platform/kernel/linux-3.10.git
9 years agosysfs, kernfs: move dir core code to fs/kernfs/dir.c
Tejun Heo [Thu, 28 Nov 2013 19:54:33 +0000 (14:54 -0500)]
sysfs, kernfs: move dir core code to fs/kernfs/dir.c

Move core dir code to fs/kernfs/dir.c.  fs/sysfs/dir.c now only
contains sysfs_warn_dup() and sysfs wrappers around kernfs interfaces.
The respective declarations in fs/sysfs/sysfs.h are moved to
fs/kernfs/kernfs-internal.h.

This is pure relocation.

v2: sysfs_symlink_target_lock was mistakenly relocated to kernfs.  It
    should remain with sysfs.  Fixed.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: move inode code to fs/kernfs/inode.c
Tejun Heo [Thu, 28 Nov 2013 19:54:32 +0000 (14:54 -0500)]
sysfs, kernfs: move inode code to fs/kernfs/inode.c

There's nothing sysfs-specific in fs/sysfs/inode.c.  Move everything
in it to fs/kernfs/inode.c.  The respective declarations in
fs/sysfs/sysfs.h are moved to fs/kernfs/kernfs-internal.h.

This is pure relocation.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: move internal decls to fs/kernfs/kernfs-internal.h
Tejun Heo [Thu, 28 Nov 2013 19:54:31 +0000 (14:54 -0500)]
sysfs, kernfs: move internal decls to fs/kernfs/kernfs-internal.h

Move data structure, constant and basic accessor declarations from
fs/sysfs/sysfs.h to fs/kernfs/kernfs-internal.h.  The two files
currently include each other.  Once kernfs / sysfs separation is
complete, the cross inclusions will be removed.  Inclusion protectors
are added to fs/sysfs/sysfs.h to allow cross-inclusion.

This patch doesn't introduce any functional changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs[_find_and]_get() and kernfs_put()
Tejun Heo [Thu, 28 Nov 2013 19:54:30 +0000 (14:54 -0500)]
sysfs, kernfs: introduce kernfs[_find_and]_get() and kernfs_put()

Introduce kernfs interface for finding, getting and putting
sysfs_dirents.

* sysfs_find_dirent() is renamed to kernfs_find_ns() and lockdep
  assertion for sysfs_mutex is added.

* sysfs_get_dirent_ns() is renamed to kernfs_find_and_get().

* Macro inline dancing around __sysfs_get/put() are removed and
  kernfs_get/put() are made proper functions implemented in
  fs/sysfs/dir.c.

While the conversions are mostly equivalent, there's one difference -
kernfs_get() doesn't return the input param as its return value.  This
change is intentional.  While passing through the input increases
writability in some areas, it is unnecessary and has been shown to
cause confusion regarding how the last ref is handled.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: revamp sysfs_dirent active_ref lockdep annotation
Tejun Heo [Thu, 28 Nov 2013 19:54:29 +0000 (14:54 -0500)]
sysfs, kernfs: revamp sysfs_dirent active_ref lockdep annotation

Currently, sysfs_dirent active_ref lockdep annotation uses
attribute->[s]key as the lockdep key, which forces
kernfs_create_file_ns() to assume that sysfs_dirent->priv is pointing
to a struct attribute which may not be true for non-sysfs users.  This
patch restructures the lockdep annotation such that

* kernfs_ops contains lockdep_key which is used by default for files
  created kernfs_create_file_ns().

* kernfs_create_file_ns_key() is introduced which takes an extra @key
  argument.  The created file will use the specified key for
  active_ref lockdep annotation.  If NULL is specified, lockdep for
  the file is disabled.

* sysfs_add_file_mode_ns() is updated to use
  kernfs_create_file_ns_key() with the appropriate key from the
  attribute or NULL if ignore_lockdep is set.

This makes the lockdep annotation properly contained in kernfs while
allowing sysfs to cleanly keep its current behavior.  This patch
doesn't introduce any behavior differences.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: reorganize SYSFS_* constants
Tejun Heo [Thu, 28 Nov 2013 19:54:28 +0000 (14:54 -0500)]
sysfs, kernfs: reorganize SYSFS_* constants

We want to add one more SYSFS_FLAG_* but we can't use the next higher
bit, 0x10000, as the flag field is 16bits wide.  The flags are
currently arranged weirdly - 8 bits are set aside for the type flags
when there are only three three used, the first flag starts at 0x1000
instead of 0x0100 and flag literals have 5 digits (20 bits) when only
4 digits can be used.

Rearrange them so that type bits are only the lowest four, flags start
at 0x0010 and similar flags are grouped.

This patch doesn't cause any behavior difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_notify()
Tejun Heo [Thu, 28 Nov 2013 19:54:27 +0000 (14:54 -0500)]
sysfs, kernfs: introduce kernfs_notify()

Introduce kernfs interface to wake up poll(2) which takes and returns
sysfs_dirents.

sysfs_notify_dirent() is renamed to kernfs_notify() and sysfs_notify()
is updated so that it doesn't directly grab sysfs_mutex but acquires
the target sysfs_dirents using sysfs_get_dirent().
sysfs_notify_dirent() is reimplemented as a dumb inline wrapper around
kernfs_notify().

This patch doesn't introduce any behavior changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: add kernfs_ops->seq_{start|next|stop}()
Tejun Heo [Thu, 28 Nov 2013 19:54:26 +0000 (14:54 -0500)]
sysfs, kernfs: add kernfs_ops->seq_{start|next|stop}()

kernfs_ops currently only supports single_open() behavior which is
pretty restrictive.  Add optional callbacks ->seq_{start|next|stop}()
which, when implemented, are invoked for seq_file traversal.  This
allows full seq_file functionality for kernfs users.  This currently
doesn't have any user and doesn't change any behavior.

v2: Refreshed on top of the updated "sysfs, kernfs: prepare read path
    for kernfs".

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: remove sysfs_add_one()
Tejun Heo [Thu, 28 Nov 2013 19:54:25 +0000 (14:54 -0500)]
sysfs, kernfs: remove sysfs_add_one()

sysfs_add_one() is a wrapper around __sysfs_add_one() which prints out
duplicate name warning if __sysfs_add_one() fails with -EEXIST.  The
previous kernfs conversions moved all dup warnings to sysfs interface
functions and sysfs_add_one() doesn't have any user left.

Remove sysfs_add_one() and update __sysfs_add_one() to take its name.

This patch doesn't make any functional changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_create_file[_ns]()
Tejun Heo [Thu, 28 Nov 2013 19:54:24 +0000 (14:54 -0500)]
sysfs, kernfs: introduce kernfs_create_file[_ns]()

Introduce kernfs interface to create a file which takes and returns
sysfs_dirents.

