Skip to contentSkip to conversation

VFS stuff, part 1

GIT PULL Al Viro 5 participants 12 messages Dec 16, 2009 – Dec 17, 2009

Subsystems (2)
lore
Share & tools
lore

Al Viro

author
#0 lore raw

This ought to get rid of the bulk of namei.c patches affecting the areas where my and Nick's series step on the same code. There's also some struct file work, plus a bunch of fixes all over the place (in particular, Eric's series dealing with IMA horrors, partially modified to merge), but most of that stuff is fighting the call graph in fs/namei.c and massaging things into more or less sane shape there.

Please, pull from git.kernel.org//pub/scm/linux/kernel/git/viro/vfs-2.6.git/

Shortlog:
Al Viro (25):
sanitize do_pipe_flags() callers in arch
switch shmem_file_setup() to alloc_file()
switched inotify_init1() to alloc_file()
unexport get_empty_filp()
reorder alloc_fd/attach_fd in socketpair()
merge sock_alloc_fd/sock_attach_fd into a new helper
9p connect fixes
switch sock_alloc_file() to alloc_file()
get rid of init_file()
switch nilfs2 to deactivate_locked_super()
switch alloc_file() to passing struct path
switch create_read_pipe() to alloc_file()
lift path_put(path) to callers of __do_follow_link() kill __link_path_walk()/link_path_walk() distinction
fix the crap in dst/dcore
switch cachefiles to kern_path()
fix mismerge with Trond's stuff (create_mnt_ns() export is gone now) Take hash recalculation into do_lookup()
Kill path_lookup_open()
Kill cached_lookup() and real_lookup()
Sanitize exec_permission_lite()
O_TRUNC open shouldn't fail after file truncation Untangling ima mess, part 1: alloc_file()
Untangling ima mess, part 2: deal with counters Untangling ima mess, part 3: kill dead code in ima

Christoph Hellwig (5):
sanitize xattr handler prototypes
make generic_acl slightly more generic
cleanup blockdev_direct_IO locking
ocfs: stop using do_sync_mapping_range
direct I/O fallback sync simplification

Eric Paris (5):
fs: move get_empty_filp() deffinition to internal.h
ima: valid return code from ima_inode_alloc
ima: only insert at inode creation time
IMA: clean up the IMA counts updating code
ima: call ima_inode_free ima_inode_free

H Hartley Sweeten (1):
libfs: move EXPORT_SYMBOL for d_alloc_name

Jeff Layton (1):
vfs: force reval of target when following LAST_BIND symlinks (try #7)

Mimi Zohar (1):
ima: limit imbalance msg

Diffstat:

