2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2007-09-20 10:31:38 +02:00
|
|
|
* fs/sysfs/file.c - sysfs regular (text) file implementation
|
|
|
|
*
|
|
|
|
* Copyright (c) 2001-3 Patrick Mochel
|
|
|
|
* Copyright (c) 2007 SUSE Linux Products GmbH
|
|
|
|
* Copyright (c) 2007 Tejun Heo <teheo@suse.de>
|
|
|
|
*
|
|
|
|
* This file is released under the GPLv2.
|
|
|
|
*
|
|
|
|
* Please see Documentation/filesystems/sysfs.txt for more information.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/kobject.h>
|
2008-03-05 00:09:07 +01:00
|
|
|
#include <linux/kallsyms.h>
|
2008-03-14 03:41:52 +01:00
|
|
|
#include <linux/slab.h>
|
2008-06-16 13:46:47 +02:00
|
|
|
#include <linux/fsnotify.h>
|
2005-06-23 09:09:12 +02:00
|
|
|
#include <linux/namei.h>
|
2006-03-20 07:53:53 +01:00
|
|
|
#include <linux/poll.h>
|
2006-12-20 10:52:44 +01:00
|
|
|
#include <linux/list.h>
|
2007-07-26 13:03:54 +02:00
|
|
|
#include <linux/mutex.h>
|
2007-08-25 01:11:54 +02:00
|
|
|
#include <linux/limits.h>
|
2005-04-17 00:20:36 +02:00
|
|
|
#include <asm/uaccess.h>
|
|
|
|
|
|
|
|
#include "sysfs.h"
|
|
|
|
|
2007-08-25 01:11:54 +02:00
|
|
|
/* used in crash dumps to help with debugging */
|
|
|
|
static char last_sysfs_file[PATH_MAX];
|
|
|
|
void sysfs_printk_last_file(void)
|
|
|
|
{
|
|
|
|
printk(KERN_EMERG "last sysfs file: %s\n", last_sysfs_file);
|
|
|
|
}
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
/*
|
|
|
|
* There's one sysfs_buffer for each open file and one
|
|
|
|
* sysfs_open_dirent for each sysfs_dirent with one or more open
|
|
|
|
* files.
|
|
|
|
*
|
|
|
|
* filp->private_data points to sysfs_buffer and
|
|
|
|
* sysfs_dirent->s_attr.open points to sysfs_open_dirent. s_attr.open
|
|
|
|
* is protected by sysfs_open_dirent_lock.
|
|
|
|
*/
|
2007-11-21 23:55:19 +01:00
|
|
|
static DEFINE_SPINLOCK(sysfs_open_dirent_lock);
|
2007-09-20 09:05:12 +02:00
|
|
|
|
|
|
|
struct sysfs_open_dirent {
|
|
|
|
atomic_t refcnt;
|
2007-09-20 09:05:12 +02:00
|
|
|
atomic_t event;
|
|
|
|
wait_queue_head_t poll;
|
2007-09-20 09:05:12 +02:00
|
|
|
struct list_head buffers; /* goes through sysfs_buffer.list */
|
|
|
|
};
|
|
|
|
|
2007-06-13 20:45:16 +02:00
|
|
|
struct sysfs_buffer {
|
|
|
|
size_t count;
|
|
|
|
loff_t pos;
|
|
|
|
char * page;
|
|
|
|
struct sysfs_ops * ops;
|
2007-07-26 13:03:54 +02:00
|
|
|
struct mutex mutex;
|
2007-06-13 20:45:16 +02:00
|
|
|
int needs_read_fill;
|
|
|
|
int event;
|
2007-09-20 09:05:12 +02:00
|
|
|
struct list_head list;
|
2007-06-13 20:45:16 +02:00
|
|
|
};
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/**
|
|
|
|
* fill_read_buffer - allocate and fill buffer from object.
|
|
|
|
* @dentry: dentry pointer.
|
|
|
|
* @buffer: data buffer for file.
|
|
|
|
*
|
|
|
|
* Allocate @buffer->page, if it hasn't been already, then call the
|
|
|
|
* kobject's show() method to fill the buffer with this attribute's
|
|
|
|
* data.
|
2007-01-02 08:48:08 +01:00
|
|
|
* This is called only once, on the file's first read unless an error
|
|
|
|
* is returned.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
static int fill_read_buffer(struct dentry * dentry, struct sysfs_buffer * buffer)
|
|
|
|
{
|
2007-06-13 20:45:16 +02:00
|
|
|
struct sysfs_dirent *attr_sd = dentry->d_fsdata;
|
2007-09-20 09:05:11 +02:00
|
|
|
struct kobject *kobj = attr_sd->s_parent->s_dir.kobj;
|
2005-04-17 00:20:36 +02:00
|
|
|
struct sysfs_ops * ops = buffer->ops;
|
|
|
|
int ret = 0;
|
|
|
|
ssize_t count;
|
|
|
|
|
|
|
|
if (!buffer->page)
|
|
|
|
buffer->page = (char *) get_zeroed_page(GFP_KERNEL);
|
|
|
|
if (!buffer->page)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
2007-06-13 20:45:16 +02:00
|
|
|
/* need attr_sd for attr and ops, its parent for kobj */
|
|
|
|
if (!sysfs_get_active_two(attr_sd))
|
|
|
|
return -ENODEV;
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
buffer->event = atomic_read(&attr_sd->s_attr.open->event);
|
2007-09-20 09:05:11 +02:00
|
|
|
count = ops->show(kobj, attr_sd->s_attr.attr, buffer->page);
|
2007-06-13 20:45:16 +02:00
|
|
|
|
|
|
|
sysfs_put_active_two(attr_sd);
|
|
|
|
|
2007-11-21 23:55:19 +01:00
|
|
|
/*
|
|
|
|
* The code works fine with PAGE_SIZE return but it's likely to
|
|
|
|
* indicate truncated result or overflow in normal use cases.
|
|
|
|
*/
|
2008-03-05 00:09:07 +01:00
|
|
|
if (count >= (ssize_t)PAGE_SIZE) {
|
|
|
|
print_symbol("fill_read_buffer: %s returned bad count\n",
|
|
|
|
(unsigned long)ops->show);
|
|
|
|
/* Try to struggle along */
|
|
|
|
count = PAGE_SIZE - 1;
|
|
|
|
}
|
2007-01-02 08:48:08 +01:00
|
|
|
if (count >= 0) {
|
|
|
|
buffer->needs_read_fill = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
buffer->count = count;
|
2007-01-02 08:48:08 +01:00
|
|
|
} else {
|
2005-04-17 00:20:36 +02:00
|
|
|
ret = count;
|
2007-01-02 08:48:08 +01:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* sysfs_read_file - read an attribute.
|
|
|
|
* @file: file pointer.
|
|
|
|
* @buf: buffer to fill.
|
|
|
|
* @count: number of bytes to read.
|
|
|
|
* @ppos: starting offset in file.
|
|
|
|
*
|
|
|
|
* Userspace wants to read an attribute file. The attribute descriptor
|
|
|
|
* is in the file's ->d_fsdata. The target object is in the directory's
|
|
|
|
* ->d_fsdata.
|
|
|
|
*
|
|
|
|
* We call fill_read_buffer() to allocate and fill the buffer from the
|
|
|
|
* object's show() method exactly once (if the read is happening from
|
|
|
|
* the beginning of the file). That should fill the entire buffer with
|
|
|
|
* all the data the object has to offer for that attribute.
