xref: /titanic_50/usr/src/uts/common/sys/fs/hsfs_node.h (revision ab4a9beb2e4d596be0b3288c7d92919e27781b57)
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  * High Sierra filesystem structure definitions
23  */
24 /*
25  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
26  * Use is subject to license terms.
27  */
28 
29 #ifndef	_SYS_FS_HSFS_NODE_H
30 #define	_SYS_FS_HSFS_NODE_H
31 
32 #pragma ident	"%Z%%M%	%I%	%E% SMI"
33 
34 #ifdef	__cplusplus
35 extern "C" {
36 #endif
37 
38 #include <sys/taskq.h>
39 
40 struct	hs_direntry {
41 	uint_t		ext_lbn;	/* LBN of start of extent */
42 	uint_t		ext_size;    	/* no. of data bytes in extent */
43 	struct timeval	cdate;		/* creation date */
44 	struct timeval	mdate;		/* last modification date */
45 	struct timeval	adate;		/* last access date */
46 	enum vtype	type;		/* file type */
47 	mode_t		mode;		/* mode and type of file (UNIX) */
48 	uint_t		nlink;		/* no. of links to file */
49 	uid_t		uid;		/* owner's user id */
50 	gid_t		gid;		/* owner's group id */
51 	ino64_t		inode;		/* inode number from rrip data */
52 	dev_t		r_dev;		/* major/minor device numbers */
53 	uint_t		xar_prot :1;	/* 1 if protection in XAR */
54 	uchar_t		xar_len;	/* no. of Logical blocks in XAR */
55 	uchar_t		intlf_sz;	/* intleaving size */
56 	uchar_t		intlf_sk;	/* intleaving skip factor */
57 	ushort_t	sym_link_flag;	/* flags for sym link */
58 	char		*sym_link; 	/* path of sym link for readlink() */
59 };
60 
61 struct	ptable {
62 	uchar_t	filler[7];		/* filler */
63 	uchar_t	dname_len;		/* length of directory name */
64 	uchar_t	dname[HS_DIR_NAMELEN+1];	/* directory name */
65 };
66 
67 struct ptable_idx {
68 	struct ptable_idx *idx_pptbl_idx; /* parent's path table index entry */
69 	struct ptable	*idx_mptbl;	/* path table entry for myself */
70 	ushort_t idx_nochild;		/* no. of children */
71 	ushort_t idx_childid;		/* directory no of first child */
72 };
73 
74 /*
75  * hsnode structure:
76  *
77  * hs_offset, hs_ptbl_idx, base  apply to VDIR type only
78  *
79  * nodeid uniquely identifies an hsnode, ISO9660 means
80  * nodeid can be very big.
81  * For directories it is the disk address of
82  * the data extent of the dir (the directory itself,
83  * ".", and ".." all point to same data extent).
84  * For non-directories, it is the disk address of the
85  * directory entry for the file; note that this does
86  * not permit hard links, as it assumes a single dir
87  * entry per file.
88  */
89 
90 struct  hsnode {
91 	struct hsnode	*hs_hash;	/* next hsnode in hash list */
92 	struct hsnode	*hs_freef;	/* next hsnode in free list */
93 	struct hsnode	*hs_freeb;	/* previous hsnode in free list */
94 	struct vnode	*hs_vnode;	/* the real vnode for the file */
95 	struct hs_direntry hs_dirent;	/* the directory entry for this file */
96 	ino64_t		hs_nodeid;	/* "inode" number for hsnode */
97 	uint_t		hs_dir_lbn;	/* LBN of directory entry */
98 	uint_t		hs_dir_off;	/* offset in LBN of directory entry */
99 	struct ptable_idx	*hs_ptbl_idx;	/* path table index */
100 	uint_t		hs_offset;	/* start offset in dir for searching */
101 	long		hs_mapcnt;	/* mappings to file pages */
102 	uint_t		hs_seq;		/* sequence number */
103 	uint_t		hs_flags;	/* (see below) */
104 	u_offset_t	hs_prev_offset; /* Last read end offset (readahead) */
105 	int		hs_num_contig;  /* Count of contiguous reads */
106 	int		hs_ra_bytes;    /* Bytes to readahead */
107 	kmutex_t	hs_contents_lock;	/* protects hsnode contents */
108 						/* 	except hs_offset */
109 };
110 
111 /* hs_flags */
112 #define	HREF	1			/* hsnode is referenced */
113 
114 /* hs_modes */
115 
116 #define	HFDIR	0040000			/* directory */
117 #define	HFREG	0100000			/* regular file */
118 
119 struct  hsfid {
120 	ushort_t	hf_len;		/* length of fid */
121 	ushort_t	hf_dir_off;	/* offset in LBN of directory entry */
122 	uint_t		hf_dir_lbn;	/* LBN of directory */
123 	uint32_t	hf_ino;		/* The inode number or HS_DUMMY_INO */
124 };
125 
126 
127 /*
128  * All of the fields in the hs_volume are read-only once they have been
129  * initialized.
