On Thu, Aug 1, 2024 at 1:30 PM Zijian Zhang <zijianzhang@xxxxxxxxxxxxx> wrote: > > On 7/31/24 3:32 PM, Willem de Bruijn wrote: > > zijianzhang@ wrote: > >> From: Zijian Zhang <zijianzhang@xxxxxxxxxxxxx> > >> > >> We update selftests/net/msg_zerocopy.c to accommodate the new mechanism, > > First of all, thanks for the detailed suggestions! > > > > > Please make commit messages stand on their own. If someone does a git > > blame, make the message self explanatory. Replace "the new mechanism" > > with sendmsg SCM_ZC_NOTIFICATION. > > > > In patch 2 or as a separate patch 4, also add a new short section on > > this API in Documentation/networking/msg_zerocopy.rst. Probably with > > the same contents as a good explanation of the feature in the commit > > message of patch 2. > > > > Agreed. > > >> cfg_notification_limit has the same semantics for both methods. Test > >> results are as follows, we update skb_orphan_frags_rx to the same as > >> skb_orphan_frags to support zerocopy in the localhost test. > >> > >> cfg_notification_limit = 1, both method get notifications after 1 calling > >> of sendmsg. In this case, the new method has around 17% cpu savings in TCP > >> and 23% cpu savings in UDP. > >> +---------------------+---------+---------+---------+---------+ > >> | Test Type / Protocol| TCP v4 | TCP v6 | UDP v4 | UDP v6 | > >> +---------------------+---------+---------+---------+---------+ > >> | ZCopy (MB) | 7523 | 7706 | 7489 | 7304 | > >> +---------------------+---------+---------+---------+---------+ > >> | New ZCopy (MB) | 8834 | 8993 | 9053 | 9228 | > >> +---------------------+---------+---------+---------+---------+ > >> | New ZCopy / ZCopy | 117.42% | 116.70% | 120.88% | 126.34% | > >> +---------------------+---------+---------+---------+---------+ > >> > >> cfg_notification_limit = 32, both get notifications after 32 calling of > >> sendmsg, which means more chances to coalesce notifications, and less > >> overhead of poll + recvmsg for the original method. In this case, the new > >> method has around 7% cpu savings in TCP and slightly better cpu usage in > >> UDP. In the env of selftest, notifications of TCP are more likely to be > >> out of order than UDP, it's easier to coalesce more notifications in UDP. > >> The original method can get one notification with range of 32 in a recvmsg > >> most of the time. In TCP, most notifications' range is around 2, so the > >> original method needs around 16 recvmsgs to get notified in one round. > >> That's the reason for the "New ZCopy / ZCopy" diff in TCP and UDP here. > >> +---------------------+---------+---------+---------+---------+ > >> | Test Type / Protocol| TCP v4 | TCP v6 | UDP v4 | UDP v6 | > >> +---------------------+---------+---------+---------+---------+ > >> | ZCopy (MB) | 8842 | 8735 | 10072 | 9380 | > >> +---------------------+---------+---------+---------+---------+ > >> | New ZCopy (MB) | 9366 | 9477 | 10108 | 9385 | > >> +---------------------+---------+---------+---------+---------+ > >> | New ZCopy / ZCopy | 106.00% | 108.28% | 100.31% | 100.01% | > >> +---------------------+---------+---------+---------+---------+ > >> > >> In conclusion, when notification interval is small or notifications are > >> hard to be coalesced, the new mechanism is highly recommended. Otherwise, > >> the performance gain from the new mechanism is very limited. > >> > >> Signed-off-by: Zijian Zhang <zijianzhang@xxxxxxxxxxxxx> > >> Signed-off-by: Xiaochun Lu <xiaochun.lu@xxxxxxxxxxxxx> > > > >> -static bool do_sendmsg(int fd, struct msghdr *msg, bool do_zerocopy, int domain) > >> +static void add_zcopy_info(struct msghdr *msg) > >> +{ > >> + struct zc_info *zc_info; > >> + struct cmsghdr *cm; > >> + > >> + if (!msg->msg_control) > >> + error(1, errno, "NULL user arg"); > > > > Don't add precondition checks for code entirely under your control. > > This is not a user API. > > > > Ack. > > >> + cm = (struct cmsghdr *)msg->msg_control; > >> + cm->cmsg_len = CMSG_LEN(ZC_INFO_SIZE); > >> + cm->cmsg_level = SOL_SOCKET; > >> + cm->cmsg_type = SCM_ZC_NOTIFICATION; > >> + > >> + zc_info = (struct zc_info *)CMSG_DATA(cm); > >> + zc_info->size = ZC_NOTIFICATION_MAX; > >> + > >> + added_zcopy_info = true; > > > > Just initialize every time? Is this here to reuse the same msg_control > > as long as metadata is returned? > > > > Yes, the same msg_control will be reused. > > The overall paradiagm is, > start: > sendmsg(..) > sendmsg(..) > ... sends_since_notify sendmsgs in total > > add_zcopy_info(..) > sendmsg(.., msg_control) > do_recv_completions_sendmsg(..) > goto start; > > if (sends_since_notify + 1 >= cfg_notification_limit), add_zcopy_info > will