pci_ahci.c revision 280731
1/*-
2 * Copyright (c) 2013  Zhixiang Yu <zcore@freebsd.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD: stable/10/usr.sbin/bhyve/pci_ahci.c 280731 2015-03-27 08:41:49Z mav $
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD: stable/10/usr.sbin/bhyve/pci_ahci.c 280731 2015-03-27 08:41:49Z mav $");
31
32#include <sys/param.h>
33#include <sys/linker_set.h>
34#include <sys/stat.h>
35#include <sys/uio.h>
36#include <sys/ioctl.h>
37#include <sys/disk.h>
38#include <sys/ata.h>
39#include <sys/endian.h>
40
41#include <errno.h>
42#include <fcntl.h>
43#include <stdio.h>
44#include <stdlib.h>
45#include <stdint.h>
46#include <string.h>
47#include <strings.h>
48#include <unistd.h>
49#include <assert.h>
50#include <pthread.h>
51#include <pthread_np.h>
52#include <inttypes.h>
53
54#include "bhyverun.h"
55#include "pci_emul.h"
56#include "ahci.h"
57#include "block_if.h"
58
59#define	MAX_PORTS	6	/* Intel ICH8 AHCI supports 6 ports */
60
61#define	PxSIG_ATA	0x00000101 /* ATA drive */
62#define	PxSIG_ATAPI	0xeb140101 /* ATAPI drive */
63
64enum sata_fis_type {
65	FIS_TYPE_REGH2D		= 0x27,	/* Register FIS - host to device */
66	FIS_TYPE_REGD2H		= 0x34,	/* Register FIS - device to host */
67	FIS_TYPE_DMAACT		= 0x39,	/* DMA activate FIS - device to host */
68	FIS_TYPE_DMASETUP	= 0x41,	/* DMA setup FIS - bidirectional */
69	FIS_TYPE_DATA		= 0x46,	/* Data FIS - bidirectional */
70	FIS_TYPE_BIST		= 0x58,	/* BIST activate FIS - bidirectional */
71	FIS_TYPE_PIOSETUP	= 0x5F,	/* PIO setup FIS - device to host */
72	FIS_TYPE_SETDEVBITS	= 0xA1,	/* Set dev bits FIS - device to host */
73};
74
75/*
76 * SCSI opcodes
77 */
78#define	TEST_UNIT_READY		0x00
79#define	REQUEST_SENSE		0x03
80#define	INQUIRY			0x12
81#define	START_STOP_UNIT		0x1B
82#define	PREVENT_ALLOW		0x1E
83#define	READ_CAPACITY		0x25
84#define	READ_10			0x28
85#define	POSITION_TO_ELEMENT	0x2B
86#define	READ_TOC		0x43
87#define	GET_EVENT_STATUS_NOTIFICATION 0x4A
88#define	MODE_SENSE_10		0x5A
89#define	READ_12			0xA8
90#define	READ_CD			0xBE
91
92/*
93 * SCSI mode page codes
94 */
95#define	MODEPAGE_RW_ERROR_RECOVERY	0x01
96#define	MODEPAGE_CD_CAPABILITIES	0x2A
97
98/*
99 * ATA commands
100 */
101#define	ATA_SF_ENAB_SATA_SF		0x10
102#define		ATA_SATA_SF_AN		0x05
103#define	ATA_SF_DIS_SATA_SF		0x90
104
105/*
106 * Debug printf
107 */
108#ifdef AHCI_DEBUG
109static FILE *dbg;
110#define DPRINTF(format, arg...)	do{fprintf(dbg, format, ##arg);fflush(dbg);}while(0)
111#else
112#define DPRINTF(format, arg...)
113#endif
114#define WPRINTF(format, arg...) printf(format, ##arg)
115
116struct ahci_ioreq {
117	struct blockif_req io_req;
118	struct ahci_port *io_pr;
119	STAILQ_ENTRY(ahci_ioreq) io_flist;
120	TAILQ_ENTRY(ahci_ioreq) io_blist;
121	uint8_t *cfis;
122	uint32_t len;
123	uint32_t done;
124	int slot;
125	int prdtl;
126};
127
128struct ahci_port {
129	struct blockif_ctxt *bctx;
130	struct pci_ahci_softc *pr_sc;
131	uint8_t *cmd_lst;
132	uint8_t *rfis;
133	int atapi;
134	int reset;
135	int mult_sectors;
136	uint8_t xfermode;
137	uint8_t sense_key;
138	uint8_t asc;
139	uint32_t pending;
140
141	uint32_t clb;
142	uint32_t clbu;
143	uint32_t fb;
144	uint32_t fbu;
145	uint32_t is;
146	uint32_t ie;
147	uint32_t cmd;
148	uint32_t unused0;
149	uint32_t tfd;
150	uint32_t sig;
151	uint32_t ssts;
152	uint32_t sctl;
153	uint32_t serr;
154	uint32_t sact;
155	uint32_t ci;
156	uint32_t sntf;
157	uint32_t fbs;
158
159	/*
160	 * i/o request info
161	 */
162	struct ahci_ioreq *ioreq;
163	int ioqsz;
164	STAILQ_HEAD(ahci_fhead, ahci_ioreq) iofhd;
165	TAILQ_HEAD(ahci_bhead, ahci_ioreq) iobhd;
166};
167
168struct ahci_cmd_hdr {
169	uint16_t flags;
170	uint16_t prdtl;
171	uint32_t prdbc;
172	uint64_t ctba;
173	uint32_t reserved[4];
174};
175
176struct ahci_prdt_entry {
177	uint64_t dba;
178	uint32_t reserved;
179#define	DBCMASK		0x3fffff
180	uint32_t dbc;
181};
182
183struct pci_ahci_softc {
184	struct pci_devinst *asc_pi;
185	pthread_mutex_t	mtx;
186	int ports;
187	uint32_t cap;
188	uint32_t ghc;
189	uint32_t is;
190	uint32_t pi;
191	uint32_t vs;
192	uint32_t ccc_ctl;
193	uint32_t ccc_pts;
194	uint32_t em_loc;
195	uint32_t em_ctl;
196	uint32_t cap2;
197	uint32_t bohc;
198	uint32_t lintr;
199	struct ahci_port port[MAX_PORTS];
200};
201#define	ahci_ctx(sc)	((sc)->asc_pi->pi_vmctx)
202
203static inline void lba_to_msf(uint8_t *buf, int lba)
204{
205	lba += 150;
206	buf[0] = (lba / 75) / 60;
207	buf[1] = (lba / 75) % 60;
208	buf[2] = lba % 75;
209}
210
211/*
212 * generate HBA intr depending on whether or not ports within
213 * the controller have an interrupt pending.