 arch/ia64/kernel/perfmon.c            |   15 +-
 arch/parisc/hpux/sys_hpux.c           |    7 +-
 arch/x86/include/asm/sys_ia32.h       |    1 -
 arch/xtensa/include/asm/syscall.h     |    1 -
 arch/xtensa/include/asm/unistd.h      |    2 +-
 arch/xtensa/kernel/syscall.c          |   18 --
 drivers/infiniband/core/uverbs_main.c |    9 +-
 drivers/staging/dst/dcore.c           |   46 +---
 fs/anon_inodes.c                      |   18 +-
 fs/btrfs/acl.c                        |   47 +---
 fs/cachefiles/bind.c                  |   11 +-
 fs/cachefiles/rdwr.c                  |    2 -
 fs/dcache.c                           |    1 +
 fs/direct-io.c                        |  129 ++++------
 fs/ecryptfs/main.c                    |    7 +-
 fs/ext2/acl.c                         |   79 ++----
 fs/ext2/xattr.c                       |   11 +-
 fs/ext2/xattr_security.c              |   16 +-
 fs/ext2/xattr_trusted.c               |   16 +-
 fs/ext2/xattr_user.c                  |   25 +-
 fs/ext3/acl.c                         |   74 ++----
 fs/ext3/xattr.c                       |   31 ++-
 fs/ext3/xattr_security.c              |   20 +-
 fs/ext3/xattr_trusted.c               |   18 +-
 fs/ext3/xattr_user.c                  |   25 +-
 fs/ext4/acl.c                         |   74 ++----
 fs/ext4/xattr.c                       |   31 ++-
 fs/ext4/xattr_security.c              |   20 +-
 fs/ext4/xattr_trusted.c               |   20 +-
 fs/ext4/xattr_user.c                  |   25 +-
 fs/file_table.c                       |   49 +---
 fs/generic_acl.c                      |  158 +++++++-----
 fs/gfs2/acl.c                         |   16 +-
 fs/gfs2/inode.c                       |    3 +-
 fs/gfs2/xattr.c                       |   69 ++---
 fs/gfs2/xattr.h                       |    7 +-
 fs/hugetlbfs/inode.c                  |   17 +-
 fs/internal.h                         |    1 +
 fs/jffs2/acl.c                        |   65 ++---
 fs/jffs2/security.c                   |   18 +-
 fs/jffs2/xattr.c                      |    6 +-
 fs/jffs2/xattr_trusted.c              |   18 +-
 fs/jffs2/xattr_user.c                 |   18 +-
 fs/libfs.c                            |    1 -
 fs/namei.c                            |  466 ++++++++++++++++----------------
 fs/namespace.c                        |    3 +-
 fs/nfs/super.c                        |    8 -
 fs/nfsd/vfs.c                         |    5 +-
 fs/nilfs2/super.c                     |    3 +-
 fs/notify/inotify/inotify_user.c      |   29 +-
 fs/ocfs2/acl.c                        |   87 ++-----
 fs/ocfs2/alloc.c                      |    4 +-
 fs/ocfs2/aops.c                       |   34 +--
 fs/ocfs2/xattr.c                      |   72 +++---
 fs/open.c                             |    4 +
 fs/pipe.c                             |   29 +--
 fs/reiserfs/xattr.c                   |   36 ++--
 fs/reiserfs/xattr_acl.c               |   69 ++----
 fs/reiserfs/xattr_security.c          |   21 +-
 fs/reiserfs/xattr_trusted.c           |   21 +-
 fs/reiserfs/xattr_user.c              |   21 +-
 fs/xattr.c                            |   28 +-
 fs/xfs/linux-2.6/xfs_acl.c            |   57 ++---
 fs/xfs/linux-2.6/xfs_aops.c           |   20 +-
 fs/xfs/linux-2.6/xfs_xattr.c          |   71 +----
 fs/xfs/xfs_acl.h                      |    3 +-
 include/linux/file.h                  |    8 +-
 include/linux/fs.h                    |   23 +-
 include/linux/generic_acl.h           |   41 +---
 include/linux/ima.h                   |   12 +-
 include/linux/mnt_namespace.h         |    1 -
 include/linux/shmem_fs.h              |   16 --
 include/linux/xattr.h                 |   13 +-
 ipc/mqueue.c                          |    2 -
 ipc/shm.c                             |   12 +-
 mm/Makefile                           |    1 -
 mm/filemap.c                          |   15 +-
 mm/shmem.c                            |   71 +++---
 mm/shmem_acl.c                        |  171 ------------
 net/9p/trans_fd.c                     |  112 ++++-----
 net/socket.c                          |  108 +++-----
 security/integrity/ima/ima.h          |    3 -
 security/integrity/ima/ima_iint.c     |   79 +-----
 security/integrity/ima/ima_main.c     |  184 ++++++++------
 84 files changed, 1243 insertions(+), 1965 deletions(-)
author
#1 lore raw
Quoted 10 lines · Al Viro

This ought to get rid of the bulk of namei.c patches affecting the areas where my and Nick's series step on the same code. There's also some struct file work, plus a bunch of fixes all over the place (in particular, Eric's series dealing with IMA horrors, partially modified to merge), but most of that stuff is fighting the call graph in fs/namei.c and massaging things into more or less sane shape there.

Please, pull from git.kernel.org//pub/scm/linux/kernel/git/viro/vfs-2.6.git/

gyah...

git.kernel.org//pub/scm/linux/kernel/git/viro/vfs-2.6.git/ master

#2 lore raw
In reply to Al Viro

2009/12/16 Al Viro <[email protected]>:

Quoted 10 lines

This ought to get rid of the bulk of namei.c patches affecting the areas where my and Nick's series step on the same code.  There's also some struct file work, plus a bunch of fixes all over the place (in particular, Eric's series dealing with IMA horrors, partially modified to merge), but most of that stuff is fighting the call graph in fs/namei.c and massaging things into more or less sane shape there.

