Mouse jerky at 100% cpu

usercsr · 06-12-2004, 10:56 PM

hi,
I'm running Slack-current with kernel 2.6.6. Earlier I used to configure everything by myself when compiling the 2.6 kernel and everything worked fine. This time I took the .config file for the 'bareacpi' kernel and used to it to compile the 2.6.6 kernel (Made changes to some sections though). Now the problem is that the mouse movement becomes jerky whenever the cpu utilization goes to the maximum (100%). I can only guess that the mouse interrupts are getting a low priority. I have a Sony Vaio with a synaptics touchpad but mostly use a Logitech usb mouse. The problem comes for both of them. Here is my '/proc/interrupts' file if it can tell something :

Code:

~$ cat /proc/interrupts
           CPU0
  0:    5083466          XT-PIC  timer
  1:       4050          XT-PIC  i8042
  2:          0          XT-PIC  cascade
  8:          1          XT-PIC  rtc
  9:     719051          XT-PIC  acpi, eth0, ohci1394, Intel 82801CA-ICH3, uhci_hcd, uhci_hcd, uhci_hcd, radeon@PCI:1:0:0
 12:       2762          XT-PIC  i8042
 14:      26458          XT-PIC  ide0
 15:         21          XT-PIC  ide1
NMI:          0
ERR:          0

Would appreciate any help

irfanhab · 06-12-2004, 11:05 PM

Well, I think that something highpriority may be jammed or something, blocked interrupts of I/O devices such as mouse input

usercsr · 06-12-2004, 11:44 PM

okay, I found out another thing. The mouse gets blocked only when the cpu utilization goes to 100% because of disk activity. I wrote a program to go into an infinite loop (which will take the processor usage to 100%) and upon running it there is no problem, the mouse works fine. Is this normal behaviour or I need to change some settings for the disk.

cratos · 06-13-2004, 09:36 AM

I had a problem similar to this when I first compiled 2.6.x. It had me wondering why people were saying that 2.6 was faster than 2.4. It turned out to be a config file problem because after staring over with a friends config file, it just went away. I also had a similar experience with an nvidia network driver. Everytime I went on the internet I would get 100% cpu usage and the transfer rate was capped by the cpu. Moving to an older nvidia driver (linux 2.4.x) or using the included forcedeth drivers in 2.6.x took care of this.

Shade · 06-13-2004, 04:48 PM

You probably have to re-enable hdparm -- hard drive tuning tool.
The mouse becomes jumpy because of i/o wait -- load up top in an xterm and do something harddrive intensive, you'll see a column for % taken by Input/Output wait.

hdparm /dev/hda -tT

will give you some handy values for comparing the performance of your drive, and whether it can be tuned more.

Check out the hdarpm HOWTO or man hdparm.

--Shade

usercsr · 06-13-2004, 05:00 PM

Here is the output of 'hdparm -tT /dev/hda'

Code:

Timing buffer-cache reads:    1460 MB in 2.00 seconds = 729.38 MB/sec
Timing buffered disk reads:    74 MB in 3.02 seconds = 24.54 MB/sec

and the output of 'hdparm /dev/hda'

Code:

/dev/hda:
 multcount    =  0 (off)
 IO_support   =  0 (default 16-bit)
 unmaskirq    =  0 (off)
 using_dma    =  1 (on)
 keepsettings =  0 (off)
 readonly     =  0 (off)
 readahead    = 256 (on)
 geometry     = 65535/16/63, sectors = 78140160, start = 0

Cany anyone suggest what parameters needs to be changed. I've read that these parameters drastically affect performance and I don't want to mess things up.

Thank you

Shade · 06-13-2004, 05:07 PM

Well, hdparm can tell you what the best settings for your drive are.

Like I said, man hdparm

There's also a giant thread around here about hdparm settings that folks use. Use the search and dig that one up.

hdparm -i /dev/hda
(it might be a capital I, can't recall)
Will give you info on what your drive is capable.

Ignore the PIO modes, as those don't affect much...

--Shade

usercsr · 06-13-2004, 05:36 PM

okay, I'll try that.
Thank you