214 */
215static void
216ahci_generate_intr(struct pci_ahci_softc *sc)
217{
218	struct pci_devinst *pi;
219	int i;
220
221	pi = sc->asc_pi;
222
223	for (i = 0; i < sc->ports; i++) {
224		struct ahci_port *pr;
225		pr = &sc->port[i];
226		if (pr->is & pr->ie)
227			sc->is |= (1 << i);
228	}
229
230	DPRINTF("%s %x\n", __func__, sc->is);
231
232	if (sc->is && (sc->ghc & AHCI_GHC_IE)) {
233		if (pci_msi_enabled(pi)) {
234			/*
235			 * Generate an MSI interrupt on every edge
236			 */
237			pci_generate_msi(pi, 0);
238		} else if (!sc->lintr) {
239			/*
240			 * Only generate a pin-based interrupt if one wasn't
241			 * in progress
242			 */
243			sc->lintr = 1;
244			pci_lintr_assert(pi);
245		}
246	} else if (sc->lintr) {
247		/*
248		 * No interrupts: deassert pin-based signal if it had
249		 * been asserted
250		 */
251		pci_lintr_deassert(pi);
252		sc->lintr = 0;
253	}
254}
255
256static void
257ahci_write_fis(struct ahci_port *p, enum sata_fis_type ft, uint8_t *fis)
258{
259	int offset, len, irq;
260
261	if (p->rfis == NULL || !(p->cmd & AHCI_P_CMD_FRE))
262		return;
263
264	switch (ft) {
265	case FIS_TYPE_REGD2H:
266		offset = 0x40;
267		len = 20;
268		irq = AHCI_P_IX_DHR;
269		break;
270	case FIS_TYPE_SETDEVBITS:
271		offset = 0x58;
272		len = 8;
273		irq = AHCI_P_IX_SDB;
274		break;
275	case FIS_TYPE_PIOSETUP:
276		offset = 0x20;
277		len = 20;
278		irq = 0;
279		break;
280	default:
281		WPRINTF("unsupported fis type %d\n", ft);
282		return;
283	}
284	memcpy(p->rfis + offset, fis, len);
285	if (irq) {
286		p->is |= irq;
287		ahci_generate_intr(p->pr_sc);
288	}
289}
290
291static void
292ahci_write_fis_piosetup(struct ahci_port *p)
293{
294	uint8_t fis[20];
295
296	memset(fis, 0, sizeof(fis));
297	fis[0] = FIS_TYPE_PIOSETUP;
298	ahci_write_fis(p, FIS_TYPE_PIOSETUP, fis);
299}
300
301static void
302ahci_write_fis_sdb(struct ahci_port *p, int slot, uint32_t tfd)
303{
304	uint8_t fis[8];
305	uint8_t error;
306
307	error = (tfd >> 8) & 0xff;
308	memset(fis, 0, sizeof(fis));
309	fis[0] = error;
310	fis[2] = tfd & 0x77;
311	*(uint32_t *)(fis + 4) = (1 << slot);
312	if (fis[2] & ATA_S_ERROR)
313		p->is |= AHCI_P_IX_TFE;
314	p->tfd = tfd;
315	ahci_write_fis(p, FIS_TYPE_SETDEVBITS, fis);
316}
317
318static void
319ahci_write_fis_d2h(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t tfd)
320{
321	uint8_t fis[20];
322	uint8_t error;
323
324	error = (tfd >> 8) & 0xff;
325	memset(fis, 0, sizeof(fis));
326	fis[0] = FIS_TYPE_REGD2H;
327	fis[1] = (1 << 6);
328	fis[2] = tfd & 0xff;
329	fis[3] = error;
330	fis[4] = cfis[4];
331	fis[5] = cfis[5];
332	fis[6] = cfis[6];
333	fis[7] = cfis[7];
334	fis[8] = cfis[8];
335	fis[9] = cfis[9];
336	fis[10] = cfis[10];
337	fis[11] = cfis[11];
338	fis[12] = cfis[12];
339	fis[13] = cfis[13];
340	if (fis[2] & ATA_S_ERROR)
341		p->is |= AHCI_P_IX_TFE;
342	else
343		p->ci &= ~(1 << slot);
344	p->tfd = tfd;
345	ahci_write_fis(p, FIS_TYPE_REGD2H, fis);
346}
347
348static void
349ahci_write_reset_fis_d2h(struct ahci_port *p)
350{
351	uint8_t fis[20];
352
353	memset(fis, 0, sizeof(fis));
354	fis[0] = FIS_TYPE_REGD2H;
355	fis[3] = 1;
356	fis[4] = 1;
357	if (p->atapi) {
358		fis[5] = 0x14;
359		fis[6] = 0xeb;
360	}
361	fis[12] = 1;
362	ahci_write_fis(p, FIS_TYPE_REGD2H, fis);
363}
364
365static void
366ahci_check_stopped(struct ahci_port *p)
367{
368	/*
369	 * If we are no longer processing the command list and nothing
370	 * is in-flight, clear the running bit, the current command
371	 * slot, the command issue and active bits.
372	 */
373	if (!(p->cmd & AHCI_P_CMD_ST)) {
374		if (p->pending == 0) {
375			p->cmd &= ~(AHCI_P_CMD_CR | AHCI_P_CMD_CCS_MASK);
376			p->ci = 0;
377			p->sact = 0;
378		}
379	}
380}
381
382static void
383ahci_port_stop(struct ahci_port *p)
384{
385	struct ahci_ioreq *aior;
386	uint8_t *cfis;
387	int slot;
388	int ncq;
389	int error;
390
391	assert(pthread_mutex_isowned_np(&p->pr_sc->mtx));
392
393	TAILQ_FOREACH(aior, &p->iobhd, io_blist) {
394		/*
395		 * Try to cancel the outstanding blockif request.
396		 */
397		error = blockif_cancel(p->bctx, &aior->io_req);
398		if (error != 0)
399			continue;
400
401		slot = aior->slot;
402		cfis = aior->cfis;
403		if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
404		    cfis[2] == ATA_READ_FPDMA_QUEUED)
405			ncq = 1;
406
407		if (ncq)
408			p->sact &= ~(1 << slot);
409		else
410			p->ci &= ~(1 << slot);
411
412		/*
413		 * This command is now done.
414		 */
415		p->pending &= ~(1 << slot);
416
417		/*
418		 * Delete the blockif request from the busy list
419		 */
420		TAILQ_REMOVE(&p->iobhd, aior, io_blist);
421
422		/*
423		 * Move the blockif request back to the free list
424		 */
425		STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
426	}
427
428	ahci_check_stopped(p);
429}
430
431static void
432ahci_port_reset(struct ahci_port *pr)
433{
434	pr->sctl = 0;
435	pr->serr = 0;
436	pr->sact = 0;
437	pr->xfermode = ATA_UDMA6;
438	pr->mult_sectors = 128;
439
440	if (!pr->bctx) {
441		pr->ssts = ATA_SS_DET_NO_DEVICE;
442		pr->sig = 0xFFFFFFFF;
443		pr->tfd = 0x7F;
444		return;
445	}
446	pr->ssts = ATA_SS_DET_PHY_ONLINE | ATA_SS_SPD_GEN2 |
447		ATA_SS_IPM_ACTIVE;
448	pr->tfd = (1 << 8) | ATA_S_DSC | ATA_S_DMA;
449	if (!pr->atapi) {
450		pr->sig = PxSIG_ATA;
451		pr->tfd |= ATA_S_READY;
452	} else
453		pr->sig = PxSIG_ATAPI;
454	ahci_write_reset_fis_d2h(pr);
455}
456
457static void
458ahci_reset(struct pci_ahci_softc *sc)
459{
460	int i;
461
462	sc->ghc = AHCI_GHC_AE;
463	sc->is = 0;
464
465	if (sc->lintr) {
466		pci_lintr_deassert(sc->asc_pi);
467		sc->lintr = 0;
468	}
469
470	for (i = 0; i < sc->ports; i++) {
471		sc->port[i].ie = 0;
472		sc->port[i].is = 0;
473		ahci_port_reset(&sc->port[i]);
474	}
475}
476
477static void
478ata_string(uint8_t *dest, const char *src, int len)
479{
480	int i;
481
482	for (i = 0; i < len; i++) {
483		if (*src)
484			dest[i ^ 1] = *src++;
485		else
486			dest[i ^ 1] = ' ';
487	}
488}
489
490static void
491atapi_string(uint8_t *dest, const char *src, int len)
492{
493	int i;
494
495	for (i = 0; i < len; i++) {
496		if (*src)
497			dest[i] = *src++;
498		else
499			dest[i] = ' ';
500	}
501}
502
503static void
504ahci_handle_dma(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t done,
505    int seek)
506{
507	struct ahci_ioreq *aior;
508	struct blockif_req *breq;
509	struct pci_ahci_softc *sc;
510	struct ahci_prdt_entry *prdt;
511	struct ahci_cmd_hdr *hdr;
512	uint64_t lba;
513	uint32_t len;
514	int i, err, iovcnt, ncq, readop;
515
516	sc = p->pr_sc;
517	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
518	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
519	ncq = 0;
520	readop = 1;
521
522	prdt += seek;
523	if (cfis[2] == ATA_WRITE_DMA || cfis[2] == ATA_WRITE_DMA48 ||
524			cfis[2] == ATA_WRITE_FPDMA_QUEUED)
525		readop = 0;
526
527	if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
528			cfis[2] == ATA_READ_FPDMA_QUEUED) {
529		lba = ((uint64_t)cfis[10] << 40) |
530			((uint64_t)cfis[9] << 32) |
531			((uint64_t)cfis[8] << 24) |
532			((uint64_t)cfis[6] << 16) |
533			((uint64_t)cfis[5] << 8) |
534			cfis[4];
535		len = cfis[11] << 8 | cfis[3];
536		if (!len)
537			len = 65536;
538		ncq = 1;
539	} else if (cfis[2] == ATA_READ_DMA48 || cfis[2] == ATA_WRITE_DMA48) {
540		lba = ((uint64_t)cfis[10] << 40) |
541			((uint64_t)cfis[9] << 32) |
542			((uint64_t)cfis[8] << 24) |
543			((uint64_t)cfis[6] << 16) |
544			((uint64_t)cfis[5] << 8) |
545			cfis[4];
546		len = cfis[13] << 8 | cfis[12];
547		if (!len)
548			len = 65536;
549	} else {
550		lba = ((cfis[7] & 0xf) << 24) | (cfis[6] << 16) |
551			(cfis[5] << 8) | cfis[4];
552		len = cfis[12];
553		if (!len)
554			len = 256;
555	}
556	lba *= blockif_sectsz(p->bctx);
557	len *= blockif_sectsz(p->bctx);
558
559	/*
560	 * Pull request off free list
561	 */
562	aior = STAILQ_FIRST(&p->iofhd);
563	assert(aior != NULL);
564	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
565	aior->cfis = cfis;
566	aior->slot = slot;
567	aior->len = len;
568	aior->done = done;
569	breq = &aior->io_req;
570	breq->br_offset = lba + done;
571	iovcnt = hdr->prdtl - seek;
572	if (iovcnt > BLOCKIF_IOV_MAX) {
573		aior->prdtl = iovcnt - BLOCKIF_IOV_MAX;
574		iovcnt = BLOCKIF_IOV_MAX;
575	} else
576		aior->prdtl = 0;
577	breq->br_iovcnt = iovcnt;
578
579	/*
580	 * Mark this command in-flight.
