xref: /openbmc/qemu/hw/ide/core.c (revision b15e402f)
1 /*
2  * QEMU IDE disk and CD/DVD-ROM Emulator
3  *
4  * Copyright (c) 2003 Fabrice Bellard
5  * Copyright (c) 2006 Openedhand Ltd.
6  *
7  * Permission is hereby granted, free of charge, to any person obtaining a copy
8  * of this software and associated documentation files (the "Software"), to deal
9  * in the Software without restriction, including without limitation the rights
10  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
11  * copies of the Software, and to permit persons to whom the Software is
12  * furnished to do so, subject to the following conditions:
13  *
14  * The above copyright notice and this permission notice shall be included in
15  * all copies or substantial portions of the Software.
16  *
17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
23  * THE SOFTWARE.
24  */
25 
26 #include "qemu/osdep.h"
27 #include "hw/isa/isa.h"
28 #include "migration/vmstate.h"
29 #include "qemu/error-report.h"
30 #include "qemu/main-loop.h"
31 #include "qemu/timer.h"
32 #include "sysemu/sysemu.h"
33 #include "sysemu/blockdev.h"
34 #include "sysemu/dma.h"
35 #include "hw/block/block.h"
36 #include "sysemu/block-backend.h"
37 #include "qapi/error.h"
38 #include "qemu/cutils.h"
39 #include "sysemu/replay.h"
40 #include "sysemu/runstate.h"
41 #include "hw/ide/internal.h"
42 #include "trace.h"
43 
44 /* These values were based on a Seagate ST3500418AS but have been modified
45    to make more sense in QEMU */
46 static const int smart_attributes[][12] = {
47     /* id,  flags, hflags, val, wrst, raw (6 bytes), threshold */
48     /* raw read error rate*/
49     { 0x01, 0x03, 0x00, 0x64, 0x64, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x06},
50     /* spin up */
51     { 0x03, 0x03, 0x00, 0x64, 0x64, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
52     /* start stop count */
53     { 0x04, 0x02, 0x00, 0x64, 0x64, 0x64, 0x00, 0x00, 0x00, 0x00, 0x00, 0x14},
54     /* remapped sectors */
55     { 0x05, 0x03, 0x00, 0x64, 0x64, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x24},
56     /* power on hours */
57     { 0x09, 0x03, 0x00, 0x64, 0x64, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
58     /* power cycle count */
59     { 0x0c, 0x03, 0x00, 0x64, 0x64, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
60     /* airflow-temperature-celsius */
61     { 190,  0x03, 0x00, 0x45, 0x45, 0x1f, 0x00, 0x1f, 0x1f, 0x00, 0x00, 0x32},
62 };
63 
64 const char *IDE_DMA_CMD_lookup[IDE_DMA__COUNT] = {
65     [IDE_DMA_READ] = "DMA READ",
66     [IDE_DMA_WRITE] = "DMA WRITE",
67     [IDE_DMA_TRIM] = "DMA TRIM",
68     [IDE_DMA_ATAPI] = "DMA ATAPI"
69 };
70 
71 static const char *IDE_DMA_CMD_str(enum ide_dma_cmd enval)
72 {
73     if ((unsigned)enval < IDE_DMA__COUNT) {
74         return IDE_DMA_CMD_lookup[enval];
75     }
76     return "DMA UNKNOWN CMD";
77 }
78 
79 static void ide_dummy_transfer_stop(IDEState *s);
80 
81 static void padstr(char *str, const char *src, int len)
82 {
83     int i, v;
84     for(i = 0; i < len; i++) {
85         if (*src)
86             v = *src++;
87         else
88             v = ' ';
89         str[i^1] = v;
90     }
91 }
92 
93 static void put_le16(uint16_t *p, unsigned int v)
94 {
95     *p = cpu_to_le16(v);
96 }
97 
98 static void ide_identify_size(IDEState *s)
99 {
100     uint16_t *p = (uint16_t *)s->identify_data;
101     put_le16(p + 60, s->nb_sectors);
102     put_le16(p + 61, s->nb_sectors >> 16);
103     put_le16(p + 100, s->nb_sectors);
104     put_le16(p + 101, s->nb_sectors >> 16);
105     put_le16(p + 102, s->nb_sectors >> 32);
106     put_le16(p + 103, s->nb_sectors >> 48);
107 }
108 
109 static void ide_identify(IDEState *s)
110 {
111     uint16_t *p;
112     unsigned int oldsize;
113     IDEDevice *dev = s->unit ? s->bus->slave : s->bus->master;
114 
115     p = (uint16_t *)s->identify_data;
116     if (s->identify_set) {
117         goto fill_buffer;
118     }
119     memset(p, 0, sizeof(s->identify_data));
120 
121     put_le16(p + 0, 0x0040);
122     put_le16(p + 1, s->cylinders);
123     put_le16(p + 3, s->heads);
124     put_le16(p + 4, 512 * s->sectors); /* XXX: retired, remove ? */
125     put_le16(p + 5, 512); /* XXX: retired, remove ? */
126     put_le16(p + 6, s->sectors);
127     padstr((char *)(p + 10), s->drive_serial_str, 20); /* serial number */
128     put_le16(p + 20, 3); /* XXX: retired, remove ? */
129     put_le16(p + 21, 512); /* cache size in sectors */
130     put_le16(p + 22, 4); /* ecc bytes */
131     padstr((char *)(p + 23), s->version, 8); /* firmware version */
132     padstr((char *)(p + 27), s->drive_model_str, 40); /* model */
133 #if MAX_MULT_SECTORS > 1
134     put_le16(p + 47, 0x8000 | MAX_MULT_SECTORS);
135 #endif
136     put_le16(p + 48, 1); /* dword I/O */
137     put_le16(p + 49, (1 << 11) | (1 << 9) | (1 << 8)); /* DMA and LBA supported */
138     put_le16(p + 51, 0x200); /* PIO transfer cycle */
139     put_le16(p + 52, 0x200); /* DMA transfer cycle */
140     put_le16(p + 53, 1 | (1 << 1) | (1 << 2)); /* words 54-58,64-70,88 are valid */
141     put_le16(p + 54, s->cylinders);
142     put_le16(p + 55, s->heads);
143     put_le16(p + 56, s->sectors);
144     oldsize = s->cylinders * s->heads * s->sectors;
145     put_le16(p + 57, oldsize);
146     put_le16(p + 58, oldsize >> 16);
147     if (s->mult_sectors)
148         put_le16(p + 59, 0x100 | s->mult_sectors);
149     /* *(p + 60) := nb_sectors       -- see ide_identify_size */
150     /* *(p + 61) := nb_sectors >> 16 -- see ide_identify_size */
151     put_le16(p + 62, 0x07); /* single word dma0-2 supported */
152     put_le16(p + 63, 0x07); /* mdma0-2 supported */
153     put_le16(p + 64, 0x03); /* pio3-4 supported */
154     put_le16(p + 65, 120);
155     put_le16(p + 66, 120);
156     put_le16(p + 67, 120);
157     put_le16(p + 68, 120);
158     if (dev && dev->conf.discard_granularity) {
159         put_le16(p + 69, (1 << 14)); /* determinate TRIM behavior */
160     }
161 
162     if (s->ncq_queues) {
163         put_le16(p + 75, s->ncq_queues - 1);
164         /* NCQ supported */
165         put_le16(p + 76, (1 << 8));
166     }
167 
168     put_le16(p + 80, 0xf0); /* ata3 -> ata6 supported */
169     put_le16(p + 81, 0x16); /* conforms to ata5 */
170     /* 14=NOP supported, 5=WCACHE supported, 0=SMART supported */
171     put_le16(p + 82, (1 << 14) | (1 << 5) | 1);
172     /* 13=flush_cache_ext,12=flush_cache,10=lba48 */
173     put_le16(p + 83, (1 << 14) | (1 << 13) | (1 <<12) | (1 << 10));
174     /* 14=set to 1, 8=has WWN, 1=SMART self test, 0=SMART error logging */
175     if (s->wwn) {
176         put_le16(p + 84, (1 << 14) | (1 << 8) | 0);
177     } else {
178         put_le16(p + 84, (1 << 14) | 0);
179     }
180     /* 14 = NOP supported, 5=WCACHE enabled, 0=SMART feature set enabled */
181     if (blk_enable_write_cache(s->blk)) {
182         put_le16(p + 85, (1 << 14) | (1 << 5) | 1);
183     } else {
184         put_le16(p + 85, (1 << 14) | 1);
185     }
186     /* 13=flush_cache_ext,12=flush_cache,10=lba48 */
187     put_le16(p + 86, (1 << 13) | (1 <<12) | (1 << 10));
188     /* 14=set to 1, 8=has WWN, 1=SMART self test, 0=SMART error logging */
189     if (s->wwn) {
190         put_le16(p + 87, (1 << 14) | (1 << 8) | 0);
191     } else {
192         put_le16(p + 87, (1 << 14) | 0);
193     }
194     put_le16(p + 88, 0x3f | (1 << 13)); /* udma5 set and supported */
195     put_le16(p + 93, 1 | (1 << 14) | 0x2000);
196     /* *(p + 100) := nb_sectors       -- see ide_identify_size */
197     /* *(p + 101) := nb_sectors >> 16 -- see ide_identify_size */
198     /* *(p + 102) := nb_sectors >> 32 -- see ide_identify_size */
199     /* *(p + 103) := nb_sectors >> 48 -- see ide_identify_size */
200 
201     if (dev && dev->conf.physical_block_size)
202         put_le16(p + 106, 0x6000 | get_physical_block_exp(&dev->conf));
203     if (s->wwn) {
204         /* LE 16-bit words 111-108 contain 64-bit World Wide Name */
205         put_le16(p + 108, s->wwn >> 48);
206         put_le16(p + 109, s->wwn >> 32);
207         put_le16(p + 110, s->wwn >> 16);
208         put_le16(p + 111, s->wwn);
209     }
210     if (dev && dev->conf.discard_granularity) {
211         put_le16(p + 169, 1); /* TRIM support */
212     }
213     if (dev) {
214         put_le16(p + 217, dev->rotation_rate); /* Nominal media rotation rate */
215     }
216 
217     ide_identify_size(s);
218     s->identify_set = 1;
219 
220 fill_buffer:
221     memcpy(s->io_buffer, p, sizeof(s->identify_data));
222 }
223 
224 static void ide_atapi_identify(IDEState *s)
225 {
226     uint16_t *p;
227 
228     p = (uint16_t *)s->identify_data;
229     if (s->identify_set) {
230         goto fill_buffer;
231     }
232     memset(p, 0, sizeof(s->identify_data));
233 
234     /* Removable CDROM, 50us response, 12 byte packets */
235     put_le16(p + 0, (2 << 14) | (5 << 8) | (1 << 7) | (2 << 5) | (0 << 0));
236     padstr((char *)(p + 10), s->drive_serial_str, 20); /* serial number */
237     put_le16(p + 20, 3); /* buffer type */
238     put_le16(p + 21, 512); /* cache size in sectors */
239     put_le16(p + 22, 4); /* ecc bytes */
240     padstr((char *)(p + 23), s->version, 8); /* firmware version */
241     padstr((char *)(p + 27), s->drive_model_str, 40); /* model */
242     put_le16(p + 48, 1); /* dword I/O (XXX: should not be set on CDROM) */
243 #ifdef USE_DMA_CDROM
244     put_le16(p + 49, 1 << 9 | 1 << 8); /* DMA and LBA supported */
245     put_le16(p + 53, 7); /* words 64-70, 54-58, 88 valid */
246     put_le16(p + 62, 7);  /* single word dma0-2 supported */
247     put_le16(p + 63, 7);  /* mdma0-2 supported */
248 #else
249     put_le16(p + 49, 1 << 9); /* LBA supported, no DMA */
250     put_le16(p + 53, 3); /* words 64-70, 54-58 valid */
251     put_le16(p + 63, 0x103); /* DMA modes XXX: may be incorrect */
252 #endif
253     put_le16(p + 64, 3); /* pio3-4 supported */
254     put_le16(p + 65, 0xb4); /* minimum DMA multiword tx cycle time */
255     put_le16(p + 66, 0xb4); /* recommended DMA multiword tx cycle time */
256     put_le16(p + 67, 0x12c); /* minimum PIO cycle time without flow control */
257     put_le16(p + 68, 0xb4); /* minimum PIO cycle time with IORDY flow control */
258 
259     put_le16(p + 71, 30); /* in ns */
260     put_le16(p + 72, 30); /* in ns */
261 
262     if (s->ncq_queues) {
263         put_le16(p + 75, s->ncq_queues - 1);
264         /* NCQ supported */
265         put_le16(p + 76, (1 << 8));
266     }
267 
268     put_le16(p + 80, 0x1e); /* support up to ATA/ATAPI-4 */
269     if (s->wwn) {
270         put_le16(p + 84, (1 << 8)); /* supports WWN for words 108-111 */
271         put_le16(p + 87, (1 << 8)); /* WWN enabled */
272     }
273 
274 #ifdef USE_DMA_CDROM
275     put_le16(p + 88, 0x3f | (1 << 13)); /* udma5 set and supported */
276 #endif
277 
278     if (s->wwn) {
279         /* LE 16-bit words 111-108 contain 64-bit World Wide Name */
280         put_le16(p + 108, s->wwn >> 48);
281         put_le16(p + 109, s->wwn >> 32);
282         put_le16(p + 110, s->wwn >> 16);
283         put_le16(p + 111, s->wwn);
284     }
285 
286     s->identify_set = 1;
287 
288 fill_buffer:
289     memcpy(s->io_buffer, p, sizeof(s->identify_data));
290 }
291 
292 static void ide_cfata_identify_size(IDEState *s)
293 {
294     uint16_t *p = (uint16_t *)s->identify_data;
295     put_le16(p + 7, s->nb_sectors >> 16);  /* Sectors per card */
296     put_le16(p + 8, s->nb_sectors);        /* Sectors per card */
297     put_le16(p + 60, s->nb_sectors);       /* Total LBA sectors */
298     put_le16(p + 61, s->nb_sectors >> 16); /* Total LBA sectors */
299 }
300 
301 static void ide_cfata_identify(IDEState *s)
302 {
303     uint16_t *p;
304     uint32_t cur_sec;
305 
306     p = (uint16_t *)s->identify_data;
307     if (s->identify_set) {
308         goto fill_buffer;
309     }
310     memset(p, 0, sizeof(s->identify_data));
311 
312     cur_sec = s->cylinders * s->heads * s->sectors;
313 
314     put_le16(p + 0, 0x848a);			/* CF Storage Card signature */
315     put_le16(p + 1, s->cylinders);		/* Default cylinders */
316     put_le16(p + 3, s->heads);			/* Default heads */
317     put_le16(p + 6, s->sectors);		/* Default sectors per track */
318     /* *(p + 7) := nb_sectors >> 16 -- see ide_cfata_identify_size */
319     /* *(p + 8) := nb_sectors       -- see ide_cfata_identify_size */
320     padstr((char *)(p + 10), s->drive_serial_str, 20); /* serial number */
