[pulseaudio-tickets] [PulseAudio] #806: Bus Errors when /dev/shm is (almost) full

[PulseAudio]

#806: Bus Errors when /dev/shm is (almost) full
----------------------+-----------------------------------------------------
  Reporter:  raboof   |       Owner:  lennart      
      Type:  defect   |      Status:  closed       
 Milestone:           |   Component:  core         
Resolution:  wontfix  |    Keywords:  bus crash shm
----------------------+-----------------------------------------------------
Changes (by lennart):

  * status:  new => closed
  * resolution:  => wontfix


Comment:

 Well, we actually do catch SIGBUS on the server side, because it can be
 triggered by evil clients simply by ftruncate()ing the shm files in
 /dev/shm, and we cannot allow that. Doing the same on the client side is
 incredibly ugly however, since this would mean fighting for SIGBUS
 handling: only one library/module can install a SIGBUS handler. People
 would hate us if we would fiddle around with that.

 I believe that out of shared memory in /dev/shm should be handled like any
 other kind of OOM: by aborting. If normal anonymous memory is lacking
 you'll be killed by the OOM killer, if address space is lacking you'll be
 killed by our aborting malloc() implementation. And if SHM memory is
 lacking you'll be killed by SIGBUS. I think this is very natural and
 sensible. And I don't see much point in trying to handle /dev/shm in any
 way special. If you have no memory, regardless of its kind you'r fucked.
 There's not much one can do about that.

 So, sorry, I am closing this, because I neither see a way nor a reason to
 handle this kind of OOM specially.

-- 
Ticket URL: <http://pulseaudio.org/ticket/806#comment:1>
PulseAudio <http://pulseaudio.org/>
The PulseAudio Sound Server