| Message ID | 88831c36-a589-429f-8e8b-2ecb66a30263@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | net: gro: encapsulation bug fix and flush checks improvements | expand |
On Sat, Mar 9, 2024 at 4:35 PM Richard Gobert <richardbgobert@gmail.com> wrote: > > {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, > iph->id, ...) against all packets in a loop. These flush checks are > relevant only to tcp flows, and as such they're used to determine whether > the packets can be merged later in tcp_gro_receive. > > These checks are not relevant to UDP packets. I do not think this claim is true. Incoming packets -> GRO -> GSO -> forwarded packets The {GRO,GSO} step must be transparent, GRO is not LRO.
Richard Gobert wrote: > {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, > iph->id, ...) against all packets in a loop. These flush checks are > relevant only to tcp flows, and as such they're used to determine whether > the packets can be merged later in tcp_gro_receive. > > These checks are not relevant to UDP packets. These are network protocol coalescing invariants. Why would they be limited to certain transport protocols only? > Furthermore, they need to be > done only once in tcp_gro_receive and only against the found p skb, since > they only affect flush and not same_flow. > > Levaraging the previous commit in the series, in which correct network > header offsets are saved for both outer and inner network headers - > allowing these checks to be done only once, in tcp_gro_receive. As a > result, NAPI_GRO_CB(p)->flush is not used at all. In addition - flush_id > checks are more declerative and contained in inet_gro_flush, thus removing declarative > the need for flush_id in napi_gro_cb. > > Signed-off-by: Richard Gobert <richardbgobert@gmail.com> > --- > +static int inet_gro_flush(const struct iphdr *iph, const struct iphdr *iph2, > + struct sk_buff *p, u32 outer) > +{ > + const u32 id = ntohl(*(__be32 *)&iph->id); > + const u32 id2 = ntohl(*(__be32 *)&iph2->id); > + const int flush_id = ntohs(id >> 16) - ntohs(id2 >> 16); > + const u16 count = NAPI_GRO_CB(p)->count; > + const u32 df = id & IP_DF; > + u32 is_atomic; > + int flush; > + > + /* All fields must match except length and checksum. */ > + flush = (iph->ttl ^ iph2->ttl) | (iph->tos ^ iph2->tos) | (df ^ (id2 & IP_DF)); > + > + /* When we receive our second frame we can make a decision on if we > + * continue this flow as an atomic flow with a fixed ID or if we use > + * an incremdfenting ID. > + */ Comment became garbled on move: incrementing > + if (count == 1) { > + is_atomic = df && flush_id == 0; > + NAPI_GRO_CB(p)->is_atomic = is_atomic; > + } else { > + is_atomic = df && NAPI_GRO_CB(p)->is_atomic; > + } > + > + /* Ignore outer IP ID value if based on atomic datagram. */ > + outer = (outer && df) - 1; > + is_atomic--; > + > + return flush | ((flush_id ^ (count & is_atomic)) & outer); > +}
Eric Dumazet wrote: > On Sat, Mar 9, 2024 at 4:35 PM Richard Gobert <richardbgobert@gmail.com> wrote: >> >> {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, >> iph->id, ...) against all packets in a loop. These flush checks are >> relevant only to tcp flows, and as such they're used to determine whether >> the packets can be merged later in tcp_gro_receive. >> >> These checks are not relevant to UDP packets. > > I do not think this claim is true. > > Incoming packets -> GRO -> GSO -> forwarded packets > > The {GRO,GSO} step must be transparent, GRO is not LRO. Sorry, I should rephrase myself. The patch preserves the current logic in GRO. These L3 checks (ttl, flags, etc.) are written to NAPI_GRO_CB(p)->{flush,flush_id}, and NAPI_GRO_CB(skb)->is_atomic - and all of these are currently used only in tcp_gro_receive.
Willem de Bruijn wrote: > Richard Gobert wrote: >> {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, >> iph->id, ...) against all packets in a loop. These flush checks are >> relevant only to tcp flows, and as such they're used to determine whether >> the packets can be merged later in tcp_gro_receive. >> >> These checks are not relevant to UDP packets. > > These are network protocol coalescing invariants. Why would they be > limited to certain transport protocols only? Thanks for the review, I'll fix the typos. I replied to Eric's comment about the relevancy of these checks for UDP.
