[Bug 42986] New: ext2_fs.h requires undefined type umode_t

[Bug 42986] New: ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986

           Summary: ext2_fs.h requires undefined type umode_t
           Product: File System
           Version: 2.5
    Kernel Version: 3.3
          Platform: All
        OS/Version: Linux
              Tree: Mainline
            Status: NEW
          Severity: normal
          Priority: P1
         Component: ext2
        AssignedTo: fs_ext2@kernel-bugs.osdl.org
        ReportedBy: allan@archlinux.org
        Regression: Yes


In the user-space headers:

/usr/include/linux/ext2_fs.h:178:41: error: unknown type name 'umode_t'

umode_t is not provided in userspace since commit 0583fcc9

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986


Theodore Tso <tytso@mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tytso@mit.edu




--- Comment #1 from Theodore Tso <tytso@mit.edu>  2012-03-24 14:04:34 ---
Why do you need ext2_fs.h?   We're planning on removing ext2_fs.h as an
exported header file very shortly (perhaps as soon as v3.4; it's already in Al
Viro's VFS git tree).

Ext 2/3/4 userspace programs ***really*** should be using libext2fs, and if for
some reason they don't want to do that, they can use the ext2_fs.h header file
that is shipped with e2fsprogs and libext2fs.

Programs written as much as 8 years ago and which used libext2fs generally were
able to support ext4 file systems have generally been able to support ext4;
that's one of the major reasons why I strongly suggest the use of that library,
instead of trying to use kernel header files to roll your own thing....

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #2 from Allan <allan@archlinux.org>  2012-03-24 14:16:19 ---
I noticed this issue while building the latest syslinux.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986


Theodore Tso <tytso@mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |INVALID




--- Comment #3 from Theodore Tso <tytso@mit.edu>  2012-03-25 01:20:58 ---
>From Peter Anvin, the maintainer of Syslinux:

   Syslinux already includes a private copy of ext2_fs.h, but I
   need to scrub the umode_t inclusion.

This is therefore something that needs to be fixed in Syslinux, and not in the
kernel, so I'm closing this bug.  Peter has been made aware of this issue, so
presumably he will be fixing and releasing a new version of Syslinux shortly.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #4 from Allan <allan@archlinux.org>  2012-03-25 02:47:29 ---
Sure, that fixes syslinux...  But what about every other user of this header? 
The answer of "don't use it" is not very good if it is being installed.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #5 from Theodore Tso <tytso@mit.edu>  2012-03-25 17:21:44 ---
It's not going to be installed.  Al Viro has patches in his git tree which is
going to make it ***disappear*** from the kernel source tree altogether.  The
definitions which are needed for fs/ext2 will be moved to fs/ext2/ext2.h.   The
userspace definitions are being deleted.  The few places in the rest of the
kernel which were using ext2_fs.h and implying knowledge of the ext2 superblock
layout are being changed to use a function exported by the ext2 code, thus
properly providing better abstraction between the various kernel components.

There should be no valid user of this header, and there should not have been
one for at least eight years.   If they do exist, we'll find out when they
scream, and then we'll tell them that they are broken, and they should be using
the ext2_fs.h being shipped with e2fsprogs, and ideally libext2fs being shipped
with e2fsprogs.

The whole point is that it's insane to be maintaining two userspace headers,
especially since the one in the kernel tree has effectively not been maintained
for years (at least for use in userspace).

Again, if you know of other actual users of the ext2_fs.h being shipped in the
kernel, please let me know, and I'll personally reach out to them (and tell
them that they are close to a decade behind the times....)

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986


H. Peter Anvin <hpa@zytor.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |REOPENED
                 CC|                            |hpa@zytor.com
         Resolution|INVALID                     |




--- Comment #6 from H. Peter Anvin <hpa@zytor.com>  2012-03-27 04:41:22 ---
Okay... I looked into this a bit more.

The reason Syslinux included <linux/ext2_fs.h> is to get the EXT2_IOC_*
constants.  If <linux/ext2_fs.h> is going away, then what is expected to export
those constants?

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #7 from Theodore Tso <tytso@mit.edu>  2012-03-27 05:02:37 ---
Which EXT2_IOC_* constants are you using?   Aside from the largely specialized
and as far as I know, not ever used, EXT2_IOC_GETRSVSZ and EXT3_IOC_SETRSVSZ,
you can also use (and in fact are encouraged to use), the generic FS_IOC_*
ioctls instead:

#define    EXT2_IOC_GETFLAGS        FS_IOC_GETFLAGS
#define    EXT2_IOC_SETFLAGS        FS_IOC_SETFLAGS
#define    EXT2_IOC_GETVERSION        FS_IOC_GETVERSION
#define    EXT2_IOC_SETVERSION        FS_IOC_SETVERSION

These ioctl started out as ext2 specific, but they are used by multiple file
systems.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #8 from H. Peter Anvin <hpa@zytor.com>  2012-03-27 05:33:47 ---
That's fine, but it's also a gratuitous change to previously working userspace.
 It sounds like the right thing would be to leave <linux/ext2_fs.h> in place
exporting only those constants as a backwards compatibility measure.

I can update Syslinux, but it isn't just about Syslinux; in general it's rude
to break userspace unnecessarily :-/

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #9 from Theodore Tso <tytso@mit.edu>  2012-03-27 05:58:48 ---
Breaking things at compile-time (as opposed to breaking already created
executables) is not something I worry that much about --- especially if they
are flags that are fairly specialized.   In fact I was quite surprised there
were programs other than lsattr/chattr that were using these ioctls.....

