Callout lockups and spin-lock held to long panic..

From: Randall Stewart <rrs_at_freebsd.org>
Date: Tue, 27 Jan 2015 10:43:25 -0800
All:

I just wanted to send a note to let folks know I have finally dug to
the bottom of the crashes that Sean Bruno has been seeing and will shortly
have a fix committed for it.

The problem was related to two callout_reset’s being run with migration
happening and that callout was executing (or waiting to execute).

The twin callout resets would in the end each remove the entry from the
linked list (twice) thus corrupting the linked list.

The software code would thus run, holding the CC_lock spinning forever
going through the linked list.. causing the crash.

I was able to reproduce this in a branch at netflix here so I can prove
that the fix I have actually fixes the issue.

It will be a couple more days of proving things out, followed by hopefully
getting interested reviewer’s to review the patch.. and then from
there I can commit it to head ..

Best wishes

R
------------------------------
Randall Stewart
803-317-4952 (cell)
Received on Tue Jan 27 2015 - 17:43:40 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:55 UTC