From: Bartosz Golaszewski <[email protected]>
Committable items in configfs are well defined and documented but unfortunately
so far never implemented.
The use-case we have over at the GPIO subsystem is using configfs in
conjunction with sysfs to replace our current gpio-mockup testing module
with one that will be much more flexible and will allow complete coverage
of the GPIO uAPI.
The current gpio-mockup module is controlled using module parameters which
forces the user to reload it everytime they need to change the chip
configuration or layout and makes it difficult to extend its functionality.
Testing module based on configfs would allow fine-grained control over dummy
GPIO chips but since GPIO devices must be configured before they are
instantiated, we need committable items.
This implements them and adds code examples to configfs_sample module. The
first two patches are just cosmetic.
Bartosz Golaszewski (4):
configfs: increase the item name length
configfs: use BIT() for internal flags
configfs: implement committable items
samples: configfs: add a committable group
Documentation/filesystems/configfs.rst | 6 +-
fs/configfs/configfs_internal.h | 22 +--
fs/configfs/dir.c | 239 ++++++++++++++++++++++++-
fs/configfs/file.c | 8 +
include/linux/configfs.h | 3 +-
samples/configfs/configfs_sample.c | 150 ++++++++++++++++
6 files changed, 408 insertions(+), 20 deletions(-)
--
2.29.1
From: Bartosz Golaszewski <[email protected]>
This implements configfs committable items. We mostly follow the
documentation except that we extend config_group_ops with uncommit_item()
callback for reverting the changes made by commit_item().
Each committable group has two sub-directories: pending and live. New
items can only be created in pending/. Attributes can only be modified
while the item is in pending/. Once it's ready to be committed, it must
be moved over to live/ using the rename() system call. This is when the
commit_item() function will be called.
Implementation-wise: we reuse the default group mechanism to elegantly
plug the new pseude-groups into configfs. The pending group inherits the
parent group's operations so that config_items can be seamlesly created
in it using the callbacks supplied by the user as part of the committable
group itself.
Signed-off-by: Bartosz Golaszewski <[email protected]>
---
Documentation/filesystems/configfs.rst | 6 +-
fs/configfs/configfs_internal.h | 2 +
fs/configfs/dir.c | 239 ++++++++++++++++++++++++-
fs/configfs/file.c | 8 +
include/linux/configfs.h | 1 +
5 files changed, 247 insertions(+), 9 deletions(-)
diff --git a/Documentation/filesystems/configfs.rst b/Documentation/filesystems/configfs.rst
index 1d3d6f4a82a9..7e0e7c356450 100644
--- a/Documentation/filesystems/configfs.rst
+++ b/Documentation/filesystems/configfs.rst
@@ -290,6 +290,7 @@ config_item_type::
struct config_group *(*make_group)(struct config_group *group,
const char *name);
int (*commit_item)(struct config_item *item);
+ int (*uncommit_item)(struct config_item *item);
void (*disconnect_notify)(struct config_group *group,
struct config_item *item);
void (*drop_item)(struct config_group *group,
@@ -490,9 +491,6 @@ pass up an error.
Committable Items
=================
-Note:
- Committable items are currently unimplemented.
-
Some config_items cannot have a valid initial state. That is, no
default values can be specified for the item's attributes such that the
item can do its work. Userspace must configure one or more attributes,
@@ -532,4 +530,4 @@ method returns zero and the item is moved to the "live" directory.
As rmdir(2) does not work in the "live" directory, an item must be
shutdown, or "uncommitted". Again, this is done via rename(2), this
time from the "live" directory back to the "pending" one. The subsystem
-is notified by the ct_group_ops->uncommit_object() method.
