[AFS] prevent double cell registration

[AFS] prevent double cell registration

[Sven Schnelle]

[-- Attachment #1: Type: text/plain, Size: 954 bytes --]


kafs doesn't check if the cell already exists - so if you do an
echo "add newcell.org 1.2.3.4" >/proc/fs/afs/cells it will try to
create this cell again. kobject will also complain about a double
registration. To prevent such problems, return -EEXIST in that case.
---
 fs/afs/cell.c |    9 +++++++++
 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 970d38f..87934b3 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -127,6 +127,15 @@ struct afs_cell *afs_cell_create(const char *name, char *vllist)
 
 	_enter("%s,%s", name, vllist);
 
+	read_lock(&afs_cells_lock);
+	list_for_each_entry(cell, &afs_cells, link) {
+		if(!strcasecmp(cell->name, name)) {
+			read_unlock(&afs_cells_lock);
+			return ERR_PTR(-EEXIST);
+		}
+	}
+	read_unlock(&afs_cells_lock);
+
 	cell = afs_cell_alloc(name, vllist);
 	if (IS_ERR(cell)) {
 		_leave(" = %ld", PTR_ERR(cell));
-- 
1.5.4.3


[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: [AFS] prevent double cell registration

[David Howells]

Sven Schnelle <svens@stackframe.org> wrote:

> kafs doesn't check if the cell already exists - so if you do an
> echo "add newcell.org 1.2.3.4" >/proc/fs/afs/cells it will try to
> create this cell again. kobject will also complain about a double
> registration. To prevent such problems, return -EEXIST in that case.

Hmmm...  I thought I had that covered.  Thanks.  Any chance of a Signed-off-by
line?

David

Re: [AFS] prevent double cell registration

[Sven Schnelle]

David Howells <dhowells@redhat.com> writes:

> Sven Schnelle <svens@stackframe.org> wrote:
>> kafs doesn't check if the cell already exists - so if you do an
>> echo "add newcell.org 1.2.3.4" >/proc/fs/afs/cells it will try to
>> create this cell again. kobject will also complain about a double
>> registration. To prevent such problems, return -EEXIST in that case.
>
> Hmmm...  I thought I had that covered.  Thanks.  Any chance of a Signed-off-by
> line?

oops, sorry. Here it is:

Signed-off-by: Sven Schnelle <svens@stackframe.org>

Sven.

Re: [AFS] prevent double cell registration

[David Howells]

Your patch needs a bit of adjustment.  The check must be done inside the write
lock on afs_cells_sem otherwise there's a race.

David
---
AFS: prevent double cell registration

From: Sven Schnelle <svens@stackframe.org>

kafs doesn't check if the cell already exists - so if you do an
echo "add newcell.org 1.2.3.4" >/proc/fs/afs/cells it will try to
create this cell again. kobject will also complain about a double
registration. To prevent such problems, return -EEXIST in that case.

Signed-off-by: Sven Schnelle <svens@stackframe.org>
Signed-off-by: David Howells <dhowells@redhat.com>
---

 fs/afs/cell.c |   15 +++++++++++++--
 1 files changed, 13 insertions(+), 2 deletions(-)


diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 970d38f..788865d 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -127,14 +127,20 @@ struct afs_cell *afs_cell_create(const char *name, char *vllist)
 
 	_enter("%s,%s", name, vllist);
 
+	down_write(&afs_cells_sem);
+	read_lock(&afs_cells_lock);
+	list_for_each_entry(cell, &afs_cells, link) {
+		if (strcasecmp(cell->name, name) == 0)
+			goto duplicate_name;
+	}
+	read_unlock(&afs_cells_lock);
+
 	cell = afs_cell_alloc(name, vllist);
 	if (IS_ERR(cell)) {
 		_leave(" = %ld", PTR_ERR(cell));
 		return cell;
 	}
 
