Difference between revisions of "Kernel Debugging Tips"

From eLinux.org
Jump to: navigation, search
(Controlling console output)
 
(29 intermediate revisions by 5 users not shown)
Line 1: Line 1:
 
Here are some miscellaneous tips for debugging a kernel:
 
Here are some miscellaneous tips for debugging a kernel:
  
== Debugging early boot problems ==
+
== Using printks ==
=== Accessing the printk buffer after a silent hang on boot ===
+
To add your own debug message to the kernel, you can place a "printk()" in the kernel code.
Sometimes, if the kernel hangs early in the boot process, you get no
+
 
messages on the console before the hangHowever, there may still
+
See [[Debugging_by_printing#Usage| Debugging by printing -> Usage]] for more details.
be messages in the printk buffer, which can give you an idea of
+
 
where the problem is.
+
=== Log levels ===
 +
Each kernel message can be pre-pended with a tag indicating the importance of the message.
 +
 
 +
See [[Debugging_by_printing#Log_Levels| Debugging by printing -> Log_Levels]] for more details.
 +
 
 +
=== Adding timing information ===
 +
Sometimes, it is useful to add timing information to the printk values, so you can see when a particular event occurred.  The kernel includes an feature for doing
 +
this called printk times.
 +
 
 +
See the help for CONFIG_PRINTK_TIMES in the file lib/Kconfig.debug for more information on this feature.  This option is found on the "Kernel Hacking" menu when configuring the kernel.
 +
 
 +
The timestamps which are inserted into the printk output consist of seconds and microseconds, as absolute values from the start of machine operation (or from the start of kernel timekeeping).
 +
 
 +
There is also tool in the kernel source which will convert the timestamp values to relative values (so you can see the interval between events).  This tools is called show_delta and is located in the kernel 'scripts' directory.
 +
 
 +
See [[Printk Times]] for more information.
 +
 
 +
=== Viewing log messages ===
 +
The <tt>klogd</tt> program will extract messages from the kernel log buffer, and send them to the system log (which winds up in /var/log/messages on most systems).  This command runs in the background on most desktop or server systems, and continually transfers messages from the kernel log buffer to the system log.
 +
 
 +
You can view the contents of the log buffer directly, using the <tt>dmesg</tt> commandNote that by default <tt>dmesg</tt> displays the messages from the buffer, but does not remove them.  So this command can be run multiple times to view the kernel printk messages.  See the dmesg man page for more things you can do with this tool.
 +
 
 +
=== Controlling console output ===
 +
In order to have the kernel boot be less "noisy", or in order to boot more quickly, it is sometimes useful to control the amount of messages displayed to the console during boot.  You can do this by setting the kernel log level at boot time via a kernel command line option. See the "loglevel=" argument in [https://www.kernel.org/doc/Documentation/kernel-parameters.txt <tt>Documentation/kernel-parameters.txt</tt>].
 +
 
 +
You can turn off all messages using the kernel command line option "quiet". See [[Disable Console]] for information on how much time this can save at boot up.
 +
 
 +
Note that even if the log level is changed, or "quiet" is used, although the printk messages are not print to console, they are still entered into the log buffer, and they can still be extracted and displayed later using the <tt>dmesg</tt> command.
  
The kernel starts putting messages into the printk buffer as soon
+
=== Changing the size of the printk buffer ===
as it starts. They stay buffered there until the console code
+
See [[Debugging_by_printing#Internals_.2F_Changing_the_size_of_the_printk_buffer | Debugging by printing -> Internals / Changing the size of the printk buffer ]]
has a chance to initialize the console device (often the serial
 
port for embedded devices).  Eve though these messages
 
are no printed before the hang, it is still possible in some circumstances
 
to dump the printk buffer and see the messages.
 
  
Quinn Jensen writes:
+
== Using kernel symbols ==
 +
You can look up the source code for a function address using your toolchain's addr2line program.
 +
See [[Find a kernel function line]] or [[Addr2line for kernel debugging]].
  
Something I've found handy when the console is
+
== Using a kernel debugger ==
silent is to dump the printk buffer from the boot
+
You can use the in-kernel debugger: [[KDB]]
loader.  To do it you have to know how your boot
 
loader maps memory compared to the kernel.
 
