Linux Kernel bug report (includes fix)

Linux Kernel bug report (includes fix)

[Joerg Schilling]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1433 bytes --]

-	Linux Kernel include files (starting with Linux-2.5) are buggy and 
	prevent compilation. Many files may be affected but let me name
	the most important files for me:

	-	/usr/src/linux/include/scsi/scsi.h depends on a nonexistant
		type "u8". The correct way to fix this would be to replace
		any "u8" by "uint8_t". A quick and dirty fix is to call:

			"change u8 __u8 /usr/src/linux/include/scsi/scsi.h"

		ftp://ftp.berlios.de/pub/change/

	-	/usr/src/linux/include/scsi/sg.h includes "extra text" "__user"
		in some structure definitions. This may be fixed by adding
		#include <linux/compiler.h> somewhere at the beginning of
		/usr/src/linux/include/scsi/sg.h

	This bug has been reported several times (starting with Linux-2.5).

	Time to fix: 5 minutes.
	
I did spend far to much time with the discussion on LKML..... so I need a cue
whether it makes sense to continue this discussion.

You now again have the bug report _and_ the fix in a single short mail.

If the bug mentioned above is not fixed in Linux-2.6.8, I will asume that it 
makes no sense to spend further time in discussions with LKML.

Best regards

Jörg

-- 
 EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
       js@cs.tu-berlin.de		(uni)  If you don't have iso-8859-1
       schilling@fokus.fraunhofer.de	(work) chars I am J"org Schilling
 URL:  http://www.fokus.fraunhofer.de/usr/schilling ftp://ftp.berlios.de/pub/schily

Re: Linux Kernel bug report (includes fix)

[Måns Rullgård]

Joerg Schilling <schilling@fokus.fraunhofer.de> writes:

> You now again have the bug report _and_ the fix in a single short mail.

I could see no patch contained in your mail.

-- 
Måns Rullgård
mru@kth.se

Re: Linux Kernel bug report (includes fix)

[Bernd Schubert]

[-- Attachment #1: Type: text/plain, Size: 305 bytes --]

On Saturday 07 August 2004 15:26, Måns Rullgård wrote:
> Joerg Schilling <schilling@fokus.fraunhofer.de> writes:
> > You now again have the bug report _and_ the fix in a single short mail.
>
> I could see no patch contained in your mail.

I guess thats what Joerg wants to have.

Cheers,
	Bernd

[-- Attachment #2: scsi_head_patch.out --]
[-- Type: text/x-csrc, Size: 1934 bytes --]

diff -ru linux-2.6.8-rc2-mm2.bak/include/scsi/scsi.h linux-2.6.8-rc2-mm2/include/scsi/scsi.h
--- linux-2.6.8-rc2-mm2.bak/include/scsi/scsi.h	2004-06-16 07:20:25.000000000 +0200
+++ linux-2.6.8-rc2-mm2/include/scsi/scsi.h	2004-08-07 20:51:27.000000000 +0200
@@ -214,25 +214,25 @@
  */
 
 struct ccs_modesel_head {
-	u8 _r1;			/* reserved */
-	u8 medium;		/* device-specific medium type */
-	u8 _r2;			/* reserved */
-	u8 block_desc_length;	/* block descriptor length */
-	u8 density;		/* device-specific density code */
-	u8 number_blocks_hi;	/* number of blocks in this block desc */
-	u8 number_blocks_med;
-	u8 number_blocks_lo;
-	u8 _r3;
-	u8 block_length_hi;	/* block length for blocks in this desc */
-	u8 block_length_med;
-	u8 block_length_lo;
+	uint8_t _r1;			/* reserved */
+	uint8_t medium;		/* device-specific medium type */
+	uint8_t _r2;			/* reserved */
+	uint8_t block_desc_length;	/* block descriptor length */
+	uint8_t density;		/* device-specific density code */
+	uint8_t number_blocks_hi;	/* number of blocks in this block desc */
+	uint8_t number_blocks_med;
+	uint8_t number_blocks_lo;
+	uint8_t _r3;
+	uint8_t block_length_hi;	/* block length for blocks in this desc */
+	uint8_t block_length_med;
+	uint8_t block_length_lo;
 };
 
