vxfs_readdir locking incorrect: add lock_kernel() or remove unlock_kernel()?

vxfs_readdir locking incorrect: add lock_kernel() or remove unlock_kernel()?

[Josh Triplett]

Commit 7b2fd697427e73c81d5fa659efd91bd07d303b0e in the historical GIT
tree stopped calling the readdir member of a file_operations struct with
the big kernel lock held, and fixed up all the readdir functions to do
their own locking.  However, that change added calls to unlock_kernel()
in vxfs_readdir (fs/freevxfs/vxfs_lookup.c), but no call to
lock_kernel().  Should vxfs_readdir call lock_kernel(), or should the
calls to unlock_kernel() go away?

- Josh Triplett

Re: vxfs_readdir locking incorrect: add lock_kernel() or remove unlock_kernel()?

[Andrew Morton]

On Mon, 24 Jul 2006 15:42:17 -0700
Josh Triplett <josht@us.ibm.com> wrote:

> Commit 7b2fd697427e73c81d5fa659efd91bd07d303b0e in the historical GIT
> tree stopped calling the readdir member of a file_operations struct with
> the big kernel lock held, and fixed up all the readdir functions to do
> their own locking.  However, that change added calls to unlock_kernel()
> in vxfs_readdir (fs/freevxfs/vxfs_lookup.c), but no call to
> lock_kernel().

That would appear to imply that nobody has used freevxfs in four years.

>  Should vxfs_readdir call lock_kernel(), or should the
> calls to unlock_kernel() go away?

I don't see anything in there which needs the locking, apart from perhaps
f_pos updates.  But it's probably best to add the lock_kernel() - this is a
bugfixing exercise, not a remove-BKL-from-freevxfs exercise.

Re: vxfs_readdir locking incorrect: add lock_kernel() or remove unlock_kernel()?

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 421 bytes --]

On Mon, 24 Jul 2006 17:20:40 PDT, Andrew Morton said:
> That would appear to imply that nobody has used freevxfs in four years.

Given that...

> I don't see anything in there which needs the locking, apart from perhaps
> f_pos updates.  But it's probably best to add the lock_kernel() - this is a
> bugfixing exercise, not a remove-BKL-from-freevxfs exercise.

Should we consider a remove-freevxfs-from-kernel exercise?

[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]