Re: [Qemu-devel] Re: segmentation fault in qemu-kvm-0.14.0

[Jan Kiszka]

On 2011-03-09 11:06, Corentin Chary wrote:
>> Probably the best way is to make vnc stop fiddling with
>> qemu_set_fd_handler2, specifically in threaded mode.
>> Why does it need to set/reset the write handler all the time?
> 
> I didn't write the original code, but it's probably to avoid calling a
> write handler when there is no data to write. That would make select()
> busy loop when there are actually no data to write.
> There is also the vnc_disconnect_start() case when the socket seems to
> be broken and we remove all handlers.
> 
>> The other question is: Who's responsible for writing to the client
>> socket in threaded mode? Only the vnc thread(s), or also other qemu
>> threads? In the former case, just avoid setting a write handler at all.
> 
> Cheap stuff is done by the main thread (cursor, etc...). The thread
> only do framebuffer updates.

And both are synchronized with a vnc-private lock only?

The problem with this model is the non-threaded qemu execution model.
Even if we acquire the global mutex to protect handler updates against
the main select loop, this won't help here. So vnc-threading likely
needs to be made dependent on --enable-io-thread.

Jan

signature.asc
Description: OpenPGP digital signature