.. _kernel_hacking_hack: ============================================Unreliable Guide To Hacking The Linux Kernel============================================ :Author: Rusty Russell Introduction============ Welcome, gentle reader, to Rusty's Remarkably Unreliable Guide to LinuxKernel Hacking. This document describes the common routines and generalrequirements for kernel code: its goal is to serve as a primer for Linuxkernel development for experienced C programmers. I avoid implementationdetails: that's what the code is for, and I ignore whole tracts ofuseful routines. Before you read this, please understand that I never wanted to writethis document, being grossly under-qualified, but I always wanted toread it, and this was the only way. I hope it will grow into acompendium of best practice, common starting points and randominformation. The Players=========== At any time each of the CPUs in a system can be: -  not associated with any process, serving a hardware interrupt; -  not associated with any process, serving a softirq or tasklet; -  running in kernel space, associated with a process (user context); -  running a process in user space. There is an ordering between these. The bottom two can preempt eachother, but above that is a strict hierarchy: each can only be preemptedby the ones above it. For example, while a softirq is running on a CPU,no other softirq will preempt it, but a hardware interrupt can. However,any other CPUs in the system execute independently. We'll see a number of ways that the user context can block interrupts,to become truly non-preemptable. User Context------------ User context is when you are coming in from a system call or other trap:like userspace, you can be preempted by more important tasks and byinterrupts. You can sleep, by calling :c:func:`schedule()`. .. note::     You are always in user context on module load and unload, and on    operations on the block device layer. In user context, the ``current`` pointer (indicating the task we arecurrently executing) is valid, and :c:func:`in_interrupt()`(``include/linux/preempt.h``) is false. .. warning::     Beware that if you have preemption or softirqs disabled (see below),    :c:func:`in_interrupt()` will return a false positive. Hardware Interrupts (Hard IRQs)------------------------------- Timer ticks, network cards and keyboard are examples of real hardwarewhich produce interrupts at any time. The kernel runs interrupthandlers, which services the hardware. The kernel guarantees that thishandler is never re-entered: if the same interrupt arrives, it is queued(or dropped). Because it disables interrupts, this handler has to befast: frequently it simply acknowledges the interrupt, marks a 'softwareinterrupt' for execution and exits. You can tell you are in a hardware interrupt, because in_hardirq() returnstrue. .. warning::     Beware that this will return a false positive if interrupts are    disabled (see below). Software Interrupt Context: Softirqs and Tasklets------------------------------------------------- Whenever a system call is about to return to userspace, or a hardwareinterrupt handler exits, any 'software interrupts' which are markedpending (usually by hardware interrupts) are run (``kernel/softirq.c``). Much of the real interrupt handling work is done here. Early in thetransition to SMP, there were only 'bottom halves' (BHs), which didn'ttake advantage of multiple CPUs. Shortly after we switched from wind-upcomputers made of match-sticks and snot, we abandoned this limitationand switched to 'softirqs'. ``include/linux/interrupt.h`` lists the different softirqs. A veryimportant softirq is the timer softirq (``include/linux/timer.h``): youcan register to have it call functions for you in a given length oftime. Softirqs are often a pain to deal with, since the same softirq will runsimultaneously on more than one CPU. For this reason, tasklets(``include/linux/interrupt.h``) are more often used: they aredynamically-registrable (meaning you can have as many as you want), andthey also guarantee that any tasklet will only run on one CPU at anytime, although different tasklets can run simultaneously. .. warning::     The name 'tasklet' is misleading: they have nothing to do with    'tasks'. You can tell you are in a softirq (or tasklet) using the:c:func:`in_softirq()` macro (``include/linux/preempt.h``). .. warning::     Beware that this will return a false positive if a    :ref:`bottom half lock <local_bh_disable>` is held. Some Basic Rules================ No memory protection    If you corrupt memory, whether in user context or interrupt context,    the whole machine will crash. Are you sure you can't do what you    want in userspace? No floating point or MMX    The FPU context is not saved; even in user context the FPU state    probably won't correspond with the current process: you would mess    with some user process' FPU state. If you really want to do this,    you would have to explicitly save/restore the full FPU state (and    avoid context switches). It is generally a bad idea; use fixed point    arithmetic first. A rigid stack limit    Depending on configuration options the kernel stack is about 3K to    6K for most 32-bit architectures: it's about 14K on most 64-bit    archs, and often shared with interrupts so you can't use it all.    