linux-stable.git/arch/s390/kernel/stacktrace.c, branch linux-rolling-stable

s390/stacktrace: Do not fallback to RA register

2025-12-14T10:03:58+00:00

The logic to fallback to the return address (RA) register value in
the topmost frame when stack tracing using back chain is broken in
multiple ways:

When assuming the RA register 14 has not been saved yet one must assume
that a new user stack frame has not been allocated either.  Therefore
the back chain would not contain the stack pointer (SP) at entry, but
the caller's SP at its entry instead.

Therefore when falling back to the RA register 14 value it would also be
necessary to fallback to the SP register 15 value.  Otherwise an invalid
combination of RA register 14 and caller's SP at its entry (from the
back chain) is used.

In the topmost frame the back chain contains either the caller's SP at
its entry (before having allocated a new stack frame in the prologue),
the SP at entry (after having allocated a new stack frame), or an
uninitialized value (during static/dynamic stack allocation).  In both
cases where the back chain is valid either the caller or prologue must
have saved its respective RA to the respective frame.  Therefore, if the
RA obtained from the frame pointed to by the back chain is invalid, this
does not indicate that the IP in the topmost frame is still early in the
prologue and the RA has not been saved.

Reviewed-by: Heiko Carstens 
Signed-off-by: Jens Remus 
Signed-off-by: Heiko Carstens

s390: Remove compat support

2025-11-17T10:10:38+00:00

There shouldn't be any 31 bit code around anymore that matters.
Remove the compat layer support required to run 31 bit code.

Reason for removal is code simplification and reduced test effort.

Note that this comes without any deprecation warnings added to config
options, or kernel messages, since most likely those would be ignored
anyway.

If it turns out there is still a reason to keep the compat layer this
can be reverted at any time in the future.

Reviewed-by: Arnd Bergmann 
Signed-off-by: Heiko Carstens

s390/asm-offsets: Remove ASM_OFFSETS_C

2025-03-31T10:20:39+00:00

Remove ASM_OFFSETS_C which is used as guard in thread_info.h to decide if
asm-offsets can be included or not.

There is no reason to include asm-offsets.h in thread_info.h anymore.
Remove the define and the not needed include. Explicitly include
asm-offsets.h in all header files which require it, and where it used
to be included implicitly via thread_info.h.

This reduces header dependencies.

Acked-by: Vasily Gorbik 
Signed-off-by: Heiko Carstens 
Signed-off-by: Vasily Gorbik

s390/stacktrace: Use break instead of return statement

2024-11-21T11:44:07+00:00

arch_stack_walk_user_common() contains a return statement instead of a
break statement in case store_ip() fails while trying to store a callchain
entry of a user space process.
This may lead to a missing pagefault_enable() call.

If this happens any subsequent page fault of the process won't be resolved
by the page fault handler and this in turn will lead to the process being
killed.

Use a break instead of a return statement to fix this.

Fixes: ebd912ff9919 ("s390/stacktrace: Merge perf_callchain_user() and arch_stack_walk_user()")
Cc: stable@vger.kernel.org
Reviewed-by: Jens Remus 
Signed-off-by: Heiko Carstens

s390/ftrace: Avoid calling unwinder in ftrace_return_address()

2024-08-27T18:16:48+00:00

ftrace_return_address() is called extremely often from
performance-critical code paths when debugging features like
CONFIG_TRACE_IRQFLAGS are enabled. For example, with debug_defconfig,
ftrace selftests on my LPAR currently execute ftrace_return_address()
as follows:

ftrace_return_address(0) - 0 times (common code uses __builtin_return_address(0) instead)
ftrace_return_address(1) - 2,986,805,401 times (with this patch applied)
ftrace_return_address(2) - 140 times
ftrace_return_address(>2) - 0 times

The use of __builtin_return_address(n) was replaced by return_address()
with an unwinder call by commit cae74ba8c295 ("s390/ftrace:
Use unwinder instead of __builtin_return_address()") because
__builtin_return_address(n) simply walks the stack backchain and doesn't
check for reaching the stack top. For shallow stacks with fewer than
"n" frames, this results in reads at low addresses and random
memory accesses.

