Linux Kernel Hackers' Guide

Go get the real thing!

Run, don't walk to get a copy of this book.

The Linux Kernel

Table of Contents

Tour of the Linux Kernel

This is a somewhat incomplete tour of the Linux Kernel, based on Linux 1.0.9 and the 1.1.x development series. Most of it is still relevant.

Device Drivers

The most common Linux kernel programming task is writing a new device driver. The great majority of the code in the kernel is new device drivers; between 1.2.13 and 2.0 the size of the source code more than doubled, and most of that was from adding device drivers.

Filesystems

Adding a filesystem to Linux doesn't have to involve magic...

Linux Memory Management

A few outdated documents, and one completely new one by David Miller on the Linux cache flush architecture.

How System Calls Work on Linux/i86

Although this was written while Linux 0.99.2 was current, it still applies. A few filenames may need updating. find is your friend--just respond with the changes and they will be added.

Other Sources of Information

The KHG is just one collection of information about the Linux kernel. There are others!

Membership and Subscription

At the bottom of the page, you will notice two hyperlinks (among several others): Subscribe and Members. Using the KHG to its fullest involves these two hyperlinks, even though you are not required to be a member to read these pages and post responses.

Membership

HyperNews membership is site-wide. That is, you only need to sign up and become a member once for the entire KHG. It doesn't take much to be a member. Each member is identified by a unique name, which can either be a nickname or an email address. We suggest using your email address; that way it will be unique and easy to remember. On the other hand, you may want to choose a nickname if you expect to be changing your email address at any time.

We also want your real name, email address, and home page (if you have one). You can give us your phone and address if you want. You will be asked to choose a password. You can change any of these items at any time by clicking on the Membership hyperlink again.

Subscription

Subscribing to a page puts you on a mailing list to be sent notification of any new responses to the page to which you are subscribed. You subscribe separately to each page in which you are interested by clicking the Subscription link on the page to which you want to subscribe. You are also subscribed, by default, to pages that you write.

When you subscribe to a page, you subscribe to that page and all of its responses.

Contributing

Please respond to these pages if you have something to add. Think of posting a response rather like posting to an email list, except that an editor might occasionally come along to clean things up and/or put them in the main documents' bodies. So if you would post it to an email list in a similar discussion, it is probably appropriate to post here.

In order to make reading these pages a pleasure for everyone, any incomprehensible, unrelated, outdated, abusive, or other completely unnecessary post may be removed by an administrator. So if you have a message that would be inappropriate on a mailing list, it's probably also inappropriate here.

The administrators have the final say on what's appropriate. We don't expect this to become an issue...

About the new KHG

The Linux Kernel Hackers' Guide has changed quite a bit since its original conception four years ago. I struggled along with the help of many other hackers to produce a document that lived primarily on paper, and was intended to document the kernel in much the same way that a program's user guide is intended to document the program for users.

It was less successful than most user guides, for a number of reasons:

• I was working on it part time, and was otherwise busy.
• The Linux kernel is a moving target.
• I am not personally capable of documenting the entire Linux kernel.
• I became far too concerned with making the typesetting pretty, getting bogged down in details and making the document typographically noisy at the same time.

There is a clear need for further development of the KHG, and it's clear that my making it a monolithic document stood in the way of progress. The KHG is now a series of more or less independent web pages, with places for readers to leave comments and corrections that can be incorporated in the document at the maintainer's leisure--and are available to readers before they are incorporated.

The KHG is now completely web-based. There will be no official paper version. You need kernel source code nearby to read the KHG anyway, and I want to shift the emphasis from officially documenting the Linux kernel to being a learning resource about the Linux kernel--one that may well be useful to other people who want to document one part or another of the Linux kernel more fully, as well as to people who just want to hack the kernel.

Enjoy!

