1 | The following changes since commit e3debd5e7d0ce031356024878a0a18b9d109354a: | 1 | The following changes since commit f8d75e10d3e0033a0a29a7a7e4777a4fbc17a016: |
---|---|---|---|
2 | 2 | ||
3 | Merge tag 'pull-request-2023-03-24' of https://gitlab.com/thuth/qemu into staging (2023-03-24 16:08:46 +0000) | 3 | Merge remote-tracking branch 'remotes/legoater/tags/pull-ppc-20220112' into staging (2022-01-13 11:18:24 +0000) |
4 | 4 | ||
5 | are available in the git repository at: | 5 | are available in the git repository at: |
6 | 6 | ||
7 | https://github.com/jasowang/qemu.git tags/net-pull-request | 7 | https://github.com/jasowang/qemu.git tags/net-pull-request |
8 | 8 | ||
9 | for you to fetch changes up to fba7c3b788dfcb99a3f9253f7d99cc0d217d6d3c: | 9 | for you to fetch changes up to 818692f0a01587d02220916b31d5bb8e7dced611: |
10 | 10 | ||
11 | igb: respect VMVIR and VMOLR for VLAN (2023-03-28 13:10:55 +0800) | 11 | net/vmnet: update MAINTAINERS list (2022-01-14 12:58:19 +0800) |
12 | 12 | ||
13 | ---------------------------------------------------------------- | 13 | ---------------------------------------------------------------- |
14 | 14 | ||
15 | Changes since V2: | ||
16 | |||
17 | - Try to make vmnet work on some old mac version | ||
18 | |||
15 | ---------------------------------------------------------------- | 19 | ---------------------------------------------------------------- |
16 | Akihiko Odaki (4): | 20 | Peter Foley (2): |
17 | igb: Save more Tx states | 21 | net/tap: Set return code on failure |
18 | igb: Fix DMA requester specification for Tx packet | 22 | net: Fix uninitialized data usage |
19 | hw/net/net_tx_pkt: Ignore ECN bit | ||
20 | hw/net/net_tx_pkt: Align l3_hdr | ||
21 | 23 | ||
22 | Sriram Yagnaraman (8): | 24 | Philippe Mathieu-Daudé (1): |
23 | MAINTAINERS: Add Sriram Yagnaraman as a igb reviewer | 25 | hw/net/vmxnet3: Log guest-triggerable errors using LOG_GUEST_ERROR |
24 | igb: handle PF/VF reset properly | ||
25 | igb: add ICR_RXDW | ||
26 | igb: implement VFRE and VFTE registers | ||
27 | igb: check oversized packets for VMDq | ||
28 | igb: respect E1000_VMOLR_RSSE | ||
29 | igb: implement VF Tx and Rx stats | ||
30 | igb: respect VMVIR and VMOLR for VLAN | ||
31 | 26 | ||
32 | MAINTAINERS | 1 + | 27 | Rao Lei (1): |
33 | hw/net/e1000e_core.c | 6 +- | 28 | net/filter: Optimize filter_send to coroutine |
34 | hw/net/e1000x_regs.h | 4 + | 29 | |
35 | hw/net/igb.c | 26 ++++-- | 30 | Vladislav Yaroshchuk (7): |
36 | hw/net/igb_core.c | 256 ++++++++++++++++++++++++++++++++++++++------------- | 31 | net/vmnet: add vmnet dependency and customizable option |
37 | hw/net/igb_core.h | 9 +- | 32 | net/vmnet: add vmnet backends to qapi/net |
38 | hw/net/igb_regs.h | 6 ++ | 33 | net/vmnet: implement shared mode (vmnet-shared) |
39 | hw/net/net_tx_pkt.c | 30 +++--- | 34 | net/vmnet: implement host mode (vmnet-host) |
40 | hw/net/net_tx_pkt.h | 3 +- | 35 | net/vmnet: implement bridged mode (vmnet-bridged) |
41 | hw/net/trace-events | 2 + | 36 | net/vmnet: update qemu-options.hx |
42 | hw/net/vmxnet3.c | 4 +- | 37 | net/vmnet: update MAINTAINERS list |
43 | 11 files changed, 254 insertions(+), 93 deletions(-) | 38 | |
39 | Zhang Chen (2): | ||
40 | net/colo-compare.c: Optimize compare order for performance | ||
41 | net/colo-compare.c: Update the default value comments | ||
42 | |||
43 | MAINTAINERS | 5 + | ||
44 | hw/net/vmxnet3.c | 4 +- | ||
45 | meson.build | 16 +- | ||
46 | meson_options.txt | 2 + | ||
47 | net/clients.h | 11 ++ | ||
48 | net/colo-compare.c | 28 ++-- | ||
49 | net/filter-mirror.c | 66 +++++++-- | ||
50 | net/meson.build | 7 + | ||
51 | net/net.c | 10 ++ | ||
52 | net/tap-linux.c | 1 + | ||
53 | net/tap.c | 1 + | ||
54 | net/vmnet-bridged.m | 120 +++++++++++++++ | ||
55 | net/vmnet-common.m | 333 ++++++++++++++++++++++++++++++++++++++++++ | ||
56 | net/vmnet-host.c | 122 ++++++++++++++++ | ||
57 | net/vmnet-shared.c | 100 +++++++++++++ | ||
58 | net/vmnet_int.h | 48 ++++++ | ||
59 | qapi/net.json | 133 ++++++++++++++++- | ||
60 | qemu-options.hx | 25 ++++ | ||
61 | scripts/meson-buildoptions.sh | 3 + | ||
62 | 19 files changed, 1004 insertions(+), 31 deletions(-) | ||
63 | create mode 100644 net/vmnet-bridged.m | ||
64 | create mode 100644 net/vmnet-common.m | ||
65 | create mode 100644 net/vmnet-host.c | ||
66 | create mode 100644 net/vmnet-shared.c | ||
67 | create mode 100644 net/vmnet_int.h | ||
68 | |||
69 | diff view generated by jsdifflib |
1 | From: Akihiko Odaki <akihiko.odaki@daynix.com> | 1 | From: Philippe Mathieu-Daudé <philmd@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | igb used to specify the PF as DMA requester when reading Tx packets. | 3 | The "Interrupt Cause" register (VMXNET3_REG_ICR) is read-only. |
4 | This made Tx requests from VFs to be performed on the address space of | 4 | Write accesses are ignored. Log them with as LOG_GUEST_ERROR |
5 | the PF, defeating the purpose of SR-IOV. Add some logic to change the | 5 | instead of aborting: |
6 | requester depending on the queue, which can be assigned to a VF. | ||
7 | 6 | ||
8 | Fixes: 3a977deebe ("Intrdocue igb device emulation") | 7 | [R +0.239743] writeq 0xe0002031 0x46291a5a55460800 |
9 | Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> | 8 | ERROR:hw/net/vmxnet3.c:1819:vmxnet3_io_bar1_write: code should not be reached |
9 | Thread 1 "qemu-system-i38" received signal SIGABRT, Aborted. | ||
10 | (gdb) bt | ||
11 | #3 0x74c397d3 in __GI_abort () at abort.c:79 | ||
12 | #4 0x76d3cd4c in g_assertion_message (domain=<optimized out>, file=<optimized out>, line=<optimized out>, func=<optimized out>, message=<optimized out>) at ../glib/gtestutils.c:3223 | ||
13 | #5 0x76d9d45f in g_assertion_message_expr | ||
14 | (domain=0x0, file=0x59fc2e53 "hw/net/vmxnet3.c", line=1819, func=0x59fc11e0 <__func__.vmxnet3_io_bar1_write> "vmxnet3_io_bar1_write", expr=<optimized out>) | ||
15 | at ../glib/gtestutils.c:3249 | ||
16 | #6 0x57e80a3a in vmxnet3_io_bar1_write (opaque=0x62814100, addr=56, val=70, size=4) at hw/net/vmxnet3.c:1819 | ||
17 | #7 0x58c2d894 in memory_region_write_accessor (mr=0x62816b90, addr=56, value=0x7fff9450, size=4, shift=0, mask=4294967295, attrs=...) at softmmu/memory.c:492 | ||
18 | #8 0x58c2d1d2 in access_with_adjusted_size (addr=56, value=0x7fff9450, size=1, access_size_min=4, access_size_max=4, access_fn= | ||
19 | 0x58c2d290 <memory_region_write_accessor>, mr=0x62816b90, attrs=...) at softmmu/memory.c:554 | ||
20 | #9 0x58c2bae7 in memory_region_dispatch_write (mr=0x62816b90, addr=56, data=70, op=MO_8, attrs=...) at softmmu/memory.c:1504 | ||
21 | #10 0x58bfd034 in flatview_write_continue (fv=0x606000181700, addr=0xe0002038, attrs=..., ptr=0x7fffb9e0, len=1, addr1=56, l=1, mr=0x62816b90) | ||
22 | at softmmu/physmem.c:2782 | ||
23 | #11 0x58beba00 in flatview_write (fv=0x606000181700, addr=0xe0002031, attrs=..., buf=0x7fffb9e0, len=8) at softmmu/physmem.c:2822 | ||
24 | #12 0x58beb589 in address_space_write (as=0x608000015f20, addr=0xe0002031, attrs=..., buf=0x7fffb9e0, len=8) at softmmu/physmem.c:2914 | ||
25 | |||
26 | Reported-by: Dike <dike199774@qq.com> | ||
27 | Reported-by: Duhao <504224090@qq.com> | ||
28 | BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=2032932 | ||
29 | Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
10 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 30 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
11 | --- | 31 | --- |
12 | hw/net/e1000e_core.c | 6 +++--- | 32 | hw/net/vmxnet3.c | 4 +++- |
13 | hw/net/igb_core.c | 13 ++++++++----- | 33 | 1 file changed, 3 insertions(+), 1 deletion(-) |
14 | hw/net/net_tx_pkt.c | 3 ++- | ||
15 | hw/net/net_tx_pkt.h | 3 ++- | ||
16 | hw/net/vmxnet3.c | 4 ++-- | ||
17 | 5 files changed, 17 insertions(+), 12 deletions(-) | ||
18 | 34 | ||
19 | diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c | ||
20 | index XXXXXXX..XXXXXXX 100644 | ||
21 | --- a/hw/net/e1000e_core.c | ||
22 | +++ b/hw/net/e1000e_core.c | ||
23 | @@ -XXX,XX +XXX,XX @@ e1000e_process_tx_desc(E1000ECore *core, | ||
24 | } | ||
25 | |||
26 | tx->skip_cp = false; | ||
27 | - net_tx_pkt_reset(tx->tx_pkt); | ||
28 | + net_tx_pkt_reset(tx->tx_pkt, core->owner); | ||
29 | |||
30 | tx->sum_needed = 0; | ||
31 | tx->cptse = 0; | ||
32 | @@ -XXX,XX +XXX,XX @@ e1000e_core_pci_uninit(E1000ECore *core) | ||
33 | qemu_del_vm_change_state_handler(core->vmstate); | ||
34 | |||
35 | for (i = 0; i < E1000E_NUM_QUEUES; i++) { | ||
36 | - net_tx_pkt_reset(core->tx[i].tx_pkt); | ||
37 | + net_tx_pkt_reset(core->tx[i].tx_pkt, core->owner); | ||
38 | net_tx_pkt_uninit(core->tx[i].tx_pkt); | ||
39 | } | ||
40 | |||
41 | @@ -XXX,XX +XXX,XX @@ static void e1000e_reset(E1000ECore *core, bool sw) | ||
42 | e1000x_reset_mac_addr(core->owner_nic, core->mac, core->permanent_mac); | ||
43 | |||
44 | for (i = 0; i < ARRAY_SIZE(core->tx); i++) { | ||
45 | - net_tx_pkt_reset(core->tx[i].tx_pkt); | ||
46 | + net_tx_pkt_reset(core->tx[i].tx_pkt, core->owner); | ||
47 | memset(&core->tx[i].props, 0, sizeof(core->tx[i].props)); | ||
48 | core->tx[i].skip_cp = false; | ||
49 | } | ||
50 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | ||
51 | index XXXXXXX..XXXXXXX 100644 | ||
52 | --- a/hw/net/igb_core.c | ||
53 | +++ b/hw/net/igb_core.c | ||
54 | @@ -XXX,XX +XXX,XX @@ igb_on_tx_done_update_stats(IGBCore *core, struct NetTxPkt *tx_pkt) | ||
55 | |||
56 | static void | ||
57 | igb_process_tx_desc(IGBCore *core, | ||
58 | + PCIDevice *dev, | ||
59 | struct igb_tx *tx, | ||
60 | union e1000_adv_tx_desc *tx_desc, | ||
61 | int queue_index) | ||
62 | @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, | ||
63 | |||
64 | tx->first = true; | ||
65 | tx->skip_cp = false; | ||
66 | - net_tx_pkt_reset(tx->tx_pkt); | ||
67 | + net_tx_pkt_reset(tx->tx_pkt, dev); | ||
68 | } | ||
69 | } | ||
70 | |||
71 | @@ -XXX,XX +XXX,XX @@ igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) | ||
72 | d = core->owner; | ||
73 | } | ||
74 | |||
75 | + net_tx_pkt_reset(txr->tx->tx_pkt, d); | ||
76 | + | ||
77 | while (!igb_ring_empty(core, txi)) { | ||
78 | base = igb_ring_head_descr(core, txi); | ||
79 | |||
80 | @@ -XXX,XX +XXX,XX @@ igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) | ||
81 | trace_e1000e_tx_descr((void *)(intptr_t)desc.read.buffer_addr, | ||
82 | desc.read.cmd_type_len, desc.wb.status); | ||
83 | |||
84 | - igb_process_tx_desc(core, txr->tx, &desc, txi->idx); | ||
85 | + igb_process_tx_desc(core, d, txr->tx, &desc, txi->idx); | ||
86 | igb_ring_advance(core, txi, 1); | ||
87 | eic |= igb_txdesc_writeback(core, base, &desc, txi); | ||
88 | } | ||
89 | @@ -XXX,XX +XXX,XX @@ igb_core_pci_realize(IGBCore *core, | ||
90 | core->vmstate = qemu_add_vm_change_state_handler(igb_vm_state_change, core); | ||
91 | |||
92 | for (i = 0; i < IGB_NUM_QUEUES; i++) { | ||
93 | - net_tx_pkt_init(&core->tx[i].tx_pkt, core->owner, E1000E_MAX_TX_FRAGS); | ||
94 | + net_tx_pkt_init(&core->tx[i].tx_pkt, NULL, E1000E_MAX_TX_FRAGS); | ||
95 | } | ||
96 | |||
97 | net_rx_pkt_init(&core->rx_pkt); | ||
98 | @@ -XXX,XX +XXX,XX @@ igb_core_pci_uninit(IGBCore *core) | ||
99 | qemu_del_vm_change_state_handler(core->vmstate); | ||
100 | |||
101 | for (i = 0; i < IGB_NUM_QUEUES; i++) { | ||
102 | - net_tx_pkt_reset(core->tx[i].tx_pkt); | ||
103 | + net_tx_pkt_reset(core->tx[i].tx_pkt, NULL); | ||
104 | net_tx_pkt_uninit(core->tx[i].tx_pkt); | ||
105 | } | ||
106 | |||
107 | @@ -XXX,XX +XXX,XX @@ static void igb_reset(IGBCore *core, bool sw) | ||
108 | |||
109 | for (i = 0; i < ARRAY_SIZE(core->tx); i++) { | ||
110 | tx = &core->tx[i]; | ||
111 | - net_tx_pkt_reset(tx->tx_pkt); | ||
112 | + net_tx_pkt_reset(tx->tx_pkt, NULL); | ||
113 | memset(tx->ctx, 0, sizeof(tx->ctx)); | ||
114 | tx->first = true; | ||
115 | tx->skip_cp = false; | ||
116 | diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c | ||
117 | index XXXXXXX..XXXXXXX 100644 | ||
118 | --- a/hw/net/net_tx_pkt.c | ||
119 | +++ b/hw/net/net_tx_pkt.c | ||
120 | @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_dump(struct NetTxPkt *pkt) | ||
121 | #endif | ||
122 | } | ||
123 | |||
124 | -void net_tx_pkt_reset(struct NetTxPkt *pkt) | ||
125 | +void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *pci_dev) | ||
126 | { | ||
127 | int i; | ||
128 | |||
129 | @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_reset(struct NetTxPkt *pkt) | ||
130 | pkt->raw[i].iov_len, DMA_DIRECTION_TO_DEVICE, 0); | ||
131 | } | ||
132 | } | ||
133 | + pkt->pci_dev = pci_dev; | ||
134 | pkt->raw_frags = 0; | ||
135 | |||
136 | pkt->hdr_len = 0; | ||
137 | diff --git a/hw/net/net_tx_pkt.h b/hw/net/net_tx_pkt.h | ||
138 | index XXXXXXX..XXXXXXX 100644 | ||
139 | --- a/hw/net/net_tx_pkt.h | ||
140 | +++ b/hw/net/net_tx_pkt.h | ||
141 | @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_dump(struct NetTxPkt *pkt); | ||
142 | * reset tx packet private context (needed to be called between packets) | ||
143 | * | ||
144 | * @pkt: packet | ||
145 | + * @dev: PCI device processing the next packet | ||
146 | * | ||
147 | */ | ||
148 | -void net_tx_pkt_reset(struct NetTxPkt *pkt); | ||
149 | +void net_tx_pkt_reset(struct NetTxPkt *pkt, PCIDevice *dev); | ||
150 | |||
151 | /** | ||
152 | * Send packet to qemu. handles sw offloads if vhdr is not supported. | ||
153 | diff --git a/hw/net/vmxnet3.c b/hw/net/vmxnet3.c | 35 | diff --git a/hw/net/vmxnet3.c b/hw/net/vmxnet3.c |
154 | index XXXXXXX..XXXXXXX 100644 | 36 | index XXXXXXX..XXXXXXX 100644 |
155 | --- a/hw/net/vmxnet3.c | 37 | --- a/hw/net/vmxnet3.c |
156 | +++ b/hw/net/vmxnet3.c | 38 | +++ b/hw/net/vmxnet3.c |
157 | @@ -XXX,XX +XXX,XX @@ static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx) | 39 | @@ -XXX,XX +XXX,XX @@ vmxnet3_io_bar1_write(void *opaque, |
158 | vmxnet3_complete_packet(s, qidx, txd_idx); | 40 | case VMXNET3_REG_ICR: |
159 | s->tx_sop = true; | 41 | VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d", |
160 | s->skip_current_tx_pkt = false; | 42 | val, size); |
161 | - net_tx_pkt_reset(s->tx_pkt); | 43 | - g_assert_not_reached(); |
162 | + net_tx_pkt_reset(s->tx_pkt, PCI_DEVICE(s)); | 44 | + qemu_log_mask(LOG_GUEST_ERROR, |
163 | } | 45 | + "%s: write to read-only register VMXNET3_REG_ICR\n", |
164 | } | 46 | + TYPE_VMXNET3); |
165 | } | 47 | break; |
166 | @@ -XXX,XX +XXX,XX @@ static void vmxnet3_deactivate_device(VMXNET3State *s) | 48 | |
167 | { | 49 | /* Event Cause Register */ |
168 | if (s->device_active) { | ||
169 | VMW_CBPRN("Deactivating vmxnet3..."); | ||
170 | - net_tx_pkt_reset(s->tx_pkt); | ||
171 | + net_tx_pkt_reset(s->tx_pkt, PCI_DEVICE(s)); | ||
172 | net_tx_pkt_uninit(s->tx_pkt); | ||
173 | net_rx_pkt_uninit(s->rx_pkt); | ||
174 | s->device_active = false; | ||
175 | -- | 50 | -- |
176 | 2.7.4 | 51 | 2.7.4 |
52 | |||
53 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Peter Foley <pefoley@google.com> |
---|---|---|---|
2 | 2 | ||
3 | Also introduce: | 3 | Match the other error handling in this function. |
4 | - Checks for RXDCTL/TXDCTL queue enable bits | ||
5 | - IGB_NUM_VM_POOLS enum (Sec 1.5: Table 1-7) | ||
6 | 4 | ||
7 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 5 | Fixes: e7b347d0bf6 ("net: detect errors from probing vnet hdr flag for TAP devices") |
6 | |||
7 | Reviewed-by: Patrick Venture <venture@google.com> | ||