|
|
|
|
* We then call flush_read_buffer() to copy the buffer to userspace
|
|
|
|
* in the increments specified.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
sysfs_read_file(struct file *file, char __user *buf, size_t count, loff_t *ppos)
|
|
|
|
{
|
|
|
|
struct sysfs_buffer * buffer = file->private_data;
|
|
|
|
ssize_t retval = 0;
|
|
|
|
|
2007-07-26 13:03:54 +02:00
|
|
|
mutex_lock(&buffer->mutex);
|
2008-04-08 00:35:01 +02:00
|
|
|
if (buffer->needs_read_fill || *ppos == 0) {
|
2007-06-13 20:45:16 +02:00
|
|
|
retval = fill_read_buffer(file->f_path.dentry,buffer);
|
2007-03-15 20:51:28 +01:00
|
|
|
if (retval)
|
2005-04-17 00:20:36 +02:00
|
|
|
goto out;
|
|
|
|
}
|
2006-10-03 10:16:06 +02:00
|
|
|
pr_debug("%s: count = %zd, ppos = %lld, buf = %s\n",
|
2008-04-30 09:55:09 +02:00
|
|
|
__func__, count, *ppos, buffer->page);
|
2007-05-09 11:33:32 +02:00
|
|
|
retval = simple_read_from_buffer(buf, count, ppos, buffer->page,
|
|
|
|
buffer->count);
|
2005-04-17 00:20:36 +02:00
|
|
|
out:
|
2007-07-26 13:03:54 +02:00
|
|
|
mutex_unlock(&buffer->mutex);
|
2005-04-17 00:20:36 +02:00
|
|
|
return retval;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fill_write_buffer - copy buffer from userspace.
|
|
|
|
* @buffer: data buffer for file.
|
2005-05-01 17:59:26 +02:00
|
|
|
* @buf: data from user.
|
2005-04-17 00:20:36 +02:00
|
|
|
* @count: number of bytes in @userbuf.
|
|
|
|
*
|
|
|
|
* Allocate @buffer->page if it hasn't been already, then
|
|
|
|
* copy the user-supplied buffer into it.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int
|
|
|
|
fill_write_buffer(struct sysfs_buffer * buffer, const char __user * buf, size_t count)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
|
|
|
|
if (!buffer->page)
|
|
|
|
buffer->page = (char *)get_zeroed_page(GFP_KERNEL);
|
|
|
|
if (!buffer->page)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
if (count >= PAGE_SIZE)
|
2006-04-01 01:37:06 +02:00
|
|
|
count = PAGE_SIZE - 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
error = copy_from_user(buffer->page,buf,count);
|
|
|
|
buffer->needs_read_fill = 1;
|
2006-10-22 19:17:47 +02:00
|
|
|
/* if buf is assumed to contain a string, terminate it by \0,
|
|
|
|
so e.g. sscanf() can scan the string easily */
|
|
|
|
buffer->page[count] = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
return error ? -EFAULT : count;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* flush_write_buffer - push buffer to kobject.
|
2005-06-24 07:05:21 +02:00
|
|
|
* @dentry: dentry to the attribute
|
2005-04-17 00:20:36 +02:00
|
|
|
* @buffer: data buffer for file.
|
2005-06-24 07:05:21 +02:00
|
|
|
* @count: number of bytes
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
|
|
|
* Get the correct pointers for the kobject and the attribute we're
|
|
|
|
* dealing with, then call the store() method for the attribute,
|
|
|
|
* passing the buffer that we acquired in fill_write_buffer().
|
|
|
|
*/
|
|
|
|
|
2007-06-13 20:45:16 +02:00
|
|
|
static int
|
2005-04-17 00:20:36 +02:00
|
|
|
flush_write_buffer(struct dentry * dentry, struct sysfs_buffer * buffer, size_t count)
|
|
|
|
{
|
2007-06-13 20:45:15 +02:00
|
|
|
struct sysfs_dirent *attr_sd = dentry->d_fsdata;
|
2007-09-20 09:05:11 +02:00
|
|
|
struct kobject *kobj = attr_sd->s_parent->s_dir.kobj;
|
2005-04-17 00:20:36 +02:00
|
|
|
struct sysfs_ops * ops = buffer->ops;
|
2007-06-13 20:45:16 +02:00
|
|
|
int rc;
|
|
|
|
|
|
|
|
/* need attr_sd for attr and ops, its parent for kobj */
|
|
|
|
if (!sysfs_get_active_two(attr_sd))
|
|
|
|
return -ENODEV;
|
|
|
|
|
2007-09-20 09:05:11 +02:00
|
|
|
rc = ops->store(kobj, attr_sd->s_attr.attr, buffer->page, count);
|
2007-06-13 20:45:16 +02:00
|
|
|
|
|
|
|
sysfs_put_active_two(attr_sd);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-06-13 20:45:16 +02:00
|
|
|
return rc;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/**
|
|
|
|
* sysfs_write_file - write an attribute.
|
|
|
|
* @file: file pointer
|
|
|
|
* @buf: data to write
|
|
|
|
* @count: number of bytes
|
|
|
|
* @ppos: starting offset
|
|
|
|
*
|
|
|
|
* Similar to sysfs_read_file(), though working in the opposite direction.
|
|
|
|
* We allocate and fill the data from the user in fill_write_buffer(),
|
|
|
|
* then push it to the kobject in flush_write_buffer().
|
|
|
|
* There is no easy way for us to know if userspace is only doing a partial
|
|
|
|
* write, so we don't support them. We expect the entire buffer to come
|
|
|
|
* on the first write.
|
|
|
|
* Hint: if you're writing a value, first read the file, modify only the
|
|
|
|
* the value you're changing, then write entire buffer back.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
sysfs_write_file(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
|
|
|
|
{
|
|
|
|
struct sysfs_buffer * buffer = file->private_data;
|
|
|
|
ssize_t len;
|
|
|
|
|
2007-07-26 13:03:54 +02:00
|
|
|
mutex_lock(&buffer->mutex);
|
2005-04-17 00:20:36 +02:00
|
|
|
len = fill_write_buffer(buffer, buf, count);
|
|
|
|
if (len > 0)
|
2006-12-08 11:36:36 +01:00
|
|
|
len = flush_write_buffer(file->f_path.dentry, buffer, len);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (len > 0)
|
|
|
|
*ppos += len;
|
2007-07-26 13:03:54 +02:00
|
|
|
mutex_unlock(&buffer->mutex);
|
2005-04-17 00:20:36 +02:00
|
|
|
return len;
|
|
|
|
}
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
/**
|
|
|
|
* sysfs_get_open_dirent - get or create sysfs_open_dirent
|
|
|
|
* @sd: target sysfs_dirent
|
|
|
|
* @buffer: sysfs_buffer for this instance of open
|
|
|
|
*
|
|
|
|
* If @sd->s_attr.open exists, increment its reference count;
|
|
|
|
* otherwise, create one. @buffer is chained to the buffers
|
|
|
|
* list.
|
|
|
|
*
|
|
|
|
* LOCKING:
|
|
|
|
* Kernel thread context (may sleep).
|
|
|
|
*
|
|
|
|
* RETURNS:
|
|
|
|
* 0 on success, -errno on failure.
|
|
|
|
*/
|
|
|
|
static int sysfs_get_open_dirent(struct sysfs_dirent *sd,
|
|
|
|
struct sysfs_buffer *buffer)
|
|
|
|
{
|
|
|
|
struct sysfs_open_dirent *od, *new_od = NULL;
|
|
|
|
|
|
|
|
retry:
|
2009-09-16 01:05:51 +02:00
|
|
|
spin_lock_irq(&sysfs_open_dirent_lock);
|
2007-09-20 09:05:12 +02:00
|
|
|
|
|
|
|
if (!sd->s_attr.open && new_od) {
|
|
|
|
sd->s_attr.open = new_od;
|
|
|
|
new_od = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
od = sd->s_attr.open;
|
|
|
|
if (od) {
|
|
|
|
atomic_inc(&od->refcnt);
|
|
|
|
list_add_tail(&buffer->list, &od->buffers);
|
|
|
|
}
|
|
|
|
|
2009-09-16 01:05:51 +02:00
|
|
|
spin_unlock_irq(&sysfs_open_dirent_lock);
|
2007-09-20 09:05:12 +02:00
|
|
|
|
|
|
|
if (od) {
|
|
|
|
kfree(new_od);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* not there, initialize a new one and retry */
|
|
|
|
new_od = kmalloc(sizeof(*new_od), GFP_KERNEL);
|
|
|
|
if (!new_od)
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
atomic_set(&new_od->refcnt, 0);
|
2007-09-20 09:05:12 +02:00
|
|
|
atomic_set(&new_od->event, 1);
|
|
|
|
init_waitqueue_head(&new_od->poll);
|
2007-09-20 09:05:12 +02:00
|
|
|
INIT_LIST_HEAD(&new_od->buffers);
|
|
|
|
goto retry;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* sysfs_put_open_dirent - put sysfs_open_dirent
|
|
|
|
* @sd: target sysfs_dirent
|
|
|
|
* @buffer: associated sysfs_buffer
|
|
|
|
*
|
|
|
|
* Put @sd->s_attr.open and unlink @buffer from the buffers list.