581	 */
582	p->pending |= 1 << slot;
583
584	/*
585	 * Stuff request onto busy list
586	 */
587	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
588
589	/*
590	 * Build up the iovec based on the prdt
591	 */
592	for (i = 0; i < iovcnt; i++) {
593		uint32_t dbcsz;
594
595		dbcsz = (prdt->dbc & DBCMASK) + 1;
596		breq->br_iov[i].iov_base = paddr_guest2host(ahci_ctx(sc),
597		    prdt->dba, dbcsz);
598		breq->br_iov[i].iov_len = dbcsz;
599		aior->done += dbcsz;
600		prdt++;
601	}
602	if (readop)
603		err = blockif_read(p->bctx, breq);
604	else
605		err = blockif_write(p->bctx, breq);
606	assert(err == 0);
607
608	if (ncq)
609		p->ci &= ~(1 << slot);
610}
611
612static void
613ahci_handle_flush(struct ahci_port *p, int slot, uint8_t *cfis)
614{
615	struct ahci_ioreq *aior;
616	struct blockif_req *breq;
617	int err;
618
619	/*
620	 * Pull request off free list
621	 */
622	aior = STAILQ_FIRST(&p->iofhd);
623	assert(aior != NULL);
624	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
625	aior->cfis = cfis;
626	aior->slot = slot;
627	aior->len = 0;
628	aior->done = 0;
629	aior->prdtl = 0;
630	breq = &aior->io_req;
631
632	/*
633	 * Mark this command in-flight.
634	 */
635	p->pending |= 1 << slot;
636
637	/*
638	 * Stuff request onto busy list
639	 */
640	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
641
642	err = blockif_flush(p->bctx, breq);
643	assert(err == 0);
644}
645
646static inline void
647read_prdt(struct ahci_port *p, int slot, uint8_t *cfis,
648		void *buf, int size)
649{
650	struct ahci_cmd_hdr *hdr;
651	struct ahci_prdt_entry *prdt;
652	void *to;
653	int i, len;
654
655	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
656	len = size;
657	to = buf;
658	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
659	for (i = 0; i < hdr->prdtl && len; i++) {
660		uint8_t *ptr;
661		uint32_t dbcsz;
662		int sublen;
663
664		dbcsz = (prdt->dbc & DBCMASK) + 1;
665		ptr = paddr_guest2host(ahci_ctx(p->pr_sc), prdt->dba, dbcsz);
666		sublen = len < dbcsz ? len : dbcsz;
667		memcpy(to, ptr, sublen);
668		len -= sublen;
669		to += sublen;
670		prdt++;
671	}
672}
673
674static void
675ahci_handle_dsm_trim(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t done)
676{
677	struct ahci_ioreq *aior;
678	struct blockif_req *breq;
679	uint8_t *entry;
680	uint64_t elba;
681	uint32_t len, elen;
682	int err;
683	uint8_t buf[512];
684
685	len = (uint16_t)cfis[13] << 8 | cfis[12];
686	len *= 512;
687	read_prdt(p, slot, cfis, buf, sizeof(buf));
688
689next:
690	entry = &buf[done];
691	elba = ((uint64_t)entry[5] << 40) |
692		((uint64_t)entry[4] << 32) |
693		((uint64_t)entry[3] << 24) |
694		((uint64_t)entry[2] << 16) |
695		((uint64_t)entry[1] << 8) |
696		entry[0];
697	elen = (uint16_t)entry[7] << 8 | entry[6];
698	done += 8;
699	if (elen == 0) {
700		if (done >= len) {
701			ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
702			p->pending &= ~(1 << slot);
703			ahci_check_stopped(p);
704			return;
705		}
706		goto next;
707	}
708
709	/*
710	 * Pull request off free list
711	 */
712	aior = STAILQ_FIRST(&p->iofhd);
713	assert(aior != NULL);
714	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
715	aior->cfis = cfis;
716	aior->slot = slot;
717	aior->len = len;
718	aior->done = done;
719	aior->prdtl = 0;
720
721	breq = &aior->io_req;
722	breq->br_offset = elba * blockif_sectsz(p->bctx);
723	breq->br_iovcnt = 1;
724	breq->br_iov[0].iov_len = elen * blockif_sectsz(p->bctx);
725
726	/*
727	 * Mark this command in-flight.
728	 */
729	p->pending |= 1 << slot;
730
731	/*
732	 * Stuff request onto busy list
733	 */
734	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
735
736	err = blockif_delete(p->bctx, breq);
737	assert(err == 0);
738}
739
740static inline void
741write_prdt(struct ahci_port *p, int slot, uint8_t *cfis,
742		void *buf, int size)
743{
744	struct ahci_cmd_hdr *hdr;
745	struct ahci_prdt_entry *prdt;
746	void *from;
747	int i, len;
748
749	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
750	len = size;
751	from = buf;
752	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
753	for (i = 0; i < hdr->prdtl && len; i++) {
754		uint8_t *ptr;
755		uint32_t dbcsz;
756		int sublen;
757
758		dbcsz = (prdt->dbc & DBCMASK) + 1;
759		ptr = paddr_guest2host(ahci_ctx(p->pr_sc), prdt->dba, dbcsz);
760		sublen = len < dbcsz ? len : dbcsz;
761		memcpy(ptr, from, sublen);
762		len -= sublen;
763		from += sublen;
764		prdt++;
765	}
766	hdr->prdbc = size - len;
767}
768
769static void
770handle_identify(struct ahci_port *p, int slot, uint8_t *cfis)
771{
772	struct ahci_cmd_hdr *hdr;
773
774	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
775	if (p->atapi || hdr->prdtl == 0) {
776		ahci_write_fis_d2h(p, slot, cfis,
777		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
778	} else {
779		uint16_t buf[256];
780		uint64_t sectors;
781		int sectsz, psectsz, psectoff, candelete, ro;
782		uint16_t cyl;
783		uint8_t sech, heads;
784
785		ro = blockif_is_ro(p->bctx);
786		candelete = blockif_candelete(p->bctx);
787		sectsz = blockif_sectsz(p->bctx);
788		sectors = blockif_size(p->bctx) / sectsz;
789		blockif_chs(p->bctx, &cyl, &heads, &sech);
790		blockif_psectsz(p->bctx, &psectsz, &psectoff);
791		memset(buf, 0, sizeof(buf));
792		buf[0] = 0x0040;
793		buf[1] = cyl;
794		buf[3] = heads;
795		buf[6] = sech;
796		/* TODO emulate different serial? */
797		ata_string((uint8_t *)(buf+10), "123456", 20);
798		ata_string((uint8_t *)(buf+23), "001", 8);
799		ata_string((uint8_t *)(buf+27), "BHYVE SATA DISK", 40);
800		buf[47] = (0x8000 | 128);
801		buf[48] = 0x1;
802		buf[49] = (1 << 8 | 1 << 9 | 1 << 11);
803		buf[50] = (1 << 14);
804		buf[53] = (1 << 1 | 1 << 2);
805		if (p->mult_sectors)
806			buf[59] = (0x100 | p->mult_sectors);
807		buf[60] = sectors;
808		buf[61] = (sectors >> 16);
809		buf[63] = 0x7;
810		if (p->xfermode & ATA_WDMA0)
811			buf[63] |= (1 << ((p->xfermode & 7) + 8));
812		buf[64] = 0x3;
813		buf[65] = 100;
814		buf[66] = 100;
815		buf[67] = 100;
816		buf[68] = 100;
817		buf[69] = 0;
818		buf[75] = 31;
819		buf[76] = (1 << 8 | 1 << 2);
820		buf[80] = 0x1f0;
821		buf[81] = 0x28;
822		buf[82] = (1 << 5 | 1 << 14);
823		buf[83] = (1 << 10 | 1 << 12 | 1 << 13 | 1 << 14);
824		buf[84] = (1 << 14);
825		buf[85] = (1 << 5 | 1 << 14);
826		buf[86] = (1 << 10 | 1 << 12 | 1 << 13);
827		buf[87] = (1 << 14);
828		buf[88] = 0x7f;
829		if (p->xfermode & ATA_UDMA0)
830			buf[88] |= (1 << ((p->xfermode & 7) + 8));
831		buf[93] = (1 | 1 <<14);
832		buf[100] = sectors;
833		buf[101] = (sectors >> 16);
834		buf[102] = (sectors >> 32);
835		buf[103] = (sectors >> 48);
836		if (candelete && !ro) {
837			buf[69] |= ATA_SUPPORT_RZAT | ATA_SUPPORT_DRAT;
838			buf[105] = 1;
839			buf[169] = ATA_SUPPORT_DSM_TRIM;
840		}
841		buf[106] = 0x4000;
842		buf[209] = 0x4000;
843		if (psectsz > sectsz) {
844			buf[106] |= 0x2000;
845			buf[106] |= ffsl(psectsz / sectsz) - 1;
846			buf[209] |= (psectoff / sectsz);
847		}
848		if (sectsz > 512) {
849			buf[106] |= 0x1000;
850			buf[117] = sectsz / 2;
851			buf[118] = ((sectsz / 2) >> 16);
852		}
853		ahci_write_fis_piosetup(p);
854		write_prdt(p, slot, cfis, (void *)buf, sizeof(buf));
855		ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
856	}
857}
858
859static void
860handle_atapi_identify(struct ahci_port *p, int slot, uint8_t *cfis)
861{
862	if (!p->atapi) {
863		ahci_write_fis_d2h(p, slot, cfis,
864		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
