xref: /linux/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c (revision 0883c2c06fb5bcf5b9e008270827e63c09a88c1e)
1 /*******************************************************************************
2 
3   Intel 10 Gigabit PCI Express Linux driver
4   Copyright(c) 1999 - 2015 Intel Corporation.
5 
6   This program is free software; you can redistribute it and/or modify it
7   under the terms and conditions of the GNU General Public License,
8   version 2, as published by the Free Software Foundation.
9 
10   This program is distributed in the hope it will be useful, but WITHOUT
11   ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12   FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13   more details.
14 
15   You should have received a copy of the GNU General Public License along with
16   this program; if not, write to the Free Software Foundation, Inc.,
17   51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18 
19   The full GNU General Public License is included in this distribution in
20   the file called "COPYING".
21 
22   Contact Information:
23   Linux NICS <linux.nics@intel.com>
24   e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
25   Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 
27 *******************************************************************************/
28 
29 #include <linux/types.h>
30 #include <linux/module.h>
31 #include <linux/pci.h>
32 #include <linux/netdevice.h>
33 #include <linux/vmalloc.h>
34 #include <linux/string.h>
35 #include <linux/in.h>
36 #include <linux/ip.h>
37 #include <linux/tcp.h>
38 #include <linux/ipv6.h>
39 #include <linux/if_bridge.h>
40 #ifdef NETIF_F_HW_VLAN_CTAG_TX
41 #include <linux/if_vlan.h>
42 #endif
43 
44 #include "ixgbe.h"
45 #include "ixgbe_type.h"
46 #include "ixgbe_sriov.h"
47 
48 #ifdef CONFIG_PCI_IOV
49 static int __ixgbe_enable_sriov(struct ixgbe_adapter *adapter)
50 {
51 	struct ixgbe_hw *hw = &adapter->hw;
52 	int num_vf_macvlans, i;
53 	struct vf_macvlans *mv_list;
54 
55 	adapter->flags |= IXGBE_FLAG_SRIOV_ENABLED;
56 	e_info(probe, "SR-IOV enabled with %d VFs\n", adapter->num_vfs);
57 
58 	/* Enable VMDq flag so device will be set in VM mode */
59 	adapter->flags |= IXGBE_FLAG_VMDQ_ENABLED;
60 	if (!adapter->ring_feature[RING_F_VMDQ].limit)
61 		adapter->ring_feature[RING_F_VMDQ].limit = 1;
62 	adapter->ring_feature[RING_F_VMDQ].offset = adapter->num_vfs;
63 
64 	num_vf_macvlans = hw->mac.num_rar_entries -
65 	(IXGBE_MAX_PF_MACVLANS + 1 + adapter->num_vfs);
66 
67 	adapter->mv_list = mv_list = kcalloc(num_vf_macvlans,
68 					     sizeof(struct vf_macvlans),
69 					     GFP_KERNEL);
70 	if (mv_list) {
71 		/* Initialize list of VF macvlans */
72 		INIT_LIST_HEAD(&adapter->vf_mvs.l);
73 		for (i = 0; i < num_vf_macvlans; i++) {
74 			mv_list->vf = -1;
75 			mv_list->free = true;
76 			list_add(&mv_list->l, &adapter->vf_mvs.l);
77 			mv_list++;
78 		}
79 	}
80 
81 	/* Initialize default switching mode VEB */
82 	IXGBE_WRITE_REG(hw, IXGBE_PFDTXGSWC, IXGBE_PFDTXGSWC_VT_LBEN);
83 	adapter->bridge_mode = BRIDGE_MODE_VEB;
84 
85 	/* If call to enable VFs succeeded then allocate memory
86 	 * for per VF control structures.
87 	 */
88 	adapter->vfinfo =
89 		kcalloc(adapter->num_vfs,
90 			sizeof(struct vf_data_storage), GFP_KERNEL);
91 	if (adapter->vfinfo) {
92 		/* limit trafffic classes based on VFs enabled */
93 		if ((adapter->hw.mac.type == ixgbe_mac_82599EB) &&
94 		    (adapter->num_vfs < 16)) {
95 			adapter->dcb_cfg.num_tcs.pg_tcs = MAX_TRAFFIC_CLASS;
96 			adapter->dcb_cfg.num_tcs.pfc_tcs = MAX_TRAFFIC_CLASS;
97 		} else if (adapter->num_vfs < 32) {
98 			adapter->dcb_cfg.num_tcs.pg_tcs = 4;
99 			adapter->dcb_cfg.num_tcs.pfc_tcs = 4;
100 		} else {
101 			adapter->dcb_cfg.num_tcs.pg_tcs = 1;
102 			adapter->dcb_cfg.num_tcs.pfc_tcs = 1;
103 		}
104 
105 		/* Disable RSC when in SR-IOV mode */
106 		adapter->flags2 &= ~(IXGBE_FLAG2_RSC_CAPABLE |
107 				     IXGBE_FLAG2_RSC_ENABLED);
108 
109 		for (i = 0; i < adapter->num_vfs; i++) {
110 			/* enable spoof checking for all VFs */
111 			adapter->vfinfo[i].spoofchk_enabled = true;
112 
113 			/* We support VF RSS querying only for 82599 and x540
114 			 * devices at the moment. These devices share RSS
115 			 * indirection table and RSS hash key with PF therefore
116 			 * we want to disable the querying by default.
117 			 */
118 			adapter->vfinfo[i].rss_query_enabled = 0;
119 
120 			/* Untrust all VFs */
121 			adapter->vfinfo[i].trusted = false;
122 
123 			/* set the default xcast mode */
124 			adapter->vfinfo[i].xcast_mode = IXGBEVF_XCAST_MODE_NONE;
125 		}
126 
127 		return 0;
128 	}
129 
130 	return -ENOMEM;
131 }
132 
133 /**
134  * ixgbe_get_vfs - Find and take references to all vf devices
135  * @adapter: Pointer to adapter struct
136  */
137 static void ixgbe_get_vfs(struct ixgbe_adapter *adapter)
138 {
139 	struct pci_dev *pdev = adapter->pdev;
140 	u16 vendor = pdev->vendor;
141 	struct pci_dev *vfdev;
142 	int vf = 0;
143 	u16 vf_id;
144 	int pos;
145 
146 	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_SRIOV);
147 	if (!pos)
148 		return;
149 	pci_read_config_word(pdev, pos + PCI_SRIOV_VF_DID, &vf_id);
150 
151 	vfdev = pci_get_device(vendor, vf_id, NULL);
152 	for (; vfdev; vfdev = pci_get_device(vendor, vf_id, vfdev)) {
153 		if (!vfdev->is_virtfn)
154 			continue;
155 		if (vfdev->physfn != pdev)
156 			continue;
157 		if (vf >= adapter->num_vfs)
158 			continue;
159 		pci_dev_get(vfdev);
160 		adapter->vfinfo[vf].vfdev = vfdev;
161 		++vf;
162 	}
163 }
164 
165 /* Note this function is called when the user wants to enable SR-IOV
166  * VFs using the now deprecated module parameter
167  */
168 void ixgbe_enable_sriov(struct ixgbe_adapter *adapter)
169 {
170 	int pre_existing_vfs = 0;
171 
172 	pre_existing_vfs = pci_num_vf(adapter->pdev);
173 	if (!pre_existing_vfs && !adapter->num_vfs)
174 		return;
175 
176 	/* If there are pre-existing VFs then we have to force
177 	 * use of that many - over ride any module parameter value.
178 	 * This may result from the user unloading the PF driver
179 	 * while VFs were assigned to guest VMs or because the VFs
180 	 * have been created via the new PCI SR-IOV sysfs interface.
181 	 */
182 	if (pre_existing_vfs) {
183 		adapter->num_vfs = pre_existing_vfs;
184 		dev_warn(&adapter->pdev->dev,
185 			 "Virtual Functions already enabled for this device - Please reload all VF drivers to avoid spoofed packet errors\n");
186 	} else {
187 		int err;
188 		/*
189 		 * The 82599 supports up to 64 VFs per physical function
190 		 * but this implementation limits allocation to 63 so that
191 		 * basic networking resources are still available to the
192 		 * physical function.  If the user requests greater than
193 		 * 63 VFs then it is an error - reset to default of zero.
194 		 */
195 		adapter->num_vfs = min_t(unsigned int, adapter->num_vfs, IXGBE_MAX_VFS_DRV_LIMIT);
196 
197 		err = pci_enable_sriov(adapter->pdev, adapter->num_vfs);
198 		if (err) {
199 			e_err(probe, "Failed to enable PCI sriov: %d\n", err);
200 			adapter->num_vfs = 0;
201 			return;
202 		}
203 	}
204 
205 	if (!__ixgbe_enable_sriov(adapter)) {
206 		ixgbe_get_vfs(adapter);
207 		return;
208 	}
209 
210 	/* If we have gotten to this point then there is no memory available
211 	 * to manage the VF devices - print message and bail.
