Wed, Dec 14, 2022 at 10:21:10AM +0100, Klaus Schmidinger wrote:
Is there an actual problem that requires this?
It has been that way for many, many years, so I'd like to see more than
"looks problematic to me" before I dare touch this ;-).
The only problem that I am currently aware of is something specific to
rpihddevice shutdown and not VDR itself, and so far it only occurred
when I compiled the code with -fsanitize=undefined.
Some 15 years ago, VDR startup with softdevice (not softhddevice) would
hang once in a few months (say, once in 100 to 300 attempts) on a
single-core system, but I don't remember trying to analyze it back then.
It should have been on a 2.6 kernel and NPTL (not LinuxThreads), that
is, not too different from what we have now.
Code around thread creation and destruction is indeed tricky, and the
current implementation could be fine, only not "in good style" according
to ThreadSanitizer. If there is not too much thread creation and
destruction during normal usage, any actual race conditions might only
affect startup or shutdown.
I could give ThreadSanitizer another try, on a PC and a USB receiver
stick, to see what it complains during normal usage, and if something
can be fixed easily. My main motivation is to learn to use the tool, in
order to make use of it in another (much larger) code base that will
require some additional instrumentation due to the use of some custom
synchronization primitives.
Marko
_______________________________________________
vdr mailing list
vdr@xxxxxxxxxxx
https://www.linuxtv.org/cgi-bin/mailman/listinfo/vdr
