On Wed, 27 Mar 2024 09:04:39 PDT (-0700), samuel.holland@xxxxxxxxxx wrote:
Here are a few changes to minimize calls to stop_machine() and flush_icache_*() in the various text patching functions, as well as to simplify the code. This series is based on "[PATCH v3 0/2] riscv: fix patching with IPI"[1]. [1]: https://lore.kernel.org/linux-riscv/20240229121056.203419-1-alexghiti@xxxxxxxxxxxx/ Changes in v2: - Remove unnecessary line wrapping - Further simplify patch_insn_set()/patch_insn_write() loop conditions - Use min() instead of min_t() since both sides are unsigned long Samuel Holland (7): riscv: jump_label: Batch icache maintenance riscv: jump_label: Simplify assembly syntax riscv: kprobes: Use patch_text_nosync() for insn slots riscv: Simplify text patching loops riscv: Pass patch_text() the length in bytes riscv: Use offset_in_page() in text patching functions riscv: Remove extra variable in patch_text_nosync() arch/riscv/include/asm/jump_label.h | 4 +- arch/riscv/include/asm/patch.h | 2 +- arch/riscv/kernel/jump_label.c | 16 +++++-- arch/riscv/kernel/patch.c | 69 ++++++++++++++--------------- arch/riscv/kernel/probes/kprobes.c | 19 ++++---- arch/riscv/net/bpf_jit_comp64.c | 7 +-- 6 files changed, 63 insertions(+), 54 deletions(-)
I don't have any issues with this, but given that we've run into some possible text patching bug with this ftrace thing I'm just going to hold off until 6.11 for these. Maybe that's a bit too conservative, but with the bug only manifesting on HW it might be tough to sort out.
