Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
/*	$NetBSD: nfs_fha.c,v 1.2 2016/12/13 22:41:46 pgoyette Exp $	*/
/*-
 * Copyright (c) 2008 Isilon Inc http://www.isilon.com/
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

#include <sys/cdefs.h>
/* __FBSDID("FreeBSD: head/sys/nfs/nfs_fha.c 267479 2014-06-14 12:26:12Z mav "); */
__RCSID("$NetBSD: nfs_fha.c,v 1.2 2016/12/13 22:41:46 pgoyette Exp $");

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/sysproto.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/vnode.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mbuf.h>
#include <sys/sbuf.h>

#include <rpc/rpc.h>
#include <fs/nfs/common/nfs_fha.h>

static MALLOC_DEFINE(M_NFS_FHA, "NFS FHA", "NFS FHA");

/*
 * XXX need to commonize definitions between old and new NFS code.  Define
 * this here so we don't include one nfsproto.h over the other.
 */
#define	NFS_PROG		100003

void
fha_init(struct fha_params *softc)
{
	char tmpstr[128];
	int i;

	for (i = 0; i < FHA_HASH_SIZE; i++)
		mtx_init(&softc->fha_hash[i].mtx, "fhalock", NULL, MTX_DEF);

	/*
	 * Set the default tuning parameters.
	 */
	softc->ctls.enable = FHA_DEF_ENABLE;
	softc->ctls.bin_shift = FHA_DEF_BIN_SHIFT;
	softc->ctls.max_nfsds_per_fh = FHA_DEF_MAX_NFSDS_PER_FH;
	softc->ctls.max_reqs_per_nfsd = FHA_DEF_MAX_REQS_PER_NFSD;

	/*
	 * Allow the user to override the defaults at boot time with
	 * tunables.
	 */
	snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.enable",
	    softc->server_name);
	TUNABLE_INT_FETCH(tmpstr, &softc->ctls.enable);
	snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.bin_shift",
	    softc->server_name);
	TUNABLE_INT_FETCH(tmpstr, &softc->ctls.bin_shift);
	snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.max_nfsds_per_fh",
	    softc->server_name);
	TUNABLE_INT_FETCH(tmpstr, &softc->ctls.max_nfsds_per_fh);
	snprintf(tmpstr, sizeof(tmpstr), "vfs.%s.fha.max_reqs_per_nfsd",
	    softc->server_name);
	TUNABLE_INT_FETCH(tmpstr, &softc->ctls.max_reqs_per_nfsd);

	/*
	 * Add sysctls so the user can change the tuning parameters at
	 * runtime.
	 */
	SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "enable", CTLFLAG_RW,
	    &softc->ctls.enable, 0, "Enable NFS File Handle Affinity (FHA)");

	SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "bin_shift", CTLFLAG_RW,
	    &softc->ctls.bin_shift, 0, "For FHA reads, no two requests will "
	    "contend if they're 2^(bin_shift) bytes apart");

	SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "max_nfsds_per_fh", CTLFLAG_RW,
	    &softc->ctls.max_nfsds_per_fh, 0, "Maximum nfsd threads that "
	    "should be working on requests for the same file handle");

	SYSCTL_ADD_UINT(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "max_reqs_per_nfsd", CTLFLAG_RW,
	    &softc->ctls.max_reqs_per_nfsd, 0, "Maximum requests that "
	    "single nfsd thread should be working on at any time");

	SYSCTL_ADD_OID(&softc->sysctl_ctx, SYSCTL_CHILDREN(softc->sysctl_tree),
	    OID_AUTO, "fhe_stats", CTLTYPE_STRING | CTLFLAG_RD, 0, 0,
	    softc->callbacks.fhe_stats_sysctl, "A", "");

}

void
fha_uninit(struct fha_params *softc)
{
	int i;

	sysctl_ctx_free(&softc->sysctl_ctx);
	for (i = 0; i < FHA_HASH_SIZE; i++)
		mtx_destroy(&softc->fha_hash[i].mtx);
}

/*
 * This just specifies that offsets should obey affinity when within
 * the same 1Mbyte (1<<20) chunk for the file (reads only for now).
 */
static void
fha_extract_info(struct svc_req *req, struct fha_info *i,
    struct fha_callbacks *cb)
{
	struct mbuf *md;
	caddr_t dpos;
	static u_int64_t random_fh = 0;
	int error;
	int v3 = (req->rq_vers == 3);
	rpcproc_t procnum;

