The following changes since commit e607bbee553cfe73072870cef458cfa4e78133e2: Merge remote-tracking branch 'remotes/edgar/tags/edgar/xilinx-next-2018-01-26.for-upstream' into staging (2018-01-26 14:24:25 +0000) are available in the git repository at: https://github.com/jasowang/qemu.git tags/net-pull-request for you to fetch changes up to bf4835a4d5338bb7424827715df22570a8adc67c: MAINTAINERS: update Dmitry Fleytman email (2018-01-29 16:05:38 +0800) ---------------------------------------------------------------- ---------------------------------------------------------------- Mao Zhongyi (2): colo: modified the payload compare function colo: compare the packet based on the tcp sequence number Philippe Mathieu-Daudé (1): MAINTAINERS: update Dmitry Fleytman email Thomas Huth (3): net: Allow hubports to connect to other netdevs net: Allow netdevs to be used with 'hostfwd_add' and 'hostfwd_remove' qemu-doc: Get rid of "vlan=X" example in the documentation MAINTAINERS | 8 +- hmp-commands.hx | 4 +- net/colo-compare.c | 411 +++++++++++++++++++++++++++++++++-------------------- net/colo.c | 9 ++ net/colo.h | 15 ++ net/hub.c | 27 +++- net/hub.h | 3 +- net/net.c | 2 +- net/slirp.c | 33 +++-- net/trace-events | 2 +- qapi/net.json | 4 +- qemu-options.hx | 12 +- 12 files changed, 347 insertions(+), 183 deletions(-) From: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> Modified the function colo_packet_compare_common to prepare for the tcp packet comparison in the next patch. Cc: Zhang Chen <zhangckid@gmail.com> Cc: Li Zhijian <lizhijian@cn.fujitsu.com> Cc: Jason Wang <jasowang@redhat.com> Signed-off-by: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com> Signed-off-by: Zhang Chen <zhangckid@gmail.com> Reviewed-by: Zhang Chen <zhangckid@gmail.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- net/colo-compare.c | 88 +++++++++++++++++++++++++++--------------------------- 1 file changed, 44 insertions(+), 44 deletions(-) diff --git a/net/colo-compare.c b/net/colo-compare.c index XXXXXXX..XXXXXXX 100644 --- a/net/colo-compare.c +++ b/net/colo-compare.c @@ -XXX,XX +XXX,XX @@ static int packet_enqueue(CompareState *s, int mode, Connection **con) * return: 0 means packet same * > 0 || < 0 means packet different */ -static int colo_packet_compare_common(Packet *ppkt, - Packet *spkt, - int poffset, - int soffset) +static int colo_compare_packet_payload(Packet *ppkt, + Packet *spkt, + uint16_t poffset, + uint16_t soffset, + uint16_t len) + { if (trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE)) { char pri_ip_src[20], pri_ip_dst[20], sec_ip_src[20], sec_ip_dst[20]; @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_common(Packet *ppkt, sec_ip_src, sec_ip_dst); } - poffset = ppkt->vnet_hdr_len + poffset; - soffset = ppkt->vnet_hdr_len + soffset; - - if (ppkt->size - poffset == spkt->size - soffset) { - return memcmp(ppkt->data + poffset, - spkt->data + soffset, - spkt->size - soffset); - } else { - trace_colo_compare_main("Net packet size are not the same"); - return -1; - } + return memcmp(ppkt->data + poffset, spkt->data + soffset, len); } /* @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_tcp(Packet *spkt, Packet *ppkt) * the secondary guest's timestamp. COLO just focus on payload, * so we just need skip this field. */ - if (ptcp->th_off > 5) { - ptrdiff_t ptcp_offset, stcp_offset; - ptcp_offset = ppkt->transport_header - (uint8_t *)ppkt->data - + (ptcp->th_off * 4) - ppkt->vnet_hdr_len; - stcp_offset = spkt->transport_header - (uint8_t *)spkt->data - + (stcp->th_off * 4) - spkt->vnet_hdr_len; + ptrdiff_t ptcp_offset, stcp_offset; - /* - * When network is busy, some tcp options(like sack) will unpredictable - * occur in primary side or secondary side. it will make packet size - * not same, but the two packet's payload is identical. colo just - * care about packet payload, so we skip the option field. - */ - res = colo_packet_compare_common(ppkt, spkt, ptcp_offset, stcp_offset); - } else if (ptcp->th_sum == stcp->th_sum) { - res = colo_packet_compare_common(ppkt, spkt, ETH_HLEN, ETH_HLEN); + ptcp_offset = ppkt->transport_header - (uint8_t *)ppkt->data + + (ptcp->th_off << 2) - ppkt->vnet_hdr_len; + stcp_offset = spkt->transport_header - (uint8_t *)spkt->data + + (stcp->th_off << 2) - spkt->vnet_hdr_len; + if (ppkt->size - ptcp_offset == spkt->size - stcp_offset) { + res = colo_compare_packet_payload(ppkt, spkt, + ptcp_offset, stcp_offset, + ppkt->size - ptcp_offset); } else { + trace_colo_compare_main("TCP: payload size of packets are different"); res = -1; } @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_tcp(Packet *spkt, Packet *ppkt) */ static int colo_packet_compare_udp(Packet *spkt, Packet *ppkt) { - int ret; - int network_header_length = ppkt->ip->ip_hl * 4; + uint16_t network_header_length = ppkt->ip->ip_hl << 2; + uint16_t offset = network_header_length + ETH_HLEN + ppkt->vnet_hdr_len; trace_colo_compare_main("compare udp"); @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_udp(Packet *spkt, Packet *ppkt) * other field like TOS,TTL,IP Checksum. we only need to compare * the ip payload here. */ - ret = colo_packet_compare_common(ppkt, spkt, - network_header_length + ETH_HLEN, - network_header_length + ETH_HLEN); - - if (ret) { + if (ppkt->size != spkt->size) { + trace_colo_compare_main("UDP: payload size of packets are different"); + return -1; + } + if (colo_compare_packet_payload(ppkt, spkt, offset, offset, + ppkt->size - offset)) { trace_colo_compare_udp_miscompare("primary pkt size", ppkt->size); trace_colo_compare_udp_miscompare("Secondary pkt size", spkt->size); if (trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE)) { @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_udp(Packet *spkt, Packet *ppkt) qemu_hexdump((char *)spkt->data, stderr, "colo-compare sec pkt", spkt->size); } + return -1; + } else { + return 0; } - - return ret; } /* @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_udp(Packet *spkt, Packet *ppkt) */ static int colo_packet_compare_icmp(Packet *spkt, Packet *ppkt) { - int network_header_length = ppkt->ip->ip_hl * 4; + uint16_t network_header_length = ppkt->ip->ip_hl << 2; + uint16_t offset = network_header_length + ETH_HLEN + ppkt->vnet_hdr_len; trace_colo_compare_main("compare icmp"); @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_icmp(Packet *spkt, Packet *ppkt) * other field like TOS,TTL,IP Checksum. we only need to compare * the ip payload here. */ - if (colo_packet_compare_common(ppkt, spkt, - network_header_length + ETH_HLEN, - network_header_length + ETH_HLEN)) { + if (ppkt->size != spkt->size) { + trace_colo_compare_main("ICMP: payload size of packets are different"); + return -1; + } + if (colo_compare_packet_payload(ppkt, spkt, offset, offset, + ppkt->size - offset)) { trace_colo_compare_icmp_miscompare("primary pkt size", ppkt->size); trace_colo_compare_icmp_miscompare("Secondary pkt size", @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_icmp(Packet *spkt, Packet *ppkt) */ static int colo_packet_compare_other(Packet *spkt, Packet *ppkt) { + uint16_t offset = ppkt->vnet_hdr_len; + trace_colo_compare_main("compare other"); if (trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE)) { char pri_ip_src[20], pri_ip_dst[20], sec_ip_src[20], sec_ip_dst[20]; @@ -XXX,XX +XXX,XX @@ static int colo_packet_compare_other(Packet *spkt, Packet *ppkt) sec_ip_src, sec_ip_dst); } - return colo_packet_compare_common(ppkt, spkt, 0, 0); + if (ppkt->size != spkt->size) { + trace_colo_compare_main("Other: payload size of packets are different"); + return -1; + } + return colo_compare_packet_payload(ppkt, spkt, offset, offset, + ppkt->size - offset); } static int colo_old_packet_check_one(Packet *pkt, int64_t *check_time) -- 2.7.4 From: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> Packet size some time different or when network is busy. Based on same payload size, but TCP protocol can not guarantee send the same one packet in the same way, like that: We send this payload: ------------------------------ | header |1|2|3|4|5|6|7|8|9|0| ------------------------------ primary: ppkt1: ---------------- | header |1|2|3| ---------------- ppkt2: ------------------------ | header |4|5|6|7|8|9|0| ------------------------ secondary: spkt1: ------------------------------ | header |1|2|3|4|5|6|7|8|9|0| ------------------------------ In the original method, ppkt1 and ppkt2 are different in size and spkt1, so they can't compare and trigger the checkpoint. I have tested FTP get 200M and 1G file many times, I found that the performance was less than 1% of the native. Now I reconstructed the comparison of TCP packets based on the TCP sequence number. first of all, ppkt1 and spkt1 have the same starting sequence number, so they can compare, even though their length is different. And then ppkt1 with a smaller payload length is used as the comparison length, if the payload is same, send out the ppkt1 and record the offset(the length of ppkt1 payload) in spkt1. The next comparison, ppkt2 and spkt1 can be compared from the recorded position of spkt1. like that: ---------------- | header |1|2|3| ppkt1 ---------|-----| | | ---------v-----v-------------- | header |1|2|3|4|5|6|7|8|9|0| spkt1 ---------------|\------------| | \offset | ---------v-------------v | header |4|5|6|7|8|9|0| ppkt2 ------------------------ In this way, the performance can reach native 20% in my multiple tests. Cc: Zhang Chen <zhangckid@gmail.com> Cc: Li Zhijian <lizhijian@cn.fujitsu.com> Cc: Jason Wang <jasowang@redhat.com> Signed-off-by: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com> Signed-off-by: Zhang Chen <zhangckid@gmail.com> Reviewed-by: Zhang Chen <zhangckid@gmail.com> Tested-by: Zhang Chen <zhangckid@gmail.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- net/colo-compare.c | 343 +++++++++++++++++++++++++++++++++++------------------ net/colo.c | 9 ++ net/colo.h | 15 +++ net/trace-events | 2 +- 4 files changed, 250 insertions(+), 119 deletions(-) diff --git a/net/colo-compare.c b/net/colo-compare.c index XXXXXXX..XXXXXXX 100644 --- a/net/colo-compare.c +++ b/net/colo-compare.c @@ -XXX,XX +XXX,XX @@ #define COMPARE_READ_LEN_MAX NET_BUFSIZE #define MAX_QUEUE_SIZE 1024 +#define COLO_COMPARE_FREE_PRIMARY 0x01 +#define COLO_COMPARE_FREE_SECONDARY 0x02 + /* TODO: Should be configurable */ #define REGULAR_PACKET_CHECK_MS 3000 @@ -XXX,XX +XXX,XX @@ static gint seq_sorter(Packet *a, Packet *b, gpointer data) return ntohl(atcp->th_seq) - ntohl(btcp->th_seq); } +static void fill_pkt_tcp_info(void *data, uint32_t *max_ack) +{ + Packet *pkt = data; + struct tcphdr *tcphd; + + tcphd = (struct tcphdr *)pkt->transport_header; + + pkt->tcp_seq = ntohl(tcphd->th_seq); + pkt->tcp_ack = ntohl(tcphd->th_ack); + *max_ack = *max_ack > pkt->tcp_ack ? *max_ack : pkt->tcp_ack; + pkt->header_size = pkt->transport_header - (uint8_t *)pkt->data + + (tcphd->th_off << 2) - pkt->vnet_hdr_len; + pkt->payload_size = pkt->size - pkt->header_size; + pkt->seq_end = pkt->tcp_seq + pkt->payload_size; + pkt->flags = tcphd->th_flags; +} + /* * Return 1 on success, if return 0 means the * packet will be dropped */ -static int colo_insert_packet(GQueue *queue, Packet *pkt) +static int colo_insert_packet(GQueue *queue, Packet *pkt, uint32_t *max_ack) { if (g_queue_get_length(queue) <= MAX_QUEUE_SIZE) { if (pkt->ip->ip_p == IPPROTO_TCP) { + fill_pkt_tcp_info(pkt, max_ack); g_queue_insert_sorted(queue, pkt, (GCompareDataFunc)seq_sorter, @@ -XXX,XX +XXX,XX @@ static int packet_enqueue(CompareState *s, int mode, Connection **con) } if (mode == PRIMARY_IN) { - if (!colo_insert_packet(&conn->primary_list, pkt)) { + if (!colo_insert_packet(&conn->primary_list, pkt, &conn->pack)) { error_report("colo compare primary queue size too big," "drop packet"); } } else { - if (!colo_insert_packet(&conn->secondary_list, pkt)) { + if (!colo_insert_packet(&conn->secondary_list, pkt, &conn->sack)) { error_report("colo compare secondary queue size too big," "drop packet"); } @@ -XXX,XX +XXX,XX @@ static int packet_enqueue(CompareState *s, int mode, Connection **con) return 0; } +static inline bool after(uint32_t seq1, uint32_t seq2) +{ + return (int32_t)(seq1 - seq2) > 0; +} + +static void colo_release_primary_pkt(CompareState *s, Packet *pkt) +{ + int ret; + ret = compare_chr_send(s, + pkt->data, + pkt->size, + pkt->vnet_hdr_len); + if (ret < 0) { + error_report("colo send primary packet failed"); + } + trace_colo_compare_main("packet same and release packet"); + packet_destroy(pkt, NULL); +} + /* * The IP packets sent by primary and secondary * will be compared in here @@ -XXX,XX +XXX,XX @@ static int colo_compare_packet_payload(Packet *ppkt, } /* - * Called from the compare thread on the primary - * for compare tcp packet - * compare_tcp copied from Dr. David Alan Gilbert's branch - */ -static int colo_packet_compare_tcp(Packet *spkt, Packet *ppkt) + * return true means that the payload is consist and + * need to make the next comparison, false means do + * the checkpoint +*/ +static bool colo_mark_tcp_pkt(Packet *ppkt, Packet *spkt, + int8_t *mark, uint32_t max_ack) { - struct tcphdr *ptcp, *stcp; - int res; + *mark = 0; + + if (ppkt->tcp_seq == spkt->tcp_seq && ppkt->seq_end == spkt->seq_end) { + if (colo_compare_packet_payload(ppkt, spkt, + ppkt->header_size, spkt->header_size, + ppkt->payload_size)) { + *mark = COLO_COMPARE_FREE_SECONDARY | COLO_COMPARE_FREE_PRIMARY; + return true; + } + } + if (ppkt->tcp_seq == spkt->tcp_seq && ppkt->seq_end == spkt->seq_end) { + if (colo_compare_packet_payload(ppkt, spkt, + ppkt->header_size, spkt->header_size, + ppkt->payload_size)) { + *mark = COLO_COMPARE_FREE_SECONDARY | COLO_COMPARE_FREE_PRIMARY; + return true; + } + } + + /* one part of secondary packet payload still need to be compared */ + if (!after(ppkt->seq_end, spkt->seq_end)) { + if (colo_compare_packet_payload(ppkt, spkt, + ppkt->header_size + ppkt->offset, + spkt->header_size + spkt->offset, + ppkt->payload_size - ppkt->offset)) { + if (!after(ppkt->tcp_ack, max_ack)) { + *mark = COLO_COMPARE_FREE_PRIMARY; + spkt->offset += ppkt->payload_size - ppkt->offset; + return true; + } else { + /* secondary guest hasn't ack the data, don't send + * out this packet + */ + return false; + } + } + } else { + /* primary packet is longer than secondary packet, compare + * the same part and mark the primary packet offset + */ + if (colo_compare_packet_payload(ppkt, spkt, + ppkt->header_size + ppkt->offset, + spkt->header_size + spkt->offset, + spkt->payload_size - spkt->offset)) { + *mark = COLO_COMPARE_FREE_SECONDARY; + ppkt->offset += spkt->payload_size - spkt->offset; + return true; + } + } - trace_colo_compare_main("compare tcp"); + return false; +} - ptcp = (struct tcphdr *)ppkt->transport_header; - stcp = (struct tcphdr *)spkt->transport_header; +static void colo_compare_tcp(CompareState *s, Connection *conn) +{ + Packet *ppkt = NULL, *spkt = NULL; + int8_t mark; /* - * The 'identification' field in the IP header is *very* random - * it almost never matches. Fudge this by ignoring differences in - * unfragmented packets; they'll normally sort themselves out if different - * anyway, and it should recover at the TCP level. - * An alternative would be to get both the primary and secondary to rewrite - * somehow; but that would need some sync traffic to sync the state - */ - if (ntohs(ppkt->ip->ip_off) & IP_DF) { - spkt->ip->ip_id = ppkt->ip->ip_id; - /* and the sum will be different if the IDs were different */ - spkt->ip->ip_sum = ppkt->ip->ip_sum; + * If ppkt and spkt have the same payload, but ppkt's ACK + * is greater than spkt's ACK, in this case we can not + * send the ppkt because it will cause the secondary guest + * to miss sending some data in the next. Therefore, we + * record the maximum ACK in the current queue at both + * primary side and secondary side. Only when the ack is + * less than the smaller of the two maximum ack, then we + * can ensure that the packet's payload is acknowledged by + * primary and secondary. + */ + uint32_t min_ack = conn->pack > conn->sack ? conn->sack : conn->pack; + +pri: + if (g_queue_is_empty(&conn->primary_list)) { + return; } + ppkt = g_queue_pop_head(&conn->primary_list); +sec: + if (g_queue_is_empty(&conn->secondary_list)) { + g_queue_push_head(&conn->primary_list, ppkt); + return; + } + spkt = g_queue_pop_head(&conn->secondary_list); - /* - * Check tcp header length for tcp option field. - * th_off > 5 means this tcp packet have options field. - * The tcp options maybe always different. - * for example: - * From RFC 7323. - * TCP Timestamps option (TSopt): - * Kind: 8 - * - * Length: 10 bytes - * - * +-------+-------+---------------------+---------------------+ - * |Kind=8 | 10 | TS Value (TSval) |TS Echo Reply (TSecr)| - * +-------+-------+---------------------+---------------------+ - * 1 1 4 4 - * - * In this case the primary guest's timestamp always different with - * the secondary guest's timestamp. COLO just focus on payload, - * so we just need skip this field. - */ + if (ppkt->tcp_seq == ppkt->seq_end) { + colo_release_primary_pkt(s, ppkt); + ppkt = NULL; + } - ptrdiff_t ptcp_offset, stcp_offset; + if (ppkt && conn->compare_seq && !after(ppkt->seq_end, conn->compare_seq)) { + trace_colo_compare_main("pri: this packet has compared"); + colo_release_primary_pkt(s, ppkt); + ppkt = NULL; + } - ptcp_offset = ppkt->transport_header - (uint8_t *)ppkt->data - + (ptcp->th_off << 2) - ppkt->vnet_hdr_len; - stcp_offset = spkt->transport_header - (uint8_t *)spkt->data - + (stcp->th_off << 2) - spkt->vnet_hdr_len; - if (ppkt->size - ptcp_offset == spkt->size - stcp_offset) { - res = colo_compare_packet_payload(ppkt, spkt, - ptcp_offset, stcp_offset, - ppkt->size - ptcp_offset); + if (spkt->tcp_seq == spkt->seq_end) { + packet_destroy(spkt, NULL); + if (!ppkt) { + goto pri; + } else { + goto sec; + } } else { - trace_colo_compare_main("TCP: payload size of packets are different"); - res = -1; + if (conn->compare_seq && !after(spkt->seq_end, conn->compare_seq)) { + trace_colo_compare_main("sec: this packet has compared"); + packet_destroy(spkt, NULL); + if (!ppkt) { + goto pri; + } else { + goto sec; + } + } + if (!ppkt) { + g_queue_push_head(&conn->secondary_list, spkt); + goto pri; + } } - if (res != 0 && - trace_event_get_state_backends(TRACE_COLO_COMPARE_MISCOMPARE)) { - char pri_ip_src[20], pri_ip_dst[20], sec_ip_src[20], sec_ip_dst[20]; - - strcpy(pri_ip_src, inet_ntoa(ppkt->ip->ip_src)); - strcpy(pri_ip_dst, inet_ntoa(ppkt->ip->ip_dst)); - strcpy(sec_ip_src, inet_ntoa(spkt->ip->ip_src)); - strcpy(sec_ip_dst, inet_ntoa(spkt->ip->ip_dst)); - - trace_colo_compare_ip_info(ppkt->size, pri_ip_src, - pri_ip_dst, spkt->size, - sec_ip_src, sec_ip_dst); - - trace_colo_compare_tcp_info("pri tcp packet", - ntohl(ptcp->th_seq), - ntohl(ptcp->th_ack), - res, ptcp->th_flags, - ppkt->size); - - trace_colo_compare_tcp_info("sec tcp packet", - ntohl(stcp->th_seq), - ntohl(stcp->th_ack), - res, stcp->th_flags, - spkt->size); + if (colo_mark_tcp_pkt(ppkt, spkt, &mark, min_ack)) { + trace_colo_compare_tcp_info("pri", + ppkt->tcp_seq, ppkt->tcp_ack, + ppkt->header_size, ppkt->payload_size, + ppkt->offset, ppkt->flags); + + trace_colo_compare_tcp_info("sec", + spkt->tcp_seq, spkt->tcp_ack, + spkt->header_size, spkt->payload_size, + spkt->offset, spkt->flags); + + if (mark == COLO_COMPARE_FREE_PRIMARY) { + conn->compare_seq = ppkt->seq_end; + colo_release_primary_pkt(s, ppkt); + g_queue_push_head(&conn->secondary_list, spkt); + goto pri; + } + if (mark == COLO_COMPARE_FREE_SECONDARY) { + conn->compare_seq = spkt->seq_end; + packet_destroy(spkt, NULL); + goto sec; + } + if (mark == (COLO_COMPARE_FREE_PRIMARY | COLO_COMPARE_FREE_SECONDARY)) { + conn->compare_seq = ppkt->seq_end; + colo_release_primary_pkt(s, ppkt); + packet_destroy(spkt, NULL); + goto pri; + } + } else { + g_queue_push_head(&conn->primary_list, ppkt); + g_queue_push_head(&conn->secondary_list, spkt); qemu_hexdump((char *)ppkt->data, stderr, "colo-compare ppkt", ppkt->size); qemu_hexdump((char *)spkt->data, stderr, "colo-compare spkt", spkt->size); - } - return res; + /* + * colo_compare_inconsistent_notify(); + * TODO: notice to checkpoint(); + */ + } } + /* * Called from the compare thread on the primary * for compare udp packet @@ -XXX,XX +XXX,XX @@ static void colo_old_packet_check(void *opaque) (GCompareFunc)colo_old_packet_check_one_conn); } -/* - * Called from the compare thread on the primary - * for compare packet with secondary list of the - * specified connection when a new packet was - * queued to it. - */ -static void colo_compare_connection(void *opaque, void *user_data) +static void colo_compare_packet(CompareState *s, Connection *conn, + int (*HandlePacket)(Packet *spkt, + Packet *ppkt)) { - CompareState *s = user_data; - Connection *conn = opaque; Packet *pkt = NULL; GList *result = NULL; - int ret; while (!g_queue_is_empty(&conn->primary_list) && !g_queue_is_empty(&conn->secondary_list)) { pkt = g_queue_pop_head(&conn->primary_list); - switch (conn->ip_proto) { - case IPPROTO_TCP: - result = g_queue_find_custom(&conn->secondary_list, - pkt, (GCompareFunc)colo_packet_compare_tcp); - break; - case IPPROTO_UDP: - result = g_queue_find_custom(&conn->secondary_list, - pkt, (GCompareFunc)colo_packet_compare_udp); - break; - case IPPROTO_ICMP: - result = g_queue_find_custom(&conn->secondary_list, - pkt, (GCompareFunc)colo_packet_compare_icmp); - break; - default: - result = g_queue_find_custom(&conn->secondary_list, - pkt, (GCompareFunc)colo_packet_compare_other); - break; - } + result = g_queue_find_custom(&conn->secondary_list, + pkt, (GCompareFunc)HandlePacket); if (result) { - ret = compare_chr_send(s, - pkt->data, - pkt->size, - pkt->vnet_hdr_len); - if (ret < 0) { - error_report("colo_send_primary_packet failed"); - } - trace_colo_compare_main("packet same and release packet"); + colo_release_primary_pkt(s, pkt); g_queue_remove(&conn->secondary_list, result->data); - packet_destroy(pkt, NULL); } else { /* * If one packet arrive late, the secondary_list or @@ -XXX,XX +XXX,XX @@ static void colo_compare_connection(void *opaque, void *user_data) } } +/* + * Called from the compare thread on the primary + * for compare packet with secondary list of the + * specified connection when a new packet was + * queued to it. + */ +static void colo_compare_connection(void *opaque, void *user_data) +{ + CompareState *s = user_data; + Connection *conn = opaque; + + switch (conn->ip_proto) { + case IPPROTO_TCP: + colo_compare_tcp(s, conn); + break; + case IPPROTO_UDP: + colo_compare_packet(s, conn, colo_packet_compare_udp); + break; + case IPPROTO_ICMP: + colo_compare_packet(s, conn, colo_packet_compare_icmp); + break; + default: + colo_compare_packet(s, conn, colo_packet_compare_other); + break; + } +} + static int compare_chr_send(CompareState *s, const uint8_t *buf, uint32_t size, diff --git a/net/colo.c b/net/colo.c index XXXXXXX..XXXXXXX 100644 --- a/net/colo.c +++ b/net/colo.c @@ -XXX,XX +XXX,XX @@ Connection *connection_new(ConnectionKey *key) conn->processing = false; conn->offset = 0; conn->syn_flag = 0; + conn->pack = 0; + conn->sack = 0; g_queue_init(&conn->primary_list); g_queue_init(&conn->secondary_list); @@ -XXX,XX +XXX,XX @@ Packet *packet_new(const void *data, int size, int vnet_hdr_len) pkt->size = size; pkt->creation_ms = qemu_clock_get_ms(QEMU_CLOCK_HOST); pkt->vnet_hdr_len = vnet_hdr_len; + pkt->tcp_seq = 0; + pkt->tcp_ack = 0; + pkt->seq_end = 0; + pkt->header_size = 0; + pkt->payload_size = 0; + pkt->offset = 0; + pkt->flags = 0; return pkt; } diff --git a/net/colo.h b/net/colo.h index XXXXXXX..XXXXXXX 100644 --- a/net/colo.h +++ b/net/colo.h @@ -XXX,XX +XXX,XX @@ typedef struct Packet { int64_t creation_ms; /* Get vnet_hdr_len from filter */ uint32_t vnet_hdr_len; + uint32_t tcp_seq; /* sequence number */ + uint32_t tcp_ack; /* acknowledgement number */ + /* the sequence number of the last byte of the packet */ + uint32_t seq_end; + uint8_t header_size; /* the header length */ + uint16_t payload_size; /* the payload length */ + /* record the payload offset(the length that has been compared) */ + uint16_t offset; + uint8_t flags; /* Flags(aka Control bits) */ } Packet; typedef struct ConnectionKey { @@ -XXX,XX +XXX,XX @@ typedef struct Connection { /* flag to enqueue unprocessed_connections */ bool processing; uint8_t ip_proto; + /* record the sequence number that has been compared */ + uint32_t compare_seq; + /* the maximum of acknowledgement number in primary_list queue */ + uint32_t pack; + /* the maximum of acknowledgement number in secondary_list queue */ + uint32_t sack; /* offset = secondary_seq - primary_seq */ tcp_seq offset; /* diff --git a/net/trace-events b/net/trace-events index XXXXXXX..XXXXXXX 100644 --- a/net/trace-events +++ b/net/trace-events @@ -XXX,XX +XXX,XX @@ colo_compare_icmp_miscompare(const char *sta, int size) ": %s = %d" colo_compare_ip_info(int psize, const char *sta, const char *stb, int ssize, const char *stc, const char *std) "ppkt size = %d, ip_src = %s, ip_dst = %s, spkt size = %d, ip_src = %s, ip_dst = %s" colo_old_packet_check_found(int64_t old_time) "%" PRId64 colo_compare_miscompare(void) "" -colo_compare_tcp_info(const char *pkt, uint32_t seq, uint32_t ack, int res, uint32_t flag, int size) "side: %s seq/ack= %u/%u res= %d flags= 0x%x pkt_size: %d\n" +colo_compare_tcp_info(const char *pkt, uint32_t seq, uint32_t ack, int hdlen, int pdlen, int offset, int flags) "%s: seq/ack= %u/%u hdlen= %d pdlen= %d offset= %d flags=%d\n" # net/filter-rewriter.c colo_filter_rewriter_debug(void) "" -- 2.7.4 From: Thomas Huth <thuth@redhat.com> QEMU can emulate hubs to connect NICs and netdevs. This is currently primarily used for the mis-named 'vlan' feature of the networking subsystem. Now the 'vlan' feature has been marked as deprecated, since its name is rather confusing and the users often rather mis-configure their network when trying to use it. But while the 'vlan' parameter should be removed at one point in time, the basic idea of emulating a hub in QEMU is still good: It's useful for bundling up the output of multiple NICs into one single l2tp netdev for example. Now to be able to use the hubport feature without 'vlan's, there is one missing piece: The possibility to connect a hubport to a netdev, too. This patch adds this possibility by introducing a new "netdev=..." parameter to the hubports. To bundle up the output of multiple NICs into one socket netdev, you can now run QEMU with these parameters for example: qemu-system-ppc64 ... -netdev socket,id=s1,connect=:11122 \ -netdev hubport,hubid=1,id=h1,netdev=s1 \ -netdev hubport,hubid=1,id=h2 -device e1000,netdev=h2 \ -netdev hubport,hubid=1,id=h3 -device virtio-net-pci,netdev=h3 For using the socket netdev, you have got to start another QEMU as the receiving side first, for example with network dumping enabled: qemu-system-x86_64 -M isapc -netdev socket,id=s0,listen=:11122 \ -device ne2k_isa,netdev=s0 \ -object filter-dump,id=f1,netdev=s0,file=/tmp/dump.dat After the ppc64 guest tried to boot from both NICs, you can see in the dump file (using Wireshark, for example), that the output of both NICs (the e1000 and the virtio-net-pci) has been successfully transfered via the socket netdev in this case. Suggested-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Thomas Huth <thuth@redhat.