> Date: Tue, 9 Sep 2014 19:27:42 +0200 > From: Ingo Schwarze <schwa...@usta.de> > > Hi Mark, > > Mark Kettenis wrote on Mon, Sep 08, 2014 at 11:35:36PM +0200: > > > The more code & documentation I read, the more I'm convinced that > > coordinating state changes between logical processors isn't necessary > > and actually is responsible for the hangs people have been seeing. > > > > So here is a diff that does away with it all. I've tested it on a few > > laptops here, but it could use testing on a somewhat wider range of > > machines. I'm especially interested in seeing this tested on a dual > > socket machine with apmd -A. > > i'm sorry to say it makes no difference for me (i'm not opposed to the > diff, though). > > On my laptop, building ports works fine, running firefox works fine, > but whenever i surf the web with firefox while building ports, > the machine locks up hard. Sometimes, the lockup already happens > when merely starting firefox while building ports. Often, it > happens not when requesting a new URI, but when merely scrolling > within the page in firefox. > > After the lockup, CapsLk and NmLk still toggle the respective LEDs, > Fn-PgUp still switches on and off the torch, but nothing else has > any effect, not even Ctrl-Alt-Esc, Ctrl-Alt-Delete, Ctrl-Alt-Backspace > or Ctrl-Alt-F1. > > Unfortunately, i cannot break into ddb because i don't have a > docking station, hence no serial console, and when going to the > PC virtual console (Ctrl-Alt-F1), setting export DISPLAY=:0, > and starting firefox from the console, i was unable to get any > lockup. Apparently, it only happens when X (or whatever) is > actually painting something onto the screen. > > Whether i run with the defaults or with apm -A doesn't appear to > make a difference.
Not sure what you mean with "defaults", but if the crashes happen even in "manual performance adjustment mode", this diff certainly won't magically fix things.