321     put_le16(p + 22, 0x0004);			/* ECC bytes */
322     padstr((char *) (p + 23), s->version, 8);	/* Firmware Revision */
323     padstr((char *) (p + 27), s->drive_model_str, 40);/* Model number */
324 #if MAX_MULT_SECTORS > 1
325     put_le16(p + 47, 0x8000 | MAX_MULT_SECTORS);
326 #else
327     put_le16(p + 47, 0x0000);
328 #endif
329     put_le16(p + 49, 0x0f00);			/* Capabilities */
330     put_le16(p + 51, 0x0002);			/* PIO cycle timing mode */
331     put_le16(p + 52, 0x0001);			/* DMA cycle timing mode */
332     put_le16(p + 53, 0x0003);			/* Translation params valid */
333     put_le16(p + 54, s->cylinders);		/* Current cylinders */
334     put_le16(p + 55, s->heads);			/* Current heads */
335     put_le16(p + 56, s->sectors);		/* Current sectors */
336     put_le16(p + 57, cur_sec);			/* Current capacity */
337     put_le16(p + 58, cur_sec >> 16);		/* Current capacity */
338     if (s->mult_sectors)			/* Multiple sector setting */
339         put_le16(p + 59, 0x100 | s->mult_sectors);
340     /* *(p + 60) := nb_sectors       -- see ide_cfata_identify_size */
341     /* *(p + 61) := nb_sectors >> 16 -- see ide_cfata_identify_size */
342     put_le16(p + 63, 0x0203);			/* Multiword DMA capability */
343     put_le16(p + 64, 0x0001);			/* Flow Control PIO support */
344     put_le16(p + 65, 0x0096);			/* Min. Multiword DMA cycle */
345     put_le16(p + 66, 0x0096);			/* Rec. Multiword DMA cycle */
346     put_le16(p + 68, 0x00b4);			/* Min. PIO cycle time */
347     put_le16(p + 82, 0x400c);			/* Command Set supported */
348     put_le16(p + 83, 0x7068);			/* Command Set supported */
349     put_le16(p + 84, 0x4000);			/* Features supported */
350     put_le16(p + 85, 0x000c);			/* Command Set enabled */
351     put_le16(p + 86, 0x7044);			/* Command Set enabled */
352     put_le16(p + 87, 0x4000);			/* Features enabled */
353     put_le16(p + 91, 0x4060);			/* Current APM level */
354     put_le16(p + 129, 0x0002);			/* Current features option */
355     put_le16(p + 130, 0x0005);			/* Reassigned sectors */
356     put_le16(p + 131, 0x0001);			/* Initial power mode */
357     put_le16(p + 132, 0x0000);			/* User signature */
358     put_le16(p + 160, 0x8100);			/* Power requirement */
359     put_le16(p + 161, 0x8001);			/* CF command set */
360 
361     ide_cfata_identify_size(s);
362     s->identify_set = 1;
363 
364 fill_buffer:
365     memcpy(s->io_buffer, p, sizeof(s->identify_data));
366 }
367 
368 static void ide_set_signature(IDEState *s)
369 {
370     s->select &= 0xf0; /* clear head */
371     /* put signature */
372     s->nsector = 1;
373     s->sector = 1;
374     if (s->drive_kind == IDE_CD) {
375         s->lcyl = 0x14;
376         s->hcyl = 0xeb;
377     } else if (s->blk) {
378         s->lcyl = 0;
379         s->hcyl = 0;
380     } else {
381         s->lcyl = 0xff;
382         s->hcyl = 0xff;
383     }
384 }
385 
386 static bool ide_sect_range_ok(IDEState *s,
387                               uint64_t sector, uint64_t nb_sectors)
388 {
389     uint64_t total_sectors;
390 
391     blk_get_geometry(s->blk, &total_sectors);
392     if (sector > total_sectors || nb_sectors > total_sectors - sector) {
393         return false;
394     }
395     return true;
396 }
397 
398 typedef struct TrimAIOCB {
399     BlockAIOCB common;
400     IDEState *s;
401     QEMUBH *bh;
402     int ret;
403     QEMUIOVector *qiov;
404     BlockAIOCB *aiocb;
405     int i, j;
406 } TrimAIOCB;
407 
408 static void trim_aio_cancel(BlockAIOCB *acb)
409 {
410     TrimAIOCB *iocb = container_of(acb, TrimAIOCB, common);
411 
412     /* Exit the loop so ide_issue_trim_cb will not continue  */
413     iocb->j = iocb->qiov->niov - 1;
414     iocb->i = (iocb->qiov->iov[iocb->j].iov_len / 8) - 1;
415 
416     iocb->ret = -ECANCELED;
417 
418     if (iocb->aiocb) {
419         blk_aio_cancel_async(iocb->aiocb);
420         iocb->aiocb = NULL;
421     }
422 }
423 
424 static const AIOCBInfo trim_aiocb_info = {
425     .aiocb_size         = sizeof(TrimAIOCB),
426     .cancel_async       = trim_aio_cancel,
427 };
428 
429 static void ide_trim_bh_cb(void *opaque)
430 {
431     TrimAIOCB *iocb = opaque;
432 
433     iocb->common.cb(iocb->common.opaque, iocb->ret);
434 
435     qemu_bh_delete(iocb->bh);
436     iocb->bh = NULL;
437     qemu_aio_unref(iocb);
438 }
439 
440 static void ide_issue_trim_cb(void *opaque, int ret)
441 {
442     TrimAIOCB *iocb = opaque;
443     IDEState *s = iocb->s;
444 
445     if (iocb->i >= 0) {
446         if (ret >= 0) {
447             block_acct_done(blk_get_stats(s->blk), &s->acct);
448         } else {
449             block_acct_failed(blk_get_stats(s->blk), &s->acct);
450         }
451     }
452 
453     if (ret >= 0) {
454         while (iocb->j < iocb->qiov->niov) {
455             int j = iocb->j;
456             while (++iocb->i < iocb->qiov->iov[j].iov_len / 8) {
457                 int i = iocb->i;
458                 uint64_t *buffer = iocb->qiov->iov[j].iov_base;
459 
460                 /* 6-byte LBA + 2-byte range per entry */
461                 uint64_t entry = le64_to_cpu(buffer[i]);
462                 uint64_t sector = entry & 0x0000ffffffffffffULL;
463                 uint16_t count = entry >> 48;
464 
465                 if (count == 0) {
466                     continue;
467                 }
468 
469                 if (!ide_sect_range_ok(s, sector, count)) {
470                     block_acct_invalid(blk_get_stats(s->blk), BLOCK_ACCT_UNMAP);
471                     iocb->ret = -EINVAL;
472                     goto done;
473                 }
474 
475                 block_acct_start(blk_get_stats(s->blk), &s->acct,
476                                  count << BDRV_SECTOR_BITS, BLOCK_ACCT_UNMAP);
477 
478                 /* Got an entry! Submit and exit.  */
479                 iocb->aiocb = blk_aio_pdiscard(s->blk,
480                                                sector << BDRV_SECTOR_BITS,
481                                                count << BDRV_SECTOR_BITS,
482                                                ide_issue_trim_cb, opaque);
483                 return;
484             }
485 
486             iocb->j++;
487             iocb->i = -1;
488         }
489     } else {
490         iocb->ret = ret;
491     }
492 
493 done:
494     iocb->aiocb = NULL;
495     if (iocb->bh) {
496         replay_bh_schedule_event(iocb->bh);
497     }
498 }
499 
500 BlockAIOCB *ide_issue_trim(
501         int64_t offset, QEMUIOVector *qiov,
502         BlockCompletionFunc *cb, void *cb_opaque, void *opaque)
503 {
504     IDEState *s = opaque;
505     TrimAIOCB *iocb;
506 
507     iocb = blk_aio_get(&trim_aiocb_info, s->blk, cb, cb_opaque);
508     iocb->s = s;
509     iocb->bh = qemu_bh_new(ide_trim_bh_cb, iocb);
510     iocb->ret = 0;
511     iocb->qiov = qiov;
512     iocb->i = -1;
513     iocb->j = 0;
514     ide_issue_trim_cb(iocb, 0);
515     return &iocb->common;
516 }
517 
518 void ide_abort_command(IDEState *s)
519 {
520     ide_transfer_stop(s);
521     s->status = READY_STAT | ERR_STAT;
522     s->error = ABRT_ERR;
523 }
524 
525 static void ide_set_retry(IDEState *s)
526 {
527     s->bus->retry_unit = s->unit;
528     s->bus->retry_sector_num = ide_get_sector(s);
529     s->bus->retry_nsector = s->nsector;
530 }
531 
532 static void ide_clear_retry(IDEState *s)
533 {
534     s->bus->retry_unit = -1;
535     s->bus->retry_sector_num = 0;
536     s->bus->retry_nsector = 0;
537 }
538 
539 /* prepare data transfer and tell what to do after */
540 bool ide_transfer_start_norecurse(IDEState *s, uint8_t *buf, int size,
541                                   EndTransferFunc *end_transfer_func)
542 {
543     s->data_ptr = buf;
544     s->data_end = buf + size;
545     ide_set_retry(s);
546     if (!(s->status & ERR_STAT)) {
547         s->status |= DRQ_STAT;
548     }
549     if (!s->bus->dma->ops->pio_transfer) {
550         s->end_transfer_func = end_transfer_func;
551         return false;
552     }
553     s->bus->dma->ops->pio_transfer(s->bus->dma);
554     return true;
555 }
556 
557 void ide_transfer_start(IDEState *s, uint8_t *buf, int size,
558                         EndTransferFunc *end_transfer_func)
559 {
560     if (ide_transfer_start_norecurse(s, buf, size, end_transfer_func)) {
561         end_transfer_func(s);
562     }
563 }
564 
565 static void ide_cmd_done(IDEState *s)
566 {
567     if (s->bus->dma->ops->cmd_done) {
568         s->bus->dma->ops->cmd_done(s->bus->dma);
569     }
570 }
571 
572 static void ide_transfer_halt(IDEState *s)
573 {
574     s->end_transfer_func = ide_transfer_stop;
575     s->data_ptr = s->io_buffer;
576     s->data_end = s->io_buffer;
577     s->status &= ~DRQ_STAT;
578 }
579 
580 void ide_transfer_stop(IDEState *s)
581 {
582     ide_transfer_halt(s);
583     ide_cmd_done(s);
584 }
585 
586 int64_t ide_get_sector(IDEState *s)
587 {
588     int64_t sector_num;
589     if (s->select & 0x40) {
590         /* lba */
591         if (!s->lba48) {
592             sector_num = ((s->select & 0x0f) << 24) | (s->hcyl << 16) |
593                 (s->lcyl << 8) | s->sector;
594         } else {
595             sector_num = ((int64_t)s->hob_hcyl << 40) |
596                 ((int64_t) s->hob_lcyl << 32) |
597                 ((int64_t) s->hob_sector << 24) |
598                 ((int64_t) s->hcyl << 16) |
599                 ((int64_t) s->lcyl << 8) | s->sector;
600         }
601     } else {
602         sector_num = ((s->hcyl << 8) | s->lcyl) * s->heads * s->sectors +
603             (s->select & 0x0f) * s->sectors + (s->sector - 1);
604     }
605     return sector_num;
606 }
607 
608 void ide_set_sector(IDEState *s, int64_t sector_num)
609 {
610     unsigned int cyl, r;
611     if (s->select & 0x40) {
612         if (!s->lba48) {
613             s->select = (s->select & 0xf0) | (sector_num >> 24);
614             s->hcyl = (sector_num >> 16);
615             s->lcyl = (sector_num >> 8);
616             s->sector = (sector_num);
617         } else {
618             s->sector = sector_num;
619             s->lcyl = sector_num >> 8;
620             s->hcyl = sector_num >> 16;
621             s->hob_sector = sector_num >> 24;
622             s->hob_lcyl = sector_num >> 32;
623             s->hob_hcyl = sector_num >> 40;
624         }
625     } else {
626         cyl = sector_num / (s->heads * s->sectors);
627         r = sector_num % (s->heads * s->sectors);
628         s->hcyl = cyl >> 8;
629         s->lcyl = cyl;
630         s->select = (s->select & 0xf0) | ((r / s->sectors) & 0x0f);
631         s->sector = (r % s->sectors) + 1;
632     }
633 }
634 
635 static void ide_rw_error(IDEState *s) {
636     ide_abort_command(s);
637     ide_set_irq(s->bus);
638 }
639 
640 static void ide_buffered_readv_cb(void *opaque, int ret)
641 {
642     IDEBufferedRequest *req = opaque;
643     if (!req->orphaned) {
644         if (!ret) {
645             assert(req->qiov.size == req->original_qiov->size);
646             qemu_iovec_from_buf(req->original_qiov, 0,
647                                 req->qiov.local_iov.iov_base,
648                                 req->original_qiov->size);
649         }
650         req->original_cb(req->original_opaque, ret);
651     }
652     QLIST_REMOVE(req, list);
653     qemu_vfree(qemu_iovec_buf(&req->qiov));
654     g_free(req);
655 }
656 
657 #define MAX_BUFFERED_REQS 16
658 
659 BlockAIOCB *ide_buffered_readv(IDEState *s, int64_t sector_num,
660                                QEMUIOVector *iov, int nb_sectors,
661                                BlockCompletionFunc *cb, void *opaque)
662 {
663     BlockAIOCB *aioreq;
664     IDEBufferedRequest *req;
665     int c = 0;
666 
667     QLIST_FOREACH(req, &s->buffered_requests, list) {
668         c++;
669     }
670     if (c > MAX_BUFFERED_REQS) {
671         return blk_abort_aio_request(s->blk, cb, opaque, -EIO);
672     }
673 
674     req = g_new0(IDEBufferedRequest, 1);
675     req->original_qiov = iov;
676     req->original_cb = cb;
677     req->original_opaque = opaque;
678     qemu_iovec_init_buf(&req->qiov, blk_blockalign(s->blk, iov->size),
679                         iov->size);
680 
681     aioreq = blk_aio_preadv(s->blk, sector_num << BDRV_SECTOR_BITS,
682                             &req->qiov, 0, ide_buffered_readv_cb, req);
683 
684     QLIST_INSERT_HEAD(&s->buffered_requests, req, list);
685     return aioreq;
686 }
687 
688 /**
689  * Cancel all pending DMA requests.
