1.. _kernel_hacking_hack:
2
3============================================
4Unreliable Guide To Hacking The Linux Kernel
5============================================
6
7:Author: Rusty Russell
8
9Introduction
10============
11
12Welcome, gentle reader, to Rusty's Remarkably Unreliable Guide to Linux
13Kernel Hacking. This document describes the common routines and general
14requirements for kernel code: its goal is to serve as a primer for Linux
15kernel development for experienced C programmers. I avoid implementation
16details: that's what the code is for, and I ignore whole tracts of
17useful routines.
18
19Before you read this, please understand that I never wanted to write
20this document, being grossly under-qualified, but I always wanted to
21read it, and this was the only way. I hope it will grow into a
22compendium of best practice, common starting points and random
23information.
24
25The Players
26===========
27
28At any time each of the CPUs in a system can be:
29
30-  not associated with any process, serving a hardware interrupt;
31
32-  not associated with any process, serving a softirq or tasklet;
33
34-  running in kernel space, associated with a process (user context);
35
36-  running a process in user space.
37
38There is an ordering between these. The bottom two can preempt each
39other, but above that is a strict hierarchy: each can only be preempted
40by the ones above it. For example, while a softirq is running on a CPU,
41no other softirq will preempt it, but a hardware interrupt can. However,
42any other CPUs in the system execute independently.
43
44We'll see a number of ways that the user context can block interrupts,
45to become truly non-preemptable.
46
47User Context
48------------
49
50User context is when you are coming in from a system call or other trap:
51like userspace, you can be preempted by more important tasks and by
52interrupts. You can sleep, by calling :c:func:`schedule()`.
53
54.. note::
55
56    You are always in user context on module load and unload, and on
57    operations on the block device layer.
58
59In user context, the ``current`` pointer (indicating the task we are
60currently executing) is valid, and :c:func:`in_interrupt()`
61(``include/linux/preempt.h``) is false.
62
63.. warning::
64
65    Beware that if you have preemption or softirqs disabled (see below),
66    :c:func:`in_interrupt()` will return a false positive.
67
68Hardware Interrupts (Hard IRQs)
69-------------------------------
70
71Timer ticks, network cards and keyboard are examples of real hardware
72which produce interrupts at any time. The kernel runs interrupt
73handlers, which services the hardware. The kernel guarantees that this
74handler is never re-entered: if the same interrupt arrives, it is queued
75(or dropped). Because it disables interrupts, this handler has to be
76fast: frequently it simply acknowledges the interrupt, marks a 'software
77interrupt' for execution and exits.
78
79You can tell you are in a hardware interrupt, because
80:c:func:`in_irq()` returns true.
81
82.. warning::
83
84    Beware that this will return a false positive if interrupts are
85    disabled (see below).
86
87Software Interrupt Context: Softirqs and Tasklets
88-------------------------------------------------
89
90Whenever a system call is about to return to userspace, or a hardware
91interrupt handler exits, any 'software interrupts' which are marked
92pending (usually by hardware interrupts) are run (``kernel/softirq.c``).
93
94Much of the real interrupt handling work is done here. Early in the
95transition to SMP, there were only 'bottom halves' (BHs), which didn't
96take advantage of multiple CPUs. Shortly after we switched from wind-up
97computers made of match-sticks and snot, we abandoned this limitation
98and switched to 'softirqs'.
99
100``include/linux/interrupt.h`` lists the different softirqs. A very
101important softirq is the timer softirq (``include/linux/timer.h``): you
102can register to have it call functions for you in a given length of
103time.
104
105Softirqs are often a pain to deal with, since the same softirq will run
106simultaneously on more than one CPU. For this reason, tasklets
107(``include/linux/interrupt.h``) are more often used: they are
108dynamically-registrable (meaning you can have as many as you want), and
109they also guarantee that any tasklet will only run on one CPU at any
110time, although different tasklets can run simultaneously.
111
112.. warning::
113
114    The name 'tasklet' is misleading: they have nothing to do with
115    'tasks', and probably more to do with some bad vodka Alexey
116    Kuznetsov had at the time.
117
118You can tell you are in a softirq (or tasklet) using the
119:c:func:`in_softirq()` macro (``include/linux/preempt.h``).
120
121.. warning::
122
123    Beware that this will return a false positive if a
124    :ref:`botton half lock <local_bh_disable>` is held.
