From b2d5b4c468568b60916bdd5e7c10ab11b09bd164 Mon Sep 17 00:00:00 2001 From: Edward Cree Date: Wed, 13 Nov 2024 12:13:11 +0000 Subject: [PATCH 01/16] selftest: include dst-ip in ethtool ntuple rules sfc hardware does not support filters with only ipproto + dst-port; adding dst-ip to the flow spec allows the rss_ctx test to be run on these devices. Signed-off-by: Edward Cree Reviewed-by: Martin Habets Link: https://patch.msgid.link/8e5d23c8f21310c23c080cc7bcd31b76f8fd3096.1731499022.git.ecree.xilinx@gmail.com Signed-off-by: Jakub Kicinski --- tools/testing/selftests/drivers/net/hw/rss_ctx.py | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/tools/testing/selftests/drivers/net/hw/rss_ctx.py b/tools/testing/selftests/drivers/net/hw/rss_ctx.py index 29995586993c..fb61dae20fd8 100755 --- a/tools/testing/selftests/drivers/net/hw/rss_ctx.py +++ b/tools/testing/selftests/drivers/net/hw/rss_ctx.py @@ -215,7 +215,7 @@ def test_rss_queue_reconfigure(cfg, main_ctx=True): defer(ethtool, f"-X {cfg.ifname} default") else: other_key = 'noise' - flow = f"flow-type tcp{cfg.addr_ipver} dst-port {port} context {ctx_id}" + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {port} context {ctx_id}" ntuple = ethtool_create(cfg, "-N", flow) defer(ethtool, f"-N {cfg.ifname} delete {ntuple}") @@ -429,7 +429,7 @@ def test_rss_context(cfg, ctx_cnt=1, create_with_cfg=None): ksft_eq(max(data['rss-indirection-table']), 2 + i * 2 + 1, "Unexpected context cfg: " + str(data)) ports.append(rand_port()) - flow = f"flow-type tcp{cfg.addr_ipver} dst-port {ports[i]} context {ctx_id}" + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {ports[i]} context {ctx_id}" ntuple = ethtool_create(cfg, "-N", flow) defer(ethtool, f"-N {cfg.ifname} delete {ntuple}") @@ -516,7 +516,7 @@ def test_rss_context_out_of_order(cfg, ctx_cnt=4): ctx.append(defer(ethtool, f"-X {cfg.ifname} context {ctx_id} delete")) ports.append(rand_port()) - flow = f"flow-type tcp{cfg.addr_ipver} dst-port {ports[i]} context {ctx_id}" + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {ports[i]} context {ctx_id}" ntuple_id = ethtool_create(cfg, "-N", flow) ntuple.append(defer(ethtool, f"-N {cfg.ifname} delete {ntuple_id}")) @@ -569,7 +569,7 @@ def test_rss_context_overlap(cfg, other_ctx=0): port = rand_port() if other_ctx: - flow = f"flow-type tcp{cfg.addr_ipver} dst-port {port} context {other_ctx}" + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {port} context {other_ctx}" ntuple_id = ethtool_create(cfg, "-N", flow) ntuple = defer(ethtool, f"-N {cfg.ifname} delete {ntuple_id}") @@ -587,7 +587,7 @@ def test_rss_context_overlap(cfg, other_ctx=0): # Now create a rule for context 1 and make sure traffic goes to a subset if other_ctx: ntuple.exec() - flow = f"flow-type tcp{cfg.addr_ipver} dst-port {port} context {ctx_id}" + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {port} context {ctx_id}" ntuple_id = ethtool_create(cfg, "-N", flow) defer(ethtool, f"-N {cfg.ifname} delete {ntuple_id}") @@ -620,7 +620,7 @@ def test_delete_rss_context_busy(cfg): # utilize context from ntuple filter port = rand_port() - flow = f"flow-type tcp{cfg.addr_ipver} dst-port {port} context {ctx_id}" + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {port} context {ctx_id}" ntuple_id = ethtool_create(cfg, "-N", flow) defer(ethtool, f"-N {cfg.ifname} delete {ntuple_id}") -- 2.51.0 From e9e8abfec214b6a47f6c21d533c43c7f3c1f8887 Mon Sep 17 00:00:00 2001 From: Edward Cree Date: Wed, 13 Nov 2024 12:13:12 +0000 Subject: [PATCH 02/16] selftest: validate RSS+ntuple filters with nonzero ring_cookie Test creates an ntuple filter with 'action 2' and an RSS context whose indirection table has entries 0 and 1. Resulting traffic should go to queues 2 and 3; verify that it never hits queues 0 and 1. Signed-off-by: Edward Cree Link: https://patch.msgid.link/114afdf4d2867f72ed27751e8e08fe8b128a8529.1731499022.git.ecree.xilinx@gmail.com Signed-off-by: Jakub Kicinski --- .../selftests/drivers/net/hw/rss_ctx.py | 41 ++++++++++++++++++- 1 file changed, 40 insertions(+), 1 deletion(-) diff --git a/tools/testing/selftests/drivers/net/hw/rss_ctx.py b/tools/testing/selftests/drivers/net/hw/rss_ctx.py index fb61dae20fd8..8f62dc29bd26 100755 --- a/tools/testing/selftests/drivers/net/hw/rss_ctx.py +++ b/tools/testing/selftests/drivers/net/hw/rss_ctx.py @@ -633,6 +633,45 @@ def test_delete_rss_context_busy(cfg): pass +def test_rss_ntuple_addition(cfg): + """ + Test that the queue offset (ring_cookie) of an ntuple rule is added + to the queue number read from the indirection table. + """ + + require_ntuple(cfg) + + queue_cnt = len(_get_rx_cnts(cfg)) + if queue_cnt < 4: + try: + ksft_pr(f"Increasing queue count {queue_cnt} -> 4") + ethtool(f"-L {cfg.ifname} combined 4") + defer(ethtool, f"-L {cfg.ifname} combined {queue_cnt}") + except: + raise KsftSkipEx("Not enough queues for the test") + + # Use queue 0 for normal traffic + ethtool(f"-X {cfg.ifname} equal 1") + defer(ethtool, f"-X {cfg.ifname} default") + + # create additional rss context + ctx_id = ethtool_create(cfg, "-X", "context new equal 2") + defer(ethtool, f"-X {cfg.ifname} context {ctx_id} delete") + + # utilize context from ntuple filter + port = rand_port() + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {port} context {ctx_id} action 2" + try: + ntuple_id = ethtool_create(cfg, "-N", flow) + except CmdExitFailure: + raise KsftSkipEx("Ntuple filter with RSS and nonzero action not supported") + defer(ethtool, f"-N {cfg.ifname} delete {ntuple_id}") + + _send_traffic_check(cfg, port, f"context {ctx_id}", { 'target': (2, 3), + 'empty' : (1,), + 'noise' : (0,) }) + + def main() -> None: with NetDrvEpEnv(__file__, nsim_test=False) as cfg: cfg.ethnl = EthtoolFamily() @@ -644,7 +683,7 @@ def main() -> None: test_rss_context_dump, test_rss_context_queue_reconfigure, test_rss_context_overlap, test_rss_context_overlap2, test_rss_context_out_of_order, test_rss_context4_create_with_cfg, - test_delete_rss_context_busy], + test_delete_rss_context_busy, test_rss_ntuple_addition], args=(cfg, )) ksft_exit() -- 2.51.0 From 29a4bc1fe961caea52a5b945be2b4267b02002d7 Mon Sep 17 00:00:00 2001 From: Edward Cree Date: Wed, 13 Nov 2024 12:13:13 +0000 Subject: [PATCH 03/16] selftest: extend test_rss_context_queue_reconfigure for action addition The combination of ntuple action (ring_cookie) and RSS context can cause an ntuple rule to target a higher queue than appears in any RSS indirection table or directly in the ntuple rule, since the two numbers are added together. Verify the logic that prevents reducing the queue count in this case. Signed-off-by: Edward Cree Link: https://patch.msgid.link/58276b800ab78c0a79c1918046ccae7fe45ba802.1731499022.git.ecree.xilinx@gmail.com Signed-off-by: Jakub Kicinski --- .../selftests/drivers/net/hw/rss_ctx.py | 26 +++++++++++++++++++ 1 file changed, 26 insertions(+) diff --git a/tools/testing/selftests/drivers/net/hw/rss_ctx.py b/tools/testing/selftests/drivers/net/hw/rss_ctx.py index 8f62dc29bd26..0b49ce7ae678 100755 --- a/tools/testing/selftests/drivers/net/hw/rss_ctx.py +++ b/tools/testing/selftests/drivers/net/hw/rss_ctx.py @@ -238,6 +238,32 @@ def test_rss_queue_reconfigure(cfg, main_ctx=True): else: raise Exception(f"Driver didn't prevent us from deactivating a used queue (context {ctx_id})") + if not main_ctx: + ethtool(f"-L {cfg.ifname} combined 4") + flow = f"flow-type tcp{cfg.addr_ipver} dst-ip {cfg.addr} dst-port {port} context {ctx_id} action 1" + try: + # this targets queue 4, which doesn't exist + ntuple2 = ethtool_create(cfg, "-N", flow) + except CmdExitFailure: + pass + else: + raise Exception(f"Driver didn't prevent us from targeting a nonexistent queue (context {ctx_id})") + # change the table to target queues 0 and 2 + ethtool(f"-X {cfg.ifname} {ctx_ref} weight 1 0 1 0") + # ntuple rule therefore targets queues 1 and 3 + ntuple2 = ethtool_create(cfg, "-N", flow) + # should replace existing filter + ksft_eq(ntuple, ntuple2) + _send_traffic_check(cfg, port, ctx_ref, { 'target': (1, 3), + 'noise' : (0, 2) }) + # Setting queue count to 3 should fail, queue 3 is used + try: + ethtool(f"-L {cfg.ifname} combined 3") + except CmdExitFailure: + pass + else: + raise Exception(f"Driver didn't prevent us from deactivating a used queue (context {ctx_id})") + def test_rss_resize(cfg): """Test resizing of the RSS table. -- 2.51.0 From 0608746f95b29402421c0e0e96005afba45178ec Mon Sep 17 00:00:00 2001 From: Guillaume Nault Date: Thu, 14 Nov 2024 17:03:21 +0100 Subject: [PATCH 04/16] netfilter: ipv4: Convert ip_route_me_harder() to dscp_t. Use ip4h_dscp()instead of reading iph->tos directly. ip4h_dscp() returns a dscp_t value which is temporarily converted back to __u8 with inet_dscp_to_dsfield(). When converting ->flowi4_tos to dscp_t in the future, we'll only have to remove that inet_dscp_to_dsfield() call. Signed-off-by: Guillaume Nault Signed-off-by: Pablo Neira Ayuso --- net/ipv4/netfilter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/ipv4/netfilter.c b/net/ipv4/netfilter.c index e0aab66cd925..08bc3f2c0078 100644 --- a/net/ipv4/netfilter.c +++ b/net/ipv4/netfilter.c @@ -44,7 +44,7 @@ int ip_route_me_harder(struct net *net, struct sock *sk, struct sk_buff *skb, un */ fl4.daddr = iph->daddr; fl4.saddr = saddr; - fl4.flowi4_tos = iph->tos & INET_DSCP_MASK; + fl4.flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)); fl4.flowi4_oif = sk ? sk->sk_bound_dev_if : 0; fl4.flowi4_l3mdev = l3mdev_master_ifindex(dev); fl4.flowi4_mark = skb->mark; -- 2.51.0 From 6f9615a6e686bc0acfb5a02050a50782a6a378b2 Mon Sep 17 00:00:00 2001 From: Guillaume Nault Date: Thu, 14 Nov 2024 17:03:31 +0100 Subject: [PATCH 05/16] netfilter: flow_offload: Convert nft_flow_route() to dscp_t. Use ip4h_dscp()instead of reading ip_hdr()->tos directly. ip4h_dscp() returns a dscp_t value which is temporarily converted back to __u8 with inet_dscp_to_dsfield(). When converting ->flowi4_tos to dscp_t in the future, we'll only have to remove that inet_dscp_to_dsfield() call. Also, remove the comment about the net/ip.h include file, since it's now required for the ip4h_dscp() helper too. Signed-off-by: Guillaume Nault Signed-off-by: Pablo Neira Ayuso --- net/netfilter/nft_flow_offload.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/net/netfilter/nft_flow_offload.c b/net/netfilter/nft_flow_offload.c index 65199c23c75c..3b474d235663 100644 --- a/net/netfilter/nft_flow_offload.c +++ b/net/netfilter/nft_flow_offload.c @@ -8,7 +8,7 @@ #include #include #include -#include /* for ipv4 options. */ +#include #include #include #include @@ -236,7 +236,7 @@ static int nft_flow_route(const struct nft_pktinfo *pkt, fl.u.ip4.saddr = ct->tuplehash[!dir].tuple.src.u3.ip; fl.u.ip4.flowi4_oif = nft_in(pkt)->ifindex; fl.u.ip4.flowi4_iif = this_dst->dev->ifindex; - fl.u.ip4.flowi4_tos = ip_hdr(pkt->skb)->tos & INET_DSCP_MASK; + fl.u.ip4.flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(ip_hdr(pkt->skb))); fl.u.ip4.flowi4_mark = pkt->skb->mark; fl.u.ip4.flowi4_flags = FLOWI_FLAG_ANYSRC; break; -- 2.51.0 From f694ce6de58930146037aa3f69a534e98b007ff3 Mon Sep 17 00:00:00 2001 From: Guillaume Nault Date: Thu, 14 Nov 2024 17:03:38 +0100 Subject: [PATCH 06/16] netfilter: rpfilter: Convert rpfilter_mt() to dscp_t. Use ip4h_dscp() instead of reading iph->tos directly. ip4h_dscp() returns a dscp_t value which is temporarily converted back to __u8 with inet_dscp_to_dsfield(). When converting ->flowi4_tos to dscp_t in the future, we'll only have to remove that inet_dscp_to_dsfield() call. Signed-off-by: Guillaume Nault Signed-off-by: Pablo Neira Ayuso --- net/ipv4/netfilter/ipt_rpfilter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/ipv4/netfilter/ipt_rpfilter.c b/net/ipv4/netfilter/ipt_rpfilter.c index 1ce7a1655b97..a27782d7653e 100644 --- a/net/ipv4/netfilter/ipt_rpfilter.c +++ b/net/ipv4/netfilter/ipt_rpfilter.c @@ -76,7 +76,7 @@ static bool rpfilter_mt(const struct sk_buff *skb, struct xt_action_param *par) flow.daddr = iph->saddr; flow.saddr = rpfilter_get_saddr(iph->daddr); flow.flowi4_mark = info->flags & XT_RPFILTER_VALID_MARK ? skb->mark : 0; - flow.flowi4_tos = iph->tos & INET_DSCP_MASK; + flow.flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)); flow.flowi4_scope = RT_SCOPE_UNIVERSE; flow.flowi4_l3mdev = l3mdev_master_ifindex_rcu(xt_in(par)); flow.flowi4_uid = sock_net_uid(xt_net(par), NULL); -- 2.51.0 From f12b67cc7d1b67fe9ecee537df5b55625889ca9f Mon Sep 17 00:00:00 2001 From: Guillaume Nault Date: Thu, 14 Nov 2024 17:03:45 +0100 Subject: [PATCH 07/16] netfilter: nft_fib: Convert nft_fib4_eval() to dscp_t. Use ip4h_dscp() instead of reading iph->tos directly. ip4h_dscp() returns a dscp_t value which is temporarily converted back to __u8 with inet_dscp_to_dsfield(). When converting ->flowi4_tos to dscp_t in the future, we'll only have to remove that inet_dscp_to_dsfield() call. Signed-off-by: Guillaume Nault Signed-off-by: Pablo Neira Ayuso --- net/ipv4/netfilter/nft_fib_ipv4.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/net/ipv4/netfilter/nft_fib_ipv4.c b/net/ipv4/netfilter/nft_fib_ipv4.c index 09fff5d424ef..625adbc42037 100644 --- a/net/ipv4/netfilter/nft_fib_ipv4.c +++ b/net/ipv4/netfilter/nft_fib_ipv4.c @@ -11,6 +11,7 @@ #include #include +#include #include #include @@ -107,7 +108,7 @@ void nft_fib4_eval(const struct nft_expr *expr, struct nft_regs *regs, if (priv->flags & NFTA_FIB_F_MARK) fl4.flowi4_mark = pkt->skb->mark; - fl4.flowi4_tos = iph->tos & INET_DSCP_MASK; + fl4.flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)); if (priv->flags & NFTA_FIB_F_DADDR) { fl4.daddr = iph->daddr; -- 2.51.0 From f0d839c13ed50175a6e8b3a5ccd591ed15307995 Mon Sep 17 00:00:00 2001 From: Guillaume Nault Date: Thu, 14 Nov 2024 17:03:52 +0100 Subject: [PATCH 08/16] netfilter: nf_dup4: Convert nf_dup_ipv4_route() to dscp_t. Use ip4h_dscp() instead of reading iph->tos directly. ip4h_dscp() returns a dscp_t value which is temporarily converted back to __u8 with inet_dscp_to_dsfield(). When converting ->flowi4_tos to dscp_t in the future, we'll only have to remove that inet_dscp_to_dsfield() call. Signed-off-by: Guillaume Nault Signed-off-by: Pablo Neira Ayuso --- net/ipv4/netfilter/nf_dup_ipv4.