xref: /titanic_41/usr/src/uts/common/sys/lvm/md_mirror.h (revision bf85a12b7c81d0745d5a8aff65baeff50006cde9)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #ifndef _SYS_MD_MIRROR_H
28 #define	_SYS_MD_MIRROR_H
29 
30 #include <sys/callb.h>
31 #include <sys/lvm/mdvar.h>
32 #include <sys/lvm/md_mirror_shared.h>
33 #include <sys/lvm/md_rename.h>
34 #ifdef	_KERNEL
35 #include <sys/sunddi.h>
36 #endif
37 
38 #ifdef	__cplusplus
39 extern "C" {
40 #endif
41 
42 /*
43  * following bits are used in status word in the common section
44  * of unit structure
45  */
46 #define	SMS_IS(sm, state) (((sm)->sm_state & (state)) != 0)
47 #define	SMS_BY_INDEX_IS(un, index, state) \
48 		(((un)->un_sm[(index)].sm_state & (state)) != 0)
49 
50 #define	SMS_BY_INDEX_IS_TARGET(un, index) \
51 		((un)->un_sm[(index)].sm_flags & MD_SM_RESYNC_TARGET)
52 
53 #define	SUBMIRROR_IS_READABLE(un, isubmirror)				\
54 	((((un)->un_sm[(isubmirror)].sm_state & SMS_IGNORE) == 0) &&	\
55 	    ((un)->un_sm[(isubmirror)].sm_state & 			\
56 	    (SMS_RUNNING | SMS_COMP_ERRED | SMS_COMP_RESYNC)))
57 
58 #define	SUBMIRROR_IS_WRITEABLE(un, isubmirror)			\
59 	((un)->un_sm[(isubmirror)].sm_state &			\
60 	    (SMS_RUNNING | SMS_COMP_ERRED | SMS_COMP_RESYNC |	\
61 	    SMS_ATTACHED_RESYNC | SMS_OFFLINE_RESYNC))
62 
63 /*
64  * Default resync block size for MN resync messages
65  */
66 #define	MD_DEF_RESYNC_BLK_SZ		8192
67 
68 /*
69  * macro to test if the current block is within the current resync region
70  */
71 #define	IN_RESYNC_REGION(un, ps) \
72 	((un->un_rs_prev_overlap != NULL) && (ps->ps_firstblk >= \
73 	    un->un_rs_prev_overlap->ps_firstblk) && \
74 	    (ps->ps_lastblk <=  un->un_rs_prev_overlap->ps_lastblk))
75 /*
76  * Default resync update interval (in minutes).
77  */
78 #define	MD_DEF_MIRROR_RESYNC_INTVL	5
79 
80 /*
81  * Defines for flags argument in function set_sm_comp_state()
82  */
83 #define	MD_STATE_NO_XMIT	0x0000 /* Local action, (sent from master) */
84 #define	MD_STATE_XMIT		0x0001 /* Non-local action, send to master */
85 #define	MD_STATE_WMUPDATE	0x0002 /* Action because of watermark update */
86 #define	MD_STATE_OCHELD		0x0004 /* open/close lock held */
87 
88 /*
89  * Defines for flags argument in function check_comp_4_hotspares()
90  */
91 #define	MD_HOTSPARE_NO_XMIT	0x0000 /* Local action, (sent from master) */
92 #define	MD_HOTSPARE_XMIT	0x0001 /* Non-local action, send to master */
93 #define	MD_HOTSPARE_WMUPDATE	0x0002 /* Action because of watermark update */
94 #define	MD_HOTSPARE_LINKHELD	0x0004 /* md_link_rw lock held */
95 
96 /*
97  * Defines for argument in function send_mn_resync_done_message()
98  */
99 #define	RESYNC_ERR		0x1
100 #define	CLEAR_OPT_NOT_DONE	0x2
101 
102 /*
103  * Defines for argument in function resync_read_blk_range()
104  */
105 #define	MD_FIRST_RESYNC_NEXT	0x1
106 #define	MD_SEND_MESS_XMIT	0x2
107 #define	MD_RESYNC_FLAG_ERR	0x4
108 
109 /*
110  * Define for argument in function wait_for_overlaps()
111  */
112 #define	MD_OVERLAP_ALLOW_REPEAT	0x1	/* Allow if ps already in tree */
