1 <!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook V3.1//EN"[]>
3 <book id="lk-hacking-guide">
5 <title>Unreliable Guide To Hacking The Linux Kernel</title>
9 <firstname>Paul</firstname>
10 <othername>Rusty</othername>
11 <surname>Russell</surname>
14 <email>rusty@rustcorp.com.au</email>
22 <holder>Rusty Russell</holder>
27 This documentation is free software; you can redistribute
28 it and/or modify it under the terms of the GNU General Public
29 License as published by the Free Software Foundation; either
30 version 2 of the License, or (at your option) any later
35 This program is distributed in the hope that it will be
36 useful, but WITHOUT ANY WARRANTY; without even the implied
37 warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
38 See the GNU General Public License for more details.
42 You should have received a copy of the GNU General Public
43 License along with this program; if not, write to the Free
44 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston,
49 For more details see the file COPYING in the source
50 distribution of Linux.
55 This is the first release of this document as part of the kernel tarball.
62 <chapter id="introduction">
63 <title>Introduction</title>
65 Welcome, gentle reader, to Rusty's Unreliable Guide to Linux
66 Kernel Hacking. This document describes the common routines and
67 general requirements for kernel code: its goal is to serve as a
68 primer for Linux kernel development for experienced C
69 programmers. I avoid implementation details: that's what the
70 code is for, and I ignore whole tracts of useful routines.
73 Before you read this, please understand that I never wanted to
74 write this document, being grossly under-qualified, but I always
75 wanted to read it, and this was the only way. I hope it will
76 grow into a compendium of best practice, common starting points
77 and random information.
81 <chapter id="basic-players">
82 <title>The Players</title>
85 At any time each of the CPUs in a system can be:
91 not associated with any process, serving a hardware interrupt;
97 not associated with any process, serving a softirq, tasklet or bh;
103 running in kernel space, associated with a process;
109 running a process in user space.
115 There is a strict ordering between these: other than the last
116 category (userspace) each can only be pre-empted by those above.
117 For example, while a softirq is running on a CPU, no other
118 softirq will pre-empt it, but a hardware interrupt can. However,
119 any other CPUs in the system execute independently.
123 We'll see a number of ways that the user context can block
124 interrupts, to become truly non-preemptable.
127 <sect1 id="basics-usercontext">
128 <title>User Context</title>
131 User context is when you are coming in from a system call or
132 other trap: you can sleep, and you own the CPU (except for
133 interrupts) until you call <function>schedule()</function>.
134 In other words, user context (unlike userspace) is not pre-emptable.
139 You are always in user context on module load and unload,
140 and on operations on the block device layer.
145 In user context, the <varname>current</varname> pointer (indicating
146 the task we are currently executing) is valid, and
147 <function>in_interrupt()</function>
148 (<filename>include/asm/hardirq.h</filename>) is <returnvalue>false
154 Beware that if you have interrupts or bottom halves disabled
155 (see below), <function>in_interrupt()</function> will return a
161 <sect1 id="basics-hardirqs">
162 <title>Hardware Interrupts (Hard IRQs)</title>
165 Timer ticks, <hardware>network cards</hardware> and
166 <hardware>keyboard</hardware> are examples of real
167 hardware which produce interrupts at any time. The kernel runs
168 interrupt handlers, which services the hardware. The kernel
169 guarantees that this handler is never re-entered: if another
170 interrupt arrives, it is queued (or dropped). Because it
171 disables interrupts, this handler has to be fast: frequently it
172 simply acknowledges the interrupt, marks a `software interrupt'
173 for execution and exits.
177 You can tell you are in a hardware interrupt, because
178 <function>in_irq()</function> returns <returnvalue>true</returnvalue>.
182 Beware that this will return a false positive if interrupts are disabled
188 <sect1 id="basics-softirqs">
189 <title>Software Interrupt Context: Bottom Halves, Tasklets, softirqs</title>
192 Whenever a system call is about to return to userspace, or a
193 hardware interrupt handler exits, any `software interrupts'
194 which are marked pending (usually by hardware interrupts) are
195 run (<filename>kernel/softirq.c</filename>).
199 Much of the real interrupt handling work is done here. Early in
200 the transition to <acronym>SMP</acronym>, there were only `bottom
201 halves' (BHs), which didn't take advantage of multiple CPUs. Shortly
202 after we switched from wind-up computers made of match-sticks and snot,
203 we abandoned this limitation.
207 <filename class=headerfile>include/linux/interrupt.h</filename> lists the
208 different BH's. No matter how many CPUs you have, no two BHs will run at
209 the same time. This made the transition to SMP simpler, but sucks hard for
210 scalable performance. A very important bottom half is the timer
211 BH (<filename class=headerfile>include/linux/timer.h</filename>): you
212 can register to have it call functions for you in a given length of time.
