The code for injecting a signal into userland if a trapped instruction
fails emulation due to a _userland_ error (like an illegal address)
will be used more often with the next patch.
Factor out the core functionality into a separate function and use
that both for the existing trap handler and for the deprecated
instructions emulation.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
[catalin.marinas@arm.com: s/set_segfault/arm64_notify_segfault/]
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Several places open-code extraction of the EC field from an ESR_ELx
value, in subtly different ways. This is unfortunate duplication and
variation, and the precise logic used to extract the field is a
distraction.
This patch adds a new macro, ESR_ELx_EC(), to extract the EC field from
an ESR_ELx value in a consistent fashion.
Existing open-coded extractions in core arm64 code are moved over to the
new helper. KVM code is left as-is for the moment.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Huang Shijie <shijie.huang@arm.com>
Cc: Dave P Martin <dave.martin@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Currently dump_mem attempts to dump memory in 64-bit chunks when
reporting a failure in 64-bit code, or 32-bit chunks when reporting a
failure in 32-bit code. We added code to handle these two cases
separately in commit e147ae6d7f ("arm64: modify the dump mem for
64 bit addresses").
However, in all cases dump_mem is called, the failing context is a
kernel rather than user context. Additionally dump_mem is assumed to
only be used for kernel contexts, as internally it switches to
KERNEL_DS, and its callers pass kernel stack bounds.
This patch removes the redundant 32-bit chunk logic and associated
compat parameter, largely reverting the aforementioned commit. For the
call in __die(), the check of in_interrupt() is removed also, as __die()
is only called in response to faults from the kernel's exception level,
and thus the !user_mode(regs) check is sufficient. Were this not the
case, the used of task_stack_page(tsk) to generate the stack bounds
would be erroneous.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
If the kernel is set to show unhandled signals, and a user task does not
handle a SIGILL as a result of an instruction abort, we will attempt to
log the offending instruction with dump_instr before killing the task.
We use dump_instr to log the encoding of the offending userspace
instruction. However, dump_instr is also used to dump instructions from
kernel space, and internally always switches to KERNEL_DS before dumping
the instruction with get_user. When both PAN and UAO are in use, reading
a user instruction via get_user while in KERNEL_DS will result in a
permission fault, which leads to an Oops.
As we have regs corresponding to the context of the original instruction
abort, we can inspect this and only flip to KERNEL_DS if the original
abort was taken from the kernel, avoiding this issue. At the same time,
remove the redundant (and incorrect) comments regarding the order
dump_mem and dump_instr are called in.
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: James Morse <james.morse@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: <stable@vger.kernel.org> #4.6+
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reported-by: Vladimir Murzin <vladimir.murzin@arm.com>
Tested-by: Vladimir Murzin <vladimir.murzin@arm.com>
Fixes: 57f4959bad ("arm64: kernel: Add support for User Access Override")
Signed-off-by: Will Deacon <will.deacon@arm.com>
If we take an exception we don't expect (e.g. SError), we report this in
the bad_mode handler with pr_crit. Depending on the configured log
level, we may or may not log additional information in functions called
subsequently. Notably, the messages in dump_stack (including the CPU
number) are printed with KERN_DEFAULT and may not appear.
Some exceptions have an IMPLEMENTATION DEFINED ESR_ELx.ISS encoding, and
knowing the CPU number is crucial to correctly decode them. To ensure
that this is always possible, we should log the CPU number along with
the ESR_ELx value, so we are not reliant on subsequent logs or
additional printk configuration options.
This patch logs the CPU number in bad_mode such that it is possible for
a developer to decode these exceptions, provided access to sufficient
documentation.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reported-by: Al Grant <Al.Grant@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Martin <dave.martin@arm.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Switching between stacks is only valid if we are tracing ourselves while on the
irq_stack, so it is only valid when in current and non-preemptible context,
otherwise is is just zeroed off.
