Buffer fluxuating with 2.6.13 - Club CDFreaks

RaZoR1394 · 04-09-2005

With my px-716a, latest version of Nero and Gentoo with the 2.6.13 kernel the buffer keeps going up and down and the burning pauses (to wait for the buffer to fill up) very often. With gentoo-sources 2.6.12-r9 I don't have the this problem. The buffer fluxuations is also present with the vanilla-sources-2.6.13.

Axllent · 05-09-2005

Burning CD or DVD, and at what burning speed? What processor? ATM I only have 2.6.13 on my laptop which has only a CD-writer, so if it happens with DVD I'll have to get 2.6.13 to my desktop to test.

**mathf** · 05-09-2005

Hum... maybe it comes (once again) from a DMA alignment change in the ide-cdrom driver. Have you checked on both side (2.6.12 and 2.6.13) that DMA is correctly enabled on your devices ? (you can check this running hdparm -d /dev/XXX)

RaZoR1394 · 05-09-2005

I'm burning dvd's both data and from iso's. My processor is an Amd athlon 64 4000+. I'm burning Verbatim 16x DVD+R at max (16x).

hdparm /dev/hda reports (hda is px716a)

Quote:

/dev/hda:
IO_support = 1 (32-bit)
unmaskirq = 1 (on)
using_dma = 1 (on)
keepsettings = 0 (off)
readonly = 0 (off)
readahead = 256 (on)
HDIO_GETGEO failed: Invalid argument

I get exact the same message with 2.6.12.

Axllent · 05-09-2005

Sorry, took me a while.. had to compile the kernel and change my system to udev.
I can confirm this (bug?). Here (800MHz) while writing a DVD-RW it ran right the way down to 0%, then filled itself up again. On the one hand I think it's probably working ... as it never left 100% here before, but on the other hand 0%? Actually it did it twice while burning a 1.4GB file at 4x.

hdparm shows: using_dma = 1 (on)

**mathf** · 06-09-2005

Okay. Will have a look at it as soon as I have time. Hummm stupid stuff, can you revert drivers/ide/ide-cd.c and drivers/ide/ide.h to version 2.6.12 and check if it is still compilable and if it works better ?

RaZoR1394 · 06-09-2005

I'll try that. I will also check if K3B has the same problems when I get back home.

Axllent What burner/brand are you using?

Axllent · 06-09-2005

Quote:

Originally Posted by mathf

Okay. Will have a look at it as soon as I have time. Hummm stupid stuff, can you revert drivers/ide/ide-cd.c and drivers/ide/ide.h to version 2.6.12 and check if it is still compilable and if it works better ?

I presume you mean drivers/ide/ide.c ;-)

There are no differences at all between both ide.c and ide-cd.h betwen those 2 versions:

Code:

gentoo_test src # diff -u linux-2.6.12/drivers/ide/ide-cd.h linux-2.6.13/drivers/ide/ide-cd.h
gentoo_test src # 
gentoo_test src # diff -u linux-2.6.12/drivers/ide/ide.c linux-2.6.13/drivers/ide/ide.c
gentoo_test src #

Any other ideas?

Axllent · 06-09-2005

Quote:

Originally Posted by RaZoR1394

Axllent What burner/brand are you using?

Asus DRW-1604P

**mathf** · 06-09-2005

Quote:

Originally Posted by Axllent

I presume you mean drivers/ide/ide.c ;-)

Hum no. Actually I wanted to say /drivers/ide/ide-cd.c and ide-cd.h. These are the source files for IDE devices. According to kernel.org, ide-cd.c has changed. If the header file is the same, then it should compile without a problem.

Axllent · 06-09-2005

I can confirm a difference between these 2 *.c files, however I have just tested this at home on 2.6.13 with the ide-cd.c file from 2.6.12 (which compiles fine btw) and there is no change. The buffer still runs right down to 0% during burning.

Other ideas?

