| Message ID | 20201215033812.145975-11-cmi@nvidia.com |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | Add offload support for sFlow | expand |
On 15 Dec 2020, at 4:38, Chris Mi wrote: > Create a dedicated thread to poll psample netlink socket, receive > sampled packet, parse it to sFlow format and send it to sFlow > monitoring host. > > Signed-off-by: Chris Mi <cmi@nvidia.com> > Reviewed-by: Eli Britstein <elibr@nvidia.com> > --- > lib/netdev-offload-tc.c | 127 > +++++++++++++++++++++++++++++++++++++++- > 1 file changed, 126 insertions(+), 1 deletion(-) > > diff --git a/lib/netdev-offload-tc.c b/lib/netdev-offload-tc.c > index 61c5a14e6..e5bf719a4 100644 > --- a/lib/netdev-offload-tc.c > +++ b/lib/netdev-offload-tc.c > @@ -19,12 +19,14 @@ > #include <errno.h> > #include <linux/if_ether.h> > #include <linux/psample.h> > +#include <poll.h> > > #include "dpif.h" > #include "hash.h" > #include "openvswitch/hmap.h" > #include "openvswitch/match.h" > #include "openvswitch/ofpbuf.h" > +#include "openvswitch/poll-loop.h" > #include "openvswitch/thread.h" > #include "openvswitch/types.h" > #include "openvswitch/util.h" > @@ -2295,6 +2297,123 @@ netdev_tc_psample_init(void) > return sock; > } > > +struct netdev_tc_psample { > + struct nlattr *packet; /* packet data */ > + int dp_group_id; /* mapping id for sFlow offload */ > + int iifindex; /* input ifindex */ > + int group_seq; /* group sequence */ > +}; > + > +static int > +netdev_tc_psample_from_ofpbuf(struct netdev_tc_psample *psample, > + const struct ofpbuf *buf) > +{ > + static const struct nl_policy ovs_psample_policy[] = { > + [PSAMPLE_ATTR_IIFINDEX] = { .type = NL_A_U16 }, > + [PSAMPLE_ATTR_SAMPLE_GROUP] = { .type = NL_A_U32 }, > + [PSAMPLE_ATTR_GROUP_SEQ] = { .type = NL_A_U32 }, > + [PSAMPLE_ATTR_DATA] = { .type = NL_A_UNSPEC }, > + }; > + struct nlattr *a[ARRAY_SIZE(ovs_psample_policy)]; > + struct genlmsghdr *genl; > + struct nlmsghdr *nlmsg; > + struct ofpbuf b; > + > + b = ofpbuf_const_initializer(buf->data, buf->size); > + nlmsg = ofpbuf_try_pull(&b, sizeof *nlmsg); > + genl = ofpbuf_try_pull(&b, sizeof *genl); > + if (!nlmsg || !genl || nlmsg->nlmsg_type != psample_family > + || !nl_policy_parse(&b, 0, ovs_psample_policy, a, > + ARRAY_SIZE(ovs_psample_policy))) { > + return EINVAL; > + } > + > + psample->iifindex = nl_attr_get_u16(a[PSAMPLE_ATTR_IIFINDEX]); > + psample->dp_group_id = > nl_attr_get_u32(a[PSAMPLE_ATTR_SAMPLE_GROUP]); > + psample->group_seq = nl_attr_get_u16(a[PSAMPLE_ATTR_GROUP_SEQ]); > + psample->packet = a[PSAMPLE_ATTR_DATA]; > + > + return 0; > +} > + > +static int > +netdev_tc_psample_parse_packet(struct netdev_tc_psample *psample, > + struct dpif_upcall_sflow *dupcall) > +{ > + const struct gid_node *node; > + > + dp_packet_use_stub(&dupcall->packet, > + CONST_CAST(struct nlattr *, > + nl_attr_get(psample->packet)) - 1, > + nl_attr_get_size(psample->packet) + > + sizeof(struct nlattr)); > + dp_packet_set_data(&dupcall->packet, > + (char *) dp_packet_data(&dupcall->packet) + > + sizeof(struct nlattr)); > + dp_packet_set_size(&dupcall->packet, > nl_attr_get_size(psample->packet)); > + > + node = gid_find(psample->dp_group_id); > + dupcall->sflow_attr = &node->sflow; > + dupcall->iifindex = psample->iifindex; > + > + return 0; > +} > + > +static int > +netdev_tc_psample_poll(struct dpif_upcall_sflow *dupcall, > + struct nl_sock *sock) > +{ > + for (;;) { > + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, > 5); > + struct netdev_tc_psample psample; > + uint64_t buf_stub[4096 / 8]; > + struct ofpbuf buf; > + int error; > + > + ofpbuf_use_stub(&buf, buf_stub, sizeof buf_stub); > + error = nl_sock_recv(sock, &buf, NULL, false); > + > + if (!error) { > + error = netdev_tc_psample_from_ofpbuf(&psample, &buf); > + if (!error) { > + ofpbuf_uninit(&buf); > + error = netdev_tc_psample_parse_packet(&psample, > dupcall); > + return