Avoid deep recursion and huge local arrays on the stack (allocate    them dynamically instead). The Linux kernel is portable    Let's keep it that way. Your code should be 64-bit clean, and    endian-independent. You should also minimize CPU specific stuff,    e.g. inline assembly should be cleanly encapsulated and minimized to    ease porting. Generally it should be restricted to the    architecture-dependent part of the kernel tree. ioctls: Not writing a new system call===================================== A system call generally looks like this::     asmlinkage long sys_mycall(int arg)    {            return 0;    }  First, in most cases you don't want to create a new system call. Youcreate a character device and implement an appropriate ioctl for it.This is much more flexible than system calls, doesn't have to be enteredin every architecture's ``include/asm/unistd.h`` and``arch/kernel/entry.S`` file, and is much more likely to be accepted byLinus. If all your routine does is read or write some parameter, considerimplementing a :c:func:`sysfs()` interface instead. Inside the ioctl you're in user context to a process. When a erroroccurs you return a negated errno (see``include/uapi/asm-generic/errno-base.h``,``include/uapi/asm-generic/errno.h`` and ``include/linux/errno.h``),otherwise you return 0. After you slept you should check if a signal occurred: the Unix/Linuxway of handling signals is to temporarily exit the system call with the``-ERESTARTSYS`` error. The system call entry code will switch back touser context, process the signal handler and then your system call willbe restarted (unless the user disabled that). So you should be preparedto process the restart, e.g. if you're in the middle of manipulatingsome data structure. ::     if (signal_pending(current))            return -ERESTARTSYS;  If you're doing longer computations: first think userspace. If you**really** want to do it in kernel you should regularly check if you needto give up the CPU (remember there is cooperative multitasking per CPU).Idiom::     cond_resched(); /* Will sleep */  A short note on interface design: the UNIX system call motto is "Providemechanism not policy". Recipes for Deadlock==================== You cannot call any routines which may sleep, unless: -  You are in user context. -  You do not own any spinlocks. -  You have interrupts enabled (actually, Andi Kleen says that the   scheduling code will enable them for you, but that's probably not   what you wanted). Note that some functions may sleep implicitly: common ones are the userspace access functions (\*_user) and memory allocation functionswithout ``GFP_ATOMIC``. You should always compile your kernel ``CONFIG_DEBUG_ATOMIC_SLEEP`` on,and it will warn you if you break these rules. If you **do** break therules, you will eventually lock up your box. Really. Common Routines=============== :c:func:`printk()`------------------ Defined in ``include/linux/printk.h`` :c:func:`printk()` feeds kernel messages to the console, dmesg, andthe syslog daemon. It is useful for debugging and reporting errors, andcan be used inside interrupt context, but use with caution: a machinewhich has its console flooded with printk messages is unusable. It usesa format string mostly compatible with ANSI C printf, and C stringconcatenation to give it a first "priority" argument::     printk(KERN_INFO "i = %u\n", i);  See ``include/linux/kern_levels.h``; for other ``KERN_`` values; these areinterpreted by syslog as the level. Special case: for printing an IPaddress use::     __be32 ipaddress;    printk(KERN_INFO "my ip: %pI4\n", &ipaddress);  :c:func:`printk()` internally uses a 1K buffer and does not catchoverruns. Make sure that will be enough. .. note::     You will know when you are a real kernel hacker when you start    typoing printf as printk in your user programs :) .. note::     Another sidenote: the original Unix Version 6 sources had a comment    on top of