212 	 */
213 	e_err(probe, "Unable to allocate memory for VF Data Storage - "
214 	      "SRIOV disabled\n");
215 	ixgbe_disable_sriov(adapter);
216 }
217 
218 #endif /* #ifdef CONFIG_PCI_IOV */
219 int ixgbe_disable_sriov(struct ixgbe_adapter *adapter)
220 {
221 	unsigned int num_vfs = adapter->num_vfs, vf;
222 	struct ixgbe_hw *hw = &adapter->hw;
223 	u32 gpie;
224 	u32 vmdctl;
225 	int rss;
226 
227 	/* set num VFs to 0 to prevent access to vfinfo */
228 	adapter->num_vfs = 0;
229 
230 	/* put the reference to all of the vf devices */
231 	for (vf = 0; vf < num_vfs; ++vf) {
232 		struct pci_dev *vfdev = adapter->vfinfo[vf].vfdev;
233 
234 		if (!vfdev)
235 			continue;
236 		adapter->vfinfo[vf].vfdev = NULL;
237 		pci_dev_put(vfdev);
238 	}
239 
240 	/* free VF control structures */
241 	kfree(adapter->vfinfo);
242 	adapter->vfinfo = NULL;
243 
244 	/* free macvlan list */
245 	kfree(adapter->mv_list);
246 	adapter->mv_list = NULL;
247 
248 	/* if SR-IOV is already disabled then there is nothing to do */
249 	if (!(adapter->flags & IXGBE_FLAG_SRIOV_ENABLED))
250 		return 0;
251 
252 #ifdef CONFIG_PCI_IOV
253 	/*
254 	 * If our VFs are assigned we cannot shut down SR-IOV
255 	 * without causing issues, so just leave the hardware
256 	 * available but disabled
257 	 */
258 	if (pci_vfs_assigned(adapter->pdev)) {
259 		e_dev_warn("Unloading driver while VFs are assigned - VFs will not be deallocated\n");
260 		return -EPERM;
261 	}
262 	/* disable iov and allow time for transactions to clear */
263 	pci_disable_sriov(adapter->pdev);
264 #endif
265 
266 	/* turn off device IOV mode */
267 	IXGBE_WRITE_REG(hw, IXGBE_GCR_EXT, 0);
268 	gpie = IXGBE_READ_REG(hw, IXGBE_GPIE);
269 	gpie &= ~IXGBE_GPIE_VTMODE_MASK;
270 	IXGBE_WRITE_REG(hw, IXGBE_GPIE, gpie);
271 
272 	/* set default pool back to 0 */
273 	vmdctl = IXGBE_READ_REG(hw, IXGBE_VT_CTL);
274 	vmdctl &= ~IXGBE_VT_CTL_POOL_MASK;
275 	IXGBE_WRITE_REG(hw, IXGBE_VT_CTL, vmdctl);
276 	IXGBE_WRITE_FLUSH(hw);
277 
278 	/* Disable VMDq flag so device will be set in VM mode */
279 	if (adapter->ring_feature[RING_F_VMDQ].limit == 1) {
280 		adapter->flags &= ~IXGBE_FLAG_VMDQ_ENABLED;
281 		adapter->flags &= ~IXGBE_FLAG_SRIOV_ENABLED;
282 		rss = min_t(int, ixgbe_max_rss_indices(adapter),
283 			    num_online_cpus());
284 	} else {
285 		rss = min_t(int, IXGBE_MAX_L2A_QUEUES, num_online_cpus());
286 	}
287 
288 	adapter->ring_feature[RING_F_VMDQ].offset = 0;
289 	adapter->ring_feature[RING_F_RSS].limit = rss;
290 
291 	/* take a breather then clean up driver data */
292 	msleep(100);
293 	return 0;
294 }
295 
296 static int ixgbe_pci_sriov_enable(struct pci_dev *dev, int num_vfs)
297 {
298 #ifdef CONFIG_PCI_IOV
299 	struct ixgbe_adapter *adapter = pci_get_drvdata(dev);
300 	int err = 0;
301 	int i;
302 	int pre_existing_vfs = pci_num_vf(dev);
303 
304 	if (pre_existing_vfs && pre_existing_vfs != num_vfs)
305 		err = ixgbe_disable_sriov(adapter);
306 	else if (pre_existing_vfs && pre_existing_vfs == num_vfs)
307 		return num_vfs;
308 
309 	if (err)
310 		return err;
311 
312 	/* While the SR-IOV capability structure reports total VFs to be 64,
313 	 * we have to limit the actual number allocated based on two factors.
314 	 * First, we reserve some transmit/receive resources for the PF.
315 	 * Second, VMDQ also uses the same pools that SR-IOV does. We need to
316 	 * account for this, so that we don't accidentally allocate more VFs
317 	 * than we have available pools. The PCI bus driver already checks for
318 	 * other values out of range.
319 	 */
320 	if ((num_vfs + adapter->num_rx_pools) > IXGBE_MAX_VF_FUNCTIONS)
321 		return -EPERM;
322 
323 	adapter->num_vfs = num_vfs;
324 
325 	err = __ixgbe_enable_sriov(adapter);
326 	if (err)
327 		return  err;
328 
329 	for (i = 0; i < adapter->num_vfs; i++)
330 		ixgbe_vf_configuration(dev, (i | 0x10000000));
331 
332 	err = pci_enable_sriov(dev, num_vfs);
333 	if (err) {
334 		e_dev_warn("Failed to enable PCI sriov: %d\n", err);
335 		return err;
336 	}
337 	ixgbe_get_vfs(adapter);
338 	ixgbe_sriov_reinit(adapter);
339 
340 	return num_vfs;
341 #else
342 	return 0;
343 #endif
344 }
345 
346 static int ixgbe_pci_sriov_disable(struct pci_dev *dev)
347 {
348 	struct ixgbe_adapter *adapter = pci_get_drvdata(dev);
349 	int err;
350 #ifdef CONFIG_PCI_IOV
351 	u32 current_flags = adapter->flags;
352 #endif
353 
354 	err = ixgbe_disable_sriov(adapter);
355 
356 	/* Only reinit if no error and state changed */
357 #ifdef CONFIG_PCI_IOV
358 	if (!err && current_flags != adapter->flags)
359 		ixgbe_sriov_reinit(adapter);
360 #endif
361 
362 	return err;
363 }
364 
365 int ixgbe_pci_sriov_configure(struct pci_dev *dev, int num_vfs)
366 {
367 	if (num_vfs == 0)
368 		return ixgbe_pci_sriov_disable(dev);
369 	else
370 		return ixgbe_pci_sriov_enable(dev, num_vfs);
371 }
372 
373 static int ixgbe_set_vf_multicasts(struct ixgbe_adapter *adapter,
374 				   u32 *msgbuf, u32 vf)
375 {
376 	int entries = (msgbuf[0] & IXGBE_VT_MSGINFO_MASK)
377 		       >> IXGBE_VT_MSGINFO_SHIFT;
378 	u16 *hash_list = (u16 *)&msgbuf[1];
379 	struct vf_data_storage *vfinfo = &adapter->vfinfo[vf];
380 	struct ixgbe_hw *hw = &adapter->hw;
381 	int i;
382 	u32 vector_bit;
383 	u32 vector_reg;
384 	u32 mta_reg;
385 	u32 vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(vf));
386 
387 	/* only so many hash values supported */
388 	entries = min(entries, IXGBE_MAX_VF_MC_ENTRIES);
389 
390 	/*
391 	 * salt away the number of multi cast addresses assigned
392 	 * to this VF for later use to restore when the PF multi cast
393 	 * list changes
394 	 */
395 	vfinfo->num_vf_mc_hashes = entries;
396 
397 	/*
398 	 * VFs are limited to using the MTA hash table for their multicast
399 	 * addresses
400 	 */
401 	for (i = 0; i < entries; i++) {
402 		vfinfo->vf_mc_hashes[i] = hash_list[i];
403 	}
404 
405 	for (i = 0; i < vfinfo->num_vf_mc_hashes; i++) {
406 		vector_reg = (vfinfo->vf_mc_hashes[i] >> 5) & 0x7F;
407 		vector_bit = vfinfo->vf_mc_hashes[i] & 0x1F;
408 		mta_reg = IXGBE_READ_REG(hw, IXGBE_MTA(vector_reg));
409 		mta_reg |= BIT(vector_bit);
410 		IXGBE_WRITE_REG(hw, IXGBE_MTA(vector_reg), mta_reg);
411 	}
412 	vmolr |= IXGBE_VMOLR_ROMPE;
413 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(vf), vmolr);
414 
415 	return 0;
416 }
417 
418 #ifdef CONFIG_PCI_IOV
419 void ixgbe_restore_vf_multicasts(struct ixgbe_adapter *adapter)
420 {
421 	struct ixgbe_hw *hw = &adapter->hw;
422 	struct vf_data_storage *vfinfo;
423 	int i, j;
424 	u32 vector_bit;
425 	u32 vector_reg;
426 	u32 mta_reg;
427 
428 	for (i = 0; i < adapter->num_vfs; i++) {
429 		u32 vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(i));
430 		vfinfo = &adapter->vfinfo[i];
431 		for (j = 0; j < vfinfo->num_vf_mc_hashes; j++) {
432 			hw->addr_ctrl.mta_in_use++;
433 			vector_reg = (vfinfo->vf_mc_hashes[j] >> 5) & 0x7F;
434 			vector_bit = vfinfo->vf_mc_hashes[j] & 0x1F;
435 			mta_reg = IXGBE_READ_REG(hw, IXGBE_MTA(vector_reg));
436 			mta_reg |= BIT(vector_bit);
437 			IXGBE_WRITE_REG(hw, IXGBE_MTA(vector_reg), mta_reg);
438 		}
439 
440 		if (vfinfo->num_vf_mc_hashes)
441 			vmolr |= IXGBE_VMOLR_ROMPE;
442 		else
443 			vmolr &= ~IXGBE_VMOLR_ROMPE;
444 		IXGBE_WRITE_REG(hw, IXGBE_VMOLR(i), vmolr);
445 	}
446 
447 	/* Restore any VF macvlans */
448 	ixgbe_full_sync_mac_table(adapter);
449 }
450 #endif
451 
452 static int ixgbe_set_vf_vlan(struct ixgbe_adapter *adapter, int add, int vid,
453 			     u32 vf)
454 {
455 	struct ixgbe_hw *hw = &adapter->hw;
456 	int err;
457 
458 	/* If VLAN overlaps with one the PF is currently monitoring make
459 	 * sure that we are able to allocate a VLVF entry.  This may be
460 	 * redundant but it guarantees PF will maintain visibility to
461 	 * the VLAN.