113 #define	MD_OVERLAP_NO_REPEAT	0	/* ps must not already be in tree */
114 
115 /*
116  * Define for max retries of mirror_owner
117  */
118 #define	MD_OWNER_RETRIES	10
119 
120 /*
121  * mm_submirror32_od and mm_unit32_od are used only for 32 bit old format
122  */
123 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
124 #pragma pack(4)
125 #endif
126 typedef struct  mm_submirror32_od {	/* submirrors */
127 	mdkey_t		sm_key;
128 	dev32_t		sm_dev;
129 	sm_state_t	sm_state;
130 	sm_flags_t	sm_flags;
131 	caddr32_t	xx_sm_shared_by_blk;	/* really void *) */
132 	caddr32_t	xx_sm_shared_by_indx;	/* really void *) */
133 	caddr32_t	xx_sm_get_component_count;
134 	caddr32_t	xx_sm_get_bcss;	/* block count skip size */
135 	md_m_shared32_od_t sm_shared;	/* used for mirroring plain devices */
136 	int		sm_hsp_id;	/* used for mirroring plain devices */
137 	struct timeval32 sm_timestamp;	/* time of last state change */
138 } mm_submirror32_od_t;
139 
140 typedef struct	mm_submirror {		/* submirrors */
141 	mdkey_t		sm_key;
142 	md_dev64_t	sm_dev;		/* 64 bit */
143 	sm_state_t	sm_state;
144 	sm_flags_t	sm_flags;
145 	md_m_shared_t	sm_shared;	/* used for mirroring plain devices */
146 	int		sm_hsp_id;	/* used for mirroring plain devices */
147 	md_timeval32_t	sm_timestamp;	/* time of last state change, 32 bit */
148 } mm_submirror_t;
149 
150 typedef struct mm_unit32_od {
151 	mdc_unit32_od_t	c;			/* common stuff */
152 
153 	int		un_last_read;		/* last submirror index read */
154 	uint_t		un_changecnt;
155 	ushort_t	un_nsm;			/* number of submirrors */
156 	mm_submirror32_od_t un_sm[NMIRROR];
157 	int		un_overlap_tree_flag;
158 	int		xx_un_overlap_tree_mx[2];	/* replaces mutex */
159 	ushort_t	xx_un_overlap_tree_cv;
160 	caddr32_t	xx_un_overlap_root;
161 	mm_rd_opt_t	un_read_option;		/* mirror read option */
162 	mm_wr_opt_t	un_write_option;	/* mirror write option */
163 	mm_pass_num_t	un_pass_num;		/* resync pass number */
164 	/*
165 	 * following used to keep dirty bitmaps
166 	 */
167 	int		xx_un_resync_mx[2];	/* replaces mutex */
168 	ushort_t	xx_un_resync_cv;
169 	uint_t		un_resync_flg;
170 	uint_t		un_waiting_to_mark;
171 	uint_t		un_waiting_to_commit;
172 	caddr32_t	xx_un_outstanding_writes;	/* outstanding write */
173 	caddr32_t	xx_un_goingclean_bm;
174 	caddr32_t	xx_un_goingdirty_bm;
175 	caddr32_t	xx_un_dirty_bm;
176 	caddr32_t	xx_un_resync_bm;
177 	uint_t		un_rrd_blksize;	/* The blocksize of the dirty bits */
178 	uint_t		un_rrd_num;	/* The number of resync regions */
179 	mddb_recid_t	un_rr_dirty_recid;	/* resync region bm record id */
180 	/*
181 	 * following stuff is private to resync process
182 	 */
183 	int		un_rs_copysize;
184 	int		un_rs_dests;	/* destinations */
185 	daddr32_t	un_rs_resync_done;	/* used for percent done */
186 	daddr32_t	un_rs_resync_2_do;	/* user for percent done */
187 	int		un_rs_dropped_lock;
188 	caddr32_t	un_rs_type;		/* type of resync in progress */
189 	/*
190 	 * Incore elements in this old structure are no longer referenced by
191 	 * current 64 bit kernel.  Comment them out for maintenance purpose.
