I can confirm this, as I have had the same problem since upgrading to
kernel-2.6.8-1.521

Before the upgrade I could record disks in SAO / DAO mode without a
problem, but now I get the error message

âcdrecord: Drive does not support SAO recording.â
âcdrecord: Illegal write mode for this drive.â

I have a scsi YAMAHA CRW2100S writer connected to an Adaptec 19160
controller (AIC7XXX). I can still read disks in the drive, it's just
writing them that is giving me problems.

I think bugs 130577 and 130414 are duplicates of this bug... or vice
versa :-)

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=130577

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=130414

  Well, since k3b and xcdroast just invoke cdrecord, I thought the
cleanest way to show the underlying bug was by calling cdrecord directly.

I wasn't being critical of you, I was just listing all the bugs in one
place so whoever fixes it can resolve the problem for them as well.

While I'm at it, here is another thread on the matter

http://lkml.org/lkml/2004/8/15/189

sounds like our problem. I tried executing cdrecord as root and I can
confirm that it does work as it should, the problem is only when
cdrecord is executed as a normal user.

  Sorry for the misunderstanding, I didn't think you were being
critical.  I can verify the behavior with the -checkdrive option,
which is interesting since I set the setuid bit on my cdrecord
executable.  Apparently the real UID has to be root as well.
  P.S. I forgot about growisofs.  I suppose 130576 and 130577 could be
considered "primary", and 130414 could be considered a dupe of both of
these jointly (is that possible?).  But the next kernel update should
fix all of them at once, so it probably doesn't matter.

*** Bug 130577 has been marked as a duplicate of this bug. ***

*** Bug 130414 has been marked as a duplicate of this bug. ***

this seems still not fixed in 2.6.9-rc1

From what I've gathered on the subject I think the problem is due to
new security features in 2.6.8.1 which have tightened up the command
filtering in order to stop some potential security risks such as any
user being able to overwrite your cd-r's firmware.

For more information on the matter there is a thread on the lkml at

http://lkml.org/lkml/2004/8/15/159

Also, if you havn't read what I wrote in bug 130414 here it is again
because its important.

Don't try to record audio cd's as root with this kernel because there
is a memory leak that will eventually crash your system.

If you need to burn audio cd's you need the following patches.

http://kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.8.1/2.6.8.1-mm3/broken-out/bio_uncopy_user-mem-leak.patch

http://kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.8.1/2.6.8.1-mm3/broken-out/bio_uncopy_user-mem-leak-fix.patch

There is more info on this at the lkml at

http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&c2coff=1&th=dbce4a618e91ea4&seekm=2vbef-2oe-11%40gated-at.bofh.it#link10

I can confirm that these patches fix the mem leak and the audio cd's
produced are not corrupt.

Created attachment 103156 [details]
this patch fixes the issue witch cdrecord but not with k3b

This patch fixes cdrecord so it can burn with 2.6.8.1 as normal user.
by default cdrecord drops it root privilleges befor burning this patch lets
cdrecord don't do this.It keeps its root privilleges.

Could it be this problem that hits cdrdao as well? When trying to run
cdrdao-1.1.9-2 (from rawhide) I'm getting the error message
ERROR: Cannot prevent/allow medium removal.
and then it just exits.

*** Bug 130864 has been marked as a duplicate of this bug. ***

*** Bug 131085 has been marked as a duplicate of this bug. ***

*** Bug 131014 has been marked as a duplicate of this bug. ***

I can confirm the problem with the memory leak and the system crash,
when trying to burn an audio CD. The crash is always reproducible.
With my -521 Kernel on FC2 only root can convince cdrecord to do
anything at all. 

The memory leak should be fixed in 2.6.8-1.533, because its based on
2.6.9-rc1-bk4.

This seems partially fixed in kernel 2.6.8-1.533.
I can use cdrecord as a normal user on my IDE-DVD-writer again. Did
not try growisofs, though.

k3b still claims that the drive does not support any writing modes,
and my SCSI-writer is still listed as a reader.