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- net/hub.c | 27 +++++++++++++++++++++------ net/hub.h | 3 ++- net/net.c | 2 +- qapi/net.json | 4 +++- qemu-options.hx | 8 +++++--- 5 files changed, 32 insertions(+), 12 deletions(-) diff --git a/net/hub.c b/net/hub.c index XXXXXXX..XXXXXXX 100644 --- a/net/hub.c +++ b/net/hub.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "qapi/error.h" #include "monitor/monitor.h" #include "net/net.h" #include "clients.h" @@ -XXX,XX +XXX,XX @@ static NetClientInfo net_hub_port_info = { .cleanup = net_hub_port_cleanup, }; -static NetHubPort *net_hub_port_new(NetHub *hub, const char *name) +static NetHubPort *net_hub_port_new(NetHub *hub, const char *name, + NetClientState *hubpeer) { NetClientState *nc; NetHubPort *port; @@ -XXX,XX +XXX,XX @@ static NetHubPort *net_hub_port_new(NetHub *hub, const char *name) name = default_name; } - nc = qemu_new_net_client(&net_hub_port_info, NULL, "hub", name); + nc = qemu_new_net_client(&net_hub_port_info, hubpeer, "hub", name); port = DO_UPCAST(NetHubPort, nc, nc); port->id = id; port->hub = hub; @@ -XXX,XX +XXX,XX @@ static NetHubPort *net_hub_port_new(NetHub *hub, const char *name) /** * Create a port on a given hub + * @hub_id: Number of the hub * @name: Net client name or NULL for default name. + * @hubpeer: Peer to use (if "netdev=id" has been specified) * * If there is no existing hub with the given id then a new hub is created. */ -NetClientState *net_hub_add_port(int hub_id, const char *name) +NetClientState *net_hub_add_port(int hub_id, const char *name, + NetClientState *hubpeer) { NetHub *hub; NetHubPort *port; @@ -XXX,XX +XXX,XX @@ NetClientState *net_hub_add_port(int hub_id, const char *name) hub = net_hub_new(hub_id); } - port = net_hub_port_new(hub, name); + port = net_hub_port_new(hub, name, hubpeer); return &port->nc; } @@ -XXX,XX +XXX,XX @@ NetClientState *net_hub_port_find(int hub_id) } } - nc = net_hub_add_port(hub_id, NULL); + nc = net_hub_add_port(hub_id, NULL, NULL); return nc; } @@ -XXX,XX +XXX,XX @@ int net_init_hubport(const Netdev *netdev, const char *name, NetClientState *peer, Error **errp) { const NetdevHubPortOptions *hubport; + NetClientState *hubpeer = NULL; assert(netdev->type == NET_CLIENT_DRIVER_HUBPORT); assert(!peer); hubport = &netdev->u.hubport; - net_hub_add_port(hubport->hubid, name); + if (hubport->has_netdev) { + hubpeer = qemu_find_netdev(hubport->netdev); + if (!hubpeer) { + error_setg(errp, "netdev '%s' not found", hubport->netdev); + return -1; + } + } + + net_hub_add_port(hubport->hubid, name, hubpeer); + return 0; } diff --git a/net/hub.h b/net/hub.h index XXXXXXX..XXXXXXX 100644 --- a/net/hub.h +++ b/net/hub.h @@ -XXX,XX +XXX,XX @@ #include "qemu-common.h" -NetClientState *net_hub_add_port(int hub_id, const char *name); +NetClientState *net_hub_add_port(int hub_id, const char *name, + NetClientState *hubpeer); NetClientState *net_hub_find_client_by_name(int hub_id, const char *name); void net_hub_info(Monitor *mon); void net_hub_check_clients(void); diff --git a/net/net.c b/net/net.c index XXXXXXX..XXXXXXX 100644 --- a/net/net.c +++ b/net/net.c @@ -XXX,XX +XXX,XX @@ static int net_client_init1(const void *object, bool is_netdev, Error **errp) /* Do not add to a vlan if it's a nic with a netdev= parameter. */ if (netdev->type != NET_CLIENT_DRIVER_NIC || !opts->u.nic.has_netdev) { - peer = net_hub_add_port(net->has_vlan ? net->vlan : 0, NULL); + peer = net_hub_add_port(net->has_vlan ? net->vlan : 0, NULL, NULL); } if (net->has_vlan && !vlan_warned) { diff --git a/qapi/net.json b/qapi/net.json index XXXXXXX..XXXXXXX 100644 --- a/qapi/net.json +++ b/qapi/net.json @@ -XXX,XX +XXX,XX @@ # Connect two or more net clients through a software hub. # # @hubid: hub identifier number +# @netdev: used to connect hub to a netdev instead of a device (since 2.12) # # Since: 1.2 ## { 'struct': 'NetdevHubPortOptions', 'data': { - 'hubid': 'int32' } } + 'hubid': 'int32', + '*netdev': 'str' } } ## # @NetdevNetmapOptions: diff --git a/qemu-options.hx b/qemu-options.hx index XXXXXXX..XXXXXXX 100644 --- a/qemu-options.hx +++ b/qemu-options.hx @@ -XXX,XX +XXX,XX @@ DEF("netdev", HAS_ARG, QEMU_OPTION_netdev, #endif "-netdev vhost-user,id=str,chardev=dev[,vhostforce=on|off]\n" " configure a vhost-user network, backed by a chardev 'dev'\n" - "-netdev hubport,id=str,hubid=n\n" + "-netdev hubport,id=str,hubid=n[,netdev=nd]\n" " configure a hub port on QEMU VLAN 'n'\n", QEMU_ARCH_ALL) DEF("net", HAS_ARG, QEMU_OPTION_net, "-net nic[,vlan=n][,netdev=nd][,macaddr=mac][,model=type][,name=str][,addr=str][,vectors=v]\n" @@ -XXX,XX +XXX,XX @@ vde_switch -F -sock /tmp/myswitch qemu-system-i386 linux.img -net nic -net vde,sock=/tmp/myswitch @end example -@item -netdev hubport,id=@var{id},hubid=@var{hubid} +@item -netdev hubport,id=@var{id},hubid=@var{hubid}[,netdev=@var{nd}] Create a hub port on QEMU "vlan" @var{hubid}. The hubport netdev lets you connect a NIC to a QEMU "vlan" instead of a single netdev. @code{-net} and @code{-device} with parameter @option{vlan} create the -required hub automatically. +required hub automatically. Alternatively, you can also connect the hubport +to another netdev with ID @var{nd} by using the @option{netdev=@var{nd}} +option. @item -netdev vhost-user,chardev=@var{id}[,vhostforce=on|off][,queues=n] -- 2.7.4 From: Thomas Huth <thuth@redhat.com> It does not make much sense to limit these commands to the legacy 'vlan' concept only, they should work with the modern netdevs, too. So now it is possible to use this command with one, two or three parameters. With one parameter, the command installs a hostfwd rule on the default "user" network: hostfwd_add tcp:... With two parameters, the command installs a hostfwd rule on a netdev (that's the new way of using this command): hostfwd_add netdev_id tcp:... With three parameters, the command installs a rule on a 'vlan' (aka hub): hostfwd_add hub_id name tcp:... Same applies to the hostfwd_remove command now. Signed-off-by: Thomas Huth <thuth@redhat.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hmp-commands.hx | 4 ++-- net/slirp.c | 33 +++++++++++++++++++++++---------- 2 files changed, 25 insertions(+), 12 deletions(-) diff --git a/hmp-commands.hx b/hmp-commands.hx index XXXXXXX..XXXXXXX 100644 --- a/hmp-commands.hx +++ b/hmp-commands.hx @@ -XXX,XX +XXX,XX @@ ETEXI { .name = "hostfwd_add", .args_type = "arg1:s,arg2:s?,arg3:s?", - .params = "[vlan_id name] [tcp|udp]:[hostaddr]:hostport-[guestaddr]:guestport", + .params = "[hub_id name]|[netdev_id] [tcp|udp]:[hostaddr]:hostport-[guestaddr]:guestport", .help = "redirect TCP or UDP connections from host to guest (requires -net user)", .cmd = hmp_hostfwd_add, }, @@ -XXX,XX +XXX,XX @@ ETEXI { .name = "hostfwd_remove", .args_type = "arg1:s,arg2:s?,arg3:s?", - .params = "[vlan_id name] [tcp|udp]:[hostaddr]:hostport", + .params = "[hub_id name]|[netdev_id] [tcp|udp]:[hostaddr]:hostport", .help = "remove host-to-guest TCP or UDP redirection", .cmd = hmp_hostfwd_remove, }, diff --git a/net/slirp.c b/net/slirp.c index XXXXXXX..XXXXXXX 100644 --- a/net/slirp.c +++ b/net/slirp.c @@ -XXX,XX +XXX,XX @@ error: return -1; } -static SlirpState *slirp_lookup(Monitor *mon, const char *vlan, - const char *stack) +static SlirpState *slirp_lookup(Monitor *mon, const char *hub_id, + const char *name) { - - if (vlan) { + if (name) { NetClientState *nc; - nc = net_hub_find_client_by_name(strtol(vlan, NULL, 0), stack); - if (!nc) { - monitor_printf(mon, "unrecognized (vlan-id, stackname) pair\n"); - return NULL; + if (hub_id) { + nc = net_hub_find_client_by_name(strtol(hub_id, NULL, 0), name); + if (!nc) { + monitor_printf(mon, "unrecognized (vlan-id, stackname) pair\n"); + return NULL; + } + } else { + nc = qemu_find_netdev(name); + if (!nc) { + monitor_printf(mon, "unrecognized netdev id '%s'\n", name); + return NULL; + } } if (strcmp(nc->model, "user")) { monitor_printf(mon, "invalid device specified\n"); @@ -XXX,XX +XXX,XX @@ void hmp_hostfwd_remove(Monitor *mon, const QDict *qdict) const char *arg2 = qdict_get_try_str(qdict, "arg2"); const char *arg3 = qdict_get_try_str(qdict, "arg3"); - if (arg2) { + if (arg3) { s = slirp_lookup(mon, arg1, arg2); src_str = arg3; + } else if (arg2) { + s = slirp_lookup(mon, NULL, arg1); + src_str = arg2; } else { s = slirp_lookup(mon, NULL, NULL); src_str = arg1; @@ -XXX,XX +XXX,XX @@ void hmp_hostfwd_add(Monitor *mon, const QDict *qdict) const char *arg2 = qdict_get_try_str(qdict, "arg2"); const char *arg3 = qdict_get_try_str(qdict, "arg3"); - if (arg2) { + if (arg3) { s = slirp_lookup(mon, arg1, arg2); redir_str = arg3; + } else if (arg2) { + s = slirp_lookup(mon, NULL, arg1); + redir_str = arg2; } else { s = slirp_lookup(mon, NULL, NULL); redir_str = arg1; -- 2.7.4 From: Thomas Huth <thuth@redhat.com> The vlan concept is marked as deprecated, so we should not use this for examples in the documentation anymore. Signed-off-by: Thomas Huth <thuth@redhat.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- qemu-options.hx | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/qemu-options.hx b/qemu-options.hx index XXXXXXX..XXXXXXX 100644 --- a/qemu-options.hx +++ b/qemu-options.hx @@ -XXX,XX +XXX,XX @@ qemu-system-i386 linux.img -net nic -net tap #launch a QEMU instance with two NICs, each one connected #to a TAP device qemu-system-i386 linux.img \ - -net nic,vlan=0 -net tap,vlan=0,ifname=tap0 \ - -net nic,vlan=1 -net tap,vlan=1,ifname=tap1 + -netdev tap,id=nd0,ifname=tap0 -device e1000,netdev=nd0 \ + -netdev tap,id=nd1,ifname=tap1 -device rtl8139,netdev=nd1 @end example @example -- 2.7.4 From: Philippe Mathieu-Daudé <f4bug@amsat.org> gently asked by his automatic reply :) Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- MAINTAINERS | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: hw/scsi/mfi.h F: tests/megasas-test.c Network packet abstractions -M: Dmitry Fleytman <dmitry@daynix.com> +M: Dmitry Fleytman <dmitry.fleytman@gmail.com> S: Maintained F: include/net/eth.h F: net/eth.c @@ -XXX,XX +XXX,XX @@ F: hw/net/net_rx_pkt* F: hw/net/net_tx_pkt* Vmware -M: Dmitry Fleytman <dmitry@daynix.com> +M: Dmitry Fleytman <dmitry.fleytman@gmail.com> S: Maintained F: hw/net/vmxnet* F: hw/scsi/vmw_pvscsi* @@ -XXX,XX +XXX,XX @@ F: hw/mem/nvdimm.c F: include/hw/mem/nvdimm.h e1000x -M: Dmitry Fleytman <dmitry@daynix.com> +M: Dmitry Fleytman <dmitry.fleytman@gmail.com> S: Maintained F: hw/net/e1000x* e1000e -M: Dmitry Fleytman <dmitry@daynix.com> +M: Dmitry Fleytman <dmitry.fleytman@gmail.com> S: Maintained F: hw/net/e1000e* -- 2.7.4 The following changes since commit 886c0453cbf10eebd42a9ccf89c3e46eb389c357: Merge tag 'pull-qapi-2023-05-17-v2' of https://repo.or.cz/qemu/armbru into staging (2023-05-22 15:54:21 -0700) are available in the git repository at: https://github.com/jasowang/qemu.git tags/net-pull-request for you to fetch changes up to 792676c165159c11412346870fd58fd243ab2166: rtl8139: fix large_send_mss divide-by-zero (2023-05-23 15:20:15 +0800) ---------------------------------------------------------------- ---------------------------------------------------------------- Akihiko Odaki (48): hw/net/net_tx_pkt: Decouple implementation from PCI hw/net/net_tx_pkt: Decouple interface from PCI e1000x: Fix BPRC and MPRC igb: Fix Rx packet type encoding igb: Do not require CTRL.VME for tx VLAN tagging igb: Clear IMS bits when committing ICR access net/net_rx_pkt: Use iovec for net_rx_pkt_set_protocols() e1000e: Always copy ethernet header igb: Always copy ethernet header Fix references to igb Avocado test tests/avocado: Remove unused imports tests/avocado: Remove test_igb_nomsi_kvm hw/net/net_tx_pkt: Remove net_rx_pkt_get_l4_info net/eth: Rename eth_setup_vlan_headers_ex e1000x: Share more Rx filtering logic e1000x: Take CRC into consideration for size check e1000x: Rename TcpIpv6 into TcpIpv6Ex e1000e: Always log status after building rx metadata igb: Always log status after building rx metadata igb: Remove goto igb: Read DCMD.VLE of the first Tx descriptor e1000e: Reset packet state after emptying Tx queue vmxnet3: Reset packet state after emptying Tx queue igb: Add more definitions for Tx descriptor igb: Share common VF constants igb: Fix igb_mac_reg_init coding style alignment igb: Clear EICR bits for delayed MSI-X interrupts e1000e: Rename a variable in e1000e_receive_internal() igb: Rename a variable in igb_receive_internal() net/eth: Use void pointers net/eth: Always add VLAN tag hw/net/net_rx_pkt: Enforce alignment for eth_header tests/qtest/libqos/igb: Set GPIE.Multiple_MSIX igb: Implement MSI-X single vector mode igb: Use UDP for RSS hash igb: Implement Rx SCTP CSO igb: Implement Tx SCTP CSO igb: Strip the second VLAN tag for extended VLAN igb: Filter with the second VLAN tag for extended VLAN igb: Implement igb-specific oversize check igb: Implement Rx PTP2 timestamp igb: Implement Tx timestamp e1000e: Notify only new interrupts igb: Notify only new interrupts igb: Clear-on-read ICR when ICR.INTA is set vmxnet3: Do not depend on PC MAINTAINERS: Add a reviewer for network packet abstractions docs/system/devices/igb: Note igb is tested for DPDK Stefan Hajnoczi (1): rtl8139: fix large_send_mss divide-by-zero timothee.cocault@gmail.com (1): e1000e: Fix tx/rx counters MAINTAINERS | 3 +- docs/system/devices/igb.rst | 14 +- hw/net/Kconfig | 2 +- hw/net/e1000.c | 46 +- hw/net/e1000e_core.c | 297 ++++----- hw/net/e1000e_core.h | 2 - hw/net/e1000x_common.c | 82 ++- hw/net/e1000x_common.h | 9 +- hw/net/e1000x_regs.h | 24 +- hw/net/igb.c | 10 +- hw/net/igb_common.h | 24 +- hw/net/igb_core.c | 722 +++++++++++---------- hw/net/igb_regs.h | 67 +- hw/net/igbvf.c | 7 - hw/net/net_rx_pkt.c | 107 +-- hw/net/net_rx_pkt.h | 38 +- hw/net/net_tx_pkt.c | 101 +-- hw/net/net_tx_pkt.h | 46 +- hw/net/rtl8139.c | 3 + hw/net/trace-events | 19 +- hw/net/virtio-net.c | 7 +- hw/net/vmxnet3.c | 22 +- include/net/eth.h | 29 +- include/qemu/crc32c.h | 1 + net/eth.c | 100 +-- scripts/ci/org.centos/stream/8/x86_64/test-avocado | 3 +- tests/avocado/netdev-ethtool.py | 17 +- tests/qtest/libqos/igb.c | 1 + util/crc32c.c | 8 + 29 files changed, 979 insertions(+), 832 deletions(-) From: "timothee.cocault@gmail.com" <timothee.cocault@gmail.com> The bytes and packets counter registers are cleared on read. Copying the "total counter" registers to the "good counter" registers has side effects. If the "total" register is never read by the OS, it only gets incremented. This leads to exponential growth of the "good" register. This commit increments the counters individually to avoid this. Signed-off-by: Timothée Cocault <timothee.cocault@gmail.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000.c | 5 ++--- hw/net/e1000e_core.c | 5 ++--- hw/net/e1000x_common.c | 5 ++--- hw/net/igb_core.c | 5 ++--- 4 files changed, 8 insertions(+), 12 deletions(-) diff --git a/hw/net/e1000.c b/hw/net/e1000.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000.c +++ b/hw/net/e1000.c @@ -XXX,XX +XXX,XX @@ xmit_seg(E1000State *s) e1000x_inc_reg_if_not_full(s->mac_reg, TPT); e1000x_grow_8reg_if_not_full(s->mac_reg, TOTL, s->tx.size + 4); - s->mac_reg[GPTC] = s->mac_reg[TPT]; - s->mac_reg[GOTCL] = s->mac_reg[TOTL]; - s->mac_reg[GOTCH] = s->mac_reg[TOTH]; + e1000x_inc_reg_if_not_full(s->mac_reg, GPTC); + e1000x_grow_8reg_if_not_full(s->mac_reg, GOTCL, s->tx.size + 4); } static void diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_on_tx_done_update_stats(E1000ECore *core, struct NetTxPkt *tx_pkt) g_assert_not_reached(); } - core->mac[GPTC] = core->mac[TPT]; - core->mac[GOTCL] = core->mac[TOTL]; - core->mac[GOTCH] = core->mac[TOTH]; + e1000x_inc_reg_if_not_full(core->mac, GPTC); + e1000x_grow_8reg_if_not_full(core->mac, GOTCL, tot_len); } static void diff --git a/hw/net/e1000x_common.c b/hw/net/e1000x_common.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_common.c +++ b/hw/net/e1000x_common.c @@ -XXX,XX +XXX,XX @@ e1000x_update_rx_total_stats(uint32_t *mac, e1000x_increase_size_stats(mac, PRCregs, data_fcs_size); e1000x_inc_reg_if_not_full(mac, TPR); - mac[GPRC] = mac[TPR]; + e1000x_inc_reg_if_not_full(mac, GPRC); /* TOR - Total Octets Received: * This register includes bytes received in a packet from the <Destination * Address> field through the <CRC> field, inclusively. * Always include FCS length (4) in size. */ e1000x_grow_8reg_if_not_full(mac, TORL, data_size + 4); - mac[GORCL] = mac[TORL]; - mac[GORCH] = mac[TORH]; + e1000x_grow_8reg_if_not_full(mac, GORCL, data_size + 4); } void diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_on_tx_done_update_stats(IGBCore *core, struct NetTxPkt *tx_pkt, int qn) g_assert_not_reached(); } - core->mac[GPTC] = core->mac[TPT]; - core->mac[GOTCL] = core->mac[TOTL]; - core->mac[GOTCH] = core->mac[TOTH]; + e1000x_inc_reg_if_not_full(core->mac, GPTC); + e1000x_grow_8reg_if_not_full(core->mac, GOTCL, tot_len); if (core->mac[MRQC] & 1) { uint16_t pool = qn % IGB_NUM_VM_POOLS; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> This is intended to be followed by another change for the interface. It also fixes the leak of memory mapping when the specified memory is partially mapped. Fixes: e263cd49c7 ("Packet abstraction for VMWARE network devices") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/net_tx_pkt.c | 53 +++++++++++++++++++++++++++++++++-------------------- hw/net/net_tx_pkt.h | 9 +++++++++ 2 files changed, 42 insertions(+), 20 deletions(-) diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.c +++ b/hw/net/net_tx_pkt.c @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_setup_vlan_header_ex(struct NetTxPkt *pkt, } } -bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa, - size_t len) +static bool net_tx_pkt_add_raw_fragment_common(struct NetTxPkt *pkt, + void *base, size_t len) { - hwaddr mapped_len = 0; struct iovec *ventry; assert(pkt); @@ -XXX,XX +XXX,XX @@ bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa, return false; } - if (!len) { - return true; - } - ventry = &pkt->raw[pkt->raw_frags]; - mapped_len = len; + ventry->iov_base = base; + ventry->iov_len = len; + pkt->raw_frags++; - ventry->iov_base = pci_dma_map(pkt->pci_dev, pa, - &mapped_len, DMA_DIRECTION_TO_DEVICE); - - if ((ventry->iov_base != NULL) && (len == mapped_len)) { - ventry->iov_len = mapped_len; - pkt->raw_frags++; - return true; - } else { - return false; - } + return true; } bool net_tx_pkt_has_fragments(struct NetTxPkt *pkt) @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *pci_dev) assert(pkt->raw); for (i = 0; i < pkt->raw_frags; i++) { assert(pkt->raw[i].iov_base); - pci_dma_unmap(pkt->pci_dev, pkt->raw[i].iov_base, - pkt->raw[i].iov_len, DMA_DIRECTION_TO_DEVICE, 0); + net_tx_pkt_unmap_frag_pci(pkt->pci_dev, + pkt->raw[i].iov_base, + pkt->raw[i].iov_len); } } pkt->pci_dev = pci_dev; @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *pci_dev) pkt->l4proto = 0; } +void net_tx_pkt_unmap_frag_pci(void *context, void *base, size_t len) +{ + pci_dma_unmap(context, base, len, DMA_DIRECTION_TO_DEVICE, 0); +} + +bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa, + size_t len) +{ + dma_addr_t mapped_len = len; + void *base = pci_dma_map(pkt->pci_dev, pa, &mapped_len, + DMA_DIRECTION_TO_DEVICE); + if (!base) { + return false; + } + + if (mapped_len != len || + !net_tx_pkt_add_raw_fragment_common(pkt, base, len)) { + net_tx_pkt_unmap_frag_pci(pkt->pci_dev, base, mapped_len); + return false; + } + + return true; +} + static void net_tx_pkt_do_sw_csum(struct NetTxPkt *pkt, struct iovec *iov, uint32_t iov_len, uint16_t csl) diff --git a/hw/net/net_tx_pkt.h b/hw/net/net_tx_pkt.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.h +++ b/hw/net/net_tx_pkt.h @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_dump(struct NetTxPkt *pkt); void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *dev); /** + * Unmap a fragment mapped from a PCI device. + * + * @context: PCI device owning fragment + * @base: pointer to fragment + * @len: length of fragment + */ +void net_tx_pkt_unmap_frag_pci(void *context, void *base, size_t len); + +/** * Send packet to qemu. handles sw offloads if vhdr is not supported. * * @pkt: packet -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> This allows to use the network packet abstractions even if PCI is not used. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 13 ++++++++----- hw/net/igb_core.c | 13 ++++++------- hw/net/net_tx_pkt.c | 36 +++++++++++++----------------------- hw/net/net_tx_pkt.h | 31 ++++++++++++++++++++----------- hw/net/vmxnet3.c | 14 +++++++------- 5 files changed, 54 insertions(+), 53 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_process_tx_desc(E1000ECore *core, addr = le64_to_cpu(dp->buffer_addr); if (!tx->skip_cp) { - if (!net_tx_pkt_add_raw_fragment(tx->tx_pkt, addr, split_size)) { + if (!net_tx_pkt_add_raw_fragment_pci(tx->tx_pkt, core->owner, + addr, split_size)) { tx->skip_cp = true; } } @@ -XXX,XX +XXX,XX @@ e1000e_process_tx_desc(E1000ECore *core, } tx->skip_cp = false; - net_tx_pkt_reset(tx->tx_pkt, core->owner); + net_tx_pkt_reset(tx->tx_pkt, net_tx_pkt_unmap_frag_pci, core->owner); tx->sum_needed = 0; tx->cptse = 0; @@ -XXX,XX +XXX,XX @@ e1000e_core_pci_realize(E1000ECore *core, qemu_add_vm_change_state_handler(e1000e_vm_state_change, core); for (i = 0; i < E1000E_NUM_QUEUES; i++) { - net_tx_pkt_init(&core->tx[i].tx_pkt, core->owner, E1000E_MAX_TX_FRAGS); + net_tx_pkt_init(&core->tx[i].tx_pkt, E1000E_MAX_TX_FRAGS); } net_rx_pkt_init(&core->rx_pkt); @@ -XXX,XX +XXX,XX @@ e1000e_core_pci_uninit(E1000ECore *core) qemu_del_vm_change_state_handler(core->vmstate); for (i = 0; i < E1000E_NUM_QUEUES; i++) { - net_tx_pkt_reset(core->tx[i].tx_pkt, core->owner); + net_tx_pkt_reset(core->tx[i].tx_pkt, + net_tx_pkt_unmap_frag_pci, core->owner); net_tx_pkt_uninit(core->tx[i].tx_pkt); } @@ -XXX,XX +XXX,XX @@ static void e1000e_reset(E1000ECore *core, bool sw) e1000x_reset_mac_addr(core->owner_nic, core->mac, core->permanent_mac); for (i = 0; i < ARRAY_SIZE(core->tx); i++) { - net_tx_pkt_reset(core->tx[i].tx_pkt, core->owner); + net_tx_pkt_reset(core->tx[i].tx_pkt, + net_tx_pkt_unmap_frag_pci, core->owner); memset(&core->tx[i].props, 0, sizeof(core->tx[i].props)); core->tx[i].skip_cp = false; } diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, length = cmd_type_len & 0xFFFF; if (!tx->skip_cp) { - if (!net_tx_pkt_add_raw_fragment(tx->tx_pkt, buffer_addr, length)) { + if (!net_tx_pkt_add_raw_fragment_pci(tx->tx_pkt, dev, + buffer_addr, length)) { tx->skip_cp = true; } } @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, tx->first = true; tx->skip_cp = false; - net_tx_pkt_reset(tx->tx_pkt, dev); + net_tx_pkt_reset(tx->tx_pkt, net_tx_pkt_unmap_frag_pci, dev); } } @@ -XXX,XX +XXX,XX @@ igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) d = core->owner; } - net_tx_pkt_reset(txr->tx->tx_pkt, d); - while (!igb_ring_empty(core, txi)) { base = igb_ring_head_descr(core, txi); @@ -XXX,XX +XXX,XX @@ igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) core->mac[EICR] |= eic; igb_set_interrupt_cause(core, E1000_ICR_TXDW); } + + net_tx_pkt_reset(txr->tx->tx_pkt, net_tx_pkt_unmap_frag_pci, d); } static uint32_t @@ -XXX,XX +XXX,XX @@ igb_core_pci_realize(IGBCore *core, core->vmstate = qemu_add_vm_change_state_handler(igb_vm_state_change, core); for (i = 0; i < IGB_NUM_QUEUES; i++) { - net_tx_pkt_init(&core->tx[i].tx_pkt, NULL, E1000E_MAX_TX_FRAGS); + net_tx_pkt_init(&core->tx[i].tx_pkt, E1000E_MAX_TX_FRAGS); } net_rx_pkt_init(&core->rx_pkt); @@ -XXX,XX +XXX,XX @@ igb_core_pci_uninit(IGBCore *core) qemu_del_vm_change_state_handler(core->vmstate); for (i = 0; i < IGB_NUM_QUEUES; i++) { - net_tx_pkt_reset(core->tx[i].tx_pkt, NULL); net_tx_pkt_uninit(core->tx[i].tx_pkt); } @@ -XXX,XX +XXX,XX @@ static void igb_reset(IGBCore *core, bool sw) for (i = 0; i < ARRAY_SIZE(core->tx); i++) { tx = &core->tx[i]; - net_tx_pkt_reset(tx->tx_pkt, NULL); memset(tx->ctx, 0, sizeof(tx->ctx)); tx->first = true; tx->skip_cp = false; diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.c +++ b/hw/net/net_tx_pkt.