zfs/module/spl/spl-proc.c

799 lines
19 KiB
C
Raw Normal View History

/*
* Copyright (C) 2007-2010 Lawrence Livermore National Security, LLC.
* Copyright (C) 2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Brian Behlendorf <behlendorf1@llnl.gov>.
* UCRL-CODE-235197
*
* This file is part of the SPL, Solaris Porting Layer.
* For details, see <http://zfsonlinux.org/>.
*
* The SPL is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* Free Software Foundation; either version 2 of the License, or (at your
* option) any later version.
*
* The SPL is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* for more details.
*
* You should have received a copy of the GNU General Public License along
* with the SPL. If not, see <http://www.gnu.org/licenses/>.
*
* Solaris Porting Layer (SPL) Proc Implementation.
*/
#include <sys/systeminfo.h>
#include <sys/kstat.h>
#include <sys/kmem.h>
#include <sys/kmem_cache.h>
#include <sys/vmem.h>
#include <sys/taskq.h>
#include <sys/proc.h>
#include <linux/ctype.h>
#include <linux/kmod.h>
#include <linux/seq_file.h>
#include <linux/uaccess.h>
PaX/GrSecurity Linux 3.8.y compat: Use __no_const on struct ctl_table The PaX team started constifying `struct ctl_table` as of their Linux 3.8.0 patchset. This lead to zfsonlinux/spl#225 and Gentoo bug #463012. While investigating our options, I learned that there is a preprocessor directive called CONSTIFY_PLUGIN that we can use to detect the presence of the PaX changes and adjust the code accordingly. The PaX Team had suggested adopting ctl_table_no_const, but supporting older kernels required declaring that whenever the CONSTIFY_PLUGIN was set. Future compiler changes could potentially cause that to break in the presence of -Werror, so instead we define our own spl_ctl_table typdef and use that. This should be compatible with all PaX kernels. This introduces a Linux kernel version number check to prevent a build failure on versions of the PaX GCC plugin that existed for kernels before Linux 3.8.0. Affected versions of the PaX plugin will trigger a compiler error when they see no_const cast on a non-constified structure. Ordinarily, we would need an autotools check to catch that. However, it is safe to do a kernel version check instead of an autotools check in this specific instance because the affected versions of the PaX GCC plugin only exist for Linux kernels before 3.8.0 and the constification of `struct ctl_table` by the PaX developers only occurs in Linux 3.8.0 and later. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #225
2013-03-27 15:33:14 +00:00
#include <linux/version.h>
#if defined(CONSTIFY_PLUGIN) && LINUX_VERSION_CODE >= KERNEL_VERSION(3, 8, 0)
PaX/GrSecurity Linux 3.8.y compat: Use __no_const on struct ctl_table The PaX team started constifying `struct ctl_table` as of their Linux 3.8.0 patchset. This lead to zfsonlinux/spl#225 and Gentoo bug #463012. While investigating our options, I learned that there is a preprocessor directive called CONSTIFY_PLUGIN that we can use to detect the presence of the PaX changes and adjust the code accordingly. The PaX Team had suggested adopting ctl_table_no_const, but supporting older kernels required declaring that whenever the CONSTIFY_PLUGIN was set. Future compiler changes could potentially cause that to break in the presence of -Werror, so instead we define our own spl_ctl_table typdef and use that. This should be compatible with all PaX kernels. This introduces a Linux kernel version number check to prevent a build failure on versions of the PaX GCC plugin that existed for kernels before Linux 3.8.0. Affected versions of the PaX plugin will trigger a compiler error when they see no_const cast on a non-constified structure. Ordinarily, we would need an autotools check to catch that. However, it is safe to do a kernel version check instead of an autotools check in this specific instance because the affected versions of the PaX GCC plugin only exist for Linux kernels before 3.8.0 and the constification of `struct ctl_table` by the PaX developers only occurs in Linux 3.8.0 and later. Signed-off-by: Richard Yao <ryao@gentoo.org> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Closes #225