462 	 */
463 	if (add && test_bit(vid, adapter->active_vlans)) {
464 		err = hw->mac.ops.set_vfta(hw, vid, VMDQ_P(0), true, false);
465 		if (err)
466 			return err;
467 	}
468 
469 	err = hw->mac.ops.set_vfta(hw, vid, vf, !!add, false);
470 
471 	if (add && !err)
472 		return err;
473 
474 	/* If we failed to add the VF VLAN or we are removing the VF VLAN
475 	 * we may need to drop the PF pool bit in order to allow us to free
476 	 * up the VLVF resources.
477 	 */
478 	if (test_bit(vid, adapter->active_vlans) ||
479 	    (adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
480 		ixgbe_update_pf_promisc_vlvf(adapter, vid);
481 
482 	return err;
483 }
484 
485 static s32 ixgbe_set_vf_lpe(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
486 {
487 	struct ixgbe_hw *hw = &adapter->hw;
488 	int max_frame = msgbuf[1];
489 	u32 max_frs;
490 
491 	/*
492 	 * For 82599EB we have to keep all PFs and VFs operating with
493 	 * the same max_frame value in order to avoid sending an oversize
494 	 * frame to a VF.  In order to guarantee this is handled correctly
495 	 * for all cases we have several special exceptions to take into
496 	 * account before we can enable the VF for receive
497 	 */
498 	if (adapter->hw.mac.type == ixgbe_mac_82599EB) {
499 		struct net_device *dev = adapter->netdev;
500 		int pf_max_frame = dev->mtu + ETH_HLEN;
501 		u32 reg_offset, vf_shift, vfre;
502 		s32 err = 0;
503 
504 #ifdef CONFIG_FCOE
505 		if (dev->features & NETIF_F_FCOE_MTU)
506 			pf_max_frame = max_t(int, pf_max_frame,
507 					     IXGBE_FCOE_JUMBO_FRAME_SIZE);
508 
509 #endif /* CONFIG_FCOE */
510 		switch (adapter->vfinfo[vf].vf_api) {
511 		case ixgbe_mbox_api_11:
512 		case ixgbe_mbox_api_12:
513 			/*
514 			 * Version 1.1 supports jumbo frames on VFs if PF has
515 			 * jumbo frames enabled which means legacy VFs are
516 			 * disabled
517 			 */
518 			if (pf_max_frame > ETH_FRAME_LEN)
519 				break;
520 		default:
521 			/*
522 			 * If the PF or VF are running w/ jumbo frames enabled
523 			 * we need to shut down the VF Rx path as we cannot
524 			 * support jumbo frames on legacy VFs
525 			 */
526 			if ((pf_max_frame > ETH_FRAME_LEN) ||
527 			    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
528 				err = -EINVAL;
529 			break;
530 		}
531 
532 		/* determine VF receive enable location */
533 		vf_shift = vf % 32;
534 		reg_offset = vf / 32;
535 
536 		/* enable or disable receive depending on error */
537 		vfre = IXGBE_READ_REG(hw, IXGBE_VFRE(reg_offset));
538 		if (err)
539 			vfre &= ~BIT(vf_shift);
540 		else
541 			vfre |= BIT(vf_shift);
542 		IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), vfre);
543 
544 		if (err) {
545 			e_err(drv, "VF max_frame %d out of range\n", max_frame);
546 			return err;
547 		}
548 	}
549 
550 	/* MTU < 68 is an error and causes problems on some kernels */
551 	if (max_frame > IXGBE_MAX_JUMBO_FRAME_SIZE) {
552 		e_err(drv, "VF max_frame %d out of range\n", max_frame);
553 		return -EINVAL;
554 	}
555 
556 	/* pull current max frame size from hardware */
557 	max_frs = IXGBE_READ_REG(hw, IXGBE_MAXFRS);
558 	max_frs &= IXGBE_MHADD_MFS_MASK;
559 	max_frs >>= IXGBE_MHADD_MFS_SHIFT;
560 
561 	if (max_frs < max_frame) {
562 		max_frs = max_frame << IXGBE_MHADD_MFS_SHIFT;
563 		IXGBE_WRITE_REG(hw, IXGBE_MAXFRS, max_frs);
564 	}
565 
566 	e_info(hw, "VF requests change max MTU to %d\n", max_frame);
567 
568 	return 0;
569 }
570 
571 static void ixgbe_set_vmolr(struct ixgbe_hw *hw, u32 vf, bool aupe)
572 {
573 	u32 vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(vf));
574 	vmolr |= IXGBE_VMOLR_BAM;
575 	if (aupe)
576 		vmolr |= IXGBE_VMOLR_AUPE;
577 	else
578 		vmolr &= ~IXGBE_VMOLR_AUPE;
579 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(vf), vmolr);
580 }
581 
582 static void ixgbe_clear_vmvir(struct ixgbe_adapter *adapter, u32 vf)
583 {
584 	struct ixgbe_hw *hw = &adapter->hw;
585 
586 	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(vf), 0);
587 }
588 
589 static void ixgbe_clear_vf_vlans(struct ixgbe_adapter *adapter, u32 vf)
590 {
591 	struct ixgbe_hw *hw = &adapter->hw;
592 	u32 vlvfb_mask, pool_mask, i;
593 
594 	/* create mask for VF and other pools */
595 	pool_mask = ~BIT(VMDQ_P(0) % 32);
596 	vlvfb_mask = BIT(vf % 32);
597 
598 	/* post increment loop, covers VLVF_ENTRIES - 1 to 0 */
599 	for (i = IXGBE_VLVF_ENTRIES; i--;) {
600 		u32 bits[2], vlvfb, vid, vfta, vlvf;
601 		u32 word = i * 2 + vf / 32;
602 		u32 mask;
603 
604 		vlvfb = IXGBE_READ_REG(hw, IXGBE_VLVFB(word));
605 
606 		/* if our bit isn't set we can skip it */
607 		if (!(vlvfb & vlvfb_mask))
608 			continue;
609 
610 		/* clear our bit from vlvfb */
611 		vlvfb ^= vlvfb_mask;
612 
613 		/* create 64b mask to chedk to see if we should clear VLVF */
614 		bits[word % 2] = vlvfb;
615 		bits[~word % 2] = IXGBE_READ_REG(hw, IXGBE_VLVFB(word ^ 1));
616 
617 		/* if other pools are present, just remove ourselves */
618 		if (bits[(VMDQ_P(0) / 32) ^ 1] ||
619 		    (bits[VMDQ_P(0) / 32] & pool_mask))
620 			goto update_vlvfb;
621 
622 		/* if PF is present, leave VFTA */
623 		if (bits[0] || bits[1])
624 			goto update_vlvf;
625 
626 		/* if we cannot determine VLAN just remove ourselves */
627 		vlvf = IXGBE_READ_REG(hw, IXGBE_VLVF(i));
628 		if (!vlvf)
629 			goto update_vlvfb;
630 
631 		vid = vlvf & VLAN_VID_MASK;
632 		mask = BIT(vid % 32);
633 
634 		/* clear bit from VFTA */
635 		vfta = IXGBE_READ_REG(hw, IXGBE_VFTA(vid / 32));
636 		if (vfta & mask)
637 			IXGBE_WRITE_REG(hw, IXGBE_VFTA(vid / 32), vfta ^ mask);
638 update_vlvf:
639 		/* clear POOL selection enable */
640 		IXGBE_WRITE_REG(hw, IXGBE_VLVF(i), 0);
641 
642 		if (!(adapter->flags2 & IXGBE_FLAG2_VLAN_PROMISC))