130  */
131 struct	hs_volume {
132 	ulong_t		vol_size; 	/* no. of Logical blocks in Volume */
133 	uint_t		lbn_size;	/* no. of bytes in a block */
134 	uint_t		lbn_shift;	/* shift to convert lbn to bytes */
135 	uint_t		lbn_secshift;	/* shift to convert lbn to sec */
136 	uint_t		lbn_maxoffset;	/* max lbn-relative offset and mask */
137 	uchar_t		file_struct_ver; /* version of directory structure */
138 	uid_t		vol_uid;	/* uid of volume */
139 	gid_t		vol_gid;	/* gid of volume */
140 	uint_t		vol_prot;	/* protection (mode) of volume */
141 	struct timeval	cre_date;	/* volume creation time */
142 	struct timeval	mod_date;	/* volume modification time */
143 	struct	hs_direntry root_dir;	/* dir entry for Root Directory */
144 	ushort_t	ptbl_len;	/* number of bytes in Path Table */
145 	uint_t		ptbl_lbn;	/* logical block no of Path Table */
146 	ushort_t	vol_set_size;	/* number of CD in this vol set */
147 	ushort_t	vol_set_seq;	/* the sequence number of this CD */
148 	char		vol_id[32];		/* volume id in PVD */
149 };
150 
151 /*
152  * The hsnode table is no longer fixed in size but grows
153  * and shrinks dynamically. However a cache of nodes is still maintained
154  * for efficiency. This cache size (nhsnode) is a tunable which
155  * is either specified in /etc/system or calculated as the number
156  * that will fit into the number of bytes defined by HS_HSNODESPACE (below).
157  */
158 #define	HS_HASHSIZE	32		/* hsnode hash table size */
159 #define	HS_HSNODESPACE	16384		/* approx. space used for hsnodes */
160 
161 /*
162  * We usually use the starting extent LBA for the inode numbers of files and
163  * directories. As this will not work for zero sized files, we assign a dummy
164  * inode number to all zero sized files. We use the number 16 as this is the
165  * LBA for the PVD, this number cannot be a valid starting extent LBA for a
166  * file. In case that the node number is the HS_DUMMY_INO, we use the LBA and
167  * offset of the directory entry of this file (which is what we used before
168  * we started to support correct hard links).
169  */
170 #define	HS_DUMMY_INO	16	/* dummy inode number for empty files */
171 
172 /*
173  * Hsfs I/O Scheduling parameters and data structures.
174  * Deadline for reads is set at 5000 usec.
175  */
176 #define	HSFS_READ_DEADLINE 5000
177 #define	HSFS_NORMAL 0x0
178 
179 /*
180  * This structure holds information for a read request that is enqueued
181  * for processing by the scheduling function. An AVL tree is used to
182  * access the read requests in a sorted manner.
183  */
184 struct hio {
185 	struct buf	*bp;		/* The buf for this read */
186 	struct hio	*contig_chain;  /* Next adjacent read if any */
187 	offset_t	io_lblkno;	/* Starting disk block of io */
188 	u_offset_t	nblocks;	/* # disk blocks */
189 	uint64_t	io_timestamp;	/* usec timestamp for deadline */
190 	ksema_t		*sema;		/* Completion flag */
191 	avl_node_t	io_offset_node; /* Avl tree requirements */
192 	avl_node_t	io_deadline_node;
193 };
194 
195 /*
196  * This structure holds information about all the read requests issued
197  * during a read-ahead invocation. This is then enqueued on a task-queue
198  * for processing by a background thread that takes this read-ahead to
199  * completion and cleans up.