8 | Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> | ||
9 | Signed-off-by: Peter Foley <pefoley@google.com> | ||
8 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 10 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
9 | --- | 11 | --- |
10 | hw/net/igb_core.c | 38 +++++++++++++++++++++++++++++++------- | 12 | net/tap.c | 1 + |
11 | hw/net/igb_core.h | 1 + | 13 | 1 file changed, 1 insertion(+) |
12 | hw/net/igb_regs.h | 3 +++ | ||
13 | 3 files changed, 35 insertions(+), 7 deletions(-) | ||
14 | 14 | ||
15 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | 15 | diff --git a/net/tap.c b/net/tap.c |
16 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
17 | --- a/hw/net/igb_core.c | 17 | --- a/net/tap.c |
18 | +++ b/hw/net/igb_core.c | 18 | +++ b/net/tap.c |
19 | @@ -XXX,XX +XXX,XX @@ igb_txdesc_writeback(IGBCore *core, dma_addr_t base, | 19 | @@ -XXX,XX +XXX,XX @@ int net_init_tap(const Netdev *netdev, const char *name, |
20 | return igb_tx_wb_eic(core, txi->idx); | 20 | if (i == 0) { |
21 | } | 21 | vnet_hdr = tap_probe_vnet_hdr(fd, errp); |
22 | 22 | if (vnet_hdr < 0) { | |
23 | +static inline bool | 23 | + ret = -1; |
24 | +igb_tx_enabled(IGBCore *core, const E1000E_RingInfo *txi) | 24 | goto free_fail; |
25 | +{ | ||
26 | + bool vmdq = core->mac[MRQC] & 1; | ||
27 | + uint16_t qn = txi->idx; | ||
28 | + uint16_t pool = qn % IGB_NUM_VM_POOLS; | ||
29 | + | ||
30 | + return (core->mac[TCTL] & E1000_TCTL_EN) && | ||
31 | + (!vmdq || core->mac[VFTE] & BIT(pool)) && | ||
32 | + (core->mac[TXDCTL0 + (qn * 16)] & E1000_TXDCTL_QUEUE_ENABLE); | ||
33 | +} | ||
34 | + | ||
35 | static void | ||
36 | igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) | ||
37 | { | ||
38 | @@ -XXX,XX +XXX,XX @@ igb_start_xmit(IGBCore *core, const IGB_TxRing *txr) | ||
39 | const E1000E_RingInfo *txi = txr->i; | ||
40 | uint32_t eic = 0; | ||
41 | |||
42 | - /* TODO: check if the queue itself is enabled too. */ | ||
43 | - if (!(core->mac[TCTL] & E1000_TCTL_EN)) { | ||
44 | + if (!igb_tx_enabled(core, txi)) { | ||
45 | trace_e1000e_tx_disabled(); | ||
46 | return; | ||
47 | } | ||
48 | @@ -XXX,XX +XXX,XX @@ igb_can_receive(IGBCore *core) | ||
49 | |||
50 | for (i = 0; i < IGB_NUM_QUEUES; i++) { | ||
51 | E1000E_RxRing rxr; | ||
52 | + if (!(core->mac[RXDCTL0 + (i * 16)] & E1000_RXDCTL_QUEUE_ENABLE)) { | ||
53 | + continue; | ||
54 | + } | ||
55 | |||
56 | igb_rx_ring_init(core, &rxr, i); | ||
57 | if (igb_ring_enabled(core, rxr.i) && igb_has_rxbufs(core, rxr.i, 1)) { | ||
58 | @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | ||
59 | |||
60 | if (core->mac[MRQC] & 1) { | ||
61 | if (is_broadcast_ether_addr(ehdr->h_dest)) { | ||
62 | - for (i = 0; i < 8; i++) { | ||
63 | + for (i = 0; i < IGB_NUM_VM_POOLS; i++) { | ||
64 | if (core->mac[VMOLR0 + i] & E1000_VMOLR_BAM) { | ||
65 | queues |= BIT(i); | ||
66 | } | 25 | } |
67 | @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | 26 | } else if (vnet_hdr != tap_probe_vnet_hdr(fd, NULL)) { |
68 | f = ta_shift[(rctl >> E1000_RCTL_MO_SHIFT) & 3]; | ||
69 | f = (((ehdr->h_dest[5] << 8) | ehdr->h_dest[4]) >> f) & 0xfff; | ||
70 | if (macp[f >> 5] & (1 << (f & 0x1f))) { | ||
71 | - for (i = 0; i < 8; i++) { | ||
72 | + for (i = 0; i < IGB_NUM_VM_POOLS; i++) { | ||
73 | if (core->mac[VMOLR0 + i] & E1000_VMOLR_ROMPE) { | ||
74 | queues |= BIT(i); | ||
75 | } | ||
76 | @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | ||
77 | } | ||
78 | } | ||
79 | } else { | ||
80 | - for (i = 0; i < 8; i++) { | ||
81 | + for (i = 0; i < IGB_NUM_VM_POOLS; i++) { | ||
82 | if (core->mac[VMOLR0 + i] & E1000_VMOLR_AUPE) { | ||
83 | mask |= BIT(i); | ||
84 | } | ||
85 | @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | ||
86 | queues = BIT(def_pl >> E1000_VT_CTL_DEFAULT_POOL_SHIFT); | ||
87 | } | ||
88 | |||
89 | + queues &= core->mac[VFRE]; | ||
90 | igb_rss_parse_packet(core, core->rx_pkt, external_tx != NULL, rss_info); | ||
91 | if (rss_info->queue & 1) { | ||
92 | queues <<= 8; | ||
93 | @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, | ||
94 | e1000x_fcs_len(core->mac); | ||
95 | |||
96 | for (i = 0; i < IGB_NUM_QUEUES; i++) { | ||
97 | - if (!(queues & BIT(i))) { | ||
98 | + if (!(queues & BIT(i)) || | ||
99 | + !(core->mac[RXDCTL0 + (i * 16)] & E1000_RXDCTL_QUEUE_ENABLE)) { | ||
100 | continue; | ||
101 | } | ||
102 | |||
103 | @@ -XXX,XX +XXX,XX @@ static void igb_set_vfmailbox(IGBCore *core, int index, uint32_t val) | ||
104 | |||
105 | static void igb_vf_reset(IGBCore *core, uint16_t vfn) | ||
106 | { | ||
107 | + uint16_t qn0 = vfn; | ||
108 | + uint16_t qn1 = vfn + IGB_NUM_VM_POOLS; | ||
109 | + | ||
110 | /* disable Rx and Tx for the VF*/ | ||
111 | - core->mac[VFTE] &= ~BIT(vfn); | ||
112 | + core->mac[RXDCTL0 + (qn0 * 16)] &= ~E1000_RXDCTL_QUEUE_ENABLE; | ||
113 | + core->mac[RXDCTL0 + (qn1 * 16)] &= ~E1000_RXDCTL_QUEUE_ENABLE; | ||
114 | + core->mac[TXDCTL0 + (qn0 * 16)] &= ~E1000_TXDCTL_QUEUE_ENABLE; | ||
115 | + core->mac[TXDCTL0 + (qn1 * 16)] &= ~E1000_TXDCTL_QUEUE_ENABLE; | ||
116 | core->mac[VFRE] &= ~BIT(vfn); | ||
117 | + core->mac[VFTE] &= ~BIT(vfn); | ||
118 | /* indicate VF reset to PF */ | ||
119 | core->mac[VFLRE] |= BIT(vfn); | ||
120 | /* VFLRE and mailbox use the same interrupt cause */ | ||
121 | @@ -XXX,XX +XXX,XX @@ igb_phy_reg_init[] = { | ||
122 | static const uint32_t igb_mac_reg_init[] = { | ||
123 | [LEDCTL] = 2 | (3 << 8) | BIT(15) | (6 << 16) | (7 << 24), | ||
124 | [EEMNGCTL] = BIT(31), | ||
125 | + [TXDCTL0] = E1000_TXDCTL_QUEUE_ENABLE, | ||
126 | [RXDCTL0] = E1000_RXDCTL_QUEUE_ENABLE | (1 << 16), | ||
127 | [RXDCTL1] = 1 << 16, | ||
128 | [RXDCTL2] = 1 << 16, | ||
129 | diff --git a/hw/net/igb_core.h b/hw/net/igb_core.h | ||
130 | index XXXXXXX..XXXXXXX 100644 | ||
131 | --- a/hw/net/igb_core.h | ||
132 | +++ b/hw/net/igb_core.h | ||
133 | @@ -XXX,XX +XXX,XX @@ | ||
134 | #define IGB_MSIX_VEC_NUM (10) | ||
135 | #define IGBVF_MSIX_VEC_NUM (3) | ||
136 | #define IGB_NUM_QUEUES (16) | ||
137 | +#define IGB_NUM_VM_POOLS (8) | ||
138 | |||
139 | typedef struct IGBCore IGBCore; | ||
140 | |||
141 | diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h | ||
142 | index XXXXXXX..XXXXXXX 100644 | ||
143 | --- a/hw/net/igb_regs.h | ||
144 | +++ b/hw/net/igb_regs.h | ||
145 | @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { | ||
146 | #define E1000_MRQC_RSS_FIELD_IPV6_UDP 0x00800000 | ||
147 | #define E1000_MRQC_RSS_FIELD_IPV6_UDP_EX 0x01000000 | ||
148 | |||
149 | +/* Additional Transmit Descriptor Control definitions */ | ||
150 | +#define E1000_TXDCTL_QUEUE_ENABLE 0x02000000 /* Enable specific Tx Queue */ | ||
151 | + | ||
152 | /* Additional Receive Descriptor Control definitions */ | ||
153 | #define E1000_RXDCTL_QUEUE_ENABLE 0x02000000 /* Enable specific Rx Queue */ | ||
154 | |||
155 | -- | 27 | -- |
156 | 2.7.4 | 28 | 2.7.4 |
29 | |||
30 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Peter Foley <pefoley@google.com> | ||
1 | 2 | ||
3 | e.g. | ||
4 | 1109 15:16:20.151506 Uninitialized bytes in ioctl_common_pre at offset 0 inside [0x7ffc516af9b8, 4) | ||
5 | 1109 15:16:20.151659 ==588974==WARNING: MemorySanitizer: use-of-uninitialized-value | ||
6 | 1109 15:16:20.312923 #0 0x5639b88acb21 in tap_probe_vnet_hdr_len third_party/qemu/net/tap-linux.c:183:9 | ||
7 | 1109 15:16:20.312952 #1 0x5639b88afd66 in net_tap_fd_init third_party/qemu/net/tap.c:409:9 | ||
8 | 1109 15:16:20.312954 #2 0x5639b88b2d1b in net_init_tap_one third_party/qemu/net/tap.c:681:19 | ||
9 | 1109 15:16:20.312956 #3 0x5639b88b16a8 in net_init_tap third_party/qemu/net/tap.c:912:13 | ||
10 | 1109 15:16:20.312957 #4 0x5639b8890175 in net_client_init1 third_party/qemu/net/net.c:1110:9 | ||
11 | 1109 15:16:20.312958 #5 0x5639b888f912 in net_client_init third_party/qemu/net/net.c:1208:15 | ||
12 | 1109 15:16:20.312960 #6 0x5639b8894aa5 in net_param_nic third_party/qemu/net/net.c:1588:11 | ||
13 | 1109 15:16:20.312961 #7 0x5639b900cd18 in qemu_opts_foreach third_party/qemu/util/qemu-option.c:1135:14 | ||
14 | 1109 15:16:20.312962 #8 0x5639b889393c in net_init_clients third_party/qemu/net/net.c:1612:9 | ||
15 | 1109 15:16:20.312964 #9 0x5639b717aaf3 in qemu_create_late_backends third_party/qemu/softmmu/vl.c:1962:5 | ||
16 | 1109 15:16:20.312965 #10 0x5639b717aaf3 in qemu_init third_party/qemu/softmmu/vl.c:3694:5 | ||
17 | 1109 15:16:20.312967 #11 0x5639b71083b8 in main third_party/qemu/softmmu/main.c:49:5 | ||
18 | 1109 15:16:20.312968 #12 0x7f464de1d8d2 in __libc_start_main (/usr/grte/v5/lib64/libc.so.6+0x628d2) | ||
19 | 1109 15:16:20.312969 #13 0x5639b6bbd389 in _start /usr/grte/v5/debug-src/src/csu/../sysdeps/x86_64/start.S:120 | ||
20 | 1109 15:16:20.312970 | ||
21 | 1109 15:16:20.312975 Uninitialized value was stored to memory at | ||
22 | 1109 15:16:20.313393 #0 0x5639b88acbee in tap_probe_vnet_hdr_len third_party/qemu/net/tap-linux.c | ||
23 | 1109 15:16:20.313396 #1 0x5639b88afd66 in net_tap_fd_init third_party/qemu/net/tap.c:409:9 | ||
24 | 1109 15:16:20.313398 #2 0x5639b88b2d1b in net_init_tap_one third_party/qemu/net/tap.c:681:19 | ||
25 | 1109 15:16:20.313399 #3 0x5639b88b16a8 in net_init_tap third_party/qemu/net/tap.c:912:13 | ||
26 | 1109 15:16:20.313400 #4 0x5639b8890175 in net_client_init1 third_party/qemu/net/net.c:1110:9 | ||
27 | 1109 15:16:20.313401 #5 0x5639b888f912 in net_client_init third_party/qemu/net/net.c:1208:15 | ||
28 | 1109 15:16:20.313403 #6 0x5639b8894aa5 in net_param_nic third_party/qemu/net/net.c:1588:11 | ||
29 | 1109 15:16:20.313404 #7 0x5639b900cd18 in qemu_opts_foreach third_party/qemu/util/qemu-option.c:1135:14 | ||
30 | 1109 15:16:20.313405 #8 0x5639b889393c in net_init_clients third_party/qemu/net/net.c:1612:9 | ||
31 | 1109 15:16:20.313407 #9 0x5639b717aaf3 in qemu_create_late_backends third_party/qemu/softmmu/vl.c:1962:5 | ||
32 | 1109 15:16:20.313408 #10 0x5639b717aaf3 in qemu_init third_party/qemu/softmmu/vl.c:3694:5 | ||
33 | 1109 15:16:20.313409 #11 0x5639b71083b8 in main third_party/qemu/softmmu/main.c:49:5 | ||
34 | 1109 15:16:20.313410 #12 0x7f464de1d8d2 in __libc_start_main (/usr/grte/v5/lib64/libc.so.6+0x628d2) | ||
35 | 1109 15:16:20.313412 #13 0x5639b6bbd389 in _start /usr/grte/v5/debug-src/src/csu/../sysdeps/x86_64/start.S:120 | ||
36 | 1109 15:16:20.313413 | ||
37 | 1109 15:16:20.313417 Uninitialized value was stored to memory at | ||
38 | 1109 15:16:20.313791 #0 0x5639b88affbd in net_tap_fd_init third_party/qemu/net/tap.c:400:26 | ||
39 | 1109 15:16:20.313826 #1 0x5639b88b2d1b in net_init_tap_one third_party/qemu/net/tap.c:681:19 | ||
40 | 1109 15:16:20.313829 #2 0x5639b88b16a8 in net_init_tap third_party/qemu/net/tap.c:912:13 | ||
41 | 1109 15:16:20.313831 #3 0x5639b8890175 in net_client_init1 third_party/qemu/net/net.c:1110:9 | ||
42 | 1109 15:16:20.313836 #4 0x5639b888f912 in net_client_init third_party/qemu/net/net.c:1208:15 | ||
43 | 1109 15:16:20.313838 #5 0x5639b8894aa5 in net_param_nic third_party/qemu/net/net.c:1588:11 | ||
44 | 1109 15:16:20.313839 #6 0x5639b900cd18 in qemu_opts_foreach third_party/qemu/util/qemu-option.c:1135:14 | ||
45 | 1109 15:16:20.313841 #7 0x5639b889393c in net_init_clients third_party/qemu/net/net.c:1612:9 | ||
46 | 1109 15:16:20.313843 #8 0x5639b717aaf3 in qemu_create_late_backends third_party/qemu/softmmu/vl.c:1962:5 | ||
47 | 1109 15:16:20.313844 #9 0x5639b717aaf3 in qemu_init third_party/qemu/softmmu/vl.c:3694:5 | ||
48 | 1109 15:16:20.313845 #10 0x5639b71083b8 in main third_party/qemu/softmmu/main.c:49:5 | ||
49 | 1109 15:16:20.313846 #11 0x7f464de1d8d2 in __libc_start_main (/usr/grte/v5/lib64/libc.so.6+0x628d2) | ||
50 | 1109 15:16:20.313847 #12 0x5639b6bbd389 in _start /usr/grte/v5/debug-src/src/csu/../sysdeps/x86_64/start.S:120 | ||
51 | 1109 15:16:20.313849 | ||
52 | 1109 15:16:20.313851 Uninitialized value was created by an allocation of 'ifr' in the stack frame of function 'tap_probe_vnet_hdr' | ||
53 | 1109 15:16:20.313855 #0 0x5639b88ac680 in tap_probe_vnet_hdr third_party/qemu/net/tap-linux.c:151 | ||
54 | 1109 15:16:20.313856 | ||
55 | 1109 15:16:20.313878 SUMMARY: MemorySanitizer: use-of-uninitialized-value third_party/qemu/net/tap-linux.c:183:9 in tap_probe_vnet_hdr_len | ||
56 | |||
57 | Fixes: dc69004c7d8 ("net: move tap_probe_vnet_hdr() to tap-linux.c") | ||
58 | Reviewed-by: Hao Wu <wuhaotsh@google.com> | ||
59 | Reviewed-by: Patrick Venture <venture@google.com> | ||
60 | Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> | ||
61 | Signed-off-by: Peter Foley <pefoley@google.com> | ||
62 | Signed-off-by: Jason Wang <jasowang@redhat.com> | ||
63 | --- | ||
64 | net/tap-linux.c | 1 + | ||
65 | 1 file changed, 1 insertion(+) | ||
66 | |||
67 | diff --git a/net/tap-linux.c b/net/tap-linux.c | ||
68 | index XXXXXXX..XXXXXXX 100644 | ||
69 | --- a/net/tap-linux.c | ||
70 | +++ b/net/tap-linux.c | ||
71 | @@ -XXX,XX +XXX,XX @@ void tap_set_sndbuf(int fd, const NetdevTapOptions *tap, Error **errp) | ||
72 | int tap_probe_vnet_hdr(int fd, Error **errp) | ||
73 | { | ||
74 | struct ifreq ifr; | ||
75 | + memset(&ifr, 0, sizeof(ifr)); | ||
76 | |||
77 | if (ioctl(fd, TUNGETIFF, &ifr) != 0) { | ||
78 | /* TUNGETIFF is available since kernel v2.6.27 */ | ||
79 | -- | ||
80 | 2.7.4 | ||
81 | |||
82 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Zhang Chen <chen.zhang@intel.com> |
---|---|---|---|
2 | 2 | ||
3 | Please note that loopback counters for VM to VM traffic is not | 3 | COLO-compare use the glib function g_queue_find_custom to dump |
4 | implemented yet: VFGOTLBC, VFGPTLBC, VFGORLBC and VFGPRLBC. | 4 | another VM's networking packet to compare. But this function always |
5 | start find from the queue->head(here is the newest packet), It will | ||
6 | reduce the success rate of comparison. So this patch reversed | ||
7 | the order of the queues for performance. | ||
5 | 8 | ||
6 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 9 | Signed-off-by: Zhang Chen <chen.zhang@intel.com> |
10 | Reported-by: leirao <lei.rao@intel.com> | ||
7 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 11 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
8 | --- | 12 | --- |
9 | hw/net/igb_core.c | 26 ++++++++++++++++++++++---- | 13 | net/colo-compare.c | 26 +++++++++++++------------- |
10 | 1 file changed, 22 insertions(+), 4 deletions(-) | 14 | 1 file changed, 13 insertions(+), 13 deletions(-) |
11 | 15 | ||
12 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | 16 | diff --git a/net/colo-compare.c b/net/colo-compare.c |
13 | index XXXXXXX..XXXXXXX 100644 | 17 | index XXXXXXX..XXXXXXX 100644 |
14 | --- a/hw/net/igb_core.c | 18 | --- a/net/colo-compare.c |
15 | +++ b/hw/net/igb_core.c | 19 | +++ b/net/colo-compare.c |
16 | @@ -XXX,XX +XXX,XX @@ igb_tx_pkt_send(IGBCore *core, struct igb_tx *tx, int queue_index) | 20 | @@ -XXX,XX +XXX,XX @@ static void colo_compare_inconsistency_notify(CompareState *s) |
21 | /* Use restricted to colo_insert_packet() */ | ||
22 | static gint seq_sorter(Packet *a, Packet *b, gpointer data) | ||
23 | { | ||
24 | - return a->tcp_seq - b->tcp_seq; | ||
25 | + return b->tcp_seq - a->tcp_seq; | ||
17 | } | 26 | } |
18 | 27 | ||
19 | static void | 28 | static void fill_pkt_tcp_info(void *data, uint32_t *max_ack) |
20 | -igb_on_tx_done_update_stats(IGBCore *core, struct NetTxPkt *tx_pkt) | 29 | @@ -XXX,XX +XXX,XX @@ pri: |