+is notified by the ct_group_ops->uncommit_item() method.
diff --git a/fs/configfs/configfs_internal.h b/fs/configfs/configfs_internal.h
index 855f00868654..22b649567b13 100644
--- a/fs/configfs/configfs_internal.h
+++ b/fs/configfs/configfs_internal.h
@@ -56,6 +56,8 @@ struct configfs_dirent {
#define CONFIGFS_USET_DROPPING BIT(8)
#define CONFIGFS_USET_IN_MKDIR BIT(9)
#define CONFIGFS_USET_CREATING BIT(10)
+#define CONFIGFS_GROUP_PENDING BIT(11)
+#define CONFIGFS_GROUP_LIVE BIT(12)
#define CONFIGFS_NOT_PINNED (CONFIGFS_ITEM_ATTR | CONFIGFS_ITEM_BIN_ATTR)
extern struct mutex configfs_symlink_mutex;
diff --git a/fs/configfs/dir.c b/fs/configfs/dir.c
index b0983e2a4e2c..73e3bd30a2ad 100644
--- a/fs/configfs/dir.c
+++ b/fs/configfs/dir.c
@@ -655,6 +655,13 @@ static void detach_groups(struct config_group *group)
inode_unlock(d_inode(child));
+ /*
+ * Free memory allocated for the pending and live directories
+ * of committable groups.
+ */
+ if (sd->s_type & (CONFIGFS_GROUP_PENDING | CONFIGFS_GROUP_LIVE))
+ kfree(sd->s_element);
+
d_delete(child);
dput(child);
}
@@ -859,6 +866,133 @@ static void configfs_detach_item(struct config_item *item)
configfs_remove_dir(item);
}
+static bool is_committable_group(struct config_item *item)
+{
+ const struct config_item_type *type = item->ci_type;
+
+ if (type && type->ct_group_ops &&
+ type->ct_group_ops->commit_item &&
+ type->ct_group_ops->uncommit_item)
+ return true;
+
+ return false;
+}
+
+struct pending_group_data {
+ struct config_group group;
+ struct config_item_type type;
+ struct configfs_group_operations group_ops;
+};
+
+struct live_group_data {
+ struct config_group group;
+ struct config_item_type type;
+};
+
+static int create_pending_group(struct config_item *parent_item,
+ struct configfs_fragment *frag)
+{
+ const struct config_item_type *parent_type = parent_item->ci_type;
+ struct pending_group_data *pending;
+ struct configfs_dirent *sd;
+ int ret;
+
+ pending = kzalloc(sizeof(*pending), GFP_KERNEL);
+ if (!pending)
+ return -ENOMEM;
+
+ /*
+ * Let's inherit the group_ops from the parent except for item
+ * committing and uncommitting.
+ */
+ memcpy(&pending->group_ops, parent_type->ct_group_ops,
+ sizeof(struct configfs_group_operations));
+ pending->type.ct_group_ops = &pending->group_ops;
+ pending->type.ct_group_ops->commit_item = NULL;
+ pending->type.ct_group_ops->uncommit_item = NULL;
+
+ /* Let's directly reuse item_ops. */
+ pending->type.ct_item_ops = parent_type->ct_item_ops;
+ pending->type.ct_owner = parent_type->ct_owner;
+
+ config_group_init_type_name(&pending->group, "pending", &pending->type);
+
+ ret = create_default_group(to_config_group(parent_item),
+ &pending->group, frag);
+ if (ret) {
+ kfree(pending);
+ return ret;
+ }
+
+ link_group(to_config_group(parent_item), &pending->group);
+
+ sd = pending->group.cg_item.ci_dentry->d_fsdata;
+ /* Allow creating config_items in 'pending' group. */
+ sd->s_type |= (CONFIGFS_GROUP_PENDING | CONFIGFS_USET_DIR);
+
+ return 0;
+}
+
+static int create_live_group(struct config_item *parent_item,
+ struct configfs_fragment *frag)
+{
+ struct live_group_data *live;
+ struct configfs_dirent *sd;
+ int ret;
+
+ live = kzalloc(sizeof(*live), GFP_KERNEL);
+ if (!live)
+ return -ENOMEM;
+
+ live->type.ct_owner = parent_item->ci_type->ct_owner;
+
+ config_group_init_type_name(&live->group, "live", &live->type);
+
+ ret = create_default_group(to_config_group(parent_item),
+ &live->group, frag);
