4 * Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 #include <sys/queue.h>
40 #include <rte_common.h>
41 #include <rte_interrupts.h>
42 #include <rte_byteorder.h>
44 #include <rte_debug.h>
46 #include <rte_ether.h>
47 #include <rte_ethdev.h>
48 #include <rte_memory.h>
49 #include <rte_memzone.h>
51 #include <rte_atomic.h>
52 #include <rte_malloc.h>
55 #include "e1000_logs.h"
56 #include "base/e1000_api.h"
57 #include "e1000_ethdev.h"
59 #define EM_EIAC 0x000DC
61 #define PMD_ROUNDUP(x,y) (((x) + (y) - 1)/(y) * (y))
64 static int eth_em_configure(struct rte_eth_dev *dev);
65 static int eth_em_start(struct rte_eth_dev *dev);
66 static void eth_em_stop(struct rte_eth_dev *dev);
67 static void eth_em_close(struct rte_eth_dev *dev);
68 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
69 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
70 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
71 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
72 static int eth_em_link_update(struct rte_eth_dev *dev,
73 int wait_to_complete);
74 static void eth_em_stats_get(struct rte_eth_dev *dev,
75 struct rte_eth_stats *rte_stats);
76 static void eth_em_stats_reset(struct rte_eth_dev *dev);
77 static void eth_em_infos_get(struct rte_eth_dev *dev,
78 struct rte_eth_dev_info *dev_info);
79 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
80 struct rte_eth_fc_conf *fc_conf);
81 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
82 struct rte_eth_fc_conf *fc_conf);
83 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
84 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
85 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
86 static int eth_em_interrupt_action(struct rte_eth_dev *dev);
87 static void eth_em_interrupt_handler(struct rte_intr_handle *handle,
90 static int em_hw_init(struct e1000_hw *hw);
91 static int em_hardware_init(struct e1000_hw *hw);
92 static void em_hw_control_acquire(struct e1000_hw *hw);
93 static void em_hw_control_release(struct e1000_hw *hw);
94 static void em_init_manageability(struct e1000_hw *hw);
95 static void em_release_manageability(struct e1000_hw *hw);
97 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
99 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
100 uint16_t vlan_id, int on);
101 static void eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
102 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
103 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
104 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
105 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
108 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
109 uint16_t vlan_id, int on);
112 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
113 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
114 static void em_lsc_intr_disable(struct e1000_hw *hw);
115 static void em_rxq_intr_enable(struct e1000_hw *hw);
116 static void em_rxq_intr_disable(struct e1000_hw *hw);
118 static int eth_em_led_on(struct rte_eth_dev *dev);
119 static int eth_em_led_off(struct rte_eth_dev *dev);
121 static int em_get_rx_buffer_size(struct e1000_hw *hw);
122 static void eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
123 uint32_t index, uint32_t pool);
124 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
126 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
127 struct ether_addr *mc_addr_set,
128 uint32_t nb_mc_addr);
130 #define EM_FC_PAUSE_TIME 0x0680
131 #define EM_LINK_UPDATE_CHECK_TIMEOUT 90 /* 9s */
132 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
134 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
137 * The set of PCI devices this driver supports
139 static const struct rte_pci_id pci_id_em_map[] = {
141 #define RTE_PCI_DEV_ID_DECL_EM(vend, dev) {RTE_PCI_DEVICE(vend, dev)},
142 #include "rte_pci_dev_ids.h"
147 static const struct eth_dev_ops eth_em_ops = {
148 .dev_configure = eth_em_configure,
149 .dev_start = eth_em_start,
150 .dev_stop = eth_em_stop,
151 .dev_close = eth_em_close,
152 .promiscuous_enable = eth_em_promiscuous_enable,
153 .promiscuous_disable = eth_em_promiscuous_disable,
154 .allmulticast_enable = eth_em_allmulticast_enable,
155 .allmulticast_disable = eth_em_allmulticast_disable,
156 .link_update = eth_em_link_update,
157 .stats_get = eth_em_stats_get,
158 .stats_reset = eth_em_stats_reset,
159 .dev_infos_get = eth_em_infos_get,
160 .mtu_set = eth_em_mtu_set,
161 .vlan_filter_set = eth_em_vlan_filter_set,
162 .vlan_offload_set = eth_em_vlan_offload_set,
163 .rx_queue_setup = eth_em_rx_queue_setup,
164 .rx_queue_release = eth_em_rx_queue_release,
165 .rx_queue_count = eth_em_rx_queue_count,
166 .rx_descriptor_done = eth_em_rx_descriptor_done,
167 .tx_queue_setup = eth_em_tx_queue_setup,
168 .tx_queue_release = eth_em_tx_queue_release,
169 .rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
170 .rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
171 .dev_led_on = eth_em_led_on,
172 .dev_led_off = eth_em_led_off,
173 .flow_ctrl_get = eth_em_flow_ctrl_get,
174 .flow_ctrl_set = eth_em_flow_ctrl_set,
175 .mac_addr_add = eth_em_rar_set,
176 .mac_addr_remove = eth_em_rar_clear,
177 .set_mc_addr_list = eth_em_set_mc_addr_list,
178 .rxq_info_get = em_rxq_info_get,
179 .txq_info_get = em_txq_info_get,
183 * Atomically reads the link status information from global
184 * structure rte_eth_dev.