  
==== Redboot example on a Freescale ADS board ====
+
You can use the in-kernel remote debugger: [[Kgdb]]
Quinn says: Here's what I do with Redboot on i.MX31:
 
  
fgrep printk_buf System.map
+
Also, you can use QEMU and gdb (and a high-level IDE like eclipse).
  
this shows the linked address of the printk_buf, e.g.:
+
See [http://issaris.blogspot.com/2007/12/download-linux-kernel-sourcecode-from.html Debugging the Linux kernel using Eclipse/CDT and Qemu] for a great article on using Eclipse (with the CDT plugin) to debug the Linux kernel.
  
c02338f0 b printk_buf.16194
+
== Debugging early boot problems ==
 +
See [[Debugging_by_printing#Debugging_early_boot_problems]]
  
The address "c02338f0" is in kernel virtual, which,
+
== Triggering a kernel event ==
in the case of i.MX31 ADS, redboot will have mapped
+
=== Overloading the sync system call ===
to 0x802338f0.  So, after resetting the target board,
+
Sometimes, it is nice to trigger an event to happen in the kernel from user space. Instead of creating infrastructure to handle a /proc event, an ioctl() or making a new syscall, it can be quick and easy to just overload an existing function.  One function not used very often is sync.  (I have found that the sync system call is not normally called by user space programs (or during standard linux booting).
but without letting it try to boot again, at the redboot
 
prompt,
 
  
dump -b 0x802338f0 -l 10000
+
It is quite easy to put a hook to your own kernel program in the sys_sync() routine (located in fs/sync.c) and cause it to execute by issuing 'sync' from the shell command line. This is handy as a temporary mechanism to test things that you have put in the kernel.
  
And you see the printk buffer that never got flushed
+
[[Category:Development Tools]]
to the UART.  Knowing what's there can be '''very'''
 
useful in debugging your console.
 
  
==== U-boot example on an OMAP OSK board ====
+
== Interpreting an Oops message ==
Tim Bird tried these steps and they worked:
+
When the kernel encounters an internal fault, it will print an Oops message.
 +
Here are some tips on using the Oops message to find the source of the problem.
 +
* See [http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=blob_plain;f=Documentation/oops-tracing.txt;hb=HEAD Documentation/oops-tracing.txt]
 +
* See [http://vmlinux.org/jocke/linux/howto-find-oops-location.shtml HOWTO find oops location] by Denis Vlasenko
  
  grep __log_buf System.map
+
== Compilation tricks for the kernel ==
 +
Sometimes, you want to modify how the compiler builds an individual kernel file.  The following are tips for doing tasks related to this.
 +
=== Build an individual file ===
 +
You can build an individual output object file, with:
 +
  make fs/buffer.o
  
or
+
This will build JUST fs/buffer.o (if it needs rebuilding) and not the entire kernel.  To force it to need re-building, use 'touch' on the associated source file:
 +
touch fs/buffer.c
  
  grep __log_buf /proc/kallsyms
+
=== Create the preprocessed file for an individual source file ===
 +
Using the same technique, you can create the preprocessed file for a C source file. This is useful if you're having trouble tracking down macro expansion or where defines/prototypes are coming from exactly.
 +
make fs/buffer.i
  
These show:
+
=== Create the assembly file for an individual source file ===
 +
Using the same technique, you can create the assembly file for a C source file.  This is useful to get an idea what actual machine instructions are generated from the C source code.
 +
make fs/buffer.s
  
  c0352d88 B __log_buf
+
Another way to get the raw assembly, is to dump the object file using 'objdump'
 +
  objdump -d fs/buffer.o > fs/buffer.disassem
  
In the case of the OSK, this address maps to
+
This will produce a disassembly of the object file, which should show how the assembly was translated into machine instructions.
0x10352d88. So I reset the target board and use the following:
 
  
OMAP5912 OSK # md 10352d88
+
If the object has been compiled with debug symbols (using '-g'), then you might get more information using the '-S' option with objdump:
10352d88: 4c3e353c 78756e69 72657620 6e6f6973    <5>Linux version
+
  objdump -S -d fs/buffer.o >fs/buffer.disassem
10352d98: 362e3220 2e32322e 612d3631 6e5f706c    2.6.22.16-alp_n
 
10352da8: 7428206c 64726962 6d697440 6b736564    l (tbird@timdesk
 
10352db8: 2e6d612e 796e6f73 6d6f632e 67282029    .am.sony.com) (g
 
10352dc8: 76206363 69737265 33206e6f 342e342e    cc version 3.4.4
 
10352dd8: 34232029 45525020 54504d45 65755420    ) #4 PREEMPT Tue
 
  ...
 
  
=== Using CONFIG_DEBUG_LL and printascii() ===
+
You can also request that the toolchain show mixed source and assembly, by passing extra flags:
If the kernel fails before the serial console is enabled, you can use CONFIG_DEBUG_LL
+
make EXTRA_CFLAGS="-g -Wa,-a,-ad -fverbose-asm" fs/buffer.o >fs/buffer.mixed
to change the way the printk code outputs characters.
 
