Re: [PATCH V10 10/11] fs: Introduce DCACHE_DONTCACHE

On Fri, Apr 24, 2020 at 08:57:34AM +1000, Dave Chinner wrote:

On Wed, Apr 22, 2020 at 02:21:01PM -0700, ira.weiny@intel.com wrote:

quoted

From: Ira Weiny <redacted>

DCACHE_DONTCACHE indicates a dentry should not be cached on final
dput().

Also add a helper function to mark DCACHE_DONTCACHE on all dentries
pointing to a specific inode when that inode is being set I_DONTCACHE.

This facilitates dropping dentry references to inodes sooner which
require eviction to swap S_DAX mode.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Ira Weiny <redacted>

Code looks fine....

quoted

--- a/fs/inode.c
+++ b/fs/inode.c

@@ -1526,6 +1526,21 @@ int generic_delete_inode(struct inode *inode)
 }
 EXPORT_SYMBOL(generic_delete_inode);
 
+void mark_inode_dontcache(struct inode *inode)
+{
+	struct dentry *de;
+
+	spin_lock(&inode->i_lock);
+	hlist_for_each_entry(de, &inode->i_dentry, d_u.d_alias) {
+		spin_lock(&de->d_lock);
+		de->d_flags |= DCACHE_DONTCACHE;
+		spin_unlock(&de->d_lock);
+	}
+	inode->i_state |= I_DONTCACHE;
+	spin_unlock(&inode->i_lock);
+}
+EXPORT_SYMBOL(mark_inode_dontcache);

Though I suspect that this should be in fs/dcache.c and not
fs/inode.c. i.e. nothing in fs/inode.c does dentry list walks, but
there are several cases in the dcache code where inode dentry walks
are done under the inode lock (e.g. d_find_alias(inode)).

So perhaps this should be d_mark_dontcache(inode), which also marks
the inode as I_DONTCACHE so that everything is evicted on last
reference...

That does follow an existing pattern.

Al?  Any preference?

Ira

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com