User Details
- User Since
- Jun 9 2018, 8:29 PM (305 w, 6 d)
Mar 12 2019
Looks like this already merged but looks good to me.
Feb 22 2019
Looks good to me. I would not be shocked if there are plenty of similar cases elsewhere in the driver.
Feb 13 2019
The change by itself, LGTM. As David described, there's a larger fix here of which this is basically a subset. The handling in general of timed out commands, and aborting of commands, is pretty broken in this driver, with this mis-bracket being the very first step in fixing. We did add some sysctls that can be used to simulate timed out commands which may be helpful to upstream to help test - basically tweaking the timeout time within the command struct such that we'll consider an IO timed out immediately the next pass of the functionality that checks for timed out IOs. It's in my experience not 100% equivalent to a drive accepting commands and going unresponsive, since those 'really failing' drives often then fail the 'next step' in their recovery in terms of aborting the IO, but ensuring as a first step we can abort an IO to an otherwise functional drive that we think timed out is a good test for this specific case.