The suspend-imx6.S is introduced recently for suspend low-level assembly
code. Since function v7_cpu_resume() is only used by suspend support,
it makes sense to move the function into suspend-imx6.S, and control the
build of the file with CONFIG_SUSPEND option.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
When system enter suspend, we can set the DDR IO to
high-Z state to save DDR IOs' power consumption, this
operation can save many power(from ~26mA@1.5V to ~15mA@1.5V,
measured on i.MX6Q SabreSD board, R25) of DDR IOs. To
achieve that, we need to copy the suspend code to ocram
and run the low level hardware related code(set DDR IOs
to high-Z state) in ocram.
If there is no ocram space available, then system will
still do suspend in external DDR, hence no DDR IOs will
be set to high-Z.
The OCRAM usage layout is as below,
ocram suspend region(4K currently):
======================== high address ======================
.
.
.
^
^
^
imx6_suspend code
PM_INFO structure(imx6_cpu_pm_info)
======================== low address =======================
Reviewed-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Anson Huang <b20788@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Add cpuidle support for i.MX6SL, currently only support
two cpuidle levels(ARM wfi and WAIT mode), and add software
workaround for WAIT mode errata as below:
ERR005311 CCM: After exit from WAIT mode, unwanted interrupt(s) taken
during WAIT mode entry process could cause cache memory
corruption.
Software workaround:
To prevent this issue from occurring, software should ensure that
the ARM to IPG clock ratio is less than 12:5 (that is < 2.4x), before
entering WAIT mode.
Signed-off-by: Anson Huang <b20788@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Bit 17 of register CCM_CGPR is called INT_MEM_CLK_LPM as per the mx6
reference manual, so use this name instead.
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
When building a kernel image with only CONFIG_CPU_IDLE but no CONFIG_PM,
we will get the following link error.
LD init/built-in.o
arch/arm/mach-imx/built-in.o: In function `imx6q_enter_wait':
platform-spi_imx.c:(.text+0x25c0): undefined reference to `imx6q_set_lpm'
platform-spi_imx.c:(.text+0x25d4): undefined reference to `imx6q_set_lpm'
arch/arm/mach-imx/built-in.o: In function `imx6q_cpuidle_init':
platform-spi_imx.c:(.init.text+0x75d4): undefined reference to `imx6q_set_chicken_bit'
make[1]: *** [vmlinux] Error 1
Since pm-imx6q.c has been a collection of library functions that access
CCM low-power registers used by not only suspend but also cpuidle and
other drivers, let's build pm-imx6q.c independently of CONFIG_PM to fix
above error.
Reported-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Cc: stable@vger.kernel.org
Acked-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
Add machine support code for the Freescale IMX50 SoC.
The IMX50 is quite similar to the Freescale IMX53, and contains many of the
same periperhal hardware modules, at the same address offsets as the IMX53.
(Notable exceptions are that the IMX50 contains no CAN bus hardware, less
GPIO, no VPU, it does contain an Electrophoretic display controller though).
This support code uses some of the IMX53 setup code to reduce duplication
of what would be identical init IO setup.
Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
There is ~10% possibility that the following emergency restart command
fails to reboot imx6q.
$ echo b > /proc/sysrq-trigger
The IMX restart routine mxc_restart() assumes that it will always run on
primary core, and will call imx_src_prepare_restart() to disable
secondary cores in order to get them come to online in the following
boot. However, the assumption is only true for normal kernel_restart()
case where migrate_to_reboot_cpu() will be called to migrate to primary
core, but not necessarily true for emergency_restart() case. So when
emergency_restart() calls into mxc_restart() on any secondary core,
system will hang immediately once imx_src_prepare_restart() is called
to disabled secondary cores. Since emergency_restart() is defined as a
function that is safe to call in interrupt context, we cannot just call
migrate_to_reboot_cpu() to fix the issue.
Fortunately, we just found that the issue can be fixed at imx6q platform
level. We used to call imx_src_prepare_restart() to disable all
secondary cores before resetting hardware. Otherwise, the secondary
will fail come to online in the reboot. However, we recently found that
after commit 6050d18 (ARM: imx: reset core along with enable/disable
operation) comes to play, we do not need to reset the secondary cores
any more. That said, mxc_restart() now can run on any core to reboot
the system, as long as we remove the imx_src_prepare_restart() call from
mxc_restart().