690  * Any buffered DMA requests are instantly canceled,
691  * but any pending unbuffered DMA requests must be waited on.
692  */
693 void ide_cancel_dma_sync(IDEState *s)
694 {
695     IDEBufferedRequest *req;
696 
697     /* First invoke the callbacks of all buffered requests
698      * and flag those requests as orphaned. Ideally there
699      * are no unbuffered (Scatter Gather DMA Requests or
700      * write requests) pending and we can avoid to drain. */
701     QLIST_FOREACH(req, &s->buffered_requests, list) {
702         if (!req->orphaned) {
703             trace_ide_cancel_dma_sync_buffered(req->original_cb, req);
704             req->original_cb(req->original_opaque, -ECANCELED);
705         }
706         req->orphaned = true;
707     }
708 
709     /*
710      * We can't cancel Scatter Gather DMA in the middle of the
711      * operation or a partial (not full) DMA transfer would reach
712      * the storage so we wait for completion instead (we behave
713      * like if the DMA was completed by the time the guest trying
714      * to cancel dma with bmdma_cmd_writeb with BM_CMD_START not
715      * set).
716      *
717      * In the future we'll be able to safely cancel the I/O if the
718      * whole DMA operation will be submitted to disk with a single
719      * aio operation with preadv/pwritev.
720      */
721     if (s->bus->dma->aiocb) {
722         trace_ide_cancel_dma_sync_remaining();
723         blk_drain(s->blk);
724         assert(s->bus->dma->aiocb == NULL);
725     }
726 }
727 
728 static void ide_sector_read(IDEState *s);
729 
730 static void ide_sector_read_cb(void *opaque, int ret)
731 {
732     IDEState *s = opaque;
733     int n;
734 
735     s->pio_aiocb = NULL;
736     s->status &= ~BUSY_STAT;
737 
738     if (ret != 0) {
739         if (ide_handle_rw_error(s, -ret, IDE_RETRY_PIO |
740                                 IDE_RETRY_READ)) {
741             return;
742         }
743     }
744 
745     block_acct_done(blk_get_stats(s->blk), &s->acct);
746 
747     n = s->nsector;
748     if (n > s->req_nb_sectors) {
749         n = s->req_nb_sectors;
750     }
751 
752     ide_set_sector(s, ide_get_sector(s) + n);
753     s->nsector -= n;
754     /* Allow the guest to read the io_buffer */
755     ide_transfer_start(s, s->io_buffer, n * BDRV_SECTOR_SIZE, ide_sector_read);
756     ide_set_irq(s->bus);
757 }
758 
759 static void ide_sector_read(IDEState *s)
760 {
761     int64_t sector_num;
762     int n;
763 
764     s->status = READY_STAT | SEEK_STAT;
765     s->error = 0; /* not needed by IDE spec, but needed by Windows */
766     sector_num = ide_get_sector(s);
767     n = s->nsector;
768 
769     if (n == 0) {
770         ide_transfer_stop(s);
771         return;
772     }
773 
774     s->status |= BUSY_STAT;
775 
776     if (n > s->req_nb_sectors) {
777         n = s->req_nb_sectors;
778     }
779 
780     trace_ide_sector_read(sector_num, n);
781 
782     if (!ide_sect_range_ok(s, sector_num, n)) {
783         ide_rw_error(s);
784         block_acct_invalid(blk_get_stats(s->blk), BLOCK_ACCT_READ);
785         return;
786     }
787 
788     qemu_iovec_init_buf(&s->qiov, s->io_buffer, n * BDRV_SECTOR_SIZE);
789 
790     block_acct_start(blk_get_stats(s->blk), &s->acct,
791                      n * BDRV_SECTOR_SIZE, BLOCK_ACCT_READ);
792     s->pio_aiocb = ide_buffered_readv(s, sector_num, &s->qiov, n,
793                                       ide_sector_read_cb, s);
794 }
795 
796 void dma_buf_commit(IDEState *s, uint32_t tx_bytes)
797 {
798     if (s->bus->dma->ops->commit_buf) {
799         s->bus->dma->ops->commit_buf(s->bus->dma, tx_bytes);
800     }
801     s->io_buffer_offset += tx_bytes;
802     qemu_sglist_destroy(&s->sg);
803 }
804 
805 void ide_set_inactive(IDEState *s, bool more)
806 {
807     s->bus->dma->aiocb = NULL;
808     ide_clear_retry(s);
809     if (s->bus->dma->ops->set_inactive) {
810         s->bus->dma->ops->set_inactive(s->bus->dma, more);
811     }
812     ide_cmd_done(s);
813 }
814 
815 void ide_dma_error(IDEState *s)
816 {
817     dma_buf_commit(s, 0);
818     ide_abort_command(s);
819     ide_set_inactive(s, false);
820     ide_set_irq(s->bus);
821 }
822 
823 int ide_handle_rw_error(IDEState *s, int error, int op)
824 {
825     bool is_read = (op & IDE_RETRY_READ) != 0;
826     BlockErrorAction action = blk_get_error_action(s->blk, is_read, error);
827 
828     if (action == BLOCK_ERROR_ACTION_STOP) {
829         assert(s->bus->retry_unit == s->unit);
830         s->bus->error_status = op;
831     } else if (action == BLOCK_ERROR_ACTION_REPORT) {
832         block_acct_failed(blk_get_stats(s->blk), &s->acct);
833         if (IS_IDE_RETRY_DMA(op)) {
834             ide_dma_error(s);
835         } else if (IS_IDE_RETRY_ATAPI(op)) {
836             ide_atapi_io_error(s, -error);
837         } else {
838             ide_rw_error(s);
839         }
840     }
841     blk_error_action(s->blk, action, is_read, error);
842     return action != BLOCK_ERROR_ACTION_IGNORE;
843 }
844 
845 static void ide_dma_cb(void *opaque, int ret)
846 {
847     IDEState *s = opaque;
848     int n;
849     int64_t sector_num;
850     uint64_t offset;
851     bool stay_active = false;
852     int32_t prep_size = 0;
853 
854     if (ret == -EINVAL) {
855         ide_dma_error(s);
856         return;
857     }
858 
859     if (ret < 0) {
860         if (ide_handle_rw_error(s, -ret, ide_dma_cmd_to_retry(s->dma_cmd))) {
861             s->bus->dma->aiocb = NULL;
862             dma_buf_commit(s, 0);
863             return;
864         }
865     }
866 
867     if (s->io_buffer_size > s->nsector * 512) {
868         /*
869          * The PRDs were longer than needed for this request.
870          * The Active bit must remain set after the request completes.
871          */
872         n = s->nsector;
873         stay_active = true;
874     } else {
875         n = s->io_buffer_size >> 9;
876     }
877 
878     sector_num = ide_get_sector(s);
879     if (n > 0) {
880         assert(n * 512 == s->sg.size);
881         dma_buf_commit(s, s->sg.size);
882         sector_num += n;
883         ide_set_sector(s, sector_num);
884         s->nsector -= n;
885     }
886 
887     /* end of transfer ? */
888     if (s->nsector == 0) {
889         s->status = READY_STAT | SEEK_STAT;
890         ide_set_irq(s->bus);
891         goto eot;
892     }
893 
894     /* launch next transfer */
895     n = s->nsector;
896     s->io_buffer_index = 0;
897     s->io_buffer_size = n * 512;
898     prep_size = s->bus->dma->ops->prepare_buf(s->bus->dma, s->io_buffer_size);
899     /* prepare_buf() must succeed and respect the limit */
900     assert(prep_size >= 0 && prep_size <= n * 512);
901 
902     /*
903      * Now prep_size stores the number of bytes in the sglist, and
904      * s->io_buffer_size stores the number of bytes described by the PRDs.
905      */
906 
907     if (prep_size < n * 512) {
908         /*
909          * The PRDs are too short for this request. Error condition!
910          * Reset the Active bit and don't raise the interrupt.
911          */
912         s->status = READY_STAT | SEEK_STAT;
913         dma_buf_commit(s, 0);
914         goto eot;
915     }
916 
917     trace_ide_dma_cb(s, sector_num, n, IDE_DMA_CMD_str(s->dma_cmd));
918 
919     if ((s->dma_cmd == IDE_DMA_READ || s->dma_cmd == IDE_DMA_WRITE) &&
920         !ide_sect_range_ok(s, sector_num, n)) {
921         ide_dma_error(s);
922         block_acct_invalid(blk_get_stats(s->blk), s->acct.type);
923         return;
924     }
925 
926     offset = sector_num << BDRV_SECTOR_BITS;
927     switch (s->dma_cmd) {
928     case IDE_DMA_READ:
929         s->bus->dma->aiocb = dma_blk_read(s->blk, &s->sg, offset,
930                                           BDRV_SECTOR_SIZE, ide_dma_cb, s);
931         break;
932     case IDE_DMA_WRITE:
933         s->bus->dma->aiocb = dma_blk_write(s->blk, &s->sg, offset,
934                                            BDRV_SECTOR_SIZE, ide_dma_cb, s);
935         break;
936     case IDE_DMA_TRIM:
937         s->bus->dma->aiocb = dma_blk_io(blk_get_aio_context(s->blk),
938                                         &s->sg, offset, BDRV_SECTOR_SIZE,
939                                         ide_issue_trim, s, ide_dma_cb, s,
940                                         DMA_DIRECTION_TO_DEVICE);
941         break;
942     default:
943         abort();
944     }
945     return;
946 
947 eot:
948     if (s->dma_cmd == IDE_DMA_READ || s->dma_cmd == IDE_DMA_WRITE) {
949         block_acct_done(blk_get_stats(s->blk), &s->acct);
950     }
951     ide_set_inactive(s, stay_active);
952 }
953 
954 static void ide_sector_start_dma(IDEState *s, enum ide_dma_cmd dma_cmd)
955 {
956     s->status = READY_STAT | SEEK_STAT | DRQ_STAT;
957     s->io_buffer_size = 0;
958     s->dma_cmd = dma_cmd;
959 
960     switch (dma_cmd) {
961     case IDE_DMA_READ:
962         block_acct_start(blk_get_stats(s->blk), &s->acct,
963                          s->nsector * BDRV_SECTOR_SIZE, BLOCK_ACCT_READ);
964         break;
965     case IDE_DMA_WRITE:
966         block_acct_start(blk_get_stats(s->blk), &s->acct,
967                          s->nsector * BDRV_SECTOR_SIZE, BLOCK_ACCT_WRITE);
968         break;
969     default:
970         break;
971     }
972 
973     ide_start_dma(s, ide_dma_cb);
974 }
975 
976 void ide_start_dma(IDEState *s, BlockCompletionFunc *cb)
977 {
978     s->io_buffer_index = 0;
979     ide_set_retry(s);
980     if (s->bus->dma->ops->start_dma) {
981         s->bus->dma->ops->start_dma(s->bus->dma, s, cb);
982     }
983 }
984 
985 static void ide_sector_write(IDEState *s);
986 
987 static void ide_sector_write_timer_cb(void *opaque)
988 {
989     IDEState *s = opaque;
990     ide_set_irq(s->bus);
991 }
992 
993 static void ide_sector_write_cb(void *opaque, int ret)
994 {
995     IDEState *s = opaque;
996     int n;
997 
998     s->pio_aiocb = NULL;
999     s->status &= ~BUSY_STAT;
1000 
1001     if (ret != 0) {
1002         if (ide_handle_rw_error(s, -ret, IDE_RETRY_PIO)) {
1003             return;
1004         }
1005     }
1006 
1007     block_acct_done(blk_get_stats(s->blk), &s->acct);
1008 
1009     n = s->nsector;
1010     if (n > s->req_nb_sectors) {
1011         n = s->req_nb_sectors;
1012     }
1013     s->nsector -= n;
1014 
1015     ide_set_sector(s, ide_get_sector(s) + n);
1016     if (s->nsector == 0) {
1017         /* no more sectors to write */
1018         ide_transfer_stop(s);
1019     } else {
1020         int n1 = s->nsector;
1021         if (n1 > s->req_nb_sectors) {
1022             n1 = s->req_nb_sectors;
1023         }
1024         ide_transfer_start(s, s->io_buffer, n1 * BDRV_SECTOR_SIZE,
1025                            ide_sector_write);
1026     }
1027 
1028     if (win2k_install_hack && ((++s->irq_count % 16) == 0)) {
1029         /* It seems there is a bug in the Windows 2000 installer HDD
1030            IDE driver which fills the disk with empty logs when the
1031            IDE write IRQ comes too early. This hack tries to correct
1032            that at the expense of slower write performances. Use this
1033            option _only_ to install Windows 2000. You must disable it
1034            for normal use. */
1035         timer_mod(s->sector_write_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
1036                   (NANOSECONDS_PER_SECOND / 1000));
1037     } else {
1038         ide_set_irq(s->bus);
1039     }
1040 }
1041 
1042 static void ide_sector_write(IDEState *s)
1043 {
1044     int64_t sector_num;
1045     int n;
1046 
1047     s->status = READY_STAT | SEEK_STAT | BUSY_STAT;
1048     sector_num = ide_get_sector(s);
1049 
1050     n = s->nsector;
1051     if (n > s->req_nb_sectors) {
1052         n = s->req_nb_sectors;
1053     }
1054 
1055     trace_ide_sector_write(sector_num, n);
1056 
1057     if (!ide_sect_range_ok(s, sector_num, n)) {
1058         ide_rw_error(s);
1059         block_acct_invalid(blk_get_stats(s->blk), BLOCK_ACCT_WRITE);
1060         return;
1061     }
1062 
1063     qemu_iovec_init_buf(&s->qiov, s->io_buffer, n * BDRV_SECTOR_SIZE);
1064 
1065     block_acct_start(blk_get_stats(s->blk), &s->acct,
1066                      n * BDRV_SECTOR_SIZE, BLOCK_ACCT_WRITE);
1067     s->pio_aiocb = blk_aio_pwritev(s->blk, sector_num << BDRV_SECTOR_BITS,
1068                                    &s->qiov, 0, ide_sector_write_cb, s);
1069 }
1070 
1071 static void ide_flush_cb(void *opaque, int ret)
1072 {
1073     IDEState *s = opaque;
1074 
1075     s->pio_aiocb = NULL;
1076 
1077     if (ret < 0) {
1078         /* XXX: What sector number to set here? */
1079         if (ide_handle_rw_error(s, -ret, IDE_RETRY_FLUSH)) {
1080             return;
1081         }
1082     }
1083 
1084     if (s->blk) {
1085         block_acct_done(blk_get_stats(s->blk), &s->acct);
1086     }
1087     s->status = READY_STAT | SEEK_STAT;
1088     ide_cmd_done(s);
1089     ide_set_irq(s->bus);
1090 }
1091 
1092 static void ide_flush_cache(IDEState *s)
1093 {
1094     if (s->blk == NULL) {
1095         ide_flush_cb(s, 0);
1096         return;
1097     }
1098 
1099     s->status |= BUSY_STAT;
1100     ide_set_retry(s);
1101     block_acct_start(blk_get_stats(s->blk), &s->acct, 0, BLOCK_ACCT_FLUSH);
1102     s->pio_aiocb = blk_aio_flush(s->blk, ide_flush_cb, s);
1103 }
1104 
1105 static void ide_cfata_metadata_inquiry(IDEState *s)
1106 {
1107     uint16_t *p;
1108     uint32_t spd;
1109 
1110     p = (uint16_t *) s->io_buffer;
1111     memset(p, 0, 0x200);
1112     spd = ((s->mdata_size - 1) >> 9) + 1;
1113 
1114     put_le16(p + 0, 0x0001);			/* Data format revision */
1115     put_le16(p + 1, 0x0000);			/* Media property: silicon */
1116     put_le16(p + 2, s->media_changed);		/* Media status */
1117     put_le16(p + 3, s->mdata_size & 0xffff);	/* Capacity in bytes (low) */
1118     put_le16(p + 4, s->mdata_size >> 16);	/* Capacity in bytes (high) */
1119     put_le16(p + 5, spd & 0xffff);		/* Sectors per device (low) */
1120     put_le16(p + 6, spd >> 16);			/* Sectors per device (high) */
1121 }
1122 
1123 static void ide_cfata_metadata_read(IDEState *s)
1124 {
1125     uint16_t *p;
1126 
1127     if (((s->hcyl << 16) | s->lcyl) << 9 > s->mdata_size + 2) {
1128         s->status = ERR_STAT;
1129         s->error = ABRT_ERR;
1130         return;
1131     }
1132 
1133     p = (uint16_t *) s->io_buffer;
1134     memset(p, 0, 0x200);
1135 
1136     put_le16(p + 0, s->media_changed);		/* Media status */
1137     memcpy(p + 1, s->mdata_storage + (((s->hcyl << 16) | s->lcyl) << 9),
1138                     MIN(MIN(s->mdata_size - (((s->hcyl << 16) | s->lcyl) << 9),
1139                                     s->nsector << 9), 0x200 - 2));
1140 }
1141 
1142 static void ide_cfata_metadata_write(IDEState *s)
1143 {
1144     if (((s->hcyl << 16) | s->lcyl) << 9 > s->mdata_size + 2) {
1145         s->status = ERR_STAT;
1146         s->error = ABRT_ERR;
1147         return;
1148     }
1149 
1150     s->media_changed = 0;
1151 
1152     memcpy(s->mdata_storage + (((s->hcyl << 16) | s->lcyl) << 9),
1153                     s->io_buffer + 2,
1154                     MIN(MIN(s->mdata_size - (((s->hcyl << 16) | s->lcyl) << 9),
1155                                     s->nsector << 9), 0x200 - 2));
1156 }
1157 
1158 /* called when the inserted state of the media has changed */
1159 static void ide_cd_change_cb(void *opaque, bool load, Error **errp)
1160 {
1161     IDEState *s = opaque;
1162     uint64_t nb_sectors;
1163 
1164     s->tray_open = !load;
1165     blk_get_geometry(s->blk, &nb_sectors);
1166     s->nb_sectors = nb_sectors;
1167 
1168     /*
1169      * First indicate to the guest that a CD has been removed.  That's
1170      * done on the next command the guest sends us.
