Hi,

On 08/03/2016 at 12:06:39 +0100, Sebastian Andrzej Siewior wrote :

I sucked this in. It seems to work. What remains that free_irq() thingy:
|WARNING: CPU: 0 PID: 0 at kernel/irq/manage.c:1541 __free_irq+0xb4/0x2e0()
|Trying to free already-free IRQ 16
|CPU: 0 PID: 0 Comm: swapper Not tainted 4.4.4-rt10+ #219
|Hardware name: Atmel SAMA5
|[<c0015bd0>] (unwind_backtrace) from [<c0013604>] (show_stack+0x10/0x14)
|[<c0013604>] (show_stack) from [<c002b850>] (warn_slowpath_common+0x80/0xac)
|[<c002b850>] (warn_slowpath_common) from [<c002b8ac>] (warn_slowpath_fmt+0x30/0x40)
|[<c002b8ac>] (warn_slowpath_fmt) from [<c005adfc>] (__free_irq+0xb4/0x2e0)
|[<c005adfc>] (__free_irq) from [<c005b0a0>] (free_irq+0x30/0x4c)
|[<c005b0a0>] (free_irq) from [<c0429034>] (pit_clkevt_shutdown+0x24/0x2c)
|[<c0429034>] (pit_clkevt_shutdown) from [<c0076430>] (clockevents_switch_state+0x60/0x130)
|---[ end trace 0000000000000001 ]---

which is a different problem and was there. The new thing is this:

|WARNING: CPU: 0 PID: 1 at kernel/locking/rtmutex.c:2062 rt_mutex_trylock+0x30/0x108)
|Modules linked in:
|CPU: 0 PID: 1 Comm: swapper Tainted: G        W       4.4.4-rt10+ #219
|Hardware name: Atmel SAMA5
|[<c0015bd0>] (unwind_backtrace) from [<c0013604>] (show_stack+0x10/0x14)
|[<c0013604>] (show_stack) from [<c002b850>] (warn_slowpath_common+0x80/0xac)
|[<c002b850>] (warn_slowpath_common) from [<c002b918>] (warn_slowpath_null+0x1c/0x24)
|[<c002b918>] (warn_slowpath_null) from [<c054a54c>] (rt_mutex_trylock+0x30/0x108)
|[<c054a54c>] (rt_mutex_trylock) from [<c054bd0c>] (rt_spin_trylock_irqsave+0x10/0x1c)
|[<c054bd0c>] (rt_spin_trylock_irqsave) from [<c043e030>] (clk_enable_lock+0x18/0x114)
|[<c043e030>] (clk_enable_lock) from [<c043f600>] (clk_disable+0x1c/0x34)
|[<c043f600>] (clk_disable) from [<c0429188>] (tc_shutdown+0x34/0x3c)
|[<c0429188>] (tc_shutdown) from [<c04291b0>] (tc_set_oneshot+0x20/0x50)
|[<c04291b0>] (tc_set_oneshot) from [<c00764a4>] (clockevents_switch_state+0xd4/0x130)
|[<c00764a4>] (clockevents_switch_state) from [<c00773e8>] (tick_switch_to_oneshot+0x48/0xb8)
|[<c00773e8>] (tick_switch_to_oneshot) from [<c006a480>] (hrtimer_run_queues+0x48/0x108)
|[<c006a480>] (hrtimer_run_queues) from [<c0068dd8>] (update_process_times+0x2c/0x64)
|[<c0068dd8>] (update_process_times) from [<c0076ad0>] (tick_handle_periodic+0x1c/0x90)
|[<c0076ad0>] (tick_handle_periodic) from [<c0429274>] (ch2_irq+0x20/0x28)
|[<c0429274>] (ch2_irq) from [<c005a39c>] (handle_irq_event_percpu+0x74/0x16c)

Is it possible to drop this disable/enable clock on the switch from pit
to one shot mode?

Both are things to work on. In the mean time, I'm using the following
patch:
https://github.com/alexandrebelloni/linux/commit/3a2eae463fce18ae815b887a5c9ca1a657b180ac

I understood from
http://lkml.iu.edu/hypermail/linux/kernel/1601.2/00941.html that this
was not your preferred course of action.

Even if that is something that is only seen when using preempt-rt, that
is probably something I should push mainline.

quoted

You may also stop at 99a81706526fb167029a940ef1f7bfbe882abd3e which
should solve the crash but it has not been tested as thoroughly.

Did you confuse the sha1 with something? This patch is called ("clk:
at91: remove IRQ handling and use polling") and is part of the series
you gave (patch #3).

Yeah, what I meant is that you could stop merging after that patch as
this is the one solving the crash.


-- 
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com