its printf function: "Printf should not be used for    chit-chat". You should follow that advice. :c:func:`copy_to_user()` / :c:func:`copy_from_user()` / :c:func:`get_user()` / :c:func:`put_user()`--------------------------------------------------------------------------------------------------- Defined in ``include/linux/uaccess.h`` / ``asm/uaccess.h`` **[SLEEPS]** :c:func:`put_user()` and :c:func:`get_user()` are used to getand put single values (such as an int, char, or long) from and touserspace. A pointer into userspace should never be simply dereferenced:data should be copied using these routines. Both return ``-EFAULT`` or0. :c:func:`copy_to_user()` and :c:func:`copy_from_user()` aremore general: they copy an arbitrary amount of data to and fromuserspace. .. warning::     Unlike :c:func:`put_user()` and :c:func:`get_user()`, they    return the amount of uncopied data (ie. 0 still means success). [Yes, this objectionable interface makes me cringe. The flamewar comesup every year or so. --RR.] The functions may sleep implicitly. This should never be called outsideuser context (it makes no sense), with interrupts disabled, or aspinlock held. :c:func:`kmalloc()`/:c:func:`kfree()`------------------------------------- Defined in ``include/linux/slab.h`` **[MAY SLEEP: SEE BELOW]** These routines are used to dynamically request pointer-aligned chunks ofmemory, like malloc and free do in userspace, but:c:func:`kmalloc()` takes an extra flag word. Important values: ``GFP_KERNEL``    May sleep and swap to free memory. Only allowed in user context, but    is the most reliable way to allocate memory. ``GFP_ATOMIC``    Don't sleep. Less reliable than ``GFP_KERNEL``, but may be called    from interrupt context. You should **really** have a good    out-of-memory error-handling strategy. ``GFP_DMA``    Allocate ISA DMA lower than 16MB. If you don't know what that is you    don't need it. Very unreliable. If you see a sleeping function called from invalid context warningmessage, then maybe you called a sleeping allocation function frominterrupt context without ``GFP_ATOMIC``. You should really fix that.Run, don't walk. If you are allocating at least ``PAGE_SIZE`` (``asm/page.h`` or``asm/page_types.h``) bytes, consider using :c:func:`__get_free_pages()`(``include/linux/gfp.h``). It takes an order argument (0 for page sized,1 for double page, 2 for four pages etc.) and the same memory priorityflag word as above. If you are allocating more than a page worth of bytes you can use:c:func:`vmalloc()`. It'll allocate virtual memory in the kernelmap. This block is not contiguous in physical memory, but the MMU makesit look like it is for you (so it'll only look contiguous to the CPUs,not to external device drivers). If you really need large physicallycontiguous memory for some weird device, you have a problem: it ispoorly supported in Linux because after some time memory fragmentationin a running kernel makes it hard. The best way is to allocate the blockearly in the boot process via the :c:func:`alloc_bootmem()`routine. Before inventing your own cache of often-used objects consider using aslab cache in ``include/linux/slab.h`` :c:macro:`current`------------------ Defined in ``include/asm/current.h`` This global variable (really a macro) contains a pointer to the currenttask structure, so is only valid in user context. For example, when aprocess makes a system call, this will point to the task structure ofthe calling process. It is **not NULL** in interrupt context. :c:func:`mdelay()`/:c:func:`udelay()`------------------------------------- Defined in ``include/asm/delay.h`` / ``include/linux/delay.h`` The :c:func:`udelay()` and :c:func:`ndelay()` functions can beused for small pauses. Do not use large values with them as you riskoverflow - the helper function :c:func:`mdelay()` is useful here, orconsider :c:func:`msleep()`. :c:func:`cpu_to_be32()`/:c:func:`be32_to_cpu()`/:c:func:`cpu_to_le32()`/:c:func:`le32_to_cpu()`----------------------------------------------------------------------------------------------- Defined in ``include/asm/byteorder.h`` The :c:func:`cpu_to_be32()` family (where the "32" can be replacedby 64 or 16, and the "be" can be replaced by "le") are the general wayto do endian conversions in the kernel: they return the converted value.All variations supply the reverse as well::c:func:`be32_to_cpu()`, etc. There are two major variations of these functions: the pointervariation, such as :c:func:`cpu_to_be32p()`, which take a pointerto the given type, and return the converted value. The other variationis the "in-situ" family, such as :c:func:`cpu_to_be32s()`, whichconvert