So let's simply remove imx_src_prepare_restart() call to fix the above
emergency restart failure.
Reported-by: Jiada Wang <jiada_wang@mentor.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
There is a defect in imx6 LPM design. When SW tries to enter low power
mode with following sequence, the chip will enter low power mode before
A9 CPU execute WFI instruction:
1. Set CCM_CLPCR[1:0] to 2'b00;
2. ARM CPU enters WFI;
3. ARM CPU wakeup from an interrupt event, which is masked by GPC or not
visible to GPC, such as interrupt from local timer;
4. Set CCM_CLPCR[1:0] to 2'b01 or 2'b10;
5. ARM CPU execute WFI.
Before the last step, the chip will enter WAIT mode if CCM_CLPCR[1:0] is
set to 2'b01, or enter STOP mode if CCM_CLPCR[1:0] is set to 2'b10.
The patch implements a recommended workaround for this issue.
1. SW triggers irq #32(IOMUX) to be always pending manually by setting
IOMUX_GPR1_GINT bit;
2. SW should then unmask it in GPC before setting CCM LPM;
3. SW should mask it right after CCM LPM is set (bit0-1 of CCM_CLPCR).
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
The LPM (Low Power Mode) code that currently sits in imx6q clock driver
will be reused by imx6sl. Let's move it into pm-imx6q.c, so that we
can keep clock driver SoC specific and reuse pm-imx6q.c on imx6sl.
In order to avoid adding another ioremap for CCM block,
imx6q_pm_set_ccm_base() is created to let clock driver set up ccm_base
for pm code.
During the move, the unused CCGR macros get removed.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Since commit 70dc8a4 (checkpatch: warn when using extern with function
prototypes in .h files), we will get checkpatch warning when updating
common.h following the existing convention which has extern for function
prototypes.
Let's change the convention to not use extern with function prototypes
in this header.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
The imx6q_restart() works fine with normal reboot but will run into
problem with emergency reboot like sysrq-b. In that case, of_iomap()
gets called from interrupt context and hence triggers the BUG_ON in
__get_vm_area_node().
Actually, since commit c1e31d1 (ARM: imx: create
mxc_arch_reset_init_dt() for DT boot), imx6q/dl should try to use
mxc_restart() by calling mxc_arch_reset_init_dt() beforehand, where
things like of_iomap() can be done.
The patch updates mxc_restart() a little bit to get it work for imx6q/dl
and kill imx6q_restart() completely.
Reported-by: Nathan Lynch <nathan_lynch@mentor.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
The patch enables soc bus infrastructure and adds a function
imx_soc_device_init() to report soc info via soc device interface for
imx6qdl and imx6sl. With the support, user space can get soc related
info by looking at sysfs like below.
$ cat /sys/devices/soc0/machine
Freescale i.MX6 Quad SABRE Smart Device Board
$ cat /sys/devices/soc0/family
Freescale i.MX
$ cat /sys/devices/soc0/soc_id
i.MX6Q
$ cat /sys/devices/soc0/revision
1.2
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
The patch creates a common function imx_init_revision_from_anatop() by
merging imx6q_init_revision() and imx_anatop_get_digprog(), so that any
SoC that encodes revision info in anatop can use it to initialize
revision.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
It calls imx_set_soc_revision() to set up soc revision in
imx6q_init_revision(), and replaces all the occurrences of
imx6q_revision() with common helper imx_get_soc_revision().
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Similar to what we do for cpu type, the patch adds helper functions
imx_set_soc_revision() and imx_get_soc_revision() to maintain
imx_soc_revision in cpu.c.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
With arch/arm calling of_clk_init(NULL) from time_init(), we can now
remove custom .init_time hooks.
Signed-off-by: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Move anatop related (For USB) from board file to anatop driver
Signed-off-by: Peter Chen <peter.chen@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Current imx53_pm_init() implementation is incomplete as it lacks calling
suspend_set_ops().