216 2.3.43 introduced softirqs, and re-implemented the (now
217 deprecated) BHs underneath them. Softirqs are fully-SMP
218 versions of BHs: they can run on as many CPUs at once as
219 required. This means they need to deal with any races in shared
220 data using their own locks. A bitmask is used to keep track of
221 which are enabled, so the 32 available softirqs should not be
222 used up lightly. (<emphasis>Yes</emphasis>, people will
227 tasklets (<filename class=headerfile>include/linux/interrupt.h</filename>)
228 are like softirqs, except they are dynamically-registrable (meaning you
229 can have as many as you want), and they also guarantee that any tasklet
230 will only run on one CPU at any time, although different tasklets can
231 run simultaneously (unlike different BHs).
235 The name `tasklet' is misleading: they have nothing to do with `tasks',
236 and probably more to do with some bad vodka Alexey Kuznetsov had at the
242 You can tell you are in a softirq (or bottom half, or tasklet)
243 using the <function>in_softirq()</function> macro
244 (<filename class=headerfile>include/asm/softirq.h</filename>).
248 Beware that this will return a false positive if a bh lock (see below)
255 <chapter id="basic-rules">
256 <title>Some Basic Rules</title>
260 <term>No memory protection</term>
263 If you corrupt memory, whether in user context or
264 interrupt context, the whole machine will crash. Are you
265 sure you can't do what you want in userspace?
271 <term>No floating point or <acronym>MMX</acronym></term>
274 The <acronym>FPU</acronym> context is not saved; even in user
275 context the <acronym>FPU</acronym> state probably won't
276 correspond with the current process: you would mess with some
277 user process' <acronym>FPU</acronym> state. If you really want
278 to do this, you would have to explicitly save/restore the full
279 <acronym>FPU</acronym> state (and avoid context switches). It
280 is generally a bad idea; use fixed point arithmetic first.
286 <term>A rigid stack limit</term>
289 The kernel stack is about 6K in 2.2 (for most
290 architectures: it's about 14K on the Alpha), and shared
291 with interrupts so you can't use it all. Avoid deep
292 recursion and huge local arrays on the stack (allocate
293 them dynamically instead).
299 <term>The Linux kernel is portable</term>
302 Let's keep it that way. Your code should be 64-bit clean,
303 and endian-independent. You should also minimize CPU
304 specific stuff, e.g. inline assembly should be cleanly
305 encapsulated and minimized to ease porting. Generally it
306 should be restricted to the architecture-dependent part of
314 <chapter id="ioctls">
315 <title>ioctls: Not writing a new system call</title>
318 A system call generally looks like this
322 asmlinkage int sys_mycall(int arg)
329 First, in most cases you don't want to create a new system call.
330 You create a character device and implement an appropriate ioctl
331 for it. This is much more flexible than system calls, doesn't have
332 to be entered in every architecture's
333 <filename class=headerfile>include/asm/unistd.h</filename> and
334 <filename>arch/kernel/entry.S</filename> file, and is much more
335 likely to be accepted by Linus.
339 If all your routine does is read or write some parameter, consider
340 implementing a <function>sysctl</function> interface instead.
344 Inside the ioctl you're in user context to a process. When a
345 error occurs you return a negated errno (see
346 <filename class=headerfile>include/linux/errno.h</filename>),
347 otherwise you return <returnvalue>0</returnvalue>.
351 After you slept you should check if a signal occurred: the
352 Unix/Linux way of handling signals is to temporarily exit the
353 system call with the <constant>-ERESTARTSYS</constant> error. The
354 system call entry code will switch back to user context, process
355 the signal handler and then your system call will be restarted
356 (unless the user disabled that). So you should be prepared to
357 process the restart, e.g. if you're in the middle of manipulating
362 if (signal_pending())
367 If you're doing longer computations: first think userspace. If you
368 <emphasis>really</emphasis> want to do it in kernel you should
369 regularly check if you need to give up the CPU (remember there is
370 cooperative multitasking per CPU). Idiom:
374 if (current->need_resched)
375 schedule(); /* Will sleep */
379 A short note on interface design: the UNIX system call motto is
380 "Provide mechanism not policy".
384 <chapter id="deadlock-recipes">
385 <title>Recipes for Deadlock</title>
388 You cannot call any routines which may sleep, unless:
393 You are in user context.
399 You do not own any spinlocks.
405 You have interrupts enabled (actually, Andi Kleen says
406 that the scheduling code will enable them for you, but
407 that's probably not what you wanted).
