2 * Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
3 * Copyright (C) 2007 The Regents of the University of California.
4 * Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
5 * Written by Brian Behlendorf <behlendorf1@llnl.gov>.
8 * This file is part of the SPL, Solaris Porting Layer.
10 * The SPL is free software; you can redistribute it and/or modify it
11 * under the terms of the GNU General Public License as published by the
12 * Free Software Foundation; either version 2 of the License, or (at your
13 * option) any later version.
15 * The SPL is distributed in the hope that it will be useful, but WITHOUT
16 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
17 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
20 * You should have received a copy of the GNU General Public License along
21 * with the SPL. If not, see <http://www.gnu.org/licenses/>.
23 * Solaris Porting Layer (SPL) Proc Implementation.
26 #include <sys/systeminfo.h>
27 #include <sys/kstat.h>
29 #include <sys/kmem_cache.h>
31 #include <sys/taskq.h>
33 #include <linux/ctype.h>
34 #include <linux/kmod.h>
35 #include <linux/seq_file.h>
36 #include <linux/uaccess.h>
37 #include <linux/version.h>
39 #if defined(CONSTIFY_PLUGIN) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
40 typedef struct ctl_table __no_const spl_ctl_table;
42 typedef struct ctl_table spl_ctl_table;
45 static unsigned long table_min = 0;
46 static unsigned long table_max = ~0;
48 static struct ctl_table_header *spl_header = NULL;
49 static struct proc_dir_entry *proc_spl = NULL;
50 static struct proc_dir_entry *proc_spl_kmem = NULL;
51 static struct proc_dir_entry *proc_spl_kmem_slab = NULL;
52 static struct proc_dir_entry *proc_spl_taskq_all = NULL;
53 static struct proc_dir_entry *proc_spl_taskq = NULL;
54 struct proc_dir_entry *proc_spl_kstat = NULL;
57 proc_copyin_string(char *kbuffer, int kbuffer_size, const char *ubuffer,
62 if (ubuffer_size > kbuffer_size)
65 if (copy_from_user((void *)kbuffer, (void *)ubuffer, ubuffer_size))
68 /* strip trailing whitespace */
69 size = strnlen(kbuffer, ubuffer_size);
71 if (!isspace(kbuffer[size]))
78 /* no space to terminate */
79 if (size == kbuffer_size)
82 kbuffer[size + 1] = 0;
87 proc_copyout_string(char *ubuffer, int ubuffer_size, const char *kbuffer,
91 * NB if 'append' != NULL, it's a single character to append to the
92 * copied out string - usually "\n", for /proc entries and
93 * (i.e. a terminating zero byte) for sysctl entries
95 int size = MIN(strlen(kbuffer), ubuffer_size);
97 if (copy_to_user(ubuffer, kbuffer, size))
100 if (append != NULL && size < ubuffer_size) {
101 if (copy_to_user(ubuffer + size, append, 1))
112 proc_domemused(struct ctl_table *table, int write,
113 void __user *buffer, size_t *lenp, loff_t *ppos)
116 unsigned long min = 0, max = ~0, val;
117 spl_ctl_table dummy = *table;
120 dummy.proc_handler = &proc_dointvec;
127 #ifdef HAVE_ATOMIC64_T
128 val = atomic64_read((atomic64_t *)table->data);
130 val = atomic_read((atomic_t *)table->data);
131 #endif /* HAVE_ATOMIC64_T */
132 rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos);
137 #endif /* DEBUG_KMEM */
140 proc_doslab(struct ctl_table *table, int write,
141 void __user *buffer, size_t *lenp, loff_t *ppos)
144 unsigned long min = 0, max = ~0, val = 0, mask;
145 spl_ctl_table dummy = *table;
146 spl_kmem_cache_t *skc = NULL;
149 dummy.proc_handler = &proc_dointvec;
156 down_read(&spl_kmem_cache_sem);
157 mask = (unsigned long)table->data;
159 list_for_each_entry(skc, &spl_kmem_cache_list, skc_list) {
161 /* Only use slabs of the correct kmem/vmem type */
162 if (!(skc->skc_flags & mask))
165 /* Sum the specified field for selected slabs */
166 switch (mask & (KMC_TOTAL | KMC_ALLOC | KMC_MAX)) {
168 val += skc->skc_slab_size * skc->skc_slab_total;
171 val += skc->skc_obj_size * skc->skc_obj_alloc;
174 val += skc->skc_obj_size * skc->skc_obj_max;
179 up_read(&spl_kmem_cache_sem);
180 rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos);
187 proc_dohostid(struct ctl_table *table, int write,
188 void __user *buffer, size_t *lenp, loff_t *ppos)
195 * We can't use proc_doulongvec_minmax() in the write
196 * case here because hostid while a hex value has no
197 * leading 0x which confuses the helper function.