 /*
  * ScsiLun: 8 byte LUN.
  */
 struct scsi_lun {
-	u8 scsi_lun[8];
+	uint8_t scsi_lun[8];
 };
 
 /*
diff -ru linux-2.6.8-rc2-mm2.bak/include/scsi/sg.h linux-2.6.8-rc2-mm2/include/scsi/sg.h
--- linux-2.6.8-rc2-mm2.bak/include/scsi/sg.h	2004-08-03 13:22:50.000000000 +0200
+++ linux-2.6.8-rc2-mm2/include/scsi/sg.h	2004-08-07 20:53:41.000000000 +0200
@@ -89,6 +89,8 @@
 
 /* New interface introduced in the 3.x SG drivers follows */
 
+#include <linux/compiler.h>
+
 typedef struct sg_iovec /* same structure as used by readv() Linux system */
 {                       /* call. It defines one scatter-gather element. */
     void __user *iov_base;      /* Starting address  */

Re: Linux Kernel bug report (includes fix)

[Horst von Brand]

Joerg Schilling <schilling@fokus.fraunhofer.de> said:
> -	Linux Kernel include files (starting with Linux-2.5) are buggy and 
> 	prevent compilation.

They do not, the kernel compiles just fine. They are _not_ to be used for
random userspace programs.
-- 
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: Linux Kernel bug report (includes fix)

[Alexander E. Patrakov]

Horst von Brand wrote:
> Joerg Schilling <schilling@fokus.fraunhofer.de> said:
> 
>>-	Linux Kernel include files (starting with Linux-2.5) are buggy and 
>>	prevent compilation.
> 
> 
> They do not, the kernel compiles just fine. They are _not_ to be used for
> random userspace programs.

You are supposed to either bring the needed "sanitized" kernel headers 
with your program, or have those provided by the linux-libc-headers 
(http://ep09.pld-linux.org/~mmazur/linux-libc-headers/) in /usr/include. 
Adding /usr/src/linux/include to the gcc search path is a bug in 
userspace programs.

-- 
Alexander E. Patrakov

Re: Linux Kernel bug report (includes fix)

[Joerg Schilling]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 723 bytes --]

>From vonbrand@inf.utfsm.cl  Sun Aug  8 05:59:28 2004

>Joerg Schilling <schilling@fokus.fraunhofer.de> said:
>> -	Linux Kernel include files (starting with Linux-2.5) are buggy and 
>> 	prevent compilation.

>They do not, the kernel compiles just fine. They are _not_ to be used for
>random userspace programs.

As you don't know how kernel/user interfaces are handled, it would be wise for 
you to keep quiet.....

Jörg

-- 
 EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
       js@cs.tu-berlin.de		(uni)  If you don't have iso-8859-1
       schilling@fokus.fraunhofer.de	(work) chars I am J"org Schilling
 URL:  http://www.fokus.fraunhofer.de/usr/schilling ftp://ftp.berlios.de/pub/schily

Re: Linux Kernel bug report (includes fix)

[Alan Cox]

On Llu, 2004-08-09 at 13:03, Joerg Schilling wrote:
> >Jörg Schilling <schilling@fokus.fraunhofer.de> said:
> >> -	Linux Kernel include files (starting with Linux-2.5) are buggy and 
> >> 	prevent compilation.
> 
> >They do not, the kernel compiles just fine. They are _not_ to be used for
> >random userspace programs.
> 
> As you don't know how kernel/user interfaces are handled, it would be wise for 
> you to keep quiet.....

Linux kernel include files are not meant to be used by user
applications. He's perfectly correct. Glibc has its own exported set.
This is intentional to seperate internals from user space.

Alan

Re: Linux Kernel bug report (includes fix)

[Joerg Schilling]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 788 bytes --]

>From: Alan Cox <alan@lxorguk.ukuu.org.uk>

>> As you don't know how kernel/user interfaces are handled, it would be wise for 
>> you to keep quiet.....

>Linux kernel include files are not meant to be used by user
>applications. He's perfectly correct. Glibc has its own exported set.
>This is intentional to seperate internals from user space.

You should know that GLIBc is unrelated to the Linux kernel interfaces we are 
talking about. Start using serious arguments please.



