From c4fcd7cabb83fcac1e83e17845b1fede5401aa25 Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 08:03:27 -0300 Subject: [PATCH 01/90] docs-rst: convert kernel-hacking to ReST Use pandoc to convert documentation to ReST by calling Documentation/sphinx/tmplcvt script. - Manually adjusted to use ..note and ..warning - Minor fixes for it to be parsed without errors - Use **bold** for emphasis. Signed-off-by: Mauro Carvalho Chehab --- Documentation/DocBook/Makefile | 2 +- Documentation/DocBook/kernel-hacking.tmpl | 1312 --------------------- Documentation/conf.py | 2 + Documentation/index.rst | 1 + Documentation/kernel-hacking/conf.py | 10 + Documentation/kernel-hacking/hacking.rst | 794 +++++++++++++ Documentation/kernel-hacking/index.rst | 8 + 7 files changed, 816 insertions(+), 1313 deletions(-) delete mode 100644 Documentation/DocBook/kernel-hacking.tmpl create mode 100644 Documentation/kernel-hacking/conf.py create mode 100644 Documentation/kernel-hacking/hacking.rst create mode 100644 Documentation/kernel-hacking/index.rst diff --git a/Documentation/DocBook/Makefile b/Documentation/DocBook/Makefile index 85916f13d330..7d7482b5ad92 100644 --- a/Documentation/DocBook/Makefile +++ b/Documentation/DocBook/Makefile @@ -7,7 +7,7 @@ # list of DOCBOOKS. DOCBOOKS := z8530book.xml \ - kernel-hacking.xml kernel-locking.xml \ + kernel-locking.xml \ networking.xml \ filesystems.xml lsm.xml kgdb.xml \ libata.xml mtdnand.xml librs.xml rapidio.xml \ diff --git a/Documentation/DocBook/kernel-hacking.tmpl b/Documentation/DocBook/kernel-hacking.tmpl deleted file mode 100644 index da5c087462b1..000000000000 --- a/Documentation/DocBook/kernel-hacking.tmpl +++ /dev/null @@ -1,1312 +0,0 @@ - - - - - - Unreliable Guide To Hacking The Linux Kernel - - - - Rusty - Russell - -
- rusty@rustcorp.com.au -
- - - - - - 2005 - Rusty Russell - - - - - This documentation is free software; you can redistribute - it and/or modify it under the terms of the GNU General Public - License as published by the Free Software Foundation; either - version 2 of the License, or (at your option) any later - version. - - - - This program is distributed in the hope that it will be - useful, but WITHOUT ANY WARRANTY; without even the implied - warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. - See the GNU General Public License for more details. - - - - You should have received a copy of the GNU General Public - License along with this program; if not, write to the Free - Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, - MA 02111-1307 USA - - - - For more details see the file COPYING in the source - distribution of Linux. - - - - - This is the first release of this document as part of the kernel tarball. - - - - - - - - Introduction - - Welcome, gentle reader, to Rusty's Remarkably Unreliable Guide to Linux - Kernel Hacking. This document describes the common routines and - general requirements for kernel code: its goal is to serve as a - primer for Linux kernel development for experienced C - programmers. I avoid implementation details: that's what the - code is for, and I ignore whole tracts of useful routines. - - - Before you read this, please understand that I never wanted to - write this document, being grossly under-qualified, but I always - wanted to read it, and this was the only way. I hope it will - grow into a compendium of best practice, common starting points - and random information. - - - - - 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 - each other, but above that is a strict hierarchy: each can only be - preempted by 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 by interrupts. You can sleep, by calling - schedule(). - - - - - 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 are currently executing) is valid, and - in_interrupt() - (include/linux/interrupt.h) is false - . - - - - - Beware that if you have preemption or softirqs disabled - (see below), in_interrupt() will return a - false positive. - - - - - - Hardware Interrupts (Hard IRQs) - - - Timer ticks, network cards and - keyboard are examples of real - hardware which produce interrupts at any time. The kernel runs - interrupt handlers, which services the hardware. The kernel - guarantees that this handler is never re-entered: if the same - interrupt arrives, it is queued (or dropped). Because it - disables interrupts, this handler has to be fast: frequently it - simply acknowledges the interrupt, marks a 'software interrupt' - for execution and exits. - - - - You can tell you are in a hardware interrupt, because - in_irq() returns true. - - - - 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 - hardware interrupt handler exits, any 'software interrupts' - which are marked pending (usually by hardware interrupts) are - run (kernel/softirq.c). - - - - Much of the real interrupt handling work is done here. Early in - the transition to SMP, there were only 'bottom - halves' (BHs), which didn't take advantage of multiple CPUs. Shortly - after we switched from wind-up computers made of match-sticks and snot, - we abandoned this limitation and switched to 'softirqs'. - - - - include/linux/interrupt.h lists the - different softirqs. A very important softirq is the - timer softirq (include/linux/timer.h): you can - register to have it call functions for you in a given length of - time. - - - - Softirqs are often a pain to deal with, since the same softirq - will run simultaneously on more than one CPU. For this reason, - tasklets (include/linux/interrupt.h) are more - often used: they are dynamically-registrable (meaning you can have - as many as you want), and they also guarantee that any tasklet - will only run on one CPU at any time, although different tasklets - can run simultaneously. - - - - The name 'tasklet' is misleading: they have nothing to do with 'tasks', - and probably more to do with some bad vodka Alexey Kuznetsov had at the - time. - - - - - You can tell you are in a softirq (or tasklet) - using the in_softirq() macro - (include/linux/interrupt.h). - - - - Beware that this will return a false positive if a bh lock (see below) - 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. - You create a character device and implement an appropriate ioctl - for it. This is much more flexible than system calls, doesn't have - to be entered in every architecture's - include/asm/unistd.h and - arch/kernel/entry.S file, and is much more - likely to be accepted by Linus. - - - - If all your routine does is read or write some parameter, consider - implementing a sysfs interface instead. - - - - Inside the ioctl you're in user context to a process. When a - error occurs you return a negated errno (see - include/linux/errno.h), - otherwise you return 0. - - - - After you slept you should check if a signal occurred: the - Unix/Linux way of handling signals is to temporarily exit the - system call with the -ERESTARTSYS error. The - system call entry code will switch back to user context, process - the signal handler and then your system call will be restarted - (unless the user disabled that). So you should be prepared to - process the restart, e.g. if you're in the middle of manipulating - some 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 need to 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 - "Provide mechanism 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 user space access functions (*_user) and memory allocation - functions without 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 - the rules, you will eventually lock up your box. - - - - Really. - - - - - Common Routines - - - - <function>printk()</function> - <filename class="headerfile">include/linux/kernel.h</filename> - - - - printk() feeds kernel messages to the - console, dmesg, and the syslog daemon. It is useful for debugging - and reporting errors, and can be used inside interrupt context, - but use with caution: a machine which has its console flooded with - printk messages is unusable. It uses a format string mostly - compatible with ANSI C printf, and C string concatenation to give - it a first "priority" argument: - - - -printk(KERN_INFO "i = %u\n", i); - - - - See include/linux/kernel.h; - for other KERN_ values; these are interpreted by syslog as the - level. Special case: for printing an IP address use - - - -__be32 ipaddress; -printk(KERN_INFO "my ip: %pI4\n", &ipaddress); - - - - printk() internally uses a 1K buffer and does - not catch overruns. Make sure that will be enough. - - - - - You will know when you are a real kernel hacker - when you start typoing printf as printk in your user programs :) - - - - - - - - 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. - - - - - - - <function>copy_[to/from]_user()</function> - / - <function>get_user()</function> - / - <function>put_user()</function> - <filename class="headerfile">include/linux/uaccess.h</filename> - - - - [SLEEPS] - - - - put_user() and get_user() - are used to get and put single values (such as an int, char, or - long) from and to userspace. A pointer into userspace should - never be simply dereferenced: data should be copied using these - routines. Both return -EFAULT or 0. - - - copy_to_user() and - copy_from_user() are more general: they copy - an arbitrary amount of data to and from userspace. - - - Unlike put_user() and - get_user(), they return the amount of - uncopied data (ie. 0 still means - success). - - - [Yes, this moronic interface makes me cringe. The flamewar comes up every year or so. --RR.] - - - The functions may sleep implicitly. This should never be called - outside user context (it makes no sense), with interrupts - disabled, or a spinlock held. - - - - - <function>kmalloc()</function>/<function>kfree()</function> - <filename class="headerfile">include/linux/slab.h</filename> - - - [MAY SLEEP: SEE BELOW] - - - - These routines are used to dynamically request pointer-aligned - chunks of memory, like malloc and free do in userspace, but - 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 warning message, then maybe you called a - sleeping allocation function from interrupt context without - GFP_ATOMIC. You should really fix that. - Run, don't walk. - - - - If you are allocating at least PAGE_SIZE - (include/asm/page.h) bytes, - consider using __get_free_pages() - - (include/linux/mm.h). It - takes an order argument (0 for page sized, 1 for double page, 2 - for four pages etc.) and the same memory priority flag word as - above. - - - - If you are allocating more than a page worth of bytes you can use - vmalloc(). It'll allocate virtual memory in - the kernel map. This block is not contiguous in physical memory, - but the MMU makes it 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 physically contiguous memory - for some weird device, you have a problem: it is poorly supported - in Linux because after some time memory fragmentation in a running - kernel makes it hard. The best way is to allocate the block early - in the boot process via the alloc_bootmem() - routine. - - - - Before inventing your own cache of often-used objects consider - using a slab cache in - include/linux/slab.h - - - - - <function>current</function> - <filename class="headerfile">include/asm/current.h</filename> - - - This global variable (really a macro) contains a pointer to - the current task structure, so is only valid in user context. - For example, when a process makes a system call, this will - point to the task structure of the calling process. It is - not NULL in interrupt context. - - - - - <function>mdelay()</function>/<function>udelay()</function> - <filename class="headerfile">include/asm/delay.h</filename> - <filename class="headerfile">include/linux/delay.h</filename> - - - - The udelay() and ndelay() functions can be used for small pauses. - Do not use large values with them as you risk - overflow - the helper function mdelay() is useful - here, or consider msleep(). - - - - - <function>cpu_to_be32()</function>/<function>be32_to_cpu()</function>/<function>cpu_to_le32()</function>/<function>le32_to_cpu()</function> - <filename class="headerfile">include/asm/byteorder.h</filename> - - - - The cpu_to_be32() family (where the "32" can - be replaced by 64 or 16, and the "be" can be replaced by "le") are - the general way to do endian conversions in the kernel: they - return the converted value. All variations supply the reverse as - well: be32_to_cpu(), etc. - - - - There are two major variations of these functions: the pointer - variation, such as cpu_to_be32p(), which take - a pointer to the given type, and return the converted value. The - other variation is the "in-situ" family, such as - cpu_to_be32s(), which convert value referred - to by the pointer, and return void. - - - - - <function>local_irq_save()</function>/<function>local_irq_restore()</function> - <filename class="headerfile">include/linux/irqflags.h</filename> - - - - These routines disable hard interrupts on the local CPU, and - restore them. They are reentrant; saving the previous state in - their one unsigned long flags argument. If you - know that interrupts are enabled, you can simply use - local_irq_disable() and - local_irq_enable(). - - - - - <function>local_bh_disable()</function>/<function>local_bh_enable()</function> - <filename class="headerfile">include/linux/interrupt.h</filename> - - - These routines disable soft interrupts on the local CPU, and - restore them. They are reentrant; if soft interrupts were - disabled before, they will still be disabled after this pair - of functions has been called. They prevent softirqs and tasklets - from running on the current CPU. - - - - - <function>smp_processor_id</function>() - <filename class="headerfile">include/asm/smp.h</filename> - - - get_cpu() disables preemption (so you won't - suddenly get moved to another CPU) and returns the current - processor number, between 0 and NR_CPUS. Note - that the CPU numbers are not necessarily continuous. You return - it again with put_cpu() when you are done. - - - If you know you cannot be preempted by another task (ie. you are - in interrupt context, or have preemption disabled) you can use - smp_processor_id(). - - - - - <type>__init</type>/<type>__exit</type>/<type>__initdata</type> - <filename class="headerfile">include/linux/init.h</filename> - - - After boot, the kernel frees up a special section; functions - marked with __init and data structures marked with - __initdata are dropped after boot is complete: similarly - modules discard this memory after initialization. __exit - is used to declare a function which is only required on exit: the - function will be dropped if this file is not compiled as a module. - See the header file for use. Note that it makes no sense for a function - marked with __init to be exported to modules with - EXPORT_SYMBOL() - this will break. - - - - - - <function>__initcall()</function>/<function>module_init()</function> - <filename class="headerfile">include/linux/init.h</filename> - - Many parts of the kernel are well served as a module - (dynamically-loadable parts of the kernel). Using the - module_init() and - module_exit() macros it is easy to write code - without #ifdefs which can operate both as a module or built into - the kernel. - - - - The module_init() macro defines which - function is to be called 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 module_init() macro - becomes equivalent to __initcall(), which - through linker magic ensures that the function is called on boot. - - - - The function can return a negative error number to cause - module loading to fail (unfortunately, this has no effect if - the module is compiled into the kernel). This function is - called in user context with interrupts enabled, so it can sleep. - - - - - <function>module_exit()</function> - <filename class="headerfile">include/linux/init.h</filename> - - - This macro defines the function to be called at module removal - time (or never, in the case of the file compiled into the - kernel). It will only be called if the module usage count has - reached zero. This function can also sleep, but cannot fail: - everything must be cleaned up by the time it returns. - - - - Note that this macro is optional: if it is not present, your - module will not be removable (except for 'rmmod -f'). - - - - - <function>try_module_get()</function>/<function>module_put()</function> - <filename class="headerfile">include/linux/module.h</filename> - - - These manipulate the module usage count, to protect against - removal (a module also can't be removed if another module uses one - of its exported symbols: see below). Before calling into module - code, you should call try_module_get() on - that module: if it fails, then the module is being removed and you - should act as if it wasn't there. Otherwise, you can safely enter - the module, and call module_put() when you're - finished. - - - - Most registerable structures have an - owner field, such as in the - file_operations structure. Set this field - to the macro THIS_MODULE. - - - - - - - - Wait Queues - <filename class="headerfile">include/linux/wait.h</filename> - - - [SLEEPS] - - - - A wait queue is used to wait for someone to wake you up when a - certain condition is true. They must be used carefully to ensure - there is no race condition. You declare a - wait_queue_head_t, and then processes which want to - wait for that condition declare a wait_queue_t - referring to themselves, and place that in the queue. - - - - Declaring - - - You declare a wait_queue_head_t using the - DECLARE_WAIT_QUEUE_HEAD() macro, or using the - init_waitqueue_head() routine in your - initialization code. - - - - - Queuing - - - Placing yourself in the waitqueue is fairly complex, because you - must put yourself in the queue before checking the condition. - There is a macro to do this: - wait_event_interruptible() - - include/linux/wait.h The - first argument is the wait queue head, and the second is an - expression which is evaluated; the macro returns - 0 when this expression is true, or - -ERESTARTSYS if a signal is received. - The wait_event() version ignores signals. - - - - - - Waking Up Queued Tasks - - - Call wake_up() - - include/linux/wait.h;, - which will wake up every process in the queue. The exception is - if one has TASK_EXCLUSIVE set, in which case - the remainder of the queue will not be woken. There are other variants - of this basic function available in the same header. - - - - - - Atomic Operations - - - Certain operations are guaranteed atomic on all platforms. The - first class of operations work on atomic_t - - include/asm/atomic.h; this - contains a signed integer (at least 32 bits long), and you must use - these functions to manipulate or read atomic_t variables. - atomic_read() and - atomic_set() get and set the counter, - atomic_add(), - atomic_sub(), - atomic_inc(), - atomic_dec(), and - atomic_dec_and_test() (returns - true if it was decremented to zero). - - - - Yes. It returns true (i.e. != 0) if the - atomic variable is zero. - - - - Note that these functions are slower than normal arithmetic, and - so should not be used unnecessarily. - - - - The second class of atomic operations is atomic bit operations on an - unsigned long, defined in - - include/linux/bitops.h. These - operations generally take a pointer to the bit pattern, and a bit - number: 0 is the least significant bit. - set_bit(), clear_bit() - and change_bit() set, clear, and flip the - given bit. test_and_set_bit(), - test_and_clear_bit() and - test_and_change_bit() do the same thing, - except return true if the bit was previously set; these are - particularly useful for atomically setting flags. - - - - It is possible to call these operations with bit indices greater - than BITS_PER_LONG. The resulting behavior is strange on big-endian - platforms though so it is a good idea not to do this. - - - - - Symbols - - - Within the kernel proper, the normal linking rules apply - (ie. unless a symbol is declared to be file scope with the - static keyword, it can be used anywhere in the - kernel). However, for modules, a special exported symbol table is - kept which limits the entry points to the kernel proper. Modules - can also export symbols. - - - - <function>EXPORT_SYMBOL()</function> - <filename class="headerfile">include/linux/export.h</filename> - - - This is the classic method of exporting a symbol: dynamically - loaded modules will be able to use the symbol as normal. - - - - - <function>EXPORT_SYMBOL_GPL()</function> - <filename class="headerfile">include/linux/export.h</filename> - - - Similar to EXPORT_SYMBOL() except that the - symbols exported by EXPORT_SYMBOL_GPL() can - only be seen by modules with a - MODULE_LICENSE() that specifies a GPL - compatible license. It implies that the function is considered - an internal implementation issue, and not really an interface. - Some maintainers and developers may however - require EXPORT_SYMBOL_GPL() when adding any new APIs or functionality. - - - - - - Routines and Conventions - - - Double-linked lists - <filename class="headerfile">include/linux/list.h</filename> - - - There used to be three sets of linked-list routines in the kernel - headers, but this one is the winner. If you don't have some - particular pressing need for a single list, it's a good choice. - - - - In particular, list_for_each_entry is useful. - - - - - Return Conventions - - - For code called in user context, it's very common to defy C - convention, and return 0 for success, - and a negative error number - (eg. -EFAULT) for failure. This can be - unintuitive at first, but it's fairly widespread in the kernel. - - - - Using ERR_PTR() - - include/linux/err.h; to - encode a negative error number into a pointer, and - IS_ERR() and PTR_ERR() - to get it back out again: avoids a separate pointer parameter for - the error number. Icky, but in a good way. - - - - - Breaking Compilation - - - Linus and the other developers sometimes change function or - structure names in development kernels; this is not done just to - keep everyone on their toes: it reflects a fundamental change - (eg. can no longer be called with interrupts on, or does extra - checks, or doesn't do checks which were caught before). Usually - this is accompanied by a fairly complete note to the linux-kernel - mailing list; search the archive. Simply doing a global replace - on the file usually makes things worse. - - - - - Initializing structure members - - - The preferred method of initializing structures is to use - designated initialisers, as defined by ISO C99, eg: - - -static struct block_device_operations opt_fops = { - .open = opt_open, - .release = opt_release, - .ioctl = opt_ioctl, - .check_media_change = opt_media_change, -}; - - - This makes it easy to grep for, and makes it clear which - structure fields are set. You should do this because it looks - cool. - - - - - GNU Extensions - - - GNU Extensions are explicitly allowed in the Linux kernel. - Note that some of the more complex ones are not very well - supported, due to lack of general use, but the following are - considered standard (see the GCC info page section "C - Extensions" for more details - Yes, really the info page, the - man page is only a short summary of the stuff in info). - - - - - Inline functions - - - - - Statement expressions (ie. the ({ and }) constructs). - - - - - Declaring attributes of a function / variable / type - (__attribute__) - - - - - typeof - - - - - Zero length arrays - - - - - Macro varargs - - - - - Arithmetic on void pointers - - - - - Non-Constant initializers - - - - - Assembler Instructions (not outside arch/ and include/asm/) - - - - - Function names as strings (__func__). - - - - - __builtin_constant_p() - - - - - - Be wary when using long long in the kernel, the code gcc generates for - it is horrible and worse: division and multiplication does not work - on i386 because the GCC runtime functions for it are missing from - the kernel environment. - - - - - - - C++ - - - Using C++ in the kernel is usually a bad idea, because the - kernel does not provide the necessary runtime environment - and the include files are not tested for it. It is still - possible, but not recommended. If you really want to do - this, forget about exceptions at least. - - - - - #if - - - It is generally considered cleaner to use macros in header files - (or at the top of .c files) to abstract away functions rather than - using `#if' pre-processor statements throughout the source code. - - - - - - Putting Your Stuff in the Kernel - - - In order to get your stuff into shape for official inclusion, or - even to make a neat patch, there's administrative work to be - done: - - - - - Figure out whose pond you've been pissing in. Look at the top of - the source files, inside the MAINTAINERS - file, and last of all in the CREDITS file. - You should coordinate with this person to make sure you're not - duplicating effort, or trying something that's already been - rejected. - - - - Make sure you put your name and EMail address at the top of - any files you create or mangle significantly. This is the - first place people will look when they find a bug, or when - they want to make a change. - - - - - - Usually you want a configuration option for your kernel hack. - Edit Kconfig in the appropriate directory. - The Config language is simple to use by cut and paste, and there's - complete documentation in - Documentation/kbuild/kconfig-language.txt. - - - - In your description of the option, make sure you address both the - expert user and the user who knows nothing about your feature. Mention - incompatibilities and issues here. Definitely - end your description with if in doubt, say N - (or, occasionally, `Y'); this is for people who have no - idea what you are talking about. - - - - - - Edit the Makefile: the CONFIG variables are - exported here so you can usually just add a "obj-$(CONFIG_xxx) += - xxx.o" line. The syntax is documented in - Documentation/kbuild/makefiles.txt. - - - - - - Put yourself in CREDITS if you've done - something noteworthy, usually beyond a single file (your name - should be at the top of the source files anyway). - MAINTAINERS means you want to be consulted - when changes are made to a subsystem, and hear about bugs; it - implies a more-than-passing commitment to some part of the code. - - - - - - Finally, don't forget to read Documentation/process/submitting-patches.rst - and possibly Documentation/process/submitting-drivers.rst. - - - - - - - Kernel Cantrips - - - Some favorites from browsing the source. Feel free to add to this - list. - - - - arch/x86/include/asm/delay.h: - - -#define ndelay(n) (__builtin_constant_p(n) ? \ - ((n) > 20000 ? __bad_ndelay() : __const_udelay((n) * 5ul)) : \ - __ndelay(n)) - - - - include/linux/fs.h: - - -/* - * Kernel pointers have redundant information, so we can use a - * scheme where we can return either an error code or a dentry - * pointer with the same return value. - * - * This should be a per-architecture thing, to allow different - * error and pointer decisions. - */ - #define ERR_PTR(err) ((void *)((long)(err))) - #define PTR_ERR(ptr) ((long)(ptr)) - #define IS_ERR(ptr) ((unsigned long)(ptr) > (unsigned long)(-1000)) - - - - arch/x86/include/asm/uaccess_32.h: - - - -#define copy_to_user(to,from,n) \ - (__builtin_constant_p(n) ? \ - __constant_copy_to_user((to),(from),(n)) : \ - __generic_copy_to_user((to),(from),(n))) - - - - arch/sparc/kernel/head.S: - - - -/* - * Sun people can't spell worth damn. "compatability" indeed. - * At least we *know* we can't spell, and use a spell-checker. - */ - -/* Uh, actually Linus it is I who cannot spell. Too much murky - * Sparc assembly will do this to ya. - */ -C_LABEL(cputypvar): - .asciz "compatibility" - -/* Tested on SS-5, SS-10. Probably someone at Sun applied a spell-checker. */ - .align 4 -C_LABEL(cputypvar_sun4m): - .asciz "compatible" - - - - arch/sparc/lib/checksum.S: - - - - /* Sun, you just can't beat me, you just can't. Stop trying, - * give up. I'm serious, I am going to kick the living shit - * out of you, game over, lights out. - */ - - - - - Thanks - - - Thanks to Andi Kleen for the idea, answering my questions, fixing - my mistakes, filling content, etc. Philipp Rumpf for more spelling - and clarity fixes, and some excellent non-obvious points. Werner - Almesberger for giving me a great summary of - disable_irq(), and Jes Sorensen and Andrea - Arcangeli added caveats. Michael Elizabeth Chastain for checking - and adding to the Configure section. Telsa Gwynne for teaching me DocBook. - - - - diff --git a/Documentation/conf.py b/Documentation/conf.py index bacf9d337c89..0ce8001a50b6 100644 --- a/Documentation/conf.py +++ b/Documentation/conf.py @@ -352,6 +352,8 @@ latex_documents = [ 'The kernel development community', 'manual'), ('kernel-documentation', 'kernel-documentation.tex', 'The Linux Kernel Documentation', 'The kernel development community', 'manual'), + ('kernel-hacking/index', 'kernel-hacking.tex', 'Kernel Hacking Guides', + 'The kernel development community', 'manual'), ('process/index', 'development-process.tex', 'Linux Kernel Development Documentation', 'The kernel development community', 'manual'), ('gpu/index', 'gpu.tex', 'Linux GPU Driver Developer\'s Guide', diff --git a/Documentation/index.rst b/Documentation/index.rst index bc67dbf76eb0..01a24486ea5f 100644 --- a/Documentation/index.rst +++ b/Documentation/index.rst @@ -51,6 +51,7 @@ merged much easier. process/index dev-tools/index doc-guide/index + kernel-hacking/index Kernel API documentation ------------------------ diff --git a/Documentation/kernel-hacking/conf.py b/Documentation/kernel-hacking/conf.py new file mode 100644 index 000000000000..3d8acf0f33ad --- /dev/null +++ b/Documentation/kernel-hacking/conf.py @@ -0,0 +1,10 @@ +# -*- coding: utf-8; mode: python -*- + +project = "Kernel Hacking Guides" + +tags.add("subproject") + +latex_documents = [ + ('index', 'kernel-hacking.tex', project, + 'The kernel development community', 'manual'), +] diff --git a/Documentation/kernel-hacking/hacking.rst b/Documentation/kernel-hacking/hacking.rst new file mode 100644 index 000000000000..75597e627791 --- /dev/null +++ b/Documentation/kernel-hacking/hacking.rst @@ -0,0 +1,794 @@ +============================================ +Unreliable Guide To Hacking The Linux Kernel +============================================ + +:Author: Rusty Russell + +Introduction +============ + +Welcome, gentle reader, to Rusty's Remarkably Unreliable Guide to Linux +Kernel Hacking. This document describes the common routines and general +requirements for kernel code: its goal is to serve as a primer for Linux +kernel development for experienced C programmers. I avoid implementation +details: that's what the code is for, and I ignore whole tracts of +useful routines. + +Before you read this, please understand that I never wanted to write +this document, being grossly under-qualified, but I always wanted to +read it, and this was the only way. I hope it will grow into a +compendium of best practice, common starting points and random +information. + +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 each +other, but above that is a strict hierarchy: each can only be preempted +by 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 by +interrupts. 