21 | +igb_on_tx_done_update_stats(IGBCore *core, struct NetTxPkt *tx_pkt, int qn) | 30 | if (g_queue_is_empty(&conn->primary_list)) { |
22 | { | 31 | return; |
23 | static const int PTCregs[6] = { PTC64, PTC127, PTC255, PTC511, | 32 | } |
24 | PTC1023, PTC1522 }; | 33 | - ppkt = g_queue_pop_head(&conn->primary_list); |
25 | @@ -XXX,XX +XXX,XX @@ igb_on_tx_done_update_stats(IGBCore *core, struct NetTxPkt *tx_pkt) | 34 | + ppkt = g_queue_pop_tail(&conn->primary_list); |
26 | core->mac[GPTC] = core->mac[TPT]; | 35 | sec: |
27 | core->mac[GOTCL] = core->mac[TOTL]; | 36 | if (g_queue_is_empty(&conn->secondary_list)) { |
28 | core->mac[GOTCH] = core->mac[TOTH]; | 37 | - g_queue_push_head(&conn->primary_list, ppkt); |
29 | + | 38 | + g_queue_push_tail(&conn->primary_list, ppkt); |
30 | + if (core->mac[MRQC] & 1) { | 39 | return; |
31 | + uint16_t pool = qn % IGB_NUM_VM_POOLS; | 40 | } |
32 | + | 41 | - spkt = g_queue_pop_head(&conn->secondary_list); |
33 | + core->mac[PVFGOTC0 + (pool * 64)] += tot_len; | 42 | + spkt = g_queue_pop_tail(&conn->secondary_list); |
34 | + core->mac[PVFGPTC0 + (pool * 64)]++; | 43 | |
35 | + } | 44 | if (ppkt->tcp_seq == ppkt->seq_end) { |
36 | } | 45 | colo_release_primary_pkt(s, ppkt); |
37 | 46 | @@ -XXX,XX +XXX,XX @@ sec: | |
38 | static void | ||
39 | @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, | ||
40 | net_tx_pkt_setup_vlan_header_ex(tx->tx_pkt, vlan, vet); | ||
41 | } | ||
42 | if (igb_tx_pkt_send(core, tx, queue_index)) { | ||
43 | - igb_on_tx_done_update_stats(core, tx->tx_pkt); | ||
44 | + igb_on_tx_done_update_stats(core, tx->tx_pkt, queue_index); | ||
45 | } | 47 | } |
46 | } | 48 | } |
47 | 49 | if (!ppkt) { | |
48 | @@ -XXX,XX +XXX,XX @@ igb_write_to_rx_buffers(IGBCore *core, | 50 | - g_queue_push_head(&conn->secondary_list, spkt); |
51 | + g_queue_push_tail(&conn->secondary_list, spkt); | ||
52 | goto pri; | ||
53 | } | ||
54 | } | ||
55 | @@ -XXX,XX +XXX,XX @@ sec: | ||
56 | if (mark == COLO_COMPARE_FREE_PRIMARY) { | ||
57 | conn->compare_seq = ppkt->seq_end; | ||
58 | colo_release_primary_pkt(s, ppkt); | ||
59 | - g_queue_push_head(&conn->secondary_list, spkt); | ||
60 | + g_queue_push_tail(&conn->secondary_list, spkt); | ||
61 | goto pri; | ||
62 | } else if (mark == COLO_COMPARE_FREE_SECONDARY) { | ||
63 | conn->compare_seq = spkt->seq_end; | ||
64 | @@ -XXX,XX +XXX,XX @@ sec: | ||
65 | goto pri; | ||
66 | } | ||
67 | } else { | ||
68 | - g_queue_push_head(&conn->primary_list, ppkt); | ||
69 | - g_queue_push_head(&conn->secondary_list, spkt); | ||
70 | + g_queue_push_tail(&conn->primary_list, ppkt); | ||
71 | + g_queue_push_tail(&conn->secondary_list, spkt); | ||
72 | |||
73 | #ifdef DEBUG_COLO_PACKETS | ||
74 | qemu_hexdump(stderr, "colo-compare ppkt", ppkt->data, ppkt->size); | ||
75 | @@ -XXX,XX +XXX,XX @@ static void colo_compare_packet(CompareState *s, Connection *conn, | ||
76 | |||
77 | while (!g_queue_is_empty(&conn->primary_list) && | ||
78 | !g_queue_is_empty(&conn->secondary_list)) { | ||
79 | - pkt = g_queue_pop_head(&conn->primary_list); | ||
80 | + pkt = g_queue_pop_tail(&conn->primary_list); | ||
81 | result = g_queue_find_custom(&conn->secondary_list, | ||
82 | pkt, (GCompareFunc)HandlePacket); | ||
83 | |||
84 | @@ -XXX,XX +XXX,XX @@ static void colo_compare_packet(CompareState *s, Connection *conn, | ||
85 | * timeout, it will trigger a checkpoint request. | ||
86 | */ | ||
87 | trace_colo_compare_main("packet different"); | ||
88 | - g_queue_push_head(&conn->primary_list, pkt); | ||
89 | + g_queue_push_tail(&conn->primary_list, pkt); | ||
90 | |||
91 | colo_compare_inconsistency_notify(s); | ||
92 | break; | ||
93 | @@ -XXX,XX +XXX,XX @@ static int compare_chr_send(CompareState *s, | ||
94 | entry->buf = g_malloc(size); | ||
95 | memcpy(entry->buf, buf, size); | ||
96 | } | ||
97 | - g_queue_push_head(&sendco->send_list, entry); | ||
98 | + g_queue_push_tail(&sendco->send_list, entry); | ||
99 | |||
100 | if (sendco->done) { | ||
101 | sendco->co = qemu_coroutine_create(_compare_chr_send, sendco); | ||
102 | @@ -XXX,XX +XXX,XX @@ static void colo_flush_packets(void *opaque, void *user_data) | ||
103 | Packet *pkt = NULL; | ||
104 | |||
105 | while (!g_queue_is_empty(&conn->primary_list)) { | ||
106 | - pkt = g_queue_pop_head(&conn->primary_list); | ||
107 | + pkt = g_queue_pop_tail(&conn->primary_list); | ||
108 | compare_chr_send(s, | ||
109 | pkt->data, | ||
110 | pkt->size, | ||
111 | @@ -XXX,XX +XXX,XX @@ static void colo_flush_packets(void *opaque, void *user_data) | ||
112 | packet_destroy_partial(pkt, NULL); | ||
113 | } | ||
114 | while (!g_queue_is_empty(&conn->secondary_list)) { | ||
115 | - pkt = g_queue_pop_head(&conn->secondary_list); | ||
116 | + pkt = g_queue_pop_tail(&conn->secondary_list); | ||
117 | packet_destroy(pkt, NULL); | ||
118 | } | ||
49 | } | 119 | } |
50 | |||
51 | static void | ||
52 | -igb_update_rx_stats(IGBCore *core, size_t data_size, size_t data_fcs_size) | ||
53 | +igb_update_rx_stats(IGBCore *core, const E1000E_RingInfo *rxi, | ||
54 | + size_t data_size, size_t data_fcs_size) | ||
55 | { | ||
56 | e1000x_update_rx_total_stats(core->mac, data_size, data_fcs_size); | ||
57 | |||
58 | @@ -XXX,XX +XXX,XX @@ igb_update_rx_stats(IGBCore *core, size_t data_size, size_t data_fcs_size) | ||
59 | default: | ||
60 | break; | ||
61 | } | ||
62 | + | ||
63 | + if (core->mac[MRQC] & 1) { | ||
64 | + uint16_t pool = rxi->idx % IGB_NUM_VM_POOLS; | ||
65 | + | ||
66 | + core->mac[PVFGORC0 + (pool * 64)] += data_size + 4; | ||
67 | + core->mac[PVFGPRC0 + (pool * 64)]++; | ||
68 | + if (net_rx_pkt_get_packet_type(core->rx_pkt) == ETH_PKT_MCAST) { | ||
69 | + core->mac[PVFMPRC0 + (pool * 64)]++; | ||
70 | + } | ||
71 | + } | ||
72 | } | ||
73 | |||
74 | static inline bool | ||
75 | @@ -XXX,XX +XXX,XX @@ igb_write_packet_to_guest(IGBCore *core, struct NetRxPkt *pkt, | ||
76 | |||
77 | } while (desc_offset < total_size); | ||
78 | |||
79 | - igb_update_rx_stats(core, size, total_size); | ||
80 | + igb_update_rx_stats(core, rxi, size, total_size); | ||
81 | } | ||
82 | |||
83 | static inline void | ||
84 | -- | 120 | -- |
85 | 2.7.4 | 121 | 2.7.4 |
122 | |||
123 | diff view generated by jsdifflib |
1 | From: Akihiko Odaki <akihiko.odaki@daynix.com> | 1 | From: Zhang Chen <chen.zhang@intel.com> |
---|---|---|---|
2 | 2 | ||
3 | The current implementation of igb uses only part of a advanced Tx | 3 | Make the comments consistent with the REGULAR_PACKET_CHECK_MS. |
4 | context descriptor and first data descriptor because it misses some | ||
5 | features and sniffs the trait of the packet instead of respecting the | ||
6 | packet type specified in the descriptor. However, we will certainly | ||
7 | need the entire Tx context descriptor when we update igb to respect | ||
8 | these ignored fields. Save the entire context descriptor and first | ||
9 | data descriptor except the buffer address to prepare for such a change. | ||
10 | 4 | ||
11 | This also introduces the distinction of contexts with different | 5 | Signed-off-by: Zhang Chen <chen.zhang@intel.com> |
12 | indexes, which was not present in e1000e but in igb. | ||
13 | |||
14 | Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> | ||
15 | Reviewed-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | ||
16 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 6 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
17 | --- | 7 | --- |
18 | hw/net/igb.c | 26 +++++++++++++++++++------- | 8 | net/colo-compare.c | 2 +- |
19 | hw/net/igb_core.c | 39 +++++++++++++++++++-------------------- | 9 | 1 file changed, 1 insertion(+), 1 deletion(-) |
20 | hw/net/igb_core.h | 8 +++----- | ||
21 | 3 files changed, 41 insertions(+), 32 deletions(-) | ||
22 | 10 | ||
23 | diff --git a/hw/net/igb.c b/hw/net/igb.c | 11 | diff --git a/net/colo-compare.c b/net/colo-compare.c |
24 | index XXXXXXX..XXXXXXX 100644 | 12 | index XXXXXXX..XXXXXXX 100644 |
25 | --- a/hw/net/igb.c | 13 | --- a/net/colo-compare.c |
26 | +++ b/hw/net/igb.c | 14 | +++ b/net/colo-compare.c |
27 | @@ -XXX,XX +XXX,XX @@ static int igb_post_load(void *opaque, int version_id) | 15 | @@ -XXX,XX +XXX,XX @@ static void colo_compare_complete(UserCreatable *uc, Error **errp) |
28 | return igb_core_post_load(&s->core); | ||
29 | } | ||
30 | |||
31 | -static const VMStateDescription igb_vmstate_tx = { | ||
32 | - .name = "igb-tx", | ||
33 | +static const VMStateDescription igb_vmstate_tx_ctx = { | ||
34 | + .name = "igb-tx-ctx", | ||
35 | .version_id = 1, | ||
36 | .minimum_version_id = 1, | ||
37 | .fields = (VMStateField[]) { | ||
38 | - VMSTATE_UINT16(vlan, struct igb_tx), | ||
39 | - VMSTATE_UINT16(mss, struct igb_tx), | ||
40 | - VMSTATE_BOOL(tse, struct igb_tx), | ||
41 | - VMSTATE_BOOL(ixsm, struct igb_tx), | ||
42 | - VMSTATE_BOOL(txsm, struct igb_tx), | ||
43 | + VMSTATE_UINT32(vlan_macip_lens, struct e1000_adv_tx_context_desc), | ||
44 | + VMSTATE_UINT32(seqnum_seed, struct e1000_adv_tx_context_desc), | ||
45 | + VMSTATE_UINT32(type_tucmd_mlhl, struct e1000_adv_tx_context_desc), | ||
46 | + VMSTATE_UINT32(mss_l4len_idx, struct e1000_adv_tx_context_desc), | ||
47 | + VMSTATE_END_OF_LIST() | ||
48 | + } | ||
49 | +}; | ||
50 | + | ||
51 | +static const VMStateDescription igb_vmstate_tx = { | ||
52 | + .name = "igb-tx", | ||
53 | + .version_id = 2, | ||
54 | + .minimum_version_id = 2, | ||
55 | + .fields = (VMStateField[]) { | ||
56 | + VMSTATE_STRUCT_ARRAY(ctx, struct igb_tx, 2, 0, igb_vmstate_tx_ctx, | ||
57 | + struct e1000_adv_tx_context_desc), | ||
58 | + VMSTATE_UINT32(first_cmd_type_len, struct igb_tx), | ||
59 | + VMSTATE_UINT32(first_olinfo_status, struct igb_tx), | ||
60 | VMSTATE_BOOL(first, struct igb_tx), | ||
61 | VMSTATE_BOOL(skip_cp, struct igb_tx), | ||
62 | VMSTATE_END_OF_LIST() | ||
63 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | ||
64 | index XXXXXXX..XXXXXXX 100644 | ||
65 | --- a/hw/net/igb_core.c | ||
66 | +++ b/hw/net/igb_core.c | ||
67 | @@ -XXX,XX +XXX,XX @@ igb_rss_parse_packet(IGBCore *core, struct NetRxPkt *pkt, bool tx, | ||
68 | static bool | ||
69 | igb_setup_tx_offloads(IGBCore *core, struct igb_tx *tx) | ||
70 | { | ||
71 | - if (tx->tse) { | ||
72 | - if (!net_tx_pkt_build_vheader(tx->tx_pkt, true, true, tx->mss)) { | ||
73 | + if (tx->first_cmd_type_len & E1000_ADVTXD_DCMD_TSE) { | ||
74 | + uint32_t idx = (tx->first_olinfo_status >> 4) & 1; | ||
75 | + uint32_t mss = tx->ctx[idx].mss_l4len_idx >> 16; | ||
76 | + if (!net_tx_pkt_build_vheader(tx->tx_pkt, true, true, mss)) { | ||
77 | return false; | ||
78 | } | ||
79 | |||
80 | @@ -XXX,XX +XXX,XX @@ igb_setup_tx_offloads(IGBCore *core, struct igb_tx *tx) | ||
81 | return true; | ||
82 | } | 16 | } |
83 | 17 | ||
84 | - if (tx->txsm) { | 18 | if (!s->expired_scan_cycle) { |
85 | + if (tx->first_olinfo_status & E1000_ADVTXD_POTS_TXSM) { | 19 | - /* Set default value to 3000 MS */ |
86 | if (!net_tx_pkt_build_vheader(tx->tx_pkt, false, true, 0)) { | 20 | + /* Set default value to 1000 MS */ |
87 | return false; | 21 | s->expired_scan_cycle = REGULAR_PACKET_CHECK_MS; |
88 | } | ||
89 | } | 22 | } |
90 | 23 | ||
91 | - if (tx->ixsm) { | ||
92 | + if (tx->first_olinfo_status & E1000_ADVTXD_POTS_IXSM) { | ||
93 | net_tx_pkt_update_ip_hdr_checksum(tx->tx_pkt); | ||
94 | } | ||
95 | |||
96 | @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, | ||
97 | { | ||
98 | struct e1000_adv_tx_context_desc *tx_ctx_desc; | ||
99 | uint32_t cmd_type_len; | ||
100 | - uint32_t olinfo_status; | ||
101 | + uint32_t idx; | ||
102 | uint64_t buffer_addr; | ||
103 | uint16_t length; | ||
104 | |||
105 | @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, | ||
106 | E1000_ADVTXD_DTYP_DATA) { | ||
107 | /* advanced transmit data descriptor */ | ||
108 | if (tx->first) { | ||
109 | - olinfo_status = le32_to_cpu(tx_desc->read.olinfo_status); | ||
110 | - | ||
111 | - tx->tse = !!(cmd_type_len & E1000_ADVTXD_DCMD_TSE); | ||
112 | - tx->ixsm = !!(olinfo_status & E1000_ADVTXD_POTS_IXSM); | ||
113 | - tx->txsm = !!(olinfo_status & E1000_ADVTXD_POTS_TXSM); | ||
114 | - | ||
115 | + tx->first_cmd_type_len = cmd_type_len; | ||
116 | + tx->first_olinfo_status = le32_to_cpu(tx_desc->read.olinfo_status); | ||
117 | tx->first = false; | ||
118 | } | ||
119 | } else if ((cmd_type_len & E1000_ADVTXD_DTYP_CTXT) == | ||
120 | E1000_ADVTXD_DTYP_CTXT) { | ||
121 | /* advanced transmit context descriptor */ | ||
122 | tx_ctx_desc = (struct e1000_adv_tx_context_desc *)tx_desc; | ||
123 | - tx->vlan = le32_to_cpu(tx_ctx_desc->vlan_macip_lens) >> 16; | ||
124 | - tx->mss = le32_to_cpu(tx_ctx_desc->mss_l4len_idx) >> 16; | ||
125 | + idx = (le32_to_cpu(tx_ctx_desc->mss_l4len_idx) >> 4) & 1; | ||
126 | + tx->ctx[idx].vlan_macip_lens = le32_to_cpu(tx_ctx_desc->vlan_macip_lens); | ||
127 | + tx->ctx[idx].seqnum_seed = le32_to_cpu(tx_ctx_desc->seqnum_seed); | ||
128 | + tx->ctx[idx].type_tucmd_mlhl = le32_to_cpu(tx_ctx_desc->type_tucmd_mlhl); | ||
129 | + tx->ctx[idx].mss_l4len_idx = le32_to_cpu(tx_ctx_desc->mss_l4len_idx); | ||
130 | return; | ||
131 | } else { | ||
132 | /* unknown descriptor type */ | ||
133 | @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, | ||
134 | if (cmd_type_len & E1000_TXD_CMD_EOP) { | ||
135 | if (!tx->skip_cp && net_tx_pkt_parse(tx->tx_pkt)) { | ||
136 | if (cmd_type_len & E1000_TXD_CMD_VLE) { | ||
137 | - net_tx_pkt_setup_vlan_header_ex(tx->tx_pkt, tx->vlan, | ||
138 | - core->mac[VET] & 0xffff); | ||
139 | + idx = (tx->first_olinfo_status >> 4) & 1; | ||
140 | + uint16_t vlan = tx->ctx[idx].vlan_macip_lens >> 16; | ||
141 | + uint16_t vet = core->mac[VET] & 0xffff; | ||
142 | + net_tx_pkt_setup_vlan_header_ex(tx->tx_pkt, vlan, vet); | ||
143 | } | ||
144 | if (igb_tx_pkt_send(core, tx, queue_index)) { | ||
145 | igb_on_tx_done_update_stats(core, tx->tx_pkt); | ||
146 | @@ -XXX,XX +XXX,XX @@ static void igb_reset(IGBCore *core, bool sw) | ||
147 | for (i = 0; i < ARRAY_SIZE(core->tx); i++) { | ||
148 | tx = &core->tx[i]; | ||
149 | net_tx_pkt_reset(tx->tx_pkt); | ||
150 | - tx->vlan = 0; | ||
151 | - tx->mss = 0; | ||
152 | - tx->tse = false; | ||
153 | - tx->ixsm = false; | ||
154 | - tx->txsm = false; | ||
155 | + memset(tx->ctx, 0, sizeof(tx->ctx)); | ||
156 | tx->first = true; | ||
157 | tx->skip_cp = false; | ||
158 | } | ||
159 | diff --git a/hw/net/igb_core.h b/hw/net/igb_core.h | ||
160 | index XXXXXXX..XXXXXXX 100644 | ||
161 | --- a/hw/net/igb_core.h | ||
162 | +++ b/hw/net/igb_core.h | ||
163 | @@ -XXX,XX +XXX,XX @@ struct IGBCore { | ||
164 | QEMUTimer *autoneg_timer; | ||
165 | |||
166 | struct igb_tx { | ||
167 | - uint16_t vlan; /* VLAN Tag */ | ||
168 | - uint16_t mss; /* Maximum Segment Size */ | ||
169 | - bool tse; /* TCP/UDP Segmentation Enable */ | ||
170 | - bool ixsm; /* Insert IP Checksum */ | ||
171 | - bool txsm; /* Insert TCP/UDP Checksum */ | ||
172 | + struct e1000_adv_tx_context_desc ctx[2]; | ||
173 | + uint32_t first_cmd_type_len; | ||
174 | + uint32_t first_olinfo_status; | ||
175 | |||
176 | bool first; | ||
177 | bool skip_cp; | ||
178 | -- | 24 | -- |
179 | 2.7.4 | 25 | 2.7.4 |
26 | |||
27 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Rao Lei <lei.rao@intel.com> |
---|---|---|---|
2 | 2 | ||
