RE: [PATCH v4 2/6] crypto/scheduler: use unified session

[Akhil Goyal <gakhil@marvell.com>]

> Have the sym sessions changes been tested with the dpdk-test-crypto-perf tool ?

I have not tested for scheduler pmd. Can you root cause the issue and send a fix? 
Fan may have tested it.

> 
> root@silpixa00401033:build# ./app/dpdk-test-crypto-perf -l 3,4 --socket-mem
> 4096,0 -a 0000:33:01.0,qat_sym_cipher_crc_enable=1 --vdev crypto_aesni_mb1
> --vdev
> "crypto_scheduler,worker=crypto_aesni_mb1,worker=0000:33:01.0_qat_sym,m
> ode=packet-size-distr,ordering=disable,mode_param=threshold:64" -n 6 --force-
> max-simd-bitwidth=512 -- --ptest throughput --silent --total-ops 3000000 --
> burst-sz 32 --buffer-sz 105,277,1301  --imix 15,10,75 --devtype crypto_scheduler
> --optype cipher-only --cipher-algo aes-docsisbpi --cipher-iv-sz 16 --cipher-op
> encrypt --cipher-key-sz 16 --docsis-hdr-sz 17
> EAL: Detected CPU lcores: 128
> EAL: Detected NUMA nodes: 2
> EAL: Detected static linkage of DPDK
> EAL: Multi-process socket /var/run/dpdk/rte/mp_socket
> EAL: Selected IOVA mode 'PA'
> EAL: VFIO support initialized
> EAL: Using IOMMU type 1 (Type 1)
> EAL: Probe PCI driver: qat (8086:37c9) device: 0000:33:01.0 (socket 0)
> CRYPTODEV: Creating cryptodev 0000:33:01.0_qat_sym
> 
> CRYPTODEV: Initialisation parameters - name: 0000:33:01.0_qat_sym,socket id:
> 0, max queue pairs: 0
> CRYPTODEV: Creating cryptodev 0000:33:01.0_qat_asym
> 
> CRYPTODEV: Initialisation parameters - name: 0000:33:01.0_qat_asym,socket
> id: 0, max queue pairs: 0
> CRYPTODEV: Creating cryptodev crypto_aesni_mb1
> 
> CRYPTODEV: Initialisation parameters - name: crypto_aesni_mb1,socket id: 0,
> max queue pairs: 8
> ipsec_mb_create() line 152: IPSec Multi-buffer library version used: 1.2.0
> 
> CRYPTODEV: Creating cryptodev crypto_scheduler
> 
> CRYPTODEV: Initialisation parameters - name: crypto_scheduler,socket id: 0,
> max queue pairs: 8
> cryptodev_scheduler_create() line 138:   Scheduling mode = packet-size-distr
> PMD:   Sched mode param (threshold = 64)
> cryptodev_scheduler_create() line 193:   Packet ordering = disable
> scheduler_attach_init_worker() line 45: Scheduler crypto_scheduler attached
> worker 0000:33:01.0_qat_sym
> scheduler_attach_init_worker() line 45: Scheduler crypto_scheduler attached
> worker crypto_aesni_mb1
> Allocated pool "sess_mp_0" on socket 0
> USER1: Test run constructor failed
> 
> > -----Original Message-----
> > From: Akhil Goyal <gakhil@marvell.com>
> > Sent: Monday, September 26, 2022 8:14 PM
> > To: dev@dpdk.org
> > Cc: thomas@monjalon.net; david.marchand@redhat.com;
> > hemant.agrawal@nxp.com; vattunuru@marvell.com;
> > ferruh.yigit@xilinx.com; andrew.rybchenko@oktetlabs.ru;
> > konstantin.v.ananyev@yandex.ru; jiawenwu@trustnetic.com;
> > yisen.zhuang@huawei.com; irusskikh@marvell.com; jerinj@marvell.com;
> > adwivedi@marvell.com; maxime.coquelin@redhat.com; chandu@amd.com;
> > ruifeng.wang@arm.com; ajit.khaparde@broadcom.com;
> > anoobj@marvell.com; De Lara Guarch, Pablo
> > <pablo.de.lara.guarch@intel.com>; matan@nvidia.com; g.singh@nxp.com;
> > Yang, Qiming <qiming.yang@intel.com>; Wu, Wenjun1
> > <wenjun1.wu@intel.com>; jianwang@trustnetic.com; Wu, Jingjing
> > <jingjing.wu@intel.com>; Xing, Beilei <beilei.xing@intel.com>;
> > ndabilpuram@marvell.com; Zhang, Roy Fan <roy.fan.zhang@intel.com>;
> > Akhil Goyal <gakhil@marvell.com>; Ji, Kai <kai.ji@intel.com>; Coyle, David
> > <david.coyle@intel.com>; O'Sullivan, Kevin <kevin.osullivan@intel.com>
> > Subject: [PATCH v4 2/6] crypto/scheduler: use unified session
> >
> > From: Fan Zhang <roy.fan.zhang@intel.com>
> >
> > This patch updates the scheduler PMD to use unified session data structure.
> > Previously thanks to the private session array in cryptodev sym session there
> > are no necessary change needed for scheduler PMD other than the way ops
> > are enqueued/dequeued. The patch inherits the same design in the original
> > session data structure to the scheduler PMD so the cryptodev sym session
> > can be as a linear buffer for both session header and driver private data.
> >
> > With the change there are inevitable extra cost on both memory
> > (64 bytes per session per driver type) and cycle count (set the correct session
> > for each cop based on the worker before enqueue, and retrieve the original
> > session after dequeue).
> >
> > Signed-off-by: Fan Zhang <roy.fan.zhang@intel.com>
> > Signed-off-by: Akhil Goyal <gakhil@marvell.com>
> > Acked-by: Kai Ji <kai.ji@intel.com>
> > Tested-by: Gagandeep Singh <g.singh@nxp.com>
> > Tested-by: David Coyle <david.coyle@intel.com>
> > Tested-by: Kevin O'Sullivan <kevin.osullivan@intel.com>
> > ---