199 rc = proc_copyin_string(str, sizeof (str), buffer, *lenp);
203 spl_hostid = simple_strtoul(str, &end, 16);
208 len = snprintf(str, sizeof (str), "%lx",
209 (unsigned long) zone_get_hostid(NULL));
213 rc = proc_copyout_string(buffer,
214 *lenp, str + *ppos, "\n");
226 taskq_seq_show_headers(struct seq_file *f)
228 seq_printf(f, "%-25s %5s %5s %5s %5s %5s %5s %12s %5s %10s\n",
229 "taskq", "act", "nthr", "spwn", "maxt", "pri",
230 "mina", "maxa", "cura", "flags");
233 /* indices into the lheads array below */
236 #define LHEAD_DELAY 2
238 #define LHEAD_ACTIVE 4
242 static unsigned int spl_max_show_tasks = 512;
243 module_param(spl_max_show_tasks, uint, 0644);
244 MODULE_PARM_DESC(spl_max_show_tasks, "Max number of tasks shown in taskq proc");
248 taskq_seq_show_impl(struct seq_file *f, void *p, boolean_t allflag)
251 taskq_thread_t *tqt = NULL;
252 spl_wait_queue_entry_t *wq;
253 struct task_struct *tsk;
256 struct list_head *lheads[LHEAD_SIZE], *lh;
257 static char *list_names[LHEAD_SIZE] =
258 {"pend", "prio", "delay", "wait", "active" };
259 int i, j, have_lheads = 0;
260 unsigned long wflags, flags;
262 spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
263 spin_lock_irqsave(&tq->tq_wait_waitq.lock, wflags);
265 /* get the various lists and check whether they're empty */
266 lheads[LHEAD_PEND] = &tq->tq_pend_list;
267 lheads[LHEAD_PRIO] = &tq->tq_prio_list;
268 lheads[LHEAD_DELAY] = &tq->tq_delay_list;
269 #ifdef HAVE_WAIT_QUEUE_HEAD_ENTRY
270 lheads[LHEAD_WAIT] = &tq->tq_wait_waitq.head;
272 lheads[LHEAD_WAIT] = &tq->tq_wait_waitq.task_list;
274 lheads[LHEAD_ACTIVE] = &tq->tq_active_list;
276 for (i = 0; i < LHEAD_SIZE; ++i) {
277 if (list_empty(lheads[i]))
283 /* early return in non-"all" mode if lists are all empty */
284 if (!allflag && !have_lheads) {
285 spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
286 spin_unlock_irqrestore(&tq->tq_lock, flags);
290 /* unlock the waitq quickly */
291 if (!lheads[LHEAD_WAIT])
292 spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
294 /* show the base taskq contents */
295 snprintf(name, sizeof (name), "%s/%d", tq->tq_name, tq->tq_instance);
296 seq_printf(f, "%-25s ", name);
297 seq_printf(f, "%5d %5d %5d %5d %5d %5d %12d %5d %10x\n",
298 tq->tq_nactive, tq->tq_nthreads, tq->tq_nspawn,
299 tq->tq_maxthreads, tq->tq_pri, tq->tq_minalloc, tq->tq_maxalloc,
