17:55 <bamse> sboyd: under what circumstance would i get a valid struct clk back but with clk->core NULL?

17:56 <sboyd> bamse: not sure how that ever happens

17:56 <sboyd> something got unregistered?

18:03 <bamse> no i don't think so

18:04 <bamse> sboyd: msm_find_best_baud() tries to round_rate and that gives me a NULL pointer dereference

18:04 <sboyd> bamse: has it been registered yet?

18:05 <bamse> i'm just hitting user space, so seems strange if gcc isn't probed yet

18:06 <bamse> but clk_get() should have returned IS_ERR() in that case, no?

18:24 <sboyd> bamse: I don't think we return EPROBE_DEFER for orphan clks still

18:24 <sboyd> we have random drivers that really want to keep using the clk even it is orphaned

18:25 <sboyd> is the NULL pointer for msm_port->clk or something else

18:25 <sboyd> presumably it's crashing on clk_core_get_boundaries()

20:36 <bamse> sboyd: msm_port->clk is valid, msm_port->clk->core is NULL

20:36 <bamse> sboyd: and yes, it crashes as i dereference core in clk_core_get_boundaries

20:45 <bamse> sboyd: sorry for the noise...the problem was in msm_serial all along

20:46 <bamse> sboyd: i screwed up my aliases so both my uarts where line 0...which we apparently handle by screwing up the data structures

20:53 <sboyd> Oh no!

20:56 <bamse> i guess we can improve that...