643 			vlvfb = 0;
644 update_vlvfb:
645 		/* clear pool bits */
646 		IXGBE_WRITE_REG(hw, IXGBE_VLVFB(word), vlvfb);
647 	}
648 }
649 
650 static inline void ixgbe_vf_reset_event(struct ixgbe_adapter *adapter, u32 vf)
651 {
652 	struct ixgbe_hw *hw = &adapter->hw;
653 	struct vf_data_storage *vfinfo = &adapter->vfinfo[vf];
654 	u8 num_tcs = netdev_get_num_tc(adapter->netdev);
655 
656 	/* remove VLAN filters beloning to this VF */
657 	ixgbe_clear_vf_vlans(adapter, vf);
658 
659 	/* add back PF assigned VLAN or VLAN 0 */
660 	ixgbe_set_vf_vlan(adapter, true, vfinfo->pf_vlan, vf);
661 
662 	/* reset offloads to defaults */
663 	ixgbe_set_vmolr(hw, vf, !vfinfo->pf_vlan);
664 
665 	/* set outgoing tags for VFs */
666 	if (!vfinfo->pf_vlan && !vfinfo->pf_qos && !num_tcs) {
667 		ixgbe_clear_vmvir(adapter, vf);
668 	} else {
669 		if (vfinfo->pf_qos || !num_tcs)
670 			ixgbe_set_vmvir(adapter, vfinfo->pf_vlan,
671 					vfinfo->pf_qos, vf);
672 		else
673 			ixgbe_set_vmvir(adapter, vfinfo->pf_vlan,
674 					adapter->default_up, vf);
675 
676 		if (vfinfo->spoofchk_enabled)
677 			hw->mac.ops.set_vlan_anti_spoofing(hw, true, vf);
678 	}
679 
680 	/* reset multicast table array for vf */
681 	adapter->vfinfo[vf].num_vf_mc_hashes = 0;
682 
683 	/* Flush and reset the mta with the new values */
684 	ixgbe_set_rx_mode(adapter->netdev);
685 
686 	ixgbe_del_mac_filter(adapter, adapter->vfinfo[vf].vf_mac_addresses, vf);
687 
688 	/* reset VF api back to unknown */
689 	adapter->vfinfo[vf].vf_api = ixgbe_mbox_api_10;
690 }
691 
692 static int ixgbe_set_vf_mac(struct ixgbe_adapter *adapter,
693 			    int vf, unsigned char *mac_addr)
694 {
695 	ixgbe_del_mac_filter(adapter, adapter->vfinfo[vf].vf_mac_addresses, vf);
696 	memcpy(adapter->vfinfo[vf].vf_mac_addresses, mac_addr, ETH_ALEN);
697 	ixgbe_add_mac_filter(adapter, adapter->vfinfo[vf].vf_mac_addresses, vf);
698 
699 	return 0;
700 }
701 
702 static int ixgbe_set_vf_macvlan(struct ixgbe_adapter *adapter,
703 				int vf, int index, unsigned char *mac_addr)
704 {
705 	struct list_head *pos;
706 	struct vf_macvlans *entry;
707 
708 	if (index <= 1) {
709 		list_for_each(pos, &adapter->vf_mvs.l) {
710 			entry = list_entry(pos, struct vf_macvlans, l);
711 			if (entry->vf == vf) {
712 				entry->vf = -1;
713 				entry->free = true;
714 				entry->is_macvlan = false;
715 				ixgbe_del_mac_filter(adapter,
716 						     entry->vf_macvlan, vf);
717 			}
718 		}
719 	}
720 
721 	/*
722 	 * If index was zero then we were asked to clear the uc list
723 	 * for the VF.  We're done.
724 	 */
725 	if (!index)
726 		return 0;
727 
728 	entry = NULL;
729 
730 	list_for_each(pos, &adapter->vf_mvs.l) {
731 		entry = list_entry(pos, struct vf_macvlans, l);
732 		if (entry->free)
733 			break;
734 	}
735 
736 	/*
737 	 * If we traversed the entire list and didn't find a free entry
738 	 * then we're out of space on the RAR table.  Also entry may
739 	 * be NULL because the original memory allocation for the list
740 	 * failed, which is not fatal but does mean we can't support
741 	 * VF requests for MACVLAN because we couldn't allocate
742 	 * memory for the list management required.
743 	 */
744 	if (!entry || !entry->free)
745 		return -ENOSPC;
746 
747 	entry->free = false;
748 	entry->is_macvlan = true;
749 	entry->vf = vf;
750 	memcpy(entry->vf_macvlan, mac_addr, ETH_ALEN);
751 
752 	ixgbe_add_mac_filter(adapter, mac_addr, vf);
753 
754 	return 0;
755 }
756 
757 int ixgbe_vf_configuration(struct pci_dev *pdev, unsigned int event_mask)
758 {
759 	struct ixgbe_adapter *adapter = pci_get_drvdata(pdev);
760 	unsigned int vfn = (event_mask & 0x3f);
761 
762 	bool enable = ((event_mask & 0x10000000U) != 0);
763 
764 	if (enable)
765 		eth_zero_addr(adapter->vfinfo[vfn].vf_mac_addresses);
766 
767 	return 0;
768 }
769 
770 static inline void ixgbe_write_qde(struct ixgbe_adapter *adapter, u32 vf,
771 				   u32 qde)
772 {
773 	struct ixgbe_hw *hw = &adapter->hw;
774 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
775 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
776 	int i;
777 
778 	for (i = vf * q_per_pool; i < ((vf + 1) * q_per_pool); i++) {
779 		u32 reg;
780 
781 		/* flush previous write */
782 		IXGBE_WRITE_FLUSH(hw);
783 
784 		/* indicate to hardware that we want to set drop enable */
785 		reg = IXGBE_QDE_WRITE | IXGBE_QDE_ENABLE;
786 		reg |= i <<  IXGBE_QDE_IDX_SHIFT;
787 		IXGBE_WRITE_REG(hw, IXGBE_QDE, reg);
788 	}
789 }
790 
791 static int ixgbe_vf_reset_msg(struct ixgbe_adapter *adapter, u32 vf)
792 {
793 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
794 	struct ixgbe_hw *hw = &adapter->hw;
795 	unsigned char *vf_mac = adapter->vfinfo[vf].vf_mac_addresses;
796 	u32 reg, reg_offset, vf_shift;
797 	u32 msgbuf[4] = {0, 0, 0, 0};
798 	u8 *addr = (u8 *)(&msgbuf[1]);
799 	u32 q_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
800 	int i;
801 
802 	e_info(probe, "VF Reset msg received from vf %d\n", vf);
803 
804 	/* reset the filters for the device */
805 	ixgbe_vf_reset_event(adapter, vf);
806 
807 	/* set vf mac address */
808 	if (!is_zero_ether_addr(vf_mac))
809 		ixgbe_set_vf_mac(adapter, vf, vf_mac);
810 
811 	vf_shift = vf % 32;
812 	reg_offset = vf / 32;
813 
814 	/* enable transmit for vf */
815 	reg = IXGBE_READ_REG(hw, IXGBE_VFTE(reg_offset));
816 	reg |= BIT(vf_shift);
817 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), reg);
818 
819 	/* force drop enable for all VF Rx queues */
820 	ixgbe_write_qde(adapter, vf, IXGBE_QDE_ENABLE);
821 
822 	/* enable receive for vf */
823 	reg = IXGBE_READ_REG(hw, IXGBE_VFRE(reg_offset));
824 	reg |= BIT(vf_shift);
825 	/*
826 	 * The 82599 cannot support a mix of jumbo and non-jumbo PF/VFs.
827 	 * For more info take a look at ixgbe_set_vf_lpe
828 	 */
829 	if (adapter->hw.mac.type == ixgbe_mac_82599EB) {
830 		struct net_device *dev = adapter->netdev;
831 		int pf_max_frame = dev->mtu + ETH_HLEN;
832 
833 #ifdef CONFIG_FCOE
834 		if (dev->features & NETIF_F_FCOE_MTU)
835 			pf_max_frame = max_t(int, pf_max_frame,
836 					     IXGBE_FCOE_JUMBO_FRAME_SIZE);
837 
838 #endif /* CONFIG_FCOE */
839 		if (pf_max_frame > ETH_FRAME_LEN)
840 			reg &= ~BIT(vf_shift);
841 	}
842 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), reg);
843 
844 	/* enable VF mailbox for further messages */
845 	adapter->vfinfo[vf].clear_to_send = true;
846 
847 	/* Enable counting of spoofed packets in the SSVPC register */
848 	reg = IXGBE_READ_REG(hw, IXGBE_VMECM(reg_offset));
849 	reg |= BIT(vf_shift);
850 	IXGBE_WRITE_REG(hw, IXGBE_VMECM(reg_offset), reg);
851 
852 	/*
853 	 * Reset the VFs TDWBAL and TDWBAH registers
854 	 * which are not cleared by an FLR
855 	 */
856 	for (i = 0; i < q_per_pool; i++) {
857 		IXGBE_WRITE_REG(hw, IXGBE_PVFTDWBAHn(q_per_pool, vf, i), 0);
858 		IXGBE_WRITE_REG(hw, IXGBE_PVFTDWBALn(q_per_pool, vf, i), 0);
859 	}
860 
861 	/* reply to reset with ack and vf mac address */
862 	msgbuf[0] = IXGBE_VF_RESET;
863 	if (!is_zero_ether_addr(vf_mac)) {
864 		msgbuf[0] |= IXGBE_VT_MSGTYPE_ACK;
865 		memcpy(addr, vf_mac, ETH_ALEN);
866 	} else {
867 		msgbuf[0] |= IXGBE_VT_MSGTYPE_NACK;
868 		dev_warn(&adapter->pdev->dev,
869 			 "VF %d has no MAC address assigned, you may have to assign one manually\n",
870 			 vf);
871 	}
872 
873 	/*
874 	 * Piggyback the multicast filter type so VF can compute the
875 	 * correct vectors
876 	 */
877 	msgbuf[3] = hw->mac.mc_filter_type;
878 	ixgbe_write_mbx(hw, msgbuf, IXGBE_VF_PERMADDR_MSG_LEN, vf);
879 
880 	return 0;
881 }
882 
883 static int ixgbe_set_vf_mac_addr(struct ixgbe_adapter *adapter,
884 				 u32 *msgbuf, u32 vf)
885 {
886 	u8 *new_mac = ((u8 *)(&msgbuf[1]));
887 
888 	if (!is_valid_ether_addr(new_mac)) {
889 		e_warn(drv, "VF %d attempted to set invalid mac\n", vf);
890 		return -1;
891 	}
892 
893 	if (adapter->vfinfo[vf].pf_set_mac && !adapter->vfinfo[vf].trusted &&
894 	    !ether_addr_equal(adapter->vfinfo[vf].vf_mac_addresses, new_mac)) {
895 		e_warn(drv,
896 		       "VF %d attempted to override administratively set MAC address\n"
897 		       "Reload the VF driver to resume operations\n",
898 		       vf);
899 		return -1;
900 	}
901 
902 	return ixgbe_set_vf_mac(adapter, vf, new_mac) < 0;
903 }
904 
905 static int ixgbe_set_vf_vlan_msg(struct ixgbe_adapter *adapter,
906 				 u32 *msgbuf, u32 vf)
907 {
908 	u32 add = (msgbuf[0] & IXGBE_VT_MSGINFO_MASK) >> IXGBE_VT_MSGINFO_SHIFT;
909 	u32 vid = (msgbuf[1] & IXGBE_VLVF_VLANID_MASK);
910 	u8 tcs = netdev_get_num_tc(adapter->netdev);
911 
912 	if (adapter->vfinfo[vf].pf_vlan || tcs) {
913 		e_warn(drv,
914 		       "VF %d attempted to override administratively set VLAN configuration\n"
915 		       "Reload the VF driver to resume operations\n",
916 		       vf);
917 		return -1;
918 	}
919 
920 	/* VLAN 0 is a special case, don't allow it to be removed */
921 	if (!vid && !add)
922 		return 0;
923 
924 	return ixgbe_set_vf_vlan(adapter, add, vid, vf);
925 }
926 
927 static int ixgbe_set_vf_macvlan_msg(struct ixgbe_adapter *adapter,
928 				    u32 *msgbuf, u32 vf)
929 {
930 	u8 *new_mac = ((u8 *)(&msgbuf[1]));
931 	int index = (msgbuf[0] & IXGBE_VT_MSGINFO_MASK) >>
932 		    IXGBE_VT_MSGINFO_SHIFT;
933 	int err;
934 
935 	if (adapter->vfinfo[vf].pf_set_mac && index > 0) {
936 		e_warn(drv,
937 		       "VF %d requested MACVLAN filter but is administratively denied\n",
938 		       vf);
939 		return -1;
940 	}
941 
942 	/* An non-zero index indicates the VF is setting a filter */
943 	if (index) {
944 		if (!is_valid_ether_addr(new_mac)) {
945 			e_warn(drv, "VF %d attempted to set invalid mac\n", vf);
946 			return -1;
947 		}
948 
949 		/*
950 		 * If the VF is allowed to set MAC filters then turn off
951 		 * anti-spoofing to avoid false positives.
952 		 */
953 		if (adapter->vfinfo[vf].spoofchk_enabled) {
954 			struct ixgbe_hw *hw = &adapter->hw;
955 
956 			hw->mac.ops.set_mac_anti_spoofing(hw, false, vf);
957 		}
958 	}
959 
960 	err = ixgbe_set_vf_macvlan(adapter, vf, index, new_mac);
961 	if (err == -ENOSPC)
962 		e_warn(drv,
963 		       "VF %d has requested a MACVLAN filter but there is no space for it\n",
964 		       vf);
965 
966 	return err < 0;
967 }
968 
969 static int ixgbe_negotiate_vf_api(struct ixgbe_adapter *adapter,
970 				  u32 *msgbuf, u32 vf)
971 {
972 	int api = msgbuf[1];
973 
974 	switch (api) {
975 	case ixgbe_mbox_api_10:
976 	case ixgbe_mbox_api_11:
977 	case ixgbe_mbox_api_12:
978 		adapter->vfinfo[vf].vf_api = api;
979 		return 0;
980 	default:
981 		break;
982 	}
983 
984 	e_info(drv, "VF %d requested invalid api version %u\n", vf, api);
985 
986 	return -1;
987 }
988 
989 static int ixgbe_get_vf_queues(struct ixgbe_adapter *adapter,
990 			       u32 *msgbuf, u32 vf)
991 {
992 	struct net_device *dev = adapter->netdev;
993 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
994 	unsigned int default_tc = 0;
995 	u8 num_tcs = netdev_get_num_tc(dev);
996 
997 	/* verify the PF is supporting the correct APIs */
998 	switch (adapter->vfinfo[vf].vf_api) {
999 	case ixgbe_mbox_api_20:
1000 	case ixgbe_mbox_api_11:
1001 	case ixgbe_mbox_api_12:
1002 		break;
1003 	default:
1004 		return -1;
1005 	}
1006 
1007 	/* only allow 1 Tx queue for bandwidth limiting */
1008 	msgbuf[IXGBE_VF_TX_QUEUES] = __ALIGN_MASK(1, ~vmdq->mask);
1009 	msgbuf[IXGBE_VF_RX_QUEUES] = __ALIGN_MASK(1, ~vmdq->mask);
1010 
1011 	/* if TCs > 1 determine which TC belongs to default user priority */
1012 	if (num_tcs > 1)
1013 		default_tc = netdev_get_prio_tc_map(dev, adapter->default_up);
1014 
1015 	/* notify VF of need for VLAN tag stripping, and correct queue */
1016 	if (num_tcs)
1017 		msgbuf[IXGBE_VF_TRANS_VLAN] = num_tcs;
1018 	else if (adapter->vfinfo[vf].pf_vlan || adapter->vfinfo[vf].pf_qos)
1019 		msgbuf[IXGBE_VF_TRANS_VLAN] = 1;
1020 	else
1021 		msgbuf[IXGBE_VF_TRANS_VLAN] = 0;
1022 
1023 	/* notify VF of default queue */
1024 	msgbuf[IXGBE_VF_DEF_QUEUE] = default_tc;
1025 
1026 	return 0;
1027 }
1028 
1029 static int ixgbe_get_vf_reta(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
1030 {
1031 	u32 i, j;
1032 	u32 *out_buf = &msgbuf[1];
1033 	const u8 *reta = adapter->rss_indir_tbl;
1034 	u32 reta_size = ixgbe_rss_indir_tbl_entries(adapter);
1035 
1036 	/* Check if operation is permitted */
1037 	if (!adapter->vfinfo[vf].rss_query_enabled)
1038 		return -EPERM;
1039 
1040 	/* verify the PF is supporting the correct API */
1041 	if (adapter->vfinfo[vf].vf_api != ixgbe_mbox_api_12)
1042 		return -EOPNOTSUPP;
1043 
1044 	/* This mailbox command is supported (required) only for 82599 and x540
1045 	 * VFs which support up to 4 RSS queues. Therefore we will compress the
1046 	 * RETA by saving only 2 bits from each entry. This way we will be able
1047 	 * to transfer the whole RETA in a single mailbox operation.