error; > + } > + } else if (error != EAGAIN) { > + VLOG_WARN_RL(&rl, "%s: error reading or parsing netlink > (%s)", > + __func__, ovs_strerror(error)); > + nl_sock_drain(sock); > + error = ENOBUFS; > + } > + > + ofpbuf_uninit(&buf); > + if (error) { > + return error; > + } > + } > +} > + > +static void * > +netdev_tc_psample_handler(void *arg) > +{ > + struct nl_sock *sock = CONST_CAST(struct nl_sock *, arg); > + > + struct dpif_upcall_sflow dupcall; > + int err; > + > + while (true) { > + err = netdev_tc_psample_poll(&dupcall, sock); > + if (!err) { I have not (fully) reviewed the patches below: [PATCH v9 03/11] dpif: Introduce register sFlow upcall [PATCH v9 04/11] ofproto: Add upcall callback to proces [PATCH v9 05/11] netdev-offload: Introduce register sFlow upcall callback API [PATCH v9 06/11] netdev-offload-tc: Implement register sFlow upcall callback API [PATCH v9 07/11] dpif-netlink: Implement register sFlow Upcall callback API Reason being that adding all this additional infrastructure does not make sense to me. We should just use the existing register_upcall_cb, and call it when we have an sflow packet via netlink. We can easily format the data such that upcall_cb()->upcall_receive()->process_upcall() will handle the sflow packet as it came from a normal upcall. What do you think? > + upcall_cb(&dupcall); > + } > + nl_sock_wait(sock, POLLIN); > + poll_block(); > + } > + > + return NULL; > +} > + > static int > netdev_tc_init_flow_api(struct netdev *netdev) > { > @@ -2324,12 +2443,18 @@ netdev_tc_init_flow_api(struct netdev *netdev) > tc_add_del_qdisc(ifindex, false, 0, hook); > > if (ovsthread_once_start(&once)) { > + struct nl_sock *sock; > + > probe_tc_block_support(ifindex); > /* Need to re-fetch block id as it depends on feature > availability. */ > block_id = get_block_id_from_netdev(netdev); > > probe_multi_mask_per_prio(ifindex); > - netdev_tc_psample_init(); > + sock = netdev_tc_psample_init(); > + if (sock) { > + ovs_thread_create("psample_handler", > netdev_tc_psample_handler, > + sock); > + } > ovsthread_once_done(&once); > } > > -- > 2.26.2
On 1/13/2021 3:53 AM, Eelco Chaudron wrote: > > > On 15 Dec 2020, at 4:38, Chris Mi wrote: > >> Create a dedicated thread to poll psample netlink socket, receive >> sampled packet, parse it to sFlow format and send it to sFlow >> monitoring host. >> >> Signed-off-by: Chris Mi <cmi@nvidia.com> >> Reviewed-by: Eli Britstein <elibr@nvidia.com> >> --- >> lib/netdev-offload-tc.c | 127 +++++++++++++++++++++++++++++++++++++++- >> 1 file changed, 126 insertions(+), 1 deletion(-) >> >> diff --git a/lib/netdev-offload-tc.c b/lib/netdev-offload-tc.c >> index 61c5a14e6..e5bf719a4 100644 >> --- a/lib/netdev-offload-tc.c >> +++ b/lib/netdev-offload-tc.c >> @@ -19,12 +19,14 @@ >> #include <errno.h> >> #include <linux/if_ether.h> >> #include <linux/psample.h> >> +#include <poll.h> >> >> #include "dpif.h" >> #include "hash.h" >> #include "openvswitch/hmap.h" >> #include "openvswitch/match.h" >> #include "openvswitch/ofpbuf.h" >> +#include "openvswitch/poll-loop.h" >> #include "openvswitch/thread.h" >> #include "openvswitch/types.h" >> #include "openvswitch/util.h" >> @@ -2295,6 +2297,123 @@ netdev_tc_psample_init(void) >> return sock; >> } >> >> +struct netdev_tc_psample { >> + struct nlattr *packet; /* packet data */ >> + int dp_group_id; /* mapping id for sFlow offload */ >> + int iifindex; /* input ifindex */ >> + int group_seq; /* group sequence */ >> +}; >> + >> +static int >> +netdev_tc_psample_from_ofpbuf(struct netdev_tc_psample *psample, >> + const struct ofpbuf *buf) >> +{ >> + static const struct nl_policy ovs_psample_policy[] = { >> + [PSAMPLE_ATTR_IIFINDEX] = { .type = NL_A_U16 }, >> + [PSAMPLE_ATTR_SAMPLE_GROUP] = { .type = NL_A_U32 }, >> + [PSAMPLE_ATTR_GROUP_SEQ] = { .type = NL_A_U32 }, >> + [PSAMPLE_ATTR_DATA] = { .type = NL_A_UNSPEC }, >> + }; >> + struct nlattr *a[ARRAY_SIZE(ovs_psample_policy)]; >> + struct genlmsghdr *genl; >> + struct nlmsghdr *nlmsg; >> + struct ofpbuf b; >> + >> + b = ofpbuf_const_initializer(buf->data, buf->size); >> + nlmsg = ofpbuf_try_pull(&b, sizeof *nlmsg); >> + genl = ofpbuf_try_pull(&b, sizeof *genl); >> + if (!nlmsg || !genl || nlmsg->nlmsg_type != psample_family >> + || !nl_policy_parse(&b, 0, ovs_psample_policy, a, >> + ARRAY_SIZE(ovs_psample_policy))) { >> + return EINVAL; >> + } >> + >> + psample->iifindex = nl_attr_get_u16(a[PSAMPLE_ATTR_IIFINDEX]); >> + psample->dp_group_id = >> nl_attr_get_u32(a[PSAMPLE_ATTR_SAMPLE_GROUP]); >> + psample->group_seq = nl_attr_get_u16(a[PSAMPLE_ATTR_GROUP_SEQ]); >> + psample->packet = a[PSAMPLE_ATTR_DATA]; >> + >> + return 0; >> +} >> + >> +static int >> +netdev_tc_psample_parse_packet(struct netdev_tc_psample *psample, >> + struct dpif_upcall_sflow *dupcall) >> +{ >> + const struct gid_node *node; >> + >> + dp_packet_use_stub(&dupcall->packet, >> + CONST_CAST(struct nlattr *, >> + nl_attr_get(psample->packet)) - 1, >> + nl_attr_get_size(psample->packet) + >> + sizeof(struct nlattr)); >> + dp_packet_set_data(&dupcall->packet, >> + (char *) dp_packet_data(&dupcall->packet) + >> + sizeof(struct nlattr)); >> + dp_packet_set_size(&dupcall->packet, >> nl_attr_get_size(psample->packet)); >> + >> + node = gid_find(psample->dp_group_id); >> + dupcall->sflow_attr = &node->sflow; >> + dupcall->iifindex = psample->iifindex; >> + >> + return 0; >> +} >> + >> +static int >> +netdev_tc_psample_poll(struct dpif_upcall_sflow *dupcall, >> + struct nl_sock *sock) >> +{ >> + for (;;) { >> + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); >> + struct netdev_tc_psample psample; >> + uint64_t buf_stub[4096 / 8]; >> + struct ofpbuf buf; >> + int error; >> + >> + ofpbuf_use_stub(&buf, buf_stub, sizeof buf_stub); >> + error = nl_sock_recv(sock, &buf, NULL, false); >> + >> + if (!error) { >> + error = netdev_tc_psample_from_ofpbuf(&psample, &buf); >> + if (!error) { >> + ofpbuf_uninit(&buf); >> + error = netdev_tc_psample_parse_packet(&psample, >> dupcall); >> + return error; >> + } >> + } else if (error != EAGAIN) { >> + VLOG_WARN_RL(&rl, "%s: error reading or parsing netlink >> (%s)", >> + __func__, ovs_strerror(error)); >> + nl_sock_drain(sock); >> + error = ENOBUFS; >> + } >> + >> + ofpbuf_uninit(&buf); >> + if (error) { >> + return error; >> + } >> + } >> +} >> + >> +static void * >> +netdev_tc_psample_handler(void *arg) >> +{ >> + struct nl_sock *sock = CONST_CAST(struct nl_sock *, arg); >> + >> + struct dpif_upcall_sflow dupcall; >> + int err; >> + >> + while (true) { >> + err = netdev_tc_psample_poll(&dupcall, sock); >> + if (!err) { > > I have not (fully) reviewed the patches below: > > [PATCH v9 03/11] dpif: Introduce register sFlow upcall > [PATCH v9 04/11] ofproto: Add upcall callback to proces > [PATCH v9 05/11] netdev-offload: Introduce register sFlow upcall > callback API > [PATCH v9 06/11] netdev-offload-tc: Implement register sFlow upcall > callback API > [PATCH v9 07/11] dpif-netlink: Implement register sFlow Upcall > callback API > > Reason being that adding all this additional infrastructure does not > make sense to me. > We should just use the existing register_upcall_cb, and call it when > we have an sflow packet via netlink. > > We can easily format the data such that > upcall_cb()->upcall_receive()->process_upcall() will > handle the sflow packet as it came from a normal upcall. > > What do you think? Actually, even if we re-use upcall_cb, patch #5, #6 and #7 are needed. Only patch #3 is not needed. I think we need more change for ofproto to make upcall_cb works for sFlow offload. And many of the arguments of upcall_cb are not needed for sFlow offload. If you let me choose, I'll think creating a new callback for sFlow offload is simple and