300 tq->tq_nalloc, tq->tq_flags);
302 /* show the active list */
303 if (lheads[LHEAD_ACTIVE]) {
305 list_for_each_entry(tqt, &tq->tq_active_list, tqt_active_list) {
307 seq_printf(f, "\t%s:",
308 list_names[LHEAD_ACTIVE]);
310 seq_printf(f, "\n\t ");
313 seq_printf(f, " [%d]%pf(%ps)",
314 tqt->tqt_thread->pid,
315 tqt->tqt_task->tqent_func,
316 tqt->tqt_task->tqent_arg);
322 for (i = LHEAD_PEND; i <= LHEAD_WAIT; ++i)
325 list_for_each(lh, lheads[i]) {
326 if (spl_max_show_tasks != 0 &&
327 j >= spl_max_show_tasks) {
328 seq_printf(f, "\n\t(truncated)");
331 /* show the wait waitq list */
332 if (i == LHEAD_WAIT) {
333 #ifdef HAVE_WAIT_QUEUE_HEAD_ENTRY
335 spl_wait_queue_entry_t, entry);
338 spl_wait_queue_entry_t, task_list);
341 seq_printf(f, "\t%s:",
344 seq_printf(f, "\n\t ");
347 seq_printf(f, " %d", tsk->pid);
348 /* pend, prio and delay lists */
350 tqe = list_entry(lh, taskq_ent_t,
353 seq_printf(f, "\t%s:",
356 seq_printf(f, "\n\t ");
358 seq_printf(f, " %pf(%ps)",
366 if (lheads[LHEAD_WAIT])
367 spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
368 spin_unlock_irqrestore(&tq->tq_lock, flags);
374 taskq_all_seq_show(struct seq_file *f, void *p)
376 return (taskq_seq_show_impl(f, p, B_TRUE));
380 taskq_seq_show(struct seq_file *f, void *p)
382 return (taskq_seq_show_impl(f, p, B_FALSE));
386 taskq_seq_start(struct seq_file *f, loff_t *pos)
391 down_read(&tq_list_sem);
393 taskq_seq_show_headers(f);
402 return (list_entry(p, taskq_t, tq_taskqs));
406 taskq_seq_next(struct seq_file *f, void *p, loff_t *pos)
411 return ((tq->tq_taskqs.next == &tq_list) ?
412 NULL : list_entry(tq->tq_taskqs.next, taskq_t, tq_taskqs));
416 slab_seq_show_headers(struct seq_file *f)
419 "--------------------- cache ----------"
420 "--------------------------------------------- "
423 "--- emergency ---\n");
426 " flags size alloc slabsize objsize "
429 "dlock alloc max\n");
433 slab_seq_show(struct seq_file *f, void *p)
435 spl_kmem_cache_t *skc = p;
437 ASSERT(skc->skc_magic == SKC_MAGIC);
439 if (skc->skc_flags & KMC_SLAB) {
441 * This cache is backed by a generic Linux kmem cache which
442 * has its own accounting. For these caches we only track
443 * the number of active allocated objects that exist within
444 * the underlying Linux slabs. For the overall statistics of
445 * the underlying Linux cache please refer to /proc/slabinfo.
