1// SPDX-License-Identifier: GPL-2.0 or BSD-3-Clause
2
3/* Authors: Bernard Metzler <bmt@zurich.ibm.com> */
4/* Copyright (c) 2008-2019, IBM Corporation */
5
6#include <linux/init.h>
7#include <linux/errno.h>
8#include <linux/netdevice.h>
9#include <linux/inetdevice.h>
10#include <net/net_namespace.h>
11#include <linux/rtnetlink.h>
12#include <linux/if_arp.h>
13#include <linux/list.h>
14#include <linux/kernel.h>
15#include <linux/sched.h>
16#include <linux/module.h>
17#include <linux/dma-mapping.h>
18
19#include <net/addrconf.h>
20#include <rdma/ib_verbs.h>
21#include <rdma/ib_user_verbs.h>
22#include <rdma/rdma_netlink.h>
23#include <linux/kthread.h>
24
25#include "siw.h"
26#include "siw_verbs.h"
27
28MODULE_AUTHOR("Bernard Metzler");
29MODULE_DESCRIPTION("Software iWARP Driver");
30MODULE_LICENSE("Dual BSD/GPL");
31
32/* transmit from user buffer, if possible */
33const bool zcopy_tx = true;
34
35/* Restrict usage of GSO, if hardware peer iwarp is unable to process
36 * large packets. try_gso = true lets siw try to use local GSO,
37 * if peer agrees.  Not using GSO severly limits siw maximum tx bandwidth.
38 */
39const bool try_gso;
40
41/* Attach siw also with loopback devices */
42const bool loopback_enabled = true;
43
44/* We try to negotiate CRC on, if true */
45const bool mpa_crc_required;
46
47/* MPA CRC on/off enforced */
48const bool mpa_crc_strict;
49
50/* Control TCP_NODELAY socket option */
51const bool siw_tcp_nagle;
52
53/* Select MPA version to be used during connection setup */
54u_char mpa_version = MPA_REVISION_2;
55
56/* Selects MPA P2P mode (additional handshake during connection
57 * setup, if true.
58 */
59const bool peer_to_peer;
60
61struct task_struct *siw_tx_thread[NR_CPUS];
62struct crypto_shash *siw_crypto_shash;
63
64static int siw_device_register(struct siw_device *sdev, const char *name)
65{
66	struct ib_device *base_dev = &sdev->base_dev;
67	static int dev_id = 1;
68	int rv;
69
70	sdev->vendor_part_id = dev_id++;
71
72	rv = ib_register_device(base_dev, name, NULL);
73	if (rv) {
74		pr_warn("siw: device registration error %d\n", rv);
75		return rv;
76	}
77
78	siw_dbg(base_dev, "HWaddr=%pM\n", sdev->netdev->dev_addr);
79
80	return 0;
81}
82
83static void siw_device_cleanup(struct ib_device *base_dev)
84{
85	struct siw_device *sdev = to_siw_dev(base_dev);
86
87	xa_destroy(&sdev->qp_xa);
88	xa_destroy(&sdev->mem_xa);
89}
90
91static int siw_create_tx_threads(void)
92{
93	int cpu, assigned = 0;
94
95	for_each_online_cpu(cpu) {
96		/* Skip HT cores */
97		if (cpu % cpumask_weight(topology_sibling_cpumask(cpu)))
98			continue;
99
100		siw_tx_thread[cpu] =
101			kthread_create(siw_run_sq, (unsigned long *)(long)cpu,
102				       "siw_tx/%d", cpu);
103		if (IS_ERR(siw_tx_thread[cpu])) {
104			siw_tx_thread[cpu] = NULL;
105			continue;
106		}
107		kthread_bind(siw_tx_thread[cpu], cpu);
108
109		wake_up_process(siw_tx_thread[cpu]);
110		assigned++;
111	}
112	return assigned;
113}
114
115static int siw_dev_qualified(struct net_device *netdev)
116{
117	/*
118	 * Additional hardware support can be added here
119	 * (e.g. ARPHRD_FDDI, ARPHRD_ATM, ...) - see
120	 * <linux/if_arp.h> for type identifiers.