**mathf** · 12-09-2005

Ok... we are currently working on this issue. One point that is, we was able to reproduce the problem with a brand new 2.6.13.1 out of kernel.org. The let's say "bug" was occuring whatever the interface was. So I suspect something deeper in the kernel

Axllent · 12-09-2005

Thanks for the feedback. I confirmed this yesterday too on 2.6.13.1

**mathf** · 12-09-2005

Have you checked with 2.6.12.6 or with another one ? The last 'stable' kernel I have on my machine is 2.6.11.10 so it's a little bit old

Axllent · 12-09-2005

I did have iirc a 2.6.12.x ... which worked quite fine, however I'm at work and can't check which "x" that was as I'm at work ;-) I can/will repost when I get home though.

From what I noticed it came with 2.6.13, but then again I don't normally try every 2.6.x.x version, and it may have been introduced in one of the last 2.6.12.x versions.

Axllent · 12-09-2005

Sorry, my bad.... 2.6.11.10 was my previous kernel on the machine with a decent burner. On my laptop I had 2.6.12 , but no subversion. If you need me to test a specific 2.6.12.x version just drop a note ;-)

**mathf** · 12-09-2005

No problem, I have compiled a 2.6.12.6 and a 2.6.13. The problem is only in 2.6.13, so I am downgrading step by step to identify the faulty file

**mathf** · 13-09-2005

Quote:

Originally Posted by RaZoR1394

With my px-716a, latest version of Nero and Gentoo with the 2.6.13 kernel the buffer keeps going up and down and the burning pauses (to wait for the buffer to fill up) very often. With gentoo-sources 2.6.12-r9 I don't have the this problem. The buffer fluxuations is also present with the vanilla-sources-2.6.13.

Ok got some good news for you... The problem seems to be related to the new timer i/o options. Which freq did you choose when you compile your kernel ?

For me, with 250 Hz, I can reproduce the bug (finally found

)

**mathf** · 13-09-2005

Yepeeah ! Setting the 'Timer frequency' in 'Processor type and features' to 1000 Hz in the kernel configuration file seems to solve our problem.

Axllent · 14-09-2005

Mathf, I salute you

That solved the problem 100%. Now is this a nero "bug" (due to a new feature in the kernel), or a kernel issue?

Bobmeister · 14-09-2005

I don't recompile kernels all that often...what is the "normal" mode that most stock kernels are compiled with? Just curious. Also what are the advantages/disadvantages to the different timer frequency in the kernel compile option....?

Axllent · 14-09-2005

This mode only seems to be in the latest 2.6.13, and I doubt there are "stock kernels" already using that. As far as the advantages / disadvantages go, I have no idea ;-) According to the help:

1000 HZ is the preferred choice for desktop systems and other systems requiring fast interactive responses to events.
250 HZ is a good compromise choice allowing server performance while also showing good interactive responsiveness even on SMP and NUMA systems.
100 HZ is a typical choice for servers, SMP and NUMA systems with lots of processors that may show reduced performance if too many timer interrupts are occurring.

I guess 1000 is the correct one indeed to use regarding desktops and where NeroLINUX is most likely to be used, however I also presume that this will be fixed in future versions of NeroLINUX to accomodate all values, considering that I think stock kernel's most likely will have the default (250).

**mathf** · 15-09-2005

Quote:

Originally Posted by Bobmeister

I don't recompile kernels all that often...what is the "normal" mode that most stock kernels are compiled with? Just curious. Also what are the advantages/disadvantages to the different timer frequency in the kernel compile option....?

In the beginning, 1000 was automatically set. Now, starting from 2.6.13-rc1, they (kernel devs) decided to let user choose its prefered value, and set the default to 250. Do not ask why!

Bobmeister · 16-09-2005

Well...that's good to know for people compiling new kernels....it apparenty could bite a person in the you know what. Thanks for the tip. I don't usually have a need to compile my own kernels, but if and when I do, I'll make sure to keep an eye out for that option and set it correctly.

RaZoR1394 · 17-09-2005

My problem went away recently and when you mention the i/o timing option I remember I tweaked it to 1000hz on all my desktop computers. Thanks for clearing that out

.