2013-03-27 15:33:14 +00:00
typedef struct ctl_table __no_const spl_ctl_table;
#else
typedef struct ctl_table spl_ctl_table;
#endif
static unsigned long table_min = 0;
static unsigned long table_max = ~0;
static struct ctl_table_header *spl_header = NULL;
static struct proc_dir_entry *proc_spl = NULL;
static struct proc_dir_entry *proc_spl_kmem = NULL;
static struct proc_dir_entry *proc_spl_kmem_slab = NULL;
static struct proc_dir_entry *proc_spl_taskq_all = NULL;
static struct proc_dir_entry *proc_spl_taskq = NULL;
struct proc_dir_entry *proc_spl_kstat = NULL;
static int
proc_copyin_string(char *kbuffer, int kbuffer_size, const char *ubuffer,
int ubuffer_size)
{
int size;
if (ubuffer_size > kbuffer_size)
return (-EOVERFLOW);
if (copy_from_user((void *)kbuffer, (void *)ubuffer, ubuffer_size))
return (-EFAULT);
/* strip trailing whitespace */
size = strnlen(kbuffer, ubuffer_size);
while (size-- >= 0)
if (!isspace(kbuffer[size]))
break;
/* empty string */
if (size < 0)
return (-EINVAL);
/* no space to terminate */
if (size == kbuffer_size)
return (-EOVERFLOW);
kbuffer[size + 1] = 0;
return (0);
}
static int
proc_copyout_string(char *ubuffer, int ubuffer_size, const char *kbuffer,
char *append)
{
/*
* NB if 'append' != NULL, it's a single character to append to the
* copied out string - usually "\n", for /proc entries and
* (i.e. a terminating zero byte) for sysctl entries
*/
int size = MIN(strlen(kbuffer), ubuffer_size);
if (copy_to_user(ubuffer, kbuffer, size))
return (-EFAULT);
if (append != NULL && size < ubuffer_size) {
if (copy_to_user(ubuffer + size, append, 1))
return (-EFAULT);
size++;
}
return (size);
}
#ifdef DEBUG_KMEM
static int
proc_domemused(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
int rc = 0;
unsigned long min = 0, max = ~0, val;
spl_ctl_table dummy = *table;
dummy.data = &val;
dummy.proc_handler = &proc_dointvec;
dummy.extra1 = &min;
dummy.extra2 = &max;
if (write) {
*ppos += *lenp;
} else {
#ifdef HAVE_ATOMIC64_T
val = atomic64_read((atomic64_t *)table->data);
#else
val = atomic_read((atomic_t *)table->data);
#endif /* HAVE_ATOMIC64_T */
rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos);
}
return (rc);
}
#endif /* DEBUG_KMEM */
static int
proc_doslab(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
int rc = 0;
unsigned long min = 0, max = ~0, val = 0, mask;
spl_ctl_table dummy = *table;
spl_kmem_cache_t *skc;
dummy.data = &val;
dummy.proc_handler = &proc_dointvec;
dummy.extra1 = &min;
dummy.extra2 = &max;
if (write) {
*ppos += *lenp;
} else {
down_read(&spl_kmem_cache_sem);
mask = (unsigned long)table->data;
list_for_each_entry(skc, &spl_kmem_cache_list, skc_list) {
/* Only use slabs of the correct kmem/vmem type */
if (!(skc->skc_flags & mask))
continue;
/* Sum the specified field for selected slabs */
switch (mask & (KMC_TOTAL | KMC_ALLOC | KMC_MAX)) {
case KMC_TOTAL:
val += skc->skc_slab_size * skc->skc_slab_total;
break;
case KMC_ALLOC:
val += skc->skc_obj_size * skc->skc_obj_alloc;
break;
case KMC_MAX:
val += skc->skc_obj_size * skc->skc_obj_max;
break;
}
}
up_read(&spl_kmem_cache_sem);
rc = proc_doulongvec_minmax(&dummy, write, buffer, lenp, ppos);
}
return (rc);
}
static int
proc_dohostid(struct ctl_table *table, int write,
void __user *buffer, size_t *lenp, loff_t *ppos)
{
int len, rc = 0;
char *end, str[32];
if (write) {
/*
* We can't use proc_doulongvec_minmax() in the write
* case here because hostid while a hex value has no
* leading 0x which confuses the helper function.