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" -#include "net_tx_pkt.h" #include "net/eth.h" #include "net/checksum.h" #include "net/tap.h" #include "net/net.h" #include "hw/pci/pci_device.h" +#include "net_tx_pkt.h" enum { NET_TX_PKT_VHDR_FRAG = 0, @@ -XXX,XX +XXX,XX @@ enum { /* TX packet private context */ struct NetTxPkt { - PCIDevice *pci_dev; - struct virtio_net_hdr virt_hdr; struct iovec *raw; @@ -XXX,XX +XXX,XX @@ struct NetTxPkt { uint8_t l4proto; }; -void net_tx_pkt_init(struct NetTxPkt **pkt, PCIDevice *pci_dev, - uint32_t max_frags) +void net_tx_pkt_init(struct NetTxPkt **pkt, uint32_t max_frags) { struct NetTxPkt *p = g_malloc0(sizeof *p); - p->pci_dev = pci_dev; - p->vec = g_new(struct iovec, max_frags + NET_TX_PKT_PL_START_FRAG); p->raw = g_new(struct iovec, max_frags); @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_setup_vlan_header_ex(struct NetTxPkt *pkt, } } -static bool net_tx_pkt_add_raw_fragment_common(struct NetTxPkt *pkt, - void *base, size_t len) +bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, void *base, size_t len) { struct iovec *ventry; assert(pkt); @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_dump(struct NetTxPkt *pkt) #endif } -void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *pci_dev) +void net_tx_pkt_reset(struct NetTxPkt *pkt, + NetTxPktFreeFrag callback, void *context) { int i; @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *pci_dev) assert(pkt->raw); for (i = 0; i < pkt->raw_frags; i++) { assert(pkt->raw[i].iov_base); - net_tx_pkt_unmap_frag_pci(pkt->pci_dev, - pkt->raw[i].iov_base, - pkt->raw[i].iov_len); + callback(context, pkt->raw[i].iov_base, pkt->raw[i].iov_len); } } - pkt->pci_dev = pci_dev; pkt->raw_frags = 0; pkt->hdr_len = 0; @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_unmap_frag_pci(void *context, void *base, size_t len) pci_dma_unmap(context, base, len, DMA_DIRECTION_TO_DEVICE, 0); } -bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa, - size_t len) +bool net_tx_pkt_add_raw_fragment_pci(struct NetTxPkt *pkt, PCIDevice *pci_dev, + dma_addr_t pa, size_t len) { dma_addr_t mapped_len = len; - void *base = pci_dma_map(pkt->pci_dev, pa, &mapped_len, - DMA_DIRECTION_TO_DEVICE); + void *base = pci_dma_map(pci_dev, pa, &mapped_len, DMA_DIRECTION_TO_DEVICE); if (!base) { return false; } - if (mapped_len != len || - !net_tx_pkt_add_raw_fragment_common(pkt, base, len)) { - net_tx_pkt_unmap_frag_pci(pkt->pci_dev, base, mapped_len); + if (mapped_len != len || !net_tx_pkt_add_raw_fragment(pkt, base, len)) { + net_tx_pkt_unmap_frag_pci(pci_dev, base, mapped_len); return false; } @@ -XXX,XX +XXX,XX @@ static void net_tx_pkt_udp_fragment_fix(struct NetTxPkt *pkt, } static bool net_tx_pkt_do_sw_fragmentation(struct NetTxPkt *pkt, - NetTxPktCallback callback, + NetTxPktSend callback, void *context) { uint8_t gso_type = pkt->virt_hdr.gso_type & ~VIRTIO_NET_HDR_GSO_ECN; @@ -XXX,XX +XXX,XX @@ bool net_tx_pkt_send(struct NetTxPkt *pkt, NetClientState *nc) } bool net_tx_pkt_send_custom(struct NetTxPkt *pkt, bool offload, - NetTxPktCallback callback, void *context) + NetTxPktSend callback, void *context) { assert(pkt); diff --git a/hw/net/net_tx_pkt.h b/hw/net/net_tx_pkt.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.h +++ b/hw/net/net_tx_pkt.h @@ -XXX,XX +XXX,XX @@ struct NetTxPkt; -typedef void (* NetTxPktCallback)(void *, const struct iovec *, int, const struct iovec *, int); +typedef void (*NetTxPktFreeFrag)(void *, void *, size_t); +typedef void (*NetTxPktSend)(void *, const struct iovec *, int, const struct iovec *, int); /** * Init function for tx packet functionality * * @pkt: packet pointer - * @pci_dev: PCI device processing this packet * @max_frags: max tx ip fragments */ -void net_tx_pkt_init(struct NetTxPkt **pkt, PCIDevice *pci_dev, - uint32_t max_frags); +void net_tx_pkt_init(struct NetTxPkt **pkt, uint32_t max_frags); /** * Clean all tx packet resources. @@ -XXX,XX +XXX,XX @@ net_tx_pkt_setup_vlan_header(struct NetTxPkt *pkt, uint16_t vlan) * populate data fragment into pkt context. * * @pkt: packet - * @pa: physical address of fragment + * @base: pointer to fragment * @len: length of fragment * */ -bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, hwaddr pa, - size_t len); +bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, void *base, size_t len); /** * Fix ip header fields and calculate IP header and pseudo header checksums. @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_dump(struct NetTxPkt *pkt); * reset tx packet private context (needed to be called between packets) * * @pkt: packet - * @dev: PCI device processing the next packet - * + * @callback: function to free the fragments + * @context: pointer to be passed to the callback */ -void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *dev); +void net_tx_pkt_reset(struct NetTxPkt *pkt, + NetTxPktFreeFrag callback, void *context); /** * Unmap a fragment mapped from a PCI device. @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *dev); void net_tx_pkt_unmap_frag_pci(void *context, void *base, size_t len); /** + * map data fragment from PCI device and populate it into pkt context. + * + * @pci_dev: PCI device owning fragment + * @pa: physical address of fragment + * @len: length of fragment + */ +bool net_tx_pkt_add_raw_fragment_pci(struct NetTxPkt *pkt, PCIDevice *pci_dev, + dma_addr_t pa, size_t len); + +/** * Send packet to qemu. handles sw offloads if vhdr is not supported. * * @pkt: packet @@ -XXX,XX +XXX,XX @@ bool net_tx_pkt_send(struct NetTxPkt *pkt, NetClientState *nc); * @ret: operation result */ bool net_tx_pkt_send_custom(struct NetTxPkt *pkt, bool offload, - NetTxPktCallback callback, void *context); + NetTxPktSend callback, void *context); /** * parse raw packet data and analyze offload requirements. diff --git a/hw/net/vmxnet3.c b/hw/net/vmxnet3.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/vmxnet3.c +++ b/hw/net/vmxnet3.c @@ -XXX,XX +XXX,XX @@ static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx) data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE; data_pa = txd.addr; - if (!net_tx_pkt_add_raw_fragment(s->tx_pkt, - data_pa, - data_len)) { + if (!net_tx_pkt_add_raw_fragment_pci(s->tx_pkt, PCI_DEVICE(s), + data_pa, data_len)) { s->skip_current_tx_pkt = true; } } @@ -XXX,XX +XXX,XX @@ static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx) vmxnet3_complete_packet(s, qidx, txd_idx); s->tx_sop = true; s->skip_current_tx_pkt = false; - net_tx_pkt_reset(s->tx_pkt, PCI_DEVICE(s)); + net_tx_pkt_reset(s->tx_pkt, + net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s)); } } } @@ -XXX,XX +XXX,XX @@ static void vmxnet3_deactivate_device(VMXNET3State *s) { if (s->device_active) { VMW_CBPRN("Deactivating vmxnet3..."); - net_tx_pkt_reset(s->tx_pkt, PCI_DEVICE(s)); + net_tx_pkt_reset(s->tx_pkt, net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s)); net_tx_pkt_uninit(s->tx_pkt); net_rx_pkt_uninit(s->rx_pkt); s->device_active = false; @@ -XXX,XX +XXX,XX @@ static void vmxnet3_activate_device(VMXNET3State *s) /* Preallocate TX packet wrapper */ VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags); - net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s), s->max_tx_frags); + net_tx_pkt_init(&s->tx_pkt, s->max_tx_frags); net_rx_pkt_init(&s->rx_pkt); /* Read rings memory locations for RX queues */ @@ -XXX,XX +XXX,XX @@ static int vmxnet3_post_load(void *opaque, int version_id) { VMXNET3State *s = opaque; - net_tx_pkt_init(&s->tx_pkt, PCI_DEVICE(s), s->max_tx_frags); + net_tx_pkt_init(&s->tx_pkt, s->max_tx_frags); net_rx_pkt_init(&s->rx_pkt); if (s->msix_used) { -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Before this change, e1000 and the common code updated BPRC and MPRC depending on the matched filter, but e1000e and igb decided to update those counters by deriving the packet type independently. This inconsistency caused a multicast packet to be counted twice. Updating BPRC and MPRC depending on are fundamentally flawed anyway as a filter can be used for different types of packets. For example, it is possible to filter broadcast packets with MTA. Always determine what counters to update by inspecting the packets. Fixes: 3b27430177 ("e1000: Implementing various counters") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000.c | 6 +++--- hw/net/e1000e_core.c | 20 +++----------------- hw/net/e1000x_common.c | 25 +++++++++++++++++++------ hw/net/e1000x_common.h | 5 +++-- hw/net/igb_core.c | 22 +++++----------------- 5 files changed, 33 insertions(+), 45 deletions(-) diff --git a/hw/net/e1000.c b/hw/net/e1000.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000.c +++ b/hw/net/e1000.c @@ -XXX,XX +XXX,XX @@ receive_filter(E1000State *s, const uint8_t *buf, int size) } if (ismcast && (rctl & E1000_RCTL_MPE)) { /* promiscuous mcast */ - e1000x_inc_reg_if_not_full(s->mac_reg, MPRC); return 1; } if (isbcast && (rctl & E1000_RCTL_BAM)) { /* broadcast enabled */ - e1000x_inc_reg_if_not_full(s->mac_reg, BPRC); return 1; } @@ -XXX,XX +XXX,XX @@ e1000_receive_iov(NetClientState *nc, const struct iovec *iov, int iovcnt) size_t desc_offset; size_t desc_size; size_t total_size; + eth_pkt_types_e pkt_type; if (!e1000x_hw_rx_enabled(s->mac_reg)) { return -1; @@ -XXX,XX +XXX,XX @@ e1000_receive_iov(NetClientState *nc, const struct iovec *iov, int iovcnt) size -= 4; } + pkt_type = get_eth_packet_type(PKT_GET_ETH_HDR(filter_buf)); rdh_start = s->mac_reg[RDH]; desc_offset = 0; total_size = size + e1000x_fcs_len(s->mac_reg); @@ -XXX,XX +XXX,XX @@ e1000_receive_iov(NetClientState *nc, const struct iovec *iov, int iovcnt) } } while (desc_offset < total_size); - e1000x_update_rx_total_stats(s->mac_reg, size, total_size); + e1000x_update_rx_total_stats(s->mac_reg, pkt_type, size, total_size); n = E1000_ICS_RXT0; if ((rdt = s->mac_reg[RDT]) < s->mac_reg[RDH]) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_write_to_rx_buffers(E1000ECore *core, } static void -e1000e_update_rx_stats(E1000ECore *core, - size_t data_size, - size_t data_fcs_size) +e1000e_update_rx_stats(E1000ECore *core, size_t pkt_size, size_t pkt_fcs_size) { - e1000x_update_rx_total_stats(core->mac, data_size, data_fcs_size); - - switch (net_rx_pkt_get_packet_type(core->rx_pkt)) { - case ETH_PKT_BCAST: - e1000x_inc_reg_if_not_full(core->mac, BPRC); - break; - - case ETH_PKT_MCAST: - e1000x_inc_reg_if_not_full(core->mac, MPRC); - break; - - default: - break; - } + eth_pkt_types_e pkt_type = net_rx_pkt_get_packet_type(core->rx_pkt); + e1000x_update_rx_total_stats(core->mac, pkt_type, pkt_size, pkt_fcs_size); } static inline bool diff --git a/hw/net/e1000x_common.c b/hw/net/e1000x_common.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_common.c +++ b/hw/net/e1000x_common.c @@ -XXX,XX +XXX,XX @@ bool e1000x_rx_group_filter(uint32_t *mac, const uint8_t *buf) f = mta_shift[(rctl >> E1000_RCTL_MO_SHIFT) & 3]; f = (((buf[5] << 8) | buf[4]) >> f) & 0xfff; if (mac[MTA + (f >> 5)] & (1 << (f & 0x1f))) { - e1000x_inc_reg_if_not_full(mac, MPRC); return true; } @@ -XXX,XX +XXX,XX @@ e1000x_rxbufsize(uint32_t rctl) void e1000x_update_rx_total_stats(uint32_t *mac, - size_t data_size, - size_t data_fcs_size) + eth_pkt_types_e pkt_type, + size_t pkt_size, + size_t pkt_fcs_size) { static const int PRCregs[6] = { PRC64, PRC127, PRC255, PRC511, PRC1023, PRC1522 }; - e1000x_increase_size_stats(mac, PRCregs, data_fcs_size); + e1000x_increase_size_stats(mac, PRCregs, pkt_fcs_size); e1000x_inc_reg_if_not_full(mac, TPR); e1000x_inc_reg_if_not_full(mac, GPRC); /* TOR - Total Octets Received: @@ -XXX,XX +XXX,XX @@ e1000x_update_rx_total_stats(uint32_t *mac, * Address> field through the <CRC> field, inclusively. * Always include FCS length (4) in size. */ - e1000x_grow_8reg_if_not_full(mac, TORL, data_size + 4); - e1000x_grow_8reg_if_not_full(mac, GORCL, data_size + 4); + e1000x_grow_8reg_if_not_full(mac, TORL, pkt_size + 4); + e1000x_grow_8reg_if_not_full(mac, GORCL, pkt_size + 4); + + switch (pkt_type) { + case ETH_PKT_BCAST: + e1000x_inc_reg_if_not_full(mac, BPRC); + break; + + case ETH_PKT_MCAST: + e1000x_inc_reg_if_not_full(mac, MPRC); + break; + + default: + break; + } } void diff --git a/hw/net/e1000x_common.h b/hw/net/e1000x_common.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_common.h +++ b/hw/net/e1000x_common.h @@ -XXX,XX +XXX,XX @@ e1000x_update_regs_on_link_up(uint32_t *mac, uint16_t *phy) } void e1000x_update_rx_total_stats(uint32_t *mac, - size_t data_size, - size_t data_fcs_size); + eth_pkt_types_e pkt_type, + size_t pkt_size, + size_t pkt_fcs_size); void e1000x_core_prepare_eeprom(uint16_t *eeprom, const uint16_t *templ, diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_write_to_rx_buffers(IGBCore *core, static void igb_update_rx_stats(IGBCore *core, const E1000E_RingInfo *rxi, - size_t data_size, size_t data_fcs_size) + size_t pkt_size, size_t pkt_fcs_size) { - e1000x_update_rx_total_stats(core->mac, data_size, data_fcs_size); - - switch (net_rx_pkt_get_packet_type(core->rx_pkt)) { - case ETH_PKT_BCAST: - e1000x_inc_reg_if_not_full(core->mac, BPRC); - break; - - case ETH_PKT_MCAST: - e1000x_inc_reg_if_not_full(core->mac, MPRC); - break; - - default: - break; - } + eth_pkt_types_e pkt_type = net_rx_pkt_get_packet_type(core->rx_pkt); + e1000x_update_rx_total_stats(core->mac, pkt_type, pkt_size, pkt_fcs_size); if (core->mac[MRQC] & 1) { uint16_t pool = rxi->idx % IGB_NUM_VM_POOLS; - core->mac[PVFGORC0 + (pool * 64)] += data_size + 4; + core->mac[PVFGORC0 + (pool * 64)] += pkt_size + 4; core->mac[PVFGPRC0 + (pool * 64)]++; - if (net_rx_pkt_get_packet_type(core->rx_pkt) == ETH_PKT_MCAST) { + if (pkt_type == ETH_PKT_MCAST) { core->mac[PVFMPRC0 + (pool * 64)]++; } } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> igb's advanced descriptor uses a packet type encoding different from one used in e1000e's extended descriptor. Fix the logic to encode Rx packet type accordingly. Fixes: 3a977deebe ("Intrdocue igb device emulation") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 38 +++++++++++++++++++------------------- hw/net/igb_regs.h | 5 +++++ 2 files changed, 24 insertions(+), 19 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, struct virtio_net_hdr *vhdr; bool hasip4, hasip6; EthL4HdrProto l4hdr_proto; - uint32_t pkt_type; *status_flags = E1000_RXD_STAT_DD; @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, trace_e1000e_rx_metadata_ack(); } - if (hasip6 && (core->mac[RFCTL] & E1000_RFCTL_IPV6_DIS)) { - trace_e1000e_rx_metadata_ipv6_filtering_disabled(); - pkt_type = E1000_RXD_PKT_MAC; - } else if (l4hdr_proto == ETH_L4_HDR_PROTO_TCP || - l4hdr_proto == ETH_L4_HDR_PROTO_UDP) { - pkt_type = hasip4 ? E1000_RXD_PKT_IP4_XDP : E1000_RXD_PKT_IP6_XDP; - } else if (hasip4 || hasip6) { - pkt_type = hasip4 ? E1000_RXD_PKT_IP4 : E1000_RXD_PKT_IP6; - } else { - pkt_type = E1000_RXD_PKT_MAC; - } + if (pkt_info) { + *pkt_info = rss_info->enabled ? rss_info->type : 0; - trace_e1000e_rx_metadata_pkt_type(pkt_type); + if (hasip4) { + *pkt_info |= E1000_ADVRXD_PKT_IP4; + } - if (pkt_info) { - if (rss_info->enabled) { - *pkt_info = rss_info->type; + if (hasip6) { + *pkt_info |= E1000_ADVRXD_PKT_IP6; } - *pkt_info |= (pkt_type << 4); - } else { - *status_flags |= E1000_RXD_PKT_TYPE(pkt_type); + switch (l4hdr_proto) { + case ETH_L4_HDR_PROTO_TCP: + *pkt_info |= E1000_ADVRXD_PKT_TCP; + break; + + case ETH_L4_HDR_PROTO_UDP: + *pkt_info |= E1000_ADVRXD_PKT_UDP; + break; + + default: + break; + } } if (hdr_info) { diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_regs.h +++ b/hw/net/igb_regs.h @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define E1000_STATUS_NUM_VFS_SHIFT 14 +#define E1000_ADVRXD_PKT_IP4 BIT(4) +#define E1000_ADVRXD_PKT_IP6 BIT(6) +#define E1000_ADVRXD_PKT_TCP BIT(8) +#define E1000_ADVRXD_PKT_UDP BIT(9) + static inline uint8_t igb_ivar_entry_rx(uint8_t i) { return i < 8 ? i * 4 : (i - 8) * 4 + 2; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> While the datasheet of e1000e says it checks CTRL.VME for tx VLAN tagging, igb's datasheet has no such statements. It also says for "CTRL.VLE": > This register only affects the VLAN Strip in Rx it does not have any > influence in the Tx path in the 82576. (Appendix A. Changes from the 82575) There is no "CTRL.VLE" so it is more likely that it is a mistake of CTRL.VME. Fixes: fba7c3b788 ("igb: respect VMVIR and VMOLR for VLAN") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_tx_insert_vlan(IGBCore *core, uint16_t qn, struct igb_tx *tx, } } - if (insert_vlan && e1000x_vlan_enabled(core->mac)) { + if (insert_vlan) { net_tx_pkt_setup_vlan_header_ex(tx->tx_pkt, vlan, core->mac[VET] & 0xffff); } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> The datasheet says contradicting statements regarding ICR accesses so it is not reliable to determine the behavior of ICR accesses. However, e1000e does clear IMS bits when reading ICR accesses and Linux also expects ICR accesses will clear IMS bits according to: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/net/ethernet/intel/igb/igb_main.c?h=v6.2#n8048 Fixes: 3a977deebe ("Intrdocue igb device emulation") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_set_ims(IGBCore *core, int index, uint32_t val) static void igb_commit_icr(IGBCore *core) { /* - * If GPIE.NSICR = 0, then the copy of IAM to IMS will occur only if at + * If GPIE.NSICR = 0, then the clear of IMS will occur only if at * least one bit is set in the IMS and there is a true interrupt as * reflected in ICR.INTA. */ if ((core->mac[GPIE] & E1000_GPIE_NSICR) || (core->mac[IMS] && (core->mac[ICR] & E1000_ICR_INT_ASSERTED))) { - igb_set_ims(core, IMS, core->mac[IAM]); - } else { - igb_update_interrupt_state(core); + igb_clear_ims_bits(core, core->mac[IAM]); } + + igb_update_interrupt_state(core); } static void igb_set_icr(IGBCore *core, int index, uint32_t val) -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> igb does not properly ensure the buffer passed to net_rx_pkt_set_protocols() is contiguous for the entire L2/L3/L4 header. Allow it to pass scattered data to net_rx_pkt_set_protocols(). Fixes: 3a977deebe ("Intrdocue igb device emulation") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 2 +- hw/net/net_rx_pkt.c | 14 +++++--------- hw/net/net_rx_pkt.h | 10 ++++++---- hw/net/virtio-net.c | 7 +++++-- hw/net/vmxnet3.c | 7 ++++++- include/net/eth.h | 6 +++--- net/eth.c | 18 ++++++++---------- 7 files changed, 34 insertions(+), 30 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, ehdr = PKT_GET_ETH_HDR(filter_buf); net_rx_pkt_set_packet_type(core->rx_pkt, get_eth_packet_type(ehdr)); - net_rx_pkt_set_protocols(core->rx_pkt, filter_buf, size); + net_rx_pkt_set_protocols(core->rx_pkt, iov, iovcnt, iov_ofs); queues = igb_receive_assign(core, ehdr, size, &rss_info, external_tx); if (!queues) { diff --git a/hw/net/net_rx_pkt.c b/hw/net/net_rx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.c +++ b/hw/net/net_rx_pkt.c @@ -XXX,XX +XXX,XX @@ net_rx_pkt_pull_data(struct NetRxPkt *pkt, iov, iovcnt, ploff, pkt->tot_len); } - eth_get_protocols(pkt->vec, pkt->vec_len, &pkt->hasip4, &pkt->hasip6, + eth_get_protocols(pkt->vec, pkt->vec_len, 0, &pkt->hasip4, &pkt->hasip6, &pkt->l3hdr_off, &pkt->l4hdr_off, &pkt->l5hdr_off, &pkt->ip6hdr_info, &pkt->ip4hdr_info, &pkt->l4hdr_info); @@ -XXX,XX +XXX,XX @@ size_t net_rx_pkt_get_total_len(struct NetRxPkt *pkt) return pkt->tot_len; } -void net_rx_pkt_set_protocols(struct NetRxPkt *pkt, const void *data, - size_t len) +void net_rx_pkt_set_protocols(struct NetRxPkt *pkt, + const struct iovec *iov, size_t iovcnt, + size_t iovoff) { - const struct iovec iov = { - .iov_base = (void *)data, - .iov_len = len - }; - assert(pkt); - eth_get_protocols(&iov, 1, &pkt->hasip4, &pkt->hasip6, + eth_get_protocols(iov, iovcnt, iovoff, &pkt->hasip4, &pkt->hasip6, &pkt->l3hdr_off, &pkt->l4hdr_off, &pkt->l5hdr_off, &pkt->ip6hdr_info, &pkt->ip4hdr_info, &pkt->l4hdr_info); } diff --git a/hw/net/net_rx_pkt.h b/hw/net/net_rx_pkt.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.h +++ b/hw/net/net_rx_pkt.h @@ -XXX,XX +XXX,XX @@ size_t net_rx_pkt_get_total_len(struct NetRxPkt *pkt); * parse and set packet analysis results * * @pkt: packet - * @data: pointer to the data buffer to be parsed - * @len: data length + * @iov: received data scatter-gather list + * @iovcnt: number of elements in iov + * @iovoff: data start offset in the iov * */ -void net_rx_pkt_set_protocols(struct NetRxPkt *pkt, const void *data, - size_t len); +void net_rx_pkt_set_protocols(struct NetRxPkt *pkt, + const struct iovec *iov, size_t iovcnt, + size_t iovoff); /** * fetches packet analysis results diff --git a/hw/net/virtio-net.c b/hw/net/virtio-net.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/virtio-net.c +++ b/hw/net/virtio-net.c @@ -XXX,XX +XXX,XX @@ static int virtio_net_process_rss(NetClientState *nc, const uint8_t *buf, VIRTIO_NET_HASH_REPORT_UDPv6, VIRTIO_NET_HASH_REPORT_UDPv6_EX }; + struct iovec iov = { + .iov_base = (void *)buf, + .iov_len = size + }; - net_rx_pkt_set_protocols(pkt, buf + n->host_hdr_len, - size - n->host_hdr_len); + net_rx_pkt_set_protocols(pkt, &iov, 1, n->host_hdr_len); net_rx_pkt_get_protocols(pkt, &hasip4, &hasip6, &l4hdr_proto); net_hash_type = virtio_net_get_hash_type(hasip4, hasip6, l4hdr_proto, n->rss_data.hash_types); diff --git a/hw/net/vmxnet3.c b/hw/net/vmxnet3.