The actual file creation part is separated out from
sysfs_add_file_mode_ns() into kernfs_create_file_ns().  The former now
only decides the kernfs_ops to use and the file's size and invokes the
latter.

This patch doesn't introduce behavior changes.

v2: Dummy implementation for !CONFIG_SYSFS updated to return -ENOSYS.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: remove SYSFS_KOBJ_BIN_ATTR
Tejun Heo [Thu, 28 Nov 2013 19:54:23 +0000 (14:54 -0500)]
sysfs, kernfs: remove SYSFS_KOBJ_BIN_ATTR

After kernfs_ops and sysfs_dirent->s_attr.size addition, the
distinction between SYSFS_KOBJ_BIN_ATTR and SYSFS_KOBJ_ATTR is only
necessary while creating files to decide which kernfs_ops to use.
Afterwards, they behave exactly the same.

This patch removes SYSFS_KOBJ_BIN_ATTR along with sysfs_is_bin().
sysfs_add_file[_mode_ns]() are updated to take bool @is_bin instead of
@type.

This patch doesn't introduce any behavior changes.  This completely
isolates the distinction between the two sysfs file types in the sysfs
layer proper.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: add sysfs_dirent->s_attr.size
Tejun Heo [Thu, 28 Nov 2013 19:54:22 +0000 (14:54 -0500)]
sysfs, kernfs: add sysfs_dirent->s_attr.size

sysfs sets the size of regular files unconditionally at PAGE_SIZE and
takes the size of bin files from bin_attribute.  The latter is a
pretty bad interface which forces bin_attribute users to create a
separate copy of bin_attribute for each instance of the file -
e.g. pci resource files.

Add sysfs_dirent->s_attr.size so that the size can be specified
separately.  This unifies inode init paths of ATTR and BIN_ATTR
identical and allows for generic size handling for kernfs.

Unfortunately, this grows the size of sysfs_dirent by sizeof(loff_t).

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_ops
Tejun Heo [Thu, 28 Nov 2013 19:54:21 +0000 (14:54 -0500)]
sysfs, kernfs: introduce kernfs_ops

We're in the process of separating out core sysfs functionality into
kernfs which will deal with sysfs_dirents directly.  This patch
introduces kernfs_ops which hosts methods kernfs users implement and
updates fs/sysfs/file.c such that sysfs_kf_*() functions populate
kernfs_ops and kernfs_file_*() functions call the matching entries
from kernfs_ops.

kernfs_ops contains the following groups of methods.

* seq_show() - for kernfs files which use seq_file for reads.

* read() - for direct read implementations.  Used iff seq_show() is
  not implemented.

* write() - for writes.

* mmap() - for mmaps.

Notes:

* sysfs_elem_attr->ops is added so that kernfs_ops can be accessed
  from sysfs_dirent.  kernfs_ops() helper is added to verify locking
  and access the field.

* SYSFS_FLAG_HAS_(SEQ_SHOW|MMAP) added.  sd->s_attr->ops is accessible
  only while holding active_ref and there are cases where we want to
  take different actions depending on which ops are implemented.
  These two flags cache whether the two ops are implemented for those.

* kernfs_file_*() no longer test sysfs type but chooses different
  behaviors depending on which methods in kernfs_ops are implemented.
  The conversions are trivial except for the open path.  As
  kernfs_file_open() now decides whether to allow read/write accesses
  depending on the kernfs_ops implemented, the presence of methods in
  kobjs and attribute_bin should be propagated to kernfs_ops.
  sysfs_add_file_mode_ns() is updated so that it propagates presence /
  absence of the callbacks through _empty, _ro, _wo, _rw kernfs_ops.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: move sysfs_open_file to include/linux/kernfs.h
Tejun Heo [Thu, 28 Nov 2013 19:54:20 +0000 (14:54 -0500)]
sysfs, kernfs: move sysfs_open_file to include/linux/kernfs.h

sysfs_open_file will be used as the primary handle for kernfs methods.
Move its definition from fs/sysfs/file.c to include/linux/kernfs.h and
mark the public and private fields.

This is pure relocation.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: prepare open, release, poll paths for kernfs
Tejun Heo [Thu, 28 Nov 2013 19:54:19 +0000 (14:54 -0500)]
sysfs, kernfs: prepare open, release, poll paths for kernfs

We're in the process of separating out core sysfs functionality into
kernfs which will deal with sysfs_dirents directly.  This patch
prepares the rest - open, release and poll.  There isn't much to do.
Just renaming is enough.  As sysfs_file_operations and
sysfs_bin_operations are identical now, use the same file_operations
for both - kernfs_file_operations.

This patch doesn't introduce any behavior changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: prepare mmap path for kernfs
Tejun Heo [Thu, 28 Nov 2013 19:54:18 +0000 (14:54 -0500)]
sysfs, kernfs: prepare mmap path for kernfs

We're in the process of separating out core sysfs functionality into
kernfs which will deal with sysfs_dirents directly.  This patch
rearranges mmap path so that the kernfs and sysfs parts are separate.

sysfs_kf_bin_mmap() which handles the interaction with bin_attribute
mmap method is factored out of sysfs_bin_mmap(), which is renamed to
kernfs_file_mmap().  All vma ops are renamed accordingly.

sysfs_bin_mmap() is updated such that it can be used for both file
types.  This will eventually allow using the same file_operations for
both file types, which is necessary to separate out kernfs.

This patch doesn't introduce any behavior changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: prepare write path for kernfs
Tejun Heo [Thu, 28 Nov 2013 19:54:17 +0000 (14:54 -0500)]
sysfs, kernfs: prepare write path for kernfs

We're in the process of separating out core sysfs functionality into
kernfs which will deal with sysfs_dirents directly.  This patch
rearranges write path so that the kernfs and sysfs parts are separate.

kernfs_file_write() handles all boilerplate work including buffer
management and locking and invokes sysfs_kf_write() or
sysfs_kf_bin_write() depending on the file type which deals with the
interaction with kobj store or bin_attribute write method.

While this patch changes the order of some operations, it shouldn't
change any visible behavior.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: prepare read path for kernfs
Tejun Heo [Thu, 28 Nov 2013 19:54:16 +0000 (14:54 -0500)]
sysfs, kernfs: prepare read path for kernfs

We're in the process of separating out core sysfs functionality into
kernfs which will deal with sysfs_dirents directly.  This patch
rearranges read path so that the kernfs and sysfs parts are separate.

* Regular file read path is refactored such that
  kernfs_seq_start/next/stop/show() handle all the boilerplate work
  including locking and updating event count for poll, while
  sysfs_kf_seq_show() deals with interaction with kobj show method.