413 Note that some functions may sleep implicitly: common ones are
414 the user space access functions (*_user) and memory allocation
415 functions without <symbol>GFP_ATOMIC</symbol>.
419 You will eventually lock up your box if you break these rules.
427 <chapter id="common-routines">
428 <title>Common Routines</title>
430 <sect1 id="routines-printk">
432 <function>printk()</function>
433 <filename class=headerfile>include/linux/kernel.h</filename>
437 <function>printk()</function> feeds kernel messages to the
438 console, dmesg, and the syslog daemon. It is useful for debugging
439 and reporting errors, and can be used inside interrupt context,
440 but use with caution: a machine which has its console flooded with
441 printk messages is unusable. It uses a format string mostly
442 compatible with ANSI C printf, and C string concatenation to give
443 it a first "priority" argument:
447 printk(KERN_INFO "i = %u\n", i);
451 See <filename class=headerfile>include/linux/kernel.h</filename>;
452 for other KERN_ values; these are interpreted by syslog as the
453 level. Special case: for printing an IP address use
458 printk(KERN_INFO "my ip: %d.%d.%d.%d\n", NIPQUAD(ipaddress));
462 <function>printk()</function> internally uses a 1K buffer and does
463 not catch overruns. Make sure that will be enough.
468 You will know when you are a real kernel hacker
469 when you start typoing printf as printk in your user programs :)
473 <!--- From the Lions book reader department -->
477 Another sidenote: the original Unix Version 6 sources had a
478 comment on top of its printf function: "Printf should not be
479 used for chit-chat". You should follow that advice.
484 <sect1 id="routines-copy">
486 <function>copy_[to/from]_user()</function>
488 <function>get_user()</function>
490 <function>put_user()</function>
491 <filename class=headerfile>include/asm/uaccess.h</filename>
495 <emphasis>[SLEEPS]</emphasis>
499 <function>put_user()</function> and <function>get_user()</function>
500 are used to get and put single values (such as an int, char, or
501 long) from and to userspace. A pointer into userspace should
502 never be simply dereferenced: data should be copied using these
503 routines. Both return <constant>-EFAULT</constant> or 0.
506 <function>copy_to_user()</function> and
507 <function>copy_from_user()</function> are more general: they copy
508 an arbitrary amount of data to and from userspace.
511 Unlike <function>put_user()</function> and
512 <function>get_user()</function>, they return the amount of
513 uncopied data (ie. <returnvalue>0</returnvalue> still means
517 [Yes, this moronic interface makes me cringe. Please submit a
518 patch and become my hero --RR.]
521 The functions may sleep implicitly. This should never be called
522 outside user context (it makes no sense), with interrupts
523 disabled, or a spinlock held.
527 <sect1 id="routines-kmalloc">
528 <title><function>kmalloc()</function>/<function>kfree()</function>
529 <filename class=headerfile>include/linux/slab.h</filename></title>
532 <emphasis>[MAY SLEEP: SEE BELOW]</emphasis>
536 These routines are used to dynamically request pointer-aligned
537 chunks of memory, like malloc and free do in userspace, but
538 <function>kmalloc()</function> takes an extra flag word.
551 May sleep and swap to free memory. Only allowed in user
552 context, but is the most reliable way to allocate memory.
565 Don't sleep. Less reliable than <constant>GFP_KERNEL</constant>,
566 but may be called from interrupt context. You should
567 <emphasis>really</emphasis> have a good out-of-memory
568 error-handling strategy.
581 Allocate ISA DMA lower than 16MB. If you don't know what that
582 is you don't need it. Very unreliable.
589 If you see a <errorname>kmem_grow: Called nonatomically from int
590 </errorname> warning message you called a memory allocation function
591 from interrupt context without <constant>GFP_ATOMIC</constant>.
592 You should really fix that. Run, don't walk.
596 If you are allocating at least <constant>PAGE_SIZE</constant>
597 (<filename class=headerfile>include/asm/page.h</filename>) bytes,
598 consider using <function>__get_free_pages()</function>
600 (<filename class=headerfile>include/linux/mm.h</filename>). It
601 takes an order argument (0 for page sized, 1 for double page, 2
602 for four pages etc.) and the same memory priority flag word as
607 If you are allocating more than a page worth of bytes you can use
608 <function>vmalloc()</function>. It'll allocate virtual memory in
609 the kernel map. This block is not contiguous in physical memory,
610 but the <acronym>MMU</acronym> makes it look like it is for you
611 (so it'll only look contiguous to the CPUs, not to external device
612 drivers). If you really need large physically contiguous memory
613 for some weird device, you have a problem: it is poorly supported
614 in Linux because after some time memory fragmentation in a running
615 kernel makes it hard. The best way is to allocate the block early
616 in the boot process via the <function>alloc_bootmem()</function>
621 Before inventing your own cache of often-used objects consider
622 using a slab cache in
623 <filename class=headerfile>include/linux/slab.h</filename>
627 <sect1 id="routines-current">
628 <title><function>current</function>
629 <filename class=headerfile>include/asm/current.h</filename></title>
632 This global variable (really a macro) contains a pointer to
633 the current task structure, so is only valid in user context.
