On 2022/9/30 17:47, Naveen N. Rao wrote:

Li Huafei wrote:

quoted

I found a null pointer reference in arch_prepare_kprobe():

Good find!

Hi Naveen,

Thank you for the review.

quoted

  # echo 'p cmdline_proc_show' > kprobe_events
  # echo 'p cmdline_proc_show+16' >> kprobe_events

I think we should extend multiple_kprobes selftest to also place
contiguous probes to catch such errors.

Yes. But each architecture implementation is different and it looks a
little difficult to decide which offsets need to be tested.

quoted

  [   67.278533][  T122] Kernel attempted to read user page (0) -
exploitattempt? (uid: 0)
  [   67.279326][  T122] BUG: Kernel NULL pointer dereference on read
at 0x00000000
  [   67.279738][  T122] Faulting instruction address: 0xc000000000050bfc
  [   67.280486][  T122] Oops: Kernel access of bad area, sig: 11 [#1]
  [   67.280846][  T122] LE PAGE_SIZE=64K MMU=Radix SMP NR_CPUS=2048
NUMA PowerNV
  [   67.281435][  T122] Modules linked in:
  [   67.281903][  T122] CPU: 0 PID: 122 Comm: sh Not tainted
6.0.0-rc3-00007-gdcf8e5633e2e #10
  [   67.282547][  T122] NIP:  c000000000050bfc LR: c000000000050bec
CTR:0000000000005bdc
  [   67.282920][  T122] REGS: c0000000348475b0 TRAP: 0300   Not
tainted (6.0.0-rc3-00007-gdcf8e5633e2e)
  [   67.283424][  T122] MSR:  9000000000009033
<SF,HV,EE,ME,IR,DR,RI,LE> CR: 88002444  XER: 20040006
  [   67.284023][  T122] CFAR: c00000000022d100 DAR: 0000000000000000
DSISR: 40000000 IRQMASK: 0
  [   67.284023][  T122] GPR00: c000000000050bec c000000034847850
c0000000013f6100 c000000001fb7718
  [   67.284023][  T122] GPR04: c000000000515c10 c000000000e5fe08
c00000000133da60 c000000004839300
  [   67.284023][  T122] GPR08: c0000000014ffb98 0000000000000000
c000000000515c0c c000000000e18576
  [   67.284023][  T122] GPR12: c000000000e60170 c0000000015a0000
00000001155e0460 0000000000000000
  [   67.284023][  T122] GPR16: 0000000000000000 00007fffe8eeb3c8
0000000116320728 0000000000000000
  [   67.284023][  T122] GPR20: 0000000116320720 0000000000000000
c0000000012fa918 0000000000000006
  [   67.284023][  T122] GPR24: c0000000014ffb98 c0000000011ed360
0000000000000000 c000000001fb7928
  [   67.284023][  T122] GPR28: 0000000000000000 0000000000000000
000000007c0802a6 c000000001fb7918
  [   67.287799][  T122] NIP [c000000000050bfc]
arch_prepare_kprobe+0x10c/0x2d0
  [   67.288490][  T122] LR [c000000000050bec]
arch_prepare_kprobe+0xfc/0x2d0
  [   67.289025][  T122] Call Trace:
  [   67.289268][  T122] [c000000034847850] [c0000000012f77a0]
0xc0000000012f77a0 (unreliable)
  [   67.289999][  T122] [c0000000348478d0] [c000000000231320]
register_kprobe+0x3c0/0x7a0
  [   67.290439][  T122] [c000000034847940] [c0000000002938c0]
__register_trace_kprobe+0x140/0x1a0
  [   67.290898][  T122] [c0000000348479b0] [c0000000002944c4]
__trace_kprobe_create+0x794/0x1040
  [   67.291330][  T122] [c000000034847b60] [c0000000002a1614]
trace_probe_create+0xc4/0xe0
  [   67.291717][  T122] [c000000034847bb0] [c00000000029363c]
create_or_delete_trace_kprobe+0x2c/0x80
  [   67.292158][  T122] [c000000034847bd0] [c000000000264420]
trace_parse_run_command+0xf0/0x210
  [   67.292611][  T122] [c000000034847c70] [c0000000002934a0]
probes_write+0x20/0x40
  [   67.292996][  T122] [c000000034847c90] [c00000000045e98c]
vfs_write+0xfc/0x450
  [   67.293356][  T122] [c000000034847d50] [c00000000045eec4]
ksys_write+0x84/0x140
  [   67.293716][  T122] [c000000034847da0] [c00000000002e4fc]
system_call_exception+0x17c/0x3a0
  [   67.294186][  T122] [c000000034847e10] [c00000000000c0e8]
system_call_vectored_common+0xe8/0x278
  [   67.294680][  T122] --- interrupt: 3000 at 0x7fffa5682de0
  [   67.294937][  T122] NIP:  00007fffa5682de0 LR: 0000000000000000
CTR:0000000000000000
  [   67.295313][  T122] REGS: c000000034847e80 TRAP: 3000   Not
tainted (6.0.0-rc3-00007-gdcf8e5633e2e)
  [   67.295725][  T122] MSR:  900000000280f033
<SF,HV,VEC,VSX,EE,PR,FP,ME,IR,DR,RI,LE>  CR: 44002408  XER: 00000000
  [   67.296291][  T122] IRQMASK: 0
  [   67.296291][  T122] GPR00: 0000000000000004 00007fffe8eeaec0
00007fffa5757300 0000000000000001
  [   67.296291][  T122] GPR04: 0000000116329c60 0000000000000017
0000000000116329 0000000000000000
  [   67.296291][  T122] GPR08: 0000000000000006 0000000000000000
0000000000000000 0000000000000000
  [   67.296291][  T122] GPR12: 0000000000000000 00007fffa580ac60
00000001155e0460 0000000000000000
  [   67.296291][  T122] GPR16: 0000000000000000 00007fffe8eeb3c8
0000000116320728 0000000000000000
  [   67.296291][  T122] GPR20: 0000000116320720 0000000000000000
0000000000000000 0000000000000002
  [   67.296291][  T122] GPR24: 00000001163206f0 0000000000000020
00007fffe8eeafa0 0000000000000001
  [   67.296291][  T122] GPR28: 0000000000000000 0000000000000017
0000000116329c60 0000000000000001
  [   67.299570][  T122] NIP [00007fffa5682de0] 0x7fffa5682de0
  [   67.299837][  T122] LR [0000000000000000] 0x0
  [   67.300072][  T122] --- interrupt: 3000
  [   67.300447][  T122] Instruction dump:
  [   67.300736][  T122] 386319d8 481342f5 60000000 60000000 60000000
e87f0028 3863fffc 481dc4d1
  [   67.301230][  T122] 60000000 2c230000 41820018 e9230058
<81290000> 552936be 2c090001 4182018c
  [   67.302102][  T122] ---[ end trace 0000000000000000 ]---
  [   67.302496][  T122]