1171      *
1172      * Then we set UNIT_ATTENTION, by which the guest will
1173      * detect a new CD in the drive.  See ide_atapi_cmd() for details.
1174      */
1175     s->cdrom_changed = 1;
1176     s->events.new_media = true;
1177     s->events.eject_request = false;
1178     ide_set_irq(s->bus);
1179 }
1180 
1181 static void ide_cd_eject_request_cb(void *opaque, bool force)
1182 {
1183     IDEState *s = opaque;
1184 
1185     s->events.eject_request = true;
1186     if (force) {
1187         s->tray_locked = false;
1188     }
1189     ide_set_irq(s->bus);
1190 }
1191 
1192 static void ide_cmd_lba48_transform(IDEState *s, int lba48)
1193 {
1194     s->lba48 = lba48;
1195 
1196     /* handle the 'magic' 0 nsector count conversion here. to avoid
1197      * fiddling with the rest of the read logic, we just store the
1198      * full sector count in ->nsector and ignore ->hob_nsector from now
1199      */
1200     if (!s->lba48) {
1201         if (!s->nsector)
1202             s->nsector = 256;
1203     } else {
1204         if (!s->nsector && !s->hob_nsector)
1205             s->nsector = 65536;
1206         else {
1207             int lo = s->nsector;
1208             int hi = s->hob_nsector;
1209 
1210             s->nsector = (hi << 8) | lo;
1211         }
1212     }
1213 }
1214 
1215 static void ide_clear_hob(IDEBus *bus)
1216 {
1217     /* any write clears HOB high bit of device control register */
1218     bus->ifs[0].select &= ~(1 << 7);
1219     bus->ifs[1].select &= ~(1 << 7);
1220 }
1221 
1222 /* IOport [W]rite [R]egisters */
1223 enum ATA_IOPORT_WR {
1224     ATA_IOPORT_WR_DATA = 0,
1225     ATA_IOPORT_WR_FEATURES = 1,
1226     ATA_IOPORT_WR_SECTOR_COUNT = 2,
1227     ATA_IOPORT_WR_SECTOR_NUMBER = 3,
1228     ATA_IOPORT_WR_CYLINDER_LOW = 4,
1229     ATA_IOPORT_WR_CYLINDER_HIGH = 5,
1230     ATA_IOPORT_WR_DEVICE_HEAD = 6,
1231     ATA_IOPORT_WR_COMMAND = 7,
1232     ATA_IOPORT_WR_NUM_REGISTERS,
1233 };
1234 
1235 const char *ATA_IOPORT_WR_lookup[ATA_IOPORT_WR_NUM_REGISTERS] = {
1236     [ATA_IOPORT_WR_DATA] = "Data",
1237     [ATA_IOPORT_WR_FEATURES] = "Features",
1238     [ATA_IOPORT_WR_SECTOR_COUNT] = "Sector Count",
1239     [ATA_IOPORT_WR_SECTOR_NUMBER] = "Sector Number",
1240     [ATA_IOPORT_WR_CYLINDER_LOW] = "Cylinder Low",
1241     [ATA_IOPORT_WR_CYLINDER_HIGH] = "Cylinder High",
1242     [ATA_IOPORT_WR_DEVICE_HEAD] = "Device/Head",
1243     [ATA_IOPORT_WR_COMMAND] = "Command"
1244 };
1245 
1246 void ide_ioport_write(void *opaque, uint32_t addr, uint32_t val)
1247 {
1248     IDEBus *bus = opaque;
1249     IDEState *s = idebus_active_if(bus);
1250     int reg_num = addr & 7;
1251 
1252     trace_ide_ioport_write(addr, ATA_IOPORT_WR_lookup[reg_num], val, bus, s);
1253 
1254     /* ignore writes to command block while busy with previous command */
1255     if (reg_num != 7 && (s->status & (BUSY_STAT|DRQ_STAT))) {
1256         return;
1257     }
1258 
1259     switch (reg_num) {
1260     case 0:
1261         break;
1262     case ATA_IOPORT_WR_FEATURES:
1263         ide_clear_hob(bus);
1264         /* NOTE: data is written to the two drives */
1265         bus->ifs[0].hob_feature = bus->ifs[0].feature;
1266         bus->ifs[1].hob_feature = bus->ifs[1].feature;
1267         bus->ifs[0].feature = val;
1268         bus->ifs[1].feature = val;
1269         break;
1270     case ATA_IOPORT_WR_SECTOR_COUNT:
1271         ide_clear_hob(bus);
1272         bus->ifs[0].hob_nsector = bus->ifs[0].nsector;
1273         bus->ifs[1].hob_nsector = bus->ifs[1].nsector;
1274         bus->ifs[0].nsector = val;
1275         bus->ifs[1].nsector = val;
1276         break;
1277     case ATA_IOPORT_WR_SECTOR_NUMBER:
1278         ide_clear_hob(bus);
1279         bus->ifs[0].hob_sector = bus->ifs[0].sector;
1280         bus->ifs[1].hob_sector = bus->ifs[1].sector;
1281         bus->ifs[0].sector = val;
1282         bus->ifs[1].sector = val;
1283         break;
1284     case ATA_IOPORT_WR_CYLINDER_LOW:
1285         ide_clear_hob(bus);
1286         bus->ifs[0].hob_lcyl = bus->ifs[0].lcyl;
1287         bus->ifs[1].hob_lcyl = bus->ifs[1].lcyl;
1288         bus->ifs[0].lcyl = val;
1289         bus->ifs[1].lcyl = val;
1290         break;
1291     case ATA_IOPORT_WR_CYLINDER_HIGH:
1292         ide_clear_hob(bus);
1293         bus->ifs[0].hob_hcyl = bus->ifs[0].hcyl;
1294         bus->ifs[1].hob_hcyl = bus->ifs[1].hcyl;
1295         bus->ifs[0].hcyl = val;
1296         bus->ifs[1].hcyl = val;
1297         break;
1298     case ATA_IOPORT_WR_DEVICE_HEAD:
1299         /* FIXME: HOB readback uses bit 7 */
1300         bus->ifs[0].select = (val & ~0x10) | 0xa0;
1301         bus->ifs[1].select = (val | 0x10) | 0xa0;
1302         /* select drive */
1303         bus->unit = (val >> 4) & 1;
1304         break;
1305     default:
1306     case ATA_IOPORT_WR_COMMAND:
1307         /* command */
1308         ide_exec_cmd(bus, val);
1309         break;
1310     }
1311 }
1312 
1313 static void ide_reset(IDEState *s)
1314 {
1315     trace_ide_reset(s);
1316 
1317     if (s->pio_aiocb) {
1318         blk_aio_cancel(s->pio_aiocb);
1319         s->pio_aiocb = NULL;
1320     }
1321 
1322     if (s->drive_kind == IDE_CFATA)
1323         s->mult_sectors = 0;
1324     else
1325         s->mult_sectors = MAX_MULT_SECTORS;
1326     /* ide regs */
1327     s->feature = 0;
1328     s->error = 0;
1329     s->nsector = 0;
1330     s->sector = 0;
1331     s->lcyl = 0;
1332     s->hcyl = 0;
1333 
1334     /* lba48 */
1335     s->hob_feature = 0;
1336     s->hob_sector = 0;
1337     s->hob_nsector = 0;
1338     s->hob_lcyl = 0;
1339     s->hob_hcyl = 0;
1340 
1341     s->select = 0xa0;
1342     s->status = READY_STAT | SEEK_STAT;
1343 
1344     s->lba48 = 0;
1345 
1346     /* ATAPI specific */
1347     s->sense_key = 0;
1348     s->asc = 0;
1349     s->cdrom_changed = 0;
1350     s->packet_transfer_size = 0;
1351     s->elementary_transfer_size = 0;
1352     s->io_buffer_index = 0;
1353     s->cd_sector_size = 0;
1354     s->atapi_dma = 0;
1355     s->tray_locked = 0;
1356     s->tray_open = 0;
1357     /* ATA DMA state */
1358     s->io_buffer_size = 0;
1359     s->req_nb_sectors = 0;
1360 
1361     ide_set_signature(s);
1362     /* init the transfer handler so that 0xffff is returned on data
1363        accesses */
1364     s->end_transfer_func = ide_dummy_transfer_stop;
1365     ide_dummy_transfer_stop(s);
1366     s->media_changed = 0;
1367 }
1368 
1369 static bool cmd_nop(IDEState *s, uint8_t cmd)
1370 {
1371     return true;
1372 }
1373 
1374 static bool cmd_device_reset(IDEState *s, uint8_t cmd)
1375 {
1376     /* Halt PIO (in the DRQ phase), then DMA */
1377     ide_transfer_halt(s);
1378     ide_cancel_dma_sync(s);
1379 
1380     /* Reset any PIO commands, reset signature, etc */
1381     ide_reset(s);
1382 
1383     /* RESET: ATA8-ACS3 7.10.4 "Normal Outputs";
1384      * ATA8-ACS3 Table 184 "Device Signatures for Normal Output" */
1385     s->status = 0x00;
1386 
1387     /* Do not overwrite status register */
1388     return false;
1389 }
1390 
1391 static bool cmd_data_set_management(IDEState *s, uint8_t cmd)
1392 {
1393     switch (s->feature) {
1394     case DSM_TRIM:
1395         if (s->blk) {
1396             ide_sector_start_dma(s, IDE_DMA_TRIM);
1397             return false;
1398         }
1399         break;
1400     }
1401 
1402     ide_abort_command(s);
1403     return true;
1404 }
1405 
1406 static bool cmd_identify(IDEState *s, uint8_t cmd)
1407 {
1408     if (s->blk && s->drive_kind != IDE_CD) {
1409         if (s->drive_kind != IDE_CFATA) {
1410             ide_identify(s);
1411         } else {
1412             ide_cfata_identify(s);
1413         }
1414         s->status = READY_STAT | SEEK_STAT;
1415         ide_transfer_start(s, s->io_buffer, 512, ide_transfer_stop);
1416         ide_set_irq(s->bus);
1417         return false;
1418     } else {
1419         if (s->drive_kind == IDE_CD) {
1420             ide_set_signature(s);
1421         }
1422         ide_abort_command(s);
1423     }
1424 
1425     return true;
1426 }
1427 
1428 static bool cmd_verify(IDEState *s, uint8_t cmd)
1429 {
1430     bool lba48 = (cmd == WIN_VERIFY_EXT);
1431 
1432     /* do sector number check ? */
1433     ide_cmd_lba48_transform(s, lba48);
1434 
1435     return true;
1436 }
1437 
1438 static bool cmd_set_multiple_mode(IDEState *s, uint8_t cmd)
1439 {
1440     if (s->drive_kind == IDE_CFATA && s->nsector == 0) {
1441         /* Disable Read and Write Multiple */
1442         s->mult_sectors = 0;
1443     } else if ((s->nsector & 0xff) != 0 &&
1444         ((s->nsector & 0xff) > MAX_MULT_SECTORS ||
1445          (s->nsector & (s->nsector - 1)) != 0)) {
1446         ide_abort_command(s);
1447     } else {
1448         s->mult_sectors = s->nsector & 0xff;
1449     }
1450 
1451     return true;
1452 }
1453 
1454 static bool cmd_read_multiple(IDEState *s, uint8_t cmd)
1455 {
1456     bool lba48 = (cmd == WIN_MULTREAD_EXT);
1457 
1458     if (!s->blk || !s->mult_sectors) {
1459         ide_abort_command(s);
1460         return true;
1461     }
1462 
1463     ide_cmd_lba48_transform(s, lba48);
1464     s->req_nb_sectors = s->mult_sectors;
1465     ide_sector_read(s);
1466     return false;
1467 }
1468 
1469 static bool cmd_write_multiple(IDEState *s, uint8_t cmd)
1470 {
1471     bool lba48 = (cmd == WIN_MULTWRITE_EXT);
1472     int n;
1473 
1474     if (!s->blk || !s->mult_sectors) {
1475         ide_abort_command(s);
1476         return true;
1477     }
1478 
1479     ide_cmd_lba48_transform(s, lba48);
1480 
1481     s->req_nb_sectors = s->mult_sectors;
1482     n = MIN(s->nsector, s->req_nb_sectors);
1483 
1484     s->status = SEEK_STAT | READY_STAT;
1485     ide_transfer_start(s, s->io_buffer, 512 * n, ide_sector_write);
1486 
1487     s->media_changed = 1;
1488 
1489     return false;
1490 }
1491 
1492 static bool cmd_read_pio(IDEState *s, uint8_t cmd)
1493 {
1494     bool lba48 = (cmd == WIN_READ_EXT);
1495 
1496     if (s->drive_kind == IDE_CD) {
1497         ide_set_signature(s); /* odd, but ATA4 8.27.5.2 requires it */
1498         ide_abort_command(s);
1499         return true;
1500     }
1501 
1502     if (!s->blk) {
1503         ide_abort_command(s);
1504         return true;
1505     }
1506 
1507     ide_cmd_lba48_transform(s, lba48);
1508     s->req_nb_sectors = 1;
1509     ide_sector_read(s);
1510 
1511     return false;
1512 }
1513 
1514 static bool cmd_write_pio(IDEState *s, uint8_t cmd)
1515 {
1516     bool lba48 = (cmd == WIN_WRITE_EXT);
1517 
1518     if (!s->blk) {
1519         ide_abort_command(s);
1520         return true;
1521     }
1522 
1523     ide_cmd_lba48_transform(s, lba48);
1524 
1525     s->req_nb_sectors = 1;
1526     s->status = SEEK_STAT | READY_STAT;
1527     ide_transfer_start(s, s->io_buffer, 512, ide_sector_write);
1528 
1529     s->media_changed = 1;
1530 
1531     return false;
1532 }
1533 
1534 static bool cmd_read_dma(IDEState *s, uint8_t cmd)
1535 {
1536     bool lba48 = (cmd == WIN_READDMA_EXT);
1537 
1538     if (!s->blk) {
1539         ide_abort_command(s);
1540         return true;
1541     }
1542 
1543     ide_cmd_lba48_transform(s, lba48);
1544     ide_sector_start_dma(s, IDE_DMA_READ);
1545 
1546     return false;
1547 }
1548 
1549 static bool cmd_write_dma(IDEState *s, uint8_t cmd)
1550 {
1551     bool lba48 = (cmd == WIN_WRITEDMA_EXT);
1552 
1553     if (!s->blk) {
1554         ide_abort_command(s);
1555         return true;
1556     }
1557 
1558     ide_cmd_lba48_transform(s, lba48);
1559     ide_sector_start_dma(s, IDE_DMA_WRITE);
1560 