192 	 *
193 	 * 	mm_submirror_ic_t	un_smic[NMIRROR];
194 	 * 	kmutex_t		un_ovrlap_chn_mx;
195 	 * 	kcondvar_t		un_ovrlap_chn_cv;
196 	 * 	struct md_mps		*un_ovrlap_chn;
197 	 * 	kmutex_t		un_resync_mx;
198 	 * 	kcondvar_t		un_resync_cv;
199 	 * 	short			*un_outstanding_writes;
200 	 * 	uchar_t			*un_goingclean_bm;
201 	 * 	uchar_t			*un_goingdirty_bm;
202 	 * 	uchar_t			*un_dirty_bm;
203 	 * 	uchar_t			*un_resync_bm;
204 	 * 	char			*un_rs_buffer;
205 	 */
206 } mm_unit32_od_t;
207 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
208 #pragma pack()
209 #endif
210 
211 /* Types of resync in progress (used for un_rs_type) */
212 #define	MD_RS_NONE		0		/* No resync */
213 #define	MD_RS_OPTIMIZED		0x0001		/* Optimized resync */
214 #define	MD_RS_COMPONENT		0x0002		/* Component resync */
215 #define	MD_RS_SUBMIRROR		0x0003		/* Submirror resync */
216 #define	MD_RS_ABR		0x0004		/* Application based resync */
217 
218 /*
219  * un_rs_type is split into the following bitfields:
220  *
221  * 0-3	Resync type (as above)
222  * 4-7	Submirror index [0..3]
223  * 8-31	Component index
224  */
225 #define	RS_TYPE_MASK	0xF
226 #define	RS_SMI_MASK	0xF0
227 #define	RS_CI_MASK	0x1FFF00
228 
229 #define	RS_TYPE(x)	((x) & RS_TYPE_MASK)
230 #define	RS_SMI(x)	(((x) & RS_SMI_MASK) >> 4)
231 #define	RS_CI(x)	(((x) & RS_CI_MASK) >> 8)
232 
233 #define	SET_RS_TYPE(x, v)	{					\
234 				    (x) &= ~RS_TYPE_MASK;		\
235 				    (x) |= ((v) & RS_TYPE_MASK);	\
236 				}
237 #define	SET_RS_TYPE_NONE(x)	{					\
238 				    (x) &= ~RS_TYPE_MASK;		\
239 				}
240 #define	SET_RS_SMI(x, v)	{					\
241 				    (x) &= ~RS_SMI_MASK; 		\
242 				    (x) |= (((v) << 4) & RS_SMI_MASK);	\
243 				}
244 #define	SET_RS_CI(x, v)		{					\
245 				    (x) &= ~RS_CI_MASK;			\
246 				    (x) |= (((v) << 8) & RS_CI_MASK);	\
247 				}
248 
249 typedef struct	mm_submirror_ic {
250 	intptr_t	(*sm_shared_by_blk)(md_dev64_t, void *,
251 				diskaddr_t, u_longlong_t *);
252 	intptr_t	(*sm_shared_by_indx)(md_dev64_t, void *, int);
253 	int		(*sm_get_component_count)(md_dev64_t, void *);
254 	int		(*sm_get_bcss)(md_dev64_t, void *, int, diskaddr_t *,
255 				size_t *, u_longlong_t *, u_longlong_t *);
256 } mm_submirror_ic_t;
257 
258 typedef struct md_mps {
259 	DAEMON_QUEUE
260 	buf_t		*ps_bp;
261 	struct mm_unit	*ps_un;
262 	mdi_unit_t	*ps_ui;
263 	uint_t		 ps_childbflags;
264 	caddr_t		 ps_addr;
265 	diskaddr_t	 ps_firstblk;
266 	diskaddr_t	 ps_lastblk;
267 	uint_t		 ps_flags;
268 	uint_t		 ps_allfrom_sm;		/* entire read came from here */
269 	uint_t		 ps_writable_sm;
270 	uint_t		 ps_current_sm;
271 	uint_t		 ps_active_cnt;
272 	int		 ps_frags;
273 	uint_t		 ps_changecnt;
274 	struct md_mps	*ps_unused1;
275 	struct md_mps	*ps_unused2;
276 	void		 (*ps_call)();
