kernel-hacking.tmpl

Linux Kernel 2.6.9 for OMAP1710

    for other KERN_ values; these are interpreted by syslog as the
    level.  Special case: for printing an IP address use
   </para>

   <programlisting>
__u32 ipaddress;
printk(KERN_INFO "my ip: %d.%d.%d.%d\n", NIPQUAD(ipaddress));
   </programlisting>

   <para>
    <function>printk()</function> internally uses a 1K buffer and does
    not catch overruns.  Make sure that will be enough.
   </para>

   <note>
    <para>
     You will know when you are a real kernel hacker
     when you start typoing printf as printk in your user programs :)
    </para>
   </note>

   <!--- From the Lions book reader department --> 

   <note>
    <para>
     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.
    </para>
   </note>
  </sect1>

  <sect1 id="routines-copy">
   <title>
    <function>copy_[to/from]_user()</function>
    /
    <function>get_user()</function>
    /
    <function>put_user()</function>
    <filename class="headerfile">include/asm/uaccess.h</filename>
   </title>  

   <para>
    <emphasis>[SLEEPS]</emphasis>
   </para>

   <para>
    <function>put_user()</function> and <function>get_user()</function>
    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 <constant>-EFAULT</constant> or 0.
   </para>
   <para>
    <function>copy_to_user()</function> and
    <function>copy_from_user()</function> are more general: they copy
    an arbitrary amount of data to and from userspace.
    <caution>
     <para>
      Unlike <function>put_user()</function> and
      <function>get_user()</function>, they return the amount of
      uncopied data (ie. <returnvalue>0</returnvalue> still means
      success).
     </para>
    </caution>
    [Yes, this moronic interface makes me cringe.  Please submit a
    patch and become my hero --RR.]
   </para>
   <para>
    The functions may sleep implicitly. This should never be called
    outside user context (it makes no sense), with interrupts
    disabled, or a spinlock held.
   </para>
  </sect1>

  <sect1 id="routines-kmalloc">
   <title><function>kmalloc()</function>/<function>kfree()</function>
    <filename class="headerfile">include/linux/slab.h</filename></title>

   <para>
    <emphasis>[MAY SLEEP: SEE BELOW]</emphasis>
   </para>

   <para>
    These routines are used to dynamically request pointer-aligned
    chunks of memory, like malloc and free do in userspace, but
    <function>kmalloc()</function> takes an extra flag word.
    Important values:
   </para>

   <variablelist>
    <varlistentry>
     <term>
      <constant>
       GFP_KERNEL
      </constant>
     </term>
     <listitem>
      <para>
       May sleep and swap to free memory. Only allowed in user
       context, but is the most reliable way to allocate memory.
      </para>
     </listitem>
    </varlistentry>
    
    <varlistentry>
     <term>
      <constant>
       GFP_ATOMIC
      </constant>
     </term>
     <listitem>
      <para>
       Don't sleep. Less reliable than <constant>GFP_KERNEL</constant>,
       but may be called from interrupt context. You should
       <emphasis>really</emphasis> have a good out-of-memory
       error-handling strategy.
      </para>
     </listitem>
    </varlistentry>
    
    <varlistentry>
     <term>
      <constant>
       GFP_DMA
      </constant>
     </term>
     <listitem>
      <para>
       Allocate ISA DMA lower than 16MB. If you don't know what that
       is you don't need it.  Very unreliable.
      </para>
     </listitem>
    </varlistentry>
   </variablelist>

   <para>
    If you see a <errorname>kmem_grow: Called nonatomically from int
    </errorname> warning message you called a memory allocation function
    from interrupt context without <constant>GFP_ATOMIC</constant>.
    You should really fix that.  Run, don't walk.
   </para>

   <para>
    If you are allocating at least <constant>PAGE_SIZE</constant>
    (<filename class="headerfile">include/asm/page.h</filename>) bytes,
    consider using <function>__get_free_pages()</function>

    (<filename class="headerfile">include/linux/mm.h</filename>).  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.
   </para>

   <para>
    If you are allocating more than a page worth of bytes you can use
    <function>vmalloc()</function>.  It'll allocate virtual memory in
    the kernel map.  This block is not contiguous in physical memory,
    but the <acronym>MMU</acronym> 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 <function>alloc_bootmem()</function>
    routine.
   </para>