Please consider trimming the backtrace to only capture the necessary
information:
https://docs.kernel.org/process/submitting-patches.html#backtraces-in-commit-mesages

Thank you for the reminder. Michael has modified it when merging in,
thanks Michael.

quoted

The address being probed has some special:

  cmdline_proc_show: Probe based on ftrace
  cmdline_proc_show+16: Probe for the next instruction at the ftrace
location

The ftrace-based kprobe does not generate kprobe::ainsn::insn, it gets
set to NULL. In arch_prepare_kprobe() it will check for:

  ...
  prev = get_kprobe(p->addr - 1);
  preempt_enable_no_resched();
  if (prev && ppc_inst_prefixed(ppc_inst_read(prev->ainsn.insn))) {
  ...

If prev is based on ftrace, 'ppc_inst_read(prev->ainsn.insn)' will occur
with a null pointer reference. At this point prev->addr will not be a
prefixed instruction, so the check can be skipped.

Check if prev is ftrace-based kprobe before reading 'prev->ainsn.insn'
to fix this problem.

Fixes: b4657f7650ba ("powerpc/kprobes: Don't allow breakpoints on
suffixes")
Signed-off-by: Li Huafei <redacted>
---
 arch/powerpc/kernel/kprobes.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/kernel/kprobes.c

b/arch/powerpc/kernel/kprobes.c
index 912d4f8a13be..9f6cbbd56809 100644

--- a/arch/powerpc/kernel/kprobes.c
+++ b/arch/powerpc/kernel/kprobes.c

@@ -161,7 +161,12 @@ int arch_prepare_kprobe(struct kprobe *p)

     preempt_disable();
     prev = get_kprobe(p->addr - 1);
     preempt_enable_no_resched();
-    if (prev && ppc_inst_prefixed(ppc_inst_read(prev->ainsn.insn))) {
+    /*
+     * When prev is a ftrace-based kprobe, we don't have an insn, and it
+     * doesn't probe for prefixed instruction.
+     */
+    if (prev && !kprobe_ftrace(prev) &&
+        ppc_inst_prefixed(ppc_inst_read(prev->ainsn.insn))) {
         printk("Cannot register a kprobe on the second word of
prefixed instruction\n");
         ret = -EINVAL;
     }

It's fine to keep the if condition on a single line.

Other than that, thanks for the fix!
Reviewed-by: Naveen N. Rao <redacted>


- Naveen

.