*/
rc = proc_copyin_string(str, sizeof (str), buffer, *lenp);
if (rc < 0)
return (rc);
spl_hostid = simple_strtoul(str, &end, 16);
if (str == end)
return (-EINVAL);
} else {
len = snprintf(str, sizeof (str), "%lx",
(unsigned long) zone_get_hostid(NULL));
if (*ppos >= len)
rc = 0;
else
rc = proc_copyout_string(buffer,
*lenp, str + *ppos, "\n");
if (rc >= 0) {
*lenp = rc;
*ppos += rc;
}
}
return (rc);
}
static void
taskq_seq_show_headers(struct seq_file *f)
{
seq_printf(f, "%-25s %5s %5s %5s %5s %5s %5s %12s %5s %10s\n",
"taskq", "act", "nthr", "spwn", "maxt", "pri",
"mina", "maxa", "cura", "flags");
}
/* indices into the lheads array below */
#define LHEAD_PEND 0
#define LHEAD_PRIO 1
#define LHEAD_DELAY 2
#define LHEAD_WAIT 3
#define LHEAD_ACTIVE 4
#define LHEAD_SIZE 5
/* BEGIN CSTYLED */
static unsigned int spl_max_show_tasks = 512;
module_param(spl_max_show_tasks, uint, 0644);
MODULE_PARM_DESC(spl_max_show_tasks, "Max number of tasks shown in taskq proc");
/* END CSTYLED */
static int
taskq_seq_show_impl(struct seq_file *f, void *p, boolean_t allflag)
{
taskq_t *tq = p;
taskq_thread_t *tqt;
spl_wait_queue_entry_t *wq;
struct task_struct *tsk;
taskq_ent_t *tqe;
char name[100];
struct list_head *lheads[LHEAD_SIZE], *lh;
static char *list_names[LHEAD_SIZE] =
{"pend", "prio", "delay", "wait", "active" };
int i, j, have_lheads = 0;
unsigned long wflags, flags;
spin_lock_irqsave_nested(&tq->tq_lock, flags, tq->tq_lock_class);
spin_lock_irqsave(&tq->tq_wait_waitq.lock, wflags);
/* get the various lists and check whether they're empty */
lheads[LHEAD_PEND] = &tq->tq_pend_list;
lheads[LHEAD_PRIO] = &tq->tq_prio_list;
lheads[LHEAD_DELAY] = &tq->tq_delay_list;
#ifdef HAVE_WAIT_QUEUE_HEAD_ENTRY
lheads[LHEAD_WAIT] = &tq->tq_wait_waitq.head;
#else
lheads[LHEAD_WAIT] = &tq->tq_wait_waitq.task_list;
#endif
lheads[LHEAD_ACTIVE] = &tq->tq_active_list;
for (i = 0; i < LHEAD_SIZE; ++i) {
if (list_empty(lheads[i]))
lheads[i] = NULL;
else
++have_lheads;
}
/* early return in non-"all" mode if lists are all empty */
if (!allflag && !have_lheads) {
spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (0);
}
/* unlock the waitq quickly */
if (!lheads[LHEAD_WAIT])
spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
/* show the base taskq contents */
snprintf(name, sizeof (name), "%s/%d", tq->tq_name, tq->tq_instance);
seq_printf(f, "%-25s ", name);
seq_printf(f, "%5d %5d %5d %5d %5d %5d %12d %5d %10x\n",
tq->tq_nactive, tq->tq_nthreads, tq->tq_nspawn,
tq->tq_maxthreads, tq->tq_pri, tq->tq_minalloc, tq->tq_maxalloc,
tq->tq_nalloc, tq->tq_flags);
/* show the active list */
if (lheads[LHEAD_ACTIVE]) {
j = 0;
list_for_each_entry(tqt, &tq->tq_active_list, tqt_active_list) {
if (j == 0)
seq_printf(f, "\t%s:",
list_names[LHEAD_ACTIVE]);
else if (j == 2) {
seq_printf(f, "\n\t ");
j = 0;
}
seq_printf(f, " [%d]%pf(%ps)",
tqt->tqt_thread->pid,
tqt->tqt_task->tqent_func,
tqt->tqt_task->tqent_arg);
++j;
}
seq_printf(f, "\n");
}
for (i = LHEAD_PEND; i <= LHEAD_WAIT; ++i)
if (lheads[i]) {
j = 0;
list_for_each(lh, lheads[i]) {
if (spl_max_show_tasks != 0 &&
j >= spl_max_show_tasks) {
seq_printf(f, "\n\t(truncated)");
break;
}
/* show the wait waitq list */
if (i == LHEAD_WAIT) {
#ifdef HAVE_WAIT_QUEUE_HEAD_ENTRY
wq = list_entry(lh,
spl_wait_queue_entry_t, entry);
#else
wq = list_entry(lh,
spl_wait_queue_entry_t, task_list);
#endif
if (j == 0)
seq_printf(f, "\t%s:",
list_names[i]);
else if (j % 8 == 0)
seq_printf(f, "\n\t ");
tsk = wq->private;
seq_printf(f, " %d", tsk->pid);
/* pend, prio and delay lists */
} else {
tqe = list_entry(lh, taskq_ent_t,
tqent_list);
if (j == 0)
seq_printf(f, "\t%s:",
list_names[i]);
else if (j % 2 == 0)
seq_printf(f, "\n\t ");
seq_printf(f, " %pf(%ps)",
tqe->tqent_func,
tqe->tqent_arg);
}
++j;
}
seq_printf(f, "\n");
}
if (lheads[LHEAD_WAIT])