634 For example, when a process makes a system call, this will
635 point to the task structure of the calling process. It is
636 <emphasis>not NULL</emphasis> in interrupt context.
640 <sect1 id="routines-udelay">
641 <title><function>udelay()</function>/<function>mdelay()</function>
642 <filename class=headerfile>include/asm/delay.h</filename>
643 <filename class=headerfile>include/linux/delay.h</filename>
647 The <function>udelay()</function> function can be used for small pauses.
648 Do not use large values with <function>udelay()</function> as you risk
649 overflow - the helper function <function>mdelay()</function> is useful
650 here, or even consider <function>schedule_timeout()</function>.
654 <sect1 id="routines-endian">
655 <title><function>cpu_to_be32()</function>/<function>be32_to_cpu()</function>/<function>cpu_to_le32()</function>/<function>le32_to_cpu()</function>
656 <filename class=headerfile>include/asm/byteorder.h</filename>
660 The <function>cpu_to_be32()</function> family (where the "32" can
661 be replaced by 64 or 16, and the "be" can be replaced by "le") are
662 the general way to do endian conversions in the kernel: they
663 return the converted value. All variations supply the reverse as
664 well: <function>be32_to_cpu()</function>, etc.
668 There are two major variations of these functions: the pointer
669 variation, such as <function>cpu_to_be32p()</function>, which take
670 a pointer to the given type, and return the converted value. The
671 other variation is the "in-situ" family, such as
672 <function>cpu_to_be32s()</function>, which convert value referred
673 to by the pointer, and return void.
677 <sect1 id="routines-local-irqs">
678 <title><function>local_irq_save()</function>/<function>local_irq_restore()</function>
679 <filename class=headerfile>include/asm/system.h</filename>
683 These routines disable hard interrupts on the local CPU, and
684 restore them. They are reentrant; saving the previous state in
685 their one <varname>unsigned long flags</varname> argument. If you
686 know that interrupts are enabled, you can simply use
687 <function>local_irq_disable()</function> and
688 <function>local_irq_enable()</function>.
692 <sect1 id="routines-softirqs">
693 <title><function>local_bh_disable()</function>/<function>local_bh_enable()</function>
694 <filename class=headerfile>include/asm/softirq.h</filename></title>
697 These routines disable soft interrupts on the local CPU, and
698 restore them. They are reentrant; if soft interrupts were
699 disabled before, they will still be disabled after this pair
700 of functions has been called. They prevent softirqs, tasklets
701 and bottom halves from running on the current CPU.
705 <sect1 id="routines-processorids">
706 <title><function>smp_processor_id</function>()/<function>cpu_[number/logical]_map()</function>
707 <filename class=headerfile>include/asm/smp.h</filename></title>
710 <function>smp_processor_id()</function> returns the current
711 processor number, between 0 and <symbol>NR_CPUS</symbol> (the
712 maximum number of CPUs supported by Linux, currently 32). These
713 values are not necessarily continuous: to get a number between 0
714 and <function>smp_num_cpus()</function> (the number of actual
715 processors in this machine), the
716 <function>cpu_number_map()</function> function is used to map the
717 processor id to a logical number.
718 <function>cpu_logical_map()</function> does the reverse.
722 <sect1 id="routines-init">
723 <title><type>__init</type>/<type>__exit</type>/<type>__initdata</type>
724 <filename class=headerfile>include/linux/init.h</filename></title>
727 After boot, the kernel frees up a special section; functions
728 marked with <type>__init</type> and data structures marked with
729 <type>__initdata</type> are dropped after boot is complete (within
730 modules this directive is currently ignored). <type>__exit</type>
731 is used to declare a function which is only required on exit: the
732 function will be dropped if this file is not compiled as a module.
733 See the header file for use. Note that it makes no sense for a function
734 marked with <type>__init</type> to be exported to modules with
735 <function>EXPORT_SYMBOL()</function> - this will break.