1048 	 */
1049 	for (i = 0; i < reta_size / 16; i++) {
1050 		out_buf[i] = 0;
1051 		for (j = 0; j < 16; j++)
1052 			out_buf[i] |= (u32)(reta[16 * i + j] & 0x3) << (2 * j);
1053 	}
1054 
1055 	return 0;
1056 }
1057 
1058 static int ixgbe_get_vf_rss_key(struct ixgbe_adapter *adapter,
1059 				u32 *msgbuf, u32 vf)
1060 {
1061 	u32 *rss_key = &msgbuf[1];
1062 
1063 	/* Check if the operation is permitted */
1064 	if (!adapter->vfinfo[vf].rss_query_enabled)
1065 		return -EPERM;
1066 
1067 	/* verify the PF is supporting the correct API */
1068 	if (adapter->vfinfo[vf].vf_api != ixgbe_mbox_api_12)
1069 		return -EOPNOTSUPP;
1070 
1071 	memcpy(rss_key, adapter->rss_key, sizeof(adapter->rss_key));
1072 
1073 	return 0;
1074 }
1075 
1076 static int ixgbe_update_vf_xcast_mode(struct ixgbe_adapter *adapter,
1077 				      u32 *msgbuf, u32 vf)
1078 {
1079 	struct ixgbe_hw *hw = &adapter->hw;
1080 	int xcast_mode = msgbuf[1];
1081 	u32 vmolr, disable, enable;
1082 
1083 	/* verify the PF is supporting the correct APIs */
1084 	switch (adapter->vfinfo[vf].vf_api) {
1085 	case ixgbe_mbox_api_12:
1086 		break;
1087 	default:
1088 		return -EOPNOTSUPP;
1089 	}
1090 
1091 	if (xcast_mode > IXGBEVF_XCAST_MODE_MULTI &&
1092 	    !adapter->vfinfo[vf].trusted) {
1093 		xcast_mode = IXGBEVF_XCAST_MODE_MULTI;
1094 	}
1095 
1096 	if (adapter->vfinfo[vf].xcast_mode == xcast_mode)
1097 		goto out;
1098 
1099 	switch (xcast_mode) {
1100 	case IXGBEVF_XCAST_MODE_NONE:
1101 		disable = IXGBE_VMOLR_BAM | IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_MPE;
1102 		enable = 0;
1103 		break;
1104 	case IXGBEVF_XCAST_MODE_MULTI:
1105 		disable = IXGBE_VMOLR_MPE;
1106 		enable = IXGBE_VMOLR_BAM | IXGBE_VMOLR_ROMPE;
1107 		break;
1108 	case IXGBEVF_XCAST_MODE_ALLMULTI:
1109 		disable = 0;
1110 		enable = IXGBE_VMOLR_BAM | IXGBE_VMOLR_ROMPE | IXGBE_VMOLR_MPE;
1111 		break;
1112 	default:
1113 		return -EOPNOTSUPP;
1114 	}
1115 
1116 	vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(vf));
1117 	vmolr &= ~disable;
1118 	vmolr |= enable;
1119 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(vf), vmolr);
1120 
1121 	adapter->vfinfo[vf].xcast_mode = xcast_mode;
1122 
1123 out:
1124 	msgbuf[1] = xcast_mode;
1125 
1126 	return 0;
1127 }
1128 
1129 static int ixgbe_rcv_msg_from_vf(struct ixgbe_adapter *adapter, u32 vf)
1130 {
1131 	u32 mbx_size = IXGBE_VFMAILBOX_SIZE;
1132 	u32 msgbuf[IXGBE_VFMAILBOX_SIZE];
1133 	struct ixgbe_hw *hw = &adapter->hw;
1134 	s32 retval;
1135 
1136 	retval = ixgbe_read_mbx(hw, msgbuf, mbx_size, vf);
1137 
1138 	if (retval) {
1139 		pr_err("Error receiving message from VF\n");
1140 		return retval;
1141 	}
1142 
1143 	/* this is a message we already processed, do nothing */
1144 	if (msgbuf[0] & (IXGBE_VT_MSGTYPE_ACK | IXGBE_VT_MSGTYPE_NACK))
1145 		return 0;
1146 
1147 	/* flush the ack before we write any messages back */
1148 	IXGBE_WRITE_FLUSH(hw);
1149 
1150 	if (msgbuf[0] == IXGBE_VF_RESET)
1151 		return ixgbe_vf_reset_msg(adapter, vf);
1152 
1153 	/*
1154 	 * until the vf completes a virtual function reset it should not be
1155 	 * allowed to start any configuration.