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/vmxnet3.c +++ b/hw/net/vmxnet3.c @@ -XXX,XX +XXX,XX @@ vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size) get_eth_packet_type(PKT_GET_ETH_HDR(buf))); if (vmxnet3_rx_filter_may_indicate(s, buf, size)) { - net_rx_pkt_set_protocols(s->rx_pkt, buf, size); + struct iovec iov = { + .iov_base = (void *)buf, + .iov_len = size + }; + + net_rx_pkt_set_protocols(s->rx_pkt, &iov, 1, 0); vmxnet3_rx_need_csum_calculate(s->rx_pkt, buf, size); net_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping); bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1; diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ eth_get_l2_hdr_length(const void *p) } static inline uint32_t -eth_get_l2_hdr_length_iov(const struct iovec *iov, int iovcnt) +eth_get_l2_hdr_length_iov(const struct iovec *iov, size_t iovcnt, size_t iovoff) { uint8_t p[sizeof(struct eth_header) + sizeof(struct vlan_header)]; - size_t copied = iov_to_buf(iov, iovcnt, 0, p, ARRAY_SIZE(p)); + size_t copied = iov_to_buf(iov, iovcnt, iovoff, p, ARRAY_SIZE(p)); if (copied < ARRAY_SIZE(p)) { return copied; @@ -XXX,XX +XXX,XX @@ typedef struct eth_l4_hdr_info_st { bool has_tcp_data; } eth_l4_hdr_info; -void eth_get_protocols(const struct iovec *iov, int iovcnt, +void eth_get_protocols(const struct iovec *iov, size_t iovcnt, size_t iovoff, bool *hasip4, bool *hasip6, size_t *l3hdr_off, size_t *l4hdr_off, diff --git a/net/eth.c b/net/eth.c index XXXXXXX..XXXXXXX 100644 --- a/net/eth.c +++ b/net/eth.c @@ -XXX,XX +XXX,XX @@ _eth_tcp_has_data(bool is_ip4, return l4len > TCP_HEADER_DATA_OFFSET(tcp); } -void eth_get_protocols(const struct iovec *iov, int iovcnt, +void eth_get_protocols(const struct iovec *iov, size_t iovcnt, size_t iovoff, bool *hasip4, bool *hasip6, size_t *l3hdr_off, size_t *l4hdr_off, @@ -XXX,XX +XXX,XX @@ void eth_get_protocols(const struct iovec *iov, int iovcnt, { int proto; bool fragment = false; - size_t l2hdr_len = eth_get_l2_hdr_length_iov(iov, iovcnt); size_t input_size = iov_size(iov, iovcnt); size_t copied; uint8_t ip_p; *hasip4 = *hasip6 = false; + *l3hdr_off = iovoff + eth_get_l2_hdr_length_iov(iov, iovcnt, iovoff); l4hdr_info->proto = ETH_L4_HDR_PROTO_INVALID; - proto = eth_get_l3_proto(iov, iovcnt, l2hdr_len); - - *l3hdr_off = l2hdr_len; + proto = eth_get_l3_proto(iov, iovcnt, *l3hdr_off); if (proto == ETH_P_IP) { struct ip_header *iphdr = &ip4hdr_info->ip4_hdr; - if (input_size < l2hdr_len) { + if (input_size < *l3hdr_off) { return; } - copied = iov_to_buf(iov, iovcnt, l2hdr_len, iphdr, sizeof(*iphdr)); + copied = iov_to_buf(iov, iovcnt, *l3hdr_off, iphdr, sizeof(*iphdr)); if (copied < sizeof(*iphdr) || IP_HEADER_VERSION(iphdr) != IP_HEADER_VERSION_4) { return; @@ -XXX,XX +XXX,XX @@ void eth_get_protocols(const struct iovec *iov, int iovcnt, *hasip4 = true; ip_p = iphdr->ip_p; ip4hdr_info->fragment = IP4_IS_FRAGMENT(iphdr); - *l4hdr_off = l2hdr_len + IP_HDR_GET_LEN(iphdr); + *l4hdr_off = *l3hdr_off + IP_HDR_GET_LEN(iphdr); fragment = ip4hdr_info->fragment; } else if (proto == ETH_P_IPV6) { - if (!eth_parse_ipv6_hdr(iov, iovcnt, l2hdr_len, ip6hdr_info)) { + if (!eth_parse_ipv6_hdr(iov, iovcnt, *l3hdr_off, ip6hdr_info)) { return; } *hasip6 = true; ip_p = ip6hdr_info->l4proto; - *l4hdr_off = l2hdr_len + ip6hdr_info->full_hdr_len; + *l4hdr_off = *l3hdr_off + ip6hdr_info->full_hdr_len; fragment = ip6hdr_info->fragment; } else { return; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> e1000e_receive_internal() used to check the iov length to determine copy the iovs to a contiguous buffer, but the check is flawed in two ways: - It does not ensure that iovcnt > 0. - It does not take virtio-net header into consideration. The size of this copy is just 18 octets, which can be even less than the code size required for checks. This (wrong) optimization is probably not worth so just remove it. Fixes: 6f3fbe4ed0 ("net: Introduce e1000e device emulation") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 26 ++++++++++---------------- 1 file changed, 10 insertions(+), 16 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ static ssize_t e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, bool has_vnet) { - static const int maximum_ethernet_hdr_len = (ETH_HLEN + 4); - uint32_t n = 0; - uint8_t min_buf[ETH_ZLEN]; + uint8_t buf[ETH_ZLEN]; struct iovec min_iov; - uint8_t *filter_buf; size_t size, orig_size; size_t iov_ofs = 0; E1000E_RxRing rxr; @@ -XXX,XX +XXX,XX @@ e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, net_rx_pkt_unset_vhdr(core->rx_pkt); } - filter_buf = iov->iov_base + iov_ofs; orig_size = iov_size(iov, iovcnt); size = orig_size - iov_ofs; /* Pad to minimum Ethernet frame length */ - if (size < sizeof(min_buf)) { - iov_to_buf(iov, iovcnt, iov_ofs, min_buf, size); - memset(&min_buf[size], 0, sizeof(min_buf) - size); + if (size < sizeof(buf)) { + iov_to_buf(iov, iovcnt, iov_ofs, buf, size); + memset(&buf[size], 0, sizeof(buf) - size); e1000x_inc_reg_if_not_full(core->mac, RUC); - min_iov.iov_base = filter_buf = min_buf; - min_iov.iov_len = size = sizeof(min_buf); + min_iov.iov_base = buf; + min_iov.iov_len = size = sizeof(buf); iovcnt = 1; iov = &min_iov; iov_ofs = 0; - } else if (iov->iov_len < maximum_ethernet_hdr_len) { - /* This is very unlikely, but may happen. */ - iov_to_buf(iov, iovcnt, iov_ofs, min_buf, maximum_ethernet_hdr_len); - filter_buf = min_buf; + } else { + iov_to_buf(iov, iovcnt, iov_ofs, buf, ETH_HLEN + 4); } /* Discard oversized packets if !LPE and !SBP. */ @@ -XXX,XX +XXX,XX @@ e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, } net_rx_pkt_set_packet_type(core->rx_pkt, - get_eth_packet_type(PKT_GET_ETH_HDR(filter_buf))); + get_eth_packet_type(PKT_GET_ETH_HDR(buf))); - if (!e1000e_receive_filter(core, filter_buf, size)) { + if (!e1000e_receive_filter(core, buf, size)) { trace_e1000e_rx_flt_dropped(); return orig_size; } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> igb_receive_internal() used to check the iov length to determine copy the iovs to a contiguous buffer, but the check is flawed in two ways: - It does not ensure that iovcnt > 0. - It does not take virtio-net header into consideration. The size of this copy is just 22 octets, which can be even less than the code size required for checks. This (wrong) optimization is probably not worth so just remove it. Removing this also allows igb to assume aligned accesses for the ethernet header. Fixes: 3a977deebe ("Intrdocue igb device emulation") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 43 +++++++++++++++++++++++-------------------- 1 file changed, 23 insertions(+), 20 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ typedef struct IGBTxPktVmdqCallbackContext { NetClientState *nc; } IGBTxPktVmdqCallbackContext; +typedef struct L2Header { + struct eth_header eth; + struct vlan_header vlan; +} L2Header; + static ssize_t igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, bool has_vnet, bool *external_tx); @@ -XXX,XX +XXX,XX @@ igb_rx_is_oversized(IGBCore *core, uint16_t qn, size_t size) return size > (lpe ? max_ethernet_lpe_size : max_ethernet_vlan_size); } -static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, +static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, size_t size, E1000E_RSSInfo *rss_info, bool *external_tx) { static const int ta_shift[] = { 4, 3, 2, 0 }; + const struct eth_header *ehdr = &l2_header->eth; uint32_t f, ra[2], *macp, rctl = core->mac[RCTL]; uint16_t queues = 0; uint16_t oversized = 0; - uint16_t vid = lduw_be_p(&PKT_GET_VLAN_HDR(ehdr)->h_tci) & VLAN_VID_MASK; + uint16_t vid = be16_to_cpu(l2_header->vlan.h_tci) & VLAN_VID_MASK; bool accepted = false; int i; @@ -XXX,XX +XXX,XX @@ static ssize_t igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, bool has_vnet, bool *external_tx) { - static const int maximum_ethernet_hdr_len = (ETH_HLEN + 4); - uint16_t queues = 0; uint32_t n = 0; - uint8_t min_buf[ETH_ZLEN]; + union { + L2Header l2_header; + uint8_t octets[ETH_ZLEN]; + } buf; struct iovec min_iov; - struct eth_header *ehdr; - uint8_t *filter_buf; size_t size, orig_size; size_t iov_ofs = 0; E1000E_RxRing rxr; @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, net_rx_pkt_unset_vhdr(core->rx_pkt); } - filter_buf = iov->iov_base + iov_ofs; orig_size = iov_size(iov, iovcnt); size = orig_size - iov_ofs; /* Pad to minimum Ethernet frame length */ - if (size < sizeof(min_buf)) { - iov_to_buf(iov, iovcnt, iov_ofs, min_buf, size); - memset(&min_buf[size], 0, sizeof(min_buf) - size); + if (size < sizeof(buf)) { + iov_to_buf(iov, iovcnt, iov_ofs, &buf, size); + memset(&buf.octets[size], 0, sizeof(buf) - size); e1000x_inc_reg_if_not_full(core->mac, RUC); - min_iov.iov_base = filter_buf = min_buf; - min_iov.iov_len = size = sizeof(min_buf); + min_iov.iov_base = &buf; + min_iov.iov_len = size = sizeof(buf); iovcnt = 1; iov = &min_iov; iov_ofs = 0; - } else if (iov->iov_len < maximum_ethernet_hdr_len) { - /* This is very unlikely, but may happen. */ - iov_to_buf(iov, iovcnt, iov_ofs, min_buf, maximum_ethernet_hdr_len); - filter_buf = min_buf; + } else { + iov_to_buf(iov, iovcnt, iov_ofs, &buf, sizeof(buf.l2_header)); } /* Discard oversized packets if !LPE and !SBP. */ @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, return orig_size; } - ehdr = PKT_GET_ETH_HDR(filter_buf); - net_rx_pkt_set_packet_type(core->rx_pkt, get_eth_packet_type(ehdr)); + net_rx_pkt_set_packet_type(core->rx_pkt, + get_eth_packet_type(&buf.l2_header.eth)); net_rx_pkt_set_protocols(core->rx_pkt, iov, iovcnt, iov_ofs); - queues = igb_receive_assign(core, ehdr, size, &rss_info, external_tx); + queues = igb_receive_assign(core, &buf.l2_header, size, + &rss_info, external_tx); if (!queues) { trace_e1000e_rx_flt_dropped(); return orig_size; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Fixes: 9f95111474 ("tests/avocado: re-factor igb test to avoid timeouts") Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- MAINTAINERS | 2 +- docs/system/devices/igb.rst | 2 +- scripts/ci/org.centos/stream/8/x86_64/test-avocado | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ R: Sriram Yagnaraman <sriram.yagnaraman@est.tech> S: Maintained F: docs/system/devices/igb.rst F: hw/net/igb* -F: tests/avocado/igb.py +F: tests/avocado/netdev-ethtool.py F: tests/qtest/igb-test.c F: tests/qtest/libqos/igb.c diff --git a/docs/system/devices/igb.rst b/docs/system/devices/igb.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/system/devices/igb.rst +++ b/docs/system/devices/igb.rst @@ -XXX,XX +XXX,XX @@ Avocado test and can be ran with the following command: .. code:: shell - make check-avocado AVOCADO_TESTS=tests/avocado/igb.py + make check-avocado AVOCADO_TESTS=tests/avocado/netdev-ethtool.py References ========== diff --git a/scripts/ci/org.centos/stream/8/x86_64/test-avocado b/scripts/ci/org.centos/stream/8/x86_64/test-avocado index XXXXXXX..XXXXXXX 100755 --- a/scripts/ci/org.centos/stream/8/x86_64/test-avocado +++ b/scripts/ci/org.centos/stream/8/x86_64/test-avocado @@ -XXX,XX +XXX,XX @@ make get-vm-images tests/avocado/cpu_queries.py:QueryCPUModelExpansion.test \ tests/avocado/empty_cpu_model.py:EmptyCPUModel.test \ tests/avocado/hotplug_cpu.py:HotPlugCPU.test \ - tests/avocado/igb.py:IGB.test \ + tests/avocado/netdev-ethtool.py:NetDevEthtool.test_igb_nomsi \ tests/avocado/info_usernet.py:InfoUsernet.test_hostfwd \ tests/avocado/intel_iommu.py:IntelIOMMU.test_intel_iommu \ tests/avocado/intel_iommu.py:IntelIOMMU.test_intel_iommu_pt \ -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- tests/avocado/netdev-ethtool.py | 1 - 1 file changed, 1 deletion(-) diff --git a/tests/avocado/netdev-ethtool.py b/tests/avocado/netdev-ethtool.py index XXXXXXX..XXXXXXX 100644 --- a/tests/avocado/netdev-ethtool.py +++ b/tests/avocado/netdev-ethtool.py @@ -XXX,XX +XXX,XX @@ from avocado import skip from avocado_qemu import QemuSystemTest -from avocado_qemu import exec_command, exec_command_and_wait_for_pattern from avocado_qemu import wait_for_console_pattern class NetDevEthtool(QemuSystemTest): -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> It is unlikely to find more bugs with KVM so remove test_igb_nomsi_kvm to save time to run it. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Thomas Huth <thuth@redhat.com> Acked-by: Alex Bennée <alex.bennee@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- tests/avocado/netdev-ethtool.py | 12 +----------- 1 file changed, 1 insertion(+), 11 deletions(-) diff --git a/tests/avocado/netdev-ethtool.py b/tests/avocado/netdev-ethtool.py index XXXXXXX..XXXXXXX 100644 --- a/tests/avocado/netdev-ethtool.py +++ b/tests/avocado/netdev-ethtool.py @@ -XXX,XX +XXX,XX @@ def get_asset(self, name, sha1): # URL into a unique one return self.fetch_asset(name=name, locations=(url), asset_hash=sha1) - def common_test_code(self, netdev, extra_args=None, kvm=False): + def common_test_code(self, netdev, extra_args=None): # This custom kernel has drivers for all the supported network # devices we can emulate in QEMU @@ -XXX,XX +XXX,XX @@ def common_test_code(self, netdev, extra_args=None, kvm=False): '-drive', drive, '-device', netdev) - if kvm: - self.vm.add_args('-accel', 'kvm') - self.vm.set_console(console_index=0) self.vm.launch() @@ -XXX,XX +XXX,XX @@ def test_igb_nomsi(self): """ self.common_test_code("igb", "pci=nomsi") - def test_igb_nomsi_kvm(self): - """ - :avocado: tags=device:igb - """ - self.require_accelerator('kvm') - self.common_test_code("igb", "pci=nomsi", True) - # It seems the other popular cards we model in QEMU currently fail # the pattern test with: # -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> This function is not used. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/net_rx_pkt.c | 5 ----- hw/net/net_rx_pkt.h | 9 --------- 2 files changed, 14 deletions(-) diff --git a/hw/net/net_rx_pkt.c b/hw/net/net_rx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.c +++ b/hw/net/net_rx_pkt.c @@ -XXX,XX +XXX,XX @@ eth_ip4_hdr_info *net_rx_pkt_get_ip4_info(struct NetRxPkt *pkt) return &pkt->ip4hdr_info; } -eth_l4_hdr_info *net_rx_pkt_get_l4_info(struct NetRxPkt *pkt) -{ - return &pkt->l4hdr_info; -} - static inline void _net_rx_rss_add_chunk(uint8_t *rss_input, size_t *bytes_written, void *ptr, size_t size) diff --git a/hw/net/net_rx_pkt.h b/hw/net/net_rx_pkt.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.h +++ b/hw/net/net_rx_pkt.h @@ -XXX,XX +XXX,XX @@ eth_ip6_hdr_info *net_rx_pkt_get_ip6_info(struct NetRxPkt *pkt); */ eth_ip4_hdr_info *net_rx_pkt_get_ip4_info(struct NetRxPkt *pkt); -/** - * fetches L4 header analysis results - * - * Return: pointer to analysis results structure which is stored in internal - * packet area. - * - */ -eth_l4_hdr_info *net_rx_pkt_get_l4_info(struct NetRxPkt *pkt); - typedef enum { NetPktRssIpV4, NetPktRssIpV4Tcp, -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> The old eth_setup_vlan_headers has no user so remove it and rename eth_setup_vlan_headers_ex. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/net_tx_pkt.c | 2 +- include/net/eth.h | 9 +-------- net/eth.c | 2 +- 3 files changed, 3 insertions(+), 10 deletions(-) diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.c +++ b/hw/net/net_tx_pkt.c @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_setup_vlan_header_ex(struct NetTxPkt *pkt, bool is_new; assert(pkt); - eth_setup_vlan_headers_ex(pkt->vec[NET_TX_PKT_L2HDR_FRAG].iov_base, + eth_setup_vlan_headers(pkt->vec[NET_TX_PKT_L2HDR_FRAG].iov_base, vlan, vlan_ethtype, &is_new); /* update l2hdrlen */ diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, uint16_t eth_get_l3_proto(const struct iovec *l2hdr_iov, int iovcnt, size_t l2hdr_len); -void eth_setup_vlan_headers_ex(struct eth_header *ehdr, uint16_t vlan_tag, +void eth_setup_vlan_headers(struct eth_header *ehdr, uint16_t vlan_tag, uint16_t vlan_ethtype, bool *is_new); -static inline void -eth_setup_vlan_headers(struct eth_header *ehdr, uint16_t vlan_tag, - bool *is_new) -{ - eth_setup_vlan_headers_ex(ehdr, vlan_tag, ETH_P_VLAN, is_new); -} - uint8_t eth_get_gso_type(uint16_t l3_proto, uint8_t *l3_hdr, uint8_t l4proto); diff --git a/net/eth.c b/net/eth.c index XXXXXXX..XXXXXXX 100644 --- a/net/eth.c +++ b/net/eth.c @@ -XXX,XX +XXX,XX @@ #include "net/checksum.h" #include "net/tap.h" -void eth_setup_vlan_headers_ex(struct eth_header *ehdr, uint16_t vlan_tag, +void eth_setup_vlan_headers(struct eth_header *ehdr, uint16_t vlan_tag, uint16_t vlan_ethtype, bool *is_new) { struct vlan_header *vhdr = PKT_GET_VLAN_HDR(ehdr); -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> This saves some code and enables tracepoint for e1000's VLAN filtering. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000.c | 35 +++++------------------------------ hw/net/e1000e_core.c | 47 +++++------------------------------------------ hw/net/e1000x_common.c | 44 ++++++++++++++++++++++++++++++++++++++------ hw/net/e1000x_common.h | 4 +++- hw/net/igb_core.c | 41 +++-------------------------------------- hw/net/trace-events | 4 ++-- 6 files changed, 56 insertions(+), 119 deletions(-) diff --git a/hw/net/e1000.c b/hw/net/e1000.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000.c +++ b/hw/net/e1000.c @@ -XXX,XX +XXX,XX @@ start_xmit(E1000State *s) } static int -receive_filter(E1000State *s, const uint8_t *buf, int size) +receive_filter(E1000State *s, const void *buf) { - uint32_t rctl = s->mac_reg[RCTL]; - int isbcast = is_broadcast_ether_addr(buf); - int ismcast = is_multicast_ether_addr(buf); - - if (e1000x_is_vlan_packet(buf, le16_to_cpu(s->mac_reg[VET])) && - e1000x_vlan_rx_filter_enabled(s->mac_reg)) { - uint16_t vid = lduw_be_p(&PKT_GET_VLAN_HDR(buf)->h_tci); - uint32_t vfta = - ldl_le_p((uint32_t *)(s->mac_reg + VFTA) + - ((vid >> E1000_VFTA_ENTRY_SHIFT) & E1000_VFTA_ENTRY_MASK)); - if ((vfta & (1 << (vid & E1000_VFTA_ENTRY_BIT_SHIFT_MASK))) == 0) { - return 0; - } - } - - if (!isbcast && !ismcast && (rctl & E1000_RCTL_UPE)) { /* promiscuous ucast */ - return 1; - } - - if (ismcast && (rctl & E1000_RCTL_MPE)) { /* promiscuous mcast */ - return 1; - } - - if (isbcast && (rctl & E1000_RCTL_BAM)) { /* broadcast enabled */ - return 1; - } - - return e1000x_rx_group_filter(s->mac_reg, buf); + return (!e1000x_is_vlan_packet(buf, s->mac_reg[VET]) || + e1000x_rx_vlan_filter(s->mac_reg, PKT_GET_VLAN_HDR(buf))) && + e1000x_rx_group_filter(s->mac_reg, buf); } static void @@ -XXX,XX +XXX,XX @@ e1000_receive_iov(NetClientState *nc, const struct iovec *iov, int iovcnt) return size; } - if (!receive_filter(s, filter_buf, size)) { + if (!receive_filter(s, filter_buf)) { return size; } diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_rx_l4_cso_enabled(E1000ECore *core) } static bool -e1000e_receive_filter(E1000ECore *core, const uint8_t *buf, int size) +e1000e_receive_filter(E1000ECore *core, const void *buf) { - uint32_t rctl = core->mac[RCTL]; - - if (e1000x_is_vlan_packet(buf, core->mac[VET]) && - e1000x_vlan_rx_filter_enabled(core->mac)) { - uint16_t vid = lduw_be_p(&PKT_GET_VLAN_HDR(buf)->h_tci); - uint32_t vfta = - ldl_le_p((uint32_t *)(core->mac + VFTA) + - ((vid >> E1000_VFTA_ENTRY_SHIFT) & E1000_VFTA_ENTRY_MASK)); - if ((vfta & (1 << (vid & E1000_VFTA_ENTRY_BIT_SHIFT_MASK))) == 0) { - trace_e1000e_rx_flt_vlan_mismatch(vid); - return false; - } else { - trace_e1000e_rx_flt_vlan_match(vid); - } - } - - switch (net_rx_pkt_get_packet_type(core->rx_pkt)) { - case ETH_PKT_UCAST: - if (rctl & E1000_RCTL_UPE) { - return true; /* promiscuous ucast */ - } - break; - - case ETH_PKT_BCAST: - if (rctl & E1000_RCTL_BAM) { - return true; /* broadcast enabled */ - } - break; - - case ETH_PKT_MCAST: - if (rctl & E1000_RCTL_MPE) { - return true; /* promiscuous mcast */ - } - break; - - default: - g_assert_not_reached(); - } - - return e1000x_rx_group_filter(core->mac, buf); + return (!e1000x_is_vlan_packet(buf, core->mac[VET]) || + e1000x_rx_vlan_filter(core->mac, PKT_GET_VLAN_HDR(buf))) && + e1000x_rx_group_filter(core->mac, buf); } static inline void @@ -XXX,XX +XXX,XX @@ e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, net_rx_pkt_set_packet_type(core->rx_pkt, get_eth_packet_type(PKT_GET_ETH_HDR(buf))); - if (!e1000e_receive_filter(core, buf, size)) { + if (!e1000e_receive_filter(core, buf)) { trace_e1000e_rx_flt_dropped(); return orig_size; } diff --git a/hw/net/e1000x_common.c b/hw/net/e1000x_common.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_common.c +++ b/hw/net/e1000x_common.c @@ -XXX,XX +XXX,XX @@ bool e1000x_is_vlan_packet(const void *buf, uint16_t vet) return res; } -bool e1000x_rx_group_filter(uint32_t *mac, const uint8_t *buf) +bool e1000x_rx_vlan_filter(uint32_t *mac, const struct vlan_header *vhdr) +{ + if (e1000x_vlan_rx_filter_enabled(mac)) { + uint16_t vid = lduw_be_p(&vhdr->h_tci); + uint32_t vfta = + ldl_le_p((uint32_t *)(mac + VFTA) + + ((vid >> E1000_VFTA_ENTRY_SHIFT) & E1000_VFTA_ENTRY_MASK)); + if ((vfta & (1 << (vid & E1000_VFTA_ENTRY_BIT_SHIFT_MASK))) == 0) { + trace_e1000x_rx_flt_vlan_mismatch(vid); + return false; + } + + trace_e1000x_rx_flt_vlan_match(vid); + } + + return true; +} + +bool e1000x_rx_group_filter(uint32_t *mac, const struct eth_header *ehdr) { static const int mta_shift[] = { 4, 3, 2, 0 }; uint32_t f, ra[2], *rp, rctl = mac[RCTL]; + if (is_broadcast_ether_addr(ehdr->h_dest)) { + if (rctl & E1000_RCTL_BAM) { + return true; + } + } else if (is_multicast_ether_addr(ehdr->h_dest)) { + if (rctl & E1000_RCTL_MPE) { + return true; + } + } else { + if (rctl & E1000_RCTL_UPE) { + return true; + } + } + for (rp = mac + RA; rp < mac + RA + 32; rp += 2) { if (!(rp[1] & E1000_RAH_AV)) { continue; } ra[0] = cpu_to_le32(rp[0]); ra[1] = cpu_to_le32(rp[1]); - if (!memcmp(buf, (uint8_t *)ra, ETH_ALEN)) { + if (!memcmp(ehdr->h_dest, (uint8_t *)ra, ETH_ALEN)) { trace_e1000x_rx_flt_ucast_match((int)(rp - mac - RA) / 2, - MAC_ARG(buf)); + MAC_ARG(ehdr->h_dest)); return true; } } - trace_e1000x_rx_flt_ucast_mismatch(MAC_ARG(buf)); + trace_e1000x_rx_flt_ucast_mismatch(MAC_ARG(ehdr->h_dest)); f = mta_shift[(rctl >> E1000_RCTL_MO_SHIFT) & 3]; - f = (((buf[5] << 8) | buf[4]) >> f) & 0xfff; + f = (((ehdr->h_dest[5] << 8) | ehdr->h_dest[4]) >> f) & 0xfff; if (mac[MTA + (f >> 5)] & (1 << (f & 0x1f))) { return true; } - trace_e1000x_rx_flt_inexact_mismatch(MAC_ARG(buf), + trace_e1000x_rx_flt_inexact_mismatch(MAC_ARG(ehdr->h_dest), (rctl >> E1000_RCTL_MO_SHIFT) & 3, f >> 5, mac[MTA + (f >> 5)]); diff --git a/hw/net/e1000x_common.h b/hw/net/e1000x_common.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_common.h +++ b/hw/net/e1000x_common.h @@ -XXX,XX +XXX,XX @@ bool e1000x_rx_ready(PCIDevice *d, uint32_t *mac); bool e1000x_is_vlan_packet(const void *buf, uint16_t vet); -bool e1000x_rx_group_filter(uint32_t *mac, const uint8_t *buf); +bool e1000x_rx_vlan_filter(uint32_t *mac, const struct vlan_header *vhdr); + +bool e1000x_rx_group_filter(uint32_t *mac, const struct eth_header *ehdr); bool e1000x_hw_rx_enabled(uint32_t *mac); diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, uint16_t queues = 0; uint16_t oversized = 0; uint16_t vid = be16_to_cpu(l2_header->vlan.h_tci) & VLAN_VID_MASK; - bool accepted = false; int i; memset(rss_info, 0, sizeof(E1000E_RSSInfo)); @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, } if (e1000x_is_vlan_packet(ehdr, core->mac[VET] & 0xffff) && - e1000x_vlan_rx_filter_enabled(core->mac)) { - uint32_t vfta = - ldl_le_p((uint32_t *)(core->mac + VFTA) + - ((vid >> E1000_VFTA_ENTRY_SHIFT) & E1000_VFTA_ENTRY_MASK)); - if ((vfta & (1 << (vid & E1000_VFTA_ENTRY_BIT_SHIFT_MASK))) == 0) { - trace_e1000e_rx_flt_vlan_mismatch(vid); - return queues; - } else { - trace_e1000e_rx_flt_vlan_match(vid); - } + !e1000x_rx_vlan_filter(core->mac, PKT_GET_VLAN_HDR(ehdr))) { + return queues; } if (core->mac[MRQC] & 1) { @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, } } } else { - switch (net_rx_pkt_get_packet_type(core->rx_pkt)) { - case ETH_PKT_UCAST: - if (rctl & E1000_RCTL_UPE) { - accepted = true; /* promiscuous ucast */ - } - break; - - case ETH_PKT_BCAST: - if (rctl & E1000_RCTL_BAM) { - accepted = true; /* broadcast enabled */ - } - break; - - case ETH_PKT_MCAST: - if (rctl & E1000_RCTL_MPE) { - accepted = true; /* promiscuous mcast */ - } - break; - - default: - g_assert_not_reached(); - } - - if (!accepted) { - accepted = e1000x_rx_group_filter(core->mac, ehdr->h_dest); - } - + bool accepted = e1000x_rx_group_filter(core->mac, ehdr); if (!accepted) { for (macp = core->mac + RA2; macp < core->mac + RA2 + 16; macp += 2) { if (!(macp[1] & E1000_RAH_AV)) { diff --git a/hw/net/trace-events b/hw/net/trace-events index XXXXXXX..XXXXXXX 100644 --- a/hw/net/trace-events +++ b/hw/net/trace-events @@ -XXX,XX +XXX,XX @@ e1000_receiver_overrun(size_t s, uint32_t rdh, uint32_t rdt) "Receiver overrun: # e1000x_common.c e1000x_rx_can_recv_disabled(bool link_up, bool rx_enabled, bool pci_master) "link_up: %d, rx_enabled %d, pci_master %d" e1000x_vlan_is_vlan_pkt(bool is_vlan_pkt, uint16_t eth_proto, uint16_t vet) "Is VLAN packet: %d, ETH proto: 0x%X, VET: 0x%X" +e1000x_rx_flt_vlan_mismatch(uint16_t vid) "VID mismatch: 0x%X" +e1000x_rx_flt_vlan_match(uint16_t vid) "VID match: 0x%X" e1000x_rx_flt_ucast_match(uint32_t idx, uint8_t b0, uint8_t b1, uint8_t b2, uint8_t b3, uint8_t b4, uint8_t b5) "unicast match[%d]: %02x:%02x:%02x:%02x:%02x:%02x" e1000x_rx_flt_ucast_mismatch(uint8_t b0, uint8_t b1, uint8_t b2, uint8_t b3, uint8_t b4, uint8_t b5) "unicast mismatch: %02x:%02x:%02x:%02x:%02x:%02x" e1000x_rx_flt_inexact_mismatch(uint8_t b0, uint8_t b1, uint8_t b2, uint8_t b3, uint8_t b4, uint8_t b5, uint32_t mo, uint32_t mta, uint32_t mta_val) "inexact mismatch: %02x:%02x:%02x:%02x:%02x:%02x MO %d MTA[%d] 0x%x" @@ -XXX,XX +XXX,XX @@ e1000e_rx_can_recv_rings_full(void) "Cannot receive: all rings are full" e1000e_rx_can_recv(void) "Can receive" e1000e_rx_has_buffers(int ridx, uint32_t free_desc, size_t total_size, uint32_t desc_buf_size) "ring #%d: free descr: %u, packet size %zu, descr buffer size %u" e1000e_rx_null_descriptor(void) "Null RX descriptor!!" -e1000e_rx_flt_vlan_mismatch(uint16_t vid) "VID mismatch: 0x%X" -e1000e_rx_flt_vlan_match(uint16_t vid) "VID match: 0x%X" e1000e_rx_desc_ps_read(uint64_t a0, uint64_t a1, uint64_t a2, uint64_t a3) "buffers: [0x%"PRIx64", 0x%"PRIx64", 0x%"PRIx64", 0x%"PRIx64"]" e1000e_rx_desc_ps_write(uint16_t a0, uint16_t a1, uint16_t a2, uint16_t a3) "bytes written: [%u, %u, %u, %u]" e1000e_rx_desc_buff_sizes(uint32_t b0, uint32_t b1, uint32_t b2, uint32_t b3) "buffer sizes: [%u, %u, %u, %u]" -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Section 13.7.15 Receive Length Error Count says: > Packets over 1522 bytes are oversized if LongPacketEnable is 0b > (RCTL.LPE). If LongPacketEnable (LPE) is 1b, then an incoming packet > is considered oversized if it exceeds 16384 bytes. > These lengths are based on bytes in the received packet from > <Destination Address> through <CRC>, inclusively. As QEMU processes packets without CRC, the number of bytes for CRC need to be subtracted. This change adds some size definitions to be used to derive the new size thresholds to eth.h. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000x_common.c | 10 +++++----- include/net/eth.h | 2 ++ 2 files changed, 7 insertions(+), 5 deletions(-) diff --git a/hw/net/e1000x_common.c b/hw/net/e1000x_common.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_common.c +++ b/hw/net/e1000x_common.c @@ -XXX,XX +XXX,XX @@ bool e1000x_hw_rx_enabled(uint32_t *mac) bool e1000x_is_oversized(uint32_t *mac, size_t size) { + size_t header_size = sizeof(struct eth_header) + sizeof(struct vlan_header); /* this is the size past which hardware will drop packets when setting LPE=0 */ - static const int maximum_ethernet_vlan_size = 1522; + size_t maximum_short_size = header_size + ETH_MTU; /* this is the size past which hardware will drop packets when setting LPE=1 */ - static const int maximum_ethernet_lpe_size = 16 * KiB; + size_t maximum_large_size = 16 * KiB - ETH_FCS_LEN; - if ((size > maximum_ethernet_lpe_size || - (size > maximum_ethernet_vlan_size - && !(mac[RCTL] & E1000_RCTL_LPE))) + if ((size > maximum_large_size || + (size > maximum_short_size && !(mac[RCTL] & E1000_RCTL_LPE))) && !(mac[RCTL] & E1000_RCTL_SBP)) { e1000x_inc_reg_if_not_full(mac, ROC); trace_e1000x_rx_oversized(size); diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ #define ETH_ALEN 6 #define ETH_HLEN 14 #define ETH_ZLEN 60 /* Min. octets in frame without FCS */ +#define ETH_FCS_LEN 4 +#define ETH_MTU 1500 struct eth_header { uint8_t h_dest[ETH_ALEN]; /* destination eth addr */ -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> e1000e and igb employs NetPktRssIpV6TcpEx for RSS hash if TcpIpv6 MRQC bit is set. Moreover, igb also has a MRQC bit for NetPktRssIpV6Tcp though it is not implemented yet. Rename it to TcpIpv6Ex to avoid confusion. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 8 ++++---- hw/net/e1000x_regs.h | 24 ++++++++++++------------ hw/net/igb_core.c | 8 ++++---- hw/net/trace-events | 2 +- 4 files changed, 21 insertions(+), 21 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_rss_get_hash_type(E1000ECore *core, struct NetRxPkt *pkt) ip6info->rss_ex_dst_valid, ip6info->rss_ex_src_valid, core->mac[MRQC], - E1000_MRQC_EN_TCPIPV6(core->mac[MRQC]), + E1000_MRQC_EN_TCPIPV6EX(core->mac[MRQC]), E1000_MRQC_EN_IPV6EX(core->mac[MRQC]), E1000_MRQC_EN_IPV6(core->mac[MRQC])); @@ -XXX,XX +XXX,XX @@ e1000e_rss_get_hash_type(E1000ECore *core, struct NetRxPkt *pkt) ip6info->rss_ex_src_valid))) { if (l4hdr_proto == ETH_L4_HDR_PROTO_TCP && - E1000_MRQC_EN_TCPIPV6(core->mac[MRQC])) { - return E1000_MRQ_RSS_TYPE_IPV6TCP; + E1000_MRQC_EN_TCPIPV6EX(core->mac[MRQC])) { + return E1000_MRQ_RSS_TYPE_IPV6TCPEX; } if (E1000_MRQC_EN_IPV6EX(core->mac[MRQC])) { @@ -XXX,XX +XXX,XX @@ e1000e_rss_calc_hash(E1000ECore *core, case E1000_MRQ_RSS_TYPE_IPV4TCP: type = NetPktRssIpV4Tcp; break; - case E1000_MRQ_RSS_TYPE_IPV6TCP: + case E1000_MRQ_RSS_TYPE_IPV6TCPEX: type = NetPktRssIpV6TcpEx; break; case E1000_MRQ_RSS_TYPE_IPV6: diff --git a/hw/net/e1000x_regs.h b/hw/net/e1000x_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000x_regs.h +++ b/hw/net/e1000x_regs.h @@ -XXX,XX +XXX,XX @@ #define E1000_RETA_IDX(hash) ((hash) & (BIT(7) - 1)) #define E1000_RETA_VAL(reta, hash) (((uint8_t *)(reta))[E1000_RETA_IDX(hash)]) -#define E1000_MRQC_EN_TCPIPV4(mrqc) ((mrqc) & BIT(16)) -#define E1000_MRQC_EN_IPV4(mrqc) ((mrqc) & BIT(17)) -#define E1000_MRQC_EN_TCPIPV6(mrqc) ((mrqc) & BIT(18)) -#define E1000_MRQC_EN_IPV6EX(mrqc) ((mrqc) & BIT(19)) -#define E1000_MRQC_EN_IPV6(mrqc) ((mrqc) & BIT(20)) - -#define E1000_MRQ_RSS_TYPE_NONE (0) -#define E1000_MRQ_RSS_TYPE_IPV4TCP (1) -#define E1000_MRQ_RSS_TYPE_IPV4 (2) -#define E1000_MRQ_RSS_TYPE_IPV6TCP (3) -#define E1000_MRQ_RSS_TYPE_IPV6EX (4) -#define E1000_MRQ_RSS_TYPE_IPV6 (5) +#define E1000_MRQC_EN_TCPIPV4(mrqc) ((mrqc) & BIT(16)) +#define E1000_MRQC_EN_IPV4(mrqc) ((mrqc) & BIT(17)) +#define E1000_MRQC_EN_TCPIPV6EX(mrqc) ((mrqc) & BIT(18)) +#define E1000_MRQC_EN_IPV6EX(mrqc) ((mrqc) & BIT(19)) +#define E1000_MRQC_EN_IPV6(mrqc) ((mrqc) & BIT(20)) + +#define E1000_MRQ_RSS_TYPE_NONE (0) +#define E1000_MRQ_RSS_TYPE_IPV4TCP (1) +#define E1000_MRQ_RSS_TYPE_IPV4 (2) +#define E1000_MRQ_RSS_TYPE_IPV6TCPEX (3) +#define E1000_MRQ_RSS_TYPE_IPV6EX (4) +#define E1000_MRQ_RSS_TYPE_IPV6 (5) #define E1000_ICR_ASSERTED BIT(31) #define E1000_EIAC_MASK 0x01F00000 diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_rss_get_hash_type(IGBCore *core, struct NetRxPkt *pkt) ip6info->rss_ex_dst_valid, ip6info->rss_ex_src_valid, core->mac[MRQC], - E1000_MRQC_EN_TCPIPV6(core->mac[MRQC]), + E1000_MRQC_EN_TCPIPV6EX(core->mac[MRQC]), E1000_MRQC_EN_IPV6EX(core->mac[MRQC]), E1000_MRQC_EN_IPV6(core->mac[MRQC])); @@ -XXX,XX +XXX,XX @@ igb_rss_get_hash_type(IGBCore *core, struct NetRxPkt *pkt) ip6info->rss_ex_src_valid))) { if (l4hdr_proto == ETH_L4_HDR_PROTO_TCP && - E1000_MRQC_EN_TCPIPV6(core->mac[MRQC])) { - return E1000_MRQ_RSS_TYPE_IPV6TCP; + E1000_MRQC_EN_TCPIPV6EX(core->mac[MRQC])) { + return E1000_MRQ_RSS_TYPE_IPV6TCPEX; } if (E1000_MRQC_EN_IPV6EX(core->mac[MRQC])) { @@ -XXX,XX +XXX,XX @@ igb_rss_calc_hash(IGBCore *core, struct NetRxPkt *pkt, E1000E_RSSInfo *info) case E1000_MRQ_RSS_TYPE_IPV4TCP: type = NetPktRssIpV4Tcp; break; - case E1000_MRQ_RSS_TYPE_IPV6TCP: + case E1000_MRQ_RSS_TYPE_IPV6TCPEX: type = NetPktRssIpV6TcpEx; break; case E1000_MRQ_RSS_TYPE_IPV6: diff --git a/hw/net/trace-events b/hw/net/trace-events index XXXXXXX..XXXXXXX 100644 --- a/hw/net/trace-events +++ b/hw/net/trace-events @@ -XXX,XX +XXX,XX @@ e1000e_rx_rss_disabled(void) "RSS is disabled" e1000e_rx_rss_type(uint32_t type) "RSS type is %u" e1000e_rx_rss_ip4(int l4hdr_proto, uint32_t mrqc, bool tcpipv4_enabled, bool ipv4_enabled) "RSS IPv4: L4 header protocol %d, mrqc 0x%X, tcpipv4 enabled %d, ipv4 enabled %d" e1000e_rx_rss_ip6_rfctl(uint32_t rfctl) "RSS IPv6: rfctl 0x%X" -e1000e_rx_rss_ip6(bool ex_dis, bool new_ex_dis, int l4hdr_proto, bool has_ext_headers, bool ex_dst_valid, bool ex_src_valid, uint32_t mrqc, bool tcpipv6_enabled, bool ipv6ex_enabled, bool ipv6_enabled) "RSS IPv6: ex_dis: %d, new_ex_dis: %d, L4 header protocol %d, has_ext_headers %d, ex_dst_valid %d, ex_src_valid %d, mrqc 0x%X, tcpipv6 enabled %d, ipv6ex enabled %d, ipv6 enabled %d" +e1000e_rx_rss_ip6(bool ex_dis, bool new_ex_dis, int l4hdr_proto, bool has_ext_headers, bool ex_dst_valid, bool ex_src_valid, uint32_t mrqc, bool tcpipv6ex_enabled, bool ipv6ex_enabled, bool ipv6_enabled) "RSS IPv6: ex_dis: %d, new_ex_dis: %d, L4 header protocol %d, has_ext_headers %d, ex_dst_valid %d, ex_src_valid %d, mrqc 0x%X, tcpipv6ex enabled %d, ipv6ex enabled %d, ipv6 enabled %d" e1000e_rx_metadata_protocols(bool hasip4, bool hasip6, int l4hdr_protocol) "protocols: ip4: %d, ip6: %d, l4hdr: %d" e1000e_rx_metadata_vlan(uint16_t vlan_tag) "VLAN tag is 0x%X" -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Without this change, the status flags may not be traced e.g. if checksum offloading is disabled. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_build_rx_metadata(E1000ECore *core, trace_e1000e_rx_metadata_l4_cso_disabled(); } - trace_e1000e_rx_metadata_status_flags(*status_flags); - func_exit: + trace_e1000e_rx_metadata_status_flags(*status_flags); *status_flags = cpu_to_le32(*status_flags); } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Without this change, the status flags may not be traced e.g. if checksum offloading is disabled. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, trace_e1000e_rx_metadata_l4_cso_disabled(); } - trace_e1000e_rx_metadata_status_flags(*status_flags); - func_exit: + trace_e1000e_rx_metadata_status_flags(*status_flags); *status_flags = cpu_to_le32(*status_flags); } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> The goto is a bit confusing as it changes the control flow only if L4 protocol is not recognized. It is also different from e1000e, and noisy when comparing e1000e and igb. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, break; default: - goto func_exit; + break; } } else { trace_e1000e_rx_metadata_l4_cso_disabled(); -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Section 7.2.2.3 Advanced Transmit Data Descriptor says: > For frames that spans multiple descriptors, all fields apart from > DCMD.EOP, DCMD.RS, DCMD.DEXT, DTALEN, Address and DTYP are valid only > in the first descriptors and are ignored in the subsequent ones. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, idx = (tx->first_olinfo_status >> 4) & 1; igb_tx_insert_vlan(core, queue_index, tx, tx->ctx[idx].vlan_macip_lens >> 16, - !!(cmd_type_len & E1000_TXD_CMD_VLE)); + !!(tx->first_cmd_type_len & E1000_TXD_CMD_VLE)); if (igb_tx_pkt_send(core, tx, queue_index)) { igb_on_tx_done_update_stats(core, tx->tx_pkt, queue_index); -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Keeping Tx packet state after the transmit queue is emptied has some problems: - The datasheet says the descriptors can be reused after the transmit queue is emptied, but the Tx packet state may keep references to them. - The Tx packet state cannot be migrated so it can be reset anytime the migration happens. Always reset Tx packet state always after the queue is emptied. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_start_xmit(E1000ECore *core, const E1000E_TxRing *txr) if (!ide || !e1000e_intrmgr_delay_tx_causes(core, &cause)) { e1000e_set_interrupt_cause(core, cause); } + + net_tx_pkt_reset(txr->tx->tx_pkt, net_tx_pkt_unmap_frag_pci, core->owner); } static bool @@ -XXX,XX +XXX,XX @@ e1000e_core_pci_uninit(E1000ECore *core) qemu_del_vm_change_state_handler(core->vmstate); for (i = 0; i < E1000E_NUM_QUEUES; i++) { - net_tx_pkt_reset(core->tx[i].tx_pkt, - net_tx_pkt_unmap_frag_pci, core->owner); net_tx_pkt_uninit(core->tx[i].tx_pkt); } @@ -XXX,XX +XXX,XX @@ static void e1000e_reset(E1000ECore *core, bool sw) e1000x_reset_mac_addr(core->owner_nic, core->mac, core->permanent_mac); for (i = 0; i < ARRAY_SIZE(core->tx); i++) { - net_tx_pkt_reset(core->tx[i].tx_pkt, - net_tx_pkt_unmap_frag_pci, core->owner); memset(&core->tx[i].props, 0, sizeof(core->tx[i].props)); core->tx[i].skip_cp = false; } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Keeping Tx packet state after the transmit queue is emptied but this behavior is unreliable as the state can be reset anytime the migration happens. Always reset Tx packet state always after the queue is emptied. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/vmxnet3.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/hw/net/vmxnet3.c b/hw/net/vmxnet3.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/vmxnet3.c +++ b/hw/net/vmxnet3.c @@ -XXX,XX +XXX,XX @@ static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx) net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s)); } } + + net_tx_pkt_reset(s->tx_pkt, net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s)); } static inline void @@ -XXX,XX +XXX,XX @@ static void vmxnet3_deactivate_device(VMXNET3State *s) { if (s->device_active) { VMW_CBPRN("Deactivating vmxnet3..."); - net_tx_pkt_reset(s->tx_pkt, net_tx_pkt_unmap_frag_pci, PCI_DEVICE(s)); net_tx_pkt_uninit(s->tx_pkt); net_rx_pkt_uninit(s->rx_pkt); s->device_active = false; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 4 ++-- hw/net/igb_regs.h | 32 +++++++++++++++++++++++++++----- 2 files changed, 29 insertions(+), 7 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_setup_tx_offloads(IGBCore *core, struct igb_tx *tx) { if (tx->first_cmd_type_len & E1000_ADVTXD_DCMD_TSE) { uint32_t idx = (tx->first_olinfo_status >> 4) & 1; - uint32_t mss = tx->ctx[idx].mss_l4len_idx >> 16; + uint32_t mss = tx->ctx[idx].mss_l4len_idx >> E1000_ADVTXD_MSS_SHIFT; if (!net_tx_pkt_build_vheader(tx->tx_pkt, true, true, mss)) { return false; } @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, if (!tx->skip_cp && net_tx_pkt_parse(tx->tx_pkt)) { idx = (tx->first_olinfo_status >> 4) & 1; igb_tx_insert_vlan(core, queue_index, tx, - tx->ctx[idx].vlan_macip_lens >> 16, + tx->ctx[idx].vlan_macip_lens >> IGB_TX_FLAGS_VLAN_SHIFT, !!(tx->first_cmd_type_len & E1000_TXD_CMD_VLE)); if (igb_tx_pkt_send(core, tx, queue_index)) { diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_regs.h +++ b/hw/net/igb_regs.h @@ -XXX,XX +XXX,XX @@ union e1000_adv_tx_desc { } wb; }; -#define E1000_ADVTXD_DTYP_CTXT 0x00200000 /* Advanced Context Descriptor */ -#define E1000_ADVTXD_DTYP_DATA 0x00300000 /* Advanced Data Descriptor */ -#define E1000_ADVTXD_DCMD_DEXT 0x20000000 /* Descriptor Extension (1=Adv) */ -#define E1000_ADVTXD_DCMD_TSE 0x80000000 /* TCP/UDP Segmentation Enable */ - #define E1000_ADVTXD_POTS_IXSM 0x00000100 /* Insert TCP/UDP Checksum */ #define E1000_ADVTXD_POTS_TXSM 0x00000200 /* Insert TCP/UDP Checksum */ @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define IGB_82576_VF_DEV_ID 0x10CA #define IGB_I350_VF_DEV_ID 0x1520 +/* VLAN info */ +#define IGB_TX_FLAGS_VLAN_MASK 0xffff0000 +#define IGB_TX_FLAGS_VLAN_SHIFT 16 + /* from igb/e1000_82575.h */ #define E1000_MRQC_ENABLE_RSS_MQ 0x00000002 @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define E1000_MRQC_RSS_FIELD_IPV6_UDP 0x00800000 #define E1000_MRQC_RSS_FIELD_IPV6_UDP_EX 0x01000000 +/* Adv Transmit Descriptor Config Masks */ +#define E1000_ADVTXD_MAC_TSTAMP 0x00080000 /* IEEE1588 Timestamp packet */ +#define E1000_ADVTXD_DTYP_CTXT 0x00200000 /* Advanced Context Descriptor */ +#define E1000_ADVTXD_DTYP_DATA 0x00300000 /* Advanced Data Descriptor */ +#define E1000_ADVTXD_DCMD_EOP 0x01000000 /* End of Packet */ +#define E1000_ADVTXD_DCMD_IFCS 0x02000000 /* Insert FCS (Ethernet CRC) */ +#define E1000_ADVTXD_DCMD_RS 0x08000000 /* Report Status */ +#define E1000_ADVTXD_DCMD_DEXT 0x20000000 /* Descriptor extension (1=Adv) */ +#define E1000_ADVTXD_DCMD_VLE 0x40000000 /* VLAN pkt enable */ +#define E1000_ADVTXD_DCMD_TSE 0x80000000 /* TCP Seg enable */ +#define E1000_ADVTXD_PAYLEN_SHIFT 14 /* Adv desc PAYLEN shift */ + +#define E1000_ADVTXD_MACLEN_SHIFT 9 /* Adv ctxt desc mac len shift */ +#define E1000_ADVTXD_TUCMD_L4T_UDP 0x00000000 /* L4 Packet TYPE of UDP */ +#define E1000_ADVTXD_TUCMD_IPV4 0x00000400 /* IP Packet Type: 1=IPv4 */ +#define E1000_ADVTXD_TUCMD_L4T_TCP 0x00000800 /* L4 Packet TYPE of TCP */ +#define E1000_ADVTXD_TUCMD_L4T_SCTP 0x00001000 /* L4 packet TYPE of SCTP */ +/* IPSec Encrypt Enable for ESP */ +#define E1000_ADVTXD_L4LEN_SHIFT 8 /* Adv ctxt L4LEN shift */ +#define E1000_ADVTXD_MSS_SHIFT 16 /* Adv ctxt MSS shift */ +/* Adv ctxt IPSec SA IDX mask */ +/* Adv ctxt IPSec ESP len mask */ + /* Additional Transmit Descriptor Control definitions */ #define E1000_TXDCTL_QUEUE_ENABLE 0x02000000 /* Enable specific Tx Queue */ -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> The constants need to be consistent between the PF and VF. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb.c | 10 +++++----- hw/net/igb_common.h | 8 ++++++++ hw/net/igbvf.c | 7 ------- 3 files changed, 13 insertions(+), 12 deletions(-) diff --git a/hw/net/igb.c b/hw/net/igb.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb.c +++ b/hw/net/igb.c @@ -XXX,XX +XXX,XX @@ static void igb_pci_realize(PCIDevice *pci_dev, Error **errp) pcie_ari_init(pci_dev, 0x150, 1); - pcie_sriov_pf_init(pci_dev, IGB_CAP_SRIOV_OFFSET, "igbvf", + pcie_sriov_pf_init(pci_dev, IGB_CAP_SRIOV_OFFSET, TYPE_IGBVF, IGB_82576_VF_DEV_ID, IGB_MAX_VF_FUNCTIONS, IGB_MAX_VF_FUNCTIONS, IGB_VF_OFFSET, IGB_VF_STRIDE); - pcie_sriov_pf_init_vf_bar(pci_dev, 0, + pcie_sriov_pf_init_vf_bar(pci_dev, IGBVF_MMIO_BAR_IDX, PCI_BASE_ADDRESS_MEM_TYPE_64 | PCI_BASE_ADDRESS_MEM_PREFETCH, - 16 * KiB); - pcie_sriov_pf_init_vf_bar(pci_dev, 3, + IGBVF_MMIO_SIZE); + pcie_sriov_pf_init_vf_bar(pci_dev, IGBVF_MSIX_BAR_IDX, PCI_BASE_ADDRESS_MEM_TYPE_64 | PCI_BASE_ADDRESS_MEM_PREFETCH, - 16 * KiB); + IGBVF_MSIX_SIZE); igb_init_net_peer(s, pci_dev, macaddr); diff --git a/hw/net/igb_common.h b/hw/net/igb_common.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_common.h +++ b/hw/net/igb_common.h @@ -XXX,XX +XXX,XX @@ #include "igb_regs.h" +#define TYPE_IGBVF "igbvf" + +#define IGBVF_MMIO_BAR_IDX (0) +#define IGBVF_MSIX_BAR_IDX (3) + +#define IGBVF_MMIO_SIZE (16 * 1024) +#define IGBVF_MSIX_SIZE (16 * 1024) + #define defreg(x) x = (E1000_##x >> 2) #define defreg_indexed(x, i) x##i = (E1000_##x(i) >> 2) #define defreg_indexeda(x, i) x##i##_A = (E1000_##x##_A(i) >> 2) diff --git a/hw/net/igbvf.c b/hw/net/igbvf.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igbvf.c +++ b/hw/net/igbvf.c @@ -XXX,XX +XXX,XX @@ #include "trace.h" #include "qapi/error.h" -#define TYPE_IGBVF "igbvf" OBJECT_DECLARE_SIMPLE_TYPE(IgbVfState, IGBVF) -#define IGBVF_MMIO_BAR_IDX (0) -#define IGBVF_MSIX_BAR_IDX (3) - -#define IGBVF_MMIO_SIZE (16 * 1024) -#define IGBVF_MSIX_SIZE (16 * 1024) - struct IgbVfState { PCIDevice parent_obj; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 96 +++++++++++++++++++++++++++---------------------------- 1 file changed, 48 insertions(+), 48 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ static const uint32_t igb_mac_reg_init[] = { [VMOLR0 ... VMOLR0 + 7] = 0x2600 | E1000_VMOLR_STRCRC, [RPLOLR] = E1000_RPLOLR_STRCRC, [RLPML] = 0x2600, - [TXCTL0] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL1] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL2] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL3] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL4] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL5] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL6] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL7] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL8] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL9] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL10] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL11] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL12] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL13] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL14] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, - [TXCTL15] = E1000_DCA_TXCTRL_DATA_RRO_EN | - E1000_DCA_TXCTRL_TX_WB_RO_EN | - E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL0] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL1] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL2] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL3] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL4] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL5] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL6] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL7] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL8] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL9] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL10] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL11] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL12] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL13] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL14] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, + [TXCTL15] = E1000_DCA_TXCTRL_DATA_RRO_EN | + E1000_DCA_TXCTRL_TX_WB_RO_EN | + E1000_DCA_TXCTRL_DESC_RRO_EN, }; static void igb_reset(IGBCore *core, bool sw) -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Section 7.3.4.1 says: > When auto-clear is enabled for an interrupt cause, the EICR bit is > set when a cause event mapped to this vector occurs. When the EITR > Counter reaches zero, the MSI-X message is sent on PCIe. Then the > EICR bit is cleared and enabled to be set by a new cause event Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_lower_legacy_irq(IGBCore *core) pci_set_irq(core->owner, 0); } -static void igb_msix_notify(IGBCore *core, unsigned int vector) +static void igb_msix_notify(IGBCore *core, unsigned int cause) { PCIDevice *dev = core->owner; uint16_t vfn; + uint32_t effective_eiac; + unsigned int vector; - vfn = 8 - (vector + 2) / IGBVF_MSIX_VEC_NUM; + vfn = 8 - (cause + 2) / IGBVF_MSIX_VEC_NUM; if (vfn < pcie_sriov_num_vfs(core->owner)) { dev = pcie_sriov_get_vf_at_index(core->owner, vfn); assert(dev); - vector = (vector + 2) % IGBVF_MSIX_VEC_NUM; - } else if (vector >= IGB_MSIX_VEC_NUM) { + vector = (cause + 2) % IGBVF_MSIX_VEC_NUM; + } else if (cause >= IGB_MSIX_VEC_NUM) { qemu_log_mask(LOG_GUEST_ERROR, "igb: Tried to use vector unavailable for PF"); return; + } else { + vector = cause; } msix_notify(dev, vector); + + trace_e1000e_irq_icr_clear_eiac(core->mac[EICR], core->mac[EIAC]); + effective_eiac = core->mac[EIAC] & BIT(cause); + core->mac[EICR] &= ~effective_eiac; } static inline void @@ -XXX,XX +XXX,XX @@ igb_eitr_should_postpone(IGBCore *core, int idx) static void igb_send_msix(IGBCore *core) { uint32_t causes = core->mac[EICR] & core->mac[EIMS]; - uint32_t effective_eiac; int vector; for (vector = 0; vector < IGB_INTR_NUM; ++vector) { @@ -XXX,XX +XXX,XX @@ static void igb_send_msix(IGBCore *core) trace_e1000e_irq_msix_notify_vec(vector); igb_msix_notify(core, vector); - - trace_e1000e_irq_icr_clear_eiac(core->mac[EICR], core->mac[EIAC]); - effective_eiac = core->mac[EIAC] & BIT(vector); - core->mac[EICR] &= ~effective_eiac; } } } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Rename variable "n" to "causes", which properly represents the content of the variable. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ static ssize_t e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, bool has_vnet) { - uint32_t n = 0; + uint32_t causes = 0; uint8_t buf[ETH_ZLEN]; struct iovec min_iov; size_t size, orig_size; @@ -XXX,XX +XXX,XX @@ e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, /* Perform small receive detection (RSRPD) */ if (total_size < core->mac[RSRPD]) { - n |= E1000_ICS_SRPD; + causes |= E1000_ICS_SRPD; } /* Perform ACK receive detection */ if (!(core->mac[RFCTL] & E1000_RFCTL_ACK_DIS) && (e1000e_is_tcp_ack(core, core->rx_pkt))) { - n |= E1000_ICS_ACK; + causes |= E1000_ICS_ACK; } /* Check if receive descriptor minimum threshold hit */ rdmts_hit = e1000e_rx_descr_threshold_hit(core, rxr.i); - n |= e1000e_rx_wb_interrupt_cause(core, rxr.i->idx, rdmts_hit); + causes |= e1000e_rx_wb_interrupt_cause(core, rxr.i->idx, rdmts_hit); trace_e1000e_rx_written_to_guest(rxr.i->idx); } else { - n |= E1000_ICS_RXO; + causes |= E1000_ICS_RXO; retval = 0; trace_e1000e_rx_not_written_to_guest(rxr.i->idx); } - if (!e1000e_intrmgr_delay_rx_causes(core, &n)) { - trace_e1000e_rx_interrupt_set(n); - e1000e_set_interrupt_cause(core, n); + if (!e1000e_intrmgr_delay_rx_causes(core, &causes)) { + trace_e1000e_rx_interrupt_set(causes); + e1000e_set_interrupt_cause(core, causes); } else { - trace_e1000e_rx_interrupt_delayed(n); + trace_e1000e_rx_interrupt_delayed(causes); } return retval; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Rename variable "n" to "causes", which properly represents the content of the variable. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, bool has_vnet, bool *external_tx) { uint16_t queues = 0; - uint32_t n = 0; + uint32_t causes = 0; union { L2Header l2_header; uint8_t octets[ETH_ZLEN]; @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, e1000x_fcs_len(core->mac); if (!igb_has_rxbufs(core, rxr.i, total_size)) { - n |= E1000_ICS_RXO; + causes |= E1000_ICS_RXO; trace_e1000e_rx_not_written_to_guest(rxr.i->idx); continue; } - n |= E1000_ICR_RXDW; + causes |= E1000_ICR_RXDW; igb_rx_fix_l4_csum(core, core->rx_pkt); igb_write_packet_to_guest(core, core->rx_pkt, &rxr, &rss_info); /* Check if receive descriptor minimum threshold hit */ if (igb_rx_descr_threshold_hit(core, rxr.i)) { - n |= E1000_ICS_RXDMT0; + causes |= E1000_ICS_RXDMT0; } core->mac[EICR] |= igb_rx_wb_eic(core, rxr.i->idx); @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, trace_e1000e_rx_written_to_guest(rxr.i->idx); } - trace_e1000e_rx_interrupt_set(n); - igb_set_interrupt_cause(core, n); + trace_e1000e_rx_interrupt_set(causes); + igb_set_interrupt_cause(core, causes); return orig_size; } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> The uses of uint8_t pointers were misleading as they are never accessed as an array of octets and it even require more strict alignment to access as struct eth_header. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- include/net/eth.h | 4 ++-- net/eth.c | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ eth_get_pkt_tci(const void *p) size_t eth_strip_vlan(const struct iovec *iov, int iovcnt, size_t iovoff, - uint8_t *new_ehdr_buf, + void *new_ehdr_buf, uint16_t *payload_offset, uint16_t *tci); size_t eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, - uint16_t vet, uint8_t *new_ehdr_buf, + uint16_t vet, void *new_ehdr_buf, uint16_t *payload_offset, uint16_t *tci); uint16_t diff --git a/net/eth.c b/net/eth.c index XXXXXXX..XXXXXXX 100644 --- a/net/eth.c +++ b/net/eth.c @@ -XXX,XX +XXX,XX @@ void eth_get_protocols(const struct iovec *iov, size_t iovcnt, size_t iovoff, size_t eth_strip_vlan(const struct iovec *iov, int iovcnt, size_t iovoff, - uint8_t *new_ehdr_buf, + void *new_ehdr_buf, uint16_t *payload_offset, uint16_t *tci) { struct vlan_header vlan_hdr; - struct eth_header *new_ehdr = (struct eth_header *) new_ehdr_buf; + struct eth_header *new_ehdr = new_ehdr_buf; size_t copied = iov_to_buf(iov, iovcnt, iovoff, new_ehdr, sizeof(*new_ehdr)); @@ -XXX,XX +XXX,XX @@ eth_strip_vlan(const struct iovec *iov, int iovcnt, size_t iovoff, size_t eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, - uint16_t vet, uint8_t *new_ehdr_buf, + uint16_t vet, void *new_ehdr_buf, uint16_t *payload_offset, uint16_t *tci) { struct vlan_header vlan_hdr; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> It is possible to have another VLAN tag even if the packet is already tagged. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/net_tx_pkt.c | 16 +++++++--------- include/net/eth.h | 4 ++-- net/eth.c | 22 ++++++---------------- 3 files changed, 15 insertions(+), 27 deletions(-) diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.c +++ b/hw/net/net_tx_pkt.c @@ -XXX,XX +XXX,XX @@ struct NetTxPkt { struct iovec *vec; - uint8_t l2_hdr[ETH_MAX_L2_HDR_LEN]; + struct { + struct eth_header eth; + struct vlan_header vlan[3]; + } l2_hdr; union { struct ip_header ip; struct ip6_header ip6; @@ -XXX,XX +XXX,XX @@ bool net_tx_pkt_build_vheader(struct NetTxPkt *pkt, bool tso_enable, void net_tx_pkt_setup_vlan_header_ex(struct NetTxPkt *pkt, uint16_t vlan, uint16_t vlan_ethtype) { - bool is_new; assert(pkt); eth_setup_vlan_headers(pkt->vec[NET_TX_PKT_L2HDR_FRAG].iov_base, - vlan, vlan_ethtype, &is_new); + &pkt->vec[NET_TX_PKT_L2HDR_FRAG].iov_len, + vlan, vlan_ethtype); - /* update l2hdrlen */ - if (is_new) { - pkt->hdr_len += sizeof(struct vlan_header); - pkt->vec[NET_TX_PKT_L2HDR_FRAG].iov_len += - sizeof(struct vlan_header); - } + pkt->hdr_len += sizeof(struct vlan_header); } bool net_tx_pkt_add_raw_fragment(struct NetTxPkt *pkt, void *base, size_t len) diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, uint16_t eth_get_l3_proto(const struct iovec *l2hdr_iov, int iovcnt, size_t l2hdr_len); -void eth_setup_vlan_headers(struct eth_header *ehdr, uint16_t vlan_tag, - uint16_t vlan_ethtype, bool *is_new); +void eth_setup_vlan_headers(struct eth_header *ehdr, size_t *ehdr_size, + uint16_t vlan_tag, uint16_t vlan_ethtype); uint8_t eth_get_gso_type(uint16_t l3_proto, uint8_t *l3_hdr, uint8_t l4proto); diff --git a/net/eth.c b/net/eth.c index XXXXXXX..XXXXXXX 100644 --- a/net/eth.c +++ b/net/eth.c @@ -XXX,XX +XXX,XX @@ #include "net/checksum.h" #include "net/tap.h" -void eth_setup_vlan_headers(struct eth_header *ehdr, uint16_t vlan_tag, - uint16_t vlan_ethtype, bool *is_new) +void eth_setup_vlan_headers(struct eth_header *ehdr, size_t *ehdr_size, + uint16_t vlan_tag, uint16_t vlan_ethtype) { struct vlan_header *vhdr = PKT_GET_VLAN_HDR(ehdr); - switch (be16_to_cpu(ehdr->h_proto)) { - case ETH_P_VLAN: - case ETH_P_DVLAN: - /* vlan hdr exists */ - *is_new = false; - break; - - default: - /* No VLAN header, put a new one */ - vhdr->h_proto = ehdr->h_proto; - ehdr->h_proto = cpu_to_be16(vlan_ethtype); - *is_new = true; - break; - } + memmove(vhdr + 1, vhdr, *ehdr_size - ETH_HLEN); vhdr->h_tci = cpu_to_be16(vlan_tag); + vhdr->h_proto = ehdr->h_proto; + ehdr->h_proto = cpu_to_be16(vlan_ethtype); + *ehdr_size += sizeof(*vhdr); } uint8_t -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> eth_strip_vlan and eth_strip_vlan_ex refers to ehdr_buf as struct eth_header. Enforce alignment for the structure. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/net_rx_pkt.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) diff --git a/hw/net/net_rx_pkt.c b/hw/net/net_rx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.c +++ b/hw/net/net_rx_pkt.c @@ -XXX,XX +XXX,XX @@ struct NetRxPkt { struct virtio_net_hdr virt_hdr; - uint8_t ehdr_buf[sizeof(struct eth_header) + sizeof(struct vlan_header)]; + struct { + struct eth_header eth; + struct vlan_header vlan; + } ehdr_buf; struct iovec *vec; uint16_t vec_len_total; uint16_t vec_len; @@ -XXX,XX +XXX,XX @@ net_rx_pkt_pull_data(struct NetRxPkt *pkt, if (pkt->ehdr_buf_len) { net_rx_pkt_iovec_realloc(pkt, iovcnt + 1); - pkt->vec[0].iov_base = pkt->ehdr_buf; + pkt->vec[0].iov_base = &pkt->ehdr_buf; pkt->vec[0].iov_len = pkt->ehdr_buf_len; pkt->tot_len = pllen + pkt->ehdr_buf_len; @@ -XXX,XX +XXX,XX @@ void net_rx_pkt_attach_iovec(struct NetRxPkt *pkt, assert(pkt); if (strip_vlan) { - pkt->ehdr_buf_len = eth_strip_vlan(iov, iovcnt, iovoff, pkt->ehdr_buf, + pkt->ehdr_buf_len = eth_strip_vlan(iov, iovcnt, iovoff, &pkt->ehdr_buf, &ploff, &tci); } else { pkt->ehdr_buf_len = 0; @@ -XXX,XX +XXX,XX @@ void net_rx_pkt_attach_iovec_ex(struct NetRxPkt *pkt, if (strip_vlan) { pkt->ehdr_buf_len = eth_strip_vlan_ex(iov, iovcnt, iovoff, vet, - pkt->ehdr_buf, + &pkt->ehdr_buf, &ploff, &tci); } else { pkt->ehdr_buf_len = 0; -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> GPIE.Multiple_MSIX is not set by default, and needs to be set to get interrupts from multiple MSI-X vectors. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- tests/qtest/libqos/igb.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/qtest/libqos/igb.c b/tests/qtest/libqos/igb.c index XXXXXXX..XXXXXXX 100644 --- a/tests/qtest/libqos/igb.c +++ b/tests/qtest/libqos/igb.c @@ -XXX,XX +XXX,XX @@ static void igb_pci_start_hw(QOSGraphObject *obj) e1000e_macreg_write(&d->e1000e, E1000_RCTL, E1000_RCTL_EN); /* Enable all interrupts */ + e1000e_macreg_write(&d->e1000e, E1000_GPIE, E1000_GPIE_MSIX_MODE); e1000e_macreg_write(&d->e1000e, E1000_IMS, 0xFFFFFFFF); e1000e_macreg_write(&d->e1000e, E1000_EIMS, 0xFFFFFFFF); -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_update_interrupt_state(IGBCore *core) icr = core->mac[ICR] & core->mac[IMS]; - if (msix_enabled(core->owner)) { + if (core->mac[GPIE] & E1000_GPIE_MSIX_MODE) { if (icr) { causes = 0; if (icr & E1000_ICR_DRSTA) { @@ -XXX,XX +XXX,XX @@ igb_update_interrupt_state(IGBCore *core) trace_e1000e_irq_pending_interrupts(core->mac[ICR] & core->mac[IMS], core->mac[ICR], core->mac[IMS]); - if (msi_enabled(core->owner)) { + if (msix_enabled(core->owner)) { + if (icr) { + trace_e1000e_irq_msix_notify_vec(0); + msix_notify(core->owner, 0); + } + } else if (msi_enabled(core->owner)) { if (icr) { msi_notify(core->owner, 0); } -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> e1000e does not support using UDP for RSS hash, but igb does. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 16 ++++++++++++++++ hw/net/igb_regs.h | 3 +++ 2 files changed, 19 insertions(+) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_rss_get_hash_type(IGBCore *core, struct NetRxPkt *pkt) return E1000_MRQ_RSS_TYPE_IPV4TCP; } + if (l4hdr_proto == ETH_L4_HDR_PROTO_UDP && + (core->mac[MRQC] & E1000_MRQC_RSS_FIELD_IPV4_UDP)) { + return E1000_MRQ_RSS_TYPE_IPV4UDP; + } + if (E1000_MRQC_EN_IPV4(core->mac[MRQC])) { return E1000_MRQ_RSS_TYPE_IPV4; } @@ -XXX,XX +XXX,XX @@ igb_rss_get_hash_type(IGBCore *core, struct NetRxPkt *pkt) return E1000_MRQ_RSS_TYPE_IPV6TCPEX; } + if (l4hdr_proto == ETH_L4_HDR_PROTO_UDP && + (core->mac[MRQC] & E1000_MRQC_RSS_FIELD_IPV6_UDP)) { + return E1000_MRQ_RSS_TYPE_IPV6UDP; + } + if (E1000_MRQC_EN_IPV6EX(core->mac[MRQC])) { return E1000_MRQ_RSS_TYPE_IPV6EX; } @@ -XXX,XX +XXX,XX @@ igb_rss_calc_hash(IGBCore *core, struct NetRxPkt *pkt, E1000E_RSSInfo *info) case E1000_MRQ_RSS_TYPE_IPV6EX: type = NetPktRssIpV6Ex; break; + case E1000_MRQ_RSS_TYPE_IPV4UDP: + type = NetPktRssIpV4Udp; + break; + case E1000_MRQ_RSS_TYPE_IPV6UDP: + type = NetPktRssIpV6Udp; + break; default: assert(false); return 0; diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_regs.h +++ b/hw/net/igb_regs.h @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define E1000_RSS_QUEUE(reta, hash) (E1000_RETA_VAL(reta, hash) & 0x0F) +#define E1000_MRQ_RSS_TYPE_IPV4UDP 7 +#define E1000_MRQ_RSS_TYPE_IPV6UDP 8 + #define E1000_STATUS_IOV_MODE 0x00040000 #define E1000_STATUS_NUM_VFS_SHIFT 14 -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 5 ++++ hw/net/igb_core.c | 15 +++++++++++- hw/net/igb_regs.h | 1 + hw/net/net_rx_pkt.c | 64 ++++++++++++++++++++++++++++++++++++++++++--------- include/net/eth.h | 4 +++- include/qemu/crc32c.h | 1 + net/eth.c | 4 ++++ util/crc32c.c | 8 +++++++ 8 files changed, 89 insertions(+), 13 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_verify_csum_in_sw(E1000ECore *core, return; } + if (l4hdr_proto != ETH_L4_HDR_PROTO_TCP && + l4hdr_proto != ETH_L4_HDR_PROTO_UDP) { + return; + } + if (!net_rx_pkt_validate_l4_csum(pkt, &csum_valid)) { trace_e1000e_rx_metadata_l4_csum_validation_failed(); return; diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, uint16_t *vlan_tag) { struct virtio_net_hdr *vhdr; - bool hasip4, hasip6; + bool hasip4, hasip6, csum_valid; EthL4HdrProto l4hdr_proto; *status_flags = E1000_RXD_STAT_DD; @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, *pkt_info |= E1000_ADVRXD_PKT_UDP; break; + case ETH_L4_HDR_PROTO_SCTP: + *pkt_info |= E1000_ADVRXD_PKT_SCTP; + break; + default: break; } @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, if (igb_rx_l4_cso_enabled(core)) { switch (l4hdr_proto) { + case ETH_L4_HDR_PROTO_SCTP: + if (!net_rx_pkt_validate_l4_csum(pkt, &csum_valid)) { + trace_e1000e_rx_metadata_l4_csum_validation_failed(); + goto func_exit; + } + if (!csum_valid) { + *status_flags |= E1000_RXDEXT_STATERR_TCPE; + } + /* fall through */ case ETH_L4_HDR_PROTO_TCP: *status_flags |= E1000_RXD_STAT_TCPCS; break; diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_regs.h +++ b/hw/net/igb_regs.h @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define E1000_ADVRXD_PKT_IP6 BIT(6) #define E1000_ADVRXD_PKT_TCP BIT(8) #define E1000_ADVRXD_PKT_UDP BIT(9) +#define E1000_ADVRXD_PKT_SCTP BIT(10) static inline uint8_t igb_ivar_entry_rx(uint8_t i) { diff --git a/hw/net/net_rx_pkt.c b/hw/net/net_rx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.c +++ b/hw/net/net_rx_pkt.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "qemu/crc32c.h" #include "trace.h" #include "net_rx_pkt.h" #include "net/checksum.h" @@ -XXX,XX +XXX,XX @@ _net_rx_pkt_calc_l4_csum(struct NetRxPkt *pkt) return csum; } -bool net_rx_pkt_validate_l4_csum(struct NetRxPkt *pkt, bool *csum_valid) +static bool +_net_rx_pkt_validate_sctp_sum(struct NetRxPkt *pkt) { - uint16_t csum; + size_t csum_off; + size_t off = pkt->l4hdr_off; + size_t vec_len = pkt->vec_len; + struct iovec *vec; + uint32_t calculated = 0; + uint32_t original; + bool valid; - trace_net_rx_pkt_l4_csum_validate_entry(); + for (vec = pkt->vec; vec->iov_len < off; vec++) { + off -= vec->iov_len; + vec_len--; + } - if (pkt->l4hdr_info.proto != ETH_L4_HDR_PROTO_TCP && - pkt->l4hdr_info.proto != ETH_L4_HDR_PROTO_UDP) { - trace_net_rx_pkt_l4_csum_validate_not_xxp(); + csum_off = off + 8; + + if (!iov_to_buf(vec, vec_len, csum_off, &original, sizeof(original))) { return false; } - if (pkt->l4hdr_info.proto == ETH_L4_HDR_PROTO_UDP && - pkt->l4hdr_info.hdr.udp.uh_sum == 0) { - trace_net_rx_pkt_l4_csum_validate_udp_with_no_checksum(); + if (!iov_from_buf(vec, vec_len, csum_off, + &calculated, sizeof(calculated))) { return false; } + calculated = crc32c(0xffffffff, + (uint8_t *)vec->iov_base + off, vec->iov_len - off); + calculated = iov_crc32c(calculated ^ 0xffffffff, vec + 1, vec_len - 1); + valid = calculated == le32_to_cpu(original); + iov_from_buf(vec, vec_len, csum_off, &original, sizeof(original)); + + return valid; +} + +bool net_rx_pkt_validate_l4_csum(struct NetRxPkt *pkt, bool *csum_valid) +{ + uint32_t csum; + + trace_net_rx_pkt_l4_csum_validate_entry(); + if (pkt->hasip4 && pkt->ip4hdr_info.fragment) { trace_net_rx_pkt_l4_csum_validate_ip4_fragment(); return false; } - csum = _net_rx_pkt_calc_l4_csum(pkt); + switch (pkt->l4hdr_info.proto) { + case ETH_L4_HDR_PROTO_UDP: + if (pkt->l4hdr_info.hdr.udp.uh_sum == 0) { + trace_net_rx_pkt_l4_csum_validate_udp_with_no_checksum(); + return false; + } + /* fall through */ + case ETH_L4_HDR_PROTO_TCP: + csum = _net_rx_pkt_calc_l4_csum(pkt); + *csum_valid = ((csum == 0) || (csum == 0xFFFF)); + break; + + case ETH_L4_HDR_PROTO_SCTP: + *csum_valid = _net_rx_pkt_validate_sctp_sum(pkt); + break; - *csum_valid = ((csum == 0) || (csum == 0xFFFF)); + default: + trace_net_rx_pkt_l4_csum_validate_not_xxp(); + return false; + } trace_net_rx_pkt_l4_csum_validate_csum(*csum_valid); diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ struct tcp_hdr { #define IP_HEADER_VERSION_6 (6) #define IP_PROTO_TCP (6) #define IP_PROTO_UDP (17) +#define IP_PROTO_SCTP (132) #define IPTOS_ECN_MASK 0x03 #define IPTOS_ECN(x) ((x) & IPTOS_ECN_MASK) #define IPTOS_ECN_CE 0x03 @@ -XXX,XX +XXX,XX @@ typedef struct eth_ip4_hdr_info_st { typedef enum EthL4HdrProto { ETH_L4_HDR_PROTO_INVALID, ETH_L4_HDR_PROTO_TCP, - ETH_L4_HDR_PROTO_UDP + ETH_L4_HDR_PROTO_UDP, + ETH_L4_HDR_PROTO_SCTP } EthL4HdrProto; typedef struct eth_l4_hdr_info_st { diff --git a/include/qemu/crc32c.h b/include/qemu/crc32c.h index XXXXXXX..XXXXXXX 100644 --- a/include/qemu/crc32c.h +++ b/include/qemu/crc32c.h @@ -XXX,XX +XXX,XX @@ uint32_t crc32c(uint32_t crc, const uint8_t *data, unsigned int length); +uint32_t iov_crc32c(uint32_t crc, const struct iovec *iov, size_t iov_cnt); #endif diff --git a/net/eth.c b/net/eth.c index XXXXXXX..XXXXXXX 100644 --- a/net/eth.c +++ b/net/eth.c @@ -XXX,XX +XXX,XX @@ void eth_get_protocols(const struct iovec *iov, size_t iovcnt, size_t iovoff, *l5hdr_off = *l4hdr_off + sizeof(l4hdr_info->hdr.udp); } break; + + case IP_PROTO_SCTP: + l4hdr_info->proto = ETH_L4_HDR_PROTO_SCTP; + break; } } diff --git a/util/crc32c.c b/util/crc32c.c index XXXXXXX..XXXXXXX 100644 --- a/util/crc32c.c +++ b/util/crc32c.c @@ -XXX,XX +XXX,XX @@ uint32_t crc32c(uint32_t crc, const uint8_t *data, unsigned int length) return crc^0xffffffff; } +uint32_t iov_crc32c(uint32_t crc, const struct iovec *iov, size_t iov_cnt) +{ + while (iov_cnt--) { + crc = crc32c(crc, iov->iov_base, iov->iov_len) ^ 0xffffffff; + iov++; + } + return crc ^ 0xffffffff; +} -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 12 +++++++----- hw/net/net_tx_pkt.c | 18 ++++++++++++++++++ hw/net/net_tx_pkt.h | 8 ++++++++ 3 files changed, 33 insertions(+), 5 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_tx_insert_vlan(IGBCore *core, uint16_t qn, struct igb_tx *tx, static bool igb_setup_tx_offloads(IGBCore *core, struct igb_tx *tx) { + uint32_t idx = (tx->first_olinfo_status >> 4) & 1; + if (tx->first_cmd_type_len & E1000_ADVTXD_DCMD_TSE) { - uint32_t idx = (tx->first_olinfo_status >> 4) & 1; uint32_t mss = tx->ctx[idx].mss_l4len_idx >> E1000_ADVTXD_MSS_SHIFT; if (!net_tx_pkt_build_vheader(tx->tx_pkt, true, true, mss)) { return false; @@ -XXX,XX +XXX,XX @@ igb_setup_tx_offloads(IGBCore *core, struct igb_tx *tx) return true; } - if (tx->first_olinfo_status & E1000_ADVTXD_POTS_TXSM) { - if (!net_tx_pkt_build_vheader(tx->tx_pkt, false, true, 0)) { - return false; - } + if ((tx->first_olinfo_status & E1000_ADVTXD_POTS_TXSM) && + !((tx->ctx[idx].type_tucmd_mlhl & E1000_ADVTXD_TUCMD_L4T_SCTP) ? + net_tx_pkt_update_sctp_checksum(tx->tx_pkt) : + net_tx_pkt_build_vheader(tx->tx_pkt, false, true, 0))) { + return false; } if (tx->first_olinfo_status & E1000_ADVTXD_POTS_IXSM) { diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.c +++ b/hw/net/net_tx_pkt.c @@ -XXX,XX +XXX,XX @@ */ #include "qemu/osdep.h" +#include "qemu/crc32c.h" #include "net/eth.h" #include "net/checksum.h" #include "net/tap.h" @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_update_ip_checksums(struct NetTxPkt *pkt) pkt->virt_hdr.csum_offset, &csum, sizeof(csum)); } +bool net_tx_pkt_update_sctp_checksum(struct NetTxPkt *pkt) +{ + uint32_t csum = 0; + struct iovec *pl_start_frag = pkt->vec + NET_TX_PKT_PL_START_FRAG; + + if (iov_from_buf(pl_start_frag, pkt->payload_frags, 8, &csum, sizeof(csum)) < sizeof(csum)) { + return false; + } + + csum = cpu_to_le32(iov_crc32c(0xffffffff, pl_start_frag, pkt->payload_frags)); + if (iov_from_buf(pl_start_frag, pkt->payload_frags, 8, &csum, sizeof(csum)) < sizeof(csum)) { + return false; + } + + return true; +} + static void net_tx_pkt_calculate_hdr_len(struct NetTxPkt *pkt) { pkt->hdr_len = pkt->vec[NET_TX_PKT_L2HDR_FRAG].iov_len + diff --git a/hw/net/net_tx_pkt.h b/hw/net/net_tx_pkt.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_tx_pkt.h +++ b/hw/net/net_tx_pkt.h @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_update_ip_checksums(struct NetTxPkt *pkt); void net_tx_pkt_update_ip_hdr_checksum(struct NetTxPkt *pkt); /** + * Calculate the SCTP checksum. + * + * @pkt: packet + * + */ +bool net_tx_pkt_update_sctp_checksum(struct NetTxPkt *pkt); + +/** * get length of all populated data. * * @pkt: packet -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 3 ++- hw/net/igb_core.c | 14 ++++++++++++-- hw/net/net_rx_pkt.c | 15 ++++++--------- hw/net/net_rx_pkt.h | 19 ++++++++++--------- include/net/eth.h | 4 ++-- net/eth.c | 52 +++++++++++++++++++++++++++++++++------------------- 6 files changed, 65 insertions(+), 42 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_receive_internal(E1000ECore *core, const struct iovec *iov, int iovcnt, } net_rx_pkt_attach_iovec_ex(core->rx_pkt, iov, iovcnt, iov_ofs, - e1000x_vlan_enabled(core->mac), core->mac[VET]); + e1000x_vlan_enabled(core->mac) ? 0 : -1, + core->mac[VET], 0); e1000e_rss_parse_packet(core, core->rx_pkt, &rss_info); e1000e_rx_ring_init(core, &rxr, rss_info.queue); diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, E1000E_RxRing rxr; E1000E_RSSInfo rss_info; size_t total_size; + int strip_vlan_index; int i; trace_e1000e_rx_receive_iov(iovcnt); @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, igb_rx_ring_init(core, &rxr, i); + if (!igb_rx_strip_vlan(core, rxr.i)) { + strip_vlan_index = -1; + } else if (core->mac[CTRL_EXT] & BIT(26)) { + strip_vlan_index = 1; + } else { + strip_vlan_index = 0; + } + net_rx_pkt_attach_iovec_ex(core->rx_pkt, iov, iovcnt, iov_ofs, - igb_rx_strip_vlan(core, rxr.i), - core->mac[VET] & 0xffff); + strip_vlan_index, + core->mac[VET] & 0xffff, + core->mac[VET] >> 16); total_size = net_rx_pkt_get_total_len(core->rx_pkt) + e1000x_fcs_len(core->mac); diff --git a/hw/net/net_rx_pkt.c b/hw/net/net_rx_pkt.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.c +++ b/hw/net/net_rx_pkt.c @@ -XXX,XX +XXX,XX @@ void net_rx_pkt_attach_iovec(struct NetRxPkt *pkt, void net_rx_pkt_attach_iovec_ex(struct NetRxPkt *pkt, const struct iovec *iov, int iovcnt, - size_t iovoff, bool strip_vlan, - uint16_t vet) + size_t iovoff, int strip_vlan_index, + uint16_t vet, uint16_t vet_ext) { uint16_t tci = 0; uint16_t ploff = iovoff; assert(pkt); - if (strip_vlan) { - pkt->ehdr_buf_len = eth_strip_vlan_ex(iov, iovcnt, iovoff, vet, - &pkt->ehdr_buf, - &ploff, &tci); - } else { - pkt->ehdr_buf_len = 0; - } + pkt->ehdr_buf_len = eth_strip_vlan_ex(iov, iovcnt, iovoff, + strip_vlan_index, vet, vet_ext, + &pkt->ehdr_buf, + &ploff, &tci); pkt->tci = tci; diff --git a/hw/net/net_rx_pkt.h b/hw/net/net_rx_pkt.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/net_rx_pkt.h +++ b/hw/net/net_rx_pkt.h @@ -XXX,XX +XXX,XX @@ void net_rx_pkt_attach_iovec(struct NetRxPkt *pkt, /** * attach scatter-gather data to rx packet * -* @pkt: packet -* @iov: received data scatter-gather list -* @iovcnt number of elements in iov -* @iovoff data start offset in the iov -* @strip_vlan: should the module strip vlan from data -* @vet: VLAN tag Ethernet type +* @pkt: packet +* @iov: received data scatter-gather list +* @iovcnt: number of elements in iov +* @iovoff: data start offset in the iov +* @strip_vlan_index: index of Q tag if it is to be stripped. negative otherwise. +* @vet: VLAN tag Ethernet type +* @vet_ext: outer VLAN tag Ethernet type * */ void net_rx_pkt_attach_iovec_ex(struct NetRxPkt *pkt, - const struct iovec *iov, int iovcnt, - size_t iovoff, bool strip_vlan, - uint16_t vet); + const struct iovec *iov, int iovcnt, + size_t iovoff, int strip_vlan_index, + uint16_t vet, uint16_t vet_ext); /** * attach data to rx packet diff --git a/include/net/eth.h b/include/net/eth.h index XXXXXXX..XXXXXXX 100644 --- a/include/net/eth.h +++ b/include/net/eth.h @@ -XXX,XX +XXX,XX @@ eth_strip_vlan(const struct iovec *iov, int iovcnt, size_t iovoff, uint16_t *payload_offset, uint16_t *tci); size_t -eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, - uint16_t vet, void *new_ehdr_buf, +eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, int index, + uint16_t vet, uint16_t vet_ext, void *new_ehdr_buf, uint16_t *payload_offset, uint16_t *tci); uint16_t diff --git a/net/eth.c b/net/eth.c index XXXXXXX..XXXXXXX 100644 --- a/net/eth.c +++ b/net/eth.c @@ -XXX,XX +XXX,XX @@ eth_strip_vlan(const struct iovec *iov, int iovcnt, size_t iovoff, } size_t -eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, - uint16_t vet, void *new_ehdr_buf, +eth_strip_vlan_ex(const struct iovec *iov, int iovcnt, size_t iovoff, int index, + uint16_t vet, uint16_t vet_ext, void *new_ehdr_buf, uint16_t *payload_offset, uint16_t *tci) { struct vlan_header vlan_hdr; - struct eth_header *new_ehdr = (struct eth_header *) new_ehdr_buf; - - size_t copied = iov_to_buf(iov, iovcnt, iovoff, - new_ehdr, sizeof(*new_ehdr)); - - if (copied < sizeof(*new_ehdr)) { - return 0; - } + uint16_t *new_ehdr_proto; + size_t new_ehdr_size; + size_t copied; - if (be16_to_cpu(new_ehdr->h_proto) == vet) { - copied = iov_to_buf(iov, iovcnt, iovoff + sizeof(*new_ehdr), - &vlan_hdr, sizeof(vlan_hdr)); + switch (index) { + case 0: + new_ehdr_proto = &PKT_GET_ETH_HDR(new_ehdr_buf)->h_proto; + new_ehdr_size = sizeof(struct eth_header); + copied = iov_to_buf(iov, iovcnt, iovoff, new_ehdr_buf, new_ehdr_size); + break; - if (copied < sizeof(vlan_hdr)) { + case 1: + new_ehdr_proto = &PKT_GET_VLAN_HDR(new_ehdr_buf)->h_proto; + new_ehdr_size = sizeof(struct eth_header) + sizeof(struct vlan_header); + copied = iov_to_buf(iov, iovcnt, iovoff, new_ehdr_buf, new_ehdr_size); + if (be16_to_cpu(PKT_GET_ETH_HDR(new_ehdr_buf)->h_proto) != vet_ext) { return 0; } + break; - new_ehdr->h_proto = vlan_hdr.h_proto; + default: + return 0; + } - *tci = be16_to_cpu(vlan_hdr.h_tci); - *payload_offset = iovoff + sizeof(*new_ehdr) + sizeof(vlan_hdr); - return sizeof(struct eth_header); + if (copied < new_ehdr_size || be16_to_cpu(*new_ehdr_proto) != vet) { + return 0; + } + + copied = iov_to_buf(iov, iovcnt, iovoff + new_ehdr_size, + &vlan_hdr, sizeof(vlan_hdr)); + if (copied < sizeof(vlan_hdr)) { + return 0; } - return 0; + *new_ehdr_proto = vlan_hdr.h_proto; + *payload_offset = iovoff + new_ehdr_size + sizeof(vlan_hdr); + *tci = be16_to_cpu(vlan_hdr.h_tci); + + return new_ehdr_size; } void -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ typedef struct IGBTxPktVmdqCallbackContext { typedef struct L2Header { struct eth_header eth; - struct vlan_header vlan; + struct vlan_header vlan[2]; } L2Header; static ssize_t @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, uint32_t f, ra[2], *macp, rctl = core->mac[RCTL]; uint16_t queues = 0; uint16_t oversized = 0; - uint16_t vid = be16_to_cpu(l2_header->vlan.h_tci) & VLAN_VID_MASK; + size_t vlan_num = 0; int i; memset(rss_info, 0, sizeof(E1000E_RSSInfo)); @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, *external_tx = true; } - if (e1000x_is_vlan_packet(ehdr, core->mac[VET] & 0xffff) && - !e1000x_rx_vlan_filter(core->mac, PKT_GET_VLAN_HDR(ehdr))) { + if (core->mac[CTRL_EXT] & BIT(26)) { + if (be16_to_cpu(ehdr->h_proto) == core->mac[VET] >> 16 && + be16_to_cpu(l2_header->vlan[0].h_proto) == (core->mac[VET] & 0xffff)) { + vlan_num = 2; + } + } else { + if (be16_to_cpu(ehdr->h_proto) == (core->mac[VET] & 0xffff)) { + vlan_num = 1; + } + } + + if (vlan_num && + !e1000x_rx_vlan_filter(core->mac, l2_header->vlan + vlan_num - 1)) { return queues; } @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, if (e1000x_vlan_rx_filter_enabled(core->mac)) { uint16_t mask = 0; - if (e1000x_is_vlan_packet(ehdr, core->mac[VET] & 0xffff)) { + if (vlan_num) { + uint16_t vid = be16_to_cpu(l2_header->vlan[vlan_num - 1].h_tci) & VLAN_VID_MASK; + for (i = 0; i < E1000_VLVF_ARRAY_SIZE; i++) { if ((core->mac[VLVF0 + i] & E1000_VLVF_VLANID_MASK) == vid && (core->mac[VLVF0 + i] & E1000_VLVF_VLANID_ENABLE)) { -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> igb has a configurable size limit for LPE, and uses different limits depending on whether the packet is treated as a VLAN packet. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 36 +++++++++++++++++++++--------------- 1 file changed, 21 insertions(+), 15 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_rx_l4_cso_enabled(IGBCore *core) return !!(core->mac[RXCSUM] & E1000_RXCSUM_TUOFLD); } -static bool -igb_rx_is_oversized(IGBCore *core, uint16_t qn, size_t size) +static bool igb_rx_is_oversized(IGBCore *core, const struct eth_header *ehdr, + size_t size, size_t vlan_num, + bool lpe, uint16_t rlpml) { - uint16_t pool = qn % IGB_NUM_VM_POOLS; - bool lpe = !!(core->mac[VMOLR0 + pool] & E1000_VMOLR_LPE); - int max_ethernet_lpe_size = - core->mac[VMOLR0 + pool] & E1000_VMOLR_RLPML_MASK; - int max_ethernet_vlan_size = 1522; - - return size > (lpe ? max_ethernet_lpe_size : max_ethernet_vlan_size); + size_t vlan_header_size = sizeof(struct vlan_header) * vlan_num; + size_t header_size = sizeof(struct eth_header) + vlan_header_size; + return lpe ? size + ETH_FCS_LEN > rlpml : size > header_size + ETH_MTU; } static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, uint16_t queues = 0; uint16_t oversized = 0; size_t vlan_num = 0; + bool lpe; + uint16_t rlpml; int i; memset(rss_info, 0, sizeof(E1000E_RSSInfo)); @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, } } + lpe = !!(core->mac[RCTL] & E1000_RCTL_LPE); + rlpml = core->mac[RLPML]; + if (!(core->mac[RCTL] & E1000_RCTL_SBP) && + igb_rx_is_oversized(core, ehdr, size, vlan_num, lpe, rlpml)) { + trace_e1000x_rx_oversized(size); + return queues; + } + if (vlan_num && !e1000x_rx_vlan_filter(core->mac, l2_header->vlan + vlan_num - 1)) { return queues; @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, queues &= core->mac[VFRE]; if (queues) { for (i = 0; i < IGB_NUM_VM_POOLS; i++) { - if ((queues & BIT(i)) && igb_rx_is_oversized(core, i, size)) { + lpe = !!(core->mac[VMOLR0 + i] & E1000_VMOLR_LPE); + rlpml = core->mac[VMOLR0 + i] & E1000_VMOLR_RLPML_MASK; + if ((queues & BIT(i)) && + igb_rx_is_oversized(core, ehdr, size, vlan_num, + lpe, rlpml)) { oversized |= BIT(i); } } @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, iov_to_buf(iov, iovcnt, iov_ofs, &buf, sizeof(buf.l2_header)); } - /* Discard oversized packets if !LPE and !SBP. */ - if (e1000x_is_oversized(core->mac, size)) { - return orig_size; - } - net_rx_pkt_set_packet_type(core->rx_pkt, get_eth_packet_type(&buf.l2_header.eth)); net_rx_pkt_set_protocols(core->rx_pkt, iov, iovcnt, iov_ofs); -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_common.h | 16 ++++--- hw/net/igb_core.c | 129 ++++++++++++++++++++++++++++++++++++++-------------- hw/net/igb_regs.h | 23 ++++++++++ 3 files changed, 127 insertions(+), 41 deletions(-) diff --git a/hw/net/igb_common.h b/hw/net/igb_common.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_common.h +++ b/hw/net/igb_common.h @@ -XXX,XX +XXX,XX @@ defreg_indexeda(x, 0), defreg_indexeda(x, 1), \ defreg_indexeda(x, 2), defreg_indexeda(x, 3) -#define defregv(x) defreg_indexed(x, 0), defreg_indexed(x, 1), \ +#define defreg8(x) defreg_indexed(x, 0), defreg_indexed(x, 1), \ defreg_indexed(x, 2), defreg_indexed(x, 3), \ defreg_indexed(x, 4), defreg_indexed(x, 5), \ defreg_indexed(x, 6), defreg_indexed(x, 7) @@ -XXX,XX +XXX,XX @@ enum { defreg(EICS), defreg(EIMS), defreg(EIMC), defreg(EIAM), defreg(EICR), defreg(IVAR_MISC), defreg(GPIE), + defreg(TSYNCRXCFG), defreg8(ETQF), + defreg(RXPBS), defregd(RDBAL), defregd(RDBAH), defregd(RDLEN), defregd(SRRCTL), defregd(RDH), defregd(RDT), defregd(RXDCTL), defregd(RXCTL), defregd(RQDPC), defreg(RA2), @@ -XXX,XX +XXX,XX @@ enum { defreg(VT_CTL), - defregv(P2VMAILBOX), defregv(V2PMAILBOX), defreg(MBVFICR), defreg(MBVFIMR), + defreg8(P2VMAILBOX), defreg8(V2PMAILBOX), defreg(MBVFICR), defreg(MBVFIMR), defreg(VFLRE), defreg(VFRE), defreg(VFTE), defreg(WVBR), defreg(QDE), defreg(DTXSWC), defreg_indexed(VLVF, 0), - defregv(VMOLR), defreg(RPLOLR), defregv(VMBMEM), defregv(VMVIR), + defreg8(VMOLR), defreg(RPLOLR), defreg8(VMBMEM), defreg8(VMVIR), - defregv(PVTCTRL), defregv(PVTEICS), defregv(PVTEIMS), defregv(PVTEIMC), - defregv(PVTEIAC), defregv(PVTEIAM), defregv(PVTEICR), defregv(PVFGPRC), - defregv(PVFGPTC), defregv(PVFGORC), defregv(PVFGOTC), defregv(PVFMPRC), - defregv(PVFGPRLBC), defregv(PVFGPTLBC), defregv(PVFGORLBC), defregv(PVFGOTLBC), + defreg8(PVTCTRL), defreg8(PVTEICS), defreg8(PVTEIMS), defreg8(PVTEIMC), + defreg8(PVTEIAC), defreg8(PVTEIAM), defreg8(PVTEICR), defreg8(PVFGPRC), + defreg8(PVFGPTC), defreg8(PVFGORC), defreg8(PVFGOTC), defreg8(PVFMPRC), + defreg8(PVFGPRLBC), defreg8(PVFGPTLBC), defreg8(PVFGORLBC), defreg8(PVFGOTLBC), defreg(MTA_A), diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ typedef struct L2Header { struct vlan_header vlan[2]; } L2Header; +typedef struct PTP2 { + uint8_t message_id_transport_specific; + uint8_t version_ptp; + uint16_t message_length; + uint8_t subdomain_number; + uint8_t reserved0; + uint16_t flags; + uint64_t correction; + uint8_t reserved1[5]; + uint8_t source_communication_technology; + uint32_t source_uuid_lo; + uint16_t source_uuid_hi; + uint16_t source_port_id; + uint16_t sequence_id; + uint8_t control; + uint8_t log_message_period; +} PTP2; + static ssize_t igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, bool has_vnet, bool *external_tx); @@ -XXX,XX +XXX,XX @@ static bool igb_rx_is_oversized(IGBCore *core, const struct eth_header *ehdr, return lpe ? size + ETH_FCS_LEN > rlpml : size > header_size + ETH_MTU; } -static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, - size_t size, E1000E_RSSInfo *rss_info, - bool *external_tx) +static uint16_t igb_receive_assign(IGBCore *core, const struct iovec *iov, + size_t iovcnt, size_t iov_ofs, + const L2Header *l2_header, size_t size, + E1000E_RSSInfo *rss_info, + uint16_t *etqf, bool *ts, bool *external_tx) { static const int ta_shift[] = { 4, 3, 2, 0 }; const struct eth_header *ehdr = &l2_header->eth; @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, uint16_t queues = 0; uint16_t oversized = 0; size_t vlan_num = 0; + PTP2 ptp2; bool lpe; uint16_t rlpml; int i; memset(rss_info, 0, sizeof(E1000E_RSSInfo)); + *ts = false; if (external_tx) { *external_tx = true; @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const L2Header *l2_header, return queues; } + for (*etqf = 0; *etqf < 8; (*etqf)++) { + if ((core->mac[ETQF0 + *etqf] & E1000_ETQF_FILTER_ENABLE) && + be16_to_cpu(ehdr->h_proto) == (core->mac[ETQF0 + *etqf] & E1000_ETQF_ETYPE_MASK)) { + if ((core->mac[ETQF0 + *etqf] & E1000_ETQF_1588) && + (core->mac[TSYNCRXCTL] & E1000_TSYNCRXCTL_ENABLED) && + !(core->mac[TSYNCRXCTL] & E1000_TSYNCRXCTL_VALID) && + iov_to_buf(iov, iovcnt, iov_ofs + ETH_HLEN, &ptp2, sizeof(ptp2)) >= sizeof(ptp2) && + (ptp2.version_ptp & 15) == 2 && + ptp2.message_id_transport_specific == ((core->mac[TSYNCRXCFG] >> 8) & 255)) { + e1000x_timestamp(core->mac, core->timadj, RXSTMPL, RXSTMPH); + *ts = true; + core->mac[TSYNCRXCTL] |= E1000_TSYNCRXCTL_VALID; + core->mac[RXSATRL] = le32_to_cpu(ptp2.source_uuid_lo); + core->mac[RXSATRH] = le16_to_cpu(ptp2.source_uuid_hi) | + (le16_to_cpu(ptp2.sequence_id) << 16); + } + break; + } + } + if (vlan_num && !e1000x_rx_vlan_filter(core->mac, l2_header->vlan + vlan_num - 1)) { return queues; @@ -XXX,XX +XXX,XX @@ static void igb_build_rx_metadata(IGBCore *core, struct NetRxPkt *pkt, bool is_eop, - const E1000E_RSSInfo *rss_info, + const E1000E_RSSInfo *rss_info, uint16_t etqf, bool ts, uint16_t *pkt_info, uint16_t *hdr_info, uint32_t *rss, uint32_t *status_flags, @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, if (pkt_info) { *pkt_info = rss_info->enabled ? rss_info->type : 0; - if (hasip4) { - *pkt_info |= E1000_ADVRXD_PKT_IP4; - } + if (etqf < 8) { + *pkt_info |= (BIT(11) | etqf) << 4; + } else { + if (hasip4) { + *pkt_info |= E1000_ADVRXD_PKT_IP4; + } - if (hasip6) { - *pkt_info |= E1000_ADVRXD_PKT_IP6; - } + if (hasip6) { + *pkt_info |= E1000_ADVRXD_PKT_IP6; + } - switch (l4hdr_proto) { - case ETH_L4_HDR_PROTO_TCP: - *pkt_info |= E1000_ADVRXD_PKT_TCP; - break; + switch (l4hdr_proto) { + case ETH_L4_HDR_PROTO_TCP: + *pkt_info |= E1000_ADVRXD_PKT_TCP; + break; - case ETH_L4_HDR_PROTO_UDP: - *pkt_info |= E1000_ADVRXD_PKT_UDP; - break; + case ETH_L4_HDR_PROTO_UDP: + *pkt_info |= E1000_ADVRXD_PKT_UDP; + break; - case ETH_L4_HDR_PROTO_SCTP: - *pkt_info |= E1000_ADVRXD_PKT_SCTP; - break; + case ETH_L4_HDR_PROTO_SCTP: + *pkt_info |= E1000_ADVRXD_PKT_SCTP; + break; - default: - break; + default: + break; + } } } @@ -XXX,XX +XXX,XX @@ igb_build_rx_metadata(IGBCore *core, *hdr_info = 0; } + if (ts) { + *status_flags |= BIT(16); + } + /* RX CSO information */ if (hasip6 && (core->mac[RFCTL] & E1000_RFCTL_IPV6_XSUM_DIS)) { trace_e1000e_rx_metadata_ipv6_sum_disabled(); @@ -XXX,XX +XXX,XX @@ func_exit: static inline void igb_write_lgcy_rx_descr(IGBCore *core, struct e1000_rx_desc *desc, struct NetRxPkt *pkt, - const E1000E_RSSInfo *rss_info, + const E1000E_RSSInfo *rss_info, uint16_t etqf, bool ts, uint16_t length) { uint32_t status_flags, rss; @@ -XXX,XX +XXX,XX @@ igb_write_lgcy_rx_descr(IGBCore *core, struct e1000_rx_desc *desc, desc->csum = 0; igb_build_rx_metadata(core, pkt, pkt != NULL, - rss_info, + rss_info, etqf, ts, NULL, NULL, &rss, &status_flags, &ip_id, &desc->special); @@ -XXX,XX +XXX,XX @@ igb_write_lgcy_rx_descr(IGBCore *core, struct e1000_rx_desc *desc, static inline void igb_write_adv_rx_descr(IGBCore *core, union e1000_adv_rx_desc *desc, struct NetRxPkt *pkt, - const E1000E_RSSInfo *rss_info, + const E1000E_RSSInfo *rss_info, uint16_t etqf, bool ts, uint16_t length) { memset(&desc->wb, 0, sizeof(desc->wb)); @@ -XXX,XX +XXX,XX @@ igb_write_adv_rx_descr(IGBCore *core, union e1000_adv_rx_desc *desc, desc->wb.upper.length = cpu_to_le16(length); igb_build_rx_metadata(core, pkt, pkt != NULL, - rss_info, + rss_info, etqf, ts, &desc->wb.lower.lo_dword.pkt_info, &desc->wb.lower.lo_dword.hdr_info, &desc->wb.lower.hi_dword.rss, @@ -XXX,XX +XXX,XX @@ igb_write_adv_rx_descr(IGBCore *core, union e1000_adv_rx_desc *desc, static inline void igb_write_rx_descr(IGBCore *core, union e1000_rx_desc_union *desc, -struct NetRxPkt *pkt, const E1000E_RSSInfo *rss_info, uint16_t length) + struct NetRxPkt *pkt, const E1000E_RSSInfo *rss_info, + uint16_t etqf, bool ts, uint16_t length) { if (igb_rx_use_legacy_descriptor(core)) { - igb_write_lgcy_rx_descr(core, &desc->legacy, pkt, rss_info, length); + igb_write_lgcy_rx_descr(core, &desc->legacy, pkt, rss_info, + etqf, ts, length); } else { - igb_write_adv_rx_descr(core, &desc->adv, pkt, rss_info, length); + igb_write_adv_rx_descr(core, &desc->adv, pkt, rss_info, + etqf, ts, length); } } @@ -XXX,XX +XXX,XX @@ igb_rx_descr_threshold_hit(IGBCore *core, const E1000E_RingInfo *rxi) static void igb_write_packet_to_guest(IGBCore *core, struct NetRxPkt *pkt, const E1000E_RxRing *rxr, - const E1000E_RSSInfo *rss_info) + const E1000E_RSSInfo *rss_info, + uint16_t etqf, bool ts) { PCIDevice *d; dma_addr_t base; @@ -XXX,XX +XXX,XX @@ igb_write_packet_to_guest(IGBCore *core, struct NetRxPkt *pkt, } igb_write_rx_descr(core, &desc, is_last ? core->rx_pkt : NULL, - rss_info, written); + rss_info, etqf, ts, written); igb_pci_dma_write_rx_desc(core, d, base, &desc, core->rx_desc_len); igb_ring_advance(core, rxi, core->rx_desc_len / E1000_MIN_RX_DESC_LEN); @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, size_t iov_ofs = 0; E1000E_RxRing rxr; E1000E_RSSInfo rss_info; + uint16_t etqf; + bool ts; size_t total_size; int strip_vlan_index; int i; @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, get_eth_packet_type(&buf.l2_header.eth)); net_rx_pkt_set_protocols(core->rx_pkt, iov, iovcnt, iov_ofs); - queues = igb_receive_assign(core, &buf.l2_header, size, - &rss_info, external_tx); + queues = igb_receive_assign(core, iov, iovcnt, iov_ofs, + &buf.l2_header, size, + &rss_info, &etqf, &ts, external_tx); if (!queues) { trace_e1000e_rx_flt_dropped(); return orig_size; @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, causes |= E1000_ICR_RXDW; igb_rx_fix_l4_csum(core, core->rx_pkt); - igb_write_packet_to_guest(core, core->rx_pkt, &rxr, &rss_info); + igb_write_packet_to_guest(core, core->rx_pkt, &rxr, &rss_info, etqf, ts); /* Check if receive descriptor minimum threshold hit */ if (igb_rx_descr_threshold_hit(core, rxr.i)) { @@ -XXX,XX +XXX,XX @@ static const readops igb_macreg_readops[] = { [EIAM] = igb_mac_readreg, [IVAR0 ... IVAR0 + 7] = igb_mac_readreg, igb_getreg(IVAR_MISC), + igb_getreg(TSYNCRXCFG), + [ETQF0 ... ETQF0 + 7] = igb_mac_readreg, igb_getreg(VT_CTL), [P2VMAILBOX0 ... P2VMAILBOX7] = igb_mac_readreg, [V2PMAILBOX0 ... V2PMAILBOX7] = igb_mac_vfmailbox_read, @@ -XXX,XX +XXX,XX @@ static const writeops igb_macreg_writeops[] = { [EIMS] = igb_set_eims, [IVAR0 ... IVAR0 + 7] = igb_mac_writereg, igb_putreg(IVAR_MISC), + igb_putreg(TSYNCRXCFG), + [ETQF0 ... ETQF0 + 7] = igb_mac_writereg, igb_putreg(VT_CTL), [P2VMAILBOX0 ... P2VMAILBOX7] = igb_set_pfmailbox, [V2PMAILBOX0 ... V2PMAILBOX7] = igb_set_vfmailbox, diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_regs.h +++ b/hw/net/igb_regs.h @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define E1000_DCA_TXCTRL_CPUID_SHIFT 24 /* Tx CPUID now in the last byte */ #define E1000_DCA_RXCTRL_CPUID_SHIFT 24 /* Rx CPUID now in the last byte */ +/* ETQF register bit definitions */ +#define E1000_ETQF_FILTER_ENABLE BIT(26) +#define E1000_ETQF_1588 BIT(30) +#define E1000_ETQF_IMM_INT BIT(29) +#define E1000_ETQF_QUEUE_ENABLE BIT(31) +#define E1000_ETQF_QUEUE_SHIFT 16 +#define E1000_ETQF_QUEUE_MASK 0x00070000 +#define E1000_ETQF_ETYPE_MASK 0x0000FFFF + #define E1000_DTXSWC_MAC_SPOOF_MASK 0x000000FF /* Per VF MAC spoof control */ #define E1000_DTXSWC_VLAN_SPOOF_MASK 0x0000FF00 /* Per VF VLAN spoof control */ #define E1000_DTXSWC_LLE_MASK 0x00FF0000 /* Per VF Local LB enables */ @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { #define E1000_FRTIMER 0x01048 /* Free Running Timer - RW */ #define E1000_FCRTV 0x02460 /* Flow Control Refresh Timer Value - RW */ +#define E1000_TSYNCRXCFG 0x05F50 /* Time Sync Rx Configuration - RW */ + +/* Filtering Registers */ +#define E1000_SAQF(_n) (0x5980 + 4 * (_n)) +#define E1000_DAQF(_n) (0x59A0 + 4 * (_n)) +#define E1000_SPQF(_n) (0x59C0 + 4 * (_n)) +#define E1000_FTQF(_n) (0x59E0 + 4 * (_n)) +#define E1000_SAQF0 E1000_SAQF(0) +#define E1000_DAQF0 E1000_DAQF(0) +#define E1000_SPQF0 E1000_SPQF(0) +#define E1000_FTQF0 E1000_FTQF(0) +#define E1000_SYNQF(_n) (0x055FC + (4 * (_n))) /* SYN Packet Queue Fltr */ +#define E1000_ETQF(_n) (0x05CB0 + (4 * (_n))) /* EType Queue Fltr */ + #define E1000_RQDPC(_n) (0x0C030 + ((_n) * 0x40)) #define E1000_RXPBS 0x02404 /* Rx Packet Buffer Size - RW */ -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 7 +++++++ hw/net/igb_regs.h | 3 +++ 2 files changed, 10 insertions(+) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, tx->ctx[idx].vlan_macip_lens >> IGB_TX_FLAGS_VLAN_SHIFT, !!(tx->first_cmd_type_len & E1000_TXD_CMD_VLE)); + if ((tx->first_cmd_type_len & E1000_ADVTXD_MAC_TSTAMP) && + (core->mac[TSYNCTXCTL] & E1000_TSYNCTXCTL_ENABLED) && + !(core->mac[TSYNCTXCTL] & E1000_TSYNCTXCTL_VALID)) { + core->mac[TSYNCTXCTL] |= E1000_TSYNCTXCTL_VALID; + e1000x_timestamp(core->mac, core->timadj, TXSTMPL, TXSTMPH); + } + if (igb_tx_pkt_send(core, tx, queue_index)) { igb_on_tx_done_update_stats(core, tx->tx_pkt, queue_index); } diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_regs.h +++ b/hw/net/igb_regs.h @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { /* E1000_EITR_CNT_IGNR is only for 82576 and newer */ #define E1000_EITR_CNT_IGNR 0x80000000 /* Don't reset counters on write */ +#define E1000_TSYNCTXCTL_VALID 0x00000001 /* tx timestamp valid */ +#define E1000_TSYNCTXCTL_ENABLED 0x00000010 /* enable tx timestampping */ + /* PCI Express Control */ #define E1000_GCR_CMPL_TMOUT_MASK 0x0000F000 #define E1000_GCR_CMPL_TMOUT_10ms 0x00001000 -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> In MSI-X mode, if there are interrupts already notified but not cleared and a new interrupt arrives, e1000e incorrectly notifies the notified ones again along with the new one. To fix this issue, replace e1000e_update_interrupt_state() with two new functions: e1000e_raise_interrupts() and e1000e_lower_interrupts(). These functions don't only raise or lower interrupts, but it also performs register writes which updates the interrupt state. Before it performs a register write, these function determines the interrupts already raised, and compares with the interrupts raised after the register write to determine the interrupts to notify. The introduction of these functions made tracepoints which assumes that the caller of e1000e_update_interrupt_state() performs register writes obsolete. These tracepoints are now removed, and alternative ones are added to the new functions. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/e1000e_core.c | 153 ++++++++++++++++++++++----------------------------- hw/net/e1000e_core.h | 2 - hw/net/trace-events | 2 + 3 files changed, 69 insertions(+), 88 deletions(-) diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.c +++ b/hw/net/e1000e_core.c @@ -XXX,XX +XXX,XX @@ e1000e_intrmgr_on_throttling_timer(void *opaque) timer->running = false; - if (msi_enabled(timer->core->owner)) { - trace_e1000e_irq_msi_notify_postponed(); - /* Clear msi_causes_pending to fire MSI eventually */ - timer->core->msi_causes_pending = 0; - e1000e_set_interrupt_cause(timer->core, 0); - } else { - trace_e1000e_irq_legacy_notify_postponed(); - e1000e_set_interrupt_cause(timer->core, 0); + if (timer->core->mac[IMS] & timer->core->mac[ICR]) { + if (msi_enabled(timer->core->owner)) { + trace_e1000e_irq_msi_notify_postponed(); + msi_notify(timer->core->owner, 0); + } else { + trace_e1000e_irq_legacy_notify_postponed(); + e1000e_raise_legacy_irq(timer->core); + } } } @@ -XXX,XX +XXX,XX @@ static void e1000e_intrmgr_fire_all_timers(E1000ECore *core) { int i; - uint32_t val = e1000e_intmgr_collect_delayed_causes(core); - - trace_e1000e_irq_adding_delayed_causes(val, core->mac[ICR]); - core->mac[ICR] |= val; if (core->itr.running) { timer_del(core->itr.timer); @@ -XXX,XX +XXX,XX @@ void(*e1000e_phyreg_writeops[E1000E_PHY_PAGES][E1000E_PHY_PAGE_SIZE]) } }; -static inline void -e1000e_clear_ims_bits(E1000ECore *core, uint32_t bits) -{ - trace_e1000e_irq_clear_ims(bits, core->mac[IMS], core->mac[IMS] & ~bits); - core->mac[IMS] &= ~bits; -} - static inline bool e1000e_postpone_interrupt(E1000IntrDelayTimer *timer) { @@ -XXX,XX +XXX,XX @@ e1000e_msix_notify_one(E1000ECore *core, uint32_t cause, uint32_t int_cfg) effective_eiac = core->mac[EIAC] & cause; core->mac[ICR] &= ~effective_eiac; - core->msi_causes_pending &= ~effective_eiac; if (!(core->mac[CTRL_EXT] & E1000_CTRL_EXT_IAME)) { core->mac[IMS] &= ~effective_eiac; @@ -XXX,XX +XXX,XX @@ e1000e_fix_icr_asserted(E1000ECore *core) trace_e1000e_irq_fix_icr_asserted(core->mac[ICR]); } -static void -e1000e_send_msi(E1000ECore *core, bool msix) +static void e1000e_raise_interrupts(E1000ECore *core, + size_t index, uint32_t causes) { - uint32_t causes = core->mac[ICR] & core->mac[IMS] & ~E1000_ICR_ASSERTED; - - core->msi_causes_pending &= causes; - causes ^= core->msi_causes_pending; - if (causes == 0) { - return; - } - core->msi_causes_pending |= causes; + bool is_msix = msix_enabled(core->owner); + uint32_t old_causes = core->mac[IMS] & core->mac[ICR]; + uint32_t raised_causes; - if (msix) { - e1000e_msix_notify(core, causes); - } else { - if (!e1000e_itr_should_postpone(core)) { - trace_e1000e_irq_msi_notify(causes); - msi_notify(core->owner, 0); - } - } -} + trace_e1000e_irq_set(index << 2, + core->mac[index], core->mac[index] | causes); -static void -e1000e_update_interrupt_state(E1000ECore *core) -{ - bool interrupts_pending; - bool is_msix = msix_enabled(core->owner); + core->mac[index] |= causes; /* Set ICR[OTHER] for MSI-X */ if (is_msix) { @@ -XXX,XX +XXX,XX @@ e1000e_update_interrupt_state(E1000ECore *core) */ core->mac[ICS] = core->mac[ICR]; - interrupts_pending = (core->mac[IMS] & core->mac[ICR]) ? true : false; - if (!interrupts_pending) { - core->msi_causes_pending = 0; - } - trace_e1000e_irq_pending_interrupts(core->mac[ICR] & core->mac[IMS], core->mac[ICR], core->mac[IMS]); - if (is_msix || msi_enabled(core->owner)) { - if (interrupts_pending) { - e1000e_send_msi(core, is_msix); - } - } else { - if (interrupts_pending) { - if (!e1000e_itr_should_postpone(core)) { - e1000e_raise_legacy_irq(core); - } + raised_causes = core->mac[IMS] & core->mac[ICR] & ~old_causes; + if (!raised_causes) { + return; + } + + if (is_msix) { + e1000e_msix_notify(core, raised_causes & ~E1000_ICR_ASSERTED); + } else if (!e1000e_itr_should_postpone(core)) { + if (msi_enabled(core->owner)) { + trace_e1000e_irq_msi_notify(raised_causes); + msi_notify(core->owner, 0); } else { - e1000e_lower_legacy_irq(core); + e1000e_raise_legacy_irq(core); } } } -static void -e1000e_set_interrupt_cause(E1000ECore *core, uint32_t val) +static void e1000e_lower_interrupts(E1000ECore *core, + size_t index, uint32_t causes) { - trace_e1000e_irq_set_cause_entry(val, core->mac[ICR]); + trace_e1000e_irq_clear(index << 2, + core->mac[index], core->mac[index] & ~causes); - val |= e1000e_intmgr_collect_delayed_causes(core); - core->mac[ICR] |= val; + core->mac[index] &= ~causes; - trace_e1000e_irq_set_cause_exit(val, core->mac[ICR]); + /* + * Make sure ICR and ICS registers have the same value. + * The spec says that the ICS register is write-only. However in practice, + * on real hardware ICS is readable, and for reads it has the same value as + * ICR (except that ICS does not have the clear on read behaviour of ICR). + * + * The VxWorks PRO/1000 driver uses this behaviour. + */ + core->mac[ICS] = core->mac[ICR]; + + trace_e1000e_irq_pending_interrupts(core->mac[ICR] & core->mac[IMS], + core->mac[ICR], core->mac[IMS]); - e1000e_update_interrupt_state(core); + if (!