Copyright (C) 1996,1997 Michael K. Johnson, [email protected]

Messages

349. Loading shared objects - How? by Wesley Terpstra

342. How can I see the current kernel configuration? by Melwin

1. My mouse no work in X windows by alfonso santana

340. The crash(1M) command in Linux? by Dmitry

338. Where can I gen detailed info on VM86 by Sebastien Plante

335. How to print floating point numbers from the kernel? by [email protected]

333. PS/2 Mouse Operating in Remote Mode by Andrei Racz

331. basic module by [email protected]

329. How to check if the user is local? by [email protected]

328. Ldt & Privileges by Ganesh

326. skb queues by Rahul Singh

323. Page locking (for DMA) and process termination? by Espen Skoglund

322. SMP code by [email protected]

319. Porting GC: Difficulties with pthreads by Talin

314. Linux for "Besta - 88"? by Dmitry

1. MVME147 Linux by Edward Tulupnikov

313. /proc/locks by Marco Morandini

310. syscall by [email protected]

308. How to run a bigger kernel ? by Kyung D. Ryu

300. Linux Terminal Device Driver by Nils Appeldoorn

1. Terminal DD by Doug McNash

297. DMA to user allocated buffer ? by Chris Read

1. allocator-example in A.Rubini's book by Thomas Sefzick

293. Patching problems by Maryam

1. Untitled by [email protected]

290. Ethernet Collision by jerome bonnet

1. Ethernet collisions by Juha Laine

289. Segmentation in Linux by Andrew Sampson

288. How can the kernel copy directly data from one process to another process? by J�rgen Zeller

1. Use the /Proc file system by [email protected]

286. Remapping Memory Buffer using vmalloc/vma_nopage by Brian W. Taylor

1. Fixed.... strncpy to blame by Brian W. Taylor

283. Does memory area assigned by "vmalloc()" get swapped to disk? by Saurabh Desai

1. Lock the pages in memory by [email protected]

-> How about assigning a fixed size array...does it get swapped too? by saurabh desai

282. Creative Lab's DVD Encore by Brandon

274. TCP sliding window by Olivier

273. Packets and default route versus direct route by Steve Resnick

269. IPv6 description - QoS Implementation - 2 IP Queues by wehrle

2. See the kernel IPv4 implementation documentation by Juha Laine

268. writing to user file directly from kernel space, How can it be done? by Johan

267. how can i increase the number of processes running? by ElmerFudd

261. How do I change the amount of time a process is allowed before it is pre-empted? by [email protected]

260. Network device stops after a while by Andrew Ordin

1. Untitled by Andrew

259. Does MMAP work with Redhat 4.2? by Guy

1. Yes, it works just fine. by Michael K. Johnson

3. What about mprotect? by Sengan Baring-Gould

2. It Works! Thanks! by Guy

256. multitasking by Dennis J Perkins

1. Answer by David Welch

-> multitasking by Dennis J Perkins

-> answer by David Welch

247. linux on sparc by darrin hodges

241. How to call a function in user space from inside the kernel ? by Ronald Tonn

1. How to call a user routine from kernel mode by David Welch

240. Can I map kernel (device driver) memory into user space ? by Ronald Tonn

237. driver x_open,x_release work, x_ioctl,x_write don't by Carl Schwartz

1. Depmod Unresolved symbols? by Carl Schwartz

235. How to sleep for x jiffies? by Trent Piepho

1. Use add_timer/del_timer (in kernel/sched.c) by Amos Shapira

234. Adding code to the Linux Kernel by Patrick

1. /dev/random by Simon Green

231. MSG_WAITALL flag by Leonard Mosescu

230. possible bug in ipc/msg.c by Michael Adda

225. scheduler Question by Arne Spetzler

1. Untitled by Ovsov

-> thanks by arne spetzler

221. File Descriptor Passing? by The Llamatron

220. Linux SMP Scheduling by Angela

2. Finding definitions in the source by Felix Rauch

1. Re: Linux SMP Scheduling by Franky

217. Difference between ELF and kernel file by Thomas Prokosch

216. How kernel communicates with outside when it's started? by Xuan Cao

1. Printing to the kernel log by Thomas Prokosch

213. The way from a kernel hackers' idea to an "official" kernel? by Roger Schreiter

1. [email protected] by Michael K. Johnson

212. Curious about sleep_on_interruptible() in ancient kernels. by Colin Howell

208. Server crashes using 2.0.32 and SMP by Steve Resnick

1. Debugging server crash by Balaji Srinivasan

-> More Information by Steve Resnick

-> it should not have happenned... by Balaji Srinivasan

207. Signals ID definitions by Franky

206. the segment D000 is not visible by [email protected]

