On 2015/07/29 22:35, Thomas Gleixner wrote:
> On Wed, 29 Jul 2015, Peter Hurley wrote:
>> On 07/29/2015 07:53 AM, Thomas Gleixner wrote:
>>> On Wed, 29 Jul 2015, Peter Hurley wrote:
>>>> On 07/29/2015 06:32 AM, Thomas Gleixner wrote:
>>>>> On Wed, 29 Jul 2015, Taichi Kageyama wrote:
>>>>>> - Keep interrupt disabled on the CPU which is used to detect
>>>>>> an interrupt during the timeout of autoconfig_irq().
>>>>>> + Kick printk() on the CPU which detects interrupt
>>>>>> from a console serial port.
>>>>>
>>>>> This is wrong to begin with. How is that supposed to work on an UP
>>>>> machine? Not at all.
>>>>>
>>>>> So no, fix the code which has interrupts disabled accross autoprobing
>>>>> and do not try to apply bandaids somewhere else.
>>>>
>>>> Like printk() from some unrelated driver?
>>>
>>> If that's the cause for the wreckage then yes, we need a way to tell
>>> the printk code not to call into the driver until that initialization
>>> step is done. It's that simple.
>>
>> Like this?
>
> Looks about right.
Thomas,
Thank you for your comments.
The status of interrupt disabled should be short,
but it cannot be controlled from autorpobing side.
That's why I was thinking fixing autoprobing itself was better.
However, I remember Peter also mentioned about your idea first.
OK, I change my mind to fix the real problem simply.
I actually have seen the problem caused only by "printk()" on RHEL.
Peter,
I know your code is sample (console.h is required),
but it is conflict with [patch v2 1/3].
I think serial8250_console_write should not touch ctrl reg during autoconfig_irq.
To resolve the real problem, I think keeping only [patch v2 1/3] is best(opt1).
What do you think?
opt1. keep [patch v2 1/3]
+ Don't touch other legacy drivers using autoprobe.
Each driver can use console_lock to fix this problem if it happens.
opt2. fix probe_irq_on/mask/off for all legacy devices which use them
+ Discard [patch v2 1/3]
+ Fix autoconfig_irq and probe_irq_*
+ Test for all these drivers is required.
--> Probably, it works but I cannot test them.
Thanks,
Taichi
>
>> --- >% --
>> Subject: [PATCH] genirq: Disable printk() during irq probe
>>
>> printk() disables interrupts for extended periods of time while
>> outputting to console drivers. This breaks irq probing since the
>> triggered irq may not be serviced (in time) if scheduled on the
>> printk() cpu.
>>
>> Claim the console_lock() which effectively disables console output.
>>
>> Signed-off-by: Peter Hurley <peter@xxxxxxxxxxxxxxxxxx>
>> ---
>> kernel/irq/autoprobe.c | 9 +++++++++
>> 1 file changed, 9 insertions(+)
>>
>> diff --git a/kernel/irq/autoprobe.c b/kernel/irq/autoprobe.c
>> index 0119b9d..47535d2 100644
>> --- a/kernel/irq/autoprobe.c
>> +++ b/kernel/irq/autoprobe.c
>> @@ -39,6 +39,13 @@ unsigned long probe_irq_on(void)
>> */
>> async_synchronize_full();
>> mutex_lock(&probing_active);
>> +
>> + /*
>> + * printk() breaks irq probing - disable printk output until probe
>> + * completes
>> + */
>> + console_lock();
>> +
>> /*
>> * something may have generated an irq long ago and we want to
>> * flush such a longstanding irq before considering it as spurious.
>> @@ -132,6 +139,7 @@ unsigned int probe_irq_mask(unsigned long val)
>> }
>> raw_spin_unlock_irq(&desc->lock);
>> }
>> + console_unlock();
>> mutex_unlock(&probing_active);
>>
>> return mask & val;
>> @@ -174,6 +182,7 @@ int probe_irq_off(unsigned long val)
>> }
>> raw_spin_unlock_irq(&desc->lock);
>> }
>> + console_unlock();
>> mutex_unlock(&probing_active);
>>
>> if (nr_of_irqs > 1)
>> --
>> 2.5.0
>>
>>
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