Please, pull from git.kernel.org//pub/scm/linux/kernel/git/viro/vfs-2.6.git/

Hi

Here is a regression between
5ac4d63 (Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block)
and
bac5e54 (Merge branch 'master' of
git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6)

[   17.462212] REISERFS (device sda1): found reiserfs format "3.6"
with standard journal
[   17.470199] REISERFS (device sda1): using ordered data mode
[   17.480645] REISERFS (device sda1): journal params: device sda1,
size 8192, journal first block 18, max trans len 1024, max batch 900,
max commit age 30, max trans age 30
[   17.497079] REISERFS (device sda1): checking transaction log (sda1)
[   17.503457] REISERFS debug (device sda1): journal-1153: found in
header: first_unflushed_offset 6839, last_flushed_trans_id 2074215
[   17.521242] REISERFS debug (device sda1): journal-1206: Starting
replay from offset 8908689884846775, trans_id 4161904640
[   17.532296] REISERFS debug (device sda1): journal-1299: Setting
newest_mount_id to 315
[   17.573874] REISERFS (device sda1): Using r5 hash to sort names
[   17.603715] VFS: Mounted root (reiserfs filesystem) readonly on device 8:1.
[   17.610972] Freeing unused kernel memory: 280k freed
[   17.790027] ohci_hcd 0000:00:02.0: auto-stop root hub
INIT: version 2.87 booting
[   18.290029] ohci_hcd 0000:00:02.1: auto-stop root hub
[   18.595537] kbd_mode used greatest stack depth: 6028 bytes left
Loading /lib/rc/console/keymap
[   18.648845] loadkeys used greatest stack depth: 5764 bytes left
[   18.655619] init-early.sh used greatest stack depth: 5276 bytes left

OpenRC 0.5.3 is starting up Gentoo Linux (i686)

* Mounting /proc...

[   18.819236] mount used greatest stack depth: 5260 bytes left
 [ ok ]

mkdir /lib/rc/init.d/starting': Operation not supported mkdir /lib/rc/init.d/started': Operation not supported mkdir /lib/rc/init.d/stopping': Operation not supported mkdir /lib/rc/init.d/inactive': Operation not supported mkdir /lib/rc/init.d/wasinactive': Operation not supported mkdir /lib/rc/init.d/failed': Operation not supported mkdir /lib/rc/init.d/hotplugged': Operation not supported mkdir /lib/rc/init.d/daemons': Operation not supported mkdir /lib/rc/init.d/options': Operation not supported mkdir /lib/rc/init.d/exclusive': Operation not supported mkdir /lib/rc/init.d/scheduled': Operation not supported mkdir /lib/rc/init.d/tmp': Operation not supported
* Caching service dependencies...

[   20.203989] gendepends.sh used greatest stack depth: 5176 bytes left
 [ ok ]
* rc: failed to cr* rc: failed to crINIT: Entering runlevel: 3
* rc: failed to create stopping dir `/lib/rc/init.d/rc.stopping':
Operation not supported
[   21.075265] device: 'vcs6': device_add
[   21.078101] device: 'vcsa6': device_add
[   21.086896] device: 'vcs4': device_add
[   21.087221] device: 'vcsa4': device_add
[   21.095348] device: 'vcs2': device_add
[   21.095696] device: 'vcsa2': device_add
[   21.103825] device: 'vcs3': device_add
[   21.104151] device: 'vcsa3': device_add
[   21.112285] device: 'vcs5': device_add
[   21.112604] device: 'vcsa5': device_add

INIT: Id "s0" respawni
INIT: Id "s0" respawni

Is it obvious for you or do I need to bisect ?

#3 lore raw
Quoted 2 lines · Alexander Beregalov

Is it obvious for you or do I need to bisect ?

It might be obvious to Al, but it would be good to bisect it anyway, just in case. Since you now have a fairly tight range of commits, it shouldn't be _too_ bad.

[ My main desktop is running that bac5e54 kernel that you have as bad
right noe, so it's not something totally obviously broken. Maybe it needs btrfs or something else specific to trigger. ]

Linus

#4 lore raw
In reply to Alexander Beregalov
Quoted 1 line · Alexander Beregalov

mkdir `/lib/rc/init.d/starting': Operation not supported

Confirmed, mkdir fails for any tmpfs filesystem. I can still create files using touch and remove them again with rm.

Regards,
Tony V.

author
#5 lore raw
Quoted 5 lines · Tony Vroon
Quoted 1 line · Alexander Beregalov

mkdir `/lib/rc/init.d/starting': Operation not supported

Confirmed, mkdir fails for any tmpfs filesystem. I can still create files using touch and remove them again with rm.

Hrm... What about your CONFIG_TMPFS_POSIX_ACL? And see if the breakage starts at 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 (make generic_acl slightly more generic).

