[dpdk-dev] [PATCH v4 2/2] test/crypto: add GMAC SGL tests

Akhil Goyal akhil.goyal at nxp.com
Fri Oct 9 20:36:19 CEST 2020


Hi Pablo,

> Hi Akhil,
> 
> > > diff --git a/app/test/test_cryptodev.h b/app/test/test_cryptodev.h
> > > index 41542e055..57003733b 100644
> > > --- a/app/test/test_cryptodev.h
> > > +++ b/app/test/test_cryptodev.h
> > > @@ -17,9 +17,9 @@
> > >  #define DEFAULT_NUM_QPS_PER_QAT_DEVICE  (2)
> > >  #define DEFAULT_BURST_SIZE              (64)
> > >  #define DEFAULT_NUM_XFORMS              (2)
> > > -#define NUM_MBUFS                       (8191)
> > > +#define NUM_MBUFS                       (4095)
> > >  #define MBUF_CACHE_SIZE                 (256)
> > > -#define MBUF_DATAPAYLOAD_SIZE		(2048 +
> > > DIGEST_BYTE_LENGTH_SHA512)
> > > +#define MBUF_DATAPAYLOAD_SIZE		(4096 +
> > > DIGEST_BYTE_LENGTH_SHA512)
> > >  #define MBUF_SIZE			(sizeof(struct rte_mbuf) + \
> > >  		RTE_PKTMBUF_HEADROOM + MBUF_DATAPAYLOAD_SIZE)
> > >
> > Why are these changes done? Any specific requirement for GMAC? I believe
> > 2048 is also good enough.
> 
> I needed to reduce the maximum number of segments for the SGL tests, since
> Intel QAT PMD does not support more than 16, and this way both PMDs can be
> tested with the new tests.
> 
What is your packet size. I believe 16 segments of 2K means 32KB of data which
Should be more than enough for testing. Right? Can we reduce the length of vector?

But by changing the MBUF_DATAPAYLOAD_SIZE to 4K would change the behavior of
Other cases. Isn't it?

Regards,
Akhil




More information about the dev mailing list