|
|
|
|
* If reference count reaches zero, disassociate and free it.
|
|
|
|
*
|
|
|
|
* LOCKING:
|
|
|
|
* None.
|
|
|
|
*/
|
|
|
|
static void sysfs_put_open_dirent(struct sysfs_dirent *sd,
|
|
|
|
struct sysfs_buffer *buffer)
|
|
|
|
{
|
|
|
|
struct sysfs_open_dirent *od = sd->s_attr.open;
|
2009-09-16 01:05:51 +02:00
|
|
|
unsigned long flags;
|
2007-09-20 09:05:12 +02:00
|
|
|
|
2009-09-16 01:05:51 +02:00
|
|
|
spin_lock_irqsave(&sysfs_open_dirent_lock, flags);
|
2007-09-20 09:05:12 +02:00
|
|
|
|
|
|
|
list_del(&buffer->list);
|
|
|
|
if (atomic_dec_and_test(&od->refcnt))
|
|
|
|
sd->s_attr.open = NULL;
|
|
|
|
else
|
|
|
|
od = NULL;
|
|
|
|
|
2009-09-16 01:05:51 +02:00
|
|
|
spin_unlock_irqrestore(&sysfs_open_dirent_lock, flags);
|
2007-09-20 09:05:12 +02:00
|
|
|
|
|
|
|
kfree(od);
|
|
|
|
}
|
|
|
|
|
2006-12-20 10:52:44 +01:00
|
|
|
static int sysfs_open_file(struct inode *inode, struct file *file)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-06-13 20:45:15 +02:00
|
|
|
struct sysfs_dirent *attr_sd = file->f_path.dentry->d_fsdata;
|
2007-09-20 09:05:11 +02:00
|
|
|
struct kobject *kobj = attr_sd->s_parent->s_dir.kobj;
|
2007-11-02 13:47:53 +01:00
|
|
|
struct sysfs_buffer *buffer;
|
|
|
|
struct sysfs_ops *ops;
|
|
|
|
int error = -EACCES;
|
2007-08-25 01:11:54 +02:00
|
|
|
char *p;
|
|
|
|
|
|
|
|
p = d_path(&file->f_path, last_sysfs_file, sizeof(last_sysfs_file));
|
|
|
|
if (p)
|
|
|
|
memmove(last_sysfs_file, p, strlen(p) + 1);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-06-13 20:45:16 +02:00
|
|
|
/* need attr_sd for attr and ops, its parent for kobj */
|
|
|
|
if (!sysfs_get_active_two(attr_sd))
|
|
|
|
return -ENODEV;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-11-02 13:47:53 +01:00
|
|
|
/* every kobject with an attribute needs a ktype assigned */
|
|
|
|
if (kobj->ktype && kobj->ktype->sysfs_ops)
|
2005-04-17 00:20:36 +02:00
|
|
|
ops = kobj->ktype->sysfs_ops;
|
2007-11-02 13:47:53 +01:00
|
|
|
else {
|
2008-07-26 04:45:41 +02:00
|
|
|
WARN(1, KERN_ERR "missing sysfs attribute operations for "
|
2007-11-02 13:47:53 +01:00
|
|
|
"kobject: %s\n", kobject_name(kobj));
|
2007-06-13 20:45:17 +02:00
|
|
|
goto err_out;
|
2007-11-02 13:47:53 +01:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* File needs write support.
|
|
|
|
* The inode's perms must say it's ok,
|
|
|
|
* and we must have a store method.
|
|
|
|
*/
|
|
|
|
if (file->f_mode & FMODE_WRITE) {
|
|
|
|
if (!(inode->i_mode & S_IWUGO) || !ops->store)
|
2007-06-13 20:45:17 +02:00
|
|
|
goto err_out;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* File needs read support.
|
|
|
|
* The inode's perms must say it's ok, and we there
|
|
|
|
* must be a show method for it.
|
|
|
|
*/
|
|
|
|
if (file->f_mode & FMODE_READ) {
|
|
|
|
if (!(inode->i_mode & S_IRUGO) || !ops->show)
|
2007-06-13 20:45:17 +02:00
|
|
|
goto err_out;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/* No error? Great, allocate a buffer for the file, and store it
|
|
|
|
* it in file->private_data for easy access.
|
|
|
|
*/
|
2007-06-13 20:45:16 +02:00
|
|
|
error = -ENOMEM;
|
2006-02-22 11:18:15 +01:00
|
|
|
buffer = kzalloc(sizeof(struct sysfs_buffer), GFP_KERNEL);
|
2007-06-13 20:45:16 +02:00
|
|
|
if (!buffer)
|
2007-06-13 20:45:17 +02:00
|
|
|
goto err_out;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-07-26 13:03:54 +02:00
|
|
|
mutex_init(&buffer->mutex);
|
2007-06-13 20:45:16 +02:00
|
|
|
buffer->needs_read_fill = 1;
|
|
|
|
buffer->ops = ops;
|
|
|
|
file->private_data = buffer;
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
/* make sure we have open dirent struct */
|
|
|
|
error = sysfs_get_open_dirent(attr_sd, buffer);
|
|
|
|
if (error)
|
|
|
|
goto err_free;
|
|
|
|
|
2007-09-20 09:05:10 +02:00
|
|
|
/* open succeeded, put active references */
|
2007-06-13 20:45:16 +02:00
|
|
|
sysfs_put_active_two(attr_sd);
|
|
|
|
return 0;
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
err_free:
|
|
|
|
kfree(buffer);
|
2007-06-13 20:45:17 +02:00
|
|
|
err_out:
|
2007-06-13 20:45:16 +02:00
|
|
|
sysfs_put_active_two(attr_sd);
|
2005-04-17 00:20:36 +02:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
static int sysfs_release(struct inode *inode, struct file *filp)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2007-09-20 09:05:12 +02:00
|
|
|
struct sysfs_dirent *sd = filp->f_path.dentry->d_fsdata;
|
2007-06-13 20:45:16 +02:00
|
|
|
struct sysfs_buffer *buffer = filp->private_data;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
sysfs_put_open_dirent(sd, buffer);
|
|
|
|
|
2007-09-20 09:05:10 +02:00
|
|
|
if (buffer->page)
|
|
|
|
free_page((unsigned long)buffer->page);
|
|
|
|
kfree(buffer);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2006-03-20 07:53:53 +01:00
|
|
|
/* Sysfs attribute files are pollable. The idea is that you read
|
|
|
|
* the content and then you use 'poll' or 'select' to wait for
|
|
|
|
* the content to change. When the content changes (assuming the
|
|
|
|
* manager for the kobject supports notification), poll will
|
|
|
|
* return POLLERR|POLLPRI, and select will return the fd whether
|
|
|
|
* it is waiting for read, write, or exceptions.
|
|
|
|
* Once poll/select indicates that the value has changed, you
|
2008-04-08 00:35:01 +02:00
|
|
|
* need to close and re-open the file, or seek to 0 and read again.
|
2006-03-20 07:53:53 +01:00
|
|
|
* Reminder: this only works for attributes which actively support
|
|
|
|
* it, and it is not possible to test an attribute from userspace
|
2007-08-10 22:51:07 +02:00
|
|
|
* to see if it supports poll (Neither 'poll' nor 'select' return
|
2006-03-20 07:53:53 +01:00
|
|
|
* an appropriate error code). When in doubt, set a suitable timeout value.
|
|
|
|
*/
|
|
|
|
static unsigned int sysfs_poll(struct file *filp, poll_table *wait)
|
|
|
|
{
|
|
|
|
struct sysfs_buffer * buffer = filp->private_data;
|
2007-06-13 20:45:16 +02:00
|
|
|
struct sysfs_dirent *attr_sd = filp->f_path.dentry->d_fsdata;
|
2007-09-20 09:05:12 +02:00
|
|
|
struct sysfs_open_dirent *od = attr_sd->s_attr.open;
|
2007-06-13 20:45:16 +02:00
|
|
|
|
|
|
|
/* need parent for the kobj, grab both */
|
|
|
|
if (!sysfs_get_active_two(attr_sd))
|
|
|
|
goto trigger;
|
2006-03-20 07:53:53 +01:00
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
poll_wait(filp, &od->poll, wait);
|
2006-03-20 07:53:53 +01:00
|
|
|
|
2007-06-13 20:45:16 +02:00
|
|
|
sysfs_put_active_two(attr_sd);
|
|
|
|
|
2007-09-20 09:05:12 +02:00
|
|
|
if (buffer->event != atomic_read(&od->event))
|
2007-06-13 20:45:16 +02:00
|
|
|
goto trigger;
|
2006-03-20 07:53:53 +01:00
|
|
|
|
sysfs: sysfs poll keep the poll rule of regular file.