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/net/ipv4/netfilter/nf_dup_ipv4.c b/net/ipv4/netfilter/nf_dup_ipv4.c index ec94ee1051c7..25e1e8eb18dd 100644 --- a/net/ipv4/netfilter/nf_dup_ipv4.c +++ b/net/ipv4/netfilter/nf_dup_ipv4.c @@ -33,7 +33,7 @@ static bool nf_dup_ipv4_route(struct net *net, struct sk_buff *skb, fl4.flowi4_oif = oif; fl4.daddr = gw->s_addr; - fl4.flowi4_tos = iph->tos & INET_DSCP_MASK; + fl4.flowi4_tos = inet_dscp_to_dsfield(ip4h_dscp(iph)); fl4.flowi4_scope = RT_SCOPE_UNIVERSE; fl4.flowi4_flags = FLOWI_FLAG_KNOWN_NH; rt = ip_route_output_key(net, &fl4); -- 2.51.0 From a12143e6084c502fc3cfaa8b717bffc8c14cf806 Mon Sep 17 00:00:00 2001 From: Jeremy Sowden Date: Thu, 14 Nov 2024 22:07:51 +0100 Subject: [PATCH 09/16] netfilter: bitwise: rename some boolean operation functions In the next patch we add support for doing AND, OR and XOR operations directly in the kernel, so rename some functions and an enum constant related to mask-and-xor boolean operations. Signed-off-by: Jeremy Sowden Signed-off-by: Pablo Neira Ayuso --- include/uapi/linux/netfilter/nf_tables.h | 10 ++++--- net/netfilter/nft_bitwise.c | 34 ++++++++++++------------ 2 files changed, 24 insertions(+), 20 deletions(-) diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h index 9e9079321380..487542234ccd 100644 --- a/include/uapi/linux/netfilter/nf_tables.h +++ b/include/uapi/linux/netfilter/nf_tables.h @@ -564,16 +564,20 @@ enum nft_immediate_attributes { /** * enum nft_bitwise_ops - nf_tables bitwise operations * - * @NFT_BITWISE_BOOL: mask-and-xor operation used to implement NOT, AND, OR and - * XOR boolean operations + * @NFT_BITWISE_MASK_XOR: mask-and-xor operation used to implement NOT, AND, OR + * and XOR boolean operations * @NFT_BITWISE_LSHIFT: left-shift operation * @NFT_BITWISE_RSHIFT: right-shift operation */ enum nft_bitwise_ops { - NFT_BITWISE_BOOL, + NFT_BITWISE_MASK_XOR, NFT_BITWISE_LSHIFT, NFT_BITWISE_RSHIFT, }; +/* + * Old name for NFT_BITWISE_MASK_XOR. Retained for backwards-compatibility. + */ +#define NFT_BITWISE_BOOL NFT_BITWISE_MASK_XOR /** * enum nft_bitwise_attributes - nf_tables bitwise expression netlink attributes diff --git a/net/netfilter/nft_bitwise.c b/net/netfilter/nft_bitwise.c index 7de95674fd8c..7f6a4f800537 100644 --- a/net/netfilter/nft_bitwise.c +++ b/net/netfilter/nft_bitwise.c @@ -25,8 +25,8 @@ struct nft_bitwise { struct nft_data data; }; -static void nft_bitwise_eval_bool(u32 *dst, const u32 *src, - const struct nft_bitwise *priv) +static void nft_bitwise_eval_mask_xor(u32 *dst, const u32 *src, + const struct nft_bitwise *priv) { unsigned int i; @@ -68,8 +68,8 @@ void nft_bitwise_eval(const struct nft_expr *expr, u32 *dst = ®s->data[priv->dreg]; switch (priv->op) { - case NFT_BITWISE_BOOL: - nft_bitwise_eval_bool(dst, src, priv); + case NFT_BITWISE_MASK_XOR: + nft_bitwise_eval_mask_xor(dst, src, priv); break; case NFT_BITWISE_LSHIFT: nft_bitwise_eval_lshift(dst, src, priv); @@ -90,8 +90,8 @@ static const struct nla_policy nft_bitwise_policy[NFTA_BITWISE_MAX + 1] = { [NFTA_BITWISE_DATA] = { .type = NLA_NESTED }, }; -static int nft_bitwise_init_bool(struct nft_bitwise *priv, - const struct nlattr *const tb[]) +static int nft_bitwise_init_mask_xor(struct nft_bitwise *priv, + const struct nlattr *const tb[]) { struct nft_data_desc mask = { .type = NFT_DATA_VALUE, @@ -185,7 +185,7 @@ static int nft_bitwise_init(const struct nft_ctx *ctx, if (tb[NFTA_BITWISE_OP]) { priv->op = ntohl(nla_get_be32(tb[NFTA_BITWISE_OP])); switch (priv->op) { - case NFT_BITWISE_BOOL: + case NFT_BITWISE_MASK_XOR: case NFT_BITWISE_LSHIFT: case NFT_BITWISE_RSHIFT: break; @@ -193,12 +193,12 @@ static int nft_bitwise_init(const struct nft_ctx *ctx, return -EOPNOTSUPP; } } else { - priv->op = NFT_BITWISE_BOOL; + priv->op = NFT_BITWISE_MASK_XOR; } switch(priv->op) { - case NFT_BITWISE_BOOL: - err = nft_bitwise_init_bool(priv, tb); + case NFT_BITWISE_MASK_XOR: + err = nft_bitwise_init_mask_xor(priv, tb); break; case NFT_BITWISE_LSHIFT: case NFT_BITWISE_RSHIFT: @@ -209,8 +209,8 @@ static int nft_bitwise_init(const struct nft_ctx *ctx, return err; } -static int nft_bitwise_dump_bool(struct sk_buff *skb, - const struct nft_bitwise *priv) +static int nft_bitwise_dump_mask_xor(struct sk_buff *skb, + const struct nft_bitwise *priv) { if (nft_data_dump(skb, NFTA_BITWISE_MASK, &priv->mask, NFT_DATA_VALUE, priv->len) < 0) @@ -248,8 +248,8 @@ static int nft_bitwise_dump(struct sk_buff *skb, return -1; switch (priv->op) { - case NFT_BITWISE_BOOL: - err = nft_bitwise_dump_bool(skb, priv); + case NFT_BITWISE_MASK_XOR: + err = nft_bitwise_dump_mask_xor(skb, priv); break; case NFT_BITWISE_LSHIFT: case NFT_BITWISE_RSHIFT: @@ -269,7 +269,7 @@ static int nft_bitwise_offload(struct nft_offload_ctx *ctx, const struct nft_bitwise *priv = nft_expr_priv(expr); struct nft_offload_reg *reg = &ctx->regs[priv->dreg]; - if (priv->op != NFT_BITWISE_BOOL) + if (priv->op != NFT_BITWISE_MASK_XOR) return -EOPNOTSUPP; if (memcmp(&priv->xor, &zero, sizeof(priv->xor)) || @@ -406,7 +406,7 @@ nft_bitwise_fast_dump(struct sk_buff *skb, return -1; if (nla_put_be32(skb, NFTA_BITWISE_LEN, htonl(sizeof(u32)))) return -1; - if (nla_put_be32(skb, NFTA_BITWISE_OP, htonl(NFT_BITWISE_BOOL))) + if (nla_put_be32(skb, NFTA_BITWISE_OP, htonl(NFT_BITWISE_MASK_XOR))) return -1; data.data[0] = priv->mask; @@ -501,7 +501,7 @@ nft_bitwise_select_ops(const struct nft_ctx *ctx, return &nft_bitwise_ops; if (tb[NFTA_BITWISE_OP] && - ntohl(nla_get_be32(tb[NFTA_BITWISE_OP])) != NFT_BITWISE_BOOL) + ntohl(nla_get_be32(tb[NFTA_BITWISE_OP])) != NFT_BITWISE_MASK_XOR) return &nft_bitwise_ops; return &nft_bitwise_fast_ops; -- 2.51.0 From b0ccf4f53d968e794a4ea579d5135cc1aaf1a53f Mon Sep 17 00:00:00 2001 From: Jeremy Sowden Date: Thu, 14 Nov 2024 22:08:13 +0100 Subject: [PATCH 10/16] netfilter: bitwise: add support for doing AND, OR and XOR directly Hitherto, these operations have been converted in user space to mask-and-xor operations on one register and two immediate values, and it is the latter which have been evaluated by the kernel. We add support for evaluating these operations directly in kernel space on one register and either an immediate value or a second register. Pablo made a few changes to the original patch: - EINVAL if NFTA_BITWISE_SREG2 is used with fast version. - Allow _AND,_OR,_XOR with _DATA != sizeof(u32) - Dump _SREG2 or _DATA with _AND,_OR,_XOR Signed-off-by: Jeremy Sowden Signed-off-by: Pablo Neira Ayuso --- include/uapi/linux/netfilter/nf_tables.h | 8 ++ net/netfilter/nft_bitwise.c | 134 +++++++++++++++++++++-- 2 files changed, 131 insertions(+), 11 deletions(-) diff --git a/include/uapi/linux/netfilter/nf_tables.h b/include/uapi/linux/netfilter/nf_tables.h index 487542234ccd..49c944e78463 100644 --- a/include/uapi/linux/netfilter/nf_tables.h +++ b/include/uapi/linux/netfilter/nf_tables.h @@ -568,11 +568,17 @@ enum nft_immediate_attributes { * and XOR boolean operations * @NFT_BITWISE_LSHIFT: left-shift operation * @NFT_BITWISE_RSHIFT: right-shift operation + * @NFT_BITWISE_AND: and operation + * @NFT_BITWISE_OR: or operation + * @NFT_BITWISE_XOR: xor operation */ enum nft_bitwise_ops { NFT_BITWISE_MASK_XOR, NFT_BITWISE_LSHIFT, NFT_BITWISE_RSHIFT, + NFT_BITWISE_AND, + NFT_BITWISE_OR, + NFT_BITWISE_XOR, }; /* * Old name for NFT_BITWISE_MASK_XOR. Retained for backwards-compatibility. @@ -590,6 +596,7 @@ enum nft_bitwise_ops { * @NFTA_BITWISE_OP: type of operation (NLA_U32: nft_bitwise_ops) * @NFTA_BITWISE_DATA: argument for non-boolean operations * (NLA_NESTED: nft_data_attributes) + * @NFTA_BITWISE_SREG2: second source register (NLA_U32: nft_registers) * * The bitwise expression supports boolean and shift operations. It implements * the boolean operations by performing the following operation: @@ -613,6 +620,7 @@ enum nft_bitwise_attributes { NFTA_BITWISE_XOR, NFTA_BITWISE_OP, NFTA_BITWISE_DATA, + NFTA_BITWISE_SREG2, __NFTA_BITWISE_MAX }; #define NFTA_BITWISE_MAX (__NFTA_BITWISE_MAX - 1) diff --git a/net/netfilter/nft_bitwise.c b/net/netfilter/nft_bitwise.c index 7f6a4f800537..d550910aabec 100644 --- a/net/netfilter/nft_bitwise.c +++ b/net/netfilter/nft_bitwise.c @@ -17,6 +17,7 @@ struct nft_bitwise { u8 sreg; + u8 sreg2; u8 dreg; enum nft_bitwise_ops op:8; u8 len; @@ -60,28 +61,72 @@ static void nft_bitwise_eval_rshift(u32 *dst, const u32 *src, } } +static void nft_bitwise_eval_and(u32 *dst, const u32 *src, const u32 *src2, + const struct nft_bitwise *priv) +{ + unsigned int i, n; + + for (i = 0, n = DIV_ROUND_UP(priv->len, sizeof(u32)); i < n; i++) + dst[i] = src[i] & src2[i]; +} + +static void nft_bitwise_eval_or(u32 *dst, const u32 *src, const u32 *src2, + const struct nft_bitwise *priv) +{ + unsigned int i, n; + + for (i = 0, n = DIV_ROUND_UP(priv->len, sizeof(u32)); i < n; i++) + dst[i] = src[i] | src2[i]; +} + +static void nft_bitwise_eval_xor(u32 *dst, const u32 *src, const u32 *src2, + const struct nft_bitwise *priv) +{ + unsigned int i, n; + + for (i = 0, n = DIV_ROUND_UP(priv->len, sizeof(u32)); i < n; i++) + dst[i] = src[i] ^ src2[i]; +} + void nft_bitwise_eval(const struct nft_expr *expr, struct nft_regs *regs, const struct nft_pktinfo *pkt) { const struct nft_bitwise *priv = nft_expr_priv(expr); - const u32 *src = ®s->data[priv->sreg]; + const u32 *src = ®s->data[priv->sreg], *src2; u32 *dst = ®s->data[priv->dreg]; - switch (priv->op) { - case NFT_BITWISE_MASK_XOR: + if (priv->op == NFT_BITWISE_MASK_XOR) { nft_bitwise_eval_mask_xor(dst, src, priv); - break; - case NFT_BITWISE_LSHIFT: + return; + } + if (priv->op == NFT_BITWISE_LSHIFT) { nft_bitwise_eval_lshift(dst, src, priv); - break; - case NFT_BITWISE_RSHIFT: + return; + } + if (priv->op == NFT_BITWISE_RSHIFT) { nft_bitwise_eval_rshift(dst, src, priv); - break; + return; + } + + src2 = priv->sreg2 ? ®s->data[priv->sreg2] : priv->data.data; + + if (priv->op == NFT_BITWISE_AND) { + nft_bitwise_eval_and(dst, src, src2, priv); + return; + } + if (priv->op == NFT_BITWISE_OR) { + nft_bitwise_eval_or(dst, src, src2, priv); + return; + } + if (priv->op == NFT_BITWISE_XOR) { + nft_bitwise_eval_xor(dst, src, src2, priv); + return; } } static const struct nla_policy nft_bitwise_policy[NFTA_BITWISE_MAX + 1] = { [NFTA_BITWISE_SREG] = { .type = NLA_U32 }, + [NFTA_BITWISE_SREG2] = { .type = NLA_U32 }, [NFTA_BITWISE_DREG] = { .type = NLA_U32 }, [NFTA_BITWISE_LEN] = { .type = NLA_U32 }, [NFTA_BITWISE_MASK] = { .type = NLA_NESTED }, @@ -105,7 +150,8 @@ static int nft_bitwise_init_mask_xor(struct nft_bitwise *priv, }; int err; - if (tb[NFTA_BITWISE_DATA]) + if (tb[NFTA_BITWISE_DATA] || + tb[NFTA_BITWISE_SREG2]) return -EINVAL; if (!tb[NFTA_BITWISE_MASK] || @@ -139,7 +185,8 @@ static int nft_bitwise_init_shift(struct nft_bitwise *priv, int err; if (tb[NFTA_BITWISE_MASK] || - tb[NFTA_BITWISE_XOR]) + tb[NFTA_BITWISE_XOR] || + tb[NFTA_BITWISE_SREG2]) return -EINVAL; if (!tb[NFTA_BITWISE_DATA]) @@ -157,6 +204,41 @@ static int nft_bitwise_init_shift(struct nft_bitwise *priv, return 0; } +static int nft_bitwise_init_bool(const struct nft_ctx *ctx, + struct nft_bitwise *priv, + const struct nlattr *const tb[]) +{ + int err; + + if (tb[NFTA_BITWISE_MASK] || + tb[NFTA_BITWISE_XOR]) + return -EINVAL; + + if ((!tb[NFTA_BITWISE_DATA] && !tb[NFTA_BITWISE_SREG2]) || + (tb[NFTA_BITWISE_DATA] && tb[NFTA_BITWISE_SREG2])) + return -EINVAL; + + if (tb[NFTA_BITWISE_DATA]) { + struct nft_data_desc desc = { + .type = NFT_DATA_VALUE, + .size = sizeof(priv->data), + .len = priv->len, + }; + + err = nft_data_init(NULL, &priv->data, &desc, + tb[NFTA_BITWISE_DATA]); + if (err < 0) + return err; + } else { + err = nft_parse_register_load(ctx, tb[NFTA_BITWISE_SREG2], + &priv->sreg2, priv->len); + if (err < 0) + return err; + } + + return 0; +} + static int nft_bitwise_init(const struct nft_ctx *ctx, const struct nft_expr *expr, const struct nlattr * const tb[]) @@ -188,6 +270,9 @@ static int nft_bitwise_init(const struct nft_ctx *ctx, case NFT_BITWISE_MASK_XOR: case NFT_BITWISE_LSHIFT: case NFT_BITWISE_RSHIFT: + case NFT_BITWISE_AND: + case NFT_BITWISE_OR: + case NFT_BITWISE_XOR: break; default: return -EOPNOTSUPP; @@ -204,6 +289,11 @@ static int nft_bitwise_init(const struct nft_ctx *ctx, case NFT_BITWISE_RSHIFT: err = nft_bitwise_init_shift(priv, tb); break; + case NFT_BITWISE_AND: + case NFT_BITWISE_OR: + case NFT_BITWISE_XOR: + err = nft_bitwise_init_bool(ctx, priv, tb); + break; } return err; @@ -232,6 +322,21 @@ static int nft_bitwise_dump_shift(struct sk_buff *skb, return 0; } +static int nft_bitwise_dump_bool(struct sk_buff *skb, + const struct nft_bitwise *priv) +{ + if (priv->sreg2) { + if (nft_dump_register(skb, NFTA_BITWISE_SREG2, priv->sreg2)) + return -1; + } else { + if (nft_data_dump(skb, NFTA_BITWISE_DATA, &priv->data, + NFT_DATA_VALUE, sizeof(u32)) < 0) + return -1; + } + + return 0; +} + static int nft_bitwise_dump(struct sk_buff *skb, const struct nft_expr *expr, bool reset) { @@ -255,6 +360,11 @@ static int nft_bitwise_dump(struct sk_buff *skb, case NFT_BITWISE_RSHIFT: err = nft_bitwise_dump_shift(skb, priv); break; + case NFT_BITWISE_AND: + case NFT_BITWISE_OR: + case NFT_BITWISE_XOR: + err = nft_bitwise_dump_bool(skb, priv); + break; } return err; @@ -299,6 +409,7 @@ static bool nft_bitwise_reduce(struct nft_regs_track *track, track->regs[priv->dreg].bitwise && track->regs[priv->dreg].bitwise->ops == expr->ops && priv->sreg == bitwise->sreg && + priv->sreg2 == bitwise->sreg2 && priv->dreg == bitwise->dreg && priv->op == bitwise->op && priv->len == bitwise->len && @@ -375,7 +486,8 @@ static int nft_bitwise_fast_init(const struct nft_ctx *ctx, if (err < 0) return err; - if (tb[NFTA_BITWISE_DATA]) + if (tb[NFTA_BITWISE_DATA] || + tb[NFTA_BITWISE_SREG2]) return -EINVAL; if (!tb[NFTA_BITWISE_MASK] || -- 2.51.0 From ed7231f56cd7f795ff3a831e32946a96661bfee9 Mon Sep 17 00:00:00 2001 From: Joe Damato Date: Thu, 14 Nov 2024 17:55:59 +0000 Subject: [PATCH 11/16] netdev-genl: Hold rcu_read_lock in napi_set Hold rcu_read_lock during netdev_nl_napi_set_doit, which calls napi_by_id and requires rcu_read_lock to be held. Closes: https://lore.kernel.org/netdev/719083c2-e277-447b-b6ea-ca3acb293a03@redhat.com/ Fixes: 1287c1ae0fc2 ("netdev-genl: Support setting per-NAPI config values") Signed-off-by: Joe Damato Link: https://patch.msgid.link/20241114175600.18882-1-jdamato@fastly.com Signed-off-by: Jakub Kicinski --- net/core/netdev-genl.