#qi-hardware IRC log for Thursday, 2013-09-26

apeletelarsc mth: are you there ?20:48
apeleteI'm trying to debug a kernel issue in the jz4740 musb driver20:49
apeletegdb log: http://paste.debian.net/46591/20:51
apeletejz4740 musb code: http://paste.debian.net/46592/20:51
apeleteaccording to gdb log, driver got stuck in platform_set_drvdata()20:52
apeletebut I can't tell why, I don't see anything suspicious20:52
apeletelarsc mth: could you please tell me what's wrong in there ?20:54
mthI'll try, looking at the code now20:54
larscI don't think it is possible to get stuck in platform_set_drvdata21:00
mthyeah, there is no loop there, so it should always exit21:05
mthclk_put is not necessary if you acquired it with devm_clk_get21:06
mthbut I don't see anything obviously wrong21:10
apeletemth: so what's wrong in the code then ? how come the kernel is panicking ?21:10
apelete<mth> but I don't see anything obviously wrong21:10
apeleteI still have gdb running right now21:10
apeletemth: I can print some variables if you suspect something, and paste it so you can see21:11
mthwhere does that SIGHUP come from?21:11
apeletegood question, I don't really see where it might come from21:12
mthI mean, why does the debugger stop there, something must have triggered it?21:13
mthdoes the kernel get stuck as well if you boot it without the debugger?21:14
mthand if so, what is the last thing that it logs?21:14
mthI tend to debug with logging rather than interactive debuggers most of the time; I still mistrust debuggers from my Turbo Pascal days21:14
apeleteok I'll recompile it right away without the debugger option and see what happens21:15
viricgdb isn't that bad :)21:15
apeleteI need it to understand how the usb code works (looking at the call stack and printing variables, and so on)21:19
apeletebut I will disable it just for now to see what happens21:19
mthviric: a big part of the problem was that the Turbo Pascal IDE ran in the same process as the program being debugged, so if you had any memory corruption the debugger itself or the data it was operating on could be invalidated21:28
mthgdb is more reliable, but still there are situations in which it is feeding nonsense21:29
pcercueiand it's generally a bad idea to trust blindly a program21:31
apeletemth: I'm recompiling, should get the log soon21:52
apeletemeanwhile, what does that SIGHUP means in kernel context ?21:52
apeletemth: boot log -> http://paste.debian.net/46609/21:59
apeletenot very helpful if you ask me :)21:59
qi-bot[commit] Paul Cercueil: Remove calls to sync() (master) http://qi-hw.com/p/gmenu2x/8452c0322:01
qi-bot[commit] Paul Cercueil: Remove calls to setInputSpeed(), as it only needs to be called once (master) http://qi-hw.com/p/gmenu2x/217a9b522:01
--- Fri Sep 27 201300:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!