Currently, following test programs don't finished.
% ruby -e '
Thread.new { sleep }
File.read("/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies")
'
strace expose the reason.
...
open("/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies", O_RDONLY|O_LARGEFILE) = 3
ioctl(3, SNDCTL_TMR_TIMEBASE or TCGETS, 0xbf9fa6b8) = -1 ENOTTY (Inappropriate ioctl for device)
fstat64(3, {st_mode=S_IFREG|0444, st_size=4096, ...}) = 0
_llseek(3, 0, [0], SEEK_CUR) = 0
select(4, [3], NULL, NULL, NULL) = 1 (in [3])
read(3, "1400000 1300000 1200000 1100000 1"..., 4096) = 62
select(4, [3], NULL, NULL, NULL
Because Ruby (the scripting language) VM assume select system-call
against regular file don't block. it because SUSv3 says "Regular files
shall always poll TRUE for reading and writing". see
http://www.opengroup.org/onlinepubs/009695399/functions/poll.html it
seems valid assumption.
But sysfs_poll() don't keep this rule although sysfs file can read and
write always.
This patch restore proper poll behavior to sysfs.
/sys/block/md*/md/sync_action polling application and another sysfs
updating sensitive application still can use POLLERR and POLLPRI.
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-04-09 06:53:22 +02:00
|
|
|
return DEFAULT_POLLMASK;
|
2007-06-13 20:45:16 +02:00
|
|
|
|
|
|
|
trigger:
|
|
|
|
buffer->needs_read_fill = 1;
|
sysfs: sysfs poll keep the poll rule of regular file.
Currently, following test programs don't finished.
% ruby -e '
Thread.new { sleep }
File.read("/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies")
'
strace expose the reason.
...
open("/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies", O_RDONLY|O_LARGEFILE) = 3
ioctl(3, SNDCTL_TMR_TIMEBASE or TCGETS, 0xbf9fa6b8) = -1 ENOTTY (Inappropriate ioctl for device)
fstat64(3, {st_mode=S_IFREG|0444, st_size=4096, ...}) = 0
_llseek(3, 0, [0], SEEK_CUR) = 0
select(4, [3], NULL, NULL, NULL) = 1 (in [3])
read(3, "1400000 1300000 1200000 1100000 1"..., 4096) = 62
select(4, [3], NULL, NULL, NULL
Because Ruby (the scripting language) VM assume select system-call
against regular file don't block. it because SUSv3 says "Regular files
shall always poll TRUE for reading and writing". see
http://www.opengroup.org/onlinepubs/009695399/functions/poll.html it
seems valid assumption.
But sysfs_poll() don't keep this rule although sysfs file can read and
write always.
This patch restore proper poll behavior to sysfs.
/sys/block/md*/md/sync_action polling application and another sysfs
updating sensitive application still can use POLLERR and POLLPRI.
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-04-09 06:53:22 +02:00
|
|
|
return DEFAULT_POLLMASK|POLLERR|POLLPRI;
|
2006-03-20 07:53:53 +01:00
|
|
|
}
|
|
|
|
|
2008-07-16 00:58:04 +02:00
|
|
|
void sysfs_notify_dirent(struct sysfs_dirent *sd)
|
|
|
|
{
|
|
|
|
struct sysfs_open_dirent *od;
|
2009-09-16 01:05:51 +02:00
|
|
|
unsigned long flags;
|
2008-07-16 00:58:04 +02:00
|
|
|
|
2009-09-16 01:05:51 +02:00
|
|
|
spin_lock_irqsave(&sysfs_open_dirent_lock, flags);
|
2008-07-16 00:58:04 +02:00
|
|
|
|
|
|
|
od = sd->s_attr.open;
|
|
|
|
if (od) {
|
|
|
|
atomic_inc(&od->event);
|
|
|
|
wake_up_interruptible(&od->poll);
|
|
|
|
}
|
|
|
|
|
2009-09-16 01:05:51 +02:00
|
|
|
spin_unlock_irqrestore(&sysfs_open_dirent_lock, flags);
|
2008-07-16 00:58:04 +02:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_notify_dirent);
|
|
|
|
|
2008-09-26 01:45:13 +02:00
|
|
|
void sysfs_notify(struct kobject *k, const char *dir, const char *attr)
|
2006-03-20 07:53:53 +01:00
|
|
|
{
|
2007-06-13 21:27:25 +02:00
|
|
|
struct sysfs_dirent *sd = k->sd;
|
2006-03-20 07:53:53 +01:00
|
|
|
|
2007-06-13 21:27:25 +02:00
|
|
|
mutex_lock(&sysfs_mutex);
|
|
|
|
|
|
|
|
if (sd && dir)
|
|
|
|
sd = sysfs_find_dirent(sd, dir);
|
|
|
|
if (sd && attr)
|
|
|
|
sd = sysfs_find_dirent(sd, attr);
|
2008-07-16 00:58:04 +02:00
|
|
|
if (sd)
|
|
|
|
sysfs_notify_dirent(sd);
|
2007-06-13 21:27:25 +02:00
|
|
|
|
|
|
|
mutex_unlock(&sysfs_mutex);
|
2006-03-20 07:53:53 +01:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_notify);
|
|
|
|
|
2006-03-28 11:56:42 +02:00
|
|
|
const struct file_operations sysfs_file_operations = {
|
2005-04-17 00:20:36 +02:00
|
|
|
.read = sysfs_read_file,
|
|
|
|
.write = sysfs_write_file,
|
|
|
|
.llseek = generic_file_llseek,
|
|
|
|
.open = sysfs_open_file,
|
|
|
|
.release = sysfs_release,
|
2006-03-20 07:53:53 +01:00
|
|
|
.poll = sysfs_poll,
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
2008-03-21 02:47:52 +01:00
|
|
|
int sysfs_add_file_mode(struct sysfs_dirent *dir_sd,
|
|
|
|
const struct attribute *attr, int type, mode_t amode)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2008-03-21 02:47:52 +01:00
|
|
|
umode_t mode = (amode & S_IALLUGO) | S_IFREG;
|
2007-06-13 21:27:24 +02:00
|
|
|
struct sysfs_addrm_cxt acxt;
|
2007-06-13 20:45:14 +02:00
|
|
|
struct sysfs_dirent *sd;
|
2007-08-02 14:38:03 +02:00
|
|
|
int rc;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-06-13 21:27:23 +02:00
|
|
|
sd = sysfs_new_dirent(attr->name, mode, type);
|
|
|
|
if (!sd)
|
|
|
|
return -ENOMEM;
|
2007-09-20 09:05:11 +02:00
|
|
|
sd->s_attr.attr = (void *)attr;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-06-13 21:27:24 +02:00
|
|
|
sysfs_addrm_start(&acxt, dir_sd);
|
2007-08-02 14:38:03 +02:00
|
|
|
rc = sysfs_add_one(&acxt, sd);
|
|
|
|
sysfs_addrm_finish(&acxt);
|
2007-06-13 20:45:14 +02:00
|
|
|
|
2007-08-02 14:38:03 +02:00
|
|
|
if (rc)
|
2007-07-18 09:38:11 +02:00
|
|
|
sysfs_put(sd);
|
2007-06-13 21:27:23 +02:00
|
|
|
|
2007-08-02 14:38:03 +02:00
|
|
|
return rc;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2008-03-21 02:47:52 +01:00
|
|
|
int sysfs_add_file(struct sysfs_dirent *dir_sd, const struct attribute *attr,
|
|
|
|
int type)
|
|
|
|
{
|
|
|
|
return sysfs_add_file_mode(dir_sd, attr, type, attr->mode);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/**
|
|
|
|
* sysfs_create_file - create an attribute file for an object.