738 Static data structures marked as <type>__initdata</type> must be initialised
739 (as opposed to ordinary static data which is zeroed BSS) and cannot be
745 <sect1 id="routines-init-again">
746 <title><function>__initcall()</function>/<function>module_init()</function>
747 <filename class=headerfile>include/linux/init.h</filename></title>
749 Many parts of the kernel are well served as a module
750 (dynamically-loadable parts of the kernel). Using the
751 <function>module_init()</function> and
752 <function>module_exit()</function> macros it is easy to write code
753 without #ifdefs which can operate both as a module or built into
758 The <function>module_init()</function> macro defines which
759 function is to be called at module insertion time (if the file is
760 compiled as a module), or at boot time: if the file is not
761 compiled as a module the <function>module_init()</function> macro
762 becomes equivalent to <function>__initcall()</function>, which
763 through linker magic ensures that the function is called on boot.
767 The function can return a negative error number to cause
768 module loading to fail (unfortunately, this has no effect if
769 the module is compiled into the kernel). For modules, this is
770 called in user context, with interrupts enabled, and the
771 kernel lock held, so it can sleep.
775 <sect1 id="routines-moduleexit">
776 <title> <function>module_exit()</function>
777 <filename class=headerfile>include/linux/init.h</filename> </title>
780 This macro defines the function to be called at module removal
781 time (or never, in the case of the file compiled into the
782 kernel). It will only be called if the module usage count has
783 reached zero. This function can also sleep, but cannot fail:
784 everything must be cleaned up by the time it returns.
788 <sect1 id="routines-module-use-counters">
789 <title> <function>MOD_INC_USE_COUNT</function>/<function>MOD_DEC_USE_COUNT</function>
790 <filename class=headerfile>include/linux/module.h</filename></title>
793 These manipulate the module usage count, to protect against
794 removal (a module also can't be removed if another module uses
795 one of its exported symbols: see below). Every reference to
796 the module from user context should be reflected by this
797 counter (e.g. for every data structure or socket) before the
798 function sleeps. To quote Tim Waugh:
808 sleep.. (might get unloaded here)
830 You can often avoid having to deal with these problems by using the
831 <structfield>owner</structfield> field of the
832 <structname>file_operations</structname> structure. Set this field
833 as the macro <symbol>THIS_MODULE</symbol>.
837 For more complicated module unload locking requirements, you can set the
838 <structfield>can_unload</structfield> function pointer to your own routine,
839 which should return <returnvalue>0</returnvalue> if the module is
840 unloadable, or <returnvalue>-EBUSY</returnvalue> otherwise.
846 <chapter id="queues">
848 <filename class=headerfile>include/linux/wait.h</filename>
851 <emphasis>[SLEEPS]</emphasis>
855 A wait queue is used to wait for someone to wake you up when a
856 certain condition is true. They must be used carefully to ensure
857 there is no race condition. You declare a
858 <type>wait_queue_head_t</type>, and then processes which want to
859 wait for that condition declare a <type>wait_queue_t</type>
860 referring to themselves, and place that in the queue.
863 <sect1 id="queue-declaring">
864 <title>Declaring</title>
867 You declare a <type>wait_queue_head_t</type> using the
868 <function>DECLARE_WAIT_QUEUE_HEAD()</function> macro, or using the
869 <function>init_waitqueue_head()</function> routine in your
874 <sect1 id="queue-waitqueue">
875 <title>Queuing</title>
878 Placing yourself in the waitqueue is fairly complex, because you
879 must put yourself in the queue before checking the condition.
880 There is a macro to do this:
881 <function>wait_event_interruptible()</function>
883 <filename class=headerfile>include/linux/sched.h</filename> The
884 first argument is the wait queue head, and the second is an
885 expression which is evaluated; the macro returns
886 <returnvalue>0</returnvalue> when this expression is true, or
887 <returnvalue>-ERESTARTSYS</returnvalue> if a signal is received.
888 The <function>wait_event()</function> version ignores signals.
891 Do not use the <function>sleep_on()</function> function family -
892 it is very easy to accidentally introduce races; almost certainly
893 one of the <function>wait_event()</function> family will do, or a
894 loop around <function>schedule_timeout()</function>. If you choose
895 to loop around <function>schedule_timeout()</function> remember
896 you must set the task state (with
897 <function>set_current_state()</function>) on each iteration to avoid
903 <sect1 id="queue-waking">
904 <title>Waking Up Queued Tasks</title>
907 Call <function>wake_up()</function>
909 <filename class=headerfile>include/linux/sched.h</filename>;,
910 which will wake up every process in the queue. The exception is
911 if one has <constant>TASK_EXCLUSIVE</constant> set, in which case
912 the remainder of the queue will not be woken.