* Bin file read path is refactored such that kernfs_file_direct_read()
  handles all the boilerplate work including buffer management and
  locking, while sysfs_kf_bin_read() deals with interaction with
  bin_attribute read method.

kernfs_file_read() is added.  It invokes either the seq_file or direct
read path depending on the file type.  This will eventually allow
using the same file_operations for both file types, which is necessary
to separate out kernfs.

While this patch changes the order of some operations, it shouldn't
change any visible behavior.

v2: Dropped unnecessary zeroing of @count from sysfs_kf_seq_show().
    Add comments explaining single_open() behavior.  Both suggested by
    Pavel.

v3: seq_stop() is called even after seq_start() failed.
    kernfs_seq_start() updated so that it doesn't unlock
    sysfs_open_file->mutex on failure so that kernfs_seq_stop()
    doesn't try to unlock an already unlocked mutex.  Reported by
    Fengguang.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_create_dir[_ns]()
Tejun Heo [Thu, 28 Nov 2013 19:54:15 +0000 (14:54 -0500)]
sysfs, kernfs: introduce kernfs_create_dir[_ns]()

Introduce kernfs interface to manipulate a directory which takes and
returns sysfs_dirents.

create_dir() is renamed to kernfs_create_dir_ns() and its argumantes
and return value are updated.  create_dir() usages are replaced with
kernfs_create_dir_ns() and sysfs_create_subdir() usages are replaced
with kernfs_create_dir().  Dup warnings are handled explicitly by
sysfs users of the kernfs interface.

sysfs_enable_ns() is renamed to kernfs_enable_ns().

This patch doesn't introduce any behavior changes.

v2: Dummy implementation for !CONFIG_SYSFS updated to return -ENOSYS.

v3: kernfs_enable_ns() added.

v4: Refreshed on top of "sysfs: drop kobj_ns_type handling, take #2"
    so that this patch removes sysfs_enable_ns().

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: replace sysfs_dirent->s_dir.kobj and ->s_attr.[bin_]attr with ->priv
Tejun Heo [Thu, 28 Nov 2013 19:54:14 +0000 (14:54 -0500)]
sysfs, kernfs: replace sysfs_dirent->s_dir.kobj and ->s_attr.[bin_]attr with ->priv

A directory sysfs_dirent points to the associated kobj.  A regular or
bin file points to the associated [bin_]attribute.  This patch
replaces sysfs_dirent->s_dir.kobj and ->s_attr.[bin_]attr with void *
->priv.

This is to prepare for kernfs interface so that sysfs can specify the
private data in the same way for directories and files.  This lower
debuggability but not by much - the whole thing was overlaid in a
union anyway.  If debuggability becomes an issue, we can later add
->priv accessors which explicitly check for the sysfs_dirent type and
performs casting.

This patch doesn't introduce any behavior difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_setattr()
Tejun Heo [Sat, 23 Nov 2013 22:21:52 +0000 (17:21 -0500)]
sysfs, kernfs: introduce kernfs_setattr()

Introduce kernfs setattr interface - kernfs_setattr().

sysfs_sd_setattr() is renamed to __kernfs_setattr() and
kernfs_setattr() is a simple wrapper around it with sysfs_mutex
locking.  sysfs_chmod_file() is updated to get an explicit ref on
kobj->sd and then invoke kernfs_setattr() so that it doesn't have to
use internal interface.

This patch doesn't introduce any behavior differences.

v2: Dummy implementation for !CONFIG_SYSFS updated to return -ENOSYS.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_rename[_ns]()
Tejun Heo [Sat, 23 Nov 2013 22:21:51 +0000 (17:21 -0500)]
sysfs, kernfs: introduce kernfs_rename[_ns]()

Introduce kernfs rename interface, krenfs_rename[_ns]().

This is just rename of sysfs_rename().  No functional changes.
Function comment is added to kernfs_rename_ns() and @new_parent_sd is
renamed to @new_parent for consistency with other kernfs interfaces.

v2: Dummy implementation for !CONFIG_SYSFS updated to return -ENOSYS.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_create_link()
Tejun Heo [Sat, 23 Nov 2013 22:21:50 +0000 (17:21 -0500)]
sysfs, kernfs: introduce kernfs_create_link()

Separate out kernfs symlink interface - kernfs_create_link() - which
takes and returns sysfs_dirents, from sysfs_do_create_link_sd().
sysfs_do_create_link_sd() now just determines the parent and target
sysfs_dirents and invokes the new interface and handles dup warning.

This patch doesn't introduce behavior changes.

v2: Dummy implementation for !CONFIG_SYSFS updated to return -ENOSYS.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: introduce kernfs_remove[_by_name[_ns]]()
Tejun Heo [Sat, 23 Nov 2013 22:21:49 +0000 (17:21 -0500)]
sysfs, kernfs: introduce kernfs_remove[_by_name[_ns]]()

Introduce kernfs removal interfaces - kernfs_remove() and
kernfs_remove_by_name[_ns]().

These are just renames of sysfs_remove() and sysfs_hash_and_remove().
No functional changes.

v2: Dummy kernfs_remove_by_name_ns() for !CONFIG_SYSFS updated to
    return -ENOSYS instead of 0.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs, kernfs: add skeletons for kernfs
Tejun Heo [Sun, 24 Nov 2013 14:54:58 +0000 (09:54 -0500)]
sysfs, kernfs: add skeletons for kernfs

Core sysfs implementation will be separated into kernfs so that it can
be used by other non-kobject users.

This patch creates fs/kernfs/ directory and makes boilerplate changes.
kernfs interface will be directly based on sysfs_dirent and its
forward declaration is moved to include/linux/kernfs.h which is
included from include/linux/sysfs.h.  sysfs core implementation will
be gradually separated out and moved to kernfs.

This patch doesn't introduce any functional changes.

v2: mount.c added.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: linux-fsdevel@vger.kernel.org
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: make __sysfs_add_one() fail if the parent isn't a directory
Tejun Heo [Sat, 23 Nov 2013 22:21:47 +0000 (17:21 -0500)]
sysfs: make __sysfs_add_one() fail if the parent isn't a directory

Currently the kobject based interface guarantees that a parent
sysfs_dirent is always a directory; however, the planned kernfs
interface will be directly based on sysfs_dirents and the caller may
specify non-directory node as the parent.  Add an explicit check in
__sysfs_add_one() so that such attempts fail with -EINVAL.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: drop kobj_ns_type handling, take #2
Tejun Heo [Sat, 23 Nov 2013 23:01:46 +0000 (18:01 -0500)]
sysfs: drop kobj_ns_type handling, take #2

The way namespace tags are implemented in sysfs is more complicated
than necessary.  As each tag is a pointer value and required to be
non-NULL under a namespace enabled parent, there's no need to record
separately what type each tag is.  If multiple namespace types are
needed, which currently aren't, we can simply compare the tag to a set
of allowed tags in the superblock assuming that the tags, being
pointers, won't have the same value across multiple types.