|
|
|
|
* @kobj: object we're creating for.
|
2007-10-20 03:14:32 +02:00
|
|
|
* @attr: attribute descriptor.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
|
|
|
|
int sysfs_create_file(struct kobject * kobj, const struct attribute * attr)
|
|
|
|
{
|
2007-06-13 21:27:22 +02:00
|
|
|
BUG_ON(!kobj || !kobj->sd || !attr);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2007-06-13 21:27:22 +02:00
|
|
|
return sysfs_add_file(kobj->sd, attr, SYSFS_KOBJ_ATTR);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-02-20 21:02:44 +01:00
|
|
|
/**
|
|
|
|
* sysfs_add_file_to_group - add an attribute file to a pre-existing group.
|
|
|
|
* @kobj: object we're acting for.
|
|
|
|
* @attr: attribute descriptor.
|
|
|
|
* @group: group name.
|
|
|
|
*/
|
|
|
|
int sysfs_add_file_to_group(struct kobject *kobj,
|
|
|
|
const struct attribute *attr, const char *group)
|
|
|
|
{
|
2007-06-13 21:27:22 +02:00
|
|
|
struct sysfs_dirent *dir_sd;
|
2007-02-20 21:02:44 +01:00
|
|
|
int error;
|
|
|
|
|
2008-01-03 01:44:05 +01:00
|
|
|
if (group)
|
|
|
|
dir_sd = sysfs_get_dirent(kobj->sd, group);
|
|
|
|
else
|
|
|
|
dir_sd = sysfs_get(kobj->sd);
|
|
|
|
|
2007-06-13 21:27:22 +02:00
|
|
|
if (!dir_sd)
|
|
|
|
return -ENOENT;
|
|
|
|
|
|
|
|
error = sysfs_add_file(dir_sd, attr, SYSFS_KOBJ_ATTR);
|
|
|
|
sysfs_put(dir_sd);
|
|
|
|
|
2007-02-20 21:02:44 +01:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_add_file_to_group);
|
|
|
|
|
2005-04-19 06:57:32 +02:00
|
|
|
/**
|
|
|
|
* sysfs_chmod_file - update the modified mode value on an object attribute.
|
|
|
|
* @kobj: object we're acting for.
|
|
|
|
* @attr: attribute descriptor.
|
|
|
|
* @mode: file permissions.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
int sysfs_chmod_file(struct kobject *kobj, struct attribute *attr, mode_t mode)
|
|
|
|
{
|
2007-06-13 21:27:25 +02:00
|
|
|
struct sysfs_dirent *victim_sd = NULL;
|
|
|
|
struct dentry *victim = NULL;
|
2005-07-29 21:13:35 +02:00
|
|
|
struct inode * inode;
|
|
|
|
struct iattr newattrs;
|
2007-06-13 21:27:25 +02:00
|
|
|
int rc;
|
|
|
|
|
|
|
|
rc = -ENOENT;
|
|
|
|
victim_sd = sysfs_get_dirent(kobj->sd, attr->name);
|
|
|
|
if (!victim_sd)
|
|
|
|
goto out;
|
|
|
|
|
2007-08-20 14:36:30 +02:00
|
|
|
mutex_lock(&sysfs_rename_mutex);
|
2007-06-13 21:27:25 +02:00
|
|
|
victim = sysfs_get_dentry(victim_sd);
|
2007-08-20 14:36:30 +02:00
|
|
|
mutex_unlock(&sysfs_rename_mutex);
|
2007-06-13 21:27:25 +02:00
|
|
|
if (IS_ERR(victim)) {
|
|
|
|
rc = PTR_ERR(victim);
|
|
|
|
victim = NULL;
|
|
|
|
goto out;
|
2005-04-19 06:57:32 +02:00
|
|
|
}
|
|
|
|
|
2007-06-13 21:27:25 +02:00
|
|
|
inode = victim->d_inode;
|
2007-09-20 09:05:10 +02:00
|
|
|
|
2007-06-13 21:27:25 +02:00
|
|
|
mutex_lock(&inode->i_mutex);
|
2007-09-20 09:05:10 +02:00
|
|
|
|
2007-06-13 21:27:25 +02:00
|
|
|
newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO);
|
2009-11-08 08:27:02 +01:00
|
|
|
newattrs.ia_valid = ATTR_MODE;
|
2008-06-16 13:46:47 +02:00
|
|
|
rc = sysfs_setattr(victim, &newattrs);
|
2007-09-20 09:05:10 +02:00
|
|
|
|
2007-06-13 21:27:25 +02:00
|
|
|
mutex_unlock(&inode->i_mutex);
|
|
|
|
out:
|
|
|
|
dput(victim);
|
|
|
|
sysfs_put(victim_sd);
|
|
|
|
return rc;
|
2005-04-19 06:57:32 +02:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_chmod_file);
|
|
|
|
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/**
|
|
|
|
* sysfs_remove_file - remove an object attribute.
|
|
|
|
* @kobj: object we're acting for.
|
|
|
|
* @attr: attribute descriptor.
|
|
|
|
*
|
|
|
|
* Hash the attribute name and kill the victim.
|
|
|
|
*/
|
|
|
|
|
|
|
|
void sysfs_remove_file(struct kobject * kobj, const struct attribute * attr)
|
|
|
|
{
|
2007-06-13 21:27:22 +02:00
|
|
|
sysfs_hash_and_remove(kobj->sd, attr->name);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
2007-02-20 21:02:44 +01:00
|
|
|
/**
|
|
|
|
* sysfs_remove_file_from_group - remove an attribute file from a group.
|
|
|
|
* @kobj: object we're acting for.
|
|
|
|
* @attr: attribute descriptor.
|
|
|
|
* @group: group name.
|
|
|
|
*/
|
|
|
|
void sysfs_remove_file_from_group(struct kobject *kobj,
|
|
|
|
const struct attribute *attr, const char *group)
|
|
|
|
{
|
2007-06-13 21:27:22 +02:00
|
|
|
struct sysfs_dirent *dir_sd;
|
2007-02-20 21:02:44 +01:00
|
|
|
|
2008-01-03 01:44:05 +01:00
|
|
|
if (group)
|
|
|
|
dir_sd = sysfs_get_dirent(kobj->sd, group);
|
|
|
|
else
|
|
|
|
dir_sd = sysfs_get(kobj->sd);
|
2007-06-13 21:27:22 +02:00
|
|
|
if (dir_sd) {
|
|
|
|
sysfs_hash_and_remove(dir_sd, attr->name);
|
|
|
|
sysfs_put(dir_sd);
|
2007-02-20 21:02:44 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_remove_file_from_group);
|
|
|
|
|
2007-03-15 20:50:34 +01:00
|
|
|
struct sysfs_schedule_callback_struct {
|
2009-03-13 19:07:36 +01:00
|
|
|
struct list_head workq_list;
|
|
|
|
struct kobject *kobj;
|
2007-03-15 20:50:34 +01:00
|
|
|
void (*func)(void *);
|
|
|
|
void *data;
|
2007-04-26 09:12:04 +02:00
|
|
|
struct module *owner;
|
2007-03-15 20:50:34 +01:00
|
|
|
struct work_struct work;
|
|
|
|
};
|
|
|
|
|
sysfs: don't use global workqueue in sysfs_schedule_callback()
A sysfs attribute using sysfs_schedule_callback() to commit suicide
may end up calling device_unregister(), which will eventually call
a driver's ->remove function.