Fixes: 132cd887b5 ("arm64: Modify stack trace and dump for use with irq_stack")
Acked-by: James Morse <james.morse@arm.com>
Tested-by: James Morse <james.morse@arm.com>
Signed-off-by: Yang Shi <yang.shi@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Commit ac7b406c1a ("arm64: Use pr_* instead of printk") was a fairly
mindless s/printk/pr_*/ change driven by a complaint from checkpatch.
As is usual with such changes, this has led to some odd behaviour on
arm64:
* syslog now picks up the "pr_emerg" line from dump_backtrace, but not
the actual trace, which leads to a bunch of "kernel:Call trace:"
lines in the log
* __{pte,pmd,pgd}_error print at KERN_CRIT, as opposed to KERN_ERR
which is used by other architectures.
This patch restores the original printk behaviour for dump_backtrace
and downgrade the pgtable error macros to KERN_ERR.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Function graph tracer modifies a return address (LR) in a stack frame
to hook a function return. This will result in many useless entries
(return_to_handler) showing up in
a) a stack tracer's output
b) perf call graph (with perf record -g)
c) dump_backtrace (at panic et al.)
For example, in case of a),
$ echo function_graph > /sys/kernel/debug/tracing/current_tracer
$ echo 1 > /proc/sys/kernel/stack_trace_enabled
$ cat /sys/kernel/debug/tracing/stack_trace
Depth Size Location (54 entries)
----- ---- --------
0) 4504 16 gic_raise_softirq+0x28/0x150
1) 4488 80 smp_cross_call+0x38/0xb8
2) 4408 48 return_to_handler+0x0/0x40
3) 4360 32 return_to_handler+0x0/0x40
...
In case of b),
$ echo function_graph > /sys/kernel/debug/tracing/current_tracer
$ perf record -e mem:XXX:x -ag -- sleep 10
$ perf report
...
| | |--0.22%-- 0x550f8
| | | 0x10888
| | | el0_svc_naked
| | | sys_openat
| | | return_to_handler
| | | return_to_handler
...
In case of c),
$ echo function_graph > /sys/kernel/debug/tracing/current_tracer
$ echo c > /proc/sysrq-trigger
...
Call trace:
[<ffffffc00044d3ac>] sysrq_handle_crash+0x24/0x30
[<ffffffc000092250>] return_to_handler+0x0/0x40
[<ffffffc000092250>] return_to_handler+0x0/0x40
...
This patch replaces such entries with real addresses preserved in
current->ret_stack[] at unwind_frame(). This way, we can cover all
the cases.
Reviewed-by: Jungseok Lee <jungseoklee85@gmail.com>
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
[will: fixed minor context changes conflicting with irq stack bits]
Signed-off-by: Will Deacon <will.deacon@arm.com>
Function graph tracer modifies a return address (LR) in a stack frame
to hook a function's return. This will result in many useless entries
(return_to_handler) showing up in a call stack list.
We will fix this problem in a later patch ("arm64: ftrace: fix a stack
tracer's output under function graph tracer"). But since real return
addresses are saved in ret_stack[] array in struct task_struct,
unwind functions need to be notified of, in addition to a stack pointer
address, which task is being traced in order to find out real return
addresses.
This patch extends unwind functions' interfaces by adding an extra
argument of a pointer to task_struct.
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
This patch allows unwind_frame() to traverse from interrupt stack to task
stack correctly. It requires data from a dummy stack frame, created
during irq_stack_entry(), added by a later patch.
A similar approach is taken to modify dump_backtrace(), which expects to
find struct pt_regs underneath any call to functions marked __exception.
When on an irq_stack, the struct pt_regs is stored on the old task stack,
the location of which is stored in the dummy stack frame.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
[james.morse: merged two patches, reworked for per_cpu irq_stacks, and
no alignment guarantees, added irq_stack definitions]
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Unlike perf callchain relying on walk_stackframe(), dump_backtrace()
has its own backtrace logic. A major difference between them is the
moment a symbol is recorded. Perf writes down a symbol *before*
calling unwind_frame(), but dump_backtrace() prints it out *after*
unwind_frame(). As a result, the last valid symbol cannot be hooked
in case of dump_backtrace(). This patch addresses the issue as
synchronising dump_backtrace() with perf callchain.
