Re: [dpdk-dev] [PATCH 1/2] eventdev: implement ABI change

[Jerin Jacob <jerinjacobk@gmail.com>]

On Thu, Sep 24, 2020 at 3:25 PM Sunil Kumar Kori <skori@marvell.com> wrote:
>
> >-----Original Message-----
> >From: dev <dev-bounces@dpdk.org> On Behalf Of Timothy McDaniel
> >Sent: Friday, September 11, 2020 10:28 PM
> >Cc: Jerin Jacob Kollanukkaran <jerinj@marvell.com>;
> >mattias.ronnblom@ericsson.com; liang.j.ma@intel.com;
> >peter.mccarthy@intel.com; nipun.gupta@nxp.com; Pavan Nikhilesh
> >Bhagavatula <pbhagavatula@marvell.com>; dev@dpdk.org;
> >erik.g.carrillo@intel.com; gage.eads@intel.com; harry.van.haaren@intel.com;
> >hemant.agrawal@nxp.com; bruce.richardson@intel.com
> >Subject: [dpdk-dev] [PATCH 1/2] eventdev: implement ABI change

Please change subject to "eventdev: express DLB PMD constraints" or
something similar.

> >
> >This commit implements the eventdev ABI changes required by the DLB PMD.
> >
> >The DLB hardware does not conform exactly to the eventdev interface.
> >1) It has a limit on the number of queues that may be linked to a port.
> >2) Some ports are further restricted to a maximum of 1 linked queue.
> >3) It does not (currently) have the ability to carry the flow_id as part of the
> >event (QE) payload.
> >
> >Due to the above, we would like to propose the following enhancements.
> >
> >1) Add new fields to the rte_event_dev_info struct. These fields allow the
> >device to advertise its capabilities so that applications can take the
> >appropriate actions based on those capabilities.
> >
> >2) Add a new field to the rte_event_dev_config struct. This field allows the
> >application to specify how many of its ports are limited to a single link, or will
> >be used in single link mode.
> >
> >3) Replace the dedicated implicit_release_disabled field with a bit field of
> >explicit port capabilities. The implicit_release_disable functionality is assigned
> >to one bit, and a port-is-single-link-only attribute is assigned to another, with
> >the remaining bits available for future assignment.
> >
> >Signed-off-by: Timothy McDaniel <timothy.mcdaniel@intel.com>
> >---
> > drivers/event/dpaa/dpaa_eventdev.c             |  3 +-
> > drivers/event/dpaa2/dpaa2_eventdev.c           |  5 +-
> > drivers/event/dsw/dsw_evdev.c                  |  3 +-
> > drivers/event/octeontx/ssovf_evdev.c           |  5 +-
> > drivers/event/octeontx2/otx2_evdev.c           |  3 +-
> > drivers/event/opdl/opdl_evdev.c                |  3 +-
> > drivers/event/skeleton/skeleton_eventdev.c     |  5 +-
> > drivers/event/sw/sw_evdev.c                    |  8 ++--
> > drivers/event/sw/sw_evdev_selftest.c           |  6 +--
> > lib/librte_eventdev/rte_event_eth_tx_adapter.c |  2 +-
> > lib/librte_eventdev/rte_eventdev.c             | 66 +++++++++++++++++++++++---
> > lib/librte_eventdev/rte_eventdev.h             | 51 ++++++++++++++++----
> > lib/librte_eventdev/rte_eventdev_pmd_pci.h     |  1 -
> > lib/librte_eventdev/rte_eventdev_trace.h       |  7 +--
> > lib/librte_eventdev/rte_eventdev_version.map   |  4 +-

Missed to remove the announced depreciation notice.


