| Message ID | 20250213122137.11830-1-alexjlzheng@tencent.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | fs: remove useless d_unhashed() retry in d_alloc_parallel() | expand |
On Thu, Feb 13, 2025 at 08:21:37PM +0800, Jinliang Zheng wrote: > After commit 45f78b0a2743 ("fs/dcache: Move the wakeup from > __d_lookup_done() to the caller."), we will only wake up > d_wait_lookup() after adding dentry to dentry_hashtable. Not true. d_lookup_done() might be called without having *ever* hashed the sucker. Just think for a moment - what, for example, should happen if ->lookup() fails? Would you have d_alloc_parallel() coming during that ->lookup() (while dentry is in in-lookup hash) hang forever? NAK.
On Thu, 13 Feb 2025 13:20:06 +0000, Al Viro wrote: > On Thu, Feb 14, 2025 at 08:21:37PM +0800, Jinliang Zheng wrote: > > After commit 45f78b0a2743 ("fs/dcache: Move the wakeup from > > __d_lookup_done() to the caller."), we will only wake up > > d_wait_lookup() after adding dentry to dentry_hashtable. > > Not true. d_lookup_done() might be called without having > *ever* hashed the sucker. > > Just think for a moment - what, for example, should happen > if ->lookup() fails? Would you have d_alloc_parallel() > coming during that ->lookup() (while dentry is in in-lookup > hash) hang forever? Haha, yes, my bad. Thanks for pointing out my mistake. Thank you, :) Jinliang Zheng > > NAK.
diff --git a/fs/dcache.c b/fs/dcache.c index e3634916ffb9..543833eedd8c 100644 --- a/fs/dcache.c +++ b/fs/dcache.c @@ -2584,8 +2584,6 @@ struct dentry *d_alloc_parallel(struct dentry *parent, goto mismatch; if (unlikely(dentry->d_parent != parent)) goto mismatch; - if (unlikely(d_unhashed(dentry))) - goto mismatch; if (unlikely(!d_same_name(dentry, parent, name))) goto mismatch; /* OK, it *is* a hashed match; return it */
After commit 45f78b0a2743 ("fs/dcache: Move the wakeup from __d_lookup_done() to the caller."), we will only wake up d_wait_lookup() after adding dentry to dentry_hashtable. Therefore, there is no need to keep retries about d_unhashed() in d_alloc_parallel(). Signed-off-by: Jinliang Zheng <alexjlzheng@tencent.com> --- fs/dcache.c | 2 -- 1 file changed, 2 deletions(-)