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 are +currently executing) is valid, and :c:func:`in_interrupt()` +(``include/linux/interrupt.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 hardware +which produce interrupts at any time. The kernel runs interrupt +handlers, which services the hardware. The kernel guarantees that this +handler is never re-entered: if the same interrupt arrives, it is queued +(or dropped). Because it disables interrupts, this handler has to be +fast: frequently it simply acknowledges the interrupt, marks a 'software +interrupt' for execution and exits. + +You can tell you are in a hardware interrupt, because +:c:func:`in_irq()` returns true. + +.. 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 hardware +interrupt handler exits, any 'software interrupts' which are marked +pending (usually by hardware interrupts) are run (``kernel/softirq.c``). + +Much of the real interrupt handling work is done here. Early in the +transition to SMP, there were only 'bottom halves' (BHs), which didn't +take advantage of multiple CPUs. Shortly after we switched from wind-up +computers made of match-sticks and snot, we abandoned this limitation +and switched to 'softirqs'. + +``include/linux/interrupt.h`` lists the different softirqs. A very +important softirq is the timer softirq (``include/linux/timer.h``): you +can register to have it call functions for you in a given length of +time. + +Softirqs are often a pain to deal with, since the same softirq will run +simultaneously on more than one CPU. For this reason, tasklets +(``include/linux/interrupt.h``) are more often used: they are +dynamically-registrable (meaning you can have as many as you want), and +they also guarantee that any tasklet will only run on one CPU at any +time, although different tasklets can run simultaneously. + +.. warning:: + + The name 'tasklet' is misleading: they have nothing to do with + 'tasks', and probably more to do with some bad vodka Alexey + Kuznetsov had at the time. + +You can tell you are in a softirq (or tasklet) using the +:c:func:`in_softirq()` macro (``include/linux/interrupt.h``). + +.. warning:: + + Beware that this will return a false positive if a bh lock (see + below) 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. You +create a character device and implement an appropriate ioctl for it. +This is much more flexible than system calls, doesn't have to be entered +in every architecture's ``include/asm/unistd.h`` and +``arch/kernel/entry.S`` file, and is much more likely to be accepted by +Linus. + +If all your routine does is read or write some parameter, consider +implementing a :c:func:`sysfs()` interface instead. + +Inside the ioctl you're in user context to a process. When a error +occurs you return a negated errno (see ``include/linux/errno.h``), +otherwise you return 0. + +After you slept you should check if a signal occurred: the Unix/Linux +way of handling signals is to temporarily exit the system call with the +``-ERESTARTSYS`` error. The system call entry code will switch back to +user context, process the signal handler and then your system call will +be restarted (unless the user disabled that). So you should be prepared +to process the restart, e.g. if you're in the middle of manipulating +some 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 need +to 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 "Provide +mechanism 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 user +space access functions (\*_user) and memory allocation functions +without ``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 the +rules, you will eventually lock up your box. + +Really. + +Common Routines +=============== + +:c:func:`printk()` ``include/linux/kernel.h`` +--------------------------------------------- + +:c:func:`printk()` feeds kernel messages to the console, dmesg, and +the syslog daemon. It is useful for debugging and reporting errors, and +can be used inside interrupt context, but use with caution: a machine +which has its console flooded with printk messages is unusable. It uses +a format string mostly compatible with ANSI C printf, and C string +concatenation to give it a first "priority" argument: + +:: + + printk(KERN_INFO "i = %u\n", i); + + +See ``include/linux/kernel.h``; for other ``KERN_`` values; these are +interpreted by syslog as the level. Special case: for printing an IP +address use + +:: + + __be32 ipaddress; + printk(KERN_INFO "my ip: %pI4\n", &ipaddress); + + +:c:func:`printk()` internally uses a 1K buffer and does not catch +overruns. 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/from]_user()` / :c:func:`get_user()` / :c:func:`put_user()` ``include/linux/uaccess.h`` +--------------------------------------------------------------------------------------------------------- + +**[SLEEPS]** + +:c:func:`put_user()` and :c:func:`get_user()` are used to get +and put single values (such as an int, char, or long) from and to +userspace. A pointer into userspace should never be simply dereferenced: +data should be copied using these routines. Both return ``-EFAULT`` or +0. + +:c:func:`copy_to_user()` and :c:func:`copy_from_user()` are +more general: they copy an arbitrary amount of data to and from +userspace. + +.. 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 moronic interface makes me cringe. The flamewar comes up +every year or so. --RR.] + +The functions may sleep implicitly. This should never be called outside +user context (it makes no sense), with interrupts disabled, or a +spinlock held. + +:c:func:`kmalloc()`/:c:func:`kfree()` ``include/linux/slab.h`` +-------------------------------------------------------------- + +**[MAY SLEEP: SEE BELOW]** + +These routines are used to dynamically request pointer-aligned chunks of +memory, 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 warning +message, then maybe you called a sleeping allocation function from +interrupt context without ``GFP_ATOMIC``. You should really fix that. +Run, don't walk. + +If you are allocating at least ``PAGE_SIZE`` (``include/asm/page.h``) +bytes, consider using :c:func:`__get_free_pages()` +(``include/linux/mm.h``). It takes an order argument (0 for page sized, +1 for double page, 2 for four pages etc.) and the same memory priority +flag 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 kernel +map. This block is not contiguous in physical memory, but the MMU makes +it 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 physically +contiguous memory for some weird device, you have a problem: it is +poorly supported in Linux because after some time memory fragmentation +in a running kernel makes it hard. The best way is to allocate the block +early in the boot process via the :c:func:`alloc_bootmem()` +routine. + +Before inventing your own cache of often-used objects consider using a +slab cache in ``include/linux/slab.h`` + +:c:func:`current()` ``include/asm/current.h`` +--------------------------------------------- + +This global variable (really a macro) contains a pointer to the current +task structure, so is only valid in user context. For example, when a +process makes a system call, this will point to the task structure of +the calling process. It is **not NULL** in interrupt context. + +:c:func:`mdelay()`/:c:func:`udelay()` ``include/asm/delay.h`` ``include/linux/delay.h`` +--------------------------------------------------------------------------------------- + +The :c:func:`udelay()` and :c:func:`ndelay()` functions can be +used for small pauses. Do not use large values with them as you risk +overflow - the helper function :c:func:`mdelay()` is useful here, or +consider :c:func:`msleep()`. + +:c:func:`cpu_to_be32()`/:c:func:`be32_to_cpu()`/:c:func:`cpu_to_le32()`/:c:func:`le32_to_cpu()` ``include/asm/byteorder.h`` +--------------------------------------------------------------------------------------------------------------------------- + +The :c:func:`cpu_to_be32()` family (where the "32" can be replaced +by 64 or 16, and the "be" can be replaced by "le") are the general way +to 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 pointer +variation, such as :c:func:`cpu_to_be32p()`, which take a pointer +to the given type, and return the converted value. The other variation +is the "in-situ" family, such as :c:func:`cpu_to_be32s()`, which +convert value referred to by the pointer, and return void. + +:c:func:`local_irq_save()`/:c:func:`local_irq_restore()` ``include/linux/irqflags.h`` +------------------------------------------------------------------------------------- + +These routines disable hard interrupts on the local CPU, and restore +them. They are reentrant; saving the previous state in their one +``unsigned long flags`` argument. If you know that interrupts are +enabled, you can simply use :c:func:`local_irq_disable()` and +:c:func:`local_irq_enable()`. + +:c:func:`local_bh_disable()`/:c:func:`local_bh_enable()` ``include/linux/interrupt.h`` +-------------------------------------------------------------------------------------- + +These routines disable soft interrupts on the local CPU, and restore +them. They are reentrant; if soft interrupts were disabled before, they +will 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()`() ``include/asm/smp.h`` +---------------------------------------------------- + +:c:func:`get_cpu()` disables preemption (so you won't suddenly get +moved to another CPU) and returns the current processor number, between +0 and ``NR_CPUS``. Note that the CPU numbers are not necessarily +continuous. You return it again with :c:func:`put_cpu()` when you +are done. + +If you know you cannot be preempted by another task (ie. you are in +interrupt context, or have preemption disabled) you can use +smp_processor_id(). + +``__init``/``__exit``/``__initdata`` ``include/linux/init.h`` +------------------------------------------------------------- + +After boot, the kernel frees up a special section; functions marked with +``__init`` and data structures marked with ``__initdata`` are dropped +after boot is complete: similarly modules discard this memory after +initialization. ``__exit`` is used to declare a function which is only +required on exit: the function will be dropped if this file is not +compiled as a module. See the header file for use. Note that it makes no +sense for a function marked with ``__init`` to be exported to modules +with :c:func:`EXPORT_SYMBOL()` - this will break. + +:c:func:`__initcall()`/:c:func:`module_init()` ``include/linux/init.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 it +is easy to write code without #ifdefs which can operate both as a module +or built into the kernel. + +The :c:func:`module_init()` macro defines which function is to be +called 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 that +the function is called on boot. + +The function can return a negative error number to cause module loading +to fail (unfortunately, this has no effect if the module is compiled +into the kernel). This function is called in user context with +interrupts enabled, so it can sleep. + +:c:func:`module_exit()` ``include/linux/init.h`` +------------------------------------------------ + +This macro defines the function to be called at module removal time (or +never, in the case of the file compiled into the kernel). It will only +be called if the module usage count has reached zero. This function can +also sleep, but cannot fail: everything must be cleaned up by the time +it returns. + +Note that this macro is optional: if it is not present, your module will +not be removable (except for 'rmmod -f'). + +:c:func:`try_module_get()`/:c:func:`module_put()` ``include/linux/module.h`` +---------------------------------------------------------------------------- + +These manipulate the module usage count, to protect against removal (a +module also can't be removed if another module uses one of its exported +symbols: see below). Before calling into module code, you should call +:c:func:`try_module_get()` on that module: if it fails, then the +module 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 ` 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 certain +condition is true. They must be used carefully to ensure there is no +race condition. You declare a ``wait_queue_head_t``, and then processes +which want to wait for that condition declare a ``wait_queue_t`` +referring to themselves, and place that in the queue. + +Declaring +--------- + +You declare a ``wait_queue_head_t`` using the +:c:func:`DECLARE_WAIT_QUEUE_HEAD()` macro, or using the +:c:func:`init_waitqueue_head()` routine in your initialization +code. + +Queuing +------- + +Placing yourself in the waitqueue is fairly complex, because you must +put yourself in the queue before checking the condition. There is a +macro to do this: :c:func:`wait_event_interruptible()` +``include/linux/wait.h`` The first argument is the wait queue head, and +the second is an expression which is evaluated; the macro returns 0 when +this expression is true, or -ERESTARTSYS if a signal is received. The +:c:func:`wait_event()` version ignores signals. + +Waking Up Queued Tasks +---------------------- + +Call :c:func:`wake_up()` ``include/linux/wait.h``;, which will wake +up every process in the queue. The exception is if one has +``TASK_EXCLUSIVE`` set, in which case the remainder of the queue will +not be woken. There are other variants of this basic function available +in the same header. + +Atomic Operations +================= + +Certain operations are guaranteed atomic on all platforms. The first +class of operations work on ``atomic_t`` ``include/asm/atomic.h``; this +contains a signed integer (at least 32 bits long), and you must use +these functions to manipulate or read atomic_t variables. +:c:func:`atomic_read()` and :c:func:`atomic_set()` get and set +the counter, :c:func:`atomic_add()`, :c:func:`atomic_sub()`, +:c:func:`atomic_inc()`, :c:func:`atomic_dec()`, and +:c:func:`atomic_dec_and_test()` (returns true if it was +decremented to zero). + +Yes. It returns true (i.e. != 0) if the atomic variable is zero. + +Note that these functions are slower than normal arithmetic, and so +should not be used unnecessarily. + +The second class of atomic operations is atomic bit operations on an +``unsigned long``, defined in ``include/linux/bitops.h``. These +operations generally take a pointer to the bit pattern, and a bit +number: 0 is the least significant bit. :c:func:`set_bit()`, +:c:func:`clear_bit()` and :c:func:`change_bit()` set, clear, +and flip the given bit. :c:func:`test_and_set_bit()`, +:c:func:`test_and_clear_bit()` and +:c:func:`test_and_change_bit()` do the same thing, except return +true if the bit was previously set; these are particularly useful for +atomically setting flags. + +It is possible to call these operations with bit indices greater than +BITS_PER_LONG. The resulting behavior is strange on big-endian +platforms though so it is a good idea not to do this. + +Symbols +======= + +Within the kernel proper, the normal linking rules apply (ie. unless a +symbol is declared to be file scope with the ``static`` keyword, it can +be used anywhere in the kernel). However, for modules, a special +exported symbol table is kept which limits the entry points to the +kernel proper. Modules can also export symbols. + +:c:func:`EXPORT_SYMBOL()` ``include/linux/export.h`` +---------------------------------------------------- + +This is the classic method of exporting a symbol: dynamically loaded +modules will be able to use the symbol as normal. + +:c:func:`EXPORT_SYMBOL_GPL()` ``include/linux/export.h`` +-------------------------------------------------------- + +Similar to :c:func:`EXPORT_SYMBOL()` except that the symbols +exported by :c:func:`EXPORT_SYMBOL_GPL()` can only be seen by +modules with a :c:func:`MODULE_LICENSE()` that specifies a GPL +compatible license. It implies that the function is considered an +internal implementation issue, and not really an interface. Some +maintainers and developers may however require EXPORT_SYMBOL_GPL() +when adding any new APIs or functionality. + +Routines and Conventions +======================== + +Double-linked lists ``include/linux/list.h`` +-------------------------------------------- + +There used to be three sets of linked-list routines in the kernel +headers, but this one is the winner. If you don't have some particular +pressing need for a single list, it's a good choice. + +In particular, :c:func:`list_for_each_entry()` is useful. + +Return Conventions +------------------ + +For code called in user context, it's very common to defy C convention, +and return 0 for success, and a negative error number (eg. -EFAULT) for +failure. This can be unintuitive at first, but it's fairly widespread in +the kernel. + +Using :c:func:`ERR_PTR()` ``include/linux/err.h``; to encode a +negative error number into a pointer, and :c:func:`IS_ERR()` and +:c:func:`PTR_ERR()` to get it back out again: avoids a separate +pointer parameter for the error number. Icky, but in a good way. + +Breaking Compilation +-------------------- + +Linus and the other developers sometimes change function or structure +names in development kernels; this is not done just to keep everyone on +their toes: it reflects a fundamental change (eg. can no longer be +called with interrupts on, or does extra checks, or doesn't do checks +which were caught before). Usually this is accompanied by a fairly +complete note to the linux-kernel mailing list; search the archive. +Simply doing a global replace on the file usually makes things **worse**. + +Initializing structure members +------------------------------ + +The preferred method of initializing structures is to use designated +initialisers, as defined by ISO C99, eg: + +:: + + static struct block_device_operations opt_fops = { + .open = opt_open, + .release = opt_release, + .ioctl = opt_ioctl, + .check_media_change = opt_media_change, + }; + + +This makes it easy to grep for, and makes it clear which structure +fields are set. You should do this because it looks cool. + +GNU Extensions +-------------- + +GNU Extensions are explicitly allowed in the Linux kernel. Note that +some of the more complex ones are not very well supported, due to lack +of general use, but the following are considered standard (see the GCC +info page section "C Extensions" for more details - Yes, really the info +page, the man page is only a short summary of the stuff in info). + +- Inline functions + +- Statement expressions (ie. the ({ and }) constructs). + +- Declaring attributes of a function / variable / type + (__attribute__) + +- typeof + +- Zero length arrays + +- Macro varargs + +- Arithmetic on void pointers + +- Non-Constant initializers + +- Assembler Instructions (not outside arch/ and include/asm/) + +- Function names as strings (__func__). + +- __builtin_constant_p() + +Be wary when using long long in the kernel, the code gcc generates for +it is horrible and worse: division and multiplication does not work on +i386 because the GCC runtime functions for it are missing from the +kernel environment. + +C++ +--- + +Using C++ in the kernel is usually a bad idea, because the kernel does +not provide the necessary runtime environment and the include files are +not tested for it. It is still possible, but not recommended. If you +really want to do this, forget about exceptions at least. + +NUMif +----- + +It is generally considered cleaner to use macros in header files (or at +the top of .c files) to abstract away functions rather than using \`#if' +pre-processor statements throughout the source code. + +Putting Your Stuff in the Kernel +================================ + +In order to get your stuff into shape for official inclusion, or even to +make a neat patch, there's administrative work to be done: + +- Figure out whose pond you've been pissing in. Look at the top of the + source files, inside the ``MAINTAINERS`` file, and last of all in the + ``CREDITS`` file. You should coordinate with this person to make sure + you're not duplicating effort, or trying something that's already + been rejected. + + Make sure you put your name and EMail address at the top of any files + you create or mangle significantly. This is the first place people + will look when they find a bug, or when **they** want to make a change. + +- Usually you want a configuration option for your kernel hack. Edit + ``Kconfig`` in the appropriate directory. The Config language is + simple to use by cut and paste, and there's complete documentation in + ``Documentation/kbuild/kconfig-language.txt``. + + In your description of the option, make sure you address both the + expert user and the user who knows nothing about your feature. + Mention incompatibilities and issues here. **Definitely** end your + description with “if in doubt, say N” (or, occasionally, \`Y'); this + is for people who have no idea what you are talking about. + +- Edit the ``Makefile``: the CONFIG variables are exported here so you + can usually just add a "obj-$(CONFIG_xxx) += xxx.o" line. The syntax + is documented in ``Documentation/kbuild/makefiles.txt``. + +- Put yourself in ``CREDITS`` if you've done something noteworthy, + usually beyond a single file (your name should be at the top of the + source files anyway). ``MAINTAINERS`` means you want to be consulted + when changes are made to a subsystem, and hear about bugs; it implies + a more-than-passing commitment to some part of the code. + +- Finally, don't forget to read + ``Documentation/process/submitting-patches.rst`` and possibly + ``Documentation/process/submitting-drivers.rst``. + +Kernel Cantrips +=============== + +Some favorites from browsing the source. Feel free to add to this list. + +``arch/x86/include/asm/delay.h:`` + +:: + + #define ndelay(n) (__builtin_constant_p(n) ? \ + ((n) > 20000 ? __bad_ndelay() : __const_udelay((n) * 5ul)) : \ + __ndelay(n)) + + +``include/linux/fs.h``: + +:: + + /* + * Kernel pointers have redundant information, so we can use a + * scheme where we can return either an error code or a dentry + * pointer with the same return value. + * + * This should be a per-architecture thing, to allow different + * error and pointer decisions. + */ + #define ERR_PTR(err) ((void *)((long)(err))) + #define PTR_ERR(ptr) ((long)(ptr)) + #define IS_ERR(ptr) ((unsigned long)(ptr) > (unsigned long)(-1000)) + +``arch/x86/include/asm/uaccess_32.h:`` + +:: + + #define copy_to_user(to,from,n) \ + (__builtin_constant_p(n) ? \ + __constant_copy_to_user((to),(from),(n)) : \ + __generic_copy_to_user((to),(from),(n))) + + +``arch/sparc/kernel/head.S:`` + +:: + + /* + * Sun people can't spell worth damn. "compatability" indeed. + * At least we *know* we can't spell, and use a spell-checker. + */ + + /* Uh, actually Linus it is I who cannot spell. Too much murky + * Sparc assembly will do this to ya. + */ + C_LABEL(cputypvar): + .asciz "compatibility" + + /* Tested on SS-5, SS-10. Probably someone at Sun applied a spell-checker. */ + .align 4 + C_LABEL(cputypvar_sun4m): + .asciz "compatible" + + +``arch/sparc/lib/checksum.S:`` + +:: + + /* Sun, you just can't beat me, you just can't. Stop trying, + * give up. I'm serious, I am going to kick the living shit + * out of you, game over, lights out. + */ + + +Thanks +====== + +Thanks to Andi Kleen for the idea, answering my questions, fixing my +mistakes, filling content, etc. Philipp Rumpf for more spelling and +clarity fixes, and some excellent non-obvious points. Werner Almesberger +for giving me a great summary of :c:func:`disable_irq()`, and Jes +Sorensen and Andrea Arcangeli added caveats. Michael Elizabeth Chastain +for checking and adding to the Configure section. Telsa Gwynne for +teaching me DocBook. diff --git a/Documentation/kernel-hacking/index.rst b/Documentation/kernel-hacking/index.rst new file mode 100644 index 000000000000..ba208bf03ce9 --- /dev/null +++ b/Documentation/kernel-hacking/index.rst @@ -0,0 +1,8 @@ +===================== +Kernel Hacking Guides +===================== + +.. toctree:: + :maxdepth: 2 + + hacking From dca1e58e3f1c82a840abaafb9328f84ae69a9926 Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 09:41:52 -0300 Subject: [PATCH 02/90] kernel-hacking: update document This document is fairly updated. Yet, some stuff moved to other kernel headers. So, update to point to the right places. While here, adjust some minor ReST markups. Signed-off-by: Mauro Carvalho Chehab --- Documentation/kernel-hacking/hacking.rst | 179 +++++++++++++---------- 1 file changed, 98 insertions(+), 81 deletions(-) diff --git a/Documentation/kernel-hacking/hacking.rst b/Documentation/kernel-hacking/hacking.rst index 75597e627791..1a456b60a7cf 100644 --- a/Documentation/kernel-hacking/hacking.rst +++ b/Documentation/kernel-hacking/hacking.rst @@ -56,7 +56,7 @@ interrupts. You can sleep, by calling :c:func:`schedule()`. In user context, the ``current`` pointer (indicating the task we are currently executing) is valid, and :c:func:`in_interrupt()` -(``include/linux/interrupt.h``) is false. +(``include/linux/preempt.h``) is false. .. warning:: @@ -114,12 +114,12 @@ time, although different tasklets can run simultaneously. Kuznetsov had at the time. You can tell you are in a softirq (or tasklet) using the -:c:func:`in_softirq()` macro (``include/linux/interrupt.h``). +:c:func:`in_softirq()` macro (``include/linux/preempt.h``). .. warning:: - Beware that this will return a false positive if a bh lock (see - below) is held. + Beware that this will return a false positive if a + :ref:`botton half lock ` is held. Some Basic Rules ================ @@ -154,9 +154,7 @@ The Linux kernel is portable ioctls: Not writing a new system call ===================================== -A system call generally looks like this - -:: +A system call generally looks like this:: asmlinkage long sys_mycall(int arg) { @@ -175,7 +173,9 @@ If all your routine does is read or write some parameter, consider implementing a :c:func:`sysfs()` interface instead. Inside the ioctl you're in user context to a process. When a error -occurs you return a negated errno (see ``include/linux/errno.h``), +occurs 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/Linux @@ -195,9 +195,7 @@ some data structure. If you're doing longer computations: first think userspace. If you **really** want to do it in kernel you should regularly check if you need to give up the CPU (remember there is cooperative multitasking per CPU). -Idiom: - -:: +Idiom:: cond_resched(); /* Will sleep */ @@ -231,26 +229,24 @@ Really. Common Routines =============== -:c:func:`printk()` ``include/linux/kernel.h`` ---------------------------------------------- +:c:func:`printk()` +------------------ + +Defined in ``include/linux/printk.h`` :c:func:`printk()` feeds kernel messages to the console, dmesg, and the syslog daemon. It is useful for debugging and reporting errors, and can be used inside interrupt context, but use with caution: a machine which has its console flooded with printk messages is unusable. It uses a format string mostly compatible with ANSI C printf, and C string -concatenation to give it a first "priority" argument: - -:: +concatenation to give it a first "priority" argument:: printk(KERN_INFO "i = %u\n", i); -See ``include/linux/kernel.h``; for other ``KERN_`` values; these are +See ``include/linux/kern_levels.h``; for other ``KERN_`` values; these are interpreted by syslog as the level. Special case: for printing an IP -address use - -:: +address use:: __be32 ipaddress; printk(KERN_INFO "my ip: %pI4\n", &ipaddress); @@ -270,8 +266,10 @@ overruns. Make sure that will be enough. on top of its printf function: "Printf should not be used for chit-chat". You should follow that advice. -:c:func:`copy_[to/from]_user()` / :c:func:`get_user()` / :c:func:`put_user()` ``include/linux/uaccess.h`` ---------------------------------------------------------------------------------------------------------- +: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]** @@ -297,8 +295,10 @@ The functions may sleep implicitly. This should never be called outside user context (it makes no sense), with interrupts disabled, or a spinlock held. -:c:func:`kmalloc()`/:c:func:`kfree()` ``include/linux/slab.h`` --------------------------------------------------------------- +:c:func:`kmalloc()`/:c:func:`kfree()` +------------------------------------- + +Defined in ``include/linux/slab.h`` **[MAY SLEEP: SEE BELOW]** @@ -324,9 +324,9 @@ message, then maybe you called a sleeping allocation function from interrupt context without ``GFP_ATOMIC``. You should really fix that. Run, don't walk. -If you are allocating at least ``PAGE_SIZE`` (``include/asm/page.h``) -bytes, consider using :c:func:`__get_free_pages()` -(``include/linux/mm.h``). It takes an order argument (0 for page sized, +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 priority flag word as above. @@ -344,24 +344,30 @@ routine. Before inventing your own cache of often-used objects consider using a slab cache in ``include/linux/slab.h`` -:c:func:`current()` ``include/asm/current.h`` ---------------------------------------------- +:c:func:`current()` +------------------- + +Defined in ``include/asm/current.h`` This global variable (really a macro) contains a pointer to the current task structure, so is only valid in user context. For example, when a process makes a system call, this will point to the task structure of the calling process. It is **not NULL** in interrupt context. -:c:func:`mdelay()`/:c:func:`udelay()` ``include/asm/delay.h`` ``include/linux/delay.h`` ---------------------------------------------------------------------------------------- +: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 be used for small pauses. Do not use large values with them as you risk overflow - the helper function :c:func:`mdelay()` is useful here, or consider :c:func:`msleep()`. -:c:func:`cpu_to_be32()`/:c:func:`be32_to_cpu()`/:c:func:`cpu_to_le32()`/:c:func:`le32_to_cpu()` ``include/asm/byteorder.h`` ---------------------------------------------------------------------------------------------------------------------------- +: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 replaced by 64 or 16, and the "be" can be replaced by "le") are the general way @@ -375,8 +381,10 @@ to the given type, and return the converted value. The other variation is the "in-situ" family, such as :c:func:`cpu_to_be32s()`, which convert value referred to by the pointer, and return void. -:c:func:`local_irq_save()`/:c:func:`local_irq_restore()` ``include/linux/irqflags.h`` -------------------------------------------------------------------------------------- +: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 restore them. They are reentrant; saving the previous state in their one @@ -384,16 +392,23 @@ them. They are reentrant; saving the previous state in their one enabled, you can simply use :c:func:`local_irq_disable()` and :c:func:`local_irq_enable()`. -:c:func:`local_bh_disable()`/:c:func:`local_bh_enable()` ``include/linux/interrupt.h`` --------------------------------------------------------------------------------------- +.. _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 restore them. They are reentrant; if soft interrupts were disabled before, they will 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()`() ``include/asm/smp.h`` ----------------------------------------------------- +:c:func:`smp_processor_id()` +---------------------------- + +Defined in ``include/linux/smp.h`` :c:func:`get_cpu()` disables preemption (so you won't suddenly get moved to another CPU) and returns the current processor number, between @@ -405,8 +420,10 @@ If you know you cannot be preempted by another task (ie. you are in interrupt context, or have preemption disabled) you can use smp_processor_id(). -``__init``/``__exit``/``__initdata`` ``include/linux/init.h`` -------------------------------------------------------------- +``__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 dropped @@ -415,10 +432,13 @@ initialization. ``__exit`` is used to declare a function which is only required on exit: the function will be dropped if this file is not compiled as a module. See the header file for use. Note that it makes no sense for a function marked with ``__init`` to be exported to modules -with :c:func:`EXPORT_SYMBOL()` - this will break. +with :c:func:`EXPORT_SYMBOL()` or :c:func:`EXPORT_SYMBOL_GPL()`- this +will break. -:c:func:`__initcall()`/:c:func:`module_init()` ``include/linux/init.h`` ------------------------------------------------------------------------ +: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 @@ -438,8 +458,11 @@ to fail (unfortunately, this has no effect if the module is compiled into the kernel). This function is called in user context with interrupts enabled, so it can sleep. -:c:func:`module_exit()` ``include/linux/init.h`` ------------------------------------------------- +:c:func:`module_exit()` +----------------------- + + +Defined in ``include/linux/module.h`` This macro defines the function to be called at module removal time (or never, in the case of the file compiled into the kernel). It will only @@ -450,8 +473,10 @@ it returns. Note that this macro is optional: if it is not present, your module will not be removable (except for 'rmmod -f'). -:c:func:`try_module_get()`/:c:func:`module_put()` ``include/linux/module.h`` ----------------------------------------------------------------------------- +: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 (a module also can't be removed if another module uses one of its exported @@ -472,8 +497,8 @@ Wait Queues ``include/linux/wait.h`` A wait queue is used to wait for someone to wake you up when a certain condition is true. They must be used carefully to ensure there is no -race condition. You declare a ``wait_queue_head_t``, and then processes -which want to wait for that condition declare a ``wait_queue_t`` +race condition. You declare a :c:type:`wait_queue_head_t`, and then processes +which want to wait for that condition declare a :c:type:`wait_queue_t` referring to themselves, and place that in the queue. Declaring @@ -490,15 +515,15 @@ Queuing Placing yourself in the waitqueue is fairly complex, because you must put yourself in the queue before checking the condition. There is a macro to do this: :c:func:`wait_event_interruptible()` -``include/linux/wait.h`` The first argument is the wait queue head, and +(``include/linux/wait.h``) The first argument is the wait queue head, and the second is an expression which is evaluated; the macro returns 0 when -this expression is true, or -ERESTARTSYS if a signal is received. The +this expression is true, or ``-ERESTARTSYS`` if a signal is received. The :c:func:`wait_event()` version ignores signals. Waking Up Queued Tasks ---------------------- -Call :c:func:`wake_up()` ``include/linux/wait.h``;, which will wake +Call :c:func:`wake_up()` (``include/linux/wait.h``);, which will wake up every process in the queue. The exception is if one has ``TASK_EXCLUSIVE`` set, in which case the remainder of the queue will not be woken. There are other variants of this basic function available @@ -508,9 +533,9 @@ Atomic Operations ================= Certain operations are guaranteed atomic on all platforms. The first -class of operations work on ``atomic_t`` ``include/asm/atomic.h``; this -contains a signed integer (at least 32 bits long), and you must use -these functions to manipulate or read atomic_t variables. +class of operations work on :c:type:`atomic_t` (``include/asm/atomic.h``); +this contains a signed integer (at least 32 bits long), and you must use +these functions to manipulate or read :c:type:`atomic_t` variables. :c:func:`atomic_read()` and :c:func:`atomic_set()` get and set the counter, :c:func:`atomic_add()`, :c:func:`atomic_sub()`, :c:func:`atomic_inc()`, :c:func:`atomic_dec()`, and @@ -534,7 +559,7 @@ true if the bit was previously set; these are particularly useful for atomically setting flags. It is possible to call these operations with bit indices greater than -BITS_PER_LONG. The resulting behavior is strange on big-endian +``BITS_PER_LONG``. The resulting behavior is strange on big-endian platforms though so it is a good idea not to do this. Symbols @@ -546,14 +571,18 @@ be used anywhere in the kernel). However, for modules, a special exported symbol table is kept which limits the entry points to the kernel proper. Modules can also export symbols. -:c:func:`EXPORT_SYMBOL()` ``include/linux/export.h`` ----------------------------------------------------- +:c:func:`EXPORT_SYMBOL()` +------------------------- + +Defined in ``include/linux/export.h`` This is the classic method of exporting a symbol: dynamically loaded modules will be able to use the symbol as normal. -:c:func:`EXPORT_SYMBOL_GPL()` ``include/linux/export.h`` --------------------------------------------------------- +:c:func:`EXPORT_SYMBOL_GPL()` +----------------------------- + +Defined in ``include/linux/export.h`` Similar to :c:func:`EXPORT_SYMBOL()` except that the symbols exported by :c:func:`EXPORT_SYMBOL_GPL()` can only be seen by @@ -579,11 +608,11 @@ Return Conventions ------------------ For code called in user context, it's very common to defy C convention, -and return 0 for success, and a negative error number (eg. -EFAULT) for +and return 0 for success, and a negative error number (eg. ``-EFAULT``) for failure. This can be unintuitive at first, but it's fairly widespread in the kernel. -Using :c:func:`ERR_PTR()` ``include/linux/err.h``; to encode a +Using :c:func:`ERR_PTR()` (``include/linux/err.h``) to encode a negative error number into a pointer, and :c:func:`IS_ERR()` and :c:func:`PTR_ERR()` to get it back out again: avoids a separate pointer parameter for the error number. Icky, but in a good way. @@ -603,9 +632,7 @@ Initializing structure members ------------------------------ The preferred method of initializing structures is to use designated -initialisers, as defined by ISO C99, eg: - -:: +initialisers, as defined by ISO C99, eg:: static struct block_device_operations opt_fops = { .open = opt_open, @@ -716,18 +743,14 @@ Kernel Cantrips Some favorites from browsing the source. Feel free to add to this list. -``arch/x86/include/asm/delay.h:`` - -:: +``arch/x86/include/asm/delay.h``:: #define ndelay(n) (__builtin_constant_p(n) ? \ ((n) > 20000 ? __bad_ndelay() : __const_udelay((n) * 5ul)) : \ __ndelay(n)) -``include/linux/fs.h``: - -:: +``include/linux/fs.h``:: /* * Kernel pointers have redundant information, so we can use a @@ -741,9 +764,7 @@ Some favorites from browsing the source. Feel free to add to this list. #define PTR_ERR(ptr) ((long)(ptr)) #define IS_ERR(ptr) ((unsigned long)(ptr) > (unsigned long)(-1000)) -``arch/x86/include/asm/uaccess_32.h:`` - -:: +``arch/x86/include/asm/uaccess_32.h:``:: #define copy_to_user(to,from,n) \ (__builtin_constant_p(n) ? \ @@ -751,9 +772,7 @@ Some favorites from browsing the source. Feel free to add to this list. __generic_copy_to_user((to),(from),(n))) -``arch/sparc/kernel/head.S:`` - -:: +``arch/sparc/kernel/head.S:``:: /* * Sun people can't spell worth damn. "compatability" indeed. @@ -772,9 +791,7 @@ Some favorites from browsing the source. Feel free to add to this list. .asciz "compatible" -``arch/sparc/lib/checksum.S:`` - -:: +``arch/sparc/lib/checksum.S:``:: /* Sun, you just can't beat me, you just can't. Stop trying, * give up. I'm serious, I am going to kick the living shit From e548cdeffcd8ab8d3551a539890e682c08ab7828 Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 09:55:30 -0300 Subject: [PATCH 03/90] docs-rst: convert kernel-locking to ReST Use pandoc to convert documentation to ReST by calling Documentation/sphinx/tmplcvt script. - Manually adjust tables with got broken by conversion Signed-off-by: Mauro Carvalho Chehab --- Documentation/DocBook/Makefile | 1 - Documentation/DocBook/kernel-locking.tmpl | 2151 --------------------- Documentation/kernel-hacking/index.rst | 1 + Documentation/kernel-hacking/locking.rst | 1453 ++++++++++++++ 4 files changed, 1454 insertions(+), 2152 deletions(-) delete mode 100644 Documentation/DocBook/kernel-locking.tmpl create mode 100644 Documentation/kernel-hacking/locking.rst diff --git a/Documentation/DocBook/Makefile b/Documentation/DocBook/Makefile index 7d7482b5ad92..9df94f7c2003 100644 --- a/Documentation/DocBook/Makefile +++ b/Documentation/DocBook/Makefile @@ -7,7 +7,6 @@ # list of DOCBOOKS. DOCBOOKS := z8530book.xml \ - kernel-locking.xml \ networking.xml \ filesystems.xml lsm.xml kgdb.xml \ libata.xml mtdnand.xml librs.xml rapidio.xml \ diff --git a/Documentation/DocBook/kernel-locking.tmpl b/Documentation/DocBook/kernel-locking.tmpl deleted file mode 100644 index 7c9cc4846cb6..000000000000 --- a/Documentation/DocBook/kernel-locking.tmpl +++ /dev/null @@ -1,2151 +0,0 @@ - - - - - - Unreliable Guide To Locking - - - - Rusty - Russell - -
- rusty@rustcorp.com.au -
- - - - - - 2003 - Rusty Russell - - - - - This documentation is free software; you can redistribute - it and/or modify it under the terms of the GNU General Public - License as published by the Free Software Foundation; either - version 2 of the License, or (at your option) any later - version. - - - - This program is distributed in the hope that it will be - useful, but WITHOUT ANY WARRANTY; without even the implied - warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. - See the GNU General Public License for more details. - - - - You should have received a copy of the GNU General Public - License along with this program; if not, write to the Free - Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, - MA 02111-1307 USA - - - - For more details see the file COPYING in the source - distribution of Linux. - - - - - - - Introduction - - Welcome, to Rusty's Remarkably Unreliable Guide to Kernel - Locking issues. This document describes the locking systems in - the Linux Kernel in 2.6. - - - With the wide availability of HyperThreading, and preemption in the Linux - Kernel, everyone hacking on the kernel needs to know the - fundamentals of concurrency and locking for - SMP. - - - - - The Problem With Concurrency - - (Skip this if you know what a Race Condition is). - - - In a normal program, you can increment a counter like so: - - - very_important_count++; - - - - This is what they would expect to happen: - - - - Expected Results - - - - - - Instance 1 - Instance 2 - - - - - - read very_important_count (5) - - - - add 1 (6) - - - - write very_important_count (6) - - - - - read very_important_count (6) - - - - add 1 (7) - - - - write very_important_count (7) - - - - -
- - - This is what might happen: - - - - Possible Results - - - - - Instance 1 - Instance 2 - - - - - - read very_important_count (5) - - - - - read very_important_count (5) - - - add 1 (6) - - - - - add 1 (6) - - - write very_important_count (6) - - - - - write very_important_count (6) - - - -
- - - Race Conditions and Critical Regions - - This overlap, where the result depends on the - relative timing of multiple tasks, is called a race condition. - The piece of code containing the concurrency issue is called a - critical region. And especially since Linux starting running - on SMP machines, they became one of the major issues in kernel - design and implementation. - - - Preemption can have the same effect, even if there is only one - CPU: by preempting one task during the critical region, we have - exactly the same race condition. In this case the thread which - preempts might run the critical region itself. - - - The solution is to recognize when these simultaneous accesses - occur, and use locks to make sure that only one instance can - enter the critical region at any time. There are many - friendly primitives in the Linux kernel to help you do this. - And then there are the unfriendly primitives, but I'll pretend - they don't exist. - - - - - - Locking in the Linux Kernel - - - If I could give you one piece of advice: never sleep with anyone - crazier than yourself. But if I had to give you advice on - locking: keep it simple. - - - - Be reluctant to introduce new locks. - - - - Strangely enough, this last one is the exact reverse of my advice when - you have slept with someone crazier than yourself. - And you should think about getting a big dog. - - - - Two Main Types of Kernel Locks: Spinlocks and Mutexes - - - There are two main types of kernel locks. The fundamental type - is the spinlock - (include/asm/spinlock.h), - which is a very simple single-holder lock: if you can't get the - spinlock, you keep trying (spinning) until you can. Spinlocks are - very small and fast, and can be used anywhere. - - - The second type is a mutex - (include/linux/mutex.h): it - is like a spinlock, but you may block holding a mutex. - If you can't lock a mutex, your task will suspend itself, and be woken - up when the mutex is released. This means the CPU can do something - else while you are waiting. There are many cases when you simply - can't sleep (see ), and so have to - use a spinlock instead. - - - Neither type of lock is recursive: see - . - - - - - Locks and Uniprocessor Kernels - - - For kernels compiled without CONFIG_SMP, and - without CONFIG_PREEMPT spinlocks do not exist at - all. This is an excellent design decision: when no-one else can - run at the same time, there is no reason to have a lock. - - - - If the kernel is compiled without CONFIG_SMP, - but CONFIG_PREEMPT is set, then spinlocks - simply disable preemption, which is sufficient to prevent any - races. For most purposes, we can think of preemption as - equivalent to SMP, and not worry about it separately. - - - - You should always test your locking code with CONFIG_SMP - and CONFIG_PREEMPT enabled, even if you don't have an SMP test box, because it - will still catch some kinds of locking bugs. - - - - Mutexes still exist, because they are required for - synchronization between user - contexts, as we will see below. - - - - - Locking Only In User Context - - - If you have a data structure which is only ever accessed from - user context, then you can use a simple mutex - (include/linux/mutex.h) to protect it. This - is the most trivial case: you initialize the mutex. Then you can - call mutex_lock_interruptible() to grab the mutex, - and mutex_unlock() to release it. There is also a - mutex_lock(), which should be avoided, because it - will not return if a signal is received. - - - - Example: net/netfilter/nf_sockopt.c allows - registration of new setsockopt() and - getsockopt() calls, with - nf_register_sockopt(). Registration and - de-registration are only done on module load and unload (and boot - time, where there is no concurrency), and the list of registrations - is only consulted for an unknown setsockopt() - or getsockopt() system call. The - nf_sockopt_mutex is perfect to protect this, - especially since the setsockopt and getsockopt calls may well - sleep. - - - - - Locking Between User Context and Softirqs - - - If a softirq shares - data with user context, you have two problems. Firstly, the current - user context can be interrupted by a softirq, and secondly, the - critical region could be entered from another CPU. This is where - spin_lock_bh() - (include/linux/spinlock.h) is - used. It disables softirqs on that CPU, then grabs the lock. - spin_unlock_bh() does the reverse. (The - '_bh' suffix is a historical reference to "Bottom Halves", the - old name for software interrupts. It should really be - called spin_lock_softirq()' in a perfect world). - - - - Note that you can also use spin_lock_irq() - or spin_lock_irqsave() here, which stop - hardware interrupts as well: see . - - - - This works perfectly for UP - as well: the spin lock vanishes, and this macro - simply becomes local_bh_disable() - (include/linux/interrupt.h), which - protects you from the softirq being run. - - - - - Locking Between User Context and Tasklets - - - This is exactly the same as above, because tasklets are actually run - from a softirq. - - - - - Locking Between User Context and Timers - - - This, too, is exactly the same as above, because timers are actually run from - a softirq. From a locking point of view, tasklets and timers - are identical. - - - - - Locking Between Tasklets/Timers - - - Sometimes a tasklet or timer might want to share data with - another tasklet or timer. - - - - The Same Tasklet/Timer - - Since a tasklet is never run on two CPUs at once, you don't - need to worry about your tasklet being reentrant (running - twice at once), even on SMP. - - - - - Different Tasklets/Timers - - If another tasklet/timer wants - to share data with your tasklet or timer , you will both need to use - spin_lock() and - spin_unlock() calls. - spin_lock_bh() is - unnecessary here, as you are already in a tasklet, and - none will be run on the same CPU. - - - - - - Locking Between Softirqs - - - Often a softirq might - want to share data with itself or a tasklet/timer. - - - - The Same Softirq - - - The same softirq can run on the other CPUs: you can use a - per-CPU array (see ) for better - performance. If you're going so far as to use a softirq, - you probably care about scalable performance enough - to justify the extra complexity. - - - - You'll need to use spin_lock() and - spin_unlock() for shared data. - - - - - Different Softirqs - - - You'll need to use spin_lock() and - spin_unlock() for shared data, whether it - be a timer, tasklet, different softirq or the same or another - softirq: any of them could be running on a different CPU. - - - - - - - Hard IRQ Context - - - Hardware interrupts usually communicate with a - tasklet or softirq. Frequently this involves putting work in a - queue, which the softirq will take out. - - - - Locking Between Hard IRQ and Softirqs/Tasklets - - - If a hardware irq handler shares data with a softirq, you have - two concerns. Firstly, the softirq processing can be - interrupted by a hardware interrupt, and secondly, the - critical region could be entered by a hardware interrupt on - another CPU. This is where spin_lock_irq() is - used. It is defined to disable interrupts on that cpu, then grab - the lock. spin_unlock_irq() does the reverse. - - - - The irq handler does not to use - spin_lock_irq(), because the softirq cannot - run while the irq handler is running: it can use - spin_lock(), which is slightly faster. The - only exception would be if a different hardware irq handler uses - the same lock: spin_lock_irq() will stop - that from interrupting us. - - - - This works perfectly for UP as well: the spin lock vanishes, - and this macro simply becomes local_irq_disable() - (include/asm/smp.h), which - protects you from the softirq/tasklet/BH being run. - - - - spin_lock_irqsave() - (include/linux/spinlock.h) is a variant - which saves whether interrupts were on or off in a flags word, - which is passed to spin_unlock_irqrestore(). This - means that the same code can be used inside an hard irq handler (where - interrupts are already off) and in softirqs (where the irq - disabling is required). - - - - Note that softirqs (and hence tasklets and timers) are run on - return from hardware interrupts, so - spin_lock_irq() also stops these. In that - sense, spin_lock_irqsave() is the most - general and powerful locking function. - - - - - Locking Between Two Hard IRQ Handlers - - It is rare to have to share data between two IRQ handlers, but - if you do, spin_lock_irqsave() should be - used: it is architecture-specific whether all interrupts are - disabled inside irq handlers themselves. - - - - - - - Cheat Sheet For Locking - - Pete Zaitcev gives the following summary: - - - - - If you are in a process context (any syscall) and want to - lock other process out, use a mutex. You can take a mutex - and sleep (copy_from_user*( or - kmalloc(x,GFP_KERNEL)). - - - - - Otherwise (== data can be touched in an interrupt), use - spin_lock_irqsave() and - spin_unlock_irqrestore(). - - - - - Avoid holding spinlock for more than 5 lines of code and - across any function call (except accessors like - readb). - - - - - - Table of Minimum Requirements - - The following table lists the minimum - locking requirements between various contexts. In some cases, - the same context can only be running on one CPU at a time, so - no locking is required for that context (eg. a particular - thread can only run on one CPU at a time, but if it needs - shares data with another thread, locking is required). - - - Remember the advice above: you can always use - spin_lock_irqsave(), which is a superset - of all other spinlock primitives. - - - -Table of Locking Requirements - - - - - -IRQ Handler A -IRQ Handler B -Softirq A -Softirq B -Tasklet A -Tasklet B -Timer A -Timer B -User Context A -User Context B - - - -IRQ Handler A -None - - - -IRQ Handler B -SLIS -None - - - -Softirq A -SLI -SLI -SL - - - -Softirq B -SLI -SLI -SL -SL - - - -Tasklet A -SLI -SLI -SL -SL -None - - - -Tasklet B -SLI -SLI -SL -SL -SL -None - - - -Timer A -SLI -SLI -SL -SL -SL -SL -None - - - -Timer B -SLI -SLI -SL -SL -SL -SL -SL -None - - - -User Context A -SLI -SLI -SLBH -SLBH -SLBH -SLBH -SLBH -SLBH -None - - - -User Context B -SLI -SLI -SLBH -SLBH -SLBH -SLBH -SLBH -SLBH -MLI -None - - - - -
- - -Legend for Locking Requirements Table - - - - -SLIS -spin_lock_irqsave - - -SLI -spin_lock_irq - - -SL -spin_lock - - -SLBH -spin_lock_bh - - -MLI -mutex_lock_interruptible - - - - -
