PROBLEM: 2.6.0-test5: ioctl.h: _IOC_TYPECHECK: Is it a bug?

DISCUSSION Sinelnikov Evgeny 2 participants 2 messages Oct 9, 2003 – Oct 19, 2003

Share & tools

Sinelnikov Evgeny

author

2003-10-09 19:34 UTC #0 lore raw

Hello!
I have problem with linux-utils.2.12 compile on 2.6.0-test5 headers. From this version instead 2.6.0-test4 was changed next file where added next string:

include/asm-i386/ioctl.h:

#define _IOC_TYPECHECK(t) \
        ((sizeof(t) == sizeof(t[1]) && \
          sizeof(t) < (1 << _IOC_SIZEBITS)) ? \
          sizeof(t) : __invalid_size_argument_for_IOC)

This string have problems with compiling files using _IOR macros from ioctl.h Here is example from util-linux-2.12:
#define BLKBSZGET _IOR(0x12,112,sizeof(int))
ioctl (fd, BLKBSZGET, &bsz) == 0
After precompiling looking like this:
ioctl (fd, (((2U) << (((0 +8)+8)+14)) | (((0x12)) << (0 +8)) | (((112)) << 0) | (((((sizeof(sizeof(int)) == size
of(sizeof(int)[1]) && sizeof(sizeof(int)) < (1 << 14)) ? sizeof(sizeof(int)) :
__invalid_size_argument_for_IOC))) << ((0 +8)
+8))), &bsz) == 0

Is it a bug or new feature? May be this is a bug of package?

PS:
Output of scripts/ver_linux on native ALT Linux:
lfs@localhost src]$ [lfs@localhost linux-2.6.0]$ sh scripts/ver_linux If some fields are empty or look unusual you may have an old version. Compare to the current minimal requirements in Documentation/Changes.

Linux localhost.localdomain 2.4.20-alt7-up #1 Fri Mar 14 14:57:05 MSK 2003 i686 unknown unknown GNU/Linux

Gnu C                  3.2.3
Gnu make               3.80
util-linux             2.11z
mount                  2.11z
module-init-tools      2.4.25
e2fsprogs              1.32
reiserfsprogs          3.6.4
pcmcia-cs              3.2.4
quota-tools            3.09.
PPP                    2.4.2b3
nfs-utils              1.0.3
Linux C Library        2.2.6
Dynamic linker (ldd)   2.2.6
Procps                 2.0.10
Net-tools              1.60
Console-tools          0.2.3
Sh-utils               5.0
Modules Loaded         sr_mod ide-scsi ide-cd scsi_mod cdrom ppp_deflate

zlib_inflate zlib_deflate bsd_comp ppp_async ppp_generic slhc appletalk ipx nvidia autofs4 snd-pcm-oss snd-mixer-oss snd-intel8x0 snd-pcm snd-timer snd-ac97-codec snd-mpu401-uart snd-rawmidi snd-seq-device snd soundcore floppy serial isa-pnp ntfs nls_koi8-r nls_cp866 vfat fat ext3 jbd ext2 rtc reiserfs

Output of scripts/ver_linux on chrooted LFS, where compiled util-linux-2.12 where was the problem:

root:/usr/src/linux-2.6.0$ sh scripts/ver_linux
If some fields are empty or look unusual you may have an old version.
Compare to the current minimal requirements in Documentation/Changes.

Linux localhost.localdomain 2.4.20-alt7-up #1 Fri Mar 14 14:57:05 MSK 2003 i686 GenuineIntel unknown GNU/Linux

Gnu C 3.3.1
Gnu make 3.80
scripts/ver_linux: line 27: fdformat: command not found

mount                  2.12
e2fsprogs              1.34
Linux C Library        2.3.2
Dynamic linker (ldd)   2.3.2
Linux C++ Library      5.0.5
Procps                 3.1.13
Net-tools              1.60
Kbd                    1.08
Sh-utils               5.0
Modules Loaded         sr_mod ide-scsi ide-cd scsi_mod cdrom ppp_deflate

test.sh:
tar -xzf util-linux-2.12.tar.gz &&
cd util-linux-2.12 &&
./configure &&
make

Sinelnikov Evgeny

Andries Brouwer

2003-10-19 21:14 UTC #1 lore raw

Quoted 4 lines · Sinelnikov Evgeny

I have problem with linux-utils.2.12 compile on 2.6.0-test5 headers. From this version instead 2.6.0-test4 was changed next file where added next string:
include/asm-i386/ioctl.h:

Quoted 8 lines

#define _IOC_TYPECHECK(t) \
        ((sizeof(t) == sizeof(t[1]) && \
          sizeof(t) < (1 << _IOC_SIZEBITS)) ? \
          sizeof(t) : __invalid_size_argument_for_IOC)
This string have problems with compiling files using _IOR macros from ioctl.h Here is example from util-linux-2.12:
#define BLKBSZGET _IOR(0x12,112,sizeof(int))

Normally, user space includes <sys/ioctl.h>, which again includes <asm/ioctl.h> which defines _IOR etc.

It sounds like you use a symlink /usr/include/asm into a kernel source tree. Doing such things is discouraged. For people that follow kernel development the kernel source tree changes from day to day. When some utility shows obscure errors it would be necessary to recall when, and with what kernel headers it was last compiled in order to debug the problem. That is not good for stability and reproducibility.

Moreover, there are no guarantees that kernel headers work in user space. Often they do for some kernel versions and don't for other kernel versions. It is the task of the distributor of include files to pick a suitable set.

Finally, when something is wrong with util-linux, please tell [email protected]. (There used to be a mailing list but I think it has been killed since the recent spam floods.)

Andries