187 * - Pointer to the structure rte_eth_dev to read from.
188 * - Pointer to the buffer to be saved with the link status.
191 * - On success, zero.
192 * - On failure, negative value.
195 rte_em_dev_atomic_read_link_status(struct rte_eth_dev *dev,
196 struct rte_eth_link *link)
198 struct rte_eth_link *dst = link;
199 struct rte_eth_link *src = &(dev->data->dev_link);
201 if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
202 *(uint64_t *)src) == 0)
209 * Atomically writes the link status information into global
210 * structure rte_eth_dev.
213 * - Pointer to the structure rte_eth_dev to read from.
214 * - Pointer to the buffer to be saved with the link status.
217 * - On success, zero.
218 * - On failure, negative value.
221 rte_em_dev_atomic_write_link_status(struct rte_eth_dev *dev,
222 struct rte_eth_link *link)
224 struct rte_eth_link *dst = &(dev->data->dev_link);
225 struct rte_eth_link *src = link;
227 if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
228 *(uint64_t *)src) == 0)
235 * eth_em_dev_is_ich8 - Check for ICH8 device
236 * @hw: pointer to the HW structure
238 * return TRUE for ICH8, otherwise FALSE
241 eth_em_dev_is_ich8(struct e1000_hw *hw)
243 DEBUGFUNC("eth_em_dev_is_ich8");
245 switch (hw->device_id) {
246 case E1000_DEV_ID_PCH_LPT_I217_LM:
247 case E1000_DEV_ID_PCH_LPT_I217_V:
248 case E1000_DEV_ID_PCH_LPTLP_I218_LM:
249 case E1000_DEV_ID_PCH_LPTLP_I218_V:
250 case E1000_DEV_ID_PCH_I218_V2:
251 case E1000_DEV_ID_PCH_I218_LM2:
252 case E1000_DEV_ID_PCH_I218_V3:
253 case E1000_DEV_ID_PCH_I218_LM3:
261 eth_em_dev_init(struct rte_eth_dev *eth_dev)
263 struct rte_pci_device *pci_dev;
264 struct e1000_adapter *adapter =
265 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
266 struct e1000_hw *hw =
267 E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
268 struct e1000_vfta * shadow_vfta =
269 E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
271 pci_dev = eth_dev->pci_dev;
273 eth_dev->dev_ops = ð_em_ops;
274 eth_dev->rx_pkt_burst = (eth_rx_burst_t)ð_em_recv_pkts;
275 eth_dev->tx_pkt_burst = (eth_tx_burst_t)ð_em_xmit_pkts;
277 /* for secondary processes, we don't initialise any further as primary
278 * has already done this work. Only check we don't need a different
280 if (rte_eal_process_type() != RTE_PROC_PRIMARY){
281 if (eth_dev->data->scattered_rx)
282 eth_dev->rx_pkt_burst =
283 (eth_rx_burst_t)ð_em_recv_scattered_pkts;
287 rte_eth_copy_pci_info(eth_dev, pci_dev);
289 hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
290 hw->device_id = pci_dev->id.device_id;
291 adapter->stopped = 0;
293 /* For ICH8 support we'll need to map the flash memory BAR */
294 if (eth_em_dev_is_ich8(hw))
295 hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
297 if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
298 em_hw_init(hw) != 0) {
299 PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
301 eth_dev->data->port_id, pci_dev->id.vendor_id,
302 pci_dev->id.device_id);
306 /* Allocate memory for storing MAC addresses */
307 eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
308 hw->mac.rar_entry_count, 0);
309 if (eth_dev->data->mac_addrs == NULL) {
310 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
311 "store MAC addresses",
312 ETHER_ADDR_LEN * hw->mac.rar_entry_count);
316 /* Copy the permanent MAC address */
317 ether_addr_copy((struct ether_addr *) hw->mac.addr,
318 eth_dev->data->mac_addrs);
320 /* initialize the vfta */
321 memset(shadow_vfta, 0, sizeof(*shadow_vfta));
323 PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
324 eth_dev->data->port_id, pci_dev->id.vendor_id,
325 pci_dev->id.device_id);
327 rte_intr_callback_register(&(pci_dev->intr_handle),
328 eth_em_interrupt_handler, (void *)eth_dev);
334 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
336 struct rte_pci_device *pci_dev;
337 struct e1000_adapter *adapter =
338 E1000_DEV_PRIVATE(eth_dev->data->dev_private);
340 PMD_INIT_FUNC_TRACE();
342 if (rte_eal_process_type() != RTE_PROC_PRIMARY)
345 pci_dev = eth_dev->pci_dev;
347 if (adapter->stopped == 0)
348 eth_em_close(eth_dev);
350 eth_dev->dev_ops = NULL;
351 eth_dev->rx_pkt_burst = NULL;