- - - - - - The trylock Functions - - There are functions that try to acquire a lock only once and immediately - return a value telling about success or failure to acquire the lock. - They can be used if you need no access to the data protected with the lock - when some other thread is holding the lock. You should acquire the lock - later if you then need access to the data protected with the lock. - - - - spin_trylock() does not spin but returns non-zero if - it acquires the spinlock on the first try or 0 if not. This function can - be used in all contexts like spin_lock: you must have - disabled the contexts that might interrupt you and acquire the spin lock. - - - - mutex_trylock() does not suspend your task - but returns non-zero if it could lock the mutex on the first try - or 0 if not. This function cannot be safely used in hardware or software - interrupt contexts despite not sleeping. - - - - - Common Examples - -Let's step through a simple example: a cache of number to name -mappings. The cache keeps a count of how often each of the objects is -used, and when it gets full, throws out the least used one. - - - - - All In User Context - -For our first example, we assume that all operations are in user -context (ie. from system calls), so we can sleep. This means we can -use a mutex to protect the cache and all the objects within -it. Here's the code: - - - -#include <linux/list.h> -#include <linux/slab.h> -#include <linux/string.h> -#include <linux/mutex.h> -#include <asm/errno.h> - -struct object -{ - struct list_head list; - int id; - char name[32]; - int popularity; -}; - -/* Protects the cache, cache_num, and the objects within it */ -static DEFINE_MUTEX(cache_lock); -static LIST_HEAD(cache); -static unsigned int cache_num = 0; -#define MAX_CACHE_SIZE 10 - -/* Must be holding cache_lock */ -static struct object *__cache_find(int id) -{ - struct object *i; - - list_for_each_entry(i, &cache, list) - if (i->id == id) { - i->popularity++; - return i; - } - return NULL; -} - -/* Must be holding cache_lock */ -static void __cache_delete(struct object *obj) -{ - BUG_ON(!obj); - list_del(&obj->list); - kfree(obj); - cache_num--; -} - -/* Must be holding cache_lock */ -static void __cache_add(struct object *obj) -{ - list_add(&obj->list, &cache); - if (++cache_num > MAX_CACHE_SIZE) { - struct object *i, *outcast = NULL; - list_for_each_entry(i, &cache, list) { - if (!outcast || i->popularity < outcast->popularity) - outcast = i; - } - __cache_delete(outcast); - } -} - -int cache_add(int id, const char *name) -{ - struct object *obj; - - if ((obj = kmalloc(sizeof(*obj), GFP_KERNEL)) == NULL) - return -ENOMEM; - - strlcpy(obj->name, name, sizeof(obj->name)); - obj->id = id; - obj->popularity = 0; - - mutex_lock(&cache_lock); - __cache_add(obj); - mutex_unlock(&cache_lock); - return 0; -} - -void cache_delete(int id) -{ - mutex_lock(&cache_lock); - __cache_delete(__cache_find(id)); - mutex_unlock(&cache_lock); -} - -int cache_find(int id, char *name) -{ - struct object *obj; - int ret = -ENOENT; - - mutex_lock(&cache_lock); - obj = __cache_find(id); - if (obj) { - ret = 0; - strcpy(name, obj->name); - } - mutex_unlock(&cache_lock); - return ret; -} - - - -Note that we always make sure we have the cache_lock when we add, -delete, or look up the cache: both the cache infrastructure itself and -the contents of the objects are protected by the lock. In this case -it's easy, since we copy the data for the user, and never let them -access the objects directly. - - -There is a slight (and common) optimization here: in -cache_add we set up the fields of the object -before grabbing the lock. This is safe, as no-one else can access it -until we put it in cache. - - - - - Accessing From Interrupt Context - -Now consider the case where cache_find can be -called from interrupt context: either a hardware interrupt or a -softirq. An example would be a timer which deletes object from the -cache. - - -The change is shown below, in standard patch format: the -- are lines which are taken away, and the -+ are lines which are added. - - ---- cache.c.usercontext 2003-12-09 13:58:54.000000000 +1100 -+++ cache.c.interrupt 2003-12-09 14:07:49.000000000 +1100 -@@ -12,7 +12,7 @@ - int popularity; - }; - --static DEFINE_MUTEX(cache_lock); -+static DEFINE_SPINLOCK(cache_lock); - static LIST_HEAD(cache); - static unsigned int cache_num = 0; - #define MAX_CACHE_SIZE 10 -@@ -55,6 +55,7 @@ - int cache_add(int id, const char *name) - { - struct object *obj; -+ unsigned long flags; - - if ((obj = kmalloc(sizeof(*obj), GFP_KERNEL)) == NULL) - return -ENOMEM; -@@ -63,30 +64,33 @@ - obj->id = id; - obj->popularity = 0; - -- mutex_lock(&cache_lock); -+ spin_lock_irqsave(&cache_lock, flags); - __cache_add(obj); -- mutex_unlock(&cache_lock); -+ spin_unlock_irqrestore(&cache_lock, flags); - return 0; - } - - void cache_delete(int id) - { -- mutex_lock(&cache_lock); -+ unsigned long flags; -+ -+ spin_lock_irqsave(&cache_lock, flags); - __cache_delete(__cache_find(id)); -- mutex_unlock(&cache_lock); -+ spin_unlock_irqrestore(&cache_lock, flags); - } - - int cache_find(int id, char *name) - { - struct object *obj; - int ret = -ENOENT; -+ unsigned long flags; - -- mutex_lock(&cache_lock); -+ spin_lock_irqsave(&cache_lock, flags); - obj = __cache_find(id); - if (obj) { - ret = 0; - strcpy(name, obj->name); - } -- mutex_unlock(&cache_lock); -+ spin_unlock_irqrestore(&cache_lock, flags); - return ret; - } - - - -Note that the spin_lock_irqsave will turn off -interrupts if they are on, otherwise does nothing (if we are already -in an interrupt handler), hence these functions are safe to call from -any context. - - -Unfortunately, cache_add calls -kmalloc with the GFP_KERNEL -flag, which is only legal in user context. I have assumed that -cache_add is still only called in user context, -otherwise this should become a parameter to -cache_add. - - - - Exposing Objects Outside This File - -If our objects contained more information, it might not be sufficient -to copy the information in and out: other parts of the code might want -to keep pointers to these objects, for example, rather than looking up -the id every time. This produces two problems. - - -The first problem is that we use the cache_lock to -protect objects: we'd need to make this non-static so the rest of the -code can use it. This makes locking trickier, as it is no longer all -in one place. - - -The second problem is the lifetime problem: if another structure keeps -a pointer to an object, it presumably expects that pointer to remain -valid. Unfortunately, this is only guaranteed while you hold the -lock, otherwise someone might call cache_delete -and even worse, add another object, re-using the same address. - - -As there is only one lock, you can't hold it forever: no-one else would -get any work done. - - -The solution to this problem is to use a reference count: everyone who -has a pointer to the object increases it when they first get the -object, and drops the reference count when they're finished with it. -Whoever drops it to zero knows it is unused, and can actually delete it. - - -Here is the code: - - - ---- cache.c.interrupt 2003-12-09 14:25:43.000000000 +1100 -+++ cache.c.refcnt 2003-12-09 14:33:05.000000000 +1100 -@@ -7,6 +7,7 @@ - struct object - { - struct list_head list; -+ unsigned int refcnt; - int id; - char name[32]; - int popularity; -@@ -17,6 +18,35 @@ - static unsigned int cache_num = 0; - #define MAX_CACHE_SIZE 10 - -+static void __object_put(struct object *obj) -+{ -+ if (--obj->refcnt == 0) -+ kfree(obj); -+} -+ -+static void __object_get(struct object *obj) -+{ -+ obj->refcnt++; -+} -+ -+void object_put(struct object *obj) -+{ -+ unsigned long flags; -+ -+ spin_lock_irqsave(&cache_lock, flags); -+ __object_put(obj); -+ spin_unlock_irqrestore(&cache_lock, flags); -+} -+ -+void object_get(struct object *obj) -+{ -+ unsigned long flags; -+ -+ spin_lock_irqsave(&cache_lock, flags); -+ __object_get(obj); -+ spin_unlock_irqrestore(&cache_lock, flags); -+} -+ - /* Must be holding cache_lock */ - static struct object *__cache_find(int id) - { -@@ -35,6 +65,7 @@ - { - BUG_ON(!obj); - list_del(&obj->list); -+ __object_put(obj); - cache_num--; - } - -@@ -63,6 +94,7 @@ - strlcpy(obj->name, name, sizeof(obj->name)); - obj->id = id; - obj->popularity = 0; -+ obj->refcnt = 1; /* The cache holds a reference */ - - spin_lock_irqsave(&cache_lock, flags); - __cache_add(obj); -@@ -79,18 +111,15 @@ - spin_unlock_irqrestore(&cache_lock, flags); - } - --int cache_find(int id, char *name) -+struct object *cache_find(int id) - { - struct object *obj; -- int ret = -ENOENT; - unsigned long flags; - - spin_lock_irqsave(&cache_lock, flags); - obj = __cache_find(id); -- if (obj) { -- ret = 0; -- strcpy(name, obj->name); -- } -+ if (obj) -+ __object_get(obj); - spin_unlock_irqrestore(&cache_lock, flags); -- return ret; -+ return obj; - } - - - -We encapsulate the reference counting in the standard 'get' and 'put' -functions. Now we can return the object itself from -cache_find which has the advantage that the user -can now sleep holding the object (eg. to -copy_to_user to name to userspace). - - -The other point to note is that I said a reference should be held for -every pointer to the object: thus the reference count is 1 when first -inserted into the cache. In some versions the framework does not hold -a reference count, but they are more complicated. - - - - Using Atomic Operations For The Reference Count - -In practice, atomic_t would usually be used for -refcnt. There are a number of atomic -operations defined in - -include/asm/atomic.h: these are -guaranteed to be seen atomically from all CPUs in the system, so no -lock is required. In this case, it is simpler than using spinlocks, -although for anything non-trivial using spinlocks is clearer. The -atomic_inc and -atomic_dec_and_test are used instead of the -standard increment and decrement operators, and the lock is no longer -used to protect the reference count itself. - - - ---- cache.c.refcnt 2003-12-09 15:00:35.000000000 +1100 -+++ cache.c.refcnt-atomic 2003-12-11 15:49:42.000000000 +1100 -@@ -7,7 +7,7 @@ - struct object - { - struct list_head list; -- unsigned int refcnt; -+ atomic_t refcnt; - int id; - char name[32]; - int popularity; -@@ -18,33 +18,15 @@ - static unsigned int cache_num = 0; - #define MAX_CACHE_SIZE 10 - --static void __object_put(struct object *obj) --{ -- if (--obj->refcnt == 0) -- kfree(obj); --} -- --static void __object_get(struct object *obj) --{ -- obj->refcnt++; --} -- - void object_put(struct object *obj) - { -- unsigned long flags; -- -- spin_lock_irqsave(&cache_lock, flags); -- __object_put(obj); -- spin_unlock_irqrestore(&cache_lock, flags); -+ if (atomic_dec_and_test(&obj->refcnt)) -+ kfree(obj); - } - - void object_get(struct object *obj) - { -- unsigned long flags; -- -- spin_lock_irqsave(&cache_lock, flags); -- __object_get(obj); -- spin_unlock_irqrestore(&cache_lock, flags); -+ atomic_inc(&obj->refcnt); - } - - /* Must be holding cache_lock */ -@@ -65,7 +47,7 @@ - { - BUG_ON(!obj); - list_del(&obj->list); -- __object_put(obj); -+ object_put(obj); - cache_num--; - } - -@@ -94,7 +76,7 @@ - strlcpy(obj->name, name, sizeof(obj->name)); - obj->id = id; - obj->popularity = 0; -- obj->refcnt = 1; /* The cache holds a reference */ -+ atomic_set(&obj->refcnt, 1); /* The cache holds a reference */ - - spin_lock_irqsave(&cache_lock, flags); - __cache_add(obj); -@@ -119,7 +101,7 @@ - spin_lock_irqsave(&cache_lock, flags); - obj = __cache_find(id); - if (obj) -- __object_get(obj); -+ object_get(obj); - spin_unlock_irqrestore(&cache_lock, flags); - return obj; - } - - - - - - Protecting The Objects Themselves - -In these examples, we assumed that the objects (except the reference -counts) never changed once they are created. If we wanted to allow -the name to change, there are three possibilities: - - - - -You can make cache_lock non-static, and tell people -to grab that lock before changing the name in any object. - - - - -You can provide a cache_obj_rename which grabs -this lock and changes the name for the caller, and tell everyone to -use that function. - - - - -You can make the cache_lock protect only the cache -itself, and use another lock to protect the name. - - - - - -Theoretically, you can make the locks as fine-grained as one lock for -every field, for every object. In practice, the most common variants -are: - - - - -One lock which protects the infrastructure (the cache -list in this example) and all the objects. This is what we have done -so far. - - - - -One lock which protects the infrastructure (including the list -pointers inside the objects), and one lock inside the object which -protects the rest of that object. - - - - -Multiple locks to protect the infrastructure (eg. one lock per hash -chain), possibly with a separate per-object lock. - - - - - -Here is the "lock-per-object" implementation: - - ---- cache.c.refcnt-atomic 2003-12-11 15:50:54.000000000 +1100 -+++ cache.c.perobjectlock 2003-12-11 17:15:03.000000000 +1100 -@@ -6,11 +6,17 @@ - - struct object - { -+ /* These two protected by cache_lock. */ - struct list_head list; -+ int popularity; -+ - atomic_t refcnt; -+ -+ /* Doesn't change once created. */ - int id; -+ -+ spinlock_t lock; /* Protects the name */ - char name[32]; -- int popularity; - }; - - static DEFINE_SPINLOCK(cache_lock); -@@ -77,6 +84,7 @@ - obj->id = id; - obj->popularity = 0; - atomic_set(&obj->refcnt, 1); /* The cache holds a reference */ -+ spin_lock_init(&obj->lock); - - spin_lock_irqsave(&cache_lock, flags); - __cache_add(obj); - - - -Note that I decide that the popularity -count should be protected by the cache_lock rather -than the per-object lock: this is because it (like the -struct list_head inside the object) is -logically part of the infrastructure. This way, I don't need to grab -the lock of every object in __cache_add when -seeking the least popular. - - - -I also decided that the id member is -unchangeable, so I don't need to grab each object lock in -__cache_find() to examine the -id: the object lock is only used by a -caller who wants to read or write the name -field. - - - -Note also that I added a comment describing what data was protected by -which locks. This is extremely important, as it describes the runtime -behavior of the code, and can be hard to gain from just reading. And -as Alan Cox says, Lock data, not code. - - - - - - Common Problems - - Deadlock: Simple and Advanced - - - There is a coding bug where a piece of code tries to grab a - spinlock twice: it will spin forever, waiting for the lock to - be released (spinlocks, rwlocks and mutexes are not - recursive in Linux). This is trivial to diagnose: not a - stay-up-five-nights-talk-to-fluffy-code-bunnies kind of - problem. - - - - For a slightly more complex case, imagine you have a region - shared by a softirq and user context. If you use a - spin_lock() call to protect it, it is - possible that the user context will be interrupted by the softirq - while it holds the lock, and the softirq will then spin - forever trying to get the same lock. - - - - Both of these are called deadlock, and as shown above, it can - occur even with a single CPU (although not on UP compiles, - since spinlocks vanish on kernel compiles with - CONFIG_SMP=n. You'll still get data corruption - in the second example). - - - - This complete lockup is easy to diagnose: on SMP boxes the - watchdog timer or compiling with DEBUG_SPINLOCK set - (include/linux/spinlock.h) will show this up - immediately when it happens. - - - - A more complex problem is the so-called 'deadly embrace', - involving two or more locks. Say you have a hash table: each - entry in the table is a spinlock, and a chain of hashed - objects. Inside a softirq handler, you sometimes want to - alter an object from one place in the hash to another: you - grab the spinlock of the old hash chain and the spinlock of - the new hash chain, and delete the object from the old one, - and insert it in the new one. - - - - There are two problems here. First, if your code ever - tries to move the object to the same chain, it will deadlock - with itself as it tries to lock it twice. Secondly, if the - same softirq on another CPU is trying to move another object - in the reverse direction, the following could happen: - - - - Consequences - - - - - - CPU 1 - CPU 2 - - - - - - Grab lock A -> OK - Grab lock B -> OK - - - Grab lock B -> spin - Grab lock A -> spin - - - -
- - - The two CPUs will spin forever, waiting for the other to give up - their lock. It will look, smell, and feel like a crash. - - - - - Preventing Deadlock - - - Textbooks will tell you that if you always lock in the same - order, you will never get this kind of deadlock. Practice - will tell you that this approach doesn't scale: when I - create a new lock, I don't understand enough of the kernel - to figure out where in the 5000 lock hierarchy it will fit. - - - - The best locks are encapsulated: they never get exposed in - headers, and are never held around calls to non-trivial - functions outside the same file. You can read through this - code and see that it will never deadlock, because it never - tries to grab another lock while it has that one. People - using your code don't even need to know you are using a - lock. - - - - A classic problem here is when you provide callbacks or - hooks: if you call these with the lock held, you risk simple - deadlock, or a deadly embrace (who knows what the callback - will do?). Remember, the other programmers are out to get - you, so don't do this. - - - - Overzealous Prevention Of Deadlocks - - - Deadlocks are problematic, but not as bad as data - corruption. Code which grabs a read lock, searches a list, - fails to find what it wants, drops the read lock, grabs a - write lock and inserts the object has a race condition. - - - - If you don't see why, please stay the fuck away from my code. - - - - - - Racing Timers: A Kernel Pastime - - - Timers can produce their own special problems with races. - Consider a collection of objects (list, hash, etc) where each - object has a timer which is due to destroy it. - - - - If you want to destroy the entire collection (say on module - removal), you might do the following: - - - - /* THIS CODE BAD BAD BAD BAD: IF IT WAS ANY WORSE IT WOULD USE - HUNGARIAN NOTATION */ - spin_lock_bh(&list_lock); - - while (list) { - struct foo *next = list->next; - del_timer(&list->timer); - kfree(list); - list = next; - } - - spin_unlock_bh(&list_lock); - - - - Sooner or later, this will crash on SMP, because a timer can - have just gone off before the spin_lock_bh(), - and it will only get the lock after we - spin_unlock_bh(), and then try to free - the element (which has already been freed!). - - - - This can be avoided by checking the result of - del_timer(): if it returns - 1, the timer has been deleted. - If 0, it means (in this - case) that it is currently running, so we can do: - - - - retry: - spin_lock_bh(&list_lock); - - while (list) { - struct foo *next = list->next; - if (!del_timer(&list->timer)) { - /* Give timer a chance to delete this */ - spin_unlock_bh(&list_lock); - goto retry; - } - kfree(list); - list = next; - } - - spin_unlock_bh(&list_lock); - - - - Another common problem is deleting timers which restart - themselves (by calling add_timer() at the end - of their timer function). Because this is a fairly common case - which is prone to races, you should use del_timer_sync() - (include/linux/timer.h) - to handle this case. It returns the number of times the timer - had to be deleted before we finally stopped it from adding itself back - in. - - - - - - - Locking Speed - - -There are three main things to worry about when considering speed of -some code which does locking. First is concurrency: how many things -are going to be waiting while someone else is holding a lock. Second -is the time taken to actually acquire and release an uncontended lock. -Third is using fewer, or smarter locks. I'm assuming that the lock is -used fairly often: otherwise, you wouldn't be concerned about -efficiency. - - -Concurrency depends on how long the lock is usually held: you should -hold the lock for as long as needed, but no longer. In the cache -example, we always create the object without the lock held, and then -grab the lock only when we are ready to insert it in the list. - - -Acquisition times depend on how much damage the lock operations do to -the pipeline (pipeline stalls) and how likely it is that this CPU was -the last one to grab the lock (ie. is the lock cache-hot for this -CPU): on a machine with more CPUs, this likelihood drops fast. -Consider a 700MHz Intel Pentium III: an instruction takes about 0.7ns, -an atomic increment takes about 58ns, a lock which is cache-hot on -this CPU takes 160ns, and a cacheline transfer from another CPU takes -an additional 170 to 360ns. (These figures from Paul McKenney's - Linux -Journal RCU article). - - -These two aims conflict: holding a lock for a short time might be done -by splitting locks into parts (such as in our final per-object-lock -example), but this increases the number of lock acquisitions, and the -results are often slower than having a single lock. This is another -reason to advocate locking simplicity. - - -The third concern is addressed below: there are some methods to reduce -the amount of locking which needs to be done. - - - - Read/Write Lock Variants - - - Both spinlocks and mutexes have read/write variants: - rwlock_t and struct rw_semaphore. - These divide users into two classes: the readers and the writers. If - you are only reading the data, you can get a read lock, but to write to - the data you need the write lock. Many people can hold a read lock, - but a writer must be sole holder. - - - - If your code divides neatly along reader/writer lines (as our - cache code does), and the lock is held by readers for - significant lengths of time, using these locks can help. They - are slightly slower than the normal locks though, so in practice - rwlock_t is not usually worthwhile. - - - - - Avoiding Locks: Read Copy Update - - - There is a special method of read/write locking called Read Copy - Update. Using RCU, the readers can avoid taking a lock - altogether: as we expect our cache to be read more often than - updated (otherwise the cache is a waste of time), it is a - candidate for this optimization. - - - - How do we get rid of read locks? Getting rid of read locks - means that writers may be changing the list underneath the - readers. That is actually quite simple: we can read a linked - list while an element is being added if the writer adds the - element very carefully. For example, adding - new to a single linked list called - list: - - - - new->next = list->next; - wmb(); - list->next = new; - - - - The wmb() is a write memory barrier. It - ensures that the first operation (setting the new element's - next pointer) is complete and will be seen by - all CPUs, before the second operation is (putting the new - element into the list). This is important, since modern - compilers and modern CPUs can both reorder instructions unless - told otherwise: we want a reader to either not see the new - element at all, or see the new element with the - next pointer correctly pointing at the rest of - the list. - - - Fortunately, there is a function to do this for standard - struct list_head lists: - list_add_rcu() - (include/linux/list.h). - - - Removing an element from the list is even simpler: we replace - the pointer to the old element with a pointer to its successor, - and readers will either see it, or skip over it. - - - list->next = old->next; - - - There is list_del_rcu() - (include/linux/list.h) which does this (the - normal version poisons the old object, which we don't want). - - - The reader must also be careful: some CPUs can look through the - next pointer to start reading the contents of - the next element early, but don't realize that the pre-fetched - contents is wrong when the next pointer changes - underneath them. Once again, there is a - list_for_each_entry_rcu() - (include/linux/list.h) to help you. Of - course, writers can just use - list_for_each_entry(), since there cannot - be two simultaneous writers. - - - Our final dilemma is this: when can we actually destroy the - removed element? Remember, a reader might be stepping through - this element in the list right now: if we free this element and - the next pointer changes, the reader will jump - off into garbage and crash. We need to wait until we know that - all the readers who were traversing the list when we deleted the - element are finished. We use call_rcu() to - register a callback which will actually destroy the object once - all pre-existing readers are finished. Alternatively, - synchronize_rcu() may be used to block until - all pre-existing are finished. - - - But how does Read Copy Update know when the readers are - finished? The method is this: firstly, the readers always - traverse the list inside - rcu_read_lock()/rcu_read_unlock() - pairs: these simply disable preemption so the reader won't go to - sleep while reading the list. - - - RCU then waits until every other CPU has slept at least once: - since readers cannot sleep, we know that any readers which were - traversing the list during the deletion are finished, and the - callback is triggered. The real Read Copy Update code is a - little more optimized than this, but this is the fundamental - idea. - - - ---- cache.c.perobjectlock 2003-12-11 17:15:03.000000000 +1100 -+++ cache.c.rcupdate 2003-12-11 17:55:14.000000000 +1100 -@@ -1,15 +1,18 @@ - #include <linux/list.h> - #include <linux/slab.h> - #include <linux/string.h> -+#include <linux/rcupdate.h> - #include <linux/mutex.h> - #include <asm/errno.h> - - struct object - { -- /* These two protected by cache_lock. */ -+ /* This is protected by RCU */ - struct list_head list; - int popularity; - -+ struct rcu_head rcu; -+ - atomic_t refcnt; - - /* Doesn't change once created. */ -@@ -40,7 +43,7 @@ - { - struct object *i; - -- list_for_each_entry(i, &cache, list) { -+ list_for_each_entry_rcu(i, &cache, list) { - if (i->id == id) { - i->popularity++; - return i; -@@ -49,19 +52,25 @@ - return NULL; - } - -+/* Final discard done once we know no readers are looking. */ -+static void cache_delete_rcu(void *arg) -+{ -+ object_put(arg); -+} -+ - /* Must be holding cache_lock */ - static void __cache_delete(struct object *obj) - { - BUG_ON(!obj); -- list_del(&obj->list); -- object_put(obj); -+ list_del_rcu(&obj->list); - cache_num--; -+ call_rcu(&obj->rcu, cache_delete_rcu); - } - - /* Must be holding cache_lock */ - static void __cache_add(struct object *obj) - { -- list_add(&obj->list, &cache); -+ list_add_rcu(&obj->list, &cache); - if (++cache_num > MAX_CACHE_SIZE) { - struct object *i, *outcast = NULL; - list_for_each_entry(i, &cache, list) { -@@ -104,12 +114,11 @@ - struct object *cache_find(int id) - { - struct object *obj; -- unsigned long flags; - -- spin_lock_irqsave(&cache_lock, flags); -+ rcu_read_lock(); - obj = __cache_find(id); - if (obj) - object_get(obj); -- spin_unlock_irqrestore(&cache_lock, flags); -+ rcu_read_unlock(); - return obj; - } - - - -Note that the reader will alter the -popularity member in -__cache_find(), and now it doesn't hold a lock. -One solution would be to make it an atomic_t, but for -this usage, we don't really care about races: an approximate result is -good enough, so I didn't change it. - - - -The result is that cache_find() requires no -synchronization with any other functions, so is almost as fast on SMP -as it would be on UP. - - - -There is a further optimization possible here: remember our original -cache code, where there were no reference counts and the caller simply -held the lock whenever using the object? This is still possible: if -you hold the lock, no one can delete the object, so you don't need to -get and put the reference count. - - - -Now, because the 'read lock' in RCU is simply disabling preemption, a -caller which always has preemption disabled between calling -cache_find() and -object_put() does not need to actually get and -put the reference count: we could expose -__cache_find() by making it non-static, and -such callers could simply call that. - - -The benefit here is that the reference count is not written to: the -object is not altered in any way, which is much faster on SMP -machines due to caching. - - - - - Per-CPU Data - - - Another technique for avoiding locking which is used fairly - widely is to duplicate information for each CPU. For example, - if you wanted to keep a count of a common condition, you could - use a spin lock and a single counter. Nice and simple. - - - - If that was too slow (it's usually not, but if you've got a - really big machine to test on and can show that it is), you - could instead use a counter for each CPU, then none of them need - an exclusive lock. See DEFINE_PER_CPU(), - get_cpu_var() and - put_cpu_var() - (include/linux/percpu.h). - - - - Of particular use for simple per-cpu counters is the - local_t type, and the - cpu_local_inc() and related functions, - which are more efficient than simple code on some architectures - (include/asm/local.h). - - - - Note that there is no simple, reliable way of getting an exact - value of such a counter, without introducing more locks. This - is not a problem for some uses. - - - - - Data Which Mostly Used By An IRQ Handler - - - If data is always accessed from within the same IRQ handler, you - don't need a lock at all: the kernel already guarantees that the - irq handler will not run simultaneously on multiple CPUs. - - - Manfred Spraul points out that you can still do this, even if - the data is very occasionally accessed in user context or - softirqs/tasklets. The irq handler doesn't use a lock, and - all other accesses are done as so: - - - - spin_lock(&lock); - disable_irq(irq); - ... - enable_irq(irq); - spin_unlock(&lock); - - - The disable_irq() prevents the irq handler - from running (and waits for it to finish if it's currently - running on other CPUs). The spinlock prevents any other - accesses happening at the same time. Naturally, this is slower - than just a spin_lock_irq() call, so it - only makes sense if this type of access happens extremely - rarely. - - - - - - What Functions Are Safe To Call From Interrupts? - - - Many functions in the kernel sleep (ie. call schedule()) - directly or indirectly: you can never call them while holding a - spinlock, or with preemption disabled. This also means you need - to be in user context: calling them from an interrupt is illegal. - - - - Some Functions Which Sleep - - - The most common ones are listed below, but you usually have to - read the code to find out if other calls are safe. If everyone - else who calls it can sleep, you probably need to be able to - sleep, too. In particular, registration and deregistration - functions usually expect to be called from user context, and can - sleep. - - - - - - Accesses to - userspace: - - - - - copy_from_user() - - - - - copy_to_user() - - - - - get_user() - - - - - put_user() - - - - - - - - kmalloc(GFP_KERNEL) - - - - - - mutex_lock_interruptible() and - mutex_lock() - - - There is a mutex_trylock() which does not - sleep. Still, it must not be used inside interrupt context since - its implementation is not safe for that. - mutex_unlock() will also never sleep. - It cannot be used in interrupt context either since a mutex - must be released by the same task that acquired it. - - - - - - - Some Functions Which Don't Sleep - - - Some functions are safe to call from any context, or holding - almost any lock. - - - - - - printk() - - - - - kfree() - - - - - add_timer() and del_timer() - - - - - - - - Mutex API reference -!Iinclude/linux/mutex.h -!Ekernel/locking/mutex.c - - - - Futex API reference -!Ikernel/futex.c - - - - Further reading - - - - - Documentation/locking/spinlocks.txt: - Linus Torvalds' spinlocking tutorial in the kernel sources. - - - - - - Unix Systems for Modern Architectures: Symmetric - Multiprocessing and Caching for Kernel Programmers: - - - - Curt Schimmel's very good introduction to kernel level - locking (not written for Linux, but nearly everything - applies). The book is expensive, but really worth every - penny to understand SMP locking. [ISBN: 0201633388] - - - - - - - Thanks - - - Thanks to Telsa Gwynne for DocBooking, neatening and adding - style. - - - - Thanks to Martin Pool, Philipp Rumpf, Stephen Rothwell, Paul - Mackerras, Ruedi Aschwanden, Alan Cox, Manfred Spraul, Tim - Waugh, Pete Zaitcev, James Morris, Robert Love, Paul McKenney, - John Ashby for proofreading, correcting, flaming, commenting. - - - - Thanks to the cabal for having no influence on this document. - - - - - Glossary - - - preemption - - - Prior to 2.5, or when CONFIG_PREEMPT is - unset, processes in user context inside the kernel would not - preempt each other (ie. you had that CPU until you gave it up, - except for interrupts). With the addition of - CONFIG_PREEMPT in 2.5.4, this changed: when - in user context, higher priority tasks can "cut in": spinlocks - were changed to disable preemption, even on UP. - - - - - - bh - - - Bottom Half: for historical reasons, functions with - '_bh' in them often now refer to any software interrupt, e.g. - spin_lock_bh() blocks any software interrupt - on the current CPU. Bottom halves are deprecated, and will - eventually be replaced by tasklets. Only one bottom half will be - running at any time. - - - - - - Hardware Interrupt / Hardware IRQ - - - Hardware interrupt request. in_irq() returns - true in a hardware interrupt handler. - - - - - - Interrupt Context - - - Not user context: processing a hardware irq or software irq. - Indicated by the in_interrupt() macro - returning true. - - - - - - SMP - - - Symmetric Multi-Processor: kernels compiled for multiple-CPU - machines. (CONFIG_SMP=y). - - - - - - Software Interrupt / softirq - - - Software interrupt handler. in_irq() returns - false; in_softirq() - returns true. Tasklets and softirqs - both fall into the category of 'software interrupts'. - - - Strictly speaking a softirq is one of up to 32 enumerated software - interrupts which can run on multiple CPUs at once. - Sometimes used to refer to tasklets as - well (ie. all software interrupts). - - - - - - tasklet - - - A dynamically-registrable software interrupt, - which is guaranteed to only run on one CPU at a time. - - - - - - timer - - - A dynamically-registrable software interrupt, which is run at - (or close to) a given time. When running, it is just like a - tasklet (in fact, they are called from the TIMER_SOFTIRQ). - - - - - - UP - - - Uni-Processor: Non-SMP. (CONFIG_SMP=n). - - - - - - User Context - - - The kernel executing on behalf of a particular process (ie. a - system call or trap) or kernel thread. You can tell which - process with the current macro.) Not to - be confused with userspace. Can be interrupted by software or - hardware interrupts. - - - - - - Userspace - - - A process executing its own code outside the kernel. - - - - - - - diff --git a/Documentation/kernel-hacking/index.rst b/Documentation/kernel-hacking/index.rst index ba208bf03ce9..fcb0eda3cca3 100644 --- a/Documentation/kernel-hacking/index.rst +++ b/Documentation/kernel-hacking/index.rst @@ -6,3 +6,4 @@ Kernel Hacking Guides :maxdepth: 2 hacking + locking diff --git a/Documentation/kernel-hacking/locking.rst b/Documentation/kernel-hacking/locking.rst new file mode 100644 index 000000000000..976b2703df75 --- /dev/null +++ b/Documentation/kernel-hacking/locking.rst @@ -0,0 +1,1453 @@ +=========================== +Unreliable Guide To Locking +=========================== + +:Author: Rusty Russell + +Introduction +============ + +Welcome, to Rusty's Remarkably Unreliable Guide to Kernel Locking +issues. This document describes the locking systems in the Linux Kernel +in 2.6. + +With the wide availability of HyperThreading, and preemption in the +Linux Kernel, everyone hacking on the kernel needs to know the +fundamentals of concurrency and locking for SMP. + +The Problem With Concurrency +============================ + +(Skip this if you know what a Race Condition is). + +In a normal program, you can increment a counter like so: + +:: + + very_important_count++; + + +This is what they would expect to happen: + ++------------------------------------+------------------------------------+ +| Instance 1 | Instance 2 | ++====================================+====================================+ +| read very_important_count (5) | | ++------------------------------------+------------------------------------+ +| add 1 (6) | | ++------------------------------------+------------------------------------+ +| write very_important_count (6) | | ++------------------------------------+------------------------------------+ +| | read very_important_count (6) | ++------------------------------------+------------------------------------+ +| | add 1 (7) | ++------------------------------------+------------------------------------+ +| | write very_important_count (7) | ++------------------------------------+------------------------------------+ + +Table: Expected Results + +This is what might happen: + ++------------------------------------+------------------------------------+ +| Instance 1 | Instance 2 | ++====================================+====================================+ +| read very_important_count (5) | | ++------------------------------------+------------------------------------+ +| | read very_important_count (5) | ++------------------------------------+------------------------------------+ +| add 1 (6) | | ++------------------------------------+------------------------------------+ +| | add 1 (6) | ++------------------------------------+------------------------------------+ +| write very_important_count (6) | | ++------------------------------------+------------------------------------+ +| | write very_important_count (6) | ++------------------------------------+------------------------------------+ + +Table: Possible Results + +Race Conditions and Critical Regions +------------------------------------ + +This overlap, where the result depends on the relative timing of +multiple tasks, is called a race condition. The piece of code containing +the concurrency issue is called a critical region. And especially since +Linux starting running on SMP machines, they became one of the major +issues in kernel design and implementation. + +Preemption can have the same effect, even if there is only one CPU: by +preempting one task during the critical region, we have exactly the same +race condition. In this case the thread which preempts might run the +critical region itself. + +The solution is to recognize when these simultaneous accesses occur, and +use locks to make sure that only one instance can enter the critical +region at any time. There are many friendly primitives in the Linux +kernel to help you do this. And then there are the unfriendly +primitives, but I'll pretend they don't exist. + +Locking in the Linux Kernel +=========================== + +If I could give you one piece of advice: never sleep with anyone crazier +than yourself. But if I had to give you advice on locking: *keep it +simple*. + +Be reluctant to introduce new locks. + +Strangely enough, this last one is the exact reverse of my advice when +you *have* slept with someone crazier than yourself. And you should +think about getting a big dog. + +Two Main Types of Kernel Locks: Spinlocks and Mutexes +----------------------------------------------------- + +There are two main types of kernel locks. The fundamental type is the +spinlock (``include/asm/spinlock.h``), which is a very simple +single-holder lock: if you can't get the spinlock, you keep trying +(spinning) until you can. Spinlocks are very small and fast, and can be +used anywhere. + +The second type is a mutex (``include/linux/mutex.h``): it is like a +spinlock, but you may block holding a mutex. If you can't lock a mutex, +your task will suspend itself, and be woken up when the mutex is +released. This means the CPU can do something else while you are +waiting. There are many cases when you simply can't sleep (see +`What Functions Are Safe To Call From Interrupts? <#sleeping-things>`__), +and so have to use a spinlock instead. + +Neither type of lock is recursive: see +`Deadlock: Simple and Advanced <#deadlock>`__. + +Locks and Uniprocessor Kernels +------------------------------ + +For kernels compiled without ``CONFIG_SMP``, and without +``CONFIG_PREEMPT`` spinlocks do not exist at all. This is an excellent +design decision: when no-one else can run at the same time, there is no +reason to have a lock. + +If the kernel is compiled without ``CONFIG_SMP``, but ``CONFIG_PREEMPT`` +is set, then spinlocks simply disable preemption, which is sufficient to +prevent any races. For most purposes, we can think of preemption as +equivalent to SMP, and not worry about it separately. + +You should always test your locking code with ``CONFIG_SMP`` and +``CONFIG_PREEMPT`` enabled, even if you don't have an SMP test box, +because it will still catch some kinds of locking bugs. + +Mutexes still exist, because they are required for synchronization +between user contexts, as we will see below. + +Locking Only In User Context +---------------------------- + +If you have a data structure which is only ever accessed from user +context, then you can use a simple mutex (``include/linux/mutex.h``) to +protect it. This is the most trivial case: you initialize the mutex. +Then you can call :c:func:`mutex_lock_interruptible()` to grab the +mutex, and :c:func:`mutex_unlock()` to release it. There is also a +:c:func:`mutex_lock()`, which should be avoided, because it will +not return if a signal is received. + +Example: ``net/netfilter/nf_sockopt.c`` allows registration of new +:c:func:`setsockopt()` and :c:func:`getsockopt()` calls, with +:c:func:`nf_register_sockopt()`. Registration and de-registration +are only done on module load and unload (and boot time, where there is +no concurrency), and the list of registrations is only consulted for an +unknown :c:func:`setsockopt()` or :c:func:`getsockopt()` system +call. The ``nf_sockopt_mutex`` is perfect to protect this, especially +since the setsockopt and getsockopt calls may well sleep. + +Locking Between User Context and Softirqs +----------------------------------------- + +If a softirq shares data with user context, you have two problems. +Firstly, the current user context can be interrupted by a softirq, and +secondly, the critical region could be entered from another CPU. This is +where :c:func:`spin_lock_bh()` (``include/linux/spinlock.h``) is +used. It disables softirqs on that CPU, then grabs the lock. +:c:func:`spin_unlock_bh()` does the reverse. (The '_bh' suffix is +a historical reference to "Bottom Halves", the old name for software +interrupts. It should really be called spin_lock_softirq()' in a +perfect world). + +Note that you can also use :c:func:`spin_lock_irq()` or +:c:func:`spin_lock_irqsave()` here, which stop hardware interrupts +as well: see `Hard IRQ Context <#hardirq-context>`__. + +This works perfectly for UP as well: the spin lock vanishes, and this +macro simply becomes :c:func:`local_bh_disable()` +(``include/linux/interrupt.h``), which protects you from the softirq +being run. + +Locking Between User Context and Tasklets +----------------------------------------- + +This is exactly the same as above, because tasklets are actually run +from a softirq. + +Locking Between User Context and Timers +--------------------------------------- + +This, too, is exactly the same as above, because timers are actually run +from a softirq. From a locking point of view, tasklets and timers are +identical. + +Locking Between Tasklets/Timers +------------------------------- + +Sometimes a tasklet or timer might want to share data with another +tasklet or timer. + +The Same Tasklet/Timer +~~~~~~~~~~~~~~~~~~~~~~ + +Since a tasklet is never run on two CPUs at once, you don't need to +worry about your tasklet being reentrant (running twice at once), even +on SMP. + +Different Tasklets/Timers +~~~~~~~~~~~~~~~~~~~~~~~~~ + +If another tasklet/timer wants to share data with your tasklet or timer +, you will both need to use :c:func:`spin_lock()` and +:c:func:`spin_unlock()` calls. :c:func:`spin_lock_bh()` is +unnecessary here, as you are already in a tasklet, and none will be run +on the same CPU. + +Locking Between Softirqs +------------------------ + +Often a softirq might want to share data with itself or a tasklet/timer. + +The Same Softirq +~~~~~~~~~~~~~~~~ + +The same softirq can run on the other CPUs: you can use a per-CPU array +(see `Per-CPU Data <#per-cpu>`__) for better performance. If you're +going so far as to use a softirq, you probably care about scalable +performance enough to justify the extra complexity. + +You'll need to use :c:func:`spin_lock()` and +:c:func:`spin_unlock()` for shared data. + +Different Softirqs +~~~~~~~~~~~~~~~~~~ + +You'll need to use :c:func:`spin_lock()` and +:c:func:`spin_unlock()` for shared data, whether it be a timer, +tasklet, different softirq or the same or another softirq: any of them +could be running on a different CPU. + +Hard IRQ Context +================ + +Hardware interrupts usually communicate with a tasklet or softirq. +Frequently this involves putting work in a queue, which the softirq will +take out. + +Locking Between Hard IRQ and Softirqs/Tasklets +---------------------------------------------- + +If a hardware irq handler shares data with a softirq, you have two +concerns. Firstly, the softirq processing can be interrupted by a +hardware interrupt, and secondly, the critical region could be entered +by a hardware interrupt on another CPU. This is where +:c:func:`spin_lock_irq()` is used. It is defined to disable +interrupts on that cpu, then grab the lock. +:c:func:`spin_unlock_irq()` does the reverse. + +The irq handler does not to use :c:func:`spin_lock_irq()`, because +the softirq cannot run while the irq handler is running: it can use +:c:func:`spin_lock()`, which is slightly faster. The only exception +would be if a different hardware irq handler uses the same lock: +:c:func:`spin_lock_irq()` will stop that from interrupting us. + +This works perfectly for UP as well: the spin lock vanishes, and this +macro simply becomes :c:func:`local_irq_disable()` +(``include/asm/smp.h``), which protects you from the softirq/tasklet/BH +being run. + +:c:func:`spin_lock_irqsave()` (``include/linux/spinlock.h``) is a +variant which saves whether interrupts were on or off in a flags word, +which is passed to :c:func:`spin_unlock_irqrestore()`. This means +that the same code can be used inside an hard irq handler (where +interrupts are already off) and in softirqs (where the irq disabling is +required). + +Note that softirqs (and hence tasklets and timers) are run on return +from hardware interrupts, so :c:func:`spin_lock_irq()` also stops +these. In that sense, :c:func:`spin_lock_irqsave()` is the most +general and powerful locking function. + +Locking Between Two Hard IRQ Handlers +------------------------------------- + +It is rare to have to share data between two IRQ handlers, but if you +do, :c:func:`spin_lock_irqsave()` should be used: it is +architecture-specific whether all interrupts are disabled inside irq +handlers themselves. + +Cheat Sheet For Locking +======================= + +Pete Zaitcev gives the following summary: + +- If you are in a process context (any syscall) and want to lock other + process out, use a mutex. You can take a mutex and sleep + (``copy_from_user*(`` or ``kmalloc(x,GFP_KERNEL)``). + +- Otherwise (== data can be touched in an interrupt), use + :c:func:`spin_lock_irqsave()` and + :c:func:`spin_unlock_irqrestore()`. + +- Avoid holding spinlock for more than 5 lines of code and across any + function call (except accessors like :c:func:`readb()`). + +Table of Minimum Requirements +----------------------------- + +The following table lists the *minimum* locking requirements between +various contexts. In some cases, the same context can only be running on +one CPU at a time, so no locking is required for that context (eg. a +particular thread can only run on one CPU at a time, but if it needs +shares data with another thread, locking is required). + +Remember the advice above: you can always use +:c:func:`spin_lock_irqsave()`, which is a superset of all other +spinlock primitives. + ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| | IRQ Handler A | IRQ Handler B | Softirq A | Softirq B | Tasklet A | Tasklet B | Timer A | Timer B | User Context A | User Context B | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| IRQ Handler A | None | | | | | | | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| IRQ Handler B | SLIS | None | | | | | | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| Softirq A | SLI | SLI | SL | | | | | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| Softirq B | SLI | SLI | SL | SL | | | | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| Tasklet A | SLI | SLI | SL | SL | None | | | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| Tasklet B | SLI | SLI | SL | SL | SL | None | | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| Timer A | SLI | SLI | SL | SL | SL | SL | None | | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| Timer B | SLI | SLI | SL | SL | SL | SL | SL | None | | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| User Context A | SLI | SLI | SLBH | SLBH | SLBH | SLBH | SLBH | SLBH | None | | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +| User Context B | SLI | SLI | SLBH | SLBH | SLBH | SLBH | SLBH | SLBH | MLI | None | ++------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ + +Table: Table of Locking Requirements + ++--------+----------------------------+ +| SLIS | spin_lock_irqsave | ++--------+----------------------------+ +| SLI | spin_lock_irq | ++--------+----------------------------+ +| SL | spin_lock | ++--------+----------------------------+ +| SLBH | spin_lock_bh | ++--------+----------------------------+ +| MLI | mutex_lock_interruptible | ++--------+----------------------------+ + +Table: Legend for Locking Requirements Table + +The trylock Functions +===================== + +There are functions that try to acquire a lock only once and immediately +return a value telling about success or failure to acquire the lock. +They can be used if you need no access to the data protected with the +lock when some other thread is holding the lock. You should acquire the +lock later if you then need access to the data protected with the lock. + +:c:func:`spin_trylock()` does not spin but returns non-zero if it +acquires the spinlock on the first try or 0 if not. This function can be +used in all contexts like :c:func:`spin_lock()`: you must have +disabled the contexts that might interrupt you and acquire the spin +lock. + +:c:func:`mutex_trylock()` does not suspend your task but returns +non-zero if it could lock the mutex on the first try or 0 if not. This +function cannot be safely used in hardware or software interrupt +contexts despite not sleeping. + +Common Examples +=============== + +Let's step through a simple example: a cache of number to name mappings. +The cache keeps a count of how often each of the objects is used, and +when it gets full, throws out the least used one. + +All In User Context +------------------- + +For our first example, we assume that all operations are in user context +(ie. from system calls), so we can sleep. This means we can use a mutex +to protect the cache and all the objects within it. Here's the code:: + + #include + #include + #include + #include + #include + + struct object + { + struct list_head list; + int id; + char name[32]; + int popularity; + }; + + /* Protects the cache, cache_num, and the objects within it */ + static DEFINE_MUTEX(cache_lock); + static LIST_HEAD(cache); + static unsigned int cache_num = 0; + #define MAX_CACHE_SIZE 10 + + /* Must be holding cache_lock */ + static struct object *__cache_find(int id) + { + struct object *i; + + list_for_each_entry(i, &cache, list) + if (i->id == id) { + i->popularity++; + return i; + } + return NULL; + } + + /* Must be holding cache_lock */ + static void __cache_delete(struct object *obj) + { + BUG_ON(!obj); + list_del(&obj->list); + kfree(obj); + cache_num--; + } + + /* Must be holding cache_lock */ + static void __cache_add(struct object *obj) + { + list_add(&obj->list, &cache); + if (++cache_num > MAX_CACHE_SIZE) { + struct object *i, *outcast = NULL; + list_for_each_entry(i, &cache, list) { + if (!outcast || i->popularity < outcast->popularity) + outcast = i; + } + __cache_delete(outcast); + } + } + + int cache_add(int id, const char *name) + { + struct object *obj; + + if ((obj = kmalloc(sizeof(*obj), GFP_KERNEL)) == NULL) + return -ENOMEM; + + strlcpy(obj->name, name, sizeof(obj->name)); + obj->id = id; + obj->popularity = 0; + + mutex_lock(&cache_lock); + __cache_add(obj); + mutex_unlock(&cache_lock); + return 0; + } + + void cache_delete(int id) + { + mutex_lock(&cache_lock); + __cache_delete(__cache_find(id)); + mutex_unlock(&cache_lock); + } + + int cache_find(int id, char *name) + { + struct object *obj; + int ret = -ENOENT; + + mutex_lock(&cache_lock); + obj = __cache_find(id); + if (obj) { + ret = 0; + strcpy(name, obj->name); + } + mutex_unlock(&cache_lock); + return ret; + } + +Note that we always make sure we have the cache_lock when we add, +delete, or look up the cache: both the cache infrastructure itself and +the contents of the objects are protected by the lock. In this case it's +easy, since we copy the data for the user, and never let them access the +objects directly. + +There is a slight (and common) optimization here: in +:c:func:`cache_add()` we set up the fields of the object before +grabbing the lock. This is safe, as no-one else can access it until we +put it in cache. + +Accessing From Interrupt Context +-------------------------------- + +Now consider the case where :c:func:`cache_find()` can be called +from interrupt context: either a hardware interrupt or a softirq. An +example would be a timer which deletes object from the cache. + +The change is shown below, in standard patch format: the ``-`` are lines +which are taken away, and the ``+`` are lines which are added. + +:: + + --- cache.c.usercontext 2003-12-09 13:58:54.000000000 +1100 + +++ cache.c.interrupt 2003-12-09 14:07:49.000000000 +1100 + @@ -12,7 +12,7 @@ + int popularity; + }; + + -static DEFINE_MUTEX(cache_lock); + +static DEFINE_SPINLOCK(cache_lock); + static LIST_HEAD(cache); + static unsigned int cache_num = 0; + #define MAX_CACHE_SIZE 10 + @@ -55,6 +55,7 @@ + int cache_add(int id, const char *name) + { + struct object *obj; + + unsigned long flags; + + if ((obj = kmalloc(sizeof(*obj), GFP_KERNEL)) == NULL) + return -ENOMEM; + @@ -63,30 +64,33 @@ + obj->id = id; + obj->popularity = 0; + + - mutex_lock(&cache_lock); + + spin_lock_irqsave(&cache_lock, flags); + __cache_add(obj); + - mutex_unlock(&cache_lock); + + spin_unlock_irqrestore(&cache_lock, flags); + return 0; + } + + void cache_delete(int id) + { + - mutex_lock(&cache_lock); + + unsigned long flags; + + + + spin_lock_irqsave(&cache_lock, flags); + __cache_delete(__cache_find(id)); + - mutex_unlock(&cache_lock); + + spin_unlock_irqrestore(&cache_lock, flags); + } + + int cache_find(int id, char *name) + { + struct object *obj; + int ret = -ENOENT; + + unsigned long flags; + + - mutex_lock(&cache_lock); + + spin_lock_irqsave(&cache_lock, flags); + obj = __cache_find(id); + if (obj) { + ret = 0; + strcpy(name, obj->name); + } + - mutex_unlock(&cache_lock); + + spin_unlock_irqrestore(&cache_lock, flags); + return ret; + } + +Note that the :c:func:`spin_lock_irqsave()` will turn off +interrupts if they are on, otherwise does nothing (if we are already in +an interrupt handler), hence these functions are safe to call from any +context. + +Unfortunately, :c:func:`cache_add()` calls :c:func:`kmalloc()` +with the ``GFP_KERNEL`` flag, which is only legal in user context. I +have assumed that :c:func:`cache_add()` is still only called in +user context, otherwise this should become a parameter to +:c:func:`cache_add()`. + +Exposing Objects Outside This File +---------------------------------- + +If our objects contained more information, it might not be sufficient to +copy the information in and out: other parts of the code might want to +keep pointers to these objects, for example, rather than looking up the +id every time. This produces two problems. + +The first problem is that we use the ``cache_lock`` to protect objects: +we'd need to make this non-static so the rest of the code can use it. +This makes locking trickier, as it is no longer all in one place. + +The second problem is the lifetime problem: if another structure keeps a +pointer to an object, it presumably expects that pointer to remain +valid. Unfortunately, this is only guaranteed while you hold the lock, +otherwise someone might call :c:func:`cache_delete()` and even +worse, add another object, re-using the same address. + +As there is only one lock, you can't hold it forever: no-one else would +get any work done. + +The solution to this problem is to use a reference count: everyone who +has a pointer to the object increases it when they first get the object, +and drops the reference count when they're finished with it. Whoever +drops it to zero knows it is unused, and can actually delete it. + +Here is the code:: + + --- cache.c.interrupt 2003-12-09 14:25:43.000000000 +1100 + +++ cache.c.refcnt 2003-12-09 14:33:05.000000000 +1100 + @@ -7,6 +7,7 @@ + struct object + { + struct list_head list; + + unsigned int refcnt; + int id; + char name[32]; + int popularity; + @@ -17,6 +18,35 @@ + static unsigned int cache_num = 0; + #define MAX_CACHE_SIZE 10 + + +static void __object_put(struct object *obj) + +{ + + if (--obj->refcnt == 0) + + kfree(obj); + +} + + + +static void __object_get(struct object *obj) + +{ + + obj->refcnt++; + +} + + + +void object_put(struct object *obj) + +{ + + unsigned long flags; + + + + spin_lock_irqsave(&cache_lock, flags); + + __object_put(obj); + + spin_unlock_irqrestore(&cache_lock, flags); + +} + + + +void object_get(struct object *obj) + +{ + + unsigned long flags; + + + + spin_lock_irqsave(&cache_lock, flags); + + __object_get(obj); + + spin_unlock_irqrestore(&cache_lock, flags); + +} + + + /* Must be holding cache_lock */ + static struct object *__cache_find(int id) + { + @@ -35,6 +65,7 @@ + { + BUG_ON(!obj); + list_del(&obj->list); + + __object_put(obj); + cache_num--; + } + + @@ -63,6 +94,7 @@ + strlcpy(obj->name, name, sizeof(obj->name)); + obj->id = id; + obj->popularity = 0; + + obj->refcnt = 1; /* The cache holds a reference */ + + spin_lock_irqsave(&cache_lock, flags); + __cache_add(obj); + @@ -79,18 +111,15 @@ + spin_unlock_irqrestore(&cache_lock, flags); + } + + -int cache_find(int id, char *name) + +struct object *cache_find(int id) + { + struct object *obj; + - int ret = -ENOENT; + unsigned long flags; + + spin_lock_irqsave(&cache_lock, flags); + obj = __cache_find(id); + - if (obj) { + - ret = 0; + - strcpy(name, obj->name); + - } + + if (obj) + + __object_get(obj); + spin_unlock_irqrestore(&cache_lock, flags); + - return ret; + + return obj; + } + +We encapsulate the reference counting in the standard 'get' and 'put' +functions. Now we can return the object itself from +:c:func:`cache_find()` which has the advantage that the user can +now sleep holding the object (eg. to :c:func:`copy_to_user()` to +name to userspace). + +The other point to note is that I said a reference should be held for +every pointer to the object: thus the reference count is 1 when first +inserted into the cache. In some versions the framework does not hold a +reference count, but they are more complicated. + +Using Atomic Operations For The Reference Count +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +In practice, ``atomic_t`` would usually be used for refcnt. There are a +number of atomic operations defined in ``include/asm/atomic.h``: these +are guaranteed to be seen atomically from all CPUs in the system, so no +lock is required. In this case, it is simpler than using spinlocks, +although for anything non-trivial using spinlocks is clearer. The +:c:func:`atomic_inc()` and :c:func:`atomic_dec_and_test()` +are used instead of the standard increment and decrement operators, and +the lock is no longer used to protect the reference count itself. + +:: + + --- cache.c.refcnt 2003-12-09 15:00:35.000000000 +1100 + +++ cache.c.refcnt-atomic 2003-12-11 15:49:42.000000000 +1100 + @@ -7,7 +7,7 @@ + struct object + { + struct list_head list; + - unsigned int refcnt; + + atomic_t refcnt; + int id; + char name[32]; + int popularity; + @@ -18,33 +18,15 @@ + static unsigned int cache_num = 0; + #define MAX_CACHE_SIZE 10 + + -static void __object_put(struct object *obj) + -{ + - if (--obj->refcnt == 0) + - kfree(obj); + -} + - + -static void __object_get(struct object *obj) + -{ + - obj->refcnt++; + -} + - + void object_put(struct object *obj) + { + - unsigned long flags; + - + - spin_lock_irqsave(&cache_lock, flags); + - __object_put(obj); + - spin_unlock_irqrestore(&cache_lock, flags); + + if (atomic_dec_and_test(&obj->refcnt)) + + kfree(obj); + } + + void object_get(struct object *obj) + { + - unsigned long flags; + - + - spin_lock_irqsave(&cache_lock, flags); + - __object_get(obj); + - spin_unlock_irqrestore(&cache_lock, flags); + + atomic_inc(&obj->refcnt); + } + + /* Must be holding cache_lock */ + @@ -65,7 +47,7 @@ + { + BUG_ON(!obj); + list_del(&obj->list); + - __object_put(obj); + + object_put(obj); + cache_num--; + } + + @@ -94,7 +76,7 @@ + strlcpy(obj->name, name, sizeof(obj->name)); + obj->id = id; + obj->popularity = 0; + - obj->refcnt = 1; /* The cache holds a reference */ + + atomic_set(&obj->refcnt, 1); /* The cache holds a reference */ + + spin_lock_irqsave(&cache_lock, flags); + __cache_add(obj); + @@ -119,7 +101,7 @@ + spin_lock_irqsave(&cache_lock, flags); + obj = __cache_find(id); + if (obj) + - __object_get(obj); + + object_get(obj); + spin_unlock_irqrestore(&cache_lock, flags); + return obj; + } + +Protecting The Objects Themselves +--------------------------------- + +In these examples, we assumed that the objects (except the reference +counts) never changed once they are created. If we wanted to allow the +name to change, there are three possibilities: + +- You can make ``cache_lock`` non-static, and tell people to grab that + lock before changing the name in any object. + +- You can provide a :c:func:`cache_obj_rename()` which grabs this + lock and changes the name for the caller, and tell everyone to use + that function. + +- You can make the ``cache_lock`` protect only the cache itself, and + use another lock to protect the name. + +Theoretically, you can make the locks as fine-grained as one lock for +every field, for every object. In practice, the most common variants +are: + +- One lock which protects the infrastructure (the ``cache`` list in + this example) and all the objects. This is what we have done so far. + +- One lock which protects the