Richard Gobert wrote: > Eric Dumazet wrote: > > On Sat, Mar 9, 2024 at 4:35 PM Richard Gobert <richardbgobert@gmail.com> wrote: > >> > >> {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, > >> iph->id, ...) against all packets in a loop. These flush checks are > >> relevant only to tcp flows, and as such they're used to determine whether > >> the packets can be merged later in tcp_gro_receive. > >> > >> These checks are not relevant to UDP packets. > > > > I do not think this claim is true. > > > > Incoming packets -> GRO -> GSO -> forwarded packets > > > > The {GRO,GSO} step must be transparent, GRO is not LRO. > > Sorry, I should rephrase myself. The patch preserves the > current logic in GRO. These L3 checks (ttl, flags, etc.) are written to > NAPI_GRO_CB(p)->{flush,flush_id}, and NAPI_GRO_CB(skb)->is_atomic - and > all of these are currently used only in tcp_gro_receive. That was perhaps an oversight when adding UDP GRO? Simply because the flush is determined in the innermost callback.
Willem de Bruijn wrote: > Richard Gobert wrote: >> Eric Dumazet wrote: >>> On Sat, Mar 9, 2024 at 4:35 PM Richard Gobert <richardbgobert@gmail.com> wrote: >>>> >>>> {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, >>>> iph->id, ...) against all packets in a loop. These flush checks are >>>> relevant only to tcp flows, and as such they're used to determine whether >>>> the packets can be merged later in tcp_gro_receive. >>>> >>>> These checks are not relevant to UDP packets. >>> >>> I do not think this claim is true. >>> >>> Incoming packets -> GRO -> GSO -> forwarded packets >>> >>> The {GRO,GSO} step must be transparent, GRO is not LRO. >> >> Sorry, I should rephrase myself. The patch preserves the >> current logic in GRO. These L3 checks (ttl, flags, etc.) are written to >> NAPI_GRO_CB(p)->{flush,flush_id}, and NAPI_GRO_CB(skb)->is_atomic - and >> all of these are currently used only in tcp_gro_receive. > > That was perhaps an oversight when adding UDP GRO? > > Simply because the flush is determined in the innermost callback. It might have been an oversight. From what I have seen it's only relevant to GRO's UDP fraglist path (it was added in 9fd1ff5d ("udp: Support UDP fraglist GRO/GSO.")). That's the only UDP path that calls skb_gro_receive - which may alter the forwarded packets and make GRO/GSO not transparent. AFAIU NAPI_GRO_CB(p)->flush value is not overwritten in encapsulation - it is determined by both outer and inner callbacks. I tried to preserve the current behaviour in GRO - if we want to change this behaviour I'll gladly do it, although I'd prefer to address it in a different patch series. What do you think? Thanks
Richard Gobert wrote: > Willem de Bruijn wrote: > > Richard Gobert wrote: > >> Eric Dumazet wrote: > >>> On Sat, Mar 9, 2024 at 4:35 PM Richard Gobert <richardbgobert@gmail.com> wrote: > >>>> > >>>> {inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, > >>>> iph->id, ...) against all packets in a loop. These flush checks are > >>>> relevant only to tcp flows, and as such they're used to determine whether > >>>> the packets can be merged later in tcp_gro_receive. > >>>> > >>>> These checks are not relevant to UDP packets. > >>> > >>> I do not think this claim is true. > >>> > >>> Incoming packets -> GRO -> GSO -> forwarded packets > >>> > >>> The {GRO,GSO} step must be transparent, GRO is not LRO. > >> > >> Sorry, I should rephrase myself. The patch preserves the > >> current logic in GRO. These L3 checks (ttl, flags, etc.) are written to > >> NAPI_GRO_CB(p)->{flush,flush_id}, and NAPI_GRO_CB(skb)->is_atomic - and > >> all of these are currently used only in tcp_gro_receive. > > > > That was perhaps an oversight when adding UDP GRO? > > > > Simply because the flush is determined in the innermost callback. > > It might have been an oversight. From what I have seen it's only relevant > to GRO's UDP fraglist path (it was added in 9fd1ff5d ("udp: Support UDP > fraglist GRO/GSO.")). That's the only UDP path that calls skb_gro_receive - > which may alter the forwarded packets and make GRO/GSO not transparent. > > AFAIU NAPI_GRO_CB(p)->flush value is not overwritten in encapsulation - it > is determined by both outer and inner callbacks. Thanks for the context > I tried to preserve the current behaviour in GRO - if we want to change > this behaviour I'll gladly do it, although I'd prefer to address it in a > different patch series. What do you think? Yes, it's entirely reasonable to leave that out of this series.
