On Fri, May 11, 2018 at 07:15:10PM +0200, Antoine Jacoutot wrote:
> On May 11, 2018 6:03:32 PM GMT+02:00, Ryan Freeman <[email protected]> wrote:
> >On Fri, May 11, 2018 at 11:24:01AM -0400, Bryan Steele wrote:
> >> Hi,
> >> 
> >> I've been noticing ~15-30 second hangs when using some OpenGL
> >> applications since the radeondrm update when in, it happens
> >frequently
> >> in chromium but is hard to reproduce, but still happens often enough
> >> during daily browsing to notice. The mouse cursor remains active, but
> >> I cannot interact with anything, or even open an xterm.
> >> 
> >> Any mouse clicks I make happen all at once after the system becomes
> >> responsive again.. which can be quite annoying if I didn't notice the
> >> hang, heh. :-)
> >> 
> >> Perhaps an unrelated issue, a similar hang happens when loading
> >ioquake3,
> >> which shows WCHAN in schto. Unlike the hang with chromium, only the
> >> ioquake process normally hangs, but occasionally the whole system
> >will
> >> lock up, including the mouse cursor, with the CPU fans spinning.. but
> >> eventually it too recovers..
> >> 
> >> Sorry for the lack of a more detailed report, I don't really
> >understand
> >> the all the things in play.
> >> 
> >> I'm using cwm without a compositing manager, but someone else on IRC
> >> said it happened to them on gnome with a different radeon card.
> >
> >Thanks for reporting this Bryan,  I have been trying to figure out
> >the exactly same style hangs on my workstation at home. 
> >
> >I've been running the full gnome desktop, which uses opengl, so it
> >has made the experience quite interesting trying to debug :-)
> >I had meant to switch over to cwm to eliminate gnome as a problem
> >but I am thinking it is not so much to blame now...
> >
> >I was flipping between stock xorg config (no config) and using
> >'glamoregl'
> >instead to see if it made a difference, I seemed to be able to play
> >games a lot longer without freezes that way but nothing measurable
> >to show for it.  Logs never seemed to indicate much for me.
> >
> >Cheers,
> >-ryan

** big snip **

> I am not sure the underlying issue is due to the radeon update. Maybe it just 
> made it more obvious. Because I have been seeing the same with Intel for a 
> few months now. It's a pain to reproduce (happens most of the time with 
> chrome ) which is why I did not  report it.
> 

I managed to come up with a couple more data points for the ones I am seeing:
- definitely seems to be gnome-related, opened up an openbox session instead
  and couldn't get it to hang the same way once, even with chrome involved.
- flipping back to gnome, i can get the 30+ second hangs just from using
  gnome-shell, though any kind of gui interaction seems to be able to make it
  happen, such as switching tabs in chrome, or pressing a button in nautilus.

(using gnome-shell, go to top-left hot corner, while cute little animations
are trying to do their thing everything stops mid-animation for a looong time,
mouse still moves but you can't interact with anything, switching to virt
console works fine though)

now, perhaps it isn't gnome's fault upfront but a combination of the radeon
update and how mutter (that is what draws most things in gnome IIRC?) is using
it?  I only played around with openbox for about 10 minutes but within 1 minute
of starting gnome i was able to get it to stuck again, so gnome might be the
best way to bring this out.

a note on chrome: chrome is sometimes a piggy and i do see it randomly hang 
itself,
but generally it doesn't interrupt my work with other programs.  example:  
close a
tab, then it takes a good 15 seconds to realize the tab is closed and finish all
the redraws.  usually a chrome.core is sitting in my homedir.

Just want to make sure we are all talking about the same hangs :-)

I know aalm on FreeNode IRC mentioned this exact type of hang as well, and they
were using cwm with a radeon of similar vintage (reports as r600) but not exact
same model.

-ryan

Reply via email to