447 spin_lock(&skc->skc_lock);
448 uint64_t objs_allocated =
449 percpu_counter_sum(&skc->skc_linux_alloc);
450 seq_printf(f, "%-36s ", skc->skc_name);
451 seq_printf(f, "0x%05lx %9s %9lu %8s %8u "
452 "%5s %5s %5s %5s %5lu %5s %5s %5s %5s\n",
453 (long unsigned)skc->skc_flags,
455 (long unsigned)(skc->skc_obj_size * objs_allocated),
457 (unsigned)skc->skc_obj_size,
459 (long unsigned)objs_allocated,
461 spin_unlock(&skc->skc_lock);
465 spin_lock(&skc->skc_lock);
466 seq_printf(f, "%-36s ", skc->skc_name);
467 seq_printf(f, "0x%05lx %9lu %9lu %8u %8u "
468 "%5lu %5lu %5lu %5lu %5lu %5lu %5lu %5lu %5lu\n",
469 (long unsigned)skc->skc_flags,
470 (long unsigned)(skc->skc_slab_size * skc->skc_slab_total),
471 (long unsigned)(skc->skc_obj_size * skc->skc_obj_alloc),
472 (unsigned)skc->skc_slab_size,
473 (unsigned)skc->skc_obj_size,
474 (long unsigned)skc->skc_slab_total,
475 (long unsigned)skc->skc_slab_alloc,
476 (long unsigned)skc->skc_slab_max,
477 (long unsigned)skc->skc_obj_total,
478 (long unsigned)skc->skc_obj_alloc,
479 (long unsigned)skc->skc_obj_max,
480 (long unsigned)skc->skc_obj_deadlock,
481 (long unsigned)skc->skc_obj_emergency,
482 (long unsigned)skc->skc_obj_emergency_max);
483 spin_unlock(&skc->skc_lock);
488 slab_seq_start(struct seq_file *f, loff_t *pos)
493 down_read(&spl_kmem_cache_sem);
495 slab_seq_show_headers(f);
497 p = spl_kmem_cache_list.next;
500 if (p == &spl_kmem_cache_list)
504 return (list_entry(p, spl_kmem_cache_t, skc_list));
508 slab_seq_next(struct seq_file *f, void *p, loff_t *pos)
510 spl_kmem_cache_t *skc = p;
513 return ((skc->skc_list.next == &spl_kmem_cache_list) ?
514 NULL : list_entry(skc->skc_list.next, spl_kmem_cache_t, skc_list));
518 slab_seq_stop(struct seq_file *f, void *v)
520 up_read(&spl_kmem_cache_sem);
523 static struct seq_operations slab_seq_ops = {
524 .show = slab_seq_show,
525 .start = slab_seq_start,
526 .next = slab_seq_next,
527 .stop = slab_seq_stop,
531 proc_slab_open(struct inode *inode, struct file *filp)
533 return (seq_open(filp, &slab_seq_ops));
536 static const kstat_proc_op_t proc_slab_operations = {
537 #ifdef HAVE_PROC_OPS_STRUCT
538 .proc_open = proc_slab_open,
539 .proc_read = seq_read,
540 .proc_lseek = seq_lseek,
541 .proc_release = seq_release,
543 .open = proc_slab_open,
546 .release = seq_release,
551 taskq_seq_stop(struct seq_file *f, void *v)
553 up_read(&tq_list_sem);
556 static struct seq_operations taskq_all_seq_ops = {
557 .show = taskq_all_seq_show,
558 .start = taskq_seq_start,
559 .next = taskq_seq_next,
560 .stop = taskq_seq_stop,
563 static struct seq_operations taskq_seq_ops = {
564 .show = taskq_seq_show,
565 .start = taskq_seq_start,
566 .next = taskq_seq_next,
567 .stop = taskq_seq_stop,
571 proc_taskq_all_open(struct inode *inode, struct file *filp)
573 return (seq_open(filp, &taskq_all_seq_ops));
577 proc_taskq_open(struct inode *inode, struct file *filp)
579 return (seq_open(filp, &taskq_seq_ops));
582 static const kstat_proc_op_t proc_taskq_all_operations = {
583 #ifdef HAVE_PROC_OPS_STRUCT
584 .proc_open = proc_taskq_all_open,
585 .proc_read = seq_read,
586 .proc_lseek = seq_lseek,
587 .proc_release = seq_release,
589 .open = proc_taskq_all_open,
592 .release = seq_release,
596 static const kstat_proc_op_t proc_taskq_operations = {