1561     s->media_changed = 1;
1562 
1563     return false;
1564 }
1565 
1566 static bool cmd_flush_cache(IDEState *s, uint8_t cmd)
1567 {
1568     ide_flush_cache(s);
1569     return false;
1570 }
1571 
1572 static bool cmd_seek(IDEState *s, uint8_t cmd)
1573 {
1574     /* XXX: Check that seek is within bounds */
1575     return true;
1576 }
1577 
1578 static bool cmd_read_native_max(IDEState *s, uint8_t cmd)
1579 {
1580     bool lba48 = (cmd == WIN_READ_NATIVE_MAX_EXT);
1581 
1582     /* Refuse if no sectors are addressable (e.g. medium not inserted) */
1583     if (s->nb_sectors == 0) {
1584         ide_abort_command(s);
1585         return true;
1586     }
1587 
1588     ide_cmd_lba48_transform(s, lba48);
1589     ide_set_sector(s, s->nb_sectors - 1);
1590 
1591     return true;
1592 }
1593 
1594 static bool cmd_check_power_mode(IDEState *s, uint8_t cmd)
1595 {
1596     s->nsector = 0xff; /* device active or idle */
1597     return true;
1598 }
1599 
1600 static bool cmd_set_features(IDEState *s, uint8_t cmd)
1601 {
1602     uint16_t *identify_data;
1603 
1604     if (!s->blk) {
1605         ide_abort_command(s);
1606         return true;
1607     }
1608 
1609     /* XXX: valid for CDROM ? */
1610     switch (s->feature) {
1611     case 0x02: /* write cache enable */
1612         blk_set_enable_write_cache(s->blk, true);
1613         identify_data = (uint16_t *)s->identify_data;
1614         put_le16(identify_data + 85, (1 << 14) | (1 << 5) | 1);
1615         return true;
1616     case 0x82: /* write cache disable */
1617         blk_set_enable_write_cache(s->blk, false);
1618         identify_data = (uint16_t *)s->identify_data;
1619         put_le16(identify_data + 85, (1 << 14) | 1);
1620         ide_flush_cache(s);
1621         return false;
1622     case 0xcc: /* reverting to power-on defaults enable */
1623     case 0x66: /* reverting to power-on defaults disable */
1624     case 0xaa: /* read look-ahead enable */
1625     case 0x55: /* read look-ahead disable */
1626     case 0x05: /* set advanced power management mode */
1627     case 0x85: /* disable advanced power management mode */
1628     case 0x69: /* NOP */
1629     case 0x67: /* NOP */
1630     case 0x96: /* NOP */
1631     case 0x9a: /* NOP */
1632     case 0x42: /* enable Automatic Acoustic Mode */
1633     case 0xc2: /* disable Automatic Acoustic Mode */
1634         return true;
1635     case 0x03: /* set transfer mode */
1636         {
1637             uint8_t val = s->nsector & 0x07;
1638             identify_data = (uint16_t *)s->identify_data;
1639 
1640             switch (s->nsector >> 3) {
1641             case 0x00: /* pio default */
1642             case 0x01: /* pio mode */
1643                 put_le16(identify_data + 62, 0x07);
1644                 put_le16(identify_data + 63, 0x07);
1645                 put_le16(identify_data + 88, 0x3f);
1646                 break;
1647             case 0x02: /* sigle word dma mode*/
1648                 put_le16(identify_data + 62, 0x07 | (1 << (val + 8)));
1649                 put_le16(identify_data + 63, 0x07);
1650                 put_le16(identify_data + 88, 0x3f);
1651                 break;
1652             case 0x04: /* mdma mode */
1653                 put_le16(identify_data + 62, 0x07);
1654                 put_le16(identify_data + 63, 0x07 | (1 << (val + 8)));
1655                 put_le16(identify_data + 88, 0x3f);
1656                 break;
1657             case 0x08: /* udma mode */
1658                 put_le16(identify_data + 62, 0x07);
1659                 put_le16(identify_data + 63, 0x07);
1660                 put_le16(identify_data + 88, 0x3f | (1 << (val + 8)));
1661                 break;
1662             default:
1663                 goto abort_cmd;
1664             }
1665             return true;
1666         }
1667     }
1668 
1669 abort_cmd:
1670     ide_abort_command(s);
1671     return true;
1672 }
1673 
1674 
1675 /*** ATAPI commands ***/
1676 
1677 static bool cmd_identify_packet(IDEState *s, uint8_t cmd)
1678 {
1679     ide_atapi_identify(s);
1680     s->status = READY_STAT | SEEK_STAT;
1681     ide_transfer_start(s, s->io_buffer, 512, ide_transfer_stop);
1682     ide_set_irq(s->bus);
1683     return false;
1684 }
1685 
1686 static bool cmd_exec_dev_diagnostic(IDEState *s, uint8_t cmd)
1687 {
1688     ide_set_signature(s);
1689 
1690     if (s->drive_kind == IDE_CD) {
1691         s->status = 0; /* ATAPI spec (v6) section 9.10 defines packet
1692                         * devices to return a clear status register
1693                         * with READY_STAT *not* set. */
1694         s->error = 0x01;
1695     } else {
1696         s->status = READY_STAT | SEEK_STAT;
1697         /* The bits of the error register are not as usual for this command!
1698          * They are part of the regular output (this is why ERR_STAT isn't set)
1699          * Device 0 passed, Device 1 passed or not present. */
1700         s->error = 0x01;
1701         ide_set_irq(s->bus);
1702     }
1703 
1704     return false;
1705 }
1706 
1707 static bool cmd_packet(IDEState *s, uint8_t cmd)
1708 {
1709     /* overlapping commands not supported */
1710     if (s->feature & 0x02) {
1711         ide_abort_command(s);
1712         return true;
1713     }
1714 
1715     s->status = READY_STAT | SEEK_STAT;
1716     s->atapi_dma = s->feature & 1;
1717     if (s->atapi_dma) {
1718         s->dma_cmd = IDE_DMA_ATAPI;
1719     }
1720     s->nsector = 1;
1721     ide_transfer_start(s, s->io_buffer, ATAPI_PACKET_SIZE,
1722                        ide_atapi_cmd);
1723     return false;
1724 }
1725 
1726 
1727 /*** CF-ATA commands ***/
1728 
1729 static bool cmd_cfa_req_ext_error_code(IDEState *s, uint8_t cmd)
1730 {
1731     s->error = 0x09;    /* miscellaneous error */
1732     s->status = READY_STAT | SEEK_STAT;
1733     ide_set_irq(s->bus);
1734 
1735     return false;
1736 }
1737 
1738 static bool cmd_cfa_erase_sectors(IDEState *s, uint8_t cmd)
1739 {
1740     /* WIN_SECURITY_FREEZE_LOCK has the same ID as CFA_WEAR_LEVEL and is
1741      * required for Windows 8 to work with AHCI */
1742 
1743     if (cmd == CFA_WEAR_LEVEL) {
1744         s->nsector = 0;
1745     }
1746 
1747     if (cmd == CFA_ERASE_SECTORS) {
1748         s->media_changed = 1;
1749     }
1750 
1751     return true;
1752 }
1753 
1754 static bool cmd_cfa_translate_sector(IDEState *s, uint8_t cmd)
1755 {
1756     s->status = READY_STAT | SEEK_STAT;
1757 
1758     memset(s->io_buffer, 0, 0x200);
1759     s->io_buffer[0x00] = s->hcyl;                   /* Cyl MSB */
1760     s->io_buffer[0x01] = s->lcyl;                   /* Cyl LSB */
1761     s->io_buffer[0x02] = s->select;                 /* Head */
1762     s->io_buffer[0x03] = s->sector;                 /* Sector */
1763     s->io_buffer[0x04] = ide_get_sector(s) >> 16;   /* LBA MSB */
1764     s->io_buffer[0x05] = ide_get_sector(s) >> 8;    /* LBA */
1765     s->io_buffer[0x06] = ide_get_sector(s) >> 0;    /* LBA LSB */
1766     s->io_buffer[0x13] = 0x00;                      /* Erase flag */
1767     s->io_buffer[0x18] = 0x00;                      /* Hot count */
1768     s->io_buffer[0x19] = 0x00;                      /* Hot count */
1769     s->io_buffer[0x1a] = 0x01;                      /* Hot count */
1770 
1771     ide_transfer_start(s, s->io_buffer, 0x200, ide_transfer_stop);
1772     ide_set_irq(s->bus);
1773 
1774     return false;
1775 }
1776 
1777 static bool cmd_cfa_access_metadata_storage(IDEState *s, uint8_t cmd)
1778 {
1779     switch (s->feature) {
1780     case 0x02:  /* Inquiry Metadata Storage */
1781         ide_cfata_metadata_inquiry(s);
1782         break;
1783     case 0x03:  /* Read Metadata Storage */
1784         ide_cfata_metadata_read(s);
1785         break;
1786     case 0x04:  /* Write Metadata Storage */
1787         ide_cfata_metadata_write(s);
1788         break;
1789     default:
1790         ide_abort_command(s);
1791         return true;
1792     }
1793 
1794     ide_transfer_start(s, s->io_buffer, 0x200, ide_transfer_stop);
1795     s->status = 0x00; /* NOTE: READY is _not_ set */
1796     ide_set_irq(s->bus);
1797 
1798     return false;
1799 }
1800 
1801 static bool cmd_ibm_sense_condition(IDEState *s, uint8_t cmd)
1802 {
1803     switch (s->feature) {
1804     case 0x01:  /* sense temperature in device */
1805         s->nsector = 0x50;      /* +20 C */
1806         break;
1807     default:
1808         ide_abort_command(s);
1809         return true;
1810     }
1811 
1812     return true;
1813 }
1814 
1815 
1816 /*** SMART commands ***/
1817 
1818 static bool cmd_smart(IDEState *s, uint8_t cmd)
1819 {
1820     int n;
1821 
1822     if (s->hcyl != 0xc2 || s->lcyl != 0x4f) {
1823         goto abort_cmd;
1824     }
1825 
1826     if (!s->smart_enabled && s->feature != SMART_ENABLE) {
1827         goto abort_cmd;
1828     }
1829 
1830     switch (s->feature) {
1831     case SMART_DISABLE:
1832         s->smart_enabled = 0;
1833         return true;
1834 
1835     case SMART_ENABLE:
1836         s->smart_enabled = 1;
1837         return true;
1838 
1839     case SMART_ATTR_AUTOSAVE:
1840         switch (s->sector) {
1841         case 0x00:
1842             s->smart_autosave = 0;
1843             break;
1844         case 0xf1:
1845             s->smart_autosave = 1;
1846             break;
1847         default:
1848             goto abort_cmd;
1849         }
1850         return true;
1851 
1852     case SMART_STATUS:
1853         if (!s->smart_errors) {
1854             s->hcyl = 0xc2;
1855             s->lcyl = 0x4f;
1856         } else {
1857             s->hcyl = 0x2c;
1858             s->lcyl = 0xf4;
1859         }
1860         return true;
1861 
1862     case SMART_READ_THRESH:
1863         memset(s->io_buffer, 0, 0x200);
1864         s->io_buffer[0] = 0x01; /* smart struct version */
1865 
1866         for (n = 0; n < ARRAY_SIZE(smart_attributes); n++) {
1867             s->io_buffer[2 + 0 + (n * 12)] = smart_attributes[n][0];
1868             s->io_buffer[2 + 1 + (n * 12)] = smart_attributes[n][11];
1869         }
1870 
1871         /* checksum */
1872         for (n = 0; n < 511; n++) {
1873             s->io_buffer[511] += s->io_buffer[n];
1874         }
1875         s->io_buffer[511] = 0x100 - s->io_buffer[511];
1876 
1877         s->status = READY_STAT | SEEK_STAT;
1878         ide_transfer_start(s, s->io_buffer, 0x200, ide_transfer_stop);
1879         ide_set_irq(s->bus);
1880         return false;
1881 
1882     case SMART_READ_DATA:
1883         memset(s->io_buffer, 0, 0x200);
1884         s->io_buffer[0] = 0x01; /* smart struct version */
1885 
1886         for (n = 0; n < ARRAY_SIZE(smart_attributes); n++) {
1887             int i;
1888             for (i = 0; i < 11; i++) {
1889                 s->io_buffer[2 + i + (n * 12)] = smart_attributes[n][i];
1890             }
1891         }
1892 
1893         s->io_buffer[362] = 0x02 | (s->smart_autosave ? 0x80 : 0x00);
1894         if (s->smart_selftest_count == 0) {
1895             s->io_buffer[363] = 0;
1896         } else {
1897             s->io_buffer[363] =
1898                 s->smart_selftest_data[3 +
1899                            (s->smart_selftest_count - 1) *
1900                            24];
1901         }
1902         s->io_buffer[364] = 0x20;
1903         s->io_buffer[365] = 0x01;
1904         /* offline data collection capacity: execute + self-test*/
1905         s->io_buffer[367] = (1 << 4 | 1 << 3 | 1);
1906         s->io_buffer[368] = 0x03; /* smart capability (1) */
