1/*-
2 * Copyright (c) 2012 Alexander Motin <mav@FreeBSD.org>
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 */
26
27#include <sys/cdefs.h>
28__FBSDID("$FreeBSD: stable/10/sys/geom/raid/tr_raid5.c 326745 2017-12-10 13:45:41Z eugen $");
29
30#include <sys/param.h>
31#include <sys/bio.h>
32#include <sys/endian.h>
33#include <sys/kernel.h>
34#include <sys/kobj.h>
35#include <sys/limits.h>
36#include <sys/lock.h>
37#include <sys/malloc.h>
38#include <sys/mutex.h>
39#include <sys/sysctl.h>
40#include <sys/systm.h>
41#include <geom/geom.h>
42#include "geom/raid/g_raid.h"
43#include "g_raid_tr_if.h"
44
45static MALLOC_DEFINE(M_TR_RAID5, "tr_raid5_data", "GEOM_RAID RAID5 data");
46
47#define TR_RAID5_NONE 0
48#define TR_RAID5_REBUILD 1
49#define TR_RAID5_RESYNC 2
50
51#define TR_RAID5_F_DOING_SOME	0x1
52#define TR_RAID5_F_LOCKED	0x2
53#define TR_RAID5_F_ABORT	0x4
54
55struct g_raid_tr_raid5_object {
56	struct g_raid_tr_object	 trso_base;
57	int			 trso_starting;
58	int			 trso_stopping;
59	int			 trso_type;
60	int			 trso_recover_slabs; /* slabs before rest */
61	int			 trso_fair_io;
62	int			 trso_meta_update;
63	int			 trso_flags;
64	struct g_raid_subdisk	*trso_failed_sd; /* like per volume */
65	void			*trso_buffer;	 /* Buffer space */
66	struct bio		 trso_bio;
67};
68
69static g_raid_tr_taste_t g_raid_tr_taste_raid5;
70static g_raid_tr_event_t g_raid_tr_event_raid5;
71static g_raid_tr_start_t g_raid_tr_start_raid5;
72static g_raid_tr_stop_t g_raid_tr_stop_raid5;
73static g_raid_tr_iostart_t g_raid_tr_iostart_raid5;
74static g_raid_tr_iodone_t g_raid_tr_iodone_raid5;
75static g_raid_tr_kerneldump_t g_raid_tr_kerneldump_raid5;
76static g_raid_tr_locked_t g_raid_tr_locked_raid5;
77static g_raid_tr_free_t g_raid_tr_free_raid5;
78
79static kobj_method_t g_raid_tr_raid5_methods[] = {
80	KOBJMETHOD(g_raid_tr_taste,	g_raid_tr_taste_raid5),
81	KOBJMETHOD(g_raid_tr_event,	g_raid_tr_event_raid5),
82	KOBJMETHOD(g_raid_tr_start,	g_raid_tr_start_raid5),
83	KOBJMETHOD(g_raid_tr_stop,	g_raid_tr_stop_raid5),
84	KOBJMETHOD(g_raid_tr_iostart,	g_raid_tr_iostart_raid5),
85	KOBJMETHOD(g_raid_tr_iodone,	g_raid_tr_iodone_raid5),
86	KOBJMETHOD(g_raid_tr_kerneldump, g_raid_tr_kerneldump_raid5),
87	KOBJMETHOD(g_raid_tr_locked,	g_raid_tr_locked_raid5),
88	KOBJMETHOD(g_raid_tr_free,	g_raid_tr_free_raid5),
89	{ 0, 0 }
90};
91
92static struct g_raid_tr_class g_raid_tr_raid5_class = {
93	"RAID5",
94	g_raid_tr_raid5_methods,
95	sizeof(struct g_raid_tr_raid5_object),
96	.trc_enable = 1,
97	.trc_priority = 100
98};
99
100static int
101g_raid_tr_taste_raid5(struct g_raid_tr_object *tr, struct g_raid_volume *vol)
102{
103	struct g_raid_tr_raid5_object *trs;
104	u_int qual;
105
106	trs = (struct g_raid_tr_raid5_object *)tr;
107	qual = tr->tro_volume->v_raid_level_qualifier;
108	if (tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAID4 &&
109	    (qual == G_RAID_VOLUME_RLQ_R4P0 ||
110	     qual == G_RAID_VOLUME_RLQ_R4PN)) {
111		/* RAID4 */
