[RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[Alexey Kardashevskiy]

This is rather a bugreport than actual revert but nevertheless.

Since 27a4c827c34a I started noticing weird traces and quite often
I could not boot my test machine which is POWER8 box with AST video
or ATI video (video adapter type does not matter much although
is happens lot more often with the OpenPOWER machine which got AST).
The system console is IPMI, not the actual video adapter.

I bisected it to this:

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id'a4c827c34ac4256a190cc9d24607f953c1c459

For some reason which I do not understand, in cursor_timer_handler(),
ops->cur_blink_jiffies is zero. Why can this happen? What is a proper
fix? Thanks.


---
 drivers/video/console/fbcon.c | 12 ++++++------
 drivers/video/console/fbcon.h |  1 -
 2 files changed, 6 insertions(+), 7 deletions(-)

diff --git a/drivers/video/console/fbcon.c b/drivers/video/console/fbcon.c
index afd3301..7e5b425 100644
--- a/drivers/video/console/fbcon.c
+++ b/drivers/video/console/fbcon.c
@@ -402,7 +402,7 @@ static void cursor_timer_handler(unsigned long dev_addr)
 	struct fbcon_ops *ops = info->fbcon_par;
 
 	queue_work(system_power_efficient_wq, &info->queue);
-	mod_timer(&ops->cursor_timer, jiffies + ops->cur_blink_jiffies);
+	mod_timer(&ops->cursor_timer, jiffies + HZ/5);
 }
 
 static void fbcon_add_cursor_timer(struct fb_info *info)
@@ -417,7 +417,7 @@ static void fbcon_add_cursor_timer(struct fb_info *info)
 
 		init_timer(&ops->cursor_timer);
 		ops->cursor_timer.function = cursor_timer_handler;
-		ops->cursor_timer.expires = jiffies + ops->cur_blink_jiffies;
+		ops->cursor_timer.expires = jiffies + HZ / 5;
 		ops->cursor_timer.data = (unsigned long ) info;
 		add_timer(&ops->cursor_timer);
 		ops->flags |= FBCON_FLAGS_CURSOR_TIMER;
@@ -709,7 +709,7 @@ static int con2fb_acquire_newinfo(struct vc_data *vc, struct fb_info *info,
 	}
 
 	if (!err) {
-		ops->cur_blink_jiffies = HZ / 5;
+//		ops->cur_blink_jiffies = HZ / 5;
 		info->fbcon_par = ops;
 
 		if (vc)
@@ -957,7 +957,7 @@ static const char *fbcon_startup(void)
 	ops->currcon = -1;
 	ops->graphics = 1;
 	ops->cur_rotate = -1;
-	ops->cur_blink_jiffies = HZ / 5;
+//	ops->cur_blink_jiffies = HZ / 5;
 	info->fbcon_par = ops;
 	p->con_rotate = initial_rotation;
 	set_blitting_type(vc, info);
@@ -1095,7 +1095,7 @@ static void fbcon_init(struct vc_data *vc, int init)
 		con_copy_unimap(vc, svc);
 
 	ops = info->fbcon_par;
-	ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms);
+//	ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms);
 	p->con_rotate = initial_rotation;
 	set_blitting_type(vc, info);
 
@@ -1309,7 +1309,7 @@ static void fbcon_cursor(struct vc_data *vc, int mode)
 	int y;
  	int c = scr_readw((u16 *) vc->vc_pos);
 
-	ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms);
+//	ops->cur_blink_jiffies = msecs_to_jiffies(vc->vc_cur_blink_ms);
 
 	if (fbcon_is_inactive(vc, info) || vc->vc_deccm != 1)
 		return;
diff --git a/drivers/video/console/fbcon.h b/drivers/video/console/fbcon.h
index 7aaa4ea..6bd2e0c 100644
--- a/drivers/video/console/fbcon.h
+++ b/drivers/video/console/fbcon.h
@@ -70,7 +70,6 @@ struct fbcon_ops {
 	struct fb_cursor cursor_state;
 	struct display *p;
         int    currcon;	                /* Current VC. */
-	int    cur_blink_jiffies;
 	int    cursor_flash;
 	int    cursor_reset;
 	int    blank_state;
-- 
2.5.0.rc3

Re: [RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[Benjamin Herrenschmidt]