A simple test and its results are as follows:
- crash trigger
$ sudo echo c > /proc/sysrq-trigger
- current status
Call trace:
[<fffffe00003dc738>] sysrq_handle_crash+0x24/0x30
[<fffffe00003dd2ac>] __handle_sysrq+0x128/0x19c
[<fffffe00003dd730>] write_sysrq_trigger+0x60/0x74
[<fffffe0000249fc4>] proc_reg_write+0x84/0xc0
[<fffffe00001f2638>] __vfs_write+0x44/0x104
[<fffffe00001f2e60>] vfs_write+0x98/0x1a8
[<fffffe00001f3730>] SyS_write+0x50/0xb0
- with this change
Call trace:
[<fffffe00003dc738>] sysrq_handle_crash+0x24/0x30
[<fffffe00003dd2ac>] __handle_sysrq+0x128/0x19c
[<fffffe00003dd730>] write_sysrq_trigger+0x60/0x74
[<fffffe0000249fc4>] proc_reg_write+0x84/0xc0
[<fffffe00001f2638>] __vfs_write+0x44/0x104
[<fffffe00001f2e60>] vfs_write+0x98/0x1a8
[<fffffe00001f3730>] SyS_write+0x50/0xb0
[<fffffe00000939ec>] el0_svc_naked+0x20/0x28
Note that this patch does not cover a case where MMU is disabled. The
last stack frame of swapper, for example, has PC in a form of physical
address. Unfortunately, a simple conversion using phys_to_virt() cannot
cover all scenarios since PC is retrieved from LR - 4, not LR. It is
a big tradeoff to change both head.S and unwind_frame() for only a few
of symbols in *.S. Thus, this hunk does not take care of the case.
Cc: AKASHI Takahiro <takahiro.akashi@linaro.org>
Cc: James Morse <james.morse@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Jungseok Lee <jungseoklee85@gmail.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The generic slowpath WARN implementation prints a backtrace, but
the report_bug() based implementation does not, opting to print the
registers instead which is generally not as useful.
Ideally, report_bug() should be fixed to make the behaviour more
consistent, but in the meantime this patch generates a backtrace
directly from the arm64 backend instead so that this functionality
is not lost with the migration to report_bug().
As a side-effect, the backtrace will be outside the oops end
marker, but that's hard to avoid without modifying generic code.
This patch can go away if report_bug() grows the ability in the
future to generate a backtrace directly or call an arch hook at the
appropriate time.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Currently, the minimal default BUG() implementation from asm-
generic is used for arm64.
This patch uses the BRK software breakpoint instruction to generate
a trap instead, similarly to most other arches, with the generic
BUG code generating the dmesg boilerplate.
This allows bug metadata to be moved to a separate table and
reduces the amount of inline code at BUG and WARN sites. This also
avoids clobbering any registers before they can be dumped.
To mitigate the size of the bug table further, this patch makes
use of the existing infrastructure for encoding addresses within
the bug table as 32-bit offsets instead of absolute pointers.
(Note that this limits the kernel size to 2GB.)
Traps are registered at arch_initcall time for aarch64, but BUG
has minimal real dependencies and it is desirable to be able to
generate bug splats as early as possible. This patch redirects
all debug exceptions caused by BRK directly to bug_handler() until
the full debug exception support has been initialised.
Signed-off-by: Dave Martin <Dave.Martin@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Nobody seems to be producing !SMP systems anymore, so this is just
becoming a source of kernel bugs, particularly if people want to use
coherent DMA with non-shared pages.
This patch forces CONFIG_SMP=y for arm64, removing a modest amount of
code in the process.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Commit 86dca36e6b introduced ratelimited usage for
'unhandled_signal' messages.
The commit checks the ratelimit irrespective of whether
the signal is handled or not, which is wrong and leads
to false reports like the below in dmesg :
__do_user_fault: 127 callbacks suppressed
Do the ratelimit check only if the signal is unhandled.
