Re: [PATCH v9 3/8] powerpc/kprobes: Mark newly allocated probes as RO

On Tue, Mar 16, 2021 at 5:44 PM Christophe Leroy
[off-list ref] wrote:

Le 16/03/2021 à 04:17, Jordan Niethe a écrit :

quoted

From: Russell Currey <redacted>

With CONFIG_STRICT_KERNEL_RWX=y and CONFIG_KPROBES=y, there will be one
W+X page at boot by default.  This can be tested with
CONFIG_PPC_PTDUMP=y and CONFIG_PPC_DEBUG_WX=y set, and checking the
kernel log during boot.

This text is confusing. I don't understand what is the status before the patch, and what is the
status after.

"there will be one ...", does it mean after the patch ?

quoted

Add an arch specific insn page allocator which returns RO pages if
STRICT_KERNEL_RWX is enabled. This page is only written to with
patch_instruction() which is able to write RO pages.

"an" or "the" arch specific insn page allocator ?

quoted

Reviewed-by: Daniel Axtens <redacted>
Signed-off-by: Russell Currey <redacted>
Signed-off-by: Christophe Leroy <redacted>
[jpn: Reword commit message, switch from vmalloc_exec(), add
       free_insn_page()]
Signed-off-by: Jordan Niethe <redacted>
---
v9: - vmalloc_exec() no longer exists
     - Set the page to RW before freeing it
---
  arch/powerpc/kernel/kprobes.c | 22 ++++++++++++++++++++++
  1 file changed, 22 insertions(+)

diff --git a/arch/powerpc/kernel/kprobes.c b/arch/powerpc/kernel/kprobes.c
index 01ab2163659e..bb7e4d321988 100644
--- a/arch/powerpc/kernel/kprobes.c
+++ b/arch/powerpc/kernel/kprobes.c

@@ -25,6 +25,8 @@
  #include <asm/sections.h>
  #include <asm/inst.h>
  #include <linux/uaccess.h>
+#include <linux/set_memory.h>
+#include <linux/vmalloc.h>

  DEFINE_PER_CPU(struct kprobe *, current_kprobe) = NULL;
  DEFINE_PER_CPU(struct kprobe_ctlblk, kprobe_ctlblk);

@@ -103,6 +105,26 @@ kprobe_opcode_t *kprobe_lookup_name(const char *name, unsigned int offset)
      return addr;
  }

+void *alloc_insn_page(void)
+{
+     void *page = vmalloc(PAGE_SIZE);

Can't do that on book3s/32, see https://github.com/linuxppc/linux/commit/6ca05532 and
https://github.com/linuxppc/linux/commit/7fbc22ce

Should do:
        return __vmalloc_node_range(size, 1, MODULES_VADDR, MODULES_END, GFP_KERNEL,
                                    PAGE_KERNEL_ROX, VM_FLUSH_RESET_PERMS, NUMA_NO_NODE,
                                    __builtin_return_address(0));


To keep it simple, you'll probably need to define MODULES_VADDR and MODULES_END as resp
VMALLOC_START and VMALLOC_END when they are not defined, maybe in asm/pgtable.h

Thank you, I had overlooked that. I will do it like that in the next revision.

quoted

+
+     if (!page)
+             return NULL;
+
+     set_memory_ro((unsigned long)page, 1);
+     set_memory_x((unsigned long)page, 1);
+
+     return page;
+}
+
+void free_insn_page(void *page)
+{
+     set_memory_nx((unsigned long)page, 1);
+     set_memory_rw((unsigned long)page, 1);
+     vfree(page);
+}
+
  int arch_prepare_kprobe(struct kprobe *p)
  {
      int ret = 0;