While calling the fully functional unwinder "works", it is very slow
for this purpose. Moreover, potentially following stack switches and
walking past IRQ context is simply wrong thing to do for
ftrace_return_address().

Reimplement return_address() to essentially be __builtin_return_address(n)
with checks for reaching the stack top. Since the ftrace_return_address(n)
argument is always a constant, keep the implementation in the header,
allowing both GCC and Clang to unroll the loop and optimize it to the
bare minimum.

Fixes: cae74ba8c295 ("s390/ftrace: Use unwinder instead of __builtin_return_address()")
Cc: stable@vger.kernel.org
Reported-by: Sumanth Korikkar 
Reviewed-by: Heiko Carstens 
Acked-by: Sumanth Korikkar 
Signed-off-by: Vasily Gorbik

s390/stackstrace: Detect vdso stack frames

2024-05-14T11:37:07+00:00

Clear the backchain of the extra stack frame added by the vdso user wrapper
code. This allows the user stack walker to detect and skip the non-standard
stack frame. Without this an incorrect instruction pointer would be added
to stack traces, and stack frame walking would be continued with a more or
less random back chain.

Fixes: aa44433ac4ee ("s390: add USER_STACKTRACE support")
Reviewed-by: Jens Remus 
Signed-off-by: Heiko Carstens 
Signed-off-by: Alexander Gordeev

s390/stacktrace: Improve detection of invalid instruction pointers

2024-05-14T11:37:06+00:00

Add basic checks to identify invalid instruction pointers when walking
stack frames:

Instruction pointers must

- have even addresses
- be larger than mmap_min_addr
- lower than the asce_limit of the process

Alternatively it would also be possible to walk page tables similar to fast
GUP and verify that the mapping of the corresponding page is executable,
however that seems to be overkill.

Fixes: aa44433ac4ee ("s390: add USER_STACKTRACE support")
Reviewed-by: Jens Remus 
Signed-off-by: Heiko Carstens 
Signed-off-by: Alexander Gordeev

s390/stacktrace: Skip first user stack frame

2024-05-14T11:37:06+00:00

When walking user stack frames the first stack frame (where the stack
pointer points to) should be skipped: the return address of the current
function is saved in the previous stack frame, not the current stack frame,
which is allocated for to be called functions.

Fixes: aa44433ac4ee ("s390: add USER_STACKTRACE support")
Reviewed-by: Jens Remus 
Signed-off-by: Heiko Carstens 
Signed-off-by: Alexander Gordeev

s390/stacktrace: Merge perf_callchain_user() and arch_stack_walk_user()

2024-05-14T11:37:06+00:00

The two functions perf_callchain_user() and arch_stack_walk_user() are
nearly identical. Reduce code duplication and add a common helper which can
be called by both functions.

Fixes: aa44433ac4ee ("s390: add USER_STACKTRACE support")
Reviewed-by: Jens Remus 
Signed-off-by: Heiko Carstens 
Signed-off-by: Alexander Gordeev

s390/ftrace: Use unwinder instead of __builtin_return_address()

2024-04-29T15:33:30+00:00

Using __builtin_return_address(n) might return undefined values
when used with values of n outside of the stack. This was noticed
when __builtin_return_address() was called in ftrace on top level
functions like the interrupt handlers.

As this behaviour cannot be fixed, use the s390 stack unwinder and
remove the ftrace compilation flags for unwind_bc.c and stacktrace.c
to prevent the unwinding function polluting function traces.

Another advantage is that this also works with clang.

Reviewed-by: Heiko Carstens 
Signed-off-by: Sven Schnelle 
Signed-off-by: Alexander Gordeev