205. ICMP - Supressing Messages in 2.1.82 Kernel by Brent Johnson

1. Change /etc/syslog.conf by Balaji Srinivasan

203. Modem bits by Franky

1. Untitled by Kostya

200. I need some way to measure the time a process spend in READY QUEUE by Leandro Gelasi

197. How to make sockets work whilst my process is in kernel mode? by Mikhail Kourinny

193. Realtime Problem by Uwe Gaethke

1. SCHED_FIFO scheduling by Balaji Srinivasan

190. inodes by Ovsov

186. Difference between SOCK_RAW SOCK_PACKET by Chris Leung

1. SOCK_PACKET by Eddie Leung

185. Need additional termcap entries for TERM=linux by Karl Bullock

184. Question on Umount or sys_umount by teddy

183. Passing file descriptors to the kernel by Pradeep Gore

1. A way to "transform" a file descriptor into a struct file* in a user process by Lorenzo Cavallaro

181. Dead Man Timer by Jody Winston

179. raw sockets by lightman

178. a kernel-hacking newbie by Bradley Lawrence

2. A place to start.

1. Modems in general by Ian Carr-de Avelon

176. How to write CD-ROM Driver ? Any Source Code ? by Madhura Upadhya

174. Measuring the scheduler overhead by Jasleen Kaur

173. Where can I find the tcpdump or snoop in linux? by wangc@taurus

1. man which by [email protected]

172. Timers don't work?? by Joshua Liew

1. Timers Work... by Balaji Srinivasan

171. problem of Linux's bridge code by wangc@taurus

170. Documention on writing kernel modules by Erik Nygren

168. How to display a clock on my console? by keco

167. Difference between SCO and Linux drivers. by M COTE

165. Changing the scheduler from round robin to shortest job first for kernel 2.0 and up by [email protected]

2. Improving the Scheduer by Lee Ingram

1. Improving the Scheduler : use QNX-like by Leandro Gelasi

1. Re: Changing the sched. from round robin to shortest job first for kernel 2.0 and up. by Pirasenna V.T.

164. meanings of file->private_data by [email protected]

162. /dev/signalprocess by flatmax

161. how to track VM page access sequence? by shawn

160. Whats the difference between dev_tint(dev) and mark_bh(NET_BH)? by Jaspreet Singh

159. PCI by [email protected]

1. RE: PCI by Armin A. Arbinger

158. Can I make syscall from inside a kernel module? by Shawn Chang

3. Re: Can I make syscall from inside a kernel module? by Massoud Asgharifard

1. Make a syscall despite of wrong fs!! by Mikhail Kourinny

2. code snip to make a sys_* call from a module by Pradeep Gore

1. Dont use system calls within kernel...(esp sys_mlock) by Balaji Srinivasan

157. Untitled by Steve Durst

154. RAW Sockets (Art)

153. use phy mem by WYB

151. HyperNews for RH Linux ? by Eigil Krogh Sorensen

1. Not really needed by Cameron

150. about raw ethernet frame: how to do it ? by [email protected]

149. process table by Blaz Novak

148. Stream drivers by Nick Egorov

3. Streams drivers

1. Stream in Solaris by [email protected]

143. Xircom External Ethernet driver anywhere? by mike head

140. interruptible_sleep_on() too slow! by Bill Blackwell

1. wrong functions by Michael K. Johnson

139. creating a kernel relocatable module by Simon Kittle

138. Up to date serial console patches by Simon Green

136. Kernel-Level Support for Checkpointing on Linux? by Argenis R. Fernandez

1. Working on it. by Jan Rychter

135. Problem creating a new system call by sauru

3. How did the file /arch/i386/kernel/entry.S do its job by Wang Ju

2. system call returns "Bad Address". Why? by sauru

1. Re:return values by C.H.Gopinath

2. Re:return values by Sameer Shah

1. possible reason for segmentation fault

1. Creating a new sytem call: solution by C.H.Gopinath

2. problem with system call slot 167 by Todd Medlock

1. Kernel Debuggers for Linux by sauru

133. Resetting interface counters by Keith Dart

130. writing/accessing modules by Jones MB

1. Use a device driver and read()/write()/ioctl() by Michael K. Johnson

-> getting to the kernel's memory by Jones MB

-> use buffers! by Rubens

124. Help with CPU scheduler! by Lerris

1. Response to "Help with CPU scheduler!" by Jeremy Impson

-> Response to "Help with CPU scheduler!" (Redux) by Jeremy Impson

117. calling interupts from linux by John J. Binder

1. You can't by Michael K. Johnson

-> Calling BIOS interrupts from Linux kernel by Ian Collier

-> Possible, but takes work by Michael K. Johnson

-> VBE video driver by Ian Collier

-> VM86 mode at which abstraction level? by Michael K. Johnson

116. DVD-ROM and Linux? (sorry if it's off topic...) by Joel Hardy

3. DVD-ROM and linux by [email protected]

2. Response to DVD and Mpeg in Linux by Mike Corrieri

1. DVD Encryption by Mark Treiber

-> Untitled by Tim

-> DVD?