1907         s->io_buffer[369] = 0x00; /* smart capability (2) */
1908         s->io_buffer[370] = 0x01; /* error logging supported */
1909         s->io_buffer[372] = 0x02; /* minutes for poll short test */
1910         s->io_buffer[373] = 0x36; /* minutes for poll ext test */
1911         s->io_buffer[374] = 0x01; /* minutes for poll conveyance */
1912 
1913         for (n = 0; n < 511; n++) {
1914             s->io_buffer[511] += s->io_buffer[n];
1915         }
1916         s->io_buffer[511] = 0x100 - s->io_buffer[511];
1917 
1918         s->status = READY_STAT | SEEK_STAT;
1919         ide_transfer_start(s, s->io_buffer, 0x200, ide_transfer_stop);
1920         ide_set_irq(s->bus);
1921         return false;
1922 
1923     case SMART_READ_LOG:
1924         switch (s->sector) {
1925         case 0x01: /* summary smart error log */
1926             memset(s->io_buffer, 0, 0x200);
1927             s->io_buffer[0] = 0x01;
1928             s->io_buffer[1] = 0x00; /* no error entries */
1929             s->io_buffer[452] = s->smart_errors & 0xff;
1930             s->io_buffer[453] = (s->smart_errors & 0xff00) >> 8;
1931 
1932             for (n = 0; n < 511; n++) {
1933                 s->io_buffer[511] += s->io_buffer[n];
1934             }
1935             s->io_buffer[511] = 0x100 - s->io_buffer[511];
1936             break;
1937         case 0x06: /* smart self test log */
1938             memset(s->io_buffer, 0, 0x200);
1939             s->io_buffer[0] = 0x01;
1940             if (s->smart_selftest_count == 0) {
1941                 s->io_buffer[508] = 0;
1942             } else {
1943                 s->io_buffer[508] = s->smart_selftest_count;
1944                 for (n = 2; n < 506; n++)  {
1945                     s->io_buffer[n] = s->smart_selftest_data[n];
1946                 }
1947             }
1948 
1949             for (n = 0; n < 511; n++) {
1950                 s->io_buffer[511] += s->io_buffer[n];
1951             }
1952             s->io_buffer[511] = 0x100 - s->io_buffer[511];
1953             break;
1954         default:
1955             goto abort_cmd;
1956         }
1957         s->status = READY_STAT | SEEK_STAT;
1958         ide_transfer_start(s, s->io_buffer, 0x200, ide_transfer_stop);
1959         ide_set_irq(s->bus);
1960         return false;
1961 
1962     case SMART_EXECUTE_OFFLINE:
1963         switch (s->sector) {
1964         case 0: /* off-line routine */
1965         case 1: /* short self test */
1966         case 2: /* extended self test */
1967             s->smart_selftest_count++;
1968             if (s->smart_selftest_count > 21) {
1969                 s->smart_selftest_count = 1;
1970             }
1971             n = 2 + (s->smart_selftest_count - 1) * 24;
1972             s->smart_selftest_data[n] = s->sector;
1973             s->smart_selftest_data[n + 1] = 0x00; /* OK and finished */
1974             s->smart_selftest_data[n + 2] = 0x34; /* hour count lsb */
1975             s->smart_selftest_data[n + 3] = 0x12; /* hour count msb */
1976             break;
1977         default:
1978             goto abort_cmd;
1979         }
1980         return true;
1981     }
1982 
1983 abort_cmd:
1984     ide_abort_command(s);
1985     return true;
1986 }
1987 
1988 #define HD_OK (1u << IDE_HD)
1989 #define CD_OK (1u << IDE_CD)
1990 #define CFA_OK (1u << IDE_CFATA)
1991 #define HD_CFA_OK (HD_OK | CFA_OK)
1992 #define ALL_OK (HD_OK | CD_OK | CFA_OK)
1993 
1994 /* Set the Disk Seek Completed status bit during completion */
1995 #define SET_DSC (1u << 8)
1996 
1997 /* See ACS-2 T13/2015-D Table B.2 Command codes */
1998 static const struct {
1999     /* Returns true if the completion code should be run */
2000     bool (*handler)(IDEState *s, uint8_t cmd);
2001     int flags;
2002 } ide_cmd_table[0x100] = {
2003     /* NOP not implemented, mandatory for CD */
2004     [CFA_REQ_EXT_ERROR_CODE]      = { cmd_cfa_req_ext_error_code, CFA_OK },
2005     [WIN_DSM]                     = { cmd_data_set_management, HD_CFA_OK },
2006     [WIN_DEVICE_RESET]            = { cmd_device_reset, CD_OK },
2007     [WIN_RECAL]                   = { cmd_nop, HD_CFA_OK | SET_DSC},
2008     [WIN_READ]                    = { cmd_read_pio, ALL_OK },
2009     [WIN_READ_ONCE]               = { cmd_read_pio, HD_CFA_OK },
2010     [WIN_READ_EXT]                = { cmd_read_pio, HD_CFA_OK },
2011     [WIN_READDMA_EXT]             = { cmd_read_dma, HD_CFA_OK },
2012     [WIN_READ_NATIVE_MAX_EXT]     = { cmd_read_native_max, HD_CFA_OK | SET_DSC },
2013     [WIN_MULTREAD_EXT]            = { cmd_read_multiple, HD_CFA_OK },
2014     [WIN_WRITE]                   = { cmd_write_pio, HD_CFA_OK },
2015     [WIN_WRITE_ONCE]              = { cmd_write_pio, HD_CFA_OK },
2016     [WIN_WRITE_EXT]               = { cmd_write_pio, HD_CFA_OK },
2017     [WIN_WRITEDMA_EXT]            = { cmd_write_dma, HD_CFA_OK },
2018     [CFA_WRITE_SECT_WO_ERASE]     = { cmd_write_pio, CFA_OK },
2019     [WIN_MULTWRITE_EXT]           = { cmd_write_multiple, HD_CFA_OK },
2020     [WIN_WRITE_VERIFY]            = { cmd_write_pio, HD_CFA_OK },
2021     [WIN_VERIFY]                  = { cmd_verify, HD_CFA_OK | SET_DSC },
2022     [WIN_VERIFY_ONCE]             = { cmd_verify, HD_CFA_OK | SET_DSC },
2023     [WIN_VERIFY_EXT]              = { cmd_verify, HD_CFA_OK | SET_DSC },
2024     [WIN_SEEK]                    = { cmd_seek, HD_CFA_OK | SET_DSC },
2025     [CFA_TRANSLATE_SECTOR]        = { cmd_cfa_translate_sector, CFA_OK },
2026     [WIN_DIAGNOSE]                = { cmd_exec_dev_diagnostic, ALL_OK },
2027     [WIN_SPECIFY]                 = { cmd_nop, HD_CFA_OK | SET_DSC },
2028     [WIN_STANDBYNOW2]             = { cmd_nop, HD_CFA_OK },
2029     [WIN_IDLEIMMEDIATE2]          = { cmd_nop, HD_CFA_OK },
2030     [WIN_STANDBY2]                = { cmd_nop, HD_CFA_OK },
2031     [WIN_SETIDLE2]                = { cmd_nop, HD_CFA_OK },
2032     [WIN_CHECKPOWERMODE2]         = { cmd_check_power_mode, HD_CFA_OK | SET_DSC },
2033     [WIN_SLEEPNOW2]               = { cmd_nop, HD_CFA_OK },
2034     [WIN_PACKETCMD]               = { cmd_packet, CD_OK },
2035     [WIN_PIDENTIFY]               = { cmd_identify_packet, CD_OK },
2036     [WIN_SMART]                   = { cmd_smart, HD_CFA_OK | SET_DSC },
2037     [CFA_ACCESS_METADATA_STORAGE] = { cmd_cfa_access_metadata_storage, CFA_OK },
2038     [CFA_ERASE_SECTORS]           = { cmd_cfa_erase_sectors, CFA_OK | SET_DSC },
2039     [WIN_MULTREAD]                = { cmd_read_multiple, HD_CFA_OK },
2040     [WIN_MULTWRITE]               = { cmd_write_multiple, HD_CFA_OK },
2041     [WIN_SETMULT]                 = { cmd_set_multiple_mode, HD_CFA_OK | SET_DSC },
2042     [WIN_READDMA]                 = { cmd_read_dma, HD_CFA_OK },
2043     [WIN_READDMA_ONCE]            = { cmd_read_dma, HD_CFA_OK },
2044     [WIN_WRITEDMA]                = { cmd_write_dma, HD_CFA_OK },
2045     [WIN_WRITEDMA_ONCE]           = { cmd_write_dma, HD_CFA_OK },
2046     [CFA_WRITE_MULTI_WO_ERASE]    = { cmd_write_multiple, CFA_OK },
2047     [WIN_STANDBYNOW1]             = { cmd_nop, HD_CFA_OK },
2048     [WIN_IDLEIMMEDIATE]           = { cmd_nop, HD_CFA_OK },
2049     [WIN_STANDBY]                 = { cmd_nop, HD_CFA_OK },
2050     [WIN_SETIDLE1]                = { cmd_nop, HD_CFA_OK },
2051     [WIN_CHECKPOWERMODE1]         = { cmd_check_power_mode, HD_CFA_OK | SET_DSC },
2052     [WIN_SLEEPNOW1]               = { cmd_nop, HD_CFA_OK },
2053     [WIN_FLUSH_CACHE]             = { cmd_flush_cache, ALL_OK },
2054     [WIN_FLUSH_CACHE_EXT]         = { cmd_flush_cache, HD_CFA_OK },
2055     [WIN_IDENTIFY]                = { cmd_identify, ALL_OK },
2056     [WIN_SETFEATURES]             = { cmd_set_features, ALL_OK | SET_DSC },
2057     [IBM_SENSE_CONDITION]         = { cmd_ibm_sense_condition, CFA_OK | SET_DSC },
2058     [CFA_WEAR_LEVEL]              = { cmd_cfa_erase_sectors, HD_CFA_OK | SET_DSC },
2059     [WIN_READ_NATIVE_MAX]         = { cmd_read_native_max, HD_CFA_OK | SET_DSC },
2060 };
2061 
2062 static bool ide_cmd_permitted(IDEState *s, uint32_t cmd)
2063 {
2064     return cmd < ARRAY_SIZE(ide_cmd_table)
2065         && (ide_cmd_table[cmd].flags & (1u << s->drive_kind));
2066 }
2067 
2068 void ide_exec_cmd(IDEBus *bus, uint32_t val)
2069 {
2070     IDEState *s;
2071     bool complete;
2072 
2073     s = idebus_active_if(bus);
2074     trace_ide_exec_cmd(bus, s, val);
2075 
2076     /* ignore commands to non existent slave */
2077     if (s != bus->ifs && !s->blk) {
2078         return;
2079     }
2080 
2081     /* Only RESET is allowed while BSY and/or DRQ are set,
2082      * and only to ATAPI devices. */
2083     if (s->status & (BUSY_STAT|DRQ_STAT)) {
2084         if (val != WIN_DEVICE_RESET || s->drive_kind != IDE_CD) {
2085             return;
2086         }
2087     }
2088 
2089     if (!ide_cmd_permitted(s, val)) {
2090         ide_abort_command(s);
2091         ide_set_irq(s->bus);
2092         return;
2093     }
2094 
2095     s->status = READY_STAT | BUSY_STAT;
2096     s->error = 0;
2097     s->io_buffer_offset = 0;
2098 
2099     complete = ide_cmd_table[val].handler(s, val);
2100     if (complete) {
2101         s->status &= ~BUSY_STAT;
2102         assert(!!s->error == !!(s->status & ERR_STAT));
2103 
2104         if ((ide_cmd_table[val].flags & SET_DSC) && !s->error) {
2105             s->status |= SEEK_STAT;
2106         }
2107 
2108         ide_cmd_done(s);
2109         ide_set_irq(s->bus);
2110     }
2111 }
2112 
2113 /* IOport [R]ead [R]egisters */
2114 enum ATA_IOPORT_RR {
2115     ATA_IOPORT_RR_DATA = 0,
2116     ATA_IOPORT_RR_ERROR = 1,
2117     ATA_IOPORT_RR_SECTOR_COUNT = 2,
2118     ATA_IOPORT_RR_SECTOR_NUMBER = 3,
2119     ATA_IOPORT_RR_CYLINDER_LOW = 4,
2120     ATA_IOPORT_RR_CYLINDER_HIGH = 5,
2121     ATA_IOPORT_RR_DEVICE_HEAD = 6,
2122     ATA_IOPORT_RR_STATUS = 7,
2123     ATA_IOPORT_RR_NUM_REGISTERS,
2124 };
2125 
2126 const char *ATA_IOPORT_RR_lookup[ATA_IOPORT_RR_NUM_REGISTERS] = {
2127     [ATA_IOPORT_RR_DATA] = "Data",
2128     [ATA_IOPORT_RR_ERROR] = "Error",
2129     [ATA_IOPORT_RR_SECTOR_COUNT] = "Sector Count",
2130     [ATA_IOPORT_RR_SECTOR_NUMBER] = "Sector Number",
2131     [ATA_IOPORT_RR_CYLINDER_LOW] = "Cylinder Low",
2132     [ATA_IOPORT_RR_CYLINDER_HIGH] = "Cylinder High",
2133     [ATA_IOPORT_RR_DEVICE_HEAD] = "Device/Head",
2134     [ATA_IOPORT_RR_STATUS] = "Status"
2135 };
2136 
2137 uint32_t ide_ioport_read(void *opaque, uint32_t addr)
2138 {
2139     IDEBus *bus = opaque;
2140     IDEState *s = idebus_active_if(bus);
2141     uint32_t reg_num;
2142     int ret, hob;
2143 
2144     reg_num = addr & 7;
2145     /* FIXME: HOB readback uses bit 7, but it's always set right now */
2146     //hob = s->select & (1 << 7);
2147     hob = 0;
2148     switch (reg_num) {
2149     case ATA_IOPORT_RR_DATA:
2150         ret = 0xff;
2151         break;
2152     case ATA_IOPORT_RR_ERROR:
2153         if ((!bus->ifs[0].blk && !bus->ifs[1].blk) ||
2154             (s != bus->ifs && !s->blk)) {
2155             ret = 0;
2156         } else if (!hob) {
2157             ret = s->error;
2158         } else {
2159             ret = s->hob_feature;
2160         }
2161         break;
2162     case ATA_IOPORT_RR_SECTOR_COUNT:
2163         if (!bus->ifs[0].blk && !bus->ifs[1].blk) {
2164             ret = 0;
2165         } else if (!hob) {
2166             ret = s->nsector & 0xff;
2167         } else {
2168             ret = s->hob_nsector;
2169         }
2170         break;
2171     case ATA_IOPORT_RR_SECTOR_NUMBER:
2172         if (!bus->ifs[0].blk && !bus->ifs[1].blk) {
