[ovs-dev] [PATCH v4 2/3] netdev-dpdk: Add custom stat for vhost tx retries.

Ilya Maximets i.maximets at samsung.com
Thu Jul 4 15:15:53 UTC 2019


On 04.07.2019 18:13, Ian Stokes wrote:
> On 7/4/2019 12:18 PM, Ilya Maximets wrote:
>> On 04.07.2019 14:06, David Marchand wrote:
>>>
>>>
>>> On Thu, Jul 4, 2019 at 12:42 PM Ilya Maximets <i.maximets at samsung.com <mailto:i.maximets at samsung.com>> wrote:
>>>
>>>      On 03.07.2019 18:03, Ian Stokes wrote:
>>>      > On 7/2/2019 1:32 AM, Kevin Traynor wrote:
>>>      >> vhost tx retries may occur, and it can be a sign that
>>>      >> the guest is not optimally configured.
>>>      >>
>>>      >> Add a custom stat so a user will know if vhost tx retries are
>>>      >> occurring and hence give a hint that guest config should be
>>>      >> examined.
>>>      >>
>>>      >
>>>      > Thanks Kevin, tests ok for me.
>>>      >
>>>      > Just a general comment on the design. In comparison to the previous approach proposed there seems to be more required with the custom stat approach below.
>>>      >
>>>      > This may be ok as the retry is very OVS DPDK specific and doesn;t come dor lets say DPDK (unlike the XTATS).
>>>      >
>>>      > @Ilya, whats your thoughts? Is this approach preferable for you rather than adding it to the general stats for all devices? (in which case I agree, they will not be used for dpdk types so will not be of use).
>>>
>>>      I think, It's better to keep this in custom stats section since
>>>      no other port types are going to use it in a near future.
>>>
>>>      Have a few style/naming comments inline.
>>>
>>>      >
>>>      > One other minor comment below.
>>>      >
>>>      >> Signed-off-by: Kevin Traynor <ktraynor at redhat.com <mailto:ktraynor at redhat.com>>
>>>      >> ---
>>>      >>   Documentation/topics/dpdk/vhost-user.rst |  5 ++++
>>>      >>   lib/netdev-dpdk.c                        | 38 +++++++++++++++++++++++-
>>>      >>   2 files changed, 42 insertions(+), 1 deletion(-)
>>>      >>
>>>      >> diff --git a/Documentation/topics/dpdk/vhost-user.rst b/Documentation/topics/dpdk/vhost-user.rst
>>>      >> index 5f393aced..368f44520 100644
>>>      >> --- a/Documentation/topics/dpdk/vhost-user.rst
>>>      >> +++ b/Documentation/topics/dpdk/vhost-user.rst
>>>      >> @@ -521,4 +521,9 @@ The guest should also have sufficient cores dedicated for consuming and
>>>      >>   processing packets at the required rate.
>>>      >>   +The amount of Tx retries on a vhost-user or vhost-user-client interface can be
>>>      >> +shown with::
>>>      >> +
>>>      >> +  $ ovs-vsctl get Interface dpdkvhostclient0 statistics:tx_retries
>>>      >> +
>>>      >>   vhost-user Dequeue Zero Copy (experimental)
>>>      >>   -------------------------------------------
>>>      >> diff --git a/lib/netdev-dpdk.c b/lib/netdev-dpdk.c
>>>      >> index a380b6ffe..d3e02d389 100644
>>>      >> --- a/lib/netdev-dpdk.c
>>>      >> +++ b/lib/netdev-dpdk.c
>>>      >> @@ -139,4 +139,10 @@ BUILD_ASSERT_DECL((MAX_NB_MBUF / ROUND_DOWN_POW2(MAX_NB_MBUF / MIN_NB_MBUF))
>>>      >>   #define XSTAT_RX_JABBER_ERRORS           "rx_jabber_errors"
>>>      >>   +/* Size of vHost custom stats. */
>>>      >> +#define VHOST_STAT_TX_RETRIES_SIZE    1
>>>
>>>      I think, it's better to rename to something like VHOST_CUSTOM_STATS_SIZE.
>>>
>>>
>>> +1
>>>
>>>
>>>
>>>      >> +
>>>      >> +/* Name of vHost custom stats. */
>>>
>>>      s/Name/Names/ ?
>>>
>>>      >> +#define VHOST_STAT_TX_RETRIES    "tx_retries" > +