+ if (ret) {
+ kfree(live);
+ return ret;
+ }
+
+ link_group(to_config_group(parent_item), &live->group);
+
+ sd = live->group.cg_item.ci_dentry->d_fsdata;
+ sd->s_type |= CONFIGFS_GROUP_LIVE;
+
+ return 0;
+}
+
+static int create_committable_groups(struct config_item *parent_item,
+ struct configfs_fragment *frag)
+{
+ struct configfs_dirent *sd;
+ int ret;
+
+ ret = create_pending_group(parent_item, frag);
+ if (ret)
+ return ret;
+
+ ret = create_live_group(parent_item, frag);
+ if (ret) {
+ detach_groups(to_config_group(parent_item));
+ return ret;
+ }
+
+ /* Disallow creating items directly in the committable group. */
+ sd = parent_item->ci_dentry->d_fsdata;
+ sd->s_type &= ~CONFIGFS_USET_DIR;
+
+ return 0;
+}
+
+static void dentry_mark_dead(struct config_item *item, struct dentry *dentry)
+{
+ configfs_detach_item(item);
+ d_inode(dentry)->i_flags |= S_DEAD;
+ dont_mount(dentry);
+}
+
static int configfs_attach_group(struct config_item *parent_item,
struct config_item *item,
struct dentry *dentry,
@@ -884,11 +1018,15 @@ static int configfs_attach_group(struct config_item *parent_item,
inode_lock_nested(d_inode(dentry), I_MUTEX_CHILD);
configfs_adjust_dir_dirent_depth_before_populate(sd);
ret = populate_groups(to_config_group(item), frag);
- if (ret) {
- configfs_detach_item(item);
- d_inode(dentry)->i_flags |= S_DEAD;
- dont_mount(dentry);
+ if (ret)
+ dentry_mark_dead(item, dentry);
+
+ if (is_committable_group(item)) {
+ ret = create_committable_groups(item, frag);
+ if (ret)
+ dentry_mark_dead(item, dentry);
}
+
configfs_adjust_dir_dirent_depth_after_populate(sd);
inode_unlock(d_inode(dentry));
if (ret)
@@ -1455,7 +1593,7 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
struct config_item *parent_item;
struct config_item *item;
struct configfs_subsystem *subsys;
- struct configfs_dirent *sd;
+ struct configfs_dirent *sd, *parent_sd;
struct configfs_fragment *frag;
struct module *subsys_owner = NULL, *dead_item_owner = NULL;
int ret;
@@ -1474,6 +1612,12 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
return -EINVAL;
}
+ parent_sd = dentry->d_parent->d_fsdata;
+ if (parent_sd->s_type & CONFIGFS_GROUP_LIVE) {
+ config_item_put(parent_item);
+ return -EPERM;
+ }
+
/* configfs_mkdir() shouldn't have allowed this */
BUG_ON(!subsys->su_group.cg_item.ci_type);
subsys_owner = subsys->su_group.cg_item.ci_type->ct_owner;
@@ -1560,9 +1704,94 @@ static int configfs_rmdir(struct inode *dir, struct dentry *dentry)
return 0;
}
+static int configfs_rename(struct inode *old_dir, struct dentry *old_dentry,
+ struct inode *new_dir, struct dentry *new_dentry,
+ unsigned int flags)
+{
+ struct configfs_dirent *sd, *old_parent_sd, *new_parent_sd;
+ struct dentry *old_parent_dentry, *new_parent_dentry;
+ struct configfs_dirent *committable_group_sd;
+ struct dentry *committable_group_dentry;
+ struct config_item *committable_group_item, *item, *new_parent_item;
+ struct configfs_subsystem *committable_group_subsys;
+ struct configfs_group_operations *committable_group_ops;
+ int ret = 0;
+
+ if (flags)
+ return -EINVAL;
+
+ old_parent_dentry = old_dentry->d_parent;
+ new_parent_dentry = new_dentry->d_parent;
+
+ sd = old_dentry->d_fsdata;
+ old_parent_sd = old_dentry->d_parent->d_fsdata;
+ new_parent_sd = new_dentry->d_parent->d_fsdata;
+
+ if (!old_parent_sd || !new_parent_sd)
+ return -EPERM;
+
+ /*
+ * Renaming must always be between a 'pending' and a 'live' group and
+ * both need to have the same parent.