2173             ret = 0;
2174         } else if (!hob) {
2175             ret = s->sector;
2176         } else {
2177             ret = s->hob_sector;
2178         }
2179         break;
2180     case ATA_IOPORT_RR_CYLINDER_LOW:
2181         if (!bus->ifs[0].blk && !bus->ifs[1].blk) {
2182             ret = 0;
2183         } else if (!hob) {
2184             ret = s->lcyl;
2185         } else {
2186             ret = s->hob_lcyl;
2187         }
2188         break;
2189     case ATA_IOPORT_RR_CYLINDER_HIGH:
2190         if (!bus->ifs[0].blk && !bus->ifs[1].blk) {
2191             ret = 0;
2192         } else if (!hob) {
2193             ret = s->hcyl;
2194         } else {
2195             ret = s->hob_hcyl;
2196         }
2197         break;
2198     case ATA_IOPORT_RR_DEVICE_HEAD:
2199         if (!bus->ifs[0].blk && !bus->ifs[1].blk) {
2200             ret = 0;
2201         } else {
2202             ret = s->select;
2203         }
2204         break;
2205     default:
2206     case ATA_IOPORT_RR_STATUS:
2207         if ((!bus->ifs[0].blk && !bus->ifs[1].blk) ||
2208             (s != bus->ifs && !s->blk)) {
2209             ret = 0;
2210         } else {
2211             ret = s->status;
2212         }
2213         qemu_irq_lower(bus->irq);
2214         break;
2215     }
2216 
2217     trace_ide_ioport_read(addr, ATA_IOPORT_RR_lookup[reg_num], ret, bus, s);
2218     return ret;
2219 }
2220 
2221 uint32_t ide_status_read(void *opaque, uint32_t addr)
2222 {
2223     IDEBus *bus = opaque;
2224     IDEState *s = idebus_active_if(bus);
2225     int ret;
2226 
2227     if ((!bus->ifs[0].blk && !bus->ifs[1].blk) ||
2228         (s != bus->ifs && !s->blk)) {
2229         ret = 0;
2230     } else {
2231         ret = s->status;
2232     }
2233 
2234     trace_ide_status_read(addr, ret, bus, s);
2235     return ret;
2236 }
2237 
2238 void ide_cmd_write(void *opaque, uint32_t addr, uint32_t val)
2239 {
2240     IDEBus *bus = opaque;
2241     IDEState *s;
2242     int i;
2243 
2244     trace_ide_cmd_write(addr, val, bus);
2245 
2246     /* common for both drives */
2247     if (!(bus->cmd & IDE_CMD_RESET) &&
2248         (val & IDE_CMD_RESET)) {
2249         /* reset low to high */
2250         for(i = 0;i < 2; i++) {
2251             s = &bus->ifs[i];
2252             s->status = BUSY_STAT | SEEK_STAT;
2253             s->error = 0x01;
2254         }
2255     } else if ((bus->cmd & IDE_CMD_RESET) &&
2256                !(val & IDE_CMD_RESET)) {
2257         /* high to low */
2258         for(i = 0;i < 2; i++) {
2259             s = &bus->ifs[i];
2260             if (s->drive_kind == IDE_CD)
2261                 s->status = 0x00; /* NOTE: READY is _not_ set */
2262             else
2263                 s->status = READY_STAT | SEEK_STAT;
2264             ide_set_signature(s);
2265         }
2266     }
2267 
2268     bus->cmd = val;
2269 }
2270 
2271 /*
2272  * Returns true if the running PIO transfer is a PIO out (i.e. data is
2273  * transferred from the device to the guest), false if it's a PIO in
2274  */
2275 static bool ide_is_pio_out(IDEState *s)
2276 {
2277     if (s->end_transfer_func == ide_sector_write ||
2278         s->end_transfer_func == ide_atapi_cmd) {
2279         return false;
2280     } else if (s->end_transfer_func == ide_sector_read ||
2281                s->end_transfer_func == ide_transfer_stop ||
2282                s->end_transfer_func == ide_atapi_cmd_reply_end ||
2283                s->end_transfer_func == ide_dummy_transfer_stop) {
2284         return true;
2285     }
2286 
2287     abort();
2288 }
2289 
2290 void ide_data_writew(void *opaque, uint32_t addr, uint32_t val)
2291 {
2292     IDEBus *bus = opaque;
2293     IDEState *s = idebus_active_if(bus);
2294     uint8_t *p;
2295 
2296     trace_ide_data_writew(addr, val, bus, s);
2297 
2298     /* PIO data access allowed only when DRQ bit is set. The result of a write
2299      * during PIO out is indeterminate, just ignore it. */
2300     if (!(s->status & DRQ_STAT) || ide_is_pio_out(s)) {
2301         return;
2302     }
2303 
2304     p = s->data_ptr;
2305     if (p + 2 > s->data_end) {
2306         return;
2307     }
2308 
2309     *(uint16_t *)p = le16_to_cpu(val);
2310     p += 2;
2311     s->data_ptr = p;
2312     if (p >= s->data_end) {
2313         s->status &= ~DRQ_STAT;
2314         s->end_transfer_func(s);
2315     }
2316 }
2317 
2318 uint32_t ide_data_readw(void *opaque, uint32_t addr)
2319 {
2320     IDEBus *bus = opaque;
2321     IDEState *s = idebus_active_if(bus);
2322     uint8_t *p;
2323     int ret;
2324 
2325     /* PIO data access allowed only when DRQ bit is set. The result of a read
2326      * during PIO in is indeterminate, return 0 and don't move forward. */
2327     if (!(s->status & DRQ_STAT) || !ide_is_pio_out(s)) {
2328         return 0;
2329     }
2330 
2331     p = s->data_ptr;
2332     if (p + 2 > s->data_end) {
2333         return 0;
2334     }
2335 
2336     ret = cpu_to_le16(*(uint16_t *)p);
2337     p += 2;
2338     s->data_ptr = p;
2339     if (p >= s->data_end) {
2340         s->status &= ~DRQ_STAT;
2341         s->end_transfer_func(s);
2342     }
2343 
2344     trace_ide_data_readw(addr, ret, bus, s);
2345     return ret;
2346 }
2347 
2348 void ide_data_writel(void *opaque, uint32_t addr, uint32_t val)
2349 {
2350     IDEBus *bus = opaque;
2351     IDEState *s = idebus_active_if(bus);
2352     uint8_t *p;
2353 
2354     trace_ide_data_writel(addr, val, bus, s);
2355 
2356     /* PIO data access allowed only when DRQ bit is set. The result of a write
2357      * during PIO out is indeterminate, just ignore it. */
2358     if (!(s->status & DRQ_STAT) || ide_is_pio_out(s)) {
2359         return;
2360     }
2361 
2362     p = s->data_ptr;
2363     if (p + 4 > s->data_end) {
2364         return;
2365     }
2366 
2367     *(uint32_t *)p = le32_to_cpu(val);
2368     p += 4;
2369     s->data_ptr = p;
2370     if (p >= s->data_end) {
2371         s->status &= ~DRQ_STAT;
2372         s->end_transfer_func(s);
2373     }
2374 }
2375 
2376 uint32_t ide_data_readl(void *opaque, uint32_t addr)
2377 {
2378     IDEBus *bus = opaque;
2379     IDEState *s = idebus_active_if(bus);
2380     uint8_t *p;
2381     int ret;
2382 
2383     /* PIO data access allowed only when DRQ bit is set. The result of a read
2384      * during PIO in is indeterminate, return 0 and don't move forward. */
2385     if (!(s->status & DRQ_STAT) || !ide_is_pio_out(s)) {
2386         ret = 0;
2387         goto out;
2388     }
2389 
2390     p = s->data_ptr;
2391     if (p + 4 > s->data_end) {
2392         return 0;
2393     }
2394 
2395     ret = cpu_to_le32(*(uint32_t *)p);
2396     p += 4;
2397     s->data_ptr = p;
2398     if (p >= s->data_end) {
2399         s->status &= ~DRQ_STAT;
2400         s->end_transfer_func(s);
2401     }
2402 
2403 out:
2404     trace_ide_data_readl(addr, ret, bus, s);
2405     return ret;
2406 }
2407 
2408 static void ide_dummy_transfer_stop(IDEState *s)
2409 {
2410     s->data_ptr = s->io_buffer;
2411     s->data_end = s->io_buffer;
2412     s->io_buffer[0] = 0xff;
2413     s->io_buffer[1] = 0xff;
2414     s->io_buffer[2] = 0xff;
2415     s->io_buffer[3] = 0xff;
2416 }
2417 
2418 void ide_bus_reset(IDEBus *bus)
2419 {
2420     bus->unit = 0;
2421     bus->cmd = 0;
2422     ide_reset(&bus->ifs[0]);
2423     ide_reset(&bus->ifs[1]);
2424     ide_clear_hob(bus);
2425 
2426     /* pending async DMA */
2427     if (bus->dma->aiocb) {
2428         trace_ide_bus_reset_aio();
2429         blk_aio_cancel(bus->dma->aiocb);
2430         bus->dma->aiocb = NULL;
2431     }
2432 
2433     /* reset dma provider too */
2434     if (bus->dma->ops->reset) {
2435         bus->dma->ops->reset(bus->dma);
2436     }
2437 }
2438 
2439 static bool ide_cd_is_tray_open(void *opaque)
2440 {
2441     return ((IDEState *)opaque)->tray_open;
2442 }
2443 
2444 static bool ide_cd_is_medium_locked(void *opaque)
2445 {
2446     return ((IDEState *)opaque)->tray_locked;
2447 }
2448 
2449 static void ide_resize_cb(void *opaque)
2450 {
2451     IDEState *s = opaque;
2452     uint64_t nb_sectors;
2453 
2454     if (!s->identify_set) {
2455         return;
2456     }
2457 
2458     blk_get_geometry(s->blk, &nb_sectors);
2459     s->nb_sectors = nb_sectors;
2460 
2461     /* Update the identify data buffer. */
2462     if (s->drive_kind == IDE_CFATA) {
2463         ide_cfata_identify_size(s);
2464     } else {
2465         /* IDE_CD uses a different set of callbacks entirely. */
2466         assert(s->drive_kind != IDE_CD);
2467         ide_identify_size(s);
2468     }
2469 }
2470 
2471 static const BlockDevOps ide_cd_block_ops = {
2472     .change_media_cb = ide_cd_change_cb,
2473     .eject_request_cb = ide_cd_eject_request_cb,
2474     .is_tray_open = ide_cd_is_tray_open,
2475     .is_medium_locked = ide_cd_is_medium_locked,
2476 };
2477 
2478 static const BlockDevOps ide_hd_block_ops = {
2479     .resize_cb = ide_resize_cb,
2480 };
2481 
2482 int ide_init_drive(IDEState *s, BlockBackend *blk, IDEDriveKind kind,
2483                    const char *version, const char *serial, const char *model,
2484                    uint64_t wwn,
2485                    uint32_t cylinders, uint32_t heads, uint32_t secs,
2486                    int chs_trans, Error **errp)
2487 {
2488     uint64_t nb_sectors;
2489 
2490     s->blk = blk;
2491     s->drive_kind = kind;
2492 
2493     blk_get_geometry(blk, &nb_sectors);
2494     s->cylinders = cylinders;
2495     s->heads = heads;
2496     s->sectors = secs;
2497     s->chs_trans = chs_trans;
2498     s->nb_sectors = nb_sectors;
2499     s->wwn = wwn;
2500     /* The SMART values should be preserved across power cycles
2501        but they aren't.  */
2502     s->smart_enabled = 1;
2503     s->smart_autosave = 1;
2504     s->smart_errors = 0;
2505     s->smart_selftest_count = 0;
2506     if (kind == IDE_CD) {
2507         blk_set_dev_ops(blk, &ide_cd_block_ops, s);
2508         blk_set_guest_block_size(blk, 2048);
2509     } else {
2510         if (!blk_is_inserted(s->blk)) {
2511             error_setg(errp, "Device needs media, but drive is empty");
2512             return -1;
2513         }
2514         if (blk_is_read_only(blk)) {
2515             error_setg(errp, "Can't use a read-only drive");
2516             return -1;
2517         }
2518         blk_set_dev_ops(blk, &ide_hd_block_ops, s);
2519     }
2520     if (serial) {
2521         pstrcpy(s->drive_serial_str, sizeof(s->drive_serial_str), serial);
2522     } else {
2523         snprintf(s->drive_serial_str, sizeof(s->drive_serial_str),
2524                  "QM%05d", s->drive_serial);
2525     }
2526     if (model) {
2527         pstrcpy(s->drive_model_str, sizeof(s->drive_model_str), model);
2528     } else {
2529         switch (kind) {
2530         case IDE_CD:
2531             strcpy(s->drive_model_str, "QEMU DVD-ROM");
2532             break;
2533         case IDE_CFATA:
2534             strcpy(s->drive_model_str, "QEMU MICRODRIVE");
2535             break;
2536         default:
2537             strcpy(s->drive_model_str, "QEMU HARDDISK");
2538             break;
2539         }
2540     }
2541 
2542     if (version) {
2543         pstrcpy(s->version, sizeof(s->version), version);
2544     } else {
2545         pstrcpy(s->version, sizeof(s->version), qemu_hw_version());
2546     }
2547 
2548     ide_reset(s);
2549     blk_iostatus_enable(blk);
2550     return 0;
2551 }
2552 
2553 static void ide_init1(IDEBus *bus, int unit)
2554 {
2555     static int drive_serial = 1;