This patch rips out kobj_ns_type handling from sysfs.  sysfs now has
an enable switch to turn on namespace under a node.  If enabled, all
children are required to have non-NULL namespace tags and filtered
against the super_block's tag.

kobject namespace determination is now performed in
lib/kobject.c::create_dir() making sysfs_read_ns_type() unnecessary.
The sanity checks are also moved.  create_dir() is restructured to
ease such addition.  This removes most kobject namespace knowledge
from sysfs proper which will enable proper separation and layering of
sysfs.

This is the second try.  The first one was cb26a311578e ("sysfs: drop
kobj_ns_type handling") which tried to automatically enable namespace
if there are children with non-NULL namespace tags; however, it was
broken for symlinks as they should inherit the target's tag iff
namespace is enabled in the parent.  This led to namespace filtering
enabled incorrectly for wireless net class devices through phy80211
symlinks and thus network configuration failure.  a1212d278c05
("Revert "sysfs: drop kobj_ns_type handling"") reverted the commit.

This shouldn't introduce any behavior changes, for real.

v2: Dummy implementation of sysfs_enable_ns() for !CONFIG_SYSFS was
    missing and caused build failure.  Reported by kbuild test robot.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Kay Sievers <kay@vrfy.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: use a separate locking class for open files depending on mmap
Tejun Heo [Sun, 17 Nov 2013 02:17:36 +0000 (11:17 +0900)]
sysfs: use a separate locking class for open files depending on mmap

The following two commits implemented mmap support in the regular file
path and merged bin file support into the regular path.

 73d9714627ad ("sysfs: copy bin mmap support from fs/sysfs/bin.c to fs/sysfs/file.c")
 3124eb1679b2 ("sysfs: merge regular and bin file handling")

After the merge, the following commands trigger a spurious lockdep
warning.  "test-mmap-read" simply mmaps the file and dumps the
content.

  $ cat /sys/block/sda/trace/act_mask
  $ test-mmap-read /sys/devices/pci0000\:00/0000\:00\:03.0/resource0 4096

  ======================================================
  [ INFO: possible circular locking dependency detected ]
  3.12.0-work+ #378 Not tainted
  -------------------------------------------------------
  test-mmap-read/567 is trying to acquire lock:
   (&of->mutex){+.+.+.}, at: [<ffffffff8120a8df>] sysfs_bin_mmap+0x4f/0x120

  but task is already holding lock:
   (&mm->mmap_sem){++++++}, at: [<ffffffff8114b399>] vm_mmap_pgoff+0x49/0xa0

  which lock already depends on the new lock.

  the existing dependency chain (in reverse order) is:

  -> #3 (&mm->mmap_sem){++++++}:
  ...
  -> #2 (sr_mutex){+.+.+.}:
  ...
  -> #1 (&bdev->bd_mutex){+.+.+.}:
  ...
  -> #0 (&of->mutex){+.+.+.}:
  ...

  other info that might help us debug this:

  Chain exists of:
   &of->mutex --> sr_mutex --> &mm->mmap_sem

   Possible unsafe locking scenario:

 CPU0                    CPU1
 ----                    ----
    lock(&mm->mmap_sem);
 lock(sr_mutex);
 lock(&mm->mmap_sem);
    lock(&of->mutex);

   *** DEADLOCK ***

  1 lock held by test-mmap-read/567:
   #0:  (&mm->mmap_sem){++++++}, at: [<ffffffff8114b399>] vm_mmap_pgoff+0x49/0xa0

  stack backtrace:
  CPU: 3 PID: 567 Comm: test-mmap-read Not tainted 3.12.0-work+ #378
  Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
   ffffffff81ed41a0 ffff880009441bc8 ffffffff81611ad2 ffffffff81eccb80
   ffff880009441c08 ffffffff8160f215 ffff880009441c60 ffff880009c75208
   0000000000000000 ffff880009c751e0 ffff880009c75208 ffff880009c74ac0
  Call Trace:
   [<ffffffff81611ad2>] dump_stack+0x4e/0x7a
   [<ffffffff8160f215>] print_circular_bug+0x2b0/0x2bf
   [<ffffffff8109ca0a>] __lock_acquire+0x1a3a/0x1e60
   [<ffffffff8109d6ba>] lock_acquire+0x9a/0x1d0
   [<ffffffff81615547>] mutex_lock_nested+0x67/0x3f0
   [<ffffffff8120a8df>] sysfs_bin_mmap+0x4f/0x120
   [<ffffffff8115d363>] mmap_region+0x3b3/0x5b0
   [<ffffffff8115d8ae>] do_mmap_pgoff+0x34e/0x3d0
   [<ffffffff8114b3ba>] vm_mmap_pgoff+0x6a/0xa0
   [<ffffffff8115be3e>] SyS_mmap_pgoff+0xbe/0x250
   [<ffffffff81008282>] SyS_mmap+0x22/0x30
   [<ffffffff8161a4d2>] system_call_fastpath+0x16/0x1b

This happens because one file nests sr_mutex, which nests mm->mmap_sem
under it, under of->mutex while mmap implementation naturally nests
of->mutex under mm->mmap_sem.  The warning is false positive as
of->mutex is per open-file and the two paths belong to two different
files.  This warning didn't trigger before regular and bin file
supports were merged because only bin file supported mmap and the
other side of locking happened only on regular files which used
equivalent but separate locking.

It'd be best if we give separate locking classes per file but we can't
easily do that.  Let's differentiate on ->mmap() for now.  Later we'll
add explicit file operations struct and can add per-ops lockdep key
there.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: use generic_file_llseek() for sysfs_file_operations
Tejun Heo [Fri, 1 Nov 2013 17:16:53 +0000 (13:16 -0400)]
sysfs: use generic_file_llseek() for sysfs_file_operations

13c589d5b0ac6 ("sysfs: use seq_file when reading regular files")
converted regular sysfs files to use seq_file.  The commit substituted
generic_file_llseek() with seq_lseek() for llseek implementation.

Before the change, all regular sysfs files were allowed to seek to any
position in [0, PAGE_SIZE] as the file size is always PAGE_SIZE and
generic_file_llseek() allows any seeking inside the range under file
size; however, seq_lseek()'s behavior is different.  It traverses the
output by repeatedly invoking ->show() until it reaches the target
offset or traversal indicates EOF.  As seq_files are fully dynamic and
may not end at all, it doesn't support seeking from the end
(SEEK_END).

Apparently, there are userland tools which uses SEEK_END to discover
the buffer size to use and the switch to seq_lseek() disturbs them as
SEEK_END fails with -EINVAL.