352 eth_dev->tx_pkt_burst = NULL;
354 rte_free(eth_dev->data->mac_addrs);
355 eth_dev->data->mac_addrs = NULL;
357 /* disable uio intr before callback unregister */
358 rte_intr_disable(&(pci_dev->intr_handle));
359 rte_intr_callback_unregister(&(pci_dev->intr_handle),
360 eth_em_interrupt_handler, (void *)eth_dev);
365 static struct eth_driver rte_em_pmd = {
367 .name = "rte_em_pmd",
368 .id_table = pci_id_em_map,
369 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
370 RTE_PCI_DRV_DETACHABLE,
372 .eth_dev_init = eth_em_dev_init,
373 .eth_dev_uninit = eth_em_dev_uninit,
374 .dev_private_size = sizeof(struct e1000_adapter),
378 rte_em_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
380 rte_eth_driver_register(&rte_em_pmd);
385 em_hw_init(struct e1000_hw *hw)
389 diag = hw->mac.ops.init_params(hw);
391 PMD_INIT_LOG(ERR, "MAC Initialization Error");
394 diag = hw->nvm.ops.init_params(hw);
396 PMD_INIT_LOG(ERR, "NVM Initialization Error");
399 diag = hw->phy.ops.init_params(hw);
401 PMD_INIT_LOG(ERR, "PHY Initialization Error");
404 (void) e1000_get_bus_info(hw);
407 hw->phy.autoneg_wait_to_complete = 0;
408 hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
410 e1000_init_script_state_82541(hw, TRUE);
411 e1000_set_tbi_compatibility_82543(hw, TRUE);
414 if (hw->phy.media_type == e1000_media_type_copper) {
415 hw->phy.mdix = 0; /* AUTO_ALL_MODES */
416 hw->phy.disable_polarity_correction = 0;
417 hw->phy.ms_type = e1000_ms_hw_default;
421 * Start from a known state, this is important in reading the nvm
426 /* Make sure we have a good EEPROM before we read from it */
427 if (e1000_validate_nvm_checksum(hw) < 0) {
429 * Some PCI-E parts fail the first check due to
430 * the link being in sleep state, call it again,
431 * if it fails a second time its a real issue.
433 diag = e1000_validate_nvm_checksum(hw);
435 PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
440 /* Read the permanent MAC address out of the EEPROM */
441 diag = e1000_read_mac_addr(hw);
443 PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
447 /* Now initialize the hardware */
448 diag = em_hardware_init(hw);
450 PMD_INIT_LOG(ERR, "Hardware initialization failed");
454 hw->mac.get_link_status = 1;
456 /* Indicate SOL/IDER usage */
457 diag = e1000_check_reset_block(hw);
459 PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
465 em_hw_control_release(hw);
470 eth_em_configure(struct rte_eth_dev *dev)
472 struct e1000_interrupt *intr =
473 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
475 PMD_INIT_FUNC_TRACE();
476 intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
477 PMD_INIT_FUNC_TRACE();
483 em_set_pba(struct e1000_hw *hw)
488 * Packet Buffer Allocation (PBA)
489 * Writing PBA sets the receive portion of the buffer
490 * the remainder is used for the transmit buffer.
491 * Devices before the 82547 had a Packet Buffer of 64K.
492 * After the 82547 the buffer was reduced to 40K.
494 switch (hw->mac.type) {
496 case e1000_82547_rev_2:
497 /* 82547: Total Packet Buffer is 40K */
498 pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
502 case e1000_80003es2lan:
503 pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
505 case e1000_82573: /* 82573: Total Packet Buffer is 32K */
506 pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
510 pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
525 pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
528 E1000_WRITE_REG(hw, E1000_PBA, pba);
532 eth_em_start(struct rte_eth_dev *dev)
534 struct e1000_adapter *adapter =
535 E1000_DEV_PRIVATE(dev->data->dev_private);
536 struct e1000_hw *hw =
537 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
538 struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
540 uint32_t intr_vector = 0;
545 PMD_INIT_FUNC_TRACE();
549 e1000_power_up_phy(hw);
551 /* Set default PBA value */
554 /* Put the address into the Receive Address Array */
555 e1000_rar_set(hw, hw->mac.addr, 0);
558 * With the 82571 adapter, RAR[0] may be overwritten
559 * when the other port is reset, we make a duplicate
560 * in RAR[14] for that eventuality, this assures
561 * the interface continues to function.