865	} else {
866		uint16_t buf[256];
867
868		memset(buf, 0, sizeof(buf));
869		buf[0] = (2 << 14 | 5 << 8 | 1 << 7 | 2 << 5);
870		/* TODO emulate different serial? */
871		ata_string((uint8_t *)(buf+10), "123456", 20);
872		ata_string((uint8_t *)(buf+23), "001", 8);
873		ata_string((uint8_t *)(buf+27), "BHYVE SATA DVD ROM", 40);
874		buf[49] = (1 << 9 | 1 << 8);
875		buf[50] = (1 << 14 | 1);
876		buf[53] = (1 << 2 | 1 << 1);
877		buf[62] = 0x3f;
878		buf[63] = 7;
879		buf[64] = 3;
880		buf[65] = 100;
881		buf[66] = 100;
882		buf[67] = 100;
883		buf[68] = 100;
884		buf[76] = (1 << 2 | 1 << 1);
885		buf[78] = (1 << 5);
886		buf[80] = (0x1f << 4);
887		buf[82] = (1 << 4);
888		buf[83] = (1 << 14);
889		buf[84] = (1 << 14);
890		buf[85] = (1 << 4);
891		buf[87] = (1 << 14);
892		buf[88] = (1 << 14 | 0x7f);
893		ahci_write_fis_piosetup(p);
894		write_prdt(p, slot, cfis, (void *)buf, sizeof(buf));
895		ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
896	}
897}
898
899static void
900atapi_inquiry(struct ahci_port *p, int slot, uint8_t *cfis)
901{
902	uint8_t buf[36];
903	uint8_t *acmd;
904	int len;
905
906	acmd = cfis + 0x40;
907
908	buf[0] = 0x05;
909	buf[1] = 0x80;
910	buf[2] = 0x00;
911	buf[3] = 0x21;
912	buf[4] = 31;
913	buf[5] = 0;
914	buf[6] = 0;
915	buf[7] = 0;
916	atapi_string(buf + 8, "BHYVE", 8);
917	atapi_string(buf + 16, "BHYVE DVD-ROM", 16);
918	atapi_string(buf + 32, "001", 4);
919
920	len = sizeof(buf);
921	if (len > acmd[4])
922		len = acmd[4];
923	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
924	write_prdt(p, slot, cfis, buf, len);
925	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
926}
927
928static void
929atapi_read_capacity(struct ahci_port *p, int slot, uint8_t *cfis)
930{
931	uint8_t buf[8];
932	uint64_t sectors;
933
934	sectors = blockif_size(p->bctx) / 2048;
935	be32enc(buf, sectors - 1);
936	be32enc(buf + 4, 2048);
937	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
938	write_prdt(p, slot, cfis, buf, sizeof(buf));
939	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
940}
941
942static void
943atapi_read_toc(struct ahci_port *p, int slot, uint8_t *cfis)
944{
945	uint8_t *acmd;
946	uint8_t format;
947	int len;
948
949	acmd = cfis + 0x40;
950
951	len = be16dec(acmd + 7);
952	format = acmd[9] >> 6;
953	switch (format) {
954	case 0:
955	{
956		int msf, size;
957		uint64_t sectors;
958		uint8_t start_track, buf[20], *bp;
959
960		msf = (acmd[1] >> 1) & 1;
961		start_track = acmd[6];
962		if (start_track > 1 && start_track != 0xaa) {
963			uint32_t tfd;
964			p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
965			p->asc = 0x24;
966			tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
967			cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
968			ahci_write_fis_d2h(p, slot, cfis, tfd);
969			return;
970		}
971		bp = buf + 2;
972		*bp++ = 1;
973		*bp++ = 1;
974		if (start_track <= 1) {
975			*bp++ = 0;
976			*bp++ = 0x14;
977			*bp++ = 1;
978			*bp++ = 0;
979			if (msf) {
980				*bp++ = 0;
981				lba_to_msf(bp, 0);
982				bp += 3;
983			} else {
984				*bp++ = 0;
985				*bp++ = 0;
986				*bp++ = 0;
987				*bp++ = 0;
988			}
989		}
990		*bp++ = 0;
991		*bp++ = 0x14;
992		*bp++ = 0xaa;
993		*bp++ = 0;
994		sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx);
995		sectors >>= 2;
996		if (msf) {
997			*bp++ = 0;
998			lba_to_msf(bp, sectors);
999			bp += 3;
1000		} else {
1001			be32enc(bp, sectors);
1002			bp += 4;
1003		}
1004		size = bp - buf;
1005		be16enc(buf, size - 2);
1006		if (len > size)
1007			len = size;
1008		write_prdt(p, slot, cfis, buf, len);
1009		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1010		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1011		break;
1012	}
1013	case 1:
1014	{
1015		uint8_t buf[12];
1016
1017		memset(buf, 0, sizeof(buf));
1018		buf[1] = 0xa;
1019		buf[2] = 0x1;
1020		buf[3] = 0x1;
1021		if (len > sizeof(buf))
1022			len = sizeof(buf);
1023		write_prdt(p, slot, cfis, buf, len);
1024		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1025		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1026		break;
1027	}
1028	case 2:
1029	{
1030		int msf, size;
1031		uint64_t sectors;
1032		uint8_t start_track, *bp, buf[50];
1033
1034		msf = (acmd[1] >> 1) & 1;
1035		start_track = acmd[6];
1036		bp = buf + 2;
1037		*bp++ = 1;
1038		*bp++ = 1;
1039
1040		*bp++ = 1;
1041		*bp++ = 0x14;
1042		*bp++ = 0;
1043		*bp++ = 0xa0;
1044		*bp++ = 0;
1045		*bp++ = 0;
1046		*bp++ = 0;
1047		*bp++ = 0;
1048		*bp++ = 1;
1049		*bp++ = 0;
1050		*bp++ = 0;
1051
1052		*bp++ = 1;
1053		*bp++ = 0x14;
1054		*bp++ = 0;
1055		*bp++ = 0xa1;
1056		*bp++ = 0;
1057		*bp++ = 0;
1058		*bp++ = 0;
1059		*bp++ = 0;
1060		*bp++ = 1;
1061		*bp++ = 0;
1062		*bp++ = 0;
1063
1064		*bp++ = 1;
1065		*bp++ = 0x14;
1066		*bp++ = 0;
1067		*bp++ = 0xa2;
1068		*bp++ = 0;
1069		*bp++ = 0;
1070		*bp++ = 0;
1071		sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx);
1072		sectors >>= 2;
1073		if (msf) {
1074			*bp++ = 0;
1075			lba_to_msf(bp, sectors);
1076			bp += 3;
1077		} else {
1078			be32enc(bp, sectors);
1079			bp += 4;
1080		}
1081
1082		*bp++ = 1;
1083		*bp++ = 0x14;
1084		*bp++ = 0;
1085		*bp++ = 1;
1086		*bp++ = 0;
1087		*bp++ = 0;
1088		*bp++ = 0;
1089		if (msf) {
1090			*bp++ = 0;
1091			lba_to_msf(bp, 0);
1092			bp += 3;
1093		} else {
1094			*bp++ = 0;
1095			*bp++ = 0;
1096			*bp++ = 0;
1097			*bp++ = 0;
1098		}
1099
1100		size = bp - buf;
1101		be16enc(buf, size - 2);
1102		if (len > size)
1103			len = size;
1104		write_prdt(p, slot, cfis, buf, len);
1105		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1106		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1107		break;
1108	}
1109	default:
1110	{
1111		uint32_t tfd;
1112
1113		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1114		p->asc = 0x24;
1115		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1116		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1117		ahci_write_fis_d2h(p, slot, cfis, tfd);
1118		break;
1119	}
1120	}
1121}
1122
1123static void
1124atapi_read(struct ahci_port *p, int slot, uint8_t *cfis,
1125		uint32_t done, int seek)
1126{
1127	struct ahci_ioreq *aior;
1128	struct ahci_cmd_hdr *hdr;
1129	struct ahci_prdt_entry *prdt;
1130	struct blockif_req *breq;
1131	struct pci_ahci_softc *sc;
1132	uint8_t *acmd;
1133	uint64_t lba;
1134	uint32_t len;
1135	int i, err, iovcnt;
1136
1137	sc = p->pr_sc;
1138	acmd = cfis + 0x40;
1139	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1140	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
1141
1142	prdt += seek;
1143	lba = be32dec(acmd + 2);
1144	if (acmd[0] == READ_10)
1145		len = be16dec(acmd + 7);
1146	else
1147		len = be32dec(acmd + 6);
1148	if (len == 0) {
1149		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1150		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1151	}
1152	lba *= 2048;
1153	len *= 2048;
1154
1155	/*
1156	 * Pull request off free list
1157	 */
1158	aior = STAILQ_FIRST(&p->iofhd);
1159	assert(aior != NULL);
1160	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
1161	aior->cfis = cfis;
1162	aior->slot = slot;
1163	aior->len = len;
1164	aior->done = done;
1165	breq = &aior->io_req;
1166	breq->br_offset = lba + done;
1167	iovcnt = hdr->prdtl - seek;
1168	if (iovcnt > BLOCKIF_IOV_MAX) {
1169		aior->prdtl = iovcnt - BLOCKIF_IOV_MAX;
1170		iovcnt = BLOCKIF_IOV_MAX;
1171	} else
1172		aior->prdtl = 0;
1173	breq->br_iovcnt = iovcnt;
1174
1175	/*
1176	 * Mark this command in-flight.