clean. > >> + upcall_cb(&dupcall); >> + } >> + nl_sock_wait(sock, POLLIN); >> + poll_block(); >> + } >> + >> + return NULL; >> +} >> + >> static int >> netdev_tc_init_flow_api(struct netdev *netdev) >> { >> @@ -2324,12 +2443,18 @@ netdev_tc_init_flow_api(struct netdev *netdev) >> tc_add_del_qdisc(ifindex, false, 0, hook); >> >> if (ovsthread_once_start(&once)) { >> + struct nl_sock *sock; >> + >> probe_tc_block_support(ifindex); >> /* Need to re-fetch block id as it depends on feature >> availability. */ >> block_id = get_block_id_from_netdev(netdev); >> >> probe_multi_mask_per_prio(ifindex); >> - netdev_tc_psample_init(); >> + sock = netdev_tc_psample_init(); >> + if (sock) { >> + ovs_thread_create("psample_handler", >> netdev_tc_psample_handler, >> + sock); >> + } >> ovsthread_once_done(&once); >> } >> >> -- >> 2.26.2 >
On 22 Jan 2021, at 4:38, Chris Mi wrote: > On 1/13/2021 3:53 AM, Eelco Chaudron wrote: >> >> >> On 15 Dec 2020, at 4:38, Chris Mi wrote: >> >>> Create a dedicated thread to poll psample netlink socket, receive >>> sampled packet, parse it to sFlow format and send it to sFlow >>> monitoring host. >>> >>> Signed-off-by: Chris Mi <cmi@nvidia.com> >>> Reviewed-by: Eli Britstein <elibr@nvidia.com> >>> --- >>> lib/netdev-offload-tc.c | 127 >>> +++++++++++++++++++++++++++++++++++++++- >>> 1 file changed, 126 insertions(+), 1 deletion(-) >>> >>> diff --git a/lib/netdev-offload-tc.c b/lib/netdev-offload-tc.c >>> index 61c5a14e6..e5bf719a4 100644 >>> --- a/lib/netdev-offload-tc.c >>> +++ b/lib/netdev-offload-tc.c >>> @@ -19,12 +19,14 @@ >>> #include <errno.h> >>> #include <linux/if_ether.h> >>> #include <linux/psample.h> >>> +#include <poll.h> >>> >>> #include "dpif.h" >>> #include "hash.h" >>> #include "openvswitch/hmap.h" >>> #include "openvswitch/match.h" >>> #include "openvswitch/ofpbuf.h" >>> +#include "openvswitch/poll-loop.h" >>> #include "openvswitch/thread.h" >>> #include "openvswitch/types.h" >>> #include "openvswitch/util.h" >>> @@ -2295,6 +2297,123 @@ netdev_tc_psample_init(void) >>> return sock; >>> } >>> >>> +struct netdev_tc_psample { >>> + struct nlattr *packet; /* packet data */ >>> + int dp_group_id; /* mapping id for >>> sFlow offload */ >>> + int iifindex; /* input ifindex >>> */ >>> + int group_seq; /* group sequence >>> */ >>> +}; >>> + >>> +static int >>> +netdev_tc_psample_from_ofpbuf(struct netdev_tc_psample *psample, >>> + const >>> struct ofpbuf *buf) >>> +{ >>> + static const struct nl_policy ovs_psample_policy[] = { >>> + [PSAMPLE_ATTR_IIFINDEX] = { .type = NL_A_U16 }, >>> + [PSAMPLE_ATTR_SAMPLE_GROUP] = { .type = NL_A_U32 }, >>> + [PSAMPLE_ATTR_GROUP_SEQ] = { .type = NL_A_U32 }, >>> + [PSAMPLE_ATTR_DATA] = { .type = NL_A_UNSPEC }, >>> + }; >>> + struct nlattr *a[ARRAY_SIZE(ovs_psample_policy)]; >>> + struct genlmsghdr *genl; >>> + struct nlmsghdr *nlmsg; >>> + struct ofpbuf b; >>> + >>> + b = ofpbuf_const_initializer(buf->data, buf->size); >>> + nlmsg = ofpbuf_try_pull(&b, sizeof *nlmsg); >>> + genl = ofpbuf_try_pull(&b, sizeof *genl); >>> + if (!nlmsg || !genl || nlmsg->nlmsg_type != psample_family >>> + || !nl_policy_parse(&b, 0, ovs_psample_policy, a, >>> + >>> ARRAY_SIZE(ovs_psample_policy))) { >>> + return EINVAL; >>> + } >>> + >>> + psample->iifindex = >>> nl_attr_get_u16(a[PSAMPLE_ATTR_IIFINDEX]); >>> + psample->dp_group_id = >>> nl_attr_get_u32(a[PSAMPLE_ATTR_SAMPLE_GROUP]); >>> + psample->group_seq = >>> nl_attr_get_u16(a[PSAMPLE_ATTR_GROUP_SEQ]); >>> + psample->packet = a[PSAMPLE_ATTR_DATA]; >>> + >>> + return 0; >>> +} >>> + >>> +static int >>> +netdev_tc_psample_parse_packet(struct netdev_tc_psample *psample, >>> + struct >>> dpif_upcall_sflow *dupcall) >>> +{ >>> + const struct gid_node *node; >>> + >>> + dp_packet_use_stub(&dupcall->packet, >>> + CONST_CAST(struct >>> nlattr *, >>> + nl_attr_get(psample->packet)) - 1, >>> + >>> nl_attr_get_size(psample->packet) + >>> + sizeof(struct >>> nlattr)); >>> + dp_packet_set_data(&dupcall->packet, >>> + (char *) >>> dp_packet_data(&dupcall->packet) + >>> + sizeof(struct >>> nlattr)); >>> + dp_packet_set_size(&dupcall->packet, >>> nl_attr_get_size(psample->packet)); >>> + >>> + node = gid_find(psample->dp_group_id); >>> + dupcall->sflow_attr = &node->sflow; >>> + dupcall->iifindex = psample->iifindex; >>> + >>> + return 0; >>> +} >>> + >>> +static int >>> +netdev_tc_psample_poll(struct dpif_upcall_sflow *dupcall, >>> + struct nl_sock *sock) >>> +{ >>> + for (;;) { >>> + static struct vlog_rate_limit rl = >>> VLOG_RATE_LIMIT_INIT(1, 5); >>> + struct netdev_tc_psample psample; >>> + uint64_t buf_stub[4096 / 8]; >>> + struct ofpbuf buf; >>> + int error; >>> + >>> + ofpbuf_use_stub(&buf, buf_stub, sizeof buf_stub); >>> + error = nl_sock_recv(sock, &buf, NULL, false); >>> + >>> + if (!error) { >>> + error = >>> netdev_tc_psample_from_ofpbuf(&psample, &buf); >>> + if (!error) { >>> + ofpbuf_uninit(&buf); >>> + error = >>> netdev_tc_psample_parse_packet(&psample, dupcall); >>> + return error; >>> + } >>> + } else if (error != EAGAIN) { >>> + VLOG_WARN_RL(&rl, "%s: error reading or >>> parsing netlink (%s)", >>> + __func__, >>> ovs_strerror(error)); >>> + nl_sock_drain(sock); >>> + error = ENOBUFS; >>> + } >>> + >>> + ofpbuf_uninit(&buf); >>> + if (error) { >>> + return error; >>> + } >>> + } >>> +} >>> + >>> +static void * >>> +netdev_tc_psample_handler(void *arg) >>> +{ >>> + struct nl_sock *sock = CONST_CAST(struct nl_sock *, arg); >>> + >>> + struct dpif_upcall_sflow dupcall; >>> + int err; >>> + >>> + while (true) { >>> + err = netdev_tc_psample_poll(&dupcall, sock); >>> + if (!err) { >> >> I have not (fully) reviewed the patches below: >> >> [PATCH v9 03/11] dpif: Introduce register sFlow upcall >> [PATCH v9 04/11] ofproto: Add upcall callback to proces >> [PATCH v9 05/11] netdev-offload: Introduce register sFlow upcall >> callback API >> [PATCH v9 06/11] netdev-offload-tc: Implement register sFlow upcall >> callback API >> [PATCH v9 07/11] dpif-netlink: Implement register sFlow Upcall >> callback API >> >> Reason being that adding all this additional infrastructure does not >> make sense to me. >> We should just use the existing register_upcall_cb, and call it when >> we have an sflow packet via netlink. >> >> We can easily format the data such that >> upcall_cb()->upcall_receive()->process_upcall() will >> handle the sflow packet as it came from a normal upcall. >> >> What do you think? > Actually, even if we re-use upcall_cb, patch #5, #6 and #7 are needed. > Only patch #3 is not needed. > I think we need more change for ofproto to make upcall_cb works for > sFlow offload. > And many of the arguments of upcall_cb are not needed for sFlow > offload. > If you let me choose, I'll think creating a new callback for sFlow > offload is simple and clean. I would like some other people’s feedback on this because if we add this callback (for a slow patch action), we might end up with more in the future. Ilya? >>> + upcall_cb(&dupcall); >>> + } >>> + nl_sock_wait(sock, POLLIN); >>> + poll_block(); >>> + } >>> + >>> + return NULL; >>> +} >>> + >>> static int >>> netdev_tc_init_flow_api(struct netdev *netdev) >>> { >>> @@ -2324,12 +2443,18 @@ netdev_tc_init_flow_api(struct netdev >>> *netdev) >>> tc_add_del_qdisc(ifindex, false, 0, hook); >>> >>> if (ovsthread_once_start(&once)) { >>> + struct nl_sock *sock; >>> + >>> probe_tc_block_support(ifindex); >>> /* Need to re-fetch block id as it depends on >>> feature availability. */ >>> block_id = get_block_id_from_netdev(netdev); >>> >>> probe_multi_mask_per_prio(ifindex); >>> - netdev_tc_psample_init(); >>> + sock = netdev_tc_psample_init(); >>> + if (sock) { >>> + ovs_thread_create("psample_handler", >>> netdev_tc_psample_handler, >>> + sock); >>> + } >>> ovsthread_once_done(&once); >>> } >>> >>> -- >>> 2.26.2 >>
