xref: /freebsd-14.2/usr.sbin/bhyve/pci_ahci.c (revision 3981cf10)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause
3  *
4  * Copyright (c) 2013  Zhixiang Yu <[email protected]>
5  * Copyright (c) 2015-2016 Alexander Motin <[email protected]>
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 #include <sys/param.h>
32 #include <sys/linker_set.h>
33 #include <sys/stat.h>
34 #include <sys/uio.h>
35 #include <sys/ioctl.h>
36 #include <sys/disk.h>
37 #include <sys/ata.h>
38 #include <sys/endian.h>
39 
40 #include <errno.h>
41 #include <fcntl.h>
42 #include <stdio.h>
43 #include <stdlib.h>
44 #include <stdint.h>
45 #include <string.h>
46 #include <strings.h>
47 #include <unistd.h>
48 #include <assert.h>
49 #include <pthread.h>
50 #include <pthread_np.h>
51 #include <inttypes.h>
52 #include <md5.h>
53 
54 #include "bhyverun.h"
55 #include "config.h"
56 #include "debug.h"
57 #include "pci_emul.h"
58 #ifdef BHYVE_SNAPSHOT
59 #include "snapshot.h"
60 #endif
61 #include "ahci.h"
62 #include "block_if.h"
63 
64 #define	DEF_PORTS	6	/* Intel ICH8 AHCI supports 6 ports */
65 #define	MAX_PORTS	32	/* AHCI supports 32 ports */
66 
67 #define	PxSIG_ATA	0x00000101 /* ATA drive */
68 #define	PxSIG_ATAPI	0xeb140101 /* ATAPI drive */
69 
70 enum sata_fis_type {
71 	FIS_TYPE_REGH2D		= 0x27,	/* Register FIS - host to device */
72 	FIS_TYPE_REGD2H		= 0x34,	/* Register FIS - device to host */
73 	FIS_TYPE_DMAACT		= 0x39,	/* DMA activate FIS - device to host */
74 	FIS_TYPE_DMASETUP	= 0x41,	/* DMA setup FIS - bidirectional */
75 	FIS_TYPE_DATA		= 0x46,	/* Data FIS - bidirectional */
76 	FIS_TYPE_BIST		= 0x58,	/* BIST activate FIS - bidirectional */
77 	FIS_TYPE_PIOSETUP	= 0x5F,	/* PIO setup FIS - device to host */
78 	FIS_TYPE_SETDEVBITS	= 0xA1,	/* Set dev bits FIS - device to host */
79 };
80 
81 /*
82  * SCSI opcodes
83  */
84 #define	TEST_UNIT_READY		0x00
85 #define	REQUEST_SENSE		0x03
86 #define	INQUIRY			0x12
87 #define	START_STOP_UNIT		0x1B
88 #define	PREVENT_ALLOW		0x1E
89 #define	READ_CAPACITY		0x25
90 #define	READ_10			0x28
91 #define	POSITION_TO_ELEMENT	0x2B
92 #define	READ_TOC		0x43
93 #define	GET_EVENT_STATUS_NOTIFICATION 0x4A
94 #define	MODE_SENSE_10		0x5A
95 #define	REPORT_LUNS		0xA0
96 #define	READ_12			0xA8
97 #define	READ_CD			0xBE
98 
99 /*
100  * SCSI mode page codes
101  */
102 #define	MODEPAGE_RW_ERROR_RECOVERY	0x01
103 #define	MODEPAGE_CD_CAPABILITIES	0x2A
104 
105 /*
106  * ATA commands
107  */
108 #define	ATA_SF_ENAB_SATA_SF		0x10
109 #define	ATA_SATA_SF_AN			0x05
110 #define	ATA_SF_DIS_SATA_SF		0x90
111 
112 /*
113  * Debug printf
114  */
115 #ifdef AHCI_DEBUG
116 static FILE *dbg;
117 #define DPRINTF(format, arg...)	do{fprintf(dbg, format, ##arg);fflush(dbg);}while(0)
118 #else
119 #define DPRINTF(format, arg...)
120 #endif
121 
122 #define AHCI_PORT_IDENT 20 + 1
123 
124 struct ahci_ioreq {
125 	struct blockif_req io_req;
126 	struct ahci_port *io_pr;
127 	STAILQ_ENTRY(ahci_ioreq) io_flist;
128 	TAILQ_ENTRY(ahci_ioreq) io_blist;
129 	uint8_t *cfis;
130 	uint8_t *dsm;
131 	uint32_t len;
132 	uint32_t done;
133 	int slot;
134 	int more;
135 	int readop;
136 };
137 
138 struct ahci_port {
139 	struct blockif_ctxt *bctx;
140 	struct pci_ahci_softc *pr_sc;
141 	struct ata_params ata_ident;
142 	uint8_t *cmd_lst;
143 	uint8_t *rfis;
144 	int port;
145 	int atapi;
146 	int reset;
147 	int waitforclear;
148 	int mult_sectors;
149 	uint8_t xfermode;
150 	uint8_t err_cfis[20];
151 	uint8_t sense_key;
152 	uint8_t asc;
153 	u_int ccs;
154 	uint32_t pending;
155 
156 	uint32_t clb;
157 	uint32_t clbu;
158 	uint32_t fb;
159 	uint32_t fbu;
160 	uint32_t is;
161 	uint32_t ie;
162 	uint32_t cmd;
163 	uint32_t unused0;
164 	uint32_t tfd;
165 	uint32_t sig;
166 	uint32_t ssts;
167 	uint32_t sctl;
168 	uint32_t serr;
169 	uint32_t sact;
170 	uint32_t ci;
171 	uint32_t sntf;
172 	uint32_t fbs;
173 
174 	/*
175 	 * i/o request info
176 	 */
177 	struct ahci_ioreq *ioreq;
178 	int ioqsz;
179 	STAILQ_HEAD(ahci_fhead, ahci_ioreq) iofhd;
180 	TAILQ_HEAD(ahci_bhead, ahci_ioreq) iobhd;
181 };
182 
183 struct ahci_cmd_hdr {
184 	uint16_t flags;
185 	uint16_t prdtl;
186 	uint32_t prdbc;
187 	uint64_t ctba;
188 	uint32_t reserved[4];
189 };
190 
191 struct ahci_prdt_entry {
192 	uint64_t dba;
193 	uint32_t reserved;
194 #define	DBCMASK		0x3fffff
195 	uint32_t dbc;
196 };
197 
198 struct pci_ahci_softc {
199 	struct pci_devinst *asc_pi;
200 	pthread_mutex_t	mtx;
201 	int ports;
202 	uint32_t cap;
203 	uint32_t ghc;
204 	uint32_t is;
205 	uint32_t pi;
206 	uint32_t vs;
207 	uint32_t ccc_ctl;
208 	uint32_t ccc_pts;
209 	uint32_t em_loc;
210 	uint32_t em_ctl;
211 	uint32_t cap2;
212 	uint32_t bohc;
213 	uint32_t lintr;
214 	struct ahci_port port[MAX_PORTS];
215 };
216 #define	ahci_ctx(sc)	((sc)->asc_pi->pi_vmctx)
217 
218 static void ahci_handle_next_trim(struct ahci_port *p, int slot, uint8_t *cfis,
219     uint8_t *buf, uint32_t len, uint32_t done);
220 static void ahci_handle_port(struct ahci_port *p);
221 
lba_to_msf(uint8_t * buf,int lba)222 static inline void lba_to_msf(uint8_t *buf, int lba)
223 {
224 	lba += 150;
225 	buf[0] = (lba / 75) / 60;
226 	buf[1] = (lba / 75) % 60;
227 	buf[2] = lba % 75;
228 }
229 
230 /*
231  * Generate HBA interrupts on global IS register write.
232  */
233 static void
ahci_generate_intr(struct pci_ahci_softc * sc,uint32_t mask)234 ahci_generate_intr(struct pci_ahci_softc *sc, uint32_t mask)
235 {
236 	struct pci_devinst *pi = sc->asc_pi;
237 	struct ahci_port *p;
238 	int i, nmsg;
239 	uint32_t mmask;
240 
241 	/* Update global IS from PxIS/PxIE. */
242 	for (i = 0; i < sc->ports; i++) {
243 		p = &sc->port[i];
244 		if (p->is & p->ie)
245 			sc->is |= (1 << i);
246 	}
247 	DPRINTF("%s(%08x) %08x", __func__, mask, sc->is);
248 
249 	/* If there is nothing enabled -- clear legacy interrupt and exit. */
250 	if (sc->is == 0 || (sc->ghc & AHCI_GHC_IE) == 0) {
251 		if (sc->lintr) {
252 			pci_lintr_deassert(pi);
253 			sc->lintr = 0;
254 		}
255 		return;
256 	}
257 
258 	/* If there is anything and no MSI -- assert legacy interrupt. */
259 	nmsg = pci_msi_maxmsgnum(pi);
260 	if (nmsg == 0) {
261 		if (!sc->lintr) {
262 			sc->lintr = 1;
263 			pci_lintr_assert(pi);
264 		}
265 		return;
266 	}
267 
268 	/* Assert respective MSIs for ports that were touched. */
269 	for (i = 0; i < nmsg; i++) {
270 		if (sc->ports <= nmsg || i < nmsg - 1)
271 			mmask = 1 << i;
272 		else
273 			mmask = 0xffffffff << i;
274 		if (sc->is & mask && mmask & mask)
275 			pci_generate_msi(pi, i);
276 	}
277 }
278 
279 /*
280  * Generate HBA interrupt on specific port event.
281  */
282 static void
ahci_port_intr(struct ahci_port * p)283 ahci_port_intr(struct ahci_port *p)
284 {
285 	struct pci_ahci_softc *sc = p->pr_sc;
286 	struct pci_devinst *pi = sc->asc_pi;
287 	int nmsg;
288 
289 	DPRINTF("%s(%d) %08x/%08x %08x", __func__,
290 	    p->port, p->is, p->ie, sc->is);
291 
292 	/* If there is nothing enabled -- we are done. */
293 	if ((p->is & p->ie) == 0)
294 		return;
295 
296 	/* In case of non-shared MSI always generate interrupt. */
297 	nmsg = pci_msi_maxmsgnum(pi);
298 	if (sc->ports <= nmsg || p->port < nmsg - 1) {
299 		sc->is |= (1 << p->port);
300 		if ((sc->ghc & AHCI_GHC_IE) == 0)
301 			return;
302 		pci_generate_msi(pi, p->port);
303 		return;
304 	}
305 
306 	/* If IS for this port is already set -- do nothing. */
307 	if (sc->is & (1 << p->port))
308 		return;
309 
310 	sc->is |= (1 << p->port);
311 
312 	/* If interrupts are enabled -- generate one. */
313 	if ((sc->ghc & AHCI_GHC_IE) == 0)
314 		return;
315 	if (nmsg > 0) {
316 		pci_generate_msi(pi, nmsg - 1);
317 	} else if (!sc->lintr) {
318 		sc->lintr = 1;
319 		pci_lintr_assert(pi);
320 	}
321 }
322 
323 static void
ahci_write_fis(struct ahci_port * p,enum sata_fis_type ft,uint8_t * fis)324 ahci_write_fis(struct ahci_port *p, enum sata_fis_type ft, uint8_t *fis)
325 {
326 	int offset, len, irq;
327 
328 	if (p->rfis == NULL || !(p->cmd & AHCI_P_CMD_FRE))
329 		return;
330 
331 	switch (ft) {
332 	case FIS_TYPE_REGD2H:
333 		offset = 0x40;
334 		len = 20;
335 		irq = (fis[1] & (1 << 6)) ? AHCI_P_IX_DHR : 0;
336 		break;
337 	case FIS_TYPE_SETDEVBITS:
338 		offset = 0x58;
339 		len = 8;
340 		irq = (fis[1] & (1 << 6)) ? AHCI_P_IX_SDB : 0;
341 		break;
342 	case FIS_TYPE_PIOSETUP:
343 		offset = 0x20;
344 		len = 20;
345 		irq = (fis[1] & (1 << 6)) ? AHCI_P_IX_PS : 0;
346 		break;
347 	default:
348 		EPRINTLN("unsupported fis type %d", ft);
349 		return;
350 	}
351 	if (fis[2] & ATA_S_ERROR) {
352 		p->waitforclear = 1;
353 		irq |= AHCI_P_IX_TFE;
354 	}
355 	memcpy(p->rfis + offset, fis, len);
356 	if (irq) {
357 		if (~p->is & irq) {
358 			p->is |= irq;
359 			ahci_port_intr(p);
360 		}
361 	}
362 }
363 
364 static void
ahci_write_fis_piosetup(struct ahci_port * p)365 ahci_write_fis_piosetup(struct ahci_port *p)
366 {
367 	uint8_t fis[20];
368 
369 	memset(fis, 0, sizeof(fis));
370 	fis[0] = FIS_TYPE_PIOSETUP;
371 	ahci_write_fis(p, FIS_TYPE_PIOSETUP, fis);
372 }
373 
374 static void
ahci_write_fis_sdb(struct ahci_port * p,int slot,uint8_t * cfis,uint32_t tfd)375 ahci_write_fis_sdb(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t tfd)
376 {
377 	uint8_t fis[8];
378 	uint8_t error;
379 
380 	error = (tfd >> 8) & 0xff;
381 	tfd &= 0x77;
382 	memset(fis, 0, sizeof(fis));
383 	fis[0] = FIS_TYPE_SETDEVBITS;
384 	fis[1] = (1 << 6);
385 	fis[2] = tfd;
386 	fis[3] = error;
387 	if (fis[2] & ATA_S_ERROR) {
388 		p->err_cfis[0] = slot;
389 		p->err_cfis[2] = tfd;
390 		p->err_cfis[3] = error;
391 		memcpy(&p->err_cfis[4], cfis + 4, 16);
392 	} else {
393 		*(uint32_t *)(fis + 4) = (1 << slot);
394 		p->sact &= ~(1 << slot);
395 	}
396 	p->tfd &= ~0x77;
397 	p->tfd |= tfd;
398 	ahci_write_fis(p, FIS_TYPE_SETDEVBITS, fis);
399 }
400 
401 static void
ahci_write_fis_d2h(struct ahci_port * p,int slot,uint8_t * cfis,uint32_t tfd)402 ahci_write_fis_d2h(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t tfd)
403 {
404 	uint8_t fis[20];
405 	uint8_t error;
406 
407 	error = (tfd >> 8) & 0xff;
408 	memset(fis, 0, sizeof(fis));
409 	fis[0] = FIS_TYPE_REGD2H;
410 	fis[1] = (1 << 6);
411 	fis[2] = tfd & 0xff;
412 	fis[3] = error;
413 	fis[4] = cfis[4];
414 	fis[5] = cfis[5];
415 	fis[6] = cfis[6];
416 	fis[7] = cfis[7];
417 	fis[8] = cfis[8];
418 	fis[9] = cfis[9];
419 	fis[10] = cfis[10];
420 	fis[11] = cfis[11];
421 	fis[12] = cfis[12];
422 	fis[13] = cfis[13];
423 	if (fis[2] & ATA_S_ERROR) {
424 		p->err_cfis[0] = 0x80;
425 		p->err_cfis[2] = tfd & 0xff;
426 		p->err_cfis[3] = error;
427 		memcpy(&p->err_cfis[4], cfis + 4, 16);
428 	} else
429 		p->ci &= ~(1 << slot);
430 	p->tfd = tfd;
431 	ahci_write_fis(p, FIS_TYPE_REGD2H, fis);
432 }
433 
434 static void
ahci_write_fis_d2h_ncq(struct ahci_port * p,int slot)435 ahci_write_fis_d2h_ncq(struct ahci_port *p, int slot)
436 {
437 	uint8_t fis[20];
438 
439 	p->tfd = ATA_S_READY | ATA_S_DSC;
440 	memset(fis, 0, sizeof(fis));
441 	fis[0] = FIS_TYPE_REGD2H;
442 	fis[1] = 0;			/* No interrupt */
443 	fis[2] = p->tfd;		/* Status */
444 	fis[3] = 0;			/* No error */
445 	p->ci &= ~(1 << slot);
446 	ahci_write_fis(p, FIS_TYPE_REGD2H, fis);
447 }
448 
449 static void
ahci_write_reset_fis_d2h(struct ahci_port * p)450 ahci_write_reset_fis_d2h(struct ahci_port *p)
451 {
452 	uint8_t fis[20];
453 
454 	memset(fis, 0, sizeof(fis));
455 	fis[0] = FIS_TYPE_REGD2H;
456 	fis[3] = 1;
457 	fis[4] = 1;
458 	if (p->atapi) {
459 		fis[5] = 0x14;
460 		fis[6] = 0xeb;
461 	}
462 	fis[12] = 1;
463 	ahci_write_fis(p, FIS_TYPE_REGD2H, fis);
464 }
465 
466 static void
ahci_check_stopped(struct ahci_port * p)467 ahci_check_stopped(struct ahci_port *p)
468 {
469 	/*
470 	 * If we are no longer processing the command list and nothing
471 	 * is in-flight, clear the running bit, the current command
472 	 * slot, the command issue and active bits.