1177	 */
1178	p->pending |= 1 << slot;
1179
1180	/*
1181	 * Stuff request onto busy list
1182	 */
1183	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
1184
1185	/*
1186	 * Build up the iovec based on the prdt
1187	 */
1188	for (i = 0; i < iovcnt; i++) {
1189		uint32_t dbcsz;
1190
1191		dbcsz = (prdt->dbc & DBCMASK) + 1;
1192		breq->br_iov[i].iov_base = paddr_guest2host(ahci_ctx(sc),
1193		    prdt->dba, dbcsz);
1194		breq->br_iov[i].iov_len = dbcsz;
1195		aior->done += dbcsz;
1196		prdt++;
1197	}
1198	err = blockif_read(p->bctx, breq);
1199	assert(err == 0);
1200}
1201
1202static void
1203atapi_request_sense(struct ahci_port *p, int slot, uint8_t *cfis)
1204{
1205	uint8_t buf[64];
1206	uint8_t *acmd;
1207	int len;
1208
1209	acmd = cfis + 0x40;
1210	len = acmd[4];
1211	if (len > sizeof(buf))
1212		len = sizeof(buf);
1213	memset(buf, 0, len);
1214	buf[0] = 0x70 | (1 << 7);
1215	buf[2] = p->sense_key;
1216	buf[7] = 10;
1217	buf[12] = p->asc;
1218	write_prdt(p, slot, cfis, buf, len);
1219	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1220	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1221}
1222
1223static void
1224atapi_start_stop_unit(struct ahci_port *p, int slot, uint8_t *cfis)
1225{
1226	uint8_t *acmd = cfis + 0x40;
1227	uint32_t tfd;
1228
1229	switch (acmd[4] & 3) {
1230	case 0:
1231	case 1:
1232	case 3:
1233		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1234		tfd = ATA_S_READY | ATA_S_DSC;
1235		break;
1236	case 2:
1237		/* TODO eject media */
1238		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1239		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1240		p->asc = 0x53;
1241		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1242		break;
1243	}
1244	ahci_write_fis_d2h(p, slot, cfis, tfd);
1245}
1246
1247static void
1248atapi_mode_sense(struct ahci_port *p, int slot, uint8_t *cfis)
1249{
1250	uint8_t *acmd;
1251	uint32_t tfd;
1252	uint8_t pc, code;
1253	int len;
1254
1255	acmd = cfis + 0x40;
1256	len = be16dec(acmd + 7);
1257	pc = acmd[2] >> 6;
1258	code = acmd[2] & 0x3f;
1259
1260	switch (pc) {
1261	case 0:
1262		switch (code) {
1263		case MODEPAGE_RW_ERROR_RECOVERY:
1264		{
1265			uint8_t buf[16];
1266
1267			if (len > sizeof(buf))
1268				len = sizeof(buf);
1269
1270			memset(buf, 0, sizeof(buf));
1271			be16enc(buf, 16 - 2);
1272			buf[2] = 0x70;
1273			buf[8] = 0x01;
1274			buf[9] = 16 - 10;
1275			buf[11] = 0x05;
1276			write_prdt(p, slot, cfis, buf, len);
1277			tfd = ATA_S_READY | ATA_S_DSC;
1278			break;
1279		}
1280		case MODEPAGE_CD_CAPABILITIES:
1281		{
1282			uint8_t buf[30];
1283
1284			if (len > sizeof(buf))
1285				len = sizeof(buf);
1286
1287			memset(buf, 0, sizeof(buf));
1288			be16enc(buf, 30 - 2);
1289			buf[2] = 0x70;
1290			buf[8] = 0x2A;
1291			buf[9] = 30 - 10;
1292			buf[10] = 0x08;
1293			buf[12] = 0x71;
1294			be16enc(&buf[18], 2);
1295			be16enc(&buf[20], 512);
1296			write_prdt(p, slot, cfis, buf, len);
1297			tfd = ATA_S_READY | ATA_S_DSC;
1298			break;
1299		}
1300		default:
1301			goto error;
1302			break;
1303		}
1304		break;
1305	case 3:
1306		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1307		p->asc = 0x39;
1308		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1309		break;
1310error:
1311	case 1:
1312	case 2:
1313		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1314		p->asc = 0x24;
1315		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1316		break;
1317	}
1318	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1319	ahci_write_fis_d2h(p, slot, cfis, tfd);
1320}
1321
1322static void
1323atapi_get_event_status_notification(struct ahci_port *p, int slot,
1324    uint8_t *cfis)
1325{
1326	uint8_t *acmd;
1327	uint32_t tfd;
1328
1329	acmd = cfis + 0x40;
1330
1331	/* we don't support asynchronous operation */
1332	if (!(acmd[1] & 1)) {
1333		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1334		p->asc = 0x24;
1335		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1336	} else {
1337		uint8_t buf[8];
1338		int len;
1339
1340		len = be16dec(acmd + 7);
1341		if (len > sizeof(buf))
1342			len = sizeof(buf);
1343
1344		memset(buf, 0, sizeof(buf));
1345		be16enc(buf, 8 - 2);
1346		buf[2] = 0x04;
1347		buf[3] = 0x10;
1348		buf[5] = 0x02;
1349		write_prdt(p, slot, cfis, buf, len);
1350		tfd = ATA_S_READY | ATA_S_DSC;
1351	}
1352	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1353	ahci_write_fis_d2h(p, slot, cfis, tfd);
1354}
1355
1356static void
1357handle_packet_cmd(struct ahci_port *p, int slot, uint8_t *cfis)
1358{
1359	uint8_t *acmd;
1360
1361	acmd = cfis + 0x40;
1362
1363#ifdef AHCI_DEBUG
1364	{
1365		int i;
1366		DPRINTF("ACMD:");
1367		for (i = 0; i < 16; i++)
1368			DPRINTF("%02x ", acmd[i]);
1369		DPRINTF("\n");
1370	}
1371#endif
1372
1373	switch (acmd[0]) {
1374	case TEST_UNIT_READY:
1375		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1376		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1377		break;
1378	case INQUIRY:
1379		atapi_inquiry(p, slot, cfis);
1380		break;
1381	case READ_CAPACITY:
1382		atapi_read_capacity(p, slot, cfis);
1383		break;
1384	case PREVENT_ALLOW:
1385		/* TODO */
1386		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1387		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1388		break;
1389	case READ_TOC:
1390		atapi_read_toc(p, slot, cfis);
1391		break;
1392	case READ_10:
1393	case READ_12:
1394		atapi_read(p, slot, cfis, 0, 0);
1395		break;
1396	case REQUEST_SENSE:
1397		atapi_request_sense(p, slot, cfis);
1398		break;
1399	case START_STOP_UNIT:
1400		atapi_start_stop_unit(p, slot, cfis);
1401		break;
1402	case MODE_SENSE_10:
1403		atapi_mode_sense(p, slot, cfis);
1404		break;
1405	case GET_EVENT_STATUS_NOTIFICATION:
1406		atapi_get_event_status_notification(p, slot, cfis);
1407		break;
1408	default:
1409		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1410		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1411		p->asc = 0x20;
1412		ahci_write_fis_d2h(p, slot, cfis, (p->sense_key << 12) |
1413				ATA_S_READY | ATA_S_ERROR);
1414		break;
1415	}
1416}
1417
1418static void
1419ahci_handle_cmd(struct ahci_port *p, int slot, uint8_t *cfis)
1420{
1421
1422	switch (cfis[2]) {
1423	case ATA_ATA_IDENTIFY:
1424		handle_identify(p, slot, cfis);
1425		break;
1426	case ATA_SETFEATURES:
1427	{
1428		switch (cfis[3]) {
1429		case ATA_SF_ENAB_SATA_SF:
1430			switch (cfis[12]) {
1431			case ATA_SATA_SF_AN:
1432				p->tfd = ATA_S_DSC | ATA_S_READY;
1433				break;
1434			default:
1435				p->tfd = ATA_S_ERROR | ATA_S_READY;
1436				p->tfd |= (ATA_ERROR_ABORT << 8);
1437				break;
1438			}
1439			break;
1440		case ATA_SF_ENAB_WCACHE:
1441		case ATA_SF_DIS_WCACHE:
1442		case ATA_SF_ENAB_RCACHE:
1443		case ATA_SF_DIS_RCACHE:
1444			p->tfd = ATA_S_DSC | ATA_S_READY;
1445			break;
1446		case ATA_SF_SETXFER:
1447		{
1448			switch (cfis[12] & 0xf8) {
1449			case ATA_PIO:
1450			case ATA_PIO0:
1451				break;
1452			case ATA_WDMA0:
1453			case ATA_UDMA0:
1454				p->xfermode = (cfis[12] & 0x7);
1455				break;
1456			}
1457			p->tfd = ATA_S_DSC | ATA_S_READY;
1458			break;
1459		}
1460		default:
1461			p->tfd = ATA_S_ERROR | ATA_S_READY;
1462			p->tfd |= (ATA_ERROR_ABORT << 8);
1463			break;
1464		}
1465		ahci_write_fis_d2h(p, slot, cfis, p->tfd);
1466		break;
1467	}
1468	case ATA_SET_MULTI:
1469		if (cfis[12] != 0 &&
1470			(cfis[12] > 128 || (cfis[12] & (cfis[12] - 1)))) {
1471			p->tfd = ATA_S_ERROR | ATA_S_READY;
1472			p->tfd |= (ATA_ERROR_ABORT << 8);
1473		} else {
1474			p->mult_sectors = cfis[12];
1475			p->tfd = ATA_S_DSC | ATA_S_READY;
1476		}
1477		ahci_write_fis_d2h(p, slot, cfis, p->tfd);
1478		break;
1479	case ATA_READ_DMA:
1480	case ATA_WRITE_DMA:
1481	case ATA_READ_DMA48:
1482	case ATA_WRITE_DMA48:
1483	case ATA_READ_FPDMA_QUEUED:
1484	case ATA_WRITE_FPDMA_QUEUED:
1485		ahci_handle_dma(p, slot, cfis, 0, 0);
1486		break;
1487	case ATA_FLUSHCACHE:
1488	case ATA_FLUSHCACHE48:
1489		ahci_handle_flush(p, slot, cfis);
1490		break;
1491	case ATA_DATA_SET_MANAGEMENT:
1492		if (cfis[11] == 0 && cfis[3] == ATA_DSM_TRIM &&
1493		    cfis[13] == 0 && cfis[12] == 1) {
1494			ahci_handle_dsm_trim(p, slot, cfis, 0);
1495			break;
1496		}
1497		ahci_write_fis_d2h(p, slot, cfis,
1498		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1499		break;
1500	case ATA_STANDBY_CMD:
1501		break;
1502	case ATA_NOP:
1503	case ATA_STANDBY_IMMEDIATE:
1504	case ATA_IDLE_IMMEDIATE:
1505	case ATA_SLEEP:
1506		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1507		break;
1508	case ATA_ATAPI_IDENTIFY:
1509		handle_atapi_identify(p, slot, cfis);
1510		break;
1511	case ATA_PACKET_CMD:
1512		if (!p->atapi) {
1513			ahci_write_fis_d2h(p, slot, cfis,
1514			    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1515		} else
1516			handle_packet_cmd(p, slot, cfis);
1517		break;
1518	default:
1519		WPRINTF("Unsupported cmd:%02x\n", cfis[2]);
1520		ahci_write_fis_d2h(p, slot, cfis,
1521		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1522		break;
1523	}
1524}
1525
1526static void
1527ahci_handle_slot(struct ahci_port *p, int slot)
1528{
1529	struct ahci_cmd_hdr *hdr;
1530	struct ahci_prdt_entry *prdt;
1531	struct pci_ahci_softc *sc;
1532	uint8_t *cfis;
1533	int cfl;
1534
1535	sc = p->pr_sc;
1536	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1537	cfl = (hdr->flags & 0x1f) * 4;
1538	cfis = paddr_guest2host(ahci_ctx(sc), hdr->ctba,
1539			0x80 + hdr->prdtl * sizeof(struct ahci_prdt_entry));
1540	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
1541
1542#ifdef AHCI_DEBUG
1543	DPRINTF("\ncfis:");
1544	for (i = 0; i < cfl; i++) {
1545		if (i % 10 == 0)
1546			DPRINTF("\n");
1547		DPRINTF("%02x ", cfis[i]);
1548	}
1549	DPRINTF("\n");
1550
1551	for (i = 0; i < hdr->prdtl; i++) {
1552		DPRINTF("%d@%08"PRIx64"\n", prdt->dbc & 0x3fffff, prdt->dba);
1553		prdt++;
1554	}
1555#endif
1556
1557	if (cfis[0] != FIS_TYPE_REGH2D) {
1558		WPRINTF("Not a H2D FIS:%02x\n", cfis[0]);
1559		return;
1560	}
1561
1562	if (cfis[1] & 0x80) {
1563		ahci_handle_cmd(p, slot, cfis);
1564	} else {
1565		if (cfis[15] & (1 << 2))
1566			p->reset = 1;
1567		else if (p->reset) {
1568			p->reset = 0;
1569			ahci_port_reset(p);
1570		}
1571		p->ci &= ~(1 << slot);
1572	}
1573}
1574
1575static void
1576ahci_handle_port(struct ahci_port *p)
1577{
1578	int i;
1579
1580	if (!(p->cmd & AHCI_P_CMD_ST))
1581		return;
1582
1583	/*
1584	 * Search for any new commands to issue ignoring those that
1585	 * are already in-flight.
1586	 */
1587	for (i = 0; (i < 32) && p->ci; i++) {
1588		if ((p->ci & (1 << i)) && !(p->pending & (1 << i))) {
1589			p->cmd &= ~AHCI_P_CMD_CCS_MASK;
1590			p->cmd |= i << AHCI_P_CMD_CCS_SHIFT;
1591			ahci_handle_slot(p, i);
1592		}
1593	}
1594}
1595
1596/*
1597 * blockif callback routine - this runs in the context of the blockif
1598 * i/o thread, so the mutex needs to be acquired.
1599 */
1600static void
1601ata_ioreq_cb(struct blockif_req *br, int err)
1602{
1603	struct ahci_cmd_hdr *hdr;
1604	struct ahci_ioreq *aior;
1605	struct ahci_port *p;
1606	struct pci_ahci_softc *sc;
1607	uint32_t tfd;
1608	uint8_t *cfis;
1609	int pending, slot, ncq, dsm;
1610
1611	DPRINTF("%s %d\n", __func__, err);
1612
1613	ncq = dsm = 0;
1614	aior = br->br_param;
1615	p = aior->io_pr;
1616	cfis = aior->cfis;
1617	slot = aior->slot;
1618	pending = aior->prdtl;
1619	sc = p->pr_sc;
1620	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1621
1622	if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
1623			cfis[2] == ATA_READ_FPDMA_QUEUED)
1624		ncq = 1;
1625	if (cfis[2] == ATA_DATA_SET_MANAGEMENT)
1626		dsm = 1;
1627
1628	pthread_mutex_lock(&sc->mtx);
1629
1630	/*
1631	 * Delete the blockif request from the busy list
1632	 */
1633	TAILQ_REMOVE(&p->iobhd, aior, io_blist);
1634
1635	/*
1636	 * Move the blockif request back to the free list
1637	 */
1638	STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
1639
1640	if (dsm) {
1641		if (aior->done != aior->len && !err) {
1642			ahci_handle_dsm_trim(p, slot, cfis, aior->done);
1643			goto out;
1644		}
1645	} else {
1646		if (pending && !err) {
1647			ahci_handle_dma(p, slot, cfis, aior->done,
1648			    hdr->prdtl - pending);
1649			goto out;
1650		}
1651	}
1652
1653	if (!err && aior->done == aior->len) {
1654		tfd = ATA_S_READY | ATA_S_DSC;
1655		if (ncq)
1656			hdr->prdbc = 0;
1657		else
1658			hdr->prdbc = aior->len;
1659	} else {
1660		tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
1661		hdr->prdbc = 0;
1662		if (ncq)
1663			p->serr |= (1 << slot);
1664	}
1665
1666	if (ncq) {
1667		p->sact &= ~(1 << slot);
1668		ahci_write_fis_sdb(p, slot, tfd);
1669	} else
1670		ahci_write_fis_d2h(p, slot, cfis, tfd);
1671
1672	/*
1673	 * This command is now complete.