value referred to by the pointer, and return void. :c:func:`local_irq_save()`/:c:func:`local_irq_restore()`-------------------------------------------------------- Defined in ``include/linux/irqflags.h`` These routines disable hard interrupts on the local CPU, and restorethem. They are reentrant; saving the previous state in their one``unsigned long flags`` argument. If you know that interrupts areenabled, you can simply use :c:func:`local_irq_disable()` and:c:func:`local_irq_enable()`. .. _local_bh_disable: :c:func:`local_bh_disable()`/:c:func:`local_bh_enable()`-------------------------------------------------------- Defined in ``include/linux/bottom_half.h``  These routines disable soft interrupts on the local CPU, and restorethem. They are reentrant; if soft interrupts were disabled before, theywill still be disabled after this pair of functions has been called.They prevent softirqs and tasklets from running on the current CPU. :c:func:`smp_processor_id()`---------------------------- Defined in ``include/linux/smp.h`` :c:func:`get_cpu()` disables preemption (so you won't suddenly getmoved to another CPU) and returns the current processor number, between0 and ``NR_CPUS``. Note that the CPU numbers are not necessarilycontinuous. You return it again with :c:func:`put_cpu()` when youare done. If you know you cannot be preempted by another task (ie. you are ininterrupt context, or have preemption disabled) you can usesmp_processor_id(). ``__init``/``__exit``/``__initdata``------------------------------------ Defined in  ``include/linux/init.h`` After boot, the kernel frees up a special section; functions marked with``__init`` and data structures marked with ``__initdata`` are droppedafter boot is complete: similarly modules discard this memory afterinitialization. ``__exit`` is used to declare a function which is onlyrequired on exit: the function will be dropped if this file is notcompiled as a module. See the header file for use. Note that it makes nosense for a function marked with ``__init`` to be exported to moduleswith :c:func:`EXPORT_SYMBOL()` or :c:func:`EXPORT_SYMBOL_GPL()`- thiswill break. :c:func:`__initcall()`/:c:func:`module_init()`---------------------------------------------- Defined in  ``include/linux/init.h`` / ``include/linux/module.h`` Many parts of the kernel are well served as a module(dynamically-loadable parts of the kernel). Using the:c:func:`module_init()` and :c:func:`module_exit()` macros itis easy to write code without #ifdefs which can operate both as a moduleor built into the kernel. The :c:func:`module_init()` macro defines which function is to becalled at module insertion time (if the file is compiled as a module),or at boot time: if the file is not compiled as a module the:c:func:`module_init()` macro becomes equivalent to:c:func:`__initcall()`, which through linker magic ensures thatthe function is called on boot. The function can return a negative error number to cause module loadingto fail (unfortunately, this has no effect if the module is compiledinto the kernel). This function is called in user context withinterrupts enabled, so it can sleep. :c:func:`module_exit()`-----------------------  Defined in  ``include/linux/module.h`` This macro defines the function to be called at module removal time (ornever, in the case of the file compiled into the kernel). It will onlybe called if the module usage count has reached zero. This function canalso sleep, but cannot fail: everything must be cleaned up by the timeit returns. Note that this macro is optional: if it is not present, your module willnot be removable (except for 'rmmod -f'). :c:func:`try_module_get()`/:c:func:`module_put()`------------------------------------------------- Defined in ``include/linux/module.h`` These manipulate the module usage count, to protect against removal (amodule also can't be removed if another module uses one of its exportedsymbols: see below). Before calling into module code, you should call:c:func:`try_module_get()` on that module: if it fails, then themodule is being removed and you should act as if it wasn't there.Otherwise, you can safely enter the module, and call:c:func:`module_put()` when you're finished. Most registerable structures have an owner field, such as in the:c:type:`struct file_operations <file_operations>` structure.Set this field to the macro ``THIS_MODULE``. Wait Queues ``include/linux/wait.h``==================================== **[SLEEPS]** A wait queue is used to wait for someone to wake you up when a certaincondition is true. They must be used carefully to ensure there is no

Lines 1–500 of 830. Load more