473 	 */
474 	if (!(p->cmd & AHCI_P_CMD_ST)) {
475 		if (p->pending == 0) {
476 			p->ccs = 0;
477 			p->cmd &= ~(AHCI_P_CMD_CR | AHCI_P_CMD_CCS_MASK);
478 			p->ci = 0;
479 			p->sact = 0;
480 			p->waitforclear = 0;
481 		}
482 	}
483 }
484 
485 static void
ahci_port_stop(struct ahci_port * p)486 ahci_port_stop(struct ahci_port *p)
487 {
488 	struct ahci_ioreq *aior;
489 	uint8_t *cfis;
490 	int slot;
491 	int error;
492 
493 	assert(pthread_mutex_isowned_np(&p->pr_sc->mtx));
494 
495 	TAILQ_FOREACH(aior, &p->iobhd, io_blist) {
496 		/*
497 		 * Try to cancel the outstanding blockif request.
498 		 */
499 		error = blockif_cancel(p->bctx, &aior->io_req);
500 		if (error != 0)
501 			continue;
502 
503 		slot = aior->slot;
504 		cfis = aior->cfis;
505 		if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
506 		    cfis[2] == ATA_READ_FPDMA_QUEUED ||
507 		    cfis[2] == ATA_SEND_FPDMA_QUEUED)
508 			p->sact &= ~(1 << slot);	/* NCQ */
509 		else
510 			p->ci &= ~(1 << slot);
511 
512 		/*
513 		 * This command is now done.
514 		 */
515 		p->pending &= ~(1 << slot);
516 
517 		/*
518 		 * Delete the blockif request from the busy list
519 		 */
520 		TAILQ_REMOVE(&p->iobhd, aior, io_blist);
521 
522 		/*
523 		 * Move the blockif request back to the free list
524 		 */
525 		STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
526 	}
527 
528 	ahci_check_stopped(p);
529 }
530 
531 static void
ahci_port_reset(struct ahci_port * pr)532 ahci_port_reset(struct ahci_port *pr)
533 {
534 	pr->serr = 0;
535 	pr->sact = 0;
536 	pr->xfermode = ATA_UDMA6;
537 	pr->mult_sectors = 128;
538 
539 	if (!pr->bctx) {
540 		pr->ssts = ATA_SS_DET_NO_DEVICE;
541 		pr->sig = 0xFFFFFFFF;
542 		pr->tfd = 0x7F;
543 		return;
544 	}
545 	pr->ssts = ATA_SS_DET_PHY_ONLINE | ATA_SS_IPM_ACTIVE;
546 	if (pr->sctl & ATA_SC_SPD_MASK)
547 		pr->ssts |= (pr->sctl & ATA_SC_SPD_MASK);
548 	else
549 		pr->ssts |= ATA_SS_SPD_GEN3;
550 	pr->tfd = (1 << 8) | ATA_S_DSC | ATA_S_DMA;
551 	if (!pr->atapi) {
552 		pr->sig = PxSIG_ATA;
553 		pr->tfd |= ATA_S_READY;
554 	} else
555 		pr->sig = PxSIG_ATAPI;
556 	ahci_write_reset_fis_d2h(pr);
557 }
558 
559 static void
ahci_reset(struct pci_ahci_softc * sc)560 ahci_reset(struct pci_ahci_softc *sc)
561 {
562 	int i;
563 
564 	sc->ghc = AHCI_GHC_AE;
565 	sc->is = 0;
566 
567 	if (sc->lintr) {
568 		pci_lintr_deassert(sc->asc_pi);
569 		sc->lintr = 0;
570 	}
571 
572 	for (i = 0; i < sc->ports; i++) {
573 		sc->port[i].ie = 0;
574 		sc->port[i].is = 0;
575 		sc->port[i].cmd = (AHCI_P_CMD_SUD | AHCI_P_CMD_POD);
576 		if (sc->port[i].bctx)
577 			sc->port[i].cmd |= AHCI_P_CMD_CPS;
578 		sc->port[i].sctl = 0;
579 		ahci_port_reset(&sc->port[i]);
580 	}
581 }
582 
583 static void
ata_string(uint8_t * dest,const char * src,int len)584 ata_string(uint8_t *dest, const char *src, int len)
585 {
586 	int i;
587 
588 	for (i = 0; i < len; i++) {
589 		if (*src)
590 			dest[i ^ 1] = *src++;
591 		else
592 			dest[i ^ 1] = ' ';
593 	}
594 }
595 
596 static void
atapi_string(uint8_t * dest,const char * src,int len)597 atapi_string(uint8_t *dest, const char *src, int len)
598 {
599 	int i;
600 
601 	for (i = 0; i < len; i++) {
602 		if (*src)
603 			dest[i] = *src++;
604 		else
605 			dest[i] = ' ';
606 	}
607 }
608 
609 /*
610  * Build up the iovec based on the PRDT, 'done' and 'len'.
611  */
612 static void
ahci_build_iov(struct ahci_port * p,struct ahci_ioreq * aior,struct ahci_prdt_entry * prdt,uint16_t prdtl)613 ahci_build_iov(struct ahci_port *p, struct ahci_ioreq *aior,
614     struct ahci_prdt_entry *prdt, uint16_t prdtl)
615 {
616 	struct blockif_req *breq = &aior->io_req;
617 	uint32_t dbcsz, extra, left, skip, todo;
618 	int i, j;
619 
620 	assert(aior->len >= aior->done);
621 
622 	/* Copy part of PRDT between 'done' and 'len' bytes into the iov. */
623 	skip = aior->done;
624 	left = aior->len - aior->done;
625 	todo = 0;
626 	for (i = 0, j = 0; i < prdtl && j < BLOCKIF_IOV_MAX && left > 0;
627 	    i++, prdt++) {
628 		dbcsz = (prdt->dbc & DBCMASK) + 1;
629 		/* Skip already done part of the PRDT */
630 		if (dbcsz <= skip) {
631 			skip -= dbcsz;
632 			continue;
633 		}
634 		dbcsz -= skip;
635 		if (dbcsz > left)
636 			dbcsz = left;
637 		breq->br_iov[j].iov_base = paddr_guest2host(ahci_ctx(p->pr_sc),
638 		    prdt->dba + skip, dbcsz);
639 		breq->br_iov[j].iov_len = dbcsz;
640 		todo += dbcsz;
641 		left -= dbcsz;
642 		skip = 0;
643 		j++;
644 	}
645 
646 	/* If we got limited by IOV length, round I/O down to sector size. */
647 	if (j == BLOCKIF_IOV_MAX) {
648 		extra = todo % blockif_sectsz(p->bctx);
649 		todo -= extra;
650 		assert(todo > 0);
651 		while (extra > 0) {
652 			if (breq->br_iov[j - 1].iov_len > extra) {
653 				breq->br_iov[j - 1].iov_len -= extra;
654 				break;
655 			}
656 			extra -= breq->br_iov[j - 1].iov_len;
657 			j--;
658 		}
659 	}
660 
661 	breq->br_iovcnt = j;
662 	breq->br_resid = todo;
663 	aior->done += todo;
664 	aior->more = (aior->done < aior->len && i < prdtl);
665 }
666 
667 static void
ahci_handle_rw(struct ahci_port * p,int slot,uint8_t * cfis,uint32_t done)668 ahci_handle_rw(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t done)
669 {
670 	struct ahci_ioreq *aior;
671 	struct blockif_req *breq;
672 	struct ahci_prdt_entry *prdt;
673 	struct ahci_cmd_hdr *hdr;
674 	uint64_t lba;
675 	uint32_t len;
676 	int err, first, ncq, readop;
677 
678 	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
679 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
680 	ncq = 0;
681 	readop = 1;
682 	first = (done == 0);
683 
684 	if (cfis[2] == ATA_WRITE || cfis[2] == ATA_WRITE48 ||
685 	    cfis[2] == ATA_WRITE_MUL || cfis[2] == ATA_WRITE_MUL48 ||
686 	    cfis[2] == ATA_WRITE_DMA || cfis[2] == ATA_WRITE_DMA48 ||
687 	    cfis[2] == ATA_WRITE_FPDMA_QUEUED)
688 		readop = 0;
689 
690 	if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
691 	    cfis[2] == ATA_READ_FPDMA_QUEUED) {
692 		lba = ((uint64_t)cfis[10] << 40) |
693 			((uint64_t)cfis[9] << 32) |
694 			((uint64_t)cfis[8] << 24) |
695 			((uint64_t)cfis[6] << 16) |
696 			((uint64_t)cfis[5] << 8) |
697 			cfis[4];
698 		len = cfis[11] << 8 | cfis[3];
699 		if (!len)
700 			len = 65536;
701 		ncq = 1;
702 	} else if (cfis[2] == ATA_READ48 || cfis[2] == ATA_WRITE48 ||
703 	    cfis[2] == ATA_READ_MUL48 || cfis[2] == ATA_WRITE_MUL48 ||
704 	    cfis[2] == ATA_READ_DMA48 || cfis[2] == ATA_WRITE_DMA48) {
705 		lba = ((uint64_t)cfis[10] << 40) |
706 			((uint64_t)cfis[9] << 32) |
707 			((uint64_t)cfis[8] << 24) |
708 			((uint64_t)cfis[6] << 16) |
709 			((uint64_t)cfis[5] << 8) |
710 			cfis[4];
711 		len = cfis[13] << 8 | cfis[12];
712 		if (!len)
713 			len = 65536;
714 	} else {
715 		lba = ((cfis[7] & 0xf) << 24) | (cfis[6] << 16) |
716 			(cfis[5] << 8) | cfis[4];
717 		len = cfis[12];
718 		if (!len)
719 			len = 256;
720 	}
721 	lba *= blockif_sectsz(p->bctx);
722 	len *= blockif_sectsz(p->bctx);
723 
724 	/* Pull request off free list */
725 	aior = STAILQ_FIRST(&p->iofhd);
726 	assert(aior != NULL);
727 	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
728 
729 	aior->cfis = cfis;
730 	aior->slot = slot;
731 	aior->len = len;
732 	aior->done = done;
733 	aior->readop = readop;
734 	breq = &aior->io_req;
735 	breq->br_offset = lba + done;
736 	ahci_build_iov(p, aior, prdt, hdr->prdtl);
737 
738 	/* Mark this command in-flight. */
739 	p->pending |= 1 << slot;
740 
741 	/* Stuff request onto busy list. */
742 	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
743 
744 	if (ncq && first)
745 		ahci_write_fis_d2h_ncq(p, slot);
746 
747 	if (readop)
748 		err = blockif_read(p->bctx, breq);
749 	else
750 		err = blockif_write(p->bctx, breq);
751 	assert(err == 0);
752 }
753 
754 static void
ahci_handle_flush(struct ahci_port * p,int slot,uint8_t * cfis)755 ahci_handle_flush(struct ahci_port *p, int slot, uint8_t *cfis)
756 {
757 	struct ahci_ioreq *aior;
758 	struct blockif_req *breq;
759 	int err;
760 
761 	/*
762 	 * Pull request off free list
763 	 */
764 	aior = STAILQ_FIRST(&p->iofhd);
765 	assert(aior != NULL);
766 	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
767 	aior->cfis = cfis;
768 	aior->slot = slot;
769 	aior->len = 0;
770 	aior->done = 0;
771 	aior->more = 0;
772 	breq = &aior->io_req;
773 
774 	/*
775 	 * Mark this command in-flight.
