From 37fdffb217a45609edccbb8b407d031143f551c0 Mon Sep 17 00:00:00 2001 From: Tariq Toukan Date: Tue, 21 Aug 2018 14:41:41 +0300 Subject: net/mlx5: WQ, fixes for fragmented WQ buffers API mlx5e netdevice used to calculate fragment edges by a call to mlx5_wq_cyc_get_frag_size(). This calculation did not give the correct indication for queues smaller than a PAGE_SIZE, (broken by default on PowerPC, where PAGE_SIZE == 64KB). Here it is replaced by the correct new calls/API. Since (TX/RX) Work Queues buffers are fragmented, here we introduce changes to the API in core driver, so that it gets a stride index and returns the index of last stride on same fragment, and an additional wrapping function that returns the number of physically contiguous strides that can be written contiguously to the work queue. This obsoletes the following API functions, and their buggy usage in EN driver: * mlx5_wq_cyc_get_frag_size() * mlx5_wq_cyc_ctr2fragix() The new API improves modularity and hides the details of such calculation for mlx5e netdevice and mlx5_ib rdma drivers. New calculation is also more efficient, and improves performance as follows: Packet rate test: pktgen, UDP / IPv4, 64byte, single ring, 8K ring size. Before: 16,477,619 pps After: 17,085,793 pps 3.7% improvement Fixes: 3a2f70331226 ("net/mlx5: Use order-0 allocations for all WQ types") Signed-off-by: Tariq Toukan Reviewed-by: Eran Ben Elisha Signed-off-by: Saeed Mahameed --- include/linux/mlx5/driver.h | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'include') diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index 66d94b4557cf..88a041b73abf 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -1032,6 +1032,14 @@ static inline void *mlx5_frag_buf_get_wqe(struct mlx5_frag_buf_ctrl *fbc, ((fbc->frag_sz_m1 & ix) << fbc->log_stride); } +static inline u32 +mlx5_frag_buf_get_idx_last_contig_stride(struct mlx5_frag_buf_ctrl *fbc, u32 ix) +{ + u32 last_frag_stride_idx = (ix + fbc->strides_offset) | fbc->frag_sz_m1; + + return min_t(u32, last_frag_stride_idx - fbc->strides_offset, fbc->sz_m1); +} + int mlx5_cmd_init(struct mlx5_core_dev *dev); void mlx5_cmd_cleanup(struct mlx5_core_dev *dev); void mlx5_cmd_use_events(struct mlx5_core_dev *dev); -- cgit v1.2.3 From f547fac624be53ad8b07e9ebca7654a7827ba61b Mon Sep 17 00:00:00 2001 From: Sabrina Dubroca Date: Fri, 12 Oct 2018 16:22:47 +0200 Subject: ipv6: rate-limit probes for neighbourless routes When commit 270972554c91 ("[IPV6]: ROUTE: Add Router Reachability Probing (RFC4191).") introduced router probing, the rt6_probe() function required that a neighbour entry existed. This neighbour entry is used to record the timestamp of the last probe via the ->updated field. Later, commit 2152caea7196 ("ipv6: Do not depend on rt->n in rt6_probe().") removed the requirement for a neighbour entry. Neighbourless routes skip the interval check and are not rate-limited. This patch adds rate-limiting for neighbourless routes, by recording the timestamp of the last probe in the fib6_info itself. Fixes: 2152caea7196 ("ipv6: Do not depend on rt->n in rt6_probe().") Signed-off-by: Sabrina Dubroca Reviewed-by: Stefano Brivio Signed-off-by: David S. Miller --- include/net/ip6_fib.h | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'include') diff --git a/include/net/ip6_fib.h b/include/net/ip6_fib.h index 3d4930528db0..2d31e22babd8 100644 --- a/include/net/ip6_fib.h +++ b/include/net/ip6_fib.h @@ -159,6 +159,10 @@ struct fib6_info { struct rt6_info * __percpu *rt6i_pcpu; struct rt6_exception_bucket __rcu *rt6i_exception_bucket; +#ifdef CONFIG_IPV6_ROUTER_PREF + unsigned long last_probe; +#endif + u32 fib6_metric; u8 fib6_protocol; u8 fib6_type; -- cgit v1.2.3 From d805397c3822d57ca3884d4bea37b2291fc40992 Mon Sep 17 00:00:00 2001 From: Xin Long Date: Mon, 15 Oct 2018 19:58:29 +0800 Subject: sctp: use the pmtu from the icmp packet to update transport pathmtu Other than asoc pmtu sync from all transports, sctp_assoc_sync_pmtu is also processing transport pmtu_pending by icmp packets. But it's meaningless to use sctp_dst_mtu(t->dst) as new pmtu for a transport. The right pmtu value should come from the icmp packet, and it would be saved into transport->mtu_info in this patch and used later when the pmtu sync happens in sctp_sendmsg_to_asoc or sctp_packet_config. Besides, without this patch, as pmtu can only be updated correctly when receiving a icmp packet and no place is holding sock lock, it will take long time if the sock is busy with sending packets. Note that it doesn't process transport->mtu_info in .release_cb(), as there is no enough information for pmtu update, like for which asoc or transport. It is not worth traversing all asocs to check pmtu_pending. So unlike tcp, sctp does this in tx path, for which mtu_info needs to be atomic_t. Signed-off-by: Xin Long Acked-by: Marcelo Ricardo Leitner Signed-off-by: David S. Miller --- include/net/sctp/structs.