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #10 from H. Peter Anvin <hpa@zytor.com>  2012-03-27 06:05:47 ---
The flags-setting ioctls aren't really all that different from chmod(), so I'm
a bit surprised at your surprise.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #11 from Theodore Tso <tytso@mit.edu>  2012-03-27 06:17:59 ---
Yeah, but chmod is POSIX; EXT2_IOC_GETFLAGS is not.  And if you are using the
EXT2_IOC_GETFLAGS variant, then you're saying that you only care about it
working on ext2, and most programs hopefully aspire to more portability than
that.  I really doubt there are _that_ many ext2-specific program out there,
and the flags that can be set by IOC_GETFLAGS really are more esoteric than the
chmod flags.

In any case, Al Viro is carrying the patch in his tree, so ultimately it will
be up to him.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986





--- Comment #12 from H. Peter Anvin <hpa@zytor.com>  2012-03-27 06:21:24 ---
It's mostly an issue with keeping compatibility with as many systems as
possible.  It is fairly painful to deal with breakages in userspace without
breaking backwards compatibility.

I'll talk to Al.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

Re: [Bug 42986] ext2_fs.h requires undefined type umode_t

[Christoph Hellwig]

On Tue, Mar 27, 2012 at 05:33:47AM +0000, bugzilla-daemon@bugzilla.kernel.org wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=42986
> 
> 
> 
> 
> 
> --- Comment #8 from H. Peter Anvin <hpa@zytor.com>  2012-03-27 05:33:47 ---
> That's fine, but it's also a gratuitous change to previously working userspace.
>  It sounds like the right thing would be to leave <linux/ext2_fs.h> in place
> exporting only those constants as a backwards compatibility measure.

That's probably a good idea, given how widespread people use the flags.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986


Denys Vlasenko <vda.linux@googlemail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vda.linux@googlemail.com




--- Comment #13 from Denys Vlasenko <vda.linux@googlemail.com>  2012-04-12 11:37:05 ---
For the record: busybox's mkfs_ext2.c uses linux/ext2_fs.h.
If its include is commented out, the following errors pop out:

error: ‘EXT2_DFL_MAX_MNT_COUNT’ undeclared (first use in this function)
error: ‘EXT2_DYNAMIC_REV’ undeclared (first use in this function)
error: ‘EXT2_ERRORS_DEFAULT’ undeclared (first use in this function)
error: ‘EXT2_FEATURE_COMPAT_DIR_INDEX’ undeclared (first use in this function)
error: ‘EXT2_FEATURE_COMPAT_RESIZE_INO’ undeclared (first use in this function)
error: ‘EXT2_FEATURE_COMPAT_SUPP’ undeclared (first use in this function)
error: ‘EXT2_FEATURE_INCOMPAT_FILETYPE’ undeclared (first use in this function)
error: ‘EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER’ undeclared (first use in this
function)
error: ‘EXT2_FT_DIR’ undeclared (first use in this function)
error: ‘EXT2_GOOD_OLD_FIRST_INO’ undeclared (first use in this function)
error: ‘EXT2_MAX_BLOCK_SIZE’ undeclared (first use in this function)
error: ‘EXT2_MIN_BLOCK_LOG_SIZE’ undeclared (first use in this function)
error: ‘EXT2_MIN_BLOCK_SIZE’ undeclared (first use in this function)
error: ‘EXT2_NDIR_BLOCKS’ undeclared (first use in this function)
error: ‘EXT2_OS_LINUX’ undeclared (first use in this function)
error: ‘EXT2_ROOT_INO’ undeclared (first use in this function)
error: ‘EXT2_SUPER_MAGIC’ undeclared (first use in this function)
error: invalid application of ‘sizeof’ to incomplete type ‘struct ext2_inode’
error: invalid use of undefined type ‘struct ext2_group_desc’

struct ext2_inode is complex, I'd prefer to use system header than maintaining
my own copy....

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986


Theodore Tso <tytso@mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
         Resolution|                            |CODE_FIX




--- Comment #14 from Theodore Tso <tytso@mit.edu>  2012-04-12 15:04:23 ---
Use the one which is shipped with e2fsprogs.   Better yet, use libext2fs to
access the file system, and busybox will be future-proofed against future
changes / improvements in the ext 2/3/4 family of file systems.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.

[Bug 42986] ext2_fs.h requires undefined type umode_t

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=42986


Artem Bityutskiy <dedekind1@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dedekind1@gmail.com




--- Comment #15 from Artem Bityutskiy <dedekind1@gmail.com>  2012-04-13 06:48:52 ---
(In reply to comment #13)
> For the record: busybox's mkfs_ext2.c uses linux/ext2_fs.h.
> If its include is commented out, the following errors pop out:

The idea is that file-system on-disk layout is stable, so it should be easy for
you to have a copy, and you do not need to ever update it unless you add
support of a new feature in busybox which requeres an up-to-date header file,
and this should be a rare event.

>From the FS'es POW - the on-disk layout is not kernel-userspace API so kernel
should not export it to user-space. This will also make things easier for the
kernel people - we can freely clean-up/rename/shuffle things in the header
without being afraid of breaking busybox or any other tool.

-- 
Configure bugmail: https://bugzilla.kernel.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching the assignee of the bug.