121	 */
122	if (netdev->type == ARPHRD_ETHER || netdev->type == ARPHRD_IEEE802 ||
123	    (netdev->type == ARPHRD_LOOPBACK && loopback_enabled))
124		return 1;
125
126	return 0;
127}
128
129static DEFINE_PER_CPU(atomic_t, siw_use_cnt);
130
131static struct {
132	struct cpumask **tx_valid_cpus;
133	int num_nodes;
134} siw_cpu_info;
135
136static int siw_init_cpulist(void)
137{
138	int i, num_nodes = nr_node_ids;
139
140	memset(siw_tx_thread, 0, sizeof(siw_tx_thread));
141
142	siw_cpu_info.num_nodes = num_nodes;
143
144	siw_cpu_info.tx_valid_cpus =
145		kcalloc(num_nodes, sizeof(struct cpumask *), GFP_KERNEL);
146	if (!siw_cpu_info.tx_valid_cpus) {
147		siw_cpu_info.num_nodes = 0;
148		return -ENOMEM;
149	}
150	for (i = 0; i < siw_cpu_info.num_nodes; i++) {
151		siw_cpu_info.tx_valid_cpus[i] =
152			kzalloc(sizeof(struct cpumask), GFP_KERNEL);
153		if (!siw_cpu_info.tx_valid_cpus[i])
154			goto out_err;
155
156		cpumask_clear(siw_cpu_info.tx_valid_cpus[i]);
157	}
158	for_each_possible_cpu(i)
159		cpumask_set_cpu(i, siw_cpu_info.tx_valid_cpus[cpu_to_node(i)]);
160
161	return 0;
162
163out_err:
164	siw_cpu_info.num_nodes = 0;
165	while (--i >= 0)
166		kfree(siw_cpu_info.tx_valid_cpus[i]);
167	kfree(siw_cpu_info.tx_valid_cpus);
168	siw_cpu_info.tx_valid_cpus = NULL;
169
170	return -ENOMEM;
171}
172
173static void siw_destroy_cpulist(void)
174{
175	int i = 0;
176
177	while (i < siw_cpu_info.num_nodes)
178		kfree(siw_cpu_info.tx_valid_cpus[i++]);
179
180	kfree(siw_cpu_info.tx_valid_cpus);
181}
182
183/*
184 * Choose CPU with least number of active QP's from NUMA node of
185 * TX interface.
186 */
187int siw_get_tx_cpu(struct siw_device *sdev)
188{
189	const struct cpumask *tx_cpumask;
190	int i, num_cpus, cpu, min_use, node = sdev->numa_node, tx_cpu = -1;
191
192	if (node < 0)
193		tx_cpumask = cpu_online_mask;
194	else
195		tx_cpumask = siw_cpu_info.tx_valid_cpus[node];
196
197	num_cpus = cpumask_weight(tx_cpumask);
198	if (!num_cpus) {
199		/* no CPU on this NUMA node */
200		tx_cpumask = cpu_online_mask;
201		num_cpus = cpumask_weight(tx_cpumask);
202	}
203	if (!num_cpus)
204		goto out;
205
206	cpu = cpumask_first(tx_cpumask);
207
208	for (i = 0, min_use = SIW_MAX_QP; i < num_cpus;
209	     i++, cpu = cpumask_next(cpu, tx_cpumask)) {
210		int usage;
211
212		/* Skip any cores which have no TX thread */
213		if (!siw_tx_thread[cpu])
214			continue;
215
216		usage = atomic_read(&per_cpu(siw_use_cnt, cpu));
217		if (usage <= min_use) {
218			tx_cpu = cpu;
219			min_use = usage;
220		}
221	}
222	siw_dbg(&sdev->base_dev,
223		"tx cpu %d, node %d, %d qp's\n", tx_cpu, node, min_use);
224
225out:
226	if (tx_cpu >= 0)
227		atomic_inc(&per_cpu(siw_use_cnt, tx_cpu));
228	else
229		pr_warn("siw: no tx cpu found\n");
230
231	return tx_cpu;
232}
233
234void siw_put_tx_cpu(int cpu)
235{
236	atomic_dec(&per_cpu(siw_use_cnt, cpu));
237}
238
239static struct ib_qp *siw_get_base_qp(struct ib_device *base_dev, int id)
