On Mon, Feb 9, 2015 at 8:23 AM, 王旭 <xuw2015@xxxxxxxxx> wrote:
> Do not cache the overlay file dentries in dentry lru. For the fs dentry of
> overlay is fake, the real operations happen either lower dentry or upper dentry.
> So it's not helpful to performance.
Lookup of an uncached dentry causes acquisition of parent's i_mutex.
This can very much degrade performance if multiple lookups are being
performed in the same directory at the same time. So I don't think
dropping the overlay dentry is a good idea.
> And for a special use, the fnctl F_SETLEASE F_WRLCK needs the dentry count is 1.
> If we cache the ovl dentry, which will hold the reference for upper dentry, and
> the fcntl will do operation on upper dentry, whose reference will be 2. And
> finally the fcntl F_SETLEASE F_WRLCK will fail with -EAGAIN. The ltp fcntl24,
> fcntl25 and fcntl26 test case cover it.
The F_WRLOCK check looks somewhat bogus because it will succeed or
fail depending on whether or not a file with multiple hard links has
more than one of it's dentries cached. I guess that needs to be
cleaned up similar to how F_RDLOCK is handled (i.e. have a separate
opencount in the inode, instead of trying to guess that based on
d_count and i_count).
Thanks,
Miklos
>
> Signed-off-by: George Wang <xuw@xxxxxxxxxx>
> ---
> fs/overlayfs/super.c | 14 +++++++++++++-
> 1 file changed, 13 insertions(+), 1 deletion(-)
>
> diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
> index f16d318..80b9ea8 100644
> --- a/fs/overlayfs/super.c
> +++ b/fs/overlayfs/super.c
> @@ -256,6 +256,17 @@ static bool ovl_is_opaquedir(struct dentry *dentry)
> return false;
> }
>
> +static int ovl_dentry_delete(struct dentry *dentry)
> +{
> + struct inode *inode = dentry->d_inode;
> +
> + //dir dentry holds ovl_dir_cache, which will speed up readdir
> + if (inode && S_ISDIR(inode->i_mode))
> + return 0;
> +
> + return 1;
> +}
> +
> static void ovl_dentry_release(struct dentry *dentry)
> {
> struct ovl_entry *oe = dentry->d_fsdata;
> @@ -268,7 +279,8 @@ static void ovl_dentry_release(struct dentry *dentry)
> }
>
> static const struct dentry_operations ovl_dentry_operations = {
> - .d_release = ovl_dentry_release,
> + .d_delete = ovl_dentry_delete,
> + .d_release = ovl_dentry_release,
> };
>
> static struct ovl_entry *ovl_alloc_entry(void)
> --
> 1.9.3
--
To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