277 	kmutex_t	 ps_mx;
278 	avl_node_t	ps_overlap_node;
279 } md_mps_t;
280 
281 #define	MD_MPS_ON_OVERLAP	0x0001
282 #define	MD_MPS_ERROR		0x0002
283 #define	MD_MPS_WRITE_AFTER_READ	0x0004
284 #define	MD_MPS_WOW		0x0008
285 #define	MD_MPS_DONTFREE		0x0010
286 #define	MD_MPS_DONE		0x0020
287 #define	MD_MPS_MAPPED		0x0040		/* re: MD_STR_MAPPED	*/
288 #define	MD_MPS_NOBLOCK		0x0080		/* re: MD_NOBLOCK	*/
289 #define	MD_MPS_ABR		0x0100		/* re: MD_STR_ABR	*/
290 #define	MD_MPS_DMR		0x0200		/* re: MD_STR_DMR	*/
291 #define	MD_MPS_WMUPDATE		0x0400		/* re: MD_STR_WMUPDATE	*/
292 #define	MD_MPS_DIRTY_RD		0x0800		/* re: MD_STR_DIRTY_RD	*/
293 #define	MD_MPS_RESYNC_READ	0x1000
294 #define	MD_MPS_FLAG_ERROR	0x2000		/* re: MD_STR_FLAG_ERR	*/
295 #define	MD_MPS_BLOCKABLE_IO	0x4000		/* re: MD_STR_BLOCK_OK  */
296 
297 #define	MPS_FREE(kc, ps)			\
298 {						\
299 	if ((ps)->ps_flags & MD_MPS_DONTFREE)	\
300 		(ps)->ps_flags |= MD_MPS_DONE;	\
301 	else					\
302 		kmem_cache_free((kc), (ps));	\
303 }
304 
305 typedef struct md_mcs {
306 	DAEMON_QUEUE
307 	md_mps_t	*cs_ps;
308 	minor_t		 cs_mdunit;
309 	/* Add new structure members HERE!! */
310 	buf_t		 cs_buf;
311 	/*  DO NOT add structure members here; cs_buf is dynamically sized */
312 } md_mcs_t;
313 
314 typedef struct  mm_mirror_ic {
315 	kmutex_t	un_overlap_tree_mx;
316 	kcondvar_t	un_overlap_tree_cv;
317 	avl_tree_t	un_overlap_root;
318 	kmutex_t	un_resync_mx;
319 	kcondvar_t	un_resync_cv;
320 	short		*un_outstanding_writes; /* outstanding write array */
321 	uchar_t		*un_goingclean_bm;
322 	uchar_t		*un_goingdirty_bm;
323 	uchar_t		*un_dirty_bm;
324 	uchar_t		*un_resync_bm;
325 	char		*un_rs_buffer;
326 	int		un_suspend_wr_flag;
327 	kmutex_t	un_suspend_wr_mx;
328 	kcondvar_t	un_suspend_wr_cv;
329 	md_mn_nodeid_t	un_mirror_owner;	/* Node which owns mirror */
330 	diskaddr_t	un_resync_startbl;	/* Start block for resync */
331 	kmutex_t	un_owner_mx;		/* Mutex for un_owner_state */
332 	uint_t		un_owner_state;		/* See below */
333 	uint_t		un_mirror_owner_status;	/* status for ioctl request */
334 	kmutex_t	un_dmr_mx;		/* mutex for DMR requests */
335 	kcondvar_t	un_dmr_cv;		/* condvar for DMR requests */
336 	int		un_dmr_last_read;	/* last DMR submirror read */
337 	callb_cpr_t	un_rs_cprinfo;		/* CPR info for resync thread */
338 	kmutex_t	un_rs_cpr_mx;		/* mutex for resync CPR info */
339 	kmutex_t	un_prr_cpr_mx;		/* mutex for prr CPR info */
340 	uint_t		un_resync_completed;	/* type of last resync */
341 	int		un_abr_count;		/* count of sp's with abr set */
342 
343 	uchar_t		*un_pernode_dirty_bm[MD_MNMAXSIDES];
344 	uchar_t		*un_pernode_dirty_sum;
345 
346 	krwlock_t	un_pernode_dirty_mx[MD_MNMAXSIDES];
347 	ushort_t	un_rr_clean_start_bit;  /* where to start next clean */