> > 15 files changed, 134 insertions(+), 38 deletions(-)
> >
> > /**
> >  * @internal
> >  * Wrapper for use by pci drivers as a .remove function to detach a event diff
> >--git a/lib/librte_eventdev/rte_eventdev_trace.h
> >b/lib/librte_eventdev/rte_eventdev_trace.h
> >index 4de6341..5ec43d8 100644
> >--- a/lib/librte_eventdev/rte_eventdev_trace.h
> >+++ b/lib/librte_eventdev/rte_eventdev_trace.h
> >@@ -34,6 +34,7 @@ RTE_TRACE_POINT(
> >       rte_trace_point_emit_u32(dev_conf-
> >>nb_event_port_dequeue_depth);
> >       rte_trace_point_emit_u32(dev_conf-
> >>nb_event_port_enqueue_depth);
> >       rte_trace_point_emit_u32(dev_conf->event_dev_cfg);
> >+      rte_trace_point_emit_u8(dev_conf-
> >>nb_single_link_event_port_queues);
> >       rte_trace_point_emit_int(rc);
> > )
> >
> I tried running testpmd with trace support with this patch and got following error:
>
> EAL: __rte_trace_point_emit_field():442 CTF field is too long
> EAL: __rte_trace_point_register():468 missing rte_trace_emit_header() in register fn
>
> It is because of memory (384 bytes) for CTF gets exhausted during eventdev_configure.
> Did you also observe the same issue ? If yes, Then please send a separate patch to increase
> Maximum buffer size TRACE_CTF_FIELD_SIZE. Recommended size is 448 bytes.

Please check this and send the next version for this and common code
change[1]. I would like to merge-common code change first and then the
DLB drivers.

[1]
http://patches.dpdk.org/patch/77466/


>
> >@@ -59,7 +60,7 @@ RTE_TRACE_POINT(
> >       rte_trace_point_emit_i32(port_conf->new_event_threshold);
> >       rte_trace_point_emit_u16(port_conf->dequeue_depth);
> >       rte_trace_point_emit_u16(port_conf->enqueue_depth);
> >-      rte_trace_point_emit_u8(port_conf->disable_implicit_release);
> >+      rte_trace_point_emit_u32(port_conf->event_port_cfg);
> >       rte_trace_point_emit_int(rc);
> > )
> >
> >@@ -165,7 +166,7 @@ RTE_TRACE_POINT(
> >       rte_trace_point_emit_i32(port_conf->new_event_threshold);
> >       rte_trace_point_emit_u16(port_conf->dequeue_depth);
> >       rte_trace_point_emit_u16(port_conf->enqueue_depth);
> >-      rte_trace_point_emit_u8(port_conf->disable_implicit_release);
> >+      rte_trace_point_emit_u32(port_conf->event_port_cfg);
> >       rte_trace_point_emit_ptr(conf_cb);
> >       rte_trace_point_emit_int(rc);
> > )
> >@@ -257,7 +258,7 @@ RTE_TRACE_POINT(
> >       rte_trace_point_emit_i32(port_conf->new_event_threshold);
> >       rte_trace_point_emit_u16(port_conf->dequeue_depth);
> >       rte_trace_point_emit_u16(port_conf->enqueue_depth);
> >-      rte_trace_point_emit_u8(port_conf->disable_implicit_release);
> >+      rte_trace_point_emit_u32(port_conf->event_port_cfg);
> > )
> >
> > RTE_TRACE_POINT(
> >diff --git a/lib/librte_eventdev/rte_eventdev_version.map
> >b/lib/librte_eventdev/rte_eventdev_version.map
> >index 3d9d0ca..2846d04 100644
> >--- a/lib/librte_eventdev/rte_eventdev_version.map
> >+++ b/lib/librte_eventdev/rte_eventdev_version.map
> >@@ -100,7 +100,6 @@ EXPERIMENTAL {
> >       # added in 20.05
> >       __rte_eventdev_trace_configure;
> >       __rte_eventdev_trace_queue_setup;
> >-      __rte_eventdev_trace_port_setup;
> >       __rte_eventdev_trace_port_link;
> >       __rte_eventdev_trace_port_unlink;
> >       __rte_eventdev_trace_start;
> >@@ -134,4 +133,7 @@ EXPERIMENTAL {
> >       __rte_eventdev_trace_crypto_adapter_queue_pair_del;
> >       __rte_eventdev_trace_crypto_adapter_start;
> >       __rte_eventdev_trace_crypto_adapter_stop;
> >+
> >+      # changed in 20.11
> >+      __rte_eventdev_trace_port_setup;

+ Ray, Techboard

Trace is still experimental. So I assume we don't need any
depreciation notice to change this.

> > };
> >--
> >2.6.4
>