563 if (hw->mac.type == e1000_82571) {
564 e1000_set_laa_state_82571(hw, TRUE);
565 e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
568 /* Initialize the hardware */
569 if (em_hardware_init(hw)) {
570 PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
574 E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
576 /* Configure for OS presence */
577 em_init_manageability(hw);
579 if (dev->data->dev_conf.intr_conf.rxq != 0) {
580 intr_vector = dev->data->nb_rx_queues;
581 if (rte_intr_efd_enable(intr_handle, intr_vector))
585 if (rte_intr_dp_is_en(intr_handle)) {
586 intr_handle->intr_vec =
587 rte_zmalloc("intr_vec",
588 dev->data->nb_rx_queues * sizeof(int), 0);
589 if (intr_handle->intr_vec == NULL) {
590 PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
591 " intr_vec\n", dev->data->nb_rx_queues);
595 /* enable rx interrupt */
596 em_rxq_intr_enable(hw);
601 ret = eth_em_rx_init(dev);
603 PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
604 em_dev_clear_queues(dev);
608 e1000_clear_hw_cntrs_base_generic(hw);
610 mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
611 ETH_VLAN_EXTEND_MASK;
612 eth_em_vlan_offload_set(dev, mask);
614 /* Set Interrupt Throttling Rate to maximum allowed value. */
615 E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
617 /* Setup link speed and duplex */
618 speeds = &dev->data->dev_conf.link_speeds;
619 if (*speeds == ETH_LINK_SPEED_AUTONEG) {
620 hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
623 autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
626 hw->phy.autoneg_advertised = 0;
628 if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
629 ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
630 ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
632 goto error_invalid_config;
634 if (*speeds & ETH_LINK_SPEED_10M_HD) {
635 hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
638 if (*speeds & ETH_LINK_SPEED_10M) {
639 hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
642 if (*speeds & ETH_LINK_SPEED_100M_HD) {
643 hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
646 if (*speeds & ETH_LINK_SPEED_100M) {
647 hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
650 if (*speeds & ETH_LINK_SPEED_1G) {
651 hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
654 if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
655 goto error_invalid_config;
658 e1000_setup_link(hw);
660 if (rte_intr_allow_others(intr_handle)) {
661 /* check if lsc interrupt is enabled */
662 if (dev->data->dev_conf.intr_conf.lsc != 0) {
663 ret = eth_em_interrupt_setup(dev);
665 PMD_INIT_LOG(ERR, "Unable to setup interrupts");
666 em_dev_clear_queues(dev);
671 rte_intr_callback_unregister(intr_handle,
672 eth_em_interrupt_handler,
674 if (dev->data->dev_conf.intr_conf.lsc != 0)
675 PMD_INIT_LOG(INFO, "lsc won't enable because of"
676 " no intr multiplex\n");
678 /* check if rxq interrupt is enabled */
679 if (dev->data->dev_conf.intr_conf.rxq != 0)
680 eth_em_rxq_interrupt_setup(dev);
682 rte_intr_enable(intr_handle);
684 adapter->stopped = 0;
686 PMD_INIT_LOG(DEBUG, "<<");
690 error_invalid_config:
691 PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
692 dev->data->dev_conf.link_speeds, dev->data->port_id);
693 em_dev_clear_queues(dev);
697 /*********************************************************************
699 * This routine disables all traffic on the adapter by issuing a
700 * global reset on the MAC.
702 **********************************************************************/
704 eth_em_stop(struct rte_eth_dev *dev)
706 struct rte_eth_link link;
707 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
708 struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
710 em_rxq_intr_disable(hw);
711 em_lsc_intr_disable(hw);
714 if (hw->mac.type >= e1000_82544)
715 E1000_WRITE_REG(hw, E1000_WUC, 0);
717 /* Power down the phy. Needed to make the link go down */
718 e1000_power_down_phy(hw);
720 em_dev_clear_queues(dev);
722 /* clear the recorded link status */
723 memset(&link, 0, sizeof(link));
724 rte_em_dev_atomic_write_link_status(dev, &link);
726 if (!rte_intr_allow_others(intr_handle))
727 /* resume to the default handler */
728 rte_intr_callback_register(intr_handle,
729 eth_em_interrupt_handler,
732 /* Clean datapath event and queue/vec mapping */
733 rte_intr_efd_disable(intr_handle);
734 if (intr_handle->intr_vec != NULL) {
735 rte_free(intr_handle->intr_vec);
736 intr_handle->intr_vec = NULL;
741 eth_em_close(struct rte_eth_dev *dev)
743 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
744 struct e1000_adapter *adapter =
745 E1000_DEV_PRIVATE(dev->data->dev_private);
748 adapter->stopped = 1;
749 em_dev_free_queues(dev);
750 e1000_phy_hw_reset(hw);
751 em_release_manageability(hw);
752 em_hw_control_release(hw);
756 em_get_rx_buffer_size(struct e1000_hw *hw)
758 uint32_t rx_buf_size;
760 rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
764 /*********************************************************************
766 * Initialize the hardware
768 **********************************************************************/
770 em_hardware_init(struct e1000_hw *hw)
772 uint32_t rx_buf_size;
775 /* Issue a global reset */
778 /* Let the firmware know the OS is in control */
779 em_hw_control_acquire(hw);
782 * These parameters control the automatic generation (Tx) and
783 * response (Rx) to Ethernet PAUSE frames.
784 * - High water mark should allow for at least two standard size (1518)
785 * frames to be received after sending an XOFF.
786 * - Low water mark works best when it is very near the high water mark.
787 * This allows the receiver to restart by sending XON when it has
788 * drained a bit. Here we use an arbitrary value of 1500 which will
789 * restart after one full frame is pulled from the buffer. There
790 * could be several smaller frames in the buffer and if so they will
791 * not trigger the XON until their total number reduces the buffer
793 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
795 rx_buf_size = em_get_rx_buffer_size(hw);
797 hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
798 hw->fc.low_water = hw->fc.high_water - 1500;
800 if (hw->mac.type == e1000_80003es2lan)
801 hw->fc.pause_time = UINT16_MAX;
803 hw->fc.pause_time = EM_FC_PAUSE_TIME;
807 /* Set Flow control, use the tunable location if sane */
808 if (em_fc_setting <= e1000_fc_full)
809 hw->fc.requested_mode = em_fc_setting;
811 hw->fc.requested_mode = e1000_fc_none;
813 /* Workaround: no TX flow ctrl for PCH */
814 if (hw->mac.type == e1000_pchlan)
815 hw->fc.requested_mode = e1000_fc_rx_pause;
817 /* Override - settings for PCH2LAN, ya its magic :) */
818 if (hw->mac.type == e1000_pch2lan) {
819 hw->fc.high_water = 0x5C20;
820 hw->fc.low_water = 0x5048;
821 hw->fc.pause_time = 0x0650;
822 hw->fc.refresh_time = 0x0400;
823 } else if (hw->mac.type == e1000_pch_lpt) {
824 hw->fc.requested_mode = e1000_fc_full;
827 diag = e1000_init_hw(hw);
830 e1000_check_for_link(hw);
834 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
836 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
838 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
839 struct e1000_hw_stats *stats =
840 E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
843 if(hw->phy.media_type == e1000_media_type_copper ||
844 (E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
845 stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
846 stats->sec += E1000_READ_REG(hw, E1000_SEC);
849 stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
850 stats->mpc += E1000_READ_REG(hw, E1000_MPC);
851 stats->scc += E1000_READ_REG(hw, E1000_SCC);
852 stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
854 stats->mcc += E1000_READ_REG(hw, E1000_MCC);
855 stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
856 stats->colc += E1000_READ_REG(hw, E1000_COLC);
857 stats->dc += E1000_READ_REG(hw, E1000_DC);
858 stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
859 stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
860 stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
863 * For watchdog management we need to know if we have been
864 * paused during the last interval, so capture that here.