On Wed, 2016-06-29 at 18:29 +1000, Benjamin Herrenschmidt wrote:
> On Wed, 2016-06-29 at 18:19 +1000, Alexey Kardashevskiy wrote:
> > This is rather a bugreport than actual revert but nevertheless.
> > 
> > Since 27a4c827c34a I started noticing weird traces and quite often
> > I could not boot my test machine which is POWER8 box with AST video
> > or ATI video (video adapter type does not matter much although
> > is happens lot more often with the OpenPOWER machine which got
> > AST).
> > The system console is IPMI, not the actual video adapter.
> > 
> > I bisected it to this:
> > 
> > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/com
> > mi
> > t/?id'a4c827c34ac4256a190cc9d24607f953c1c459
> > 
> > For some reason which I do not understand, in
> > cursor_timer_handler(),
> > ops->cur_blink_jiffies is zero. Why can this happen? What is a
> > proper
> > fix? Thanks.
> 
> Rings a bell... do you have fbcon: initialize blink interval before
> calling fb_set_par (f235f664a8afabccf863a5dee4777d2d7b676fda) ?

Also check if you have:

> Cheers,
> Ben.

Re: [RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[Benjamin Herrenschmidt]

On Wed, 2016-06-29 at 18:29 +1000, Benjamin Herrenschmidt wrote:
> On Wed, 2016-06-29 at 18:19 +1000, Alexey Kardashevskiy wrote:
> > This is rather a bugreport than actual revert but nevertheless.
> > 
> > Since 27a4c827c34a I started noticing weird traces and quite often
> > I could not boot my test machine which is POWER8 box with AST video
> > or ATI video (video adapter type does not matter much although
> > is happens lot more often with the OpenPOWER machine which got
> > AST).
> > The system console is IPMI, not the actual video adapter.
> > 
> > I bisected it to this:
> > 
> > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/com
> > mi
> > t/?id'a4c827c34ac4256a190cc9d24607f953c1c459
> > 
> > For some reason which I do not understand, in
> > cursor_timer_handler(),
> > ops->cur_blink_jiffies is zero. Why can this happen? What is a
> > proper
> > fix? Thanks.
> 
> Rings a bell... do you have fbcon: initialize blink interval before
> calling fb_set_par (f235f664a8afabccf863a5dee4777d2d7b676fda) ?

Also check if you have:

"fbcon: set a default value to blink interval"
(upstream ID a1e533ec07d583d01349ef13c0c965b8633e1b91).

> Cheers,
> Ben.

Re: [RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[Alexey Kardashevskiy]

On 29/06/16 18:31, Benjamin Herrenschmidt wrote:
> On Wed, 2016-06-29 at 18:29 +1000, Benjamin Herrenschmidt wrote:
>> On Wed, 2016-06-29 at 18:19 +1000, Alexey Kardashevskiy wrote:
>>> This is rather a bugreport than actual revert but nevertheless.
>>>
>>> Since 27a4c827c34a I started noticing weird traces and quite often
>>> I could not boot my test machine which is POWER8 box with AST video
>>> or ATI video (video adapter type does not matter much although
>>> is happens lot more often with the OpenPOWER machine which got
>>> AST).
>>> The system console is IPMI, not the actual video adapter.
>>>
>>> I bisected it to this:
>>>
>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/com
>>> mi
>>> t/?id'a4c827c34ac4256a190cc9d24607f953c1c459
>>>
>>> For some reason which I do not understand, in
>>> cursor_timer_handler(),
>>> ops->cur_blink_jiffies is zero. Why can this happen? What is a
>>> proper
>>> fix? Thanks.
>>
>> Rings a bell... do you have fbcon: initialize blink interval before
>> calling fb_set_par (f235f664a8afabccf863a5dee4777d2d7b676fda) ?
> 
> Also check if you have:
> 
> "fbcon: set a default value to blink interval"
> (upstream ID a1e533ec07d583d01349ef13c0c965b8633e1b91).


Yes, both. The problem I see appears from v4.4 till the very recent
upstream kernel which I am debugging now and which is v4.7-rc5.