Jörg

-- 
 EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
       js@cs.tu-berlin.de		(uni)  If you don't have iso-8859-1
       schilling@fokus.fraunhofer.de	(work) chars I am J"org Schilling
 URL:  http://www.fokus.fraunhofer.de/usr/schilling ftp://ftp.berlios.de/pub/schily

Re: Linux Kernel bug report (includes fix)

[Paolo Ciarrocchi]

On Mon, 9 Aug 2004 16:20:11 +0200 (CEST), Joerg Schilling
<schilling@fokus.fraunhofer.de> wrote:
[..]
> 
> You should know that GLIBc is unrelated to the Linux kernel interfaces we are
> talking about. Start using serious arguments please.

brave man...


-- 
ever tried. ever failed. no matter. try again. fail again. fail better.
Personal home page: paoloc.doesntexist.org

Re: Linux Kernel bug report (includes fix)

[Matt Mackall]

On Mon, Aug 09, 2004 at 04:20:11PM +0200, Joerg Schilling wrote:
> >From: Alan Cox <alan@lxorguk.ukuu.org.uk>
> 
> >> As you don't know how kernel/user interfaces are handled, it would be wise for 
> >> you to keep quiet.....
> 
> >Linux kernel include files are not meant to be used by user
> >applications. He's perfectly correct. Glibc has its own exported set.
> >This is intentional to seperate internals from user space.
> 
> You should know that GLIBc is unrelated to the Linux kernel interfaces we are> talking about. Start using serious arguments please.

If you had any inkling, you'd have caught on by now that using kernel
headers in userspace programs has been deprecated for about six years.

They don't compile in userspace because it's not supported. Phased
out, outmoded, obsolete, passe, renounced, defunct, disused,
abandoned. The plumage doesn't enter into it. It's stone dead.

-- 
Mathematics is the supreme nostalgia of our time.

Re: Linux Kernel bug report (includes fix)

[Helge Hafting]

Joerg Schilling wrote:

>
>>Linux kernel include files are not meant to be used by user
>>applications. He's perfectly correct. Glibc has its own exported set.
>>This is intentional to seperate internals from user space.
>>    
>>
>
>You should know that GLIBc is unrelated to the Linux kernel interfaces we are 
>talking about. Start using serious arguments please.
>  
>
The kernel headers are _still_ not meant to be used by userspace,
so don't try that.  If you don't want to use glibc headers - you get to 
write
your own headers.  You can extract info from the kernel headers, but they
cannot be used directly from userspace programs.  They are not designed to
be used that way.

Helge Hafting

Re: Linux Kernel bug report (includes fix)

[Joerg Schilling]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1135 bytes --]


>From: Matt Mackall <mpm@selenic.com>

>> You should know that GLIBc is unrelated to the Linux kernel interfaces we are> talking about. Start using serious arguments please.

>If you had any inkling, you'd have caught on by now that using kernel
>headers in userspace programs has been deprecated for about six years.

Well, everybody has the right to make mistakes and trying to force people
not to use the official header is a big mistake.


If Linux was a complete OS and not only a Kernel and if it was always released 
with a full set of /usr/include files, libc, utilities,.... this could work.
But even then only if somebody would test the consistence of everything.

Releasing the kernel separately requires the kernel distribution to contain
a usable set of include files that match the interfaces inside the kernel.


Jörg

-- 
 EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
       js@cs.tu-berlin.de		(uni)  If you don't have iso-8859-1
       schilling@fokus.fraunhofer.de	(work) chars I am J"org Schilling
 URL:  http://www.fokus.fraunhofer.de/usr/schilling ftp://ftp.berlios.de/pub/schily

Re: Linux Kernel bug report (includes fix)

[Jim Gifford]

[-- Attachment #1: Type: text/plain, Size: 2058 bytes --]

There is a simple fix to your problem, I have sent this patch to you a 
few times, but never got an answer.