866 pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
867 stats->xoffrxc += pause_frames;
868 stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
869 stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
870 stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
871 stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
872 stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
873 stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
874 stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
875 stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
876 stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
877 stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
878 stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
879 stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
882 * For the 64-bit byte counters the low dword must be read first.
883 * Both registers clear on the read of the high dword.
886 stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
887 stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
888 stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
889 stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
891 stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
892 stats->ruc += E1000_READ_REG(hw, E1000_RUC);
893 stats->rfc += E1000_READ_REG(hw, E1000_RFC);
894 stats->roc += E1000_READ_REG(hw, E1000_ROC);
895 stats->rjc += E1000_READ_REG(hw, E1000_RJC);
897 stats->tor += E1000_READ_REG(hw, E1000_TORH);
898 stats->tot += E1000_READ_REG(hw, E1000_TOTH);
900 stats->tpr += E1000_READ_REG(hw, E1000_TPR);
901 stats->tpt += E1000_READ_REG(hw, E1000_TPT);
902 stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
903 stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
904 stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
905 stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
906 stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
907 stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
908 stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
909 stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
911 /* Interrupt Counts */
913 if (hw->mac.type >= e1000_82571) {
914 stats->iac += E1000_READ_REG(hw, E1000_IAC);
915 stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
916 stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
917 stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
918 stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
919 stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
920 stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
921 stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
922 stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
925 if (hw->mac.type >= e1000_82543) {
926 stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
927 stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
928 stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
929 stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
930 stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
931 stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
934 if (rte_stats == NULL)
938 rte_stats->imissed = stats->mpc;
939 rte_stats->ierrors = stats->crcerrs +
940 stats->rlec + stats->ruc + stats->roc +
941 stats->rxerrc + stats->algnerrc + stats->cexterr;
944 rte_stats->oerrors = stats->ecol + stats->latecol;
946 rte_stats->ipackets = stats->gprc;
947 rte_stats->opackets = stats->gptc;
948 rte_stats->ibytes = stats->gorc;
949 rte_stats->obytes = stats->gotc;
953 eth_em_stats_reset(struct rte_eth_dev *dev)
955 struct e1000_hw_stats *hw_stats =
956 E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
958 /* HW registers are cleared on read */
959 eth_em_stats_get(dev, NULL);
961 /* Reset software totals */
962 memset(hw_stats, 0, sizeof(*hw_stats));
966 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
968 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
970 em_rxq_intr_enable(hw);
971 rte_intr_enable(&dev->pci_dev->intr_handle);
977 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
979 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
981 em_rxq_intr_disable(hw);
987 em_get_max_pktlen(const struct e1000_hw *hw)
989 switch (hw->mac.type) {
997 case e1000_80003es2lan: /* 9K Jumbo Frame size */
1002 /* Adapters that do not support jumbo frames */
1004 return ETHER_MAX_LEN;
1006 return MAX_JUMBO_FRAME_SIZE;
1011 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1013 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1015 dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1016 dev_info->max_rx_pktlen = em_get_max_pktlen(hw);
1017 dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1020 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1021 * Unfortunatelly, all these nics have just one TX context.
1022 * So we have few choises for TX:
1023 * - Use just one TX queue.
1024 * - Allow cksum offload only for one TX queue.
1025 * - Don't allow TX cksum offload at all.
1026 * For now, option #1 was chosen.
1027 * To use second RX queue we have to use extended RX descriptor
1028 * (Multiple Receive Queues are mutually exclusive with UDP
1029 * fragmentation and are not supported when a legacy receive
1030 * descriptor format is used).
1031 * Which means separate RX routinies - as legacy nics (82540, 82545)
1032 * don't support extended RXD.
1033 * To avoid it we support just one RX queue for now (no RSS).