Fixes: 86dca36e6b ("arm64: use private ratelimit state along with show_unhandled_signals")
Cc: Vladimir Murzin <Vladimir.Murzin@arm.com>
Signed-off-by: Suzuki K. Poulose <suzuki.poulose@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
printk_ratelimit() shares the ratelimiting state with other callers what
may lead to scenarios where at the time we want to print out debug
information we already limited, so nothing appears in the dmesg - this
makes exception-trace quite poor helper in debugging.
Additionally, we have imbalance with some messages limited with global
ratelimit state and other messages limited with their private state
defined via pr_*_ratelimited().
To address this inconsistency show_unhandled_signals_ratelimited()
macro is introduced and caller sites are converted to use it.
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
To aid the developer when something triggers an unexpected exception,
decode the ESR_ELx.EC field when logging an ESR_ELx value. This doesn't
tell the developer the specifics of the exception encoded in the
remaining IL and ISS bits, but it can be helpful to distinguish between
exception classes (e.g. SError and a data abort) without having to
manually decode the field, which can be tiresome.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Will Deacon <will.deacon@arm.com>
Add support to register hooks for undefined instructions. The handlers
will be called when the undefined instruction and the processor state
(as contained in pstate) match criteria used at registration.
Signed-off-by: Punit Agrawal <punit.agrawal@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Use the global current_stack_pointer to get the value of the stack pointer.
This change supports being able to compile the kernel with both gcc and clang.
Signed-off-by: Behan Webster <behanw@converseincode.com>
Signed-off-by: Mark Charlebois <charlebm@gmail.com>
Reviewed-by: Olof Johansson <olof@lixom.net>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
This patch implements 4 levels of translation tables since 3 levels
of page tables with 4KB pages cannot support 40-bit physical address
space described in [1] due to the following issue.
It is a restriction that kernel logical memory map with 4KB + 3 levels
(0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
mapping for this region in map_mem function since __phys_to_virt for
this region reaches to address overflow.
If SoC design follows the document, [1], over 32GB RAM would be placed
from 544GB. Even 64GB system is supposed to use the region from 544GB
to 576GB for only 32GB RAM. Naturally, it would reach to enable 4 levels
of page tables to avoid hacking __virt_to_phys and __phys_to_virt.
However, it is recommended 4 levels of page table should be only enabled
if memory map is too sparse or there is about 512GB RAM.
References
----------
[1]: Principles of ARM Memory Maps, White Paper, Issue C
Signed-off-by: Jungseok Lee <jays.lee@samsung.com>
Reviewed-by: Sungjinn Chung <sungjinn.chung@samsung.com>
Acked-by: Kukjin Kim <kgene.kim@samsung.com>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
Reviewed-by: Steve Capper <steve.capper@linaro.org>
[catalin.marinas@arm.com: MEMBLOCK_INITIAL_LIMIT removed, same as PUD_SIZE]
[catalin.marinas@arm.com: early_ioremap_init() updated for 4 levels]
[catalin.marinas@arm.com: 48-bit VA depends on BROKEN until KVM is fixed]
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Tested-by: Jungseok Lee <jungseoklee85@gmail.com>
This patch fixed the following checkpatch complaint as using pr_*
instead of printk.
WARNING: printk() should include KERN_ facility level
Signed-off-by: Jungseok Lee <jays.lee@samsung.com>
Reviewed-by: Sungjinn Chung <sungjinn.chung@samsung.com>
Acked-by: Kukjin Kim <kgene.kim@samsung.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
For AArch32, bit 11 (WnR) of the FSR/ESR register is set when the fault
was caused by a write access and applications like Qemu rely on such
information being provided in sigcontext. This patch introduces the
ESR_EL1 tracking for the arm64 kernel faults and sets bit 11 accordingly
in compat sigcontext.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
The software breakpoint handlers are hooked in directly from ptrace,
which makes it difficult to add additional handlers for things like
kprobes and kgdb.