>>>      >
>>>      > The alignment of the defines above look odd as they are not aligned with the existing DPDK XSTAT name definitions. It's only minor, as Kevin is on leave I think this will be ok to change on commit if there are no objections.
>>>
>>>      Sure.
>>>
>>>      >
>>>      > Regards
>>>      > Ian
>>>      >>   #define SOCKET0              0
>>>      >>   @@ -434,7 +440,9 @@ struct netdev_dpdk {
>>>      >>       PADDED_MEMBERS(CACHE_LINE_SIZE,
>>>      >>           struct netdev_stats stats;
>>>      >> +        /* Custom stat for retries when unable to transmit. */
>>>      >> +        uint64_t tx_retries;
>>>      >>           /* Protects stats */
>>>      >>           rte_spinlock_t stats_lock;
>>>      >> -        /* 44 pad bytes here. */
>>>      >> +        /* 4 pad bytes here. */
>>>      >>       );
>>>      >>   @@ -1190,4 +1198,6 @@ common_construct(struct netdev *netdev, dpdk_port_t port_no,
>>>      >>       dev->rte_xstats_ids_size = 0;
>>>      >>   +    dev->tx_retries = 0;
>>>      >> +
>>>      >>       return 0;
>>>      >>   }
>>>      >> @@ -2381,4 +2391,5 @@ __netdev_dpdk_vhost_send(struct netdev *netdev, int qid,
>>>      >>       netdev_dpdk_vhost_update_tx_counters(&dev->stats, pkts, total_pkts,
>>>      >>                                            cnt + dropped);
>>>      >> +    dev->tx_retries += MIN(retries, VHOST_ENQ_RETRY_NUM);
>>>      >>       rte_spinlock_unlock(&dev->stats_lock);
>>>      >>   @@ -2818,4 +2829,27 @@ netdev_dpdk_get_custom_stats(const struct netdev *netdev,
>>>      >>   }
>>>      >>   +static int
>>>      >> +netdev_dpdk_vhost_get_custom_stats(const struct netdev *netdev,
>>>      >> +                                   struct netdev_custom_stats *custom_stats)
>>>      >> +{
>>>      >> +    struct netdev_dpdk *dev = netdev_dpdk_cast(netdev);
>>>      >> +
>>>      >> +    ovs_mutex_lock(&dev->mutex);
>>>      >> +
>>>      >> +    custom_stats->size = VHOST_STAT_TX_RETRIES_SIZE;
>>>      >> +    custom_stats->counters = (struct netdev_custom_counter *)
>>>      >> +                                 xzalloc(sizeof(struct netdev_custom_counter));
>>>
>>>      Since we have a _SIZE, we, probably, need to use it while allocation.
>>>      With this and a few coding-style fixes, this line should look like:
>>>
>>>          custom_stats->counters = xcalloc(VHOST_STAT_TX_RETRIES_SIZE,
>>>                                           sizeof *custom_stats->counters);
>>>
>>>
>>> Can I suggest?
>>>      custom_stats->counters = xcalloc(custom_stats->size,
>>>                                       sizeof *custom_stats->counters);
>>
>> Sure. This looks fine for me too.
>>
>>>
>>>      >> +    ovs_strlcpy(custom_stats->counters->name, VHOST_STAT_TX_RETRIES,
>>>      >> +                NETDEV_CUSTOM_STATS_NAME_SIZE);
>>>      >> +
>>>      >> +    rte_spinlock_lock(&dev->stats_lock);
>>>      >> +    custom_stats->counters->value = dev->tx_retries;
>>>
>>>
>>>
>>> I have a patch after this that will add another stats, so I'd rather see custom_stats->counters[0]. than custom_stats->counters->
>>
>> +1.
>> This is also in line with my previous comment about having a _SIZE.
> 
> Just to clarify above, maybe I misunderstood what you meant, but are you suggesting
> 
> custom_stats->counters[0] = dev->tx_retries;
> 
> I don't think that will work as it will be an incorrect type assignment right?

custom_stats->counters[0].value = dev->tx_retries;


More information about the dev mailing list