1036 dev_info->max_rx_queues = 1;
1037 dev_info->max_tx_queues = 1;
1039 dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1040 .nb_max = E1000_MAX_RING_DESC,
1041 .nb_min = E1000_MIN_RING_DESC,
1042 .nb_align = EM_RXD_ALIGN,
1045 dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1046 .nb_max = E1000_MAX_RING_DESC,
1047 .nb_min = E1000_MIN_RING_DESC,
1048 .nb_align = EM_TXD_ALIGN,
1051 dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1052 ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1056 /* return 0 means link status changed, -1 means not changed */
1058 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1060 struct e1000_hw *hw =
1061 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1062 struct rte_eth_link link, old;
1063 int link_check, count;
1066 hw->mac.get_link_status = 1;
1068 /* possible wait-to-complete in up to 9 seconds */
1069 for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1070 /* Read the real link status */
1071 switch (hw->phy.media_type) {
1072 case e1000_media_type_copper:
1073 /* Do the work to read phy */
1074 e1000_check_for_link(hw);
1075 link_check = !hw->mac.get_link_status;
1078 case e1000_media_type_fiber:
1079 e1000_check_for_link(hw);
1080 link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1084 case e1000_media_type_internal_serdes:
1085 e1000_check_for_link(hw);
1086 link_check = hw->mac.serdes_has_link;
1092 if (link_check || wait_to_complete == 0)
1094 rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1096 memset(&link, 0, sizeof(link));
1097 rte_em_dev_atomic_read_link_status(dev, &link);
1100 /* Now we check if a transition has happened */
1101 if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1102 uint16_t duplex, speed;
1103 hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1104 link.link_duplex = (duplex == FULL_DUPLEX) ?
1105 ETH_LINK_FULL_DUPLEX :
1106 ETH_LINK_HALF_DUPLEX;
1107 link.link_speed = speed;
1108 link.link_status = ETH_LINK_UP;
1109 link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1110 ETH_LINK_SPEED_FIXED);
1111 } else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1112 link.link_speed = 0;
1113 link.link_duplex = ETH_LINK_HALF_DUPLEX;
1114 link.link_status = ETH_LINK_DOWN;
1115 link.link_autoneg = ETH_LINK_SPEED_FIXED;
1117 rte_em_dev_atomic_write_link_status(dev, &link);
1120 if (old.link_status == link.link_status)
1128 * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1129 * For ASF and Pass Through versions of f/w this means
1130 * that the driver is loaded. For AMT version type f/w
1131 * this means that the network i/f is open.
1134 em_hw_control_acquire(struct e1000_hw *hw)
1136 uint32_t ctrl_ext, swsm;
1138 /* Let firmware know the driver has taken over */
1139 if (hw->mac.type == e1000_82573) {
1140 swsm = E1000_READ_REG(hw, E1000_SWSM);
1141 E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1144 ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1145 E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1146 ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1151 * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1152 * For ASF and Pass Through versions of f/w this means that the
1153 * driver is no longer loaded. For AMT versions of the
1154 * f/w this means that the network i/f is closed.
1157 em_hw_control_release(struct e1000_hw *hw)
1159 uint32_t ctrl_ext, swsm;
1161 /* Let firmware taken over control of h/w */
1162 if (hw->mac.type == e1000_82573) {
1163 swsm = E1000_READ_REG(hw, E1000_SWSM);
1164 E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1166 ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1167 E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1168 ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1173 * Bit of a misnomer, what this really means is
1174 * to enable OS management of the system... aka
1175 * to disable special hardware management features.
1178 em_init_manageability(struct e1000_hw *hw)
1180 if (e1000_enable_mng_pass_thru(hw)) {
1181 uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1182 uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1184 /* disable hardware interception of ARP */
1185 manc &= ~(E1000_MANC_ARP_EN);
1187 /* enable receiving management packets to the host */
1188 manc |= E1000_MANC_EN_MNG2HOST;
1189 manc2h |= 1 << 5; /* Mng Port 623 */
1190 manc2h |= 1 << 6; /* Mng Port 664 */
1191 E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1192 E1000_WRITE_REG(hw, E1000_MANC, manc);
1197 * Give control back to hardware management
1198 * controller if there is one.