112	} else if ((tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAID5 ||
113	     tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAID5E ||
114	     tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAID5EE ||
115	     tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAID5R ||
116	     tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAID6 ||
117	     tr->tro_volume->v_raid_level == G_RAID_VOLUME_RL_RAIDMDF) &&
118	    (qual == G_RAID_VOLUME_RLQ_R5RA ||
119	     qual == G_RAID_VOLUME_RLQ_R5RS ||
120	     qual == G_RAID_VOLUME_RLQ_R5LA ||
121	     qual == G_RAID_VOLUME_RLQ_R5LS)) {
122		/* RAID5/5E/5EE/5R/6/MDF */
123	} else
124		return (G_RAID_TR_TASTE_FAIL);
125	trs->trso_starting = 1;
126	return (G_RAID_TR_TASTE_SUCCEED);
127}
128
129static int
130g_raid_tr_update_state_raid5(struct g_raid_volume *vol,
131    struct g_raid_subdisk *sd)
132{
133	struct g_raid_tr_raid5_object *trs;
134	struct g_raid_softc *sc;
135	u_int s;
136	int na, ns, nu;
137
138	sc = vol->v_softc;
139	trs = (struct g_raid_tr_raid5_object *)vol->v_tr;
140	if (trs->trso_stopping &&
141	    (trs->trso_flags & TR_RAID5_F_DOING_SOME) == 0)
142		s = G_RAID_VOLUME_S_STOPPED;
143	else if (trs->trso_starting)
144		s = G_RAID_VOLUME_S_STARTING;
145	else {
146		na = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_ACTIVE);
147		ns = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_STALE) +
148		     g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_RESYNC);
149		nu = g_raid_nsubdisks(vol, G_RAID_SUBDISK_S_UNINITIALIZED);
150		if (na == vol->v_disks_count)
151			s = G_RAID_VOLUME_S_OPTIMAL;
152		else if (na + ns == vol->v_disks_count ||
153		    na + ns + nu == vol->v_disks_count /* XXX: Temporary. */)
154			s = G_RAID_VOLUME_S_SUBOPTIMAL;
155		else if (na == vol->v_disks_count - 1 ||
156		    na + ns + nu == vol->v_disks_count)
157			s = G_RAID_VOLUME_S_DEGRADED;
158		else
159			s = G_RAID_VOLUME_S_BROKEN;
160	}
161	if (s != vol->v_state) {
162		g_raid_event_send(vol, G_RAID_VOLUME_S_ALIVE(s) ?
163		    G_RAID_VOLUME_E_UP : G_RAID_VOLUME_E_DOWN,
164		    G_RAID_EVENT_VOLUME);
165		g_raid_change_volume_state(vol, s);
166		if (!trs->trso_starting && !trs->trso_stopping)
167			g_raid_write_metadata(sc, vol, NULL, NULL);
168	}
169	return (0);
170}
171
172static int
173g_raid_tr_event_raid5(struct g_raid_tr_object *tr,
174    struct g_raid_subdisk *sd, u_int event)
175{
176
177	g_raid_tr_update_state_raid5(tr->tro_volume, sd);
178	return (0);
179}
180
181static int
182g_raid_tr_start_raid5(struct g_raid_tr_object *tr)
183{
184	struct g_raid_tr_raid5_object *trs;
185	struct g_raid_volume *vol;
186
187	trs = (struct g_raid_tr_raid5_object *)tr;
188	trs->trso_starting = 0;
189	vol = tr->tro_volume;
190	vol->v_read_only = 1;
191	g_raid_tr_update_state_raid5(vol, NULL);
192	return (0);
193}
194
195static int
196g_raid_tr_stop_raid5(struct g_raid_tr_object *tr)
197{
198	struct g_raid_tr_raid5_object *trs;
199	struct g_raid_volume *vol;
200
201	trs = (struct g_raid_tr_raid5_object *)tr;