#6 lore raw
Quoted 1 line · Al Viro

Hrm... What about your CONFIG_TMPFS_POSIX_ACL?

That is off:
chainsaw@prometheus /cvs/linux-2.6 $ zgrep -i tmpfs /proc/config.gz # CONFIG_DEVTMPFS is not set
CONFIG_TMPFS=y
# CONFIG_TMPFS_POSIX_ACL is not set

Quoted 3 lines

And see if the breakage
starts at 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 (make generic_acl slightly more generic).

Do you want me to revert 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 locally or try linux-2.6 old enough not to have this?

Regards,
Tony V.

author
#7 lore raw
Quoted 9 lines · Tony Vroon
Quoted 1 line · Al Viro

Hrm... What about your CONFIG_TMPFS_POSIX_ACL?

That is off:
chainsaw@prometheus /cvs/linux-2.6 $ zgrep -i tmpfs /proc/config.gz # CONFIG_DEVTMPFS is not set
CONFIG_TMPFS=y
# CONFIG_TMPFS_POSIX_ACL is not set

Quoted 6 lines
Quoted 3 lines

And see if the breakage
starts at 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 (make generic_acl slightly more generic).

Do you want me to revert 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 locally or try linux-2.6 old enough not to have this?

The former. Or try that kernel and compare with grandparent commit (ef26ca97e83052790940cbc444b01b0d17a813c1)

#8 lore raw
Quoted 3 lines · Al Viro
Quoted 2 lines

Do you want me to revert 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 locally or try linux-2.6 old enough not to have this?

The former.

That resolves the issue completely. I am now able to boot.

Regards,
Tony V.

author
#9 lore raw
Quoted 6 lines · Tony Vroon
Quoted 3 lines · Al Viro
Quoted 2 lines

Do you want me to revert 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 locally or try linux-2.6 old enough not to have this?

The former.

That resolves the issue completely. I am now able to boot.

Oh, fsck... I see. Linus, the fix is added to the same branch (and I'll throw the fix for removed alloc_file export there in a few minutes). In the meanwhile, see if this fixes it

8f4cb3a4252a053d2c7299d7d251389bddf0bcd4

mm/shmem.c +2 -0
index f848506..eef4ebe 100644
@@ -1830,6 +1830,8 @@ shmem_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
 			iput(inode);
 			return error;
 		}
+#else
+		error = 0;
 #endif
 		if (dir->i_mode & S_ISGID) {
 			inode->i_gid = dir->i_gid;
author
#10 lore raw
Quoted 10 lines · Al Viro
Quoted 6 lines · Tony Vroon
Quoted 3 lines · Al Viro
Quoted 2 lines

Do you want me to revert 1c7c474c31aea6d5cb2fb35f31d9e9e91ae466b1 locally or try linux-2.6 old enough not to have this?

The former.

That resolves the issue completely. I am now able to boot.

Oh, fsck... I see. Linus, the fix is added to the same branch (and I'll throw the fix for removed alloc_file export there in a few minutes).

Nevermind, alloc_file export had been restored in mainline already. So just pull from the same place, same branch ( git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6.git/ master ), there's just that one commit on top of mainline.

What happened: in absence of CONFIG_TPMFS_POSIX_ACL, shmem_mknod() had code that boiled down to

	error = 0; /* shmem_acl_init(...), actually */
	if (error)
		cleanup

and Christoph's patch had put all that chunk under ifdef. Which was not an equivalent transformation, since we could arrive there with non-zero error. Old code used to set it to 0, new one left as-is.

#11 lore raw
In reply to Al Viro

mm/shmem.c

Quoted 3 lines · Al Viro

In the meanwhile, see if this fixes it
8f4cb3a4252a053d2c7299d7d251389bddf0bcd4

mm/shmem.c +0 -0

I can confirm that this fixes the problem. (When applied to a virgin linux-2.6 checkout, without reverting anything)

Thank you :)

mm/shmem.c shmem_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev) · line 1837

Quoted 12 lines

index f848506..eef4ebe 100644

mm/shmem.c +2 -0
@@ -1830,6 +1830,8 @@ shmem_mknod(struct inode *dir, struct dentry *dentry, int mode, dev_t dev)
 			iput(inode);
 			return error;
 		}
+#else
+		error = 0;
 #endif
 		if (dir->i_mode & S_ISGID) {
 			inode->i_gid = dir->i_gid;

Regards,
Tony V.