[dpdk-stable] [PATCH v2] net/i40e: fix flow RSS queue index check error
Zhang, Qi Z
qi.z.zhang at intel.com
Tue Apr 10 12:45:00 CEST 2018
> -----Original Message-----
> From: Zhao1, Wei
> Sent: Tuesday, April 10, 2018 10:19 AM
> To: Zhang, Qi Z <qi.z.zhang at intel.com>; dev at dpdk.org
> Cc: stable at dpdk.org
> Subject: RE: [PATCH v2] net/i40e: fix flow RSS queue index check error
>
>
>
> > -----Original Message-----
> > From: Zhang, Qi Z
> > Sent: Monday, April 9, 2018 9:57 PM
> > To: Zhao1, Wei <wei.zhao1 at intel.com>; dev at dpdk.org
> > Cc: stable at dpdk.org
> > Subject: RE: [PATCH v2] net/i40e: fix flow RSS queue index check error
> >
> >
> >
> > > -----Original Message-----
> > > From: Zhao1, Wei
> > > Sent: Monday, April 9, 2018 4:38 PM
> > > To: dev at dpdk.org
> > > Cc: stable at dpdk.org; Zhang, Qi Z <qi.z.zhang at intel.com>; Zhao1, Wei
> > > <wei.zhao1 at intel.com>
> > > Subject: [PATCH v2] net/i40e: fix flow RSS queue index check error
> > >
> > > There is an error in queue index check for RSS queue region
> > > configuration.If the queue index is not continuous sequence for RSS,
> > > but queue region index is continuous sequence, in this case we can
> > > not use the old method for queue index check.
> > > This patch also add comment for flow rss parse function in order to
> > > explain some important info.
> > >
> > > Fixes: ecad87d22383 ("net/i40e: move RSS to flow API")
> > > Signed-off-by: Wei Zhao <wei.zhao1 at intel.com>
> > > Tested-by: Peng Yuan <yuan.peng at intel.com>
> > >
> > > ---
> > >
> > > v2:
> > > -merge this with another patch for function comment, add git log info.
> > > ---
> > > drivers/net/i40e/i40e_flow.c | 32 +++++++++++++++++++++-----------
> > > 1 file changed, 21 insertions(+), 11 deletions(-)
> > >
> > > diff --git a/drivers/net/i40e/i40e_flow.c
> > > b/drivers/net/i40e/i40e_flow.c index
> > > f4d08bb..40f15c2 100644
> > > --- a/drivers/net/i40e/i40e_flow.c
> > > +++ b/drivers/net/i40e/i40e_flow.c
> > > @@ -4171,6 +4171,19 @@ i40e_flow_parse_rss_pattern(__rte_unused
> > struct
> > > rte_eth_dev *dev,
> > > return 0;
> > > }
> > >
> > > +/**
> > > + * This function is used to parse rss queue index, total queue
> > > +number and
> > > + * hash functions, If the purpose of this configuration is for
> > > +queue region
> > > + * configuration, it will set queue_region_conf flag to TRUE, else to FALSE.
> > > + * In queue region configuration, it also need to parse hardware
> > > +flowtype
> > > + * and user_priority from configuration, it will also cheeck the
> > > +validity
> > > + * of these parameters. For example, The queue region sizes should
> > > + * be any of the following values: 1, 2, 4, 8, 16, 32, 64, the
> > > + * hw_flowtype or PCTYPE max index should be 63, the user priority
> > > + * max index should be 7, and so on. And also, queue index should
> > > +be
> > > + * continuous sequence and queue region index should be part of rss
> > > + * queue index for this port.
> > > + */
> > > static int
> > > i40e_flow_parse_rss_action(struct rte_eth_dev *dev,
> > > const struct rte_flow_action *actions, @@ -4240,6
> > > +4253,14 @@ i40e_flow_parse_rss_action(struct rte_eth_dev *dev,
> > > return -rte_errno;
> > > }
> > > }
> > > +
> > > + if (rss_info->num < rss->num) {
> > > + rte_flow_error_set(error, EINVAL,
> > > + RTE_FLOW_ERROR_TYPE_ACTION,
> > > + act,
> > > + "no valid queues");
> > > + return -rte_errno;
> > > + }
> > > }
> > >
> > > for (n = 0; n < conf_info->queue_region_number; n++) { @@ -
> > 4264,17
> > > +4285,6 @@ i40e_flow_parse_rss_action(struct rte_eth_dev *dev,
> > > return -rte_errno;
> > > }
> > >
> > > - if (rss_info->num < rss->num ||
> > > - rss->queue[0] < rss_info->queue[0] ||
> > > - (rss->queue[0] + rss->num >
> > > - rss_info->num + rss_info->queue[0]))
> > {
> > > - rte_flow_error_set(error, EINVAL,
> > > - RTE_FLOW_ERROR_TYPE_ACTION,
> > > - act,
> > > - "no valid queues");
> > > - return -rte_errno;
> > > - }
> >
> > Is this a code clean?
> > I didn't see anything wrong with the code you removed, but they just
> > looks redundant.
> > Btw why you still check (rss_info->num < rss->num) ?, previous, I
> > think you already iterate all rss queue to make sure they are matched
> > by a rss_info queue so it already proved rss_info->num >= rss->num.
>
> After thinking it over, I think we had better keep this check for (rss_info->num
> < rss->num) because customer may input a wrong number for this parameter,
> although queue index maybe rightly match rss span. So, v3 may be not need.
If rss->num > rss_info->num, it should already failed in below code, so I still think it's no necessary.
/* each rss queue must match a rss_info queue */
for (i = 0; i < rss->num; i++) {
for (j = 0; j < rss_info->num; j++) {
if (rss->queue[i] == rss_info->queue[j])
break;
}
if (j == rss_info->num) {
rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ACTION,
act,
"no valid queues");
return -rte_errno;
}
}
/* rss queue must be continues */
for (i = 0; i < rss->num - 1; i++) {
if (rss->queue[i + 1] != rss->queue[i] + 1) {
rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ACTION,
act,
"no valid queues");
return -rte_errno;
}
}
>
>
> >
> >
> > > for (i = 0; i < info->queue_region_number; i++) {
> > > if (info->region[i].queue_num == rss->num
> > &&
> > > info->region[i].queue_start_index ==
> > > --
> > > 2.7.5
More information about the stable
mailing list