+ */
+ if (!((old_parent_sd->s_type & CONFIGFS_GROUP_PENDING) &&
+ (new_parent_sd->s_type & CONFIGFS_GROUP_LIVE)) &&
+ !((old_parent_sd->s_type & CONFIGFS_GROUP_LIVE) &&
+ (new_parent_sd->s_type & CONFIGFS_GROUP_PENDING)))
+ return -EPERM;
+
+ if (old_parent_dentry->d_parent != new_parent_dentry->d_parent)
+ return -EPERM;
+
+ committable_group_dentry = old_parent_dentry->d_parent;
+ committable_group_sd = committable_group_dentry->d_fsdata;
+ /*
+ * Grab a reference to the committable group for the duration of
+ * this function.
+ */
+ committable_group_item =
+ configfs_get_config_item(committable_group_dentry);
+ committable_group_subsys =
+ to_config_group(committable_group_item)->cg_subsys;
+ committable_group_ops = committable_group_item->ci_type->ct_group_ops;
+
+ item = sd->s_element;
+ new_parent_item = new_parent_sd->s_element;
+
+ if (WARN_ON(!is_committable_group(committable_group_item))) {
+ /* This would be a result of a programming error in configfs. */
+ config_item_put(committable_group_item);
+ return -EPERM;
+ }
+
+ mutex_lock(&committable_group_subsys->su_mutex);
+ spin_lock(&configfs_dirent_lock);
+
+ if ((old_parent_sd->s_type & CONFIGFS_GROUP_PENDING) &&
+ (new_parent_sd->s_type & CONFIGFS_GROUP_LIVE))
+ ret = committable_group_ops->commit_item(item);
+ else
+ ret = committable_group_ops->uncommit_item(item);
+ if (ret)
+ goto out;
+
+ new_dentry->d_fsdata = sd;
+ list_move(&sd->s_sibling, &new_parent_sd->s_children);
+ item->ci_parent = new_parent_item;
+ d_move(old_dentry, new_dentry);
+
+out:
+ spin_unlock(&configfs_dirent_lock);
+ mutex_unlock(&committable_group_subsys->su_mutex);
+ config_item_put(committable_group_item);
+
+ return ret;
+}
+
const struct inode_operations configfs_dir_inode_operations = {
.mkdir = configfs_mkdir,
.rmdir = configfs_rmdir,
+ .rename = configfs_rename,
.symlink = configfs_symlink,
.unlink = configfs_unlink,
.lookup = configfs_lookup,
diff --git a/fs/configfs/file.c b/fs/configfs/file.c
index 1f0270229d7b..a20e55fd05e8 100644
--- a/fs/configfs/file.c
+++ b/fs/configfs/file.c
@@ -243,9 +243,17 @@ fill_write_buffer(struct configfs_buffer * buffer, const char __user * buf, size
static int
flush_write_buffer(struct file *file, struct configfs_buffer *buffer, size_t count)
{
+ struct config_item *parent_item = buffer->item->ci_parent;
struct configfs_fragment *frag = to_frag(file);
+ struct configfs_dirent *sd;
int res = -ENOENT;
+ if (parent_item && parent_item->ci_dentry) {
+ sd = parent_item->ci_dentry->d_fsdata;
+ if (sd->s_type & CONFIGFS_GROUP_LIVE)
+ return -EPERM;
+ }
+
down_read(&frag->frag_sem);
if (!frag->frag_dead)
res = buffer->attr->store(buffer->item, buffer->page, count);
diff --git a/include/linux/configfs.h b/include/linux/configfs.h
index 4f76dcc08134..ff6b0e408136 100644
--- a/include/linux/configfs.h
+++ b/include/linux/configfs.h
@@ -219,6 +219,7 @@ struct configfs_group_operations {
struct config_item *(*make_item)(struct config_group *group, const char *name);
struct config_group *(*make_group)(struct config_group *group, const char *name);
int (*commit_item)(struct config_item *item);
+ int (*uncommit_item)(struct config_item *item);
void (*disconnect_notify)(struct config_group *group, struct config_item *item);
void (*drop_item)(struct config_group *group, struct config_item *item);
};
--
2.29.1
On Wed, Nov 25, 2020 at 04:22:46PM +0100, Bartosz Golaszewski wrote:
> From: Bartosz Golaszewski <[email protected]>
>
> This implements configfs committable items. We mostly follow the
> documentation except that we extend config_group_ops with uncommit_item()
> callback for reverting the changes made by commit_item().