348 
349 #ifdef	_KERNEL
350 	ddi_taskq_t	*un_drl_task;		/* deferred RR_CLEAN taskq */
351 #else
352 	void		*un_drl_task;		/* deferred RR_CLEAN taskq */
353 #endif	/* _KERNEL */
354 	uint_t		un_waiting_to_clear;	/* Blocked waiting to clear */
355 
356 }mm_mirror_ic_t;
357 
358 #define	MM_MN_OWNER_SENT	0x0001		/* RPC in progress */
359 #define	MM_MN_BECOME_OWNER	0x0002		/* Ownership change in prog. */
360 #define	MM_MN_PREVENT_CHANGE	0x0004		/* Disallow ownership change */
361 
362 typedef struct mm_unit {
363 	mdc_unit_t	c;			/* common stuff */
364 
365 	int		un_last_read;		/* last submirror index read */
366 	uint_t		un_changecnt;
367 	ushort_t	un_nsm;			/* number of submirrors */
368 	mm_submirror_t	un_sm[NMIRROR];
369 	int		un_overlap_tree_flag;
370 	mm_rd_opt_t	un_read_option;		/* mirror read option */
371 	mm_wr_opt_t	un_write_option;	/* mirror write option */
372 	mm_pass_num_t	un_pass_num;		/* resync pass number */
373 	/*
374 	 * following used to keep dirty bitmaps
375 	 */
376 	uint_t		un_resync_flg;
377 	uint_t		un_waiting_to_mark;
378 	uint_t		un_waiting_to_commit;
379 	uint_t		un_rrd_blksize;	  /* The blocksize of the dirty bits */
380 	uint_t		un_rrd_num;	  /* The number of resync regions */
381 	mddb_recid_t	un_rr_dirty_recid; /* resync region bm db record id */
382 	/*
383 	 * following stuff is private to resync process
384 	 */
385 	int 		un_rs_copysize;
386 	int 		un_rs_dests;		/* destinations */
387 	diskaddr_t	un_rs_resync_done;	/* used for percent done */
388 	diskaddr_t	un_rs_resync_2_do;	/* user for percent done */
389 	int		un_rs_dropped_lock;
390 	uint_t		un_rs_type;		/* type of resync */
391 	/*
392 	 * Incore only elements
393 	 */
394 	mm_submirror_ic_t un_smic[NMIRROR];	/* NMIRROR elements array */
395 	mm_mirror_ic_t	un_mmic;
396 	kmutex_t	un_rrp_inflight_mx;
397 	/*
398 	 * resync thread control
399 	 */
400 	kthread_t	*un_rs_thread;		/* Resync thread ID */
401 	kmutex_t	un_rs_thread_mx;	/* Thread cv mutex */
402 	kcondvar_t	un_rs_thread_cv;	/* Cond. Var. for thread */
403 	uint_t		un_rs_thread_flags;	/* Thread control flags */
404 	md_mps_t	*un_rs_prev_overlap;	/* existing overlap request */
405 	timeout_id_t	un_rs_resync_to_id;	/* resync progress timeout */
406 	kmutex_t	un_rs_progress_mx;	/* Resync progress mutex */
407 	kcondvar_t	un_rs_progress_cv;	/* Cond. Var. for progress */
408 	uint_t		un_rs_progress_flags;	/* Thread control flags */
409 	void		*un_rs_msg;		/* Intra-node resync message */
410 } mm_unit_t;
411 
412 #define	un_overlap_tree_mx	un_mmic.un_overlap_tree_mx
413 #define	un_overlap_tree_cv	un_mmic.un_overlap_tree_cv
414 #define	un_overlap_root		un_mmic.un_overlap_root
415 #define	un_resync_mx		un_mmic.un_resync_mx
416 #define	un_resync_cv		un_mmic.un_resync_cv