1201 em_release_manageability(struct e1000_hw *hw)
1205 if (e1000_enable_mng_pass_thru(hw)) {
1206 manc = E1000_READ_REG(hw, E1000_MANC);
1208 /* re-enable hardware interception of ARP */
1209 manc |= E1000_MANC_ARP_EN;
1210 manc &= ~E1000_MANC_EN_MNG2HOST;
1212 E1000_WRITE_REG(hw, E1000_MANC, manc);
1217 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1219 struct e1000_hw *hw =
1220 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1223 rctl = E1000_READ_REG(hw, E1000_RCTL);
1224 rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1225 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1229 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1231 struct e1000_hw *hw =
1232 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1235 rctl = E1000_READ_REG(hw, E1000_RCTL);
1236 rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1237 if (dev->data->all_multicast == 1)
1238 rctl |= E1000_RCTL_MPE;
1240 rctl &= (~E1000_RCTL_MPE);
1241 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1245 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1247 struct e1000_hw *hw =
1248 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1251 rctl = E1000_READ_REG(hw, E1000_RCTL);
1252 rctl |= E1000_RCTL_MPE;
1253 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1257 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1259 struct e1000_hw *hw =
1260 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1263 if (dev->data->promiscuous == 1)
1264 return; /* must remain in all_multicast mode */
1265 rctl = E1000_READ_REG(hw, E1000_RCTL);
1266 rctl &= (~E1000_RCTL_MPE);
1267 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1271 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1273 struct e1000_hw *hw =
1274 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1275 struct e1000_vfta * shadow_vfta =
1276 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1281 vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1282 E1000_VFTA_ENTRY_MASK);
1283 vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1284 vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1289 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1291 /* update local VFTA copy */
1292 shadow_vfta->vfta[vid_idx] = vfta;
1298 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1300 struct e1000_hw *hw =
1301 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1304 /* Filter Table Disable */
1305 reg = E1000_READ_REG(hw, E1000_RCTL);
1306 reg &= ~E1000_RCTL_CFIEN;
1307 reg &= ~E1000_RCTL_VFE;
1308 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1312 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1314 struct e1000_hw *hw =
1315 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1316 struct e1000_vfta * shadow_vfta =
1317 E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1321 /* Filter Table Enable, CFI not used for packet acceptance */
1322 reg = E1000_READ_REG(hw, E1000_RCTL);
1323 reg &= ~E1000_RCTL_CFIEN;
1324 reg |= E1000_RCTL_VFE;
1325 E1000_WRITE_REG(hw, E1000_RCTL, reg);
1327 /* restore vfta from local copy */
1328 for (i = 0; i < IGB_VFTA_SIZE; i++)
1329 E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1333 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1335 struct e1000_hw *hw =
1336 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1339 /* VLAN Mode Disable */
1340 reg = E1000_READ_REG(hw, E1000_CTRL);
1341 reg &= ~E1000_CTRL_VME;
1342 E1000_WRITE_REG(hw, E1000_CTRL, reg);
1347 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1349 struct e1000_hw *hw =
1350 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1353 /* VLAN Mode Enable */
1354 reg = E1000_READ_REG(hw, E1000_CTRL);
1355 reg |= E1000_CTRL_VME;
1356 E1000_WRITE_REG(hw, E1000_CTRL, reg);
1360 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1362 if(mask & ETH_VLAN_STRIP_MASK){
1363 if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1364 em_vlan_hw_strip_enable(dev);
1366 em_vlan_hw_strip_disable(dev);
1369 if(mask & ETH_VLAN_FILTER_MASK){
1370 if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1371 em_vlan_hw_filter_enable(dev);
1373 em_vlan_hw_filter_disable(dev);
1378 * It enables the interrupt mask and then enable the interrupt.
1381 * Pointer to struct rte_eth_dev.
1384 * - On success, zero.
1385 * - On failure, a negative value.
1388 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1391 struct e1000_hw *hw =
1392 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1394 /* clear interrupt */
1395 E1000_READ_REG(hw, E1000_ICR);
1396 regval = E1000_READ_REG(hw, E1000_IMS);
1397 E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1402 * It clears the interrupt causes and enables the interrupt.
1403 * It will be called once only during nic initialized.
1406 * Pointer to struct rte_eth_dev.
1409 * - On success, zero.
1410 * - On failure, a negative value.
1413 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1415 struct e1000_hw *hw =
1416 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1418 E1000_READ_REG(hw, E1000_ICR);
1419 em_rxq_intr_enable(hw);
1424 * It enable receive packet interrupt.
1426 * Pointer to struct e1000_hw
1431 em_rxq_intr_enable(struct e1000_hw *hw)
1433 E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1434 E1000_WRITE_FLUSH(hw);
1438 * It disabled lsc interrupt.
1440 * Pointer to struct e1000_hw
1445 em_lsc_intr_disable(struct e1000_hw *hw)
1447 E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1448 E1000_WRITE_FLUSH(hw);
1452 * It disabled receive packet interrupt.
1454 * Pointer to struct e1000_hw
1459 em_rxq_intr_disable(struct e1000_hw *hw)
1461 E1000_READ_REG(hw, E1000_ICR);
1462 E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1463 E1000_WRITE_FLUSH(hw);
1467 * It reads ICR and gets interrupt causes, check it and set a bit flag
1468 * to update link status.
1471 * Pointer to struct rte_eth_dev.
1474 * - On success, zero.
1475 * - On failure, a negative value.
1478 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1481 struct e1000_hw *hw =
1482 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1483 struct e1000_interrupt *intr =
1484 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1486 /* read-on-clear nic registers here */
1487 icr = E1000_READ_REG(hw, E1000_ICR);
1488 if (icr & E1000_ICR_LSC) {
1489 intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1496 * It executes link_update after knowing an interrupt is prsent.
1499 * Pointer to struct rte_eth_dev.
1502 * - On success, zero.
1503 * - On failure, a negative value.
1506 eth_em_interrupt_action(struct rte_eth_dev *dev)
1508 struct e1000_hw *hw =
1509 E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1510 struct e1000_interrupt *intr =
1511 E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1512 uint32_t tctl, rctl;
1513 struct rte_eth_link link;
1516 if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1519 intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1520 rte_intr_enable(&(dev->pci_dev->intr_handle));
1522 /* set get_link_status to check register later */
1523 hw->mac.get_link_status = 1;
1524 ret = eth_em_link_update(dev, 0);
1526 /* check if link has changed */
1530 memset(&link, 0, sizeof(link));
1531 rte_em_dev_atomic_read_link_status(dev, &link);
1532 if (link.link_status) {
1533 PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1534 dev->data->port_id, (unsigned)link.link_speed,
1535 link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1536 "full-duplex" : "half-duplex");
1538 PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1540 PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1541 dev->pci_dev->addr.domain, dev->pci_dev->addr.bus,
1542 dev->pci_dev->addr.devid, dev->pci_dev->addr.function);
1544 tctl = E1000_READ_REG(hw, E1000_TCTL);
1545 rctl = E1000_READ_REG(hw, E1000_RCTL);
1546 if (link.link_status) {
1548 tctl |= E1000_TCTL_EN;
1549 rctl |= E1000_RCTL_EN;
1552 tctl &= ~E1000_TCTL_EN;
1553 rctl &= ~E1000_RCTL_EN;
1555 E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1556 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1557 E1000_WRITE_FLUSH(hw);
1563 * Interrupt handler which shall be registered at first.