240{
241	struct siw_qp *qp = siw_qp_id2obj(to_siw_dev(base_dev), id);
242
243	if (qp) {
244		/*
245		 * siw_qp_id2obj() increments object reference count
246		 */
247		siw_qp_put(qp);
248		return &qp->base_qp;
249	}
250	return NULL;
251}
252
253static const struct ib_device_ops siw_device_ops = {
254	.owner = THIS_MODULE,
255	.uverbs_abi_ver = SIW_ABI_VERSION,
256	.driver_id = RDMA_DRIVER_SIW,
257
258	.alloc_mr = siw_alloc_mr,
259	.alloc_pd = siw_alloc_pd,
260	.alloc_ucontext = siw_alloc_ucontext,
261	.create_cq = siw_create_cq,
262	.create_qp = siw_create_qp,
263	.create_srq = siw_create_srq,
264	.dealloc_driver = siw_device_cleanup,
265	.dealloc_pd = siw_dealloc_pd,
266	.dealloc_ucontext = siw_dealloc_ucontext,
267	.dereg_mr = siw_dereg_mr,
268	.destroy_cq = siw_destroy_cq,
269	.destroy_qp = siw_destroy_qp,
270	.destroy_srq = siw_destroy_srq,
271	.get_dma_mr = siw_get_dma_mr,
272	.get_port_immutable = siw_get_port_immutable,
273	.iw_accept = siw_accept,
274	.iw_add_ref = siw_qp_get_ref,
275	.iw_connect = siw_connect,
276	.iw_create_listen = siw_create_listen,
277	.iw_destroy_listen = siw_destroy_listen,
278	.iw_get_qp = siw_get_base_qp,
279	.iw_reject = siw_reject,
280	.iw_rem_ref = siw_qp_put_ref,
281	.map_mr_sg = siw_map_mr_sg,
282	.mmap = siw_mmap,
283	.mmap_free = siw_mmap_free,
284	.modify_qp = siw_verbs_modify_qp,
285	.modify_srq = siw_modify_srq,
286	.poll_cq = siw_poll_cq,
287	.post_recv = siw_post_receive,
288	.post_send = siw_post_send,
289	.post_srq_recv = siw_post_srq_recv,
290	.query_device = siw_query_device,
291	.query_gid = siw_query_gid,
292	.query_port = siw_query_port,
293	.query_qp = siw_query_qp,
294	.query_srq = siw_query_srq,
295	.req_notify_cq = siw_req_notify_cq,
296	.reg_user_mr = siw_reg_user_mr,
297
298	INIT_RDMA_OBJ_SIZE(ib_cq, siw_cq, base_cq),
299	INIT_RDMA_OBJ_SIZE(ib_pd, siw_pd, base_pd),
300	INIT_RDMA_OBJ_SIZE(ib_srq, siw_srq, base_srq),
301	INIT_RDMA_OBJ_SIZE(ib_ucontext, siw_ucontext, base_ucontext),
302};
303
304static struct siw_device *siw_device_create(struct net_device *netdev)
305{
306	struct siw_device *sdev = NULL;
307	struct ib_device *base_dev;
308	int rv;
309
310	sdev = ib_alloc_device(siw_device, base_dev);
311	if (!sdev)
312		return NULL;
313
314	base_dev = &sdev->base_dev;
315
316	sdev->netdev = netdev;
317
318	if (netdev->type != ARPHRD_LOOPBACK) {
319		addrconf_addr_eui48((unsigned char *)&base_dev->node_guid,
320				    netdev->dev_addr);
321	} else {
322		/*
323		 * The loopback device does not have a HW address,
324		 * but connection mangagement lib expects gid != 0
325		 */
326		size_t len = min_t(size_t, strlen(base_dev->name), 6);
327		char addr[6] = { };
328
329		memcpy(addr, base_dev->name, len);
330		addrconf_addr_eui48((unsigned char *)&base_dev->node_guid,
331				    addr);
332	}
333	base_dev->uverbs_cmd_mask =
334		(1ull << IB_USER_VERBS_CMD_QUERY_DEVICE) |
335		(1ull << IB_USER_VERBS_CMD_QUERY_PORT) |
