Bug 8882 - 100% CPU for iowait in USB storage devices
Summary: 100% CPU for iowait in USB storage devices
Status: REJECTED INVALID
Alias: None
Product: Drivers
Classification: Unclassified
Component: USB (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: Greg Kroah-Hartman
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2007-08-12 05:04 UTC by Ivan Zaera Avellon
Modified: 2007-08-12 09:49 UTC (History)
0 users

See Also:
Kernel Version: 2.6
Subsystem:
Regression: ---
Bisected commit-id:


Attachments

Description Ivan Zaera Avellon 2007-08-12 05:04:50 UTC
Most recent kernel where this bug did not occur:
Distribution: Debian testing
Hardware Environment: happens in three different computers
Software Environment: 
Problem Description: 
When I read from USB disks, CPU usage goes to near 100% for iowait, as shown in top. It happens in an AMD64 desktop with USB 2.0, in a Pentium III desktop with USB 1.0, and also in a Centrino laptop with USB 2.0.

It happens with an SD card USB adapter and also with a SATA disk USB adapter. 

I don't know if this is the expected behaviour for USB device in Linux but seems strange to me that I cannot use an USB disk to, for example, store DivX movies and play them.

I have Googled this problem but cannot find any information of help. Please tell me if this is the expected behaviour or if it is a bug.

In case of a bug I can try any patches or solutions you tell me.

Thanks in advance.

Steps to reproduce: plug in an USB disk and copy a big file to, for example, /dev/null
Comment 1 Ivan Zaera Avellon 2007-08-12 09:49:37 UTC
OK, forget about the bug. I have looked for more info about I/O wait and have found that this behaviour is normal.

I suppose that cp should increase a lot iowait as it is copying as fast as it can, saturating the bus. In USB 2.0 systems, playing DivXs inside the USB does not increase iowait, while in USB 1.0 it does because the full USB bandwith is used, thus the CPU sits idle waiting for USB to end.

Sorry again for filing the bug. I only wish there was more info about I/O wait :-(.

Note You need to log in before you can comment on or make changes to this bug.