1156 	 */
1157 	if (!adapter->vfinfo[vf].clear_to_send) {
1158 		msgbuf[0] |= IXGBE_VT_MSGTYPE_NACK;
1159 		ixgbe_write_mbx(hw, msgbuf, 1, vf);
1160 		return 0;
1161 	}
1162 
1163 	switch ((msgbuf[0] & 0xFFFF)) {
1164 	case IXGBE_VF_SET_MAC_ADDR:
1165 		retval = ixgbe_set_vf_mac_addr(adapter, msgbuf, vf);
1166 		break;
1167 	case IXGBE_VF_SET_MULTICAST:
1168 		retval = ixgbe_set_vf_multicasts(adapter, msgbuf, vf);
1169 		break;
1170 	case IXGBE_VF_SET_VLAN:
1171 		retval = ixgbe_set_vf_vlan_msg(adapter, msgbuf, vf);
1172 		break;
1173 	case IXGBE_VF_SET_LPE:
1174 		retval = ixgbe_set_vf_lpe(adapter, msgbuf, vf);
1175 		break;
1176 	case IXGBE_VF_SET_MACVLAN:
1177 		retval = ixgbe_set_vf_macvlan_msg(adapter, msgbuf, vf);
1178 		break;
1179 	case IXGBE_VF_API_NEGOTIATE:
1180 		retval = ixgbe_negotiate_vf_api(adapter, msgbuf, vf);
1181 		break;
1182 	case IXGBE_VF_GET_QUEUES:
1183 		retval = ixgbe_get_vf_queues(adapter, msgbuf, vf);
1184 		break;
1185 	case IXGBE_VF_GET_RETA:
1186 		retval = ixgbe_get_vf_reta(adapter, msgbuf, vf);
1187 		break;
1188 	case IXGBE_VF_GET_RSS_KEY:
1189 		retval = ixgbe_get_vf_rss_key(adapter, msgbuf, vf);
1190 		break;
1191 	case IXGBE_VF_UPDATE_XCAST_MODE:
1192 		retval = ixgbe_update_vf_xcast_mode(adapter, msgbuf, vf);
1193 		break;
1194 	default:
1195 		e_err(drv, "Unhandled Msg %8.8x\n", msgbuf[0]);
1196 		retval = IXGBE_ERR_MBX;
1197 		break;
1198 	}
1199 
1200 	/* notify the VF of the results of what it sent us */
1201 	if (retval)
1202 		msgbuf[0] |= IXGBE_VT_MSGTYPE_NACK;
1203 	else
1204 		msgbuf[0] |= IXGBE_VT_MSGTYPE_ACK;
1205 
1206 	msgbuf[0] |= IXGBE_VT_MSGTYPE_CTS;
1207 
1208 	ixgbe_write_mbx(hw, msgbuf, mbx_size, vf);
1209 
1210 	return retval;
1211 }
1212 
1213 static void ixgbe_rcv_ack_from_vf(struct ixgbe_adapter *adapter, u32 vf)
1214 {
1215 	struct ixgbe_hw *hw = &adapter->hw;
1216 	u32 msg = IXGBE_VT_MSGTYPE_NACK;
1217 
1218 	/* if device isn't clear to send it shouldn't be reading either */
1219 	if (!adapter->vfinfo[vf].clear_to_send)
1220 		ixgbe_write_mbx(hw, &msg, 1, vf);
1221 }
1222 
1223 void ixgbe_msg_task(struct ixgbe_adapter *adapter)
1224 {
1225 	struct ixgbe_hw *hw = &adapter->hw;
1226 	u32 vf;
1227 
1228 	for (vf = 0; vf < adapter->num_vfs; vf++) {
1229 		/* process any reset requests */
1230 		if (!ixgbe_check_for_rst(hw, vf))
1231 			ixgbe_vf_reset_event(adapter, vf);
1232 
1233 		/* process any messages pending */
1234 		if (!ixgbe_check_for_msg(hw, vf))
1235 			ixgbe_rcv_msg_from_vf(adapter, vf);
1236 
1237 		/* process any acks */
1238 		if (!ixgbe_check_for_ack(hw, vf))
1239 			ixgbe_rcv_ack_from_vf(adapter, vf);
1240 	}
1241 }
1242 
1243 void ixgbe_disable_tx_rx(struct ixgbe_adapter *adapter)
1244 {
1245 	struct ixgbe_hw *hw = &adapter->hw;
1246 
1247 	/* disable transmit and receive for all vfs */
1248 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(0), 0);
1249 	IXGBE_WRITE_REG(hw, IXGBE_VFTE(1), 0);
1250 
1251 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(0), 0);
1252 	IXGBE_WRITE_REG(hw, IXGBE_VFRE(1), 0);
1253 }
1254 
1255 static inline void ixgbe_ping_vf(struct ixgbe_adapter *adapter, int vf)
1256 {
1257 	struct ixgbe_hw *hw = &adapter->hw;
1258 	u32 ping;
1259 
1260 	ping = IXGBE_PF_CONTROL_MSG;
1261 	if (adapter->vfinfo[vf].clear_to_send)
1262 		ping |= IXGBE_VT_MSGTYPE_CTS;
1263 	ixgbe_write_mbx(hw, &ping, 1, vf);
1264 }
1265 
1266 void ixgbe_ping_all_vfs(struct ixgbe_adapter *adapter)
1267 {
1268 	struct ixgbe_hw *hw = &adapter->hw;
1269 	u32 ping;
1270 	int i;
1271 
1272 	for (i = 0 ; i < adapter->num_vfs; i++) {
1273 		ping = IXGBE_PF_CONTROL_MSG;
1274 		if (adapter->vfinfo[i].clear_to_send)
1275 			ping |= IXGBE_VT_MSGTYPE_CTS;
1276 		ixgbe_write_mbx(hw, &ping, 1, i);
1277 	}
1278 }
1279 
1280 int ixgbe_ndo_set_vf_mac(struct net_device *netdev, int vf, u8 *mac)
1281 {
1282 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1283 	if (!is_valid_ether_addr(mac) || (vf >= adapter->num_vfs))
1284 		return -EINVAL;
1285 	adapter->vfinfo[vf].pf_set_mac = true;
1286 	dev_info(&adapter->pdev->dev, "setting MAC %pM on VF %d\n", mac, vf);
1287 	dev_info(&adapter->pdev->dev, "Reload the VF driver to make this"
1288 				      " change effective.");
1289 	if (test_bit(__IXGBE_DOWN, &adapter->state)) {
1290 		dev_warn(&adapter->pdev->dev, "The VF MAC address has been set,"
1291 			 " but the PF device is not up.\n");
1292 		dev_warn(&adapter->pdev->dev, "Bring the PF device up before"
1293 			 " attempting to use the VF device.\n");
1294 	}
1295 	return ixgbe_set_vf_mac(adapter, vf, mac);
1296 }
1297 
1298 static int ixgbe_enable_port_vlan(struct ixgbe_adapter *adapter, int vf,
1299 				  u16 vlan, u8 qos)
1300 {
1301 	struct ixgbe_hw *hw = &adapter->hw;
1302 	int err;
1303 
1304 	err = ixgbe_set_vf_vlan(adapter, true, vlan, vf);
1305 	if (err)
1306 		goto out;
1307 
1308 	/* Revoke tagless access via VLAN 0 */
1309 	ixgbe_set_vf_vlan(adapter, false, 0, vf);
1310 
1311 	ixgbe_set_vmvir(adapter, vlan, qos, vf);
1312 	ixgbe_set_vmolr(hw, vf, false);
1313 
1314 	/* enable hide vlan on X550 */
1315 	if (hw->mac.type >= ixgbe_mac_X550)
1316 		ixgbe_write_qde(adapter, vf, IXGBE_QDE_ENABLE |
1317 				IXGBE_QDE_HIDE_VLAN);
1318 
1319 	adapter->vfinfo[vf].pf_vlan = vlan;
1320 	adapter->vfinfo[vf].pf_qos = qos;
1321 	dev_info(&adapter->pdev->dev,
1322 		 "Setting VLAN %d, QOS 0x%x on VF %d\n", vlan, qos, vf);
1323 	if (test_bit(__IXGBE_DOWN, &adapter->state)) {
1324 		dev_warn(&adapter->pdev->dev,
1325 			 "The VF VLAN has been set, but the PF device is not up.\n");
1326 		dev_warn(&adapter->pdev->dev,
1327 			 "Bring the PF device up before attempting to use the VF device.\n");
1328 	}
1329 
1330 out:
1331 	return err;
1332 }
1333 
1334 static int ixgbe_disable_port_vlan(struct ixgbe_adapter *adapter, int vf)
1335 {
1336 	struct ixgbe_hw *hw = &adapter->hw;
1337 	int err;
1338 
1339 	err = ixgbe_set_vf_vlan(adapter, false,
1340 				adapter->vfinfo[vf].pf_vlan, vf);
1341 	/* Restore tagless access via VLAN 0 */
1342 	ixgbe_set_vf_vlan(adapter, true, 0, vf);
1343 	ixgbe_clear_vmvir(adapter, vf);
1344 	ixgbe_set_vmolr(hw, vf, true);
1345 
1346 	/* disable hide VLAN on X550 */
1347 	if (hw->mac.type >= ixgbe_mac_X550)
1348 		ixgbe_write_qde(adapter, vf, IXGBE_QDE_ENABLE);
1349 
1350 	adapter->vfinfo[vf].pf_vlan = 0;
1351 	adapter->vfinfo[vf].pf_qos = 0;
1352 
1353 	return err;
1354 }
1355 
1356 int ixgbe_ndo_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
1357 {
1358 	int err = 0;
1359 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1360 
1361 	if ((vf >= adapter->num_vfs) || (vlan > 4095) || (qos > 7))
1362 		return -EINVAL;
1363 	if (vlan || qos) {
1364 		/* Check if there is already a port VLAN set, if so
1365 		 * we have to delete the old one first before we
1366 		 * can set the new one.  The usage model had
1367 		 * previously assumed the user would delete the
1368 		 * old port VLAN before setting a new one but this
1369 		 * is not necessarily the case.