Hi Ilya, On 1/25/2021 5:04 PM, Eelco Chaudron wrote: > > > On 22 Jan 2021, at 4:38, Chris Mi wrote: > >> On 1/13/2021 3:53 AM, Eelco Chaudron wrote: >>> >>> >>> On 15 Dec 2020, at 4:38, Chris Mi wrote: >>> >>>> Create a dedicated thread to poll psample netlink socket, receive >>>> sampled packet, parse it to sFlow format and send it to sFlow >>>> monitoring host. >>>> >>>> Signed-off-by: Chris Mi <cmi@nvidia.com> >>>> Reviewed-by: Eli Britstein <elibr@nvidia.com> >>>> --- >>>> lib/netdev-offload-tc.c | 127 >>>> +++++++++++++++++++++++++++++++++++++++- >>>> 1 file changed, 126 insertions(+), 1 deletion(-) >>>> >>>> diff --git a/lib/netdev-offload-tc.c b/lib/netdev-offload-tc.c >>>> index 61c5a14e6..e5bf719a4 100644 >>>> --- a/lib/netdev-offload-tc.c >>>> +++ b/lib/netdev-offload-tc.c >>>> @@ -19,12 +19,14 @@ >>>> #include <errno.h> >>>> #include <linux/if_ether.h> >>>> #include <linux/psample.h> >>>> +#include <poll.h> >>>> >>>> #include "dpif.h" >>>> #include "hash.h" >>>> #include "openvswitch/hmap.h" >>>> #include "openvswitch/match.h" >>>> #include "openvswitch/ofpbuf.h" >>>> +#include "openvswitch/poll-loop.h" >>>> #include "openvswitch/thread.h" >>>> #include "openvswitch/types.h" >>>> #include "openvswitch/util.h" >>>> @@ -2295,6 +2297,123 @@ netdev_tc_psample_init(void) >>>> return sock; >>>> } >>>> >>>> +struct netdev_tc_psample { >>>> + struct nlattr *packet; /* packet data */ >>>> + int dp_group_id; /* mapping id for sFlow offload */ >>>> + int iifindex; /* input ifindex */ >>>> + int group_seq; /* group sequence */ >>>> +}; >>>> + >>>> +static int >>>> +netdev_tc_psample_from_ofpbuf(struct netdev_tc_psample *psample, >>>> + const struct ofpbuf *buf) >>>> +{ >>>> + static const struct nl_policy ovs_psample_policy[] = { >>>> + [PSAMPLE_ATTR_IIFINDEX] = { .type = NL_A_U16 }, >>>> + [PSAMPLE_ATTR_SAMPLE_GROUP] = { .type = NL_A_U32 }, >>>> + [PSAMPLE_ATTR_GROUP_SEQ] = { .type = NL_A_U32 }, >>>> + [PSAMPLE_ATTR_DATA] = { .type = NL_A_UNSPEC }, >>>> + }; >>>> + struct nlattr *a[ARRAY_SIZE(ovs_psample_policy)]; >>>> + struct genlmsghdr *genl; >>>> + struct nlmsghdr *nlmsg; >>>> + struct ofpbuf b; >>>> + >>>> + b = ofpbuf_const_initializer(buf->data, buf->size); >>>> + nlmsg = ofpbuf_try_pull(&b, sizeof *nlmsg); >>>> + genl = ofpbuf_try_pull(&b, sizeof *genl); >>>> + if (!nlmsg || !genl || nlmsg->nlmsg_type != psample_family >>>> + || !nl_policy_parse(&b, 0, ovs_psample_policy, a, >>>> + ARRAY_SIZE(ovs_psample_policy))) { >>>> + return EINVAL; >>>> + } >>>> + >>>> + psample->iifindex = nl_attr_get_u16(a[PSAMPLE_ATTR_IIFINDEX]); >>>> + psample->dp_group_id = >>>> nl_attr_get_u32(a[PSAMPLE_ATTR_SAMPLE_GROUP]); >>>> + psample->group_seq = nl_attr_get_u16(a[PSAMPLE_ATTR_GROUP_SEQ]); >>>> + psample->packet = a[PSAMPLE_ATTR_DATA]; >>>> + >>>> + return 0; >>>> +} >>>> + >>>> +static int >>>> +netdev_tc_psample_parse_packet(struct netdev_tc_psample *psample, >>>> + struct dpif_upcall_sflow *dupcall) >>>> +{ >>>> + const struct gid_node *node; >>>> + >>>> + dp_packet_use_stub(&dupcall->packet, >>>> + CONST_CAST(struct nlattr *, >>>> + nl_attr_get(psample->packet)) - 1, >>>> + nl_attr_get_size(psample->packet) + >>>> + sizeof(struct nlattr)); >>>> + dp_packet_set_data(&dupcall->packet, >>>> + (char *) dp_packet_data(&dupcall->packet) + >>>> + sizeof(struct nlattr)); >>>> + dp_packet_set_size(&dupcall->packet, >>>> nl_attr_get_size(psample->packet)); >>>> + >>>> + node = gid_find(psample->dp_group_id); >>>> + dupcall->sflow_attr = &node->sflow; >>>> + dupcall->iifindex = psample->iifindex; >>>> + >>>> + return 0; >>>> +} >>>> + >>>> +static int >>>> +netdev_tc_psample_poll(struct dpif_upcall_sflow *dupcall, >>>> + struct nl_sock *sock) >>>> +{ >>>> + for (;;) { >>>> + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, >>>> 5); >>>> + struct netdev_tc_psample psample; >>>> + uint64_t