417 #define	un_outstanding_writes	un_mmic.un_outstanding_writes
418 #define	un_goingclean_bm	un_mmic.un_goingclean_bm
419 #define	un_goingdirty_bm	un_mmic.un_goingdirty_bm
420 #define	un_dirty_bm		un_mmic.un_dirty_bm
421 #define	un_resync_bm		un_mmic.un_resync_bm
422 #define	un_rs_buffer		un_mmic.un_rs_buffer
423 #define	un_suspend_wr_mx	un_mmic.un_suspend_wr_mx
424 #define	un_suspend_wr_cv	un_mmic.un_suspend_wr_cv
425 #define	un_suspend_wr_flag	un_mmic.un_suspend_wr_flag
426 #define	un_mirror_owner		un_mmic.un_mirror_owner
427 #define	un_resync_startbl	un_mmic.un_resync_startbl
428 #define	un_owner_mx		un_mmic.un_owner_mx
429 #define	un_owner_state		un_mmic.un_owner_state
430 #define	un_mirror_reqs		un_mmic.un_mirror_reqs
431 #define	un_mirror_reqs_done	un_mmic.un_mirror_reqs_done
432 #define	un_mirror_owner_status	un_mmic.un_mirror_owner_status
433 #define	un_dmr_mx		un_mmic.un_dmr_mx
434 #define	un_dmr_cv		un_mmic.un_dmr_cv
435 #define	un_dmr_last_read	un_mmic.un_dmr_last_read
436 #define	un_rs_cprinfo		un_mmic.un_rs_cprinfo
437 #define	un_rs_cpr_mx		un_mmic.un_rs_cpr_mx
438 #define	un_prr_cpr_mx		un_mmic.un_prr_cpr_mx
439 #define	un_resync_completed	un_mmic.un_resync_completed
440 #define	un_abr_count		un_mmic.un_abr_count
441 #define	un_pernode_dirty_bm	un_mmic.un_pernode_dirty_bm
442 #define	un_pernode_dirty_sum	un_mmic.un_pernode_dirty_sum
443 #define	un_pernode_dirty_mx	un_mmic.un_pernode_dirty_mx
444 #define	un_rr_clean_start_bit	un_mmic.un_rr_clean_start_bit
445 #define	un_drl_task		un_mmic.un_drl_task
446 #define	un_waiting_to_clear	un_mmic.un_waiting_to_clear
447 
448 #define	MM_RF_GATECLOSED	0x0001
449 #define	MM_RF_COMMIT_NEEDED	0x0002
450 #define	MM_RF_COMMITING		0x0004
451 #define	MM_RF_STALL_CLEAN	(MM_RF_COMMITING | \
452 				    MM_RF_COMMIT_NEEDED | \
453 				    MM_RF_GATECLOSED)
454 
455 
456 #define	MD_MN_MIRROR_UNOWNED	0
457 #define	MD_MN_MIRROR_OWNER(un)	 (un->un_mirror_owner == md_mn_mynode_id)
458 #define	MD_MN_NO_MIRROR_OWNER(un)	\
459 	(un->un_mirror_owner == MD_MN_MIRROR_UNOWNED)
460 
461 typedef struct err_comp {
462 	struct err_comp	*ec_next;
463 	int		ec_smi;
464 	int		ec_ci;
465 } err_comp_t;
466 
467 extern	int	md_min_rr_size;
468 extern	int	md_def_num_rr;
469 
470 /* Optimized resync records controllers */
471 #define	MD_MIN_RR_SIZE		(md_min_rr_size)
472 #define	MD_DEF_NUM_RR		(md_def_num_rr)
473 #define	MD_MAX_NUM_RR		(4192*NBBY - sizeof (struct optim_resync))
474 
475 /* default resync buffer size */
476 #define	MD_DEF_RESYNC_BUF_SIZE	(1024)
477 
478 /* Structure for optimized resync records */
479 #define	OR_MAGIC	0xFECA	/* Only missing the L */
480 typedef struct optim_resync {
481 	uint_t	or_revision;
482 	uint_t	or_magic;
483 	uint_t	or_blksize;