1674	 */
1675	p->pending &= ~(1 << slot);
1676
1677	ahci_check_stopped(p);
1678out:
1679	pthread_mutex_unlock(&sc->mtx);
1680	DPRINTF("%s exit\n", __func__);
1681}
1682
1683static void
1684atapi_ioreq_cb(struct blockif_req *br, int err)
1685{
1686	struct ahci_cmd_hdr *hdr;
1687	struct ahci_ioreq *aior;
1688	struct ahci_port *p;
1689	struct pci_ahci_softc *sc;
1690	uint8_t *cfis;
1691	uint32_t tfd;
1692	int pending, slot;
1693
1694	DPRINTF("%s %d\n", __func__, err);
1695
1696	aior = br->br_param;
1697	p = aior->io_pr;
1698	cfis = aior->cfis;
1699	slot = aior->slot;
1700	pending = aior->prdtl;
1701	sc = p->pr_sc;
1702	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + aior->slot * AHCI_CL_SIZE);
1703
1704	pthread_mutex_lock(&sc->mtx);
1705
1706	/*
1707	 * Delete the blockif request from the busy list
1708	 */
1709	TAILQ_REMOVE(&p->iobhd, aior, io_blist);
1710
1711	/*
1712	 * Move the blockif request back to the free list
1713	 */
1714	STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
1715
1716	if (pending && !err) {
1717		atapi_read(p, slot, cfis, aior->done, hdr->prdtl - pending);
1718		goto out;
1719	}
1720
1721	if (!err && aior->done == aior->len) {
1722		tfd = ATA_S_READY | ATA_S_DSC;
1723		hdr->prdbc = aior->len;
1724	} else {
1725		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1726		p->asc = 0x21;
1727		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1728		hdr->prdbc = 0;
1729	}
1730
1731	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1732	ahci_write_fis_d2h(p, slot, cfis, tfd);
1733
1734	/*
1735	 * This command is now complete.
1736	 */
1737	p->pending &= ~(1 << slot);
1738
1739	ahci_check_stopped(p);
1740out:
1741	pthread_mutex_unlock(&sc->mtx);
1742	DPRINTF("%s exit\n", __func__);
1743}
1744
1745static void
1746pci_ahci_ioreq_init(struct ahci_port *pr)
1747{
1748	struct ahci_ioreq *vr;
1749	int i;
1750
1751	pr->ioqsz = blockif_queuesz(pr->bctx);
1752	pr->ioreq = calloc(pr->ioqsz, sizeof(struct ahci_ioreq));
1753	STAILQ_INIT(&pr->iofhd);
1754
1755	/*
1756	 * Add all i/o request entries to the free queue
1757	 */
1758	for (i = 0; i < pr->ioqsz; i++) {
1759		vr = &pr->ioreq[i];
1760		vr->io_pr = pr;
1761		if (!pr->atapi)
1762			vr->io_req.br_callback = ata_ioreq_cb;
1763		else
1764			vr->io_req.br_callback = atapi_ioreq_cb;
1765		vr->io_req.br_param = vr;
1766		STAILQ_INSERT_TAIL(&pr->iofhd, vr, io_flist);
1767	}
1768
1769	TAILQ_INIT(&pr->iobhd);
1770}
1771
1772static void
1773pci_ahci_port_write(struct pci_ahci_softc *sc, uint64_t offset, uint64_t value)
1774{
1775	int port = (offset - AHCI_OFFSET) / AHCI_STEP;
1776	offset = (offset - AHCI_OFFSET) % AHCI_STEP;
1777	struct ahci_port *p = &sc->port[port];
1778
1779	DPRINTF("pci_ahci_port %d: write offset 0x%"PRIx64" value 0x%"PRIx64"\n",
1780		port, offset, value);
1781
1782	switch (offset) {
1783	case AHCI_P_CLB:
1784		p->clb = value;
1785		break;
1786	case AHCI_P_CLBU:
1787		p->clbu = value;
1788		break;
1789	case AHCI_P_FB:
1790		p->fb = value;
1791		break;
1792	case AHCI_P_FBU:
1793		p->fbu = value;
1794		break;
1795	case AHCI_P_IS:
1796		p->is &= ~value;
1797		break;
1798	case AHCI_P_IE:
1799		p->ie = value & 0xFDC000FF;
1800		ahci_generate_intr(sc);
1801		break;
1802	case AHCI_P_CMD:
1803	{
1804		p->cmd = value;
1805
1806		if (!(value & AHCI_P_CMD_ST)) {
1807			ahci_port_stop(p);
1808		} else {
1809			uint64_t clb;
1810
1811			p->cmd |= AHCI_P_CMD_CR;
1812			clb = (uint64_t)p->clbu << 32 | p->clb;
1813			p->cmd_lst = paddr_guest2host(ahci_ctx(sc), clb,
1814					AHCI_CL_SIZE * AHCI_MAX_SLOTS);
1815		}
1816
1817		if (value & AHCI_P_CMD_FRE) {
1818			uint64_t fb;
1819
1820			p->cmd |= AHCI_P_CMD_FR;
1821			fb = (uint64_t)p->fbu << 32 | p->fb;
1822			/* we don't support FBSCP, so rfis size is 256Bytes */
1823			p->rfis = paddr_guest2host(ahci_ctx(sc), fb, 256);
1824		} else {
1825			p->cmd &= ~AHCI_P_CMD_FR;
1826		}
1827
1828		if (value & AHCI_P_CMD_CLO) {
1829			p->tfd = 0;
1830			p->cmd &= ~AHCI_P_CMD_CLO;
1831		}
1832
1833		ahci_handle_port(p);
1834		break;
1835	}
1836	case AHCI_P_TFD:
1837	case AHCI_P_SIG:
1838	case AHCI_P_SSTS:
1839		WPRINTF("pci_ahci_port: read only registers 0x%"PRIx64"\n", offset);
1840		break;
1841	case AHCI_P_SCTL:
1842		if (!(p->cmd & AHCI_P_CMD_ST)) {
1843			if (value & ATA_SC_DET_RESET)
1844				ahci_port_reset(p);
1845			p->sctl = value;
1846		}
1847		break;
1848	case AHCI_P_SERR:
1849		p->serr &= ~value;
1850		break;
1851	case AHCI_P_SACT:
1852		p->sact |= value;
1853		break;
1854	case AHCI_P_CI:
1855		p->ci |= value;
1856		ahci_handle_port(p);
1857		break;
1858	case AHCI_P_SNTF:
1859	case AHCI_P_FBS:
1860	default:
1861		break;
1862	}
1863}
1864
1865static void
1866pci_ahci_host_write(struct pci_ahci_softc *sc, uint64_t offset, uint64_t value)
1867{
1868	DPRINTF("pci_ahci_host: write offset 0x%"PRIx64" value 0x%"PRIx64"\n",
1869		offset, value);
1870
1871	switch (offset) {
1872	case AHCI_CAP:
1873	case AHCI_PI:
1874	case AHCI_VS:
1875	case AHCI_CAP2:
1876		DPRINTF("pci_ahci_host: read only registers 0x%"PRIx64"\n", offset);
1877		break;
1878	case AHCI_GHC:
1879		if (value & AHCI_GHC_HR)
1880			ahci_reset(sc);
1881		else if (value & AHCI_GHC_IE) {
1882			sc->ghc |= AHCI_GHC_IE;
1883			ahci_generate_intr(sc);
1884		}
1885		break;
1886	case AHCI_IS:
1887		sc->is &= ~value;
1888		ahci_generate_intr(sc);
1889		break;
1890	default:
1891		break;
1892	}
1893}
1894
1895static void
1896pci_ahci_write(struct vmctx *ctx, int vcpu, struct pci_devinst *pi,
1897		int baridx, uint64_t offset, int size, uint64_t value)
1898{
1899	struct pci_ahci_softc *sc = pi->pi_arg;
1900
1901	assert(baridx == 5);
1902	assert(size == 4);
1903
1904	pthread_mutex_lock(&sc->mtx);
1905
1906	if (offset < AHCI_OFFSET)
1907		pci_ahci_host_write(sc, offset, value);
