From: Ido Goshen <Ido@cgstowernetworks.com>
To: Ferruh Yigit <ferruh.yigit@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH v3 1/1] net/pcap: imissed stats support
Date: Thu, 4 Feb 2021 07:56:53 +0000 [thread overview]
Message-ID: <DB7PR09MB2345C76D4173776F978ECB5DD6B39@DB7PR09MB2345.eurprd09.prod.outlook.com> (raw)
In-Reply-To: <add64905-4f9f-394d-e059-e63bb43c709f@intel.com>
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@intel.com>
> Sent: Thursday, 4 February 2021 2:13
> To: Ido Goshen <Ido@cgstowernetworks.com>
> Cc: dev@dpdk.org
> Subject: Re: [PATCH v3 1/1] net/pcap: imissed stats support
>
> On 2/3/2021 11:07 PM, Ido Goshen wrote:
> > get value from pcap_stats.ps_drop (see man pcap_stats) the value is
> > adjusted in this cases:
> > - port stop - pcap is closed and will lose count
> > - stats reset - pcap doesn't provide reset api
> > - rollover - pcap counter size is u_32 only
> >
> > Signed-off-by: Ido Goshen <ido@cgstowernetworks.com>
> > ---
> > v3:
> > * code cleanup by dedicated struct and functions extraction
> > * multi stop support by menmonic+= accumulation
> > * rollover fixup
> >
> > v2:
> > * sum all queues (rx_missed_total += fix)
> > * null pcap protection
> > * inter stop/start persistancy (counter won't reset on stop)
> >
> > drivers/net/pcap/rte_eth_pcap.c | 59
> +++++++++++++++++++++++++++++++++
> > 1 file changed, 59 insertions(+)
> >
> > diff --git a/drivers/net/pcap/rte_eth_pcap.c
> > b/drivers/net/pcap/rte_eth_pcap.c index a32b1f3f3..16e8752f3 100644
> > --- a/drivers/net/pcap/rte_eth_pcap.c
> > +++ b/drivers/net/pcap/rte_eth_pcap.c
> > @@ -60,11 +60,21 @@ struct queue_stat {
> > volatile unsigned long err_pkts;
> > };
> >
> > +struct queue_missed_stat {
> > + /* last value retrieved from pcap */
> > + volatile unsigned int pcap;
> > + /* stores values lost by pcap stop or rollover */
> > + volatile unsigned long mnemonic;
> > + /* value on last reset */
> > + volatile unsigned long reset;
> > +};
>
> I am aware other stats has 'volatile' keyword, but as far as I can see it is not
> needed, since these are new ones can you please drop the 'volatile'?
ok
>
> > +
> > struct pcap_rx_queue {
> > uint16_t port_id;
> > uint16_t queue_id;
> > struct rte_mempool *mb_pool;
> > struct queue_stat rx_stat;
> > + struct queue_missed_stat missed_stat;
> > char name[PATH_MAX];
> > char type[ETH_PCAP_ARG_MAXLEN];
> >
> > @@ -144,6 +154,49 @@ RTE_LOG_REGISTER(eth_pcap_logtype,
> pmd.net.pcap, NOTICE);
> > rte_log(RTE_LOG_ ## level, eth_pcap_logtype, \
> > "%s(): " fmt "\n", __func__, ##args)
> >
> > +static struct queue_missed_stat*
> > +queue_missed_stat_update(struct rte_eth_dev *dev, unsigned int qid) {
> > + struct pmd_internals *internals = dev->data->dev_private;
> > + struct queue_missed_stat *missed_stat =
> > + &internals->rx_queue[qid].missed_stat;
> > + const struct pmd_process_private *pp = dev->process_private;
> > + pcap_t *pcap = pp->rx_pcap[qid];
> > + struct pcap_stat stat;
>
> Can you please put an empty line after variable declarations, and before return.
ok
>
> > + if (!pcap || (pcap_stats(pcap, &stat) != 0))
> > + return missed_stat;
> > + /* rollover check - best effort fixup assuming single rollover */
> > + if (stat.ps_drop < missed_stat->pcap)
> > + missed_stat->mnemonic += UINT_MAX;
> > + missed_stat->pcap = stat.ps_drop;
>
> here.