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/net/core/netdev-genl.c b/net/core/netdev-genl.c index 765ce7c9d73b..fa119ff68698 100644 --- a/net/core/netdev-genl.c +++ b/net/core/netdev-genl.c @@ -346,6 +346,7 @@ int netdev_nl_napi_set_doit(struct sk_buff *skb, struct genl_info *info) napi_id = nla_get_u32(info->attrs[NETDEV_A_NAPI_ID]); rtnl_lock(); + rcu_read_lock(); napi = napi_by_id(napi_id); if (napi) { @@ -355,6 +356,7 @@ int netdev_nl_napi_set_doit(struct sk_buff *skb, struct genl_info *info) err = -ENOENT; } + rcu_read_unlock(); rtnl_unlock(); return err; -- 2.51.0 From c7a21af711e846a844095ae474f0f7e0ea8c6967 Mon Sep 17 00:00:00 2001 From: Vadim Fedorenko Date: Thu, 14 Nov 2024 03:48:20 -0800 Subject: [PATCH 12/16] bnxt_en: optimize gettimex64 Current implementation of gettimex64() makes at least 3 PCIe reads to get current PHC time. It takes at least 2.2us to get this value back to userspace. At the same time there is cached value of upper bits of PHC available for packet timestamps already. This patch reuses cached value to speed up reading of PHC time. Signed-off-by: Vadim Fedorenko Reviewed-by: Michael Chan Reviewed-by: Jacob Keller Link: https://patch.msgid.link/20241114114820.1411660-1-vadfed@meta.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c | 32 +++++++++++++++---- drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.h | 11 +++++++ 2 files changed, 37 insertions(+), 6 deletions(-) diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c index 91e7e08fabb1..075ccd589845 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.c @@ -112,6 +112,28 @@ static int bnxt_refclk_read(struct bnxt *bp, struct ptp_system_timestamp *sts, return rc; } +static int bnxt_refclk_read_low(struct bnxt *bp, struct ptp_system_timestamp *sts, + u32 *low) +{ + struct bnxt_ptp_cfg *ptp = bp->ptp_cfg; + unsigned long flags; + + /* We have to serialize reg access and FW reset */ + read_seqlock_excl_irqsave(&ptp->ptp_lock, flags); + + if (test_bit(BNXT_STATE_IN_FW_RESET, &bp->state)) { + read_sequnlock_excl_irqrestore(&ptp->ptp_lock, flags); + return -EIO; + } + + ptp_read_system_prets(sts); + *low = readl(bp->bar0 + ptp->refclk_mapped_regs[0]); + ptp_read_system_postts(sts); + + read_sequnlock_excl_irqrestore(&ptp->ptp_lock, flags); + return 0; +} + static void bnxt_ptp_get_current_time(struct bnxt *bp) { struct bnxt_ptp_cfg *ptp = bp->ptp_cfg; @@ -163,12 +185,14 @@ static int bnxt_ptp_gettimex(struct ptp_clock_info *ptp_info, struct bnxt_ptp_cfg *ptp = container_of(ptp_info, struct bnxt_ptp_cfg, ptp_info); u64 ns, cycles; + u32 low; int rc; - rc = bnxt_refclk_read(ptp->bp, sts, &cycles); + rc = bnxt_refclk_read_low(ptp->bp, sts, &low); if (rc) return rc; + cycles = bnxt_extend_cycles_32b_to_48b(ptp, low); ns = bnxt_timecounter_cyc2time(ptp, cycles); *ts = ns_to_timespec64(ns); @@ -801,15 +825,11 @@ void bnxt_get_tx_ts_p5(struct bnxt *bp, struct sk_buff *skb, u16 prod) int bnxt_get_rx_ts_p5(struct bnxt *bp, u64 *ts, u32 pkt_ts) { struct bnxt_ptp_cfg *ptp = bp->ptp_cfg; - u64 time; if (!ptp) return -ENODEV; - time = (u64)READ_ONCE(ptp->old_time) << BNXT_HI_TIMER_SHIFT; - *ts = (time & BNXT_HI_TIMER_MASK) | pkt_ts; - if (pkt_ts < (time & BNXT_LO_TIMER_MASK)) - *ts += BNXT_LO_TIMER_MASK + 1; + *ts = bnxt_extend_cycles_32b_to_48b(ptp, pkt_ts); return 0; } diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.h b/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.h index 4df4c2f373e0..c7851f8c971c 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.h +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ptp.h @@ -182,4 +182,15 @@ static inline u64 bnxt_timecounter_cyc2time(struct bnxt_ptp_cfg *ptp, u64 ts) return ns; } + +static inline u64 bnxt_extend_cycles_32b_to_48b(struct bnxt_ptp_cfg *ptp, u32 ts) +{ + u64 time, cycles; + + time = (u64)READ_ONCE(ptp->old_time) << BNXT_HI_TIMER_SHIFT; + cycles = (time & BNXT_HI_TIMER_MASK) | ts; + if (ts < (time & BNXT_LO_TIMER_MASK)) + cycles += BNXT_LO_TIMER_MASK + 1; + return cycles; +} #endif -- 2.51.0 From 6bbdb903db082fbb43bc1c2bbc7d2f6921cf031b Mon Sep 17 00:00:00 2001 From: "Rob Herring (Arm)" Date: Wed, 13 Nov 2024 16:56:43 -0600 Subject: [PATCH 13/16] dt-bindings: net: dsa: microchip,ksz: Drop undocumented "id" "id" is not a documented property, so drop it. Signed-off-by: Rob Herring (Arm) Acked-by: Conor Dooley Link: https://patch.msgid.link/20241113225642.1783485-2-robh@kernel.org Signed-off-by: Jakub Kicinski --- Documentation/devicetree/bindings/net/dsa/microchip,ksz.yaml | 1 - 1 file changed, 1 deletion(-) diff --git a/Documentation/devicetree/bindings/net/dsa/microchip,ksz.yaml b/Documentation/devicetree/bindings/net/dsa/microchip,ksz.yaml index 4722847dc862..62ca63e8a26f 100644 --- a/Documentation/devicetree/bindings/net/dsa/microchip,ksz.yaml +++ b/Documentation/devicetree/bindings/net/dsa/microchip,ksz.yaml @@ -159,7 +159,6 @@ examples: pinctrl-0 = <&pinctrl_spi_ksz>; cs-gpios = <&pioC 25 0>; - id = <1>; ksz9477: switch@0 { compatible = "microchip,ksz9477"; -- 2.51.0 From 025b2bbc5ab1c9ccd50189cae028d1b2bf3d433e Mon Sep 17 00:00:00 2001 From: Tarun Alle Date: Thu, 14 Nov 2024 15:49:51 +0530 Subject: [PATCH 14/16] net: phy: microchip_t1: Clause-45 PHY loopback support for LAN887x Adds support for clause-45 PHY loopback for the Microchip LAN887x driver. Signed-off-by: Tarun Alle Reviewed-by: Andrew Lunn Link: https://patch.msgid.link/20241114101951.382996-1-Tarun.Alle@microchip.com Signed-off-by: Jakub Kicinski --- drivers/net/phy/microchip_t1.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/net/phy/microchip_t1.c b/drivers/net/phy/microchip_t1.c index 71d6050b2833..b17bf6708003 100644 --- a/drivers/net/phy/microchip_t1.c +++ b/drivers/net/phy/microchip_t1.c @@ -2113,6 +2113,7 @@ static struct phy_driver microchip_t1_phy_driver[] = { .handle_interrupt = lan887x_handle_interrupt, .get_sqi = lan887x_get_sqi, .get_sqi_max = lan87xx_get_sqi_max, + .set_loopback = genphy_c45_loopback, } }; -- 2.51.0 From b67609c9315397789a3c497e08779b00d8cf4621 Mon Sep 17 00:00:00 2001 From: Nelson Escobar Date: Wed, 13 Nov 2024 23:56:33 +0000 Subject: [PATCH 15/16] enic: Create enic_wq/rq structures to bundle per wq/rq data Bundling the wq/rq specific data into dedicated enic_wq/rq structures cleans up the enic structure and simplifies future changes related to wq/rq. Co-developed-by: John Daley Signed-off-by: John Daley Co-developed-by: Satish Kharat Signed-off-by: Satish Kharat Signed-off-by: Nelson Escobar Reviewed-by: Simon Horman Reviewed-by: Vadim Fedorenko Link: https://patch.msgid.link/20241113-remove_vic_resource_limits-v4-1-a34cf8570c67@cisco.