printk change in 2.6.18?

printk change in 2.6.18?

[Dennis Munsie]

I just migrated my development over to 2.6.18-rc2, and I've now  
noticed that some of the printk's that used to work under 2.6.17  
stopped working under 2.6.18.  The printk's that I'm referring to are  
in intelfbhw_mode_to_hw() which is ultimately called by  
intelfb_set_par() which is called by fb_set_var().

I can put a printk() at the top of fb_set_var(), and it shows up.   
But anything else in there fails to show up.

Have there been any changes in the framebuffer that would suppress  
printk's in fb_set_var()?  I didn't see anything obvious, but I'm  
probably missing something.

dennis

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: printk change in 2.6.18?

[Antonino A. Daplas]

Dennis Munsie wrote:
> I just migrated my development over to 2.6.18-rc2, and I've now  
> noticed that some of the printk's that used to work under 2.6.17  
> stopped working under 2.6.18.  The printk's that I'm referring to are  
> in intelfbhw_mode_to_hw() which is ultimately called by  
> intelfb_set_par() which is called by fb_set_var().

Does the printk's show up in syslog (dmesg)? How about when intelfb
and fbcon are just loaded? How about while doing an fbset?

> 
> I can put a printk() at the top of fb_set_var(), and it shows up.   
> But anything else in there fails to show up.
> 
> Have there been any changes in the framebuffer that would suppress  
> printk's in fb_set_var()?  I didn't see anything obvious, but I'm  
> probably missing something.

It's possible that the changes in the vt layer may have caused this
unintentionally.

Tony

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: printk change in 2.6.18?

[Dennis Munsie]

The weirdness just goes on and on.... it seems that even though  
intelfb_mode_to_hw() is only called by intelfb_set_par() and that is  
only (as far as I can tell) called by fb_set_var(), the calls to  
fb_set_var() that I am seeing are not the ones calling intelfb_set_par 
().  Those ones either seem to be masked out completely, or they  
don't seem to be happening at all.

Except, that I know that they are happening, because I have some test  
code in there that I know very well is working and that if I comment  
that code out, everything breaks.

Is it possible for intelfb_set_par() to be called from anywhere  
else?  Or is it only called by fb_set_var()?

On Jul 18, 2006, at 7:46 PM, Antonino A. Daplas wrote:

> Does the printk's show up in syslog (dmesg)? How about when intelfb
> and fbcon are just loaded? How about while doing an fbset?

the printk's don't seem to show up anywhere... I've tried all kinds  
of variations -- using the DBG_MSG() macro in the intelfb, using  
printk() with KERN_ERR, KERN_CRIT, and without anything (which should  
be KERN_WARN).

I have a serial connection with earlyprintk turned on -- I can see  
all of the messages back to the first line in the boot.  None of my  
messages show up.  I also have loglevel set to a ridiculous 255 -- it  
was at 8 before.  I even have debug turned on in the intelfb driver.   
At one point, I had verbose turned on as well, but that gets way too  
chatty.

I haven't tried the other methods -- to be honest, I've been using  
printk's without problem now for the past couple of months, and I  
never bothered trying to use modules yet.  For my purposes, making  
some changes, recompiling, and rebooting the box hasn't been too big  
of a hassle.

So, for now, I'm sidestepping this entire episode by just hacking in  
some sysfs attributes to get the info out that I need.  It's not  
ideal, but it'll work.  I might have another look at it once I get  
the intelfb driver issues that I'm working on sorted out.

dennis

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: printk change in 2.6.18?

[Antonino A. Daplas]

Dennis Munsie wrote:
> The weirdness just goes on and on.... it seems that even though
> intelfb_mode_to_hw() is only called by intelfb_set_par() and that is
> only (as far as I can tell) called by fb_set_var(), the calls to
> fb_set_var() that I am seeing are not the ones calling
> intelfb_set_par().  Those ones either seem to be masked out completely,
> or they don't seem to be happening at all.
> 
> Except, that I know that they are happening, because I have some test
> code in there that I know very well is working and that if I comment
> that code out, everything breaks.
> 
> Is it possible for intelfb_set_par() to be called from anywhere else? 

Yes, fb_set_par() can be called directly by fbcon. See
drivers/video/console/fbcon.c

> Or is it only called by fb_set_var()?

No.

> 
> On Jul 18, 2006, at 7:46 PM, Antonino A. Daplas wrote:
> 
>> Does the printk's show up in syslog (dmesg)? How about when intelfb
>> and fbcon are just loaded? How about while doing an fbset?
> 
> the printk's don't seem to show up anywhere... I've tried all kinds of
> variations -- using the DBG_MSG() macro in the intelfb, using printk()
> with KERN_ERR, KERN_CRIT, and without anything (which should be KERN_WARN).
> 
> I have a serial connection with earlyprintk turned on -- I can see all
> of the messages back to the first line in the boot.  None of my messages
> show up.  I also have loglevel set to a ridiculous 255 -- it was at 8
> before.  I even have debug turned on in the intelfb driver.  At one
> point, I had verbose turned on as well, but that gets way too chatty.

You mean, if you type 'dmesg', you don't see the printk's? All printk's
should at least reach the logbuffer, but if it doesn't even reach the
logbuffer level, then the problem is very high up.

Tony

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV

Re: printk change in 2.6.18?

[Dennis Munsie]

On Jul 19, 2006, at 7:57 AM, Antonino A. Daplas wrote:

> You mean, if you type 'dmesg', you don't see the printk's? All  
> printk's
> should at least reach the logbuffer, but if it doesn't even reach the
> logbuffer level, then the problem is very high up.

ok, i just took a look at dmesg -- i had to expand the buffer for  
dmesg, but the printk's are showing up there.  So, something is  
causing them to not be sent to earlyprintk (i don't have any kernel  
output going to the console).

either way, i'm okay for now -- i'll take a look at this again once I  
have some spare cycles.

dennis

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys -- and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV