aoecmd.c 22.7 KB
Newer Older
E
Ed L. Cashin 已提交
1
/* Copyright (c) 2007 Coraid, Inc.  See COPYING for GPL terms. */
L
Linus Torvalds 已提交
2 3 4 5 6
/*
 * aoecmd.c
 * Filesystem request handling methods
 */

7
#include <linux/ata.h>
L
Linus Torvalds 已提交
8 9 10 11
#include <linux/hdreg.h>
#include <linux/blkdev.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
12
#include <linux/genhd.h>
13
#include <linux/moduleparam.h>
14
#include <net/net_namespace.h>
15
#include <asm/unaligned.h>
L
Linus Torvalds 已提交
16 17
#include "aoe.h"

18 19 20
static int aoe_deadsecs = 60 * 3;
module_param(aoe_deadsecs, int, 0644);
MODULE_PARM_DESC(aoe_deadsecs, "After aoe_deadsecs seconds, give up and fail dev.");
L
Linus Torvalds 已提交
21

22 23 24 25 26
static int aoe_maxout = 16;
module_param(aoe_maxout, int, 0644);
MODULE_PARM_DESC(aoe_maxout,
	"Only aoe_maxout outstanding packets for every MAC on eX.Y.");

27
static struct sk_buff *
28
new_skb(ulong len)
L
Linus Torvalds 已提交
29 30 31 32 33
{
	struct sk_buff *skb;

	skb = alloc_skb(len, GFP_ATOMIC);
	if (skb) {
34
		skb_reset_mac_header(skb);
35
		skb_reset_network_header(skb);
L
Linus Torvalds 已提交
36 37 38 39 40 41 42 43 44 45 46 47 48
		skb->protocol = __constant_htons(ETH_P_AOE);
		skb->priority = 0;
		skb->next = skb->prev = NULL;

		/* tell the network layer not to perform IP checksums
		 * or to get the NIC to do it
		 */
		skb->ip_summed = CHECKSUM_NONE;
	}
	return skb;
}

static struct frame *
49
getframe(struct aoetgt *t, int tag)
L
Linus Torvalds 已提交
50 51 52
{
	struct frame *f, *e;

53 54
	f = t->frames;
	e = f + t->nframes;
L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63 64 65 66
	for (; f<e; f++)
		if (f->tag == tag)
			return f;
	return NULL;
}

/*
 * Leave the top bit clear so we have tagspace for userland.
 * The bottom 16 bits are the xmit tick for rexmit/rttavg processing.
 * This driver reserves tag -1 to mean "unused frame."
 */
static int
67
newtag(struct aoetgt *t)
L
Linus Torvalds 已提交
68 69 70 71
{
	register ulong n;

	n = jiffies & 0xffff;
72
	return n |= (++t->lasttag & 0x7fff) << 16;
L
Linus Torvalds 已提交
73 74 75
}

static int
76
aoehdr_atainit(struct aoedev *d, struct aoetgt *t, struct aoe_hdr *h)
L
Linus Torvalds 已提交
77
{
78
	u32 host_tag = newtag(t);
L
Linus Torvalds 已提交
79

80 81
	memcpy(h->src, t->ifp->nd->dev_addr, sizeof h->src);
	memcpy(h->dst, t->addr, sizeof h->dst);
82
	h->type = __constant_cpu_to_be16(ETH_P_AOE);
L
Linus Torvalds 已提交
83
	h->verfl = AOE_HVER;
84
	h->major = cpu_to_be16(d->aoemajor);
L
Linus Torvalds 已提交
85 86
	h->minor = d->aoeminor;
	h->cmd = AOECMD_ATA;
87
	h->tag = cpu_to_be32(host_tag);
L
Linus Torvalds 已提交
88 89 90 91

	return host_tag;
}

92 93 94 95 96 97 98 99 100 101 102
static inline void
put_lba(struct aoe_atahdr *ah, sector_t lba)
{
	ah->lba0 = lba;
	ah->lba1 = lba >>= 8;
	ah->lba2 = lba >>= 8;
	ah->lba3 = lba >>= 8;
	ah->lba4 = lba >>= 8;
	ah->lba5 = lba >>= 8;
}

L
Linus Torvalds 已提交
103
static void
104 105 106 107 108 109 110 111 112 113 114
ifrotate(struct aoetgt *t)
{
	t->ifp++;
	if (t->ifp >= &t->ifs[NAOEIFS] || t->ifp->nd == NULL)
		t->ifp = t->ifs;
	if (t->ifp->nd == NULL) {
		printk(KERN_INFO "aoe: no interface to rotate to\n");
		BUG();
	}
}

115 116 117
static void
skb_pool_put(struct aoedev *d, struct sk_buff *skb)
{
118
	__skb_queue_tail(&d->skbpool, skb);
119 120 121 122 123
}

static struct sk_buff *
skb_pool_get(struct aoedev *d)
{
124
	struct sk_buff *skb = skb_peek(&d->skbpool);
125 126

	if (skb && atomic_read(&skb_shinfo(skb)->dataref) == 1) {
127
		__skb_unlink(skb, &d->skbpool);
128 129
		return skb;
	}
130 131
	if (skb_queue_len(&d->skbpool) < NSKBPOOLMAX &&
	    (skb = new_skb(ETH_ZLEN)))
132
		return skb;
133

134 135 136 137
	return NULL;
}

/* freeframe is where we do our load balancing so it's a little hairy. */
138 139
static struct frame *
freeframe(struct aoedev *d)
L
Linus Torvalds 已提交
140
{
141
	struct frame *f, *e, *rf;
142
	struct aoetgt **t;
143
	struct sk_buff *skb;
144 145 146 147 148

	if (d->targets[0] == NULL) {	/* shouldn't happen, but I'm paranoid */
		printk(KERN_ERR "aoe: NULL TARGETS!\n");
		return NULL;
	}
149 150 151 152 153 154 155 156 157
	t = d->tgt;
	t++;
	if (t >= &d->targets[NTARGETS] || !*t)
		t = d->targets;
	for (;;) {
		if ((*t)->nout < (*t)->maxout
		&& t != d->htgt
		&& (*t)->ifp->nd) {
			rf = NULL;
158
			f = (*t)->frames;
159
			e = f + (*t)->nframes;
160 161 162
			for (; f < e; f++) {
				if (f->tag != FREETAG)
					continue;
163 164 165 166 167
				skb = f->skb;
				if (!skb
				&& !(f->skb = skb = new_skb(ETH_ZLEN)))
					continue;
				if (atomic_read(&skb_shinfo(skb)->dataref)
168
					!= 1) {
169 170
					if (!rf)
						rf = f;
171 172
					continue;
				}
173 174
gotone:				skb_shinfo(skb)->nr_frags = skb->data_len = 0;
				skb_trim(skb, 0);
175 176 177 178
				d->tgt = t;
				ifrotate(*t);
				return f;
			}
179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
			/* Work can be done, but the network layer is
			   holding our precious packets.  Try to grab
			   one from the pool. */
			f = rf;
			if (f == NULL) {	/* more paranoia */
				printk(KERN_ERR
					"aoe: freeframe: %s.\n",
					"unexpected null rf");
				d->flags |= DEVFL_KICKME;
				return NULL;
			}
			skb = skb_pool_get(d);
			if (skb) {
				skb_pool_put(d, f->skb);
				f->skb = skb;
				goto gotone;
			}
			(*t)->dataref++;
			if ((*t)->nout == 0)
198 199
				d->flags |= DEVFL_KICKME;
		}
200 201
		if (t == d->tgt)	/* we've looped and found nada */
			break;
202
		t++;
203 204 205
		if (t >= &d->targets[NTARGETS] || !*t)
			t = d->targets;
	}
206 207 208 209 210 211 212
	return NULL;
}

static int
aoecmd_ata_rw(struct aoedev *d)
{
	struct frame *f;
L
Linus Torvalds 已提交
213 214 215
	struct aoe_hdr *h;
	struct aoe_atahdr *ah;
	struct buf *buf;
216 217
	struct bio_vec *bv;
	struct aoetgt *t;
L
Linus Torvalds 已提交
218 219 220 221 222 223 224
	struct sk_buff *skb;
	ulong bcnt;
	char writebit, extbit;

	writebit = 0x10;
	extbit = 0x4;

225 226 227 228
	f = freeframe(d);
	if (f == NULL)
		return 0;
	t = *d->tgt;
L
Linus Torvalds 已提交
229
	buf = d->inprocess;
230 231 232 233 234 235
	bv = buf->bv;
	bcnt = t->ifp->maxbcnt;
	if (bcnt == 0)
		bcnt = DEFAULTBCNT;
	if (bcnt > buf->bv_resid)
		bcnt = buf->bv_resid;
L
Linus Torvalds 已提交
236
	/* initialize the headers & frame */
237
	skb = f->skb;
238
	h = (struct aoe_hdr *) skb_mac_header(skb);
L
Linus Torvalds 已提交
239
	ah = (struct aoe_atahdr *) (h+1);
240 241
	skb_put(skb, sizeof *h + sizeof *ah);
	memset(h, 0, skb->len);
242 243
	f->tag = aoehdr_atainit(d, t, h);
	t->nout++;
L
Linus Torvalds 已提交
244 245
	f->waited = 0;
	f->buf = buf;
246
	f->bufaddr = page_address(bv->bv_page) + buf->bv_off;
247
	f->bcnt = bcnt;
248
	f->lba = buf->sector;
L
Linus Torvalds 已提交
249 250 251

	/* set up ata header */
	ah->scnt = bcnt >> 9;
252
	put_lba(ah, buf->sector);
L
Linus Torvalds 已提交
253 254 255 256 257 258 259 260
	if (d->flags & DEVFL_EXT) {
		ah->aflags |= AOEAFL_EXT;
	} else {
		extbit = 0;
		ah->lba3 &= 0x0f;
		ah->lba3 |= 0xe0;	/* LBA bit + obsolete 0xa0 */
	}
	if (bio_data_dir(buf->bio) == WRITE) {
261
		skb_fill_page_desc(skb, 0, bv->bv_page, buf->bv_off, bcnt);
L
Linus Torvalds 已提交
262
		ah->aflags |= AOEAFL_WRITE;
263 264
		skb->len += bcnt;
		skb->data_len = bcnt;
265
		t->wpkts++;
L
Linus Torvalds 已提交
266
	} else {
267
		t->rpkts++;
L
Linus Torvalds 已提交
268 269 270
		writebit = 0;
	}

271
	ah->cmdstat = ATA_CMD_PIO_READ | writebit | extbit;
L
Linus Torvalds 已提交
272 273 274

	/* mark all tracking fields and load out */
	buf->nframesout += 1;
275
	buf->bv_off += bcnt;
L
Linus Torvalds 已提交
276 277 278 279 280 281
	buf->bv_resid -= bcnt;
	buf->resid -= bcnt;
	buf->sector += bcnt >> 9;
	if (buf->resid == 0) {
		d->inprocess = NULL;
	} else if (buf->bv_resid == 0) {
282 283 284 285
		buf->bv = ++bv;
		buf->bv_resid = bv->bv_len;
		WARN_ON(buf->bv_resid == 0);
		buf->bv_off = bv->bv_offset;
L
Linus Torvalds 已提交
286 287
	}

288
	skb->dev = t->ifp->nd;
289
	skb = skb_clone(skb, GFP_ATOMIC);
290 291
	if (skb)
		__skb_queue_tail(&d->sendq, skb);
292
	return 1;
L
Linus Torvalds 已提交
293 294
}

295 296 297
/* some callers cannot sleep, and they can call this function,
 * transmitting the packets later, when interrupts are on
 */
298 299
static void
aoecmd_cfg_pkts(ushort aoemajor, unsigned char aoeminor, struct sk_buff_head *queue)
300 301 302
{
	struct aoe_hdr *h;
	struct aoe_cfghdr *ch;
303
	struct sk_buff *skb;
304 305 306
	struct net_device *ifp;

	read_lock(&dev_base_lock);
307
	for_each_netdev(&init_net, ifp) {
308 309
		dev_hold(ifp);
		if (!is_aoe_netif(ifp))
310
			goto cont;
311

312
		skb = new_skb(sizeof *h + sizeof *ch);
313
		if (skb == NULL) {
E
Ed L. Cashin 已提交
314
			printk(KERN_INFO "aoe: skb alloc failure\n");
315
			goto cont;
316
		}
317
		skb_put(skb, sizeof *h + sizeof *ch);
318
		skb->dev = ifp;
319
		__skb_queue_tail(queue, skb);
320
		h = (struct aoe_hdr *) skb_mac_header(skb);
321 322 323 324 325 326 327 328 329 330
		memset(h, 0, sizeof *h + sizeof *ch);

		memset(h->dst, 0xff, sizeof h->dst);
		memcpy(h->src, ifp->dev_addr, sizeof h->src);
		h->type = __constant_cpu_to_be16(ETH_P_AOE);
		h->verfl = AOE_HVER;
		h->major = cpu_to_be16(aoemajor);
		h->minor = aoeminor;
		h->cmd = AOECMD_CFG;

331 332
cont:
		dev_put(ifp);
333 334 335 336
	}
	read_unlock(&dev_base_lock);
}

L
Linus Torvalds 已提交
337
static void
338
resend(struct aoedev *d, struct aoetgt *t, struct frame *f)
L
Linus Torvalds 已提交
339 340 341
{
	struct sk_buff *skb;
	struct aoe_hdr *h;
342
	struct aoe_atahdr *ah;
L
Linus Torvalds 已提交
343 344 345
	char buf[128];
	u32 n;

346 347 348 349 350
	ifrotate(t);
	n = newtag(t);
	skb = f->skb;
	h = (struct aoe_hdr *) skb_mac_header(skb);
	ah = (struct aoe_atahdr *) (h+1);
L
Linus Torvalds 已提交
351 352

	snprintf(buf, sizeof buf,
353
		"%15s e%ld.%d oldtag=%08x@%08lx newtag=%08x s=%pm d=%pm nout=%d\n",
354
		"retransmit", d->aoemajor, d->aoeminor, f->tag, jiffies, n,
355
		h->src, h->dst, t->nout);
L
Linus Torvalds 已提交
356 357 358
	aoechr_error(buf);

	f->tag = n;
359
	h->tag = cpu_to_be32(n);
360 361 362 363 364 365
	memcpy(h->dst, t->addr, sizeof h->dst);
	memcpy(h->src, t->ifp->nd->dev_addr, sizeof h->src);

	switch (ah->cmdstat) {
	default:
		break;
366 367 368 369
	case ATA_CMD_PIO_READ:
	case ATA_CMD_PIO_READ_EXT:
	case ATA_CMD_PIO_WRITE:
	case ATA_CMD_PIO_WRITE_EXT:
370 371 372 373 374 375
		put_lba(ah, f->lba);

		n = f->bcnt;
		if (n > DEFAULTBCNT)
			n = DEFAULTBCNT;
		ah->scnt = n >> 9;
376
		if (ah->aflags & AOEAFL_WRITE) {
377
			skb_fill_page_desc(skb, 0, virt_to_page(f->bufaddr),
378 379 380
				offset_in_page(f->bufaddr), n);
			skb->len = sizeof *h + sizeof *ah + n;
			skb->data_len = n;
381 382
		}
	}
383
	skb->dev = t->ifp->nd;
384 385 386
	skb = skb_clone(skb, GFP_ATOMIC);
	if (skb == NULL)
		return;
387
	__skb_queue_tail(&d->sendq, skb);
L
Linus Torvalds 已提交
388 389 390 391 392 393 394 395 396 397 398 399 400 401
}

static int
tsince(int tag)
{
	int n;

	n = jiffies & 0xffff;
	n -= tag & 0xffff;
	if (n < 0)
		n += 1<<16;
	return n;
}

402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481
static struct aoeif *
getif(struct aoetgt *t, struct net_device *nd)
{
	struct aoeif *p, *e;

	p = t->ifs;
	e = p + NAOEIFS;
	for (; p < e; p++)
		if (p->nd == nd)
			return p;
	return NULL;
}

static struct aoeif *
addif(struct aoetgt *t, struct net_device *nd)
{
	struct aoeif *p;

	p = getif(t, NULL);
	if (!p)
		return NULL;
	p->nd = nd;
	p->maxbcnt = DEFAULTBCNT;
	p->lost = 0;
	p->lostjumbo = 0;
	return p;
}

static void
ejectif(struct aoetgt *t, struct aoeif *ifp)
{
	struct aoeif *e;
	ulong n;

	e = t->ifs + NAOEIFS - 1;
	n = (e - ifp) * sizeof *ifp;
	memmove(ifp, ifp+1, n);
	e->nd = NULL;
}

static int
sthtith(struct aoedev *d)
{
	struct frame *f, *e, *nf;
	struct sk_buff *skb;
	struct aoetgt *ht = *d->htgt;

	f = ht->frames;
	e = f + ht->nframes;
	for (; f < e; f++) {
		if (f->tag == FREETAG)
			continue;
		nf = freeframe(d);
		if (!nf)
			return 0;
		skb = nf->skb;
		*nf = *f;
		f->skb = skb;
		f->tag = FREETAG;
		nf->waited = 0;
		ht->nout--;
		(*d->tgt)->nout++;
		resend(d, *d->tgt, nf);
	}
	/* he's clean, he's useless.  take away his interfaces */
	memset(ht->ifs, 0, sizeof ht->ifs);
	d->htgt = NULL;
	return 1;
}

static inline unsigned char
ata_scnt(unsigned char *packet) {
	struct aoe_hdr *h;
	struct aoe_atahdr *ah;

	h = (struct aoe_hdr *) packet;
	ah = (struct aoe_atahdr *) (h+1);
	return ah->scnt;
}

L
Linus Torvalds 已提交
482 483 484
static void
rexmit_timer(ulong vp)
{
485
	struct sk_buff_head queue;
L
Linus Torvalds 已提交
486
	struct aoedev *d;
487 488
	struct aoetgt *t, **tt, **te;
	struct aoeif *ifp;
L
Linus Torvalds 已提交
489 490 491 492 493 494 495 496 497 498 499 500 501
	struct frame *f, *e;
	register long timeout;
	ulong flags, n;

	d = (struct aoedev *) vp;

	/* timeout is always ~150% of the moving average */
	timeout = d->rttavg;
	timeout += timeout >> 1;

	spin_lock_irqsave(&d->lock, flags);

	if (d->flags & DEVFL_TKILL) {
502
		spin_unlock_irqrestore(&d->lock, flags);
L
Linus Torvalds 已提交
503 504
		return;
	}
505 506 507 508 509 510 511 512 513 514
	tt = d->targets;
	te = tt + NTARGETS;
	for (; tt < te && *tt; tt++) {
		t = *tt;
		f = t->frames;
		e = f + t->nframes;
		for (; f < e; f++) {
			if (f->tag == FREETAG
			|| tsince(f->tag) < timeout)
				continue;
L
Linus Torvalds 已提交
515 516
			n = f->waited += timeout;
			n /= HZ;
517 518
			if (n > aoe_deadsecs) {
				/* waited too long.  device failure. */
L
Linus Torvalds 已提交
519
				aoedev_downdev(d);
520
				break;
L
Linus Torvalds 已提交
521
			}
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545

			if (n > HELPWAIT /* see if another target can help */
			&& (tt != d->targets || d->targets[1]))
				d->htgt = tt;

			if (t->nout == t->maxout) {
				if (t->maxout > 1)
					t->maxout--;
				t->lastwadj = jiffies;
			}

			ifp = getif(t, f->skb->dev);
			if (ifp && ++ifp->lost > (t->nframes << 1)
			&& (ifp != t->ifs || t->ifs[1].nd)) {
				ejectif(t, ifp);
				ifp = NULL;
			}

			if (ata_scnt(skb_mac_header(f->skb)) > DEFAULTBCNT / 512
			&& ifp && ++ifp->lostjumbo > (t->nframes << 1)
			&& ifp->maxbcnt != DEFAULTBCNT) {
				printk(KERN_INFO
					"aoe: e%ld.%d: "
					"too many lost jumbo on "
546
					"%s:%pm - "
547 548
					"falling back to %d frames.\n",
					d->aoemajor, d->aoeminor,
549
					ifp->nd->name, t->addr,
550 551 552 553 554 555 556 557 558 559 560 561
					DEFAULTBCNT);
				ifp->maxbcnt = 0;
			}
			resend(d, t, f);
		}

		/* window check */
		if (t->nout == t->maxout
		&& t->maxout < t->nframes
		&& (jiffies - t->lastwadj)/HZ > 10) {
			t->maxout++;
			t->lastwadj = jiffies;
L
Linus Torvalds 已提交
562 563
		}
	}
564

565
	if (!skb_queue_empty(&d->sendq)) {
566 567 568 569 570 571
		n = d->rttavg <<= 1;
		if (n > MAXTIMER)
			d->rttavg = MAXTIMER;
	}

	if (d->flags & DEVFL_KICKME || d->htgt) {
572 573 574
		d->flags &= ~DEVFL_KICKME;
		aoecmd_work(d);
	}
L
Linus Torvalds 已提交
575

576 577
	__skb_queue_head_init(&queue);
	skb_queue_splice_init(&d->sendq, &queue);
L
Linus Torvalds 已提交
578 579 580 581 582 583

	d->timer.expires = jiffies + TIMERTICK;
	add_timer(&d->timer);

	spin_unlock_irqrestore(&d->lock, flags);

584
	aoenet_xmit(&queue);
L
Linus Torvalds 已提交
585 586
}

587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605
/* enters with d->lock held */
void
aoecmd_work(struct aoedev *d)
{
	struct buf *buf;
loop:
	if (d->htgt && !sthtith(d))
		return;
	if (d->inprocess == NULL) {
		if (list_empty(&d->bufq))
			return;
		buf = container_of(d->bufq.next, struct buf, bufs);
		list_del(d->bufq.next);
		d->inprocess = buf;
	}
	if (aoecmd_ata_rw(d))
		goto loop;
}

606 607 608
/* this function performs work that has been deferred until sleeping is OK
 */
void
609
aoecmd_sleepwork(struct work_struct *work)
610
{
611
	struct aoedev *d = container_of(work, struct aoedev, work);
612 613 614 615 616 617 618 619 620

	if (d->flags & DEVFL_GDALLOC)
		aoeblk_gdalloc(d);

	if (d->flags & DEVFL_NEWSIZE) {
		struct block_device *bd;
		unsigned long flags;
		u64 ssize;

621
		ssize = get_capacity(d->gd);
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
		bd = bdget_disk(d->gd, 0);

		if (bd) {
			mutex_lock(&bd->bd_inode->i_mutex);
			i_size_write(bd->bd_inode, (loff_t)ssize<<9);
			mutex_unlock(&bd->bd_inode->i_mutex);
			bdput(bd);
		}
		spin_lock_irqsave(&d->lock, flags);
		d->flags |= DEVFL_UP;
		d->flags &= ~DEVFL_NEWSIZE;
		spin_unlock_irqrestore(&d->lock, flags);
	}
}

L
Linus Torvalds 已提交
637
static void
638
ataid_complete(struct aoedev *d, struct aoetgt *t, unsigned char *id)
L
Linus Torvalds 已提交
639 640 641 642 643
{
	u64 ssize;
	u16 n;

	/* word 83: command set supported */
644
	n = get_unaligned_le16(&id[83 << 1]);
L
Linus Torvalds 已提交
645 646

	/* word 86: command set/feature enabled */
647
	n |= get_unaligned_le16(&id[86 << 1]);
L
Linus Torvalds 已提交
648 649 650 651 652

	if (n & (1<<10)) {	/* bit 10: LBA 48 */
		d->flags |= DEVFL_EXT;

		/* word 100: number lba48 sectors */
653
		ssize = get_unaligned_le64(&id[100 << 1]);
L
Linus Torvalds 已提交
654 655 656 657 658 659 660 661 662 663

		/* set as in ide-disk.c:init_idedisk_capacity */
		d->geo.cylinders = ssize;
		d->geo.cylinders /= (255 * 63);
		d->geo.heads = 255;
		d->geo.sectors = 63;
	} else {
		d->flags &= ~DEVFL_EXT;

		/* number lba28 sectors */
664
		ssize = get_unaligned_le32(&id[60 << 1]);
L
Linus Torvalds 已提交
665 666

		/* NOTE: obsolete in ATA 6 */
667 668 669
		d->geo.cylinders = get_unaligned_le16(&id[54 << 1]);
		d->geo.heads = get_unaligned_le16(&id[55 << 1]);
		d->geo.sectors = get_unaligned_le16(&id[56 << 1]);
L
Linus Torvalds 已提交
670
	}
671 672

	if (d->ssize != ssize)
673
		printk(KERN_INFO
674 675
			"aoe: %pm e%ld.%d v%04x has %llu sectors\n",
			t->addr,
676 677
			d->aoemajor, d->aoeminor,
			d->fw_ver, (long long)ssize);
L
Linus Torvalds 已提交
678 679
	d->ssize = ssize;
	d->geo.start = 0;
680 681
	if (d->flags & (DEVFL_GDALLOC|DEVFL_NEWSIZE))
		return;
L
Linus Torvalds 已提交
682
	if (d->gd != NULL) {
683
		set_capacity(d->gd, ssize);
684
		d->flags |= DEVFL_NEWSIZE;
685
	} else
686
		d->flags |= DEVFL_GDALLOC;
L
Linus Torvalds 已提交
687 688 689 690 691 692 693 694 695
	schedule_work(&d->work);
}

static void
calc_rttavg(struct aoedev *d, int rtt)
{
	register long n;

	n = rtt;
696 697 698 699 700 701 702 703 704
	if (n < 0) {
		n = -rtt;
		if (n < MINTIMER)
			n = MINTIMER;
		else if (n > MAXTIMER)
			n = MAXTIMER;
		d->mintimer += (n - d->mintimer) >> 1;
	} else if (n < d->mintimer)
		n = d->mintimer;
L
Linus Torvalds 已提交
705 706 707 708 709 710 711 712
	else if (n > MAXTIMER)
		n = MAXTIMER;

	/* g == .25; cf. Congestion Avoidance and Control, Jacobson & Karels; 1988 */
	n -= d->rttavg;
	d->rttavg += n >> 2;
}

713 714 715 716 717 718 719 720 721 722 723 724 725 726
static struct aoetgt *
gettgt(struct aoedev *d, char *addr)
{
	struct aoetgt **t, **e;

	t = d->targets;
	e = t + NTARGETS;
	for (; t < e && *t; t++)
		if (memcmp((*t)->addr, addr, sizeof((*t)->addr)) == 0)
			return *t;
	return NULL;
}

static inline void
727
diskstats(struct gendisk *disk, struct bio *bio, ulong duration, sector_t sector)
728 729 730
{
	unsigned long n_sect = bio->bi_size >> 9;
	const int rw = bio_data_dir(bio);
731
	struct hd_struct *part;
T
Tejun Heo 已提交
732
	int cpu;
733

734
	cpu = part_stat_lock();
735 736
	part = disk_map_sector_rcu(disk, sector);

737 738 739 740
	part_stat_inc(cpu, part, ios[rw]);
	part_stat_add(cpu, part, ticks[rw], duration);
	part_stat_add(cpu, part, sectors[rw], n_sect);
	part_stat_add(cpu, part, io_ticks, duration);
T
Tejun Heo 已提交
741

742
	part_stat_unlock();
743 744
}

L
Linus Torvalds 已提交
745 746 747
void
aoecmd_ata_rsp(struct sk_buff *skb)
{
748
	struct sk_buff_head queue;
L
Linus Torvalds 已提交
749
	struct aoedev *d;
750
	struct aoe_hdr *hin, *hout;
L
Linus Torvalds 已提交
751 752 753
	struct aoe_atahdr *ahin, *ahout;
	struct frame *f;
	struct buf *buf;
754 755
	struct aoetgt *t;
	struct aoeif *ifp;
L
Linus Torvalds 已提交
756 757 758
	register long n;
	ulong flags;
	char ebuf[128];
759 760
	u16 aoemajor;

761
	hin = (struct aoe_hdr *) skb_mac_header(skb);
762
	aoemajor = get_unaligned_be16(&hin->major);
763
	d = aoedev_by_aoeaddr(aoemajor, hin->minor);
L
Linus Torvalds 已提交
764 765 766
	if (d == NULL) {
		snprintf(ebuf, sizeof ebuf, "aoecmd_ata_rsp: ata response "
			"for unknown device %d.%d\n",
767
			 aoemajor, hin->minor);
L
Linus Torvalds 已提交
768 769 770 771 772 773
		aoechr_error(ebuf);
		return;
	}

	spin_lock_irqsave(&d->lock, flags);

774
	n = get_unaligned_be32(&hin->tag);
775 776
	t = gettgt(d, hin->src);
	if (t == NULL) {
777 778
		printk(KERN_INFO "aoe: can't find target e%ld.%d:%pm\n",
			d->aoemajor, d->aoeminor, hin->src);
779 780 781 782
		spin_unlock_irqrestore(&d->lock, flags);
		return;
	}
	f = getframe(t, n);
L
Linus Torvalds 已提交
783
	if (f == NULL) {
784
		calc_rttavg(d, -tsince(n));
L
Linus Torvalds 已提交
785 786 787 788
		spin_unlock_irqrestore(&d->lock, flags);
		snprintf(ebuf, sizeof ebuf,
			"%15s e%d.%d    tag=%08x@%08lx\n",
			"unexpected rsp",
789
			get_unaligned_be16(&hin->major),
L
Linus Torvalds 已提交
790
			hin->minor,
791
			get_unaligned_be32(&hin->tag),
L
Linus Torvalds 已提交
792 793 794 795 796 797 798 799
			jiffies);
		aoechr_error(ebuf);
		return;
	}

	calc_rttavg(d, tsince(f->tag));

	ahin = (struct aoe_atahdr *) (hin+1);
800
	hout = (struct aoe_hdr *) skb_mac_header(f->skb);
801
	ahout = (struct aoe_atahdr *) (hout+1);
L
Linus Torvalds 已提交
802 803 804
	buf = f->buf;

	if (ahin->cmdstat & 0xa9) {	/* these bits cleared on success */
E
Ed L. Cashin 已提交
805
		printk(KERN_ERR
806
			"aoe: ata error cmd=%2.2Xh stat=%2.2Xh from e%ld.%d\n",
L
Linus Torvalds 已提交
807 808 809 810 811
			ahout->cmdstat, ahin->cmdstat,
			d->aoemajor, d->aoeminor);
		if (buf)
			buf->flags |= BUFFL_FAIL;
	} else {
812 813
		if (d->htgt && t == *d->htgt) /* I'll help myself, thank you. */
			d->htgt = NULL;
814
		n = ahout->scnt << 9;
L
Linus Torvalds 已提交
815
		switch (ahout->cmdstat) {
816 817
		case ATA_CMD_PIO_READ:
		case ATA_CMD_PIO_READ_EXT:
L
Linus Torvalds 已提交
818
			if (skb->len - sizeof *hin - sizeof *ahin < n) {
E
Ed L. Cashin 已提交
819
				printk(KERN_ERR
820 821
					"aoe: %s.  skb->len=%d need=%ld\n",
					"runt data size in read", skb->len, n);
L
Linus Torvalds 已提交
822 823 824 825 826
				/* fail frame f?  just returning will rexmit. */
				spin_unlock_irqrestore(&d->lock, flags);
				return;
			}
			memcpy(f->bufaddr, ahin+1, n);
827 828
		case ATA_CMD_PIO_WRITE:
		case ATA_CMD_PIO_WRITE_EXT:
829 830 831 832 833 834
			ifp = getif(t, skb->dev);
			if (ifp) {
				ifp->lost = 0;
				if (n > DEFAULTBCNT)
					ifp->lostjumbo = 0;
			}
835
			if (f->bcnt -= n) {
836
				f->lba += n >> 9;
837
				f->bufaddr += n;
838 839
				resend(d, t, f);
				goto xmit;
840
			}
L
Linus Torvalds 已提交
841
			break;
842
		case ATA_CMD_ID_ATA:
L
Linus Torvalds 已提交
843
			if (skb->len - sizeof *hin - sizeof *ahin < 512) {
E
Ed L. Cashin 已提交
844 845
				printk(KERN_INFO
					"aoe: runt data size in ataid.  skb->len=%d\n",
846
					skb->len);
L
Linus Torvalds 已提交
847 848 849
				spin_unlock_irqrestore(&d->lock, flags);
				return;
			}
850
			ataid_complete(d, t, (char *) (ahin+1));
L
Linus Torvalds 已提交
851 852
			break;
		default:
E
Ed L. Cashin 已提交
853 854
			printk(KERN_INFO
				"aoe: unrecognized ata command %2.2Xh for %d.%d\n",
855
				ahout->cmdstat,
856
				get_unaligned_be16(&hin->major),
857
				hin->minor);
L
Linus Torvalds 已提交
858 859 860
		}
	}

861
	if (buf && --buf->nframesout == 0 && buf->resid == 0) {
862
		diskstats(d->gd, buf->bio, jiffies - buf->stime, buf->sector);
863 864 865
		n = (buf->flags & BUFFL_FAIL) ? -EIO : 0;
		bio_endio(buf->bio, n);
		mempool_free(buf, d->bufpool);
L
Linus Torvalds 已提交
866 867 868 869
	}

	f->buf = NULL;
	f->tag = FREETAG;
870
	t->nout--;
L
Linus Torvalds 已提交
871 872

	aoecmd_work(d);
873
xmit:
874 875
	__skb_queue_head_init(&queue);
	skb_queue_splice_init(&d->sendq, &queue);
L
Linus Torvalds 已提交
876 877

	spin_unlock_irqrestore(&d->lock, flags);
878
	aoenet_xmit(&queue);
L
Linus Torvalds 已提交
879 880 881 882 883
}

void
aoecmd_cfg(ushort aoemajor, unsigned char aoeminor)
{
884
	struct sk_buff_head queue;
L
Linus Torvalds 已提交
885

886 887 888
	__skb_queue_head_init(&queue);
	aoecmd_cfg_pkts(aoemajor, aoeminor, &queue);
	aoenet_xmit(&queue);
L
Linus Torvalds 已提交
889 890
}
 
891
struct sk_buff *
L
Linus Torvalds 已提交
892 893 894 895 896 897
aoecmd_ata_id(struct aoedev *d)
{
	struct aoe_hdr *h;
	struct aoe_atahdr *ah;
	struct frame *f;
	struct sk_buff *skb;
898
	struct aoetgt *t;
L
Linus Torvalds 已提交
899

900
	f = freeframe(d);
901
	if (f == NULL)
L
Linus Torvalds 已提交
902
		return NULL;
903 904

	t = *d->tgt;
L
Linus Torvalds 已提交
905 906

	/* initialize the headers & frame */
907
	skb = f->skb;
908
	h = (struct aoe_hdr *) skb_mac_header(skb);
L
Linus Torvalds 已提交
909
	ah = (struct aoe_atahdr *) (h+1);
910 911
	skb_put(skb, sizeof *h + sizeof *ah);
	memset(h, 0, skb->len);
912 913
	f->tag = aoehdr_atainit(d, t, h);
	t->nout++;
L
Linus Torvalds 已提交
914 915 916 917
	f->waited = 0;

	/* set up ata header */
	ah->scnt = 1;
918
	ah->cmdstat = ATA_CMD_ID_ATA;
L
Linus Torvalds 已提交
919 920
	ah->lba3 = 0xa0;

921
	skb->dev = t->ifp->nd;
L
Linus Torvalds 已提交
922

923
	d->rttavg = MAXTIMER;
L
Linus Torvalds 已提交
924 925
	d->timer.function = rexmit_timer;

926
	return skb_clone(skb, GFP_ATOMIC);
L
Linus Torvalds 已提交
927 928
}
 
929 930 931 932 933 934 935 936 937 938 939
static struct aoetgt *
addtgt(struct aoedev *d, char *addr, ulong nframes)
{
	struct aoetgt *t, **tt, **te;
	struct frame *f, *e;

	tt = d->targets;
	te = tt + NTARGETS;
	for (; tt < te && *tt; tt++)
		;

940 941 942
	if (tt == te) {
		printk(KERN_INFO
			"aoe: device addtgt failure; too many targets\n");
943
		return NULL;
944
	}
945 946
	t = kcalloc(1, sizeof *t, GFP_ATOMIC);
	f = kcalloc(nframes, sizeof *f, GFP_ATOMIC);
947 948
	if (!t || !f) {
		kfree(f);
949
		kfree(t);
950
		printk(KERN_INFO "aoe: cannot allocate memory to add target\n");
951 952 953
		return NULL;
	}

954 955 956
	t->nframes = nframes;
	t->frames = f;
	e = f + nframes;
957
	for (; f < e; f++)
958 959 960 961 962 963 964
		f->tag = FREETAG;
	memcpy(t->addr, addr, sizeof t->addr);
	t->ifp = t->ifs;
	t->maxout = t->nframes;
	return *tt = t;
}

L
Linus Torvalds 已提交
965 966 967 968 969 970
void
aoecmd_cfg_rsp(struct sk_buff *skb)
{
	struct aoedev *d;
	struct aoe_hdr *h;
	struct aoe_cfghdr *ch;
971 972
	struct aoetgt *t;
	struct aoeif *ifp;
973
	ulong flags, sysminor, aoemajor;
L
Linus Torvalds 已提交
974
	struct sk_buff *sl;
975
	u16 n;
L
Linus Torvalds 已提交
976

977
	h = (struct aoe_hdr *) skb_mac_header(skb);
L
Linus Torvalds 已提交
978 979 980 981 982 983
	ch = (struct aoe_cfghdr *) (h+1);

	/*
	 * Enough people have their dip switches set backwards to
	 * warrant a loud message for this special case.
	 */
984
	aoemajor = get_unaligned_be16(&h->major);
L
Linus Torvalds 已提交
985
	if (aoemajor == 0xfff) {
E
Ed L. Cashin 已提交
986
		printk(KERN_ERR "aoe: Warning: shelf address is all ones.  "
987
			"Check shelf dip switches.\n");
L
Linus Torvalds 已提交
988 989 990 991
		return;
	}

	sysminor = SYSMINOR(aoemajor, h->minor);
992
	if (sysminor * AOE_PARTITIONS + AOE_PARTITIONS > MINORMASK) {
E
Ed L. Cashin 已提交
993
		printk(KERN_INFO "aoe: e%ld.%d: minor number too large\n",
994
			aoemajor, (int) h->minor);
L
Linus Torvalds 已提交
995 996 997
		return;
	}

998
	n = be16_to_cpu(ch->bufcnt);
999 1000
	if (n > aoe_maxout)	/* keep it reasonable */
		n = aoe_maxout;
L
Linus Torvalds 已提交
1001

1002
	d = aoedev_by_sysminor_m(sysminor);
L
Linus Torvalds 已提交
1003
	if (d == NULL) {
E
Ed L. Cashin 已提交
1004
		printk(KERN_INFO "aoe: device sysminor_m failure\n");
L
Linus Torvalds 已提交
1005 1006 1007 1008 1009
		return;
	}

	spin_lock_irqsave(&d->lock, flags);

1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030
	t = gettgt(d, h->src);
	if (!t) {
		t = addtgt(d, h->src, n);
		if (!t) {
			spin_unlock_irqrestore(&d->lock, flags);
			return;
		}
	}
	ifp = getif(t, skb->dev);
	if (!ifp) {
		ifp = addif(t, skb->dev);
		if (!ifp) {
			printk(KERN_INFO
				"aoe: device addif failure; "
				"too many interfaces?\n");
			spin_unlock_irqrestore(&d->lock, flags);
			return;
		}
	}
	if (ifp->maxbcnt) {
		n = ifp->nd->mtu;
1031 1032 1033 1034
		n -= sizeof (struct aoe_hdr) + sizeof (struct aoe_atahdr);
		n /= 512;
		if (n > ch->scnt)
			n = ch->scnt;
1035
		n = n ? n * 512 : DEFAULTBCNT;
1036
		if (n != ifp->maxbcnt) {
E
Ed L. Cashin 已提交
1037
			printk(KERN_INFO
1038
				"aoe: e%ld.%d: setting %d%s%s:%pm\n",
1039 1040
				d->aoemajor, d->aoeminor, n,
				" byte data frames on ", ifp->nd->name,
1041
				t->addr);
1042
			ifp->maxbcnt = n;
1043
		}
1044
	}
1045 1046

	/* don't change users' perspective */
1047
	if (d->nopen) {
L
Linus Torvalds 已提交
1048 1049 1050
		spin_unlock_irqrestore(&d->lock, flags);
		return;
	}
1051
	d->fw_ver = be16_to_cpu(ch->fwver);
L
Linus Torvalds 已提交
1052

1053
	sl = aoecmd_ata_id(d);
L
Linus Torvalds 已提交
1054 1055 1056

	spin_unlock_irqrestore(&d->lock, flags);

1057 1058 1059 1060 1061 1062
	if (sl) {
		struct sk_buff_head queue;
		__skb_queue_head_init(&queue);
		__skb_queue_tail(&queue, sl);
		aoenet_xmit(&queue);
	}
L
Linus Torvalds 已提交
1063 1064
}

1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
void
aoecmd_cleanslate(struct aoedev *d)
{
	struct aoetgt **t, **te;
	struct aoeif *p, *e;

	d->mintimer = MINTIMER;

	t = d->targets;
	te = t + NTARGETS;
	for (; t < te && *t; t++) {
		(*t)->maxout = (*t)->nframes;
		p = (*t)->ifs;
		e = p + NAOEIFS;
		for (; p < e; p++) {
			p->lostjumbo = 0;
			p->lost = 0;
			p->maxbcnt = DEFAULTBCNT;
		}
	}
}
反馈
建议
客服 返回
顶部