[dpdk-dev] [RFC v2, 1/2] cryptodev: add support to set session private data

[Gujjar, Abhinandan S]

Hi Pablo & Declan,

> -----Original Message-----
> From: De Lara Guarch, Pablo
> Sent: Thursday, January 25, 2018 1:17 AM
> To: Gujjar, Abhinandan S <abhinandan.gujjar at intel.com>; Doherty, Declan
> <declan.doherty at intel.com>; akhil.goyal at nxp.com;
> Jerin.JacobKollanukkaran at cavium.com
> Cc: dev at dpdk.org; Vangati, Narender <narender.vangati at intel.com>; Rao,
> Nikhil <nikhil.rao at intel.com>
> Subject: RE: [RFC v2, 1/2] cryptodev: add support to set session private data
> 
> 
> 
> > -----Original Message-----
> > From: Gujjar, Abhinandan S
> > Sent: Tuesday, January 23, 2018 8:54 AM
> > To: Doherty, Declan <declan.doherty at intel.com>; akhil.goyal at nxp.com;
> > De Lara Guarch, Pablo <pablo.de.lara.guarch at intel.com>;
> > Jerin.JacobKollanukkaran at cavium.com
> > Cc: dev at dpdk.org; Vangati, Narender <narender.vangati at intel.com>;
> > Gujjar, Abhinandan S <abhinandan.gujjar at intel.com>; Rao, Nikhil
> > <nikhil.rao at intel.com>
> > Subject: [RFC v2, 1/2] cryptodev: add support to set session private
> > data
> >
> > Update rte_crypto_op to indicate private data offset.
> >
> > The application may want to store private data along with the
> > rte_cryptodev that is transparent to the rte_cryptodev layer.
> > For e.g., If an eventdev based application is submitting a
> > rte_cryptodev_sym_session operation and wants to indicate event
> > information required to construct a new event that will be enqueued to
> > eventdev after completion of the rte_cryptodev_sym_session operation.
> > This patch provides a mechanism for the application to associate this
> > information with the rte_cryptodev_sym_session session.
> > The application can set the private data using
> > rte_cryptodev_sym_session_set_private_data() and retrieve it using
> > rte_cryptodev_sym_session_get_private_data().
> 
> Hi Abhinandan,
> 
> >
> > Signed-off-by: Abhinandan Gujjar <abhinandan.gujjar at intel.com>
> > Signed-off-by: Nikhil Rao <nikhil.rao at intel.com>
> > ---
> > Notes:
> >         V2:
> > 	1. Removed enum rte_crypto_op_private_data_type
> > 	2. Corrected formatting
> >
> >  lib/librte_cryptodev/rte_crypto.h    |  8 ++++++--
> >  lib/librte_cryptodev/rte_cryptodev.h | 32
> > ++++++++++++++++++++++++++++++++
> >  2 files changed, 38 insertions(+), 2 deletions(-)
> >
> > diff --git a/lib/librte_cryptodev/rte_crypto.h
> > b/lib/librte_cryptodev/rte_crypto.h
> > index 95cf861..14c87c8 100644
> > --- a/lib/librte_cryptodev/rte_crypto.h
> > +++ b/lib/librte_cryptodev/rte_crypto.h
> > @@ -84,8 +84,12 @@ struct rte_crypto_op {
> >  	 */
> >  	uint8_t sess_type;
> >  	/**< operation session type */
> > -
> > -	uint8_t reserved[5];
> > +	uint16_t private_data_offset;
> > +	/**< Offset to indicate start of private data (if any). The private
> > +	 * data may be used by the application to store information which
> > +	 * should remain untouched in the library/driver
> 
> Is this the offset for the private data after the crypto operation?
Yes. This is private date is meant for sessionless case.
> From your title, it looks like it is for the session private data, but then, this
> shouldn't be here.
Agree.
> If it is for the crypto operation, I suggest you to separate it in another patch.
> Also, you should indicate where the offset starts from. For the IV, the offset is
> counted from the start of the rte_crypto_op, so I think it should be the same, to
> keep consistency.
Sure. I will make a separate patch for this changes. Add some more information to make it clear.
> 
> For the session private data, we see two options:
> 
> 1 - Add a  "valid" private data field in the rte_cryptodev_sym_session structure,
> so when it is set, it indicates that the session contains private data (a single bit
> would be enough, 1 to indicate there is, and 0 to indicate there is not).
> This would go into the beginning of the structure, so this would require an ABI
> deprecation notice.
> This also assumes that the private data starts just after the session header
> 
> 2 -  Do not add an extra "valid" private data field in rte_cryptodev_sym_session
> structure, and add a small header in the private data, which contains the "valid"
> bit.
> Then, when calling sym_session_get_private_data, this bit should be checked.
> Note that the object that holds the session structure needs to be big enough to
> hold this value.
> If the object has only space for the sess_private_data array, then the session has
> no private data.
> Therefore, this approach might be less performant, but with no ABI deprecation
> required.
I am with option 2 with slight changes as below:
rte_cryptodev_sym_session_create() will have a flag as below
indicating private data exits or not.
{ 
- memset(sess, 0, (sizeof(void *) * nb_drivers));
+memset(sess, 0, (sizeof(void *) * nb_drivers ) + sizeof(private_data_flag));
}
and in
rte_cryptodev_get_header_session_size(void)
{
  /*
   * Header contains pointers to the private data
   * of all registered drivers
   */
  -return (sizeof(void *) * nb_drivers);
  +return ((sizeof(void *) * nb_drivers) + sizeof(private_data_flag));
}
With this, a flag indicating private data exists or not will always have valid value.

> 
> I would recommend you to send a deprecation notice for option 1, then check
> the performance of both option, and if needed, make the change in the
> structure, in 18.05.
> 
> Regards,
> Pablo