Woohoo! A long time coming, but thank you for working on the
implementation!
> Each committable group has two sub-directories: pending and live. New
> items can only be created in pending/. Attributes can only be modified
> while the item is in pending/. Once it's ready to be committed, it must
> be moved over to live/ using the rename() system call. This is when the
> commit_item() function will be called.
The original API intended for live items to still be modifyable. The
live/ path forbids mkdir()/rmdir(), but it allows store(). Otherwise,
items can't be adjusted at all while in use, which is severely limiting.
Obviously the store() handler must not allow transitions from
valid-value->invalid-value, but the handler would reject invalid values
anyway, wouldn't it?
> diff --git a/fs/configfs/file.c b/fs/configfs/file.c
> index 1f0270229d7b..a20e55fd05e8 100644
> --- a/fs/configfs/file.c
> +++ b/fs/configfs/file.c
> @@ -243,9 +243,17 @@ fill_write_buffer(struct configfs_buffer * buffer, const char __user * buf, size
> static int
> flush_write_buffer(struct file *file, struct configfs_buffer *buffer, size_t count)
> {
> + struct config_item *parent_item = buffer->item->ci_parent;
> struct configfs_fragment *frag = to_frag(file);
> + struct configfs_dirent *sd;
> int res = -ENOENT;
>
> + if (parent_item && parent_item->ci_dentry) {
> + sd = parent_item->ci_dentry->d_fsdata;
> + if (sd->s_type & CONFIGFS_GROUP_LIVE)
> + return -EPERM;
> + }
> +
> down_read(&frag->frag_sem);
> if (!frag->frag_dead)
> res = buffer->attr->store(buffer->item, buffer->page, count);
Basically, I would just leave this hunk out.
Thanks,
Joel
--
"Now Someone's on the telephone, desperate in his pain.
Someone's on the bathroom floor doing her cocaine.
Someone's got his finger on the button in some room.
No one can convince me we aren't gluttons for our doom."
http://www.jlbec.org/
[email protected]
On Thu, Jan 14, 2021 at 12:46 AM Joel Becker <[email protected]> wrote:
>
> On Wed, Nov 25, 2020 at 04:22:46PM +0100, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <[email protected]>
> >
> > This implements configfs committable items. We mostly follow the
> > documentation except that we extend config_group_ops with uncommit_item()
> > callback for reverting the changes made by commit_item().
>
> Woohoo! A long time coming, but thank you for working on the
> implementation!
>
> > Each committable group has two sub-directories: pending and live. New
> > items can only be created in pending/. Attributes can only be modified
> > while the item is in pending/. Once it's ready to be committed, it must
> > be moved over to live/ using the rename() system call. This is when the
> > commit_item() function will be called.
>
> The original API intended for live items to still be modifyable. The
> live/ path forbids mkdir()/rmdir(), but it allows store(). Otherwise,
> items can't be adjusted at all while in use, which is severely limiting.
> Obviously the store() handler must not allow transitions from
> valid-value->invalid-value, but the handler would reject invalid values
> anyway, wouldn't it?
