xref: /netbsd-src/sys/uvm/uvm_readahead.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: uvm_readahead.c,v 1.6 2009/06/10 01:54:08 yamt Exp $	*/
2 
3 /*-
4  * Copyright (c)2003, 2005, 2009 YAMAMOTO Takashi,
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  *
16  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26  * SUCH DAMAGE.
27  */
28 
29 /*
30  * uvm_object read-ahead
31  *
32  * TODO:
33  *	- tune.
34  *	- handle multiple streams.
35  *	- find a better way to deal with PGO_LOCKED pager requests.
36  *	  (currently just ignored)
37  *	- consider the amount of memory in the system.
38  *	- consider the speed of the underlying device.
39  *	- consider filesystem block size / block layout.
40  */
41 
42 #include <sys/cdefs.h>
43 __KERNEL_RCSID(0, "$NetBSD: uvm_readahead.c,v 1.6 2009/06/10 01:54:08 yamt Exp $");
44 
45 #include <sys/param.h>
46 #include <sys/pool.h>
47 
48 #include <uvm/uvm.h>
49 #include <uvm/uvm_readahead.h>
50 
51 #if defined(READAHEAD_DEBUG)
52 #define	DPRINTF(a)	printf a
53 #else /* defined(READAHEAD_DEBUG) */
54 #define	DPRINTF(a)	/* nothing */
55 #endif /* defined(READAHEAD_DEBUG) */
56 
57 /*
58  * uvm_ractx: read-ahead context.
59  */
60 
61 struct uvm_ractx {
62 	int ra_flags;
63 #define	RA_VALID	1
64 	off_t ra_winstart;	/* window start offset */
65 	size_t ra_winsize;	/* window size */
66 	off_t ra_next;		/* next offset to read-ahead */
67 };
68 
69 #if defined(sun2) || (defined(sun3) && defined(_SUN3_))
70 /* XXX: on sun2 and sun3 (but not sun3x) MAXPHYS is 0xe000 */
71 #undef MAXPHYS
72 #define MAXPHYS		0x8000	/* XXX */
73 #endif
74 
75 #define	RA_WINSIZE_INIT	MAXPHYS			/* initial window size */
76 #define	RA_WINSIZE_MAX	(MAXPHYS * 8)		/* max window size */
77 #define	RA_WINSIZE_SEQENTIAL	RA_WINSIZE_MAX	/* fixed window size used for
78 						   SEQUENTIAL hint */
79 #define	RA_MINSIZE	(MAXPHYS * 2)		/* min size to start i/o */
80 #define	RA_IOCHUNK	MAXPHYS			/* read-ahead i/o chunk size */
81 
82 static off_t ra_startio(struct uvm_object *, off_t, size_t);
83 static struct uvm_ractx *ra_allocctx(void);
84 static void ra_freectx(struct uvm_ractx *);
85 
86 static struct pool_cache ractx_cache;
87 
88 /*
89  * uvm_ra_init: initialize readahead module.
90  */
91 
92 void
93 uvm_ra_init(void)
94 {
95 
96 	pool_cache_bootstrap(&ractx_cache, sizeof(struct uvm_ractx), 0, 0, 0,
97 	    "ractx", NULL, IPL_NONE, NULL, NULL, NULL);
98 }
99 
100 static struct uvm_ractx *
101 ra_allocctx(void)
102 {
103 
104 	return pool_cache_get(&ractx_cache, PR_NOWAIT);
105 }
106 
107 static void
108 ra_freectx(struct uvm_ractx *ra)
109 {
110 
111 	pool_cache_put(&ractx_cache, ra);
112 }
113 
114 /*
115  * ra_startio: start i/o for read-ahead.
116  *
117  * => start i/o for each RA_IOCHUNK sized chunk.
118  * => return offset to which we started i/o.