336		(1ull << IB_USER_VERBS_CMD_GET_CONTEXT) |
337		(1ull << IB_USER_VERBS_CMD_ALLOC_PD) |
338		(1ull << IB_USER_VERBS_CMD_DEALLOC_PD) |
339		(1ull << IB_USER_VERBS_CMD_REG_MR) |
340		(1ull << IB_USER_VERBS_CMD_DEREG_MR) |
341		(1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) |
342		(1ull << IB_USER_VERBS_CMD_CREATE_CQ) |
343		(1ull << IB_USER_VERBS_CMD_POLL_CQ) |
344		(1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ) |
345		(1ull << IB_USER_VERBS_CMD_DESTROY_CQ) |
346		(1ull << IB_USER_VERBS_CMD_CREATE_QP) |
347		(1ull << IB_USER_VERBS_CMD_QUERY_QP) |
348		(1ull << IB_USER_VERBS_CMD_MODIFY_QP) |
349		(1ull << IB_USER_VERBS_CMD_DESTROY_QP) |
350		(1ull << IB_USER_VERBS_CMD_POST_SEND) |
351		(1ull << IB_USER_VERBS_CMD_POST_RECV) |
352		(1ull << IB_USER_VERBS_CMD_CREATE_SRQ) |
353		(1ull << IB_USER_VERBS_CMD_POST_SRQ_RECV) |
354		(1ull << IB_USER_VERBS_CMD_MODIFY_SRQ) |
355		(1ull << IB_USER_VERBS_CMD_QUERY_SRQ) |
356		(1ull << IB_USER_VERBS_CMD_DESTROY_SRQ);
357
358	base_dev->node_type = RDMA_NODE_RNIC;
359	memcpy(base_dev->node_desc, SIW_NODE_DESC_COMMON,
360	       sizeof(SIW_NODE_DESC_COMMON));
361
362	/*
363	 * Current model (one-to-one device association):
364	 * One Softiwarp device per net_device or, equivalently,
365	 * per physical port.
366	 */
367	base_dev->phys_port_cnt = 1;
368	base_dev->num_comp_vectors = num_possible_cpus();
369
370	xa_init_flags(&sdev->qp_xa, XA_FLAGS_ALLOC1);
371	xa_init_flags(&sdev->mem_xa, XA_FLAGS_ALLOC1);
372
373	ib_set_device_ops(base_dev, &siw_device_ops);
374	rv = ib_device_set_netdev(base_dev, netdev, 1);
375	if (rv)
376		goto error;
377
378	memcpy(base_dev->iw_ifname, netdev->name,
379	       sizeof(base_dev->iw_ifname));
380
381	/* Disable TCP port mapping */
382	base_dev->iw_driver_flags = IW_F_NO_PORT_MAP,
383
384	sdev->attrs.max_qp = SIW_MAX_QP;
385	sdev->attrs.max_qp_wr = SIW_MAX_QP_WR;
386	sdev->attrs.max_ord = SIW_MAX_ORD_QP;
387	sdev->attrs.max_ird = SIW_MAX_IRD_QP;
388	sdev->attrs.max_sge = SIW_MAX_SGE;
389	sdev->attrs.max_sge_rd = SIW_MAX_SGE_RD;
390	sdev->attrs.max_cq = SIW_MAX_CQ;
391	sdev->attrs.max_cqe = SIW_MAX_CQE;
392	sdev->attrs.max_mr = SIW_MAX_MR;
393	sdev->attrs.max_pd = SIW_MAX_PD;
394	sdev->attrs.max_mw = SIW_MAX_MW;
395	sdev->attrs.max_srq = SIW_MAX_SRQ;
396	sdev->attrs.max_srq_wr = SIW_MAX_SRQ_WR;
397	sdev->attrs.max_srq_sge = SIW_MAX_SGE;
398
399	INIT_LIST_HEAD(&sdev->cep_list);
400	INIT_LIST_HEAD(&sdev->qp_list);
401
402	atomic_set(&sdev->num_ctx, 0);
403	atomic_set(&sdev->num_srq, 0);
404	atomic_set(&sdev->num_qp, 0);
405	atomic_set(&sdev->num_cq, 0);
406	atomic_set(&sdev->num_mr, 0);
407	atomic_set(&sdev->num_pd, 0);
408
409	sdev->numa_node = dev_to_node(&netdev->dev);
410	spin_lock_init(&sdev->lock);
411
412	return sdev;
413error:
414	ib_dealloc_device(base_dev);
415
416	return NULL;
417}
418
419/*
420 * Network link becomes unavailable. Mark all
421 * affected QP's accordingly.
