Hi Akhil, Some replies inline.
Thanks, Ciara >-----Original Message----- >From: Akhil Goyal <gak...@marvell.com> >Sent: Monday 7 February 2022 09:05 >To: Power, Ciara <ciara.po...@intel.com>; dev@dpdk.org; >tho...@monjalon.net >Cc: Zhang, Roy Fan <roy.fan.zh...@intel.com>; Anoob Joseph ><ano...@marvell.com>; m...@ashroe.eu; Doherty, Declan ><declan.dohe...@intel.com> >Subject: RE: [EXT] [PATCH v3 4/4] crypto: modify return value for asym session >create > >> diff --git a/doc/guides/prog_guide/cryptodev_lib.rst >> b/doc/guides/prog_guide/cryptodev_lib.rst >> index 62bd3577f5..8e16461dc6 100644 >> --- a/doc/guides/prog_guide/cryptodev_lib.rst >> +++ b/doc/guides/prog_guide/cryptodev_lib.rst >> @@ -1236,10 +1236,10 @@ crypto operations is similar except change to >> respective op and xform setup). >> * Create asym crypto session and initialize it for the crypto device. >> * The session structure is hidden from the app, so void * is used. >> */ >> - void *asym_session; >> - asym_session = >> rte_cryptodev_asym_session_create(asym_session_pool, >> + void *asym_session = NULL; >> + ret = rte_cryptodev_asym_session_create(&asym_session, >> asym_session_pool, >> cdev_id, &modex_xform); >> - if (asym_session == NULL) >> + if (ret < 0) >> rte_exit(EXIT_FAILURE, "Session could not be created\n"); > >Sample Code in the rst files is no more added. @Thomas: Could you please >confirm? >Probably a separate patch is required to clean this up. > [CP] I see Thomas replied on this - thanks. Will try find a section of example/test code that does the same thing as being shown here. >> diff --git a/lib/cryptodev/rte_cryptodev.c >> b/lib/cryptodev/rte_cryptodev.c index 0d816ed4a9..005f0e7952 100644 >> --- a/lib/cryptodev/rte_cryptodev.c >> +++ b/lib/cryptodev/rte_cryptodev.c >> @@ -1912,9 +1912,9 @@ rte_cryptodev_sym_session_create(struct >> rte_mempool *mp) >> return sess; >> } >> >> -void * >> -rte_cryptodev_asym_session_create(struct rte_mempool *mp, uint8_t >> dev_id, >> - struct rte_crypto_asym_xform *xforms) >> +int >> +rte_cryptodev_asym_session_create(void **session, struct rte_mempool >> *mp, >> + uint8_t dev_id, struct rte_crypto_asym_xform *xforms) > >Do you really need a double pointer for the session handle? > [CP] Yes I believe so, the return value used to be session, but now that we have an int return value, the session needs to be passed in as a parameter somehow. We need the double pointer because we need the call to rte_mempool_get() to set the original session pointer that can be accessed outside of this function, rather than just the local copy if it were a singular session pointer passed in as a parameter. <snip>