3 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 3 | This patch is to improve the logic of QEMU main thread sleep code in |
4 | qemu_chr_write_buffer() where it can be blocked and can't run other | ||
5 | coroutines during COLO IO stress test. | ||
6 | |||
7 | Our approach is to put filter_send() in a coroutine. In this way, | ||
8 | filter_send() will call qemu_coroutine_yield() in qemu_co_sleep_ns(), | ||
9 | so that it can be scheduled out and QEMU main thread has opportunity to | ||
10 | run other tasks. | ||
11 | |||
12 | Signed-off-by: Lei Rao <lei.rao@intel.com> | ||
13 | Signed-off-by: Zhang Chen <chen.zhang@intel.com> | ||
14 | Reviewed-by: Li Zhijian <lizhijian@fujitsu.com> | ||
15 | Reviewed-by: Zhang Chen <chen.zhang@intel.com> | ||
4 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 16 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
5 | --- | 17 | --- |
6 | hw/net/igb_core.c | 41 ++++++++++++++++++++++++++++++++++++----- | 18 | net/filter-mirror.c | 66 ++++++++++++++++++++++++++++++++++++++++++----------- |
7 | 1 file changed, 36 insertions(+), 5 deletions(-) | 19 | 1 file changed, 53 insertions(+), 13 deletions(-) |
8 | 20 | ||
9 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | 21 | diff --git a/net/filter-mirror.c b/net/filter-mirror.c |
10 | index XXXXXXX..XXXXXXX 100644 | 22 | index XXXXXXX..XXXXXXX 100644 |
11 | --- a/hw/net/igb_core.c | 23 | --- a/net/filter-mirror.c |
12 | +++ b/hw/net/igb_core.c | 24 | +++ b/net/filter-mirror.c |
13 | @@ -XXX,XX +XXX,XX @@ igb_rx_l4_cso_enabled(IGBCore *core) | 25 | @@ -XXX,XX +XXX,XX @@ |
14 | return !!(core->mac[RXCSUM] & E1000_RXCSUM_TUOFLD); | 26 | #include "chardev/char-fe.h" |
27 | #include "qemu/iov.h" | ||
28 | #include "qemu/sockets.h" | ||
29 | +#include "block/aio-wait.h" | ||
30 | |||
31 | #define TYPE_FILTER_MIRROR "filter-mirror" | ||
32 | typedef struct MirrorState MirrorState; | ||
33 | @@ -XXX,XX +XXX,XX @@ struct MirrorState { | ||
34 | bool vnet_hdr; | ||
35 | }; | ||
36 | |||
37 | -static int filter_send(MirrorState *s, | ||
38 | - const struct iovec *iov, | ||
39 | - int iovcnt) | ||
40 | +typedef struct FilterSendCo { | ||
41 | + MirrorState *s; | ||
42 | + char *buf; | ||
43 | + ssize_t size; | ||
44 | + bool done; | ||
45 | + int ret; | ||
46 | +} FilterSendCo; | ||
47 | + | ||
48 | +static int _filter_send(MirrorState *s, | ||
49 | + char *buf, | ||
50 | + ssize_t size) | ||
51 | { | ||
52 | NetFilterState *nf = NETFILTER(s); | ||
53 | int ret = 0; | ||
54 | - ssize_t size = 0; | ||
55 | uint32_t len = 0; | ||
56 | - char *buf; | ||
57 | - | ||
58 | - size = iov_size(iov, iovcnt); | ||
59 | - if (!size) { | ||
60 | - return 0; | ||
61 | - } | ||
62 | |||
63 | len = htonl(size); | ||
64 | ret = qemu_chr_fe_write_all(&s->chr_out, (uint8_t *)&len, sizeof(len)); | ||
65 | @@ -XXX,XX +XXX,XX @@ static int filter_send(MirrorState *s, | ||
66 | } | ||
67 | } | ||
68 | |||
69 | - buf = g_malloc(size); | ||
70 | - iov_to_buf(iov, iovcnt, 0, buf, size); | ||
71 | ret = qemu_chr_fe_write_all(&s->chr_out, (uint8_t *)buf, size); | ||
72 | - g_free(buf); | ||
73 | if (ret != size) { | ||
74 | goto err; | ||
75 | } | ||
76 | @@ -XXX,XX +XXX,XX @@ err: | ||
77 | return ret < 0 ? ret : -EIO; | ||
15 | } | 78 | } |
16 | 79 | ||
17 | +static bool | 80 | +static void coroutine_fn filter_send_co(void *opaque) |
18 | +igb_rx_is_oversized(IGBCore *core, uint16_t qn, size_t size) | ||
19 | +{ | 81 | +{ |
20 | + uint16_t pool = qn % IGB_NUM_VM_POOLS; | 82 | + FilterSendCo *data = opaque; |
21 | + bool lpe = !!(core->mac[VMOLR0 + pool] & E1000_VMOLR_LPE); | ||
22 | + int max_ethernet_lpe_size = | ||
23 | + core->mac[VMOLR0 + pool] & E1000_VMOLR_RLPML_MASK; | ||
24 | + int max_ethernet_vlan_size = 1522; | ||
25 | + | 83 | + |
26 | + return size > (lpe ? max_ethernet_lpe_size : max_ethernet_vlan_size); | 84 | + data->ret = _filter_send(data->s, data->buf, data->size); |
85 | + data->done = true; | ||
86 | + g_free(data->buf); | ||
87 | + aio_wait_kick(); | ||
27 | +} | 88 | +} |
28 | + | 89 | + |
29 | static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | 90 | +static int filter_send(MirrorState *s, |
30 | - E1000E_RSSInfo *rss_info, bool *external_tx) | 91 | + const struct iovec *iov, |
31 | + size_t size, E1000E_RSSInfo *rss_info, | 92 | + int iovcnt) |
32 | + bool *external_tx) | 93 | +{ |
33 | { | 94 | + ssize_t size = iov_size(iov, iovcnt); |
34 | static const int ta_shift[] = { 4, 3, 2, 0 }; | 95 | + char *buf = NULL; |
35 | uint32_t f, ra[2], *macp, rctl = core->mac[RCTL]; | ||
36 | uint16_t queues = 0; | ||
37 | + uint16_t oversized = 0; | ||
38 | uint16_t vid = lduw_be_p(&PKT_GET_VLAN_HDR(ehdr)->h_tci) & VLAN_VID_MASK; | ||
39 | bool accepted = false; | ||
40 | int i; | ||
41 | @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | ||
42 | } | ||
43 | |||
44 | queues &= core->mac[VFRE]; | ||
45 | - igb_rss_parse_packet(core, core->rx_pkt, external_tx != NULL, rss_info); | ||
46 | - if (rss_info->queue & 1) { | ||
47 | - queues <<= 8; | ||
48 | + if (queues) { | ||
49 | + for (i = 0; i < IGB_NUM_VM_POOLS; i++) { | ||
50 | + if ((queues & BIT(i)) && igb_rx_is_oversized(core, i, size)) { | ||
51 | + oversized |= BIT(i); | ||
52 | + } | ||
53 | + } | ||
54 | + /* 8.19.37 increment ROC if packet is oversized for all queues */ | ||
55 | + if (oversized == queues) { | ||
56 | + trace_e1000x_rx_oversized(size); | ||
57 | + e1000x_inc_reg_if_not_full(core->mac, ROC); | ||
58 | + } | ||
59 | + queues &= ~oversized; | ||
60 | + } | ||
61 | + | 96 | + |
62 | + if (queues) { | 97 | + if (!size) { |
63 | + igb_rss_parse_packet(core, core->rx_pkt, | 98 | + return 0; |
64 | + external_tx != NULL, rss_info); | 99 | + } |
65 | + if (rss_info->queue & 1) { | 100 | + |
66 | + queues <<= 8; | 101 | + buf = g_malloc(size); |
67 | + } | 102 | + iov_to_buf(iov, iovcnt, 0, buf, size); |
68 | } | 103 | + |
69 | } else { | 104 | + FilterSendCo data = { |
70 | switch (net_rx_pkt_get_packet_type(core->rx_pkt)) { | 105 | + .s = s, |
71 | @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, | 106 | + .size = size, |
72 | e1000x_vlan_enabled(core->mac), | 107 | + .buf = buf, |
73 | core->mac[VET] & 0xffff); | 108 | + .ret = 0, |
74 | 109 | + }; | |
75 | - queues = igb_receive_assign(core, ehdr, &rss_info, external_tx); | 110 | + |
76 | + queues = igb_receive_assign(core, ehdr, size, &rss_info, external_tx); | 111 | + Coroutine *co = qemu_coroutine_create(filter_send_co, &data); |
77 | if (!queues) { | 112 | + qemu_coroutine_enter(co); |
78 | trace_e1000e_rx_flt_dropped(); | 113 | + |
79 | return orig_size; | 114 | + while (!data.done) { |
115 | + aio_poll(qemu_get_aio_context(), true); | ||
116 | + } | ||
117 | + | ||
118 | + return data.ret; | ||
119 | +} | ||
120 | + | ||
121 | static void redirector_to_filter(NetFilterState *nf, | ||
122 | const uint8_t *buf, | ||
123 | int len) | ||
80 | -- | 124 | -- |
81 | 2.7.4 | 125 | 2.7.4 |
126 | |||
127 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | RSS for VFs is only enabled if VMOLR[n].RSSE is set. | 3 | vmnet.framework dependency is added with 'vmnet' option |
4 | to enable or disable it. Default value is 'auto'. | ||
4 | 5 | ||
5 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 6 | vmnet features to be used are available since macOS 11.0, |
7 | corresponding probe is created into meson.build. | ||
8 | |||
9 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
6 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 10 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
7 | --- | 11 | --- |
8 | hw/net/igb_core.c | 9 ++++++++- | 12 | meson.build | 16 +++++++++++++++- |
9 | 1 file changed, 8 insertions(+), 1 deletion(-) | 13 | meson_options.txt | 2 ++ |
14 | scripts/meson-buildoptions.sh | 3 +++ | ||
15 | 3 files changed, 20 insertions(+), 1 deletion(-) | ||
10 | 16 | ||
11 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | 17 | diff --git a/meson.build b/meson.build |
12 | index XXXXXXX..XXXXXXX 100644 | 18 | index XXXXXXX..XXXXXXX 100644 |
13 | --- a/hw/net/igb_core.c | 19 | --- a/meson.build |
14 | +++ b/hw/net/igb_core.c | 20 | +++ b/meson.build |
15 | @@ -XXX,XX +XXX,XX @@ static uint16_t igb_receive_assign(IGBCore *core, const struct eth_header *ehdr, | 21 | @@ -XXX,XX +XXX,XX @@ if cocoa.found() and get_option('gtk').enabled() |
16 | if (queues) { | 22 | error('Cocoa and GTK+ cannot be enabled at the same time') |
17 | igb_rss_parse_packet(core, core->rx_pkt, | 23 | endif |
18 | external_tx != NULL, rss_info); | 24 | |
19 | + /* Sec 8.26.1: PQn = VFn + VQn*8 */ | 25 | +vmnet = dependency('appleframeworks', modules: 'vmnet', required: get_option('vmnet')) |
20 | if (rss_info->queue & 1) { | 26 | +if vmnet.found() and not cc.has_header_symbol('vmnet/vmnet.h', |
21 | - queues <<= 8; | 27 | + 'VMNET_BRIDGED_MODE', |
22 | + for (i = 0; i < IGB_NUM_VM_POOLS; i++) { | 28 | + dependencies: vmnet) |
23 | + if ((queues & BIT(i)) && | 29 | + vmnet = not_found |
24 | + (core->mac[VMOLR0 + i] & E1000_VMOLR_RSSE)) { | 30 | + if get_option('vmnet').enabled() |
25 | + queues |= BIT(i + IGB_NUM_VM_POOLS); | 31 | + error('vmnet.framework API is outdated') |
26 | + queues &= ~BIT(i); | 32 | + else |
27 | + } | 33 | + warning('vmnet.framework API is outdated, disabling') |
28 | + } | 34 | + endif |
29 | } | 35 | +endif |
30 | } | 36 | + |
31 | } else { | 37 | seccomp = not_found |
38 | if not get_option('seccomp').auto() or have_system or have_tools | ||
39 | seccomp = dependency('libseccomp', version: '>=2.3.0', | ||
40 | @@ -XXX,XX +XXX,XX @@ config_host_data.set('CONFIG_SECCOMP', seccomp.found()) | ||
41 | config_host_data.set('CONFIG_SNAPPY', snappy.found()) | ||
42 | config_host_data.set('CONFIG_USB_LIBUSB', libusb.found()) | ||
43 | config_host_data.set('CONFIG_VDE', vde.found()) | ||
44 | +config_host_data.set('CONFIG_VMNET', vmnet.found()) | ||
45 | config_host_data.set('CONFIG_VHOST_USER_BLK_SERVER', have_vhost_user_blk_server) | ||
46 | config_host_data.set('CONFIG_VNC', vnc.found()) | ||
47 | config_host_data.set('CONFIG_VNC_JPEG', jpeg.found()) | ||
48 | @@ -XXX,XX +XXX,XX @@ summary(summary_info, bool_yn: true, section: 'Crypto') | ||
49 | # Libraries | ||
50 | summary_info = {} | ||
51 | if targetos == 'darwin' | ||
52 | - summary_info += {'Cocoa support': cocoa} | ||
53 | + summary_info += {'Cocoa support': cocoa} | ||
54 | + summary_info += {'vmnet.framework support': vmnet} | ||
55 | endif | ||
56 | summary_info += {'SDL support': sdl} | ||
57 | summary_info += {'SDL image support': sdl_image} | ||
58 | diff --git a/meson_options.txt b/meson_options.txt | ||
59 | index XXXXXXX..XXXXXXX 100644 | ||
60 | --- a/meson_options.txt | ||
61 | +++ b/meson_options.txt | ||
62 | @@ -XXX,XX +XXX,XX @@ option('netmap', type : 'feature', value : 'auto', | ||
63 | description: 'netmap network backend support') | ||
64 | option('vde', type : 'feature', value : 'auto', | ||
65 | description: 'vde network backend support') | ||
66 | +option('vmnet', type : 'feature', value : 'auto', | ||
67 | + description: 'vmnet.framework network backend support') | ||
68 | option('virglrenderer', type : 'feature', value : 'auto', | ||
69 | description: 'virgl rendering support') | ||
70 | option('vnc', type : 'feature', value : 'auto', | ||
71 | diff --git a/scripts/meson-buildoptions.sh b/scripts/meson-buildoptions.sh | ||
72 | index XXXXXXX..XXXXXXX 100644 | ||
73 | --- a/scripts/meson-buildoptions.sh | ||
74 | +++ b/scripts/meson-buildoptions.sh | ||
75 | @@ -XXX,XX +XXX,XX @@ meson_options_help() { | ||
76 | printf "%s\n" ' u2f U2F emulation support' | ||
77 | printf "%s\n" ' usb-redir libusbredir support' | ||
78 | printf "%s\n" ' vde vde network backend support' | ||
79 | + printf "%s\n" ' vmnet vmnet.framework network backend support' | ||
80 | printf "%s\n" ' vhost-user-blk-server' | ||
81 | printf "%s\n" ' build vhost-user-blk server' | ||
82 | printf "%s\n" ' virglrenderer virgl rendering support' | ||
83 | @@ -XXX,XX +XXX,XX @@ _meson_option_parse() { | ||
84 | --disable-usb-redir) printf "%s" -Dusb_redir=disabled ;; | ||
85 | --enable-vde) printf "%s" -Dvde=enabled ;; | ||
86 | --disable-vde) printf "%s" -Dvde=disabled ;; | ||
87 | + --enable-vmnet) printf "%s" -Dvmnet=enabled ;; | ||
88 | + --disable-vmnet) printf "%s" -Dvmnet=disabled ;; | ||
89 | --enable-vhost-user-blk-server) printf "%s" -Dvhost_user_blk_server=enabled ;; | ||
90 | --disable-vhost-user-blk-server) printf "%s" -Dvhost_user_blk_server=disabled ;; | ||
91 | --enable-virglrenderer) printf "%s" -Dvirglrenderer=enabled ;; | ||
32 | -- | 92 | -- |
33 | 2.7.4 | 93 | 2.7.4 |
94 | |||
95 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | IGB uses RXDW ICR bit to indicate that rx descriptor has been written | 3 | Create separate netdevs for each vmnet operating mode: |
4 | back. This is the same as RXT0 bit in older HW. | 4 | - vmnet-host |
5 | - vmnet-shared | ||
6 | - vmnet-bridged | ||
5 | 7 | ||
6 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 8 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
7 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 9 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
8 | --- | 10 | --- |
9 | hw/net/e1000x_regs.h | 4 ++++ | 11 | net/clients.h | 11 +++++ |
10 | hw/net/igb_core.c | 2 +- | 12 | net/meson.build | 7 +++ |
11 | 2 files changed, 5 insertions(+), 1 deletion(-) | 13 | net/net.c | 10 ++++ |
14 | net/vmnet-bridged.m | 25 ++++++++++ | ||
15 | net/vmnet-common.m | 19 ++++++++ | ||
16 | net/vmnet-host.c | 24 ++++++++++ | ||
17 | net/vmnet-shared.c | 25 ++++++++++ | ||
18 | net/vmnet_int.h | 25 ++++++++++ | ||
19 | qapi/net.json | 133 +++++++++++++++++++++++++++++++++++++++++++++++++++- | ||
20 | 9 files changed, 277 insertions(+), 2 deletions(-) | ||
21 | create mode 100644 net/vmnet-bridged.m | ||
22 | create mode 100644 net/vmnet-common.m | ||
23 | create mode 100644 net/vmnet-host.c | ||
24 | create mode 100644 net/vmnet-shared.c | ||
25 | create mode 100644 net/vmnet_int.h | ||
12 | 26 | ||
13 | diff --git a/hw/net/e1000x_regs.h b/hw/net/e1000x_regs.h | 27 | diff --git a/net/clients.h b/net/clients.h |
14 | index XXXXXXX..XXXXXXX 100644 | 28 | index XXXXXXX..XXXXXXX 100644 |
15 | --- a/hw/net/e1000x_regs.h | 29 | --- a/net/clients.h |
16 | +++ b/hw/net/e1000x_regs.h | 30 | +++ b/net/clients.h |
17 | @@ -XXX,XX +XXX,XX @@ | 31 | @@ -XXX,XX +XXX,XX @@ int net_init_vhost_user(const Netdev *netdev, const char *name, |
18 | #define E1000_ICR_RXDMT0 0x00000010 /* rx desc min. threshold (0) */ | 32 | |
19 | #define E1000_ICR_RXO 0x00000040 /* rx overrun */ | 33 | int net_init_vhost_vdpa(const Netdev *netdev, const char *name, |
20 | #define E1000_ICR_RXT0 0x00000080 /* rx timer intr (ring 0) */ | 34 | NetClientState *peer, Error **errp); |
21 | +#define E1000_ICR_RXDW 0x00000080 /* rx desc written back */ | 35 | +#ifdef CONFIG_VMNET |
22 | #define E1000_ICR_MDAC 0x00000200 /* MDIO access complete */ | 36 | +int net_init_vmnet_host(const Netdev *netdev, const char *name, |
23 | #define E1000_ICR_RXCFG 0x00000400 /* RX /c/ ordered set */ | 37 | + NetClientState *peer, Error **errp); |
24 | #define E1000_ICR_GPI_EN0 0x00000800 /* GP Int 0 */ | 38 | + |
25 | @@ -XXX,XX +XXX,XX @@ | 39 | +int net_init_vmnet_shared(const Netdev *netdev, const char *name, |
26 | #define E1000_ICS_RXDMT0 E1000_ICR_RXDMT0 /* rx desc min. threshold */ | 40 | + NetClientState *peer, Error **errp); |