The only benefits of using seq_lseek() instead of
generic_file_llseek() are

* Early failure.  If traversing to certain file position should fail,
  seq_lseek() will report such failures on lseek(2) instead of the
  following read/write operations.

* EOF detection.  While SEEK_END is not supported, SEEK_SET/CUR +
  large offset can be used to detect eof - eof at the time of the seek
  anyway as the file size may change dynamically.

Both aren't necessary for sysfs or prospect kernfs users.  Revert to
genefic_file_llseek() and preserve the original behavior.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Link: https://lkml.kernel.org/r/20131031114358.GA5551@osiris
Tested-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: rename sysfs_assoc_lock and explain what it's about
Tejun Heo [Wed, 30 Oct 2013 14:28:36 +0000 (10:28 -0400)]
sysfs: rename sysfs_assoc_lock and explain what it's about

sysfs_assoc_lock is an odd piece of locking.  In general, whoever owns
a kobject is responsible for synchronizing sysfs operations and sysfs
proper assumes that, for example, removal won't race with any other
operation; however, this doesn't work for symlinking because an entity
performing symlink doesn't usually own the target kobject and thus has
no control over its removal.

sysfs_assoc_lock synchronizes symlink operations against kobj->sd
disassociation so that symlink code doesn't end up dereferencing
already freed sysfs_dirent by racing with removal of the target
kobject.

This is quite obscure and the generic name of the lock and lack of
comments make it difficult to understand its role.  Let's rename it to
sysfs_symlink_target_lock and add comments explaining what's going on.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: separate out dup filename warning into a separate function
Tejun Heo [Thu, 24 Oct 2013 15:49:11 +0000 (11:49 -0400)]
sysfs: separate out dup filename warning into a separate function

Separate out sysfs_warn_dup() out of sysfs_add_one().  This will help
separating out the core sysfs functionalities into kernfs so that it
can be used by non-sysfs users too.

This doesn't make any functional changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: move sysfs_hash_and_remove() to fs/sysfs/dir.c
Tejun Heo [Thu, 24 Oct 2013 15:49:10 +0000 (11:49 -0400)]
sysfs: move sysfs_hash_and_remove() to fs/sysfs/dir.c

Most removal related logic is implemented in fs/sysfs/dir.c.  Move
sysfs_hash_and_remove() to fs/sysfs/dir.c so that __sysfs_remove()
doesn't have to be public.

This is pure relocation.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: remove unused sysfs_get_dentry() prototype
Tejun Heo [Thu, 24 Oct 2013 15:49:09 +0000 (11:49 -0400)]
sysfs: remove unused sysfs_get_dentry() prototype

sysfs_get_dentry() has been gone for years now.  Remove the left-over
prototype.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: honor bin_attr.attr.ignore_lockdep
Tejun Heo [Thu, 24 Oct 2013 15:49:08 +0000 (11:49 -0400)]
sysfs: honor bin_attr.attr.ignore_lockdep

ignore_lockdep is currently honored only for regular files.  There's
no reason to ignore it for bin files.  Update sysfs_ignore_lockdep()
so that bin_attr.attr.ignore_lockdep works too.

While this doesn't have any in-kernel user, this unifies the behaviors
between regular and bin files and will help later changes.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: merge sysfs_elem_bin_attr into sysfs_elem_attr
Tejun Heo [Thu, 24 Oct 2013 15:49:07 +0000 (11:49 -0400)]
sysfs: merge sysfs_elem_bin_attr into sysfs_elem_attr

3124eb1679 ("sysfs: merge regular and bin file handling") folded bin
file handling into regular file handling.  Among other things, bin
file now shares the same open path including sysfs_open_dirent
association using sysfs_dirent->s_attr.open.  This is buggy because
->s_bin_attr lives in the same union and doesn't have the field.  This
bug doesn't trigger because sysfs_elem_bin_attr doesn't have an active
field at the conflicting position.  It does have a field "buffers" but
it isn't used anymore.

This patch collapses sysfs_elem_bin_attr into sysfs_elem_attr so that
the bin_attr is accessed through ->s_attr.bin_attr which lives with
->s_attr.attr in an anonymous union.  The code paths already assume
bin_attr contains attr as the first element, so this doesn't add any
more assumptions while making it explicit that the two types are
handled together.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: make sysfs_file_ops() follow ignore_lockdep flag
Tejun Heo [Mon, 14 Oct 2013 13:27:11 +0000 (09:27 -0400)]
sysfs: make sysfs_file_ops() follow ignore_lockdep flag

375b611e60 ("sysfs: remove sysfs_buffer->ops") introduced
sysfs_file_ops() which determines the associated file operation of a
given sysfs_dirent.  As file ops access should be protected by an
active reference, the new function includes a lockdep assertion on the
sysfs_dirent; unfortunately, I forgot to take attr->ignore_lockdep
flag into account and the lockdep assertion trips spuriously for files
which opt out from active reference lockdep checking.

# cat /sys/devices/pci0000:00/0000:00:01.2/usb1/authorized

 ------------[ cut here ]------------
 WARNING: CPU: 1 PID: 540 at /work/os/work/fs/sysfs/file.c:79 sysfs_file_ops+0x4e/0x60()
 Modules linked in:
 CPU: 1 PID: 540 Comm: cat Not tainted 3.11.0-work+ #3
 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
  0000000000000009 ffff880016205c08 ffffffff81ca0131 0000000000000000
  ffff880016205c40 ffffffff81096d0d ffff8800166cb898 ffff8800166f6f60
  ffffffff8125a220 ffff880011ab1ec0 ffff88000aff0c78 ffff880016205c50
 Call Trace:
  [<ffffffff81ca0131>] dump_stack+0x4e/0x82
  [<ffffffff81096d0d>] warn_slowpath_common+0x7d/0xa0
  [<ffffffff81096dea>] warn_slowpath_null+0x1a/0x20
  [<ffffffff8125994e>] sysfs_file_ops+0x4e/0x60
  [<ffffffff8125a274>] sysfs_open_file+0x54/0x300
  [<ffffffff811df612>] do_dentry_open.isra.17+0x182/0x280
  [<ffffffff811df820>] finish_open+0x30/0x40
  [<ffffffff811f0623>] do_last+0x503/0xd90
  [<ffffffff811f0f6b>] path_openat+0xbb/0x6d0
  [<ffffffff811f23ba>] do_filp_open+0x3a/0x90
  [<ffffffff811e09a9>] do_sys_open+0x129/0x220
  [<ffffffff811e0abe>] SyS_open+0x1e/0x20
  [<ffffffff81caf3c2>] system_call_fastpath+0x16/0x1b
 ---[ end trace aa48096b111dafdb ]---

Rename fs/sysfs/dir.c::ignore_lockdep() to sysfs_ignore_lockdep() and
move it to fs/sysfs/sysfs.h and make sysfs_file_ops() skip lockdep
assertion if sysfs_ignore_lockdep() is true.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: merge regular and bin file handling
Tejun Heo [Tue, 1 Oct 2013 21:42:09 +0000 (17:42 -0400)]
sysfs: merge regular and bin file handling

With the previous changes, sysfs regular file code is ready to handle
bin files too.  This patch makes bin files share the regular file
path.

