> On Oct 31, 2022, at 1:36 PM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
>
> On Mon, 2022-10-31 at 17:28 +0000, Chuck Lever III wrote:
>>
>>> On Oct 31, 2022, at 12:43 PM, Jeff Layton <jlayton@xxxxxxxxxx> wrote:
>>>
>>> On Fri, 2022-10-28 at 10:47 -0400, Chuck Lever wrote:
>>>> Remove the call to find_file_locked() in insert_nfs4_file(). Tracing
>>>> shows that over 99% of these calls return NULL. Thus it is not worth
>>>> the expense of the extra bucket list traversal. insert_file() already
>>>> deals correctly with the case where the item is already in the hash
>>>> bucket.
>>>>
>>>> Since nfsd4_file_hash_insert() is now just a wrapper around
>>>> insert_file(), move the meat of insert_file() into
>>>> nfsd4_file_hash_insert() and get rid of it.
>>>>
>>>> Signed-off-by: Chuck Lever <chuck.lever@xxxxxxxxxx>
>>>> Reviewed-by: NeilBrown <neilb@xxxxxxx>
>>>> ---
>>>> fs/nfsd/nfs4state.c | 64 ++++++++++++++++++++++-----------------------------
>>>> 1 file changed, 28 insertions(+), 36 deletions(-)
>>>>
>>>> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
>>>> index 504455cb80e9..b1988a46fb9b 100644
>>>> --- a/fs/nfsd/nfs4state.c
>>>> +++ b/fs/nfsd/nfs4state.c
>>>> @@ -4683,24 +4683,42 @@ find_file_locked(const struct svc_fh *fh, unsigned int hashval)
>>>> return NULL;
>>>> }
>>>>
>>>> -static struct nfs4_file *insert_file(struct nfs4_file *new, struct svc_fh *fh,
>>>> - unsigned int hashval)
>>>> +static struct nfs4_file * find_file(struct svc_fh *fh)
>>>> {
>>>> struct nfs4_file *fp;
>>>> + unsigned int hashval = file_hashval(fh);
>>>> +
>>>> + rcu_read_lock();
>>>> + fp = find_file_locked(fh, hashval);
>>>> + rcu_read_unlock();
>>>> + return fp;
>>>> +}
>>>> +
>>>> +/*
>>>> + * On hash insertion, identify entries with the same inode but
>>>> + * distinct filehandles. They will all be in the same hash bucket
>>>> + * because nfs4_file's are hashed by the address in the fi_inode
>>>> + * field.
>>>> + */
>>>> +static noinline_for_stack struct nfs4_file *
>>>> +nfsd4_file_hash_insert(struct nfs4_file *new, const struct svc_fh *fhp)
>>>> +{
>>>> + unsigned int hashval = file_hashval(fhp);
>>>> struct nfs4_file *ret = NULL;
>>>> bool alias_found = false;
>>>> + struct nfs4_file *fi;
>>>>
>>>> spin_lock(&state_lock);
>>>> - hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash,
>>>> + hlist_for_each_entry_rcu(fi, &file_hashtbl[hashval], fi_hash,
>>>> lockdep_is_held(&state_lock)) {
>>>> - if (fh_match(&fp->fi_fhandle, &fh->fh_handle)) {
>>>> - if (refcount_inc_not_zero(&fp->fi_ref))
>>>> - ret = fp;
>>>> - } else if (d_inode(fh->fh_dentry) == fp->fi_inode)
>>>> - fp->fi_aliased = alias_found = true;
>>>> + if (fh_match(&fi->fi_fhandle, &fhp->fh_handle)) {
>>>> + if (refcount_inc_not_zero(&fi->fi_ref))
>>>> + ret = fi;
>>>> + } else if (d_inode(fhp->fh_dentry) == fi->fi_inode)
>>>> + fi->fi_aliased = alias_found = true;
>>>
>>> Would it not be better to do the inode comparison first? That's just a
>>> pointer check vs. a full memcmp. That would allow you to quickly rule
>>> out any entries that match different inodes but that are on the same
>>> hash chain.