Drivers may call flush_scheduled_work() in their shutdown routines,
in which case lockdep will complain with something like the following:
=============================================
[ INFO: possible recursive locking detected ]
2.6.29-rc8-kk #1
---------------------------------------------
events/4/56 is trying to acquire lock:
(events){--..}, at: [<ffffffff80257fc0>] flush_workqueue+0x0/0xa0
but task is already holding lock:
(events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
other info that might help us debug this:
3 locks held by events/4/56:
#0: (events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#1: (&ss->work){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#2: (pci_remove_rescan_mutex){--..}, at: [<ffffffff803c10d1>] remove_callback+0x21/0x40
stack backtrace:
Pid: 56, comm: events/4 Not tainted 2.6.29-rc8-kk #1
Call Trace:
[<ffffffff8026dfcd>] validate_chain+0xb7d/0x1260
[<ffffffff8026eade>] __lock_acquire+0x42e/0xa40
[<ffffffff8026f148>] lock_acquire+0x58/0x80
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff8025800d>] flush_workqueue+0x4d/0xa0
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff80258070>] flush_scheduled_work+0x10/0x20
[<ffffffffa0144065>] e1000_remove+0x55/0xfe [e1000e]
[<ffffffff8033ee30>] ? sysfs_schedule_callback_work+0x0/0x50
[<ffffffff803bfeb2>] pci_device_remove+0x32/0x70
[<ffffffff80441da9>] __device_release_driver+0x59/0x90
[<ffffffff80441edb>] device_release_driver+0x2b/0x40
[<ffffffff804419d6>] bus_remove_device+0xa6/0x120
[<ffffffff8043e46b>] device_del+0x12b/0x190
[<ffffffff8043e4f6>] device_unregister+0x26/0x70
[<ffffffff803ba969>] pci_stop_dev+0x49/0x60
[<ffffffff803baab0>] pci_remove_bus_device+0x40/0xc0
[<ffffffff803c10d9>] remove_callback+0x29/0x40
[<ffffffff8033ee4f>] sysfs_schedule_callback_work+0x1f/0x50
[<ffffffff8025769a>] run_workqueue+0x15a/0x230
[<ffffffff80257648>] ? run_workqueue+0x108/0x230
[<ffffffff8025846f>] worker_thread+0x9f/0x100
[<ffffffff8025bce0>] ? autoremove_wake_function+0x0/0x40
[<ffffffff802583d0>] ? worker_thread+0x0/0x100
[<ffffffff8025b89d>] kthread+0x4d/0x80
[<ffffffff8020d4ba>] child_rip+0xa/0x20
[<ffffffff8020cebc>] ? restore_args+0x0/0x30
[<ffffffff8025b850>] ? kthread+0x0/0x80
[<ffffffff8020d4b0>] ? child_rip+0x0/0x20
Although we know that the device_unregister path will never acquire
a lock that a driver might try to acquire in its ->remove, in general
we should never attempt to flush a workqueue from within the same
workqueue, and lockdep rightly complains.
So as long as sysfs attributes cannot commit suicide directly and we
are stuck with this callback mechanism, put the sysfs callbacks on
their own workqueue instead of the global one.
This has the side benefit that if a suicidal sysfs attribute kicks
off a long chain of ->remove callbacks, we no longer induce a long
delay on the global queue.
This also fixes a missing module_put in the error path introduced
by sysfs-only-allow-one-scheduled-removal-callback-per-kobj.patch.
We never destroy the workqueue, but I'm not sure that's a
problem.
Reported-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Tested-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-03-25 22:11:36 +01:00
|
|
|
static struct workqueue_struct *sysfs_workqueue;
|
2009-03-13 19:07:36 +01:00
|
|
|
static DEFINE_MUTEX(sysfs_workq_mutex);
|
|
|
|
static LIST_HEAD(sysfs_workq);
|
2007-03-15 20:50:34 +01:00
|
|
|
static void sysfs_schedule_callback_work(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct sysfs_schedule_callback_struct *ss = container_of(work,
|
|
|
|
struct sysfs_schedule_callback_struct, work);
|
|
|
|
|
|
|
|
(ss->func)(ss->data);
|
|
|
|
kobject_put(ss->kobj);
|
2007-04-26 09:12:04 +02:00
|
|
|
module_put(ss->owner);
|
2009-03-13 19:07:36 +01:00
|
|
|
mutex_lock(&sysfs_workq_mutex);
|
|
|
|
list_del(&ss->workq_list);
|
|
|
|
mutex_unlock(&sysfs_workq_mutex);
|
2007-03-15 20:50:34 +01:00
|
|
|
kfree(ss);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* sysfs_schedule_callback - helper to schedule a callback for a kobject
|
|
|
|
* @kobj: object we're acting for.
|
|
|
|
* @func: callback function to invoke later.
|
|
|
|
* @data: argument to pass to @func.
|
2007-04-26 09:12:04 +02:00
|
|
|
* @owner: module owning the callback code
|
2007-03-15 20:50:34 +01:00
|
|
|
*
|
|
|
|
* sysfs attribute methods must not unregister themselves or their parent
|
|
|
|
* kobject (which would amount to the same thing). Attempts to do so will
|
|
|
|
* deadlock, since unregistration is mutually exclusive with driver
|
|
|
|
* callbacks.
|
|
|
|
*
|
|
|
|
* Instead methods can call this routine, which will attempt to allocate
|
|
|
|
* and schedule a workqueue request to call back @func with @data as its
|
|
|
|
* argument in the workqueue's process context. @kobj will be pinned
|
|
|
|
* until @func returns.
|
|
|
|
*
|
|
|
|
* Returns 0 if the request was submitted, -ENOMEM if storage could not
|
2009-03-13 19:07:36 +01:00
|
|
|
* be allocated, -ENODEV if a reference to @owner isn't available,
|
|
|
|
* -EAGAIN if a callback has already been scheduled for @kobj.
|
2007-03-15 20:50:34 +01:00
|
|
|
*/
|
|
|
|
int sysfs_schedule_callback(struct kobject *kobj, void (*func)(void *),
|
2007-04-26 09:12:04 +02:00
|
|
|
void *data, struct module *owner)
|
2007-03-15 20:50:34 +01:00
|
|
|
{
|
2009-03-13 19:07:36 +01:00
|
|
|
struct sysfs_schedule_callback_struct *ss, *tmp;
|
2007-03-15 20:50:34 +01:00
|
|
|
|
2007-04-26 09:12:04 +02:00
|
|
|
if (!try_module_get(owner))
|
|
|
|
return -ENODEV;
|
2009-03-13 19:07:36 +01:00
|
|
|
|
|
|
|
mutex_lock(&sysfs_workq_mutex);
|
|
|
|
list_for_each_entry_safe(ss, tmp, &sysfs_workq, workq_list)
|
|
|
|
if (ss->kobj == kobj) {
|
sysfs: don't use global workqueue in sysfs_schedule_callback()
A sysfs attribute using sysfs_schedule_callback() to commit suicide
may end up calling device_unregister(), which will eventually call
a driver's ->remove function.