1370 		 */
1371 		if (adapter->vfinfo[vf].pf_vlan)
1372 			err = ixgbe_disable_port_vlan(adapter, vf);
1373 		if (err)
1374 			goto out;
1375 		err = ixgbe_enable_port_vlan(adapter, vf, vlan, qos);
1376 	} else {
1377 		err = ixgbe_disable_port_vlan(adapter, vf);
1378 	}
1379 
1380 out:
1381 	return err;
1382 }
1383 
1384 int ixgbe_link_mbps(struct ixgbe_adapter *adapter)
1385 {
1386 	switch (adapter->link_speed) {
1387 	case IXGBE_LINK_SPEED_100_FULL:
1388 		return 100;
1389 	case IXGBE_LINK_SPEED_1GB_FULL:
1390 		return 1000;
1391 	case IXGBE_LINK_SPEED_10GB_FULL:
1392 		return 10000;
1393 	default:
1394 		return 0;
1395 	}
1396 }
1397 
1398 static void ixgbe_set_vf_rate_limit(struct ixgbe_adapter *adapter, int vf)
1399 {
1400 	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
1401 	struct ixgbe_hw *hw = &adapter->hw;
1402 	u32 bcnrc_val = 0;
1403 	u16 queue, queues_per_pool;
1404 	u16 tx_rate = adapter->vfinfo[vf].tx_rate;
1405 
1406 	if (tx_rate) {
1407 		/* start with base link speed value */
1408 		bcnrc_val = adapter->vf_rate_link_speed;
1409 
1410 		/* Calculate the rate factor values to set */
1411 		bcnrc_val <<= IXGBE_RTTBCNRC_RF_INT_SHIFT;
1412 		bcnrc_val /= tx_rate;
1413 
1414 		/* clear everything but the rate factor */
1415 		bcnrc_val &= IXGBE_RTTBCNRC_RF_INT_MASK |
1416 			     IXGBE_RTTBCNRC_RF_DEC_MASK;
1417 
1418 		/* enable the rate scheduler */
1419 		bcnrc_val |= IXGBE_RTTBCNRC_RS_ENA;
1420 	}
1421 
1422 	/*
1423 	 * Set global transmit compensation time to the MMW_SIZE in RTTBCNRM
1424 	 * register. Typically MMW_SIZE=0x014 if 9728-byte jumbo is supported
1425 	 * and 0x004 otherwise.
1426 	 */
1427 	switch (hw->mac.type) {
1428 	case ixgbe_mac_82599EB:
1429 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, 0x4);
1430 		break;
1431 	case ixgbe_mac_X540:
1432 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRM, 0x14);
1433 		break;
1434 	default:
1435 		break;
1436 	}
1437 
1438 	/* determine how many queues per pool based on VMDq mask */
1439 	queues_per_pool = __ALIGN_MASK(1, ~vmdq->mask);
1440 
1441 	/* write value for all Tx queues belonging to VF */
1442 	for (queue = 0; queue < queues_per_pool; queue++) {
1443 		unsigned int reg_idx = (vf * queues_per_pool) + queue;
1444 
1445 		IXGBE_WRITE_REG(hw, IXGBE_RTTDQSEL, reg_idx);
1446 		IXGBE_WRITE_REG(hw, IXGBE_RTTBCNRC, bcnrc_val);
1447 	}
1448 }
1449 
1450 void ixgbe_check_vf_rate_limit(struct ixgbe_adapter *adapter)
1451 {
1452 	int i;
1453 
1454 	/* VF Tx rate limit was not set */
1455 	if (!adapter->vf_rate_link_speed)
1456 		return;
1457 
1458 	if (ixgbe_link_mbps(adapter) != adapter->vf_rate_link_speed) {
1459 		adapter->vf_rate_link_speed = 0;
1460 		dev_info(&adapter->pdev->dev,
1461 			 "Link speed has been changed. VF Transmit rate is disabled\n");
1462 	}
1463 
1464 	for (i = 0; i < adapter->num_vfs; i++) {
1465 		if (!adapter->vf_rate_link_speed)
1466 			adapter->vfinfo[i].tx_rate = 0;
1467 
1468 		ixgbe_set_vf_rate_limit(adapter, i);
1469 	}
1470 }
1471 
1472 int ixgbe_ndo_set_vf_bw(struct net_device *netdev, int vf, int min_tx_rate,
1473 			int max_tx_rate)
1474 {
1475 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1476 	int link_speed;
1477 
1478 	/* verify VF is active */
1479 	if (vf >= adapter->num_vfs)
1480 		return -EINVAL;
1481 
1482 	/* verify link is up */
1483 	if (!adapter->link_up)
1484 		return -EINVAL;
1485 
1486 	/* verify we are linked at 10Gbps */
1487 	link_speed = ixgbe_link_mbps(adapter);
1488 	if (link_speed != 10000)
1489 		return -EINVAL;
1490 
1491 	if (min_tx_rate)
1492 		return -EINVAL;
1493 
1494 	/* rate limit cannot be less than 10Mbs or greater than link speed */
1495 	if (max_tx_rate && ((max_tx_rate <= 10) || (max_tx_rate > link_speed)))
1496 		return -EINVAL;
1497 
1498 	/* store values */
1499 	adapter->vf_rate_link_speed = link_speed;
1500 	adapter->vfinfo[vf].tx_rate = max_tx_rate;
1501 
1502 	/* update hardware configuration */
1503 	ixgbe_set_vf_rate_limit(adapter, vf);
1504 
1505 	return 0;
1506 }
1507 
1508 int ixgbe_ndo_set_vf_spoofchk(struct net_device *netdev, int vf, bool setting)
1509 {
1510 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1511 	struct ixgbe_hw *hw = &adapter->hw;
1512 
1513 	if (vf >= adapter->num_vfs)
1514 		return -EINVAL;
1515 
1516 	adapter->vfinfo[vf].spoofchk_enabled = setting;
1517 
1518 	/* configure MAC spoofing */
1519 	hw->mac.ops.set_mac_anti_spoofing(hw, setting, vf);
1520 
1521 	/* configure VLAN spoofing */
1522 	hw->mac.ops.set_vlan_anti_spoofing(hw, setting, vf);
1523 
1524 	/* Ensure LLDP and FC is set for Ethertype Antispoofing if we will be
1525 	 * calling set_ethertype_anti_spoofing for each VF in loop below
1526 	 */
1527 	if (hw->mac.ops.set_ethertype_anti_spoofing) {
1528 		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_LLDP),
1529 				(IXGBE_ETQF_FILTER_EN    |
1530 				 IXGBE_ETQF_TX_ANTISPOOF |
1531 				 IXGBE_ETH_P_LLDP));
1532 
1533 		IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_FC),
1534 				(IXGBE_ETQF_FILTER_EN |
1535 				 IXGBE_ETQF_TX_ANTISPOOF |
1536 				 ETH_P_PAUSE));
1537 
1538 		hw->mac.ops.set_ethertype_anti_spoofing(hw, setting, vf);
1539 	}
1540 
1541 	return 0;
1542 }
1543 
1544 int ixgbe_ndo_set_vf_rss_query_en(struct net_device *netdev, int vf,
1545 				  bool setting)
1546 {
1547 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1548 
1549 	/* This operation is currently supported only for 82599 and x540
1550 	 * devices.
1551 	 */
1552 	if (adapter->hw.mac.type < ixgbe_mac_82599EB ||
1553 	    adapter->hw.mac.type >= ixgbe_mac_X550)
1554 		return -EOPNOTSUPP;
1555 
1556 	if (vf >= adapter->num_vfs)
1557 		return -EINVAL;
1558 
1559 	adapter->vfinfo[vf].rss_query_enabled = setting;
1560 
1561 	return 0;
1562 }
1563 
1564 int ixgbe_ndo_set_vf_trust(struct net_device *netdev, int vf, bool setting)
1565 {
1566 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1567 
1568 	if (vf >= adapter->num_vfs)
1569 		return -EINVAL;
1570 
1571 	/* nothing to do */
1572 	if (adapter->vfinfo[vf].trusted == setting)
1573 		return 0;
1574 
1575 	adapter->vfinfo[vf].trusted = setting;
1576 
1577 	/* reset VF to reconfigure features */
1578 	adapter->vfinfo[vf].clear_to_send = false;
1579 	ixgbe_ping_vf(adapter, vf);
1580 
1581 	e_info(drv, "VF %u is %strusted\n", vf, setting ? "" : "not ");
1582 
1583 	return 0;
1584 }
1585 
1586 int ixgbe_ndo_get_vf_config(struct net_device *netdev,
1587 			    int vf, struct ifla_vf_info *ivi)
1588 {
1589 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
1590 	if (vf >= adapter->num_vfs)
1591 		return -EINVAL;
1592 	ivi->vf = vf;
1593 	memcpy(&ivi->mac, adapter->vfinfo[vf].vf_mac_addresses, ETH_ALEN);
1594 	ivi->max_tx_rate = adapter->vfinfo[vf].tx_rate;
1595 	ivi->min_tx_rate = 0;
1596 	ivi->vlan = adapter->vfinfo[vf].pf_vlan;
1597 	ivi->qos = adapter->vfinfo[vf].pf_qos;
1598 	ivi->spoofchk = adapter->vfinfo[vf].spoofchk_enabled;
1599 	ivi->rss_query_en = adapter->vfinfo[vf].rss_query_enabled;
1600 	ivi->trusted = adapter->vfinfo[vf].trusted;
1601 	return 0;
1602 }
1603