422 */
423static void siw_netdev_down(struct work_struct *work)
424{
425	struct siw_device *sdev =
426		container_of(work, struct siw_device, netdev_down);
427
428	struct siw_qp_attrs qp_attrs;
429	struct list_head *pos, *tmp;
430
431	memset(&qp_attrs, 0, sizeof(qp_attrs));
432	qp_attrs.state = SIW_QP_STATE_ERROR;
433
434	list_for_each_safe(pos, tmp, &sdev->qp_list) {
435		struct siw_qp *qp = list_entry(pos, struct siw_qp, devq);
436
437		down_write(&qp->state_lock);
438		WARN_ON(siw_qp_modify(qp, &qp_attrs, SIW_QP_ATTR_STATE));
439		up_write(&qp->state_lock);
440	}
441	ib_device_put(&sdev->base_dev);
442}
443
444static void siw_device_goes_down(struct siw_device *sdev)
445{
446	if (ib_device_try_get(&sdev->base_dev)) {
447		INIT_WORK(&sdev->netdev_down, siw_netdev_down);
448		schedule_work(&sdev->netdev_down);
449	}
450}
451
452static int siw_netdev_event(struct notifier_block *nb, unsigned long event,
453			    void *arg)
454{
455	struct net_device *netdev = netdev_notifier_info_to_dev(arg);
456	struct ib_device *base_dev;
457	struct siw_device *sdev;
458
459	dev_dbg(&netdev->dev, "siw: event %lu\n", event);
460
461	base_dev = ib_device_get_by_netdev(netdev, RDMA_DRIVER_SIW);
462	if (!base_dev)
463		return NOTIFY_OK;
464
465	sdev = to_siw_dev(base_dev);
466
467	switch (event) {
468	case NETDEV_UP:
469		sdev->state = IB_PORT_ACTIVE;
470		siw_port_event(sdev, 1, IB_EVENT_PORT_ACTIVE);
471		break;
472
473	case NETDEV_GOING_DOWN:
474		siw_device_goes_down(sdev);
475		break;
476
477	case NETDEV_DOWN:
478		sdev->state = IB_PORT_DOWN;
479		siw_port_event(sdev, 1, IB_EVENT_PORT_ERR);
480		break;
481
482	case NETDEV_REGISTER:
483		/*
484		 * Device registration now handled only by
485		 * rdma netlink commands. So it shall be impossible
486		 * to end up here with a valid siw device.
487		 */
488		siw_dbg(base_dev, "unexpected NETDEV_REGISTER event\n");
489		break;
490
491	case NETDEV_UNREGISTER:
492		ib_unregister_device_queued(&sdev->base_dev);
493		break;
494
495	case NETDEV_CHANGEADDR:
496		siw_port_event(sdev, 1, IB_EVENT_LID_CHANGE);
497		break;
498	/*
499	 * Todo: Below netdev events are currently not handled.