	/*
	 * We start off with a random fh.  If we get a reasonable
	 * procnum, we set the fh.  If there's a concept of offset
	 * that we're interested in, we set that.
	 */
	i->fh = ++random_fh;
	i->offset = 0;
	i->locktype = LK_EXCLUSIVE;

	/*
	 * Extract the procnum and convert to v3 form if necessary,
	 * taking care to deal with out-of-range procnums.  Caller will
	 * ensure that rq_vers is either 2 or 3.
	 */
	procnum = req->rq_proc;
	if (!v3) {
		rpcproc_t tmp_procnum;

		tmp_procnum = cb->get_procnum(procnum);
		if (tmp_procnum == -1)
			goto out;
		procnum = tmp_procnum;
	}

	/*
	 * We do affinity for most.  However, we divide a realm of affinity
	 * by file offset so as to allow for concurrent random access.  We
	 * only do this for reads today, but this may change when IFS supports
	 * efficient concurrent writes.
	 */
	if (cb->no_offset(procnum))
		goto out;

	error = cb->realign(&req->rq_args, M_NOWAIT);
	if (error)
		goto out;
	md = req->rq_args;
	dpos = mtod(md, caddr_t);

	/* Grab the filehandle. */
	error = cb->get_fh(&i->fh, v3, &md, &dpos);
	if (error)
		goto out;

	/* Content ourselves with zero offset for all but reads. */
	if (cb->is_read(procnum) || cb->is_write(procnum))
		cb->get_offset(&md, &dpos, v3, i);

out:
	cb->set_locktype(procnum, i);
}

static struct fha_hash_entry *
fha_hash_entry_new(u_int64_t fh)
{
	struct fha_hash_entry *e;

	e = malloc(sizeof(*e), M_NFS_FHA, M_WAITOK);
	e->fh = fh;
	e->num_rw = 0;
	e->num_exclusive = 0;
	e->num_threads = 0;
	LIST_INIT(&e->threads);

	return (e);
}

static void
fha_hash_entry_destroy(struct fha_hash_entry *e)
{

	mtx_assert(e->mtx, MA_OWNED);
	KASSERT(e->num_rw == 0,
	    ("%d reqs on destroyed fhe %p", e->num_rw, e));
	KASSERT(e->num_exclusive == 0,
	    ("%d exclusive reqs on destroyed fhe %p", e->num_exclusive, e));
	KASSERT(e->num_threads == 0,
	    ("%d threads on destroyed fhe %p", e->num_threads, e));
	free(e, M_NFS_FHA);
}

static void
fha_hash_entry_remove(struct fha_hash_entry *e)
{

	mtx_assert(e->mtx, MA_OWNED);
	LIST_REMOVE(e, link);
	fha_hash_entry_destroy(e);
}

static struct fha_hash_entry *
fha_hash_entry_lookup(struct fha_params *softc, u_int64_t fh)
{
	SVCPOOL *pool;
	struct fha_hash_slot *fhs;
	struct fha_hash_entry *fhe, *new_fhe;

	pool = *softc->pool;
	fhs = &softc->fha_hash[fh % FHA_HASH_SIZE];
	new_fhe = fha_hash_entry_new(fh);
	new_fhe->mtx = &fhs->mtx;
	mtx_lock(&fhs->mtx);
	LIST_FOREACH(fhe, &fhs->list, link)
		if (fhe->fh == fh)
			break;
	if (!fhe) {
		fhe = new_fhe;
		LIST_INSERT_HEAD(&fhs->list, fhe, link);
	} else
		fha_hash_entry_destroy(new_fhe);
	return (fhe);
}

static void
fha_hash_entry_add_thread(struct fha_hash_entry *fhe, SVCTHREAD *thread)
{

	mtx_assert(fhe->mtx, MA_OWNED);
	thread->st_p2 = 0;
	LIST_INSERT_HEAD(&fhe->threads, thread, st_alink);
	fhe->num_threads++;
}

static void
fha_hash_entry_remove_thread(struct fha_hash_entry *fhe, SVCTHREAD *thread)
{

	mtx_assert(fhe->mtx, MA_OWNED);
	KASSERT(thread->st_p2 == 0,
	    ("%d reqs on removed thread %p", thread->st_p2, thread));
	LIST_REMOVE(thread, st_alink);
	fhe->num_threads--;
}