be invoked, and the right next sendmsg will have the msg_control > passed in. > > If (added_zcopy_info), do_recv_completions_sendmsg will be invoked, > and added_zcopy_info will be set to false in it. This does not seem like it would need a global variable? > >> +} > >> + > >> +static bool do_sendmsg(int fd, struct msghdr *msg, > >> + enum notification_type do_zerocopy, int domain) > >> { > >> int ret, len, i, flags; > >> static uint32_t cookie; > >> @@ -200,6 +233,12 @@ static bool do_sendmsg(int fd, struct msghdr *msg, bool do_zerocopy, int domain) > >> msg->msg_controllen = CMSG_SPACE(sizeof(cookie)); > >> msg->msg_control = (struct cmsghdr *)ckbuf; > >> add_zcopy_cookie(msg, ++cookie); > >> + } else if (do_zerocopy == MSG_ZEROCOPY_NOTIFY_SENDMSG && > >> + sends_since_notify + 1 >= cfg_notification_limit) { > >> + memset(&msg->msg_control, 0, sizeof(msg->msg_control)); > >> + msg->msg_controllen = CMSG_SPACE(ZC_INFO_SIZE); > >> + msg->msg_control = (struct cmsghdr *)zc_ckbuf; > >> + add_zcopy_info(msg); > >> } > >> } > >> > >> @@ -218,7 +257,7 @@ static bool do_sendmsg(int fd, struct msghdr *msg, bool do_zerocopy, int domain) > >> if (do_zerocopy && ret) > >> expected_completions++; > >> } > >> - if (do_zerocopy && domain == PF_RDS) { > >> + if (msg->msg_control) { > >> msg->msg_control = NULL; > >> msg->msg_controllen = 0; > >> } > >> @@ -466,6 +505,44 @@ static void do_recv_completions(int fd, int domain) > >> sends_since_notify = 0; > >> } > >> > >> +static void do_recv_completions2(void) > > > > functionname2 is very uninformative. > > > > do_recv_completions_sendmsg or so. > > > > Ack. > > >> +{ > >> + struct cmsghdr *cm = (struct cmsghdr *)zc_ckbuf; > >> + struct zc_info *zc_info; > >> + __u32 hi, lo, range; > >> + __u8 zerocopy; > >> + int i; > >> + > >> + zc_info = (struct zc_info *)CMSG_DATA(cm); > >> + for (i = 0; i < zc_info->size; i++) { > >> + hi = zc_info->arr[i].hi; > >> + lo = zc_info->arr[i].lo; > >> + zerocopy = zc_info->arr[i].zerocopy; > >> + range = hi - lo + 1; > >> + > >> + if (cfg_verbose && lo != next_completion) > >> + fprintf(stderr, "gap: %u..%u does not append to %u\n", > >> + lo, hi, next_completion); > >> + next_completion = hi + 1; > >> + > >> + if (zerocopied == -1) { > >> + zerocopied = zerocopy; > >> + } else if (zerocopied != zerocopy) { > >> + fprintf(stderr, "serr: inconsistent\n"); > >> + zerocopied = zerocopy; > >> + } > >> + > >> + completions += range; > >> + sends_since_notify -= range; > >> + > >> + if (cfg_verbose >= 2) > >> + fprintf(stderr, "completed: %u (h=%u l=%u)\n", > >> + range, hi, lo); > >> + } > >> + > >> + added_zcopy_info = false; > >> +} > >> + > >> /* Wait for all remaining completions on the errqueue */ > >> static void do_recv_remaining_completions(int fd, int domain) > >> { > >> @@ -553,11 +630,16 @@ static void do_tx(int domain, int type, int protocol) > >> else > >> do_sendmsg(fd, &msg, cfg_zerocopy, domain); > >> > >> - if (cfg_zerocopy && sends_since_notify >= cfg_notification_limit) > >> + if (cfg_zerocopy == MSG_ZEROCOPY_NOTIFY_ERRQUEUE && > >> + sends_since_notify >= cfg_notification_limit) > >> do_recv_completions(fd, domain); > >> > >> + if (cfg_zerocopy == MSG_ZEROCOPY_NOTIFY_SENDMSG && > >> + added_zcopy_info) > >> + do_recv_completions2(); > >> + > >> while (!do_poll(fd, POLLOUT)) { > >> - if (cfg_zerocopy) > >> + if (cfg_zerocopy == MSG_ZEROCOPY_NOTIFY_ERRQUEUE) > >> do_recv_completions(fd, domain); > >> } > >> > >> @@ -715,7 +797,7 @@ static void parse_opts(int argc, char **argv) > >> > >> cfg_payload_len = max_payload_len; > >> > >> - while ((c = getopt(argc, argv, "46c:C:D:i:l:mp:rs:S:t:vz")) != -1) { > >> + while ((c = getopt(argc, argv, "46c:C:D:i:l:mnp:rs:S:t:vz")) != -1) { > >> switch (c) { > >> case '4': > >> if (cfg_family != PF_UNSPEC) > >> @@ -749,6 +831,9 @@ static void parse_opts(int argc, char **argv) > >> case 'm': > >> cfg_cork_mixed = true; > >> break; > >> + case 'n': > >> + cfg_zerocopy = MSG_ZEROCOPY_NOTIFY_SENDMSG; > >> + break; > > > > How about -Z to make clear that this is still MSG_ZEROCOPY, just with > > a different notification mechanism. > > > > And perhaps add a testcase that exercises both this mechanism and > > existing recvmsg MSG_ERRQUEUE. As they should work in parallel and > > concurrently in a multithreaded environment. > > > > -Z is more clear, and the hybrid testcase will be helpful. > > Btw, before I put some efforts to solve the current issues, I think > I should wait for comments about api change from linux-api@xxxxxxxxxxxxxxx? I'm not sure whether anyone on that list will give feedback. I would continue with revisions at a normal schedule, as long as that stays in the Cc.