* sysfs_create/remove_bin_file() are moved to fs/sysfs/file.c.

* sysfs_init_inode() is updated to use the new sysfs_bin_operations
  instead of bin_fops for bin files.

* fs/sysfs/bin.c and the related pieces are removed.

This patch shouldn't introduce any behavior difference to bin file
accesses.

Overall, this unification reduces the amount of duplicate logic, makes
behaviors more consistent and paves the road for building simpler and
more versatile interface which will allow other subsystems to make use
of sysfs for their pseudo filesystems.

v2: Stale fs/sysfs/bin.c reference dropped from
    Documentation/DocBook/filesystems.tmpl.  Reported by kbuild test
    robot.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kay Sievers <kay@vrfy.org>
Cc: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: prepare open path for unified regular / bin file handling
Tejun Heo [Tue, 1 Oct 2013 21:42:08 +0000 (17:42 -0400)]
sysfs: prepare open path for unified regular / bin file handling

sysfs bin file handling will be merged into the regular file support.
This patch prepares the open path.

This patch updates sysfs_open_file() such that it can handle both
regular and bin files.

This is a preparation and the new bin file path isn't used yet.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: copy bin mmap support from fs/sysfs/bin.c to fs/sysfs/file.c
Tejun Heo [Tue, 1 Oct 2013 21:42:07 +0000 (17:42 -0400)]
sysfs: copy bin mmap support from fs/sysfs/bin.c to fs/sysfs/file.c

sysfs bin file handling will be merged into the regular file support.
This patch copies mmap support from bin so that fs/sysfs/file.c can
handle mmapping bin files.

The code is copied mostly verbatim with the following updates.

* ->mmapped and ->vm_ops are added to sysfs_open_file and bin_buffer
  references are replaced with sysfs_open_file ones.

* Symbols are prefixed with sysfs_.

* sysfs_unmap_bin_file() grabs sysfs_open_dirent and traverses
  ->files.  Invocation of this function is added to
  sysfs_addrm_finish().

* sysfs_bin_mmap() is added to sysfs_bin_operations.

This is a preparation and the new mmap path isn't used yet.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: add sysfs_bin_read()
Tejun Heo [Tue, 1 Oct 2013 21:42:06 +0000 (17:42 -0400)]
sysfs: add sysfs_bin_read()

sysfs bin file handling will be merged into the regular file support.
This patch prepares the read path.

Copy fs/sysfs/bin.c::read() to fs/sysfs/file.c and make it use
sysfs_open_file instead of bin_buffer.  The function is identical copy
except for the use of sysfs_open_file.

The new function is added to sysfs_bin_operations.  This isn't used
yet but will eventually replace fs/sysfs/bin.c.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: prepare path write for unified regular / bin file handling
Tejun Heo [Tue, 1 Oct 2013 21:42:05 +0000 (17:42 -0400)]
sysfs: prepare path write for unified regular / bin file handling

sysfs bin file handling will be merged into the regular file support.
This patch prepares the write path.

bin file write is almost identical to regular file write except that
the write length is capped by the inode size and @off is passed to the
write method.  This patch adds bin file handling to sysfs_write_file()
so that it can handle both regular and bin files.

A new file_operations struct sysfs_bin_operations is added, which
currently only hosts sysfs_write_file() and generic_file_llseek().
This isn't used yet but will eventually replace fs/sysfs/bin.c.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: collapse fs/sysfs/bin.c::fill_read() into read()
Tejun Heo [Tue, 1 Oct 2013 21:42:04 +0000 (17:42 -0400)]
sysfs: collapse fs/sysfs/bin.c::fill_read() into read()

read() is simple enough and fill_read() being in a separate function
doesn't add anything.  Let's collapse it into read().  This will make
merging bin file handling with regular file.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: skip bin_buffer->buffer while reading
Tejun Heo [Tue, 1 Oct 2013 21:42:03 +0000 (17:42 -0400)]
sysfs: skip bin_buffer->buffer while reading

After b31ca3f5dfc ("sysfs: fix deadlock"), bin read() first writes
data to bb->buffer and bounces it to a transient kernel buffer which
is then copied out to userland.  The double bouncing doesn't add
anything.  Let's just use the transient buffer directly.

While at it, rename @temp to @buf for clarity.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: use seq_file when reading regular files
Tejun Heo [Tue, 1 Oct 2013 21:42:02 +0000 (17:42 -0400)]
sysfs: use seq_file when reading regular files

sysfs read path implements its own buffering scheme between userland
and kernel callbacks, which essentially is a degenerate duplicate of
seq_file.  This patch replaces the custom read buffering
implementation in sysfs with seq_file.

While the amount of code reduction is small, this reduces low level
hairiness and enables future development of a new versatile API based
on seq_file so that sysfs features can be shared with other
subsystems.

As write path was already converted to not use sysfs_open_file->page,
this patch makes ->page and ->count unused and removes them.

Userland behavior remains the same except for some extreme corner
cases - e.g. sysfs will now regenerate the content each time a file is
read after a non-contiguous seek whereas the original code would keep
using the same content.  While this is a userland visible behavior
change, it is extremely unlikely to be noticeable and brings sysfs
behavior closer to that of procfs.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kay Sievers <kay@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: use transient write buffer
Tejun Heo [Tue, 1 Oct 2013 21:42:01 +0000 (17:42 -0400)]
sysfs: use transient write buffer

There isn't much to be gained by keeping around kernel buffer while a
file is open especially as the read path planned to be converted to
use seq_file and won't use the buffer.  This patch makes
sysfs_write_file() use per-write transient buffer instead of
sysfs_open_file->page.

This simplifies the write path, enables removing sysfs_open_file->page
once read path is updated and will help merging bin file write path
which already requires the use of a transient buffer due to a locking
order issue.

As the function comments of flush_write_buffer() and
sysfs_write_buffer() are being updated anyway, reformat them so that
they're more conventional.

v2: Use min_t() instead of min() in sysfs_write_file() to avoid build
    warning on arm.  Reported by build test robot.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: add sysfs_open_file->sd and ->file
Tejun Heo [Tue, 1 Oct 2013 21:42:00 +0000 (17:42 -0400)]
sysfs: add sysfs_open_file->sd and ->file

sysfs will be converted to use seq_file for read path, which will make
it difficult to pass around multiple pointers directly.  This patch
adds sysfs_open_file->sd and ->file so that we can reach all the
necessary data structures from sysfs_open_file.

flush_write_buffer() is updated to drop @dentry which was used to
discover the sysfs_dirent as it's now available through
sysfs_open_file->sd.