27 | #define E1000_ICS_RXO E1000_ICR_RXO /* rx overrun */ | 41 | + |
28 | #define E1000_ICS_RXT0 E1000_ICR_RXT0 /* rx timer intr */ | 42 | +int net_init_vmnet_bridged(const Netdev *netdev, const char *name, |
29 | +#define E1000_ICS_RXDW E1000_ICR_RXDW /* rx desc written back */ | 43 | + NetClientState *peer, Error **errp); |
30 | #define E1000_ICS_MDAC E1000_ICR_MDAC /* MDIO access complete */ | 44 | +#endif /* CONFIG_VMNET */ |
31 | #define E1000_ICS_RXCFG E1000_ICR_RXCFG /* RX /c/ ordered set */ | 45 | + |
32 | #define E1000_ICS_GPI_EN0 E1000_ICR_GPI_EN0 /* GP Int 0 */ | 46 | #endif /* QEMU_NET_CLIENTS_H */ |
33 | @@ -XXX,XX +XXX,XX @@ | 47 | diff --git a/net/meson.build b/net/meson.build |
34 | #define E1000_IMS_RXDMT0 E1000_ICR_RXDMT0 /* rx desc min. threshold */ | ||
35 | #define E1000_IMS_RXO E1000_ICR_RXO /* rx overrun */ | ||
36 | #define E1000_IMS_RXT0 E1000_ICR_RXT0 /* rx timer intr */ | ||
37 | +#define E1000_IMS_RXDW E1000_ICR_RXDW /* rx desc written back */ | ||
38 | #define E1000_IMS_MDAC E1000_ICR_MDAC /* MDIO access complete */ | ||
39 | #define E1000_IMS_RXCFG E1000_ICR_RXCFG /* RX /c/ ordered set */ | ||
40 | #define E1000_IMS_GPI_EN0 E1000_ICR_GPI_EN0 /* GP Int 0 */ | ||
41 | @@ -XXX,XX +XXX,XX @@ | ||
42 | #define E1000_IMC_RXDMT0 E1000_ICR_RXDMT0 /* rx desc min. threshold */ | ||
43 | #define E1000_IMC_RXO E1000_ICR_RXO /* rx overrun */ | ||
44 | #define E1000_IMC_RXT0 E1000_ICR_RXT0 /* rx timer intr */ | ||
45 | +#define E1000_IMC_RXDW E1000_ICR_RXDW /* rx desc written back */ | ||
46 | #define E1000_IMC_MDAC E1000_ICR_MDAC /* MDIO access complete */ | ||
47 | #define E1000_IMC_RXCFG E1000_ICR_RXCFG /* RX /c/ ordered set */ | ||
48 | #define E1000_IMC_GPI_EN0 E1000_ICR_GPI_EN0 /* GP Int 0 */ | ||
49 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | ||
50 | index XXXXXXX..XXXXXXX 100644 | 48 | index XXXXXXX..XXXXXXX 100644 |
51 | --- a/hw/net/igb_core.c | 49 | --- a/net/meson.build |
52 | +++ b/hw/net/igb_core.c | 50 | +++ b/net/meson.build |
53 | @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, | 51 | @@ -XXX,XX +XXX,XX @@ softmmu_ss.add(when: 'CONFIG_POSIX', if_true: files(tap_posix)) |
54 | continue; | 52 | softmmu_ss.add(when: 'CONFIG_WIN32', if_true: files('tap-win32.c')) |
55 | } | 53 | softmmu_ss.add(when: 'CONFIG_VHOST_NET_VDPA', if_true: files('vhost-vdpa.c')) |
56 | 54 | ||
57 | - n |= E1000_ICR_RXT0; | 55 | +vmnet_files = files( |
58 | + n |= E1000_ICR_RXDW; | 56 | + 'vmnet-common.m', |
59 | 57 | + 'vmnet-bridged.m', | |
60 | igb_rx_fix_l4_csum(core, core->rx_pkt); | 58 | + 'vmnet-host.c', |
61 | igb_write_packet_to_guest(core, core->rx_pkt, &rxr, &rss_info); | 59 | + 'vmnet-shared.c' |
60 | +) | ||
61 | +softmmu_ss.add(when: vmnet, if_true: vmnet_files) | ||
62 | subdir('can') | ||
63 | diff --git a/net/net.c b/net/net.c | ||
64 | index XXXXXXX..XXXXXXX 100644 | ||
65 | --- a/net/net.c | ||
66 | +++ b/net/net.c | ||
67 | @@ -XXX,XX +XXX,XX @@ static int (* const net_client_init_fun[NET_CLIENT_DRIVER__MAX])( | ||
68 | #ifdef CONFIG_L2TPV3 | ||
69 | [NET_CLIENT_DRIVER_L2TPV3] = net_init_l2tpv3, | ||
70 | #endif | ||
71 | +#ifdef CONFIG_VMNET | ||
72 | + [NET_CLIENT_DRIVER_VMNET_HOST] = net_init_vmnet_host, | ||
73 | + [NET_CLIENT_DRIVER_VMNET_SHARED] = net_init_vmnet_shared, | ||
74 | + [NET_CLIENT_DRIVER_VMNET_BRIDGED] = net_init_vmnet_bridged, | ||
75 | +#endif /* CONFIG_VMNET */ | ||
76 | }; | ||
77 | |||
78 | |||
79 | @@ -XXX,XX +XXX,XX @@ void show_netdevs(void) | ||
80 | #ifdef CONFIG_VHOST_VDPA | ||
81 | "vhost-vdpa", | ||
82 | #endif | ||
83 | +#ifdef CONFIG_VMNET | ||
84 | + "vmnet-host", | ||
85 | + "vmnet-shared", | ||
86 | + "vmnet-bridged", | ||
87 | +#endif | ||
88 | }; | ||
89 | |||
90 | qemu_printf("Available netdev backend types:\n"); | ||
91 | diff --git a/net/vmnet-bridged.m b/net/vmnet-bridged.m | ||
92 | new file mode 100644 | ||
93 | index XXXXXXX..XXXXXXX | ||
94 | --- /dev/null | ||
95 | +++ b/net/vmnet-bridged.m | ||
96 | @@ -XXX,XX +XXX,XX @@ | ||
97 | +/* | ||
98 | + * vmnet-bridged.m | ||
99 | + * | ||
100 | + * Copyright(c) 2021 Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
101 | + * | ||
102 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
103 | + * See the COPYING file in the top-level directory. | ||
104 | + * | ||
105 | + */ | ||
106 | + | ||
107 | +#include "qemu/osdep.h" | ||
108 | +#include "qapi/qapi-types-net.h" | ||
109 | +#include "vmnet_int.h" | ||
110 | +#include "clients.h" | ||
111 | +#include "qemu/error-report.h" | ||
112 | +#include "qapi/error.h" | ||
113 | + | ||
114 | +#include <vmnet/vmnet.h> | ||
115 | + | ||
116 | +int net_init_vmnet_bridged(const Netdev *netdev, const char *name, | ||
117 | + NetClientState *peer, Error **errp) | ||
118 | +{ | ||
119 | + error_setg(errp, "vmnet-bridged is not implemented yet"); | ||
120 | + return -1; | ||
121 | +} | ||
122 | diff --git a/net/vmnet-common.m b/net/vmnet-common.m | ||
123 | new file mode 100644 | ||
124 | index XXXXXXX..XXXXXXX | ||
125 | --- /dev/null | ||
126 | +++ b/net/vmnet-common.m | ||
127 | @@ -XXX,XX +XXX,XX @@ | ||
128 | +/* | ||
129 | + * vmnet-common.m - network client wrapper for Apple vmnet.framework | ||
130 | + * | ||
131 | + * Copyright(c) 2021 Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
132 | + * Copyright(c) 2021 Phillip Tennen <phillip@axleos.com> | ||
133 | + * | ||
134 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
135 | + * See the COPYING file in the top-level directory. | ||
136 | + * | ||
137 | + */ | ||
138 | + | ||
139 | +#include "qemu/osdep.h" | ||
140 | +#include "qapi/qapi-types-net.h" | ||
141 | +#include "vmnet_int.h" | ||
142 | +#include "clients.h" | ||
143 | +#include "qemu/error-report.h" | ||
144 | +#include "qapi/error.h" | ||
145 | + | ||
146 | +#include <vmnet/vmnet.h> | ||
147 | diff --git a/net/vmnet-host.c b/net/vmnet-host.c | ||
148 | new file mode 100644 | ||
149 | index XXXXXXX..XXXXXXX | ||
150 | --- /dev/null | ||
151 | +++ b/net/vmnet-host.c | ||
152 | @@ -XXX,XX +XXX,XX @@ | ||
153 | +/* | ||
154 | + * vmnet-host.c | ||
155 | + * | ||
156 | + * Copyright(c) 2021 Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
157 | + * | ||
158 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
159 | + * See the COPYING file in the top-level directory. | ||
160 | + * | ||
161 | + */ | ||
162 | + | ||
163 | +#include "qemu/osdep.h" | ||
164 | +#include "qapi/qapi-types-net.h" | ||
165 | +#include "vmnet_int.h" | ||
166 | +#include "clients.h" | ||
167 | +#include "qemu/error-report.h" | ||
168 | +#include "qapi/error.h" | ||
169 | + | ||
170 | +#include <vmnet/vmnet.h> | ||
171 | + | ||
172 | +int net_init_vmnet_host(const Netdev *netdev, const char *name, | ||
173 | + NetClientState *peer, Error **errp) { | ||
174 | + error_setg(errp, "vmnet-host is not implemented yet"); | ||
175 | + return -1; | ||
176 | +} | ||
177 | diff --git a/net/vmnet-shared.c b/net/vmnet-shared.c | ||
178 | new file mode 100644 | ||
179 | index XXXXXXX..XXXXXXX | ||
180 | --- /dev/null | ||
181 | +++ b/net/vmnet-shared.c | ||
182 | @@ -XXX,XX +XXX,XX @@ | ||
183 | +/* | ||
184 | + * vmnet-shared.c | ||
185 | + * | ||
186 | + * Copyright(c) 2021 Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
187 | + * | ||
188 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
189 | + * See the COPYING file in the top-level directory. | ||
190 | + * | ||
191 | + */ | ||
192 | + | ||
193 | +#include "qemu/osdep.h" | ||
194 | +#include "qapi/qapi-types-net.h" | ||
195 | +#include "vmnet_int.h" | ||
196 | +#include "clients.h" | ||
197 | +#include "qemu/error-report.h" | ||
198 | +#include "qapi/error.h" | ||
199 | + | ||
200 | +#include <vmnet/vmnet.h> | ||
201 | + | ||
202 | +int net_init_vmnet_shared(const Netdev *netdev, const char *name, | ||
203 | + NetClientState *peer, Error **errp) | ||
204 | +{ | ||
205 | + error_setg(errp, "vmnet-shared is not implemented yet"); | ||
206 | + return -1; | ||
207 | +} | ||
208 | diff --git a/net/vmnet_int.h b/net/vmnet_int.h | ||
209 | new file mode 100644 | ||
210 | index XXXXXXX..XXXXXXX | ||
211 | --- /dev/null | ||
212 | +++ b/net/vmnet_int.h | ||
213 | @@ -XXX,XX +XXX,XX @@ | ||
214 | +/* | ||
215 | + * vmnet_int.h | ||
216 | + * | ||
217 | + * Copyright(c) 2021 Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
218 | + * | ||
219 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
220 | + * See the COPYING file in the top-level directory. | ||
221 | + * | ||
222 | + */ | ||
223 | +#ifndef VMNET_INT_H | ||
224 | +#define VMNET_INT_H | ||
225 | + | ||
226 | +#include "qemu/osdep.h" | ||
227 | +#include "vmnet_int.h" | ||
228 | +#include "clients.h" | ||
229 | + | ||
230 | +#include <vmnet/vmnet.h> | ||
231 | + | ||
232 | +typedef struct VmnetCommonState { | ||
233 | + NetClientState nc; | ||
234 | + | ||
235 | +} VmnetCommonState; | ||
236 | + | ||
237 | + | ||
238 | +#endif /* VMNET_INT_H */ | ||
239 | diff --git a/qapi/net.json b/qapi/net.json | ||
240 | index XXXXXXX..XXXXXXX 100644 | ||
241 | --- a/qapi/net.json | ||
242 | +++ b/qapi/net.json | ||
243 | @@ -XXX,XX +XXX,XX @@ | ||
244 | '*queues': 'int' } } | ||
245 | |||
246 | ## | ||
247 | +# @NetdevVmnetHostOptions: | ||
248 | +# | ||
249 | +# vmnet (host mode) network backend. | ||
250 | +# | ||
251 | +# Allows the vmnet interface to communicate with other vmnet | ||
252 | +# interfaces that are in host mode and also with the host. | ||
253 | +# | ||
254 | +# @start-address: The starting IPv4 address to use for the interface. | ||
255 | +# Must be in the private IP range (RFC 1918). Must be | ||
256 | +# specified along with @end-address and @subnet-mask. | ||
257 | +# This address is used as the gateway address. The | ||
258 | +# subsequent address up to and including end-address are | ||
259 | +# placed in the DHCP pool. | ||
260 | +# | ||
261 | +# @end-address: The DHCP IPv4 range end address to use for the | ||
262 | +# interface. Must be in the private IP range (RFC 1918). | ||
263 | +# Must be specified along with @start-address and | ||
264 | +# @subnet-mask. | ||
265 | +# | ||
266 | +# @subnet-mask: The IPv4 subnet mask to use on the interface. Must | ||
267 | +# be specified along with @start-address and @subnet-mask. | ||
268 | +# | ||
269 | +# @isolated: Enable isolation for this interface. Interface isolation | ||
270 | +# ensures that vmnet interface is not able to communicate | ||
271 | +# with any other vmnet interfaces. Only communication with | ||
272 | +# host is allowed. Available since macOS Big Sur 11.0. | ||
273 | +# | ||
274 | +# @net-uuid: The identifier (UUID) to uniquely identify the isolated | ||
275 | +# network vmnet interface should be added to. If | ||
276 | +# set, no DHCP service is provided for this interface and | ||
277 | +# network communication is allowed only with other interfaces | ||
278 | +# added to this network identified by the UUID. Available | ||
279 | +# since macOS Big Sur 11.0. | ||
280 | +# | ||
281 | +# Since: 7.0 | ||
282 | +## | ||
283 | +{ 'struct': 'NetdevVmnetHostOptions', | ||
284 | + 'data': { | ||
285 | + '*start-address': 'str', | ||
286 | + '*end-address': 'str', | ||
287 | + '*subnet-mask': 'str', | ||
288 | + '*isolated': 'bool', | ||
289 | + '*net-uuid': 'str' }, | ||
290 | + 'if': 'CONFIG_VMNET' } | ||
291 | + | ||
292 | +## | ||
293 | +# @NetdevVmnetSharedOptions: | ||
294 | +# | ||
295 | +# vmnet (shared mode) network backend. | ||
296 | +# | ||
297 | +# Allows traffic originating from the vmnet interface to reach the | ||
298 | +# Internet through a network address translator (NAT). | ||
299 | +# The vmnet interface can communicate with the host and with | ||
300 | +# other shared mode interfaces on the same subnet. If no DHCP | ||
301 | +# settings, subnet mask and IPv6 prefix specified, the interface can | ||
302 | +# communicate with any of other interfaces in shared mode. | ||
303 | +# | ||
304 | +# @start-address: The starting IPv4 address to use for the interface. | ||
305 | +# Must be in the private IP range (RFC 1918). Must be | ||
306 | +# specified along with @end-address and @subnet-mask. | ||
307 | +# This address is used as the gateway address. The | ||
308 | +# subsequent address up to and including end-address are | ||
309 | +# placed in the DHCP pool. | ||
310 | +# | ||
311 | +# @end-address: The DHCP IPv4 range end address to use for the | ||
312 | +# interface. Must be in the private IP range (RFC 1918). | ||
313 | +# Must be specified along with @start-address and @subnet-mask. | ||
314 | +# | ||
315 | +# @subnet-mask: The IPv4 subnet mask to use on the interface. Must | ||
316 | +# be specified along with @start-address and @subnet-mask. | ||
317 | +# | ||
318 | +# @isolated: Enable isolation for this interface. Interface isolation | ||
319 | +# ensures that vmnet interface is not able to communicate | ||
320 | +# with any other vmnet interfaces. Only communication with | ||
321 | +# host is allowed. Available since macOS Big Sur 11.0. | ||
322 | +# | ||
323 | +# @nat66-prefix: The IPv6 prefix to use into guest network. Must be a | ||
324 | +# unique local address i.e. start with fd00::/8 and have | ||
325 | +# length of 64. | ||
326 | +# | ||
327 | +# Since: 7.0 | ||
328 | +## | ||
329 | +{ 'struct': 'NetdevVmnetSharedOptions', | ||
330 | + 'data': { | ||
331 | + '*start-address': 'str', | ||
332 | + '*end-address': 'str', | ||
333 | + '*subnet-mask': 'str', | ||
334 | + '*isolated': 'bool', | ||
335 | + '*nat66-prefix': 'str' }, | ||
336 | + 'if': 'CONFIG_VMNET' } | ||
337 | + | ||
338 | +## | ||
339 | +# @NetdevVmnetBridgedOptions: | ||
340 | +# | ||
341 | +# vmnet (bridged mode) network backend. | ||
342 | +# | ||
343 | +# Bridges the vmnet interface with a physical network interface. | ||
344 | +# | ||
345 | +# @ifname: The name of the physical interface to be bridged. | ||
346 | +# | ||
347 | +# @isolated: Enable isolation for this interface. Interface isolation | ||
348 | +# ensures that vmnet interface is not able to communicate | ||
349 | +# with any other vmnet interfaces. Only communication with | ||
350 | +# host is allowed. Available since macOS Big Sur 11.0. | ||
351 | +# | ||
352 | +# Since: 7.0 | ||
353 | +## | ||
354 | +{ 'struct': 'NetdevVmnetBridgedOptions', | ||
355 | + 'data': { | ||
356 | + 'ifname': 'str', | ||
357 | + '*isolated': 'bool' }, | ||
358 | + 'if': 'CONFIG_VMNET' } | ||
359 | + | ||
360 | +## | ||
361 | # @NetClientDriver: | ||
362 | # | ||
363 | # Available netdev drivers. | ||
364 | @@ -XXX,XX +XXX,XX @@ | ||
365 | # Since: 2.7 | ||
366 | # | ||
367 | # @vhost-vdpa since 5.1 | ||
368 | +# @vmnet-host since 7.0 | ||
369 | +# @vmnet-shared since 7.0 | ||
370 | +# @vmnet-bridged since 7.0 | ||
371 | ## | ||
372 | { 'enum': 'NetClientDriver', | ||
373 | 'data': [ 'none', 'nic', 'user', 'tap', 'l2tpv3', 'socket', 'vde', | ||
374 | - 'bridge', 'hubport', 'netmap', 'vhost-user', 'vhost-vdpa' ] } | ||
375 | + 'bridge', 'hubport', 'netmap', 'vhost-user', 'vhost-vdpa', | ||
376 | + { 'name': 'vmnet-host', 'if': 'CONFIG_VMNET' }, | ||
377 | + { 'name': 'vmnet-shared', 'if': 'CONFIG_VMNET' }, | ||
378 | + { 'name': 'vmnet-bridged', 'if': 'CONFIG_VMNET' }] } | ||
379 | |||
380 | ## | ||
381 | # @Netdev: | ||
382 | @@ -XXX,XX +XXX,XX @@ | ||
383 | # Since: 1.2 | ||