776 	 */
777 	p->pending |= 1 << slot;
778 
779 	/*
780 	 * Stuff request onto busy list
781 	 */
782 	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
783 
784 	err = blockif_flush(p->bctx, breq);
785 	assert(err == 0);
786 }
787 
788 static inline unsigned int
read_prdt(struct ahci_port * p,int slot,uint8_t * cfis,void * buf,unsigned int size)789 read_prdt(struct ahci_port *p, int slot, uint8_t *cfis, void *buf,
790     unsigned int size)
791 {
792 	struct ahci_cmd_hdr *hdr;
793 	struct ahci_prdt_entry *prdt;
794 	uint8_t *to;
795 	unsigned int len;
796 	int i;
797 
798 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
799 	len = size;
800 	to = buf;
801 	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
802 	for (i = 0; i < hdr->prdtl && len; i++) {
803 		uint8_t *ptr;
804 		uint32_t dbcsz;
805 		unsigned int sublen;
806 
807 		dbcsz = (prdt->dbc & DBCMASK) + 1;
808 		ptr = paddr_guest2host(ahci_ctx(p->pr_sc), prdt->dba, dbcsz);
809 		sublen = MIN(len, dbcsz);
810 		memcpy(to, ptr, sublen);
811 		len -= sublen;
812 		to += sublen;
813 		prdt++;
814 	}
815 	return (size - len);
816 }
817 
818 static void
ahci_handle_dsm_trim(struct ahci_port * p,int slot,uint8_t * cfis)819 ahci_handle_dsm_trim(struct ahci_port *p, int slot, uint8_t *cfis)
820 {
821 	uint32_t len;
822 	int ncq;
823 	uint8_t *buf;
824 	unsigned int nread;
825 
826 	buf = NULL;
827 	if (cfis[2] == ATA_DATA_SET_MANAGEMENT) {
828 		len = (uint16_t)cfis[13] << 8 | cfis[12];
829 		len *= 512;
830 		ncq = 0;
831 	} else { /* ATA_SEND_FPDMA_QUEUED */
832 		len = (uint16_t)cfis[11] << 8 | cfis[3];
833 		len *= 512;
834 		ncq = 1;
835 	}
836 
837 	/* Support for only a single block is advertised via IDENTIFY. */
838 	if (len > 512) {
839 		goto invalid_command;
840 	}
841 
842 	buf = malloc(len);
843 	nread = read_prdt(p, slot, cfis, buf, len);
844 	if (nread != len) {
845 		goto invalid_command;
846 	}
847 	ahci_handle_next_trim(p, slot, cfis, buf, len, 0);
848 	return;
849 
850 invalid_command:
851 	free(buf);
852 	if (ncq) {
853 		ahci_write_fis_d2h_ncq(p, slot);
854 		ahci_write_fis_sdb(p, slot, cfis,
855 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
856 	} else {
857 		ahci_write_fis_d2h(p, slot, cfis,
858 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
859 	}
860 }
861 
862 static void
ahci_handle_next_trim(struct ahci_port * p,int slot,uint8_t * cfis,uint8_t * buf,uint32_t len,uint32_t done)863 ahci_handle_next_trim(struct ahci_port *p, int slot, uint8_t *cfis,
864     uint8_t *buf, uint32_t len, uint32_t done)
865 {
866 	struct ahci_ioreq *aior;
867 	struct blockif_req *breq;
868 	uint8_t *entry;
869 	uint64_t elba;
870 	uint32_t elen;
871 	int err;
872 	bool first, ncq;
873 
874 	first = (done == 0);
875 	if (cfis[2] == ATA_DATA_SET_MANAGEMENT) {
876 		ncq = false;
877 	} else { /* ATA_SEND_FPDMA_QUEUED */
878 		ncq = true;
879 	}
880 
881 	/* Find the next range to TRIM. */
882 	while (done < len) {
883 		entry = &buf[done];
884 		elba = ((uint64_t)entry[5] << 40) |
885 		    ((uint64_t)entry[4] << 32) |
886 		    ((uint64_t)entry[3] << 24) |
887 		    ((uint64_t)entry[2] << 16) |
888 		    ((uint64_t)entry[1] << 8) |
889 		    entry[0];
890 		elen = (uint16_t)entry[7] << 8 | entry[6];
891 		done += 8;
892 		if (elen != 0)
893 			break;
894 	}
895 
896 	/* All remaining ranges were empty. */
897 	if (done == len) {
898 		free(buf);
899 		if (ncq) {
900 			if (first)
901 				ahci_write_fis_d2h_ncq(p, slot);
902 			ahci_write_fis_sdb(p, slot, cfis,
903 			    ATA_S_READY | ATA_S_DSC);
904 		} else {
905 			ahci_write_fis_d2h(p, slot, cfis,
906 			    ATA_S_READY | ATA_S_DSC);
907 		}
908 		if (!first) {
909 			p->pending &= ~(1 << slot);
910 			ahci_check_stopped(p);
911 			ahci_handle_port(p);
912 		}
913 		return;
914 	}
915 
916 	/*
917 	 * Pull request off free list
918 	 */
919 	aior = STAILQ_FIRST(&p->iofhd);
920 	assert(aior != NULL);
921 	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
922 	aior->cfis = cfis;
923 	aior->slot = slot;
924 	aior->len = len;
925 	aior->done = done;
926 	aior->dsm = buf;
927 	aior->more = (len != done);
928 
929 	breq = &aior->io_req;
930 	breq->br_offset = elba * blockif_sectsz(p->bctx);
931 	breq->br_resid = elen * blockif_sectsz(p->bctx);
932 
933 	/*
934 	 * Mark this command in-flight.
935 	 */
936 	p->pending |= 1 << slot;
937 
938 	/*
939 	 * Stuff request onto busy list
940 	 */
941 	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
942 
943 	if (ncq && first)
944 		ahci_write_fis_d2h_ncq(p, slot);
945 
946 	err = blockif_delete(p->bctx, breq);
947 	assert(err == 0);
948 }
949 
950 static inline void
write_prdt(struct ahci_port * p,int slot,uint8_t * cfis,void * buf,unsigned int size)951 write_prdt(struct ahci_port *p, int slot, uint8_t *cfis, void *buf,
952     unsigned int size)
953 {
954 	struct ahci_cmd_hdr *hdr;
955 	struct ahci_prdt_entry *prdt;
956 	uint8_t *from;
957 	unsigned int len;
958 	int i;
959 
960 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
961 	len = size;
962 	from = buf;
963 	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
964 	for (i = 0; i < hdr->prdtl && len; i++) {
965 		uint8_t *ptr;
966 		uint32_t dbcsz;
967 		int sublen;
968 
969 		dbcsz = (prdt->dbc & DBCMASK) + 1;
970 		ptr = paddr_guest2host(ahci_ctx(p->pr_sc), prdt->dba, dbcsz);
971 		sublen = MIN(len, dbcsz);
972 		memcpy(ptr, from, sublen);
973 		len -= sublen;
974 		from += sublen;
975 		prdt++;
976 	}
977 	hdr->prdbc = size - len;
978 }
979 
980 static void
ahci_checksum(uint8_t * buf,int size)981 ahci_checksum(uint8_t *buf, int size)
982 {
983 	int i;
984 	uint8_t sum = 0;
985 
986 	for (i = 0; i < size - 1; i++)
987 		sum += buf[i];
988 	buf[size - 1] = 0x100 - sum;
989 }
990 
991 static void
ahci_handle_read_log(struct ahci_port * p,int slot,uint8_t * cfis)992 ahci_handle_read_log(struct ahci_port *p, int slot, uint8_t *cfis)
993 {
994 	struct ahci_cmd_hdr *hdr;
995 	uint32_t buf[128];
996 	uint8_t *buf8 = (uint8_t *)buf;
997 	uint16_t *buf16 = (uint16_t *)buf;
998 
999 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1000 	if (p->atapi || hdr->prdtl == 0 || cfis[5] != 0 ||
1001 	    cfis[9] != 0 || cfis[12] != 1 || cfis[13] != 0) {
1002 		ahci_write_fis_d2h(p, slot, cfis,
1003 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1004 		return;
1005 	}
1006 
1007 	memset(buf, 0, sizeof(buf));
1008 	if (cfis[4] == 0x00) {	/* Log directory */
1009 		buf16[0x00] = 1; /* Version -- 1 */
1010 		buf16[0x10] = 1; /* NCQ Command Error Log -- 1 page */
1011 		buf16[0x13] = 1; /* SATA NCQ Send and Receive Log -- 1 page */
1012 	} else if (cfis[4] == 0x10) {	/* NCQ Command Error Log */
1013 		memcpy(buf8, p->err_cfis, sizeof(p->err_cfis));
1014 		ahci_checksum(buf8, sizeof(buf));
1015 	} else if (cfis[4] == 0x13) {	/* SATA NCQ Send and Receive Log */
1016 		if (blockif_candelete(p->bctx) && !blockif_is_ro(p->bctx)) {
1017 			buf[0x00] = 1;	/* SFQ DSM supported */
1018 			buf[0x01] = 1;	/* SFQ DSM TRIM supported */
1019 		}
1020 	} else {
1021 		ahci_write_fis_d2h(p, slot, cfis,
1022 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1023 		return;
1024 	}
1025 
1026 	if (cfis[2] == ATA_READ_LOG_EXT)
1027 		ahci_write_fis_piosetup(p);
1028 	write_prdt(p, slot, cfis, (void *)buf, sizeof(buf));
1029 	ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
1030 }
1031 
1032 static void
handle_identify(struct ahci_port * p,int slot,uint8_t * cfis)1033 handle_identify(struct ahci_port *p, int slot, uint8_t *cfis)
1034 {
1035 	struct ahci_cmd_hdr *hdr;
1036 
1037 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1038 	if (p->atapi || hdr->prdtl == 0) {
1039 		ahci_write_fis_d2h(p, slot, cfis,
1040 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1041 	} else {
1042 		ahci_write_fis_piosetup(p);
1043 		write_prdt(p, slot, cfis, (void*)&p->ata_ident, sizeof(struct ata_params));
1044 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
1045 	}
1046 }
1047 
1048 static void
ata_identify_init(struct ahci_port * p,int atapi)1049 ata_identify_init(struct ahci_port* p, int atapi)
1050 {
1051 	struct ata_params* ata_ident = &p->ata_ident;
1052 
1053 	if (atapi) {
1054 		ata_ident->config = ATA_PROTO_ATAPI | ATA_ATAPI_TYPE_CDROM |
1055 		    ATA_ATAPI_REMOVABLE | ATA_DRQ_FAST;
1056 		ata_ident->capabilities1 = ATA_SUPPORT_LBA |
1057 			ATA_SUPPORT_DMA;
1058 		ata_ident->capabilities2 = (1 << 14 | 1);
1059 		ata_ident->atavalid = ATA_FLAG_64_70 | ATA_FLAG_88;
1060 		ata_ident->obsolete62 = 0x3f;
1061 		ata_ident->mwdmamodes = 7;
1062 		if (p->xfermode & ATA_WDMA0)
1063 			ata_ident->mwdmamodes |= (1 << ((p->xfermode & 7) + 8));
1064 		ata_ident->apiomodes = 3;
1065 		ata_ident->mwdmamin = 0x0078;
1066 		ata_ident->mwdmarec = 0x0078;
1067 		ata_ident->pioblind = 0x0078;
1068 		ata_ident->pioiordy = 0x0078;
1069 		ata_ident->satacapabilities = (ATA_SATA_GEN1 | ATA_SATA_GEN2 | ATA_SATA_GEN3);
1070 		ata_ident->satacapabilities2 = ((p->ssts & ATA_SS_SPD_MASK) >> 3);
1071 		ata_ident->satasupport = ATA_SUPPORT_NCQ_STREAM;
1072 		ata_ident->version_major = 0x3f0;
1073 		ata_ident->support.command1 = (ATA_SUPPORT_POWERMGT | ATA_SUPPORT_PACKET |
1074 			ATA_SUPPORT_RESET | ATA_SUPPORT_NOP);
1075 		ata_ident->support.command2 = (1 << 14);
1076 		ata_ident->support.extension = (1 << 14);
1077 		ata_ident->enabled.command1 = (ATA_SUPPORT_POWERMGT | ATA_SUPPORT_PACKET |
1078 			ATA_SUPPORT_RESET | ATA_SUPPORT_NOP);
1079 		ata_ident->enabled.extension = (1 << 14);
1080 		ata_ident->udmamodes = 0x7f;
1081 		if (p->xfermode & ATA_UDMA0)
1082 			ata_ident->udmamodes |= (1 << ((p->xfermode & 7) + 8));
1083 		ata_ident->transport_major = 0x1020;
1084 		ata_ident->integrity = 0x00a5;
1085 	} else {
1086 		uint64_t sectors;
1087 		int sectsz, psectsz, psectoff, candelete, ro;
1088 		uint16_t cyl;
1089 		uint8_t sech, heads;
1090 
1091 		ro = blockif_is_ro(p->bctx);
1092 		candelete = blockif_candelete(p->bctx);
1093 		sectsz = blockif_sectsz(p->bctx);
1094 		sectors = blockif_size(p->bctx) / sectsz;
1095 		blockif_chs(p->bctx, &cyl, &heads, &sech);
1096 		blockif_psectsz(p->bctx, &psectsz, &psectoff);
1097 		ata_ident->config = ATA_DRQ_FAST;
1098 		ata_ident->cylinders = cyl;
1099 		ata_ident->heads = heads;
1100 		ata_ident->sectors = sech;
1101 
1102 		ata_ident->sectors_intr = (0x8000 | 128);
1103 		ata_ident->tcg = 0;
1104 
1105 		ata_ident->capabilities1 = ATA_SUPPORT_DMA |
1106 			ATA_SUPPORT_LBA | ATA_SUPPORT_IORDY;
1107 		ata_ident->capabilities2 = (1 << 14);
1108 		ata_ident->atavalid = ATA_FLAG_64_70 | ATA_FLAG_88;
1109 		if (p->mult_sectors)
1110 			ata_ident->multi = (ATA_MULTI_VALID | p->mult_sectors);
1111 		if (sectors <= 0x0fffffff) {
1112 			ata_ident->lba_size_1 = sectors;
1113 			ata_ident->lba_size_2 = (sectors >> 16);
1114 		} else {
1115 			ata_ident->lba_size_1 = 0xffff;
1116 			ata_ident->lba_size_2 = 0x0fff;
1117 		}
1118 		ata_ident->mwdmamodes = 0x7;
1119 		if (p->xfermode & ATA_WDMA0)
1120 			ata_ident->mwdmamodes |= (1 << ((p->xfermode & 7) + 8));
1121 		ata_ident->apiomodes = 0x3;
1122 		ata_ident->mwdmamin = 0x0078;
1123 		ata_ident->mwdmarec = 0x0078;
1124 		ata_ident->pioblind = 0x0078;
1125 		ata_ident->pioiordy = 0x0078;
1126 		ata_ident->support3 = 0;
1127 		ata_ident->queue = 31;
1128 		ata_ident->satacapabilities = (ATA_SATA_GEN1 | ATA_SATA_GEN2 | ATA_SATA_GEN3 |