diff --git a/include/net/gro.h b/include/net/gro.h index 9d1389269509..34e50f77f744 100644 --- a/include/net/gro.h +++ b/include/net/gro.h @@ -35,15 +35,15 @@ struct napi_gro_cb { /* This is non-zero if the packet cannot be merged with the new skb. */ u16 flush; - /* Save the IP ID here and check when we get to the transport layer */ - u16 flush_id; - /* Number of segments aggregated. */ u16 count; /* Used in ipv6_gro_receive() and foo-over-udp and esp-in-udp */ u16 proto; + /* used to support CHECKSUM_COMPLETE for tunneling protocols */ + __wsum csum; + /* Used in napi_gro_cb::free */ #define NAPI_GRO_FREE 1 #define NAPI_GRO_FREE_STOLEN_HEAD 2 @@ -84,9 +84,6 @@ struct napi_gro_cb { u8 is_flist:1; ); - /* used to support CHECKSUM_COMPLETE for tunneling protocols */ - __wsum csum; - /* L3 offsets */ union { struct { diff --git a/net/core/gro.c b/net/core/gro.c index 2b42138f816c..128d7b9c8dfb 100644 --- a/net/core/gro.c +++ b/net/core/gro.c @@ -332,8 +332,6 @@ static void gro_list_prepare(const struct list_head *head, list_for_each_entry(p, head, list) { unsigned long diffs; - NAPI_GRO_CB(p)->flush = 0; - if (hash != skb_get_hash_raw(p)) { NAPI_GRO_CB(p)->same_flow = 0; continue; @@ -473,7 +471,6 @@ static enum gro_result dev_gro_receive(struct napi_struct *napi, struct sk_buff sizeof(u32))); /* Avoid slow unaligned acc */ *(u32 *)&NAPI_GRO_CB(skb)->zeroed = 0; NAPI_GRO_CB(skb)->flush = skb_has_frag_list(skb); - NAPI_GRO_CB(skb)->is_atomic = 1; NAPI_GRO_CB(skb)->count = 1; if (unlikely(skb_is_gso(skb))) { NAPI_GRO_CB(skb)->count = skb_shinfo(skb)->gso_segs; diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c index c6bb21c27aee..5b74c6d2ed8b 100644 --- a/net/ipv4/af_inet.c +++ b/net/ipv4/af_inet.c @@ -1512,7 +1512,6 @@ struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) list_for_each_entry(p, head, list) { struct iphdr *iph2; - u16 flush_id; if (!NAPI_GRO_CB(p)->same_flow) continue; @@ -1529,43 +1528,8 @@ struct sk_buff *inet_gro_receive(struct list_head *head, struct sk_buff *skb) NAPI_GRO_CB(p)->same_flow = 0; continue; } - - /* All fields must match except length and checksum. */ - NAPI_GRO_CB(p)->flush |= - (iph->ttl ^ iph2->ttl) | - (iph->tos ^ iph2->tos) | - ((iph->frag_off ^ iph2->frag_off) & htons(IP_DF)); - - NAPI_GRO_CB(p)->flush |= flush; - - /* We need to store of the IP ID check to be included later - * when we can verify that this packet does in fact belong - * to a given flow. - */ - flush_id = (u16)(id - ntohs(iph2->id)); - - /* This bit of code makes it much easier for us to identify - * the cases where we are doing atomic vs non-atomic IP ID - * checks. Specifically an atomic check can return IP ID - * values 0 - 0xFFFF, while a non-atomic check can only - * return 0 or 0xFFFF. - */ - if (!NAPI_GRO_CB(p)->is_atomic || - !(iph->frag_off & htons(IP_DF))) { - flush_id ^= NAPI_GRO_CB(p)->count; - flush_id = flush_id ? 0xFFFF : 0; - } - - /* If the previous IP ID value was based on an atomic - * datagram we can overwrite the value and ignore it. - */ - if (NAPI_GRO_CB(skb)->is_atomic) - NAPI_GRO_CB(p)->flush_id = flush_id; - else - NAPI_GRO_CB(p)->flush_id |= flush_id; } - NAPI_GRO_CB(skb)->is_atomic = !!