(core->mac[IMS] & core->mac[ICR]) && + !msix_enabled(core->owner) && !msi_enabled(core->owner)) { + e1000e_lower_legacy_irq(core); + } +} + +static void +e1000e_set_interrupt_cause(E1000ECore *core, uint32_t val) +{ + val |= e1000e_intmgr_collect_delayed_causes(core); + e1000e_raise_interrupts(core, ICR, val); } static inline void @@ -XXX,XX +XXX,XX @@ e1000e_set_ics(E1000ECore *core, int index, uint32_t val) static void e1000e_set_icr(E1000ECore *core, int index, uint32_t val) { - uint32_t icr = 0; if ((core->mac[ICR] & E1000_ICR_ASSERTED) && (core->mac[CTRL_EXT] & E1000_CTRL_EXT_IAME)) { trace_e1000e_irq_icr_process_iame(); - e1000e_clear_ims_bits(core, core->mac[IAM]); + e1000e_lower_interrupts(core, IMS, core->mac[IAM]); } - icr = core->mac[ICR] & ~val; /* * Windows driver expects that the "receive overrun" bit and other * ones to be cleared when the "Other" bit (#24) is cleared. */ - icr = (val & E1000_ICR_OTHER) ? (icr & ~E1000_ICR_OTHER_CAUSES) : icr; - trace_e1000e_irq_icr_write(val, core->mac[ICR], icr); - core->mac[ICR] = icr; - e1000e_update_interrupt_state(core); + if (val & E1000_ICR_OTHER) { + val |= E1000_ICR_OTHER_CAUSES; + } + e1000e_lower_interrupts(core, ICR, val); } static void e1000e_set_imc(E1000ECore *core, int index, uint32_t val) { trace_e1000e_irq_ims_clear_set_imc(val); - e1000e_clear_ims_bits(core, val); - e1000e_update_interrupt_state(core); + e1000e_lower_interrupts(core, IMS, val); } static void @@ -XXX,XX +XXX,XX @@ e1000e_set_ims(E1000ECore *core, int index, uint32_t val) uint32_t valid_val = val & ims_valid_mask; - trace_e1000e_irq_set_ims(val, core->mac[IMS], core->mac[IMS] | valid_val); - core->mac[IMS] |= valid_val; - if ((valid_val & ims_ext_mask) && (core->mac[CTRL_EXT] & E1000_CTRL_EXT_PBA_CLR) && msix_enabled(core->owner)) { @@ -XXX,XX +XXX,XX @@ e1000e_set_ims(E1000ECore *core, int index, uint32_t val) e1000e_intrmgr_fire_all_timers(core); } - e1000e_update_interrupt_state(core); + e1000e_raise_interrupts(core, IMS, valid_val); } static void @@ -XXX,XX +XXX,XX @@ static uint32_t e1000e_mac_icr_read(E1000ECore *core, int index) { uint32_t ret = core->mac[ICR]; - trace_e1000e_irq_icr_read_entry(ret); if (core->mac[IMS] == 0) { trace_e1000e_irq_icr_clear_zero_ims(); - core->mac[ICR] = 0; + e1000e_lower_interrupts(core, ICR, 0xffffffff); } if (!msix_enabled(core->owner)) { trace_e1000e_irq_icr_clear_nonmsix_icr_read(); - core->mac[ICR] = 0; + e1000e_lower_interrupts(core, ICR, 0xffffffff); } if ((core->mac[ICR] & E1000_ICR_ASSERTED) && (core->mac[CTRL_EXT] & E1000_CTRL_EXT_IAME)) { trace_e1000e_irq_icr_clear_iame(); - core->mac[ICR] = 0; + e1000e_lower_interrupts(core, ICR, 0xffffffff); trace_e1000e_irq_icr_process_iame(); - e1000e_clear_ims_bits(core, core->mac[IAM]); + e1000e_lower_interrupts(core, IMS, core->mac[IAM]); } - trace_e1000e_irq_icr_read_exit(core->mac[ICR]); - e1000e_update_interrupt_state(core); return ret; } diff --git a/hw/net/e1000e_core.h b/hw/net/e1000e_core.h index XXXXXXX..XXXXXXX 100644 --- a/hw/net/e1000e_core.h +++ b/hw/net/e1000e_core.h @@ -XXX,XX +XXX,XX @@ struct E1000Core { PCIDevice *owner; void (*owner_start_recv)(PCIDevice *d); - uint32_t msi_causes_pending; - int64_t timadj; }; diff --git a/hw/net/trace-events b/hw/net/trace-events index XXXXXXX..XXXXXXX 100644 --- a/hw/net/trace-events +++ b/hw/net/trace-events @@ -XXX,XX +XXX,XX @@ e1000e_irq_msix_notify_postponed_vec(int idx) "Sending MSI-X postponed by EITR[% e1000e_irq_legacy_notify(bool level) "IRQ line state: %d" e1000e_irq_msix_notify_vec(uint32_t vector) "MSI-X notify vector 0x%x" e1000e_irq_postponed_by_xitr(uint32_t reg) "Interrupt postponed by [E]ITR register 0x%x" +e1000e_irq_clear(uint32_t offset, uint32_t old, uint32_t new) "Clearing interrupt register 0x%x: 0x%x --> 0x%x" +e1000e_irq_set(uint32_t offset, uint32_t old, uint32_t new) "Setting interrupt register 0x%x: 0x%x --> 0x%x" e1000e_irq_clear_ims(uint32_t bits, uint32_t old_ims, uint32_t new_ims) "Clearing IMS bits 0x%x: 0x%x --> 0x%x" e1000e_irq_set_ims(uint32_t bits, uint32_t old_ims, uint32_t new_ims) "Setting IMS bits 0x%x: 0x%x --> 0x%x" e1000e_irq_fix_icr_asserted(uint32_t new_val) "ICR_ASSERTED bit fixed: 0x%x" -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> This follows the corresponding change for e1000e. This fixes: tests/avocado/netdev-ethtool.py:NetDevEthtool.test_igb Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 201 +++++++++------------ hw/net/trace-events | 11 +- scripts/ci/org.centos/stream/8/x86_64/test-avocado | 1 + tests/avocado/netdev-ethtool.py | 4 - 4 files changed, 87 insertions(+), 130 deletions(-) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ static ssize_t igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, bool has_vnet, bool *external_tx); -static inline void -igb_set_interrupt_cause(IGBCore *core, uint32_t val); - -static void igb_update_interrupt_state(IGBCore *core); +static void igb_raise_interrupts(IGBCore *core, size_t index, uint32_t causes); static void igb_reset(IGBCore *core, bool sw); static inline void @@ -XXX,XX +XXX,XX @@ igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) } if (eic) { - core->mac[EICR] |= eic; - igb_set_interrupt_cause(core, E1000_ICR_TXDW); + igb_raise_interrupts(core, EICR, eic); + igb_raise_interrupts(core, ICR, E1000_ICR_TXDW); } net_tx_pkt_reset(txr->tx->tx_pkt, net_tx_pkt_unmap_frag_pci, d); @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, { uint16_t queues = 0; uint32_t causes = 0; + uint32_t ecauses = 0; union { L2Header l2_header; uint8_t octets[ETH_ZLEN]; @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, causes |= E1000_ICS_RXDMT0; } - core->mac[EICR] |= igb_rx_wb_eic(core, rxr.i->idx); + ecauses |= igb_rx_wb_eic(core, rxr.i->idx); trace_e1000e_rx_written_to_guest(rxr.i->idx); } trace_e1000e_rx_interrupt_set(causes); - igb_set_interrupt_cause(core, causes); + igb_raise_interrupts(core, EICR, ecauses); + igb_raise_interrupts(core, ICR, causes); return orig_size; } @@ -XXX,XX +XXX,XX @@ void igb_core_set_link_status(IGBCore *core) } if (core->mac[STATUS] != old_status) { - igb_set_interrupt_cause(core, E1000_ICR_LSC); + igb_raise_interrupts(core, ICR, E1000_ICR_LSC); } } @@ -XXX,XX +XXX,XX @@ igb_set_rx_control(IGBCore *core, int index, uint32_t val) } } -static inline void -igb_clear_ims_bits(IGBCore *core, uint32_t bits) -{ - trace_e1000e_irq_clear_ims(bits, core->mac[IMS], core->mac[IMS] & ~bits); - core->mac[IMS] &= ~bits; -} - static inline bool igb_postpone_interrupt(IGBIntrDelayTimer *timer) { @@ -XXX,XX +XXX,XX @@ igb_eitr_should_postpone(IGBCore *core, int idx) return igb_postpone_interrupt(&core->eitr[idx]); } -static void igb_send_msix(IGBCore *core) +static void igb_send_msix(IGBCore *core, uint32_t causes) { - uint32_t causes = core->mac[EICR] & core->mac[EIMS]; int vector; for (vector = 0; vector < IGB_INTR_NUM; ++vector) { @@ -XXX,XX +XXX,XX @@ igb_fix_icr_asserted(IGBCore *core) trace_e1000e_irq_fix_icr_asserted(core->mac[ICR]); } -static void -igb_update_interrupt_state(IGBCore *core) +static void igb_raise_interrupts(IGBCore *core, size_t index, uint32_t causes) { - uint32_t icr; - uint32_t causes; + uint32_t old_causes = core->mac[ICR] & core->mac[IMS]; + uint32_t old_ecauses = core->mac[EICR] & core->mac[EIMS]; + uint32_t raised_causes; + uint32_t raised_ecauses; uint32_t int_alloc; - icr = core->mac[ICR] & core->mac[IMS]; + trace_e1000e_irq_set(index << 2, + core->mac[index], core->mac[index] | causes); + + core->mac[index] |= causes; if (core->mac[GPIE] & E1000_GPIE_MSIX_MODE) { - if (icr) { - causes = 0; - if (icr & E1000_ICR_DRSTA) { - int_alloc = core->mac[IVAR_MISC] & 0xff; - if (int_alloc & E1000_IVAR_VALID) { - causes |= BIT(int_alloc & 0x1f); - } + raised_causes = core->mac[ICR] & core->mac[IMS] & ~old_causes; + + if (raised_causes & E1000_ICR_DRSTA) { + int_alloc = core->mac[IVAR_MISC] & 0xff; + if (int_alloc & E1000_IVAR_VALID) { + core->mac[EICR] |= BIT(int_alloc & 0x1f); } - /* Check if other bits (excluding the TCP Timer) are enabled. */ - if (icr & ~E1000_ICR_DRSTA) { - int_alloc = (core->mac[IVAR_MISC] >> 8) & 0xff; - if (int_alloc & E1000_IVAR_VALID) { - causes |= BIT(int_alloc & 0x1f); - } - trace_e1000e_irq_add_msi_other(core->mac[EICR]); + } + /* Check if other bits (excluding the TCP Timer) are enabled. */ + if (raised_causes & ~E1000_ICR_DRSTA) { + int_alloc = (core->mac[IVAR_MISC] >> 8) & 0xff; + if (int_alloc & E1000_IVAR_VALID) { + core->mac[EICR] |= BIT(int_alloc & 0x1f); } - core->mac[EICR] |= causes; } - if ((core->mac[EICR] & core->mac[EIMS])) { - igb_send_msix(core); + raised_ecauses = core->mac[EICR] & core->mac[EIMS] & ~old_ecauses; + if (!raised_ecauses) { + return; } + + igb_send_msix(core, raised_ecauses); } else { igb_fix_icr_asserted(core); - if (icr) { - core->mac[EICR] |= (icr & E1000_ICR_DRSTA) | E1000_EICR_OTHER; - } else { - core->mac[EICR] &= ~E1000_EICR_OTHER; + raised_causes = core->mac[ICR] & core->mac[IMS] & ~old_causes; + if (!raised_causes) { + return; } - trace_e1000e_irq_pending_interrupts(core->mac[ICR] & core->mac[IMS], - core->mac[ICR], core->mac[IMS]); + core->mac[EICR] |= (raised_causes & E1000_ICR_DRSTA) | E1000_EICR_OTHER; if (msix_enabled(core->owner)) { - if (icr) { - trace_e1000e_irq_msix_notify_vec(0); - msix_notify(core->owner, 0); - } + trace_e1000e_irq_msix_notify_vec(0); + msix_notify(core->owner, 0); } else if (msi_enabled(core->owner)) { - if (icr) { - msi_notify(core->owner, 0); - } + trace_e1000e_irq_msi_notify(raised_causes); + msi_notify(core->owner, 0); } else { - if (icr) { - igb_raise_legacy_irq(core); - } else { - igb_lower_legacy_irq(core); - } + igb_raise_legacy_irq(core); } } } -static void -igb_set_interrupt_cause(IGBCore *core, uint32_t val) +static void igb_lower_interrupts(IGBCore *core, size_t index, uint32_t causes) { - trace_e1000e_irq_set_cause_entry(val, core->mac[ICR]); + trace_e1000e_irq_clear(index << 2, + core->mac[index], core->mac[index] & ~causes); + + core->mac[index] &= ~causes; - core->mac[ICR] |= val; + trace_e1000e_irq_pending_interrupts(core->mac[ICR] & core->mac[IMS], + core->mac[ICR], core->mac[IMS]); - trace_e1000e_irq_set_cause_exit(val, core->mac[ICR]); + if (!(core->mac[ICR] & core->mac[IMS]) && + !(core->mac[GPIE] & E1000_GPIE_MSIX_MODE)) { + core->mac[EICR] &= ~E1000_EICR_OTHER; - igb_update_interrupt_state(core); + if (!msix_enabled(core->owner) && !msi_enabled(core->owner)) { + igb_lower_legacy_irq(core); + } + } } static void igb_set_eics(IGBCore *core, int index, uint32_t val) { bool msix = !!(core->mac[GPIE] & E1000_GPIE_MSIX_MODE); + uint32_t mask = msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK; trace_igb_irq_write_eics(val, msix); - - core->mac[EICS] |= - val & (msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK); - - /* - * TODO: Move to igb_update_interrupt_state if EICS is modified in other - * places. - */ - core->mac[EICR] = core->mac[EICS]; - - igb_update_interrupt_state(core); + igb_raise_interrupts(core, EICR, val & mask); } static void igb_set_eims(IGBCore *core, int index, uint32_t val) { bool msix = !!(core->mac[GPIE] & E1000_GPIE_MSIX_MODE); + uint32_t mask = msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK; trace_igb_irq_write_eims(val, msix); - - core->mac[EIMS] |= - val & (msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK); - - igb_update_interrupt_state(core); + igb_raise_interrupts(core, EIMS, val & mask); } static void mailbox_interrupt_to_vf(IGBCore *core, uint16_t vfn) { uint32_t ent = core->mac[VTIVAR_MISC + vfn]; + uint32_t causes; if ((ent & E1000_IVAR_VALID)) { - core->mac[EICR] |= (ent & 0x3) << (22 - vfn * IGBVF_MSIX_VEC_NUM); - igb_update_interrupt_state(core); + causes = (ent & 0x3) << (22 - vfn * IGBVF_MSIX_VEC_NUM); + igb_raise_interrupts(core, EICR, causes); } } static void mailbox_interrupt_to_pf(IGBCore *core) { - igb_set_interrupt_cause(core, E1000_ICR_VMMB); + igb_raise_interrupts(core, ICR, E1000_ICR_VMMB); } static void igb_set_pfmailbox(IGBCore *core, int index, uint32_t val) @@ -XXX,XX +XXX,XX @@ static void igb_w1c(IGBCore *core, int index, uint32_t val) static void igb_set_eimc(IGBCore *core, int index, uint32_t val) { bool msix = !!(core->mac[GPIE] & E1000_GPIE_MSIX_MODE); + uint32_t mask = msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK; - /* Interrupts are disabled via a write to EIMC and reflected in EIMS. */ - core->mac[EIMS] &= - ~(val & (msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK)); + trace_igb_irq_write_eimc(val, msix); - trace_igb_irq_write_eimc(val, core->mac[EIMS], msix); - igb_update_interrupt_state(core); + /* Interrupts are disabled via a write to EIMC and reflected in EIMS. */ + igb_lower_interrupts(core, EIMS, val & mask); } static void igb_set_eiac(IGBCore *core, int index, uint32_t val) @@ -XXX,XX +XXX,XX @@ static void igb_set_eicr(IGBCore *core, int index, uint32_t val) * TODO: In IOV mode, only bit zero of this vector is available for the PF * function. */ - core->mac[EICR] &= - ~(val & (msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK)); + uint32_t mask = msix ? E1000_EICR_MSIX_MASK : E1000_EICR_LEGACY_MASK; trace_igb_irq_write_eicr(val, msix); - igb_update_interrupt_state(core); + igb_lower_interrupts(core, EICR, val & mask); } static void igb_set_vtctrl(IGBCore *core, int index, uint32_t val) @@ -XXX,XX +XXX,XX @@ igb_autoneg_timer(void *opaque) igb_update_flowctl_status(core); /* signal link status change to the guest */ - igb_set_interrupt_cause(core, E1000_ICR_LSC); + igb_raise_interrupts(core, ICR, E1000_ICR_LSC); } } @@ -XXX,XX +XXX,XX @@ igb_set_mdic(IGBCore *core, int index, uint32_t val) core->mac[MDIC] = val | E1000_MDIC_READY; if (val & E1000_MDIC_INT_EN) { - igb_set_interrupt_cause(core, E1000_ICR_MDAC); + igb_raise_interrupts(core, ICR, E1000_ICR_MDAC); } } @@ -XXX,XX +XXX,XX @@ static void igb_set_ics(IGBCore *core, int index, uint32_t val) { trace_e1000e_irq_write_ics(val); - igb_set_interrupt_cause(core, val); + igb_raise_interrupts(core, ICR, val); } static void igb_set_imc(IGBCore *core, int index, uint32_t val) { trace_e1000e_irq_ims_clear_set_imc(val); - igb_clear_ims_bits(core, val); - igb_update_interrupt_state(core); + igb_lower_interrupts(core, IMS, val); } static void igb_set_ims(IGBCore *core, int index, uint32_t val) { - uint32_t valid_val = val & 0x77D4FBFD; - - trace_e1000e_irq_set_ims(val, core->mac[IMS], core->mac[IMS] | valid_val); - core->mac[IMS] |= valid_val; - igb_update_interrupt_state(core); + igb_raise_interrupts(core, IMS, val & 0x77D4FBFD); } -static void igb_commit_icr(IGBCore *core) +static void igb_nsicr(IGBCore *core) { /* * If GPIE.NSICR = 0, then the clear of IMS will occur only if at @@ -XXX,XX +XXX,XX @@ static void igb_commit_icr(IGBCore *core) */ if ((core->mac[GPIE] & E1000_GPIE_NSICR) || (core->mac[IMS] && (core->mac[ICR] & E1000_ICR_INT_ASSERTED))) { - igb_clear_ims_bits(core, core->mac[IAM]); + igb_lower_interrupts(core, IMS, core->mac[IAM]); } - - igb_update_interrupt_state(core); } static void igb_set_icr(IGBCore *core, int index, uint32_t val) { - uint32_t icr = core->mac[ICR] & ~val; - - trace_igb_irq_icr_write(val, core->mac[ICR], icr); - core->mac[ICR] = icr; - igb_commit_icr(core); + igb_nsicr(core); + igb_lower_interrupts(core, ICR, val); } static uint32_t @@ -XXX,XX +XXX,XX @@ static uint32_t igb_mac_icr_read(IGBCore *core, int index) { uint32_t ret = core->mac[ICR]; - trace_e1000e_irq_icr_read_entry(ret); if (core->mac[GPIE] & E1000_GPIE_NSICR) { trace_igb_irq_icr_clear_gpie_nsicr(); - core->mac[ICR] = 0; + igb_lower_interrupts(core, ICR, 0xffffffff); } else if (core->mac[IMS] == 0) { trace_e1000e_irq_icr_clear_zero_ims(); - core->mac[ICR] = 0; + igb_lower_interrupts(core, ICR, 0xffffffff); } else if (!msix_enabled(core->owner)) { trace_e1000e_irq_icr_clear_nonmsix_icr_read(); - core->mac[ICR] = 0; + igb_lower_interrupts(core, ICR, 0xffffffff); } - trace_e1000e_irq_icr_read_exit(core->mac[ICR]); - igb_commit_icr(core); + igb_nsicr(core); return ret; } diff --git a/hw/net/trace-events b/hw/net/trace-events index XXXXXXX..XXXXXXX 100644 --- a/hw/net/trace-events +++ b/hw/net/trace-events @@ -XXX,XX +XXX,XX @@ e1000e_irq_msix_notify_vec(uint32_t vector) "MSI-X notify vector 0x%x" e1000e_irq_postponed_by_xitr(uint32_t reg) "Interrupt postponed by [E]ITR register 0x%x" e1000e_irq_clear(uint32_t offset, uint32_t old, uint32_t new) "Clearing interrupt register 0x%x: 0x%x --> 0x%x" e1000e_irq_set(uint32_t offset, uint32_t old, uint32_t new) "Setting interrupt register 0x%x: 0x%x --> 0x%x" -e1000e_irq_clear_ims(uint32_t bits, uint32_t old_ims, uint32_t new_ims) "Clearing IMS bits 0x%x: 0x%x --> 0x%x" -e1000e_irq_set_ims(uint32_t bits, uint32_t old_ims, uint32_t new_ims) "Setting IMS bits 0x%x: 0x%x --> 0x%x" e1000e_irq_fix_icr_asserted(uint32_t new_val) "ICR_ASSERTED bit fixed: 0x%x" e1000e_irq_add_msi_other(uint32_t new_val) "ICR_OTHER bit added: 0x%x" e1000e_irq_pending_interrupts(uint32_t pending, uint32_t icr, uint32_t ims) "ICR PENDING: 0x%x (ICR: 0x%x, IMS: 0x%x)" -e1000e_irq_set_cause_entry(uint32_t val, uint32_t icr) "Going to set IRQ cause 0x%x, ICR: 0x%x" -e1000e_irq_set_cause_exit(uint32_t val, uint32_t icr) "Set IRQ cause 0x%x, ICR: 0x%x" -e1000e_irq_icr_write(uint32_t bits, uint32_t old_icr, uint32_t new_icr) "Clearing ICR bits 0x%x: 0x%x --> 0x%x" e1000e_irq_write_ics(uint32_t val) "Adding ICR bits 0x%x" e1000e_irq_icr_process_iame(void) "Clearing IMS bits due to IAME" e1000e_irq_read_ics(uint32_t ics) "Current ICS: 0x%x" e1000e_irq_read_ims(uint32_t ims) "Current IMS: 0x%x" e1000e_irq_icr_clear_nonmsix_icr_read(void) "Clearing ICR on read due to non MSI-X int" -e1000e_irq_icr_read_entry(uint32_t icr) "Starting ICR read. Current ICR: 0x%x" -e1000e_irq_icr_read_exit(uint32_t icr) "Ending ICR read. Current ICR: 0x%x" e1000e_irq_icr_clear_zero_ims(void) "Clearing ICR on read due to zero IMS" e1000e_irq_icr_clear_iame(void) "Clearing ICR on read due to IAME" e1000e_irq_iam_clear_eiame(uint32_t iam, uint32_t cause) "Clearing IMS due to EIAME, IAM: 0x%X, cause: 0x%X" @@ -XXX,XX +XXX,XX @@ e1000e_irq_tidv_fpd_not_running(void) "FPD written while TIDV was not running" e1000e_irq_eitr_set(uint32_t eitr_num, uint32_t val) "EITR[%u] = %u" e1000e_irq_itr_set(uint32_t val) "ITR = %u" e1000e_irq_fire_all_timers(uint32_t val) "Firing all delay/throttling timers on all interrupts enable (0x%X written to IMS)" -e1000e_irq_adding_delayed_causes(uint32_t val, uint32_t icr) "Merging delayed causes 0x%X to ICR 0x%X" e1000e_irq_msix_pending_clearing(uint32_t cause, uint32_t int_cfg, uint32_t vec) "Clearing MSI-X pending bit for cause 0x%x, IVAR config 0x%x, vector %u" e1000e_wrn_msix_vec_wrong(uint32_t cause, uint32_t cfg) "Invalid configuration for cause 0x%x: 0x%x" @@ -XXX,XX +XXX,XX @@ igb_rx_desc_buff_write(uint64_t addr, uint16_t offset, const void* source, uint3 igb_rx_metadata_rss(uint32_t rss) "RSS data: 0x%X" igb_irq_icr_clear_gpie_nsicr(void) "Clearing ICR on read due to GPIE.NSICR enabled" -igb_irq_icr_write(uint32_t bits, uint32_t old_icr, uint32_t new_icr) "Clearing ICR bits 0x%x: 0x%x --> 0x%x" igb_irq_set_iam(uint32_t icr) "Update IAM: 0x%x" igb_irq_read_iam(uint32_t icr) "Current IAM: 0x%x" igb_irq_write_eics(uint32_t val, bool msix) "Update EICS: 0x%x MSI-X: %d" igb_irq_write_eims(uint32_t val, bool msix) "Update EIMS: 0x%x MSI-X: %d" -igb_irq_write_eimc(uint32_t val, uint32_t eims, bool msix) "Update EIMC: 0x%x EIMS: 0x%x MSI-X: %d" +igb_irq_write_eimc(uint32_t val, bool msix) "Update EIMC: 0x%x MSI-X: %d" igb_irq_write_eiac(uint32_t val) "Update EIAC: 0x%x" igb_irq_write_eiam(uint32_t val, bool msix) "Update EIAM: 0x%x MSI-X: %d" igb_irq_write_eicr(uint32_t val, bool msix) "Update EICR: 0x%x MSI-X: %d" diff --git a/scripts/ci/org.centos/stream/8/x86_64/test-avocado b/scripts/ci/org.centos/stream/8/x86_64/test-avocado index XXXXXXX..XXXXXXX 100755 --- a/scripts/ci/org.centos/stream/8/x86_64/test-avocado +++ b/scripts/ci/org.centos/stream/8/x86_64/test-avocado @@ -XXX,XX +XXX,XX @@ make get-vm-images tests/avocado/cpu_queries.py:QueryCPUModelExpansion.test \ tests/avocado/empty_cpu_model.py:EmptyCPUModel.test \ tests/avocado/hotplug_cpu.py:HotPlugCPU.test \ + tests/avocado/netdev-ethtool.py:NetDevEthtool.test_igb \ tests/avocado/netdev-ethtool.py:NetDevEthtool.test_igb_nomsi \ tests/avocado/info_usernet.py:InfoUsernet.test_hostfwd \ tests/avocado/intel_iommu.py:IntelIOMMU.test_intel_iommu \ diff --git a/tests/avocado/netdev-ethtool.py b/tests/avocado/netdev-ethtool.py index XXXXXXX..XXXXXXX 100644 --- a/tests/avocado/netdev-ethtool.py +++ b/tests/avocado/netdev-ethtool.py @@ -XXX,XX +XXX,XX @@ def common_test_code(self, netdev, extra_args=None): # no need to gracefully shutdown, just finish self.vm.kill() - # Skip testing for MSI for now. Allegedly it was fixed by: - # 28e96556ba (igb: Allocate MSI-X vector when testing) - # but I'm seeing oops in the kernel - @skip("Kernel bug with MSI enabled") def test_igb(self): """ :avocado: tags=device:igb -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> For GPIE.NSICR, Section 7.3.2.1.2 says: > ICR bits are cleared on register read. If GPIE.NSICR = 0b, then the > clear on read occurs only if no bit is set in the IMS or at least one > bit is set in the IMS and there is a true interrupt as reflected in > ICR.INTA. e1000e does similar though it checks for CTRL_EXT.IAME, which does not exist on igb. Suggested-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/igb_core.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/igb_core.c +++ b/hw/net/igb_core.c @@ -XXX,XX +XXX,XX @@ igb_mac_icr_read(IGBCore *core, int index) } else if (core->mac[IMS] == 0) { trace_e1000e_irq_icr_clear_zero_ims(); igb_lower_interrupts(core, ICR, 0xffffffff); + } else if (core->mac[ICR] & E1000_ICR_INT_ASSERTED) { + igb_lower_interrupts(core, ICR, 0xffffffff); } else if (!msix_enabled(core->owner)) { trace_e1000e_irq_icr_clear_nonmsix_icr_read(); igb_lower_interrupts(core, ICR, 0xffffffff); -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> vmxnet3 has no dependency on PC, and VMware Fusion actually makes it available on Apple Silicon according to: https://kb.vmware.com/s/article/90364 Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/hw/net/Kconfig b/hw/net/Kconfig index XXXXXXX..XXXXXXX 100644 --- a/hw/net/Kconfig +++ b/hw/net/Kconfig @@ -XXX,XX +XXX,XX @@ config RTL8139_PCI config VMXNET3_PCI bool - default y if PCI_DEVICES && PC_PCI + default y if PCI_DEVICES depends on PCI config SMC91C111 -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> I have made significant changes for network packet abstractions so add me as a reviewer. Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Signed-off-by: Jason Wang <jasowang@redhat.com> --- MAINTAINERS | 1 + 1 file changed, 1 insertion(+) diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: tests/qtest/fuzz-megasas-test.c Network packet abstractions M: Dmitry Fleytman <dmitry.fleytman@gmail.com> +R: Akihiko Odaki <akihiko.odaki@daynix.com> S: Maintained F: include/net/eth.h F: net/eth.c -- 2.7.4 From: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- docs/system/devices/igb.rst | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/docs/system/devices/igb.rst b/docs/system/devices/igb.rst index XXXXXXX..XXXXXXX 100644 --- a/docs/system/devices/igb.rst +++ b/docs/system/devices/igb.rst @@ -XXX,XX +XXX,XX @@ Limitations =========== This igb implementation was tested with Linux Test Project [2]_ and Windows HLK -[3]_ during the initial development. The command used when testing with LTP is: +[3]_ during the initial development. Later it was also tested with DPDK Test +Suite [4]_. The command used when testing with LTP is: .. code-block:: shell @@ -XXX,XX +XXX,XX @@ This igb implementation was tested with Linux Test Project [2]_ and Windows HLK Be aware that this implementation lacks many functionalities available with the actual hardware, and you may experience various failures if you try to use it -with a different operating system other than Linux and Windows or if you try -functionalities not covered by the tests. +with a different operating system other than DPDK, Linux, and Windows or if you +try functionalities not covered by the tests. Using igb ========= @@ -XXX,XX +XXX,XX @@ Using igb should be nothing different from using another network device. See :ref:`Network_emulation` in general. However, you may also need to perform additional steps to activate SR-IOV -feature on your guest. For Linux, refer to [4]_. +feature on your guest. For Linux, refer to [5]_. Developing igb ============== @@ -XXX,XX +XXX,XX @@ References .. [1] https://www.intel.com/content/dam/www/public/us/en/documents/datasheets/82576eb-gigabit-ethernet-controller-datasheet.pdf .. [2] https://github.com/linux-test-project/ltp .. [3] https://learn.microsoft.com/en-us/windows-hardware/test/hlk/ -.. [4] https://docs.kernel.org/PCI/pci-iov-howto.html +.. [4] https://doc.dpdk.org/dts/gsg/ +.. [5] https://docs.kernel.org/PCI/pci-iov-howto.html -- 2.7.4 From: Stefan Hajnoczi <stefanha@redhat.com> If the driver sets large_send_mss to 0 then a divide-by-zero occurs. Even if the division wasn't a problem, the for loop that emits MSS-sized packets would never terminate. Solve these issues by skipping offloading when large_send_mss=0. This issue was found by OSS-Fuzz as part of Alexander Bulekov's device fuzzing work. The reproducer is: $ cat << EOF | ./qemu-system-i386 -display none -machine accel=qtest, -m \ 512M,slots=1,maxmem=0xffff000000000000 -machine q35 -nodefaults -device \ rtl8139,netdev=net0 -netdev user,id=net0 -device \ pc-dimm,id=nv1,memdev=mem1,addr=0xb800a64602800000 -object \ memory-backend-ram,id=mem1,size=2M -qtest stdio outl 0xcf8 0x80000814 outl 0xcfc 0xe0000000 outl 0xcf8 0x80000804 outw 0xcfc 0x06 write 0xe0000037 0x1 0x04 write 0xe00000e0 0x2 0x01 write 0x1 0x1 0x04 write 0x3 0x1 0x98 write 0xa 0x1 0x8c write 0xb 0x1 0x02 write 0xc 0x1 0x46 write 0xd 0x1 0xa6 write 0xf 0x1 0xb8 write 0xb800a646028c000c 0x1 0x08 write 0xb800a646028c000e 0x1 0x47 write 0xb800a646028c0010 0x1 0x02 write 0xb800a646028c0017 0x1 0x06 write 0xb800a646028c0036 0x1 0x80 write 0xe00000d9 0x1 0x40 EOF Buglink: https://gitlab.com/qemu-project/qemu/-/issues/1582 Closes: https://gitlab.com/qemu-project/qemu/-/issues/1582 Cc: qemu-stable@nongnu.org Cc: Peter Maydell <peter.maydell@linaro.org> Fixes: 6d71357a3b65 ("rtl8139: honor large send MSS value") Reported-by: Alexander Bulekov <alxndr@bu.edu> Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org> Tested-by: Alexander Bulekov <alxndr@bu.edu> Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/net/rtl8139.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/hw/net/rtl8139.c b/hw/net/rtl8139.c index XXXXXXX..XXXXXXX 100644 --- a/hw/net/rtl8139.c +++ b/hw/net/rtl8139.c @@ -XXX,XX +XXX,XX @@ static int rtl8139_cplus_transmit_one(RTL8139State *s) int large_send_mss = (txdw0 >> CP_TC_LGSEN_MSS_SHIFT) & CP_TC_LGSEN_MSS_MASK; + if (large_send_mss == 0) { + goto skip_offload; + } DPRINTF("+++ C+ mode offloaded task TSO IP data %d " "frame data %d specified MSS=%d\n", -- 2.7.4