484 	uint_t	or_num;
485 	uchar_t	or_rr[1];
486 } optim_resync_t;
487 
488 /* Type 2 for mirror records */
489 #define	MIRROR_REC	1
490 #define	RESYNC_REC	2
491 
492 #ifdef _KERNEL
493 
494 #define	NO_SUBMIRRORS	(0)
495 #define	ALL_SUBMIRRORS	(0xFFF)
496 #define	SMI2BIT(smi)	(1 << (smi))
497 
498 /* For use with mirror_other_sources() */
499 #define	WHOLE_SM	(-1)
500 
501 #define	BLK_TO_RR(i, b, un)  {\
502 	(i) = ((b) / ((un))->un_rrd_blksize); \
503 	if ((i) > ((un))->un_rrd_num) \
504 		{ panic("md: BLK_TO_RR"); } \
505 }
506 
507 #define	RR_TO_BLK(b, i, un) \
508 	(b) = ((i) * ((un))->un_rrd_blksize)
509 
510 #define	IS_GOING_DIRTY(i, un)	(isset((un)->un_goingdirty_bm, (i)))
511 #define	CLR_GOING_DIRTY(i, un)	(clrbit((un)->un_goingdirty_bm, (i)))
512 #define	SET_GOING_DIRTY(i, un)	(setbit((un)->un_goingdirty_bm, (i)))
513 
514 #define	IS_GOING_CLEAN(i, un)	(isset((un)->un_goingclean_bm, (i)))
515 #define	CLR_GOING_CLEAN(i, un)	(clrbit((un)->un_goingclean_bm, (i)))
516 #define	SET_GOING_CLEAN(i, un)	(setbit((un)->un_goingclean_bm, (i)))
517 
518 #define	IS_REGION_DIRTY(i, un)	(isset((un)->un_dirty_bm, (i)))
519 #define	CLR_REGION_DIRTY(i, un)	(clrbit((un)->un_dirty_bm, (i)))
520 #define	SET_REGION_DIRTY(i, un)	(setbit((un)->un_dirty_bm, (i)))
521 
522 #define	IS_KEEPDIRTY(i, un)	(isset((un)->un_resync_bm, (i)))
523 #define	CLR_KEEPDIRTY(i, un)	(clrbit((un)->un_resync_bm, (i)))
524 
525 #define	IS_PERNODE_DIRTY(n, i, un) \
526 	(isset((un)->un_pernode_dirty_bm[(n)-1], (i)))
527 #define	CLR_PERNODE_DIRTY(n, i, un) \
528 	(clrbit((un)->un_pernode_dirty_bm[(n)-1], (i)))
529 #define	SET_PERNODE_DIRTY(n, i, un) \
530 	(setbit((un)->un_pernode_dirty_bm[(n)-1], (i)))
531 
532 /*
533  * Write-On-Write handling.
534  *   flags for md_mirror_wow_flg
535  *   structure for quing copy-writes
536  *   macros for relative locating of header and buffer
537  */
538 #define	WOW_DISABLE	0x0001	/* turn off WOW detection */
539 #define	WOW_PHYS_ENABLE	0x0020	/* turn on WOW for PHYS */
540 #define	WOW_LOGIT	0x0002	/* log non-disabled WOW detections */
541 #define	WOW_NOCOPY	0x0004	/* repeat normal write on WOW detection */
542 
543 typedef	struct wowhdr {
544 	DAEMON_QUEUE
545 	md_mps_t	*wow_ps;
546 	int		wow_offset;
547 } wowhdr_t;
548 
549 #define	WOWBUF_HDR(wowbuf)	((void *)(wowbuf-sizeof (wowhdr_t)))
550 #define	WOWHDR_BUF(wowhdr)	((char *)wowhdr+sizeof (wowhdr_t))
551 
552 /*
553  * Structure used to to save information about DMR reads.  Used to save
554  * the count of all DMR reads and the timestamp of the last one executed.
555  * We declare a global with this structure and it can be read by a debugger to
556  * verify that the DMR ioctl has been executed and the number of times that it
557  * has been executed.
