On 8/24/05, Stefan Richter [off-list ref] wrote:

quoted

quoted

Timers appear to operate in an atomic context, so timers should not be
allowed to call scsi_remove_device, which eventually schedules.

Any suggestions on the best way to fix this?

Workqueue, perhaps.

Perhaps.  Actually, of course :)

The reason these aren't working is because they have never been
tested.  I sent in my not-entirely-finished patches the night before I
left for China for one month.

When I get back to Waterloo (Ontario) in September, I should send in
revised versions of these patches with the following fixes:

- mod_timer instead of delete_timer/change timeout/add_timer
- bunch of code cleanups
- proper error handling
- actually making the patches work.

For now, check the list archives for the very first send of the
patches (with a resend of patch #3), and use those.  Those patches do
not use a debounce timer, so can be prone to some failures.  However,
I wasn't able to get those failures while testing, so at least you can
help me test out the overall logic and robustness of my work.

Luke Kosewski