This patch doesn't cause any behavior difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: rename sysfs_buffer to sysfs_open_file
Tejun Heo [Tue, 1 Oct 2013 21:41:59 +0000 (17:41 -0400)]
sysfs: rename sysfs_buffer to sysfs_open_file

sysfs read path will be converted to use seq_file which will handle
buffering making sysfs_buffer a misnomer.  Rename sysfs_buffer to
sysfs_open_file, and sysfs_open_dirent->buffers to ->files.

This path is pure rename.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: add sysfs_open_file_mutex
Tejun Heo [Tue, 1 Oct 2013 21:41:58 +0000 (17:41 -0400)]
sysfs: add sysfs_open_file_mutex

Add a separate mutex to protect sysfs_open_dirent->buffers list.  This
will allow performing sleepable operations while traversing
sysfs_buffers, which will be renamed to sysfs_open_file.

Note that currently sysfs_open_dirent->buffers list isn't being used
for anything and this patch doesn't make any functional difference.
It will be used to merge regular and bin file supports.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: remove sysfs_buffer->ops
Tejun Heo [Tue, 1 Oct 2013 21:41:57 +0000 (17:41 -0400)]
sysfs: remove sysfs_buffer->ops

Currently, sysfs_ops is fetched during sysfs_open_file() and cached in
sysfs_buffer->ops to be used while the file is open.  This patch
removes the caching and makes each operation directly fetch sysfs_ops.

This patch doesn't introduce any behavior difference and is to prepare
for merging regular and bin file supports.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: remove sysfs_buffer->needs_read_fill
Tejun Heo [Tue, 1 Oct 2013 21:41:56 +0000 (17:41 -0400)]
sysfs: remove sysfs_buffer->needs_read_fill

->needs_read_fill is used to implement the following behaviors.

1. Ensure buffer filling on the first read.
2. Force buffer filling after a write.
3. Force buffer filling after a successful poll.

However, #2 and #3 don't really work as sysfs doesn't reset file
position.  While the read buffer would be refilled, the next read
would continue from the position after the last read or write,
requiring an explicit seek to the start for it to be useful, which
makes ->needs_read_fill superflous as read buffer is always refilled
if f_pos == 0.

Update sysfs_read_file() to test buffer->page for #1 instead and
remove ->needs_read_fill.  While this changes behavior in extreme
corner cases - e.g. re-reading a sysfs file after seeking to non-zero
position after a write or poll, it's highly unlikely to lead to actual
breakage.  This change is to prepare for using seq_file in the read
path.

While at it, reformat a comment in fill_write_buffer().

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Kay Sievers <kay@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: remove unused sysfs_buffer->pos
Tejun Heo [Tue, 1 Oct 2013 21:41:55 +0000 (17:41 -0400)]
sysfs: remove unused sysfs_buffer->pos

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: introduce [__]sysfs_remove()
Tejun Heo [Wed, 18 Sep 2013 21:15:38 +0000 (17:15 -0400)]
sysfs: introduce [__]sysfs_remove()

Given a sysfs_dirent, there is no reason to have multiple versions of
removal functions.  A function which removes the specified
sysfs_dirent and its descendants is enough.

This patch intorduces [__}sysfs_remove() which replaces all internal
variations of removal functions.  This will be the only removal
function in the planned new sysfs_dirent based interface.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: make __sysfs_remove_dir() recursive
Tejun Heo [Wed, 18 Sep 2013 21:15:37 +0000 (17:15 -0400)]
sysfs: make __sysfs_remove_dir() recursive

Currently, sysfs directory removal is inconsistent in that it would
remove any files directly under it but wouldn't recurse into
directories.  Thanks to group subdirectories, this doesn't even match
with kobject boundaries.  sysfs is in the process of being separated
out so that it can be used by multiple subsystems and we want to have
a consistent behavior - either removal of a sysfs_dirent should remove
every descendant entries or none instead of something inbetween.

This patch implements proper recursive removal in
__sysfs_remove_dir().  The function now walks its subtree in a
post-order walk to remove all descendants.

This is a behavior change but kobject / driver layer, which currently
is the only consumer, has already been updated to handle duplicate
removal attempts, so nothing should be broken after this change.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agokobject: grab an extra reference on kobject->sd to allow duplicate deletes
Tejun Heo [Wed, 18 Sep 2013 21:15:36 +0000 (17:15 -0400)]
kobject: grab an extra reference on kobject->sd to allow duplicate deletes

sysfs currently has a rather weird behavior regarding removals.  A
directory removal would delete all files directly under it but
wouldn't recurse into subdirectories, which, while a bit inconsistent,
seems to make sense at the first glance as each directory is
supposedly associated with a kobject and each kobject can take care of
the directory deletion; however, this doesn't really hold as we have
groups which can be directories without a kobject associated with it
and require explicit deletions.

We're in the process of separating out sysfs from kboject / driver
core and want a consistent behavior.  A removal should delete either
only the specified node or everything under it.  I think it is helpful
to support recursive atomic removal and later patches will implement
it.

Such change means that a sysfs_dirent associated with kobject may be
deleted before the kobject itself is removed if one of its ancestor
gets removed before it.  As sysfs_remove_dir() puts the base ref, we
may end up with dangling pointer on descendants.  This can be solved
by holding an extra reference on the sd from kobject.

Acquire an extra reference on the associated sysfs_dirent on directory
creation and put it after removal.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agosysfs: remove sysfs_addrm_cxt->parent_sd
Tejun Heo [Wed, 18 Sep 2013 21:15:35 +0000 (17:15 -0400)]
sysfs: remove sysfs_addrm_cxt->parent_sd

sysfs_addrm_start/finish() enclose sysfs_dirent additions and
deletions and sysfs_addrm_cxt is used to record information necessary
to finish the operations.  Currently, sysfs_addrm_start() takes
@parent_sd, records it in sysfs_addrm_cxt, and assumes that all
operations in the block are performed under that @parent_sd.

This assumption has been fine until now but we want to make some
operations behave recursively and, while having @parent_sd recorded in
sysfs_addrm_cxt doesn't necessarily prevents that, it becomes
confusing.

This patch removes sysfs_addrm_cxt->parent_sd and makes
sysfs_add_one() take an explicit @parent_sd parameter.  Note that
sysfs_remove_one() doesn't need the extra argument as its parent is
always known from the target @sd.

While at it, add __acquires/releases() notations to
sysfs_addrm_start/finish() respectively.

