[dpdk-dev] mbuf offload flags

[dpdk-dev] mbuf offload flags

[Thomas Monjalon]

Helin,

In commit http://dpdk.org/browse/dpdk/commit/?id=c22265f6fd4cdc, some fake flags
were added:

#define PKT_RX_EIP_CKSUM_BAD (0ULL << 0)  /**< External IP header checksum error. */                             
#define PKT_RX_OVERSIZE      (0ULL << 0)  /**< Num of desc of an RX pkt oversize. */
#define PKT_RX_HBUF_OVERFLOW (0ULL << 0)  /**< Header buffer overflow. */
#define PKT_RX_RECIP_ERR     (0ULL << 0)  /**< Hardware processing error. */
#define PKT_RX_MAC_ERR       (0ULL << 0)  /**< MAC error. */

Can we remove them?

In a tunnel case, what means PKT_RX_IP_CKSUM_BAD and PKT_RX_L4_CKSUM_BAD?
Inner or outer?
The API comment must be updated.

Re: [dpdk-dev] mbuf offload flags

[Zhang, Helin]

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> Sent: Thursday, July 16, 2015 2:39 AM
> To: Zhang, Helin
> Cc: dev@dpdk.org; olivier.matz@6wind.com
> Subject: mbuf offload flags
> 
> Helin,
> 
> In commit http://dpdk.org/browse/dpdk/commit/?id=c22265f6fd4cdc, some
> fake flags were added:
> 
> #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0)  /**< External IP header
> checksum error. */
> #define PKT_RX_OVERSIZE      (0ULL << 0)  /**< Num of desc of an RX pkt
> oversize. */
> #define PKT_RX_HBUF_OVERFLOW (0ULL << 0)  /**< Header buffer overflow.
> */
> #define PKT_RX_RECIP_ERR     (0ULL << 0)  /**< Hardware processing error.
> */
> #define PKT_RX_MAC_ERR       (0ULL << 0)  /**< MAC error. */
> 
> Can we remove them?
Yes, I agree with you, except PKT_RX_EIP_CKSUM_BAD.

> 
> In a tunnel case, what means PKT_RX_IP_CKSUM_BAD and
> PKT_RX_L4_CKSUM_BAD?
> Inner or outer?
> The API comment must be updated.

Currently PKT_RX_EIP_CKSUM_BAD means outer IP checksum error.
We may need to re-think it?
PKT_RX_IP_CKSUM_BAD for outer for tunnel case, and add a new one for inner IP checksum error case?

For L4, do we need both outer and inner for tunnel case? One might be enough.
We can add one more for L4 checksum error, when it is really needed. For now, I don't see any case.

Regards,
Helin

Re: [dpdk-dev] mbuf offload flags

[Thomas Monjalon]

2015-07-16 15:50, Zhang, Helin:
> From: Thomas Monjalon [mailto:thomas.monjalon@6wind.com]
> > Helin,
> > 
> > In commit http://dpdk.org/browse/dpdk/commit/?id=c22265f6fd4cdc, some
> > fake flags were added:
> > 
> > #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0)  /**< External IP header
> > checksum error. */
> > #define PKT_RX_OVERSIZE      (0ULL << 0)  /**< Num of desc of an RX pkt
> > oversize. */
> > #define PKT_RX_HBUF_OVERFLOW (0ULL << 0)  /**< Header buffer overflow.
> > */
> > #define PKT_RX_RECIP_ERR     (0ULL << 0)  /**< Hardware processing error.
> > */
> > #define PKT_RX_MAC_ERR       (0ULL << 0)  /**< MAC error. */
> > 
> > Can we remove them?
> 
> Yes, I agree with you, except PKT_RX_EIP_CKSUM_BAD.

It is 0. Removing it shouldn't hurt.

> > In a tunnel case, what means PKT_RX_IP_CKSUM_BAD and
> > PKT_RX_L4_CKSUM_BAD?
> > Inner or outer?
> > The API comment must be updated.
> 
> Currently PKT_RX_EIP_CKSUM_BAD means outer IP checksum error.
> We may need to re-think it?
> PKT_RX_IP_CKSUM_BAD for outer for tunnel case, and add a new one for inner IP checksum error case?

Yes.
Maybe that having CKSUM_OK would be better to be sure it has been checked.

> For L4, do we need both outer and inner for tunnel case? One might be enough.
> We can add one more for L4 checksum error, when it is really needed. For now, I don't see any case.

I don't know.