1129 			ATA_SUPPORT_NCQ);
1130 		ata_ident->satacapabilities2 = (ATA_SUPPORT_RCVSND_FPDMA_QUEUED |
1131 			(p->ssts & ATA_SS_SPD_MASK) >> 3);
1132 		ata_ident->version_major = 0x3f0;
1133 		ata_ident->version_minor = 0x28;
1134 		ata_ident->support.command1 = (ATA_SUPPORT_POWERMGT | ATA_SUPPORT_WRITECACHE |
1135 			ATA_SUPPORT_LOOKAHEAD | ATA_SUPPORT_NOP);
1136 		ata_ident->support.command2 = (ATA_SUPPORT_ADDRESS48 | ATA_SUPPORT_FLUSHCACHE |
1137 			ATA_SUPPORT_FLUSHCACHE48 | 1 << 14);
1138 		ata_ident->support.extension = (1 << 14);
1139 		ata_ident->enabled.command1 = (ATA_SUPPORT_POWERMGT | ATA_SUPPORT_WRITECACHE |
1140 			ATA_SUPPORT_LOOKAHEAD | ATA_SUPPORT_NOP);
1141 		ata_ident->enabled.command2 = (ATA_SUPPORT_ADDRESS48 | ATA_SUPPORT_FLUSHCACHE |
1142 			ATA_SUPPORT_FLUSHCACHE48 | 1 << 15);
1143 		ata_ident->enabled.extension = (1 << 14);
1144 		ata_ident->udmamodes = 0x7f;
1145 		if (p->xfermode & ATA_UDMA0)
1146 			ata_ident->udmamodes |= (1 << ((p->xfermode & 7) + 8));
1147 		ata_ident->lba_size48_1 = sectors;
1148 		ata_ident->lba_size48_2 = (sectors >> 16);
1149 		ata_ident->lba_size48_3 = (sectors >> 32);
1150 		ata_ident->lba_size48_4 = (sectors >> 48);
1151 
1152 		if (candelete && !ro) {
1153 			ata_ident->support3 |= ATA_SUPPORT_RZAT | ATA_SUPPORT_DRAT;
1154 			ata_ident->max_dsm_blocks = 1;
1155 			ata_ident->support_dsm = ATA_SUPPORT_DSM_TRIM;
1156 		}
1157 		ata_ident->pss = ATA_PSS_VALID_VALUE;
1158 		ata_ident->lsalign = 0x4000;
1159 		if (psectsz > sectsz) {
1160 			ata_ident->pss |= ATA_PSS_MULTLS;
1161 			ata_ident->pss |= ffsl(psectsz / sectsz) - 1;
1162 			ata_ident->lsalign |= (psectoff / sectsz);
1163 		}
1164 		if (sectsz > 512) {
1165 			ata_ident->pss |= ATA_PSS_LSSABOVE512;
1166 			ata_ident->lss_1 = sectsz / 2;
1167 			ata_ident->lss_2 = ((sectsz / 2) >> 16);
1168 		}
1169 		ata_ident->support2 = (ATA_SUPPORT_RWLOGDMAEXT | 1 << 14);
1170 		ata_ident->enabled2 = (ATA_SUPPORT_RWLOGDMAEXT | 1 << 14);
1171 		ata_ident->transport_major = 0x1020;
1172 		ata_ident->integrity = 0x00a5;
1173 	}
1174 	ahci_checksum((uint8_t*)ata_ident, sizeof(struct ata_params));
1175 }
1176 
1177 static void
handle_atapi_identify(struct ahci_port * p,int slot,uint8_t * cfis)1178 handle_atapi_identify(struct ahci_port *p, int slot, uint8_t *cfis)
1179 {
1180 	if (!p->atapi) {
1181 		ahci_write_fis_d2h(p, slot, cfis,
1182 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1183 	} else {
1184 		ahci_write_fis_piosetup(p);
1185 		write_prdt(p, slot, cfis, (void *)&p->ata_ident, sizeof(struct ata_params));
1186 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_DSC | ATA_S_READY);
1187 	}
1188 }
1189 
1190 static void
atapi_inquiry(struct ahci_port * p,int slot,uint8_t * cfis)1191 atapi_inquiry(struct ahci_port *p, int slot, uint8_t *cfis)
1192 {
1193 	uint8_t buf[36];
1194 	uint8_t *acmd;
1195 	unsigned int len;
1196 	uint32_t tfd;
1197 
1198 	acmd = cfis + 0x40;
1199 
1200 	if (acmd[1] & 1) {		/* VPD */
1201 		if (acmd[2] == 0) {	/* Supported VPD pages */
1202 			buf[0] = 0x05;
1203 			buf[1] = 0;
1204 			buf[2] = 0;
1205 			buf[3] = 1;
1206 			buf[4] = 0;
1207 			len = 4 + buf[3];
1208 		} else {
1209 			p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1210 			p->asc = 0x24;
1211 			tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1212 			cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1213 			ahci_write_fis_d2h(p, slot, cfis, tfd);
1214 			return;
1215 		}
1216 	} else {
1217 		buf[0] = 0x05;
1218 		buf[1] = 0x80;
1219 		buf[2] = 0x00;
1220 		buf[3] = 0x21;
1221 		buf[4] = 31;
1222 		buf[5] = 0;
1223 		buf[6] = 0;
1224 		buf[7] = 0;
1225 		atapi_string(buf + 8, "BHYVE", 8);
1226 		atapi_string(buf + 16, "BHYVE DVD-ROM", 16);
1227 		atapi_string(buf + 32, "001", 4);
1228 		len = sizeof(buf);
1229 	}
1230 
1231 	if (len > acmd[4])
1232 		len = acmd[4];
1233 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1234 	write_prdt(p, slot, cfis, buf, len);
1235 	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1236 }
1237 
1238 static void
atapi_read_capacity(struct ahci_port * p,int slot,uint8_t * cfis)1239 atapi_read_capacity(struct ahci_port *p, int slot, uint8_t *cfis)
1240 {
1241 	uint8_t buf[8];
1242 	uint64_t sectors;
1243 
1244 	sectors = blockif_size(p->bctx) / 2048;
1245 	be32enc(buf, sectors - 1);
1246 	be32enc(buf + 4, 2048);
1247 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1248 	write_prdt(p, slot, cfis, buf, sizeof(buf));
1249 	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1250 }
1251 
1252 static void
atapi_read_toc(struct ahci_port * p,int slot,uint8_t * cfis)1253 atapi_read_toc(struct ahci_port *p, int slot, uint8_t *cfis)
1254 {
1255 	uint8_t *acmd;
1256 	uint8_t format;
1257 	unsigned int len;
1258 
1259 	acmd = cfis + 0x40;
1260 
1261 	len = be16dec(acmd + 7);
1262 	format = acmd[9] >> 6;
1263 	switch (format) {
1264 	case 0:
1265 	{
1266 		size_t size;
1267 		int msf;
1268 		uint64_t sectors;
1269 		uint8_t start_track, buf[20], *bp;
1270 
1271 		msf = (acmd[1] >> 1) & 1;
1272 		start_track = acmd[6];
1273 		if (start_track > 1 && start_track != 0xaa) {
1274 			uint32_t tfd;
1275 			p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1276 			p->asc = 0x24;
1277 			tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1278 			cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1279 			ahci_write_fis_d2h(p, slot, cfis, tfd);
1280 			return;
1281 		}
1282 		bp = buf + 2;
1283 		*bp++ = 1;
1284 		*bp++ = 1;
1285 		if (start_track <= 1) {
1286 			*bp++ = 0;
1287 			*bp++ = 0x14;
1288 			*bp++ = 1;
1289 			*bp++ = 0;
1290 			if (msf) {
1291 				*bp++ = 0;
1292 				lba_to_msf(bp, 0);
1293 				bp += 3;
1294 			} else {
1295 				*bp++ = 0;
1296 				*bp++ = 0;
1297 				*bp++ = 0;
1298 				*bp++ = 0;
1299 			}
1300 		}
1301 		*bp++ = 0;
1302 		*bp++ = 0x14;
1303 		*bp++ = 0xaa;
1304 		*bp++ = 0;
1305 		sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx);
1306 		sectors >>= 2;
1307 		if (msf) {
1308 			*bp++ = 0;
1309 			lba_to_msf(bp, sectors);
1310 			bp += 3;
1311 		} else {
1312 			be32enc(bp, sectors);
1313 			bp += 4;
1314 		}
1315 		size = bp - buf;
1316 		be16enc(buf, size - 2);
1317 		if (len > size)
1318 			len = size;
1319 		write_prdt(p, slot, cfis, buf, len);
1320 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1321 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1322 		break;
1323 	}
1324 	case 1:
1325 	{
1326 		uint8_t buf[12];
1327 
1328 		memset(buf, 0, sizeof(buf));
1329 		buf[1] = 0xa;
1330 		buf[2] = 0x1;
1331 		buf[3] = 0x1;
1332 		if (len > sizeof(buf))
1333 			len = sizeof(buf);
1334 		write_prdt(p, slot, cfis, buf, len);
1335 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1336 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1337 		break;
1338 	}
1339 	case 2:
1340 	{
1341 		size_t size;
1342 		int msf;
1343 		uint64_t sectors;
1344 		uint8_t *bp, buf[50];
1345 
1346 		msf = (acmd[1] >> 1) & 1;
1347 		bp = buf + 2;
1348 		*bp++ = 1;
1349 		*bp++ = 1;
1350 
1351 		*bp++ = 1;
1352 		*bp++ = 0x14;
1353 		*bp++ = 0;
1354 		*bp++ = 0xa0;
1355 		*bp++ = 0;
1356 		*bp++ = 0;
1357 		*bp++ = 0;
1358 		*bp++ = 0;
1359 		*bp++ = 1;
1360 		*bp++ = 0;
1361 		*bp++ = 0;
1362 
1363 		*bp++ = 1;
1364 		*bp++ = 0x14;
1365 		*bp++ = 0;
1366 		*bp++ = 0xa1;
1367 		*bp++ = 0;
1368 		*bp++ = 0;
1369 		*bp++ = 0;
1370 		*bp++ = 0;
1371 		*bp++ = 1;
1372 		*bp++ = 0;
1373 		*bp++ = 0;
1374 
1375 		*bp++ = 1;
1376 		*bp++ = 0x14;
1377 		*bp++ = 0;
1378 		*bp++ = 0xa2;
1379 		*bp++ = 0;
1380 		*bp++ = 0;
1381 		*bp++ = 0;
1382 		sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx);
1383 		sectors >>= 2;
1384 		if (msf) {
1385 			*bp++ = 0;
1386 			lba_to_msf(bp, sectors);
1387 			bp += 3;
1388 		} else {
1389 			be32enc(bp, sectors);
1390 			bp += 4;
1391 		}
1392 
1393 		*bp++ = 1;
1394 		*bp++ = 0x14;
1395 		*bp++ = 0;
1396 		*bp++ = 1;
1397 		*bp++ = 0;
1398 		*bp++ = 0;
1399 		*bp++ = 0;
1400 		if (msf) {
1401 			*bp++ = 0;
1402 			lba_to_msf(bp, 0);
1403 			bp += 3;
1404 		} else {
1405 			*bp++ = 0;
1406 			*bp++ = 0;
1407 			*bp++ = 0;
1408 			*bp++ = 0;
1409 		}
1410 
1411 		size = bp - buf;
1412 		be16enc(buf, size - 2);
1413 		if (len > size)
1414 			len = size;
1415 		write_prdt(p, slot, cfis, buf, len);
1416 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1417 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1418 		break;
1419 	}
1420 	default:
1421 	{
1422 		uint32_t tfd;
1423 
1424 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1425 		p->asc = 0x24;
1426 		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1427 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1428 		ahci_write_fis_d2h(p, slot, cfis, tfd);
1429 		break;
1430 	}
1431 	}
1432 }
1433 
1434 static void
atapi_report_luns(struct ahci_port * p,int slot,uint8_t * cfis)1435 atapi_report_luns(struct ahci_port *p, int slot, uint8_t *cfis)
1436 {
1437 	uint8_t buf[16];
1438 
1439 	memset(buf, 0, sizeof(buf));
1440 	buf[3] = 8;
1441 
1442 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1443 	write_prdt(p, slot, cfis, buf, sizeof(buf));
1444 	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1445 }
1446 
1447 static void
atapi_read(struct ahci_port * p,int slot,uint8_t * cfis,uint32_t done)1448 atapi_read(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t done)
1449 {
1450 	struct ahci_ioreq *aior;
1451 	struct ahci_cmd_hdr *hdr;
1452 	struct ahci_prdt_entry *prdt;
1453 	struct blockif_req *breq;
1454 	uint8_t *acmd;
1455 	uint64_t lba;
1456 	uint32_t len;
1457 	int err;
1458 
1459 	acmd = cfis + 0x40;
1460 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1461 	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
1462 
1463 	lba = be32dec(acmd + 2);
1464 	if (acmd[0] == READ_10)
1465 		len = be16dec(acmd + 7);
1466 	else
1467 		len = be32dec(acmd + 6);
1468 	if (len == 0) {
1469 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1470 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1471 	}
1472 	lba *= 2048;
1473 	len *= 2048;
1474 
1475 	/*
1476 	 * Pull request off free list
1477 	 */
1478 	aior = STAILQ_FIRST(&p->iofhd);
1479 	assert(aior != NULL);
1480 	STAILQ_REMOVE_HEAD(&p->iofhd, io_flist);
1481 	aior->cfis = cfis;
1482 	aior->slot = slot;
1483 	aior->len = len;
1484 	aior->done = done;
1485 	aior->readop = 1;
1486 	breq = &aior->io_req;
1487 	breq->br_offset = lba + done;
1488 	ahci_build_iov(p, aior, prdt, hdr->prdtl);
1489 
1490 	/* Mark this command in-flight. */
1491 	p->pending |= 1 << slot;
1492 
1493 	/* Stuff request onto busy list. */
1494 	TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist);
1495 
1496 	err = blockif_read(p->bctx, breq);
1497 	assert(err == 0);
1498 }
1499 
1500 static void
atapi_request_sense(struct ahci_port * p,int slot,uint8_t * cfis)1501 atapi_request_sense(struct ahci_port *p, int slot, uint8_t *cfis)
1502 {
1503 	uint8_t buf[64];
1504 	uint8_t *acmd;
1505 	unsigned int len;
1506 
1507 	acmd = cfis + 0x40;
1508 	len = acmd[4];
1509 	if (len > sizeof(buf))
1510 		len = sizeof(buf);
1511 	memset(buf, 0, len);
1512 	buf[0] = 0x70 | (1 << 7);
1513 	buf[2] = p->sense_key;
1514 	buf[7] = 10;
1515 	buf[12] = p->asc;
1516 	write_prdt(p, slot, cfis, buf, len);
1517 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1518 	ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1519 }
1520 
1521 static void