I tested this kernel (2.6.8-1.533) witb a modifized k3b version.
i changed this :
------------------
int K3bCdDevice::openDevice( const char* name )
{
  int fd = -1;
#ifdef HAVE_RESMGR
  // first try resmgr
  fd = ::rsm_open_device( name, O_RDONLY | O_NONBLOCK );
  kdDebug() << "(K3bCdDevice) resmgr open: " << fd << endl;
#endif

  if( fd < 0 )
    fd = ::open( name, O_RDONLY | O_NONBLOCK );

  if( fd < 0 )
    fd = -1;

  return fd;
}
-------------
to
--------------
int K3bCdDevice::openDevice( const char* name )
{
  int fd = -1;
#ifdef HAVE_RESMGR
  // first try resmgr
  fd = ::rsm_open_device( name, O_RDWR | O_NONBLOCK );
  kdDebug() << "(K3bCdDevice) resmgr open: " << fd << endl;
#endif

  if( fd < 0 )
    fd = ::open( name, O_RDWR | O_NONBLOCK );

  if( fd < 0 )
    fd = -1;

  return fd;
}
--------------
recompiled and it worked!

i forgot to say that the file was src/device/k3bdevice.cpp

With kernel-smp-2.6.8-1.541 using cdrdao, I get lots of lines like
ERROR: Read buffer capacity failed.
printed by cdrdao, but it seems like the CD burned works after all...

I have had similar problems with both the 2.6.8-1.541 kernel, and a 
compiled 2.6.8.1 kernel. 
However, if you load the ide-scsi module and run the dvd+rw (in my
case) via scsi emulation, things work again, but at about 1/10-th of
the speed. 

In my case the proble was not the inability to write, but the eventual
"out-of-memory" crash of the kernel when writing about 1/2 of a CD in 
ATAPI mode. 

I reverted to kernel 2.6.7 and things work fine. 
It is definitely a bug in the kernel and not in cdrecord, xcdroast, etc. 

I also had memory-leaks in Mozilla inder 2.6.8, (and not under 2.6.7). 

With kernel kernel-smp-2.6.8-1.541, growisofs (from
dvd+rw-tools-5.17.4.8.6-2) dies with the error message
:-( unable to PREVENT MEDIA REMOVAL: Operation not permitted

Created attachment 104561 [details]
cdrecord -v -dummy -audio MY_WAV_FILES 2>&1 > cdr.log

cdrecord output

Forgot to include in my comment #23
"cdrecord -v -dummy -audio MY_WAV_FILES 2>&1 > cdr.log" 
...
Running on FC2 Kernel 2.6.8-1.521
Always, after trying to write track 04, nothing but OOMs for me.

Read Comment #9 and patch your kernel with the patches linked to...
this will fix the memory leak.

FYI: Harald's cdrecord, k3b-0.11.14-0.FC2.1 and friends updates for
FC2 as of today do _not_ fix this problem here (only k3b tested).

FWIW, I have a LG GSA-4120B (DVD{-RW,+RW,-RAM},CDRW).  And everything
seems to work when run as root; non-root remains borked.

Potential duplicates: bug 131031, bug 131692, bug 131789, bug 133367.

hmm, does FC3 .541 have anything different?

I cannot do a full FC3testX upgrade at the moment, but will see if I
can get FC2 to run the FC3 kernel if you think it's doable and worth
checking.

should do no harm... just ignore eventual mkinitrd deps when
installing the new kernel

Ok, results: does not work with 2.6.8-1.521, works with 2.6.8-1.541
and 2.6.8-1.590.

My simple test case was to run k3b, check the devices info from its
config UI for sanity, and try to erase a CDRW.

Correct device info from k3b (got this with .541 and .590):
- Writes CDs, CD/RWs, Reads DVDs, Writes DVD-R(W)s, DVD+R(W)s: all yes
Incorrect info from .521:
- Writes CDs and Writes DVD+R(W)s: no, others yes

When not working, k3b's "Tools->CD->Erase CD-RW" has an empty Burning
Device drop-down box.