   <para>
    Before inventing your own cache of often-used objects consider
    using a slab cache in
    <filename class="headerfile">include/linux/slab.h</filename>
   </para>
  </sect1>

  <sect1 id="routines-current">
   <title><function>current</function>
    <filename class="headerfile">include/asm/current.h</filename></title>

   <para>
    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
    <emphasis>not NULL</emphasis> in interrupt context.
   </para>
  </sect1>

  <sect1 id="routines-udelay">
   <title><function>udelay()</function>/<function>mdelay()</function>
     <filename class="headerfile">include/asm/delay.h</filename>
     <filename class="headerfile">include/linux/delay.h</filename>
   </title>

   <para>
    The <function>udelay()</function> function can be used for small pauses.
    Do not use large values with <function>udelay()</function> as you risk
    overflow - the helper function <function>mdelay()</function> is useful
    here, or even consider <function>schedule_timeout()</function>.
   </para> 
  </sect1>
 
  <sect1 id="routines-endian">
   <title><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>
   </title>

   <para>
    The <function>cpu_to_be32()</function> 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: <function>be32_to_cpu()</function>, etc.
   </para>

   <para>
    There are two major variations of these functions: the pointer
    variation, such as <function>cpu_to_be32p()</function>, which take
    a pointer to the given type, and return the converted value.  The
    other variation is the "in-situ" family, such as
    <function>cpu_to_be32s()</function>, which convert value referred
    to by the pointer, and return void.
   </para> 
  </sect1>

  <sect1 id="routines-local-irqs">
   <title><function>local_irq_save()</function>/<function>local_irq_restore()</function>
    <filename class="headerfile">include/asm/system.h</filename>
   </title>

   <para>
    These routines disable hard interrupts on the local CPU, and
    restore them.  They are reentrant; saving the previous state in
    their one <varname>unsigned long flags</varname> argument.  If you
    know that interrupts are enabled, you can simply use
    <function>local_irq_disable()</function> and
    <function>local_irq_enable()</function>.
   </para>
  </sect1>

  <sect1 id="routines-softirqs">
   <title><function>local_bh_disable()</function>/<function>local_bh_enable()</function>
    <filename class="headerfile">include/linux/interrupt.h</filename></title>

   <para>
    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, tasklets
    and bottom halves from running on the current CPU.
   </para>
  </sect1>

  <sect1 id="routines-processorids">
   <title><function>smp_processor_id</function>()
    <filename class="headerfile">include/asm/smp.h</filename></title>
   
   <para>
    <function>smp_processor_id()</function> returns the current
    processor number, between 0 and <symbol>NR_CPUS</symbol> (the
    maximum number of CPUs supported by Linux, currently 32).  These
    values are not necessarily continuous.
   </para>
  </sect1>

  <sect1 id="routines-init">
   <title><type>__init</type>/<type>__exit</type>/<type>__initdata</type>
    <filename class="headerfile">include/linux/init.h</filename></title>

   <para>
    After boot, the kernel frees up a special section; functions
    marked with <type>__init</type> and data structures marked with
    <type>__initdata</type> are dropped after boot is complete (within
    modules this directive is currently ignored).  <type>__exit</type>
    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 <type>__init</type> to be exported to modules with 
    <function>EXPORT_SYMBOL()</function> - this will break.
   </para>
   <para>
   Static data structures marked as <type>__initdata</type> must be initialised
   (as opposed to ordinary static data which is zeroed BSS) and cannot be 
   <type>const</type>.
   </para> 

  </sect1>

  <sect1 id="routines-init-again">
   <title><function>__initcall()</function>/<function>module_init()</function>
    <filename class="headerfile">include/linux/init.h</filename></title>
   <para>
    Many parts of the kernel are well served as a module
    (dynamically-loadable parts of the kernel).  Using the
    <function>module_init()</function> and
    <function>module_exit()</function> macros it is easy to write code
    without #ifdefs which can operate both as a module or built into
    the kernel.
   </para>

   <para>
    The <function>module_init()</function> 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 <function>module_init()</function> macro
    becomes equivalent to <function>__initcall()</function>, which
    through linker magic ensures that the function is called on boot.
   </para>

