On 01/24/2013 05:50 PM, Florian Vaussard wrote: >>> I did it at first, but the led API executes in atomic context, where the >>> pwm-twl-led driver uses i2c communication. Setting a trigger will result in a >>> kernel panic. >> >> Now that you mentioned it, this might be true. >> > > [<c0013204>] (unwind_backtrace+0x0/0xec) from [<c00348ac>] > (warn_slowpath_common+0x4c/0x64) > [<c00348ac>] (warn_slowpath_common+0x4c/0x64) from [<c00348e0>] > (warn_slowpath_null+0x1c/0x24) > [<c00348e0>] (warn_slowpath_null+0x1c/0x24) from [<c054d384>] > (__mutex_lock_slowpath+0x6c/0x26c) > [<c054d384>] (__mutex_lock_slowpath+0x6c/0x26c) from [<c054d590>] > (mutex_lock+0xc/0x20) > [<c054d590>] (mutex_lock+0xc/0x20) from [<c02d740c>] > (regmap_bulk_write+0x48/0x138) > [<c02d740c>] (regmap_bulk_write+0x48/0x138) from [<c02de2c0>] > (twl_i2c_write+0xa4/0xf0) > [<c02de2c0>] (twl_i2c_write+0xa4/0xf0) from [<c0299e34>] > (twl4030_pwmled_config+0x70/0x9c) > [<c0299e34>] (twl4030_pwmled_config+0x70/0x9c) from [<c029875c>] > (pwm_config+0x5c/0x6c) > [<c029875c>] (pwm_config+0x5c/0x6c) from [<c039dc04>] (led_pwm_set+0x28/0x64) > [<c039dc04>] (led_pwm_set+0x28/0x64) from [<c039e27c>] > (led_heartbeat_function+0x10c/0x134) > [<c039e27c>] (led_heartbeat_function+0x10c/0x134) from [<c004359c>] > (call_timer_fn+0x90/0x178) > [<c004359c>] (call_timer_fn+0x90/0x178) from [<c0043994>] > (run_timer_softirq+0x250/0x2c8) > [<c0043994>] (run_timer_softirq+0x250/0x2c8) from [<c003cf78>] > (__do_softirq+0xf8/0x248) > [<c003cf78>] (__do_softirq+0xf8/0x248) from [<c003d154>] (irq_exit+0x44/0x98) > [<c003d154>] (irq_exit+0x44/0x98) from [<c000e338>] (handle_IRQ+0x68/0x8c) > [<c000e338>] (handle_IRQ+0x68/0x8c) from [<c000870c>] > (omap3_intc_handle_irq+0x58/0x70) > [<c000870c>] (omap3_intc_handle_irq+0x58/0x70) from [<c054f8c0>] > (__irq_svc+0x40/0x70) > Exception stack(0xc077df60 to 0xc077dfa8) > > :-) > >>> I am working on a patch for pwm-twl-led to defer using a workqueue right now. >> >> Great! >> The only thing I worry about is the latency we are going to get with the >> workqueue. >> > > If the latency becomes critical, we can create our own workqueue. Hrm, when we handled the led via gpio-leds it was also going through the same path at the end, via i2c to twl4030. I think the fix for this is going to be needed in the pwm core level. Just need to look at the gpio code to have similar handling of might_sleep interfaces. > Do we merge anyway this patchset, or do we wait until the trigger has been fixed? I think it can go and later when we have the fix for the slow path you can add the default trigger. Reviewed-by: Peter Ujfalusi <peter.ujfalusi@xxxxxx> -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
