[PATCH] powerpc/kprobes: Fix null pointer reference in arch_prepare_kprobe()

From: Li Huafei
Date: Fri Sep 23 2022 - 05:39:01 EST

I found a null pointer reference in arch_prepare_kprobe():

# echo 'p cmdline_proc_show' > kprobe_events
# echo 'p cmdline_proc_show+16' >> kprobe_events
[ 67.278533][ T122] Kernel attempted to read user page (0) - exploit attempt? (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]

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 <lihuafei1@xxxxxxxxxx>
---
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;
}
--
2.17.1