>>
>> Marginally better: The usual case after the rhltable changes are
>> applied is that the returned list contains only entries that match
>> d_inode(fhp->fh_dentry), and usually that list has just one entry
>> in it that matches the FH.
>>
>
> That depends entirely on workload. Given that you start with 512
> buckets, you'd need to be working with at least that many inodes
> concurrently to make that happen, but it could easily happen with a
> large enough working set.
I take it back. Neil pointed this out during a recent review: the
rhltable_lookup() will return a list that contains _only_ items that
match the inode. So that check is no longer needed at all. Patch
14/14 has this hunk:
+ list = rhltable_lookup(&nfs4_file_rhltable, &inode,
+ nfs4_file_rhash_params);
+ rhl_for_each_entry_rcu(fi, tmp, list, fi_rlist) {
if (fh_match(&fi->fi_fhandle, &fhp->fh_handle)) {
if (refcount_inc_not_zero(&fi->fi_ref))
ret = fi;
- } else if (d_inode(fhp->fh_dentry) == fi->fi_inode)
+ } else
fi->fi_aliased = alias_found = true;
}
which removes that check.
>> Since 11/14 is billed as a clean-up, I'd prefer to put that kind
>> of optimization in a separate patch that can be mechanically
>> reverted if need be. I'll post something that goes on top of this
>> series.
>>
>
> Fair enough. You can add my Reviewed-by: and we can address it later.
>
>>
>>>> }
>>>> if (likely(ret == NULL)) {
>>>> - nfsd4_file_init(fh, new);
>>>> + nfsd4_file_init(fhp, new);
>>>> hlist_add_head_rcu(&new->fi_hash, &file_hashtbl[hashval]);
>>>> new->fi_aliased = alias_found;
>>>> ret = new;
>>>> @@ -4709,32 +4727,6 @@ static struct nfs4_file *insert_file(struct nfs4_file *new, struct svc_fh *fh,
>>>> return ret;
>>>> }
>>>>
>>>> -static struct nfs4_file * find_file(struct svc_fh *fh)
>>>> -{
>>>> - struct nfs4_file *fp;
>>>> - unsigned int hashval = file_hashval(fh);
>>>> -
>>>> - rcu_read_lock();
>>>> - fp = find_file_locked(fh, hashval);
>>>> - rcu_read_unlock();
>>>> - return fp;
>>>> -}
>>>> -
>>>> -static struct nfs4_file *
>>>> -find_or_add_file(struct nfs4_file *new, struct svc_fh *fh)
>>>> -{
>>>> - struct nfs4_file *fp;
>>>> - unsigned int hashval = file_hashval(fh);
>>>> -
>>>> - rcu_read_lock();
>>>> - fp = find_file_locked(fh, hashval);
>>>> - rcu_read_unlock();
>>>> - if (fp)
>>>> - return fp;
>>>> -
>>>> - return insert_file(new, fh, hashval);
>>>> -}
>>>> -
>>>> static noinline_for_stack void nfsd4_file_hash_remove(struct nfs4_file *fi)
>>>> {
>>>> hlist_del_rcu(&fi->fi_hash);
>>>> @@ -5625,7 +5617,7 @@ nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nf
>>>> * and check for delegations in the process of being recalled.
>>>> * If not found, create the nfs4_file struct
>>>> */
>>>> - fp = find_or_add_file(open->op_file, current_fh);
>>>> + fp = nfsd4_file_hash_insert(open->op_file, current_fh);
>>>> if (fp != open->op_file) {
>>>> status = nfs4_check_deleg(cl, open, &dp);
>>>> if (status)
>>>>
>>>>
>>>
>>> --
>>> Jeff Layton <jlayton@xxxxxxxxxx>
>>
>> --
>> Chuck Lever
>>
>>
>>
>
> --
> Jeff Layton <jlayton@xxxxxxxxxx>
--
Chuck Lever