buf_stub[4096 / 8]; >>>> + struct ofpbuf buf; >>>> + int error; >>>> + >>>> + ofpbuf_use_stub(&buf, buf_stub, sizeof buf_stub); >>>> + error = nl_sock_recv(sock, &buf, NULL, false); >>>> + >>>> + if (!error) { >>>> + error = netdev_tc_psample_from_ofpbuf(&psample, &buf); >>>> + if (!error) { >>>> + ofpbuf_uninit(&buf); >>>> + error = >>>> netdev_tc_psample_parse_packet(&psample, dupcall); >>>> + return error; >>>> + } >>>> + } else if (error != EAGAIN) { >>>> + VLOG_WARN_RL(&rl, "%s: error reading or parsing >>>> netlink (%s)", >>>> + __func__, ovs_strerror(error)); >>>> + nl_sock_drain(sock); >>>> + error = ENOBUFS; >>>> + } >>>> + >>>> + ofpbuf_uninit(&buf); >>>> + if (error) { >>>> + return error; >>>> + } >>>> + } >>>> +} >>>> + >>>> +static void * >>>> +netdev_tc_psample_handler(void *arg) >>>> +{ >>>> + struct nl_sock *sock = CONST_CAST(struct nl_sock *, arg); >>>> + >>>> + struct dpif_upcall_sflow dupcall; >>>> + int err; >>>> + >>>> + while (true) { >>>> + err = netdev_tc_psample_poll(&dupcall, sock); >>>> + if (!err) { >>> >>> I have not (fully) reviewed the patches below: >>> >>> [PATCH v9 03/11] dpif: Introduce register sFlow upcall >>> [PATCH v9 04/11] ofproto: Add upcall callback to proces >>> [PATCH v9 05/11] netdev-offload: Introduce register sFlow upcall >>> callback API >>> [PATCH v9 06/11] netdev-offload-tc: Implement register sFlow upcall >>> callback API >>> [PATCH v9 07/11] dpif-netlink: Implement register sFlow Upcall >>> callback API >>> >>> Reason being that adding all this additional infrastructure does not >>> make sense to me. >>> We should just use the existing register_upcall_cb, and call it when >>> we have an sflow packet via netlink. >>> >>> We can easily format the data such that >>> upcall_cb()->upcall_receive()->process_upcall() will >>> handle the sflow packet as it came from a normal upcall. >>> >>> What do you think? >> Actually, even if we re-use upcall_cb, patch #5, #6 and #7 are >> needed. Only patch #3 is not needed. >> I think we need more change for ofproto to make upcall_cb works for >> sFlow offload. >> And many of the arguments of upcall_cb are not needed for sFlow offload. >> If you let me choose, I'll think creating a new callback for sFlow >> offload is simple and clean. > > I would like some other people’s feedback on this because if we add > this callback (for a slow patch action), we might end up with more in > the future. Ilya? Could I know what's your opinion about it? And I'm wondering if you have time to review other patches? Thanks, Chris > >>>> + upcall_cb(&dupcall); >>>> + } >>>> + nl_sock_wait(sock, POLLIN); >>>> + poll_block(); >>>> + } >>>> + >>>> + return NULL; >>>> +} >>>> + >>>> static int >>>> netdev_tc_init_flow_api(struct netdev *netdev) >>>> { >>>> @@ -2324,12 +2443,18 @@ netdev_tc_init_flow_api(struct netdev *netdev) >>>> tc_add_del_qdisc(ifindex, false, 0, hook); >>>> >>>> if (ovsthread_once_start(&once)) { >>>> + struct nl_sock *sock; >>>> + >>>> probe_tc_block_support(ifindex); >>>> /* Need to re-fetch block id as it depends on feature >>>> availability. */ >>>> block_id = get_block_id_from_netdev(netdev); >>>> >>>> probe_multi_mask_per_prio(ifindex); >>>> - netdev_tc_psample_init(); >>>> + sock = netdev_tc_psample_init(); >>>> + if (sock) { >>>> + ovs_thread_create("psample_handler", >>>> netdev_tc_psample_handler, >>>> + sock); >>>> + } >>>> ovsthread_once_done(&once); >>>> } >>>> >>>> -- >>>> 2.26.2 >>> >