Drivers may call flush_scheduled_work() in their shutdown routines,
in which case lockdep will complain with something like the following:
=============================================
[ INFO: possible recursive locking detected ]
2.6.29-rc8-kk #1
---------------------------------------------
events/4/56 is trying to acquire lock:
(events){--..}, at: [<ffffffff80257fc0>] flush_workqueue+0x0/0xa0
but task is already holding lock:
(events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
other info that might help us debug this:
3 locks held by events/4/56:
#0: (events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#1: (&ss->work){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#2: (pci_remove_rescan_mutex){--..}, at: [<ffffffff803c10d1>] remove_callback+0x21/0x40
stack backtrace:
Pid: 56, comm: events/4 Not tainted 2.6.29-rc8-kk #1
Call Trace:
[<ffffffff8026dfcd>] validate_chain+0xb7d/0x1260
[<ffffffff8026eade>] __lock_acquire+0x42e/0xa40
[<ffffffff8026f148>] lock_acquire+0x58/0x80
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff8025800d>] flush_workqueue+0x4d/0xa0
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff80258070>] flush_scheduled_work+0x10/0x20
[<ffffffffa0144065>] e1000_remove+0x55/0xfe [e1000e]
[<ffffffff8033ee30>] ? sysfs_schedule_callback_work+0x0/0x50
[<ffffffff803bfeb2>] pci_device_remove+0x32/0x70
[<ffffffff80441da9>] __device_release_driver+0x59/0x90
[<ffffffff80441edb>] device_release_driver+0x2b/0x40
[<ffffffff804419d6>] bus_remove_device+0xa6/0x120
[<ffffffff8043e46b>] device_del+0x12b/0x190
[<ffffffff8043e4f6>] device_unregister+0x26/0x70
[<ffffffff803ba969>] pci_stop_dev+0x49/0x60
[<ffffffff803baab0>] pci_remove_bus_device+0x40/0xc0
[<ffffffff803c10d9>] remove_callback+0x29/0x40
[<ffffffff8033ee4f>] sysfs_schedule_callback_work+0x1f/0x50
[<ffffffff8025769a>] run_workqueue+0x15a/0x230
[<ffffffff80257648>] ? run_workqueue+0x108/0x230
[<ffffffff8025846f>] worker_thread+0x9f/0x100
[<ffffffff8025bce0>] ? autoremove_wake_function+0x0/0x40
[<ffffffff802583d0>] ? worker_thread+0x0/0x100
[<ffffffff8025b89d>] kthread+0x4d/0x80
[<ffffffff8020d4ba>] child_rip+0xa/0x20
[<ffffffff8020cebc>] ? restore_args+0x0/0x30
[<ffffffff8025b850>] ? kthread+0x0/0x80
[<ffffffff8020d4b0>] ? child_rip+0x0/0x20
Although we know that the device_unregister path will never acquire
a lock that a driver might try to acquire in its ->remove, in general
we should never attempt to flush a workqueue from within the same
workqueue, and lockdep rightly complains.
So as long as sysfs attributes cannot commit suicide directly and we
are stuck with this callback mechanism, put the sysfs callbacks on
their own workqueue instead of the global one.
This has the side benefit that if a suicidal sysfs attribute kicks
off a long chain of ->remove callbacks, we no longer induce a long
delay on the global queue.
This also fixes a missing module_put in the error path introduced
by sysfs-only-allow-one-scheduled-removal-callback-per-kobj.patch.
We never destroy the workqueue, but I'm not sure that's a
problem.
Reported-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Tested-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-03-25 22:11:36 +01:00
|
|
|
module_put(owner);
|
2009-03-13 19:07:36 +01:00
|
|
|
mutex_unlock(&sysfs_workq_mutex);
|
|
|
|
return -EAGAIN;
|
|
|
|
}
|
|
|
|
mutex_unlock(&sysfs_workq_mutex);
|
|
|
|
|
sysfs: don't use global workqueue in sysfs_schedule_callback()
A sysfs attribute using sysfs_schedule_callback() to commit suicide
may end up calling device_unregister(), which will eventually call
a driver's ->remove function.
Drivers may call flush_scheduled_work() in their shutdown routines,
in which case lockdep will complain with something like the following:
=============================================
[ INFO: possible recursive locking detected ]
2.6.29-rc8-kk #1
---------------------------------------------
events/4/56 is trying to acquire lock:
(events){--..}, at: [<ffffffff80257fc0>] flush_workqueue+0x0/0xa0
but task is already holding lock:
(events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
other info that might help us debug this:
3 locks held by events/4/56:
#0: (events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#1: (&ss->work){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#2: (pci_remove_rescan_mutex){--..}, at: [<ffffffff803c10d1>] remove_callback+0x21/0x40
stack backtrace:
Pid: 56, comm: events/4 Not tainted 2.6.29-rc8-kk #1
Call Trace:
[<ffffffff8026dfcd>] validate_chain+0xb7d/0x1260
[<ffffffff8026eade>] __lock_acquire+0x42e/0xa40
[<ffffffff8026f148>] lock_acquire+0x58/0x80
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff8025800d>] flush_workqueue+0x4d/0xa0
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff80258070>] flush_scheduled_work+0x10/0x20
[<ffffffffa0144065>] e1000_remove+0x55/0xfe [e1000e]
[<ffffffff8033ee30>] ? sysfs_schedule_callback_work+0x0/0x50
[<ffffffff803bfeb2>] pci_device_remove+0x32/0x70
[<ffffffff80441da9>] __device_release_driver+0x59/0x90
[<ffffffff80441edb>] device_release_driver+0x2b/0x40
[<ffffffff804419d6>] bus_remove_device+0xa6/0x120
[<ffffffff8043e46b>] device_del+0x12b/0x190
[<ffffffff8043e4f6>] device_unregister+0x26/0x70
[<ffffffff803ba969>] pci_stop_dev+0x49/0x60
[<ffffffff803baab0>] pci_remove_bus_device+0x40/0xc0
[<ffffffff803c10d9>] remove_callback+0x29/0x40
[<ffffffff8033ee4f>] sysfs_schedule_callback_work+0x1f/0x50
[<ffffffff8025769a>] run_workqueue+0x15a/0x230
[<ffffffff80257648>] ? run_workqueue+0x108/0x230
[<ffffffff8025846f>] worker_thread+0x9f/0x100
[<ffffffff8025bce0>] ? autoremove_wake_function+0x0/0x40
[<ffffffff802583d0>] ? worker_thread+0x0/0x100
[<ffffffff8025b89d>] kthread+0x4d/0x80
[<ffffffff8020d4ba>] child_rip+0xa/0x20
[<ffffffff8020cebc>] ? restore_args+0x0/0x30
[<ffffffff8025b850>] ? kthread+0x0/0x80
[<ffffffff8020d4b0>] ? child_rip+0x0/0x20
Although we know that the device_unregister path will never acquire
a lock that a driver might try to acquire in its ->remove, in general
we should never attempt to flush a workqueue from within the same
workqueue, and lockdep rightly complains.
So as long as sysfs attributes cannot commit suicide directly and we
are stuck with this callback mechanism, put the sysfs callbacks on
their own workqueue instead of the global one.
This has the side benefit that if a suicidal sysfs attribute kicks
off a long chain of ->remove callbacks, we no longer induce a long
delay on the global queue.
This also fixes a missing module_put in the error path introduced
by sysfs-only-allow-one-scheduled-removal-callback-per-kobj.patch.
We never destroy the workqueue, but I'm not sure that's a
problem.
Reported-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Tested-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-03-25 22:11:36 +01:00
|
|
|
if (sysfs_workqueue == NULL) {
|
2009-05-07 21:36:53 +02:00
|
|
|
sysfs_workqueue = create_singlethread_workqueue("sysfsd");
|
sysfs: don't use global workqueue in sysfs_schedule_callback()
A sysfs attribute using sysfs_schedule_callback() to commit suicide
may end up calling device_unregister(), which will eventually call
a driver's ->remove function.