This patch moves the handling code into debug-monitors.c, where we can
dispatch to different debug subsystems more easily.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Rather than completely killing the kernel if we receive an esr value we
can't deal with in the el0 handlers, send the process a SIGILL and log
the esr value in the hope that we can debug it. If we receive a bad esr
from el1, we'll die() as before.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: stable@vger.kernel.org
Currently user faults (page, undefined instruction) are always reported
even though the user may have a signal handler for them. This patch adds
unhandled_signal() check together with printk_ratelimit() for these
cases.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Both dump_stack() and show_stack() are currently implemented by each
architecture. show_stack(NULL, NULL) dumps the backtrace for the
current task as does dump_stack(). On some archs, dump_stack() prints
extra information - pid, utsname and so on - in addition to the
backtrace while the two are identical on other archs.
The usages in arch-independent code of the two functions indicate
show_stack(NULL, NULL) should print out bare backtrace while
dump_stack() is used for debugging purposes when something went wrong,
so it does make sense to print additional information on the task which
triggered dump_stack().
There's no reason to require archs to implement two separate but mostly
identical functions. It leads to unnecessary subtle information.
This patch expands the dummy fallback dump_stack() implementation in
lib/dump_stack.c such that it prints out debug information (taken from
x86) and invokes show_stack(NULL, NULL) and drops arch-specific
dump_stack() implementations in all archs except blackfin. Blackfin's
dump_stack() does something wonky that I don't understand.
Debug information can be printed separately by calling
dump_stack_print_info() so that arch-specific dump_stack()
implementation can still emit the same debug information. This is used
in blackfin.
This patch brings the following behavior changes.
* On some archs, an extra level in backtrace for show_stack() could be
printed. This is because the top frame was determined in
dump_stack() on those archs while generic dump_stack() can't do that
reliably. It can be compensated by inlining dump_stack() but not
sure whether that'd be necessary.
* Most archs didn't use to print debug info on dump_stack(). They do
now.
An example WARN dump follows.
WARNING: at kernel/workqueue.c:4841 init_workqueues+0x35/0x505()
Hardware name: empty
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.9.0-rc1-work+ #9
0000000000000009 ffff88007c861e08 ffffffff81c614dc ffff88007c861e48
ffffffff8108f50f ffffffff82228240 0000000000000040 ffffffff8234a03c
0000000000000000 0000000000000000 0000000000000000 ffff88007c861e58
Call Trace:
[<ffffffff81c614dc>] dump_stack+0x19/0x1b
[<ffffffff8108f50f>] warn_slowpath_common+0x7f/0xc0
[<ffffffff8108f56a>] warn_slowpath_null+0x1a/0x20
[<ffffffff8234a071>] init_workqueues+0x35/0x505
...
v2: CPU number added to the generic debug info as requested by s390
folks and dropped the s390 specific dump_stack(). This loses %ksp
from the debug message which the maintainers think isn't important
enough to keep the s390-specific dump_stack() implementation.
dump_stack_print_info() is moved to kernel/printk.c from
lib/dump_stack.c. Because linkage is per objecct file,
dump_stack_print_info() living in the same lib file as generic
dump_stack() means that archs which implement custom dump_stack()
- at this point, only blackfin - can't use dump_stack_print_info()
as that will bring in the generic version of dump_stack() too. v1
The v1 patch broke build on blackfin due to this issue. The build
breakage was reported by Fengguang Wu.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Acked-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com> [s390 bits]
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Acked-by: Richard Kuo <rkuo@codeaurora.org> [hexagon bits]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix up all callers as they were before, with make one change: an
unsigned module taints the kernel, but doesn't turn off lockdep.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
The patch contains the exception entry code (kernel/entry.S), pt_regs
structure and related accessors, undefined instruction trapping and
stack tracing.
AArch64 Linux kernel (including kernel threads) runs in EL1 mode using
the SP1 stack. The vectors don't have a fixed address, only alignment
(2^11) requirements.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