Use a single imx5_pm_init() function to handle both mx51 and mx53.
This allows mx53 to enter in low-power mode.
Tested on a mx53qsb:
root@freescale /$ echo mem > /sys/power/state
PM: Syncing filesystems ... done.
mmc0: card e624 removed
Freezing user space processes ... (elapsed 0.001 seconds) done.
Freezing remaining freezable tasks ... (elapsed 0.001 seconds) done.
Suspending console(s) (use no_console_suspend to debug)
... (Press Power button)
PM: suspend of devices complete after 17.067 msecs
PM: suspend devices took 0.020 seconds
PM: late suspend of devices complete after 0.954 msecs
PM: noirq suspend of devices complete after 1.288 msecs
Disabling non-boot CPUs ...
PM: noirq resume of devices complete after 0.680 msecs
PM: early resume of devices complete after 0.914 msecs
PM: resume of devices complete after 44.955 msecs
PM: resume devices took 0.050 seconds
Restarting tasks ... done.
mmc0: host does not support reading read-only switch. assuming write-enable.
mmc0: new SDHC card at address e624
mmcblk0: mmc0:e624 SU04G 3.69 GiB
mmcblk0: p1 p2 p3
libphy: 63fec000.etherne:00 - Link is Down
libphy: 63fec000.etherne:00 - Link is Up - 100/Full
root@freescale /$
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Move imx6q L2 initialization function imx6q_init_l2cache() into
system.c, and rename it imx_init_l2cache(), so that other platforms
other than imx6q can also use the function.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Dirk Behme <dirk.behme@de.bosch.com>
Preparing to move the parsing of reboot= to generic kernel code forces
the change in reboot_mode handling to use the enum.
[akpm@linux-foundation.org: fix arch/arm/mach-socfpga/socfpga.c]
Signed-off-by: Robin Holt <holt@sgi.com>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Russ Anderson <rja@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Guan Xuetao <gxt@mprc.pku.edu.cn>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Instead of explicitly calling clock initialization functions, we can
declare the functions with CLK_OF_DECLARE() and then call common
of_clk_init() to have them invoked properly.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
The mxc_arch_reset_init() uses static mapping and calls clk_get_sys() to
get clock. It's suitable for non-DT boot but not for DT boot where
dynamic mapping and of_clk_get() should be used instead. Create
mxc_arch_reset_init_dt() as the DT variant of mxc_arch_reset_init(),
and change DT platforms to use it.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Commit bca7a5a (ARM: cpu hotplug: remove majority of cache flushing
from platforms) removes include of <asm/cacheflush.h> and hence
discovers a few indirect inclusion and declaration problems as below.
CC arch/arm/mach-imx/hotplug.o
In file included from arch/arm/mach-imx/hotplug.c:16:0:
arch/arm/mach-imx/common.h💯29: warning: ‘struct pt_regs’ declared inside parameter list [enabled by default]
arch/arm/mach-imx/common.h💯29: warning: its scope is only this definition or declaration, which is probably not what you want [enabled by default]
arch/arm/mach-imx/common.h:101:29: warning: ‘struct pt_regs’ declared inside parameter list [enabled by default]
arch/arm/mach-imx/hotplug.c: In function ‘imx_cpu_die’:
arch/arm/mach-imx/hotplug.c:53:2: error: implicit declaration of function ‘cpu_do_idle’ [-Werror=implicit-function-declaration]
arch/arm/mach-imx/hotplug.c: In function ‘imx_cpu_kill’:
arch/arm/mach-imx/hotplug.c:58:26: error: ‘jiffies’ undeclared (first use in this function)
arch/arm/mach-imx/hotplug.c:58:26: note: each undeclared identifier is reported only once for each function it appears in
arch/arm/mach-imx/hotplug.c:58:2: error: implicit declaration of function ‘msecs_to_jiffies’ [-Werror=implicit-function-declaration]
arch/arm/mach-imx/hotplug.c:61:3: error: implicit declaration of function ‘time_after’ [-Werror=implicit-function-declaration]
Fix them by adding the needed inclusion and declaration.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Olof Johansson <olof@lixom.net>
So it can be used in clk-imx6q.c for revision dependent clock tree setup.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
RBC is to control whether some ANATOP sub modules
can enter lpm mode when SOC is into STOP mode, if
RBC is enabled and PMIC_VSTBY_REQ is set, ANATOP
will have below behaviors:
1. Digital LDOs(CORE, SOC and PU) are bypassed;
2. Analog LDOs(1P1, 2P5, 3P0) are disabled;
As the 2P5 is necessary for DRAM IO pre-drive in
STOP mode, so we need to enable weak 2P5 in STOP
mode when 2P5 LDO is disabled.