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/cisco/enic/enic.h | 18 ++- .../net/ethernet/cisco/enic/enic_ethtool.c | 4 +- drivers/net/ethernet/cisco/enic/enic_main.c | 120 +++++++++--------- drivers/net/ethernet/cisco/enic/enic_res.c | 12 +- 4 files changed, 81 insertions(+), 73 deletions(-) diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h index 0cc3644ee855..07459eac2592 100644 --- a/drivers/net/ethernet/cisco/enic/enic.h +++ b/drivers/net/ethernet/cisco/enic/enic.h @@ -162,6 +162,17 @@ struct enic_rq_stats { u64 desc_skip; /* Rx pkt went into later buffer */ }; +struct enic_wq { + spinlock_t lock; /* spinlock for wq */ + struct vnic_wq vwq; + struct enic_wq_stats stats; +} ____cacheline_aligned; + +struct enic_rq { + struct vnic_rq vrq; + struct enic_rq_stats stats; +} ____cacheline_aligned; + /* Per-instance private data structure */ struct enic { struct net_device *netdev; @@ -194,16 +205,13 @@ struct enic { struct enic_port_profile *pp; /* work queue cache line section */ - ____cacheline_aligned struct vnic_wq wq[ENIC_WQ_MAX]; - spinlock_t wq_lock[ENIC_WQ_MAX]; - struct enic_wq_stats wq_stats[ENIC_WQ_MAX]; + ____cacheline_aligned struct enic_wq wq[ENIC_WQ_MAX]; unsigned int wq_count; u16 loop_enable; u16 loop_tag; /* receive queue cache line section */ - ____cacheline_aligned struct vnic_rq rq[ENIC_RQ_MAX]; - struct enic_rq_stats rq_stats[ENIC_RQ_MAX]; + ____cacheline_aligned struct enic_rq rq[ENIC_RQ_MAX]; unsigned int rq_count; struct vxlan_offload vxlan; struct napi_struct napi[ENIC_RQ_MAX + ENIC_WQ_MAX]; diff --git a/drivers/net/ethernet/cisco/enic/enic_ethtool.c b/drivers/net/ethernet/cisco/enic/enic_ethtool.c index 4fe85780a950..de07fa9374d2 100644 --- a/drivers/net/ethernet/cisco/enic/enic_ethtool.c +++ b/drivers/net/ethernet/cisco/enic/enic_ethtool.c @@ -337,7 +337,7 @@ static void enic_get_ethtool_stats(struct net_device *netdev, for (i = 0; i < NUM_ENIC_GEN_STATS; i++) *(data++) = ((u64 *)&enic->gen_stats)[enic_gen_stats[i].index]; for (i = 0; i < enic->rq_count; i++) { - struct enic_rq_stats *rqstats = &enic->rq_stats[i]; + struct enic_rq_stats *rqstats = &enic->rq[i].stats; int index; for (j = 0; j < NUM_ENIC_PER_RQ_STATS; j++) { @@ -346,7 +346,7 @@ static void enic_get_ethtool_stats(struct net_device *netdev, } } for (i = 0; i < enic->wq_count; i++) { - struct enic_wq_stats *wqstats = &enic->wq_stats[i]; + struct enic_wq_stats *wqstats = &enic->wq[i].stats; int index; for (j = 0; j < NUM_ENIC_PER_WQ_STATS; j++) { diff --git a/drivers/net/ethernet/cisco/enic/enic_main.c b/drivers/net/ethernet/cisco/enic/enic_main.c index ffed14b63d41..eb00058b6c68 100644 --- a/drivers/net/ethernet/cisco/enic/enic_main.c +++ b/drivers/net/ethernet/cisco/enic/enic_main.c @@ -342,8 +342,8 @@ static void enic_wq_free_buf(struct vnic_wq *wq, { struct enic *enic = vnic_dev_priv(wq->vdev); - enic->wq_stats[wq->index].cq_work++; - enic->wq_stats[wq->index].cq_bytes += buf->len; + enic->wq[wq->index].stats.cq_work++; + enic->wq[wq->index].stats.cq_bytes += buf->len; enic_free_wq_buf(wq, buf); } @@ -352,20 +352,20 @@ static int enic_wq_service(struct vnic_dev *vdev, struct cq_desc *cq_desc, { struct enic *enic = vnic_dev_priv(vdev); - spin_lock(&enic->wq_lock[q_number]); + spin_lock(&enic->wq[q_number].lock); - vnic_wq_service(&enic->wq[q_number], cq_desc, + vnic_wq_service(&enic->wq[q_number].vwq, cq_desc, completed_index, enic_wq_free_buf, opaque); if (netif_tx_queue_stopped(netdev_get_tx_queue(enic->netdev, q_number)) && - vnic_wq_desc_avail(&enic->wq[q_number]) >= + vnic_wq_desc_avail(&enic->wq[q_number].vwq) >= (MAX_SKB_FRAGS + ENIC_DESC_MAX_SPLITS)) { netif_wake_subqueue(enic->netdev, q_number); - enic->wq_stats[q_number].wake++; + enic->wq[q_number].stats.wake++; } - spin_unlock(&enic->wq_lock[q_number]); + spin_unlock(&enic->wq[q_number].lock); return 0; } @@ -377,7 +377,7 @@ static bool enic_log_q_error(struct enic *enic) bool err = false; for (i = 0; i < enic->wq_count; i++) { - error_status = vnic_wq_error_status(&enic->wq[i]); + error_status = vnic_wq_error_status(&enic->wq[i].vwq); err |= error_status; if (error_status) netdev_err(enic->netdev, "WQ[%d] error_status %d\n", @@ -385,7 +385,7 @@ static bool enic_log_q_error(struct enic *enic) } for (i = 0; i < enic->rq_count; i++) { - error_status = vnic_rq_error_status(&enic->rq[i]); + error_status = vnic_rq_error_status(&enic->rq[i].vrq); err |= error_status; if (error_status) netdev_err(enic->netdev, "RQ[%d] error_status %d\n", @@ -598,9 +598,9 @@ static int enic_queue_wq_skb_vlan(struct enic *enic, struct vnic_wq *wq, err = enic_queue_wq_skb_cont(enic, wq, skb, len_left, loopback); /* The enic_queue_wq_desc() above does not do HW checksum */ - enic->wq_stats[wq->index].csum_none++; - enic->wq_stats[wq->index].packets++; - enic->wq_stats[wq->index].bytes += skb->len; + enic->wq[wq->index].stats.csum_none++; + enic->wq[wq->index].stats.packets++; + enic->wq[wq->index].stats.bytes += skb->len; return err; } @@ -634,9 +634,9 @@ static int enic_queue_wq_skb_csum_l4(struct enic *enic, struct vnic_wq *wq, if (!eop) err = enic_queue_wq_skb_cont(enic, wq, skb, len_left, loopback); - enic->wq_stats[wq->index].csum_partial++; - enic->wq_stats[wq->index].packets++; - enic->wq_stats[wq->index].bytes += skb->len; + enic->wq[wq->index].stats.csum_partial++; + enic->wq[wq->index].stats.packets++; + enic->wq[wq->index].stats.bytes += skb->len; return err; } @@ -699,11 +699,11 @@ static int enic_queue_wq_skb_tso(struct enic *enic, struct vnic_wq *wq, if (skb->encapsulation) { hdr_len = skb_inner_tcp_all_headers(skb); enic_preload_tcp_csum_encap(skb); - enic->wq_stats[wq->index].encap_tso++; + enic->wq[wq->index].stats.encap_tso++; } else { hdr_len = skb_tcp_all_headers(skb); enic_preload_tcp_csum(skb); - enic->wq_stats[wq->index].tso++; + enic->wq[wq->index].stats.tso++; } /* Queue WQ_ENET_MAX_DESC_LEN length descriptors @@ -757,8 +757,8 @@ tso_out_stats: pkts = len / mss; if ((len % mss) > 0) pkts++; - enic->wq_stats[wq->index].packets += pkts; - enic->wq_stats[wq->index].bytes += (len + (pkts * hdr_len)); + enic->wq[wq->index].stats.packets += pkts; + enic->wq[wq->index].stats.bytes += (len + (pkts * hdr_len)); return 0; } @@ -792,9 +792,9 @@ static inline int enic_queue_wq_skb_encap(struct enic *enic, struct vnic_wq *wq, if (!eop) err = enic_queue_wq_skb_cont(enic, wq, skb, len_left, loopback); - enic->wq_stats[wq->index].encap_csum++; - enic->wq_stats[wq->index].packets++; - enic->wq_stats[wq->index].bytes += skb->len; + enic->wq[wq->index].stats.encap_csum++; + enic->wq[wq->index].stats.packets++; + enic->wq[wq->index].stats.bytes += skb->len; return err; } @@ -812,7 +812,7 @@ static inline int enic_queue_wq_skb(struct enic *enic, /* VLAN tag from trunking driver */ vlan_tag_insert = 1; vlan_tag = skb_vlan_tag_get(skb); - enic->wq_stats[wq->index].add_vlan++; + enic->wq[wq->index].stats.add_vlan++; } else if (enic->loop_enable) { vlan_tag = enic->loop_tag; loopback = 1; @@ -859,11 +859,11 @@ static netdev_tx_t enic_hard_start_xmit(struct sk_buff *skb, struct netdev_queue *txq; txq_map = skb_get_queue_mapping(skb) % enic->wq_count; - wq = &enic->wq[txq_map]; + wq = &enic->wq[txq_map].vwq; if (skb->len <= 0) { dev_kfree_skb_any(skb); - enic->wq_stats[wq->index].null_pkt++; + enic->wq[wq->index].stats.null_pkt++; return NETDEV_TX_OK; } @@ -878,19 +878,19 @@ static netdev_tx_t enic_hard_start_xmit(struct sk_buff *skb, skb_shinfo(skb)->nr_frags + 1 > ENIC_NON_TSO_MAX_DESC && skb_linearize(skb)) { dev_kfree_skb_any(skb); - enic->wq_stats[wq->index].skb_linear_fail++; + enic->wq[wq->index].stats.skb_linear_fail++; return NETDEV_TX_OK; } - spin_lock(&enic->wq_lock[txq_map]); + spin_lock(&enic->wq[txq_map].lock); if (vnic_wq_desc_avail(wq) < skb_shinfo(skb)->nr_frags + ENIC_DESC_MAX_SPLITS) { netif_tx_stop_queue(txq); /* This is a hard error, log it */ netdev_err(netdev, "BUG! Tx ring full when queue awake!