This patch doesn't make any functional difference.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
9 years agoconstify ->actor
Al Viro [Thu, 23 May 2013 02:22:04 +0000 (22:22 -0400)]
constify ->actor

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years ago->readdir() is gone
Al Viro [Thu, 23 May 2013 01:44:23 +0000 (21:44 -0400)]
->readdir() is gone

everything's converted to ->iterate()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert ecryptfs
Al Viro [Thu, 23 May 2013 01:23:40 +0000 (21:23 -0400)]
convert ecryptfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert coda
Al Viro [Thu, 23 May 2013 01:15:30 +0000 (21:15 -0400)]
convert coda

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert ocfs2
Al Viro [Thu, 23 May 2013 01:06:00 +0000 (21:06 -0400)]
convert ocfs2

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert fatfs
Al Viro [Wed, 22 May 2013 22:37:16 +0000 (18:37 -0400)]
convert fatfs

... pox upon the idiotic ioctls; life would be much easier without
those.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert xfs
Al Viro [Wed, 22 May 2013 21:07:56 +0000 (17:07 -0400)]
convert xfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert btrfs
Al Viro [Wed, 22 May 2013 20:48:09 +0000 (16:48 -0400)]
convert btrfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert hostfs
Al Viro [Wed, 22 May 2013 20:34:19 +0000 (16:34 -0400)]
convert hostfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert afs
Al Viro [Wed, 22 May 2013 20:31:14 +0000 (16:31 -0400)]
convert afs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert ncpfs
Al Viro [Wed, 22 May 2013 19:11:27 +0000 (15:11 -0400)]
convert ncpfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert hfsplus
Al Viro [Wed, 22 May 2013 18:59:39 +0000 (14:59 -0400)]
convert hfsplus

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert hfs
Al Viro [Wed, 22 May 2013 18:29:35 +0000 (14:29 -0400)]
convert hfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert befs
Al Viro [Wed, 22 May 2013 17:44:05 +0000 (13:44 -0400)]
convert befs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert cifs
Al Viro [Wed, 22 May 2013 20:17:25 +0000 (16:17 -0400)]
convert cifs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert freevxfs
Al Viro [Sat, 18 May 2013 07:15:00 +0000 (03:15 -0400)]
convert freevxfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert fuse
Al Viro [Sat, 18 May 2013 07:03:58 +0000 (03:03 -0400)]
convert fuse

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert hpfs
Al Viro [Sat, 18 May 2013 06:58:57 +0000 (02:58 -0400)]
convert hpfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoreiserfs: switch reiserfs_readdir_dentry to inode
Al Viro [Sat, 18 May 2013 02:58:58 +0000 (22:58 -0400)]
reiserfs: switch reiserfs_readdir_dentry to inode

... and clean the callers up a bit

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoreiserfs: is_privroot_deh() needs only directory inode, actually
Al Viro [Sat, 18 May 2013 02:45:29 +0000 (22:45 -0400)]
reiserfs: is_privroot_deh() needs only directory inode, actually

... and that - only to get the superblock.  Privroot is a directory
and we don't allow hardlinks to those...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert reiserfs
Al Viro [Sat, 18 May 2013 02:42:17 +0000 (22:42 -0400)]
convert reiserfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert ntfs
Al Viro [Sat, 18 May 2013 01:22:31 +0000 (21:22 -0400)]
convert ntfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert isofs
Al Viro [Sat, 18 May 2013 01:11:59 +0000 (21:11 -0400)]
convert isofs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert jffs2
Al Viro [Fri, 17 May 2013 22:08:49 +0000 (18:08 -0400)]
convert jffs2

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert f2fs
Al Viro [Fri, 17 May 2013 22:02:17 +0000 (18:02 -0400)]
convert f2fs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert 9p
Al Viro [Fri, 17 May 2013 21:51:41 +0000 (17:51 -0400)]
convert 9p

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert affs
Al Viro [Fri, 17 May 2013 21:44:42 +0000 (17:44 -0400)]
convert affs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert adfs
Al Viro [Fri, 17 May 2013 21:30:10 +0000 (17:30 -0400)]
convert adfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert logfs
Al Viro [Fri, 17 May 2013 21:06:34 +0000 (17:06 -0400)]
convert logfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert jfs
Al Viro [Fri, 17 May 2013 21:00:34 +0000 (17:00 -0400)]
convert jfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert ceph
Al Viro [Fri, 17 May 2013 20:52:26 +0000 (16:52 -0400)]
convert ceph

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert nfs
Al Viro [Fri, 17 May 2013 20:34:50 +0000 (16:34 -0400)]
convert nfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert ext4
Al Viro [Fri, 17 May 2013 20:08:53 +0000 (16:08 -0400)]
convert ext4

and trim the living hell out bogosities in inline dir case

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert qnx6
Al Viro [Fri, 17 May 2013 19:32:10 +0000 (15:32 -0400)]
convert qnx6

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert qnx4
Al Viro [Fri, 17 May 2013 19:17:59 +0000 (15:17 -0400)]
convert qnx4

... and use strnlen() instead of strlen() - it's done on untrusted data,
after all.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert omfs
Al Viro [Fri, 17 May 2013 19:05:25 +0000 (15:05 -0400)]
convert omfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert nilfs2
Al Viro [Thu, 16 May 2013 18:36:14 +0000 (14:36 -0400)]
convert nilfs2

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert sysfs
Al Viro [Thu, 16 May 2013 18:31:02 +0000 (14:31 -0400)]
convert sysfs

get rid of the kludges in sysfs_readdir()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert gfs2
Al Viro [Thu, 16 May 2013 18:14:48 +0000 (14:14 -0400)]
convert gfs2

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert exofs
Al Viro [Thu, 16 May 2013 17:48:17 +0000 (13:48 -0400)]
convert exofs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert bfs
Al Viro [Thu, 16 May 2013 17:41:48 +0000 (13:41 -0400)]
convert bfs

... and get rid of that ridiculous mutex in bfs_readdir()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert procfs
Al Viro [Thu, 16 May 2013 16:07:31 +0000 (12:07 -0400)]
convert procfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert openpromfs
Al Viro [Thu, 16 May 2013 05:52:12 +0000 (01:52 -0400)]
convert openpromfs

what the hell is op_mutex for, BTW?

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert efs
Al Viro [Thu, 16 May 2013 05:41:10 +0000 (01:41 -0400)]
convert efs

* sanity checks belong before risky operation, not after it
* don't quit as soon as we'd found an entry

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoconvert configfs
Al Viro [Thu, 16 May 2013 05:28:34 +0000 (01:28 -0400)]
convert configfs

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>