From: "De Lara Guarch, Pablo" <pablo.de.lara.guarch@intel.com>
To: "Ji, Kai" <kai.ji@intel.com>, "dev@dpdk.org" <dev@dpdk.org>
Cc: "gakhil@marvell.com" <gakhil@marvell.com>,
"Burakov, Anatoly" <anatoly.burakov@intel.com>
Subject: RE: [dpdk-dev v3] crypto/ipsec_mb: multi-process IPC request handler
Date: Wed, 26 Oct 2022 08:17:09 +0000 [thread overview]
Message-ID: <DM8PR11MB559190B7454A1569C5DE0D4484309@DM8PR11MB5591.namprd11.prod.outlook.com> (raw)
In-Reply-To: <20221025214800.44779-1-kai.ji@intel.com>
Hi Kai,
A few comments below
Thanks,
Pablo
> -----Original Message-----
> From: Ji, Kai <kai.ji@intel.com>
> Sent: Tuesday, October 25, 2022 10:48 PM
> To: dev@dpdk.org
> Cc: gakhil@marvell.com; Ji, Kai <kai.ji@intel.com>; De Lara Guarch, Pablo
> <pablo.de.lara.guarch@intel.com>; Burakov, Anatoly
> <anatoly.burakov@intel.com>
> Subject: [dpdk-dev v3] crypto/ipsec_mb: multi-process IPC request handler
>
> As the secondary process needs queue-pair to be configured by the primary
> process before use.
This sentence looks incomplete: as the secondary process needs...., then what?
> This patch adds an IPC register function to allow
> secondary process to setup cryptodev queue-pair via IPC messages during
> the runtime. A new "qp_in_used_pid" param stores the PID to provide the
> ownership of the queue-pair so that only the PID matched queue-pair can be
> free'd in the request.
>
> Signed-off-by: Kai Ji <kai.ji@intel.com>
> ---
>
> v3:
> - remove shared memzone as qp_conf params can be passed directly from
> ipc message.
>
> v2:
> - add in shared memzone for data exchange between multi-process
> ---
> drivers/crypto/ipsec_mb/ipsec_mb_ops.c | 130
> ++++++++++++++++++++-
> drivers/crypto/ipsec_mb/ipsec_mb_private.c | 24 +++-
> drivers/crypto/ipsec_mb/ipsec_mb_private.h | 46 ++++++++
> 3 files changed, 196 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> b/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> index cedcaa2742..c29656b746 100644
> --- a/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> +++ b/drivers/crypto/ipsec_mb/ipsec_mb_ops.c
> @@ -3,6 +3,7 @@
> */
>
> #include <string.h>
> +#include <unistd.h>
>
> #include <rte_common.h>
> #include <rte_malloc.h>
> @@ -93,6 +94,46 @@ ipsec_mb_info_get(struct rte_cryptodev *dev,
> }
> }
>
> +static int
> +ipsec_mb_secondary_qp_op(int dev_id, int qp_id,
> + const struct rte_cryptodev_qp_conf *qp_conf,
> + int socket_id, uint16_t pid, enum ipsec_mb_mp_req_type
> op_type) {
> + int ret;
> + struct rte_mp_msg qp_req_msg;
> + struct rte_mp_msg *qp_resp_msg;
> + struct rte_mp_reply qp_resp;
> + struct ipsec_mb_mp_param *req_param;
> + struct ipsec_mb_mp_param *resp_param;
> + struct timespec ts = {.tv_sec = 1, .tv_nsec = 0};
> +
> + memset(&qp_req_msg, 0, sizeof(IPSEC_MB_MP_MSG));
> + memcpy(qp_req_msg.name, IPSEC_MB_MP_MSG,
> sizeof(IPSEC_MB_MP_MSG));
> + req_param = (struct ipsec_mb_mp_param *)&qp_req_msg.param;
> +
> + qp_req_msg.len_param = sizeof(struct ipsec_mb_mp_param);
> + req_param->type = op_type;
> + req_param->dev_id = dev_id;
> + req_param->qp_id = qp_id;
> + req_param->socket_id = socket_id;
> + req_param->process_id = pid;
I think you can use "get_pid()" here and not have "pid" argument in the function, as I believe it is always called within the same process.
> + if (qp_conf) {
> + req_param->nb_descriptors = qp_conf->nb_descriptors;
> + req_param->mp_session = (void *)qp_conf->mp_session;
> + }
> +
> + qp_req_msg.num_fds = 0;
> + ret = rte_mp_request_sync(&qp_req_msg, &qp_resp, &ts);
> + if (ret) {
> + RTE_LOG(ERR, USER1, "Create MR request to primary
> process failed.");
> + return -1;
> + }
> + qp_resp_msg = &qp_resp.msgs[0];
> + resp_param = (struct ipsec_mb_mp_param *)qp_resp_msg->param;
> +
> + return resp_param->result;
> +}
> +
..
> diff --git a/drivers/crypto/ipsec_mb/ipsec_mb_private.h
> b/drivers/crypto/ipsec_mb/ipsec_mb_private.h
> index b56eaf061e..a2b52f808e 100644
> --- a/drivers/crypto/ipsec_mb/ipsec_mb_private.h
> +++ b/drivers/crypto/ipsec_mb/ipsec_mb_private.h
...
> +
> /** All supported device types */
> enum ipsec_mb_pmd_types {
> IPSEC_MB_PMD_TYPE_AESNI_MB = 0,
> @@ -149,11 +155,51 @@ struct ipsec_mb_qp {
> IMB_MGR *mb_mgr;
> /* Multi buffer manager */
> const struct rte_memzone *mb_mgr_mz;
> + /** The process id used for queue pairs **/
> + uint16_t qp_in_used_by_pid;
I would rename this to "qp_in_use_by_pid" or "qp_used_by_pid".
> /* Shared memzone for storing mb_mgr */
Check these comments. This last one "Shared memzone..." applies to the field "mb_mgr_mz".
Also, they should use "/**< ".
> __extension__ uint8_t additional_data[];
> /**< Storing PMD specific additional data */ };
>
next prev parent reply other threads:[~2022-10-26 8:17 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-10-20 23:51 [dpdk-dev v1] " Kai Ji
2022-10-24 23:38 ` [dpdk-dev v2] " Kai Ji
2022-10-25 21:48 ` [dpdk-dev v3] " Kai Ji
2022-10-26 8:17 ` De Lara Guarch, Pablo [this message]
2022-10-26 10:01 ` [dpdk-dev v4] " Kai Ji
2022-10-26 10:22 ` Kai Ji
2022-10-26 10:27 ` Kai Ji
2022-10-26 11:07 ` Kusztal, ArkadiuszX
2022-10-26 12:32 ` De Lara Guarch, Pablo
2022-10-26 12:48 ` Kai Ji
2022-10-26 13:04 ` De Lara Guarch, Pablo
2022-10-26 13:19 ` [dpdk-dev v5] " Kai Ji
2022-10-27 9:54 ` [EXT] " Akhil Goyal
2022-10-27 17:57 ` Konstantin Ananyev
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=DM8PR11MB559190B7454A1569C5DE0D4484309@DM8PR11MB5591.namprd11.prod.outlook.com \
--to=pablo.de.lara.guarch@intel.com \
--cc=anatoly.burakov@intel.com \
--cc=dev@dpdk.org \
--cc=gakhil@marvell.com \
--cc=kai.ji@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).