917 <chapter id="atomic-ops">
918 <title>Atomic Operations</title>
921 Certain operations are guaranteed atomic on all platforms. The
922 first class of operations work on <type>atomic_t</type>
924 <filename class=headerfile>include/asm/atomic.h</filename>; this
925 contains a signed integer (at least 24 bits long), and you must use
926 these functions to manipulate or read atomic_t variables.
927 <function>atomic_read()</function> and
928 <function>atomic_set()</function> get and set the counter,
929 <function>atomic_add()</function>,
930 <function>atomic_sub()</function>,
931 <function>atomic_inc()</function>,
932 <function>atomic_dec()</function>, and
933 <function>atomic_dec_and_test()</function> (returns
934 <returnvalue>true</returnvalue> if it was decremented to zero).
938 Yes. It returns <returnvalue>true</returnvalue> (i.e. != 0) if the
939 atomic variable is zero.
943 Note that these functions are slower than normal arithmetic, and
944 so should not be used unnecessarily. On some platforms they
945 are much slower, like 32-bit Sparc where they use a spinlock.
949 The second class of atomic operations is atomic bit operations on a
950 <type>long</type>, defined in
952 <filename class=headerfile>include/asm/bitops.h</filename>. These
953 operations generally take a pointer to the bit pattern, and a bit
954 number: 0 is the least significant bit.
955 <function>set_bit()</function>, <function>clear_bit()</function>
956 and <function>change_bit()</function> set, clear, and flip the
957 given bit. <function>test_and_set_bit()</function>,
958 <function>test_and_clear_bit()</function> and
959 <function>test_and_change_bit()</function> do the same thing,
960 except return true if the bit was previously set; these are
961 particularly useful for very simple locking.
965 It is possible to call these operations with bit indices greater
966 than BITS_PER_LONG. The resulting behavior is strange on big-endian
967 platforms though so it is a good idea not to do this.
971 Note that the order of bits depends on the architecture, and in
972 particular, the bitfield passed to these operations must be at
973 least as large as a <type>long</type>.
977 <chapter id="symbols">
978 <title>Symbols</title>
981 Within the kernel proper, the normal linking rules apply
982 (ie. unless a symbol is declared to be file scope with the
983 <type>static</type> keyword, it can be used anywhere in the
984 kernel). However, for modules, a special exported symbol table is
985 kept which limits the entry points to the kernel proper. Modules
986 can also export symbols.
989 <sect1 id="sym-exportsymbols">
990 <title><function>EXPORT_SYMBOL()</function>
991 <filename class=headerfile>include/linux/module.h</filename></title>
994 This is the classic method of exporting a symbol, and it works
995 for both modules and non-modules. In the kernel all these
996 declarations are often bundled into a single file to help
997 genksyms (which searches source files for these declarations).
998 See the comment on genksyms and Makefiles below.
1002 <sect1 id="sym-exportnosymbols">
1003 <title><symbol>EXPORT_NO_SYMBOLS</symbol>
1004 <filename class=headerfile>include/linux/module.h</filename></title>
1007 If a module exports no symbols then you can specify
1011 anywhere in the module.
1012 In kernel 2.4 and earlier, if a module contains neither
1013 <function>EXPORT_SYMBOL()</function> nor
1014 <symbol>EXPORT_NO_SYMBOLS</symbol> then the module defaults to
1015 exporting all non-static global symbols.
1016 In kernel 2.5 onwards you must explicitly specify whether a module
1017 exports symbols or not.
1021 <sect1 id="sym-exportsymbols-gpl">
1022 <title><function>EXPORT_SYMBOL_GPL()</function>
1023 <filename class=headerfile>include/linux/module.h</filename></title>
1026 Similar to <function>EXPORT_SYMBOL()</function> except that the
1027 symbols exported by <function>EXPORT_SYMBOL_GPL()</function> can
1028 only be seen by modules with a
1029 <function>MODULE_LICENSE()</function> that specifies a GPL
1035 <chapter id="conventions">
1036 <title>Routines and Conventions</title>
1038 <sect1 id="conventions-doublelinkedlist">
1039 <title>Double-linked lists
1040 <filename class=headerfile>include/linux/list.h</filename></title>
1043 There are three sets of linked-list routines in the kernel
1044 headers, but this one seems to be winning out (and Linus has
1045 used it). If you don't have some particular pressing need for
1046 a single list, it's a good choice. In fact, I don't care
1047 whether it's a good choice or not, just use it so we can get
1052 <sect1 id="convention-returns">
1053 <title>Return Conventions</title>
1056 For code called in user context, it's very common to defy C
1057 convention, and return <returnvalue>0</returnvalue> for success,
1058 and a negative error number
1059 (eg. <returnvalue>-EFAULT</returnvalue>) for failure. This can be
1060 unintuitive at first, but it's fairly widespread in the networking
1065 The filesystem code uses <function>ERR_PTR()</function>
1067 <filename class=headerfile>include/linux/fs.h</filename>; to
1068 encode a negative error number into a pointer, and
1069 <function>IS_ERR()</function> and <function>PTR_ERR()</function>
1070 to get it back out again: avoids a separate pointer parameter for
1071 the error number. Icky, but in a good way.