/*
 * Account for an ongoing operation associated with this file.
 */
static void
fha_hash_entry_add_op(struct fha_hash_entry *fhe, int locktype, int count)
{

	mtx_assert(fhe->mtx, MA_OWNED);
	if (LK_EXCLUSIVE == locktype)
		fhe->num_exclusive += count;
	else
		fhe->num_rw += count;
}

/*
 * Get the service thread currently associated with the fhe that is
 * appropriate to handle this operation.
 */
static SVCTHREAD *
fha_hash_entry_choose_thread(struct fha_params *softc,
    struct fha_hash_entry *fhe, struct fha_info *i, SVCTHREAD *this_thread)
{
	SVCTHREAD *thread, *min_thread = NULL;
	SVCPOOL *pool;
	int req_count, min_count = 0;
	off_t offset1, offset2;

	pool = *softc->pool;

	LIST_FOREACH(thread, &fhe->threads, st_alink) {
		req_count = thread->st_p2;

		/* If there are any writes in progress, use the first thread. */
		if (fhe->num_exclusive) {
#if 0
			ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
			    "fha: %p(%d)w", thread, req_count);
#endif
			return (thread);
		}

		/*
		 * Check for read locality, making sure that we won't
		 * exceed our per-thread load limit in the process.
		 */
		offset1 = i->offset;
		offset2 = thread->st_p3;

		if (((offset1 >= offset2)
		  && ((offset1 - offset2) < (1 << softc->ctls.bin_shift)))
		 || ((offset2 > offset1)
		  && ((offset2 - offset1) < (1 << softc->ctls.bin_shift)))) {
			if ((softc->ctls.max_reqs_per_nfsd == 0) ||
			    (req_count < softc->ctls.max_reqs_per_nfsd)) {
#if 0
				ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
				    "fha: %p(%d)r", thread, req_count);
#endif
				return (thread);
			}
		}

		/*
		 * We don't have a locality match, so skip this thread,
		 * but keep track of the most attractive thread in case
		 * we need to come back to it later.
		 */
#if 0
		ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
		    "fha: %p(%d)s off1 %llu off2 %llu", thread,
		    req_count, offset1, offset2);
#endif
		if ((min_thread == NULL) || (req_count < min_count)) {
			min_count = req_count;
			min_thread = thread;
		}
	}

	/*
	 * We didn't find a good match yet.  See if we can add
	 * a new thread to this file handle entry's thread list.
	 */
	if ((softc->ctls.max_nfsds_per_fh == 0) ||
	    (fhe->num_threads < softc->ctls.max_nfsds_per_fh)) {
		thread = this_thread;
#if 0
		ITRACE_CURPROC(ITRACE_NFS, ITRACE_INFO,
		    "fha: %p(%d)t", thread, thread->st_p2);
#endif
		fha_hash_entry_add_thread(fhe, thread);
	} else {
		/*
		 * We don't want to use any more threads for this file, so
		 * go back to the most attractive nfsd we're already using.
		 */
		thread = min_thread;
	}

	return (thread);
}

/*
 * After getting a request, try to assign it to some thread.  Usually we
 * handle it ourselves.
 */
SVCTHREAD *
fha_assign(SVCTHREAD *this_thread, struct svc_req *req,
    struct fha_params *softc)
{
	SVCTHREAD *thread;
	struct fha_info i;
	struct fha_hash_entry *fhe;
	struct fha_callbacks *cb;

	cb = &softc->callbacks;

	/* Check to see whether we're enabled. */
	if (softc->ctls.enable == 0)
		goto thist;

	/*
	 * Only do placement if this is an NFS request.
	 */
	if (req->rq_prog != NFS_PROG)
		goto thist;

	if (req->rq_vers != 2 && req->rq_vers != 3)
		goto thist;

	fha_extract_info(req, &i, cb);

	/*
	 * We save the offset associated with this request for later
	 * nfsd matching.
	 */
	fhe = fha_hash_entry_lookup(softc, i.fh);
	req->rq_p1 = fhe;
	req->rq_p2 = i.locktype;
	req->rq_p3 = i.offset;

