On 6/11/25 12:24 PM, Sergey Bashirov wrote: > I also have some doubts about this code: > if (xdr_stream_decode_u64(&xdr, &bex.len)) > return -NFS4ERR_BADXDR; > if (bex.len & (block_size - 1)) > return -NFS4ERR_BADXDR; > > The first error code is clear to me, it is all about decoding. But should > not we return -NFS4ERR_EINVAL in the second check? On one hand, we > encountered an invalid value after successful decoding, but on the other > hand, we stopped decoding the extent array, so we can say that this is > also a decoding error. On first read of Section 2.3 of RFC 5663, there's no mandated alignment requirement for bex_length. IMO this is a case where the implementation is deciding that a decoded value is not valid, so NFS4ERR_INVAL might be a better choice here. -- Chuck Lever