119  */
120 
121 static off_t
122 ra_startio(struct uvm_object *uobj, off_t off, size_t sz)
123 {
124 	const off_t endoff = off + sz;
125 
126 	DPRINTF(("%s: uobj=%p, off=%" PRIu64 ", endoff=%" PRIu64 "\n",
127 	    __func__, uobj, off, endoff));
128 	off = trunc_page(off);
129 	while (off < endoff) {
130 		const size_t chunksize = RA_IOCHUNK;
131 		int error;
132 		size_t donebytes;
133 		int npages;
134 		int orignpages;
135 		size_t bytelen;
136 
137 		KASSERT((chunksize & (chunksize - 1)) == 0);
138 		KASSERT((off & PAGE_MASK) == 0);
139 		bytelen = ((off + chunksize) & -(off_t)chunksize) - off;
140 		KASSERT((bytelen & PAGE_MASK) == 0);
141 		npages = orignpages = bytelen >> PAGE_SHIFT;
142 		KASSERT(npages != 0);
143 
144 		/*
145 		 * use UVM_ADV_RANDOM to avoid recursion.
146 		 */
147 
148 		mutex_enter(&uobj->vmobjlock);
149 		error = (*uobj->pgops->pgo_get)(uobj, off, NULL,
150 		    &npages, 0, VM_PROT_READ, UVM_ADV_RANDOM, 0);
151 		DPRINTF(("%s:  off=%" PRIu64 ", bytelen=%zu -> %d\n",
152 		    __func__, off, bytelen, error));
153 		if (error != 0 && error != EBUSY) {
154 			if (error != EINVAL) { /* maybe past EOF */
155 				DPRINTF(("%s: error=%d\n", __func__, error));
156 			}
157 			break;
158 		}
159 		KASSERT(orignpages == npages);
160 		donebytes = orignpages << PAGE_SHIFT;
161 		off += donebytes;
162 	}
163 
164 	return off;
165 }
166 
167 /* ------------------------------------------------------------ */
168 
169 /*
170  * uvm_ra_allocctx: allocate a context.
171  */
172 
173 struct uvm_ractx *
174 uvm_ra_allocctx(void)
175 {
176 	struct uvm_ractx *ra;
177 
178 	ra = ra_allocctx();
179 	if (ra != NULL) {
180 		ra->ra_flags = 0;
181 	}
182 
183 	return ra;
184 }
185 
186 /*
187  * uvm_ra_freectx: free a context.
188  */
189 
190 void
191 uvm_ra_freectx(struct uvm_ractx *ra)
192 {
193 
194 	KASSERT(ra != NULL);
195 	ra_freectx(ra);
196 }
197 
198 /*
199  * uvm_ra_request: update a read-ahead context and start i/o if appropriate.
200  *
201  * => called when [reqoff, reqoff+reqsize) is requested.
202  * => object must be locked by caller, will return locked.
203  */
204 
205 void
206 uvm_ra_request(struct uvm_ractx *ra, int advice, struct uvm_object *uobj,
207     off_t reqoff, size_t reqsize)
208 {
209 
210 	KASSERT(mutex_owned(&uobj->vmobjlock));
211 
212 	if (ra == NULL || advice == UVM_ADV_RANDOM) {
213 		return;
214 	}
215 
216 	if (advice == UVM_ADV_SEQUENTIAL) {
217 
218 		/*
219 		 * always do read-ahead with a large window.
220 		 */
221 
222 		if ((ra->ra_flags & RA_VALID) == 0) {
223 			ra->ra_winstart = ra->ra_next = 0;
224 			ra->ra_flags |= RA_VALID;
225 		}
226 		if (reqoff < ra->ra_winstart) {
227 			ra->ra_next = reqoff;
228 		}
229 		ra->ra_winsize = RA_WINSIZE_SEQENTIAL;
230 		goto do_readahead;
231 	}
232 
233 	/*
234 	 * a request with UVM_ADV_NORMAL hint.  (ie. no hint)
235 	 *
236 	 * we keep a sliding window in order to determine:
237 	 *	- if the previous read-ahead was successful or not.