infrastructure (including the list + pointers inside the objects), and one lock inside the object which + protects the rest of that object. + +- Multiple locks to protect the infrastructure (eg. one lock per hash + chain), possibly with a separate per-object lock. + +Here is the "lock-per-object" implementation: + +:: + + --- cache.c.refcnt-atomic 2003-12-11 15:50:54.000000000 +1100 + +++ cache.c.perobjectlock 2003-12-11 17:15:03.000000000 +1100 + @@ -6,11 +6,17 @@ + + struct object + { + + /* These two protected by cache_lock. */ + struct list_head list; + + int popularity; + + + atomic_t refcnt; + + + + /* Doesn't change once created. */ + int id; + + + + spinlock_t lock; /* Protects the name */ + char name[32]; + - int popularity; + }; + + static DEFINE_SPINLOCK(cache_lock); + @@ -77,6 +84,7 @@ + obj->id = id; + obj->popularity = 0; + atomic_set(&obj->refcnt, 1); /* The cache holds a reference */ + + spin_lock_init(&obj->lock); + + spin_lock_irqsave(&cache_lock, flags); + __cache_add(obj); + +Note that I decide that the popularity count should be protected by the +``cache_lock`` rather than the per-object lock: this is because it (like +the :c:type:`struct list_head ` inside the object) +is logically part of the infrastructure. This way, I don't need to grab +the lock of every object in :c:func:`__cache_add()` when seeking +the least popular. + +I also decided that the id member is unchangeable, so I don't need to +grab each object lock in :c:func:`__cache_find()` to examine the +id: the object lock is only used by a caller who wants to read or write +the name field. + +Note also that I added a comment describing what data was protected by +which locks. This is extremely important, as it describes the runtime +behavior of the code, and can be hard to gain from just reading. And as +Alan Cox says, “Lock data, not code”. + +Common Problems +=============== + +Deadlock: Simple and Advanced +----------------------------- + +There is a coding bug where a piece of code tries to grab a spinlock +twice: it will spin forever, waiting for the lock to be released +(spinlocks, rwlocks and mutexes are not recursive in Linux). This is +trivial to diagnose: not a +stay-up-five-nights-talk-to-fluffy-code-bunnies kind of problem. + +For a slightly more complex case, imagine you have a region shared by a +softirq and user context. If you use a :c:func:`spin_lock()` call +to protect it, it is possible that the user context will be interrupted +by the softirq while it holds the lock, and the softirq will then spin +forever trying to get the same lock. + +Both of these are called deadlock, and as shown above, it can occur even +with a single CPU (although not on UP compiles, since spinlocks vanish +on kernel compiles with ``CONFIG_SMP``\ =n. You'll still get data +corruption in the second example). + +This complete lockup is easy to diagnose: on SMP boxes the watchdog +timer or compiling with ``DEBUG_SPINLOCK`` set +(``include/linux/spinlock.h``) will show this up immediately when it +happens. + +A more complex problem is the so-called 'deadly embrace', involving two +or more locks. Say you have a hash table: each entry in the table is a +spinlock, and a chain of hashed objects. Inside a softirq handler, you +sometimes want to alter an object from one place in the hash to another: +you grab the spinlock of the old hash chain and the spinlock of the new +hash chain, and delete the object from the old one, and insert it in the +new one. + +There are two problems here. First, if your code ever tries to move the +object to the same chain, it will deadlock with itself as it tries to +lock it twice. Secondly, if the same softirq on another CPU is trying to +move another object in the reverse direction, the following could +happen: + ++-----------------------+-----------------------+ +| CPU 1 | CPU 2 | ++=======================+=======================+ +| Grab lock A -> OK | Grab lock B -> OK | ++-----------------------+-----------------------+ +| Grab lock B -> spin | Grab lock A -> spin | ++-----------------------+-----------------------+ + +Table: Consequences + +The two CPUs will spin forever, waiting for the other to give up their +lock. It will look, smell, and feel like a crash. + +Preventing Deadlock +------------------- + +Textbooks will tell you that if you always lock in the same order, you +will never get this kind of deadlock. Practice will tell you that this +approach doesn't scale: when I create a new lock, I don't understand +enough of the kernel to figure out where in the 5000 lock hierarchy it +will fit. + +The best locks are encapsulated: they never get exposed in headers, and +are never held around calls to non-trivial functions outside the same +file. You can read through this code and see that it will never +deadlock, because it never tries to grab another lock while it has that +one. People using your code don't even need to know you are using a +lock. + +A classic problem here is when you provide callbacks or hooks: if you +call these with the lock held, you risk simple deadlock, or a deadly +embrace (who knows what the callback will do?). Remember, the other +programmers are out to get you, so don't do this. + +Overzealous Prevention Of Deadlocks +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +Deadlocks are problematic, but not as bad as data corruption. Code which +grabs a read lock, searches a list, fails to find what it wants, drops +the read lock, grabs a write lock and inserts the object has a race +condition. + +If you don't see why, please stay the fuck away from my code. + +Racing Timers: A Kernel Pastime +------------------------------- + +Timers can produce their own special problems with races. Consider a +collection of objects (list, hash, etc) where each object has a timer +which is due to destroy it. + +If you want to destroy the entire collection (say on module removal), +you might do the following:: + + /* THIS CODE BAD BAD BAD BAD: IF IT WAS ANY WORSE IT WOULD USE + HUNGARIAN NOTATION */ + spin_lock_bh(&list_lock); + + while (list) { + struct foo *next = list->next; + del_timer(&list->timer); + kfree(list); + list = next; + } + + spin_unlock_bh(&list_lock); + + +Sooner or later, this will crash on SMP, because a timer can have just +gone off before the :c:func:`spin_lock_bh()`, and it will only get +the lock after we :c:func:`spin_unlock_bh()`, and then try to free +the element (which has already been freed!). + +This can be avoided by checking the result of +:c:func:`del_timer()`: if it returns 1, the timer has been deleted. +If 0, it means (in this case) that it is currently running, so we can +do:: + + retry: + spin_lock_bh(&list_lock); + + while (list) { + struct foo *next = list->next; + if (!del_timer(&list->timer)) { + /* Give timer a chance to delete this */ + spin_unlock_bh(&list_lock); + goto retry; + } + kfree(list); + list = next; + } + + spin_unlock_bh(&list_lock); + + +Another common problem is deleting timers which restart themselves (by +calling :c:func:`add_timer()` at the end of their timer function). +Because this is a fairly common case which is prone to races, you should +use :c:func:`del_timer_sync()` (``include/linux/timer.h``) to +handle this case. It returns the number of times the timer had to be +deleted before we finally stopped it from adding itself back in. + +Locking Speed +============= + +There are three main things to worry about when considering speed of +some code which does locking. First is concurrency: how many things are +going to be waiting while someone else is holding a lock. Second is the +time taken to actually acquire and release an uncontended lock. Third is +using fewer, or smarter locks. I'm assuming that the lock is used fairly +often: otherwise, you wouldn't be concerned about efficiency. + +Concurrency depends on how long the lock is usually held: you should +hold the lock for as long as needed, but no longer. In the cache +example, we always create the object without the lock held, and then +grab the lock only when we are ready to insert it in the list. + +Acquisition times depend on how much damage the lock operations do to +the pipeline (pipeline stalls) and how likely it is that this CPU was +the last one to grab the lock (ie. is the lock cache-hot for this CPU): +on a machine with more CPUs, this likelihood drops fast. Consider a +700MHz Intel Pentium III: an instruction takes about 0.7ns, an atomic +increment takes about 58ns, a lock which is cache-hot on this CPU takes +160ns, and a cacheline transfer from another CPU takes an additional 170 +to 360ns. (These figures from Paul McKenney's `Linux Journal RCU +article `__). + +These two aims conflict: holding a lock for a short time might be done +by splitting locks into parts (such as in our final per-object-lock +example), but this increases the number of lock acquisitions, and the +results are often slower than having a single lock. This is another +reason to advocate locking simplicity. + +The third concern is addressed below: there are some methods to reduce +the amount of locking which needs to be done. + +Read/Write Lock Variants +------------------------ + +Both spinlocks and mutexes have read/write variants: ``rwlock_t`` and +:c:type:`struct rw_semaphore `. These divide +users into two classes: the readers and the writers. If you are only +reading the data, you can get a read lock, but to write to the data you +need the write lock. Many people can hold a read lock, but a writer must +be sole holder. + +If your code divides neatly along reader/writer lines (as our cache code +does), and the lock is held by readers for significant lengths of time, +using these locks can help. They are slightly slower than the normal +locks though, so in practice ``rwlock_t`` is not usually worthwhile. + +Avoiding Locks: Read Copy Update +-------------------------------- + +There is a special method of read/write locking called Read Copy Update. +Using RCU, the readers can avoid taking a lock altogether: as we expect +our cache to be read more often than updated (otherwise the cache is a +waste of time), it is a candidate for this optimization. + +How do we get rid of read locks? Getting rid of read locks means that +writers may be changing the list underneath the readers. That is +actually quite simple: we can read a linked list while an element is +being added if the writer adds the element very carefully. For example, +adding ``new`` to a single linked list called ``list``:: + + new->next = list->next; + wmb(); + list->next = new; + + +The :c:func:`wmb()` is a write memory barrier. It ensures that the +first operation (setting the new element's ``next`` pointer) is complete +and will be seen by all CPUs, before the second operation is (putting +the new element into the list). This is important, since modern +compilers and modern CPUs can both reorder instructions unless told +otherwise: we want a reader to either not see the new element at all, or +see the new element with the ``next`` pointer correctly pointing at the +rest of the list. + +Fortunately, there is a function to do this for standard +:c:type:`struct list_head ` lists: +:c:func:`list_add_rcu()` (``include/linux/list.h``). + +Removing an element from the list is even simpler: we replace the +pointer to the old element with a pointer to its successor, and readers +will either see it, or skip over it. + +:: + + list->next = old->next; + + +There is :c:func:`list_del_rcu()` (``include/linux/list.h``) which +does this (the normal version poisons the old object, which we don't +want). + +The reader must also be careful: some CPUs can look through the ``next`` +pointer to start reading the contents of the next element early, but +don't realize that the pre-fetched contents is wrong when the ``next`` +pointer changes underneath them. Once again, there is a +:c:func:`list_for_each_entry_rcu()` (``include/linux/list.h``) +to help you. Of course, writers can just use +:c:func:`list_for_each_entry()`, since there cannot be two +simultaneous writers. + +Our final dilemma is this: when can we actually destroy the removed +element? Remember, a reader might be stepping through this element in +the list right now: if we free this element and the ``next`` pointer +changes, the reader will jump off into garbage and crash. We need to +wait until we know that all the readers who were traversing the list +when we deleted the element are finished. We use +:c:func:`call_rcu()` to register a callback which will actually +destroy the object once all pre-existing readers are finished. +Alternatively, :c:func:`synchronize_rcu()` may be used to block +until all pre-existing are finished. + +But how does Read Copy Update know when the readers are finished? The +method is this: firstly, the readers always traverse the list inside +:c:func:`rcu_read_lock()`/:c:func:`rcu_read_unlock()` pairs: +these simply disable preemption so the reader won't go to sleep while +reading the list. + +RCU then waits until every other CPU has slept at least once: since +readers cannot sleep, we know that any readers which were traversing the +list during the deletion are finished, and the callback is triggered. +The real Read Copy Update code is a little more optimized than this, but +this is the fundamental idea. + +:: + + --- cache.c.perobjectlock 2003-12-11 17:15:03.000000000 +1100 + +++ cache.c.rcupdate 2003-12-11 17:55:14.000000000 +1100 + @@ -1,15 +1,18 @@ + #include + #include + #include + +#include + #include + #include + + struct object + { + - /* These two protected by cache_lock. */ + + /* This is protected by RCU */ + struct list_head list; + int popularity; + + + struct rcu_head rcu; + + + atomic_t refcnt; + + /* Doesn't change once created. */ + @@ -40,7 +43,7 @@ + { + struct object *i; + + - list_for_each_entry(i, &cache, list) { + + list_for_each_entry_rcu(i, &cache, list) { + if (i->id == id) { + i->popularity++; + return i; + @@ -49,19 +52,25 @@ + return NULL; + } + + +/* Final discard done once we know no readers are looking. */ + +static void cache_delete_rcu(void *arg) + +{ + + object_put(arg); + +} + + + /* Must be holding cache_lock */ + static void __cache_delete(struct object *obj) + { + BUG_ON(!obj); + - list_del(&obj->list); + - object_put(obj); + + list_del_rcu(&obj->list); + cache_num--; + + call_rcu(&obj->rcu, cache_delete_rcu); + } + + /* Must be holding cache_lock */ + static void __cache_add(struct object *obj) + { + - list_add(&obj->list, &cache); + + list_add_rcu(&obj->list, &cache); + if (++cache_num > MAX_CACHE_SIZE) { + struct object *i, *outcast = NULL; + list_for_each_entry(i, &cache, list) { + @@ -104,12 +114,11 @@ + struct object *cache_find(int id) + { + struct object *obj; + - unsigned long flags; + + - spin_lock_irqsave(&cache_lock, flags); + + rcu_read_lock(); + obj = __cache_find(id); + if (obj) + object_get(obj); + - spin_unlock_irqrestore(&cache_lock, flags); + + rcu_read_unlock(); + return obj; + } + +Note that the reader will alter the popularity member in +:c:func:`__cache_find()`, and now it doesn't hold a lock. One +solution would be to make it an ``atomic_t``, but for this usage, we +don't really care about races: an approximate result is good enough, so +I didn't change it. + +The result is that :c:func:`cache_find()` requires no +synchronization with any other functions, so is almost as fast on SMP as +it would be on UP. + +There is a further optimization possible here: remember our original +cache code, where there were no reference counts and the caller simply +held the lock whenever using the object? This is still possible: if you +hold the lock, no one can delete the object, so you don't need to get +and put the reference count. + +Now, because the 'read lock' in RCU is simply disabling preemption, a +caller which always has preemption disabled between calling +:c:func:`cache_find()` and :c:func:`object_put()` does not +need to actually get and put the reference count: we could expose +:c:func:`__cache_find()` by making it non-static, and such +callers could simply call that. + +The benefit here is that the reference count is not written to: the +object is not altered in any way, which is much faster on SMP machines +due to caching. + +Per-CPU Data +------------ + +Another technique for avoiding locking which is used fairly widely is to +duplicate information for each CPU. For example, if you wanted to keep a +count of a common condition, you could use a spin lock and a single +counter. Nice and simple. + +If that was too slow (it's usually not, but if you've got a really big +machine to test on and can show that it is), you could instead use a +counter for each CPU, then none of them need an exclusive lock. See +:c:func:`DEFINE_PER_CPU()`, :c:func:`get_cpu_var()` and +:c:func:`put_cpu_var()` (``include/linux/percpu.h``). + +Of particular use for simple per-cpu counters is the ``local_t`` type, +and the :c:func:`cpu_local_inc()` and related functions, which are +more efficient than simple code on some architectures +(``include/asm/local.h``). + +Note that there is no simple, reliable way of getting an exact value of +such a counter, without introducing more locks. This is not a problem +for some uses. + +Data Which Mostly Used By An IRQ Handler +---------------------------------------- + +If data is always accessed from within the same IRQ handler, you don't +need a lock at all: the kernel already guarantees that the irq handler +will not run simultaneously on multiple CPUs. + +Manfred Spraul points out that you can still do this, even if the data +is very occasionally accessed in user context or softirqs/tasklets. The +irq handler doesn't use a lock, and all other accesses are done as so:: + + spin_lock(&lock); + disable_irq(irq); + ... + enable_irq(irq); + spin_unlock(&lock); + +The :c:func:`disable_irq()` prevents the irq handler from running +(and waits for it to finish if it's currently running on other CPUs). +The spinlock prevents any other accesses happening at the same time. +Naturally, this is slower than just a :c:func:`spin_lock_irq()` +call, so it only makes sense if this type of access happens extremely +rarely. + +What Functions Are Safe To Call From Interrupts? +================================================ + +Many functions in the kernel sleep (ie. call schedule()) directly or +indirectly: you can never call them while holding a spinlock, or with +preemption disabled. This also means you need to be in user context: +calling them from an interrupt is illegal. + +Some Functions Which Sleep +-------------------------- + +The most common ones are listed below, but you usually have to read the +code to find out if other calls are safe. If everyone else who calls it +can sleep, you probably need to be able to sleep, too. In particular, +registration and deregistration functions usually expect to be called +from user context, and can sleep. + +- Accesses to userspace: + + - :c:func:`copy_from_user()` + + - :c:func:`copy_to_user()` + + - :c:func:`get_user()` + + - :c:func:`put_user()` + +- ``kmalloc(GFP_KERNEL)`` + +- :c:func:`mutex_lock_interruptible()` and + :c:func:`mutex_lock()` + + There is a :c:func:`mutex_trylock()` which does not sleep. + Still, it must not be used inside interrupt context since its + implementation is not safe for that. :c:func:`mutex_unlock()` + will also never sleep. It cannot be used in interrupt context either + since a mutex must be released by the same task that acquired it. + +Some Functions Which Don't Sleep +-------------------------------- + +Some functions are safe to call from any context, or holding almost any +lock. + +- :c:func:`printk()` + +- :c:func:`kfree()` + +- :c:func:`add_timer()` and :c:func:`del_timer()` + +Mutex API reference +=================== + +.. kernel-doc:: include/linux/mutex.h + :internal: + +.. kernel-doc:: kernel/locking/mutex.c + :export: + +Futex API reference +=================== + +.. kernel-doc:: kernel/futex.c + :internal: + +Further reading +=============== + +- ``Documentation/locking/spinlocks.txt``: Linus Torvalds' spinlocking + tutorial in the kernel sources. + +- Unix Systems for Modern Architectures: Symmetric Multiprocessing and + Caching for Kernel Programmers: + + Curt Schimmel's very good introduction to kernel level locking (not + written for Linux, but nearly everything applies). The book is + expensive, but really worth every penny to understand SMP locking. + [ISBN: 0201633388] + +Thanks +====== + +Thanks to Telsa Gwynne for DocBooking, neatening and adding style. + +Thanks to Martin Pool, Philipp Rumpf, Stephen Rothwell, Paul Mackerras, +Ruedi Aschwanden, Alan Cox, Manfred Spraul, Tim Waugh, Pete Zaitcev, +James Morris, Robert Love, Paul McKenney, John Ashby for proofreading, +correcting, flaming, commenting. + +Thanks to the cabal for having no influence on this document. + +Glossary +======== + +preemption + Prior to 2.5, or when ``CONFIG_PREEMPT`` is unset, processes in user + context inside the kernel would not preempt each other (ie. you had that + CPU until you gave it up, except for interrupts). With the addition of + ``CONFIG_PREEMPT`` in 2.5.4, this changed: when in user context, higher + priority tasks can "cut in": spinlocks were changed to disable + preemption, even on UP. + +bh + Bottom Half: for historical reasons, functions with '_bh' in them often + now refer to any software interrupt, e.g. :c:func:`spin_lock_bh()` + blocks any software interrupt on the current CPU. Bottom halves are + deprecated, and will eventually be replaced by tasklets. Only one bottom + half will be running at any time. + +Hardware Interrupt / Hardware IRQ + Hardware interrupt request. :c:func:`in_irq()` returns true in a + hardware interrupt handler. + +Interrupt Context + Not user context: processing a hardware irq or software irq. Indicated + by the :c:func:`in_interrupt()` macro returning true. + +SMP + Symmetric Multi-Processor: kernels compiled for multiple-CPU machines. + (``CONFIG_SMP=y``). + +Software Interrupt / softirq + Software interrupt handler. :c:func:`in_irq()` returns false; + :c:func:`in_softirq()` returns true. Tasklets and softirqs both + fall into the category of 'software interrupts'. + + Strictly speaking a softirq is one of up to 32 enumerated software + interrupts which can run on multiple CPUs at once. Sometimes used to + refer to tasklets as well (ie. all software interrupts). + +tasklet + A dynamically-registrable software interrupt, which is guaranteed to + only run on one CPU at a time. + +timer + A dynamically-registrable software interrupt, which is run at (or close + to) a given time. When running, it is just like a tasklet (in fact, they + are called from the TIMER_SOFTIRQ). + +UP + Uni-Processor: Non-SMP. (CONFIG_SMP=n). + +User Context + The kernel executing on behalf of a particular process (ie. a system + call or trap) or kernel thread. You can tell which process with the + ``current`` macro.) Not to be confused with userspace. Can be + interrupted by software or hardware interrupts. + +Userspace + A process executing its own code outside the kernel. From 7b4ff1adb57ad96d8f12a05d8c661a3d8c4d2be1 Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 10:17:45 -0300 Subject: [PATCH 04/90] mutex, futex: adjust kernel-doc markups to generate ReST There are a few issues on some kernel-doc markups that was causing troubles with kernel-doc output on ReST format: ./kernel/futex.c:492: WARNING: Inline emphasis start-string without end-string. ./kernel/futex.c:1264: WARNING: Block quote ends without a blank line; unexpected unindent. ./kernel/futex.c:1721: WARNING: Block quote ends without a blank line; unexpected unindent. ./kernel/futex.c:2338: WARNING: Block quote ends without a blank line; unexpected unindent. ./kernel/futex.c:2426: WARNING: Block quote ends without a blank line; unexpected unindent. ./kernel/futex.c:2899: WARNING: Block quote ends without a blank line; unexpected unindent. ./kernel/futex.c:2972: WARNING: Block quote ends without a blank line; unexpected unindent. Fix them. No functional changes. Acked-by: Darren Hart (VMware) Signed-off-by: Mauro Carvalho Chehab --- include/linux/mutex.h | 6 +++--- kernel/futex.c | 40 ++++++++++++++++++++-------------------- kernel/locking/mutex.c | 6 +++--- 3 files changed, 26 insertions(+), 26 deletions(-) diff --git a/include/linux/mutex.h b/include/linux/mutex.h index 1127fe31645d..ffcba1f337da 100644 --- a/include/linux/mutex.h +++ b/include/linux/mutex.h @@ -214,9 +214,9 @@ enum mutex_trylock_recursive_enum { * raisins, and once those are gone this will be removed. * * Returns: - * MUTEX_TRYLOCK_FAILED - trylock failed, - * MUTEX_TRYLOCK_SUCCESS - lock acquired, - * MUTEX_TRYLOCK_RECURSIVE - we already owned the lock. + * - MUTEX_TRYLOCK_FAILED - trylock failed, + * - MUTEX_TRYLOCK_SUCCESS - lock acquired, + * - MUTEX_TRYLOCK_RECURSIVE - we already owned the lock. */ static inline /* __deprecated */ __must_check enum mutex_trylock_recursive_enum mutex_trylock_recursive(struct mutex *lock) diff --git a/kernel/futex.c b/kernel/futex.c index 357348a6cf6b..b8ae87d227da 100644 --- a/kernel/futex.c +++ b/kernel/futex.c @@ -488,7 +488,7 @@ static void drop_futex_key_refs(union futex_key *key) * * Return: a negative error code or 0 * - * The key words are stored in *key on success. + * The key words are stored in @key on success. * * For shared mappings, it's (page->index, file_inode(vma->vm_file), * offset_within_page). For private mappings, it's (uaddr, current->mm). @@ -1259,9 +1259,9 @@ static int lock_pi_update_atomic(u32 __user *uaddr, u32 uval, u32 newval) * @set_waiters: force setting the FUTEX_WAITERS bit (1) or not (0) * * Return: - * 0 - ready to wait; - * 1 - acquired the lock; - * <0 - error + * - 0 - ready to wait; + * - 1 - acquired the lock; + * - <0 - error * * The hb->lock and futex_key refs shall be held by the caller. */ @@ -1717,9 +1717,9 @@ void requeue_pi_wake_futex(struct futex_q *q, union futex_key *key, * hb1 and hb2 must be held by the caller. * * Return: - * 0 - failed to acquire the lock atomically; - * >0 - acquired the lock, return value is vpid of the top_waiter - * <0 - error + * - 0 - failed to acquire the lock atomically; + * - >0 - acquired the lock, return value is vpid of the top_waiter + * - <0 - error */ static int futex_proxy_trylock_atomic(u32 __user *pifutex, struct futex_hash_bucket *hb1, @@ -1785,8 +1785,8 @@ static int futex_proxy_trylock_atomic(u32 __user *pifutex, * uaddr2 atomically on behalf of the top waiter. * * Return: - * >=0 - on success, the number of tasks requeued or woken; - * <0 - on error + * - >=0 - on success, the number of tasks requeued or woken; + * - <0 - on error */ static int futex_requeue(u32 __user *uaddr1, unsigned int flags, u32 __user *uaddr2, int nr_wake, int nr_requeue, @@ -2142,8 +2142,8 @@ static inline void queue_me(struct futex_q *q, struct futex_hash_bucket *hb) * be paired with exactly one earlier call to queue_me(). * * Return: - * 1 - if the futex_q was still queued (and we removed unqueued it); - * 0 - if the futex_q was already removed by the waking thread + * - 1 - if the futex_q was still queued (and we removed unqueued it); + * - 0 - if the futex_q was already removed by the waking thread */ static int unqueue_me(struct futex_q *q) { @@ -2333,9 +2333,9 @@ static long futex_wait_restart(struct restart_block *restart); * acquire the lock. Must be called with the hb lock held. * * Return: - * 1 - success, lock taken; - * 0 - success, lock not taken; - * <0 - on error (-EFAULT) + * - 1 - success, lock taken; + * - 0 - success, lock not taken; + * - <0 - on error (-EFAULT) */ static int fixup_owner(u32 __user *uaddr, struct futex_q *q, int locked) { @@ -2422,8 +2422,8 @@ static void futex_wait_queue_me(struct futex_hash_bucket *hb, struct futex_q *q, * with no q.key reference on failure. * * Return: - * 0 - uaddr contains val and hb has been locked; - * <1 - -EFAULT or -EWOULDBLOCK (uaddr does not contain val) and hb is unlocked + * - 0 - uaddr contains val and hb has been locked; + * - <1 - -EFAULT or -EWOULDBLOCK (uaddr does not contain val) and hb is unlocked */ static int futex_wait_setup(u32 __user *uaddr, u32 val, unsigned int flags, struct futex_q *q, struct futex_hash_bucket **hb) @@ -2895,8 +2895,8 @@ pi_faulted: * called with the hb lock held. * * Return: - * 0 = no early wakeup detected; - * <0 = -ETIMEDOUT or -ERESTARTNOINTR + * - 0 = no early wakeup detected; + * - <0 = -ETIMEDOUT or -ERESTARTNOINTR */ static inline int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb, @@ -2968,8 +2968,8 @@ int handle_early_requeue_pi_wakeup(struct futex_hash_bucket *hb, * If 4 or 7, we cleanup and return with -ETIMEDOUT. * * Return: - * 0 - On success; - * <0 - On error + * - 0 - On success; + * - <0 - On error */ static int futex_wait_requeue_pi(u32 __user *uaddr, unsigned int flags, u32 val, ktime_t *abs_time, u32 bitset, diff --git a/kernel/locking/mutex.c b/kernel/locking/mutex.c index 198527a62149..858a07590e39 100644 --- a/kernel/locking/mutex.c +++ b/kernel/locking/mutex.c @@ -227,9 +227,9 @@ static void __sched __mutex_lock_slowpath(struct mutex *lock); * (or statically defined) before it can be locked. memset()-ing * the mutex to 0 is not allowed. * - * ( The CONFIG_DEBUG_MUTEXES .config option turns on debugging - * checks that will enforce the restrictions and will also do - * deadlock debugging. ) + * (The CONFIG_DEBUG_MUTEXES .config option turns on debugging + * checks that will enforce the restrictions and will also do + * deadlock debugging) * * This function is similar to (but not equivalent to) down(). */ From 5b9fd1d3be33c37c17d305752217f6099bff8667 Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 