125
126Some Basic Rules
127================
128
129No memory protection
130    If you corrupt memory, whether in user context or interrupt context,
131    the whole machine will crash. Are you sure you can't do what you
132    want in userspace?
133
134No floating point or MMX
135    The FPU context is not saved; even in user context the FPU state
136    probably won't correspond with the current process: you would mess
137    with some user process' FPU state. If you really want to do this,
138    you would have to explicitly save/restore the full FPU state (and
139    avoid context switches). It is generally a bad idea; use fixed point
140    arithmetic first.
141
142A rigid stack limit
143    Depending on configuration options the kernel stack is about 3K to
144    6K for most 32-bit architectures: it's about 14K on most 64-bit
145    archs, and often shared with interrupts so you can't use it all.
146    Avoid deep recursion and huge local arrays on the stack (allocate
147    them dynamically instead).
148
149The Linux kernel is portable
150    Let's keep it that way. Your code should be 64-bit clean, and
151    endian-independent. You should also minimize CPU specific stuff,
152    e.g. inline assembly should be cleanly encapsulated and minimized to
153    ease porting. Generally it should be restricted to the
154    architecture-dependent part of the kernel tree.
155
156ioctls: Not writing a new system call
157=====================================
158
159A system call generally looks like this::
160
161    asmlinkage long sys_mycall(int arg)
162    {
163            return 0;
164    }
165
166
167First, in most cases you don't want to create a new system call. You
168create a character device and implement an appropriate ioctl for it.
169This is much more flexible than system calls, doesn't have to be entered
170in every architecture's ``include/asm/unistd.h`` and
171``arch/kernel/entry.S`` file, and is much more likely to be accepted by
172Linus.
173
174If all your routine does is read or write some parameter, consider
175implementing a :c:func:`sysfs()` interface instead.
176
177Inside the ioctl you're in user context to a process. When a error
178occurs you return a negated errno (see
179``include/uapi/asm-generic/errno-base.h``,
180``include/uapi/asm-generic/errno.h`` and ``include/linux/errno.h``),
181otherwise you return 0.
182
183After you slept you should check if a signal occurred: the Unix/Linux
184way of handling signals is to temporarily exit the system call with the
185``-ERESTARTSYS`` error. The system call entry code will switch back to
186user context, process the signal handler and then your system call will
187be restarted (unless the user disabled that). So you should be prepared
188to process the restart, e.g. if you're in the middle of manipulating
189some data structure.
190
191::
192
193    if (signal_pending(current))
194            return -ERESTARTSYS;
195
196
197If you're doing longer computations: first think userspace. If you
198**really** want to do it in kernel you should regularly check if you need
199to give up the CPU (remember there is cooperative multitasking per CPU).
200Idiom::
201
202    cond_resched(); /* Will sleep */
203
204
205A short note on interface design: the UNIX system call motto is "Provide
206mechanism not policy".
207
208Recipes for Deadlock
209====================
210
211You cannot call any routines which may sleep, unless:
212
213-  You are in user context.
214
215-  You do not own any spinlocks.
216
217-  You have interrupts enabled (actually, Andi Kleen says that the
218   scheduling code will enable them for you, but that's probably not
219   what you wanted).
220
221Note that some functions may sleep implicitly: common ones are the user
222space access functions (\*_user) and memory allocation functions
223without ``GFP_ATOMIC``.
224
225You should always compile your kernel ``CONFIG_DEBUG_ATOMIC_SLEEP`` on,
226and it will warn you if you break these rules. If you **do** break the
227rules, you will eventually lock up your box.
228
229Really.
230
231Common Routines
232===============
233
234:c:func:`printk()`
235------------------
236
237Defined in ``include/linux/printk.h``
238
239:c:func:`printk()` feeds kernel messages to the console, dmesg, and
240the syslog daemon. It is useful for debugging and reporting errors, and
241can be used inside interrupt context, but use with caution: a machine
242which has its console flooded with printk messages is unusable. It uses
243a format string mostly compatible with ANSI C printf, and C string
244concatenation to give it a first "priority" argument::
245
246    printk(KERN_INFO "i = %u\n", i);
247
248
249See ``include/linux/kern_levels.h``; for other ``KERN_`` values; these are
250interpreted by syslog as the level. Special case: for printing an IP
251address use::
252
253    __be32 ipaddress;
254    printk(KERN_INFO "my ip: %pI4\n", &ipaddress);
255
256
257:c:func:`printk()` internally uses a 1K buffer and does not catch
258overruns. Make sure that will be enough.