spin_unlock_irqrestore(&tq->tq_wait_waitq.lock, wflags);
spin_unlock_irqrestore(&tq->tq_lock, flags);
return (0);
}
static int
taskq_all_seq_show(struct seq_file *f, void *p)
{
return (taskq_seq_show_impl(f, p, B_TRUE));
}
static int
taskq_seq_show(struct seq_file *f, void *p)
{
return (taskq_seq_show_impl(f, p, B_FALSE));
}
static void *
taskq_seq_start(struct seq_file *f, loff_t *pos)
{
struct list_head *p;
loff_t n = *pos;
down_read(&tq_list_sem);
if (!n)
taskq_seq_show_headers(f);
p = tq_list.next;
while (n--) {
p = p->next;
if (p == &tq_list)
return (NULL);
}
return (list_entry(p, taskq_t, tq_taskqs));
}
static void *
taskq_seq_next(struct seq_file *f, void *p, loff_t *pos)
{
taskq_t *tq = p;
++*pos;
return ((tq->tq_taskqs.next == &tq_list) ?
NULL : list_entry(tq->tq_taskqs.next, taskq_t, tq_taskqs));
}
static void
slab_seq_show_headers(struct seq_file *f)
{
seq_printf(f,
"--------------------- cache ----------"
"--------------------------------------------- "
"----- slab ------ "
"---- object ----- "
"--- emergency ---\n");
seq_printf(f,
"name "
" flags size alloc slabsize objsize "
"total alloc max "
"total alloc max "
"dlock alloc max\n");
}
static int
slab_seq_show(struct seq_file *f, void *p)
{
spl_kmem_cache_t *skc = p;
ASSERT(skc->skc_magic == SKC_MAGIC);
2019-10-18 17:24:28 +00:00
if (skc->skc_flags & KMC_SLAB) {
/*
* This cache is backed by a generic Linux kmem cache which
* has its own accounting. For these caches we only track
* the number of active allocated objects that exist within
* the underlying Linux slabs. For the overall statistics of
* the underlying Linux cache please refer to /proc/slabinfo.
*/
spin_lock(&skc->skc_lock);
seq_printf(f, "%-36s ", skc->skc_name);
seq_printf(f, "0x%05lx %9s %9lu %8s %8u "
"%5s %5s %5s %5s %5lu %5s %5s %5s %5s\n",
(long unsigned)skc->skc_flags,
"-",
(long unsigned)(skc->skc_obj_size * skc->skc_obj_alloc),
"-",
(unsigned)skc->skc_obj_size,
"-", "-", "-", "-",
(long unsigned)skc->skc_obj_alloc,
"-", "-", "-", "-");
spin_unlock(&skc->skc_lock);
Add KMC_SLAB cache type For small objects the Linux slab allocator has several advantages over its counterpart in the SPL. These include: 1) It is more memory-efficient and packs objects more tightly. 2) It is continually tuned to maximize performance. Therefore it makes sense to layer the SPLs slab allocator on top of the Linux slab allocator. This allows us to leverage the advantages above while preserving the Illumos semantics we depend on. However, there are some things we need to be careful of: 1) The Linux slab allocator was never designed to work well with large objects. Because the SPL slab must still handle this use case a cut off limit was added to transition from Linux slab backed objects to kmem or vmem backed slabs. spl_kmem_cache_slab_limit - Objects less than or equal to this size in bytes will be backed by the Linux slab. By default this value is zero which disables the Linux slab functionality. Reasonable values for this cut off limit are in the range of 4096-16386 bytes. spl_kmem_cache_kmem_limit - Objects less than or equal to this size in bytes will be backed by a kmem slab. Objects over this size will be vmem backed instead. This value defaults to 1/8 a page, or 512 bytes on an x86_64 architecture. 2) Be aware that using the Linux slab may inadvertently introduce new deadlocks. Care has been taken previously to ensure that all allocations which occur in the write path use GFP_NOIO. However, there may be internal allocations performed in the Linux slab which do not honor these flags. If this is the case a deadlock may occur. The path forward is definitely to start relying on the Linux slab. But for that to happen we need to start building confidence that there aren't any unexpected surprises lurking for us. And ideally need to move completely away from using the SPLs slab for large memory allocations. This patch is a first step. NOTES: 1) The KMC_NOMAGAZINE flag was leveraged to support the Linux slab backed caches but it is not supported for kmem/vmem backed caches. 