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'include') diff --git a/include/net/sctp/structs.h b/include/net/sctp/structs.h index 28a7c8e44636..a11f93790476 100644 --- a/include/net/sctp/structs.h +++ b/include/net/sctp/structs.h @@ -876,6 +876,8 @@ struct sctp_transport { unsigned long sackdelay; __u32 sackfreq; + atomic_t mtu_info; + /* When was the last time that we heard from this transport? We use * this to pick new active and retran paths. */ -- cgit v1.2.3 From 0ac1077e3a549bf8d35971613e2be05bdbb41a00 Mon Sep 17 00:00:00 2001 From: Xin Long Date: Tue, 16 Oct 2018 15:52:02 +0800 Subject: sctp: get pr_assoc and pr_stream all status with SCTP_PR_SCTP_ALL instead According to rfc7496 section 4.3 or 4.4: sprstat_policy: This parameter indicates for which PR-SCTP policy the user wants the information. It is an error to use SCTP_PR_SCTP_NONE in sprstat_policy. If SCTP_PR_SCTP_ALL is used, the counters provided are aggregated over all supported policies. We change to dump pr_assoc and pr_stream all status by SCTP_PR_SCTP_ALL instead, and return error for SCTP_PR_SCTP_NONE, as it also said "It is an error to use SCTP_PR_SCTP_NONE in sprstat_policy. " Fixes: 826d253d57b1 ("sctp: add SCTP_PR_ASSOC_STATUS on sctp sockopt") Fixes: d229d48d183f ("sctp: add SCTP_PR_STREAM_STATUS sockopt for prsctp") Reported-by: Ying Xu Signed-off-by: Xin Long Acked-by: Neil Horman Signed-off-by: David S. Miller --- include/uapi/linux/sctp.h | 1 + 1 file changed, 1 insertion(+) (limited to 'include') diff --git a/include/uapi/linux/sctp.h b/include/uapi/linux/sctp.h index b479db5c71d9..34dd3d497f2c 100644 --- a/include/uapi/linux/sctp.h +++ b/include/uapi/linux/sctp.h @@ -301,6 +301,7 @@ enum sctp_sinfo_flags { SCTP_SACK_IMMEDIATELY = (1 << 3), /* SACK should be sent without delay. */ /* 2 bits here have been used by SCTP_PR_SCTP_MASK */ SCTP_SENDALL = (1 << 6), + SCTP_PR_SCTP_ALL = (1 << 7), SCTP_NOTIFICATION = MSG_NOTIFICATION, /* Next message is not user msg but notification. */ SCTP_EOF = MSG_FIN, /* Initiate graceful shutdown process. */ }; -- cgit v1.2.3 From 6b4f92af3d59e882d3ba04c44a815266890d188f Mon Sep 17 00:00:00 2001 From: Stefano Brivio Date: Fri, 12 Oct 2018 23:53:59 +0200 Subject: geneve, vxlan: Don't set exceptions if skb->len < mtu We shouldn't abuse exceptions: if the destination MTU is already higher than what we're transmitting, no exception should be created. Fixes: 52a589d51f10 ("geneve: update skb dst pmtu on tx path") Fixes: a93bf0ff4490 ("vxlan: update skb dst pmtu on tx path") Signed-off-by: Stefano Brivio Reviewed-by: Sabrina Dubroca Signed-off-by: David S. Miller --- include/net/dst.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'include') diff --git a/include/net/dst.h b/include/net/dst.h index 7f735e76ca73..6cf0870414c7 100644 --- a/include/net/dst.h +++ b/include/net/dst.h @@ -527,4 +527,14 @@ static inline void skb_dst_update_pmtu(struct sk_buff *skb, u32 mtu) dst->ops->update_pmtu(dst, NULL, skb, mtu); } +static inline void skb_tunnel_check_pmtu(struct sk_buff *skb, + struct dst_entry *encap_dst, + int headroom) +{ + u32 encap_mtu = dst_mtu(encap_dst); + + if (skb->len > encap_mtu - headroom) + skb_dst_update_pmtu(skb, encap_mtu - headroom); +} + #endif /* _NET_DST_H */ -- cgit v1.2.3 From 5660b9d9d6a29c2c3cc12f62ae44bfb56b0a15a9 Mon Sep 17 00:00:00 2001 From: Xin Long Date: Wed, 17 Oct 2018 21:11:27 +0800 Subject: sctp: fix the data size calculation in sctp_data_size sctp data size should be calculated by subtracting data chunk header's length from chunk_hdr->length, not just data header. Fixes: 668c9beb9020 ("sctp: implement assign_number for sctp_stream_interleave") Signed-off-by: Xin Long Acked-by: Marcelo Ricardo Leitner Signed-off-by: David S. Miller --- include/net/sctp/sm.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include') diff --git a/include/net/sctp/sm.h b/include/net/sctp/sm.h index 5ef1bad81ef5..9e3d32746430 100644 --- a/include/net/sctp/sm.h +++ b/include/net/sctp/sm.h @@ -347,7 +347,7 @@ static inline __u16 sctp_data_size(struct sctp_chunk *chunk) __u16 size; size = ntohs(chunk->chunk_hdr->length); - size -= sctp_datahdr_len(&chunk->asoc->stream); + size -= sctp_datachk_len(&chunk->asoc->stream); return size; } -- cgit v1.2.3