259
260.. note::
261
262    You will know when you are a real kernel hacker when you start
263    typoing printf as printk in your user programs :)
264
265.. note::
266
267    Another sidenote: the original Unix Version 6 sources had a comment
268    on top of its printf function: "Printf should not be used for
269    chit-chat". You should follow that advice.
270
271:c:func:`copy_to_user()` / :c:func:`copy_from_user()` / :c:func:`get_user()` / :c:func:`put_user()`
272---------------------------------------------------------------------------------------------------
273
274Defined in ``include/linux/uaccess.h`` / ``asm/uaccess.h``
275
276**[SLEEPS]**
277
278:c:func:`put_user()` and :c:func:`get_user()` are used to get
279and put single values (such as an int, char, or long) from and to
280userspace. A pointer into userspace should never be simply dereferenced:
281data should be copied using these routines. Both return ``-EFAULT`` or
2820.
283
284:c:func:`copy_to_user()` and :c:func:`copy_from_user()` are
285more general: they copy an arbitrary amount of data to and from
286userspace.
287
288.. warning::
289
290    Unlike :c:func:`put_user()` and :c:func:`get_user()`, they
291    return the amount of uncopied data (ie. 0 still means success).
292
293[Yes, this moronic interface makes me cringe. The flamewar comes up
294every year or so. --RR.]
295
296The functions may sleep implicitly. This should never be called outside
297user context (it makes no sense), with interrupts disabled, or a
298spinlock held.
299
300:c:func:`kmalloc()`/:c:func:`kfree()`
301-------------------------------------
302
303Defined in ``include/linux/slab.h``
304
305**[MAY SLEEP: SEE BELOW]**
306
307These routines are used to dynamically request pointer-aligned chunks of
308memory, like malloc and free do in userspace, but
309:c:func:`kmalloc()` takes an extra flag word. Important values:
310
311``GFP_KERNEL``
312    May sleep and swap to free memory. Only allowed in user context, but
313    is the most reliable way to allocate memory.
314
315``GFP_ATOMIC``
316    Don't sleep. Less reliable than ``GFP_KERNEL``, but may be called
317    from interrupt context. You should **really** have a good
318    out-of-memory error-handling strategy.
319
320``GFP_DMA``
321    Allocate ISA DMA lower than 16MB. If you don't know what that is you
322    don't need it. Very unreliable.
323
324If you see a sleeping function called from invalid context warning
325message, then maybe you called a sleeping allocation function from
326interrupt context without ``GFP_ATOMIC``. You should really fix that.
327Run, don't walk.
328
329If you are allocating at least ``PAGE_SIZE`` (``asm/page.h`` or
330``asm/page_types.h``) bytes, consider using :c:func:`__get_free_pages()`
331(``include/linux/gfp.h``). It takes an order argument (0 for page sized,
3321 for double page, 2 for four pages etc.) and the same memory priority
333flag word as above.
334
335If you are allocating more than a page worth of bytes you can use
336:c:func:`vmalloc()`. It'll allocate virtual memory in the kernel
337map. This block is not contiguous in physical memory, but the MMU makes
338it look like it is for you (so it'll only look contiguous to the CPUs,
339not to external device drivers). If you really need large physically
340contiguous memory for some weird device, you have a problem: it is
341poorly supported in Linux because after some time memory fragmentation
342in a running kernel makes it hard. The best way is to allocate the block
343early in the boot process via the :c:func:`alloc_bootmem()`
344routine.
345
346Before inventing your own cache of often-used objects consider using a
347slab cache in ``include/linux/slab.h``
348
349:c:macro:`current`
350------------------
351
352Defined in ``include/asm/current.h``
353
354This global variable (really a macro) contains a pointer to the current
355task structure, so is only valid in user context. For example, when a
356process makes a system call, this will point to the task structure of
357the calling process. It is **not NULL** in interrupt context.
358
359:c:func:`mdelay()`/:c:func:`udelay()`
360-------------------------------------
361
362Defined in ``include/asm/delay.h`` / ``include/linux/delay.h``
363
364The :c:func:`udelay()` and :c:func:`ndelay()` functions can be
365used for small pauses. Do not use large values with them as you risk
366overflow - the helper function :c:func:`mdelay()` is useful here, or
367consider :c:func:`msleep()`.