1075 <sect1 id="conventions-borkedcompile">
1076 <title>Breaking Compilation</title>
1079 Linus and the other developers sometimes change function or
1080 structure names in development kernels; this is not done just to
1081 keep everyone on their toes: it reflects a fundamental change
1082 (eg. can no longer be called with interrupts on, or does extra
1083 checks, or doesn't do checks which were caught before). Usually
1084 this is accompanied by a fairly complete note to the linux-kernel
1085 mailing list; search the archive. Simply doing a global replace
1086 on the file usually makes things <emphasis>worse</emphasis>.
1090 <sect1 id="conventions-initialising">
1091 <title>Initializing structure members</title>
1094 The preferred method of initializing structures is to use
1095 designated initialisers, as defined by ISO C99, eg:
1098 static struct block_device_operations opt_fops = {
1100 .release = opt_release,
1102 .check_media_change = opt_media_change,
1106 This makes it easy to grep for, and makes it clear which
1107 structure fields are set. You should do this because it looks
1112 <sect1 id="conventions-gnu-extns">
1113 <title>GNU Extensions</title>
1116 GNU Extensions are explicitly allowed in the Linux kernel.
1117 Note that some of the more complex ones are not very well
1118 supported, due to lack of general use, but the following are
1119 considered standard (see the GCC info page section "C
1120 Extensions" for more details - Yes, really the info page, the
1121 man page is only a short summary of the stuff in info):
1131 Statement expressions (ie. the ({ and }) constructs).
1136 Declaring attributes of a function / variable / type
1157 Arithmetic on void pointers
1162 Non-Constant initializers
1167 Assembler Instructions (not outside arch/ and include/asm/)
1172 Function names as strings (__FUNCTION__)
1177 __builtin_constant_p()
1183 Be wary when using long long in the kernel, the code gcc generates for
1184 it is horrible and worse: division and multiplication does not work
1185 on i386 because the GCC runtime functions for it are missing from
1186 the kernel environment.
1189 <!-- FIXME: add a note about ANSI aliasing cleanness -->
1192 <sect1 id="conventions-cplusplus">
1196 Using C++ in the kernel is usually a bad idea, because the
1197 kernel does not provide the necessary runtime environment
1198 and the include files are not tested for it. It is still
1199 possible, but not recommended. If you really want to do
1200 this, forget about exceptions at least.
1204 <sect1 id="conventions-ifdef">
1205 <title>#if</title>
1208 It is generally considered cleaner to use macros in header files
1209 (or at the top of .c files) to abstract away functions rather than
1210 using `#if' pre-processor statements throughout the source code.
1215 <chapter id="submitting">
1216 <title>Putting Your Stuff in the Kernel</title>
1219 In order to get your stuff into shape for official inclusion, or
1220 even to make a neat patch, there's administrative work to be
1226 Figure out whose pond you've been pissing in. Look at the top of
1227 the source files, inside the <filename>MAINTAINERS</filename>
1228 file, and last of all in the <filename>CREDITS</filename> file.
1229 You should coordinate with this person to make sure you're not
1230 duplicating effort, or trying something that's already been
1235 Make sure you put your name and EMail address at the top of
1236 any files you create or mangle significantly. This is the
1237 first place people will look when they find a bug, or when
1238 <emphasis>they</emphasis> want to make a change.
1244 Usually you want a configuration option for your kernel hack.
1245 Edit <filename>Config.in</filename> in the appropriate directory
1246 (but under <filename>arch/</filename> it's called
1247 <filename>config.in</filename>). The Config Language used is not
1248 bash, even though it looks like bash; the safe way is to use only
1249 the constructs that you already see in
1250 <filename>Config.in</filename> files (see
1251 <filename>Documentation/kbuild/config-language.txt</filename>).
1252 It's good to run "make xconfig" at least once to test (because
1253 it's the only one with a static parser).