>
> > + return missed_stat;
> > +}
> > +
> > +static void
> > +queue_missed_stat_on_stop_update(struct rte_eth_dev *dev, unsigned
> > +int qid) {
> > + struct queue_missed_stat *missed_stat =
> > + queue_missed_stat_update(dev, qid);
> > + missed_stat->mnemonic += missed_stat->pcap;
>
> Better to reset 'missed_stat->pcap' afterwards, in case stats requested before
> port started again:
> missed_stat->pcap = 0;
right, should be careful not to double count it
but maybe better to set it to 0 in queue_missed_stat_update in the stop return case
if (!pcap || (pcap_stats(pcap, &stat) != 0))
{
missed_stat->pcap = 0;
return missed_stat;
}
this way the missed_stat->pcap will always represent the current value from pcap and not hold old value
specifically in case port is stopped it will be 0 and not re-added
agree?
>
> > +}
> > +
> > +static void
> > +queue_missed_stat_reset(struct rte_eth_dev *dev, unsigned int qid) {
> > + struct queue_missed_stat *missed_stat =
> > + queue_missed_stat_update(dev, qid);
> > + missed_stat->reset = missed_stat->pcap;
>
> I guess this should be:
> "missed_stat->reset = missed_stat->pcap + missed_stat->mnemonic;"
I don't think so
reset should only remember where pcap was at the reset point and not store old values
trying it immediately results in
testpmd> show port stats 0
######################## NIC statistics for port 0 ########################
RX-packets: 0 RX-missed: 1940 RX-bytes: 0
RX-errors: 0
RX-nombuf: 0
TX-packets: 0 TX-errors: 0 TX-bytes: 0
Throughput (since last show)
Rx-pps: 0 Rx-bps: 0
Tx-pps: 0 Tx-bps: 0
############################################################################
testpmd> clear port stats 0
NIC statistics for port 0 cleared
testpmd> show port stats 0
######################## NIC statistics for port 0 ########################
RX-packets: 0 RX-missed: 18446744073709550646 RX-bytes: 0
RX-errors: 0
RX-nombuf: 0
TX-packets: 0 TX-errors: 0 TX-bytes: 0
Throughput (since last show)
Rx-pps: 0 Rx-bps: 0
Tx-pps: 0 Tx-bps: 0
############################################################################
> > + missed_stat->mnemonic = 0;
> > +}
> > +
> > +static unsigned long
> > +queue_missed_stat_get(struct rte_eth_dev *dev, unsigned int qid) {
> > + const struct queue_missed_stat *missed_stat =
> > + queue_missed_stat_update(dev, qid);
> > + return missed_stat->pcap + missed_stat->mnemonic -
> > +missed_stat->reset; }
> > +
>
> <...>
next prev parent reply other threads:[~2021-02-04 7:56 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-25 17:58 [dpdk-dev] [PATCH " Ido Goshen
2021-01-28 18:10 ` Ferruh Yigit
2021-02-01 8:30 ` [dpdk-dev] [PATCH v2] " Ido Goshen
2021-02-01 11:48 ` Ferruh Yigit
2021-02-01 14:02 ` Ido Goshen
2021-02-01 16:21 ` Ferruh Yigit
2021-02-03 23:07 ` [dpdk-dev] [PATCH v3 1/1] " Ido Goshen
2021-02-04 0:13 ` Ferruh Yigit
2021-02-04 7:56 ` Ido Goshen [this message]
2021-02-04 9:26 ` Ferruh Yigit
2021-02-04 10:02 ` Ido Goshen
2021-02-04 10:27 ` Ferruh Yigit
2021-02-04 10:33 ` [dpdk-dev] [PATCH v4 " Ido Goshen
2021-02-04 18:31 ` Ferruh Yigit
2021-02-22 18:13 ` Ferruh Yigit
2021-01-28 18:20 ` [dpdk-dev] [PATCH " Ferruh Yigit
2021-02-01 8:53 ` Ido Goshen
2021-02-01 9:25 ` Ferruh Yigit
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=DB7PR09MB2345C76D4173776F978ECB5DD6B39@DB7PR09MB2345.eurprd09.prod.outlook.com \
--to=ido@cgstowernetworks.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@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).