patch to fix early printks with PREMPT_RT_FULL

patch to fix early printks with PREMPT_RT_FULL

[Venkat Subbiah]

[-- Attachment #1: Type: text/plain, Size: 403 bytes --]

Hello,

With PREMPT_RT_FULL I notice couple of issues when early printk is turned on

1)  printks show up only after the full console is enabled
2)  There are double prints for each of the prinkts before the full 
console was enabled.

The attached patch seem to fix the issue, but I am not sure whether this 
is completely with
the lock being acquired in console_trylock_for_printk().

Thanks,
Venkat


[-- Attachment #2: 0001-printk-Allow-early-console-prints-with-PREEMPT_RT_FU.patch --]
[-- Type: text/x-patch, Size: 1465 bytes --]

>From 876c0653e4482357823d1d36858aed7b5cd8d1fa Mon Sep 17 00:00:00 2001
From: Venkat Subbiah <venkat.subbiah@cavium.com>
Date: Fri, 4 May 2012 17:53:16 -0700
Subject: [PATCH] printk: Allow early console prints with PREEMPT_RT_FULL

Signed-off-by: Venkat Subbiah <venkat.subbiah@cavium.com>
---
 kernel/printk.c |    5 ++++-
 1 files changed, 4 insertions(+), 1 deletions(-)

diff --git a/kernel/printk.c b/kernel/printk.c
index d3f22ec..d2dda2f 100644
--- a/kernel/printk.c
+++ b/kernel/printk.c
@@ -832,6 +832,8 @@ static inline int can_use_console(unsigned int cpu)
 	return cpu_online(cpu) || have_callable_console();
 }
 
+static int boot_console_active = 1;
+
 /*
  * Try to get console ownership to actually show the kernel
  * messages from a 'printk'. Return true (and with the
@@ -848,7 +850,7 @@ static int console_trylock_for_printk(unsigned int cpu, unsigned long flags)
 	int retval = 0, wake = 0;
 #ifdef CONFIG_PREEMPT_RT_FULL
 	int lock = (!early_boot_irqs_disabled && !irqs_disabled_flags(flags) &&
-		!preempt_count()) || sysrq_in_progress;
+		!preempt_count()) || sysrq_in_progress || boot_console_active;
 #else
 	int lock = 1;
 #endif
@@ -1653,6 +1655,7 @@ void register_console(struct console *newcon)
 		 */
 		printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
 			newcon->name, newcon->index);
+		boot_console_active = 0;
 		for_each_console(bcon)
 			if (bcon->flags & CON_BOOT)
 				unregister_console(bcon);
-- 
1.7.7.6

Re: patch to fix early printks with PREMPT_RT_FULL

[Venkat Subbiah]

Thomas,
Is this patch going to make it to a release or should this sent be sent 
to more folks for review?
Thanks,
Venkat

On 05/04/2012 06:07 PM, Venkat Subbiah wrote:
> Hello,
>
> With PREMPT_RT_FULL I notice couple of issues when early printk is turned on
>
> 1)  printks show up only after the full console is enabled
> 2)  There are double prints for each of the prinkts before the full
> console was enabled.
>
> The attached patch seem to fix the issue, but I am not sure whether this
> is completely with
> the lock being acquired in console_trylock_for_printk().
>
> Thanks,
> Venkat
>

Re: patch to fix early printks with PREMPT_RT_FULL

[Frank Rowand]

On 05/17/12 14:33, Venkat Subbiah wrote:
> Thomas,
> Is this patch going to make it to a release or should this sent be sent 
> to more folks for review?
> Thanks,
> Venkat
> 
> On 05/04/2012 06:07 PM, Venkat Subbiah wrote:
>> Hello,
>>
>> With PREMPT_RT_FULL I notice couple of issues when early printk is turned on
>>
>> 1)  printks show up only after the full console is enabled
>> 2)  There are double prints for each of the prinkts before the full
>> console was enabled.

Venkat,

Do my patches from yesterday solve problem #1 for you?  (I can't
test for your problem, because I don't see it on my ARM system.)

   http://marc.info/?l=linux-rt-users&m=133721694105947&w=2

   http://marc.info/?l=linux-rt-users&m=133721717706004&w=2

For problem #2, does that also exist without the RT patches?

-Frank

Re: patch to fix early printks with PREMPT_RT_FULL

[Venkat Subbiah]

On 05/17/2012 06:30 PM, Frank Rowand wrote:
> On 05/17/12 14:33, Venkat Subbiah wrote:
>> Thomas,
>> Is this patch going to make it to a release or should this sent be sent
>> to more folks for review?
>> Thanks,
>> Venkat
>>
>> On 05/04/2012 06:07 PM, Venkat Subbiah wrote:
>>> Hello,
>>>
>>> With PREMPT_RT_FULL I notice couple of issues when early printk is turned on
>>>
>>> 1)  printks show up only after the full console is enabled
>>> 2)  There are double prints for each of the prinkts before the full
>>> console was enabled.
> Venkat,
>
> Do my patches from yesterday solve problem #1 for you?  (I can't
> test for your problem, because I don't see it on my ARM system.)
The patch doe solve both #1 and #2. Thanks.
-Venkat

>
>     http://marc.info/?l=linux-rt-users&m=133721694105947&w=2
>
>     http://marc.info/?l=linux-rt-users&m=133721717706004&w=2
>
> For problem #2, does that also exist without the RT patches?
>
> -Frank
>
>

Re: patch to fix early printks with PREMPT_RT_FULL

[Venkat Subbiah]

On 05/17/2012 06:30 PM, Frank Rowand wrote:
> On 05/17/12 14:33, Venkat Subbiah wrote:
>> Thomas,
>> Is this patch going to make it to a release or should this sent be sent
>> to more folks for review?
>> Thanks,
>> Venkat
>>
>> On 05/04/2012 06:07 PM, Venkat Subbiah wrote:
>>> Hello,
>>>
>>> With PREMPT_RT_FULL I notice couple of issues when early printk is turned on
>>>
>>> 1)  printks show up only after the full console is enabled
>>> 2)  There are double prints for each of the prinkts before the full
>>> console was enabled.
> Venkat,
>
> Do my patches from yesterday solve problem #1 for you?  (I can't
> test for your problem, because I don't see it on my ARM system.)
Frank,

Looks like the fix is not architecture specific. Curious as to why it 
doesn't show up
on ARM?

Thanks,
Venkat
>
>     http://marc.info/?l=linux-rt-users&m=133721694105947&w=2
>
>     http://marc.info/?l=linux-rt-users&m=133721717706004&w=2
>
> For problem #2, does that also exist without the RT patches?
>
> -Frank
>
>