Drivers may call flush_scheduled_work() in their shutdown routines,
in which case lockdep will complain with something like the following:
=============================================
[ INFO: possible recursive locking detected ]
2.6.29-rc8-kk #1
---------------------------------------------
events/4/56 is trying to acquire lock:
(events){--..}, at: [<ffffffff80257fc0>] flush_workqueue+0x0/0xa0
but task is already holding lock:
(events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
other info that might help us debug this:
3 locks held by events/4/56:
#0: (events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#1: (&ss->work){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#2: (pci_remove_rescan_mutex){--..}, at: [<ffffffff803c10d1>] remove_callback+0x21/0x40
stack backtrace:
Pid: 56, comm: events/4 Not tainted 2.6.29-rc8-kk #1
Call Trace:
[<ffffffff8026dfcd>] validate_chain+0xb7d/0x1260
[<ffffffff8026eade>] __lock_acquire+0x42e/0xa40
[<ffffffff8026f148>] lock_acquire+0x58/0x80
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff8025800d>] flush_workqueue+0x4d/0xa0
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff80258070>] flush_scheduled_work+0x10/0x20
[<ffffffffa0144065>] e1000_remove+0x55/0xfe [e1000e]
[<ffffffff8033ee30>] ? sysfs_schedule_callback_work+0x0/0x50
[<ffffffff803bfeb2>] pci_device_remove+0x32/0x70
[<ffffffff80441da9>] __device_release_driver+0x59/0x90
[<ffffffff80441edb>] device_release_driver+0x2b/0x40
[<ffffffff804419d6>] bus_remove_device+0xa6/0x120
[<ffffffff8043e46b>] device_del+0x12b/0x190
[<ffffffff8043e4f6>] device_unregister+0x26/0x70
[<ffffffff803ba969>] pci_stop_dev+0x49/0x60
[<ffffffff803baab0>] pci_remove_bus_device+0x40/0xc0
[<ffffffff803c10d9>] remove_callback+0x29/0x40
[<ffffffff8033ee4f>] sysfs_schedule_callback_work+0x1f/0x50
[<ffffffff8025769a>] run_workqueue+0x15a/0x230
[<ffffffff80257648>] ? run_workqueue+0x108/0x230
[<ffffffff8025846f>] worker_thread+0x9f/0x100
[<ffffffff8025bce0>] ? autoremove_wake_function+0x0/0x40
[<ffffffff802583d0>] ? worker_thread+0x0/0x100
[<ffffffff8025b89d>] kthread+0x4d/0x80
[<ffffffff8020d4ba>] child_rip+0xa/0x20
[<ffffffff8020cebc>] ? restore_args+0x0/0x30
[<ffffffff8025b850>] ? kthread+0x0/0x80
[<ffffffff8020d4b0>] ? child_rip+0x0/0x20
Although we know that the device_unregister path will never acquire
a lock that a driver might try to acquire in its ->remove, in general
we should never attempt to flush a workqueue from within the same
workqueue, and lockdep rightly complains.
So as long as sysfs attributes cannot commit suicide directly and we
are stuck with this callback mechanism, put the sysfs callbacks on
their own workqueue instead of the global one.
This has the side benefit that if a suicidal sysfs attribute kicks
off a long chain of ->remove callbacks, we no longer induce a long
delay on the global queue.
This also fixes a missing module_put in the error path introduced
by sysfs-only-allow-one-scheduled-removal-callback-per-kobj.patch.
We never destroy the workqueue, but I'm not sure that's a
problem.
Reported-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Tested-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-03-25 22:11:36 +01:00
|
|
|
if (sysfs_workqueue == NULL) {
|
|
|
|
module_put(owner);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-03-15 20:50:34 +01:00
|
|
|
ss = kmalloc(sizeof(*ss), GFP_KERNEL);
|
2007-04-26 09:12:04 +02:00
|
|
|
if (!ss) {
|
|
|
|
module_put(owner);
|
2007-03-15 20:50:34 +01:00
|
|
|
return -ENOMEM;
|
2007-04-26 09:12:04 +02:00
|
|
|
}
|
2007-03-15 20:50:34 +01:00
|
|
|
kobject_get(kobj);
|
|
|
|
ss->kobj = kobj;
|
|
|
|
ss->func = func;
|
|
|
|
ss->data = data;
|
2007-04-26 09:12:04 +02:00
|
|
|
ss->owner = owner;
|
2007-03-15 20:50:34 +01:00
|
|
|
INIT_WORK(&ss->work, sysfs_schedule_callback_work);
|
2009-03-13 19:07:36 +01:00
|
|
|
INIT_LIST_HEAD(&ss->workq_list);
|
|
|
|
mutex_lock(&sysfs_workq_mutex);
|
|
|
|
list_add_tail(&ss->workq_list, &sysfs_workq);
|
|
|
|
mutex_unlock(&sysfs_workq_mutex);
|
sysfs: don't use global workqueue in sysfs_schedule_callback()
A sysfs attribute using sysfs_schedule_callback() to commit suicide
may end up calling device_unregister(), which will eventually call
a driver's ->remove function.
Drivers may call flush_scheduled_work() in their shutdown routines,
in which case lockdep will complain with something like the following:
=============================================
[ INFO: possible recursive locking detected ]
2.6.29-rc8-kk #1
---------------------------------------------
events/4/56 is trying to acquire lock:
(events){--..}, at: [<ffffffff80257fc0>] flush_workqueue+0x0/0xa0
but task is already holding lock:
(events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
other info that might help us debug this:
3 locks held by events/4/56:
#0: (events){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#1: (&ss->work){--..}, at: [<ffffffff80257648>] run_workqueue+0x108/0x230
#2: (pci_remove_rescan_mutex){--..}, at: [<ffffffff803c10d1>] remove_callback+0x21/0x40
stack backtrace:
Pid: 56, comm: events/4 Not tainted 2.6.29-rc8-kk #1
Call Trace:
[<ffffffff8026dfcd>] validate_chain+0xb7d/0x1260
[<ffffffff8026eade>] __lock_acquire+0x42e/0xa40
[<ffffffff8026f148>] lock_acquire+0x58/0x80
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff8025800d>] flush_workqueue+0x4d/0xa0
[<ffffffff80257fc0>] ? flush_workqueue+0x0/0xa0
[<ffffffff80258070>] flush_scheduled_work+0x10/0x20
[<ffffffffa0144065>] e1000_remove+0x55/0xfe [e1000e]
[<ffffffff8033ee30>] ? sysfs_schedule_callback_work+0x0/0x50
[<ffffffff803bfeb2>] pci_device_remove+0x32/0x70
[<ffffffff80441da9>] __device_release_driver+0x59/0x90
[<ffffffff80441edb>] device_release_driver+0x2b/0x40
[<ffffffff804419d6>] bus_remove_device+0xa6/0x120
[<ffffffff8043e46b>] device_del+0x12b/0x190
[<ffffffff8043e4f6>] device_unregister+0x26/0x70
[<ffffffff803ba969>] pci_stop_dev+0x49/0x60
[<ffffffff803baab0>] pci_remove_bus_device+0x40/0xc0
[<ffffffff803c10d9>] remove_callback+0x29/0x40
[<ffffffff8033ee4f>] sysfs_schedule_callback_work+0x1f/0x50
[<ffffffff8025769a>] run_workqueue+0x15a/0x230
[<ffffffff80257648>] ? run_workqueue+0x108/0x230
[<ffffffff8025846f>] worker_thread+0x9f/0x100
[<ffffffff8025bce0>] ? autoremove_wake_function+0x0/0x40
[<ffffffff802583d0>] ? worker_thread+0x0/0x100
[<ffffffff8025b89d>] kthread+0x4d/0x80
[<ffffffff8020d4ba>] child_rip+0xa/0x20
[<ffffffff8020cebc>] ? restore_args+0x0/0x30
[<ffffffff8025b850>] ? kthread+0x0/0x80
[<ffffffff8020d4b0>] ? child_rip+0x0/0x20
Although we know that the device_unregister path will never acquire
a lock that a driver might try to acquire in its ->remove, in general
we should never attempt to flush a workqueue from within the same
workqueue, and lockdep rightly complains.
So as long as sysfs attributes cannot commit suicide directly and we
are stuck with this callback mechanism, put the sysfs callbacks on
their own workqueue instead of the global one.
This has the side benefit that if a suicidal sysfs attribute kicks
off a long chain of ->remove callbacks, we no longer induce a long
delay on the global queue.
This also fixes a missing module_put in the error path introduced
by sysfs-only-allow-one-scheduled-removal-callback-per-kobj.patch.
We never destroy the workqueue, but I'm not sure that's a
problem.
Reported-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Tested-by: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
Signed-off-by: Alex Chiang <achiang@hp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2009-03-25 22:11:36 +01:00
|
|
|
queue_work(sysfs_workqueue, &ss->work);
|
2007-03-15 20:50:34 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_schedule_callback);
|
|
|
|
|
2007-02-20 21:02:44 +01:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
EXPORT_SYMBOL_GPL(sysfs_create_file);
|
|
|
|
EXPORT_SYMBOL_GPL(sysfs_remove_file);
|