368
369:c:func:`cpu_to_be32()`/:c:func:`be32_to_cpu()`/:c:func:`cpu_to_le32()`/:c:func:`le32_to_cpu()`
370-----------------------------------------------------------------------------------------------
371
372Defined in ``include/asm/byteorder.h``
373
374The :c:func:`cpu_to_be32()` family (where the "32" can be replaced
375by 64 or 16, and the "be" can be replaced by "le") are the general way
376to do endian conversions in the kernel: they return the converted value.
377All variations supply the reverse as well:
378:c:func:`be32_to_cpu()`, etc.
379
380There are two major variations of these functions: the pointer
381variation, such as :c:func:`cpu_to_be32p()`, which take a pointer
382to the given type, and return the converted value. The other variation
383is the "in-situ" family, such as :c:func:`cpu_to_be32s()`, which
384convert value referred to by the pointer, and return void.
385
386:c:func:`local_irq_save()`/:c:func:`local_irq_restore()`
387--------------------------------------------------------
388
389Defined in ``include/linux/irqflags.h``
390
391These routines disable hard interrupts on the local CPU, and restore
392them. They are reentrant; saving the previous state in their one
393``unsigned long flags`` argument. If you know that interrupts are
394enabled, you can simply use :c:func:`local_irq_disable()` and
395:c:func:`local_irq_enable()`.
396
397.. _local_bh_disable:
398
399:c:func:`local_bh_disable()`/:c:func:`local_bh_enable()`
400--------------------------------------------------------
401
402Defined in ``include/linux/bottom_half.h``
403
404
405These routines disable soft interrupts on the local CPU, and restore
406them. They are reentrant; if soft interrupts were disabled before, they
407will still be disabled after this pair of functions has been called.
408They prevent softirqs and tasklets from running on the current CPU.
409
410:c:func:`smp_processor_id()`
411----------------------------
412
413Defined in ``include/linux/smp.h``
414
415:c:func:`get_cpu()` disables preemption (so you won't suddenly get
416moved to another CPU) and returns the current processor number, between
4170 and ``NR_CPUS``. Note that the CPU numbers are not necessarily
418continuous. You return it again with :c:func:`put_cpu()` when you
419are done.
420
421If you know you cannot be preempted by another task (ie. you are in
422interrupt context, or have preemption disabled) you can use
423smp_processor_id().
424
425``__init``/``__exit``/``__initdata``
426------------------------------------
427
428Defined in  ``include/linux/init.h``
429
430After boot, the kernel frees up a special section; functions marked with
431``__init`` and data structures marked with ``__initdata`` are dropped
432after boot is complete: similarly modules discard this memory after
433initialization. ``__exit`` is used to declare a function which is only
434required on exit: the function will be dropped if this file is not
435compiled as a module. See the header file for use. Note that it makes no
436sense for a function marked with ``__init`` to be exported to modules
437with :c:func:`EXPORT_SYMBOL()` or :c:func:`EXPORT_SYMBOL_GPL()`- this
438will break.
439
440:c:func:`__initcall()`/:c:func:`module_init()`
441----------------------------------------------
442
443Defined in  ``include/linux/init.h`` / ``include/linux/module.h``
444
445Many parts of the kernel are well served as a module
446(dynamically-loadable parts of the kernel). Using the
447:c:func:`module_init()` and :c:func:`module_exit()` macros it
448is easy to write code without #ifdefs which can operate both as a module
449or built into the kernel.
450
451The :c:func:`module_init()` macro defines which function is to be
452called at module insertion time (if the file is compiled as a module),
453or at boot time: if the file is not compiled as a module the
454:c:func:`module_init()` macro becomes equivalent to
455:c:func:`__initcall()`, which through linker magic ensures that
456the function is called on boot.
457
458The function can return a negative error number to cause module loading
459to fail (unfortunately, this has no effect if the module is compiled
460into the kernel). This function is called in user context with
461interrupts enabled, so it can sleep.
462
463:c:func:`module_exit()`
464-----------------------
465
466
467Defined in  ``include/linux/module.h``
468
469This macro defines the function to be called at module removal time (or
470never, in the case of the file compiled into the kernel). It will only
471be called if the module usage count has reached zero. This function can
472also sleep, but cannot fail: everything must be cleaned up by the time
473it returns.
474
475Note that this macro is optional: if it is not present, your module will
476not be removable (except for 'rmmod -f').