115. Kernel Makefile Configuration: how? by Simon Green

2. How to add a driver to the kernel ? by jacek Radajewski

1. See include/linux/autoconf.h by Balaji Srinivasan

113. Multiprocessor Linux by Davis Terrell

1. Building an SMP kernel by Michael K. Johnson

-> SMP and module versions by [email protected]

111. Improving event timers? by [email protected]

109. measuring time to load a virtual mem page from disk by kandr

94. using cli/sti() and save_flags/restore_flags() by george

93. Protected Mode by ac

2. 'Developers manual' from Intel(download)... by Mats Odman

1. Advanced 80386 Programming Techniques by Michael K. Johnson

92. DMA buffer sizes by [email protected]

2. DMA limits by Albert Cahalan <acahalan at cs.uml.edu>

1. Not page size, page order by Michael K. Johnson

91. Problem Getting the Kernel small enough by [email protected]

2. Check it's the right file, zImage not vmlinux by Cameron

1. Usually easy, but.... by Ian Carr-de Avelon

89. How to create /proc/sys variables? by Orlando Cantieni

88. Linux for NeXT black? by Dale Amon

87. vremap() in kernel modules? by Liam Wickins

86. giveing compatiblity to win95 for ext2 partitions (for programmers forced to deal with both) by pharos

2. Well, What's the status of the Windows / Dos driver for Ext2? by Brock Lynn

1. Working on it! by ibaird

1. revision by ibarid

-> Untitled by Olaf

84. setsockopt() error when triying to use ipfwadm for masquerading by [email protected]

1. Re: masquerading by Charles Barrasso

83. reset the irq 0 timer after APM suspend by Dong Chen

1. Re: fixed, patch for kernel 2.0.30 by Dong Chen

77. Source Code in C for make Linux partitions. by Limbert Sanabria

1. Untitled by lolley

76. How can I "cheat" and change the IP address (src,dest) in the sent socket? by Rami

5. Transparent Proxy by Zygo Blaxell

4. Untitled by [email protected]

3. Untitled by [email protected]

2. Changing your IP address is easy, but... by Zygo Blaxell

1. You have to know a bit of C (if u wanna learn) ;) by Lorenzo Cavallaro

2. Untitled

1. Do it in the kernel by Michael K. Johnson

74. Where is the source file for accept() by [email protected]

1. Here, in /usr/src/linux/net/socket.c by [email protected]

72. How can I use RAW SOCKETS in UNIX? by Rami

1. Re: Raw sockets by [email protected]

69. the KHG in spanish? by Jorge Alvarado Revatta

2. Si tenga preguntas, quisa yo pueda ayudarte. by KernelJock

3. Tengo una pregunta by [email protected]

2. Espa�ol by LL2

1. No esta aqui! Pero... by Michael K. Johnson

67. How to get a Memory snapshot ? by Manuel Porras Brand

1. Why not to get a memory snapshot? by Jukka Santala

-> Why you would want to get a memory snapshot by Dave M.

66. resources hard limits by [email protected]

