Hi Neil,
On 8/7/24 07:58, NeilBrown wrote:
On Mon, 05 Aug 2024, Guoqing Jiang wrote:
It is not needed given cache_check already calls cache_put on failure,
otherwise we call cache_put twice in case of -ETIMEDOUT.
cache_check() was called on a different *item.
This cache_put() puts the *item returned by lookup_fn() two lines
earlier.
The current code is correct.
Thanks for point it out! So this cache_put is paired with lookup_fn which
calls cache_get on a new item.
BTW, I checked idmap_id_to_name which didn't put cache if there is no
enough buffer, not sure if we need it here.
diff --git a/fs/nfsd/nfs4idmap.c b/fs/nfsd/nfs4idmap.c
index 7a806ac13e31..7abddf7d8f6d 100644
--- a/fs/nfsd/nfs4idmap.c
+++ b/fs/nfsd/nfs4idmap.c
@@ -594,8 +594,10 @@ static __be32 idmap_id_to_name(struct xdr_stream *xdr,
ret = strlen(item->name);
WARN_ON_ONCE(ret > IDMAP_NAMESZ);
p = xdr_reserve_space(xdr, ret + 4);
- if (!p)
+ if (!p) {
+ cache_put(&item->h, nn->idtoname_cache);
return nfserr_resource;
+ }
p = xdr_encode_opaque(p, item->name, ret);
cache_put(&item->h, nn->idtoname_cache);
return 0;
Thanks,
Guoqing