2556     IDEState *s = &bus->ifs[unit];
2557 
2558     s->bus = bus;
2559     s->unit = unit;
2560     s->drive_serial = drive_serial++;
2561     /* we need at least 2k alignment for accessing CDROMs using O_DIRECT */
2562     s->io_buffer_total_len = IDE_DMA_BUF_SECTORS*512 + 4;
2563     s->io_buffer = qemu_memalign(2048, s->io_buffer_total_len);
2564     memset(s->io_buffer, 0, s->io_buffer_total_len);
2565 
2566     s->smart_selftest_data = blk_blockalign(s->blk, 512);
2567     memset(s->smart_selftest_data, 0, 512);
2568 
2569     s->sector_write_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
2570                                            ide_sector_write_timer_cb, s);
2571 }
2572 
2573 static int ide_nop_int(const IDEDMA *dma, bool is_write)
2574 {
2575     return 0;
2576 }
2577 
2578 static void ide_nop(const IDEDMA *dma)
2579 {
2580 }
2581 
2582 static int32_t ide_nop_int32(const IDEDMA *dma, int32_t l)
2583 {
2584     return 0;
2585 }
2586 
2587 static const IDEDMAOps ide_dma_nop_ops = {
2588     .prepare_buf    = ide_nop_int32,
2589     .restart_dma    = ide_nop,
2590     .rw_buf         = ide_nop_int,
2591 };
2592 
2593 static void ide_restart_dma(IDEState *s, enum ide_dma_cmd dma_cmd)
2594 {
2595     s->unit = s->bus->retry_unit;
2596     ide_set_sector(s, s->bus->retry_sector_num);
2597     s->nsector = s->bus->retry_nsector;
2598     s->bus->dma->ops->restart_dma(s->bus->dma);
2599     s->io_buffer_size = 0;
2600     s->dma_cmd = dma_cmd;
2601     ide_start_dma(s, ide_dma_cb);
2602 }
2603 
2604 static void ide_restart_bh(void *opaque)
2605 {
2606     IDEBus *bus = opaque;
2607     IDEState *s;
2608     bool is_read;
2609     int error_status;
2610 
2611     qemu_bh_delete(bus->bh);
2612     bus->bh = NULL;
2613 
2614     error_status = bus->error_status;
2615     if (bus->error_status == 0) {
2616         return;
2617     }
2618 
2619     s = idebus_active_if(bus);
2620     is_read = (bus->error_status & IDE_RETRY_READ) != 0;
2621 
2622     /* The error status must be cleared before resubmitting the request: The
2623      * request may fail again, and this case can only be distinguished if the
2624      * called function can set a new error status. */
2625     bus->error_status = 0;
2626 
2627     /* The HBA has generically asked to be kicked on retry */
2628     if (error_status & IDE_RETRY_HBA) {
2629         if (s->bus->dma->ops->restart) {
2630             s->bus->dma->ops->restart(s->bus->dma);
2631         }
2632     } else if (IS_IDE_RETRY_DMA(error_status)) {
2633         if (error_status & IDE_RETRY_TRIM) {
2634             ide_restart_dma(s, IDE_DMA_TRIM);
2635         } else {
2636             ide_restart_dma(s, is_read ? IDE_DMA_READ : IDE_DMA_WRITE);
2637         }
2638     } else if (IS_IDE_RETRY_PIO(error_status)) {
2639         if (is_read) {
2640             ide_sector_read(s);
2641         } else {
2642             ide_sector_write(s);
2643         }
2644     } else if (error_status & IDE_RETRY_FLUSH) {
2645         ide_flush_cache(s);
2646     } else if (IS_IDE_RETRY_ATAPI(error_status)) {
2647         assert(s->end_transfer_func == ide_atapi_cmd);
2648         ide_atapi_dma_restart(s);
2649     } else {
2650         abort();
2651     }
2652 }
2653 
2654 static void ide_restart_cb(void *opaque, int running, RunState state)
2655 {
2656     IDEBus *bus = opaque;
2657 
2658     if (!running)
2659         return;
2660 
2661     if (!bus->bh) {
2662         bus->bh = qemu_bh_new(ide_restart_bh, bus);
2663         qemu_bh_schedule(bus->bh);
2664     }
2665 }
2666 
2667 void ide_register_restart_cb(IDEBus *bus)
2668 {
2669     if (bus->dma->ops->restart_dma) {
2670         bus->vmstate = qemu_add_vm_change_state_handler(ide_restart_cb, bus);
2671     }
2672 }
2673 
2674 static IDEDMA ide_dma_nop = {
2675     .ops = &ide_dma_nop_ops,
2676     .aiocb = NULL,
2677 };
2678 
2679 void ide_init2(IDEBus *bus, qemu_irq irq)
2680 {
2681     int i;
2682 
2683     for(i = 0; i < 2; i++) {
2684         ide_init1(bus, i);
2685         ide_reset(&bus->ifs[i]);
2686     }
2687     bus->irq = irq;
2688     bus->dma = &ide_dma_nop;
2689 }
2690 
2691 void ide_exit(IDEState *s)
2692 {
2693     timer_del(s->sector_write_timer);
2694     timer_free(s->sector_write_timer);
2695     qemu_vfree(s->smart_selftest_data);
2696     qemu_vfree(s->io_buffer);
2697 }
2698 
2699 static bool is_identify_set(void *opaque, int version_id)
2700 {
2701     IDEState *s = opaque;
2702 
2703     return s->identify_set != 0;
2704 }
2705 
2706 static EndTransferFunc* transfer_end_table[] = {
2707         ide_sector_read,
2708         ide_sector_write,
2709         ide_transfer_stop,
2710         ide_atapi_cmd_reply_end,
2711         ide_atapi_cmd,
2712         ide_dummy_transfer_stop,
2713 };
2714 
2715 static int transfer_end_table_idx(EndTransferFunc *fn)
2716 {
2717     int i;
2718 
2719     for (i = 0; i < ARRAY_SIZE(transfer_end_table); i++)
2720         if (transfer_end_table[i] == fn)
2721             return i;
2722 
2723     return -1;
2724 }
2725 
2726 static int ide_drive_post_load(void *opaque, int version_id)
2727 {
2728     IDEState *s = opaque;
2729 
2730     if (s->blk && s->identify_set) {
2731         blk_set_enable_write_cache(s->blk, !!(s->identify_data[85] & (1 << 5)));
2732     }
2733     return 0;
2734 }
2735 
2736 static int ide_drive_pio_post_load(void *opaque, int version_id)
2737 {
2738     IDEState *s = opaque;
2739 
2740     if (s->end_transfer_fn_idx >= ARRAY_SIZE(transfer_end_table)) {
2741         return -EINVAL;
2742     }
2743     s->end_transfer_func = transfer_end_table[s->end_transfer_fn_idx];
2744     s->data_ptr = s->io_buffer + s->cur_io_buffer_offset;
2745     s->data_end = s->data_ptr + s->cur_io_buffer_len;
2746     s->atapi_dma = s->feature & 1; /* as per cmd_packet */
2747 
2748     return 0;
2749 }
2750 
2751 static int ide_drive_pio_pre_save(void *opaque)
2752 {
2753     IDEState *s = opaque;
2754     int idx;
2755 
2756     s->cur_io_buffer_offset = s->data_ptr - s->io_buffer;
2757     s->cur_io_buffer_len = s->data_end - s->data_ptr;
2758 
2759     idx = transfer_end_table_idx(s->end_transfer_func);
2760     if (idx == -1) {
2761         fprintf(stderr, "%s: invalid end_transfer_func for DRQ_STAT\n",
2762                         __func__);
2763         s->end_transfer_fn_idx = 2;
2764     } else {
2765         s->end_transfer_fn_idx = idx;
2766     }
2767 
2768     return 0;
2769 }
2770 
2771 static bool ide_drive_pio_state_needed(void *opaque)
2772 {
2773     IDEState *s = opaque;
2774 
2775     return ((s->status & DRQ_STAT) != 0)
2776         || (s->bus->error_status & IDE_RETRY_PIO);
2777 }
2778 
2779 static bool ide_tray_state_needed(void *opaque)
2780 {
2781     IDEState *s = opaque;
2782 
2783     return s->tray_open || s->tray_locked;
2784 }
2785 
2786 static bool ide_atapi_gesn_needed(void *opaque)
2787 {
2788     IDEState *s = opaque;
2789 
2790     return s->events.new_media || s->events.eject_request;
2791 }
2792 
2793 static bool ide_error_needed(void *opaque)
2794 {
2795     IDEBus *bus = opaque;
2796 
2797     return (bus->error_status != 0);
2798 }
2799 
2800 /* Fields for GET_EVENT_STATUS_NOTIFICATION ATAPI command */
2801 static const VMStateDescription vmstate_ide_atapi_gesn_state = {
2802     .name ="ide_drive/atapi/gesn_state",
2803     .version_id = 1,
2804     .minimum_version_id = 1,
2805     .needed = ide_atapi_gesn_needed,
2806     .fields = (VMStateField[]) {
2807         VMSTATE_BOOL(events.new_media, IDEState),
2808         VMSTATE_BOOL(events.eject_request, IDEState),
2809         VMSTATE_END_OF_LIST()
2810     }
2811 };
2812 
2813 static const VMStateDescription vmstate_ide_tray_state = {
2814     .name = "ide_drive/tray_state",
2815     .version_id = 1,
2816     .minimum_version_id = 1,
2817     .needed = ide_tray_state_needed,
2818     .fields = (VMStateField[]) {
2819         VMSTATE_BOOL(tray_open, IDEState),
2820         VMSTATE_BOOL(tray_locked, IDEState),
2821         VMSTATE_END_OF_LIST()
2822     }
2823 };
2824 
2825 static const VMStateDescription vmstate_ide_drive_pio_state = {
2826     .name = "ide_drive/pio_state",
2827     .version_id = 1,
2828     .minimum_version_id = 1,
2829     .pre_save = ide_drive_pio_pre_save,
2830     .post_load = ide_drive_pio_post_load,
2831     .needed = ide_drive_pio_state_needed,
2832     .fields = (VMStateField[]) {
2833         VMSTATE_INT32(req_nb_sectors, IDEState),
2834         VMSTATE_VARRAY_INT32(io_buffer, IDEState, io_buffer_total_len, 1,
2835                              vmstate_info_uint8, uint8_t),
2836         VMSTATE_INT32(cur_io_buffer_offset, IDEState),
2837         VMSTATE_INT32(cur_io_buffer_len, IDEState),
2838         VMSTATE_UINT8(end_transfer_fn_idx, IDEState),
2839         VMSTATE_INT32(elementary_transfer_size, IDEState),
2840         VMSTATE_INT32(packet_transfer_size, IDEState),
2841         VMSTATE_END_OF_LIST()
2842     }
2843 };
2844 
2845 const VMStateDescription vmstate_ide_drive = {
2846     .name = "ide_drive",
2847     .version_id = 3,
2848     .minimum_version_id = 0,
2849     .post_load = ide_drive_post_load,
2850     .fields = (VMStateField[]) {
2851         VMSTATE_INT32(mult_sectors, IDEState),
2852         VMSTATE_INT32(identify_set, IDEState),
2853         VMSTATE_BUFFER_TEST(identify_data, IDEState, is_identify_set),
2854         VMSTATE_UINT8(feature, IDEState),
2855         VMSTATE_UINT8(error, IDEState),
2856         VMSTATE_UINT32(nsector, IDEState),
2857         VMSTATE_UINT8(sector, IDEState),
2858         VMSTATE_UINT8(lcyl, IDEState),
2859         VMSTATE_UINT8(hcyl, IDEState),
2860         VMSTATE_UINT8(hob_feature, IDEState),
2861         VMSTATE_UINT8(hob_sector, IDEState),
2862         VMSTATE_UINT8(hob_nsector, IDEState),
2863         VMSTATE_UINT8(hob_lcyl, IDEState),
2864         VMSTATE_UINT8(hob_hcyl, IDEState),
2865         VMSTATE_UINT8(select, IDEState),
2866         VMSTATE_UINT8(status, IDEState),
2867         VMSTATE_UINT8(lba48, IDEState),
2868         VMSTATE_UINT8(sense_key, IDEState),
2869         VMSTATE_UINT8(asc, IDEState),
2870         VMSTATE_UINT8_V(cdrom_changed, IDEState, 3),
2871         VMSTATE_END_OF_LIST()
2872     },
2873     .subsections = (const VMStateDescription*[]) {
2874         &vmstate_ide_drive_pio_state,
2875         &vmstate_ide_tray_state,
2876         &vmstate_ide_atapi_gesn_state,
2877         NULL
2878     }
2879 };
2880 
2881 static const VMStateDescription vmstate_ide_error_status = {
2882     .name ="ide_bus/error",
2883     .version_id = 2,
2884     .minimum_version_id = 1,
2885     .needed = ide_error_needed,
2886     .fields = (VMStateField[]) {
2887         VMSTATE_INT32(error_status, IDEBus),
2888         VMSTATE_INT64_V(retry_sector_num, IDEBus, 2),
2889         VMSTATE_UINT32_V(retry_nsector, IDEBus, 2),
2890         VMSTATE_UINT8_V(retry_unit, IDEBus, 2),
2891         VMSTATE_END_OF_LIST()
2892     }
2893 };
2894 
2895 const VMStateDescription vmstate_ide_bus = {
2896     .name = "ide_bus",
2897     .version_id = 1,
2898     .minimum_version_id = 1,
2899     .fields = (VMStateField[]) {
2900         VMSTATE_UINT8(cmd, IDEBus),
2901         VMSTATE_UINT8(unit, IDEBus),
2902         VMSTATE_END_OF_LIST()
2903     },
2904     .subsections = (const VMStateDescription*[]) {
2905         &vmstate_ide_error_status,
2906         NULL
2907     }
2908 };
2909 
2910 void ide_drive_get(DriveInfo **hd, int n)
2911 {
2912     int i;
2913 
2914     for (i = 0; i < n; i++) {
2915         hd[i] = drive_get_by_index(IF_IDE, i);
2916     }
2917 }
2918