  
Here is an e-mail exchange seen on the linux-embedded mailing list (with answer by George Davis):
+
=== Alter the flags for a compilation ===
<pre>
+
Sometimes, you need to alter the compilation flags for an individual file. There are two ways to do this. One is to add the extra flags on the make command line:
> When we try to boot a 2.6.21 kernel after uncompressing the kernel the boot process dies somehow.
 
> We've figured out so far that the kernel dies somewhere between  the gunzip and start_kernel.
 
  
Try enabling DEBUG_LL to see if it's an machine ID error. If you see:
+
make EXTRA_CFLAGS="-g -finstrument-functions" fs/buffer.o
  
Error: unrecognized/unsupported processor variant.
+
This will work if the flags can be appended to the regular set of C flags used for compiling the object.
  
or:
+
However, if you need to do something more complicated, like removing or modifying flags, then you can build your own command line by hand.  To do this, it is easiest to have 'make' produce the default compilation command (which will be several lines long), then copy, paste and edit it, to run on the command line directly.  To see the exact compile commands used to compile a particular object, use the V=1 option with the kernel build system:
  
Error: unrecognized/unsupported machine ID...
+
make V=1 fs/buffer.o
  
Then you either don't have proper processor support enabled for your target
+
For me, this produced something like this:
or your bootloader is passing in the wrong machine number.
 
  
If you still don't see anything, try hacking printk.c to call
+
mipsel-linux-gcc -Wp,-MD,fs/.buffer.o.d -nostdinc -isystem /home/usr/local/mipsel-linux-glibc/bin/../lib/gcc/mipsisa32el-linux/3.4.3/include -D__KERNEL__ -Iinclude -Iinclude2 -I/home/tbird/work/linux/include -I/home/tbird/work/linux/fs -Ifs -Wall -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -ffreestanding -O2 -fomit-frame-pointer -g -I/home/tbird/work/linux/ -I /home/tbird/work/linux/include/asm/gcc -G 0 -mno-abicalls -fno-pic -pipe -finline-limit=100000 -mabi=32 -march=mips32r2 -Wa,-32 -Wa,-march=mips32r2 -Wa,-mips32r2 -Wa,--trap -I/home/tbird/work/linux/include/asm-mips/ati -Iinclude/asm-mips/ati -I/home/tbird/work/linux/include/asm-mips/mach-generic -Iinclude/asm-mips/mach-generic -Wdeclaration-after-statement  -DKBUILD_BASENAME=buffer -DKBUILD_MODNAME=buffer -c -o fs/buffer.o /home/tbird/work/linux/fs/buffer.c
printascii() (enabled for the DEBUG_LL case) to print directly to the
 
serial port w/o a driver, etc.,. You can find more details on these
 
low-level debugging hacks via a little googling...
 
</pre>
 
 
[[Category:Development Tools]]
 

Latest revision as of 04:21, 19 April 2015

Here are some miscellaneous tips for debugging a kernel:

Using printks

To add your own debug message to the kernel, you can place a "printk()" in the kernel code.

See Debugging by printing -> Usage for more details.

Log levels

Each kernel message can be pre-pended with a tag indicating the importance of the message.

See Debugging by printing -> Log_Levels for more details.

Adding timing information

Sometimes, it is useful to add timing information to the printk values, so you can see when a particular event occurred. The kernel includes an feature for doing this called printk times.

See the help for CONFIG_PRINTK_TIMES in the file lib/Kconfig.debug for more information on this feature. This option is found on the "Kernel Hacking" menu when configuring the kernel.

The timestamps which are inserted into the printk output consist of seconds and microseconds, as absolute values from the start of machine operation (or from the start of kernel timekeeping).

There is also tool in the kernel source which will convert the timestamp values to relative values (so you can see the interval between events). This tools is called show_delta and is located in the kernel 'scripts' directory.

See Printk Times for more information.

Viewing log messages

The klogd program will extract messages from the kernel log buffer, and send them to the system log (which winds up in /var/log/messages on most systems). This command runs in the background on most desktop or server systems, and continually transfers messages from the kernel log buffer to the system log.

You can view the contents of the log buffer directly, using the dmesg command. Note that by default dmesg displays the messages from the buffer, but does not remove them. So this command can be run multiple times to view the kernel printk messages. See the dmesg man page for more things you can do with this tool.

Controlling console output

In order to have the kernel boot be less "noisy", or in order to boot more quickly, it is sometimes useful to control the amount of messages displayed to the console during boot. You can do this by setting the kernel log level at boot time via a kernel command line option. See the "loglevel=" argument in Documentation/kernel-parameters.txt.

You can turn off all messages using the kernel command line option "quiet". See Disable Console for information on how much time this can save at boot up.

Note that even if the log level is changed, or "quiet" is used, although the printk messages are not print to console, they are still entered into the log buffer, and they can still be extracted and displayed later using the dmesg command.