384 | # | ||
385 | # 'l2tpv3' - since 2.1 | ||
386 | +# 'vmnet-host' - since 7.0 | ||
387 | +# 'vmnet-shared' - since 7.0 | ||
388 | +# 'vmnet-bridged' - since 7.0 | ||
389 | ## | ||
390 | { 'union': 'Netdev', | ||
391 | 'base': { 'id': 'str', 'type': 'NetClientDriver' }, | ||
392 | @@ -XXX,XX +XXX,XX @@ | ||
393 | 'hubport': 'NetdevHubPortOptions', | ||
394 | 'netmap': 'NetdevNetmapOptions', | ||
395 | 'vhost-user': 'NetdevVhostUserOptions', | ||
396 | - 'vhost-vdpa': 'NetdevVhostVDPAOptions' } } | ||
397 | + 'vhost-vdpa': 'NetdevVhostVDPAOptions', | ||
398 | + 'vmnet-host': { 'type': 'NetdevVmnetHostOptions', | ||
399 | + 'if': 'CONFIG_VMNET' }, | ||
400 | + 'vmnet-shared': { 'type': 'NetdevVmnetSharedOptions', | ||
401 | + 'if': 'CONFIG_VMNET' }, | ||
402 | + 'vmnet-bridged': { 'type': 'NetdevVmnetBridgedOptions', | ||
403 | + 'if': 'CONFIG_VMNET' } } } | ||
404 | |||
405 | ## | ||
406 | # @RxState: | ||
62 | -- | 407 | -- |
63 | 2.7.4 | 408 | 2.7.4 |
409 | |||
410 | diff view generated by jsdifflib |
1 | From: Akihiko Odaki <akihiko.odaki@daynix.com> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | No segmentation should be performed if gso type is | 3 | Interaction with vmnet.framework in different modes |
4 | VIRTIO_NET_HDR_GSO_NONE even if ECN bit is set. | 4 | differs only on configuration stage, so we can create |
5 | 5 | common `send`, `receive`, etc. procedures and reuse them. | |
6 | Fixes: e263cd49c7 ("Packet abstraction for VMWARE network devices") | 6 | |
7 | Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1544 | 7 | vmnet.framework supports iov, but writing more than |
8 | Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> | 8 | one iov into vmnet interface fails with |
9 | 'VMNET_INVALID_ARGUMENT'. Collecting provided iovs into | ||
10 | one and passing it to vmnet works fine. That's the | ||
11 | reason why receive_iov() left unimplemented. But it still | ||
12 | works with good enough performance having .receive() | ||
13 | implemented only. | ||
14 | |||
15 | Also, there is no way to unsubscribe from vmnet packages | ||
16 | receiving except registering and unregistering event | ||
17 | callback or simply drop packages just ignoring and | ||
18 | not processing them when related flag is set. Here we do | ||
19 | using the second way. | ||
20 | |||
21 | Signed-off-by: Phillip Tennen <phillip@axleos.com> | ||
22 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 23 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
10 | --- | 24 | --- |
11 | hw/net/net_tx_pkt.c | 6 ++++-- | 25 | net/vmnet-common.m | 314 +++++++++++++++++++++++++++++++++++++++++++++++++++++ |
12 | 1 file changed, 4 insertions(+), 2 deletions(-) | 26 | net/vmnet-shared.c | 83 +++++++++++++- |
13 | 27 | net/vmnet_int.h | 23 ++++ | |
14 | diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c | 28 | 3 files changed, 416 insertions(+), 4 deletions(-) |
29 | |||
30 | diff --git a/net/vmnet-common.m b/net/vmnet-common.m | ||
15 | index XXXXXXX..XXXXXXX 100644 | 31 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/hw/net/net_tx_pkt.c | 32 | --- a/net/vmnet-common.m |
17 | +++ b/hw/net/net_tx_pkt.c | 33 | +++ b/net/vmnet-common.m |
18 | @@ -XXX,XX +XXX,XX @@ bool net_tx_pkt_send_custom(struct NetTxPkt *pkt, bool offload, | 34 | @@ -XXX,XX +XXX,XX @@ |
35 | */ | ||
36 | |||
37 | #include "qemu/osdep.h" | ||
38 | +#include "qemu/main-loop.h" | ||
39 | +#include "qemu/log.h" | ||
40 | #include "qapi/qapi-types-net.h" | ||
41 | #include "vmnet_int.h" | ||
42 | #include "clients.h" | ||
43 | @@ -XXX,XX +XXX,XX @@ | ||
44 | #include "qapi/error.h" | ||
45 | |||
46 | #include <vmnet/vmnet.h> | ||
47 | +#include <dispatch/dispatch.h> | ||
48 | + | ||
49 | +#ifdef DEBUG | ||
50 | +#define D(x) x | ||
51 | +#define D_LOG(...) qemu_log(__VA_ARGS__) | ||
52 | +#else | ||
53 | +#define D(x) do { } while (0) | ||
54 | +#define D_LOG(...) do { } while (0) | ||
55 | +#endif | ||
56 | + | ||
57 | +typedef struct vmpktdesc vmpktdesc_t; | ||
58 | +typedef struct iovec iovec_t; | ||
59 | + | ||
60 | +static void vmnet_set_send_enabled(VmnetCommonState *s, bool enable) | ||
61 | +{ | ||
62 | + s->send_enabled = enable; | ||
63 | +} | ||
64 | + | ||
65 | + | ||
66 | +static void vmnet_send_completed(NetClientState *nc, ssize_t len) | ||
67 | +{ | ||
68 | + VmnetCommonState *s = DO_UPCAST(VmnetCommonState, nc, nc); | ||
69 | + vmnet_set_send_enabled(s, true); | ||
70 | +} | ||
71 | + | ||
72 | + | ||
73 | +static void vmnet_send(NetClientState *nc, | ||
74 | + interface_event_t event_id, | ||
75 | + xpc_object_t event) | ||
76 | +{ | ||
77 | + assert(event_id == VMNET_INTERFACE_PACKETS_AVAILABLE); | ||
78 | + | ||
79 | + VmnetCommonState *s; | ||
80 | + uint64_t packets_available; | ||
81 | + | ||
82 | + struct iovec *iov; | ||
83 | + struct vmpktdesc *packets; | ||
84 | + int pkt_cnt; | ||
85 | + int i; | ||
86 | + | ||
87 | + vmnet_return_t if_status; | ||
88 | + ssize_t size; | ||
89 | + | ||
90 | + s = DO_UPCAST(VmnetCommonState, nc, nc); | ||
91 | + | ||
92 | + packets_available = xpc_dictionary_get_uint64( | ||
93 | + event, | ||
94 | + vmnet_estimated_packets_available_key | ||
95 | + ); | ||
96 | + | ||
97 | + pkt_cnt = (packets_available < VMNET_PACKETS_LIMIT) ? | ||
98 | + packets_available : | ||
99 | + VMNET_PACKETS_LIMIT; | ||
100 | + | ||
101 | + | ||
102 | + iov = s->iov_buf; | ||
103 | + packets = s->packets_buf; | ||
104 | + | ||
105 | + for (i = 0; i < pkt_cnt; ++i) { | ||
106 | + packets[i].vm_pkt_size = s->max_packet_size; | ||
107 | + packets[i].vm_pkt_iovcnt = 1; | ||
108 | + packets[i].vm_flags = 0; | ||
109 | + } | ||
110 | + | ||
111 | + if_status = vmnet_read(s->vmnet_if, packets, &pkt_cnt); | ||
112 | + if (if_status != VMNET_SUCCESS) { | ||
113 | + error_printf("vmnet: read failed: %s\n", | ||
114 | + vmnet_status_map_str(if_status)); | ||
115 | + } | ||
116 | + qemu_mutex_lock_iothread(); | ||
117 | + for (i = 0; i < pkt_cnt; ++i) { | ||
118 | + size = qemu_send_packet_async(nc, | ||
119 | + iov[i].iov_base, | ||
120 | + packets[i].vm_pkt_size, | ||
121 | + vmnet_send_completed); | ||
122 | + if (size == 0) { | ||
123 | + vmnet_set_send_enabled(s, false); | ||
124 | + } else if (size < 0) { | ||
125 | + break; | ||
126 | + } | ||
127 | + } | ||
128 | + qemu_mutex_unlock_iothread(); | ||
129 | + | ||
130 | +} | ||
131 | + | ||
132 | + | ||
133 | +static void vmnet_register_event_callback(VmnetCommonState *s) | ||
134 | +{ | ||
135 | + dispatch_queue_t avail_pkt_q = dispatch_queue_create( | ||
136 | + "org.qemu.vmnet.if_queue", | ||
137 | + DISPATCH_QUEUE_SERIAL | ||
138 | + ); | ||
139 | + | ||
140 | + vmnet_interface_set_event_callback( | ||
141 | + s->vmnet_if, | ||
142 | + VMNET_INTERFACE_PACKETS_AVAILABLE, | ||
143 | + avail_pkt_q, | ||
144 | + ^(interface_event_t event_id, xpc_object_t event) { | ||
145 | + if (s->send_enabled) { | ||
146 | + vmnet_send(&s->nc, event_id, event); | ||
147 | + } | ||
148 | + }); | ||
149 | +} | ||
150 | + | ||
151 | + | ||
152 | +static void vmnet_bufs_init(VmnetCommonState *s) | ||
153 | +{ | ||
154 | + int i; | ||
155 | + struct vmpktdesc *packets; | ||
156 | + struct iovec *iov; | ||
157 | + | ||
158 | + packets = s->packets_buf; | ||
159 | + iov = s->iov_buf; | ||
160 | + | ||
161 | + for (i = 0; i < VMNET_PACKETS_LIMIT; ++i) { | ||
162 | + iov[i].iov_len = s->max_packet_size; | ||
163 | + iov[i].iov_base = g_malloc0(iov[i].iov_len); | ||
164 | + packets[i].vm_pkt_iov = iov + i; | ||
165 | + } | ||
166 | +} | ||
167 | + | ||
168 | + | ||
169 | +const char *vmnet_status_map_str(vmnet_return_t status) | ||
170 | +{ | ||
171 | + switch (status) { | ||
172 | + case VMNET_SUCCESS: | ||
173 | + return "success"; | ||
174 | + case VMNET_FAILURE: | ||
175 | + return "general failure"; | ||
176 | + case VMNET_MEM_FAILURE: | ||
177 | + return "memory allocation failure"; | ||
178 | + case VMNET_INVALID_ARGUMENT: | ||
179 | + return "invalid argument specified"; | ||
180 | + case VMNET_SETUP_INCOMPLETE: | ||
181 | + return "interface setup is not complete"; | ||
182 | + case VMNET_INVALID_ACCESS: | ||
183 | + return "invalid access, permission denied"; | ||
184 | + case VMNET_PACKET_TOO_BIG: | ||
185 | + return "packet size is larger than MTU"; | ||
186 | + case VMNET_BUFFER_EXHAUSTED: | ||
187 | + return "buffers exhausted in kernel"; | ||
188 | + case VMNET_TOO_MANY_PACKETS: | ||
189 | + return "packet count exceeds limit"; | ||
190 | +#if defined(MAC_OS_VERSION_11_0) && \ | ||
191 | + MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0 | ||
192 | + case VMNET_SHARING_SERVICE_BUSY: | ||
193 | + return "conflict, sharing service is in use"; | ||
194 | +#endif | ||
195 | + default: | ||
196 | + return "unknown vmnet error"; | ||
197 | + } | ||
198 | +} | ||
199 | + | ||
200 | + | ||
201 | +int vmnet_if_create(NetClientState *nc, | ||
202 | + xpc_object_t if_desc, | ||
203 | + Error **errp, | ||
204 | + void (*completion_callback)(xpc_object_t interface_param)) | ||
205 | +{ | ||
206 | + VmnetCommonState *s; | ||
207 | + | ||
208 | + dispatch_queue_t if_create_q; | ||
209 | + dispatch_semaphore_t if_created_sem; | ||
210 | + | ||
211 | + __block vmnet_return_t if_status; | ||
212 | + | ||
213 | + if_create_q = dispatch_queue_create("org.qemu.vmnet.create", | ||
214 | + DISPATCH_QUEUE_SERIAL); | ||
215 | + if_created_sem = dispatch_semaphore_create(0); | ||
216 | + | ||
217 | + xpc_dictionary_set_bool( | ||
218 | + if_desc, | ||
219 | + vmnet_allocate_mac_address_key, | ||
220 | + false | ||
221 | + ); | ||
222 | + | ||
223 | + D(D_LOG("vmnet.start.interface_desc:\n"); | ||
224 | + xpc_dictionary_apply(if_desc, | ||
225 | + ^bool(const char *k, xpc_object_t v) { | ||
226 | + char *desc = xpc_copy_description(v); | ||
227 | + D_LOG(" %s=%s\n", k, desc); | ||
228 | + free(desc); | ||
229 | + return true; | ||
230 | + })); | ||
231 | + | ||
232 | + s = DO_UPCAST(VmnetCommonState, nc, nc); | ||
233 | + s->vmnet_if = vmnet_start_interface( | ||
234 | + if_desc, | ||
235 | + if_create_q, | ||
236 | + ^(vmnet_return_t status, xpc_object_t interface_param) { | ||
237 | + if_status = status; | ||
238 | + if (status != VMNET_SUCCESS || !interface_param) { | ||
239 | + dispatch_semaphore_signal(if_created_sem); | ||
240 | + return; | ||
241 | + } | ||
242 | + | ||
243 | + D(D_LOG("vmnet.start.interface_param:\n"); | ||
244 | + xpc_dictionary_apply(interface_param, | ||
245 | + ^bool(const char *k, xpc_object_t v) { | ||
246 | + char *desc = xpc_copy_description(v); | ||
247 | + D_LOG(" %s=%s\n", k, desc); | ||
248 | + free(desc); | ||
249 | + return true; | ||
250 | + })); | ||
251 | + | ||
252 | + s->mtu = xpc_dictionary_get_uint64( | ||
253 | + interface_param, | ||
254 | + vmnet_mtu_key); | ||
255 | + s->max_packet_size = xpc_dictionary_get_uint64( | ||
256 | + interface_param, | ||
257 | + vmnet_max_packet_size_key); | ||
258 | + | ||
259 | + if (completion_callback) { | ||
260 | + completion_callback(interface_param); | ||
261 | + } | ||
262 | + dispatch_semaphore_signal(if_created_sem); | ||
263 | + }); | ||
264 | + | ||
265 | + if (s->vmnet_if == NULL) { | ||
266 | + error_setg(errp, "unable to create interface with requested params"); | ||
267 | + return -1; | ||
268 | + } | ||
269 | + | ||
270 | + dispatch_semaphore_wait(if_created_sem, DISPATCH_TIME_FOREVER); | ||
271 | + dispatch_release(if_create_q); | ||
272 | + | ||
273 | + if (if_status != VMNET_SUCCESS) { | ||
274 | + error_setg(errp, | ||
275 | + "cannot create vmnet interface: %s", | ||
276 | + vmnet_status_map_str(if_status)); | ||
277 | + return -1; | ||
278 | + } | ||
279 | + | ||
280 | + vmnet_register_event_callback(s); | ||
281 | + vmnet_bufs_init(s); | ||
282 | + vmnet_set_send_enabled(s, true); | ||
283 | + | ||
284 | + return 0; | ||
285 | +} | ||
286 | + | ||
287 | + | ||
288 | +ssize_t vmnet_receive_common(NetClientState *nc, | ||
289 | + const uint8_t *buf, | ||
290 | + size_t size) | ||
291 | +{ | ||
292 | + VmnetCommonState *s; | ||
293 | + vmpktdesc_t packet; | ||
294 | + iovec_t iov; | ||
295 | + int pkt_cnt; | ||
296 | + vmnet_return_t if_status; | ||
297 | + | ||
298 | + s = DO_UPCAST(VmnetCommonState, nc, nc); | ||
299 | + | ||
300 | + if (size > s->max_packet_size) { | ||
301 | + warn_report("vmnet: packet is too big, %zu > %llu\n", | ||
302 | + packet.vm_pkt_size, | ||
303 | + s->max_packet_size); | ||
304 | + return -1; | ||
305 | + } | ||
306 | + | ||
307 | + iov.iov_base = (char *) buf; | ||
308 | + iov.iov_len = size; | ||
309 | + | ||
310 | + packet.vm_pkt_iovcnt = 1; | ||
311 | + packet.vm_flags = 0; | ||
312 | + packet.vm_pkt_size = size; | ||
313 | + packet.vm_pkt_iov = &iov; | ||
314 | + | ||
315 | + pkt_cnt = 1; | ||
316 | + if_status = vmnet_write(s->vmnet_if, &packet, &pkt_cnt); | ||
317 | + | ||
318 | + if (if_status != VMNET_SUCCESS) { | ||
319 | + error_report("vmnet: write error: %s\n", | ||
320 | + vmnet_status_map_str(if_status)); | ||
321 | + } | ||
322 | + | ||
323 | + if (if_status == VMNET_SUCCESS && pkt_cnt) { | ||
324 | + return size; | ||
325 | + } | ||
326 | + return 0; | ||
327 | +} | ||
328 | + | ||
329 | + | ||
330 | +void vmnet_cleanup_common(NetClientState *nc) | ||
331 | +{ | ||
332 | + VmnetCommonState *s; | ||
333 | + dispatch_queue_t if_destroy_q; | ||
334 | + | ||
335 | + s = DO_UPCAST(VmnetCommonState, nc, nc); | ||
336 | + | ||
337 | + qemu_purge_queued_packets(nc); | ||
338 | + vmnet_set_send_enabled(s, false); | ||
339 | + | ||
340 | + if (s->vmnet_if == NULL) { | ||
341 | + return; | ||
342 | + } | ||
343 | + | ||
344 | + if_destroy_q = dispatch_queue_create( | ||
345 | + "org.qemu.vmnet.destroy", | ||
346 | + DISPATCH_QUEUE_SERIAL | ||
347 | + ); | ||
348 | + | ||
349 | + vmnet_stop_interface( | ||
350 | + s->vmnet_if, | ||
351 | + if_destroy_q, | ||
352 | + ^(vmnet_return_t status) { | ||
353 | + }); | ||
354 | + | ||
355 | + for (int i = 0; i < VMNET_PACKETS_LIMIT; ++i) { | ||
356 | + g_free(s->iov_buf[i].iov_base); | ||
357 | + } | ||
358 | +} | ||
359 | diff --git a/net/vmnet-shared.c b/net/vmnet-shared.c | ||
360 | index XXXXXXX..XXXXXXX 100644 | ||
361 | --- a/net/vmnet-shared.c | ||
362 | +++ b/net/vmnet-shared.c | ||
363 | @@ -XXX,XX +XXX,XX @@ | ||
364 | |||
365 | #include "qemu/osdep.h" | ||
366 | #include "qapi/qapi-types-net.h" | ||
367 | +#include "qapi/error.h" | ||
368 | #include "vmnet_int.h" | ||
369 | #include "clients.h" | ||
370 | -#include "qemu/error-report.h" | ||
371 | -#include "qapi/error.h" | ||
372 | |||
373 | #include <vmnet/vmnet.h> | ||
374 | |||
375 | +typedef struct VmnetSharedState { | ||
376 | + VmnetCommonState cs; | ||
377 | +} VmnetSharedState; | ||
378 | + | ||
379 | + | ||
380 | +static xpc_object_t create_if_desc(const Netdev *netdev, Error **errp) | ||
381 | +{ | ||
382 | + const NetdevVmnetSharedOptions *options = &(netdev->u.vmnet_shared); | ||
383 | + xpc_object_t if_desc = xpc_dictionary_create(NULL, NULL, 0); | ||
384 | + | ||
385 | + xpc_dictionary_set_uint64( | ||
386 | + if_desc, | ||
387 | + vmnet_operation_mode_key, | ||
388 | + VMNET_SHARED_MODE | ||
389 | + ); | ||
390 | + | ||
391 | +#if defined(MAC_OS_VERSION_11_0) && \ | ||
392 | + MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0 | ||
393 | + xpc_dictionary_set_bool( | ||
394 | + if_desc, | ||
395 | + vmnet_enable_isolation_key, | ||