04-09-2005	#1 (permalink)
RaZoR1394 CD Freaks Rookie Join Date: Jan 2005 Location: Sweden - Stockholm Posts: 41	Buffer fluxuating with 2.6.13 <a href='http://adstats.cdfreaks.com/www/delivery/ck.php?n=a13febf7&cb=INSERT_RANDOM_NUMBER_HERE' target='_blank'><img src='http://adstats.cdfreaks.com/www/delivery/avw.php?zoneid=30&cb=INSERT_RANDOM_NUMBER_HERE&n=a13febf7' border='0' alt='' /></a> With my px-716a, latest version of Nero and Gentoo with the 2.6.13 kernel the buffer keeps going up and down and the burning pauses (to wait for the buffer to fill up) very often. With gentoo-sources 2.6.12-r9 I don't have the this problem. The buffer fluxuations is also present with the vanilla-sources-2.6.13. Last edited by RaZoR1394; 04-09-2005 at 16:24.

05-09-2005	#2 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 Burning CD or DVD, and at what burning speed? What processor? ATM I only have 2.6.13 on my laptop which has only a CD-writer, so if it happens with DVD I'll have to get 2.6.13 to my desktop to test.

05-09-2005	#3 (permalink)
mathf NeroLinux Moderator Join Date: Jun 2004 Posts: 433	Re: Buffer fluxuating with 2.6.13 Hum... maybe it comes (once again) from a DMA alignment change in the ide-cdrom driver. Have you checked on both side (2.6.12 and 2.6.13) that DMA is correctly enabled on your devices ? (you can check this running hdparm -d /dev/XXX)

05-09-2005	#5 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 Sorry, took me a while.. had to compile the kernel and change my system to udev. I can confirm this (bug?). Here (800MHz) while writing a DVD-RW it ran right the way down to 0%, then filled itself up again. On the one hand I think it's probably working ... as it never left 100% here before, but on the other hand 0%? Actually it did it twice while burning a 1.4GB file at 4x. hdparm shows: using_dma = 1 (on)

06-09-2005	#6 (permalink)
mathf NeroLinux Moderator Join Date: Jun 2004 Posts: 433	Re: Buffer fluxuating with 2.6.13 Okay. Will have a look at it as soon as I have time. Hummm stupid stuff, can you revert drivers/ide/ide-cd.c and drivers/ide/ide.h to version 2.6.12 and check if it is still compilable and if it works better ?

Commercial message
<a href='http://adstats.cdfreaks.com/www/delivery/ck.php?n=a5c5554b&cb=INSERT_RANDOM_NUMBER_HERE' target='_blank'><img src='http://adstats.cdfreaks.com/www/delivery/avw.php?zoneid=29&cb=INSERT_RANDOM_NUMBER_HERE&n=a5c5554b' border='0' alt='' /></a>

06-09-2005	#7 (permalink)
RaZoR1394 CD Freaks Rookie Join Date: Jan 2005 Location: Sweden - Stockholm Posts: 41	Re: Buffer fluxuating with 2.6.13 I'll try that. I will also check if K3B has the same problems when I get back home. Axllent What burner/brand are you using?

06-09-2005	#11 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 I can confirm a difference between these 2 *.c files, however I have just tested this at home on 2.6.13 with the ide-cd.c file from 2.6.12 (which compiles fine btw) and there is no change. The buffer still runs right down to 0% during burning. Other ideas?

12-09-2005	#12 (permalink)
mathf NeroLinux Moderator Join Date: Jun 2004 Posts: 433	Re: Buffer fluxuating with 2.6.13 Ok... we are currently working on this issue. One point that is, we was able to reproduce the problem with a brand new 2.6.13.1 out of kernel.org. The let's say "bug" was occuring whatever the interface was. So I suspect something deeper in the kernel

12-09-2005	#13 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 Thanks for the feedback. I confirmed this yesterday too on 2.6.13.1

12-09-2005	#14 (permalink)
mathf NeroLinux Moderator Join Date: Jun 2004 Posts: 433	Re: Buffer fluxuating with 2.6.13 Have you checked with 2.6.12.6 or with another one ? The last 'stable' kernel I have on my machine is 2.6.11.10 so it's a little bit old

12-09-2005	#15 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 I did have iirc a 2.6.12.x ... which worked quite fine, however I'm at work and can't check which "x" that was as I'm at work ;-) I can/will repost when I get home though. From what I noticed it came with 2.6.13, but then again I don't normally try every 2.6.x.x version, and it may have been introduced in one of the last 2.6.12.x versions.