I bet eveyone here will agree, this is the proper way to fix the issue

diff -Naur cdrtools-2.01.orig/DEFAULTS/Defaults.gnu 
cdrtools-2.01/DEFAULTS/Defaults.gnu
--- cdrtools-2.01.orig/DEFAULTS/Defaults.gnu    2003-02-16 
00:01:47.000000000 +0000
+++ cdrtools-2.01/DEFAULTS/Defaults.gnu    2004-05-07 00:52:58.111410726 
+0000
@@ -18,7 +18,7 @@
 ###########################################################################
 CWARNOPTS=
 
-DEFINCDIRS=    $(SRCROOT)/include /usr/src/linux/include
+DEFINCDIRS=    $(SRCROOT)/include
 LDPATH=        -L/opt/schily/lib
 RUNPATH=    -R $(INS_BASE)/lib -R /opt/schily/lib -R $(OLIBSDIR)
 
diff -Naur cdrtools-2.01.orig/DEFAULTS/Defaults.linux 
cdrtools-2.01/DEFAULTS/Defaults.linux
--- cdrtools-2.01.orig/DEFAULTS/Defaults.linux    2003-02-16 
00:01:48.000000000 +0000
+++ cdrtools-2.01/DEFAULTS/Defaults.linux    2004-05-07 
00:53:05.850026970 +0000
@@ -18,7 +18,7 @@
 ###########################################################################
 CWARNOPTS=
 
-DEFINCDIRS=    $(SRCROOT)/include /usr/src/linux/include
+DEFINCDIRS=    $(SRCROOT)/include
 LDPATH=        -L/opt/schily/lib
 RUNPATH=    -R $(INS_BASE)/lib -R /opt/schily/lib -R $(OLIBSDIR)
 
diff -Naur cdrtools-2.01.orig/libscg/scsi-linux-sg.c 
cdrtools-2.01/libscg/scsi-linux-sg.c
--- cdrtools-2.01.orig/libscg/scsi-linux-sg.c    2004-04-18 
10:26:44.000000000 +0000
+++ cdrtools-2.01/libscg/scsi-linux-sg.c    2004-05-07 
00:52:47.953227014 +0000
@@ -65,6 +65,14 @@
 
 #if LINUX_VERSION_CODE >= 0x01031a /* <linux/scsi.h> introduced in 
1.3.26 */
 #if LINUX_VERSION_CODE >= 0x020000 /* <scsi/scsi.h> introduced 
somewhere. */
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
+    #define __KERNEL__
+    #include <asm/types.h>
+    #include <asm/byteorder.h>
+    #undef __KERNEL__
+#endif
+
 /* Need to fine tune the ifdef so we get the transition point right. */
 #include <scsi/scsi.h>
 #else

-- 
----
Jim Gifford
maillist@jg555.com


[-- Attachment #2: cdrtools-2.01-kernel_2.6-3[1].patch --]
[-- Type: text/x-patch, Size: 2013 bytes --]

Submitted By: Jim Gifford (jim at linuxfromscratch dot org)
Date: 2004-05-06
Initial Package Version: 2.01
Origin: Jim Gifford and http://kerneltrap.org/node/view/879
Upstream Status: Sent
Description: Fixes CDRTools Compile with the 2.6 Kernel
 
 
diff -Naur cdrtools-2.01.orig/DEFAULTS/Defaults.gnu cdrtools-2.01/DEFAULTS/Defaults.gnu
--- cdrtools-2.01.orig/DEFAULTS/Defaults.gnu	2003-02-16 00:01:47.000000000 +0000
+++ cdrtools-2.01/DEFAULTS/Defaults.gnu	2004-05-07 00:52:58.111410726 +0000
@@ -18,7 +18,7 @@
 ###########################################################################
 CWARNOPTS=
 
-DEFINCDIRS=	$(SRCROOT)/include /usr/src/linux/include
+DEFINCDIRS=	$(SRCROOT)/include
 LDPATH=		-L/opt/schily/lib
 RUNPATH=	-R $(INS_BASE)/lib -R /opt/schily/lib -R $(OLIBSDIR)
 
diff -Naur cdrtools-2.01.orig/DEFAULTS/Defaults.linux cdrtools-2.01/DEFAULTS/Defaults.linux
--- cdrtools-2.01.orig/DEFAULTS/Defaults.linux	2003-02-16 00:01:48.000000000 +0000
+++ cdrtools-2.01/DEFAULTS/Defaults.linux	2004-05-07 00:53:05.850026970 +0000
@@ -18,7 +18,7 @@
 ###########################################################################
 CWARNOPTS=
 