Let me know if you need more info.

BTW, probably an upstream k3b bug, but being curious, clicking "Start"
in that dialog without burning devices anyway starts *something* and
displays a "Erasing CD-RW" dialog, but it certainly isn't erasing the
media nor seemingly touching the drive at all (and the dialog can't be
dismissed).

The k3b fix only works with kernel 2.6.9x not with 2.6.8.
Because .540 and .590 are based on 2.6.9x it works. .521 is based on
2.6.8.
This means we also need an updated kernel.

*** Bug 131031 has been marked as a duplicate of this bug. ***

*** Bug 131692 has been marked as a duplicate of this bug. ***

*** Bug 131789 has been marked as a duplicate of this bug. ***

*** Bug 133367 has been marked as a duplicate of this bug. ***

Just out of curiosity, where do you find the .590 kernel? At
ftp://download.fedora.redhat.com:/pub/fedora/linux/core/development/SRPMS
all I can see is the .541 kernel.

No longer there either, but I got it from
http://www.funet.fi/pub/mirrors/ftp.redhat.com/pub/fedora/linux/core/development/i386/Fedora/RPMS/

Wonder why it was pulled...
Anyway, this discussion is trailing away and with the latest errata
releases of cdrtools, xcdroast, dvd+rw-tools, k3b and cdrdao I believe
the problem is fixed so maybe we should close this kernel bug.

The problem isn't fixed by the errata of cdrtools et al, it needs a
kernel update to work.

There also seems to be a problem with the new k3b (and al) and kernel
2.6.8-1.521 when doing multi-session CDs. The burning process appears
to work but the CD is not usable. I'll try to confirm it tomorrow with
the original version of k3b from FC2 (which seemed to work as far as I
remember). Could anyone confirm the problem ? And if so, does anyone
know if kernel 2.6.8-1.541 solves this problem because I can't find
.590 version either ?

I've checked my previous post by applying selective updates to a
scratch Fedora Core 2 distrib.
I've upgraded to mkisofs-2.01.1-0.FC2.1, everything still worked.
Then I tried to upgrade to cdrecord-2.01.1-0.FC2.1 and multi-session
is now broken : with k3b (ran as root), if you create a single session
CD with joliet extensions, it works ; if you try a multi-session CD
with joliet extensions, you can't use the resulting CD (you have an
"attempt to access beyond end of device" errors)
Updating to k3b-0.11.14-0.FC2.1 doesn't improve or change anything
except for the error messages where you can read things like "hdc:
media error (bad sector) : status=0x51 { DriveReady SeekComplete Error
}". Single session CDs still work.

I've filed Bug 134642 about this.

I am completely unable to burn audio cds with the new version.  After
about 500 mb burned, it hangs up.  The cd burner light flashes
normally as in burning, while the hd led appears to stay on
continuously.  System response goes to zero.  The only recovery I've
been able to do is to punch out the system manually with the reset
button.  Not a good experience.  I did manage to write one data cd,
which paused similarly, but eventually recovered.

After upgrading to 2.6.9 burning as a normal user functions properly
again.
'bug' solved?

no we still need a kernel update for fc2

*** Bug 136851 has been marked as a duplicate of this bug. ***

*** Bug 136968 has been marked as a duplicate of this bug. ***

*** Bug 136066 has been marked as a duplicate of this bug. ***

k3b now works for me with the 2.6.9 kernel released for FC 2 this
week.  I'm using the test 2.6.9-1.6_FC2 kernel from yesterday that
will, presumably, be released very soon (since it fixes a new SATA
bug). I have not tested multi-session cds, just cd and dvd writing.  I
have these versions of s/w:

$ rpm -q cdrecord mkisofs k3b
cdrecord-2.01.1-0.FC2.1
mkisofs-2.01.1-0.FC2.1
k3b-0.11.14-0.FC2.2

This problem is still occuring to me with kernel-2.6.9-1.6_FC2.
cdrecord is installed as chmod +s.