12-09-2005	#16 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 Sorry, my bad.... 2.6.11.10 was my previous kernel on the machine with a decent burner. On my laptop I had 2.6.12 , but no subversion. If you need me to test a specific 2.6.12.x version just drop a note ;-)

12-09-2005	#17 (permalink)
mathf NeroLinux Moderator Join Date: Jun 2004 Posts: 433	Re: Buffer fluxuating with 2.6.13 No problem, I have compiled a 2.6.12.6 and a 2.6.13. The problem is only in 2.6.13, so I am downgrading step by step to identify the faulty file

13-09-2005	#19 (permalink)
mathf NeroLinux Moderator Join Date: Jun 2004 Posts: 433	Re: Buffer fluxuating with 2.6.13 Yepeeah ! Setting the 'Timer frequency' in 'Processor type and features' to 1000 Hz in the kernel configuration file seems to solve our problem.

14-09-2005	#20 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 Mathf, I salute you That solved the problem 100%. Now is this a nero "bug" (due to a new feature in the kernel), or a kernel issue?

14-09-2005	#21 (permalink)
Bobmeister CD Freaks Junior Member Join Date: Mar 2005 Location: Mt.Prospect, ILLINOIS Posts: 80	Re: Buffer fluxuating with 2.6.13 I don't recompile kernels all that often...what is the "normal" mode that most stock kernels are compiled with? Just curious. Also what are the advantages/disadvantages to the different timer frequency in the kernel compile option....?

14-09-2005	#22 (permalink)
Axllent CD Freaks Junior Member Join Date: May 2005 Location: Rotterdam Posts: 56	Re: Buffer fluxuating with 2.6.13 This mode only seems to be in the latest 2.6.13, and I doubt there are "stock kernels" already using that. As far as the advantages / disadvantages go, I have no idea ;-) According to the help: 1000 HZ is the preferred choice for desktop systems and other systems requiring fast interactive responses to events. 250 HZ is a good compromise choice allowing server performance while also showing good interactive responsiveness even on SMP and NUMA systems. 100 HZ is a typical choice for servers, SMP and NUMA systems with lots of processors that may show reduced performance if too many timer interrupts are occurring. I guess 1000 is the correct one indeed to use regarding desktops and where NeroLINUX is most likely to be used, however I also presume that this will be fixed in future versions of NeroLINUX to accomodate all values, considering that I think stock kernel's most likely will have the default (250).

16-09-2005	#24 (permalink)
Bobmeister CD Freaks Junior Member Join Date: Mar 2005 Location: Mt.Prospect, ILLINOIS Posts: 80	Re: Buffer fluxuating with 2.6.13 Well...that's good to know for people compiling new kernels....it apparenty could bite a person in the you know what. Thanks for the tip. I don't usually have a need to compile my own kernels, but if and when I do, I'll make sure to keep an eye out for that option and set it correctly.

17-09-2005	#25 (permalink)
RaZoR1394 CD Freaks Rookie Join Date: Jan 2005 Location: Sweden - Stockholm Posts: 41	Re: Buffer fluxuating with 2.6.13 My problem went away recently and when you mention the i/o timing option I remember I tweaked it to 1000hz on all my desktop computers. Thanks for clearing that out .

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Don't understand buffer and device buffer in DVD decryptor	gretagun	Newbie Forum	4	04-09-2005 07:35
NEC ND-3500 Writer buffer empty but memory buffer full?	mannen	NEC / Optiarc Burner	5	21-07-2005 00:17
Marlboro $13.95/ Camel $13.95 with delivery !	mcity4	CD Freaks Living Room	23	14-09-2004 19:24