kupdated using all CPU
hi,
ive searched all over and found many references to this problem, but never found an actual solution.
the problem is that during heavy disk I/O, kupdated will periodically take up ALL the cpu. like the mouse will stop responding, nothing moves. the system is frozen until its done doing whatever its doing. the problem is mostly evident on large file transfers or doing something like copying a large file or untarring a large tarball, etc.
i am running 2.4.17 on an IBM R30 thinkpad. problem occurs on 2.4.16 also.
so is there already an answer about this that i just cant find? any help appreciated, thanks.
-j
Quoted 5 lines · john
hi,
ive searched all over and found many references to this problem, but never found an actual solution. the problem is that during heavy disk I/O, kupdated will periodically take up ALL the cpu.
I've seen a couple of reports of this, nothing to indicate that it's a common problem?
In the other reports, it was related to extremely low disk throughput. What does `hdparm -t /dev/hda' say?
-
[Mon, Feb 18, 2002 at 02:13:06PM -0800] [email protected] wrote:
+ john wrote:
+ >
+ > hi,
+ > ive searched all over and found many references to this problem, but
+ > never found an actual solution. the problem is that during heavy
+ > disk I/O, kupdated will periodically take up ALL the cpu. +
+ I've seen a couple of reports of this, nothing to indicate that it's
+ a common problem?
+
+ In the other reports, it was related to extremely low disk throughput.
+ What does `hdparm -t /dev/hda' say?
root@doom:~# hdparm -t /dev/hda
/dev/hda:
Timing buffered disk reads: 64 MB in 25.60 seconds = 2.50 MB/sec
Quoted 15 lines · john
[Mon, Feb 18, 2002 at 02:13:06PM -0800] [email protected] wrote:
+ john wrote:
+ >
+ > hi,
+ > ive searched all over and found many references to this problem, but
+ > never found an actual solution. the problem is that during heavy
+ > disk I/O, kupdated will periodically take up ALL the cpu.(quoted diff) +6
-0+ + I've seen a couple of reports of this, nothing to indicate that it's + a common problem? + + In the other reports, it was related to extremely low disk throughput. + What does `hdparm -t /dev/hda' say?
Quoted 4 lines
root@doom:~# hdparm -t /dev/hda
/dev/hda:
Timing buffered disk reads: 64 MB in 25.60 seconds = 2.50 MB/sec
ugh. OK, I'll see if I cen reproduce this - there's no reason why disk suckiness should cause high CPU load. But you need to pay some attention to your IDE settings in kernel config, and possibly tuning.
-
Quoted 18 lines · Andrew Morton
Quoted 17 lines · john
[Mon, Feb 18, 2002 at 02:13:06PM -0800] [email protected] wrote:
+ john wrote:
+ >
+ > hi,
+ > ive searched all over and found many references to this problem, but
+ > never found an actual solution. the problem is that during heavy
+ > disk I/O, kupdated will periodically take up ALL the cpu.(quoted diff) +6
-0+ + I've seen a couple of reports of this, nothing to indicate that it's + a common problem? + + In the other reports, it was related to extremely low disk throughput. + What does `hdparm -t /dev/hda' say?root@doom:~# hdparm -t /dev/hda
Quoted 7 lines
Quoted 2 lines
/dev/hda:
Timing buffered disk reads: 64 MB in 25.60 seconds = 2.50 MB/secugh. OK, I'll see if I cen reproduce this - there's no reason why disk suckiness should cause high CPU load. But you need to pay some attention to your IDE settings in kernel config, and possibly tuning.
Another possibility with the same symptoms is that there are simply very large I/O request starving everything else out. When some crucial data needs to be paged back in. john can you confirm you looked in top and saw kupdate taking 100% of the cpu?
Eric
[Mon, Feb 18, 2002 at 09:53:08PM -0700] [email protected] wrote:
+ Andrew Morton <[email protected]> writes:
+
+ > john wrote:
+ > >
+ > > [Mon, Feb 18, 2002 at 02:13:06PM -0800] [email protected] wrote:
+ > > + john wrote:
+ > > + >
+ > > + > hi,
+ > > + > ive searched all over and found many references to this problem, but
+ > > + > never found an actual solution. the problem is that during heavy
+ > > + > disk I/O, kupdated will periodically take up ALL the cpu.
+ > > +
+ > > + I've seen a couple of reports of this, nothing to indicate that it's
+ > > + a common problem?
+ > > +
+ > > + In the other reports, it was related to extremely low disk throughput.
+ > > + What does `hdparm -t /dev/hda' say?
+ > >
+ > > root@doom:~# hdparm -t /dev/hda
+ > >
+ > > /dev/hda:
+ > > Timing buffered disk reads: 64 MB in 25.60 seconds = 2.50 MB/sec
+ >
+ > ugh. OK, I'll see if I cen reproduce this - there's no reason
+ > why disk suckiness should cause high CPU load. But you need to
+ > pay some attention to your IDE settings in kernel config, and
+ > possibly tuning.
+
+
+ Another possibility with the same symptoms is that there are simply
+ very large I/O request starving everything else out. When some
+ crucial data needs to be paged back in. john can you confirm you
+ looked in top and saw kupdate taking 100% of the cpu?
yes, i did. another thing that makes it pretty obvious is the cpu time reported by ps. kupdated cpu time will stay 0:00 until i run into the condition i described. then kupdated cpu time will jump as much as 0:45.
but i have fixed the issue by using the correct ide driver. though, someone may want to look into this anyway...dunno.. -j