[PATCH v2 0/4] crypto/ccp cleanup

[Uttarwar, Sunil Prakashrao]

[AMD Official Use Only - General]

Hi David,

Please find the below update.

Thanks
Sunil

-----Original Message-----
From: David Marchand <david.marchand at redhat.com> 
Sent: Wednesday, January 11, 2023 8:44 PM
To: Uttarwar, Sunil Prakashrao <SunilPrakashrao.Uttarwar at amd.com>
Cc: Yigit, Ferruh <Ferruh.Yigit at amd.com>; Akhil Goyal <gakhil at marvell.com>; Namburu, Chandu-babu <chandu at amd.com>; Sebastian, Selwin <Selwin.Sebastian at amd.com>; dev <dev at dpdk.org>; Thomas Monjalon <thomas at monjalon.net>
Subject: Re: [PATCH v2 0/4] crypto/ccp cleanup

Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.


On Fri, Nov 18, 2022 at 12:58 PM Uttarwar, Sunil Prakashrao <SunilPrakashrao.Uttarwar at amd.com> wrote:
> Hi David,
>
> Please find the below update
>
> - only one DPDK application can use ccp crypto engines (PCI bus 
> allow/blocklist is not respected, right?), Yes, only one crypto device can be used in a DPDK application for the crypto operations. This is introduced from the patch crypto/ccp: convert driver from vdev to PCI. This is implemented as per community suggestion.

Community suggested to have this driver a standard PCI driver, not a vdev one.
I don't remember anything about the limitation.
Can you point at the discussion that leaded to this?

Sunil -> Please find the discussion about CCP driver as a PCI driver @ https://patches.dpdk.org/project/dpdk/patch/20201225080358.366162-1-asomalap@amd.com/

>
> - since only one crypto device is exposed, there is no way for the application to dedicate/decide how to distribute crypto operations over the different ccp crypto engines available on the system.
>
> When there is no ccp device passed from the application dpdk-test-crypto-perf, it tries to probe all CCP devices present on a system and only one device can be used. It seems this is bug in the patch implemented for crypto/ccp: convert driver from vdev to PCI and we are looking into this.

Indeed.


So how should we proceed?
Patches 1 to 3 are ready and can be merged.

I don't mind dropping patch 4 if you have a better solution/alternative.

Sunil->

As mentioned earlier, there is some issue with the patch https://patches.dpdk.org/project/dpdk/patch/20221004095132.198777-5-david.marchand@redhat.com/.

I worked on this patch and found issue. Did change in this patch and observing no issues. 

Still need some more time test, confirm changes and will share.

For now, we can merge below changes, I will ack it.
https://patches.dpdk.org/project/dpdk/patch/20221004095132.198777-2-david.marchand@redhat.com/
https://patches.dpdk.org/project/dpdk/patch/20221004095132.198777-3-david.marchand@redhat.com/
https://patches.dpdk.org/project/dpdk/patch/20221004095132.198777-4-david.marchand@redhat.com/

We will merge the below patch once we fix it with modifications 
https://patches.dpdk.org/project/dpdk/patch/20221004095132.198777-5-david.marchand@redhat.com/

Please let me know if any suggestions.

--
David Marchand