(iph->frag_off & htons(IP_DF)); NAPI_GRO_CB(skb)->flush |= flush; /* Note : No need to call skb_gro_postpull_rcsum() here, diff --git a/net/ipv4/tcp_offload.c b/net/ipv4/tcp_offload.c index fde800179b2e..c165e72555e1 100644 --- a/net/ipv4/tcp_offload.c +++ b/net/ipv4/tcp_offload.c @@ -178,6 +178,55 @@ struct sk_buff *tcp_gso_segment(struct sk_buff *skb, return segs; } +static int inet_gro_flush(const struct iphdr *iph, const struct iphdr *iph2, + struct sk_buff *p, u32 outer) +{ + const u32 id = ntohl(*(__be32 *)&iph->id); + const u32 id2 = ntohl(*(__be32 *)&iph2->id); + const int flush_id = ntohs(id >> 16) - ntohs(id2 >> 16); + const u16 count = NAPI_GRO_CB(p)->count; + const u32 df = id & IP_DF; + u32 is_atomic; + int flush; + + /* All fields must match except length and checksum. */ + flush = (iph->ttl ^ iph2->ttl) | (iph->tos ^ iph2->tos) | (df ^ (id2 & IP_DF)); + + /* When we receive our second frame we can make a decision on if we + * continue this flow as an atomic flow with a fixed ID or if we use + * an incremdfenting ID. + */ + if (count == 1) { + is_atomic = df && flush_id == 0; + NAPI_GRO_CB(p)->is_atomic = is_atomic; + } else { + is_atomic = df && NAPI_GRO_CB(p)->is_atomic; + } + + /* Ignore outer IP ID value if based on atomic datagram. */ + outer = (outer && df) - 1; + is_atomic--; + + return flush | ((flush_id ^ (count & is_atomic)) & outer); +} + +static int ipv6_gro_flush(const struct ipv6hdr *iph, const struct ipv6hdr *iph2) +{ + /* <Version:4><Traffic_Class:8><Flow_Label:20> */ + __be32 first_word = *(__be32 *)iph ^ *(__be32 *)iph2; + + /* Flush if Traffic Class fields are different. */ + return (first_word & htonl(0x0FF00000)) | + (__force __be32)(iph->hop_limit ^ iph2->hop_limit); +} + +static int gro_network_flush(const void *nh, const void *nh2, + struct sk_buff *p, u32 outer) +{ + return (((struct iphdr *)nh)->version == 6) ? ipv6_gro_flush(nh, nh2) : + inet_gro_flush(nh, nh2, p, outer); +} + struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb) { struct sk_buff *pp = NULL; @@ -190,6 +239,7 @@ struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb) unsigned int mss = 1; unsigned int hlen; unsigned int off; + bool encap_mark; int flush = 1; int i; @@ -232,9 +282,7 @@ struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb) goto out_check_final; found: - /* Include the IP ID check below from the inner most IP hdr */ - flush = NAPI_GRO_CB(p)->flush; - flush |= (__force int)(flags & TCP_FLAG_CWR); + flush = (__force int)(flags & TCP_FLAG_CWR); flush |= (__force int)((flags ^ tcp_flag_word(th2)) & ~(TCP_FLAG_CWR | TCP_FLAG_FIN | TCP_FLAG_PSH)); flush |= (__force int)(th->ack_seq ^ th2->ack_seq); @@ -242,16 +290,14 @@ struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb) flush |= *(u32 *)((u8 *)th + i) ^ *(u32 *)((u8 *)th2 + i); - /* When we receive our second frame we can made a decision on if we - * continue this flow as an atomic flow