2) Regardless of the spl_kmem_cache_*_limit settings a cache may be explicitly set to a given type by passed the KMC_KMEM, KMC_VMEM, or KMC_SLAB flags during cache creation. 3) The constructors, destructors, and reclaim callbacks are all functional and will be called regardless of the cache type. 4) KMC_SLAB caches will not appear in /proc/spl/kmem/slab due to the issues involved in presenting correct object accounting. Instead they will appear in /proc/slabinfo under the same names. 5) Several kmem SPLAT tests needed to be fixed because they relied incorrectly on internal kmem slab accounting. With the updated test cases all the SPLAT tests pass as expected. 6) An autoconf test was added to ensure that the __GFP_COMP flag was correctly added to the default flags used when allocating a slab. This is required to ensure all pages in higher order slabs are properly refcounted, see ae16ed9. 7) When using the SLUB allocator there is no need to attempt to set the __GFP_COMP flag. This has been the default behavior for the SLUB since Linux 2.6.25. 8) When using the SLUB it may be desirable to set the slub_nomerge kernel parameter to prevent caches from being merged. Original-patch-by: DHE <git@dehacked.net> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Prakash Surya <surya1@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: DHE <git@dehacked.net> Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Closes #356
2013-12-08 22:01:45 +00:00
return (0);
2019-10-18 17:24:28 +00:00
}
Add KMC_SLAB cache type For small objects the Linux slab allocator has several advantages over its counterpart in the SPL. These include: 1) It is more memory-efficient and packs objects more tightly. 2) It is continually tuned to maximize performance. Therefore it makes sense to layer the SPLs slab allocator on top of the Linux slab allocator. This allows us to leverage the advantages above while preserving the Illumos semantics we depend on. However, there are some things we need to be careful of: 1) The Linux slab allocator was never designed to work well with large objects. Because the SPL slab must still handle this use case a cut off limit was added to transition from Linux slab backed objects to kmem or vmem backed slabs. spl_kmem_cache_slab_limit - Objects less than or equal to this size in bytes will be backed by the Linux slab. By default this value is zero which disables the Linux slab functionality. Reasonable values for this cut off limit are in the range of 4096-16386 bytes. spl_kmem_cache_kmem_limit - Objects less than or equal to this size in bytes will be backed by a kmem slab. Objects over this size will be vmem backed instead. This value defaults to 1/8 a page, or 512 bytes on an x86_64 architecture. 2) Be aware that using the Linux slab may inadvertently introduce new deadlocks. Care has been taken previously to ensure that all allocations which occur in the write path use GFP_NOIO. However, there may be internal allocations performed in the Linux slab which do not honor these flags. If this is the case a deadlock may occur. The path forward is definitely to start relying on the Linux slab. But for that to happen we need to start building confidence that there aren't any unexpected surprises lurking for us. And ideally need to move completely away from using the SPLs slab for large memory allocations. This patch is a first step. NOTES: 1) The KMC_NOMAGAZINE flag was leveraged to support the Linux slab backed caches but it is not supported for kmem/vmem backed caches. 