atapi_start_stop_unit(struct ahci_port * p,int slot,uint8_t * cfis)1522 atapi_start_stop_unit(struct ahci_port *p, int slot, uint8_t *cfis)
1523 {
1524 	uint8_t *acmd = cfis + 0x40;
1525 	uint32_t tfd;
1526 
1527 	switch (acmd[4] & 3) {
1528 	case 0:
1529 	case 1:
1530 	case 3:
1531 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1532 		tfd = ATA_S_READY | ATA_S_DSC;
1533 		break;
1534 	case 2:
1535 		/* TODO eject media */
1536 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1537 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1538 		p->asc = 0x53;
1539 		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1540 		break;
1541 	}
1542 	ahci_write_fis_d2h(p, slot, cfis, tfd);
1543 }
1544 
1545 static void
atapi_mode_sense(struct ahci_port * p,int slot,uint8_t * cfis)1546 atapi_mode_sense(struct ahci_port *p, int slot, uint8_t *cfis)
1547 {
1548 	uint8_t *acmd;
1549 	uint32_t tfd;
1550 	uint8_t pc, code;
1551 	unsigned int len;
1552 
1553 	acmd = cfis + 0x40;
1554 	len = be16dec(acmd + 7);
1555 	pc = acmd[2] >> 6;
1556 	code = acmd[2] & 0x3f;
1557 
1558 	switch (pc) {
1559 	case 0:
1560 		switch (code) {
1561 		case MODEPAGE_RW_ERROR_RECOVERY:
1562 		{
1563 			uint8_t buf[16];
1564 
1565 			if (len > sizeof(buf))
1566 				len = sizeof(buf);
1567 
1568 			memset(buf, 0, sizeof(buf));
1569 			be16enc(buf, 16 - 2);
1570 			buf[2] = 0x70;
1571 			buf[8] = 0x01;
1572 			buf[9] = 16 - 10;
1573 			buf[11] = 0x05;
1574 			write_prdt(p, slot, cfis, buf, len);
1575 			tfd = ATA_S_READY | ATA_S_DSC;
1576 			break;
1577 		}
1578 		case MODEPAGE_CD_CAPABILITIES:
1579 		{
1580 			uint8_t buf[30];
1581 
1582 			if (len > sizeof(buf))
1583 				len = sizeof(buf);
1584 
1585 			memset(buf, 0, sizeof(buf));
1586 			be16enc(buf, 30 - 2);
1587 			buf[2] = 0x70;
1588 			buf[8] = 0x2A;
1589 			buf[9] = 30 - 10;
1590 			buf[10] = 0x08;
1591 			buf[12] = 0x71;
1592 			be16enc(&buf[18], 2);
1593 			be16enc(&buf[20], 512);
1594 			write_prdt(p, slot, cfis, buf, len);
1595 			tfd = ATA_S_READY | ATA_S_DSC;
1596 			break;
1597 		}
1598 		default:
1599 			goto error;
1600 			break;
1601 		}
1602 		break;
1603 	case 3:
1604 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1605 		p->asc = 0x39;
1606 		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1607 		break;
1608 error:
1609 	case 1:
1610 	case 2:
1611 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1612 		p->asc = 0x24;
1613 		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1614 		break;
1615 	}
1616 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1617 	ahci_write_fis_d2h(p, slot, cfis, tfd);
1618 }
1619 
1620 static void
atapi_get_event_status_notification(struct ahci_port * p,int slot,uint8_t * cfis)1621 atapi_get_event_status_notification(struct ahci_port *p, int slot,
1622     uint8_t *cfis)
1623 {
1624 	uint8_t *acmd;
1625 	uint32_t tfd;
1626 
1627 	acmd = cfis + 0x40;
1628 
1629 	/* we don't support asynchronous operation */
1630 	if (!(acmd[1] & 1)) {
1631 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1632 		p->asc = 0x24;
1633 		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
1634 	} else {
1635 		uint8_t buf[8];
1636 		unsigned int len;
1637 
1638 		len = be16dec(acmd + 7);
1639 		if (len > sizeof(buf))
1640 			len = sizeof(buf);
1641 
1642 		memset(buf, 0, sizeof(buf));
1643 		be16enc(buf, 8 - 2);
1644 		buf[2] = 0x04;
1645 		buf[3] = 0x10;
1646 		buf[5] = 0x02;
1647 		write_prdt(p, slot, cfis, buf, len);
1648 		tfd = ATA_S_READY | ATA_S_DSC;
1649 	}
1650 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1651 	ahci_write_fis_d2h(p, slot, cfis, tfd);
1652 }
1653 
1654 static void
handle_packet_cmd(struct ahci_port * p,int slot,uint8_t * cfis)1655 handle_packet_cmd(struct ahci_port *p, int slot, uint8_t *cfis)
1656 {
1657 	uint8_t *acmd;
1658 
1659 	acmd = cfis + 0x40;
1660 
1661 #ifdef AHCI_DEBUG
1662 	{
1663 		int i;
1664 		DPRINTF("ACMD:");
1665 		for (i = 0; i < 16; i++)
1666 			DPRINTF("%02x ", acmd[i]);
1667 		DPRINTF("");
1668 	}
1669 #endif
1670 
1671 	switch (acmd[0]) {
1672 	case TEST_UNIT_READY:
1673 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1674 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1675 		break;
1676 	case INQUIRY:
1677 		atapi_inquiry(p, slot, cfis);
1678 		break;
1679 	case READ_CAPACITY:
1680 		atapi_read_capacity(p, slot, cfis);
1681 		break;
1682 	case PREVENT_ALLOW:
1683 		/* TODO */
1684 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1685 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1686 		break;
1687 	case READ_TOC:
1688 		atapi_read_toc(p, slot, cfis);
1689 		break;
1690 	case REPORT_LUNS:
1691 		atapi_report_luns(p, slot, cfis);
1692 		break;
1693 	case READ_10:
1694 	case READ_12:
1695 		atapi_read(p, slot, cfis, 0);
1696 		break;
1697 	case REQUEST_SENSE:
1698 		atapi_request_sense(p, slot, cfis);
1699 		break;
1700 	case START_STOP_UNIT:
1701 		atapi_start_stop_unit(p, slot, cfis);
1702 		break;
1703 	case MODE_SENSE_10:
1704 		atapi_mode_sense(p, slot, cfis);
1705 		break;
1706 	case GET_EVENT_STATUS_NOTIFICATION:
1707 		atapi_get_event_status_notification(p, slot, cfis);
1708 		break;
1709 	default:
1710 		cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
1711 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
1712 		p->asc = 0x20;
1713 		ahci_write_fis_d2h(p, slot, cfis, (p->sense_key << 12) |
1714 				ATA_S_READY | ATA_S_ERROR);
1715 		break;
1716 	}
1717 }
1718 
1719 static void
ahci_handle_cmd(struct ahci_port * p,int slot,uint8_t * cfis)1720 ahci_handle_cmd(struct ahci_port *p, int slot, uint8_t *cfis)
1721 {
1722 
1723 	p->tfd |= ATA_S_BUSY;
1724 	switch (cfis[2]) {
1725 	case ATA_ATA_IDENTIFY:
1726 		handle_identify(p, slot, cfis);
1727 		break;
1728 	case ATA_SETFEATURES:
1729 	{
1730 		switch (cfis[3]) {
1731 		case ATA_SF_ENAB_SATA_SF:
1732 			switch (cfis[12]) {
1733 			case ATA_SATA_SF_AN:
1734 				p->tfd = ATA_S_DSC | ATA_S_READY;
1735 				break;
1736 			default:
1737 				p->tfd = ATA_S_ERROR | ATA_S_READY;
1738 				p->tfd |= (ATA_ERROR_ABORT << 8);
1739 				break;
1740 			}
1741 			break;
1742 		case ATA_SF_ENAB_WCACHE:
1743 		case ATA_SF_DIS_WCACHE:
1744 		case ATA_SF_ENAB_RCACHE:
1745 		case ATA_SF_DIS_RCACHE:
1746 			p->tfd = ATA_S_DSC | ATA_S_READY;
1747 			break;
1748 		case ATA_SF_SETXFER:
1749 		{
1750 			switch (cfis[12] & 0xf8) {
1751 			case ATA_PIO:
1752 			case ATA_PIO0:
1753 				break;
1754 			case ATA_WDMA0:
1755 			case ATA_UDMA0:
1756 				p->xfermode = (cfis[12] & 0x7);
1757 				break;
1758 			}
1759 			p->tfd = ATA_S_DSC | ATA_S_READY;
1760 			break;
1761 		}
1762 		default:
1763 			p->tfd = ATA_S_ERROR | ATA_S_READY;
1764 			p->tfd |= (ATA_ERROR_ABORT << 8);
1765 			break;
1766 		}
1767 		ahci_write_fis_d2h(p, slot, cfis, p->tfd);
1768 		break;
1769 	}
1770 	case ATA_SET_MULTI:
1771 		if (cfis[12] != 0 &&
1772 			(cfis[12] > 128 || (cfis[12] & (cfis[12] - 1)))) {
1773 			p->tfd = ATA_S_ERROR | ATA_S_READY;
1774 			p->tfd |= (ATA_ERROR_ABORT << 8);
1775 		} else {
1776 			p->mult_sectors = cfis[12];
1777 			p->tfd = ATA_S_DSC | ATA_S_READY;
1778 		}
1779 		ahci_write_fis_d2h(p, slot, cfis, p->tfd);
1780 		break;
1781 	case ATA_READ:
1782 	case ATA_WRITE:
1783 	case ATA_READ48:
1784 	case ATA_WRITE48:
1785 	case ATA_READ_MUL:
1786 	case ATA_WRITE_MUL:
1787 	case ATA_READ_MUL48:
1788 	case ATA_WRITE_MUL48:
1789 	case ATA_READ_DMA:
1790 	case ATA_WRITE_DMA:
1791 	case ATA_READ_DMA48:
1792 	case ATA_WRITE_DMA48:
1793 	case ATA_READ_FPDMA_QUEUED:
1794 	case ATA_WRITE_FPDMA_QUEUED:
1795 		ahci_handle_rw(p, slot, cfis, 0);
1796 		break;
1797 	case ATA_FLUSHCACHE:
1798 	case ATA_FLUSHCACHE48:
1799 		ahci_handle_flush(p, slot, cfis);
1800 		break;
1801 	case ATA_DATA_SET_MANAGEMENT:
1802 		if (cfis[11] == 0 && cfis[3] == ATA_DSM_TRIM &&
1803 		    cfis[13] == 0 && cfis[12] == 1) {
1804 			ahci_handle_dsm_trim(p, slot, cfis);
1805 			break;
1806 		}
1807 		ahci_write_fis_d2h(p, slot, cfis,
1808 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1809 		break;
1810 	case ATA_SEND_FPDMA_QUEUED:
1811 		if ((cfis[13] & 0x1f) == ATA_SFPDMA_DSM &&
1812 		    cfis[17] == 0 && cfis[16] == ATA_DSM_TRIM &&
1813 		    cfis[11] == 0 && cfis[3] == 1) {
1814 			ahci_handle_dsm_trim(p, slot, cfis);
1815 			break;
1816 		}
1817 		ahci_write_fis_d2h(p, slot, cfis,
1818 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1819 		break;
1820 	case ATA_READ_LOG_EXT:
1821 	case ATA_READ_LOG_DMA_EXT:
1822 		ahci_handle_read_log(p, slot, cfis);
1823 		break;
1824 	case ATA_SECURITY_FREEZE_LOCK:
1825 	case ATA_SMART_CMD:
1826 	case ATA_NOP:
1827 		ahci_write_fis_d2h(p, slot, cfis,
1828 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1829 		break;
1830 	case ATA_CHECK_POWER_MODE:
1831 		cfis[12] = 0xff;	/* always on */
1832 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1833 		break;
1834 	case ATA_STANDBY_CMD:
1835 	case ATA_STANDBY_IMMEDIATE:
1836 	case ATA_IDLE_CMD:
1837 	case ATA_IDLE_IMMEDIATE:
1838 	case ATA_SLEEP:
1839 	case ATA_READ_VERIFY:
1840 	case ATA_READ_VERIFY48:
1841 		ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC);
1842 		break;
1843 	case ATA_ATAPI_IDENTIFY:
1844 		handle_atapi_identify(p, slot, cfis);
1845 		break;
1846 	case ATA_PACKET_CMD:
1847 		if (!p->atapi) {
1848 			ahci_write_fis_d2h(p, slot, cfis,
1849 			    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1850 		} else
1851 			handle_packet_cmd(p, slot, cfis);
1852 		break;
1853 	default:
1854 		EPRINTLN("Unsupported cmd:%02x", cfis[2]);
1855 		ahci_write_fis_d2h(p, slot, cfis,
1856 		    (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR);
1857 		break;
1858 	}
1859 }
1860 
1861 static void
ahci_handle_slot(struct ahci_port * p,int slot)1862 ahci_handle_slot(struct ahci_port *p, int slot)
1863 {
1864 	struct ahci_cmd_hdr *hdr;
1865 #ifdef AHCI_DEBUG
1866 	struct ahci_prdt_entry *prdt;
1867 #endif
1868 	struct pci_ahci_softc *sc;
1869 	uint8_t *cfis;
1870 #ifdef AHCI_DEBUG
1871 	int cfl, i;
1872 #endif
1873 
1874 	sc = p->pr_sc;
1875 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1876 #ifdef AHCI_DEBUG
1877 	cfl = (hdr->flags & 0x1f) * 4;
1878 #endif
1879 	cfis = paddr_guest2host(ahci_ctx(sc), hdr->ctba,
1880 			0x80 + hdr->prdtl * sizeof(struct ahci_prdt_entry));
1881 #ifdef AHCI_DEBUG
1882 	prdt = (struct ahci_prdt_entry *)(cfis + 0x80);
1883 
1884 	DPRINTF("cfis:");
1885 	for (i = 0; i < cfl; i++) {
1886 		if (i % 10 == 0)
1887 			DPRINTF("");
1888 		DPRINTF("%02x ", cfis[i]);
1889 	}
1890 	DPRINTF("");
1891 
1892 	for (i = 0; i < hdr->prdtl; i++) {
1893 		DPRINTF("%d@%08"PRIx64"", prdt->dbc & 0x3fffff, prdt->dba);
1894 		prdt++;
1895 	}
1896 #endif
1897 
1898 	if (cfis[0] != FIS_TYPE_REGH2D) {
1899 		EPRINTLN("Not a H2D FIS:%02x", cfis[0]);
1900 		return;
1901 	}
1902 
1903 	if (cfis[1] & 0x80) {
1904 		ahci_handle_cmd(p, slot, cfis);
1905 	} else {
1906 		if (cfis[15] & (1 << 2))
1907 			p->reset = 1;
1908 		else if (p->reset) {
1909 			p->reset = 0;
1910 			ahci_port_reset(p);
1911 		}
1912 		p->ci &= ~(1 << slot);
1913 	}
1914 }
1915 
1916 static void
ahci_handle_port(struct ahci_port * p)1917 ahci_handle_port(struct ahci_port *p)
1918 {
1919 
1920 	if (!(p->cmd & AHCI_P_CMD_ST))
1921 		return;
1922 
1923 	/*
1924 	 * Search for any new commands to issue ignoring those that
1925 	 * are already in-flight.  Stop if device is busy or in error.