   <para>
    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).  For modules, this is
    called in user context, with interrupts enabled, and the
    kernel lock held, so it can sleep.
   </para>
  </sect1>
  
  <sect1 id="routines-moduleexit">
   <title> <function>module_exit()</function>
    <filename class="headerfile">include/linux/init.h</filename> </title>

   <para>
    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.
   </para>
  </sect1>

 <!-- add info on new-style module refcounting here -->
 </chapter>

 <chapter id="queues">
  <title>Wait Queues
   <filename class="headerfile">include/linux/wait.h</filename>
  </title>
  <para>
   <emphasis>[SLEEPS]</emphasis>
  </para>

  <para>
   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
   <type>wait_queue_head_t</type>, and then processes which want to
   wait for that condition declare a <type>wait_queue_t</type>
   referring to themselves, and place that in the queue.
  </para>

  <sect1 id="queue-declaring">
   <title>Declaring</title>
   
   <para>
    You declare a <type>wait_queue_head_t</type> using the
    <function>DECLARE_WAIT_QUEUE_HEAD()</function> macro, or using the
    <function>init_waitqueue_head()</function> routine in your
    initialization code.
   </para>
  </sect1>
  
  <sect1 id="queue-waitqueue">
   <title>Queuing</title>
   
   <para>
    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:
    <function>wait_event_interruptible()</function>

    <filename class="headerfile">include/linux/sched.h</filename> The
    first argument is the wait queue head, and the second is an
    expression which is evaluated; the macro returns
    <returnvalue>0</returnvalue> when this expression is true, or
    <returnvalue>-ERESTARTSYS</returnvalue> if a signal is received.
    The <function>wait_event()</function> version ignores signals.
   </para>
   <para>
   Do not use the <function>sleep_on()</function> function family -
   it is very easy to accidentally introduce races; almost certainly
   one of the <function>wait_event()</function> family will do, or a
   loop around <function>schedule_timeout()</function>. If you choose
   to loop around <function>schedule_timeout()</function> remember
   you must set the task state (with 
   <function>set_current_state()</function>) on each iteration to avoid
   busy-looping.
   </para>
 
  </sect1>

  <sect1 id="queue-waking">
   <title>Waking Up Queued Tasks</title>
   
   <para>
    Call <function>wake_up()</function>

    <filename class="headerfile">include/linux/sched.h</filename>;,
    which will wake up every process in the queue.  The exception is
    if one has <constant>TASK_EXCLUSIVE</constant> set, in which case
    the remainder of the queue will not be woken.
   </para>
  </sect1>
 </chapter>

 <chapter id="atomic-ops">
  <title>Atomic Operations</title>

  <para>
   Certain operations are guaranteed atomic on all platforms.  The
   first class of operations work on <type>atomic_t</type>

   <filename class="headerfile">include/asm/atomic.h</filename>; this
   contains a signed integer (at least 24 bits long), and you must use
   these functions to manipulate or read atomic_t variables.
   <function>atomic_read()</function> and
   <function>atomic_set()</function> get and set the counter,
   <function>atomic_add()</function>,
   <function>atomic_sub()</function>,
   <function>atomic_inc()</function>,
   <function>atomic_dec()</function>, and
   <function>atomic_dec_and_test()</function> (returns
   <returnvalue>true</returnvalue> if it was decremented to zero).
  </para>

  <para>
   Yes.  It returns <returnvalue>true</returnvalue> (i.e. != 0) if the
   atomic variable is zero.
  </para>

  <para>
   Note that these functions are slower than normal arithmetic, and
   so should not be used unnecessarily.  On some platforms they
   are much slower, like 32-bit Sparc where they use a spinlock.
  </para>

  <para>
   The second class of atomic operations is atomic bit operations on a
   <type>long</type>, defined in

   <filename class="headerfile">include/asm/bitops.h</filename>.  These
   operations generally take a pointer to the bit pattern, and a bit
   number: 0 is the least significant bit.
   <function>set_bit()</function>, <function>clear_bit()</function>
   and <function>change_bit()</function> set, clear, and flip the
   given bit.  <function>test_and_set_bit()</function>,
   <function>test_and_clear_bit()</function> and
   <function>test_and_change_bit()</function> do the same thing,
   except return true if the bit was previously set; these are
   particularly useful for very simple locking.