[1/2] event/octeontx2: remove selftest from dev args
Checks
Commit Message
From: Pavan Nikhilesh <pbhagavatula@marvell.com>
Since selftest now depends on dynamic mbuf fields it is not
feasible to run selftest on device probe.
Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
---
drivers/event/octeontx2/otx2_evdev.c | 8 --------
drivers/event/octeontx2/otx2_evdev.h | 1 -
2 files changed, 9 deletions(-)
Comments
On Thu, Nov 19, 2020 at 1:24 PM <pbhagavatula@marvell.com> wrote:
>
> From: Pavan Nikhilesh <pbhagavatula@marvell.com>
>
> Since selftest now depends on dynamic mbuf fields it is not
> feasible to run selftest on device probe.
Indeed, this is something that I missed with the seqn rework.
Can't you call rte_event_dev_selftest() from within the pmd?
>On Thu, Nov 19, 2020 at 1:24 PM <pbhagavatula@marvell.com> wrote:
>>
>> From: Pavan Nikhilesh <pbhagavatula@marvell.com>
>>
>> Since selftest now depends on dynamic mbuf fields it is not
>> feasible to run selftest on device probe.
>
>Indeed, this is something that I missed with the seqn rework.
>Can't you call rte_event_dev_selftest() from within the pmd?
>
We can but generally selftest is run as a safety net before an
custom application runs but now since selftest uses dynamic
mbuf fields the custom application would be starved of that
field since there is no way to unregister it.
Regards,
Pavan.
>
>--
>David Marchand
On Thu, Nov 19, 2020 at 2:31 PM Pavan Nikhilesh Bhagavatula
<pbhagavatula@marvell.com> wrote:
> We can but generally selftest is run as a safety net before an
> custom application runs but now since selftest uses dynamic
> mbuf fields the custom application would be starved of that
> field since there is no way to unregister it.
Then please update the documentation accordingly.
doc/guides/eventdevs/octeontx.rst: --vdev="event_octeontx,selftest=1"
doc/guides/eventdevs/octeontx2.rst: -a 0002:0e:00.0,selftest=1
>
>On Thu, Nov 19, 2020 at 2:31 PM Pavan Nikhilesh Bhagavatula
><pbhagavatula@marvell.com> wrote:
>> We can but generally selftest is run as a safety net before an
>> custom application runs but now since selftest uses dynamic
>> mbuf fields the custom application would be starved of that
>> field since there is no way to unregister it.
>
>Then please update the documentation accordingly.
>doc/guides/eventdevs/octeontx.rst: --
>vdev="event_octeontx,selftest=1"
>doc/guides/eventdevs/octeontx2.rst: -a 0002:0e:00.0,selftest=1
>
Ah! Yes, forgot about docs will send v2.
Thanks,
Pavan.
>
>--
>David Marchand
@@ -1606,7 +1606,6 @@ static struct rte_eventdev_ops otx2_sso_ops = {
#define OTX2_SSO_XAE_CNT "xae_cnt"
#define OTX2_SSO_SINGLE_WS "single_ws"
#define OTX2_SSO_GGRP_QOS "qos"
-#define OTX2_SSO_SELFTEST "selftest"
static void
parse_queue_param(char *value, void *opaque)
@@ -1696,8 +1695,6 @@ sso_parse_devargs(struct otx2_sso_evdev *dev, struct rte_devargs *devargs)
if (kvlist == NULL)
return;
- rte_kvargs_process(kvlist, OTX2_SSO_SELFTEST, &parse_kvargs_flag,
- &dev->selftest);
rte_kvargs_process(kvlist, OTX2_SSO_XAE_CNT, &parse_kvargs_value,
&dev->xae_cnt);
rte_kvargs_process(kvlist, OTX2_SSO_SINGLE_WS, &parse_kvargs_flag,
@@ -1813,10 +1810,6 @@ otx2_sso_init(struct rte_eventdev *event_dev)
otx2_sso_dbg("Initializing %s max_queues=%d max_ports=%d",
event_dev->data->name, dev->max_event_queues,
dev->max_event_ports);
- if (dev->selftest) {
- event_dev->dev->driver = &pci_sso.driver;
- event_dev->dev_ops->dev_selftest();
- }
otx2_tim_init(pci_dev, (struct otx2_dev *)dev);
@@ -1866,5 +1859,4 @@ RTE_PMD_REGISTER_KMOD_DEP(event_octeontx2, "vfio-pci");
RTE_PMD_REGISTER_PARAM_STRING(event_octeontx2, OTX2_SSO_XAE_CNT "=<int>"
OTX2_SSO_SINGLE_WS "=1"
OTX2_SSO_GGRP_QOS "=<string>"
- OTX2_SSO_SELFTEST "=1"
OTX2_NPA_LOCK_MASK "=<1-65535>");
@@ -147,7 +147,6 @@ struct otx2_sso_evdev {
uint64_t *timer_adptr_sz;
/* Dev args */
uint8_t dual_ws;
- uint8_t selftest;
uint32_t xae_cnt;
uint8_t qos_queue_cnt;
struct otx2_sso_qos *qos_parse_data;