10:45:47 -0300 Subject: [PATCH 05/90] locking.rst: reformat locking table Use a different markup for this table, in order to make it smaller when seeing in text mode. Signed-off-by: Mauro Carvalho Chehab --- Documentation/kernel-hacking/locking.rst | 37 +++++++++--------------- 1 file changed, 14 insertions(+), 23 deletions(-) diff --git a/Documentation/kernel-hacking/locking.rst b/Documentation/kernel-hacking/locking.rst index 976b2703df75..17774b1f7587 100644 --- a/Documentation/kernel-hacking/locking.rst +++ b/Documentation/kernel-hacking/locking.rst @@ -319,29 +319,20 @@ Remember the advice above: you can always use :c:func:`spin_lock_irqsave()`, which is a superset of all other spinlock primitives. -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| | IRQ Handler A | IRQ Handler B | Softirq A | Softirq B | Tasklet A | Tasklet B | Timer A | Timer B | User Context A | User Context B | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| IRQ Handler A | None | | | | | | | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| IRQ Handler B | SLIS | None | | | | | | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| Softirq A | SLI | SLI | SL | | | | | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| Softirq B | SLI | SLI | SL | SL | | | | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| Tasklet A | SLI | SLI | SL | SL | None | | | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| Tasklet B | SLI | SLI | SL | SL | SL | None | | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| Timer A | SLI | SLI | SL | SL | SL | SL | None | | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| Timer B | SLI | SLI | SL | SL | SL | SL | SL | None | | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| User Context A | SLI | SLI | SLBH | SLBH | SLBH | SLBH | SLBH | SLBH | None | | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ -| User Context B | SLI | SLI | SLBH | SLBH | SLBH | SLBH | SLBH | SLBH | MLI | None | -+------------------+-----------------+-----------------+-------------+-------------+-------------+-------------+-----------+-----------+------------------+------------------+ +============== ============= ============= ========= ========= ========= ========= ======= ======= ============== ============== +. IRQ Handler A IRQ Handler B Softirq A Softirq B Tasklet A Tasklet B Timer A Timer B User Context A User Context B +============== ============= ============= ========= ========= ========= ========= ======= ======= ============== ============== +IRQ Handler A None +IRQ Handler B SLIS None +Softirq A SLI SLI SL +Softirq B SLI SLI SL SL +Tasklet A SLI SLI SL SL None +Tasklet B SLI SLI SL SL SL None +Timer A SLI SLI SL SL SL SL None +Timer B SLI SLI SL SL SL SL SL None +User Context A SLI SLI SLBH SLBH SLBH SLBH SLBH SLBH None +User Context B SLI SLI SLBH SLBH SLBH SLBH SLBH SLBH MLI None +============== ============= ============= ========= ========= ========= ========= ======= ======= ============== ============== Table: Table of Locking Requirements From 475c5ef83d26c399a55b106398e9dcb38e6d888b Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 16:15:07 -0300 Subject: [PATCH 06/90] locking.rst: add captions to two tables Those tables have a "caption" at the end, but ReST actually expects it on a different way. Signed-off-by: Mauro Carvalho Chehab --- Documentation/kernel-hacking/locking.rst | 66 ++++++++++++------------ 1 file changed, 34 insertions(+), 32 deletions(-) diff --git a/Documentation/kernel-hacking/locking.rst b/Documentation/kernel-hacking/locking.rst index 17774b1f7587..b70c2c4eb147 100644 --- a/Documentation/kernel-hacking/locking.rst +++ b/Documentation/kernel-hacking/locking.rst @@ -29,43 +29,45 @@ In a normal program, you can increment a counter like so: This is what they would expect to happen: -+------------------------------------+------------------------------------+ -| Instance 1 | Instance 2 | -+====================================+====================================+ -| read very_important_count (5) | | -+------------------------------------+------------------------------------+ -| add 1 (6) | | -+------------------------------------+------------------------------------+ -| write very_important_count (6) | | -+------------------------------------+------------------------------------+ -| | read very_important_count (6) | -+------------------------------------+------------------------------------+ -| | add 1 (7) | -+------------------------------------+------------------------------------+ -| | write very_important_count (7) | -+------------------------------------+------------------------------------+ -Table: Expected Results +.. table:: Expected Results + + +------------------------------------+------------------------------------+ + | Instance 1 | Instance 2 | + +====================================+====================================+ + | read very_important_count (5) | | + +------------------------------------+------------------------------------+ + | add 1 (6) | | + +------------------------------------+------------------------------------+ + | write very_important_count (6) | | + +------------------------------------+------------------------------------+ + | | read very_important_count (6) | + +------------------------------------+------------------------------------+ + | | add 1 (7) | + +------------------------------------+------------------------------------+ + | | write very_important_count (7) | + +------------------------------------+------------------------------------+ This is what might happen: -+------------------------------------+------------------------------------+ -| Instance 1 | Instance 2 | -+====================================+====================================+ -| read very_important_count (5) | | -+------------------------------------+------------------------------------+ -| | read very_important_count (5) | -+------------------------------------+------------------------------------+ -| add 1 (6) | | -+------------------------------------+------------------------------------+ -| | add 1 (6) | -+------------------------------------+------------------------------------+ -| write very_important_count (6) | | -+------------------------------------+------------------------------------+ -| | write very_important_count (6) | -+------------------------------------+------------------------------------+ +.. table:: Possible Results + + +------------------------------------+------------------------------------+ + | Instance 1 | Instance 2 | + +====================================+====================================+ + | read very_important_count (5) | | + +------------------------------------+------------------------------------+ + | | read very_important_count (5) | + +------------------------------------+------------------------------------+ + | add 1 (6) | | + +------------------------------------+------------------------------------+ + | | add 1 (6) | + +------------------------------------+------------------------------------+ + | write very_important_count (6) | | + +------------------------------------+------------------------------------+ + | | write very_important_count (6) | + +------------------------------------+------------------------------------+ -Table: Possible Results Race Conditions and Critical Regions ------------------------------------ From dc89fca93ec05a2125b1dfa78efe0e8c7668c8fb Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 16:15:16 -0300 Subject: [PATCH 07/90] locking.rst: Update some ReST markups Correct a few minor issues with ReST notation used on this file (produced by an automatic tool). Signed-off-by: Mauro Carvalho Chehab --- Documentation/kernel-hacking/locking.rst | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/Documentation/kernel-hacking/locking.rst b/Documentation/kernel-hacking/locking.rst index b70c2c4eb147..f937c0fd11aa 100644 --- a/Documentation/kernel-hacking/locking.rst +++ b/Documentation/kernel-hacking/locking.rst @@ -93,13 +93,13 @@ Locking in the Linux Kernel =========================== If I could give you one piece of advice: never sleep with anyone crazier -than yourself. But if I had to give you advice on locking: *keep it -simple*. +than yourself. But if I had to give you advice on locking: **keep it +simple**. Be reluctant to introduce new locks. Strangely enough, this last one is the exact reverse of my advice when -you *have* slept with someone crazier than yourself. And you should +you **have** slept with someone crazier than yourself. And you should think about getting a big dog. Two Main Types of Kernel Locks: Spinlocks and Mutexes @@ -311,7 +311,7 @@ Pete Zaitcev gives the following summary: Table of Minimum Requirements ----------------------------- -The following table lists the *minimum* locking requirements between +The following table lists the **minimum** locking requirements between various contexts. In some cases, the same context can only be running on one CPU at a time, so no locking is required for that context (eg. a particular thread can only run on one CPU at a time, but if it needs @@ -703,7 +703,7 @@ reference count, but they are more complicated. Using Atomic Operations For The Reference Count ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -In practice, ``atomic_t`` would usually be used for refcnt. There are a +In practice, :c:type:`atomic_t` would usually be used for refcnt. There are a number of atomic operations defined in ``include/asm/atomic.h``: these are guaranteed to be seen atomically from all CPUs in the system, so no lock is required. In this case, it is simpler than using spinlocks, @@ -1321,7 +1321,7 @@ from user context, and can sleep. - :c:func:`put_user()` -- ``kmalloc(GFP_KERNEL)`` +- :c:func:`kmalloc(GFP_KERNEL) ` - :c:func:`mutex_lock_interruptible()` and :c:func:`mutex_lock()` @@ -1431,10 +1431,10 @@ tasklet timer A dynamically-registrable software interrupt, which is run at (or close to) a given time. When running, it is just like a tasklet (in fact, they - are called from the TIMER_SOFTIRQ). + are called from the ``TIMER_SOFTIRQ``). UP - Uni-Processor: Non-SMP. (CONFIG_SMP=n). + Uni-Processor: Non-SMP. (``CONFIG_SMP=n``). User Context The kernel executing on behalf of a particular process (ie. a system From 7fb2e8a49037099c56ea209aaa8527e5f3e742ba Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Thu, 11 May 2017 16:23:50 -0300 Subject: [PATCH 08/90] docs-rst: convert kgdb DocBook to ReST Use pandoc to convert documentation to ReST by calling Documentation/sphinx/tmplcvt script. Signed-off-by: Mauro Carvalho Chehab --- Documentation/DocBook/Makefile | 2 +- Documentation/DocBook/kgdb.tmpl | 918 ----------------------------- Documentation/dev-tools/index.rst | 1 + Documentation/dev-tools/kgdb.rst | 930 ++++++++++++++++++++++++++++++ 4 files changed, 932 insertions(+), 919 deletions(-) delete mode 100644 Documentation/DocBook/kgdb.tmpl create mode 100644 Documentation/dev-tools/kgdb.rst diff --git a/Documentation/DocBook/Makefile b/Documentation/DocBook/Makefile index 9df94f7c2003..b9d2b88b9905 100644 --- a/Documentation/DocBook/Makefile +++ b/Documentation/DocBook/Makefile @@ -8,7 +8,7 @@ DOCBOOKS := z8530book.xml \ networking.xml \ - filesystems.xml lsm.xml kgdb.xml \ + filesystems.xml lsm.xml \ libata.xml mtdnand.xml librs.xml rapidio.xml \ s390-drivers.xml scsi.xml \ sh.xml w1.xml diff --git a/Documentation/DocBook/kgdb.tmpl b/Documentation/DocBook/kgdb.tmpl deleted file mode 100644 index 856ac20bf367..000000000000 --- a/Documentation/DocBook/kgdb.tmpl +++ /dev/null @@ -1,918 +0,0 @@ - - - - - - Using kgdb, kdb and the kernel debugger internals - - - - Jason - Wessel - -
- jason.wessel@windriver.com -
- - - - - 2008,2010 - Wind River Systems, Inc. - - - 2004-2005 - MontaVista Software, Inc. - - - 2004 - Amit S. Kale - - - - - This file is licensed under the terms of the GNU General Public License - version 2. This program is licensed "as is" without any warranty of any - kind, whether express or implied. - - - - - - - - Introduction - - The kernel has two different debugger front ends (kdb and kgdb) - which interface to the debug core. It is possible to use either - of the debugger front ends and dynamically transition between them - if you configure the kernel properly at compile and runtime. - - - Kdb is simplistic shell-style interface which you can use on a - system console with a keyboard or serial console. You can use it - to inspect memory, registers, process lists, dmesg, and even set - breakpoints to stop in a certain location. Kdb is not a source - level debugger, although you can set breakpoints and execute some - basic kernel run control. Kdb is mainly aimed at doing some - analysis to aid in development or diagnosing kernel problems. You - can access some symbols by name in kernel built-ins or in kernel - modules if the code was built - with CONFIG_KALLSYMS. - - - Kgdb is intended to be used as a source level debugger for the - Linux kernel. It is used along with gdb to debug a Linux kernel. - The expectation is that gdb can be used to "break in" to the - kernel to inspect memory, variables and look through call stack - information similar to the way an application developer would use - gdb to debug an application. It is possible to place breakpoints - in kernel code and perform some limited execution stepping. - - - Two machines are required for using kgdb. One of these machines is - a development machine and the other is the target machine. The - kernel to be debugged runs on the target machine. The development - machine runs an instance of gdb against the vmlinux file which - contains the symbols (not a boot image such as bzImage, zImage, - uImage...). In gdb the developer specifies the connection - parameters and connects to kgdb. The type of connection a - developer makes with gdb depends on the availability of kgdb I/O - modules compiled as built-ins or loadable kernel modules in the test - machine's kernel. - - - - Compiling a kernel - - - In order to enable compilation of kdb, you must first enable kgdb. - The kgdb test compile options are described in the kgdb test suite chapter. - - - - Kernel config options for kgdb - - To enable CONFIG_KGDB you should look under - "Kernel hacking" / "Kernel debugging" and select "KGDB: kernel debugger". - - - While it is not a hard requirement that you have symbols in your - vmlinux file, gdb tends not to be very useful without the symbolic - data, so you will want to turn - on CONFIG_DEBUG_INFO which is called "Compile the - kernel with debug info" in the config menu. - - - It is advised, but not required, that you turn on the - CONFIG_FRAME_POINTER kernel option which is called "Compile the - kernel with frame pointers" in the config menu. This option - inserts code to into the compiled executable which saves the frame - information in registers or on the stack at different points which - allows a debugger such as gdb to more accurately construct - stack back traces while debugging the kernel. - - - If the architecture that you are using supports the kernel option - CONFIG_STRICT_KERNEL_RWX, you should consider turning it off. This - option will prevent the use of software breakpoints because it - marks certain regions of the kernel's memory space as read-only. - If kgdb supports it for the architecture you are using, you can - use hardware breakpoints if you desire to run with the - CONFIG_STRICT_KERNEL_RWX option turned on, else you need to turn off - this option. - - - Next you should choose one of more I/O drivers to interconnect - debugging host and debugged target. Early boot debugging requires - a KGDB I/O driver that supports early debugging and the driver - must be built into the kernel directly. Kgdb I/O driver - configuration takes place via kernel or module parameters which - you can learn more about in the in the section that describes the - parameter "kgdboc". - - Here is an example set of .config symbols to enable or - disable for kgdb: - - # CONFIG_STRICT_KERNEL_RWX is not set - CONFIG_FRAME_POINTER=y - CONFIG_KGDB=y - CONFIG_KGDB_SERIAL_CONSOLE=y - - - - - Kernel config options for kdb - Kdb is quite a bit more complex than the simple gdbstub - sitting on top of the kernel's debug core. Kdb must implement a - shell, and also adds some helper functions in other parts of the - kernel, responsible for printing out interesting data such as what - you would see if you ran "lsmod", or "ps". In order to build kdb - into the kernel you follow the same steps as you would for kgdb. - - The main config option for kdb - is CONFIG_KGDB_KDB which is called "KGDB_KDB: - include kdb frontend for kgdb" in the config menu. In theory you - would have already also selected an I/O driver such as the - CONFIG_KGDB_SERIAL_CONSOLE interface if you plan on using kdb on a - serial port, when you were configuring kgdb. - - If you want to use a PS/2-style keyboard with kdb, you would - select CONFIG_KDB_KEYBOARD which is called "KGDB_KDB: keyboard as - input device" in the config menu. The CONFIG_KDB_KEYBOARD option - is not used for anything in the gdb interface to kgdb. The - CONFIG_KDB_KEYBOARD option only works with kdb. - - Here is an example set of .config symbols to enable/disable kdb: - - # CONFIG_STRICT_KERNEL_RWX is not set - CONFIG_FRAME_POINTER=y - CONFIG_KGDB=y - CONFIG_KGDB_SERIAL_CONSOLE=y - CONFIG_KGDB_KDB=y - CONFIG_KDB_KEYBOARD=y - - - - - - Kernel Debugger Boot Arguments - This section describes the various runtime kernel - parameters that affect the configuration of the kernel debugger. - The following chapter covers using kdb and kgdb as well as - providing some examples of the configuration parameters. - - Kernel parameter: kgdboc - The kgdboc driver was originally an abbreviation meant to - stand for "kgdb over console". Today it is the primary mechanism - to configure how to communicate from gdb to kgdb as well as the - devices you want to use to interact with the kdb shell. - - For kgdb/gdb, kgdboc is designed to work with a single serial - port. It is intended to cover the circumstance where you want to - use a serial console as your primary console as well as using it to - perform kernel debugging. It is also possible to use kgdb on a - serial port which is not designated as a system console. Kgdboc - may be configured as a kernel built-in or a kernel loadable module. - You can only make use of kgdbwait and early - debugging if you build kgdboc into the kernel as a built-in. - - Optionally you can elect to activate kms (Kernel Mode - Setting) integration. When you use kms with kgdboc and you have a - video driver that has atomic mode setting hooks, it is possible to - enter the debugger on the graphics console. When the kernel - execution is resumed, the previous graphics mode will be restored. - This integration can serve as a useful tool to aid in diagnosing - crashes or doing analysis of memory with kdb while allowing the - full graphics console applications to run. - - - kgdboc arguments - Usage: kgdboc=[kms][[,]kbd][[,]serial_device][,baud] - The order listed above must be observed if you use any of the - optional configurations together. - - Abbreviations: - - kms = Kernel Mode Setting - kbd = Keyboard - - - You can configure kgdboc to use the keyboard, and/or a serial - device depending on if you are using kdb and/or kgdb, in one of the - following scenarios. The order listed above must be observed if - you use any of the optional configurations together. Using kms + - only gdb is generally not a useful combination. - - Using loadable module or built-in - - - As a kernel built-in: - Use the kernel boot argument: kgdboc=<tty-device>,[baud] - - As a kernel loadable module: - Use the command: modprobe kgdboc kgdboc=<tty-device>,[baud] - Here are two examples of how you might format the kgdboc - string. The first is for an x86 target using the first serial port. - The second example is for the ARM Versatile AB using the second - serial port. - - kgdboc=ttyS0,115200 - kgdboc=ttyAMA1,115200 - - - - - - - Configure kgdboc at runtime with sysfs - At run time you can enable or disable kgdboc by echoing a - parameters into the sysfs. Here are two examples: - - Enable kgdboc on ttyS0 - echo ttyS0 > /sys/module/kgdboc/parameters/kgdboc - Disable kgdboc - echo "" > /sys/module/kgdboc/parameters/kgdboc - - NOTE: You do not need to specify the baud if you are - configuring the console on tty which is already configured or - open. - - - More examples - You can configure kgdboc to use the keyboard, and/or a serial device - depending on if you are using kdb and/or kgdb, in one of the - following scenarios. - - kdb and kgdb over only a serial port - kgdboc=<serial_device>[,baud] - Example: kgdboc=ttyS0,115200 - - kdb and kgdb with keyboard and a serial port - kgdboc=kbd,<serial_device>[,baud] - Example: kgdboc=kbd,ttyS0,115200 - - kdb with a keyboard - kgdboc=kbd - - kdb with kernel mode setting - kgdboc=kms,kbd - - kdb with kernel mode setting and kgdb over a serial port - kgdboc=kms,kbd,ttyS0,115200 - - - - NOTE: Kgdboc does not support interrupting the target via the - gdb remote protocol. You must manually send a sysrq-g unless you - have a proxy that splits console output to a terminal program. - A console proxy has a separate TCP port for the debugger and a separate - TCP port for the "human" console. The proxy can take care of sending - the sysrq-g for you. - - When using kgdboc with no debugger proxy, you can end up - connecting the debugger at one of two entry points. If an - exception occurs after you have loaded kgdboc, a message should - print on the console stating it is waiting for the debugger. In - this case you disconnect your terminal program and then connect the - debugger in its place. If you want to interrupt the target system - and forcibly enter a debug session you have to issue a Sysrq - sequence and then type the letter g. Then - you disconnect the terminal session and connect gdb. Your options - if you don't like this are to hack gdb to send the sysrq-g for you - as well as on the initial connect, or to use a debugger proxy that - allows an unmodified gdb to do the debugging. - - - - - - Kernel parameter: kgdbwait - - The Kernel command line option kgdbwait makes - kgdb wait for a debugger connection during booting of a kernel. You - can only use this option if you compiled a kgdb I/O driver into the - kernel and you specified the I/O driver configuration as a kernel - command line option. The kgdbwait parameter should always follow the - configuration parameter for the kgdb I/O driver in the kernel - command line else the I/O driver will not be configured prior to - asking the kernel to use it to wait. - - - The kernel will stop and wait as early as the I/O driver and - architecture allows when you use this option. If you build the - kgdb I/O driver as a loadable kernel module kgdbwait will not do - anything. - - - - Kernel parameter: kgdbcon - The kgdbcon feature allows you to see printk() messages - inside gdb while gdb is connected to the kernel. Kdb does not make - use of the kgdbcon feature. - - Kgdb supports using the gdb serial protocol to send console - messages to the debugger when the debugger is connected and running. - There are two ways to activate this feature. - - Activate with the kernel command line option: - kgdbcon - - Use sysfs before configuring an I/O driver - - echo 1 > /sys/module/kgdb/parameters/kgdb_use_con - - - NOTE: If you do this after you configure the kgdb I/O driver, the - setting will not take effect until the next point the I/O is - reconfigured. - - - - - IMPORTANT NOTE: You cannot use kgdboc + kgdbcon on a tty that is an - active system console. An example of incorrect usage is console=ttyS0,115200 kgdboc=ttyS0 kgdbcon - - It is possible to use this option with kgdboc on a tty that is not a system console. - - - - Run time parameter: kgdbreboot - The kgdbreboot feature allows you to change how the debugger - deals with the reboot notification. You have 3 choices for the - behavior. The default behavior is always set to 0. - - echo -1 > /sys/module/debug_core/parameters/kgdbreboot - Ignore the reboot notification entirely. - - echo 0 > /sys/module/debug_core/parameters/kgdbreboot - Send the detach message to any attached debugger client. - - echo 1 > /sys/module/debug_core/parameters/kgdbreboot - Enter the debugger on reboot notify. - - - - - - Using kdb - - - - Quick start for kdb on a serial port - This is a quick example of how to use kdb. - - Configure kgdboc at boot using kernel parameters: - - console=ttyS0,115200 kgdboc=ttyS0,115200 - - OR - Configure kgdboc after the kernel has booted; assuming you are using a serial port console: - - echo ttyS0 > /sys/module/kgdboc/parameters/kgdboc - - - - Enter the kernel debugger manually or by waiting for an oops or fault. There are several ways you can enter the kernel debugger manually; all involve using the sysrq-g, which means you must have enabled CONFIG_MAGIC_SYSRQ=y in your kernel config. - - When logged in as root or with a super user session you can run: - echo g > /proc/sysrq-trigger - Example using minicom 2.2 - Press: Control-a - Press: f - Press: g - - When you have telneted to a terminal server that supports sending a remote break - Press: Control-] - Type in:send break - Press: Enter - Press: g - - - - From the kdb prompt you can run the "help" command to see a complete list of the commands that are available. - Some useful commands in kdb include: - - lsmod -- Shows where kernel modules are loaded - ps -- Displays only the active processes - ps A -- Shows all the processes - summary -- Shows kernel version info and memory usage - bt -- Get a backtrace of the current process using dump_stack() - dmesg -- View the kernel syslog buffer - go -- Continue the system - - - - - When you are done using kdb you need to consider rebooting the - system or using the "go" command to resuming normal kernel - execution. If you have paused the kernel for a lengthy period of - time, applications that rely on timely networking or anything to do - with real wall clock time could be adversely affected, so you - should take this into consideration when using the kernel - debugger. - - - - - Quick start for kdb using a keyboard connected console - This is a quick example of how to use kdb with a keyboard. - - Configure kgdboc at boot using kernel parameters: - - kgdboc=kbd - - OR - Configure kgdboc after the kernel has booted: - - echo kbd > /sys/module/kgdboc/parameters/kgdboc - - - - Enter the kernel debugger manually or by waiting for an oops or fault. There are several ways you can enter the kernel debugger manually; all involve using the sysrq-g, which means you must have enabled CONFIG_MAGIC_SYSRQ=y in your kernel config. - - When logged in as root or with a super user session you can run: - echo g > /proc/sysrq-trigger - Example using a laptop keyboard - Press and hold down: Alt - Press and hold down: Fn - Press and release the key with the label: SysRq - Release: Fn - Press and release: g - Release: Alt - - Example using a PS/2 101-key keyboard - Press and hold down: Alt - Press and release the key with the label: SysRq - Press and release: g - Release: Alt - - - - - Now type in a kdb command such as "help", "dmesg", "bt" or "go" to continue kernel execution. - - - - - - Using kgdb / gdb - In order to use kgdb you must activate it by passing - configuration information to one of the kgdb I/O drivers. If you - do not pass any configuration information kgdb will not do anything - at all. Kgdb will only actively hook up to the kernel trap hooks - if a kgdb I/O driver is loaded and configured. If you unconfigure - a kgdb I/O driver, kgdb will unregister all the kernel hook points. - - All kgdb I/O drivers can be reconfigured at run time, if - CONFIG_SYSFS and CONFIG_MODULES - are enabled, by echo'ing a new config string to - /sys/module/<driver>/parameter/<option>. - The driver can be unconfigured by passing an empty string. You cannot - change the configuration while the debugger is attached. Make sure - to detach the debugger with the detach command - prior to trying to unconfigure a kgdb I/O driver. - - - Connecting with gdb to a serial port - - Configure kgdboc - Configure kgdboc at boot using kernel parameters: - - kgdboc=ttyS0,115200 - - OR - Configure kgdboc after the kernel has booted: - - echo ttyS0 > /sys/module/kgdboc/parameters/kgdboc - - - - Stop kernel execution (break into the debugger) - In order to connect to gdb via kgdboc, the kernel must - first be stopped. There are several ways to stop the kernel which - include using kgdbwait as a boot argument, via a sysrq-g, or running - the kernel until it takes an exception where it waits for the - debugger to attach. - - When logged in as root or with a super user session you can run: - echo g > /proc/sysrq-trigger - Example using minicom 2.2 - Press: Control-a - Press: f - Press: g - - When you have telneted to a terminal server that supports sending a remote break - Press: Control-] - Type in:send break - Press: Enter - Press: g - - - - - - Connect from gdb - - Example (using a directly connected port): - - - % gdb ./vmlinux - (gdb) set remotebaud 115200 - (gdb) target remote /dev/ttyS0 - - - Example (kgdb to a terminal server on TCP port 2012): - - - % gdb ./vmlinux - (gdb) target remote 192.168.2.2:2012 - - - Once connected, you can debug a kernel the way you would debug an - application program. - - - If you are having problems connecting or something is going - seriously wrong while debugging, it will most often be the case - that you want to enable gdb to be verbose about its target - communications. You do this prior to issuing the target - remote command by typing in: set debug remote 1 - - - - Remember if you continue in gdb, and need to "break in" again, - you need to issue an other sysrq-g. It is easy to create a simple - entry point by putting a breakpoint at sys_sync - and then you can run "sync" from a shell or script to break into the - debugger. - - - - kgdb and kdb interoperability - It is possible to transition between kdb and kgdb dynamically. - The debug core will remember which you used the last time and - automatically start in the same mode. - - Switching between kdb and kgdb - - Switching from kgdb to kdb - - There are two ways to switch from kgdb to kdb: you can use gdb to - issue a maintenance packet, or you can blindly type the command $3#33. - Whenever the kernel debugger stops in kgdb mode it will print the - message KGDB or $3#33 for KDB. It is important - to note that you have to type the sequence correctly in one pass. - You cannot type a backspace or delete because kgdb will interpret - that as part of the debug stream. - - Change from kgdb to kdb by blindly typing: - $3#33 - Change from kgdb to kdb with gdb - maintenance packet 3 - NOTE: Now you must kill gdb. Typically you press control-z and - issue the command: kill -9 % - - - - - Change from kdb to kgdb - There are two ways you can change from kdb to kgdb. You can - manually enter kgdb mode by issuing the kgdb command from the kdb - shell prompt, or you can connect gdb while the kdb shell prompt is - active. The kdb shell looks for the typical first commands that gdb - would issue with the gdb remote protocol and if it sees one of those - commands it automatically changes into kgdb mode. - - From kdb issue the command: - kgdb - Now disconnect your terminal program and connect gdb in its place - At the kdb prompt, disconnect the terminal program and connect gdb in its place. - - - - - Running kdb commands from gdb - It is possible to run a limited set of kdb commands from gdb, - using the gdb monitor command. You don't want to execute any of the - run control or breakpoint operations, because it can disrupt the - state of the kernel debugger. You should be using gdb for - breakpoints and run control operations if you have gdb connected. - The more useful commands to run are things like lsmod, dmesg, ps or - possibly some of the memory information commands. To see all the kdb - commands you can run monitor help. - Example: - -(gdb) monitor ps -1 idle process (state I) and -27 sleeping system daemon (state M) processes suppressed, -use 'ps A' to see all. -Task Addr Pid Parent [*] cpu State Thread Command - -0xc78291d0 1 0 0 0 S 0xc7829404 init -0xc7954150 942 1 0 0 S 0xc7954384 dropbear -0xc78789c0 944 1 0 0 S 0xc7878bf4 sh -(gdb) - - - - - - kgdb Test Suite - - When kgdb is enabled in the kernel config you can also elect to - enable the config parameter KGDB_TESTS. Turning this on will - enable a special kgdb I/O module which is designed to test the - kgdb internal functions. - - - The kgdb tests are mainly intended for developers to test the kgdb - internals as well as a tool for developing a new kgdb architecture - specific implementation. These tests are not really for end users - of the Linux kernel. The primary source of documentation would be - to look in the drivers/misc/kgdbts.c file. - - - The kgdb test suite can also be configured at compile time to run - the core set of tests by setting the kernel config parameter - KGDB_TESTS_ON_BOOT. This particular option is aimed at automated - regression testing and does not require modifying the kernel boot - config arguments. If this is turned on, the kgdb test suite can - be disabled by specifying "kgdbts=" as a kernel boot argument. - - - - Kernel Debugger Internals - - Architecture Specifics - - The kernel debugger is organized into a number of components: - - The debug core - - The debug core is found in kernel/debugger/debug_core.c. It contains: - - A generic OS exception handler which includes - sync'ing the processors into a stopped state on an multi-CPU - system. - The API to talk to the kgdb I/O drivers - The API to make calls to the arch-specific kgdb implementation - The logic to perform safe memory reads and writes to memory while using the debugger - A full implementation for software breakpoints unless overridden by the arch - The API to invoke either the kdb or kgdb frontend to the debug core. - The structures and callback API for atomic kernel mode setting. - NOTE: kgdboc is where the kms callbacks are invoked. - - - - kgdb arch-specific implementation - - This implementation is generally found in arch/*/kernel/kgdb.c. - As an example, arch/x86/kernel/kgdb.c contains the specifics to - implement HW breakpoint as well as the initialization to - dynamically register and unregister for the trap handlers on - this architecture. The arch-specific portion implements: - - contains an arch-specific trap catcher which - invokes kgdb_handle_exception() to start kgdb about doing its - work - translation to and from gdb specific packet format to pt_regs - Registration and unregistration of architecture specific trap hooks - Any special exception handling and cleanup - NMI exception handling and cleanup - (optional) HW breakpoints - - - - gdbstub frontend (aka kgdb) - The gdbstub is located in kernel/debug/gdbstub.c. It contains: - - All the logic to implement the gdb serial protocol - - - kdb frontend - The kdb debugger shell is broken down into a number of - components. The kdb core is located in kernel/debug/kdb. There - are a number of helper functions in some of the other kernel - components to make it possible for kdb to examine and report - information about the kernel without taking locks that could - cause a kernel deadlock. The kdb core contains implements the following functionality. - - A simple shell - The kdb core command set - A registration API to register additional kdb shell commands. - - A good example of a self-contained kdb module - is the "ftdump" command for dumping the ftrace buffer. See: - kernel/trace/trace_kdb.c - For an example of how to dynamically register - a new kdb command you can build the kdb_hello.ko kernel module - from samples/kdb/kdb_hello.c. To build this example you can - set CONFIG_SAMPLES=y and CONFIG_SAMPLE_KDB=m in your kernel - config. Later run "modprobe kdb_hello" and the next time you - enter the kdb shell, you can run the "hello" - command. - - The implementation for kdb_printf() which - emits messages directly to I/O drivers, bypassing the kernel - log. - SW / HW breakpoint management for the kdb shell - - - kgdb I/O driver - - Each kgdb I/O driver has to provide an implementation for the following: - - configuration via built-in or module - dynamic configuration and kgdb hook registration calls - read and write character interface - A cleanup handler for unconfiguring from the kgdb core - (optional) Early debug methodology - - Any given kgdb I/O driver has to operate very closely with the - hardware and must do it in such a way that does not enable - interrupts or change other parts of the system context without - completely restoring them. The kgdb core will repeatedly "poll" - a kgdb I/O driver for characters when it needs input. The I/O - driver is expected to return immediately if there is no data - available. Doing so allows for the future possibility to touch - watchdog hardware in such a way as to have a target system not - reset when these are enabled. - - - - - - If you are intent on adding kgdb architecture specific support - for a new architecture, the architecture should define - HAVE_ARCH_KGDB in the architecture specific - Kconfig file. This will enable kgdb for the architecture, and - at that point you must create an architecture specific kgdb - implementation. - - - There are a few flags which must be set on every architecture in - their <asm/kgdb.h> file. These are: - - - - NUMREGBYTES: The size in bytes of all of the registers, so - that we can ensure they will all fit into a packet. - - - - - BUFMAX: The size in bytes of the buffer GDB will read into. - This must be larger than NUMREGBYTES. - - - - - CACHE_FLUSH_IS_SAFE: Set to 1 if it is always safe to call - flush_cache_range or flush_icache_range. On some architectures, - these functions may not be safe to call on SMP since we keep other - CPUs in a holding pattern. - - - - - - There are also the following functions for the common backend, - found in kernel/kgdb.c, that must be supplied by the - architecture-specific backend unless marked as (optional), in - which case a default function maybe used if the architecture - does not need to provide a specific implementation. - -!Iinclude/linux/kgdb.h - - - kgdboc internals - - kgdboc and uarts - - The kgdboc driver is actually a very thin driver that relies on the - underlying low level to the hardware driver having "polling hooks" - to which the tty driver is attached. In the initial - implementation of kgdboc the serial_core was changed to expose a - low level UART hook for doing polled mode reading and writing of a - single character while in an atomic context. When kgdb makes an I/O - request to the debugger, kgdboc invokes a callback in the serial - core which in turn uses the callback in the UART driver. - - When using kgdboc with a UART, the UART driver must implement two callbacks in the struct uart_ops. Example from drivers/8250.c: -#ifdef CONFIG_CONSOLE_POLL - .poll_get_char = serial8250_get_poll_char, - .poll_put_char = serial8250_put_poll_char, -#endif - - Any implementation specifics around creating a polling driver use the - #ifdef CONFIG_CONSOLE_POLL, as shown above. - Keep in mind that polling hooks have to be implemented in such a way - that they can be called from an atomic context and have to restore - the state of the UART chip on return such that the system can return - to normal when the debugger detaches. You need to be very careful - with any kind of lock you consider, because failing here is most likely - going to mean pressing the reset button. - - - - kgdboc and keyboards - The kgdboc driver contains logic to configure communications - with an attached keyboard. The keyboard infrastructure is only - compiled into the kernel when CONFIG_KDB_KEYBOARD=y is set in the - kernel configuration. - The core polled keyboard driver driver for PS/2 type keyboards - is in drivers/char/kdb_keyboard.c. This driver is hooked into the - debug core when kgdboc populates the callback in the array - called kdb_poll_funcs[]. The - kdb_get_kbd_char() is the top-level function which polls hardware - for single character input. - - - - kgdboc and kms - The kgdboc driver contains logic to request the graphics - display to switch to a text context when you are using - "kgdboc=kms,kbd", provided that you have a video driver which has a - frame buffer console and atomic kernel mode setting support. - - Every time the kernel - debugger is entered it calls kgdboc_pre_exp_handler() which in turn - calls con_debug_enter() in the virtual console layer. On resuming kernel - execution, the kernel debugger calls kgdboc_post_exp_handler() which - in turn calls con_debug_leave(). - Any video driver that wants to be compatible with the kernel - debugger and the atomic kms callbacks must implement the - mode_set_base_atomic, fb_debug_enter and fb_debug_leave operations. - For the fb_debug_enter and fb_debug_leave the option exists to use - the generic drm fb helper functions or implement something custom for - the hardware. The following example shows the initialization of the - .mode_set_base_atomic operation in - drivers/gpu/drm/i915/intel_display.c: - - -static const struct drm_crtc_helper_funcs intel_helper_funcs = { -[...] - .mode_set_base_atomic = intel_pipe_set_base_atomic, -[...] -}; - - - - Here is an example of how the i915 driver initializes the fb_debug_enter and fb_debug_leave functions to use the generic drm helpers in - drivers/gpu/drm/i915/intel_fb.c: - - -static struct fb_ops intelfb_ops = { -[...] - .fb_debug_enter = drm_fb_helper_debug_enter, - .fb_debug_leave = drm_fb_helper_debug_leave, -[...] -}; - - - - - - - - Credits - - The following people have contributed to this document: - - Amit Kaleamitkale@linsyssoft.com - Tom Rinitrini@kernel.crashing.org - - In March 2008 this document was completely rewritten by: - - Jason Wesseljason.wessel@windriver.com - - In Jan 2010 this document was updated to include kdb. - - Jason Wesseljason.wessel@windriver.com - - - - - diff --git a/Documentation/dev-tools/index.rst b/Documentation/dev-tools/index.rst index 07d881147ef3..4ac991dbddb7 100644 --- a/Documentation/dev-tools/index.rst +++ b/Documentation/dev-tools/index.rst @@ -23,6 +23,7 @@ whole; patches welcome! kmemleak kmemcheck gdb-kernel-debugging + kgdb .. only:: subproject and html diff --git a/Documentation/dev-tools/kgdb.rst b/Documentation/dev-tools/kgdb.rst new file mode 100644 index 000000000000..ea01541806c8 --- /dev/null +++ b/Documentation/dev-tools/kgdb.rst @@ -0,0 +1,930 @@ +================================================= +Using kgdb, kdb and the kernel debugger internals +================================================= + +:Author: Jason Wessel + +Introduction +============ + +The kernel has two different debugger front ends (kdb and kgdb) which +interface to the debug core. It is possible to use either of the +debugger front ends and dynamically transition between them if you +configure the kernel properly at compile and runtime. + +Kdb is simplistic shell-style interface which you can use on a system +console with a keyboard or serial console. You can use it to inspect +memory, registers, process lists, dmesg, and even set breakpoints to +stop in a certain location. Kdb is not a source level debugger, although +you can set breakpoints and execute some basic kernel run control. Kdb +is mainly aimed at doing some analysis to aid in development or +diagnosing kernel problems. You can access some symbols by name in +kernel built-ins or in kernel modules if the code was built with +``CONFIG_KALLSYMS``. + +Kgdb is intended to be used as a source level debugger for the Linux +kernel. It is used along with gdb to debug a Linux kernel. The +expectation is that gdb can be used to "break in" to the kernel to +inspect memory, variables and look through call stack information +similar to the way an application developer would use gdb to debug an +application. It is possible to place breakpoints in kernel code and +perform some limited execution stepping. + +Two machines are required for using kgdb. One of these machines is a +development machine and the other is the target machine. The kernel to +be debugged runs on the target machine. The development machine runs an +instance of gdb against the vmlinux file which contains the symbols (not +a boot image such as bzImage, zImage, uImage...). In gdb the developer +specifies the connection parameters and connects to kgdb. The type of +connection a developer makes with gdb depends on the availability of +kgdb I/O modules compiled as built-ins or loadable kernel modules in the +test machine's kernel. + +Compiling a kernel +================== + +- In order to enable compilation of kdb, you must first enable kgdb. + +- The kgdb test compile options are described in the kgdb test suite + chapter. + +Kernel config options for kgdb +------------------------------ + +To enable ``CONFIG_KGDB`` you should look under "Kernel hacking" / +"Kernel debugging" and select "KGDB: kernel debugger". + +While it is not a hard requirement that you have symbols in your vmlinux +file, gdb tends not to be very useful without the symbolic data, so you +will want to turn on ``CONFIG_DEBUG_INFO`` which is called "Compile the +kernel with debug info" in the config menu. + +It is advised, but not required, that you turn on the +``CONFIG_FRAME_POINTER`` kernel option which is called "Compile the +kernel with frame pointers" in the config menu. This option inserts code +to into the compiled executable which saves the frame information in +registers or on the stack at different points which allows a debugger +such as gdb to more accurately construct stack back traces while +debugging the kernel. + +If the architecture that you are using supports the kernel option +CONFIG_STRICT_KERNEL_RWX, you should consider turning it off. This +option will prevent the use of software breakpoints because it marks +certain regions of the kernel's memory space as read-only. If kgdb +supports it for the architecture you are using, you can use hardware +breakpoints if you desire to run with the CONFIG_STRICT_KERNEL_RWX +option turned on, else you need to turn off this option. + +Next you should choose one of more I/O drivers to interconnect debugging +host and debugged target. Early boot debugging requires a KGDB I/O +driver that supports early debugging and the driver must be built into +the kernel directly. Kgdb I/O driver configuration takes place via +kernel or module parameters which you can learn more about in the in the +section that describes the parameter "kgdboc". + +Here is an example set of .config symbols to enable or disable for kgdb: + +- # CONFIG_STRICT_KERNEL_RWX is not set + +- CONFIG_FRAME_POINTER=y + +- CONFIG_KGDB=y + +- CONFIG_KGDB_SERIAL_CONSOLE=y + +Kernel config options for kdb +----------------------------- + +Kdb is quite a bit more complex than the simple gdbstub sitting on top +of the kernel's debug core. Kdb must implement a shell, and also adds +some helper functions in other parts of the kernel, responsible for +printing out interesting data such as what you would see if you ran +"lsmod", or "ps". In order to build kdb into the kernel you follow the +same steps as you would for kgdb. + +The main config option for kdb is ``CONFIG_KGDB_KDB`` which is called +"KGDB_KDB: include kdb frontend for kgdb" in the config menu. In theory +you would have already also selected an I/O driver such as the +CONFIG_KGDB_SERIAL_CONSOLE interface if you plan on using kdb on a +serial port, when you were configuring kgdb. + +If you want to use a PS/2-style keyboard with kdb, you would select +CONFIG_KDB_KEYBOARD which is called "KGDB_KDB: keyboard as input +device" in the config menu. The CONFIG_KDB_KEYBOARD option is not used +for anything in the gdb interface to kgdb. The CONFIG_KDB_KEYBOARD +option only works with kdb. + +Here is an example set of .config symbols to enable/disable kdb: + +- # CONFIG_STRICT_KERNEL_RWX is not set + +- CONFIG_FRAME_POINTER=y + +- CONFIG_KGDB=y + +- CONFIG_KGDB_SERIAL_CONSOLE=y + +- CONFIG_KGDB_KDB=y + +- CONFIG_KDB_KEYBOARD=y + +Kernel Debugger Boot Arguments +============================== + +This section describes the various runtime kernel parameters that affect +the configuration of the kernel debugger. The following chapter covers +using kdb and kgdb as well as providing some examples of the +configuration parameters. + +Kernel parameter: kgdboc +------------------------ + +The kgdboc driver was originally an abbreviation meant to stand for +"kgdb over console". Today it is the primary mechanism to configure how +to communicate from gdb to kgdb as well as the devices you want to use +to interact with the kdb shell. + +For kgdb/gdb, kgdboc is designed to work with a single serial port. It +is intended to cover the circumstance where you want to use a serial +console as your primary console as well as using it to perform kernel +debugging. It is also possible to use kgdb on a serial port which is not +designated as a system console. Kgdboc may be configured as a kernel +built-in or a kernel loadable module. You can only make use of +``kgdbwait`` and early debugging if you build kgdboc into the kernel as +a built-in. + +Optionally you can elect to activate kms (Kernel Mode Setting) +integration. When you use kms with kgdboc and you have a video driver +that has atomic mode setting hooks, it is possible to enter the debugger +on the graphics console. When the kernel execution is resumed, the +previous graphics mode will be restored. This integration can serve as a +useful tool to aid in diagnosing crashes or doing analysis of memory +with kdb while allowing the full graphics console applications to run. + +kgdboc arguments +~~~~~~~~~~~~~~~~ + +Usage: ``kgdboc=[kms][[,]kbd][[,]serial_device][,baud]`` + +The order listed above must be observed if you use any of the optional +configurations together. + +Abbreviations: + +- kms = Kernel Mode Setting + +- kbd = Keyboard + +You can configure kgdboc to use the keyboard, and/or a serial device +depending on if you are using kdb and/or kgdb, in one of the following +scenarios. The order listed above must be observed if you use any of the +optional configurations together. Using kms + only gdb is generally not +a useful combination. + +Using loadable module or built-in +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +1. As a kernel built-in: + + Use the kernel boot argument: ``kgdboc=,[baud]`` + +2. As a kernel loadable module: + + Use the command: ``modprobe kgdboc kgdboc=,[baud]`` + + Here are two examples of how you might format the kgdboc string. The + first is for an x86 target using the first serial port. The second + example is for the ARM Versatile AB using the second serial port. + + 1. ``kgdboc=ttyS0,115200`` + + 2. ``kgdboc=ttyAMA1,115200`` + +Configure kgdboc at runtime with sysfs +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +At run time you can enable or disable kgdboc by echoing a parameters +into the sysfs. Here are two examples: + +1. Enable kgdboc on ttyS0 + + ``echo ttyS0 > /sys/module/kgdboc/parameters/kgdboc`` + +2. Disable kgdboc + + ``echo "" > /sys/module/kgdboc/parameters/kgdboc`` + +NOTE: You do not need to specify the baud if you are configuring the +console on tty which is already configured or open. + +More examples +^^^^^^^^^^^^^ + +You can configure kgdboc to use the keyboard, and/or a serial device +depending on if you are using kdb and/or kgdb, in one of the following +scenarios. + +1. kdb and kgdb over only a serial port + + ``kgdboc=[,baud]`` + + Example: ``kgdboc=ttyS0,115200`` + +2. kdb and kgdb with keyboard and a serial port + + ``kgdboc=kbd,[,baud]`` + + Example: ``kgdboc=kbd,ttyS0,115200`` + +3. kdb with a keyboard + + ``kgdboc=kbd`` + +4. kdb with kernel mode setting + + ``kgdboc=kms,kbd`` + +5. kdb with kernel mode setting and kgdb over a serial port + + ``kgdboc=kms,kbd,ttyS0,115200`` + +NOTE: Kgdboc does not support interrupting the target via the gdb remote +protocol. You must manually send a sysrq-g unless you have a proxy that +splits console output to a terminal program. A console proxy has a +separate TCP port for the debugger and a separate TCP port for the +"human" console. The proxy can take care of sending the sysrq-g for you. + +When using kgdboc with no debugger proxy, you can end up connecting the +debugger at one of two entry points. If an exception occurs after you +have loaded kgdboc, a message should print on the console stating it is +waiting for the debugger. In this case you disconnect your terminal +program and then connect the debugger in its place. If you want to +interrupt the target system and forcibly enter a debug session you have +to issue a Sysrq sequence and then type the letter ``g``. Then you +disconnect the terminal session and connect gdb. Your options if you +don't like this are to hack gdb to send the sysrq-g for you as well as +on the initial connect, or to use a debugger proxy that allows an +unmodified gdb to do the debugging. + +Kernel parameter: kgdbwait +-------------------------- + +The Kernel command line option ``kgdbwait`` makes kgdb wait for a +debugger connection during booting of a kernel. You can only use this +option if you compiled a kgdb I/O driver into the kernel and you +specified the I/O driver configuration as a kernel command line option. +The kgdbwait parameter should always follow the configuration parameter +for the kgdb I/O driver in the kernel command line else the I/O driver +will not be configured prior to asking the kernel to use it to wait. + +The kernel will stop and wait as early as the I/O driver and +architecture allows when you use this option. If you build the kgdb I/O +driver as a loadable kernel module kgdbwait will not do anything. + +Kernel parameter: kgdbcon +------------------------- + +The kgdbcon feature allows you to see printk() messages inside gdb while +gdb is connected to the kernel. Kdb does not make use of the kgdbcon +feature. + +Kgdb supports using the gdb serial protocol to send console messages to +the debugger when the debugger is connected and running. There are two +ways to activate this feature. + +1. Activate with the kernel command line option: + + ``kgdbcon`` + +2. Use sysfs before configuring an I/O driver + + ``echo 1 > /sys/module/kgdb/parameters/kgdb_use_con`` + + NOTE: If you do this after you configure the kgdb I/O driver, the + setting will not take effect until the next point the I/O is + reconfigured. + +IMPORTANT NOTE: You cannot use kgdboc + kgdbcon on a tty that is an +active system console. An example of incorrect usage is +``console=ttyS0,115200 kgdboc=ttyS0 kgdbcon`` + +It is possible to use this option with kgdboc on a tty that is not a +system console. + +Run time parameter: kgdbreboot +------------------------------ + +The kgdbreboot feature allows you to change how the debugger deals with +the reboot notification. You have 3 choices for the behavior. The +default behavior is always set to 0. + +1. echo -1 > /sys/module/debug_core/parameters/kgdbreboot + + Ignore the reboot notification entirely. + +2. echo 0 > /sys/module/debug_core/parameters/kgdbreboot + + Send the detach message to any attached debugger client. + +3. echo 1 > /sys/module/debug_core/parameters/kgdbreboot + + Enter the debugger on reboot notify. + +Using kdb +========= + +Quick start for kdb on a serial port +------------------------------------ + +This is a quick example of how to use kdb. + +1. Configure kgdboc at boot using kernel parameters: + + - ``console=ttyS0,115200 kgdboc=ttyS0,115200`` + + OR + + Configure kgdboc after the kernel has booted; assuming you are using + a serial port console: + + - ``echo ttyS0 > /sys/module/kgdboc/parameters/kgdboc`` + +2. Enter the kernel debugger manually or by waiting for an oops or + fault. There are several ways you can enter the kernel debugger + manually; all involve using the sysrq-g, which means you must have + enabled CONFIG_MAGIC_SYSRQ=y in your kernel config. + + - When logged in as root or with a super user session you can run: + + ``echo g > /proc/sysrq-trigger`` + + - Example using minicom 2.2 + + Press: ``Control-a`` + + Press: ``f`` + + Press: ``g`` + + - When you have telneted to a terminal server that supports sending + a remote break + + Press: ``Control-]`` + + Type in:\ ``send break`` + + Press: ``Enter`` + + Press: ``g`` + +3. From the kdb prompt you can run the "help" command to see a complete + list of the commands that are available. + + Some useful commands in kdb include: + + - lsmod -- Shows where kernel modules are loaded + + - ps -- Displays only the active processes + + - ps A -- Shows all the processes + + - summary -- Shows kernel version info and memory usage + + - bt -- Get a backtrace of the current process using dump_stack() + + - dmesg -- View the kernel syslog buffer + + - go -- Continue the system + +4. When you are done using kdb you need to consider rebooting the system + or using the "go" command to resuming normal kernel execution. If you + have paused the kernel for a lengthy period of time, applications + that rely on timely networking or anything to do with real wall clock + time could be adversely affected, so you should take this into + consideration when using the kernel debugger. + +Quick start for kdb using a keyboard connected console +------------------------------------------------------ + +This is a quick example of how to use kdb with a keyboard. + +1. Configure kgdboc at boot using kernel parameters: + + - ``kgdboc=kbd`` + + OR + + Configure kgdboc after the kernel has booted: + + - ``echo kbd > /sys/module/kgdboc/parameters/kgdboc`` + +2. Enter the kernel debugger manually or by waiting for an oops or + fault. There are several ways you can enter the kernel debugger + manually; all involve using the sysrq-g, which means you must have + enabled CONFIG_MAGIC_SYSRQ=y in your kernel config. + + - When logged in as root or with a super user session you can run: + + ``echo g > /proc/sysrq-trigger`` + + - Example using a laptop keyboard + + Press and hold down: ``Alt`` + + Press and hold down: ``Fn`` + + Press and release the key with the label: ``SysRq`` + + Release: ``Fn`` + + Press and release: ``g`` + + Release: ``Alt`` + + - Example using a PS/2 101-key keyboard + + Press and hold down: ``Alt`` + + Press and release the key with the label: ``SysRq`` + + Press and release: ``g`` + + Release: ``Alt`` + +3. Now type in a kdb command such as "help", "dmesg", "bt" or "go" to + continue kernel execution. + +Using kgdb / gdb +================ + +In order to use kgdb you must activate it by passing configuration +information to one of the kgdb I/O drivers. If you do not pass any +configuration information kgdb will not do anything at all. Kgdb will +only actively hook up to the kernel trap hooks if a kgdb I/O driver is +loaded and configured. If you unconfigure a kgdb I/O driver, kgdb will +unregister all the kernel hook points. + +All kgdb I/O drivers can be reconfigured at run time, if +``CONFIG_SYSFS`` and ``CONFIG_MODULES`` are enabled, by echo'ing a new +config string to ``/sys/module//parameter/