-	down_write(&afs_cells_sem);
-
 	/* add a proc directory for this cell */
 	ret = afs_proc_cell_setup(cell);
 	if (ret < 0)
@@ -167,6 +173,11 @@ error:
 	kfree(cell);
 	_leave(" = %d", ret);
 	return ERR_PTR(ret);
+
+duplicate_name:
+	read_unlock(&afs_cells_lock);
+	up_write(&afs_cells_sem);
+	return ERR_PTR(-EEXIST);
 }
 
 /*

Re: [AFS] prevent double cell registration

[Sven Schnelle]

Hi David,

David Howells <dhowells@redhat.com> writes:

> Your patch needs a bit of adjustment.  The check must be done inside the write
> lock on afs_cells_sem otherwise there's a race.
>
> David
> ---
> AFS: prevent double cell registration
>
> From: Sven Schnelle <svens@stackframe.org>
>
> kafs doesn't check if the cell already exists - so if you do an
> echo "add newcell.org 1.2.3.4" >/proc/fs/afs/cells it will try to
> create this cell again. kobject will also complain about a double
> registration. To prevent such problems, return -EEXIST in that case.
>
> Signed-off-by: Sven Schnelle <svens@stackframe.org>
> Signed-off-by: David Howells <dhowells@redhat.com>
> ---
>
>  fs/afs/cell.c |   15 +++++++++++++--
>  1 files changed, 13 insertions(+), 2 deletions(-)
>
>
> diff --git a/fs/afs/cell.c b/fs/afs/cell.c
> index 970d38f..788865d 100644
> --- a/fs/afs/cell.c
> +++ b/fs/afs/cell.c
> @@ -127,14 +127,20 @@ struct afs_cell *afs_cell_create(const char *name, char *vllist)
>  
>  	_enter("%s,%s", name, vllist);
>  
> +	down_write(&afs_cells_sem);

> +	list_for_each_entry(cell, &afs_cells, link) {
> +		if (strcasecmp(cell->name, name) == 0)
> +			goto duplicate_name;
> +	}
> +	read_unlock(&afs_cells_lock);
> +
>  	cell = afs_cell_alloc(name, vllist);
>  	if (IS_ERR(cell)) {

Just digged again into afs, and stumbled upon this return. I think there should
be some up_write(&afs_cells_sem); before the return, or am i on the wrong track?

>  		_leave(" = %ld", PTR_ERR(cell));
>  		return cell;
>  	}
>  
> -	down_write(&afs_cells_sem);
> -
>  	/* add a proc directory for this cell */
>  	ret = afs_proc_cell_setup(cell);
>  	if (ret < 0)
> @@ -167,6 +173,11 @@ error:
>  	kfree(cell);
>  	_leave(" = %d", ret);
>  	return ERR_PTR(ret);
> +
> +duplicate_name:
> +	read_unlock(&afs_cells_lock);
> +	up_write(&afs_cells_sem);
> +	return ERR_PTR(-EEXIST);
>  }
>  
>  /*

Re: [AFS] prevent double cell registration

[David Howells]

Sven Schnelle <svens@stackframe.org> wrote:

> Just digged again into afs, and stumbled upon this return. I think there
> should be some up_write(&afs_cells_sem); before the return, or am i on the
> wrong track?

Unfortunately not:-(

David

[PATCH] [PATCH] afs: add missing up_write() on return

[Sven Schnelle]

if afs_cell_alloc() fails, afs_cells_sem doesn't get unlocked, which
leads to a deadlock. Unlock it before returning.

Signed-off-by: Sven Schnelle <svens@stackframe.org>
---
 fs/afs/cell.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/fs/afs/cell.c b/fs/afs/cell.c
index 788865d..584bb0f 100644
--- a/fs/afs/cell.c
+++ b/fs/afs/cell.c
@@ -138,6 +138,7 @@ struct afs_cell *afs_cell_create(const char *name, char *vllist)
 	cell = afs_cell_alloc(name, vllist);
 	if (IS_ERR(cell)) {
 		_leave(" = %ld", PTR_ERR(cell));
+		up_write(&afs_cells_sem);
 		return cell;
 	}
 
-- 
1.5.4.5