1908	else if (offset < AHCI_OFFSET + sc->ports * AHCI_STEP)
1909		pci_ahci_port_write(sc, offset, value);
1910	else
1911		WPRINTF("pci_ahci: unknown i/o write offset 0x%"PRIx64"\n", offset);
1912
1913	pthread_mutex_unlock(&sc->mtx);
1914}
1915
1916static uint64_t
1917pci_ahci_host_read(struct pci_ahci_softc *sc, uint64_t offset)
1918{
1919	uint32_t value;
1920
1921	switch (offset) {
1922	case AHCI_CAP:
1923	case AHCI_GHC:
1924	case AHCI_IS:
1925	case AHCI_PI:
1926	case AHCI_VS:
1927	case AHCI_CCCC:
1928	case AHCI_CCCP:
1929	case AHCI_EM_LOC:
1930	case AHCI_EM_CTL:
1931	case AHCI_CAP2:
1932	{
1933		uint32_t *p = &sc->cap;
1934		p += (offset - AHCI_CAP) / sizeof(uint32_t);
1935		value = *p;
1936		break;
1937	}
1938	default:
1939		value = 0;
1940		break;
1941	}
1942	DPRINTF("pci_ahci_host: read offset 0x%"PRIx64" value 0x%x\n",
1943		offset, value);
1944
1945	return (value);
1946}
1947
1948static uint64_t
1949pci_ahci_port_read(struct pci_ahci_softc *sc, uint64_t offset)
1950{
1951	uint32_t value;
1952	int port = (offset - AHCI_OFFSET) / AHCI_STEP;
1953	offset = (offset - AHCI_OFFSET) % AHCI_STEP;
1954
1955	switch (offset) {
1956	case AHCI_P_CLB:
1957	case AHCI_P_CLBU:
1958	case AHCI_P_FB:
1959	case AHCI_P_FBU:
1960	case AHCI_P_IS:
1961	case AHCI_P_IE:
1962	case AHCI_P_CMD:
1963	case AHCI_P_TFD:
1964	case AHCI_P_SIG:
1965	case AHCI_P_SSTS:
1966	case AHCI_P_SCTL:
1967	case AHCI_P_SERR:
1968	case AHCI_P_SACT:
1969	case AHCI_P_CI:
1970	case AHCI_P_SNTF:
1971	case AHCI_P_FBS:
1972	{
1973		uint32_t *p= &sc->port[port].clb;
1974		p += (offset - AHCI_P_CLB) / sizeof(uint32_t);
1975		value = *p;
1976		break;
1977	}
1978	default:
1979		value = 0;
1980		break;
1981	}
1982
1983	DPRINTF("pci_ahci_port %d: read offset 0x%"PRIx64" value 0x%x\n",
1984		port, offset, value);
1985
1986	return value;
1987}
1988
1989static uint64_t
1990pci_ahci_read(struct vmctx *ctx, int vcpu, struct pci_devinst *pi, int baridx,
1991    uint64_t offset, int size)
1992{
1993	struct pci_ahci_softc *sc = pi->pi_arg;
1994	uint32_t value;
1995
1996	assert(baridx == 5);
1997	assert(size == 4);
1998
1999	pthread_mutex_lock(&sc->mtx);
2000
2001	if (offset < AHCI_OFFSET)
2002		value = pci_ahci_host_read(sc, offset);
2003	else if (offset < AHCI_OFFSET + sc->ports * AHCI_STEP)
2004		value = pci_ahci_port_read(sc, offset);
2005	else {
2006		value = 0;
2007		WPRINTF("pci_ahci: unknown i/o read offset 0x%"PRIx64"\n", offset);
2008	}
2009
2010	pthread_mutex_unlock(&sc->mtx);
2011
2012	return (value);
2013}
2014
2015static int
2016pci_ahci_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts, int atapi)
2017{
2018	char bident[sizeof("XX:X:X")];
2019	struct blockif_ctxt *bctxt;
2020	struct pci_ahci_softc *sc;
2021	int ret, slots;
2022
2023	ret = 0;
2024
2025	if (opts == NULL) {
2026		fprintf(stderr, "pci_ahci: backing device required\n");
2027		return (1);
2028	}
2029
2030#ifdef AHCI_DEBUG
2031	dbg = fopen("/tmp/log", "w+");
2032#endif
2033
2034	sc = calloc(1, sizeof(struct pci_ahci_softc));
2035	pi->pi_arg = sc;
2036	sc->asc_pi = pi;
2037	sc->ports = MAX_PORTS;
2038
2039	/*
2040	 * Only use port 0 for a backing device. All other ports will be
2041	 * marked as unused
2042	 */
2043	sc->port[0].atapi = atapi;
2044
2045	/*
2046	 * Attempt to open the backing image. Use the PCI
2047	 * slot/func for the identifier string.
2048	 */
2049	snprintf(bident, sizeof(bident), "%d:%d", pi->pi_slot, pi->pi_func);
2050	bctxt = blockif_open(opts, bident);
2051	if (bctxt == NULL) {
2052		ret = 1;
2053		goto open_fail;
2054	}
2055	sc->port[0].bctx = bctxt;
2056	sc->port[0].pr_sc = sc;
2057
2058	/*
2059	 * Allocate blockif request structures and add them
2060	 * to the free list
2061	 */
2062	pci_ahci_ioreq_init(&sc->port[0]);
2063
2064	pthread_mutex_init(&sc->mtx, NULL);
2065
2066	/* Intel ICH8 AHCI */
2067	slots = sc->port[0].ioqsz;
2068	if (slots > 32)
2069		slots = 32;
2070	--slots;
2071	sc->cap = AHCI_CAP_64BIT | AHCI_CAP_SNCQ | AHCI_CAP_SSNTF |
2072	    AHCI_CAP_SMPS | AHCI_CAP_SSS | AHCI_CAP_SALP |
2073	    AHCI_CAP_SAL | AHCI_CAP_SCLO | (0x3 << AHCI_CAP_ISS_SHIFT)|
2074	    AHCI_CAP_PMD | AHCI_CAP_SSC | AHCI_CAP_PSC |
2075	    (slots << AHCI_CAP_NCS_SHIFT) | AHCI_CAP_SXS | (sc->ports - 1);
2076
2077	/* Only port 0 implemented */
2078	sc->pi = 1;
2079	sc->vs = 0x10300;
2080	sc->cap2 = AHCI_CAP2_APST;
2081	ahci_reset(sc);
2082
2083	pci_set_cfgdata16(pi, PCIR_DEVICE, 0x2821);
2084	pci_set_cfgdata16(pi, PCIR_VENDOR, 0x8086);
2085	pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE);
2086	pci_set_cfgdata8(pi, PCIR_SUBCLASS, PCIS_STORAGE_SATA);
2087	pci_set_cfgdata8(pi, PCIR_PROGIF, PCIP_STORAGE_SATA_AHCI_1_0);
2088	pci_emul_add_msicap(pi, 1);
2089	pci_emul_alloc_bar(pi, 5, PCIBAR_MEM32,
2090	    AHCI_OFFSET + sc->ports * AHCI_STEP);
2091
2092	pci_lintr_request(pi);
2093
2094open_fail:
2095	if (ret) {
2096		blockif_close(sc->port[0].bctx);
2097		free(sc);
2098	}
2099
2100	return (ret);
2101}
2102
2103static int
2104pci_ahci_hd_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts)
2105{
2106
2107	return (pci_ahci_init(ctx, pi, opts, 0));
2108}
2109
2110static int
2111pci_ahci_atapi_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts)
2112{
2113
2114	return (pci_ahci_init(ctx, pi, opts, 1));
2115}
2116
2117/*
2118 * Use separate emulation names to distinguish drive and atapi devices
2119 */
2120struct pci_devemu pci_de_ahci_hd = {
2121	.pe_emu =	"ahci-hd",
2122	.pe_init =	pci_ahci_hd_init,
2123	.pe_barwrite =	pci_ahci_write,
2124	.pe_barread =	pci_ahci_read
2125};
2126PCI_EMUL_SET(pci_de_ahci_hd);
2127
2128struct pci_devemu pci_de_ahci_cd = {
2129	.pe_emu =	"ahci-cd",
2130	.pe_init =	pci_ahci_atapi_init,
2131	.pe_barwrite =	pci_ahci_write,
2132	.pe_barread =	pci_ahci_read
2133};
2134PCI_EMUL_SET(pci_de_ahci_cd);
2135