1926 	 */
1927 	for (; (p->ci & ~p->pending) != 0; p->ccs = ((p->ccs + 1) & 31)) {
1928 		if ((p->tfd & (ATA_S_BUSY | ATA_S_DRQ)) != 0)
1929 			break;
1930 		if (p->waitforclear)
1931 			break;
1932 		if ((p->ci & ~p->pending & (1 << p->ccs)) != 0) {
1933 			p->cmd &= ~AHCI_P_CMD_CCS_MASK;
1934 			p->cmd |= p->ccs << AHCI_P_CMD_CCS_SHIFT;
1935 			ahci_handle_slot(p, p->ccs);
1936 		}
1937 	}
1938 }
1939 
1940 /*
1941  * blockif callback routine - this runs in the context of the blockif
1942  * i/o thread, so the mutex needs to be acquired.
1943  */
1944 static void
ata_ioreq_cb(struct blockif_req * br,int err)1945 ata_ioreq_cb(struct blockif_req *br, int err)
1946 {
1947 	struct ahci_cmd_hdr *hdr;
1948 	struct ahci_ioreq *aior;
1949 	struct ahci_port *p;
1950 	struct pci_ahci_softc *sc;
1951 	uint32_t tfd;
1952 	uint8_t *cfis, *dsm;
1953 	int slot, ncq;
1954 
1955 	DPRINTF("%s %d", __func__, err);
1956 
1957 	ncq = 0;
1958 	aior = br->br_param;
1959 	p = aior->io_pr;
1960 	cfis = aior->cfis;
1961 	slot = aior->slot;
1962 	sc = p->pr_sc;
1963 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE);
1964 
1965 	if (cfis[2] == ATA_WRITE_FPDMA_QUEUED ||
1966 	    cfis[2] == ATA_READ_FPDMA_QUEUED ||
1967 	    cfis[2] == ATA_SEND_FPDMA_QUEUED)
1968 		ncq = 1;
1969 	dsm = aior->dsm;
1970 	aior->dsm = NULL;
1971 
1972 	pthread_mutex_lock(&sc->mtx);
1973 
1974 	/*
1975 	 * Delete the blockif request from the busy list
1976 	 */
1977 	TAILQ_REMOVE(&p->iobhd, aior, io_blist);
1978 
1979 	/*
1980 	 * Move the blockif request back to the free list
1981 	 */
1982 	STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
1983 
1984 	if (!err)
1985 		hdr->prdbc = aior->done;
1986 
1987 	if (!err && aior->more) {
1988 		if (dsm != NULL)
1989 			ahci_handle_next_trim(p, slot, cfis, dsm,
1990 			    aior->len, aior->done);
1991 		else
1992 			ahci_handle_rw(p, slot, cfis, aior->done);
1993 		goto out;
1994 	}
1995 
1996 	if (!err)
1997 		tfd = ATA_S_READY | ATA_S_DSC;
1998 	else
1999 		tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR;
2000 	if (ncq)
2001 		ahci_write_fis_sdb(p, slot, cfis, tfd);
2002 	else
2003 		ahci_write_fis_d2h(p, slot, cfis, tfd);
2004 
2005 	/*
2006 	 * This command is now complete.
2007 	 */
2008 	p->pending &= ~(1 << slot);
2009 
2010 	ahci_check_stopped(p);
2011 	ahci_handle_port(p);
2012 	free(dsm);
2013 out:
2014 	pthread_mutex_unlock(&sc->mtx);
2015 	DPRINTF("%s exit", __func__);
2016 }
2017 
2018 static void
atapi_ioreq_cb(struct blockif_req * br,int err)2019 atapi_ioreq_cb(struct blockif_req *br, int err)
2020 {
2021 	struct ahci_cmd_hdr *hdr;
2022 	struct ahci_ioreq *aior;
2023 	struct ahci_port *p;
2024 	struct pci_ahci_softc *sc;
2025 	uint8_t *cfis;
2026 	uint32_t tfd;
2027 	int slot;
2028 
2029 	DPRINTF("%s %d", __func__, err);
2030 
2031 	aior = br->br_param;
2032 	p = aior->io_pr;
2033 	cfis = aior->cfis;
2034 	slot = aior->slot;
2035 	sc = p->pr_sc;
2036 	hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + aior->slot * AHCI_CL_SIZE);
2037 
2038 	pthread_mutex_lock(&sc->mtx);
2039 
2040 	/*
2041 	 * Delete the blockif request from the busy list
2042 	 */
2043 	TAILQ_REMOVE(&p->iobhd, aior, io_blist);
2044 
2045 	/*
2046 	 * Move the blockif request back to the free list
2047 	 */
2048 	STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist);
2049 
2050 	if (!err)
2051 		hdr->prdbc = aior->done;
2052 
2053 	if (!err && aior->more) {
2054 		atapi_read(p, slot, cfis, aior->done);
2055 		goto out;
2056 	}
2057 
2058 	if (!err) {
2059 		tfd = ATA_S_READY | ATA_S_DSC;
2060 	} else {
2061 		p->sense_key = ATA_SENSE_ILLEGAL_REQUEST;
2062 		p->asc = 0x21;
2063 		tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR;
2064 	}
2065 	cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN;
2066 	ahci_write_fis_d2h(p, slot, cfis, tfd);
2067 
2068 	/*
2069 	 * This command is now complete.
2070 	 */
2071 	p->pending &= ~(1 << slot);
2072 
2073 	ahci_check_stopped(p);
2074 	ahci_handle_port(p);
2075 out:
2076 	pthread_mutex_unlock(&sc->mtx);
2077 	DPRINTF("%s exit", __func__);
2078 }
2079 
2080 static void
pci_ahci_ioreq_init(struct ahci_port * pr)2081 pci_ahci_ioreq_init(struct ahci_port *pr)
2082 {
2083 	struct ahci_ioreq *vr;
2084 	int i;
2085 
2086 	pr->ioqsz = blockif_queuesz(pr->bctx);
2087 	pr->ioreq = calloc(pr->ioqsz, sizeof(struct ahci_ioreq));
2088 	STAILQ_INIT(&pr->iofhd);
2089 
2090 	/*
2091 	 * Add all i/o request entries to the free queue
2092 	 */
2093 	for (i = 0; i < pr->ioqsz; i++) {
2094 		vr = &pr->ioreq[i];
2095 		vr->io_pr = pr;
2096 		if (!pr->atapi)
2097 			vr->io_req.br_callback = ata_ioreq_cb;
2098 		else
2099 			vr->io_req.br_callback = atapi_ioreq_cb;
2100 		vr->io_req.br_param = vr;
2101 		STAILQ_INSERT_TAIL(&pr->iofhd, vr, io_flist);
2102 	}
2103 
2104 	TAILQ_INIT(&pr->iobhd);
2105 }
2106 
2107 static void
pci_ahci_port_write(struct pci_ahci_softc * sc,uint64_t offset,uint64_t value)2108 pci_ahci_port_write(struct pci_ahci_softc *sc, uint64_t offset, uint64_t value)
2109 {
2110 	int port = (offset - AHCI_OFFSET) / AHCI_STEP;
2111 	offset = (offset - AHCI_OFFSET) % AHCI_STEP;
2112 	struct ahci_port *p = &sc->port[port];
2113 
2114 	DPRINTF("pci_ahci_port %d: write offset 0x%"PRIx64" value 0x%"PRIx64"",
2115 		port, offset, value);
2116 
2117 	switch (offset) {
2118 	case AHCI_P_CLB:
2119 		p->clb = value;
2120 		break;
2121 	case AHCI_P_CLBU:
2122 		p->clbu = value;
2123 		break;
2124 	case AHCI_P_FB:
2125 		p->fb = value;
2126 		break;
2127 	case AHCI_P_FBU:
2128 		p->fbu = value;
2129 		break;
2130 	case AHCI_P_IS:
2131 		p->is &= ~value;
2132 		ahci_port_intr(p);
2133 		break;
2134 	case AHCI_P_IE:
2135 		p->ie = value & 0xFDC000FF;
2136 		ahci_port_intr(p);
2137 		break;
2138 	case AHCI_P_CMD:
2139 	{
2140 		p->cmd &= ~(AHCI_P_CMD_ST | AHCI_P_CMD_SUD | AHCI_P_CMD_POD |
2141 		    AHCI_P_CMD_CLO | AHCI_P_CMD_FRE | AHCI_P_CMD_APSTE |
2142 		    AHCI_P_CMD_ATAPI | AHCI_P_CMD_DLAE | AHCI_P_CMD_ALPE |
2143 		    AHCI_P_CMD_ASP | AHCI_P_CMD_ICC_MASK);
2144 		p->cmd |= (AHCI_P_CMD_ST | AHCI_P_CMD_SUD | AHCI_P_CMD_POD |
2145 		    AHCI_P_CMD_CLO | AHCI_P_CMD_FRE | AHCI_P_CMD_APSTE |
2146 		    AHCI_P_CMD_ATAPI | AHCI_P_CMD_DLAE | AHCI_P_CMD_ALPE |
2147 		    AHCI_P_CMD_ASP | AHCI_P_CMD_ICC_MASK) & value;
2148 
2149 		if (!(value & AHCI_P_CMD_ST)) {
2150 			ahci_port_stop(p);
2151 		} else {
2152 			uint64_t clb;
2153 
2154 			p->cmd |= AHCI_P_CMD_CR;
2155 			clb = (uint64_t)p->clbu << 32 | p->clb;
2156 			p->cmd_lst = paddr_guest2host(ahci_ctx(sc), clb,
2157 					AHCI_CL_SIZE * AHCI_MAX_SLOTS);
2158 		}
2159 
2160 		if (value & AHCI_P_CMD_FRE) {
2161 			uint64_t fb;
2162 
2163 			p->cmd |= AHCI_P_CMD_FR;
2164 			fb = (uint64_t)p->fbu << 32 | p->fb;
2165 			/* we don't support FBSCP, so rfis size is 256Bytes */
2166 			p->rfis = paddr_guest2host(ahci_ctx(sc), fb, 256);
2167 		} else {
2168 			p->cmd &= ~AHCI_P_CMD_FR;
2169 		}
2170 
2171 		if (value & AHCI_P_CMD_CLO) {
2172 			p->tfd &= ~(ATA_S_BUSY | ATA_S_DRQ);
2173 			p->cmd &= ~AHCI_P_CMD_CLO;
2174 		}
2175 
2176 		if (value & AHCI_P_CMD_ICC_MASK) {
2177 			p->cmd &= ~AHCI_P_CMD_ICC_MASK;
2178 		}
2179 
2180 		ahci_handle_port(p);
2181 		break;
2182 	}
2183 	case AHCI_P_TFD:
2184 	case AHCI_P_SIG:
2185 	case AHCI_P_SSTS:
2186 		EPRINTLN("pci_ahci_port: read only registers 0x%"PRIx64"", offset);
2187 		break;
2188 	case AHCI_P_SCTL:
2189 		p->sctl = value;
2190 		if (!(p->cmd & AHCI_P_CMD_ST)) {
2191 			if (value & ATA_SC_DET_RESET)
2192 				ahci_port_reset(p);
2193 		}
2194 		break;
2195 	case AHCI_P_SERR:
2196 		p->serr &= ~value;
2197 		break;
2198 	case AHCI_P_SACT:
2199 		p->sact |= value;
2200 		break;
2201 	case AHCI_P_CI:
2202 		p->ci |= value;
2203 		ahci_handle_port(p);
2204 		break;
2205 	case AHCI_P_SNTF:
2206 	case AHCI_P_FBS:
2207 	default:
2208 		break;
2209 	}
2210 }
2211 
2212 static void
pci_ahci_host_write(struct pci_ahci_softc * sc,uint64_t offset,uint64_t value)2213 pci_ahci_host_write(struct pci_ahci_softc *sc, uint64_t offset, uint64_t value)
2214 {
2215 	DPRINTF("pci_ahci_host: write offset 0x%"PRIx64" value 0x%"PRIx64"",
2216 		offset, value);
2217 
2218 	switch (offset) {
2219 	case AHCI_CAP:
2220 	case AHCI_PI:
2221 	case AHCI_VS:
2222 	case AHCI_CAP2:
2223 		DPRINTF("pci_ahci_host: read only registers 0x%"PRIx64"", offset);
2224 		break;
2225 	case AHCI_GHC:
2226 		if (value & AHCI_GHC_HR) {
2227 			ahci_reset(sc);
2228 			break;
2229 		}
2230 		if (value & AHCI_GHC_IE)
2231 			sc->ghc |= AHCI_GHC_IE;
2232 		else
2233 			sc->ghc &= ~AHCI_GHC_IE;
2234 		ahci_generate_intr(sc, 0xffffffff);
2235 		break;
2236 	case AHCI_IS:
2237 		sc->is &= ~value;
2238 		ahci_generate_intr(sc, value);
2239 		break;
2240 	default:
2241 		break;
2242 	}
2243 }
2244 
2245 static void
pci_ahci_write(struct pci_devinst * pi,int baridx,uint64_t offset,int size,uint64_t value)2246 pci_ahci_write(struct pci_devinst *pi, int baridx, uint64_t offset, int size,
2247     uint64_t value)
2248 {
2249 	struct pci_ahci_softc *sc = pi->pi_arg;
2250 
2251 	assert(baridx == 5);
2252 	assert((offset % 4) == 0 && size == 4);
2253 
2254 	pthread_mutex_lock(&sc->mtx);
2255 
2256 	if (offset < AHCI_OFFSET)
2257 		pci_ahci_host_write(sc, offset, value);
2258 	else if (offset < (uint64_t)AHCI_OFFSET + sc->ports * AHCI_STEP)
2259 		pci_ahci_port_write(sc, offset, value);
2260 	else
2261 		EPRINTLN("pci_ahci: unknown i/o write offset 0x%"PRIx64"", offset);
2262 
2263 	pthread_mutex_unlock(&sc->mtx);
2264 }
2265 
2266 static uint64_t
pci_ahci_host_read(struct pci_ahci_softc * sc,uint64_t offset)2267 pci_ahci_host_read(struct pci_ahci_softc *sc, uint64_t offset)
2268 {
2269 	uint32_t value;
2270 
2271 	switch (offset) {
2272 	case AHCI_CAP:
2273 	case AHCI_GHC:
2274 	case AHCI_IS:
2275 	case AHCI_PI:
2276 	case AHCI_VS:
2277 	case AHCI_CCCC:
2278 	case AHCI_CCCP:
2279 	case AHCI_EM_LOC:
2280 	case AHCI_EM_CTL:
2281 	case AHCI_CAP2:
2282 	{
2283 		uint32_t *p = &sc->cap;
2284 		p += (offset - AHCI_CAP) / sizeof(uint32_t);
2285 		value = *p;
2286 		break;
2287 	}
2288 	default:
2289 		value = 0;
2290 		break;
2291 	}
2292 	DPRINTF("pci_ahci_host: read offset 0x%"PRIx64" value 0x%x",
2293 		offset, value);
2294 
2295 	return (value);
2296 }
2297 
2298 static uint64_t
pci_ahci_port_read(struct pci_ahci_softc * sc,uint64_t offset)2299 pci_ahci_port_read(struct pci_ahci_softc *sc, uint64_t offset)
2300 {
2301 	uint32_t value;
2302 	int port = (offset - AHCI_OFFSET) / AHCI_STEP;
2303 	offset = (offset - AHCI_OFFSET) % AHCI_STEP;
2304 
2305 	switch (offset) {
2306 	case AHCI_P_CLB:
2307 	case AHCI_P_CLBU:
2308 	case AHCI_P_FB:
2309 	case AHCI_P_FBU:
2310 	case AHCI_P_IS:
2311 	case AHCI_P_IE:
2312 	case AHCI_P_CMD:
2313 	case AHCI_P_TFD:
2314 	case AHCI_P_SIG:
2315 	case AHCI_P_SSTS:
2316 	case AHCI_P_SCTL:
2317 	case AHCI_P_SERR:
2318 	case AHCI_P_SACT:
2319 	case AHCI_P_CI:
2320 	case AHCI_P_SNTF:
2321 	case AHCI_P_FBS:
2322 	{
2323 		uint32_t *p= &sc->port[port].clb;
2324 		p += (offset - AHCI_P_CLB) / sizeof(uint32_t);
2325 		value = *p;
2326 		break;
2327 	}
2328 	default:
2329 		value = 0;
2330 		break;
2331 	}
2332 
2333 	DPRINTF("pci_ahci_port %d: read offset 0x%"PRIx64" value 0x%x",
2334 		port, offset, value);
2335 
2336 	return value;
2337 }
2338 
2339 static uint64_t
pci_ahci_read(struct pci_devinst * pi,int baridx,uint64_t regoff,int size)2340 pci_ahci_read(struct pci_devinst *pi, int baridx, uint64_t regoff, int size)
2341 {
2342 	struct pci_ahci_softc *sc = pi->pi_arg;
2343 	uint64_t offset;
2344 	uint32_t value;
2345 
2346 	assert(baridx == 5);
2347 	assert(size == 1 || size == 2 || size == 4);
2348 	assert((regoff & (size - 1)) == 0);
2349 
2350 	pthread_mutex_lock(&sc->mtx);
2351 
2352 	offset = regoff & ~0x3;	    /* round down to a multiple of 4 bytes */
2353 	if (offset < AHCI_OFFSET)
2354 		value = pci_ahci_host_read(sc, offset);
2355 	else if (offset < (uint64_t)AHCI_OFFSET + sc->ports * AHCI_STEP)
2356 		value = pci_ahci_port_read(sc, offset);
2357 	else {
2358 		value = 0;
2359 		EPRINTLN("pci_ahci: unknown i/o read offset 0x%"PRIx64"",
2360 		    regoff);
2361 	}
2362 	value >>= 8 * (regoff & 0x3);
2363 
2364 	pthread_mutex_unlock(&sc->mtx);
2365 
2366 	return (value);
2367 }
2368 
2369 /*
2370  * Each AHCI controller has a "port" node which contains nodes for
2371  * each port named after the decimal number of the port (no leading
2372  * zeroes).  Port nodes contain a "type" ("hd" or "cd"), as well as
2373  * options for blockif.  For example:
2374  *
2375  * pci.0.1.0
2376  *          .device="ahci"
2377  *          .port
2378  *               .0
2379  *                 .type="hd"
2380  *                 .path="/path/to/image"
2381  */
2382 static int
pci_ahci_legacy_config_port(nvlist_t * nvl,int port,const char * type,const char * opts)2383 pci_ahci_legacy_config_port(nvlist_t *nvl, int port, const char *type,
2384     const char *opts)
2385 {
2386 	char node_name[sizeof("XX")];
2387 	nvlist_t *port_nvl;
2388 
2389 	snprintf(node_name, sizeof(node_name), "%d", port);
2390 	port_nvl = create_relative_config_node(nvl, node_name);
2391 	set_config_value_node(port_nvl, "type", type);
2392 	return (blockif_legacy_config(port_nvl, opts));
2393 }
2394 
2395 static int
pci_ahci_legacy_config(nvlist_t * nvl,const char * opts)2396 pci_ahci_legacy_config(nvlist_t *nvl, const char *opts)
2397 {
2398 	nvlist_t *ports_nvl;
2399 	const char *type;
2400 	char *next, *next2, *str, *tofree;
2401 	int p, ret;
2402 
2403 	if (opts == NULL)
2404 		return (0);
2405 
2406 	ports_nvl = create_relative_config_node(nvl, "port");
2407 	ret = 1;
2408 	tofree = str = strdup(opts);
2409 	for (p = 0; p < MAX_PORTS && str != NULL; p++, str = next) {
2410 		/* Identify and cut off type of present port. */
2411 		if (strncmp(str, "hd:", 3) == 0) {
2412 			type = "hd";
2413 			str += 3;
2414 		} else if (strncmp(str, "cd:", 3) == 0) {
2415 			type = "cd";
2416 			str += 3;
2417 		} else
2418 			type = NULL;
2419 
2420 		/* Find and cut off the next port options. */
2421 		next = strstr(str, ",hd:");
2422 		next2 = strstr(str, ",cd:");
2423 		if (next == NULL || (next2 != NULL && next2 < next))
2424 			next = next2;
2425 		if (next != NULL) {
2426 			next[0] = 0;
2427 			next++;
2428 		}
2429 
2430 		if (str[0] == 0)
2431 			continue;
2432 
2433 		if (type == NULL) {
2434 			EPRINTLN("Missing or invalid type for port %d: \"%s\"",
2435 			    p, str);
2436 			goto out;
2437 		}
2438 
2439 		if (pci_ahci_legacy_config_port(ports_nvl, p, type, str) != 0)
2440 			goto out;
2441 	}
2442 	ret = 0;
2443 out:
2444 	free(tofree);
2445 	return (ret);
2446 }
2447 
2448 static int
pci_ahci_cd_legacy_config(nvlist_t * nvl,const char * opts)2449 pci_ahci_cd_legacy_config(nvlist_t *nvl, const char *opts)
2450 {
2451 	nvlist_t *ports_nvl;
2452 
2453 	ports_nvl = create_relative_config_node(nvl, "port");
2454 	return (pci_ahci_legacy_config_port(ports_nvl, 0, "cd", opts));
2455 }
2456 
2457 static int
pci_ahci_hd_legacy_config(nvlist_t * nvl,const char * opts)2458 pci_ahci_hd_legacy_config(nvlist_t *nvl, const char *opts)
2459 {
2460 	nvlist_t *ports_nvl;
2461 
2462 	ports_nvl = create_relative_config_node(nvl, "port");
2463 	return (pci_ahci_legacy_config_port(ports_nvl, 0, "hd", opts));
2464 }
2465 
2466 static int
pci_ahci_init(struct pci_devinst * pi,nvlist_t * nvl)2467 pci_ahci_init(struct pci_devinst *pi, nvlist_t *nvl)
2468 {
2469 	char bident[sizeof("XXX:XXX:XXX")];
2470 	char node_name[sizeof("XX")];
2471 	struct blockif_ctxt *bctxt;
2472 	struct pci_ahci_softc *sc;
2473 	int atapi, ret, slots, p;
2474 	MD5_CTX mdctx;
2475 	u_char digest[16];
2476 	const char *path, *type, *value;
2477 	nvlist_t *ports_nvl, *port_nvl;
2478 
2479 	ret = 0;
2480 
2481 #ifdef AHCI_DEBUG
2482 	dbg = fopen("/tmp/log", "w+");
2483 #endif
2484 
2485 	sc = calloc(1, sizeof(struct pci_ahci_softc));
2486 	pi->pi_arg = sc;
2487 	sc->asc_pi = pi;
2488 	pthread_mutex_init(&sc->mtx, NULL);
2489 	sc->ports = 0;
2490 	sc->pi = 0;
2491 	slots = 32;
2492 
2493 	ports_nvl = find_relative_config_node(nvl, "port");
2494 	for (p = 0; ports_nvl != NULL && p < MAX_PORTS; p++) {
2495 		struct ata_params *ata_ident = &sc->port[p].ata_ident;
2496 		char ident[AHCI_PORT_IDENT];
2497 
2498 		snprintf(node_name, sizeof(node_name), "%d", p);
2499 		port_nvl = find_relative_config_node(ports_nvl, node_name);
2500 		if (port_nvl == NULL)
2501 			continue;
2502 
2503 		type = get_config_value_node(port_nvl, "type");
2504 		if (type == NULL)
2505 			continue;
2506 
2507 		if (strcmp(type, "hd") == 0)
2508 			atapi = 0;
2509 		else
2510 			atapi = 1;
2511 
2512 		/*
2513 		 * Attempt to open the backing image. Use the PCI slot/func
2514 		 * and the port number for the identifier string.
