On Fri, Mar 31, 2023 at 11:35:42AM +0100, Edward Cree wrote: > On 31/03/2023 07:54, Dan Carpenter wrote: > > Hello Edward Cree, > > > > The patch 2245eb0086d8: "sfc: add functions to insert encap matches > > into the MAE" from Mar 27, 2023, leads to the following Smatch static > > checker warning: > > > > drivers/net/ethernet/sfc/mae.c:1002 efx_mae_register_encap_match() > > warn: this cast is a no-op > ... > > 998 MCDI_STRUCT_SET_WORD_BE(match_crit, MAE_ENC_FIELD_PAIRS_ENC_ETHER_TYPE_BE_MASK, > > 999 ~(__be16)0); > > > > But for these ones the u16 is type promoted to int and so ~0 is negative > > one and will be sign extended. The cast does nothing. > > > > 1000 MCDI_STRUCT_SET_WORD_BE(match_crit, MAE_ENC_FIELD_PAIRS_ENC_L4_DPORT_BE, > > 1001 encap->udp_dport); > > --> 1002 MCDI_STRUCT_SET_WORD_BE(match_crit, MAE_ENC_FIELD_PAIRS_ENC_L4_DPORT_BE_MASK, > > 1003 ~(__be16)0); > > > > Same. > > The casts are there for Sparse reasons (endianness), rather than > particularly caring about the size (although that conveniently > documents to the reader the width of the MCDI field). > We could instead use (__be16)~0 but I believe some versions of > sparse will warn on this (I forget the details). > > To the compiler (i.e. outside of sparse) the cast is indeed a > no-op, but this is fine. I would consider this Smatch warning > a false positive; lmk if you disagree. Fair enough... I didn't know that Sparse printed a warning for this. regards, dan carpenter
