Re: [PATCH v2 1/9] RDMA/rxe: Move ICRC checking to a subroutine

[Jason Gunthorpe <jgg@nvidia.com>]

On Tue, Jul 06, 2021 at 11:00:33PM -0500, Bob Pearson wrote:
> Move the code in rxe_recv() that checks the ICRC on incoming packets to a
> subroutine rxe_check_icrc() and move that to rxe_icrc.c.
> 
> Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
> Reviewed-by: Zhu Yanjun <zyjzyj2000@gmail.com>
>  drivers/infiniband/sw/rxe/rxe_icrc.c | 38 ++++++++++++++++++++++++++++
>  drivers/infiniband/sw/rxe/rxe_loc.h  |  2 ++
>  drivers/infiniband/sw/rxe/rxe_recv.c | 23 ++---------------
>  3 files changed, 42 insertions(+), 21 deletions(-)
> 
> diff --git a/drivers/infiniband/sw/rxe/rxe_icrc.c b/drivers/infiniband/sw/rxe/rxe_icrc.c
> index 66b2aad54bb7..d067841214be 100644
> +++ b/drivers/infiniband/sw/rxe/rxe_icrc.c
> @@ -67,3 +67,41 @@ u32 rxe_icrc_hdr(struct rxe_pkt_info *pkt, struct sk_buff *skb)
>  			rxe_opcode[pkt->opcode].length - RXE_BTH_BYTES);
>  	return crc;
>  }
> +
> +/**
> + * rxe_icrc_check() - Compute ICRC for a packet and compare to the ICRC
> + *		      delivered in the packet.
> + * @skb: packet buffer
> + * @pkt: packet info
> + *
> + * Return: 0 if the values match else an error
> + */
> +int rxe_icrc_check(struct sk_buff *skb, struct rxe_pkt_info *pkt)
> +{
> +	__be32 *icrcp;
> +	u32 pkt_icrc;
> +	u32 icrc;
> +
> +	icrcp = (__be32 *)(pkt->hdr + pkt->paylen - RXE_ICRC_SIZE);
> +	pkt_icrc = be32_to_cpu(*icrcp);
> +
> +	icrc = rxe_icrc_hdr(pkt, skb);
> +	icrc = rxe_crc32(pkt->rxe, icrc, (u8 *)payload_addr(pkt),
> +				payload_size(pkt) + bth_pad(pkt));
> +	icrc = (__force u32)cpu_to_be32(~icrc);

This isn't right, the return value of rxe_crc32 is a __b32 so icrc
here should be a __b32 too

> +	if (unlikely(icrc != pkt_icrc)) {

And this should be written

if (unlikely(be32_to_cpu(~icrc) != pkt_icrc)) {

Or, alternatively you can skip both of the be32_to_cpu's and just
directly compare the __be32 types

Also, I'm not sure the rxe_crc32 byteswapping is right either, the
crc32 shash implementation is storing a __le32 in the shash_desc_ctx()
but this code reads it out as a __be32 ? On BE that will gain a
byteswap, is it OK?

Jason