with a fixed ID or if we use - * an incrementing ID. - */ - if (NAPI_GRO_CB(p)->flush_id != 1 || - NAPI_GRO_CB(p)->count != 1 || - !NAPI_GRO_CB(p)->is_atomic) - flush |= NAPI_GRO_CB(p)->flush_id; - else - NAPI_GRO_CB(p)->is_atomic = false; + encap_mark = NAPI_GRO_CB(skb)->encap_mark; + for (i = 0; i <= encap_mark; i++) { + const u16 diff = off - NAPI_GRO_CB(skb)->network_offsets[i]; + + flush |= gro_network_flush((void *)th - diff, + (void *)th2 - diff, + p, i != encap_mark); + } mss = skb_shinfo(p)->gso_size; diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c index d9d3a6bed510..b1850c20d799 100644 --- a/net/ipv6/ip6_offload.c +++ b/net/ipv6/ip6_offload.c @@ -288,19 +288,8 @@ INDIRECT_CALLABLE_SCOPE struct sk_buff *ipv6_gro_receive(struct list_head *head, nlen - sizeof(struct ipv6hdr))) goto not_same_flow; } - /* flush if Traffic Class fields are different */ - NAPI_GRO_CB(p)->flush |= !!((first_word & htonl(0x0FF00000)) | - (__force __be32)(iph->hop_limit ^ iph2->hop_limit)); - NAPI_GRO_CB(p)->flush |= flush; - - /* If the previous IP ID value was based on an atomic - * datagram we can overwrite the value and ignore it. - */ - if (NAPI_GRO_CB(skb)->is_atomic) - NAPI_GRO_CB(p)->flush_id = 0; } - NAPI_GRO_CB(skb)->is_atomic = true; NAPI_GRO_CB(skb)->flush |= flush; skb_gro_postpull_rcsum(skb, iph, nlen);
{inet,ipv6}_gro_receive functions perform flush checks (ttl, flags, iph->id, ...) against all packets in a loop. These flush checks are relevant only to tcp flows, and as such they're used to determine whether the packets can be merged later in tcp_gro_receive. These checks are not relevant to UDP packets. Furthermore, they need to be done only once in tcp_gro_receive and only against the found p skb, since they only affect flush and not same_flow. Levaraging the previous commit in the series, in which correct network header offsets are saved for both outer and inner network headers - allowing these checks to be done only once, in tcp_gro_receive. As a result, NAPI_GRO_CB(p)->flush is not used at all. In addition - flush_id checks are more declerative and contained in inet_gro_flush, thus removing the need for flush_id in napi_gro_cb. This results in less parsing code for UDP flows and non-loop flush tests for TCP flows. For example, running 40 IP/UDP netperf connections: ./super_netperf.sh 40 -H 1.1.1.2 -t UDP_STREAM -l 120 Running perf top for 90s we can see that relatively less time is spent on inet_gro_receive when GRO is not coalescing UDP: net-next: 1.26% [kernel] [k] inet_gro_receive patch applied: 0.85% [kernel] [k] inet_gro_receive udpgro_bench.sh single connection GRO improvement: net-next: 0.76% [kernel] [k] inet_gro_receive patch applied: 0.61% [kernel] [k] inet_gro_receive Signed-off-by: Richard Gobert <richardbgobert@gmail.com> --- include/net/gro.h | 9 ++---- net/core/gro.c | 3 -- net/ipv4/af_inet.c | 36 --------------------- net/ipv4/tcp_offload.c | 72 ++++++++++++++++++++++++++++++++++-------- net/ipv6/ip6_offload.c | 11 ------- 5 files changed, 62 insertions(+), 69 deletions(-)