1. Setting resource limits by Jukka Santala

65. How to invalidate a chache page by Gerhard Uttenthaler

1. Read the rest of the KHG! by Michael K. Johnson

64. Where are the tunable parameters of the kernel? by [email protected]

1. Kernel tunable parameters by Jukka Santala

62. How can my device driver access data structures in user space? by Stephan Theil

1. Forced Cast data type by Wang Ju

61. Problem in doing RAW SOCKET Programming by anjali sharma

1. Problem with ICMP echo-request/reply by Raghavendra Bhat

59. Tunable Kernel Parameters? by [email protected]

2. Increasing number of files in system by Simon Cooper

1. Increasing number of open files parameter by Simon Cooper

1. sysctl in Linux by Jukka Santala

1. Setting and getting kernel vars by [email protected]

58. ELF matters by Carlos Munoz

1. Information about ELF Internals by Pat Ekman

57. Droping Packets by Charles Barrasso

1. [Selectively] Droping Packets by Jose R. cordones

56. The /proc/profile by Charles Barrasso

1. readprofile systool by Jukka Santala

55. Can you block or ignore ICMP packets? by [email protected]

4. ICMP send rate limit / ignoring by Jukka Santala

1. Omission in earlier rate-limit... by Jukka Santala

-> Patch worked... by Jukka Santala

3. Using ipfwadm by Charles Barrasso

1. ipfwadm configuration utility by Sonny Parlin

1. Icmp.c and kernal ping replies by Don Thomas

52. encaps documentation by Kuang-chun Cheng

51. Mounting Caldrea OpenDOS formatted fs's by Trey Childs

49. finding the address that caused a SIGSEGV. by Ben Shelef

47. sti() called too late. by Erik Thiele

1. sti() called too late. by Gadi Oxman

38. Module Development Info? by Mark S. Mathews

1. Needed here too by ajay

2. Help needed here too! by ajay

35. Need quicker timer than 100 ms in kernel-module by Erik Thiele

1. 10 ms timer patch by Reinhold J. Gerharz

2. please send me 10 ms timer patch by Tolga Ayav

1. Please send me the patch by Jin Hwang

1. UTIME: Microsecond Resolution Timers by BalajiSrinivasan

34. Need help with finding the linked list of unacked sk_buffs in TCP by Vijay Gupta

31. Partition Type by Suman Ball

30. New document on exception handling by Michael K. Johnson

29. How to make paralelism in to the kernel? by Delian Dlechev

27. readv/writev & other sock funcs by Dave Wreski

25. I'd like to see the scheduler chapter by Tim Bird

1. Untitled by Vijay Gupta

3. Go ahead! by Michael K. Johnson

21. Unable to access KHG, port 8080 giving problem. by Srihari Nelakuditi

1. Get a proxy by Michael K. Johnson

20. proc fs docs? by David Woodruff

1. Examples code as documentation by Jeremy Impson

18. What is SOCK_RAW and how do I use it? by arkane

1. What raw sockets are for. by Cameron MacKinnon

15. Linux kernel debugging by [email protected]

2. GDB for Linux by David Grothe

2. Another kernel debugging tool by David Hinds

2. Kernel debugging with breakpoints by Keith Owens

-> Need help for debugging by C.H.Gopinath

1. gdb debugging of kernel now available by David Grothe

1. Device debugging by alombard�iiic.ethz.ch

9. Realtime mods anyone? by bill duncan

7. Summary of Linux Real-Time Status by Markus Kuhn

6. Hard real-time now available by Michael K. Johnson

2. Shortcomings of RT-Linux by Balaji Srinivasan

1. Firm Realtime available by Balaji Srinivasan

5. found some hacks ?!? by Mayk Langer

2. I want to know how to hack Red Hat Linux Release 5.0 by Kevin

4. POSIX.4 scheduler by Peter Monta

1. cli()/sti() latency, hard numbers by Ingo Molnar

2. Realtime is already done(!) by Kai Harrekilde-Petersen

1. 100 ms real time should be easy by jeff millar

1. Real-Time Applications with Linux POSIX.4 Scheduling by P. Woolley

7. Why can't we incorporate new changes in linux kernel in KHG ? by Praveen Kumar Dwivedi

1. You can! by Michael K. Johnson

3. Kernel source code by Gabor J.Toth

1. The sounds of silence... by Gabor J.Toth

1. Breaking the silence :) by Kyle Ferrio

1. Scribbling in the margins by Michael K. Johnson

2. It requires thought... by Michael K. Johnson

2. Kernel source is already browsable online by Axel Boldt

2. Need easy way to download whole KHG

5. KHG being mirrored nightly for download! by Michael K. Johnson

2. postscript version of these documents? by Michael Stiller

1. Sure! by Michael K. Johnson

-> Not so Sure! by jeff millar

-> Enough already! by Michael K. Johnson

1. Mirror packages are available, but that's not really enough by Michael K. Johnson

4. Mirror whole KHG package, off line reading and Post to this site by Kim In-Sung

2. Untitled by Jim Van Zandt

1. That works. (using it now). Two tips: by Richard Braakman

2. Appears to be a bug in getwww, though... by Michael K. Johnson

-> Sucking up to the wrong site... ;) by Jukka Santala

1. Help make the new KHG a success by Michael K. Johnson