1566 * Pointer to interrupt handle.
1568 * The address of parameter (struct rte_eth_dev *) regsitered before.
1574 eth_em_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
1577 struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1579 eth_em_interrupt_get_status(dev);
1580 eth_em_interrupt_action(dev);
1581 _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
1585 eth_em_led_on(struct rte_eth_dev *dev)
1587 struct e1000_hw *hw;
1589 hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1590 return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1594 eth_em_led_off(struct rte_eth_dev *dev)
1596 struct e1000_hw *hw;
1598 hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1599 return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1603 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1605 struct e1000_hw *hw;
1610 hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1611 fc_conf->pause_time = hw->fc.pause_time;
1612 fc_conf->high_water = hw->fc.high_water;
1613 fc_conf->low_water = hw->fc.low_water;
1614 fc_conf->send_xon = hw->fc.send_xon;
1615 fc_conf->autoneg = hw->mac.autoneg;
1618 * Return rx_pause and tx_pause status according to actual setting of
1619 * the TFCE and RFCE bits in the CTRL register.
1621 ctrl = E1000_READ_REG(hw, E1000_CTRL);
1622 if (ctrl & E1000_CTRL_TFCE)
1627 if (ctrl & E1000_CTRL_RFCE)
1632 if (rx_pause && tx_pause)
1633 fc_conf->mode = RTE_FC_FULL;
1635 fc_conf->mode = RTE_FC_RX_PAUSE;
1637 fc_conf->mode = RTE_FC_TX_PAUSE;
1639 fc_conf->mode = RTE_FC_NONE;
1645 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1647 struct e1000_hw *hw;
1649 enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1655 uint32_t rx_buf_size;
1656 uint32_t max_high_water;
1659 hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1660 if (fc_conf->autoneg != hw->mac.autoneg)
1662 rx_buf_size = em_get_rx_buffer_size(hw);
1663 PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1665 /* At least reserve one Ethernet frame for watermark */
1666 max_high_water = rx_buf_size - ETHER_MAX_LEN;
1667 if ((fc_conf->high_water > max_high_water) ||
1668 (fc_conf->high_water < fc_conf->low_water)) {
1669 PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1670 PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1674 hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1675 hw->fc.pause_time = fc_conf->pause_time;
1676 hw->fc.high_water = fc_conf->high_water;
1677 hw->fc.low_water = fc_conf->low_water;
1678 hw->fc.send_xon = fc_conf->send_xon;
1680 err = e1000_setup_link_generic(hw);
1681 if (err == E1000_SUCCESS) {
1683 /* check if we want to forward MAC frames - driver doesn't have native
1684 * capability to do that, so we'll write the registers ourselves */
1686 rctl = E1000_READ_REG(hw, E1000_RCTL);
1688 /* set or clear MFLCN.PMCF bit depending on configuration */
1689 if (fc_conf->mac_ctrl_frame_fwd != 0)
1690 rctl |= E1000_RCTL_PMCF;
1692 rctl &= ~E1000_RCTL_PMCF;
1694 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1695 E1000_WRITE_FLUSH(hw);
1700 PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1705 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1706 uint32_t index, __rte_unused uint32_t pool)
1708 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1710 e1000_rar_set(hw, mac_addr->addr_bytes, index);
1714 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1716 uint8_t addr[ETHER_ADDR_LEN];
1717 struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1719 memset(addr, 0, sizeof(addr));
1721 e1000_rar_set(hw, addr, index);
1725 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1727 struct rte_eth_dev_info dev_info;
1728 struct e1000_hw *hw;
1729 uint32_t frame_size;
1732 eth_em_infos_get(dev, &dev_info);
1733 frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1735 /* check that mtu is within the allowed range */
1736 if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1739 /* refuse mtu that requires the support of scattered packets when this
1740 * feature has not been enabled before. */
1741 if (!dev->data->scattered_rx &&
1742 frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1745 hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1746 rctl = E1000_READ_REG(hw, E1000_RCTL);
1748 /* switch to jumbo mode if needed */
1749 if (frame_size > ETHER_MAX_LEN) {
1750 dev->data->dev_conf.rxmode.jumbo_frame = 1;
1751 rctl |= E1000_RCTL_LPE;
1753 dev->data->dev_conf.rxmode.jumbo_frame = 0;
1754 rctl &= ~E1000_RCTL_LPE;
1756 E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1758 /* update max frame size */
1759 dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1764 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1765 struct ether_addr *mc_addr_set,
1766 uint32_t nb_mc_addr)
1768 struct e1000_hw *hw;
1770 hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1771 e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1775 struct rte_driver em_pmd_drv = {
1777 .init = rte_em_pmd_init,
1780 PMD_REGISTER_DRIVER(em_pmd_drv);