	/*
	 * Choose a thread, taking into consideration locality, thread load,
	 * and the number of threads already working on this file.
	 */
	thread = fha_hash_entry_choose_thread(softc, fhe, &i, this_thread);
	KASSERT(thread, ("fha_assign: NULL thread!"));
	fha_hash_entry_add_op(fhe, i.locktype, 1);
	thread->st_p2++;
	thread->st_p3 = i.offset;

	/*
	 * Grab the pool lock here to not let chosen thread go away before
	 * the new request inserted to its queue while we drop fhe lock.
	 */
	mtx_lock(&thread->st_lock);
	mtx_unlock(fhe->mtx);

	return (thread);
thist:
	req->rq_p1 = NULL;
	mtx_lock(&this_thread->st_lock);
	return (this_thread);
}

/*
 * Called when we're done with an operation.  The request has already
 * been de-queued.
 */
void
fha_nd_complete(SVCTHREAD *thread, struct svc_req *req)
{
	struct fha_hash_entry *fhe = req->rq_p1;
	struct mtx *mtx;

	/*
	 * This may be called for reqs that didn't go through
	 * fha_assign (e.g. extra NULL ops used for RPCSEC_GSS.
	 */
	if (!fhe)
		return;

	mtx = fhe->mtx;
	mtx_lock(mtx);
	fha_hash_entry_add_op(fhe, req->rq_p2, -1);
	thread->st_p2--;
	KASSERT(thread->st_p2 >= 0, ("Negative request count %d on %p",
	    thread->st_p2, thread));
	if (thread->st_p2 == 0) {
		fha_hash_entry_remove_thread(fhe, thread);
		if (0 == fhe->num_rw + fhe->num_exclusive)
			fha_hash_entry_remove(fhe);
	}
	mtx_unlock(mtx);
}

int
fhe_stats_sysctl(SYSCTL_HANDLER_ARGS, struct fha_params *softc)
{
	int error, i;
	struct sbuf sb;
	struct fha_hash_entry *fhe;
	bool_t first, hfirst;
	SVCTHREAD *thread;
	SVCPOOL *pool;

	sbuf_new(&sb, NULL, 65536, SBUF_FIXEDLEN);

	pool = NULL;

	if (!*softc->pool) {
		sbuf_printf(&sb, "NFSD not running\n");
		goto out;
	}
	pool = *softc->pool;

	for (i = 0; i < FHA_HASH_SIZE; i++)
		if (!LIST_EMPTY(&softc->fha_hash[i].list))
			break;

	if (i == FHA_HASH_SIZE) {
		sbuf_printf(&sb, "No file handle entries.\n");
		goto out;
	}

	hfirst = TRUE;
	for (; i < FHA_HASH_SIZE; i++) {
		mtx_lock(&softc->fha_hash[i].mtx);
		if (LIST_EMPTY(&softc->fha_hash[i].list)) {
			mtx_unlock(&softc->fha_hash[i].mtx);
			continue;
		}
		sbuf_printf(&sb, "%shash %d: {\n", hfirst ? "" : ", ", i);
		first = TRUE;
		LIST_FOREACH(fhe, &softc->fha_hash[i].list, link) {
			sbuf_printf(&sb, "%sfhe %p: {\n", first ? "  " : ", ", fhe);

			sbuf_printf(&sb, "    fh: %ju\n", (uintmax_t) fhe->fh);
			sbuf_printf(&sb, "    num_rw/exclusive: %d/%d\n",
			    fhe->num_rw, fhe->num_exclusive);
			sbuf_printf(&sb, "    num_threads: %d\n", fhe->num_threads);

			LIST_FOREACH(thread, &fhe->threads, st_alink) {
				sbuf_printf(&sb, "      thread %p offset %ju "
				    "reqs %d\n", thread,
				    thread->st_p3, thread->st_p2);
			}

			sbuf_printf(&sb, "  }");
			first = FALSE;
		}
		sbuf_printf(&sb, "\n}");
		mtx_unlock(&softc->fha_hash[i].mtx);
		hfirst = FALSE;
	}

 out:
	sbuf_trim(&sb);
	sbuf_finish(&sb);
	error = sysctl_handle_string(oidp, sbuf_data(&sb), sbuf_len(&sb), req);
	sbuf_delete(&sb);
	return (error);
}