-DEFINCDIRS=	$(SRCROOT)/include /usr/src/linux/include
+DEFINCDIRS=	$(SRCROOT)/include
 LDPATH=		-L/opt/schily/lib
 RUNPATH=	-R $(INS_BASE)/lib -R /opt/schily/lib -R $(OLIBSDIR)
 
diff -Naur cdrtools-2.01.orig/libscg/scsi-linux-sg.c cdrtools-2.01/libscg/scsi-linux-sg.c
--- cdrtools-2.01.orig/libscg/scsi-linux-sg.c	2004-04-18 10:26:44.000000000 +0000
+++ cdrtools-2.01/libscg/scsi-linux-sg.c	2004-05-07 00:52:47.953227014 +0000
@@ -65,6 +65,14 @@
 
 #if LINUX_VERSION_CODE >= 0x01031a /* <linux/scsi.h> introduced in 1.3.26 */
 #if LINUX_VERSION_CODE >= 0x020000 /* <scsi/scsi.h> introduced somewhere. */
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
+	#define __KERNEL__
+	#include <asm/types.h>
+	#include <asm/byteorder.h>
+	#undef __KERNEL__
+#endif
+
 /* Need to fine tune the ifdef so we get the transition point right. */
 #include <scsi/scsi.h>
 #else

Re: Linux Kernel bug report (includes fix)

[Horst von Brand]

Joerg Schilling <schilling@fokus.fraunhofer.de> said:
> >From vonbrand@inf.utfsm.cl  Sun Aug  8 05:59:28 2004
> >Joerg Schilling <schilling@fokus.fraunhofer.de> said:
> >> -	Linux Kernel include files (starting with Linux-2.5) are buggy and 
> >> 	prevent compilation.

> >They do not, the kernel compiles just fine. They are _not_ to be used for
> >random userspace programs.

> As you don't know how kernel/user interfaces are handled,

Funny... How would you know what I do or don't know?

>                                                           it would be
> wise for you to keep quiet.....

I _do_ know about kernel/user interfaces (in Linux); you clearly haven't
got a clue. Please keep quiet.
-- 
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: Linux Kernel bug report (includes fix)

[Joerg Schilling]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1350 bytes --]

>From: Jim Gifford <maillist@jg555.com>


>There is a simple fix to your problem, I have sent this patch to you a 
>few times, but never got an answer.

>I bet eveyone here will agree, this is the proper way to fix the issue

Of course, what you like to do is not the proper way....

>-DEFINCDIRS=    $(SRCROOT)/include /usr/src/linux/include
>+DEFINCDIRS=    $(SRCROOT)/include

/usr/src/linux/include is definitely needed on Linux-2.2 and before



> #if LINUX_VERSION_CODE >= 0x01031a /* <linux/scsi.h> introduced in 
>1.3.26 */
> #if LINUX_VERSION_CODE >= 0x020000 /* <scsi/scsi.h> introduced 
>somewhere. */
>+
>+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 0)
>+    #define __KERNEL__
>+    #include <asm/types.h>
>+    #include <asm/byteorder.h>
>+    #undef __KERNEL__
>+#endif

If an application would need to #define __KERNEL__ and to include 
inofficial include files, the official include files are broken 
and should be fixed. This is exactly the reason why I send a bug report
against include/scsi/scsi.h & include/scsi/sg.h



Jörg

-- 
 EMail:joerg@schily.isdn.cs.tu-berlin.de (home) Jörg Schilling D-13353 Berlin
       js@cs.tu-berlin.de		(uni)  If you don't have iso-8859-1
       schilling@fokus.fraunhofer.de	(work) chars I am J"org Schilling
 URL:  http://www.fokus.fraunhofer.de/usr/schilling ftp://ftp.berlios.de/pub/schily

Re: Linux Kernel bug report (includes fix)

[Bill Davidsen]

Jim Gifford wrote:
> There is a simple fix to your problem, I have sent this patch to you a 
> few times, but never got an answer.
> 
> I bet eveyone here will agree, this is the proper way to fix the issue

Everyone has agreed that not using kernel headers is the proper way to 
fix the issue. Joerg can include the extracted and cleansed headers 
released regularly and include them in his software.

The whole ide that no one can compile his application unless they have 
the kernel source online is a non-starter in the first place.

-- 
    -bill davidsen (davidsen@tmr.com)
"The secret to procrastination is to put things off until the
  last possible moment - but no longer"  -me