Changing the size of the printk buffer

See Debugging by printing -> Internals / Changing the size of the printk buffer

Using kernel symbols

You can look up the source code for a function address using your toolchain's addr2line program. See Find a kernel function line or Addr2line for kernel debugging.

Using a kernel debugger

You can use the in-kernel debugger: KDB

You can use the in-kernel remote debugger: Kgdb

Also, you can use QEMU and gdb (and a high-level IDE like eclipse).

See Debugging the Linux kernel using Eclipse/CDT and Qemu for a great article on using Eclipse (with the CDT plugin) to debug the Linux kernel.

Debugging early boot problems

See Debugging_by_printing#Debugging_early_boot_problems

Triggering a kernel event

Overloading the sync system call

Sometimes, it is nice to trigger an event to happen in the kernel from user space. Instead of creating infrastructure to handle a /proc event, an ioctl() or making a new syscall, it can be quick and easy to just overload an existing function. One function not used very often is sync. (I have found that the sync system call is not normally called by user space programs (or during standard linux booting).

It is quite easy to put a hook to your own kernel program in the sys_sync() routine (located in fs/sync.c) and cause it to execute by issuing 'sync' from the shell command line. This is handy as a temporary mechanism to test things that you have put in the kernel.

Interpreting an Oops message

When the kernel encounters an internal fault, it will print an Oops message. Here are some tips on using the Oops message to find the source of the problem.

Compilation tricks for the kernel

Sometimes, you want to modify how the compiler builds an individual kernel file. The following are tips for doing tasks related to this.

Build an individual file

You can build an individual output object file, with:

make fs/buffer.o

This will build JUST fs/buffer.o (if it needs rebuilding) and not the entire kernel. To force it to need re-building, use 'touch' on the associated source file:

touch fs/buffer.c

Create the preprocessed file for an individual source file

Using the same technique, you can create the preprocessed file for a C source file. This is useful if you're having trouble tracking down macro expansion or where defines/prototypes are coming from exactly.

make fs/buffer.i

Create the assembly file for an individual source file

Using the same technique, you can create the assembly file for a C source file. This is useful to get an idea what actual machine instructions are generated from the C source code.

make fs/buffer.s

Another way to get the raw assembly, is to dump the object file using 'objdump'

objdump -d fs/buffer.o > fs/buffer.disassem

This will produce a disassembly of the object file, which should show how the assembly was translated into machine instructions.

If the object has been compiled with debug symbols (using '-g'), then you might get more information using the '-S' option with objdump:

objdump -S -d fs/buffer.o >fs/buffer.disassem

You can also request that the toolchain show mixed source and assembly, by passing extra flags:

make EXTRA_CFLAGS="-g -Wa,-a,-ad -fverbose-asm" fs/buffer.o >fs/buffer.mixed

Alter the flags for a compilation

Sometimes, you need to alter the compilation flags for an individual file. There are two ways to do this. One is to add the extra flags on the make command line:

make EXTRA_CFLAGS="-g -finstrument-functions" fs/buffer.o

This will work if the flags can be appended to the regular set of C flags used for compiling the object.

However, if you need to do something more complicated, like removing or modifying flags, then you can build your own command line by hand. To do this, it is easiest to have 'make' produce the default compilation command (which will be several lines long), then copy, paste and edit it, to run on the command line directly. To see the exact compile commands used to compile a particular object, use the V=1 option with the kernel build system:

make V=1 fs/buffer.o

For me, this produced something like this:

mipsel-linux-gcc -Wp,-MD,fs/.buffer.o.d -nostdinc -isystem /home/usr/local/mipsel-linux-glibc/bin/../lib/gcc/mipsisa32el-linux/3.4.3/include -D__KERNEL__ -Iinclude -Iinclude2 -I/home/tbird/work/linux/include -I/home/tbird/work/linux/fs -Ifs -Wall -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -ffreestanding -O2 -fomit-frame-pointer -g -I/home/tbird/work/linux/ -I /home/tbird/work/linux/include/asm/gcc -G 0 -mno-abicalls -fno-pic -pipe -finline-limit=100000 -mabi=32 -march=mips32r2 -Wa,-32 -Wa,-march=mips32r2 -Wa,-mips32r2 -Wa,--trap -I/home/tbird/work/linux/include/asm-mips/ati -Iinclude/asm-mips/ati -I/home/tbird/work/linux/include/asm-mips/mach-generic -Iinclude/asm-mips/mach-generic -Wdeclaration-after-statement -DKBUILD_BASENAME=buffer -DKBUILD_MODNAME=buffer -c -o fs/buffer.o /home/tbird/work/linux/fs/buffer.c