396 | + options->isolated | ||
397 | + ); | ||
398 | +#else | ||
399 | + if (options->has_isolated) { | ||
400 | + error_setg(errp, | ||
401 | + "vmnet-shared.isolated feature is " | ||
402 | + "unavailable: outdated vmnet.framework API"); | ||
403 | + } | ||
404 | +#endif | ||
405 | + | ||
406 | + if (options->has_nat66_prefix) { | ||
407 | + xpc_dictionary_set_string(if_desc, | ||
408 | + vmnet_nat66_prefix_key, | ||
409 | + options->nat66_prefix); | ||
410 | + } | ||
411 | + | ||
412 | + if (options->has_start_address || | ||
413 | + options->has_end_address || | ||
414 | + options->has_subnet_mask) { | ||
415 | + | ||
416 | + if (options->has_start_address && | ||
417 | + options->has_end_address && | ||
418 | + options->has_subnet_mask) { | ||
419 | + | ||
420 | + xpc_dictionary_set_string(if_desc, | ||
421 | + vmnet_start_address_key, | ||
422 | + options->start_address); | ||
423 | + xpc_dictionary_set_string(if_desc, | ||
424 | + vmnet_end_address_key, | ||
425 | + options->end_address); | ||
426 | + xpc_dictionary_set_string(if_desc, | ||
427 | + vmnet_subnet_mask_key, | ||
428 | + options->subnet_mask); | ||
429 | + } else { | ||
430 | + error_setg( | ||
431 | + errp, | ||
432 | + "'start-address', 'end-address', 'subnet-mask' " | ||
433 | + "should be provided together" | ||
434 | + ); | ||
435 | + } | ||
436 | + } | ||
437 | + | ||
438 | + return if_desc; | ||
439 | +} | ||
440 | + | ||
441 | +static NetClientInfo net_vmnet_shared_info = { | ||
442 | + .type = NET_CLIENT_DRIVER_VMNET_SHARED, | ||
443 | + .size = sizeof(VmnetSharedState), | ||
444 | + .receive = vmnet_receive_common, | ||
445 | + .cleanup = vmnet_cleanup_common, | ||
446 | +}; | ||
447 | + | ||
448 | int net_init_vmnet_shared(const Netdev *netdev, const char *name, | ||
449 | NetClientState *peer, Error **errp) | ||
19 | { | 450 | { |
20 | assert(pkt); | 451 | - error_setg(errp, "vmnet-shared is not implemented yet"); |
21 | 452 | - return -1; | |
22 | + uint8_t gso_type = pkt->virt_hdr.gso_type & ~VIRTIO_NET_HDR_GSO_ECN; | 453 | + NetClientState *nc = qemu_new_net_client(&net_vmnet_shared_info, |
23 | + | 454 | + peer, "vmnet-shared", name); |
24 | /* | 455 | + xpc_object_t if_desc = create_if_desc(netdev, errp); |
25 | * Since underlying infrastructure does not support IP datagrams longer | 456 | + |
26 | * than 64K we should drop such packets and don't even try to send | 457 | + return vmnet_if_create(nc, if_desc, errp, NULL); |
27 | */ | 458 | } |
28 | - if (VIRTIO_NET_HDR_GSO_NONE != pkt->virt_hdr.gso_type) { | 459 | diff --git a/net/vmnet_int.h b/net/vmnet_int.h |
29 | + if (VIRTIO_NET_HDR_GSO_NONE != gso_type) { | 460 | index XXXXXXX..XXXXXXX 100644 |
30 | if (pkt->payload_len > | 461 | --- a/net/vmnet_int.h |
31 | ETH_MAX_IP_DGRAM_LEN - | 462 | +++ b/net/vmnet_int.h |
32 | pkt->vec[NET_TX_PKT_L3HDR_FRAG].iov_len) { | 463 | @@ -XXX,XX +XXX,XX @@ |
33 | @@ -XXX,XX +XXX,XX @@ bool net_tx_pkt_send_custom(struct NetTxPkt *pkt, bool offload, | 464 | |
34 | } | 465 | #include <vmnet/vmnet.h> |
35 | } | 466 | |
36 | 467 | +#define VMNET_PACKETS_LIMIT 50 | |
37 | - if (offload || pkt->virt_hdr.gso_type == VIRTIO_NET_HDR_GSO_NONE) { | 468 | + |
38 | + if (offload || gso_type == VIRTIO_NET_HDR_GSO_NONE) { | 469 | typedef struct VmnetCommonState { |
39 | if (!offload && pkt->virt_hdr.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { | 470 | NetClientState nc; |
40 | net_tx_pkt_do_sw_csum(pkt, &pkt->vec[NET_TX_PKT_L2HDR_FRAG], | 471 | + interface_ref vmnet_if; |
41 | pkt->payload_frags + NET_TX_PKT_PL_START_FRAG - 1, | 472 | + |
473 | + bool send_enabled; | ||
474 | + | ||
475 | + uint64_t mtu; | ||
476 | + uint64_t max_packet_size; | ||
477 | + | ||
478 | + struct vmpktdesc packets_buf[VMNET_PACKETS_LIMIT]; | ||
479 | + struct iovec iov_buf[VMNET_PACKETS_LIMIT]; | ||
480 | |||
481 | } VmnetCommonState; | ||
482 | |||
483 | +const char *vmnet_status_map_str(vmnet_return_t status); | ||
484 | + | ||
485 | +int vmnet_if_create(NetClientState *nc, | ||
486 | + xpc_object_t if_desc, | ||
487 | + Error **errp, | ||
488 | + void (*completion_callback)(xpc_object_t interface_param)); | ||
489 | + | ||
490 | +ssize_t vmnet_receive_common(NetClientState *nc, | ||
491 | + const uint8_t *buf, | ||
492 | + size_t size); | ||
493 | + | ||
494 | +void vmnet_cleanup_common(NetClientState *nc); | ||
495 | |||
496 | #endif /* VMNET_INT_H */ | ||
42 | -- | 497 | -- |
43 | 2.7.4 | 498 | 2.7.4 |
499 | |||
500 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | Add support for stripping/inserting VLAN for VFs. | 3 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
4 | |||
5 | Had to move CSUM calculation back into the for loop, since packet data | ||
6 | is pulled inside the loop based on strip VLAN decision for every VF. | ||
7 | |||
8 | net_rx_pkt_fix_l4_csum should be extended to accept a buffer instead for | ||
9 | igb. Work for a future patch. | ||
10 | |||
11 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | ||
12 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 4 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
13 | --- | 5 | --- |
14 | hw/net/igb_core.c | 62 +++++++++++++++++++++++++++++++++++++++++++------------ | 6 | net/vmnet-host.c | 110 ++++++++++++++++++++++++++++++++++++++++++++++++++++--- |
15 | 1 file changed, 49 insertions(+), 13 deletions(-) | 7 | 1 file changed, 104 insertions(+), 6 deletions(-) |
16 | 8 | ||
17 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | 9 | diff --git a/net/vmnet-host.c b/net/vmnet-host.c |
18 | index XXXXXXX..XXXXXXX 100644 | 10 | index XXXXXXX..XXXXXXX 100644 |
19 | --- a/hw/net/igb_core.c | 11 | --- a/net/vmnet-host.c |
20 | +++ b/hw/net/igb_core.c | 12 | +++ b/net/vmnet-host.c |
21 | @@ -XXX,XX +XXX,XX @@ igb_rss_parse_packet(IGBCore *core, struct NetRxPkt *pkt, bool tx, | 13 | @@ -XXX,XX +XXX,XX @@ |
22 | info->queue = E1000_RSS_QUEUE(&core->mac[RETA], info->hash); | 14 | */ |
23 | } | 15 | |
24 | 16 | #include "qemu/osdep.h" | |
25 | +static void | 17 | +#include "qemu/uuid.h" |
26 | +igb_tx_insert_vlan(IGBCore *core, uint16_t qn, struct igb_tx *tx, | 18 | #include "qapi/qapi-types-net.h" |
27 | + uint16_t vlan, bool insert_vlan) | 19 | -#include "vmnet_int.h" |
20 | -#include "clients.h" | ||
21 | -#include "qemu/error-report.h" | ||
22 | #include "qapi/error.h" | ||
23 | +#include "clients.h" | ||
24 | +#include "vmnet_int.h" | ||
25 | |||
26 | #include <vmnet/vmnet.h> | ||
27 | |||
28 | +typedef struct VmnetHostState { | ||
29 | + VmnetCommonState cs; | ||
30 | + QemuUUID network_uuid; | ||
31 | +} VmnetHostState; | ||
32 | + | ||
33 | +static xpc_object_t create_if_desc(const Netdev *netdev, | ||
34 | + NetClientState *nc, | ||
35 | + Error **errp) | ||
28 | +{ | 36 | +{ |
29 | + if (core->mac[MRQC] & 1) { | 37 | + const NetdevVmnetHostOptions *options = &(netdev->u.vmnet_host); |
30 | + uint16_t pool = qn % IGB_NUM_VM_POOLS; | ||
31 | + | 38 | + |
32 | + if (core->mac[VMVIR0 + pool] & E1000_VMVIR_VLANA_DEFAULT) { | 39 | + xpc_object_t if_desc = xpc_dictionary_create(NULL, NULL, 0); |
33 | + /* always insert default VLAN */ | 40 | + |
34 | + insert_vlan = true; | 41 | + xpc_dictionary_set_uint64( |
35 | + vlan = core->mac[VMVIR0 + pool] & 0xffff; | 42 | + if_desc, |
36 | + } else if (core->mac[VMVIR0 + pool] & E1000_VMVIR_VLANA_NEVER) { | 43 | + vmnet_operation_mode_key, |
37 | + insert_vlan = false; | 44 | + VMNET_HOST_MODE |
45 | + ); | ||
46 | + | ||
47 | +#if defined(MAC_OS_VERSION_11_0) && \ | ||
48 | + MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0 | ||
49 | + | ||
50 | + VmnetCommonState *cs = DO_UPCAST(VmnetCommonState, nc, nc); | ||
51 | + VmnetHostState *hs = DO_UPCAST(VmnetHostState, cs, cs); | ||
52 | + | ||
53 | + xpc_dictionary_set_bool( | ||
54 | + if_desc, | ||
55 | + vmnet_enable_isolation_key, | ||
56 | + options->isolated | ||
57 | + ); | ||
58 | + | ||
59 | + if (options->has_net_uuid) { | ||
60 | + if (qemu_uuid_parse(options->net_uuid, &hs->network_uuid) < 0) { | ||
61 | + error_setg(errp, "Invalid UUID provided in 'net-uuid'"); | ||
62 | + } | ||
63 | + | ||
64 | + xpc_dictionary_set_uuid( | ||
65 | + if_desc, | ||
66 | + vmnet_network_identifier_key, | ||
67 | + hs->network_uuid.data | ||
68 | + ); | ||
69 | + } | ||
70 | +#else | ||
71 | + if (options->has_isolated) { | ||
72 | + error_setg(errp, | ||
73 | + "vmnet-host.isolated feature is " | ||
74 | + "unavailable: outdated vmnet.framework API"); | ||
75 | + } | ||
76 | + | ||
77 | + if (options->has_net_uuid) { | ||
78 | + error_setg(errp, | ||
79 | + "vmnet-host.net-uuid feature is " | ||
80 | + "unavailable: outdated vmnet.framework API"); | ||
81 | + } | ||
82 | +#endif | ||
83 | + | ||
84 | + if (options->has_start_address || | ||
85 | + options->has_end_address || | ||
86 | + options->has_subnet_mask) { | ||
87 | + | ||
88 | + if (options->has_start_address && | ||
89 | + options->has_end_address && | ||
90 | + options->has_subnet_mask) { | ||
91 | + | ||
92 | + xpc_dictionary_set_string(if_desc, | ||
93 | + vmnet_start_address_key, | ||
94 | + options->start_address); | ||
95 | + xpc_dictionary_set_string(if_desc, | ||
96 | + vmnet_end_address_key, | ||
97 | + options->end_address); | ||
98 | + xpc_dictionary_set_string(if_desc, | ||
99 | + vmnet_subnet_mask_key, | ||
100 | + options->subnet_mask); | ||
101 | + } else { | ||
102 | + error_setg( | ||
103 | + errp, | ||
104 | + "'start-address', 'end-address', 'subnet-mask' " | ||
105 | + "should be provided together" | ||
106 | + ); | ||
38 | + } | 107 | + } |
39 | + } | 108 | + } |
40 | + | 109 | + |
41 | + if (insert_vlan && e1000x_vlan_enabled(core->mac)) { | 110 | + return if_desc; |
42 | + net_tx_pkt_setup_vlan_header_ex(tx->tx_pkt, vlan, | ||
43 | + core->mac[VET] & 0xffff); | ||
44 | + } | ||
45 | +} | 111 | +} |
46 | + | 112 | + |
47 | static bool | 113 | +static NetClientInfo net_vmnet_host_info = { |
48 | igb_setup_tx_offloads(IGBCore *core, struct igb_tx *tx) | 114 | + .type = NET_CLIENT_DRIVER_VMNET_HOST, |
49 | { | 115 | + .size = sizeof(VmnetHostState), |
50 | @@ -XXX,XX +XXX,XX @@ igb_process_tx_desc(IGBCore *core, | 116 | + .receive = vmnet_receive_common, |
51 | 117 | + .cleanup = vmnet_cleanup_common, | |
52 | if (cmd_type_len & E1000_TXD_CMD_EOP) { | 118 | +}; |
53 | if (!tx->skip_cp && net_tx_pkt_parse(tx->tx_pkt)) { | ||
54 | - if (cmd_type_len & E1000_TXD_CMD_VLE) { | ||
55 | - idx = (tx->first_olinfo_status >> 4) & 1; | ||
56 | - uint16_t vlan = tx->ctx[idx].vlan_macip_lens >> 16; | ||
57 | - uint16_t vet = core->mac[VET] & 0xffff; | ||
58 | - net_tx_pkt_setup_vlan_header_ex(tx->tx_pkt, vlan, vet); | ||
59 | - } | ||
60 | + idx = (tx->first_olinfo_status >> 4) & 1; | ||
61 | + igb_tx_insert_vlan(core, queue_index, tx, | ||
62 | + tx->ctx[idx].vlan_macip_lens >> 16, | ||
63 | + !!(cmd_type_len & E1000_TXD_CMD_VLE)); | ||
64 | + | 119 | + |
65 | if (igb_tx_pkt_send(core, tx, queue_index)) { | 120 | int net_init_vmnet_host(const Netdev *netdev, const char *name, |
66 | igb_on_tx_done_update_stats(core, tx->tx_pkt, queue_index); | 121 | - NetClientState *peer, Error **errp) { |
67 | } | 122 | - error_setg(errp, "vmnet-host is not implemented yet"); |
68 | @@ -XXX,XX +XXX,XX @@ igb_write_packet_to_guest(IGBCore *core, struct NetRxPkt *pkt, | 123 | - return -1; |
69 | igb_update_rx_stats(core, rxi, size, total_size); | 124 | + NetClientState *peer, Error **errp) |
125 | +{ | ||
126 | + NetClientState *nc; | ||
127 | + xpc_object_t if_desc; | ||
128 | + | ||
129 | + nc = qemu_new_net_client(&net_vmnet_host_info, | ||
130 | + peer, "vmnet-host", name); | ||
131 | + if_desc = create_if_desc(netdev, nc, errp); | ||
132 | + return vmnet_if_create(nc, if_desc, errp, NULL); | ||
70 | } | 133 | } |
71 | |||
72 | +static bool | ||
73 | +igb_rx_strip_vlan(IGBCore *core, const E1000E_RingInfo *rxi) | ||
74 | +{ | ||
75 | + if (core->mac[MRQC] & 1) { | ||
76 | + uint16_t pool = rxi->idx % IGB_NUM_VM_POOLS; | ||
77 | + /* Sec 7.10.3.8: CTRL.VME is ignored, only VMOLR/RPLOLR is used */ | ||
78 | + return (net_rx_pkt_get_packet_type(core->rx_pkt) == ETH_PKT_MCAST) ? | ||
79 | + core->mac[RPLOLR] & E1000_RPLOLR_STRVLAN : | ||
80 | + core->mac[VMOLR0 + pool] & E1000_VMOLR_STRVLAN; | ||
81 | + } | ||
82 | + | ||
83 | + return e1000x_vlan_enabled(core->mac); | ||
84 | +} | ||
85 | + | ||
86 | static inline void | ||
87 | igb_rx_fix_l4_csum(IGBCore *core, struct NetRxPkt *pkt) | ||
88 | { | ||
89 | @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, | ||
90 | |||
91 | ehdr = PKT_GET_ETH_HDR(filter_buf); | ||
92 | net_rx_pkt_set_packet_type(core->rx_pkt, get_eth_packet_type(ehdr)); | ||
93 | - | ||
94 | - net_rx_pkt_attach_iovec_ex(core->rx_pkt, iov, iovcnt, iov_ofs, | ||
95 | - e1000x_vlan_enabled(core->mac), | ||
96 | - core->mac[VET] & 0xffff); | ||
97 | + net_rx_pkt_set_protocols(core->rx_pkt, filter_buf, size); | ||
98 | |||
99 | queues = igb_receive_assign(core, ehdr, size, &rss_info, external_tx); | ||
100 | if (!queues) { | ||
101 | @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, | ||
102 | return orig_size; | ||
103 | } | ||
104 | |||
105 | - total_size = net_rx_pkt_get_total_len(core->rx_pkt) + | ||
106 | - e1000x_fcs_len(core->mac); | ||
107 | - | ||
108 | for (i = 0; i < IGB_NUM_QUEUES; i++) { | ||
109 | if (!(queues & BIT(i)) || | ||
110 | !(core->mac[RXDCTL0 + (i * 16)] & E1000_RXDCTL_QUEUE_ENABLE)) { | ||
111 | @@ -XXX,XX +XXX,XX @@ igb_receive_internal(IGBCore *core, const struct iovec *iov, int iovcnt, | ||
112 | |||
113 | igb_rx_ring_init(core, &rxr, i); | ||
114 | |||
115 | + net_rx_pkt_attach_iovec_ex(core->rx_pkt, iov, iovcnt, iov_ofs, | ||
116 | + igb_rx_strip_vlan(core, rxr.i), | ||
117 | + core->mac[VET] & 0xffff); | ||
118 | + | ||
119 | + total_size = net_rx_pkt_get_total_len(core->rx_pkt) + | ||
120 | + e1000x_fcs_len(core->mac); | ||
121 | + | ||
122 | if (!igb_has_rxbufs(core, rxr.i, total_size)) { | ||
123 | n |= E1000_ICS_RXO; | ||
124 | trace_e1000e_rx_not_written_to_guest(rxr.i->idx); | ||
125 | -- | 134 | -- |
126 | 2.7.4 | 135 | 2.7.4 |
136 | |||
137 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | Use PFRSTD to reset RSTI bit for VFs, and raise VFLRE interrupt when VF | 3 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
4 | is reset. | ||
5 | |||
6 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | ||
7 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 4 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
8 | --- | 5 | --- |
9 | hw/net/igb_core.c | 38 ++++++++++++++++++++++++++------------ | 6 | net/vmnet-bridged.m | 105 +++++++++++++++++++++++++++++++++++++++++++++++++--- |
10 | hw/net/igb_regs.h | 3 +++ | 7 | 1 file changed, 100 insertions(+), 5 deletions(-) |
11 | hw/net/trace-events | 2 ++ | ||
12 | 3 files changed, 31 insertions(+), 12 deletions(-) | ||
13 | 8 | ||
14 | diff --git a/hw/net/igb_core.c b/hw/net/igb_core.c | 9 | diff --git a/net/vmnet-bridged.m b/net/vmnet-bridged.m |
15 | index XXXXXXX..XXXXXXX 100644 | 10 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/hw/net/igb_core.c | 11 | --- a/net/vmnet-bridged.m |
17 | +++ b/hw/net/igb_core.c | 12 | +++ b/net/vmnet-bridged.m |
18 | @@ -XXX,XX +XXX,XX @@ static void igb_set_eims(IGBCore *core, int index, uint32_t val) | 13 | @@ -XXX,XX +XXX,XX @@ |
19 | igb_update_interrupt_state(core); | 14 | |