-- 
Alexey

Re: [RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[Scot Doyle]

On Wed, 29 Jun 2016, Alexey Kardashevskiy wrote:
> On 29/06/16 18:31, Benjamin Herrenschmidt wrote:
> > On Wed, 2016-06-29 at 18:29 +1000, Benjamin Herrenschmidt wrote:
> >> On Wed, 2016-06-29 at 18:19 +1000, Alexey Kardashevskiy wrote:
> >>> This is rather a bugreport than actual revert but nevertheless.
> >>>
> >>> Since 27a4c827c34a I started noticing weird traces and quite often
> >>> I could not boot my test machine which is POWER8 box with AST video
> >>> or ATI video (video adapter type does not matter much although
> >>> is happens lot more often with the OpenPOWER machine which got
> >>> AST).
> >>> The system console is IPMI, not the actual video adapter.
> >>>
> >>> I bisected it to this:
> >>>
> >>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/com
> >>> mi
> >>> t/?id'a4c827c34ac4256a190cc9d24607f953c1c459
> >>>
> >>> For some reason which I do not understand, in
> >>> cursor_timer_handler(),
> >>> ops->cur_blink_jiffies is zero. Why can this happen? What is a
> >>> proper
> >>> fix? Thanks.
> >>
> >> Rings a bell... do you have fbcon: initialize blink interval before
> >> calling fb_set_par (f235f664a8afabccf863a5dee4777d2d7b676fda) ?
> > 
> > Also check if you have:
> > 
> > "fbcon: set a default value to blink interval"
> > (upstream ID a1e533ec07d583d01349ef13c0c965b8633e1b91).
> 
> 
> Yes, both. The problem I see appears from v4.4 till the very recent
> upstream kernel which I am debugging now and which is v4.7-rc5.

Hi Alexey,

Have you tried commit 1b45996d2ebf "tty: vt: Fix soft lockup in 
fbcon cursor blink timer." from the tty-linus branch of Greg's tty tree?

Or if that doesn't work, then perhaps
"fbcon: warn on invalid cursor blink intervals" from 
http://thread.gmane.org/gmane.linux.kernel.stable/178675/focus"26587 ?

Thanks,
Scot

Re: [RFC PATCH kernel] Revert "fbcon: use the cursor blink interval provided by vt"

[Alexey Kardashevskiy]

On 29/06/16 23:55, Scot Doyle wrote:
> On Wed, 29 Jun 2016, Alexey Kardashevskiy wrote:
>> On 29/06/16 18:31, Benjamin Herrenschmidt wrote:
>>> On Wed, 2016-06-29 at 18:29 +1000, Benjamin Herrenschmidt wrote:
>>>> On Wed, 2016-06-29 at 18:19 +1000, Alexey Kardashevskiy wrote:
>>>>> This is rather a bugreport than actual revert but nevertheless.
>>>>>
>>>>> Since 27a4c827c34a I started noticing weird traces and quite often
>>>>> I could not boot my test machine which is POWER8 box with AST video
>>>>> or ATI video (video adapter type does not matter much although
>>>>> is happens lot more often with the OpenPOWER machine which got
>>>>> AST).
>>>>> The system console is IPMI, not the actual video adapter.
>>>>>
>>>>> I bisected it to this:
>>>>>
>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/com
>>>>> mi
>>>>> t/?id'a4c827c34ac4256a190cc9d24607f953c1c459
>>>>>
>>>>> For some reason which I do not understand, in
>>>>> cursor_timer_handler(),
>>>>> ops->cur_blink_jiffies is zero. Why can this happen? What is a
>>>>> proper
>>>>> fix? Thanks.
>>>>
>>>> Rings a bell... do you have fbcon: initialize blink interval before
>>>> calling fb_set_par (f235f664a8afabccf863a5dee4777d2d7b676fda) ?
>>>
>>> Also check if you have:
>>>
>>> "fbcon: set a default value to blink interval"
>>> (upstream ID a1e533ec07d583d01349ef13c0c965b8633e1b91).
>>
>>
>> Yes, both. The problem I see appears from v4.4 till the very recent
>> upstream kernel which I am debugging now and which is v4.7-rc5.
> 
> Hi Alexey,
> 
> Have you tried commit 1b45996d2ebf "tty: vt: Fix soft lockup in 
> fbcon cursor blink timer." from the tty-linus branch of Greg's tty tree?


This one does fix the problem, thanks! I'll run my reboot test for little
longer and update if it fails.


Tested-by: Alexey Kardashevskiy <aik@ozlabs.ru>

> 
> Or if that doesn't work, then perhaps
> "fbcon: warn on invalid cursor blink intervals" from 
> http://thread.gmane.org/gmane.linux.kernel.stable/178675/focus"26587 ?



-- 
Alexey