200  */
201 struct hio_info {
202 	struct buf	*bufs;	/* array of bufs issued for this R/A */
203 	caddr_t		*vas;	/* The kmem_alloced chunk for the bufs */
204 	ksema_t		*sema;	/* Semaphores used in the bufs */
205 	uint_t		bufsused; /* # of bufs actually used */
206 	uint_t		bufcnt;   /* Tot bufs allocated. */
207 	struct page	*pp;	  /* The list of I/O locked pages */
208 	struct hsfs	*fsp; /* The filesystem structure */
209 };
210 
211 /*
212  * This is per-filesystem structure that stores toplevel data structures for
213  * the I/O scheduler.
214  */
215 struct hsfs_queue {
216 	/*
217 	 * A dummy hio holding the LBN of the last read processed. Easy
218 	 * to use in AVL_NEXT for Circular Look behavior.
219 	 */
220 	struct hio	*next;
221 
222 	/*
223 	 * A pre-allocated buf for issuing coalesced reads. The scheduling
224 	 * function is mostly single threaded by necessity.
225 	 */
226 	struct buf	*nbuf;
227 	kmutex_t	hsfs_queue_lock; /* Protects the AVL trees */
228 
229 	/*
230 	 * Makes most of the scheduling function Single-threaded.
231 	 */
232 	kmutex_t	strategy_lock;
233 	avl_tree_t	read_tree;	 /* Reads ordered by LBN */
234 	avl_tree_t	deadline_tree;	 /* Reads ordered by timestamp */
235 	taskq_t		*ra_task;	 /* Read-ahead Q */
236 	int		max_ra_bytes;	 /* Max read-ahead quantum */
237 
238 	/* Device Max Transfer size in DEV_BSIZE */
239 	uint_t		dev_maxtransfer;
240 };
241 
242 /*
243  * High Sierra filesystem structure.
244  * There is one of these for each mounted High Sierra filesystem.
245  */
246 enum hs_vol_type {
247 	HS_VOL_TYPE_HS = 0, HS_VOL_TYPE_ISO = 1, HS_VOL_TYPE_ISO_V2 = 2,
248 	HS_VOL_TYPE_JOLIET = 3
249 };
250 #define	HSFS_MAGIC 0x03095500
251 struct hsfs {
252 	struct hsfs	*hsfs_next;	/* ptr to next entry in linked list */
253 	long		hsfs_magic;	/* should be HSFS_MAGIC */
254 	struct vfs	*hsfs_vfs;	/* vfs for this fs */
255 	struct vnode	*hsfs_rootvp;	/* vnode for root of filesystem */
256 	struct vnode	*hsfs_devvp;	/* device mounted on */
257 	enum hs_vol_type hsfs_vol_type; /* see above */
258 	struct hs_volume hsfs_vol;	/* File Structure Volume Descriptor */
259 	struct ptable	*hsfs_ptbl;	/* pointer to incore Path Table */
260 	int		hsfs_ptbl_size;	/* size of incore path table */
261 	struct ptable_idx *hsfs_ptbl_idx; /* pointer to path table index */
262 	int		hsfs_ptbl_idx_size;	/* no. of path table index */
263 	ulong_t		hsfs_ext_impl;	/* ext. information bits */
264 	ushort_t	hsfs_sua_off;	/* the SUA offset */
265 	ushort_t	hsfs_namemax;	/* maximum file name length */
266 	ushort_t	hsfs_namelen;	/* "official" max. file name length */
267 	ulong_t		hsfs_err_flags;	/* ways in which fs is non-conformant */
268 	char		*hsfs_fsmnt;	/* name mounted on */
269 	ulong_t		hsfs_flags;	/* hsfs-specific mount flags */
270 	krwlock_t	hsfs_hash_lock;	/* protect hash table & hst_nohsnode */
271 	struct hsnode	*hsfs_hash[HS_HASHSIZE]; /* head of hash lists */
272 	uint32_t	hsfs_nohsnode;	/* no. of allocated hsnodes */
273 	kmutex_t	hsfs_free_lock;	/* protects free list */
274 	struct hsnode	*hsfs_free_f;	/* first entry of free list */
275 	struct hsnode	*hsfs_free_b;	/* last entry of free list */
276 
277 	/*
278 	 * Counters exported through kstats.
