xref: /kernel/linux/linux-6.6/drivers/nvme/target/fc.c (revision 62306a36)
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * Copyright (c) 2016 Avago Technologies.  All rights reserved.
4 */
5#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
6#include <linux/module.h>
7#include <linux/slab.h>
8#include <linux/blk-mq.h>
9#include <linux/parser.h>
10#include <linux/random.h>
11#include <uapi/scsi/fc/fc_fs.h>
12#include <uapi/scsi/fc/fc_els.h>
13
14#include "nvmet.h"
15#include <linux/nvme-fc-driver.h>
16#include <linux/nvme-fc.h>
17#include "../host/fc.h"
18
19
20/* *************************** Data Structures/Defines ****************** */
21
22
23#define NVMET_LS_CTX_COUNT		256
24
25struct nvmet_fc_tgtport;
26struct nvmet_fc_tgt_assoc;
27
28struct nvmet_fc_ls_iod {		/* for an LS RQST RCV */
29	struct nvmefc_ls_rsp		*lsrsp;
30	struct nvmefc_tgt_fcp_req	*fcpreq;	/* only if RS */
31
32	struct list_head		ls_rcv_list; /* tgtport->ls_rcv_list */
33
34	struct nvmet_fc_tgtport		*tgtport;
35	struct nvmet_fc_tgt_assoc	*assoc;
36	void				*hosthandle;
37
38	union nvmefc_ls_requests	*rqstbuf;
39	union nvmefc_ls_responses	*rspbuf;
40	u16				rqstdatalen;
41	dma_addr_t			rspdma;
42
43	struct scatterlist		sg[2];
44
45	struct work_struct		work;
46} __aligned(sizeof(unsigned long long));
47
48struct nvmet_fc_ls_req_op {		/* for an LS RQST XMT */
49	struct nvmefc_ls_req		ls_req;
50
51	struct nvmet_fc_tgtport		*tgtport;
52	void				*hosthandle;
53
54	int				ls_error;
55	struct list_head		lsreq_list; /* tgtport->ls_req_list */
56	bool				req_queued;
57};
58
59
60/* desired maximum for a single sequence - if sg list allows it */
61#define NVMET_FC_MAX_SEQ_LENGTH		(256 * 1024)
62
63enum nvmet_fcp_datadir {
64	NVMET_FCP_NODATA,
65	NVMET_FCP_WRITE,
66	NVMET_FCP_READ,
67	NVMET_FCP_ABORTED,
68};
69
70struct nvmet_fc_fcp_iod {
71	struct nvmefc_tgt_fcp_req	*fcpreq;
72
73	struct nvme_fc_cmd_iu		cmdiubuf;
74	struct nvme_fc_ersp_iu		rspiubuf;
75	dma_addr_t			rspdma;
76	struct scatterlist		*next_sg;
77	struct scatterlist		*data_sg;
78	int				data_sg_cnt;
79	u32				offset;
80	enum nvmet_fcp_datadir		io_dir;
81	bool				active;
82	bool				abort;
83	bool				aborted;
84	bool				writedataactive;
85	spinlock_t			flock;
86
87	struct nvmet_req		req;
88	struct work_struct		defer_work;
89
90	struct nvmet_fc_tgtport		*tgtport;
91	struct nvmet_fc_tgt_queue	*queue;
92
93	struct list_head		fcp_list;	/* tgtport->fcp_list */
94};
95
96struct nvmet_fc_tgtport {
97	struct nvmet_fc_target_port	fc_target_port;
98
99	struct list_head		tgt_list; /* nvmet_fc_target_list */
100	struct device			*dev;	/* dev for dma mapping */
101	struct nvmet_fc_target_template	*ops;
102
103	struct nvmet_fc_ls_iod		*iod;
104	spinlock_t			lock;
105	struct list_head		ls_rcv_list;
106	struct list_head		ls_req_list;
107	struct list_head		ls_busylist;
108	struct list_head		assoc_list;
109	struct list_head		host_list;
110	struct ida			assoc_cnt;
111	struct nvmet_fc_port_entry	*pe;
112	struct kref			ref;
113	u32				max_sg_cnt;
114
115	struct work_struct		put_work;
116};
117
118struct nvmet_fc_port_entry {
119	struct nvmet_fc_tgtport		*tgtport;
120	struct nvmet_port		*port;
121	u64				node_name;
122	u64				port_name;
123	struct list_head		pe_list;
124};
125
126struct nvmet_fc_defer_fcp_req {
127	struct list_head		req_list;
128	struct nvmefc_tgt_fcp_req	*fcp_req;
129};
130
131struct nvmet_fc_tgt_queue {
132	bool				ninetypercent;
133	u16				qid;
134	u16				sqsize;
135	u16				ersp_ratio;
136	__le16				sqhd;
137	atomic_t			connected;
138	atomic_t			sqtail;
139	atomic_t			zrspcnt;
140	atomic_t			rsn;
141	spinlock_t			qlock;
142	struct nvmet_cq			nvme_cq;
143	struct nvmet_sq			nvme_sq;
144	struct nvmet_fc_tgt_assoc	*assoc;
145	struct list_head		fod_list;
146	struct list_head		pending_cmd_list;
147	struct list_head		avail_defer_list;
148	struct workqueue_struct		*work_q;
149	struct kref			ref;
150	struct rcu_head			rcu;
151	struct nvmet_fc_fcp_iod		fod[];		/* array of fcp_iods */
152} __aligned(sizeof(unsigned long long));
153
154struct nvmet_fc_hostport {
155	struct nvmet_fc_tgtport		*tgtport;
156	void				*hosthandle;
157	struct list_head		host_list;
158	struct kref			ref;
159	u8				invalid;
160};
161
162struct nvmet_fc_tgt_assoc {
163	u64				association_id;
164	u32				a_id;
165	atomic_t			terminating;
166	struct nvmet_fc_tgtport		*tgtport;
167	struct nvmet_fc_hostport	*hostport;
168	struct nvmet_fc_ls_iod		*rcv_disconn;
169	struct list_head		a_list;
170	struct nvmet_fc_tgt_queue 	*queues[NVMET_NR_QUEUES + 1];
171	struct kref			ref;
172	struct work_struct		del_work;
173	struct rcu_head			rcu;
174};
175
176
177static inline int
178nvmet_fc_iodnum(struct nvmet_fc_ls_iod *iodptr)
179{
180	return (iodptr - iodptr->tgtport->iod);
181}
182
183static inline int
184nvmet_fc_fodnum(struct nvmet_fc_fcp_iod *fodptr)
185{
186	return (fodptr - fodptr->queue->fod);
187}
188
189
190/*
191 * Association and Connection IDs:
192 *
193 * Association ID will have random number in upper 6 bytes and zero
194 *   in lower 2 bytes
195 *
196 * Connection IDs will be Association ID with QID or'd in lower 2 bytes
197 *
198 * note: Association ID = Connection ID for queue 0
199 */
200#define BYTES_FOR_QID			sizeof(u16)
201#define BYTES_FOR_QID_SHIFT		(BYTES_FOR_QID * 8)
202#define NVMET_FC_QUEUEID_MASK		((u64)((1 << BYTES_FOR_QID_SHIFT) - 1))
203
204static inline u64
205nvmet_fc_makeconnid(struct nvmet_fc_tgt_assoc *assoc, u16 qid)
206{
207	return (assoc->association_id | qid);
208}
209
210static inline u64
211nvmet_fc_getassociationid(u64 connectionid)
212{
213	return connectionid & ~NVMET_FC_QUEUEID_MASK;
214}
215
216static inline u16
217nvmet_fc_getqueueid(u64 connectionid)
218{
219	return (u16)(connectionid & NVMET_FC_QUEUEID_MASK);
220}
221
222static inline struct nvmet_fc_tgtport *
223targetport_to_tgtport(struct nvmet_fc_target_port *targetport)
224{
225	return container_of(targetport, struct nvmet_fc_tgtport,
226				 fc_target_port);
227}
228
229static inline struct nvmet_fc_fcp_iod *
230nvmet_req_to_fod(struct nvmet_req *nvme_req)
231{
232	return container_of(nvme_req, struct nvmet_fc_fcp_iod, req);
233}
234
235
236/* *************************** Globals **************************** */
237
238
239static DEFINE_SPINLOCK(nvmet_fc_tgtlock);
240
241static LIST_HEAD(nvmet_fc_target_list);
242static DEFINE_IDA(nvmet_fc_tgtport_cnt);
243static LIST_HEAD(nvmet_fc_portentry_list);
244
245
246static void nvmet_fc_handle_ls_rqst_work(struct work_struct *work);
247static void nvmet_fc_fcp_rqst_op_defer_work(struct work_struct *work);
248static void nvmet_fc_tgt_a_put(struct nvmet_fc_tgt_assoc *assoc);
249static int nvmet_fc_tgt_a_get(struct nvmet_fc_tgt_assoc *assoc);
250static void nvmet_fc_tgt_q_put(struct nvmet_fc_tgt_queue *queue);
251static int nvmet_fc_tgt_q_get(struct nvmet_fc_tgt_queue *queue);
252static void nvmet_fc_tgtport_put(struct nvmet_fc_tgtport *tgtport);
253static void nvmet_fc_put_tgtport_work(struct work_struct *work)
254{
255	struct nvmet_fc_tgtport *tgtport =
256		container_of(work, struct nvmet_fc_tgtport, put_work);
257
258	nvmet_fc_tgtport_put(tgtport);
259}
260static int nvmet_fc_tgtport_get(struct nvmet_fc_tgtport *tgtport);
261static void nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport,
262					struct nvmet_fc_fcp_iod *fod);
263static void nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc);
264static void nvmet_fc_xmt_ls_rsp(struct nvmet_fc_tgtport *tgtport,
265				struct nvmet_fc_ls_iod *iod);
266
267
268/* *********************** FC-NVME DMA Handling **************************** */
269
270/*
271 * The fcloop device passes in a NULL device pointer. Real LLD's will
272 * pass in a valid device pointer. If NULL is passed to the dma mapping
273 * routines, depending on the platform, it may or may not succeed, and
274 * may crash.
275 *
276 * As such:
277 * Wrapper all the dma routines and check the dev pointer.
278 *
279 * If simple mappings (return just a dma address, we'll noop them,
280 * returning a dma address of 0.
281 *
282 * On more complex mappings (dma_map_sg), a pseudo routine fills
283 * in the scatter list, setting all dma addresses to 0.
284 */
285
286static inline dma_addr_t
287fc_dma_map_single(struct device *dev, void *ptr, size_t size,
288		enum dma_data_direction dir)
289{
290	return dev ? dma_map_single(dev, ptr, size, dir) : (dma_addr_t)0L;
291}
292
293static inline int
294fc_dma_mapping_error(struct device *dev, dma_addr_t dma_addr)
295{
296	return dev ? dma_mapping_error(dev, dma_addr) : 0;
297}
298
299static inline void
300fc_dma_unmap_single(struct device *dev, dma_addr_t addr, size_t size,
301	enum dma_data_direction dir)
302{
303	if (dev)
304		dma_unmap_single(dev, addr, size, dir);
305}
306
307static inline void
308fc_dma_sync_single_for_cpu(struct device *dev, dma_addr_t addr, size_t size,
309		enum dma_data_direction dir)
310{
311	if (dev)
312		dma_sync_single_for_cpu(dev, addr, size, dir);
313}
314
315static inline void
316fc_dma_sync_single_for_device(struct device *dev, dma_addr_t addr, size_t size,
317		enum dma_data_direction dir)
318{
319	if (dev)
320		dma_sync_single_for_device(dev, addr, size, dir);
321}
322
323/* pseudo dma_map_sg call */
324static int
325fc_map_sg(struct scatterlist *sg, int nents)
326{
327	struct scatterlist *s;
328	int i;
329
330	WARN_ON(nents == 0 || sg[0].length == 0);
331
332	for_each_sg(sg, s, nents, i) {
333		s->dma_address = 0L;
334#ifdef CONFIG_NEED_SG_DMA_LENGTH
335		s->dma_length = s->length;
336#endif
337	}
338	return nents;
339}
340
341static inline int
342fc_dma_map_sg(struct device *dev, struct scatterlist *sg, int nents,
343		enum dma_data_direction dir)
344{
345	return dev ? dma_map_sg(dev, sg, nents, dir) : fc_map_sg(sg, nents);
346}
347
348static inline void
349fc_dma_unmap_sg(struct device *dev, struct scatterlist *sg, int nents,
350		enum dma_data_direction dir)
351{
352	if (dev)
353		dma_unmap_sg(dev, sg, nents, dir);
354}
355
356
357/* ********************** FC-NVME LS XMT Handling ************************* */
358
359
360static void
361__nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)
362{
363	struct nvmet_fc_tgtport *tgtport = lsop->tgtport;
364	struct nvmefc_ls_req *lsreq = &lsop->ls_req;
365	unsigned long flags;
366
367	spin_lock_irqsave(&tgtport->lock, flags);
368
369	if (!lsop->req_queued) {
370		spin_unlock_irqrestore(&tgtport->lock, flags);
371		goto out_putwork;
372	}
373
374	list_del(&lsop->lsreq_list);
375
376	lsop->req_queued = false;
377
378	spin_unlock_irqrestore(&tgtport->lock, flags);
379
380	fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma,
381				  (lsreq->rqstlen + lsreq->rsplen),
382				  DMA_BIDIRECTIONAL);
383
384out_putwork:
385	queue_work(nvmet_wq, &tgtport->put_work);
386}
387
388static int
389__nvmet_fc_send_ls_req(struct nvmet_fc_tgtport *tgtport,
390		struct nvmet_fc_ls_req_op *lsop,
391		void (*done)(struct nvmefc_ls_req *req, int status))
392{
393	struct nvmefc_ls_req *lsreq = &lsop->ls_req;
394	unsigned long flags;
395	int ret = 0;
396
397	if (!tgtport->ops->ls_req)
398		return -EOPNOTSUPP;
399
400	if (!nvmet_fc_tgtport_get(tgtport))
401		return -ESHUTDOWN;
402
403	lsreq->done = done;
404	lsop->req_queued = false;
405	INIT_LIST_HEAD(&lsop->lsreq_list);
406
407	lsreq->rqstdma = fc_dma_map_single(tgtport->dev, lsreq->rqstaddr,
408				  lsreq->rqstlen + lsreq->rsplen,
409				  DMA_BIDIRECTIONAL);
410	if (fc_dma_mapping_error(tgtport->dev, lsreq->rqstdma)) {
411		ret = -EFAULT;
412		goto out_puttgtport;
413	}
414	lsreq->rspdma = lsreq->rqstdma + lsreq->rqstlen;
415
416	spin_lock_irqsave(&tgtport->lock, flags);
417
418	list_add_tail(&lsop->lsreq_list, &tgtport->ls_req_list);
419
420	lsop->req_queued = true;
421
422	spin_unlock_irqrestore(&tgtport->lock, flags);
423
424	ret = tgtport->ops->ls_req(&tgtport->fc_target_port, lsop->hosthandle,
425				   lsreq);
426	if (ret)
427		goto out_unlink;
428
429	return 0;
430
431out_unlink:
432	lsop->ls_error = ret;
433	spin_lock_irqsave(&tgtport->lock, flags);
434	lsop->req_queued = false;
435	list_del(&lsop->lsreq_list);
436	spin_unlock_irqrestore(&tgtport->lock, flags);
437	fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma,
438				  (lsreq->rqstlen + lsreq->rsplen),
439				  DMA_BIDIRECTIONAL);
440out_puttgtport:
441	nvmet_fc_tgtport_put(tgtport);
442
443	return ret;
444}
445
446static int
447nvmet_fc_send_ls_req_async(struct nvmet_fc_tgtport *tgtport,
448		struct nvmet_fc_ls_req_op *lsop,
449		void (*done)(struct nvmefc_ls_req *req, int status))
450{
451	/* don't wait for completion */
452
453	return __nvmet_fc_send_ls_req(tgtport, lsop, done);
454}
455
456static void
457nvmet_fc_disconnect_assoc_done(struct nvmefc_ls_req *lsreq, int status)
458{
459	struct nvmet_fc_ls_req_op *lsop =
460		container_of(lsreq, struct nvmet_fc_ls_req_op, ls_req);
461
462	__nvmet_fc_finish_ls_req(lsop);
463
464	/* fc-nvme target doesn't care about success or failure of cmd */
465
466	kfree(lsop);
467}
468
469/*
470 * This routine sends a FC-NVME LS to disconnect (aka terminate)
471 * the FC-NVME Association.  Terminating the association also
472 * terminates the FC-NVME connections (per queue, both admin and io
473 * queues) that are part of the association. E.g. things are torn
474 * down, and the related FC-NVME Association ID and Connection IDs
475 * become invalid.
476 *
477 * The behavior of the fc-nvme target is such that it's
478 * understanding of the association and connections will implicitly
479 * be torn down. The action is implicit as it may be due to a loss of
480 * connectivity with the fc-nvme host, so the target may never get a
481 * response even if it tried.  As such, the action of this routine
482 * is to asynchronously send the LS, ignore any results of the LS, and
483 * continue on with terminating the association. If the fc-nvme host
484 * is present and receives the LS, it too can tear down.
485 */
486static void
487nvmet_fc_xmt_disconnect_assoc(struct nvmet_fc_tgt_assoc *assoc)
488{
489	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
490	struct fcnvme_ls_disconnect_assoc_rqst *discon_rqst;
491	struct fcnvme_ls_disconnect_assoc_acc *discon_acc;
492	struct nvmet_fc_ls_req_op *lsop;
493	struct nvmefc_ls_req *lsreq;
494	int ret;
495
496	/*
497	 * If ls_req is NULL or no hosthandle, it's an older lldd and no
498	 * message is normal. Otherwise, send unless the hostport has
499	 * already been invalidated by the lldd.
500	 */
501	if (!tgtport->ops->ls_req || !assoc->hostport ||
502	    assoc->hostport->invalid)
503		return;
504
505	lsop = kzalloc((sizeof(*lsop) +
506			sizeof(*discon_rqst) + sizeof(*discon_acc) +
507			tgtport->ops->lsrqst_priv_sz), GFP_KERNEL);
508	if (!lsop) {
509		dev_info(tgtport->dev,
510			"{%d:%d} send Disconnect Association failed: ENOMEM\n",
511			tgtport->fc_target_port.port_num, assoc->a_id);
512		return;
513	}
514
515	discon_rqst = (struct fcnvme_ls_disconnect_assoc_rqst *)&lsop[1];
516	discon_acc = (struct fcnvme_ls_disconnect_assoc_acc *)&discon_rqst[1];
517	lsreq = &lsop->ls_req;
518	if (tgtport->ops->lsrqst_priv_sz)
519		lsreq->private = (void *)&discon_acc[1];
520	else
521		lsreq->private = NULL;
522
523	lsop->tgtport = tgtport;
524	lsop->hosthandle = assoc->hostport->hosthandle;
525
526	nvmefc_fmt_lsreq_discon_assoc(lsreq, discon_rqst, discon_acc,
527				assoc->association_id);
528
529	ret = nvmet_fc_send_ls_req_async(tgtport, lsop,
530				nvmet_fc_disconnect_assoc_done);
531	if (ret) {
532		dev_info(tgtport->dev,
533			"{%d:%d} XMT Disconnect Association failed: %d\n",
534			tgtport->fc_target_port.port_num, assoc->a_id, ret);
535		kfree(lsop);
536	}
537}
538
539
540/* *********************** FC-NVME Port Management ************************ */
541
542
543static int
544nvmet_fc_alloc_ls_iodlist(struct nvmet_fc_tgtport *tgtport)
545{
546	struct nvmet_fc_ls_iod *iod;
547	int i;
548
549	iod = kcalloc(NVMET_LS_CTX_COUNT, sizeof(struct nvmet_fc_ls_iod),
550			GFP_KERNEL);
551	if (!iod)
552		return -ENOMEM;
553
554	tgtport->iod = iod;
555
556	for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++) {
557		INIT_WORK(&iod->work, nvmet_fc_handle_ls_rqst_work);
558		iod->tgtport = tgtport;
559		list_add_tail(&iod->ls_rcv_list, &tgtport->ls_rcv_list);
560
561		iod->rqstbuf = kzalloc(sizeof(union nvmefc_ls_requests) +
562				       sizeof(union nvmefc_ls_responses),
563				       GFP_KERNEL);
564		if (!iod->rqstbuf)
565			goto out_fail;
566
567		iod->rspbuf = (union nvmefc_ls_responses *)&iod->rqstbuf[1];
568
569		iod->rspdma = fc_dma_map_single(tgtport->dev, iod->rspbuf,
570						sizeof(*iod->rspbuf),
571						DMA_TO_DEVICE);
572		if (fc_dma_mapping_error(tgtport->dev, iod->rspdma))
573			goto out_fail;
574	}
575
576	return 0;
577
578out_fail:
579	kfree(iod->rqstbuf);
580	list_del(&iod->ls_rcv_list);
581	for (iod--, i--; i >= 0; iod--, i--) {
582		fc_dma_unmap_single(tgtport->dev, iod->rspdma,
583				sizeof(*iod->rspbuf), DMA_TO_DEVICE);
584		kfree(iod->rqstbuf);
585		list_del(&iod->ls_rcv_list);
586	}
587
588	kfree(iod);
589
590	return -EFAULT;
591}
592
593static void
594nvmet_fc_free_ls_iodlist(struct nvmet_fc_tgtport *tgtport)
595{
596	struct nvmet_fc_ls_iod *iod = tgtport->iod;
597	int i;
598
599	for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++) {
600		fc_dma_unmap_single(tgtport->dev,
601				iod->rspdma, sizeof(*iod->rspbuf),
602				DMA_TO_DEVICE);
603		kfree(iod->rqstbuf);
604		list_del(&iod->ls_rcv_list);
605	}
606	kfree(tgtport->iod);
607}
608
609static struct nvmet_fc_ls_iod *
610nvmet_fc_alloc_ls_iod(struct nvmet_fc_tgtport *tgtport)
611{
612	struct nvmet_fc_ls_iod *iod;
613	unsigned long flags;
614
615	spin_lock_irqsave(&tgtport->lock, flags);
616	iod = list_first_entry_or_null(&tgtport->ls_rcv_list,
617					struct nvmet_fc_ls_iod, ls_rcv_list);
618	if (iod)
619		list_move_tail(&iod->ls_rcv_list, &tgtport->ls_busylist);
620	spin_unlock_irqrestore(&tgtport->lock, flags);
621	return iod;
622}
623
624
625static void
626nvmet_fc_free_ls_iod(struct nvmet_fc_tgtport *tgtport,
627			struct nvmet_fc_ls_iod *iod)
628{
629	unsigned long flags;
630
631	spin_lock_irqsave(&tgtport->lock, flags);
632	list_move(&iod->ls_rcv_list, &tgtport->ls_rcv_list);
633	spin_unlock_irqrestore(&tgtport->lock, flags);
634}
635
636static void
637nvmet_fc_prep_fcp_iodlist(struct nvmet_fc_tgtport *tgtport,
638				struct nvmet_fc_tgt_queue *queue)
639{
640	struct nvmet_fc_fcp_iod *fod = queue->fod;
641	int i;
642
643	for (i = 0; i < queue->sqsize; fod++, i++) {
644		INIT_WORK(&fod->defer_work, nvmet_fc_fcp_rqst_op_defer_work);
645		fod->tgtport = tgtport;
646		fod->queue = queue;
647		fod->active = false;
648		fod->abort = false;
649		fod->aborted = false;
650		fod->fcpreq = NULL;
651		list_add_tail(&fod->fcp_list, &queue->fod_list);
652		spin_lock_init(&fod->flock);
653
654		fod->rspdma = fc_dma_map_single(tgtport->dev, &fod->rspiubuf,
655					sizeof(fod->rspiubuf), DMA_TO_DEVICE);
656		if (fc_dma_mapping_error(tgtport->dev, fod->rspdma)) {
657			list_del(&fod->fcp_list);
658			for (fod--, i--; i >= 0; fod--, i--) {
659				fc_dma_unmap_single(tgtport->dev, fod->rspdma,
660						sizeof(fod->rspiubuf),
661						DMA_TO_DEVICE);
662				fod->rspdma = 0L;
663				list_del(&fod->fcp_list);
664			}
665
666			return;
667		}
668	}
669}
670
671static void
672nvmet_fc_destroy_fcp_iodlist(struct nvmet_fc_tgtport *tgtport,
673				struct nvmet_fc_tgt_queue *queue)
674{
675	struct nvmet_fc_fcp_iod *fod = queue->fod;
676	int i;
677
678	for (i = 0; i < queue->sqsize; fod++, i++) {
679		if (fod->rspdma)
680			fc_dma_unmap_single(tgtport->dev, fod->rspdma,
681				sizeof(fod->rspiubuf), DMA_TO_DEVICE);
682	}
683}
684
685static struct nvmet_fc_fcp_iod *
686nvmet_fc_alloc_fcp_iod(struct nvmet_fc_tgt_queue *queue)
687{
688	struct nvmet_fc_fcp_iod *fod;
689
690	lockdep_assert_held(&queue->qlock);
691
692	fod = list_first_entry_or_null(&queue->fod_list,
693					struct nvmet_fc_fcp_iod, fcp_list);
694	if (fod) {
695		list_del(&fod->fcp_list);
696		fod->active = true;
697		/*
698		 * no queue reference is taken, as it was taken by the
699		 * queue lookup just prior to the allocation. The iod
700		 * will "inherit" that reference.
701		 */
702	}
703	return fod;
704}
705
706
707static void
708nvmet_fc_queue_fcp_req(struct nvmet_fc_tgtport *tgtport,
709		       struct nvmet_fc_tgt_queue *queue,
710		       struct nvmefc_tgt_fcp_req *fcpreq)
711{
712	struct nvmet_fc_fcp_iod *fod = fcpreq->nvmet_fc_private;
713
714	/*
715	 * put all admin cmds on hw queue id 0. All io commands go to
716	 * the respective hw queue based on a modulo basis
717	 */
718	fcpreq->hwqid = queue->qid ?
719			((queue->qid - 1) % tgtport->ops->max_hw_queues) : 0;
720
721	nvmet_fc_handle_fcp_rqst(tgtport, fod);
722}
723
724static void
725nvmet_fc_fcp_rqst_op_defer_work(struct work_struct *work)
726{
727	struct nvmet_fc_fcp_iod *fod =
728		container_of(work, struct nvmet_fc_fcp_iod, defer_work);
729
730	/* Submit deferred IO for processing */
731	nvmet_fc_queue_fcp_req(fod->tgtport, fod->queue, fod->fcpreq);
732
733}
734
735static void
736nvmet_fc_free_fcp_iod(struct nvmet_fc_tgt_queue *queue,
737			struct nvmet_fc_fcp_iod *fod)
738{
739	struct nvmefc_tgt_fcp_req *fcpreq = fod->fcpreq;
740	struct nvmet_fc_tgtport *tgtport = fod->tgtport;
741	struct nvmet_fc_defer_fcp_req *deferfcp;
742	unsigned long flags;
743
744	fc_dma_sync_single_for_cpu(tgtport->dev, fod->rspdma,
745				sizeof(fod->rspiubuf), DMA_TO_DEVICE);
746
747	fcpreq->nvmet_fc_private = NULL;
748
749	fod->active = false;
750	fod->abort = false;
751	fod->aborted = false;
752	fod->writedataactive = false;
753	fod->fcpreq = NULL;
754
755	tgtport->ops->fcp_req_release(&tgtport->fc_target_port, fcpreq);
756
757	/* release the queue lookup reference on the completed IO */
758	nvmet_fc_tgt_q_put(queue);
759
760	spin_lock_irqsave(&queue->qlock, flags);
761	deferfcp = list_first_entry_or_null(&queue->pending_cmd_list,
762				struct nvmet_fc_defer_fcp_req, req_list);
763	if (!deferfcp) {
764		list_add_tail(&fod->fcp_list, &fod->queue->fod_list);
765		spin_unlock_irqrestore(&queue->qlock, flags);
766		return;
767	}
768
769	/* Re-use the fod for the next pending cmd that was deferred */
770	list_del(&deferfcp->req_list);
771
772	fcpreq = deferfcp->fcp_req;
773
774	/* deferfcp can be reused for another IO at a later date */
775	list_add_tail(&deferfcp->req_list, &queue->avail_defer_list);
776
777	spin_unlock_irqrestore(&queue->qlock, flags);
778
779	/* Save NVME CMD IO in fod */
780	memcpy(&fod->cmdiubuf, fcpreq->rspaddr, fcpreq->rsplen);
781
782	/* Setup new fcpreq to be processed */
783	fcpreq->rspaddr = NULL;
784	fcpreq->rsplen  = 0;
785	fcpreq->nvmet_fc_private = fod;
786	fod->fcpreq = fcpreq;
787	fod->active = true;
788
789	/* inform LLDD IO is now being processed */
790	tgtport->ops->defer_rcv(&tgtport->fc_target_port, fcpreq);
791
792	/*
793	 * Leave the queue lookup get reference taken when
794	 * fod was originally allocated.
795	 */
796
797	queue_work(queue->work_q, &fod->defer_work);
798}
799
800static struct nvmet_fc_tgt_queue *
801nvmet_fc_alloc_target_queue(struct nvmet_fc_tgt_assoc *assoc,
802			u16 qid, u16 sqsize)
803{
804	struct nvmet_fc_tgt_queue *queue;
805	int ret;
806
807	if (qid > NVMET_NR_QUEUES)
808		return NULL;
809
810	queue = kzalloc(struct_size(queue, fod, sqsize), GFP_KERNEL);
811	if (!queue)
812		return NULL;
813
814	queue->work_q = alloc_workqueue("ntfc%d.%d.%d", 0, 0,
815				assoc->tgtport->fc_target_port.port_num,
816				assoc->a_id, qid);
817	if (!queue->work_q)
818		goto out_free_queue;
819
820	queue->qid = qid;
821	queue->sqsize = sqsize;
822	queue->assoc = assoc;
823	INIT_LIST_HEAD(&queue->fod_list);
824	INIT_LIST_HEAD(&queue->avail_defer_list);
825	INIT_LIST_HEAD(&queue->pending_cmd_list);
826	atomic_set(&queue->connected, 0);
827	atomic_set(&queue->sqtail, 0);
828	atomic_set(&queue->rsn, 1);
829	atomic_set(&queue->zrspcnt, 0);
830	spin_lock_init(&queue->qlock);
831	kref_init(&queue->ref);
832
833	nvmet_fc_prep_fcp_iodlist(assoc->tgtport, queue);
834
835	ret = nvmet_sq_init(&queue->nvme_sq);
836	if (ret)
837		goto out_fail_iodlist;
838
839	WARN_ON(assoc->queues[qid]);
840	assoc->queues[qid] = queue;
841
842	return queue;
843
844out_fail_iodlist:
845	nvmet_fc_destroy_fcp_iodlist(assoc->tgtport, queue);
846	destroy_workqueue(queue->work_q);
847out_free_queue:
848	kfree(queue);
849	return NULL;
850}
851
852
853static void
854nvmet_fc_tgt_queue_free(struct kref *ref)
855{
856	struct nvmet_fc_tgt_queue *queue =
857		container_of(ref, struct nvmet_fc_tgt_queue, ref);
858
859	nvmet_fc_destroy_fcp_iodlist(queue->assoc->tgtport, queue);
860
861	destroy_workqueue(queue->work_q);
862
863	kfree_rcu(queue, rcu);
864}
865
866static void
867nvmet_fc_tgt_q_put(struct nvmet_fc_tgt_queue *queue)
868{
869	kref_put(&queue->ref, nvmet_fc_tgt_queue_free);
870}
871
872static int
873nvmet_fc_tgt_q_get(struct nvmet_fc_tgt_queue *queue)
874{
875	return kref_get_unless_zero(&queue->ref);
876}
877
878
879static void
880nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue)
881{
882	struct nvmet_fc_tgtport *tgtport = queue->assoc->tgtport;
883	struct nvmet_fc_fcp_iod *fod = queue->fod;
884	struct nvmet_fc_defer_fcp_req *deferfcp, *tempptr;
885	unsigned long flags;
886	int i;
887	bool disconnect;
888
889	disconnect = atomic_xchg(&queue->connected, 0);
890
891	/* if not connected, nothing to do */
892	if (!disconnect)
893		return;
894
895	spin_lock_irqsave(&queue->qlock, flags);
896	/* abort outstanding io's */
897	for (i = 0; i < queue->sqsize; fod++, i++) {
898		if (fod->active) {
899			spin_lock(&fod->flock);
900			fod->abort = true;
901			/*
902			 * only call lldd abort routine if waiting for
903			 * writedata. other outstanding ops should finish
904			 * on their own.
905			 */
906			if (fod->writedataactive) {
907				fod->aborted = true;
908				spin_unlock(&fod->flock);
909				tgtport->ops->fcp_abort(
910					&tgtport->fc_target_port, fod->fcpreq);
911			} else
912				spin_unlock(&fod->flock);
913		}
914	}
915
916	/* Cleanup defer'ed IOs in queue */
917	list_for_each_entry_safe(deferfcp, tempptr, &queue->avail_defer_list,
918				req_list) {
919		list_del(&deferfcp->req_list);
920		kfree(deferfcp);
921	}
922
923	for (;;) {
924		deferfcp = list_first_entry_or_null(&queue->pending_cmd_list,
925				struct nvmet_fc_defer_fcp_req, req_list);
926		if (!deferfcp)
927			break;
928
929		list_del(&deferfcp->req_list);
930		spin_unlock_irqrestore(&queue->qlock, flags);
931
932		tgtport->ops->defer_rcv(&tgtport->fc_target_port,
933				deferfcp->fcp_req);
934
935		tgtport->ops->fcp_abort(&tgtport->fc_target_port,
936				deferfcp->fcp_req);
937
938		tgtport->ops->fcp_req_release(&tgtport->fc_target_port,
939				deferfcp->fcp_req);
940
941		/* release the queue lookup reference */
942		nvmet_fc_tgt_q_put(queue);
943
944		kfree(deferfcp);
945
946		spin_lock_irqsave(&queue->qlock, flags);
947	}
948	spin_unlock_irqrestore(&queue->qlock, flags);
949
950	flush_workqueue(queue->work_q);
951
952	nvmet_sq_destroy(&queue->nvme_sq);
953
954	nvmet_fc_tgt_q_put(queue);
955}
956
957static struct nvmet_fc_tgt_queue *
958nvmet_fc_find_target_queue(struct nvmet_fc_tgtport *tgtport,
959				u64 connection_id)
960{
961	struct nvmet_fc_tgt_assoc *assoc;
962	struct nvmet_fc_tgt_queue *queue;
963	u64 association_id = nvmet_fc_getassociationid(connection_id);
964	u16 qid = nvmet_fc_getqueueid(connection_id);
965
966	if (qid > NVMET_NR_QUEUES)
967		return NULL;
968
969	rcu_read_lock();
970	list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
971		if (association_id == assoc->association_id) {
972			queue = assoc->queues[qid];
973			if (queue &&
974			    (!atomic_read(&queue->connected) ||
975			     !nvmet_fc_tgt_q_get(queue)))
976				queue = NULL;
977			rcu_read_unlock();
978			return queue;
979		}
980	}
981	rcu_read_unlock();
982	return NULL;
983}
984
985static void
986nvmet_fc_hostport_free(struct kref *ref)
987{
988	struct nvmet_fc_hostport *hostport =
989		container_of(ref, struct nvmet_fc_hostport, ref);
990	struct nvmet_fc_tgtport *tgtport = hostport->tgtport;
991	unsigned long flags;
992
993	spin_lock_irqsave(&tgtport->lock, flags);
994	list_del(&hostport->host_list);
995	spin_unlock_irqrestore(&tgtport->lock, flags);
996	if (tgtport->ops->host_release && hostport->invalid)
997		tgtport->ops->host_release(hostport->hosthandle);
998	kfree(hostport);
999	nvmet_fc_tgtport_put(tgtport);
1000}
1001
1002static void
1003nvmet_fc_hostport_put(struct nvmet_fc_hostport *hostport)
1004{
1005	kref_put(&hostport->ref, nvmet_fc_hostport_free);
1006}
1007
1008static int
1009nvmet_fc_hostport_get(struct nvmet_fc_hostport *hostport)
1010{
1011	return kref_get_unless_zero(&hostport->ref);
1012}
1013
1014static void
1015nvmet_fc_free_hostport(struct nvmet_fc_hostport *hostport)
1016{
1017	/* if LLDD not implemented, leave as NULL */
1018	if (!hostport || !hostport->hosthandle)
1019		return;
1020
1021	nvmet_fc_hostport_put(hostport);
1022}
1023
1024static struct nvmet_fc_hostport *
1025nvmet_fc_match_hostport(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
1026{
1027	struct nvmet_fc_hostport *host;
1028
1029	lockdep_assert_held(&tgtport->lock);
1030
1031	list_for_each_entry(host, &tgtport->host_list, host_list) {
1032		if (host->hosthandle == hosthandle && !host->invalid) {
1033			if (nvmet_fc_hostport_get(host))
1034				return (host);
1035		}
1036	}
1037
1038	return NULL;
1039}
1040
1041static struct nvmet_fc_hostport *
1042nvmet_fc_alloc_hostport(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
1043{
1044	struct nvmet_fc_hostport *newhost, *match = NULL;
1045	unsigned long flags;
1046
1047	/* if LLDD not implemented, leave as NULL */
1048	if (!hosthandle)
1049		return NULL;
1050
1051	/*
1052	 * take reference for what will be the newly allocated hostport if
1053	 * we end up using a new allocation
1054	 */
1055	if (!nvmet_fc_tgtport_get(tgtport))
1056		return ERR_PTR(-EINVAL);
1057
1058	spin_lock_irqsave(&tgtport->lock, flags);
1059	match = nvmet_fc_match_hostport(tgtport, hosthandle);
1060	spin_unlock_irqrestore(&tgtport->lock, flags);
1061
1062	if (match) {
1063		/* no new allocation - release reference */
1064		nvmet_fc_tgtport_put(tgtport);
1065		return match;
1066	}
1067
1068	newhost = kzalloc(sizeof(*newhost), GFP_KERNEL);
1069	if (!newhost) {
1070		/* no new allocation - release reference */
1071		nvmet_fc_tgtport_put(tgtport);
1072		return ERR_PTR(-ENOMEM);
1073	}
1074
1075	spin_lock_irqsave(&tgtport->lock, flags);
1076	match = nvmet_fc_match_hostport(tgtport, hosthandle);
1077	if (match) {
1078		/* new allocation not needed */
1079		kfree(newhost);
1080		newhost = match;
1081	} else {
1082		newhost->tgtport = tgtport;
1083		newhost->hosthandle = hosthandle;
1084		INIT_LIST_HEAD(&newhost->host_list);
1085		kref_init(&newhost->ref);
1086
1087		list_add_tail(&newhost->host_list, &tgtport->host_list);
1088	}
1089	spin_unlock_irqrestore(&tgtport->lock, flags);
1090
1091	return newhost;
1092}
1093
1094static void
1095nvmet_fc_delete_assoc(struct nvmet_fc_tgt_assoc *assoc)
1096{
1097	nvmet_fc_delete_target_assoc(assoc);
1098	nvmet_fc_tgt_a_put(assoc);
1099}
1100
1101static void
1102nvmet_fc_delete_assoc_work(struct work_struct *work)
1103{
1104	struct nvmet_fc_tgt_assoc *assoc =
1105		container_of(work, struct nvmet_fc_tgt_assoc, del_work);
1106	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
1107
1108	nvmet_fc_delete_assoc(assoc);
1109	nvmet_fc_tgtport_put(tgtport);
1110}
1111
1112static void
1113nvmet_fc_schedule_delete_assoc(struct nvmet_fc_tgt_assoc *assoc)
1114{
1115	nvmet_fc_tgtport_get(assoc->tgtport);
1116	queue_work(nvmet_wq, &assoc->del_work);
1117}
1118
1119static struct nvmet_fc_tgt_assoc *
1120nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
1121{
1122	struct nvmet_fc_tgt_assoc *assoc, *tmpassoc;
1123	unsigned long flags;
1124	u64 ran;
1125	int idx;
1126	bool needrandom = true;
1127
1128	if (!tgtport->pe)
1129		return NULL;
1130
1131	assoc = kzalloc(sizeof(*assoc), GFP_KERNEL);
1132	if (!assoc)
1133		return NULL;
1134
1135	idx = ida_alloc(&tgtport->assoc_cnt, GFP_KERNEL);
1136	if (idx < 0)
1137		goto out_free_assoc;
1138
1139	if (!nvmet_fc_tgtport_get(tgtport))
1140		goto out_ida;
1141
1142	assoc->hostport = nvmet_fc_alloc_hostport(tgtport, hosthandle);
1143	if (IS_ERR(assoc->hostport))
1144		goto out_put;
1145
1146	assoc->tgtport = tgtport;
1147	assoc->a_id = idx;
1148	INIT_LIST_HEAD(&assoc->a_list);
1149	kref_init(&assoc->ref);
1150	INIT_WORK(&assoc->del_work, nvmet_fc_delete_assoc_work);
1151	atomic_set(&assoc->terminating, 0);
1152
1153	while (needrandom) {
1154		get_random_bytes(&ran, sizeof(ran) - BYTES_FOR_QID);
1155		ran = ran << BYTES_FOR_QID_SHIFT;
1156
1157		spin_lock_irqsave(&tgtport->lock, flags);
1158		needrandom = false;
1159		list_for_each_entry(tmpassoc, &tgtport->assoc_list, a_list) {
1160			if (ran == tmpassoc->association_id) {
1161				needrandom = true;
1162				break;
1163			}
1164		}
1165		if (!needrandom) {
1166			assoc->association_id = ran;
1167			list_add_tail_rcu(&assoc->a_list, &tgtport->assoc_list);
1168		}
1169		spin_unlock_irqrestore(&tgtport->lock, flags);
1170	}
1171
1172	return assoc;
1173
1174out_put:
1175	nvmet_fc_tgtport_put(tgtport);
1176out_ida:
1177	ida_free(&tgtport->assoc_cnt, idx);
1178out_free_assoc:
1179	kfree(assoc);
1180	return NULL;
1181}
1182
1183static void
1184nvmet_fc_target_assoc_free(struct kref *ref)
1185{
1186	struct nvmet_fc_tgt_assoc *assoc =
1187		container_of(ref, struct nvmet_fc_tgt_assoc, ref);
1188	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
1189	struct nvmet_fc_ls_iod	*oldls;
1190	unsigned long flags;
1191	int i;
1192
1193	for (i = NVMET_NR_QUEUES; i >= 0; i--) {
1194		if (assoc->queues[i])
1195			nvmet_fc_delete_target_queue(assoc->queues[i]);
1196	}
1197
1198	/* Send Disconnect now that all i/o has completed */
1199	nvmet_fc_xmt_disconnect_assoc(assoc);
1200
1201	nvmet_fc_free_hostport(assoc->hostport);
1202	spin_lock_irqsave(&tgtport->lock, flags);
1203	oldls = assoc->rcv_disconn;
1204	spin_unlock_irqrestore(&tgtport->lock, flags);
1205	/* if pending Rcv Disconnect Association LS, send rsp now */
1206	if (oldls)
1207		nvmet_fc_xmt_ls_rsp(tgtport, oldls);
1208	ida_free(&tgtport->assoc_cnt, assoc->a_id);
1209	dev_info(tgtport->dev,
1210		"{%d:%d} Association freed\n",
1211		tgtport->fc_target_port.port_num, assoc->a_id);
1212	kfree_rcu(assoc, rcu);
1213	nvmet_fc_tgtport_put(tgtport);
1214}
1215
1216static void
1217nvmet_fc_tgt_a_put(struct nvmet_fc_tgt_assoc *assoc)
1218{
1219	kref_put(&assoc->ref, nvmet_fc_target_assoc_free);
1220}
1221
1222static int
1223nvmet_fc_tgt_a_get(struct nvmet_fc_tgt_assoc *assoc)
1224{
1225	return kref_get_unless_zero(&assoc->ref);
1226}
1227
1228static void
1229nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc)
1230{
1231	struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
1232	unsigned long flags;
1233	int i, terminating;
1234
1235	terminating = atomic_xchg(&assoc->terminating, 1);
1236
1237	/* if already terminating, do nothing */
1238	if (terminating)
1239		return;
1240
1241	spin_lock_irqsave(&tgtport->lock, flags);
1242	list_del_rcu(&assoc->a_list);
1243	spin_unlock_irqrestore(&tgtport->lock, flags);
1244
1245	synchronize_rcu();
1246
1247	/* ensure all in-flight I/Os have been processed */
1248	for (i = NVMET_NR_QUEUES; i >= 0; i--) {
1249		if (assoc->queues[i])
1250			flush_workqueue(assoc->queues[i]->work_q);
1251	}
1252
1253	dev_info(tgtport->dev,
1254		"{%d:%d} Association deleted\n",
1255		tgtport->fc_target_port.port_num, assoc->a_id);
1256}
1257
1258static struct nvmet_fc_tgt_assoc *
1259nvmet_fc_find_target_assoc(struct nvmet_fc_tgtport *tgtport,
1260				u64 association_id)
1261{
1262	struct nvmet_fc_tgt_assoc *assoc;
1263	struct nvmet_fc_tgt_assoc *ret = NULL;
1264
1265	rcu_read_lock();
1266	list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
1267		if (association_id == assoc->association_id) {
1268			ret = assoc;
1269			if (!nvmet_fc_tgt_a_get(assoc))
1270				ret = NULL;
1271			break;
1272		}
1273	}
1274	rcu_read_unlock();
1275
1276	return ret;
1277}
1278
1279static void
1280nvmet_fc_portentry_bind(struct nvmet_fc_tgtport *tgtport,
1281			struct nvmet_fc_port_entry *pe,
1282			struct nvmet_port *port)
1283{
1284	lockdep_assert_held(&nvmet_fc_tgtlock);
1285
1286	pe->tgtport = tgtport;
1287	tgtport->pe = pe;
1288
1289	pe->port = port;
1290	port->priv = pe;
1291
1292	pe->node_name = tgtport->fc_target_port.node_name;
1293	pe->port_name = tgtport->fc_target_port.port_name;
1294	INIT_LIST_HEAD(&pe->pe_list);
1295
1296	list_add_tail(&pe->pe_list, &nvmet_fc_portentry_list);
1297}
1298
1299static void
1300nvmet_fc_portentry_unbind(struct nvmet_fc_port_entry *pe)
1301{
1302	unsigned long flags;
1303
1304	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1305	if (pe->tgtport)
1306		pe->tgtport->pe = NULL;
1307	list_del(&pe->pe_list);
1308	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1309}
1310
1311/*
1312 * called when a targetport deregisters. Breaks the relationship
1313 * with the nvmet port, but leaves the port_entry in place so that
1314 * re-registration can resume operation.
1315 */
1316static void
1317nvmet_fc_portentry_unbind_tgt(struct nvmet_fc_tgtport *tgtport)
1318{
1319	struct nvmet_fc_port_entry *pe;
1320	unsigned long flags;
1321
1322	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1323	pe = tgtport->pe;
1324	if (pe)
1325		pe->tgtport = NULL;
1326	tgtport->pe = NULL;
1327	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1328}
1329
1330/*
1331 * called when a new targetport is registered. Looks in the
1332 * existing nvmet port_entries to see if the nvmet layer is
1333 * configured for the targetport's wwn's. (the targetport existed,
1334 * nvmet configured, the lldd unregistered the tgtport, and is now
1335 * reregistering the same targetport).  If so, set the nvmet port
1336 * port entry on the targetport.
1337 */
1338static void
1339nvmet_fc_portentry_rebind_tgt(struct nvmet_fc_tgtport *tgtport)
1340{
1341	struct nvmet_fc_port_entry *pe;
1342	unsigned long flags;
1343
1344	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1345	list_for_each_entry(pe, &nvmet_fc_portentry_list, pe_list) {
1346		if (tgtport->fc_target_port.node_name == pe->node_name &&
1347		    tgtport->fc_target_port.port_name == pe->port_name) {
1348			WARN_ON(pe->tgtport);
1349			tgtport->pe = pe;
1350			pe->tgtport = tgtport;
1351			break;
1352		}
1353	}
1354	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1355}
1356
1357/**
1358 * nvmet_fc_register_targetport - transport entry point called by an
1359 *                              LLDD to register the existence of a local
1360 *                              NVME subystem FC port.
1361 * @pinfo:     pointer to information about the port to be registered
1362 * @template:  LLDD entrypoints and operational parameters for the port
1363 * @dev:       physical hardware device node port corresponds to. Will be
1364 *             used for DMA mappings
1365 * @portptr:   pointer to a local port pointer. Upon success, the routine
1366 *             will allocate a nvme_fc_local_port structure and place its
1367 *             address in the local port pointer. Upon failure, local port
1368 *             pointer will be set to NULL.
1369 *
1370 * Returns:
1371 * a completion status. Must be 0 upon success; a negative errno
1372 * (ex: -ENXIO) upon failure.
1373 */
1374int
1375nvmet_fc_register_targetport(struct nvmet_fc_port_info *pinfo,
1376			struct nvmet_fc_target_template *template,
1377			struct device *dev,
1378			struct nvmet_fc_target_port **portptr)
1379{
1380	struct nvmet_fc_tgtport *newrec;
1381	unsigned long flags;
1382	int ret, idx;
1383
1384	if (!template->xmt_ls_rsp || !template->fcp_op ||
1385	    !template->fcp_abort ||
1386	    !template->fcp_req_release || !template->targetport_delete ||
1387	    !template->max_hw_queues || !template->max_sgl_segments ||
1388	    !template->max_dif_sgl_segments || !template->dma_boundary) {
1389		ret = -EINVAL;
1390		goto out_regtgt_failed;
1391	}
1392
1393	newrec = kzalloc((sizeof(*newrec) + template->target_priv_sz),
1394			 GFP_KERNEL);
1395	if (!newrec) {
1396		ret = -ENOMEM;
1397		goto out_regtgt_failed;
1398	}
1399
1400	idx = ida_alloc(&nvmet_fc_tgtport_cnt, GFP_KERNEL);
1401	if (idx < 0) {
1402		ret = -ENOSPC;
1403		goto out_fail_kfree;
1404	}
1405
1406	if (!get_device(dev) && dev) {
1407		ret = -ENODEV;
1408		goto out_ida_put;
1409	}
1410
1411	newrec->fc_target_port.node_name = pinfo->node_name;
1412	newrec->fc_target_port.port_name = pinfo->port_name;
1413	if (template->target_priv_sz)
1414		newrec->fc_target_port.private = &newrec[1];
1415	else
1416		newrec->fc_target_port.private = NULL;
1417	newrec->fc_target_port.port_id = pinfo->port_id;
1418	newrec->fc_target_port.port_num = idx;
1419	INIT_LIST_HEAD(&newrec->tgt_list);
1420	newrec->dev = dev;
1421	newrec->ops = template;
1422	spin_lock_init(&newrec->lock);
1423	INIT_LIST_HEAD(&newrec->ls_rcv_list);
1424	INIT_LIST_HEAD(&newrec->ls_req_list);
1425	INIT_LIST_HEAD(&newrec->ls_busylist);
1426	INIT_LIST_HEAD(&newrec->assoc_list);
1427	INIT_LIST_HEAD(&newrec->host_list);
1428	kref_init(&newrec->ref);
1429	ida_init(&newrec->assoc_cnt);
1430	newrec->max_sg_cnt = template->max_sgl_segments;
1431	INIT_WORK(&newrec->put_work, nvmet_fc_put_tgtport_work);
1432
1433	ret = nvmet_fc_alloc_ls_iodlist(newrec);
1434	if (ret) {
1435		ret = -ENOMEM;
1436		goto out_free_newrec;
1437	}
1438
1439	nvmet_fc_portentry_rebind_tgt(newrec);
1440
1441	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1442	list_add_tail(&newrec->tgt_list, &nvmet_fc_target_list);
1443	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1444
1445	*portptr = &newrec->fc_target_port;
1446	return 0;
1447
1448out_free_newrec:
1449	put_device(dev);
1450out_ida_put:
1451	ida_free(&nvmet_fc_tgtport_cnt, idx);
1452out_fail_kfree:
1453	kfree(newrec);
1454out_regtgt_failed:
1455	*portptr = NULL;
1456	return ret;
1457}
1458EXPORT_SYMBOL_GPL(nvmet_fc_register_targetport);
1459
1460
1461static void
1462nvmet_fc_free_tgtport(struct kref *ref)
1463{
1464	struct nvmet_fc_tgtport *tgtport =
1465		container_of(ref, struct nvmet_fc_tgtport, ref);
1466	struct device *dev = tgtport->dev;
1467	unsigned long flags;
1468
1469	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1470	list_del(&tgtport->tgt_list);
1471	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1472
1473	nvmet_fc_free_ls_iodlist(tgtport);
1474
1475	/* let the LLDD know we've finished tearing it down */
1476	tgtport->ops->targetport_delete(&tgtport->fc_target_port);
1477
1478	ida_free(&nvmet_fc_tgtport_cnt,
1479			tgtport->fc_target_port.port_num);
1480
1481	ida_destroy(&tgtport->assoc_cnt);
1482
1483	kfree(tgtport);
1484
1485	put_device(dev);
1486}
1487
1488static void
1489nvmet_fc_tgtport_put(struct nvmet_fc_tgtport *tgtport)
1490{
1491	kref_put(&tgtport->ref, nvmet_fc_free_tgtport);
1492}
1493
1494static int
1495nvmet_fc_tgtport_get(struct nvmet_fc_tgtport *tgtport)
1496{
1497	return kref_get_unless_zero(&tgtport->ref);
1498}
1499
1500static void
1501__nvmet_fc_free_assocs(struct nvmet_fc_tgtport *tgtport)
1502{
1503	struct nvmet_fc_tgt_assoc *assoc;
1504
1505	rcu_read_lock();
1506	list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
1507		if (!nvmet_fc_tgt_a_get(assoc))
1508			continue;
1509		nvmet_fc_schedule_delete_assoc(assoc);
1510		nvmet_fc_tgt_a_put(assoc);
1511	}
1512	rcu_read_unlock();
1513}
1514
1515/**
1516 * nvmet_fc_invalidate_host - transport entry point called by an LLDD
1517 *                       to remove references to a hosthandle for LS's.
1518 *
1519 * The nvmet-fc layer ensures that any references to the hosthandle
1520 * on the targetport are forgotten (set to NULL).  The LLDD will
1521 * typically call this when a login with a remote host port has been
1522 * lost, thus LS's for the remote host port are no longer possible.
1523 *
1524 * If an LS request is outstanding to the targetport/hosthandle (or
1525 * issued concurrently with the call to invalidate the host), the
1526 * LLDD is responsible for terminating/aborting the LS and completing
1527 * the LS request. It is recommended that these terminations/aborts
1528 * occur after calling to invalidate the host handle to avoid additional
1529 * retries by the nvmet-fc transport. The nvmet-fc transport may
1530 * continue to reference host handle while it cleans up outstanding
1531 * NVME associations. The nvmet-fc transport will call the
1532 * ops->host_release() callback to notify the LLDD that all references
1533 * are complete and the related host handle can be recovered.
1534 * Note: if there are no references, the callback may be called before
1535 * the invalidate host call returns.
1536 *
1537 * @target_port: pointer to the (registered) target port that a prior
1538 *              LS was received on and which supplied the transport the
1539 *              hosthandle.
1540 * @hosthandle: the handle (pointer) that represents the host port
1541 *              that no longer has connectivity and that LS's should
1542 *              no longer be directed to.
1543 */
1544void
1545nvmet_fc_invalidate_host(struct nvmet_fc_target_port *target_port,
1546			void *hosthandle)
1547{
1548	struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port);
1549	struct nvmet_fc_tgt_assoc *assoc, *next;
1550	unsigned long flags;
1551	bool noassoc = true;
1552
1553	spin_lock_irqsave(&tgtport->lock, flags);
1554	list_for_each_entry_safe(assoc, next,
1555				&tgtport->assoc_list, a_list) {
1556		if (!assoc->hostport ||
1557		    assoc->hostport->hosthandle != hosthandle)
1558			continue;
1559		if (!nvmet_fc_tgt_a_get(assoc))
1560			continue;
1561		assoc->hostport->invalid = 1;
1562		noassoc = false;
1563		nvmet_fc_schedule_delete_assoc(assoc);
1564		nvmet_fc_tgt_a_put(assoc);
1565	}
1566	spin_unlock_irqrestore(&tgtport->lock, flags);
1567
1568	/* if there's nothing to wait for - call the callback */
1569	if (noassoc && tgtport->ops->host_release)
1570		tgtport->ops->host_release(hosthandle);
1571}
1572EXPORT_SYMBOL_GPL(nvmet_fc_invalidate_host);
1573
1574/*
1575 * nvmet layer has called to terminate an association
1576 */
1577static void
1578nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl)
1579{
1580	struct nvmet_fc_tgtport *tgtport, *next;
1581	struct nvmet_fc_tgt_assoc *assoc;
1582	struct nvmet_fc_tgt_queue *queue;
1583	unsigned long flags;
1584	bool found_ctrl = false;
1585
1586	/* this is a bit ugly, but don't want to make locks layered */
1587	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1588	list_for_each_entry_safe(tgtport, next, &nvmet_fc_target_list,
1589			tgt_list) {
1590		if (!nvmet_fc_tgtport_get(tgtport))
1591			continue;
1592		spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1593
1594		rcu_read_lock();
1595		list_for_each_entry_rcu(assoc, &tgtport->assoc_list, a_list) {
1596			queue = assoc->queues[0];
1597			if (queue && queue->nvme_sq.ctrl == ctrl) {
1598				if (nvmet_fc_tgt_a_get(assoc))
1599					found_ctrl = true;
1600				break;
1601			}
1602		}
1603		rcu_read_unlock();
1604
1605		nvmet_fc_tgtport_put(tgtport);
1606
1607		if (found_ctrl) {
1608			nvmet_fc_schedule_delete_assoc(assoc);
1609			nvmet_fc_tgt_a_put(assoc);
1610			return;
1611		}
1612
1613		spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
1614	}
1615	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
1616}
1617
1618/**
1619 * nvmet_fc_unregister_targetport - transport entry point called by an
1620 *                              LLDD to deregister/remove a previously
1621 *                              registered a local NVME subsystem FC port.
1622 * @target_port: pointer to the (registered) target port that is to be
1623 *               deregistered.
1624 *
1625 * Returns:
1626 * a completion status. Must be 0 upon success; a negative errno
1627 * (ex: -ENXIO) upon failure.
1628 */
1629int
1630nvmet_fc_unregister_targetport(struct nvmet_fc_target_port *target_port)
1631{
1632	struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port);
1633
1634	nvmet_fc_portentry_unbind_tgt(tgtport);
1635
1636	/* terminate any outstanding associations */
1637	__nvmet_fc_free_assocs(tgtport);
1638
1639	flush_workqueue(nvmet_wq);
1640
1641	/*
1642	 * should terminate LS's as well. However, LS's will be generated
1643	 * at the tail end of association termination, so they likely don't
1644	 * exist yet. And even if they did, it's worthwhile to just let
1645	 * them finish and targetport ref counting will clean things up.
1646	 */
1647
1648	nvmet_fc_tgtport_put(tgtport);
1649
1650	return 0;
1651}
1652EXPORT_SYMBOL_GPL(nvmet_fc_unregister_targetport);
1653
1654
1655/* ********************** FC-NVME LS RCV Handling ************************* */
1656
1657
1658static void
1659nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport,
1660			struct nvmet_fc_ls_iod *iod)
1661{
1662	struct fcnvme_ls_cr_assoc_rqst *rqst = &iod->rqstbuf->rq_cr_assoc;
1663	struct fcnvme_ls_cr_assoc_acc *acc = &iod->rspbuf->rsp_cr_assoc;
1664	struct nvmet_fc_tgt_queue *queue;
1665	int ret = 0;
1666
1667	memset(acc, 0, sizeof(*acc));
1668
1669	/*
1670	 * FC-NVME spec changes. There are initiators sending different
1671	 * lengths as padding sizes for Create Association Cmd descriptor
1672	 * was incorrect.
1673	 * Accept anything of "minimum" length. Assume format per 1.15
1674	 * spec (with HOSTID reduced to 16 bytes), ignore how long the
1675	 * trailing pad length is.
1676	 */
1677	if (iod->rqstdatalen < FCNVME_LSDESC_CRA_RQST_MINLEN)
1678		ret = VERR_CR_ASSOC_LEN;
1679	else if (be32_to_cpu(rqst->desc_list_len) <
1680			FCNVME_LSDESC_CRA_RQST_MIN_LISTLEN)
1681		ret = VERR_CR_ASSOC_RQST_LEN;
1682	else if (rqst->assoc_cmd.desc_tag !=
1683			cpu_to_be32(FCNVME_LSDESC_CREATE_ASSOC_CMD))
1684		ret = VERR_CR_ASSOC_CMD;
1685	else if (be32_to_cpu(rqst->assoc_cmd.desc_len) <
1686			FCNVME_LSDESC_CRA_CMD_DESC_MIN_DESCLEN)
1687		ret = VERR_CR_ASSOC_CMD_LEN;
1688	else if (!rqst->assoc_cmd.ersp_ratio ||
1689		 (be16_to_cpu(rqst->assoc_cmd.ersp_ratio) >=
1690				be16_to_cpu(rqst->assoc_cmd.sqsize)))
1691		ret = VERR_ERSP_RATIO;
1692
1693	else {
1694		/* new association w/ admin queue */
1695		iod->assoc = nvmet_fc_alloc_target_assoc(
1696						tgtport, iod->hosthandle);
1697		if (!iod->assoc)
1698			ret = VERR_ASSOC_ALLOC_FAIL;
1699		else {
1700			queue = nvmet_fc_alloc_target_queue(iod->assoc, 0,
1701					be16_to_cpu(rqst->assoc_cmd.sqsize));
1702			if (!queue) {
1703				ret = VERR_QUEUE_ALLOC_FAIL;
1704				nvmet_fc_tgt_a_put(iod->assoc);
1705			}
1706		}
1707	}
1708
1709	if (ret) {
1710		dev_err(tgtport->dev,
1711			"Create Association LS failed: %s\n",
1712			validation_errors[ret]);
1713		iod->lsrsp->rsplen = nvme_fc_format_rjt(acc,
1714				sizeof(*acc), rqst->w0.ls_cmd,
1715				FCNVME_RJT_RC_LOGIC,
1716				FCNVME_RJT_EXP_NONE, 0);
1717		return;
1718	}
1719
1720	queue->ersp_ratio = be16_to_cpu(rqst->assoc_cmd.ersp_ratio);
1721	atomic_set(&queue->connected, 1);
1722	queue->sqhd = 0;	/* best place to init value */
1723
1724	dev_info(tgtport->dev,
1725		"{%d:%d} Association created\n",
1726		tgtport->fc_target_port.port_num, iod->assoc->a_id);
1727
1728	/* format a response */
1729
1730	iod->lsrsp->rsplen = sizeof(*acc);
1731
1732	nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
1733			fcnvme_lsdesc_len(
1734				sizeof(struct fcnvme_ls_cr_assoc_acc)),
1735			FCNVME_LS_CREATE_ASSOCIATION);
1736	acc->associd.desc_tag = cpu_to_be32(FCNVME_LSDESC_ASSOC_ID);
1737	acc->associd.desc_len =
1738			fcnvme_lsdesc_len(
1739				sizeof(struct fcnvme_lsdesc_assoc_id));
1740	acc->associd.association_id =
1741			cpu_to_be64(nvmet_fc_makeconnid(iod->assoc, 0));
1742	acc->connectid.desc_tag = cpu_to_be32(FCNVME_LSDESC_CONN_ID);
1743	acc->connectid.desc_len =
1744			fcnvme_lsdesc_len(
1745				sizeof(struct fcnvme_lsdesc_conn_id));
1746	acc->connectid.connection_id = acc->associd.association_id;
1747}
1748
1749static void
1750nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport,
1751			struct nvmet_fc_ls_iod *iod)
1752{
1753	struct fcnvme_ls_cr_conn_rqst *rqst = &iod->rqstbuf->rq_cr_conn;
1754	struct fcnvme_ls_cr_conn_acc *acc = &iod->rspbuf->rsp_cr_conn;
1755	struct nvmet_fc_tgt_queue *queue;
1756	int ret = 0;
1757
1758	memset(acc, 0, sizeof(*acc));
1759
1760	if (iod->rqstdatalen < sizeof(struct fcnvme_ls_cr_conn_rqst))
1761		ret = VERR_CR_CONN_LEN;
1762	else if (rqst->desc_list_len !=
1763			fcnvme_lsdesc_len(
1764				sizeof(struct fcnvme_ls_cr_conn_rqst)))
1765		ret = VERR_CR_CONN_RQST_LEN;
1766	else if (rqst->associd.desc_tag != cpu_to_be32(FCNVME_LSDESC_ASSOC_ID))
1767		ret = VERR_ASSOC_ID;
1768	else if (rqst->associd.desc_len !=
1769			fcnvme_lsdesc_len(
1770				sizeof(struct fcnvme_lsdesc_assoc_id)))
1771		ret = VERR_ASSOC_ID_LEN;
1772	else if (rqst->connect_cmd.desc_tag !=
1773			cpu_to_be32(FCNVME_LSDESC_CREATE_CONN_CMD))
1774		ret = VERR_CR_CONN_CMD;
1775	else if (rqst->connect_cmd.desc_len !=
1776			fcnvme_lsdesc_len(
1777				sizeof(struct fcnvme_lsdesc_cr_conn_cmd)))
1778		ret = VERR_CR_CONN_CMD_LEN;
1779	else if (!rqst->connect_cmd.ersp_ratio ||
1780		 (be16_to_cpu(rqst->connect_cmd.ersp_ratio) >=
1781				be16_to_cpu(rqst->connect_cmd.sqsize)))
1782		ret = VERR_ERSP_RATIO;
1783
1784	else {
1785		/* new io queue */
1786		iod->assoc = nvmet_fc_find_target_assoc(tgtport,
1787				be64_to_cpu(rqst->associd.association_id));
1788		if (!iod->assoc)
1789			ret = VERR_NO_ASSOC;
1790		else {
1791			queue = nvmet_fc_alloc_target_queue(iod->assoc,
1792					be16_to_cpu(rqst->connect_cmd.qid),
1793					be16_to_cpu(rqst->connect_cmd.sqsize));
1794			if (!queue)
1795				ret = VERR_QUEUE_ALLOC_FAIL;
1796
1797			/* release get taken in nvmet_fc_find_target_assoc */
1798			nvmet_fc_tgt_a_put(iod->assoc);
1799		}
1800	}
1801
1802	if (ret) {
1803		dev_err(tgtport->dev,
1804			"Create Connection LS failed: %s\n",
1805			validation_errors[ret]);
1806		iod->lsrsp->rsplen = nvme_fc_format_rjt(acc,
1807				sizeof(*acc), rqst->w0.ls_cmd,
1808				(ret == VERR_NO_ASSOC) ?
1809					FCNVME_RJT_RC_INV_ASSOC :
1810					FCNVME_RJT_RC_LOGIC,
1811				FCNVME_RJT_EXP_NONE, 0);
1812		return;
1813	}
1814
1815	queue->ersp_ratio = be16_to_cpu(rqst->connect_cmd.ersp_ratio);
1816	atomic_set(&queue->connected, 1);
1817	queue->sqhd = 0;	/* best place to init value */
1818
1819	/* format a response */
1820
1821	iod->lsrsp->rsplen = sizeof(*acc);
1822
1823	nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
1824			fcnvme_lsdesc_len(sizeof(struct fcnvme_ls_cr_conn_acc)),
1825			FCNVME_LS_CREATE_CONNECTION);
1826	acc->connectid.desc_tag = cpu_to_be32(FCNVME_LSDESC_CONN_ID);
1827	acc->connectid.desc_len =
1828			fcnvme_lsdesc_len(
1829				sizeof(struct fcnvme_lsdesc_conn_id));
1830	acc->connectid.connection_id =
1831			cpu_to_be64(nvmet_fc_makeconnid(iod->assoc,
1832				be16_to_cpu(rqst->connect_cmd.qid)));
1833}
1834
1835/*
1836 * Returns true if the LS response is to be transmit
1837 * Returns false if the LS response is to be delayed
1838 */
1839static int
1840nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
1841			struct nvmet_fc_ls_iod *iod)
1842{
1843	struct fcnvme_ls_disconnect_assoc_rqst *rqst =
1844						&iod->rqstbuf->rq_dis_assoc;
1845	struct fcnvme_ls_disconnect_assoc_acc *acc =
1846						&iod->rspbuf->rsp_dis_assoc;
1847	struct nvmet_fc_tgt_assoc *assoc = NULL;
1848	struct nvmet_fc_ls_iod *oldls = NULL;
1849	unsigned long flags;
1850	int ret = 0;
1851
1852	memset(acc, 0, sizeof(*acc));
1853
1854	ret = nvmefc_vldt_lsreq_discon_assoc(iod->rqstdatalen, rqst);
1855	if (!ret) {
1856		/* match an active association - takes an assoc ref if !NULL */
1857		assoc = nvmet_fc_find_target_assoc(tgtport,
1858				be64_to_cpu(rqst->associd.association_id));
1859		iod->assoc = assoc;
1860		if (!assoc)
1861			ret = VERR_NO_ASSOC;
1862	}
1863
1864	if (ret || !assoc) {
1865		dev_err(tgtport->dev,
1866			"Disconnect LS failed: %s\n",
1867			validation_errors[ret]);
1868		iod->lsrsp->rsplen = nvme_fc_format_rjt(acc,
1869				sizeof(*acc), rqst->w0.ls_cmd,
1870				(ret == VERR_NO_ASSOC) ?
1871					FCNVME_RJT_RC_INV_ASSOC :
1872					FCNVME_RJT_RC_LOGIC,
1873				FCNVME_RJT_EXP_NONE, 0);
1874		return true;
1875	}
1876
1877	/* format a response */
1878
1879	iod->lsrsp->rsplen = sizeof(*acc);
1880
1881	nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
1882			fcnvme_lsdesc_len(
1883				sizeof(struct fcnvme_ls_disconnect_assoc_acc)),
1884			FCNVME_LS_DISCONNECT_ASSOC);
1885
1886	/*
1887	 * The rules for LS response says the response cannot
1888	 * go back until ABTS's have been sent for all outstanding
1889	 * I/O and a Disconnect Association LS has been sent.
1890	 * So... save off the Disconnect LS to send the response
1891	 * later. If there was a prior LS already saved, replace
1892	 * it with the newer one and send a can't perform reject
1893	 * on the older one.
1894	 */
1895	spin_lock_irqsave(&tgtport->lock, flags);
1896	oldls = assoc->rcv_disconn;
1897	assoc->rcv_disconn = iod;
1898	spin_unlock_irqrestore(&tgtport->lock, flags);
1899
1900	if (oldls) {
1901		dev_info(tgtport->dev,
1902			"{%d:%d} Multiple Disconnect Association LS's "
1903			"received\n",
1904			tgtport->fc_target_port.port_num, assoc->a_id);
1905		/* overwrite good response with bogus failure */
1906		oldls->lsrsp->rsplen = nvme_fc_format_rjt(oldls->rspbuf,
1907						sizeof(*iod->rspbuf),
1908						/* ok to use rqst, LS is same */
1909						rqst->w0.ls_cmd,
1910						FCNVME_RJT_RC_UNAB,
1911						FCNVME_RJT_EXP_NONE, 0);
1912		nvmet_fc_xmt_ls_rsp(tgtport, oldls);
1913	}
1914
1915	nvmet_fc_schedule_delete_assoc(assoc);
1916	nvmet_fc_tgt_a_put(assoc);
1917
1918	return false;
1919}
1920
1921
1922/* *********************** NVME Ctrl Routines **************************** */
1923
1924
1925static void nvmet_fc_fcp_nvme_cmd_done(struct nvmet_req *nvme_req);
1926
1927static const struct nvmet_fabrics_ops nvmet_fc_tgt_fcp_ops;
1928
1929static void
1930nvmet_fc_xmt_ls_rsp_done(struct nvmefc_ls_rsp *lsrsp)
1931{
1932	struct nvmet_fc_ls_iod *iod = lsrsp->nvme_fc_private;
1933	struct nvmet_fc_tgtport *tgtport = iod->tgtport;
1934
1935	fc_dma_sync_single_for_cpu(tgtport->dev, iod->rspdma,
1936				sizeof(*iod->rspbuf), DMA_TO_DEVICE);
1937	nvmet_fc_free_ls_iod(tgtport, iod);
1938	nvmet_fc_tgtport_put(tgtport);
1939}
1940
1941static void
1942nvmet_fc_xmt_ls_rsp(struct nvmet_fc_tgtport *tgtport,
1943				struct nvmet_fc_ls_iod *iod)
1944{
1945	int ret;
1946
1947	fc_dma_sync_single_for_device(tgtport->dev, iod->rspdma,
1948				  sizeof(*iod->rspbuf), DMA_TO_DEVICE);
1949
1950	ret = tgtport->ops->xmt_ls_rsp(&tgtport->fc_target_port, iod->lsrsp);
1951	if (ret)
1952		nvmet_fc_xmt_ls_rsp_done(iod->lsrsp);
1953}
1954
1955/*
1956 * Actual processing routine for received FC-NVME LS Requests from the LLD
1957 */
1958static void
1959nvmet_fc_handle_ls_rqst(struct nvmet_fc_tgtport *tgtport,
1960			struct nvmet_fc_ls_iod *iod)
1961{
1962	struct fcnvme_ls_rqst_w0 *w0 = &iod->rqstbuf->rq_cr_assoc.w0;
1963	bool sendrsp = true;
1964
1965	iod->lsrsp->nvme_fc_private = iod;
1966	iod->lsrsp->rspbuf = iod->rspbuf;
1967	iod->lsrsp->rspdma = iod->rspdma;
1968	iod->lsrsp->done = nvmet_fc_xmt_ls_rsp_done;
1969	/* Be preventative. handlers will later set to valid length */
1970	iod->lsrsp->rsplen = 0;
1971
1972	iod->assoc = NULL;
1973
1974	/*
1975	 * handlers:
1976	 *   parse request input, execute the request, and format the
1977	 *   LS response
1978	 */
1979	switch (w0->ls_cmd) {
1980	case FCNVME_LS_CREATE_ASSOCIATION:
1981		/* Creates Association and initial Admin Queue/Connection */
1982		nvmet_fc_ls_create_association(tgtport, iod);
1983		break;
1984	case FCNVME_LS_CREATE_CONNECTION:
1985		/* Creates an IO Queue/Connection */
1986		nvmet_fc_ls_create_connection(tgtport, iod);
1987		break;
1988	case FCNVME_LS_DISCONNECT_ASSOC:
1989		/* Terminate a Queue/Connection or the Association */
1990		sendrsp = nvmet_fc_ls_disconnect(tgtport, iod);
1991		break;
1992	default:
1993		iod->lsrsp->rsplen = nvme_fc_format_rjt(iod->rspbuf,
1994				sizeof(*iod->rspbuf), w0->ls_cmd,
1995				FCNVME_RJT_RC_INVAL, FCNVME_RJT_EXP_NONE, 0);
1996	}
1997
1998	if (sendrsp)
1999		nvmet_fc_xmt_ls_rsp(tgtport, iod);
2000}
2001
2002/*
2003 * Actual processing routine for received FC-NVME LS Requests from the LLD
2004 */
2005static void
2006nvmet_fc_handle_ls_rqst_work(struct work_struct *work)
2007{
2008	struct nvmet_fc_ls_iod *iod =
2009		container_of(work, struct nvmet_fc_ls_iod, work);
2010	struct nvmet_fc_tgtport *tgtport = iod->tgtport;
2011
2012	nvmet_fc_handle_ls_rqst(tgtport, iod);
2013}
2014
2015
2016/**
2017 * nvmet_fc_rcv_ls_req - transport entry point called by an LLDD
2018 *                       upon the reception of a NVME LS request.
2019 *
2020 * The nvmet-fc layer will copy payload to an internal structure for
2021 * processing.  As such, upon completion of the routine, the LLDD may
2022 * immediately free/reuse the LS request buffer passed in the call.
2023 *
2024 * If this routine returns error, the LLDD should abort the exchange.
2025 *
2026 * @target_port: pointer to the (registered) target port the LS was
2027 *              received on.
2028 * @hosthandle: pointer to the host specific data, gets stored in iod.
2029 * @lsrsp:      pointer to a lsrsp structure to be used to reference
2030 *              the exchange corresponding to the LS.
2031 * @lsreqbuf:   pointer to the buffer containing the LS Request
2032 * @lsreqbuf_len: length, in bytes, of the received LS request
2033 */
2034int
2035nvmet_fc_rcv_ls_req(struct nvmet_fc_target_port *target_port,
2036			void *hosthandle,
2037			struct nvmefc_ls_rsp *lsrsp,
2038			void *lsreqbuf, u32 lsreqbuf_len)
2039{
2040	struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port);
2041	struct nvmet_fc_ls_iod *iod;
2042	struct fcnvme_ls_rqst_w0 *w0 = (struct fcnvme_ls_rqst_w0 *)lsreqbuf;
2043
2044	if (lsreqbuf_len > sizeof(union nvmefc_ls_requests)) {
2045		dev_info(tgtport->dev,
2046			"RCV %s LS failed: payload too large (%d)\n",
2047			(w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ?
2048				nvmefc_ls_names[w0->ls_cmd] : "",
2049			lsreqbuf_len);
2050		return -E2BIG;
2051	}
2052
2053	if (!nvmet_fc_tgtport_get(tgtport)) {
2054		dev_info(tgtport->dev,
2055			"RCV %s LS failed: target deleting\n",
2056			(w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ?
2057				nvmefc_ls_names[w0->ls_cmd] : "");
2058		return -ESHUTDOWN;
2059	}
2060
2061	iod = nvmet_fc_alloc_ls_iod(tgtport);
2062	if (!iod) {
2063		dev_info(tgtport->dev,
2064			"RCV %s LS failed: context allocation failed\n",
2065			(w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ?
2066				nvmefc_ls_names[w0->ls_cmd] : "");
2067		nvmet_fc_tgtport_put(tgtport);
2068		return -ENOENT;
2069	}
2070
2071	iod->lsrsp = lsrsp;
2072	iod->fcpreq = NULL;
2073	memcpy(iod->rqstbuf, lsreqbuf, lsreqbuf_len);
2074	iod->rqstdatalen = lsreqbuf_len;
2075	iod->hosthandle = hosthandle;
2076
2077	queue_work(nvmet_wq, &iod->work);
2078
2079	return 0;
2080}
2081EXPORT_SYMBOL_GPL(nvmet_fc_rcv_ls_req);
2082
2083
2084/*
2085 * **********************
2086 * Start of FCP handling
2087 * **********************
2088 */
2089
2090static int
2091nvmet_fc_alloc_tgt_pgs(struct nvmet_fc_fcp_iod *fod)
2092{
2093	struct scatterlist *sg;
2094	unsigned int nent;
2095
2096	sg = sgl_alloc(fod->req.transfer_len, GFP_KERNEL, &nent);
2097	if (!sg)
2098		goto out;
2099
2100	fod->data_sg = sg;
2101	fod->data_sg_cnt = nent;
2102	fod->data_sg_cnt = fc_dma_map_sg(fod->tgtport->dev, sg, nent,
2103				((fod->io_dir == NVMET_FCP_WRITE) ?
2104					DMA_FROM_DEVICE : DMA_TO_DEVICE));
2105				/* note: write from initiator perspective */
2106	fod->next_sg = fod->data_sg;
2107
2108	return 0;
2109
2110out:
2111	return NVME_SC_INTERNAL;
2112}
2113
2114static void
2115nvmet_fc_free_tgt_pgs(struct nvmet_fc_fcp_iod *fod)
2116{
2117	if (!fod->data_sg || !fod->data_sg_cnt)
2118		return;
2119
2120	fc_dma_unmap_sg(fod->tgtport->dev, fod->data_sg, fod->data_sg_cnt,
2121				((fod->io_dir == NVMET_FCP_WRITE) ?
2122					DMA_FROM_DEVICE : DMA_TO_DEVICE));
2123	sgl_free(fod->data_sg);
2124	fod->data_sg = NULL;
2125	fod->data_sg_cnt = 0;
2126}
2127
2128
2129static bool
2130queue_90percent_full(struct nvmet_fc_tgt_queue *q, u32 sqhd)
2131{
2132	u32 sqtail, used;
2133
2134	/* egad, this is ugly. And sqtail is just a best guess */
2135	sqtail = atomic_read(&q->sqtail) % q->sqsize;
2136
2137	used = (sqtail < sqhd) ? (sqtail + q->sqsize - sqhd) : (sqtail - sqhd);
2138	return ((used * 10) >= (((u32)(q->sqsize - 1) * 9)));
2139}
2140
2141/*
2142 * Prep RSP payload.
2143 * May be a NVMET_FCOP_RSP or NVMET_FCOP_READDATA_RSP op
2144 */
2145static void
2146nvmet_fc_prep_fcp_rsp(struct nvmet_fc_tgtport *tgtport,
2147				struct nvmet_fc_fcp_iod *fod)
2148{
2149	struct nvme_fc_ersp_iu *ersp = &fod->rspiubuf;
2150	struct nvme_common_command *sqe = &fod->cmdiubuf.sqe.common;
2151	struct nvme_completion *cqe = &ersp->cqe;
2152	u32 *cqewd = (u32 *)cqe;
2153	bool send_ersp = false;
2154	u32 rsn, rspcnt, xfr_length;
2155
2156	if (fod->fcpreq->op == NVMET_FCOP_READDATA_RSP)
2157		xfr_length = fod->req.transfer_len;
2158	else
2159		xfr_length = fod->offset;
2160
2161	/*
2162	 * check to see if we can send a 0's rsp.
2163	 *   Note: to send a 0's response, the NVME-FC host transport will
2164	 *   recreate the CQE. The host transport knows: sq id, SQHD (last
2165	 *   seen in an ersp), and command_id. Thus it will create a
2166	 *   zero-filled CQE with those known fields filled in. Transport
2167	 *   must send an ersp for any condition where the cqe won't match
2168	 *   this.
2169	 *
2170	 * Here are the FC-NVME mandated cases where we must send an ersp:
2171	 *  every N responses, where N=ersp_ratio
2172	 *  force fabric commands to send ersp's (not in FC-NVME but good
2173	 *    practice)
2174	 *  normal cmds: any time status is non-zero, or status is zero
2175	 *     but words 0 or 1 are non-zero.
2176	 *  the SQ is 90% or more full
2177	 *  the cmd is a fused command
2178	 *  transferred data length not equal to cmd iu length
2179	 */
2180	rspcnt = atomic_inc_return(&fod->queue->zrspcnt);
2181	if (!(rspcnt % fod->queue->ersp_ratio) ||
2182	    nvme_is_fabrics((struct nvme_command *) sqe) ||
2183	    xfr_length != fod->req.transfer_len ||
2184	    (le16_to_cpu(cqe->status) & 0xFFFE) || cqewd[0] || cqewd[1] ||
2185	    (sqe->flags & (NVME_CMD_FUSE_FIRST | NVME_CMD_FUSE_SECOND)) ||
2186	    queue_90percent_full(fod->queue, le16_to_cpu(cqe->sq_head)))
2187		send_ersp = true;
2188
2189	/* re-set the fields */
2190	fod->fcpreq->rspaddr = ersp;
2191	fod->fcpreq->rspdma = fod->rspdma;
2192
2193	if (!send_ersp) {
2194		memset(ersp, 0, NVME_FC_SIZEOF_ZEROS_RSP);
2195		fod->fcpreq->rsplen = NVME_FC_SIZEOF_ZEROS_RSP;
2196	} else {
2197		ersp->iu_len = cpu_to_be16(sizeof(*ersp)/sizeof(u32));
2198		rsn = atomic_inc_return(&fod->queue->rsn);
2199		ersp->rsn = cpu_to_be32(rsn);
2200		ersp->xfrd_len = cpu_to_be32(xfr_length);
2201		fod->fcpreq->rsplen = sizeof(*ersp);
2202	}
2203
2204	fc_dma_sync_single_for_device(tgtport->dev, fod->rspdma,
2205				  sizeof(fod->rspiubuf), DMA_TO_DEVICE);
2206}
2207
2208static void nvmet_fc_xmt_fcp_op_done(struct nvmefc_tgt_fcp_req *fcpreq);
2209
2210static void
2211nvmet_fc_abort_op(struct nvmet_fc_tgtport *tgtport,
2212				struct nvmet_fc_fcp_iod *fod)
2213{
2214	struct nvmefc_tgt_fcp_req *fcpreq = fod->fcpreq;
2215
2216	/* data no longer needed */
2217	nvmet_fc_free_tgt_pgs(fod);
2218
2219	/*
2220	 * if an ABTS was received or we issued the fcp_abort early
2221	 * don't call abort routine again.
2222	 */
2223	/* no need to take lock - lock was taken earlier to get here */
2224	if (!fod->aborted)
2225		tgtport->ops->fcp_abort(&tgtport->fc_target_port, fcpreq);
2226
2227	nvmet_fc_free_fcp_iod(fod->queue, fod);
2228}
2229
2230static void
2231nvmet_fc_xmt_fcp_rsp(struct nvmet_fc_tgtport *tgtport,
2232				struct nvmet_fc_fcp_iod *fod)
2233{
2234	int ret;
2235
2236	fod->fcpreq->op = NVMET_FCOP_RSP;
2237	fod->fcpreq->timeout = 0;
2238
2239	nvmet_fc_prep_fcp_rsp(tgtport, fod);
2240
2241	ret = tgtport->ops->fcp_op(&tgtport->fc_target_port, fod->fcpreq);
2242	if (ret)
2243		nvmet_fc_abort_op(tgtport, fod);
2244}
2245
2246static void
2247nvmet_fc_transfer_fcp_data(struct nvmet_fc_tgtport *tgtport,
2248				struct nvmet_fc_fcp_iod *fod, u8 op)
2249{
2250	struct nvmefc_tgt_fcp_req *fcpreq = fod->fcpreq;
2251	struct scatterlist *sg = fod->next_sg;
2252	unsigned long flags;
2253	u32 remaininglen = fod->req.transfer_len - fod->offset;
2254	u32 tlen = 0;
2255	int ret;
2256
2257	fcpreq->op = op;
2258	fcpreq->offset = fod->offset;
2259	fcpreq->timeout = NVME_FC_TGTOP_TIMEOUT_SEC;
2260
2261	/*
2262	 * for next sequence:
2263	 *  break at a sg element boundary
2264	 *  attempt to keep sequence length capped at
2265	 *    NVMET_FC_MAX_SEQ_LENGTH but allow sequence to
2266	 *    be longer if a single sg element is larger
2267	 *    than that amount. This is done to avoid creating
2268	 *    a new sg list to use for the tgtport api.
2269	 */
2270	fcpreq->sg = sg;
2271	fcpreq->sg_cnt = 0;
2272	while (tlen < remaininglen &&
2273	       fcpreq->sg_cnt < tgtport->max_sg_cnt &&
2274	       tlen + sg_dma_len(sg) < NVMET_FC_MAX_SEQ_LENGTH) {
2275		fcpreq->sg_cnt++;
2276		tlen += sg_dma_len(sg);
2277		sg = sg_next(sg);
2278	}
2279	if (tlen < remaininglen && fcpreq->sg_cnt == 0) {
2280		fcpreq->sg_cnt++;
2281		tlen += min_t(u32, sg_dma_len(sg), remaininglen);
2282		sg = sg_next(sg);
2283	}
2284	if (tlen < remaininglen)
2285		fod->next_sg = sg;
2286	else
2287		fod->next_sg = NULL;
2288
2289	fcpreq->transfer_length = tlen;
2290	fcpreq->transferred_length = 0;
2291	fcpreq->fcp_error = 0;
2292	fcpreq->rsplen = 0;
2293
2294	/*
2295	 * If the last READDATA request: check if LLDD supports
2296	 * combined xfr with response.
2297	 */
2298	if ((op == NVMET_FCOP_READDATA) &&
2299	    ((fod->offset + fcpreq->transfer_length) == fod->req.transfer_len) &&
2300	    (tgtport->ops->target_features & NVMET_FCTGTFEAT_READDATA_RSP)) {
2301		fcpreq->op = NVMET_FCOP_READDATA_RSP;
2302		nvmet_fc_prep_fcp_rsp(tgtport, fod);
2303	}
2304
2305	ret = tgtport->ops->fcp_op(&tgtport->fc_target_port, fod->fcpreq);
2306	if (ret) {
2307		/*
2308		 * should be ok to set w/o lock as its in the thread of
2309		 * execution (not an async timer routine) and doesn't
2310		 * contend with any clearing action
2311		 */
2312		fod->abort = true;
2313
2314		if (op == NVMET_FCOP_WRITEDATA) {
2315			spin_lock_irqsave(&fod->flock, flags);
2316			fod->writedataactive = false;
2317			spin_unlock_irqrestore(&fod->flock, flags);
2318			nvmet_req_complete(&fod->req, NVME_SC_INTERNAL);
2319		} else /* NVMET_FCOP_READDATA or NVMET_FCOP_READDATA_RSP */ {
2320			fcpreq->fcp_error = ret;
2321			fcpreq->transferred_length = 0;
2322			nvmet_fc_xmt_fcp_op_done(fod->fcpreq);
2323		}
2324	}
2325}
2326
2327static inline bool
2328__nvmet_fc_fod_op_abort(struct nvmet_fc_fcp_iod *fod, bool abort)
2329{
2330	struct nvmefc_tgt_fcp_req *fcpreq = fod->fcpreq;
2331	struct nvmet_fc_tgtport *tgtport = fod->tgtport;
2332
2333	/* if in the middle of an io and we need to tear down */
2334	if (abort) {
2335		if (fcpreq->op == NVMET_FCOP_WRITEDATA) {
2336			nvmet_req_complete(&fod->req, NVME_SC_INTERNAL);
2337			return true;
2338		}
2339
2340		nvmet_fc_abort_op(tgtport, fod);
2341		return true;
2342	}
2343
2344	return false;
2345}
2346
2347/*
2348 * actual done handler for FCP operations when completed by the lldd
2349 */
2350static void
2351nvmet_fc_fod_op_done(struct nvmet_fc_fcp_iod *fod)
2352{
2353	struct nvmefc_tgt_fcp_req *fcpreq = fod->fcpreq;
2354	struct nvmet_fc_tgtport *tgtport = fod->tgtport;
2355	unsigned long flags;
2356	bool abort;
2357
2358	spin_lock_irqsave(&fod->flock, flags);
2359	abort = fod->abort;
2360	fod->writedataactive = false;
2361	spin_unlock_irqrestore(&fod->flock, flags);
2362
2363	switch (fcpreq->op) {
2364
2365	case NVMET_FCOP_WRITEDATA:
2366		if (__nvmet_fc_fod_op_abort(fod, abort))
2367			return;
2368		if (fcpreq->fcp_error ||
2369		    fcpreq->transferred_length != fcpreq->transfer_length) {
2370			spin_lock_irqsave(&fod->flock, flags);
2371			fod->abort = true;
2372			spin_unlock_irqrestore(&fod->flock, flags);
2373
2374			nvmet_req_complete(&fod->req, NVME_SC_INTERNAL);
2375			return;
2376		}
2377
2378		fod->offset += fcpreq->transferred_length;
2379		if (fod->offset != fod->req.transfer_len) {
2380			spin_lock_irqsave(&fod->flock, flags);
2381			fod->writedataactive = true;
2382			spin_unlock_irqrestore(&fod->flock, flags);
2383
2384			/* transfer the next chunk */
2385			nvmet_fc_transfer_fcp_data(tgtport, fod,
2386						NVMET_FCOP_WRITEDATA);
2387			return;
2388		}
2389
2390		/* data transfer complete, resume with nvmet layer */
2391		fod->req.execute(&fod->req);
2392		break;
2393
2394	case NVMET_FCOP_READDATA:
2395	case NVMET_FCOP_READDATA_RSP:
2396		if (__nvmet_fc_fod_op_abort(fod, abort))
2397			return;
2398		if (fcpreq->fcp_error ||
2399		    fcpreq->transferred_length != fcpreq->transfer_length) {
2400			nvmet_fc_abort_op(tgtport, fod);
2401			return;
2402		}
2403
2404		/* success */
2405
2406		if (fcpreq->op == NVMET_FCOP_READDATA_RSP) {
2407			/* data no longer needed */
2408			nvmet_fc_free_tgt_pgs(fod);
2409			nvmet_fc_free_fcp_iod(fod->queue, fod);
2410			return;
2411		}
2412
2413		fod->offset += fcpreq->transferred_length;
2414		if (fod->offset != fod->req.transfer_len) {
2415			/* transfer the next chunk */
2416			nvmet_fc_transfer_fcp_data(tgtport, fod,
2417						NVMET_FCOP_READDATA);
2418			return;
2419		}
2420
2421		/* data transfer complete, send response */
2422
2423		/* data no longer needed */
2424		nvmet_fc_free_tgt_pgs(fod);
2425
2426		nvmet_fc_xmt_fcp_rsp(tgtport, fod);
2427
2428		break;
2429
2430	case NVMET_FCOP_RSP:
2431		if (__nvmet_fc_fod_op_abort(fod, abort))
2432			return;
2433		nvmet_fc_free_fcp_iod(fod->queue, fod);
2434		break;
2435
2436	default:
2437		break;
2438	}
2439}
2440
2441static void
2442nvmet_fc_xmt_fcp_op_done(struct nvmefc_tgt_fcp_req *fcpreq)
2443{
2444	struct nvmet_fc_fcp_iod *fod = fcpreq->nvmet_fc_private;
2445
2446	nvmet_fc_fod_op_done(fod);
2447}
2448
2449/*
2450 * actual completion handler after execution by the nvmet layer
2451 */
2452static void
2453__nvmet_fc_fcp_nvme_cmd_done(struct nvmet_fc_tgtport *tgtport,
2454			struct nvmet_fc_fcp_iod *fod, int status)
2455{
2456	struct nvme_common_command *sqe = &fod->cmdiubuf.sqe.common;
2457	struct nvme_completion *cqe = &fod->rspiubuf.cqe;
2458	unsigned long flags;
2459	bool abort;
2460
2461	spin_lock_irqsave(&fod->flock, flags);
2462	abort = fod->abort;
2463	spin_unlock_irqrestore(&fod->flock, flags);
2464
2465	/* if we have a CQE, snoop the last sq_head value */
2466	if (!status)
2467		fod->queue->sqhd = cqe->sq_head;
2468
2469	if (abort) {
2470		nvmet_fc_abort_op(tgtport, fod);
2471		return;
2472	}
2473
2474	/* if an error handling the cmd post initial parsing */
2475	if (status) {
2476		/* fudge up a failed CQE status for our transport error */
2477		memset(cqe, 0, sizeof(*cqe));
2478		cqe->sq_head = fod->queue->sqhd;	/* echo last cqe sqhd */
2479		cqe->sq_id = cpu_to_le16(fod->queue->qid);
2480		cqe->command_id = sqe->command_id;
2481		cqe->status = cpu_to_le16(status);
2482	} else {
2483
2484		/*
2485		 * try to push the data even if the SQE status is non-zero.
2486		 * There may be a status where data still was intended to
2487		 * be moved
2488		 */
2489		if ((fod->io_dir == NVMET_FCP_READ) && (fod->data_sg_cnt)) {
2490			/* push the data over before sending rsp */
2491			nvmet_fc_transfer_fcp_data(tgtport, fod,
2492						NVMET_FCOP_READDATA);
2493			return;
2494		}
2495
2496		/* writes & no data - fall thru */
2497	}
2498
2499	/* data no longer needed */
2500	nvmet_fc_free_tgt_pgs(fod);
2501
2502	nvmet_fc_xmt_fcp_rsp(tgtport, fod);
2503}
2504
2505
2506static void
2507nvmet_fc_fcp_nvme_cmd_done(struct nvmet_req *nvme_req)
2508{
2509	struct nvmet_fc_fcp_iod *fod = nvmet_req_to_fod(nvme_req);
2510	struct nvmet_fc_tgtport *tgtport = fod->tgtport;
2511
2512	__nvmet_fc_fcp_nvme_cmd_done(tgtport, fod, 0);
2513}
2514
2515
2516/*
2517 * Actual processing routine for received FC-NVME I/O Requests from the LLD
2518 */
2519static void
2520nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport,
2521			struct nvmet_fc_fcp_iod *fod)
2522{
2523	struct nvme_fc_cmd_iu *cmdiu = &fod->cmdiubuf;
2524	u32 xfrlen = be32_to_cpu(cmdiu->data_len);
2525	int ret;
2526
2527	/*
2528	 * Fused commands are currently not supported in the linux
2529	 * implementation.
2530	 *
2531	 * As such, the implementation of the FC transport does not
2532	 * look at the fused commands and order delivery to the upper
2533	 * layer until we have both based on csn.
2534	 */
2535
2536	fod->fcpreq->done = nvmet_fc_xmt_fcp_op_done;
2537
2538	if (cmdiu->flags & FCNVME_CMD_FLAGS_WRITE) {
2539		fod->io_dir = NVMET_FCP_WRITE;
2540		if (!nvme_is_write(&cmdiu->sqe))
2541			goto transport_error;
2542	} else if (cmdiu->flags & FCNVME_CMD_FLAGS_READ) {
2543		fod->io_dir = NVMET_FCP_READ;
2544		if (nvme_is_write(&cmdiu->sqe))
2545			goto transport_error;
2546	} else {
2547		fod->io_dir = NVMET_FCP_NODATA;
2548		if (xfrlen)
2549			goto transport_error;
2550	}
2551
2552	fod->req.cmd = &fod->cmdiubuf.sqe;
2553	fod->req.cqe = &fod->rspiubuf.cqe;
2554	if (!tgtport->pe)
2555		goto transport_error;
2556	fod->req.port = tgtport->pe->port;
2557
2558	/* clear any response payload */
2559	memset(&fod->rspiubuf, 0, sizeof(fod->rspiubuf));
2560
2561	fod->data_sg = NULL;
2562	fod->data_sg_cnt = 0;
2563
2564	ret = nvmet_req_init(&fod->req,
2565				&fod->queue->nvme_cq,
2566				&fod->queue->nvme_sq,
2567				&nvmet_fc_tgt_fcp_ops);
2568	if (!ret) {
2569		/* bad SQE content or invalid ctrl state */
2570		/* nvmet layer has already called op done to send rsp. */
2571		return;
2572	}
2573
2574	fod->req.transfer_len = xfrlen;
2575
2576	/* keep a running counter of tail position */
2577	atomic_inc(&fod->queue->sqtail);
2578
2579	if (fod->req.transfer_len) {
2580		ret = nvmet_fc_alloc_tgt_pgs(fod);
2581		if (ret) {
2582			nvmet_req_complete(&fod->req, ret);
2583			return;
2584		}
2585	}
2586	fod->req.sg = fod->data_sg;
2587	fod->req.sg_cnt = fod->data_sg_cnt;
2588	fod->offset = 0;
2589
2590	if (fod->io_dir == NVMET_FCP_WRITE) {
2591		/* pull the data over before invoking nvmet layer */
2592		nvmet_fc_transfer_fcp_data(tgtport, fod, NVMET_FCOP_WRITEDATA);
2593		return;
2594	}
2595
2596	/*
2597	 * Reads or no data:
2598	 *
2599	 * can invoke the nvmet_layer now. If read data, cmd completion will
2600	 * push the data
2601	 */
2602	fod->req.execute(&fod->req);
2603	return;
2604
2605transport_error:
2606	nvmet_fc_abort_op(tgtport, fod);
2607}
2608
2609/**
2610 * nvmet_fc_rcv_fcp_req - transport entry point called by an LLDD
2611 *                       upon the reception of a NVME FCP CMD IU.
2612 *
2613 * Pass a FC-NVME FCP CMD IU received from the FC link to the nvmet-fc
2614 * layer for processing.
2615 *
2616 * The nvmet_fc layer allocates a local job structure (struct
2617 * nvmet_fc_fcp_iod) from the queue for the io and copies the
2618 * CMD IU buffer to the job structure. As such, on a successful
2619 * completion (returns 0), the LLDD may immediately free/reuse
2620 * the CMD IU buffer passed in the call.
2621 *
2622 * However, in some circumstances, due to the packetized nature of FC
2623 * and the api of the FC LLDD which may issue a hw command to send the
2624 * response, but the LLDD may not get the hw completion for that command
2625 * and upcall the nvmet_fc layer before a new command may be
2626 * asynchronously received - its possible for a command to be received
2627 * before the LLDD and nvmet_fc have recycled the job structure. It gives
2628 * the appearance of more commands received than fits in the sq.
2629 * To alleviate this scenario, a temporary queue is maintained in the
2630 * transport for pending LLDD requests waiting for a queue job structure.
2631 * In these "overrun" cases, a temporary queue element is allocated
2632 * the LLDD request and CMD iu buffer information remembered, and the
2633 * routine returns a -EOVERFLOW status. Subsequently, when a queue job
2634 * structure is freed, it is immediately reallocated for anything on the
2635 * pending request list. The LLDDs defer_rcv() callback is called,
2636 * informing the LLDD that it may reuse the CMD IU buffer, and the io
2637 * is then started normally with the transport.
2638 *
2639 * The LLDD, when receiving an -EOVERFLOW completion status, is to treat
2640 * the completion as successful but must not reuse the CMD IU buffer
2641 * until the LLDD's defer_rcv() callback has been called for the
2642 * corresponding struct nvmefc_tgt_fcp_req pointer.
2643 *
2644 * If there is any other condition in which an error occurs, the
2645 * transport will return a non-zero status indicating the error.
2646 * In all cases other than -EOVERFLOW, the transport has not accepted the
2647 * request and the LLDD should abort the exchange.
2648 *
2649 * @target_port: pointer to the (registered) target port the FCP CMD IU
2650 *              was received on.
2651 * @fcpreq:     pointer to a fcpreq request structure to be used to reference
2652 *              the exchange corresponding to the FCP Exchange.
2653 * @cmdiubuf:   pointer to the buffer containing the FCP CMD IU
2654 * @cmdiubuf_len: length, in bytes, of the received FCP CMD IU
2655 */
2656int
2657nvmet_fc_rcv_fcp_req(struct nvmet_fc_target_port *target_port,
2658			struct nvmefc_tgt_fcp_req *fcpreq,
2659			void *cmdiubuf, u32 cmdiubuf_len)
2660{
2661	struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port);
2662	struct nvme_fc_cmd_iu *cmdiu = cmdiubuf;
2663	struct nvmet_fc_tgt_queue *queue;
2664	struct nvmet_fc_fcp_iod *fod;
2665	struct nvmet_fc_defer_fcp_req *deferfcp;
2666	unsigned long flags;
2667
2668	/* validate iu, so the connection id can be used to find the queue */
2669	if ((cmdiubuf_len != sizeof(*cmdiu)) ||
2670			(cmdiu->format_id != NVME_CMD_FORMAT_ID) ||
2671			(cmdiu->fc_id != NVME_CMD_FC_ID) ||
2672			(be16_to_cpu(cmdiu->iu_len) != (sizeof(*cmdiu)/4)))
2673		return -EIO;
2674
2675	queue = nvmet_fc_find_target_queue(tgtport,
2676				be64_to_cpu(cmdiu->connection_id));
2677	if (!queue)
2678		return -ENOTCONN;
2679
2680	/*
2681	 * note: reference taken by find_target_queue
2682	 * After successful fod allocation, the fod will inherit the
2683	 * ownership of that reference and will remove the reference
2684	 * when the fod is freed.
2685	 */
2686
2687	spin_lock_irqsave(&queue->qlock, flags);
2688
2689	fod = nvmet_fc_alloc_fcp_iod(queue);
2690	if (fod) {
2691		spin_unlock_irqrestore(&queue->qlock, flags);
2692
2693		fcpreq->nvmet_fc_private = fod;
2694		fod->fcpreq = fcpreq;
2695
2696		memcpy(&fod->cmdiubuf, cmdiubuf, cmdiubuf_len);
2697
2698		nvmet_fc_queue_fcp_req(tgtport, queue, fcpreq);
2699
2700		return 0;
2701	}
2702
2703	if (!tgtport->ops->defer_rcv) {
2704		spin_unlock_irqrestore(&queue->qlock, flags);
2705		/* release the queue lookup reference */
2706		nvmet_fc_tgt_q_put(queue);
2707		return -ENOENT;
2708	}
2709
2710	deferfcp = list_first_entry_or_null(&queue->avail_defer_list,
2711			struct nvmet_fc_defer_fcp_req, req_list);
2712	if (deferfcp) {
2713		/* Just re-use one that was previously allocated */
2714		list_del(&deferfcp->req_list);
2715	} else {
2716		spin_unlock_irqrestore(&queue->qlock, flags);
2717
2718		/* Now we need to dynamically allocate one */
2719		deferfcp = kmalloc(sizeof(*deferfcp), GFP_KERNEL);
2720		if (!deferfcp) {
2721			/* release the queue lookup reference */
2722			nvmet_fc_tgt_q_put(queue);
2723			return -ENOMEM;
2724		}
2725		spin_lock_irqsave(&queue->qlock, flags);
2726	}
2727
2728	/* For now, use rspaddr / rsplen to save payload information */
2729	fcpreq->rspaddr = cmdiubuf;
2730	fcpreq->rsplen  = cmdiubuf_len;
2731	deferfcp->fcp_req = fcpreq;
2732
2733	/* defer processing till a fod becomes available */
2734	list_add_tail(&deferfcp->req_list, &queue->pending_cmd_list);
2735
2736	/* NOTE: the queue lookup reference is still valid */
2737
2738	spin_unlock_irqrestore(&queue->qlock, flags);
2739
2740	return -EOVERFLOW;
2741}
2742EXPORT_SYMBOL_GPL(nvmet_fc_rcv_fcp_req);
2743
2744/**
2745 * nvmet_fc_rcv_fcp_abort - transport entry point called by an LLDD
2746 *                       upon the reception of an ABTS for a FCP command
2747 *
2748 * Notify the transport that an ABTS has been received for a FCP command
2749 * that had been given to the transport via nvmet_fc_rcv_fcp_req(). The
2750 * LLDD believes the command is still being worked on
2751 * (template_ops->fcp_req_release() has not been called).
2752 *
2753 * The transport will wait for any outstanding work (an op to the LLDD,
2754 * which the lldd should complete with error due to the ABTS; or the
2755 * completion from the nvmet layer of the nvme command), then will
2756 * stop processing and call the nvmet_fc_rcv_fcp_req() callback to
2757 * return the i/o context to the LLDD.  The LLDD may send the BA_ACC
2758 * to the ABTS either after return from this function (assuming any
2759 * outstanding op work has been terminated) or upon the callback being
2760 * called.
2761 *
2762 * @target_port: pointer to the (registered) target port the FCP CMD IU
2763 *              was received on.
2764 * @fcpreq:     pointer to the fcpreq request structure that corresponds
2765 *              to the exchange that received the ABTS.
2766 */
2767void
2768nvmet_fc_rcv_fcp_abort(struct nvmet_fc_target_port *target_port,
2769			struct nvmefc_tgt_fcp_req *fcpreq)
2770{
2771	struct nvmet_fc_fcp_iod *fod = fcpreq->nvmet_fc_private;
2772	struct nvmet_fc_tgt_queue *queue;
2773	unsigned long flags;
2774
2775	if (!fod || fod->fcpreq != fcpreq)
2776		/* job appears to have already completed, ignore abort */
2777		return;
2778
2779	queue = fod->queue;
2780
2781	spin_lock_irqsave(&queue->qlock, flags);
2782	if (fod->active) {
2783		/*
2784		 * mark as abort. The abort handler, invoked upon completion
2785		 * of any work, will detect the aborted status and do the
2786		 * callback.
2787		 */
2788		spin_lock(&fod->flock);
2789		fod->abort = true;
2790		fod->aborted = true;
2791		spin_unlock(&fod->flock);
2792	}
2793	spin_unlock_irqrestore(&queue->qlock, flags);
2794}
2795EXPORT_SYMBOL_GPL(nvmet_fc_rcv_fcp_abort);
2796
2797
2798struct nvmet_fc_traddr {
2799	u64	nn;
2800	u64	pn;
2801};
2802
2803static int
2804__nvme_fc_parse_u64(substring_t *sstr, u64 *val)
2805{
2806	u64 token64;
2807
2808	if (match_u64(sstr, &token64))
2809		return -EINVAL;
2810	*val = token64;
2811
2812	return 0;
2813}
2814
2815/*
2816 * This routine validates and extracts the WWN's from the TRADDR string.
2817 * As kernel parsers need the 0x to determine number base, universally
2818 * build string to parse with 0x prefix before parsing name strings.
2819 */
2820static int
2821nvme_fc_parse_traddr(struct nvmet_fc_traddr *traddr, char *buf, size_t blen)
2822{
2823	char name[2 + NVME_FC_TRADDR_HEXNAMELEN + 1];
2824	substring_t wwn = { name, &name[sizeof(name)-1] };
2825	int nnoffset, pnoffset;
2826
2827	/* validate if string is one of the 2 allowed formats */
2828	if (strnlen(buf, blen) == NVME_FC_TRADDR_MAXLENGTH &&
2829			!strncmp(buf, "nn-0x", NVME_FC_TRADDR_OXNNLEN) &&
2830			!strncmp(&buf[NVME_FC_TRADDR_MAX_PN_OFFSET],
2831				"pn-0x", NVME_FC_TRADDR_OXNNLEN)) {
2832		nnoffset = NVME_FC_TRADDR_OXNNLEN;
2833		pnoffset = NVME_FC_TRADDR_MAX_PN_OFFSET +
2834						NVME_FC_TRADDR_OXNNLEN;
2835	} else if ((strnlen(buf, blen) == NVME_FC_TRADDR_MINLENGTH &&
2836			!strncmp(buf, "nn-", NVME_FC_TRADDR_NNLEN) &&
2837			!strncmp(&buf[NVME_FC_TRADDR_MIN_PN_OFFSET],
2838				"pn-", NVME_FC_TRADDR_NNLEN))) {
2839		nnoffset = NVME_FC_TRADDR_NNLEN;
2840		pnoffset = NVME_FC_TRADDR_MIN_PN_OFFSET + NVME_FC_TRADDR_NNLEN;
2841	} else
2842		goto out_einval;
2843
2844	name[0] = '0';
2845	name[1] = 'x';
2846	name[2 + NVME_FC_TRADDR_HEXNAMELEN] = 0;
2847
2848	memcpy(&name[2], &buf[nnoffset], NVME_FC_TRADDR_HEXNAMELEN);
2849	if (__nvme_fc_parse_u64(&wwn, &traddr->nn))
2850		goto out_einval;
2851
2852	memcpy(&name[2], &buf[pnoffset], NVME_FC_TRADDR_HEXNAMELEN);
2853	if (__nvme_fc_parse_u64(&wwn, &traddr->pn))
2854		goto out_einval;
2855
2856	return 0;
2857
2858out_einval:
2859	pr_warn("%s: bad traddr string\n", __func__);
2860	return -EINVAL;
2861}
2862
2863static int
2864nvmet_fc_add_port(struct nvmet_port *port)
2865{
2866	struct nvmet_fc_tgtport *tgtport;
2867	struct nvmet_fc_port_entry *pe;
2868	struct nvmet_fc_traddr traddr = { 0L, 0L };
2869	unsigned long flags;
2870	int ret;
2871
2872	/* validate the address info */
2873	if ((port->disc_addr.trtype != NVMF_TRTYPE_FC) ||
2874	    (port->disc_addr.adrfam != NVMF_ADDR_FAMILY_FC))
2875		return -EINVAL;
2876
2877	/* map the traddr address info to a target port */
2878
2879	ret = nvme_fc_parse_traddr(&traddr, port->disc_addr.traddr,
2880			sizeof(port->disc_addr.traddr));
2881	if (ret)
2882		return ret;
2883
2884	pe = kzalloc(sizeof(*pe), GFP_KERNEL);
2885	if (!pe)
2886		return -ENOMEM;
2887
2888	ret = -ENXIO;
2889	spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
2890	list_for_each_entry(tgtport, &nvmet_fc_target_list, tgt_list) {
2891		if ((tgtport->fc_target_port.node_name == traddr.nn) &&
2892		    (tgtport->fc_target_port.port_name == traddr.pn)) {
2893			/* a FC port can only be 1 nvmet port id */
2894			if (!tgtport->pe) {
2895				nvmet_fc_portentry_bind(tgtport, pe, port);
2896				ret = 0;
2897			} else
2898				ret = -EALREADY;
2899			break;
2900		}
2901	}
2902	spin_unlock_irqrestore(&nvmet_fc_tgtlock, flags);
2903
2904	if (ret)
2905		kfree(pe);
2906
2907	return ret;
2908}
2909
2910static void
2911nvmet_fc_remove_port(struct nvmet_port *port)
2912{
2913	struct nvmet_fc_port_entry *pe = port->priv;
2914
2915	nvmet_fc_portentry_unbind(pe);
2916
2917	/* terminate any outstanding associations */
2918	__nvmet_fc_free_assocs(pe->tgtport);
2919
2920	kfree(pe);
2921}
2922
2923static void
2924nvmet_fc_discovery_chg(struct nvmet_port *port)
2925{
2926	struct nvmet_fc_port_entry *pe = port->priv;
2927	struct nvmet_fc_tgtport *tgtport = pe->tgtport;
2928
2929	if (tgtport && tgtport->ops->discovery_event)
2930		tgtport->ops->discovery_event(&tgtport->fc_target_port);
2931}
2932
2933static const struct nvmet_fabrics_ops nvmet_fc_tgt_fcp_ops = {
2934	.owner			= THIS_MODULE,
2935	.type			= NVMF_TRTYPE_FC,
2936	.msdbd			= 1,
2937	.add_port		= nvmet_fc_add_port,
2938	.remove_port		= nvmet_fc_remove_port,
2939	.queue_response		= nvmet_fc_fcp_nvme_cmd_done,
2940	.delete_ctrl		= nvmet_fc_delete_ctrl,
2941	.discovery_chg		= nvmet_fc_discovery_chg,
2942};
2943
2944static int __init nvmet_fc_init_module(void)
2945{
2946	return nvmet_register_transport(&nvmet_fc_tgt_fcp_ops);
2947}
2948
2949static void __exit nvmet_fc_exit_module(void)
2950{
2951	/* ensure any shutdown operation, e.g. delete ctrls have finished */
2952	flush_workqueue(nvmet_wq);
2953
2954	/* sanity check - all lports should be removed */
2955	if (!list_empty(&nvmet_fc_target_list))
2956		pr_warn("%s: targetport list not empty\n", __func__);
2957
2958	nvmet_unregister_transport(&nvmet_fc_tgt_fcp_ops);
2959
2960	ida_destroy(&nvmet_fc_tgtport_cnt);
2961}
2962
2963module_init(nvmet_fc_init_module);
2964module_exit(nvmet_fc_exit_module);
2965
2966MODULE_LICENSE("GPL v2");
2967