477
478:c:func:`try_module_get()`/:c:func:`module_put()`
479-------------------------------------------------
480
481Defined in ``include/linux/module.h``
482
483These manipulate the module usage count, to protect against removal (a
484module also can't be removed if another module uses one of its exported
485symbols: see below). Before calling into module code, you should call
486:c:func:`try_module_get()` on that module: if it fails, then the
487module is being removed and you should act as if it wasn't there.
488Otherwise, you can safely enter the module, and call
489:c:func:`module_put()` when you're finished.
490
491Most registerable structures have an owner field, such as in the
492:c:type:`struct file_operations <file_operations>` structure.
493Set this field to the macro ``THIS_MODULE``.
494
495Wait Queues ``include/linux/wait.h``
496====================================
497
498**[SLEEPS]**
499
500A wait queue is used to wait for someone to wake you up when a certain
501condition is true. They must be used carefully to ensure there is no
502race condition. You declare a :c:type:`wait_queue_head_t`, and then processes
503which want to wait for that condition declare a :c:type:`wait_queue_entry_t`
504referring to themselves, and place that in the queue.
505
506Declaring
507---------
508
509You declare a ``wait_queue_head_t`` using the
510:c:func:`DECLARE_WAIT_QUEUE_HEAD()` macro, or using the
511:c:func:`init_waitqueue_head()` routine in your initialization
512code.
513
514Queuing
515-------
516
517Placing yourself in the waitqueue is fairly complex, because you must
518put yourself in the queue before checking the condition. There is a
519macro to do this: :c:func:`wait_event_interruptible()`
520(``include/linux/wait.h``) The first argument is the wait queue head, and
521the second is an expression which is evaluated; the macro returns 0 when
522this expression is true, or ``-ERESTARTSYS`` if a signal is received. The
523:c:func:`wait_event()` version ignores signals.
524
525Waking Up Queued Tasks
526----------------------
527
528Call :c:func:`wake_up()` (``include/linux/wait.h``), which will wake
529up every process in the queue. The exception is if one has
530``TASK_EXCLUSIVE`` set, in which case the remainder of the queue will
531not be woken. There are other variants of this basic function available
532in the same header.
533
534Atomic Operations
535=================
536
537Certain operations are guaranteed atomic on all platforms. The first
538class of operations work on :c:type:`atomic_t` (``include/asm/atomic.h``);
539this contains a signed integer (at least 32 bits long), and you must use
540these functions to manipulate or read :c:type:`atomic_t` variables.
541:c:func:`atomic_read()` and :c:func:`atomic_set()` get and set
542the counter, :c:func:`atomic_add()`, :c:func:`atomic_sub()`,
543:c:func:`atomic_inc()`, :c:func:`atomic_dec()`, and
544:c:func:`atomic_dec_and_test()` (returns true if it was
545decremented to zero).
546
547Yes. It returns true (i.e. != 0) if the atomic variable is zero.
548
549Note that these functions are slower than normal arithmetic, and so
550should not be used unnecessarily.
551
552The second class of atomic operations is atomic bit operations on an
553``unsigned long``, defined in ``include/linux/bitops.h``. These
554operations generally take a pointer to the bit pattern, and a bit
555number: 0 is the least significant bit. :c:func:`set_bit()`,
556:c:func:`clear_bit()` and :c:func:`change_bit()` set, clear,
557and flip the given bit. :c:func:`test_and_set_bit()`,
558:c:func:`test_and_clear_bit()` and
559:c:func:`test_and_change_bit()` do the same thing, except return
560true if the bit was previously set; these are particularly useful for
561atomically setting flags.
562
563It is possible to call these operations with bit indices greater than
564``BITS_PER_LONG``. The resulting behavior is strange on big-endian
565platforms though so it is a good idea not to do this.
566
567Symbols
568=======
569
570Within the kernel proper, the normal linking rules apply (ie. unless a
571symbol is declared to be file scope with the ``static`` keyword, it can
572be used anywhere in the kernel). However, for modules, a special
573exported symbol table is kept which limits the entry points to the
574kernel proper. Modules can also export symbols.
575
576:c:func:`EXPORT_SYMBOL()`
577-------------------------
578
579Defined in ``include/linux/export.h``
580
581This is the classic method of exporting a symbol: dynamically loaded
582modules will be able to use the symbol as normal.