diff --git a/lib/netdev-offload-tc.c b/lib/netdev-offload-tc.c index 61c5a14e6..e5bf719a4 100644 --- a/lib/netdev-offload-tc.c +++ b/lib/netdev-offload-tc.c @@ -19,12 +19,14 @@ #include <errno.h> #include <linux/if_ether.h> #include <linux/psample.h> +#include <poll.h> #include "dpif.h" #include "hash.h" #include "openvswitch/hmap.h" #include "openvswitch/match.h" #include "openvswitch/ofpbuf.h" +#include "openvswitch/poll-loop.h" #include "openvswitch/thread.h" #include "openvswitch/types.h" #include "openvswitch/util.h" @@ -2295,6 +2297,123 @@ netdev_tc_psample_init(void) return sock; } +struct netdev_tc_psample { + struct nlattr *packet; /* packet data */ + int dp_group_id; /* mapping id for sFlow offload */ + int iifindex; /* input ifindex */ + int group_seq; /* group sequence */ +}; + +static int +netdev_tc_psample_from_ofpbuf(struct netdev_tc_psample *psample, + const struct ofpbuf *buf) +{ + static const struct nl_policy ovs_psample_policy[] = { + [PSAMPLE_ATTR_IIFINDEX] = { .type = NL_A_U16 }, + [PSAMPLE_ATTR_SAMPLE_GROUP] = { .type = NL_A_U32 }, + [PSAMPLE_ATTR_GROUP_SEQ] = { .type = NL_A_U32 }, + [PSAMPLE_ATTR_DATA] = { .type = NL_A_UNSPEC }, + }; + struct nlattr *a[ARRAY_SIZE(ovs_psample_policy)]; + struct genlmsghdr *genl; + struct nlmsghdr *nlmsg; + struct ofpbuf b; + + b = ofpbuf_const_initializer(buf->data, buf->size); + nlmsg = ofpbuf_try_pull(&b, sizeof *nlmsg); + genl = ofpbuf_try_pull(&b, sizeof *genl); + if (!nlmsg || !genl || nlmsg->nlmsg_type != psample_family + || !nl_policy_parse(&b, 0, ovs_psample_policy, a, + ARRAY_SIZE(ovs_psample_policy))) { + return EINVAL; + } + + psample->iifindex = nl_attr_get_u16(a[PSAMPLE_ATTR_IIFINDEX]); + psample->dp_group_id = nl_attr_get_u32(a[PSAMPLE_ATTR_SAMPLE_GROUP]); + psample->group_seq = nl_attr_get_u16(a[PSAMPLE_ATTR_GROUP_SEQ]); + psample->packet = a[PSAMPLE_ATTR_DATA]; + + return 0; +} + +static int +netdev_tc_psample_parse_packet(struct netdev_tc_psample *psample, + struct dpif_upcall_sflow *dupcall) +{ + const struct gid_node *node; + + dp_packet_use_stub(&dupcall->packet, + CONST_CAST(struct nlattr *, + nl_attr_get(psample->packet)) - 1, + nl_attr_get_size(psample->packet) + + sizeof(struct nlattr)); + dp_packet_set_data(&dupcall->packet, + (char *) dp_packet_data(&dupcall->packet) + + sizeof(struct nlattr)); + dp_packet_set_size(&dupcall->packet, nl_attr_get_size(psample->packet)); + + node = gid_find(psample->dp_group_id); + dupcall->sflow_attr = &node->sflow; + dupcall->iifindex = psample->iifindex; + + return 0; +} + +static int +netdev_tc_psample_poll(struct dpif_upcall_sflow *dupcall, + struct nl_sock *sock) +{ + for (;;) { + static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 5); + struct netdev_tc_psample psample; + uint64_t buf_stub[4096 / 8]; + struct ofpbuf buf; + int error; + + ofpbuf_use_stub(&buf, buf_stub, sizeof buf_stub); + error = nl_sock_recv(sock, &buf, NULL, false); + + if (!error) { + error = netdev_tc_psample_from_ofpbuf(&psample, &buf); + if (!error) { + ofpbuf_uninit(&buf); + error = netdev_tc_psample_parse_packet(&psample, dupcall); + return error; + } + } else if (error != EAGAIN) { + VLOG_WARN_RL(&rl, "%s: error reading or parsing netlink (%s)", + __func__, ovs_strerror(error)); + nl_sock_drain(sock); + error = ENOBUFS; + } + + ofpbuf_uninit(&buf); + if (error) { + return error; + } + } +} + +static void * +netdev_tc_psample_handler(void *arg) +{ + struct nl_sock *sock = CONST_CAST(struct nl_sock *, arg); + + struct dpif_upcall_sflow dupcall; + int err; + + while (true) { + err = netdev_tc_psample_poll(&dupcall, sock); + if (!err) { + upcall_cb(&dupcall); + } + nl_sock_wait(sock, POLLIN); + poll_block(); + } + + return NULL; +} + static int netdev_tc_init_flow_api(struct netdev *netdev) { @@ -2324,12 +2443,18 @@ netdev_tc_init_flow_api(struct netdev *netdev) tc_add_del_qdisc(ifindex, false, 0, hook); if (ovsthread_once_start(&once)) { + struct nl_sock *sock; + probe_tc_block_support(ifindex); /* Need to re-fetch block id as it depends on feature availability. */ block_id = get_block_id_from_netdev(netdev); probe_multi_mask_per_prio(ifindex); - netdev_tc_psample_init(); + sock = netdev_tc_psample_init(); + if (sock) { + ovs_thread_create("psample_handler", netdev_tc_psample_handler, + sock); + } ovsthread_once_done(&once); }