20 | } | 15 | #include "qemu/osdep.h" |
21 | 16 | #include "qapi/qapi-types-net.h" | |
22 | -static void igb_vf_reset(IGBCore *core, uint16_t vfn) | 17 | -#include "vmnet_int.h" |
23 | -{ | 18 | -#include "clients.h" |
24 | - /* TODO: Reset of the queue enable and the interrupt registers of the VF. */ | 19 | -#include "qemu/error-report.h" |
25 | - | 20 | #include "qapi/error.h" |
26 | - core->mac[V2PMAILBOX0 + vfn] &= ~E1000_V2PMAILBOX_RSTI; | 21 | +#include "clients.h" |
27 | - core->mac[V2PMAILBOX0 + vfn] = E1000_V2PMAILBOX_RSTD; | 22 | +#include "vmnet_int.h" |
28 | -} | 23 | |
29 | - | 24 | #include <vmnet/vmnet.h> |
30 | static void mailbox_interrupt_to_vf(IGBCore *core, uint16_t vfn) | 25 | |
31 | { | 26 | +typedef struct VmnetBridgedState { |
32 | uint32_t ent = core->mac[VTIVAR_MISC + vfn]; | 27 | + VmnetCommonState cs; |
33 | @@ -XXX,XX +XXX,XX @@ static void igb_set_vfmailbox(IGBCore *core, int index, uint32_t val) | 28 | +} VmnetBridgedState; |
34 | } | 29 | + |
35 | } | 30 | +static bool validate_ifname(const char *ifname) |
36 | |||
37 | +static void igb_vf_reset(IGBCore *core, uint16_t vfn) | ||
38 | +{ | 31 | +{ |
39 | + /* disable Rx and Tx for the VF*/ | 32 | + xpc_object_t shared_if_list = vmnet_copy_shared_interface_list(); |
40 | + core->mac[VFTE] &= ~BIT(vfn); | 33 | + __block bool match = false; |
41 | + core->mac[VFRE] &= ~BIT(vfn); | 34 | + |
42 | + /* indicate VF reset to PF */ | 35 | + xpc_array_apply( |
43 | + core->mac[VFLRE] |= BIT(vfn); | 36 | + shared_if_list, |
44 | + /* VFLRE and mailbox use the same interrupt cause */ | 37 | + ^bool(size_t index, xpc_object_t value) { |
45 | + mailbox_interrupt_to_pf(core); | 38 | + if (strcmp(xpc_string_get_string_ptr(value), ifname) == 0) { |
39 | + match = true; | ||
40 | + return false; | ||
41 | + } | ||
42 | + return true; | ||
43 | + }); | ||
44 | + | ||
45 | + return match; | ||
46 | +} | 46 | +} |
47 | + | 47 | + |
48 | static void igb_w1c(IGBCore *core, int index, uint32_t val) | 48 | +static const char *get_valid_ifnames(void) |
49 | +{ | ||
50 | + xpc_object_t shared_if_list = vmnet_copy_shared_interface_list(); | ||
51 | + __block char *if_list = NULL; | ||
52 | + | ||
53 | + xpc_array_apply( | ||
54 | + shared_if_list, | ||
55 | + ^bool(size_t index, xpc_object_t value) { | ||
56 | + if_list = g_strconcat(xpc_string_get_string_ptr(value), | ||
57 | + " ", | ||
58 | + if_list, | ||
59 | + NULL); | ||
60 | + return true; | ||
61 | + }); | ||
62 | + | ||
63 | + if (if_list) { | ||
64 | + return if_list; | ||
65 | + } | ||
66 | + return "[no interfaces]"; | ||
67 | +} | ||
68 | + | ||
69 | +static xpc_object_t create_if_desc(const Netdev *netdev, Error **errp) | ||
70 | +{ | ||
71 | + const NetdevVmnetBridgedOptions *options = &(netdev->u.vmnet_bridged); | ||
72 | + xpc_object_t if_desc = xpc_dictionary_create(NULL, NULL, 0); | ||
73 | + | ||
74 | + xpc_dictionary_set_uint64( | ||
75 | + if_desc, | ||
76 | + vmnet_operation_mode_key, | ||
77 | + VMNET_BRIDGED_MODE | ||
78 | + ); | ||
79 | + | ||
80 | +#if defined(MAC_OS_VERSION_11_0) && \ | ||
81 | + MAC_OS_X_VERSION_MIN_REQUIRED >= MAC_OS_VERSION_11_0 | ||
82 | + xpc_dictionary_set_bool( | ||
83 | + if_desc, | ||
84 | + vmnet_enable_isolation_key, | ||
85 | + options->isolated | ||
86 | + ); | ||
87 | +#else | ||
88 | + if (options->has_isolated) { | ||
89 | + error_setg(errp, | ||
90 | + "vmnet-bridged.isolated feature is " | ||
91 | + "unavailable: outdated vmnet.framework API"); | ||
92 | + } | ||
93 | +#endif | ||
94 | + | ||
95 | + if (validate_ifname(options->ifname)) { | ||
96 | + xpc_dictionary_set_string(if_desc, | ||
97 | + vmnet_shared_interface_name_key, | ||
98 | + options->ifname); | ||
99 | + } else { | ||
100 | + return NULL; | ||
101 | + } | ||
102 | + return if_desc; | ||
103 | +} | ||
104 | + | ||
105 | +static NetClientInfo net_vmnet_bridged_info = { | ||
106 | + .type = NET_CLIENT_DRIVER_VMNET_BRIDGED, | ||
107 | + .size = sizeof(VmnetBridgedState), | ||
108 | + .receive = vmnet_receive_common, | ||
109 | + .cleanup = vmnet_cleanup_common, | ||
110 | +}; | ||
111 | + | ||
112 | int net_init_vmnet_bridged(const Netdev *netdev, const char *name, | ||
113 | NetClientState *peer, Error **errp) | ||
49 | { | 114 | { |
50 | core->mac[index] &= ~val; | 115 | - error_setg(errp, "vmnet-bridged is not implemented yet"); |
51 | @@ -XXX,XX +XXX,XX @@ igb_set_status(IGBCore *core, int index, uint32_t val) | 116 | - return -1; |
52 | static void | 117 | + NetClientState *nc = qemu_new_net_client(&net_vmnet_bridged_info, |
53 | igb_set_ctrlext(IGBCore *core, int index, uint32_t val) | 118 | + peer, "vmnet-bridged", name); |
54 | { | 119 | + xpc_object_t if_desc = create_if_desc(netdev, errp);; |
55 | - trace_e1000e_link_set_ext_params(!!(val & E1000_CTRL_EXT_ASDCHK), | ||
56 | - !!(val & E1000_CTRL_EXT_SPD_BYPS)); | ||
57 | - | ||
58 | - /* TODO: PFRSTD */ | ||
59 | + trace_igb_link_set_ext_params(!!(val & E1000_CTRL_EXT_ASDCHK), | ||
60 | + !!(val & E1000_CTRL_EXT_SPD_BYPS), | ||
61 | + !!(val & E1000_CTRL_EXT_PFRSTD)); | ||
62 | |||
63 | /* Zero self-clearing bits */ | ||
64 | val &= ~(E1000_CTRL_EXT_ASDCHK | E1000_CTRL_EXT_EE_RST); | ||
65 | core->mac[CTRL_EXT] = val; | ||
66 | + | 120 | + |
67 | + if (core->mac[CTRL_EXT] & E1000_CTRL_EXT_PFRSTD) { | 121 | + if (!if_desc) { |
68 | + for (int vfn = 0; vfn < IGB_MAX_VF_FUNCTIONS; vfn++) { | 122 | + error_setg(errp, |
69 | + core->mac[V2PMAILBOX0 + vfn] &= ~E1000_V2PMAILBOX_RSTI; | 123 | + "unsupported ifname, should be one of: %s", |
70 | + core->mac[V2PMAILBOX0 + vfn] |= E1000_V2PMAILBOX_RSTD; | 124 | + get_valid_ifnames()); |
71 | + } | 125 | + return -1; |
72 | + } | ||
73 | } | ||
74 | |||
75 | static void | ||
76 | @@ -XXX,XX +XXX,XX @@ static void igb_reset(IGBCore *core, bool sw) | ||
77 | |||
78 | e1000x_reset_mac_addr(core->owner_nic, core->mac, core->permanent_mac); | ||
79 | |||
80 | + for (int vfn = 0; vfn < IGB_MAX_VF_FUNCTIONS; vfn++) { | ||
81 | + /* Set RSTI, so VF can identify a PF reset is in progress */ | ||
82 | + core->mac[V2PMAILBOX0 + vfn] |= E1000_V2PMAILBOX_RSTI; | ||
83 | + } | 126 | + } |
84 | + | 127 | + |
85 | for (i = 0; i < ARRAY_SIZE(core->tx); i++) { | 128 | + return vmnet_if_create(nc, if_desc, errp, NULL); |
86 | tx = &core->tx[i]; | 129 | } |
87 | net_tx_pkt_reset(tx->tx_pkt, NULL); | ||
88 | diff --git a/hw/net/igb_regs.h b/hw/net/igb_regs.h | ||
89 | index XXXXXXX..XXXXXXX 100644 | ||
90 | --- a/hw/net/igb_regs.h | ||
91 | +++ b/hw/net/igb_regs.h | ||
92 | @@ -XXX,XX +XXX,XX @@ union e1000_adv_rx_desc { | ||
93 | |||
94 | /* from igb/e1000_defines.h */ | ||
95 | |||
96 | +/* Physical Func Reset Done Indication */ | ||
97 | +#define E1000_CTRL_EXT_PFRSTD 0x00004000 | ||
98 | + | ||
99 | #define E1000_IVAR_VALID 0x80 | ||
100 | #define E1000_GPIE_NSICR 0x00000001 | ||
101 | #define E1000_GPIE_MSIX_MODE 0x00000010 | ||
102 | diff --git a/hw/net/trace-events b/hw/net/trace-events | ||
103 | index XXXXXXX..XXXXXXX 100644 | ||
104 | --- a/hw/net/trace-events | ||
105 | +++ b/hw/net/trace-events | ||
106 | @@ -XXX,XX +XXX,XX @@ igb_core_mdic_read_unhandled(uint32_t addr) "MDIC READ: PHY[%u] UNHANDLED" | ||
107 | igb_core_mdic_write(uint32_t addr, uint32_t data) "MDIC WRITE: PHY[%u] = 0x%x" | ||
108 | igb_core_mdic_write_unhandled(uint32_t addr) "MDIC WRITE: PHY[%u] UNHANDLED" | ||
109 | |||
110 | +igb_link_set_ext_params(bool asd_check, bool speed_select_bypass, bool pfrstd) "Set extended link params: ASD check: %d, Speed select bypass: %d, PF reset done: %d" | ||
111 | + | ||
112 | igb_rx_desc_buff_size(uint32_t b) "buffer size: %u" | ||
113 | igb_rx_desc_buff_write(uint64_t addr, uint16_t offset, const void* source, uint32_t len) "addr: 0x%"PRIx64", offset: %u, from: %p, length: %u" | ||
114 | |||
115 | -- | 130 | -- |
116 | 2.7.4 | 131 | 2.7.4 |
132 | |||
133 | diff view generated by jsdifflib |
1 | From: Akihiko Odaki <akihiko.odaki@daynix.com> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | Align the l3_hdr member of NetTxPkt by defining it as a union of | 3 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
4 | ip_header, ip6_header, and an array of octets. | ||
5 | |||
6 | Fixes: e263cd49c7 ("Packet abstraction for VMWARE network devices") | ||
7 | Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1544 | ||
8 | Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com> | ||
9 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 4 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
10 | --- | 5 | --- |
11 | hw/net/net_tx_pkt.c | 21 +++++++++++---------- | 6 | qemu-options.hx | 25 +++++++++++++++++++++++++ |
12 | 1 file changed, 11 insertions(+), 10 deletions(-) | 7 | 1 file changed, 25 insertions(+) |
13 | 8 | ||
14 | diff --git a/hw/net/net_tx_pkt.c b/hw/net/net_tx_pkt.c | 9 | diff --git a/qemu-options.hx b/qemu-options.hx |
15 | index XXXXXXX..XXXXXXX 100644 | 10 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/hw/net/net_tx_pkt.c | 11 | --- a/qemu-options.hx |
17 | +++ b/hw/net/net_tx_pkt.c | 12 | +++ b/qemu-options.hx |
18 | @@ -XXX,XX +XXX,XX @@ struct NetTxPkt { | 13 | @@ -XXX,XX +XXX,XX @@ DEF("netdev", HAS_ARG, QEMU_OPTION_netdev, |
19 | struct iovec *vec; | 14 | "-netdev vhost-vdpa,id=str,vhostdev=/path/to/dev\n" |
20 | 15 | " configure a vhost-vdpa network,Establish a vhost-vdpa netdev\n" | |
21 | uint8_t l2_hdr[ETH_MAX_L2_HDR_LEN]; | 16 | #endif |
22 | - uint8_t l3_hdr[ETH_MAX_IP_DGRAM_LEN]; | 17 | +#ifdef CONFIG_VMNET |
23 | + union { | 18 | + "-netdev vmnet-host,id=str[,isolated=on|off][,net-uuid=uuid]\n" |
24 | + struct ip_header ip; | 19 | + " [,start-address=addr,end-address=addr,subnet-mask=mask]\n" |
25 | + struct ip6_header ip6; | 20 | + " configure a vmnet network backend in host mode with ID 'str',\n" |
26 | + uint8_t octets[ETH_MAX_IP_DGRAM_LEN]; | 21 | + " isolate this interface from others with 'isolated',\n" |
27 | + } l3_hdr; | 22 | + " configure the address range and choose a subnet mask,\n" |
28 | 23 | + " specify network UUID 'uuid' to disable DHCP and interact with\n" | |
29 | uint32_t payload_len; | 24 | + " vmnet-host interfaces within this isolated network\n" |
30 | 25 | + "-netdev vmnet-shared,id=str[,isolated=on|off][,nat66-prefix=addr]\n" | |
31 | @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_update_ip_hdr_checksum(struct NetTxPkt *pkt) | 26 | + " [,start-address=addr,end-address=addr,subnet-mask=mask]\n" |
32 | { | 27 | + " configure a vmnet network backend in shared mode with ID 'str',\n" |
33 | uint16_t csum; | 28 | + " configure the address range and choose a subnet mask,\n" |
34 | assert(pkt); | 29 | + " set IPv6 ULA prefix (of length 64) to use for internal network,\n" |
35 | - struct ip_header *ip_hdr; | 30 | + " isolate this interface from others with 'isolated'\n" |
36 | - ip_hdr = pkt->vec[NET_TX_PKT_L3HDR_FRAG].iov_base; | 31 | + "-netdev vmnet-bridged,id=str,ifname=name[,isolated=on|off]\n" |
37 | 32 | + " configure a vmnet network backend in bridged mode with ID 'str',\n" | |
38 | - ip_hdr->ip_len = cpu_to_be16(pkt->payload_len + | 33 | + " use 'ifname=name' to select a physical network interface to be bridged,\n" |
39 | + pkt->l3_hdr.ip.ip_len = cpu_to_be16(pkt->payload_len + | 34 | + " isolate this interface from others with 'isolated'\n" |
40 | pkt->vec[NET_TX_PKT_L3HDR_FRAG].iov_len); | 35 | +#endif |
41 | 36 | "-netdev hubport,id=str,hubid=n[,netdev=nd]\n" | |
42 | - ip_hdr->ip_sum = 0; | 37 | " configure a hub port on the hub with ID 'n'\n", QEMU_ARCH_ALL) |
43 | - csum = net_raw_checksum((uint8_t *)ip_hdr, | 38 | DEF("nic", HAS_ARG, QEMU_OPTION_nic, |
44 | + pkt->l3_hdr.ip.ip_sum = 0; | 39 | @@ -XXX,XX +XXX,XX @@ DEF("nic", HAS_ARG, QEMU_OPTION_nic, |
45 | + csum = net_raw_checksum(pkt->l3_hdr.octets, | 40 | #ifdef CONFIG_POSIX |
46 | pkt->vec[NET_TX_PKT_L3HDR_FRAG].iov_len); | 41 | "vhost-user|" |
47 | - ip_hdr->ip_sum = cpu_to_be16(csum); | 42 | #endif |
48 | + pkt->l3_hdr.ip.ip_sum = cpu_to_be16(csum); | 43 | +#ifdef CONFIG_VMNET |
49 | } | 44 | + "vmnet-host|vmnet-shared|vmnet-bridged|" |
50 | 45 | +#endif | |
51 | void net_tx_pkt_update_ip_checksums(struct NetTxPkt *pkt) | 46 | "socket][,option][,...][mac=macaddr]\n" |
52 | @@ -XXX,XX +XXX,XX @@ void net_tx_pkt_fix_ip6_payload_len(struct NetTxPkt *pkt) | 47 | " initialize an on-board / default host NIC (using MAC address\n" |
53 | { | 48 | " macaddr) and connect it to the given host network backend\n" |
54 | struct iovec *l2 = &pkt->vec[NET_TX_PKT_L2HDR_FRAG]; | 49 | @@ -XXX,XX +XXX,XX @@ DEF("net", HAS_ARG, QEMU_OPTION_net, |
55 | if (eth_get_l3_proto(l2, 1, l2->iov_len) == ETH_P_IPV6) { | 50 | #ifdef CONFIG_NETMAP |
56 | - struct ip6_header *ip6 = (struct ip6_header *) pkt->l3_hdr; | 51 | "netmap|" |
57 | /* | 52 | #endif |
58 | * TODO: if qemu would support >64K packets - add jumbo option check | 53 | +#ifdef CONFIG_VMNET |
59 | * something like that: | 54 | + "vmnet-host|vmnet-shared|vmnet-bridged|" |
60 | * 'if (ip6->ip6_plen == 0 && !has_jumbo_option(ip6)) {' | 55 | +#endif |
61 | */ | 56 | "socket][,option][,option][,...]\n" |
62 | - if (ip6->ip6_plen == 0) { | 57 | " old way to initialize a host network interface\n" |
63 | + if (pkt->l3_hdr.ip6.ip6_plen == 0) { | 58 | " (use the -netdev option if possible instead)\n", QEMU_ARCH_ALL) |
64 | if (pkt->payload_len <= ETH_MAX_IP_DGRAM_LEN) { | ||
65 | - ip6->ip6_plen = htons(pkt->payload_len); | ||
66 | + pkt->l3_hdr.ip6.ip6_plen = htons(pkt->payload_len); | ||
67 | } | ||
68 | /* | ||
69 | * TODO: if qemu would support >64K packets | ||
70 | -- | 59 | -- |
71 | 2.7.4 | 60 | 2.7.4 |
61 | |||
62 | diff view generated by jsdifflib |
1 | From: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | 1 | From: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
---|---|---|---|
2 | 2 | ||
3 | I would like to review and be informed on changes to igb device | 3 | Signed-off-by: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> |
4 | |||
5 | Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | ||
6 | Signed-off-by: Jason Wang <jasowang@redhat.com> | 4 | Signed-off-by: Jason Wang <jasowang@redhat.com> |
7 | --- | 5 | --- |
8 | MAINTAINERS | 1 + | 6 | MAINTAINERS | 5 +++++ |
9 | 1 file changed, 1 insertion(+) | 7 | 1 file changed, 5 insertions(+) |
10 | 8 | ||
11 | diff --git a/MAINTAINERS b/MAINTAINERS | 9 | diff --git a/MAINTAINERS b/MAINTAINERS |
12 | index XXXXXXX..XXXXXXX 100644 | 10 | index XXXXXXX..XXXXXXX 100644 |
13 | --- a/MAINTAINERS | 11 | --- a/MAINTAINERS |
14 | +++ b/MAINTAINERS | 12 | +++ b/MAINTAINERS |
15 | @@ -XXX,XX +XXX,XX @@ F: tests/qtest/libqos/e1000e.* | 13 | @@ -XXX,XX +XXX,XX @@ W: http://info.iet.unipi.it/~luigi/netmap/ |
16 | |||
17 | igb | ||
18 | M: Akihiko Odaki <akihiko.odaki@daynix.com> | ||
19 | +R: Sriram Yagnaraman <sriram.yagnaraman@est.tech> | ||
20 | S: Maintained | 14 | S: Maintained |
21 | F: docs/system/devices/igb.rst | 15 | F: net/netmap.c |
22 | F: hw/net/igb* | 16 | |
17 | +Apple vmnet network backends | ||
18 | +M: Vladislav Yaroshchuk <yaroshchuk2000@gmail.com> | ||
19 | +S: Maintained | ||
20 | +F: net/vmnet* | ||
21 | + | ||
22 | Host Memory Backends | ||
23 | M: David Hildenbrand <david@redhat.com> | ||
24 | M: Igor Mammedov <imammedo@redhat.com> | ||
23 | -- | 25 | -- |
24 | 2.7.4 | 26 | 2.7.4 |
27 | |||
28 | diff view generated by jsdifflib |