[dpdk-dev] [PATCH v4 1/3] compress/isal: enable checksum support in driver

Varghese, Vipin vipin.varghese at intel.com
Mon Dec 17 13:33:04 CET 2018


Thanks for the update

> -----Original Message-----
> From: Daly, Lee
> Sent: Monday, December 17, 2018 2:50 PM
> To: Varghese, Vipin <vipin.varghese at intel.com>; akhil.goyal at nxp.com
> Cc: dev at dpdk.org; Trahe, Fiona <fiona.trahe at intel.com>
> Subject: RE: [dpdk-dev] [PATCH v4 1/3] compress/isal: enable checksum
> support in driver
> 
> Hi Vipin, thanks for the question.
> 
> > -----Original Message-----
> > From: Varghese, Vipin
> > Sent: Monday, December 17, 2018 8:20 AM
> > To: Daly, Lee <lee.daly at intel.com>; akhil.goyal at nxp.com
> > Cc: dev at dpdk.org; Daly, Lee <lee.daly at intel.com>
> > Subject: RE: [dpdk-dev] [PATCH v4 1/3] compress/isal: enable checksum
> > support in driver
> >
> > Hi Lee,
> >
> > Apologies if the logic is already done for my query
> >
> > snipped
> > > +		/* Set private xform checksum */
> > > +		switch (xform->compress.chksum) {
> > > +		/* Raw deflate by default */
> > Does the user have to choose RTE_COMP_CHECKSUM_NONE while creating
> > compress/isal instance?
> 
> At the moment, yes the application must fill out the xform/instance.
> >
> > > +		case(RTE_COMP_CHECKSUM_NONE):
> > > +			priv_xform->compress.chksum = IGZIP_DEFLATE;
> > > +			break;
> > > +		case(RTE_COMP_CHECKSUM_CRC32):
> > > +			priv_xform->compress.chksum =
> > IGZIP_GZIP_NO_HDR;
> > > +			break;
> > > +		case(RTE_COMP_CHECKSUM_ADLER32):
> > > +			priv_xform->compress.chksum =
> > IGZIP_ZLIB_NO_HDR;
> > > +			break;
> > > +		case(RTE_COMP_CHECKSUM_CRC32_ADLER32):
> > > +			ISAL_PMD_LOG(ERR, "Combined CRC and ADLER
> > > checksum not"
> > > +					" supported\n");
> > > +			return -ENOTSUP;
> > > +		default:
> > > +			ISAL_PMD_LOG(ERR, "Checksum type not
> > > supported\n");
> > > +			return -ENOTSUP;
> > Do we not fall back to RTE_COMP_CHECKSUM_NONE if the configuration is
> > wrong and report warning?
> >
> 
> Right now, the xform must be filled out correctly before compression can be
> executed.
> Perhaps we could do as you say and fall back to a default config if the non-
> essential params are incorrect.
> > snipped


More information about the dev mailing list