238 	 *	- how many bytes to read-ahead.
239 	 */
240 
241 	/*
242 	 * if it's the first request for this context,
243 	 * initialize context and return.
244 	 */
245 
246 	if ((ra->ra_flags & RA_VALID) == 0) {
247 initialize:
248 		ra->ra_winstart = ra->ra_next = reqoff + reqsize;
249 		ra->ra_winsize = RA_WINSIZE_INIT;
250 		ra->ra_flags |= RA_VALID;
251 		goto done;
252 	}
253 
254 	/*
255 	 * if it isn't in our window,
256 	 * initialize context and return.
257 	 * (read-ahead miss)
258 	 */
259 
260 	if (reqoff < ra->ra_winstart ||
261 	    ra->ra_winstart + ra->ra_winsize < reqoff) {
262 
263 		/*
264 		 * ... unless we seem to be reading the same chunk repeatedly.
265 		 *
266 		 * XXX should have some margin?
267 		 */
268 
269 		if (reqoff + reqsize == ra->ra_winstart) {
270 			DPRINTF(("%s: %p: same block: off=%" PRIu64
271 			    ", size=%zd, winstart=%" PRIu64 "\n",
272 			    __func__, ra, reqoff, reqsize, ra->ra_winstart));
273 			goto done;
274 		}
275 		goto initialize;
276 	}
277 
278 	/*
279 	 * it's in our window. (read-ahead hit)
280 	 *	- start read-ahead i/o if appropriate.
281 	 *	- advance and enlarge window.
282 	 */
283 
284 do_readahead:
285 
286 	/*
287 	 * don't bother to read-ahead behind current request.
288 	 */
289 
290 	if (reqoff > ra->ra_next) {
291 		ra->ra_next = reqoff;
292 	}
293 
294 	/*
295 	 * try to make [reqoff, reqoff+ra_winsize) in-core.
296 	 * note that [reqoff, ra_next) is considered already done.
297 	 */
298 
299 	if (reqoff + ra->ra_winsize > ra->ra_next) {
300 		off_t raoff = MAX(reqoff, ra->ra_next);
301 		size_t rasize = reqoff + ra->ra_winsize - ra->ra_next;
302 
303 #if defined(DIAGNOSTIC)
304 		if (rasize > RA_WINSIZE_MAX) {
305 			printf("%s: corrupted context", __func__);
306 			rasize = RA_WINSIZE_MAX;
307 		}
308 #endif /* defined(DIAGNOSTIC) */
309 
310 		/*
311 		 * issue read-ahead only if we can start big enough i/o.
312 		 * otherwise we end up with a stream of small i/o.
313 		 */
314 
315 		if (rasize >= RA_MINSIZE) {
316 			off_t next;
317 
318 			mutex_exit(&uobj->vmobjlock);
319 			next = ra_startio(uobj, raoff, rasize);
320 			mutex_enter(&uobj->vmobjlock);
321 			ra->ra_next = next;
322 		}
323 	}
324 
325 	/*
326 	 * update window.
327 	 *
328 	 * enlarge window by reqsize, so that it grows in a predictable manner
329 	 * regardless of the size of each read(2).
330 	 */
331 
332 	ra->ra_winstart = reqoff + reqsize;
333 	ra->ra_winsize = MIN(RA_WINSIZE_MAX, ra->ra_winsize + reqsize);
334 
335 done:;
336 }
337 
338 int
339 uvm_readahead(struct uvm_object *uobj, off_t off, off_t size)
340 {
341 
342 	/*
343 	 * don't allow too much read-ahead.
344 	 */
345 	if (size > RA_WINSIZE_MAX) {
346 		size = RA_WINSIZE_MAX;
347 	}
348 	ra_startio(uobj, off, size);
349 	return 0;
350 }
351