 2.6.15.1 kernel on an athlon xp 2400+ with 1GB RAM, VIA
KT400A/VT8237 chipset.

/proc/crypto for aes:
name : aes
module : kernel
type : cipher
blocksize : 16
min keysize : 16
max keysize : 32

i searched for the aes-i586 optimization (that should be in the
kernel?) but i found no other reference to it than
CONFIG_CRYPTO_AES_586=y. guess i'm using the asm optimized code as
there is no other AES option in the kernel.


if i copy files from one encrypted disk to another my cpu usage goes up
to 100% and stays there for the whole copy process. renicing cp itself
doesn't help.

hdparm settings are the same for all disks except the SATA ones.
multcount = 16 (on)
IO_support = 3 (32-bit w/sync)
unmaskirq = 1 (on)
using_dma = 1 (on)
keepsettings = 0 (off)
readonly = 0 (off)
readahead = 256 (on)


kjournald and pdflush use all the cpu time, where kjournald is usually
at about 40-70% and pdflush goes up to 50-60%. this is while copying
large files. the cpu usage of kcryptd is negligible.

if i copy many small files, kcryptd will eat much more cpu, about
30-40% and the rest will be pdflush and kjournald.

all disks are mounted with -o defaults,noatime.

doing a 'mke2fs -j -m0 -O dir_index /dev/mapper/name' will also rise the
cpu usage to its maximum where pdflush takes about 60-70%.
(mke2fs 1.38 (30-Jun-2005) Using EXT2FS Library version 1.38)


is anyone else seeing this and is there a solution (besides using a
non-journalling fs) or do i have to live with it that encryption eats
all my cpu?

tx,
JG
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2006-02-01 22:10    [W:0.123 / U:0.168 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site