500	 */
501	case NETDEV_CHANGEMTU:
502	case NETDEV_CHANGE:
503		break;
504
505	default:
506		break;
507	}
508	ib_device_put(&sdev->base_dev);
509
510	return NOTIFY_OK;
511}
512
513static struct notifier_block siw_netdev_nb = {
514	.notifier_call = siw_netdev_event,
515};
516
517static int siw_newlink(const char *basedev_name, struct net_device *netdev)
518{
519	struct ib_device *base_dev;
520	struct siw_device *sdev = NULL;
521	int rv = -ENOMEM;
522
523	if (!siw_dev_qualified(netdev))
524		return -EINVAL;
525
526	base_dev = ib_device_get_by_netdev(netdev, RDMA_DRIVER_SIW);
527	if (base_dev) {
528		ib_device_put(base_dev);
529		return -EEXIST;
530	}
531	sdev = siw_device_create(netdev);
532	if (sdev) {
533		dev_dbg(&netdev->dev, "siw: new device\n");
534
535		if (netif_running(netdev) && netif_carrier_ok(netdev))
536			sdev->state = IB_PORT_ACTIVE;
537		else
538			sdev->state = IB_PORT_DOWN;
539
540		rv = siw_device_register(sdev, basedev_name);
541		if (rv)
542			ib_dealloc_device(&sdev->base_dev);
543	}
544	return rv;
545}
546
547static struct rdma_link_ops siw_link_ops = {
548	.type = "siw",
549	.newlink = siw_newlink,
550};
551
552/*
553 * siw_init_module - Initialize Softiwarp module and register with netdev
554 *                   subsystem.
555 */
556static __init int siw_init_module(void)
557{
558	int rv;
559	int nr_cpu;
560
561	if (SENDPAGE_THRESH < SIW_MAX_INLINE) {
562		pr_info("siw: sendpage threshold too small: %u\n",
563			(int)SENDPAGE_THRESH);
564		rv = -EINVAL;
565		goto out_error;
566	}
567	rv = siw_init_cpulist();
568	if (rv)
569		goto out_error;
570
571	rv = siw_cm_init();
572	if (rv)
573		goto out_error;
574
575	if (!siw_create_tx_threads()) {
576		pr_info("siw: Could not start any TX thread\n");
577		rv = -ENOMEM;
578		goto out_error;
579	}
580	/*
581	 * Locate CRC32 algorithm. If unsuccessful, fail
582	 * loading siw only, if CRC is required.
583	 */
584	siw_crypto_shash = crypto_alloc_shash("crc32c", 0, 0);
585	if (IS_ERR(siw_crypto_shash)) {
586		pr_info("siw: Loading CRC32c failed: %ld\n",
587			PTR_ERR(siw_crypto_shash));
588		siw_crypto_shash = NULL;
589		if (mpa_crc_required) {
590			rv = -EOPNOTSUPP;
591			goto out_error;
592		}
593	}
594	rv = register_netdevice_notifier(&siw_netdev_nb);
595	if (rv)
596		goto out_error;
597
598	rdma_link_register(&siw_link_ops);
599
600	pr_info("SoftiWARP attached\n");
601	return 0;
602
603out_error:
604	for (nr_cpu = 0; nr_cpu < nr_cpu_ids; nr_cpu++) {
605		if (siw_tx_thread[nr_cpu]) {
606			siw_stop_tx_thread(nr_cpu);
607			siw_tx_thread[nr_cpu] = NULL;
608		}
609	}
610	if (siw_crypto_shash)
611		crypto_free_shash(siw_crypto_shash);
612
613	pr_info("SoftIWARP attach failed. Error: %d\n", rv);
614
615	siw_cm_exit();
616	siw_destroy_cpulist();
617
618	return rv;
619}
620
621static void __exit siw_exit_module(void)
622{
623	int cpu;
624
625	for_each_possible_cpu(cpu) {
626		if (siw_tx_thread[cpu]) {
627			siw_stop_tx_thread(cpu);
628			siw_tx_thread[cpu] = NULL;
629		}
630	}
631	unregister_netdevice_notifier(&siw_netdev_nb);
632	rdma_link_unregister(&siw_link_ops);
633	ib_unregister_driver(RDMA_DRIVER_SIW);
634
635	siw_cm_exit();
636
637	siw_destroy_cpulist();
638
639	if (siw_crypto_shash)
640		crypto_free_shash(siw_crypto_shash);
641
642	pr_info("SoftiWARP detached\n");
643}
644
645module_init(siw_init_module);
646module_exit(siw_exit_module);
647
648MODULE_ALIAS_RDMA_LINK("siw");
649