For RBC settings, there are some rules as below
due to hardware design:
1. All interrupts must be masked during operating
RBC registers;
2. At least 2 CKIL(32K) cycles is needed after the
RBC setting is changed.
Signed-off-by: Anson Huang <b20788@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Anatop module have sereval configurations for user
to reduce the power consumption in suspend, provide
suspend/resume interface for further use and enable
fet_odrive to reduce CORE LDO leakage during suspend.
As we have a common anatop file, remove all the operations
of anatop module in other files, use anatop interfaces to
do that.
Signed-off-by: Anson Huang <b20788@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
* Clean up a couple of unneeded function declarations
* Remove imx specific cpufreq driver as generic cpufreq-cpu0 works well
as the replacement
* Remove platform ahci support
* Clean up unused ARCH/MACH Kconfig symbols
* Remove a couple of unused files
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
iQEcBAABAgAGBQJRZBpzAAoJEFBXWFqHsHzOfyAH/RdvhORI0rNhBAOrSxleWlMy
AMP9x6G5tXNK/fgtxopvCiS4VXw5MB01uqJWMVM9+hx6676oCXWc/L3I6iT5l87B
qPY/e9LyGJa5aOYczPuUT2sG+Ga8989GhOksA+L6kWHZ2df0BIB9rBRZnoPezEAx
u2ATdioyzfj2rAnDNFOxsw6oai265mowt/f8J/7EQfxgkAMcKov7/BpVac0po55R
tv+Hr3T7OBg3p/h5PYi5iAehVCToCKOD31QpnlhS+dmKS52w7tEO6T75INt7HzC8
CJT0zi5jJ3/tDJsdYJ1SS4MHK4U43y0oV7cw7sO7TzNreyOB6bzTWDqhdJY2oMM=
=N7xl
-----END PGP SIGNATURE-----
Merge tag 'imx-cleanup-3.10' into imx/soc
The imx cleanup for 3.10:
* Clean up a couple of unneeded function declarations
* Remove imx specific cpufreq driver as generic cpufreq-cpu0 works well
as the replacement
* Remove platform ahci support
* Clean up unused ARCH/MACH Kconfig symbols
* Remove a couple of unused files
Commit a1f1c7e (arm/imx6q: add suspend/resume support) added
declaration for a non-existing function pl310_get_save_ptr() by
mistake. Remove it.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
There is a sync issue with hotplug operation. It's possible that when
imx_cpu_kill gets running on primary core, the imx_cpu_die execution
on the core which is to be killed hasn't been finished yet. The problem
will very likely be hit when running suspend without no_console_suspend
setting on kernel cmdline.
It uses cpu jumping argument register to sync imx_cpu_die and
imx_cpu_kill. The register will be set in imx_cpu_die and imx_cpu_kill
will wait for the register being cleared to actually kill the cpu.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Cc: <stable@vger.kernel.org>
- It's based on imx-cleanup-3.9 to avoid conflicts.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
iQEcBAABAgAGBQJRCRzjAAoJEFBXWFqHsHzOOTYH/2YiAhCXDoovJ++leTTBKoFO
kRx2gvV0oxQYAjvRX4IiiwcuWjbCQI8ugiKxeH9Vn+jwHD8jKqzTINEUVQfKUud6
s8AI1/EqKwKnEFYFs/Fg6WxZHuHpoL/clpRB/5wZw1KoChmte2OnVU7d9vOt67iN
l0P4nPR18iwU5LFoDCnqMX50Dp9C56KEF2JhzBCQKW9TtRclwpcBNy6AsHb/TNOe
/nHN+Ku6/3xQEz2MPQ9QdmgFxnp9Pr8R6leJNg1Vg6UGRYrKxy0zVo+oFEXAZuUw
v6gnhdlvO6Ki/IKHcWkw7Q7pUDTuB6SttirQE9mBlPdc0V3kTU82SVtX9WAQhSs=
=rVol
-----END PGP SIGNATURE-----
Merge tag 'imx6q-cpudile-3.9' of git://git.linaro.org/people/shawnguo/linux-2.6 into next/soc
From Shawn Guo:
imx6q cpuidle support for 3.9
- It's based on imx-cleanup-3.9 to avoid conflicts.
