| Message ID | 20231124110511.2795958-1-mark.rutland@arm.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | arm64: stacktrace: add kunwind_stack_walk() | expand |
On Fri, Nov 24, 2023 at 11:05:09AM +0000, Mark Rutland wrote: > Puranjay has a need for this for BPF: > https://lore.kernel.org/linux-arm-kernel/20230917000045.56377-1-puranjay12@gmail.com/ > Hence I'm sending this as-is as preparatory rework. This will also be useful for joining GCS up with the stacktrace infrastructure (for reliability checks if nothing else): Reviewed-by: Mark Brown <broonie@kernel.org>
On 11/24/23 05:05, Mark Rutland wrote: > Currently arm64 uses the generic arch_stack_walk() interface for all > stack walking code. This only passes a PC value and cookie to the unwind > callback, whereas we'd like to pass some additional information in some > cases. For example, the BPF exception unwinder wants the FP, for > reliable stacktrace we'll want to perform additional checks on other > portions of unwind state, and we'd like to expand the information > printed by dump_backtrace() to include provenance and reliability > information. > > These patches refactor arm64's stacktrace code into a new > kunwind_stack_walk() function that provides the full unwind state to > callback functions. The existing arch_stack_walk() interface is > unchanged, and is implemented atop kunwind_stack_walk(). > > I had originally intended to send this with additional patches that > would have dump_backtrace() use this to identify and report exception > boundaries and fgraph/kretprobes PC recovery, but due to LPC and bug > hunting over the last few weeks I haven't managed to get all of that > ready just yet. > > Puranjay has a need for this for BPF: > > https://lore.kernel.org/linux-arm-kernel/20230917000045.56377-1-puranjay12@gmail.com/ > > Hence I'm sending this as-is as preparatory rework. > > Thanks, > Mark. > > Mark Rutland (2): > arm64: stacktrace: factor out kernel unwind state > arm64: stacktrace: factor out kunwind_stack_walk() > > arch/arm64/include/asm/stacktrace/common.h | 19 +-- > arch/arm64/include/asm/stacktrace/nvhe.h | 2 +- > arch/arm64/kernel/stacktrace.c | 146 ++++++++++++++------- > 3 files changed, 104 insertions(+), 63 deletions(-) > Reviewed-by: Madhavan T. Venkataraman <madvenka@linux.microsoft.com>
On Fri, 24 Nov 2023 11:05:09 +0000, Mark Rutland wrote: > Currently arm64 uses the generic arch_stack_walk() interface for all > stack walking code. This only passes a PC value and cookie to the unwind > callback, whereas we'd like to pass some additional information in some > cases. For example, the BPF exception unwinder wants the FP, for > reliable stacktrace we'll want to perform additional checks on other > portions of unwind state, and we'd like to expand the information > printed by dump_backtrace() to include provenance and reliability > information. > > [...] Applied to arm64 (for-next/stacktrace), thanks! [1/2] arm64: stacktrace: factor out kernel unwind state https://git.kernel.org/arm64/c/1beef60e7d6b [2/2] arm64: stacktrace: factor out kunwind_stack_walk() https://git.kernel.org/arm64/c/1aba06e7b2b4 Cheers,