\n"); - spin_unlock(&enic->wq_lock[txq_map]); - enic->wq_stats[wq->index].desc_full_awake++; + spin_unlock(&enic->wq[txq_map].lock); + enic->wq[wq->index].stats.desc_full_awake++; return NETDEV_TX_BUSY; } @@ -899,14 +899,14 @@ static netdev_tx_t enic_hard_start_xmit(struct sk_buff *skb, if (vnic_wq_desc_avail(wq) < MAX_SKB_FRAGS + ENIC_DESC_MAX_SPLITS) { netif_tx_stop_queue(txq); - enic->wq_stats[wq->index].stopped++; + enic->wq[wq->index].stats.stopped++; } skb_tx_timestamp(skb); if (!netdev_xmit_more() || netif_xmit_stopped(txq)) vnic_wq_doorbell(wq); error: - spin_unlock(&enic->wq_lock[txq_map]); + spin_unlock(&enic->wq[txq_map].lock); return NETDEV_TX_OK; } @@ -941,9 +941,9 @@ static void enic_get_stats(struct net_device *netdev, net_stats->multicast = stats->rx.rx_multicast_frames_ok; for (i = 0; i < ENIC_RQ_MAX; i++) { - struct enic_rq_stats *rqs = &enic->rq_stats[i]; + struct enic_rq_stats *rqs = &enic->rq[i].stats; - if (!enic->rq->ctrl) + if (!enic->rq[i].vrq.ctrl) break; pkt_truncated += rqs->pkt_truncated; bad_fcs += rqs->bad_fcs; @@ -1313,7 +1313,7 @@ static int enic_rq_alloc_buf(struct vnic_rq *rq) } skb = netdev_alloc_skb_ip_align(netdev, len); if (!skb) { - enic->rq_stats[rq->index].no_skb++; + enic->rq[rq->index].stats.no_skb++; return -ENOMEM; } @@ -1366,7 +1366,7 @@ static void enic_rq_indicate_buf(struct vnic_rq *rq, struct net_device *netdev = enic->netdev; struct sk_buff *skb; struct vnic_cq *cq = &enic->cq[enic_cq_rq(enic, rq->index)]; - struct enic_rq_stats *rqstats = &enic->rq_stats[rq->index]; + struct enic_rq_stats *rqstats = &enic->rq[rq->index].stats; u8 type, color, eop, sop, ingress_port, vlan_stripped; u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof; @@ -1512,7 +1512,7 @@ static int enic_rq_service(struct vnic_dev *vdev, struct cq_desc *cq_desc, { struct enic *enic = vnic_dev_priv(vdev); - vnic_rq_service(&enic->rq[q_number], cq_desc, + vnic_rq_service(&enic->rq[q_number].vrq, cq_desc, completed_index, VNIC_RQ_RETURN_DESC, enic_rq_indicate_buf, opaque); @@ -1609,7 +1609,7 @@ static int enic_poll(struct napi_struct *napi, int budget) 0 /* don't unmask intr */, 0 /* don't reset intr timer */); - err = vnic_rq_fill(&enic->rq[0], enic_rq_alloc_buf); + err = vnic_rq_fill(&enic->rq[0].vrq, enic_rq_alloc_buf); /* Buffer allocation failed. Stay in polling * mode so we can try to fill the ring again. @@ -1621,7 +1621,7 @@ static int enic_poll(struct napi_struct *napi, int budget) /* Call the function which refreshes the intr coalescing timer * value based on the traffic. */ - enic_calc_int_moderation(enic, &enic->rq[0]); + enic_calc_int_moderation(enic, &enic->rq[0].vrq); if ((rq_work_done < budget) && napi_complete_done(napi, rq_work_done)) { @@ -1630,11 +1630,11 @@ static int enic_poll(struct napi_struct *napi, int budget) */ if (enic->rx_coalesce_setting.use_adaptive_rx_coalesce) - enic_set_int_moderation(enic, &enic->rq[0]); + enic_set_int_moderation(enic, &enic->rq[0].vrq); vnic_intr_unmask(&enic->intr[intr]); - enic->rq_stats[0].napi_complete++; + enic->rq[0].stats.napi_complete++; } else { - enic->rq_stats[0].napi_repoll++; + enic->rq[0].stats.napi_repoll++; } return rq_work_done; @@ -1683,7 +1683,7 @@ static int enic_poll_msix_wq(struct napi_struct *napi, int budget) struct net_device *netdev = napi->dev; struct enic *enic = netdev_priv(netdev); unsigned int wq_index = (napi - &enic->napi[0]) - enic->rq_count; - struct vnic_wq *wq = &enic->wq[wq_index]; + struct vnic_wq *wq = &enic->wq[wq_index].vwq; unsigned int cq; unsigned int intr; unsigned int wq_work_to_do = ENIC_WQ_NAPI_BUDGET; @@ -1737,7 +1737,7 @@ static int enic_poll_msix_rq(struct napi_struct *napi, int budget) 0 /* don't unmask intr */, 0 /* don't reset intr timer */); - err = vnic_rq_fill(&enic->rq[rq], enic_rq_alloc_buf); + err = vnic_rq_fill(&enic->rq[rq].vrq, enic_rq_alloc_buf); /* Buffer allocation failed. Stay in polling mode * so we can try to fill the ring again. @@ -1749,7 +1749,7 @@ static int enic_poll_msix_rq(struct napi_struct *napi, int budget) /* Call the function which refreshes the intr coalescing timer * value based on the traffic. */ - enic_calc_int_moderation(enic, &enic->rq[rq]); + enic_calc_int_moderation(enic, &enic->rq[rq].vrq); if ((work_done < budget) && napi_complete_done(napi, work_done)) { @@ -1758,11 +1758,11 @@ static int enic_poll_msix_rq(struct napi_struct *napi, int budget) */ if (enic->rx_coalesce_setting.use_adaptive_rx_coalesce) - enic_set_int_moderation(enic, &enic->rq[rq]); + enic_set_int_moderation(enic, &enic->rq[rq].vrq); vnic_intr_unmask(&enic->intr[intr]); - enic->rq_stats[rq].napi_complete++; + enic->rq[rq].stats.napi_complete++; } else { - enic->rq_stats[rq].napi_repoll++; + enic->rq[rq].stats.napi_repoll++; } return work_done; @@ -1989,10 +1989,10 @@ static int enic_open(struct net_device *netdev) for (i = 0; i < enic->rq_count; i++) { /* enable rq before updating rq desc */ - vnic_rq_enable(&enic->rq[i]); - vnic_rq_fill(&enic->rq[i], enic_rq_alloc_buf); + vnic_rq_enable(&enic->rq[i].vrq); + vnic_rq_fill(&enic->rq[i].vrq, enic_rq_alloc_buf); /* Need at least one buffer on ring to get going */ - if (vnic_rq_desc_used(&enic->rq[i]) == 0) { + if (vnic_rq_desc_used(&enic->rq[i].vrq) == 0) { netdev_err(netdev, "Unable to alloc receive buffers\n"); err = -ENOMEM; goto err_out_free_rq; @@ -2000,7 +2000,7 @@ static int enic_open(struct net_device *netdev) } for (i = 0; i < enic->wq_count; i++) - vnic_wq_enable(&enic->wq[i]); + vnic_wq_enable(&enic->wq[i].vwq); if (!enic_is_dynamic(enic) && !enic_is_sriov_vf(enic)) enic_dev_add_station_addr(enic); @@ -2027,9 +2027,9 @@ static int enic_open(struct net_device *netdev) err_out_free_rq: for (i = 0; i < enic->rq_count; i++) { - ret = vnic_rq_disable(&enic->rq[i]); + ret = vnic_rq_disable(&enic->rq[i].vrq); if (!ret) - vnic_rq_clean(&enic->rq[i], enic_free_rq_buf); + vnic_rq_clean(&enic->rq[i].vrq, enic_free_rq_buf); } enic_dev_notify_unset(enic); err_out_free_intr: @@ -2071,12 +2071,12 @@ static int enic_stop(struct net_device *netdev) enic_dev_del_station_addr(enic); for (i = 0; i < enic->wq_count; i++) { - err = vnic_wq_disable(&enic->wq[i]); + err = vnic_wq_disable(&enic->wq[i].vwq); if (err) return err; } for (i = 0; i < enic->rq_count; i++) { - err = vnic_rq_disable(&enic->rq[i]); + err = vnic_rq_disable(&enic->rq[i].vrq); if (err) return err; } @@ -2086,9 +2086,9 @@ static int enic_stop(struct net_device *netdev) enic_free_intr(enic); for (i = 0; i < enic->wq_count; i++) - vnic_wq_clean(&enic->wq[i], enic_free_wq_buf); + vnic_wq_clean(&enic->wq[i].vwq, enic_free_wq_buf); for (i = 0; i < enic->rq_count; i++) - vnic_rq_clean(&enic->rq[i], enic_free_rq_buf); + vnic_rq_clean(&enic->rq[i].vrq, enic_free_rq_buf); for (i = 0; i < enic->cq_count; i++) vnic_cq_clean(&enic->cq[i]); for (i = 0; i < enic->intr_count; i++) @@ -2576,7 +2576,7 @@ static void enic_get_queue_stats_rx(struct net_device *dev, int idx, struct netdev_queue_stats_rx *rxs) { struct enic *enic = netdev_priv(dev); - struct enic_rq_stats *rqstats = &enic->rq_stats[idx]; + struct enic_rq_stats *rqstats = &enic->rq[idx].stats; rxs->bytes = rqstats->bytes; rxs->packets = rqstats->packets; @@ -2590,7 +2590,7 @@ static void enic_get_queue_stats_tx(struct net_device *dev, int idx, struct netdev_queue_stats_tx *txs) { struct enic *enic = netdev_priv(dev); - struct enic_wq_stats *wqstats = &enic->wq_stats[idx]; + struct enic_wq_stats *wqstats = &enic->wq[idx].stats; txs->bytes = wqstats->bytes; txs->packets = wqstats->packets; @@ -2993,7 +2993,7 @@ static int enic_probe(struct pci_dev *pdev, const struct pci_device_id *ent) INIT_WORK(&enic->change_mtu_work, enic_change_mtu_work); for (i = 0; i < enic->wq_count; i++) - spin_lock_init(&enic->wq_lock[i]); + spin_lock_init(&enic->wq[i].lock); /* Register net device */ diff --git a/drivers/net/ethernet/cisco/enic/enic_res.c b/drivers/net/ethernet/cisco/enic/enic_res.c index 1c48aebdbab0..60be09acb9fd 100644 --- a/drivers/net/ethernet/cisco/enic/enic_res.c +++ b/drivers/net/ethernet/cisco/enic/enic_res.c @@ -176,9 +176,9 @@ void enic_free_vnic_resources(struct enic *enic) unsigned int i; for (i = 0; i < enic->wq_count; i++) - vnic_wq_free(&enic->wq[i]); + vnic_wq_free(&enic->wq[i].vwq); for (i = 0; i < enic->rq_count; i++) - vnic_rq_free(&enic->rq[i]); + vnic_rq_free(&enic->rq[i].vrq); for (i = 0; i < enic->cq_count; i++) vnic_cq_free(&enic->cq[i]); for (i = 0; i < enic->intr_count; i++) @@ -233,7 +233,7 @@ void enic_init_vnic_resources(struct enic *enic) for (i = 0; i < enic->rq_count; i++) { cq_index = i; - vnic_rq_init(&enic->rq[i], + vnic_rq_init(&enic->rq[i].vrq, cq_index, error_interrupt_enable, error_interrupt_offset); @@ -241,7 +241,7 @@ void enic_init_vnic_resources(struct enic *enic) for (i = 0; i < enic->wq_count; i++) { cq_index = enic->rq_count + i; - vnic_wq_init(&enic->wq[i], + vnic_wq_init(&enic->wq[i].vwq, cq_index, error_interrupt_enable, error_interrupt_offset); @@ -322,7 +322,7 @@ int enic_alloc_vnic_resources(struct enic *enic) */ for (i = 0; i < enic->wq_count; i++) { - err = vnic_wq_alloc(enic->vdev, &enic->wq[i], i, + err = vnic_wq_alloc(enic->vdev, &enic->wq[i].vwq, i, enic->config.wq_desc_count, sizeof(struct wq_enet_desc)); if (err) @@ -330,7 +330,7 @@ int enic_alloc_vnic_resources(struct enic *enic) } for (i = 0; i < enic->rq_count; i++) { - err = vnic_rq_alloc(enic->vdev, &enic->rq[i], i, + err = vnic_rq_alloc(enic->vdev, &enic->rq[i].vrq, i, enic->config.rq_desc_count, sizeof(struct rq_enet_desc)); if (err) -- 2.51.0 From 231646cb6a8cc4e94943daf223d72aefe242ec23 Mon Sep 17 00:00:00 2001 From: Nelson Escobar Date: Wed, 13 Nov 2024 23:56:34 +0000 Subject: [PATCH 16/16] enic: Make MSI-X I/O interrupts come after the other required ones The VIC hardware has a constraint that the MSIX interrupt used for errors be specified as a 7 bit number. Before this patch, it was allocated after the I/O interrupts, which would cause a problem if 128 or more I/O interrupts are in use. So make the required interrupts come before the I/O interrupts to guarantee the error interrupt offset never exceeds 7 bits. Co-developed-by: John Daley Signed-off-by: John Daley Co-developed-by: Satish Kharat Signed-off-by: Satish Kharat Reviewed-by: Simon Horman Signed-off-by: Nelson Escobar Reviewed-by: Vadim Fedorenko Link: https://patch.msgid.link/20241113-remove_vic_resource_limits-v4-2-a34cf8570c67@cisco.com Signed-off-by: Jakub Kicinski --- drivers/net/ethernet/cisco/enic/enic.h | 20 +++++++++++++++----- drivers/net/ethernet/cisco/enic/enic_res.c | 11 +++++++---- 2 files changed, 22 insertions(+), 9 deletions(-) diff --git a/drivers/net/ethernet/cisco/enic/enic.h b/drivers/net/ethernet/cisco/enic/enic.h index 07459eac2592..ec83a273d1ca 100644 --- a/drivers/net/ethernet/cisco/enic/enic.h +++ b/drivers/net/ethernet/cisco/enic/enic.h @@ -280,18 +280,28 @@ static inline unsigned int enic_msix_wq_intr(struct enic *enic, return enic->cq[enic_cq_wq(enic, wq)].interrupt_offset; } -static inline unsigned int enic_msix_err_intr(struct enic *enic) -{ - return enic->rq_count + enic->wq_count; -} +/* MSIX interrupts are organized as the error interrupt, then the notify + * interrupt followed by all the I/O interrupts. The error interrupt needs + * to fit in 7 bits due to hardware constraints + */ +#define ENIC_MSIX_RESERVED_INTR 2 +#define ENIC_MSIX_ERR_INTR 0 +#define ENIC_MSIX_NOTIFY_INTR 1 +#define ENIC_MSIX_IO_INTR_BASE ENIC_MSIX_RESERVED_INTR +#define ENIC_MSIX_MIN_INTR (ENIC_MSIX_RESERVED_INTR + 2) #define ENIC_LEGACY_IO_INTR 0 #define ENIC_LEGACY_ERR_INTR 1 #define ENIC_LEGACY_NOTIFY_INTR 2 +static inline unsigned int enic_msix_err_intr(struct enic *enic) +{ + return ENIC_MSIX_ERR_INTR; +} + static inline unsigned int enic_msix_notify_intr(struct enic *enic) { - return enic->rq_count + enic->wq_count + 1; + return ENIC_MSIX_NOTIFY_INTR; } static inline bool enic_is_err_intr(struct enic *enic, int intr) diff --git a/drivers/net/ethernet/cisco/enic/enic_res.c b/drivers/net/ethernet/cisco/enic/enic_res.c index 60be09acb9fd..72b51e9d8d1a 100644 --- a/drivers/net/ethernet/cisco/enic/enic_res.c +++ b/drivers/net/ethernet/cisco/enic/enic_res.c @@ -221,9 +221,12 @@ void enic_init_vnic_resources(struct enic *enic) switch (intr_mode) { case VNIC_DEV_INTR_MODE_INTX: + error_interrupt_enable = 1; + error_interrupt_offset = ENIC_LEGACY_ERR_INTR; + break; case VNIC_DEV_INTR_MODE_MSIX: error_interrupt_enable = 1; - error_interrupt_offset = enic->intr_count - 2; + error_interrupt_offset = enic_msix_err_intr(enic); break; default: error_interrupt_enable = 0; @@ -249,15 +252,15 @@ void enic_init_vnic_resources(struct enic *enic) /* Init CQ resources * - * CQ[0 - n+m-1] point to INTR[0] for INTx, MSI - * CQ[0 - n+m-1] point to INTR[0 - n+m-1] for MSI-X + * All CQs point to INTR[0] for INTx, MSI + * CQ[i] point to INTR[ENIC_MSIX_IO_INTR_BASE + i] for MSI-X */ for (i = 0; i < enic->cq_count; i++) { switch (intr_mode) { case VNIC_DEV_INTR_MODE_MSIX: - interrupt_offset = i; + interrupt_offset = ENIC_MSIX_IO_INTR_BASE + i; break; default: interrupt_offset = 0; -- 2.51.0