>
> > diff --git a/fs/configfs/file.c b/fs/configfs/file.c
> > index 1f0270229d7b..a20e55fd05e8 100644
> > --- a/fs/configfs/file.c
> > +++ b/fs/configfs/file.c
> > @@ -243,9 +243,17 @@ fill_write_buffer(struct configfs_buffer * buffer, const char __user * buf, size
> > static int
> > flush_write_buffer(struct file *file, struct configfs_buffer *buffer, size_t count)
> > {
> > + struct config_item *parent_item = buffer->item->ci_parent;
> > struct configfs_fragment *frag = to_frag(file);
> > + struct configfs_dirent *sd;
> > int res = -ENOENT;
> >
> > + if (parent_item && parent_item->ci_dentry) {
> > + sd = parent_item->ci_dentry->d_fsdata;
> > + if (sd->s_type & CONFIGFS_GROUP_LIVE)
> > + return -EPERM;
> > + }
> > +
> > down_read(&frag->frag_sem);
> > if (!frag->frag_dead)
> > res = buffer->attr->store(buffer->item, buffer->page, count);
>
> Basically, I would just leave this hunk out.
>
I would make this configurable per-attribute because for the use-case
I need this for we definitely don't want the items to be modifiable
once they're "live".
Bartosz
On Thu, Jan 14, 2021 at 9:40 AM Bartosz Golaszewski
<[email protected]> wrote:
>
> On Thu, Jan 14, 2021 at 12:46 AM Joel Becker <[email protected]> wrote:
> >
> > On Wed, Nov 25, 2020 at 04:22:46PM +0100, Bartosz Golaszewski wrote:
> > > From: Bartosz Golaszewski <[email protected]>
> > >
> > > This implements configfs committable items. We mostly follow the
> > > documentation except that we extend config_group_ops with uncommit_item()
> > > callback for reverting the changes made by commit_item().
> >
> > Woohoo! A long time coming, but thank you for working on the
> > implementation!
> >
> > > Each committable group has two sub-directories: pending and live. New
> > > items can only be created in pending/. Attributes can only be modified
> > > while the item is in pending/. Once it's ready to be committed, it must
> > > be moved over to live/ using the rename() system call. This is when the
> > > commit_item() function will be called.
> >
> > The original API intended for live items to still be modifyable. The
> > live/ path forbids mkdir()/rmdir(), but it allows store(). Otherwise,
> > items can't be adjusted at all while in use, which is severely limiting.
> > Obviously the store() handler must not allow transitions from
> > valid-value->invalid-value, but the handler would reject invalid values
> > anyway, wouldn't it?
> >
> > > diff --git a/fs/configfs/file.c b/fs/configfs/file.c
> > > index 1f0270229d7b..a20e55fd05e8 100644
> > > --- a/fs/configfs/file.c
> > > +++ b/fs/configfs/file.c
> > > @@ -243,9 +243,17 @@ fill_write_buffer(struct configfs_buffer * buffer, const char __user * buf, size
> > > static int
> > > flush_write_buffer(struct file *file, struct configfs_buffer *buffer, size_t count)
> > > {
> > > + struct config_item *parent_item = buffer->item->ci_parent;
> > > struct configfs_fragment *frag = to_frag(file);
> > > + struct configfs_dirent *sd;
> > > int res = -ENOENT;
> > >
> > > + if (parent_item && parent_item->ci_dentry) {
> > > + sd = parent_item->ci_dentry->d_fsdata;
> > > + if (sd->s_type & CONFIGFS_GROUP_LIVE)
> > > + return -EPERM;
> > > + }
> > > +
> > > down_read(&frag->frag_sem);
> > > if (!frag->frag_dead)
> > > res = buffer->attr->store(buffer->item, buffer->page, count);
> >
> > Basically, I would just leave this hunk out.
> >
>
> I would make this configurable per-attribute because for the use-case
> I need this for we definitely don't want the items to be modifiable
> once they're "live".
>
> Bartosz
After a second thought: I agree this can be left to the user's
discretion so I'll just remove this and enforce it in the sample code.
One thing I'm unsure about is whether we should allow to change the
group's name when moving it from pending to live. Any thoughts?
Bartosz