202	vol = tr->tro_volume;
203	trs->trso_starting = 0;
204	trs->trso_stopping = 1;
205	g_raid_tr_update_state_raid5(vol, NULL);
206	return (0);
207}
208
209static void
210g_raid_tr_iostart_raid5_read(struct g_raid_tr_object *tr, struct bio *bp)
211{
212	struct g_raid_volume *vol;
213	struct g_raid_subdisk *sd;
214	struct bio_queue_head queue;
215	struct bio *cbp;
216	char *addr;
217	off_t offset, start, length, nstripe, remain;
218	int no, pno, ddisks, pdisks, protate, pleft;
219	u_int strip_size, lvl, qual;
220
221	vol = tr->tro_volume;
222	addr = bp->bio_data;
223	strip_size = vol->v_strip_size;
224	lvl = tr->tro_volume->v_raid_level;
225	qual = tr->tro_volume->v_raid_level_qualifier;
226	protate = tr->tro_volume->v_rotate_parity;
227
228	/* Stripe number. */
229	nstripe = bp->bio_offset / strip_size;
230	/* Start position in stripe. */
231	start = bp->bio_offset % strip_size;
232	/* Number of data and parity disks. */
233	if (lvl == G_RAID_VOLUME_RL_RAIDMDF)
234		pdisks = tr->tro_volume->v_mdf_pdisks;
235	else if (lvl == G_RAID_VOLUME_RL_RAID5EE ||
236	    lvl == G_RAID_VOLUME_RL_RAID6)
237		pdisks = 2;
238	else
239		pdisks = 1;
240	ddisks = vol->v_disks_count - pdisks;
241	/* Parity disk number. */
242	if (lvl == G_RAID_VOLUME_RL_RAID4) {
243		if (qual == 0)		/* P0 */
244			pno = 0;
245		else			/* PN */
246			pno = ddisks;
247		pleft = -1;
248	} else {
249		pno = (nstripe / (ddisks * protate)) % vol->v_disks_count;
250		pleft = protate - (nstripe / ddisks) % protate;
251		if (qual >= 2) {	/* PN/Left */
252			pno = ddisks - pno;
253			if (pno < 0)
254				pno += vol->v_disks_count;
255		}
256	}
257	/* Data disk number. */
258	no = nstripe % ddisks;
259	if (lvl == G_RAID_VOLUME_RL_RAID4) {
260		if (qual == 0)
261			no += pdisks;
262	} else if (qual & 1) {	/* Continuation/Symmetric */
263		no = (pno + pdisks + no) % vol->v_disks_count;
264	} else if (no >= pno)	/* Restart/Asymmetric */
265		no += pdisks;
266	else
267		no += imax(0, pno + pdisks - vol->v_disks_count);
268	/* Stripe start position in disk. */
269	offset = (nstripe / ddisks) * strip_size;
270	/* Length of data to operate. */
271	remain = bp->bio_length;
272
273	bioq_init(&queue);
274	do {
275		length = MIN(strip_size - start, remain);
276		cbp = g_clone_bio(bp);
277		if (cbp == NULL)
278			goto failure;
279		cbp->bio_offset = offset + start;
280		cbp->bio_data = addr;
281		cbp->bio_length = length;
282		cbp->bio_caller1 = &vol->v_subdisks[no];
283		bioq_insert_tail(&queue, cbp);
284		no++;
285		if (lvl == G_RAID_VOLUME_RL_RAID4) {
286			no %= vol->v_disks_count;
287			if (no == pno)
288				no = (no + pdisks) % vol->v_disks_count;
289		} else if (qual & 1) {	/* Continuation/Symmetric */
290			no %= vol->v_disks_count;
291			if (no == pno) {
292				if ((--pleft) <= 0) {
293					pleft += protate;
294					if (qual < 2)	/* P0/Right */
295						pno++;
296					else		/* PN/Left */
297						pno += vol->v_disks_count - 1;
298					pno %= vol->v_disks_count;
299				}