Cdrecord-Clone 2.01-dvd (i686-pc-linux-gnu) Copyright (C) 1995-2004
Jörg Schilling
Note: This version is an unofficial (modified) version with DVD support
Note: and therefore may have bugs that are not present in the original.
Note: Please send bug reports or support requests to
http://bugzilla.redhat.com/bugzilla
Note: The author of cdrecord should not be bothered with problems in
this version.
TOC Type: 1 = CD-ROM
scsidev: 'ATA:1,0,0'
devname: 'ATA'
scsibus: 1 target: 0 lun: 0
Linux sg driver version: 3.5.27
Using libscg version 'schily-0.8'.
cdrecord: Warning: using inofficial libscg transport code version
(schily - Red Hat-scsi-linux-sg.c-1.83-RH '@(#)scsi-linux-sg.c       
1.83 04/05/20 Copyright 1997 J. Schilling').
SCSI buffer size: 64512
cdrecord: Cannot allocate memory. Cannot get SCSI I/O buffer.

xpc19:~> rpm -q cdrecord
cdrecord-2.01.1-0.FC2.1
xpc19:~> uname -a
Linux xpc19.ast.cam.ac.uk 2.6.9-1.6_FC2 #1 Thu Nov 18 22:03:19 EST
2004 i686 i686 i386 GNU/Linux

mass update for old bugs:

Is this still a problem with the 2.6.9 based update kernel ?

I can burn without any problems as normal user....
2.6.9-1.681_FC3 and cdrecord-2.01.1-5

Comment 52 mentions a FC3 kernel which, according to earlier comments
also, seems to not have this problem.  However, comment 50 mentions
things still being broken with the latest 2.6.9-based FC2 update
kernel.  I don't have FC2 any more so I can't test/verify (and will
remove myself from Cc), but it looks to me like this bug should be
reopened.

On k3b.org is this in the news section:

"Cdrecord won't run anymore when configured suid root. The reason is that the
new kernel forbids suid root programs to use the SCSI subsystem (or something
like that, correct me if I am wrong :)
So Cdrecord needs to be configured without suid root and the next version of
K3bSetup will do that.
Cdrdao does not have this problem. I think that's becasue it drops the root
privileges in an earlier stage but I am not sure about that."

I have cdrecord without setuid root and it seems it works nice under normal user
(kernel 2.6.9-1.6_FC2, cdrecord 2.01.1-0.FC2.1).

I still have the problem with latest kernel update

# rpm -q cdrecord mkisofs kernel-smp
cdrecord-2.01.1-0.FC2.1
mkisofs-2.01.1-0.FC2.1
kernel-smp-2.6.9-1.6_FC2

the problem occurs with & without SUID on cdrecord.  I run cdrecord
from nautilus-cd-burner:


cdrecord: Cannot allocate memory. Cannot get SCSI I/O buffer.
SCSI buffer size: 64512
cdrecord: Warning: using inofficial libscg transport code version
(schily - Red Hat-scsi-linux-sg.c-1.83-RH '@(#)scsi-linux-sg.c    1.83
04/05/20 Copyright 1997 J. Schilling').
Linux sg driver version: 3.5.31
scsibus: 1 target: 0 lun: 0
scsidev: '1,0,0'
cdrecord: Continuing in 5 seconds...
cdrecord: Future versions of cdrecord may have different drive
dependent defaults.
cdrecord: Asuming -tao mode.
cdrecord: No write mode specified. 

Having fetched a 2.6.7-based kernel (the Stanford CCRMA one), I
confirm that the problem does not appear with 2.6.7.

Any bug-fix that may have come affecting the IDE layer in latest 2.6.9
is of no help for me, as I am using an USB CD Writer.

Can I hope in any way have some fix on FC2?  Looking at the mail
thread, I don't see why FC3 would be unaffected, so...