Description of problem:

Acer Travelmate 230 is doing a more or less correct suspend
provided acpi_sleep=s3_bios was used as a kernel parameter
AND ehci_hcd module unloaded before a sleep attempt.

The above works, up to extent, with this problem that attempts
to play sounds after a resume repeat in an infinite loop,
which can be broken with ^C, roughly the first second of a sound
to be played.  This happens at least with 2.6.6-1.391,
2.6.6-1.394 and 2.6.6-1.397 kernels.

Unloading sound modules before sleeping and reloading them
after does not change the picture.  An output of 'lscpi -vv'
before and after sleeping is the same.

The catch is that the sleep procedure described above works
also with 2.6.5-1.358 and then sound is not adversly affected.
One does not need to bother even with unloading sound modules.

How reproducible:
Always

Additional info:
This laptop "goes to work" and I will see it back approximate
in a month.