]> Erik (J.A.K.) Mouw Delft University of Technology Faculty of Information Technology and Systems

[email protected] PO BOX 5031 2600 GA Delft The Netherlands

1.0  May 30, 2001 Initial revision posted to linux-kernel 1.1  June 3, 2001 Revised after comments from linux-kernel 2001 Erik Mouw This documentation 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. This documentation 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 this program; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA For more details see the file COPYING in the source distribution of Linux. This guide describes the use of the procfs file system from within the Linux kernel. The idea to write this guide came up on the #kernelnewbies IRC channel (see http://www.kernelnewbies.org/), when Jeff Garzik explained the use of procfs and forwarded me a message Alexander Viro wrote to the linux-kernel mailing list. I agreed to write it up nicely, so here it is. I'd like to thank Jeff Garzik [email protected] and Alexander Viro [email protected] for their input, Tim Waugh [email protected] for his Selfdocbook, and Marc Joosen [email protected] for proofreading. This documentation was written while working on the LART computing board (http://www.lart.tudelft.nl/), which is sponsored by the Mobile Multi-media Communications (http://www.mmc.tudelft.nl/) and Ubiquitous Communications (http://www.ubicom.tudelft.nl/) projects. Erik The /proc file system (procfs) is a special file system in the linux kernel. It's a virtual file system: it is not associated with a block device but exists only in memory. The files in the procfs are there to allow userland programs access to certain information from the kernel (like process information in /proc/[0-9]+/), but also for debug purposes (like /proc/ksyms). This guide describes the use of the procfs file system from within the Linux kernel. It starts by introducing all relevant functions to manage the files within the file system. After that it shows how to communicate with userland, and some tips and tricks will be pointed out. Finally a complete example will be shown. Note that the files in /proc/sys are sysctl files: they don't belong to procfs and are governed by a completely different API described in the Kernel API book. This chapter describes the functions that various kernel components use to populate the procfs with files, symlinks, device nodes, and directories. A minor note before we start: if you want to use any of the procfs functions, be sure to include the correct header file! This should be one of the first lines in your code: #include <linux/proc_fs.h> struct proc_dir_entry* create_proc_entry const char* name mode_t mode struct proc_dir_entry* parent This function creates a regular file with the name name, file mode mode in the directory parent. To create a file in the root of the procfs, use NULL as parent parameter. When successful, the function will return a pointer to the freshly created struct proc_dir_entry; otherwise it will return NULL. describes how to do something useful with regular files. Note that it is specifically supported that you can pass a path that spans multiple directories. For example create_proc_entry("drivers/via0/info") will create the via0 directory if necessary, with standard 0755 permissions. If you only want to be able to read the file, the function create_proc_read_entry described in may be used to create and initialise the procfs entry in one single call. struct proc_dir_entry* proc_symlink const char* name struct proc_dir_entry* parent const char* dest This creates a symlink in the procfs directory parent that points from name to dest. This translates in userland to ln -s dest name. struct proc_dir_entry* proc_mknod const char* name mode_t mode struct proc_dir_entry* parent kdev_t rdev Creates a device file name with mode mode in the procfs directory parent. The device file will work on the device rdev, which can be generated by using the MKDEV macro from linux/kdev_t.h. The mode parameter must contain S_IFBLK or S_IFCHR to create a device node. Compare with userland mknod --mode=mode name rdev. struct proc_dir_entry* proc_mkdir const char* name struct proc_dir_entry* parent Create a directory name in the procfs directory parent. void remove_proc_entry const char* name struct proc_dir_entry* parent Removes the entry name in the directory parent from the procfs. Entries are removed by their name, not by the struct proc_dir_entry returned by the various create functions. Note that this function doesn't recursively remove entries. Be sure to free the data entry from the struct proc_dir_entry before remove_proc_entry is called (that is: if there was some data allocated, of course). See for more information on using the data entry. Instead of reading (or writing) information directly from kernel memory, procfs works with call back functions for files: functions that are called when a specific file is being read or written. Such functions have to be initialised after the procfs file is created by setting the read_proc and/or write_proc fields in the struct proc_dir_entry* that the function create_proc_entry returned: struct proc_dir_entry* entry; entry->read_proc = read_proc_foo; entry->write_proc = write_proc_foo; If you only want to use a the read_proc, the function create_proc_read_entry described in may be used to create and initialise the procfs entry in one single call. The read function is a call back function that allows userland processes to read data from the kernel. The read function should have the following format: int read_func char* page char** start off_t off int count int* eof void* data The read function should write its information into the page. For proper use, the function should start writing at an offset of off in page and write at most count bytes, but because most read functions are quite simple and only return a small amount of information, these two parameters are usually ignored (it breaks pagers like more and less, but cat still works). If the off and count parameters are properly used, eof should be used to signal that the end of the file has been reached by writing 1 to the memory location eof points to. The parameter start doesn't seem to be used anywhere in the kernel. The data parameter can be used to create a single call back function for several files, see . The read_func function must return the number of bytes written into the page. shows how to use a read call back function. The write call back function allows a userland process to write data to the kernel, so it has some kind of control over the kernel. The write function should have the following format: int write_func struct file* file const char* buffer unsigned long count void* data The write function should read count bytes at maximum from the buffer. Note that the buffer doesn't live in the kernel's memory space, so it should first be copied to kernel space with copy_from_user. The file parameter is usually ignored. shows how to use the data parameter. Again, shows how to use this call back function. When a large number of almost identical files is used, it's quite inconvenient to use a separate call back function for each file. A better approach is to have a single call back function that distinguishes between the files by using the data field in struct proc_dir_entry. First of all, the data field has to be initialised: struct proc_dir_entry* entry; struct my_file_data *file_data; file_data = kmalloc(sizeof(struct my_file_data), GFP_KERNEL); entry->data = file_data; The data field is a void *, so it can be initialised with anything. Now that the data field is set, the read_proc and write_proc can use it to distinguish between files because they get it passed into their data parameter: int foo_read_func(char *page, char **start, off_t off, int count, int *eof, void *data) { int len; if(data == file_data) { /* special case for this file */ } else { /* normal processing */ } return len; } Be sure to free the data data field when removing the procfs entry. struct proc_dir_entry* create_proc_read_entry const char* name mode_t mode struct proc_dir_entry* parent read_proc_t* read_proc void* data This function creates a regular file in exactly the same way as create_proc_entry from does, but also allows to set the read function read_proc in one call. This function can set the data as well, like explained in . If procfs is being used from within a module, be sure to set the owner field in the struct proc_dir_entry to THIS_MODULE. struct proc_dir_entry* entry; entry->owner = THIS_MODULE; Sometimes it is useful to change the mode and/or ownership of a procfs entry. Here is an example that shows how to achieve that: struct proc_dir_entry* entry; entry->mode = S_IWUSR |S_IRUSR | S_IRGRP | S_IROTH; entry->uid = 0; entry->gid = 100; &procfsexample;