300				no = (pno + pdisks) % vol->v_disks_count;
301				offset += strip_size;
302			}
303		} else {		/* Restart/Asymmetric */
304			if (no == pno)
305				no += pdisks;
306			if (no >= vol->v_disks_count) {
307				no -= vol->v_disks_count;
308				if ((--pleft) <= 0) {
309					pleft += protate;
310					if (qual < 2)	/* P0/Right */
311						pno++;
312					else		/* PN/Left */
313						pno += vol->v_disks_count - 1;
314					pno %= vol->v_disks_count;
315				}
316				if (no == pno)
317					no += pdisks;
318				else
319					no += imax(0, pno + pdisks - vol->v_disks_count);
320				offset += strip_size;
321			}
322		}
323		remain -= length;
324		addr += length;
325		start = 0;
326	} while (remain > 0);
327	while ((cbp = bioq_takefirst(&queue)) != NULL) {
328		sd = cbp->bio_caller1;
329		cbp->bio_caller1 = NULL;
330		g_raid_subdisk_iostart(sd, cbp);
331	}
332	return;
333failure:
334	while ((cbp = bioq_takefirst(&queue)) != NULL)
335		g_destroy_bio(cbp);
336	if (bp->bio_error == 0)
337		bp->bio_error = ENOMEM;
338	g_raid_iodone(bp, bp->bio_error);
339}
340
341static void
342g_raid_tr_iostart_raid5(struct g_raid_tr_object *tr, struct bio *bp)
343{
344	struct g_raid_volume *vol;
345	struct g_raid_tr_raid5_object *trs;
346
347	vol = tr->tro_volume;
348	trs = (struct g_raid_tr_raid5_object *)tr;
349	if (vol->v_state < G_RAID_VOLUME_S_SUBOPTIMAL) {
350		g_raid_iodone(bp, EIO);
351		return;
352	}
353	switch (bp->bio_cmd) {
354	case BIO_READ:
355		g_raid_tr_iostart_raid5_read(tr, bp);
356		break;
357	case BIO_WRITE:
358	case BIO_DELETE:
359	case BIO_FLUSH:
360		g_raid_iodone(bp, ENODEV);
361		break;
362	default:
363		KASSERT(1 == 0, ("Invalid command here: %u (volume=%s)",
364		    bp->bio_cmd, vol->v_name));
365		break;
366	}
367}
368
369static void
370g_raid_tr_iodone_raid5(struct g_raid_tr_object *tr,
371    struct g_raid_subdisk *sd, struct bio *bp)
372{
373	struct bio *pbp;
374
375	pbp = bp->bio_parent;
376	if (pbp->bio_error == 0)
377		pbp->bio_error = bp->bio_error;
378	pbp->bio_inbed++;
379	g_destroy_bio(bp);
380	if (pbp->bio_children == pbp->bio_inbed) {
381		pbp->bio_completed = pbp->bio_length;
382		g_raid_iodone(pbp, pbp->bio_error);
383	}
384}
385
386static int
387g_raid_tr_kerneldump_raid5(struct g_raid_tr_object *tr,
388    void *virtual, vm_offset_t physical, off_t offset, size_t length)
389{
390
391	return (ENODEV);
392}
393
394static int
395g_raid_tr_locked_raid5(struct g_raid_tr_object *tr, void *argp)
396{
397	struct bio *bp;
398	struct g_raid_subdisk *sd;
399
400	bp = (struct bio *)argp;
401	sd = (struct g_raid_subdisk *)bp->bio_caller1;
402	g_raid_subdisk_iostart(sd, bp);
403
404	return (0);
405}
406
407static int
408g_raid_tr_free_raid5(struct g_raid_tr_object *tr)
409{
410	struct g_raid_tr_raid5_object *trs;
411
412	trs = (struct g_raid_tr_raid5_object *)tr;
413
414	if (trs->trso_buffer != NULL) {
415		free(trs->trso_buffer, M_TR_RAID5);
416		trs->trso_buffer = NULL;
417	}
418	return (0);
419}
420
421G_RAID_TR_DECLARE(raid5, "RAID5");
422