279 	 */
280 	uint64_t	physical_read_bytes;
281 	uint64_t	cache_read_pages;
282 	uint64_t	readahead_bytes;
283 	uint64_t	coalesced_bytes;
284 	uint64_t	total_pages_requested;
285 	kstat_t		*hsfs_kstats;
286 
287 	struct hsfs_queue *hqueue;	/* I/O Scheduling parameters */
288 };
289 
290 /*
291  * Error types: bit offsets into hsfs_err_flags.
292  * Also serves as index into hsfs_error[], so must be
293  * kept in sync with that data structure.
294  */
295 #define	HSFS_ERR_TRAILING_JUNK		0
296 #define	HSFS_ERR_LOWER_CASE_NM		1
297 #define	HSFS_ERR_BAD_ROOT_DIR		2
298 #define	HSFS_ERR_UNSUP_TYPE		3
299 #define	HSFS_ERR_BAD_FILE_LEN		4
300 #define	HSFS_ERR_BAD_JOLIET_FILE_LEN	5
301 #define	HSFS_ERR_TRUNC_JOLIET_FILE_LEN	6
302 #define	HSFS_ERR_BAD_DIR_ENTRY		7
303 #define	HSFS_ERR_NEG_SUA_LEN		8
304 #define	HSFS_ERR_BAD_SUA_LEN		9
305 
306 #define	HSFS_HAVE_LOWER_CASE(fsp) \
307 	((fsp)->hsfs_err_flags & (1 << HSFS_ERR_LOWER_CASE_NM))
308 
309 
310 /*
311  * File system parameter macros
312  */
313 #define	hs_blksize(HSFS, HSP, OFF)	/* file system block size */ \
314 	((HSP)->hs_vn.v_flag & VROOT ? \
315 	    ((OFF) >= \
316 		((HSFS)->hsfs_rdirsec & ~((HSFS)->hsfs_spcl - 1))*HS_SECSIZE ?\
317 		((HSFS)->hsfs_rdirsec & ((HSFS)->hsfs_spcl - 1))*HS_SECSIZE :\
318 		(HSFS)->hsfs_clsize): \
319 	    (HSFS)->hsfs_clsize)
320 #define	hs_blkoff(OFF)		/* offset within block */ \
321 	((OFF) & (HS_SECSIZE - 1))
322 
323 /*
324  * Conversion macros
325  */
326 #define	VFS_TO_HSFS(VFSP)	((struct hsfs *)(VFSP)->vfs_data)
327 #define	HSFS_TO_VFS(FSP)	((FSP)->hsfs_vfs)
328 
329 #define	VTOH(VP)		((struct hsnode *)(VP)->v_data)
330 #define	HTOV(HP)		(((HP)->hs_vnode))
331 
332 /*
333  * Convert between Logical Block Number and Sector Number.
334  */
335 #define	LBN_TO_SEC(lbn, vfsp)	((lbn)>>((struct hsfs *)((vfsp)->vfs_data))->  \
336 				hsfs_vol.lbn_secshift)
337 
338 #define	SEC_TO_LBN(sec, vfsp)	((sec)<<((struct hsfs *)((vfsp)->vfs_data))->  \
339 				hsfs_vol.lbn_secshift)
340 
341 #define	LBN_TO_BYTE(lbn, vfsp)	((lbn)<<((struct hsfs *)((vfsp)->vfs_data))->  \
342 				hsfs_vol.lbn_shift)
343 #define	BYTE_TO_LBN(boff, vfsp)	((boff)>>((struct hsfs *)((vfsp)->vfs_data))-> \
344 				hsfs_vol.lbn_shift)
345 
346 #ifdef	__cplusplus
347 }
348 #endif
349 
350 #endif	/* _SYS_FS_HSFS_NODE_H */
351