2) Regardless of the spl_kmem_cache_*_limit settings a cache may be explicitly set to a given type by passed the KMC_KMEM, KMC_VMEM, or KMC_SLAB flags during cache creation. 3) The constructors, destructors, and reclaim callbacks are all functional and will be called regardless of the cache type. 4) KMC_SLAB caches will not appear in /proc/spl/kmem/slab due to the issues involved in presenting correct object accounting. Instead they will appear in /proc/slabinfo under the same names. 5) Several kmem SPLAT tests needed to be fixed because they relied incorrectly on internal kmem slab accounting. With the updated test cases all the SPLAT tests pass as expected. 6) An autoconf test was added to ensure that the __GFP_COMP flag was correctly added to the default flags used when allocating a slab. This is required to ensure all pages in higher order slabs are properly refcounted, see ae16ed9. 7) When using the SLUB allocator there is no need to attempt to set the __GFP_COMP flag. This has been the default behavior for the SLUB since Linux 2.6.25. 8) When using the SLUB it may be desirable to set the slub_nomerge kernel parameter to prevent caches from being merged. Original-patch-by: DHE <git@dehacked.net> Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Prakash Surya <surya1@llnl.gov> Signed-off-by: Tim Chase <tim@chase2k.com> Signed-off-by: DHE <git@dehacked.net> Signed-off-by: Chunwei Chen <tuxoko@gmail.com> Closes #356
2013-12-08 22:01:45 +00:00
spin_lock(&skc->skc_lock);
seq_printf(f, "%-36s ", skc->skc_name);
seq_printf(f, "0x%05lx %9lu %9lu %8u %8u "
"%5lu %5lu %5lu %5lu %5lu %5lu %5lu %5lu %5lu\n",
(long unsigned)skc->skc_flags,
(long unsigned)(skc->skc_slab_size * skc->skc_slab_total),
(long unsigned)(skc->skc_obj_size * skc->skc_obj_alloc),
(unsigned)skc->skc_slab_size,
(unsigned)skc->skc_obj_size,
(long unsigned)skc->skc_slab_total,
(long unsigned)skc->skc_slab_alloc,
(long unsigned)skc->skc_slab_max,
(long unsigned)skc->skc_obj_total,
(long unsigned)skc->skc_obj_alloc,
(long unsigned)skc->skc_obj_max,
(long unsigned)skc->skc_obj_deadlock,
(long unsigned)skc->skc_obj_emergency,
(long unsigned)skc->skc_obj_emergency_max);
spin_unlock(&skc->skc_lock);
return (0);
}
static void *
slab_seq_start(struct seq_file *f, loff_t *pos)
{
struct list_head *p;
loff_t n = *pos;
down_read(&spl_kmem_cache_sem);
if (!n)
slab_seq_show_headers(f);
p = spl_kmem_cache_list.next;
while (n--) {
p = p->next;
if (p == &spl_kmem_cache_list)
return (NULL);
}
return (list_entry(p, spl_kmem_cache_t, skc_list));
}
static void *
slab_seq_next(struct seq_file *f, void *p, loff_t *pos)
{
spl_kmem_cache_t *skc = p;
++*pos;
return ((skc->skc_list.next == &spl_kmem_cache_list) ?
NULL : list_entry(skc->skc_list.next, spl_kmem_cache_t, skc_list));
}
static void
slab_seq_stop(struct seq_file *f, void *v)
{
up_read(&spl_kmem_cache_sem);
}
static struct seq_operations slab_seq_ops = {
.show = slab_seq_show,
.start = slab_seq_start,
.next = slab_seq_next,
.stop = slab_seq_stop,
};
static int
proc_slab_open(struct inode *inode, struct file *filp)
{
return (seq_open(filp, &slab_seq_ops));
}
static struct file_operations proc_slab_operations = {
.open = proc_slab_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
};
static void
taskq_seq_stop(struct seq_file *f, void *v)
{
up_read(&tq_list_sem);
}
static struct seq_operations taskq_all_seq_ops = {
.show = taskq_all_seq_show,
.start = taskq_seq_start,
.next = taskq_seq_next,
.stop = taskq_seq_stop,
};
static struct seq_operations taskq_seq_ops = {
.show = taskq_seq_show,
.start = taskq_seq_start,
.next = taskq_seq_next,
.stop = taskq_seq_stop,
};
static int
proc_taskq_all_open(struct inode *inode, struct file *filp)
{
return (seq_open(filp, &taskq_all_seq_ops));
}
static int
proc_taskq_open(struct inode *inode, struct file *filp)
{
return (seq_open(filp, &taskq_seq_ops));
}
static struct file_operations proc_taskq_all_operations = {