* tag 'imx6q-cpudile-3.9' of git://git.linaro.org/people/shawnguo/linux-2.6:
ARM: imx6q: support WAIT mode using cpuidle
ARM: imx: move imx6q_cpuidle_driver into a separate file
ARM: imx: mask gpc interrupts initially
ARM: imx: return zero in case next event gets a large increment
ARM: imx: Remove mx508 support
ARM: imx: Remove mach-mx51_3ds board
ARM: imx: use debug_ll_io_init() for imx6q
ARM: imx: remove unused imx6q_clock_map_io()
ARM: mach-imx: Kconfig: Do not select Babbage for MACH_IMX51_DT
Signed-off-by: Olof Johansson <olof@lixom.net>
Still, two delete/change conflicts caused by imx/cleanup:
arch/arm/mach-imx/mach-mx50_rdp.c
arch/arm/mach-imx/mach-mx51_3ds.c
Add WAIT mode (ARM core clock gating) support to imx6q cpuidle driver.
As WAIT mode is broken on imx6q TO 1.0 and 1.1, it only enables the
support for revision 1.2 with chicken bit set.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Only mx508 based board is mach-mx50_rdp and it has been marked as BROKEN
for several releases.
mx508 currently lacks clock support.
In case someone needs to add mx508 support back, then the recommended approach
is to use device tree.
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Use debug_ll_io_init() to map low level debug port for imx6q, so that
arch/arm/mach-imx/lluart.c can be removed.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
imx6q_clock_map_io() becomes an empty function since imx6q clock driver
is moved to common clock framework. It's used nowhere now. Remove it.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
It's buggy to disable the cpu that is being hot-unplugged in .cpu_die
hook which runs on the cpu itself. Instead, it should be done in
.cpu_kill which runs on the thread (another cpu) that asks for shutting
down the cpu. Move imx_enable_cpu(cpu, false) call into .cpu_kill
hook, and leave the cpu to be hot-unplugged in WFI within .cpu_die,
so that we can get a more stable cpu hot-plug operation.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Now that the GIC initialization sets up the handle_arch_irq pointer, we
can remove it for all machines and make it static.
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Anton Vorontsov <avorontsov@mvista.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Sascha Hauer <kernel@pengutronix.de>
Cc: David Brown <davidb@codeaurora.org>
Cc: Daniel Walker <dwalker@fifo99.com>
Cc: Bryan Huntsman <bryanh@codeaurora.org>
Acked-by: Tony Lindgren <tony@atomide.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Magnus Damm <magnus.damm@gmail.com>
Cc: Dinh Nguyen <dinguyen@altera.com>
Cc: Shiraz Hashim <shiraz.hashim@st.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Cc: Srinidhi Kasagar <srinidhi.kasagar@stericsson.com>
Cc: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Kukjin Kim <kgene.kim@samsung.com>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Arnd Bergmann <arnd@arndb.de>
This adds a i.MX25 dt machine descriptor and changes the clock
support to optionally initialize from dt.
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
mxc_device_init() is a core_initcall function used to register devices
for mxc_aips_bus and mxc_ahb_bus, which are needed by gpio and dma
device registration.
Instead of being a core_initcall function, we have it called in soc
specific initialization function before gpio and dma devices get
registered, so that it will not be called for other platforms when
we enable multi-platform support for imx.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Rename mach-imx/include/mach/common.h to mach-imx/common.h and update
all users to include common.h rather than mach/common.h.
It also removes an unneeded inclusion to common.h in
mach-imx/devices/devices.c.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>