558  */
559 typedef struct dmr_stats {
560 	uint_t		dmr_count;
561 	struct timeval	dmr_timestamp;
562 } dmr_stats_t;
563 
564 /* Externals from mirror.c */
565 extern mddb_recid_t	mirror_get_sm_unit(md_dev64_t);
566 extern void		mirror_release_sm_unit(md_dev64_t);
567 
568 extern void		mirror_set_sm_state(mm_submirror_t *,
569 				mm_submirror_ic_t *, sm_state_t, int);
570 
571 extern void		mirror_commit(mm_unit_t *, int, mddb_recid_t *);
572 extern int		poke_hotspares(void);
573 extern void		build_submirror(mm_unit_t *, int, int);
574 extern int		mirror_build_incore(mm_unit_t *, int);
575 extern void		reset_mirror(mm_unit_t *, minor_t, int);
576 extern int		mirror_internal_open(minor_t, int, int, int, IOLOCK *);
577 extern int		mirror_internal_close(minor_t, int, int, IOLOCK *);
578 extern void		set_sm_comp_state(mm_unit_t *, int, int, int,
579 			    mddb_recid_t *, uint_t, IOLOCK *);
580 extern int		mirror_other_sources(mm_unit_t *, int, int, int);
581 extern int		mirror_resync_message(md_mn_rs_params_t *, IOLOCK *);
582 extern void		md_mirror_strategy(buf_t *, int, void *);
583 extern int		mirror_directed_read(dev_t, vol_directed_rd_t *, int);
584 extern void		mirror_check_failfast(minor_t mnum);
585 extern int		check_comp_4_hotspares(mm_unit_t *, int, int, uint_t,
586 			    mddb_recid_t, IOLOCK *);
587 extern void		mirror_overlap_tree_remove(md_mps_t *ps);
588 extern void		mirror_child_init(md_mcs_t *cs);
589 
590 /* Externals from mirror_ioctl.c */
591 extern void		reset_comp_states(mm_submirror_t *,
592 			    mm_submirror_ic_t *);
593 extern int		mirror_grow_unit(mm_unit_t *un, md_error_t *ep);
594 extern int		md_mirror_ioctl(dev_t dev, int cmd, void *data,
595 			    int mode, IOLOCK *lockp);
596 extern mm_unit_t	*mirror_getun(minor_t, md_error_t *, int, IOLOCK *);
597 extern void		mirror_get_status(mm_unit_t *un, IOLOCK *lockp);
598 extern int		mirror_choose_owner(mm_unit_t *un, md_mn_req_owner_t *);
599 
600 /* rename named service functions */
601 md_ren_list_svc_t	mirror_rename_listkids;
602 md_ren_svc_t		mirror_rename_check;
603 md_ren_roleswap_svc_t	mirror_renexch_update_kids;
604 md_ren_roleswap_svc_t	mirror_exchange_parent_update_to;
605 md_ren_roleswap_svc_t	mirror_exchange_self_update_from_down;
606 
607 /* Externals from mirror_resync.c */
608 extern int		unit_setup_resync(mm_unit_t *, int);
609 extern int		mirror_resync_unit(minor_t mnum, md_resync_ioctl_t *ri,
610 			    md_error_t *ep, IOLOCK *);
611 extern int		mirror_ioctl_resync(md_resync_ioctl_t *p, IOLOCK *);
612 extern int		mirror_mark_resync_region(mm_unit_t *, diskaddr_t,
613 				diskaddr_t, md_mn_nodeid_t);
614 extern void		resync_start_timeout(set_t setno);
615 extern int		mirror_resize_resync_regions(mm_unit_t *, diskaddr_t);
616 extern int		mirror_add_resync_regions(mm_unit_t *, diskaddr_t);
617 extern int		mirror_probedevs(md_probedev_t *, IOLOCK *);
618 extern void		mirror_copy_rr(int, uchar_t *, uchar_t *);
619 extern void		mirror_process_unit_resync(mm_unit_t *);
620 extern int		mirror_set_dirty_rr(md_mn_rr_dirty_params_t *);
621 extern int		mirror_set_clean_rr(md_mn_rr_clean_params_t *);
622 #endif	/* _KERNEL */
623 
624 #ifdef	__cplusplus
625 }
626 #endif
627 
628 #endif	/* _SYS_MD_MIRROR_H */
629