2515 		 */
2516 		snprintf(bident, sizeof(bident), "%u:%u:%u", pi->pi_slot,
2517 		    pi->pi_func, p);
2518 
2519 		bctxt = blockif_open(port_nvl, bident);
2520 		if (bctxt == NULL) {
2521 			sc->ports = p;
2522 			ret = 1;
2523 			goto open_fail;
2524 		}
2525 
2526 		ret = blockif_add_boot_device(pi, bctxt);
2527 		if (ret) {
2528 			sc->ports = p;
2529 			goto open_fail;
2530 		}
2531 
2532 		sc->port[p].bctx = bctxt;
2533 		sc->port[p].pr_sc = sc;
2534 		sc->port[p].port = p;
2535 		sc->port[p].atapi = atapi;
2536 
2537 		/*
2538 		 * Create an identifier for the backing file.
2539 		 * Use parts of the md5 sum of the filename
2540 		 */
2541 		path = get_config_value_node(port_nvl, "path");
2542 		MD5Init(&mdctx);
2543 		MD5Update(&mdctx, path, strlen(path));
2544 		MD5Final(digest, &mdctx);
2545 		snprintf(ident, AHCI_PORT_IDENT,
2546 			"BHYVE-%02X%02X-%02X%02X-%02X%02X",
2547 			digest[0], digest[1], digest[2], digest[3], digest[4],
2548 			digest[5]);
2549 
2550 		memset(ata_ident, 0, sizeof(struct ata_params));
2551 		ata_string((uint8_t*)&ata_ident->serial, ident, 20);
2552 		ata_string((uint8_t*)&ata_ident->revision, "001", 8);
2553 		if (atapi)
2554 			ata_string((uint8_t*)&ata_ident->model, "BHYVE SATA DVD ROM", 40);
2555 		else
2556 			ata_string((uint8_t*)&ata_ident->model, "BHYVE SATA DISK", 40);
2557 		value = get_config_value_node(port_nvl, "nmrr");
2558 		if (value != NULL)
2559 			ata_ident->media_rotation_rate = atoi(value);
2560 		value = get_config_value_node(port_nvl, "ser");
2561 		if (value != NULL)
2562 			ata_string((uint8_t*)(&ata_ident->serial), value, 20);
2563 		value = get_config_value_node(port_nvl, "rev");
2564 		if (value != NULL)
2565 			ata_string((uint8_t*)(&ata_ident->revision), value, 8);
2566 		value = get_config_value_node(port_nvl, "model");
2567 		if (value != NULL)
2568 			ata_string((uint8_t*)(&ata_ident->model), value, 40);
2569 		ata_identify_init(&sc->port[p], atapi);
2570 
2571 		/*
2572 		 * Allocate blockif request structures and add them
2573 		 * to the free list
2574 		 */
2575 		pci_ahci_ioreq_init(&sc->port[p]);
2576 
2577 		sc->pi |= (1 << p);
2578 		if (sc->port[p].ioqsz < slots)
2579 			slots = sc->port[p].ioqsz;
2580 	}
2581 	sc->ports = p;
2582 
2583 	/* Intel ICH8 AHCI */
2584 	--slots;
2585 	if (sc->ports < DEF_PORTS)
2586 		sc->ports = DEF_PORTS;
2587 	sc->cap = AHCI_CAP_64BIT | AHCI_CAP_SNCQ | AHCI_CAP_SSNTF |
2588 	    AHCI_CAP_SMPS | AHCI_CAP_SSS | AHCI_CAP_SALP |
2589 	    AHCI_CAP_SAL | AHCI_CAP_SCLO | (0x3 << AHCI_CAP_ISS_SHIFT)|
2590 	    AHCI_CAP_PMD | AHCI_CAP_SSC | AHCI_CAP_PSC |
2591 	    (slots << AHCI_CAP_NCS_SHIFT) | AHCI_CAP_SXS | (sc->ports - 1);
2592 
2593 	sc->vs = 0x10300;
2594 	sc->cap2 = AHCI_CAP2_APST;
2595 	ahci_reset(sc);
2596 
2597 	pci_set_cfgdata16(pi, PCIR_DEVICE, 0x2821);
2598 	pci_set_cfgdata16(pi, PCIR_VENDOR, 0x8086);
2599 	pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE);
2600 	pci_set_cfgdata8(pi, PCIR_SUBCLASS, PCIS_STORAGE_SATA);
2601 	pci_set_cfgdata8(pi, PCIR_PROGIF, PCIP_STORAGE_SATA_AHCI_1_0);
2602 	p = MIN(sc->ports, 16);
2603 	p = flsl(p) - ((p & (p - 1)) ? 0 : 1);
2604 	pci_emul_add_msicap(pi, 1 << p);
2605 	pci_emul_alloc_bar(pi, 5, PCIBAR_MEM32,
2606 	    AHCI_OFFSET + sc->ports * AHCI_STEP);
2607 
2608 	pci_lintr_request(pi);
2609 
2610 open_fail:
2611 	if (ret) {
2612 		for (p = 0; p < sc->ports; p++) {
2613 			if (sc->port[p].bctx != NULL)
2614 				blockif_close(sc->port[p].bctx);
2615 		}
2616 		free(sc);
2617 	}
2618 
2619 	return (ret);
2620 }
2621 
2622 #ifdef BHYVE_SNAPSHOT
2623 static int
pci_ahci_snapshot(struct vm_snapshot_meta * meta)2624 pci_ahci_snapshot(struct vm_snapshot_meta *meta)
2625 {
2626 	int i, ret;
2627 	void *bctx;
2628 	struct pci_devinst *pi;
2629 	struct pci_ahci_softc *sc;
2630 	struct ahci_port *port;
2631 
2632 	pi = meta->dev_data;
2633 	sc = pi->pi_arg;
2634 
2635 	/* TODO: add mtx lock/unlock */
2636 
2637 	SNAPSHOT_VAR_OR_LEAVE(sc->ports, meta, ret, done);
2638 	SNAPSHOT_VAR_OR_LEAVE(sc->cap, meta, ret, done);
2639 	SNAPSHOT_VAR_OR_LEAVE(sc->ghc, meta, ret, done);
2640 	SNAPSHOT_VAR_OR_LEAVE(sc->is, meta, ret, done);
2641 	SNAPSHOT_VAR_OR_LEAVE(sc->pi, meta, ret, done);
2642 	SNAPSHOT_VAR_OR_LEAVE(sc->vs, meta, ret, done);
2643 	SNAPSHOT_VAR_OR_LEAVE(sc->ccc_ctl, meta, ret, done);
2644 	SNAPSHOT_VAR_OR_LEAVE(sc->ccc_pts, meta, ret, done);
2645 	SNAPSHOT_VAR_OR_LEAVE(sc->em_loc, meta, ret, done);
2646 	SNAPSHOT_VAR_OR_LEAVE(sc->em_ctl, meta, ret, done);
2647 	SNAPSHOT_VAR_OR_LEAVE(sc->cap2, meta, ret, done);
2648 	SNAPSHOT_VAR_OR_LEAVE(sc->bohc, meta, ret, done);
2649 	SNAPSHOT_VAR_OR_LEAVE(sc->lintr, meta, ret, done);
2650 
2651 	for (i = 0; i < MAX_PORTS; i++) {
2652 		port = &sc->port[i];
2653 
2654 		if (meta->op == VM_SNAPSHOT_SAVE)
2655 			bctx = port->bctx;
2656 
2657 		SNAPSHOT_VAR_OR_LEAVE(bctx, meta, ret, done);
2658 		SNAPSHOT_VAR_OR_LEAVE(port->port, meta, ret, done);
2659 
2660 		/* Mostly for restore; save is ensured by the lines above. */
2661 		if (((bctx == NULL) && (port->bctx != NULL)) ||
2662 		    ((bctx != NULL) && (port->bctx == NULL))) {
2663 			EPRINTLN("%s: ports not matching", __func__);
2664 			ret = EINVAL;
2665 			goto done;
2666 		}
2667 
2668 		if (port->bctx == NULL)
2669 			continue;
2670 
2671 		if (port->port != i) {
2672 			EPRINTLN("%s: ports not matching: "
2673 			    "actual: %d expected: %d", __func__, port->port, i);
2674 			ret = EINVAL;
2675 			goto done;
2676 		}
2677 
2678 		SNAPSHOT_GUEST2HOST_ADDR_OR_LEAVE(pi->pi_vmctx, port->cmd_lst,
2679 			AHCI_CL_SIZE * AHCI_MAX_SLOTS, false, meta, ret, done);
2680 		SNAPSHOT_GUEST2HOST_ADDR_OR_LEAVE(pi->pi_vmctx, port->rfis, 256,
2681 		    false, meta, ret, done);
2682 
2683 		SNAPSHOT_VAR_OR_LEAVE(port->ata_ident, meta, ret, done);
2684 		SNAPSHOT_VAR_OR_LEAVE(port->atapi, meta, ret, done);
2685 		SNAPSHOT_VAR_OR_LEAVE(port->reset, meta, ret, done);
2686 		SNAPSHOT_VAR_OR_LEAVE(port->waitforclear, meta, ret, done);
2687 		SNAPSHOT_VAR_OR_LEAVE(port->mult_sectors, meta, ret, done);
2688 		SNAPSHOT_VAR_OR_LEAVE(port->xfermode, meta, ret, done);
2689 		SNAPSHOT_VAR_OR_LEAVE(port->err_cfis, meta, ret, done);
2690 		SNAPSHOT_VAR_OR_LEAVE(port->sense_key, meta, ret, done);
2691 		SNAPSHOT_VAR_OR_LEAVE(port->asc, meta, ret, done);
2692 		SNAPSHOT_VAR_OR_LEAVE(port->ccs, meta, ret, done);
2693 		SNAPSHOT_VAR_OR_LEAVE(port->pending, meta, ret, done);
2694 
2695 		SNAPSHOT_VAR_OR_LEAVE(port->clb, meta, ret, done);
2696 		SNAPSHOT_VAR_OR_LEAVE(port->clbu, meta, ret, done);
2697 		SNAPSHOT_VAR_OR_LEAVE(port->fb, meta, ret, done);
2698 		SNAPSHOT_VAR_OR_LEAVE(port->fbu, meta, ret, done);
2699 		SNAPSHOT_VAR_OR_LEAVE(port->ie, meta, ret, done);
2700 		SNAPSHOT_VAR_OR_LEAVE(port->cmd, meta, ret, done);
2701 		SNAPSHOT_VAR_OR_LEAVE(port->unused0, meta, ret, done);
2702 		SNAPSHOT_VAR_OR_LEAVE(port->tfd, meta, ret, done);
2703 		SNAPSHOT_VAR_OR_LEAVE(port->sig, meta, ret, done);
2704 		SNAPSHOT_VAR_OR_LEAVE(port->ssts, meta, ret, done);
2705 		SNAPSHOT_VAR_OR_LEAVE(port->sctl, meta, ret, done);
2706 		SNAPSHOT_VAR_OR_LEAVE(port->serr, meta, ret, done);
2707 		SNAPSHOT_VAR_OR_LEAVE(port->sact, meta, ret, done);
2708 		SNAPSHOT_VAR_OR_LEAVE(port->ci, meta, ret, done);
2709 		SNAPSHOT_VAR_OR_LEAVE(port->sntf, meta, ret, done);
2710 		SNAPSHOT_VAR_OR_LEAVE(port->fbs, meta, ret, done);
2711 		SNAPSHOT_VAR_OR_LEAVE(port->ioqsz, meta, ret, done);
2712 
2713 		assert(TAILQ_EMPTY(&port->iobhd));
2714 	}
2715 
2716 done:
2717 	return (ret);
2718 }
2719 
2720 static int
pci_ahci_pause(struct pci_devinst * pi)2721 pci_ahci_pause(struct pci_devinst *pi)
2722 {
2723 	struct pci_ahci_softc *sc;
2724 	struct blockif_ctxt *bctxt;
2725 	int i;
2726 
2727 	sc = pi->pi_arg;
2728 
2729 	for (i = 0; i < MAX_PORTS; i++) {
2730 		bctxt = sc->port[i].bctx;
2731 		if (bctxt == NULL)
2732 			continue;
2733 
2734 		blockif_pause(bctxt);
2735 	}
2736 
2737 	return (0);
2738 }
2739 
2740 static int
pci_ahci_resume(struct pci_devinst * pi)2741 pci_ahci_resume(struct pci_devinst *pi)
2742 {
2743 	struct pci_ahci_softc *sc;
2744 	struct blockif_ctxt *bctxt;
2745 	int i;
2746 
2747 	sc = pi->pi_arg;
2748 
2749 	for (i = 0; i < MAX_PORTS; i++) {
2750 		bctxt = sc->port[i].bctx;
2751 		if (bctxt == NULL)
2752 			continue;
2753 
2754 		blockif_resume(bctxt);
2755 	}
2756 
2757 	return (0);
2758 }
2759 #endif	/* BHYVE_SNAPSHOT */
2760 
2761 /*
2762  * Use separate emulation names to distinguish drive and atapi devices
2763  */
2764 static const struct pci_devemu pci_de_ahci = {
2765 	.pe_emu =	"ahci",
2766 	.pe_init =	pci_ahci_init,
2767 	.pe_legacy_config = pci_ahci_legacy_config,
2768 	.pe_barwrite =	pci_ahci_write,
2769 	.pe_barread =	pci_ahci_read,
2770 #ifdef BHYVE_SNAPSHOT
2771 	.pe_snapshot =	pci_ahci_snapshot,
2772 	.pe_pause =	pci_ahci_pause,
2773 	.pe_resume =	pci_ahci_resume,
2774 #endif
2775 };
2776 PCI_EMUL_SET(pci_de_ahci);
2777 
2778 static const struct pci_devemu pci_de_ahci_hd = {
2779 	.pe_emu =	"ahci-hd",
2780 	.pe_legacy_config = pci_ahci_hd_legacy_config,
2781 	.pe_alias =	"ahci",
2782 };
2783 PCI_EMUL_SET(pci_de_ahci_hd);
2784 
2785 static const struct pci_devemu pci_de_ahci_cd = {
2786 	.pe_emu =	"ahci-cd",
2787 	.pe_legacy_config = pci_ahci_cd_legacy_config,
2788 	.pe_alias =	"ahci",
2789 };
2790 PCI_EMUL_SET(pci_de_ahci_cd);
2791