1257 Variables which can be Y or N use <type>bool</type> followed by a
1258 tagline and the config define name (which must start with
1259 CONFIG_). The <type>tristate</type> function is the same, but
1260 allows the answer M (which defines
1261 <symbol>CONFIG_foo_MODULE</symbol> in your source, instead of
1262 <symbol>CONFIG_FOO</symbol>) if <symbol>CONFIG_MODULES</symbol>
1267 You may well want to make your CONFIG option only visible if
1268 <symbol>CONFIG_EXPERIMENTAL</symbol> is enabled: this serves as a
1269 warning to users. There many other fancy things you can do: see
1270 the various <filename>Config.in</filename> files for ideas.
1276 Edit the <filename>Makefile</filename>: the CONFIG variables are
1277 exported here so you can conditionalize compilation with `ifeq'.
1278 If your file exports symbols then add the names to
1279 <varname>export-objs</varname> so that genksyms will find them.
1282 There is a restriction on the kernel build system that objects
1283 which export symbols must have globally unique names.
1284 If your object does not have a globally unique name then the
1285 standard fix is to move the
1286 <function>EXPORT_SYMBOL()</function> statements to their own
1287 object with a unique name.
1288 This is why several systems have separate exporting objects,
1289 usually suffixed with ksyms.
1297 Document your option in Documentation/Configure.help. Mention
1298 incompatibilities and issues here. <emphasis> Definitely
1299 </emphasis> end your description with <quote> if in doubt, say N
1300 </quote> (or, occasionally, `Y'); this is for people who have no
1301 idea what you are talking about.
1307 Put yourself in <filename>CREDITS</filename> if you've done
1308 something noteworthy, usually beyond a single file (your name
1309 should be at the top of the source files anyway).
1310 <filename>MAINTAINERS</filename> means you want to be consulted
1311 when changes are made to a subsystem, and hear about bugs; it
1312 implies a more-than-passing commitment to some part of the code.
1318 Finally, don't forget to read <filename>Documentation/SubmittingPatches</filename>
1319 and possibly <filename>Documentation/SubmittingDrivers</filename>.
1325 <chapter id="cantrips">
1326 <title>Kernel Cantrips</title>
1329 Some favorites from browsing the source. Feel free to add to this
1334 <filename>include/linux/brlock.h:</filename>
1337 extern inline void br_read_lock (enum brlock_indices idx)
1340 * This causes a link-time bug message if an
1341 * invalid index is used:
1343 if (idx >= __BR_END)
1344 __br_lock_usage_bug();
1346 read_lock(&__brlock_array[smp_processor_id()][idx]);
1351 <filename>include/linux/fs.h</filename>:
1355 * Kernel pointers have redundant information, so we can use a
1356 * scheme where we can return either an error code or a dentry
1357 * pointer with the same return value.
1359 * This should be a per-architecture thing, to allow different
1360 * error and pointer decisions.
1362 #define ERR_PTR(err) ((void *)((long)(err)))
1363 #define PTR_ERR(ptr) ((long)(ptr))
1364 #define IS_ERR(ptr) ((unsigned long)(ptr) > (unsigned long)(-1000))
1368 <filename>include/asm-i386/uaccess.h:</filename>
1372 #define copy_to_user(to,from,n) \
1373 (__builtin_constant_p(n) ? \
1374 __constant_copy_to_user((to),(from),(n)) : \
1375 __generic_copy_to_user((to),(from),(n)))
1379 <filename>arch/sparc/kernel/head.S:</filename>
1384 * Sun people can't spell worth damn. "compatability" indeed.
1385 * At least we *know* we can't spell, and use a spell-checker.
1388 /* Uh, actually Linus it is I who cannot spell. Too much murky
1389 * Sparc assembly will do this to ya.
1392 .asciz "compatability"
1394 /* Tested on SS-5, SS-10. Probably someone at Sun applied a spell-checker. */
1396 C_LABEL(cputypvar_sun4m):
1401 <filename>arch/sparc/lib/checksum.S:</filename>
1405 /* Sun, you just can't beat me, you just can't. Stop trying,
1406 * give up. I'm serious, I am going to kick the living shit
1407 * out of you, game over, lights out.
1412 <chapter id="credits">
1413 <title>Thanks</title>
1416 Thanks to Andi Kleen for the idea, answering my questions, fixing
1417 my mistakes, filling content, etc. Philipp Rumpf for more spelling
1418 and clarity fixes, and some excellent non-obvious points. Werner
1419 Almesberger for giving me a great summary of
1420 <function>disable_irq()</function>, and Jes Sorensen and Andrea
1421 Arcangeli added caveats. Michael Elizabeth Chastain for checking
1422 and adding to the Configure section. <!-- Rusty insisted on this
1423 bit; I didn't do it! --> Telsa Gwynne for teaching me DocBook.