583
584:c:func:`EXPORT_SYMBOL_GPL()`
585-----------------------------
586
587Defined in ``include/linux/export.h``
588
589Similar to :c:func:`EXPORT_SYMBOL()` except that the symbols
590exported by :c:func:`EXPORT_SYMBOL_GPL()` can only be seen by
591modules with a :c:func:`MODULE_LICENSE()` that specifies a GPL
592compatible license. It implies that the function is considered an
593internal implementation issue, and not really an interface. Some
594maintainers and developers may however require EXPORT_SYMBOL_GPL()
595when adding any new APIs or functionality.
596
597:c:func:`EXPORT_SYMBOL_NS()`
598----------------------------
599
600Defined in ``include/linux/export.h``
601
602This is the variant of `EXPORT_SYMBOL()` that allows specifying a symbol
603namespace. Symbol Namespaces are documented in
604Documentation/core-api/symbol-namespaces.rst
605
606:c:func:`EXPORT_SYMBOL_NS_GPL()`
607--------------------------------
608
609Defined in ``include/linux/export.h``
610
611This is the variant of `EXPORT_SYMBOL_GPL()` that allows specifying a symbol
612namespace. Symbol Namespaces are documented in
613Documentation/core-api/symbol-namespaces.rst
614
615Routines and Conventions
616========================
617
618Double-linked lists ``include/linux/list.h``
619--------------------------------------------
620
621There used to be three sets of linked-list routines in the kernel
622headers, but this one is the winner. If you don't have some particular
623pressing need for a single list, it's a good choice.
624
625In particular, :c:func:`list_for_each_entry()` is useful.
626
627Return Conventions
628------------------
629
630For code called in user context, it's very common to defy C convention,
631and return 0 for success, and a negative error number (eg. ``-EFAULT``) for
632failure. This can be unintuitive at first, but it's fairly widespread in
633the kernel.
634
635Using :c:func:`ERR_PTR()` (``include/linux/err.h``) to encode a
636negative error number into a pointer, and :c:func:`IS_ERR()` and
637:c:func:`PTR_ERR()` to get it back out again: avoids a separate
638pointer parameter for the error number. Icky, but in a good way.
639
640Breaking Compilation
641--------------------
642
643Linus and the other developers sometimes change function or structure
644names in development kernels; this is not done just to keep everyone on
645their toes: it reflects a fundamental change (eg. can no longer be
646called with interrupts on, or does extra checks, or doesn't do checks
647which were caught before). Usually this is accompanied by a fairly
648complete note to the linux-kernel mailing list; search the archive.
649Simply doing a global replace on the file usually makes things **worse**.
650
651Initializing structure members
652------------------------------
653
654The preferred method of initializing structures is to use designated
655initialisers, as defined by ISO C99, eg::
656
657    static struct block_device_operations opt_fops = {
658            .open               = opt_open,
659            .release            = opt_release,
660            .ioctl              = opt_ioctl,
661            .check_media_change = opt_media_change,
662    };
663
664
665This makes it easy to grep for, and makes it clear which structure
666fields are set. You should do this because it looks cool.
667
668GNU Extensions
669--------------
670
671GNU Extensions are explicitly allowed in the Linux kernel. Note that
672some of the more complex ones are not very well supported, due to lack
673of general use, but the following are considered standard (see the GCC
674info page section "C Extensions" for more details - Yes, really the info
675page, the man page is only a short summary of the stuff in info).
676
677-  Inline functions
678
679-  Statement expressions (ie. the ({ and }) constructs).
680
681-  Declaring attributes of a function / variable / type
682   (__attribute__)
683
684-  typeof
685
686-  Zero length arrays
687
688-  Macro varargs
689
690-  Arithmetic on void pointers
691
692-  Non-Constant initializers
693
694-  Assembler Instructions (not outside arch/ and include/asm/)
695
696-  Function names as strings (__func__).
697
698-  __builtin_constant_p()
699
700Be wary when using long long in the kernel, the code gcc generates for
701it is horrible and worse: division and multiplication does not work on
702i386 because the GCC runtime functions for it are missing from the
703kernel environment.
704
705C++
706---
707
708Using C++ in the kernel is usually a bad idea, because the kernel does
709not provide the necessary runtime environment and the include files are
710not tested for it. It is still possible, but not recommended. If you
711really want to do this, forget about exceptions at least.
712
713#if
714---
715
716It is generally considered cleaner to use macros in header files (or at
717the top of .c files) to abstract away functions rather than using \`#if'
718pre-processor statements throughout the source code.
