On Wed, Aug 1, 2018 at 11:51 AM, Amir Goldstein <amir73il@xxxxxxxxx> wrote:
> On Tue, Jul 31, 2018 at 12:37 PM, zhangyi (F) <yi.zhang@xxxxxxxxxx> wrote:
>> overlayfs will want to add the upper layer's feature bitset if some
>> overlay file system features were set. For example: redirect dir feature
>> will be set when user rename a lower/merge dir if redirect dir feature
>> is enabled. So add helper functions to set feature sets to the upper
>> layer for future use.
>>
>> This patch introduce helper functions only, does not set any feature
>> bits.
>>
>> Signed-off-by: zhangyi (F) <yi.zhang@xxxxxxxxxx>
>> ---
>> fs/overlayfs/feature.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++++
>> fs/overlayfs/overlayfs.h | 46 ++++++++++++++++++++++++++
>> fs/overlayfs/ovl_entry.h | 2 ++
>> fs/overlayfs/super.c | 1 +
>> 4 files changed, 134 insertions(+)
>>
>> diff --git a/fs/overlayfs/feature.c b/fs/overlayfs/feature.c
>> index fac4d7544475..7c66e71bdd98 100644
>> --- a/fs/overlayfs/feature.c
>> +++ b/fs/overlayfs/feature.c
>> @@ -10,6 +10,8 @@
>>
>> #include <linux/fs.h>
>> #include <linux/xattr.h>
>> +#include <linux/mount.h>
>> +#include <linux/ratelimit.h>
>> #include "overlayfs.h"
>>
>> /*
>> @@ -63,6 +65,89 @@ static struct ovl_d_feature *ovl_get_feature(struct dentry *realroot)
>> }
>>
>> /*
>> + * Set overlay features to the upper root dir.
>> + *
>> + * @upper_layer: overlay upper layer information,
>> + * @upper: upper root dentry.
>> + *
>> + * Return 0 if success, error number otherwise.
>> + */
>> +static int ovl_set_feature(struct ovl_layer *upper_layer,
>> + struct dentry *upper)
>> +{
>> + struct vfsmount *upper_mnt = upper_layer->mnt;
>> + struct ovl_d_feature odf = {};
>> + int err;
>> +
>> + odf.magic = OVL_FEATURE_MAGIC;
>> + odf.compat = cpu_to_be64(upper_layer->compat);
>> + odf.ro_compat = cpu_to_be64(upper_layer->ro_compat);
>> + odf.incompat = cpu_to_be64(upper_layer->incompat);
>> +
>> + err = mnt_want_write(upper_mnt);
>> + if (err)
>> + return err;
>> +
>> + err = ovl_do_setxattr(upper, OVL_XATTR_FEATURE,
>> + &odf, sizeof(odf), 0);
>> + if (err)
>> + pr_err_ratelimited("overlayfs: failed to set features (%i)\n", err);
>> +
>> + mnt_drop_write(upper_mnt);
>> + return err;
>> +}
>> +
>> +/*
>> + * Add feature bit mask to the upper root dir.
>> + *
>> + * @ofs: overlay filesystem information
>> + * @type: feature set type, compat, ro compat or incompat
>> + * @mask: new features mask want to add
>> + *
>> + * Return 0 if success, error number otherwise.
>> + */
>> +int ovl_set_upper_feature(struct ovl_fs *ofs,
>> + enum ovl_feature_type type,
>> + u64 mask)
>> +{
>> + struct ovl_layer *upper_layer = ofs->upper_layer;
>> + u64 *features;
>> + int err;
>> +
>> + if (!upper_layer)
>> + return -EINVAL;
>> +
>> + switch (type) {
>> + case OVL_FEATURE_COMPAT:
>> + features = &upper_layer->compat;
>> + break;
>> + case OVL_FEATURE_RO_COMPAT:
>> + features = &upper_layer->ro_compat;
>> + break;
>> + case OVL_FEATURE_INCOMPAT:
>> + features = &upper_layer->incompat;
>> + break;
>> + default:
>> + return -EINVAL;
>> + }
>> +
>> + spin_lock(&upper_layer->lock);
>> + if ((*features) & mask) {
>> + spin_unlock(&upper_layer->lock);
>> + return 0;
>> + }
>> +
>> + (*features) |= mask;
>> + spin_unlock(&upper_layer->lock);
>> +
>
> Seems like you can move the above test and set into
> ovl_set_feature() and you can also use upperdir->d_lock
> and don't need to introduce a new spinlock for this.
>
Also, since there is no clear bit functionality, it is worth
testing the bit before the lock and then repeat the test under
lock if first test fails.
Thanks,
Amir.
--
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