597 #ifdef HAVE_PROC_OPS_STRUCT
598 .proc_open = proc_taskq_open,
599 .proc_read = seq_read,
600 .proc_lseek = seq_lseek,
601 .proc_release = seq_release,
603 .open = proc_taskq_open,
606 .release = seq_release,
610 static struct ctl_table spl_kmem_table[] = {
613 .procname = "kmem_used",
614 .data = &kmem_alloc_used,
615 #ifdef HAVE_ATOMIC64_T
616 .maxlen = sizeof (atomic64_t),
618 .maxlen = sizeof (atomic_t),
619 #endif /* HAVE_ATOMIC64_T */
621 .proc_handler = &proc_domemused,
624 .procname = "kmem_max",
625 .data = &kmem_alloc_max,
626 .maxlen = sizeof (unsigned long),
627 .extra1 = &table_min,
628 .extra2 = &table_max,
630 .proc_handler = &proc_doulongvec_minmax,
632 #endif /* DEBUG_KMEM */
634 .procname = "slab_kvmem_total",
635 .data = (void *)(KMC_KVMEM | KMC_TOTAL),
636 .maxlen = sizeof (unsigned long),
637 .extra1 = &table_min,
638 .extra2 = &table_max,
640 .proc_handler = &proc_doslab,
643 .procname = "slab_kvmem_alloc",
644 .data = (void *)(KMC_KVMEM | KMC_ALLOC),
645 .maxlen = sizeof (unsigned long),
646 .extra1 = &table_min,
647 .extra2 = &table_max,
649 .proc_handler = &proc_doslab,
652 .procname = "slab_kvmem_max",
653 .data = (void *)(KMC_KVMEM | KMC_MAX),
654 .maxlen = sizeof (unsigned long),
655 .extra1 = &table_min,
656 .extra2 = &table_max,
658 .proc_handler = &proc_doslab,
663 static struct ctl_table spl_kstat_table[] = {
667 static struct ctl_table spl_table[] = {
669 * NB No .strategy entries have been provided since
670 * sysctl(8) prefers to go via /proc for portability.
673 .procname = "gitrev",
675 .maxlen = sizeof (spl_gitrev),
677 .proc_handler = &proc_dostring,
680 .procname = "hostid",
682 .maxlen = sizeof (unsigned long),
684 .proc_handler = &proc_dohostid,
689 .child = spl_kmem_table,
694 .child = spl_kstat_table,
699 static struct ctl_table spl_dir[] = {
708 static struct ctl_table spl_root[] = {
710 .procname = "kernel",
722 spl_header = register_sysctl_table(spl_root);
723 if (spl_header == NULL)
726 proc_spl = proc_mkdir("spl", NULL);
727 if (proc_spl == NULL) {
732 proc_spl_taskq_all = proc_create_data("taskq-all", 0444, proc_spl,
733 &proc_taskq_all_operations, NULL);
734 if (proc_spl_taskq_all == NULL) {
739 proc_spl_taskq = proc_create_data("taskq", 0444, proc_spl,
740 &proc_taskq_operations, NULL);
741 if (proc_spl_taskq == NULL) {
746 proc_spl_kmem = proc_mkdir("kmem", proc_spl);
747 if (proc_spl_kmem == NULL) {
752 proc_spl_kmem_slab = proc_create_data("slab", 0444, proc_spl_kmem,
753 &proc_slab_operations, NULL);
754 if (proc_spl_kmem_slab == NULL) {
759 proc_spl_kstat = proc_mkdir("kstat", proc_spl);
760 if (proc_spl_kstat == NULL) {
766 remove_proc_entry("kstat", proc_spl);
767 remove_proc_entry("slab", proc_spl_kmem);
768 remove_proc_entry("kmem", proc_spl);
769 remove_proc_entry("taskq-all", proc_spl);
770 remove_proc_entry("taskq", proc_spl);
771 remove_proc_entry("spl", NULL);
772 unregister_sysctl_table(spl_header);
781 remove_proc_entry("kstat", proc_spl);
782 remove_proc_entry("slab", proc_spl_kmem);
783 remove_proc_entry("kmem", proc_spl);
784 remove_proc_entry("taskq-all", proc_spl);
785 remove_proc_entry("taskq", proc_spl);
786 remove_proc_entry("spl", NULL);
788 ASSERT(spl_header != NULL);
789 unregister_sysctl_table(spl_header);