719
720Putting Your Stuff in the Kernel
721================================
722
723In order to get your stuff into shape for official inclusion, or even to
724make a neat patch, there's administrative work to be done:
725
726-  Figure out whose pond you've been pissing in. Look at the top of the
727   source files, inside the ``MAINTAINERS`` file, and last of all in the
728   ``CREDITS`` file. You should coordinate with this person to make sure
729   you're not duplicating effort, or trying something that's already
730   been rejected.
731
732   Make sure you put your name and EMail address at the top of any files
733   you create or mangle significantly. This is the first place people
734   will look when they find a bug, or when **they** want to make a change.
735
736-  Usually you want a configuration option for your kernel hack. Edit
737   ``Kconfig`` in the appropriate directory. The Config language is
738   simple to use by cut and paste, and there's complete documentation in
739   ``Documentation/kbuild/kconfig-language.rst``.
740
741   In your description of the option, make sure you address both the
742   expert user and the user who knows nothing about your feature.
743   Mention incompatibilities and issues here. **Definitely** end your
744   description with “if in doubt, say N” (or, occasionally, \`Y'); this
745   is for people who have no idea what you are talking about.
746
747-  Edit the ``Makefile``: the CONFIG variables are exported here so you
748   can usually just add a "obj-$(CONFIG_xxx) += xxx.o" line. The syntax
749   is documented in ``Documentation/kbuild/makefiles.rst``.
750
751-  Put yourself in ``CREDITS`` if you've done something noteworthy,
752   usually beyond a single file (your name should be at the top of the
753   source files anyway). ``MAINTAINERS`` means you want to be consulted
754   when changes are made to a subsystem, and hear about bugs; it implies
755   a more-than-passing commitment to some part of the code.
756
757-  Finally, don't forget to read
758   ``Documentation/process/submitting-patches.rst`` and possibly
759   ``Documentation/process/submitting-drivers.rst``.
760
761Kernel Cantrips
762===============
763
764Some favorites from browsing the source. Feel free to add to this list.
765
766``arch/x86/include/asm/delay.h``::
767
768    #define ndelay(n) (__builtin_constant_p(n) ? \
769            ((n) > 20000 ? __bad_ndelay() : __const_udelay((n) * 5ul)) : \
770            __ndelay(n))
771
772
773``include/linux/fs.h``::
774
775    /*
776     * Kernel pointers have redundant information, so we can use a
777     * scheme where we can return either an error code or a dentry
778     * pointer with the same return value.
779     *
780     * This should be a per-architecture thing, to allow different
781     * error and pointer decisions.
782     */
783     #define ERR_PTR(err)    ((void *)((long)(err)))
784     #define PTR_ERR(ptr)    ((long)(ptr))
785     #define IS_ERR(ptr)     ((unsigned long)(ptr) > (unsigned long)(-1000))
786
787``arch/x86/include/asm/uaccess_32.h:``::
788
789    #define copy_to_user(to,from,n)                         \
790            (__builtin_constant_p(n) ?                      \
791             __constant_copy_to_user((to),(from),(n)) :     \
792             __generic_copy_to_user((to),(from),(n)))
793
794
795``arch/sparc/kernel/head.S:``::
796
797    /*
798     * Sun people can't spell worth damn. "compatability" indeed.
799     * At least we *know* we can't spell, and use a spell-checker.
800     */
801
802    /* Uh, actually Linus it is I who cannot spell. Too much murky
803     * Sparc assembly will do this to ya.
804     */
805    C_LABEL(cputypvar):
806            .asciz "compatibility"
807
808    /* Tested on SS-5, SS-10. Probably someone at Sun applied a spell-checker. */
809            .align 4
810    C_LABEL(cputypvar_sun4m):
811            .asciz "compatible"
812
813
814``arch/sparc/lib/checksum.S:``::
815
816            /* Sun, you just can't beat me, you just can't.  Stop trying,
817             * give up.  I'm serious, I am going to kick the living shit
818             * out of you, game over, lights out.
819             */
820
821
822Thanks
823======
824
825Thanks to Andi Kleen for the idea, answering my questions, fixing my
826mistakes, filling content, etc. Philipp Rumpf for more spelling and
827clarity fixes, and some excellent non-obvious points. Werner Almesberger
828for giving me a great summary of :c:func:`disable_irq()`, and Jes
829Sorensen and Andrea Arcangeli added caveats. Michael Elizabeth Chastain
830for checking and adding to the Configure section. Telsa Gwynne for
831teaching me DocBook.
832