.open = proc_taskq_all_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
};
static struct file_operations proc_taskq_operations = {
.open = proc_taskq_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release,
};
static struct ctl_table spl_kmem_table[] = {
#ifdef DEBUG_KMEM
{
.procname = "kmem_used",
.data = &kmem_alloc_used,
#ifdef HAVE_ATOMIC64_T
.maxlen = sizeof (atomic64_t),
#else
.maxlen = sizeof (atomic_t),
#endif /* HAVE_ATOMIC64_T */
.mode = 0444,
.proc_handler = &proc_domemused,
},
{
.procname = "kmem_max",
.data = &kmem_alloc_max,
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doulongvec_minmax,
},
#endif /* DEBUG_KMEM */
{
.procname = "slab_kmem_total",
.data = (void *)(KMC_KMEM | KMC_TOTAL),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_kmem_alloc",
.data = (void *)(KMC_KMEM | KMC_ALLOC),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_kmem_max",
.data = (void *)(KMC_KMEM | KMC_MAX),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_vmem_total",
.data = (void *)(KMC_VMEM | KMC_TOTAL),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_vmem_alloc",
.data = (void *)(KMC_VMEM | KMC_ALLOC),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{
.procname = "slab_vmem_max",
.data = (void *)(KMC_VMEM | KMC_MAX),
.maxlen = sizeof (unsigned long),
.extra1 = &table_min,
.extra2 = &table_max,
.mode = 0444,
.proc_handler = &proc_doslab,
},
{},
};
static struct ctl_table spl_kstat_table[] = {
{},
};
static struct ctl_table spl_table[] = {
/*
* NB No .strategy entries have been provided since
* sysctl(8) prefers to go via /proc for portability.
*/
{
.procname = "gitrev",
.data = spl_gitrev,
.maxlen = sizeof (spl_gitrev),
.mode = 0444,
.proc_handler = &proc_dostring,
},
{
.procname = "hostid",
.data = &spl_hostid,
.maxlen = sizeof (unsigned long),
.mode = 0644,
.proc_handler = &proc_dohostid,
},
{
.procname = "kmem",
.mode = 0555,
.child = spl_kmem_table,
},
{
.procname = "kstat",
.mode = 0555,
.child = spl_kstat_table,
},
{},
};
static struct ctl_table spl_dir[] = {
{
.procname = "spl",
.mode = 0555,
.child = spl_table,
},
{}
};
static struct ctl_table spl_root[] = {
{
#ifdef HAVE_CTL_NAME
.ctl_name = CTL_KERN,
#endif
.procname = "kernel",
.mode = 0555,
.child = spl_dir,
},
{}
};
int
spl_proc_init(void)
{
int rc = 0;
spl_header = register_sysctl_table(spl_root);
if (spl_header == NULL)
return (-EUNATCH);
proc_spl = proc_mkdir("spl", NULL);
if (proc_spl == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_taskq_all = proc_create_data("taskq-all", 0444, proc_spl,
&proc_taskq_all_operations, NULL);
if (proc_spl_taskq_all == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_taskq = proc_create_data("taskq", 0444, proc_spl,
&proc_taskq_operations, NULL);
if (proc_spl_taskq == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_kmem = proc_mkdir("kmem", proc_spl);
if (proc_spl_kmem == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_kmem_slab = proc_create_data("slab", 0444, proc_spl_kmem,
&proc_slab_operations, NULL);
if (proc_spl_kmem_slab == NULL) {
rc = -EUNATCH;
goto out;
}
proc_spl_kstat = proc_mkdir("kstat", proc_spl);
if (proc_spl_kstat == NULL) {
rc = -EUNATCH;
goto out;
}
out:
if (rc) {
remove_proc_entry("kstat", proc_spl);
remove_proc_entry("slab", proc_spl_kmem);
remove_proc_entry("kmem", proc_spl);
remove_proc_entry("taskq-all", proc_spl);
remove_proc_entry("taskq", proc_spl);
remove_proc_entry("spl", NULL);
unregister_sysctl_table(spl_header);
}
return (rc);
}
void
spl_proc_fini(void)
{
remove_proc_entry("kstat", proc_spl);
remove_proc_entry("slab", proc_spl_kmem);
remove_proc_entry("kmem", proc_spl);
remove_proc_entry("taskq-all", proc_spl);
remove_proc_entry("taskq", proc_spl);
remove_proc_entry("spl", NULL);
ASSERT(spl_header != NULL);
unregister_sysctl_table(spl_header);
}