ATA READ command timeout (and worse)

[Stephen McKay]

On Tuesday, 17th June 2003, Jaw-Shiang Joshua Huang wrote:

>Because your machine will reboot automatically when the disk driver operation
>is abnormal, it makes me want to know more.
>
>Is your kernel compiled with DDB?  If not, it will reboot after 15 seconds
>while hitting panic.  If it's reproducable, would you mind to compile a new 
>kernel and try to find out where it panic or page fault?  I just want to know
>this bug will make FreeBSD kernel reboot or just hit panic or page fault.

It is not compiled with DDB.  I will add that before I do any more
experimenting.

I've only seen it reboot once.  I don't know if it was a panic or something
worse (like power dropped to the CPU and it reset) since I was not watching
the console at the time.

>Do you think it's possible that sometimes the disk is not working properly?
>In my memory, the disk DMA limitation checks the ATA identify command report
>from the HDD or the register of IDE controller.  I'll doubt the HDD first. :)

What do you mean by "not working properly"?  The disks are initially probed
with their correct speeds (and the cables are correct too).  One of the
disks then appears to be using the wrong cable after recovery from the
timeouts.  I can't see how that could be anything other than a driver bug.

>Also, I guess that 230W is not enough to provide the current for 4 disks
>concurrently.  Can you try your experiment with only 3 disks connected power
>line again?  If possible, two will be better.

This seems like a good idea.  I'll do this when I get back to testing.

>Sorry that I did not provide you any suggest but only questions.  I have
>not had a SiI IDE controller, either.

I'm sure someone, somewhere, is using them. :-)

Stephen.