It's been quite a while since the last iteration, there were a few things to hunt down and fix.
The first was the way that I was updating PTEs was using set_pte_at() unsafely - now, each page is updated by clearing -> flushing -> setting. This should be generic across all MMUs, I know that there are some potential inefficiencies - for example, Hash flushes the entire PID regardless of the given page range - but I don't think it's a very big deal. The next was that there is an errant page that was tricky to hunt down, it turned out to be that kprobes never get marked RO after creation, leading to (at least) one W+X page present in the kernel, even with both STRICT_KERNEL_RWX and STRICT_MODULE_RWX on. I added a debugfs handler to call ptdump_check_wx() to facilitate making sure module RWX continues to work after boot. There's more detail in the final patch about exactly how "on by default" module RWX is, but it doesn't really matter until STRICT_KERNEL_RWX is on by default too. Thanks to Nick Piggin, Michael Ellerman, Daniel Axtens, and others for helping. Christophe, I did test this in qemu mac99 so hopefully it works for all 32bit, I'm sure you'll let me know if it doesn't. Would appreciate an ack from Joel to enable this in skiroot_defconfig. Russell Currey (4): powerpc/mm: Implement set_memory() routines powerpc/kprobes: Mark newly allocated probes as RO powerpc/mm/ptdump: debugfs handler for W+X checks at runtime powerpc: Enable STRICT_MODULE_RWX arch/powerpc/Kconfig | 2 + arch/powerpc/configs/skiroot_defconfig | 1 + arch/powerpc/include/asm/set_memory.h | 32 ++++++++++++++ arch/powerpc/kernel/kprobes.c | 3 ++ arch/powerpc/mm/Makefile | 1 + arch/powerpc/mm/pageattr.c | 60 ++++++++++++++++++++++++++ arch/powerpc/mm/ptdump/ptdump.c | 31 ++++++++++--- 7 files changed, 124 insertions(+), 6 deletions(-) create mode 100644 arch/powerpc/include/asm/set_memory.h create mode 100644 arch/powerpc/mm/pageattr.c -- 2.23.0
