Linux: arbitrary kernel read into dmesg via missing address check in segfault handler
There is a missing address check in both show_opcodes() callers.
show_opcodes() is mostly used by the kernel to print the raw
instruction bytes surrounding an instruction that generated an
unexpected exception; however, sometimes it is also used to
print userspace instructions.
Because the userspace address isn't checked against TASK_SIZE_MAX, if userspace
faults on a kernel address, the kernel can dump data from a user-controlled
address into dmesg.
show_opcodes() has two callers:
- since commit ba54d856a9d8 (first in 4.18): show_signal_msg() shows userspace
instructions when userspace e.g. segfaults
- show_ip() is used when the kernel detects some sort of bug; this means that
to trigger it, you need some way to at least trigger a WARN() or so
Repro for the first variant:
=========================
user@debian:~/segfault$ sudo grep core_pattern /proc/kallsyms
ffffffff9ae34180 D core_pattern
ffffffff9be99500 t _GLOBAL__sub_I_65535_1_core_pattern
ffffffff9bff2860 t _GLOBAL__sub_D_65535_0_core_pattern
user@debian:~/segfault$ cat segfault.c
int main(void) {
void (*fn)(void) = (void*)0xffffffff9ae34180;
fn();
}
user@debian:~/segfault$ gcc -o segfault segfault.c
user@debian:~/segfault$ ./segfault
Segmentation fault
user@debian:~/segfault$ sudo dmesg | tail -n2
[19511.957855] segfault[2622]: segfault at ffffffff9ae34180 ip ffffffff9ae34180 sp 00007ffe0adf1568 error 15
[19511.962055] Code: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 <63> 6f 72 65 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
=========================
Note the "<63> 6f 72 65 00" - that's "core\0".
Repro for the second variant:
Patch the kernel like this to get an easy way to trigger a WARN() in the right
context:
=========================
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index b9123c497e0a..fab40edd4c9e 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -891,6 +891,7 @@ __bad_area_nosemaphore(struct pt_regs *regs, unsigned long error_code,
tsk->thread.trap_nr = X86_TRAP_PF;
force_sig_info_fault(SIGSEGV, si_code, address, tsk, pkey, 0);
+ WARN(1, "TESTING WARN()");
return;
}
=========================
Then run the same repro code as before (with the core_pattern address fixed up).
Result:
=========================
[ 125.564041] segfault[1602]: segfault at ffffffff854340c0 ip ffffffff854340c0 sp 00007ffd4cc7a568 error 15
[ 125.569923] Code: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 <63> 6f 72 65 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 125.576859] ------------[ cut here ]------------
[ 125.578406] TESTING WARN()
[ 125.578439] WARNING: CPU: 6 PID: 1602 at arch/x86/mm/fault.c:894 __bad_area_nosemaphore+0x147/0x270
[ 125.582172] Modules linked in: bpfilter
[ 125.583394] CPU: 6 PID: 1602 Comm: segfault Tainted: G W 4.18.0+ #108
[ 125.585811] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
[ 125.588410] RIP: 0010:__bad_area_nosemaphore+0x147/0x270
[ 125.590078] Code: 48 89 d9 48 89 ea 44 89 e6 48 c7 83 30 0b 00 00 0e 00 00 00 bf 0b 00 00 00 e8 f5 eb ff ff 48 c7 c7 00 61 66 84 e8 79 11 05 00 <0f> 0b 48 83 c4 28 5b 5d 41 5c 41 5d 41 5e 41 5f c3 48 83 c4 28 4c
[ 125.595779] RSP: 0018:ffff8801cb3b7e18 EFLAGS: 00010286
[ 125.597426] RAX: 0000000000000000 RBX: ffff8801cbb9e000 RCX: 0000000000000000
[ 125.599605] RDX: 0000000000000001 RSI: dffffc0000000000 RDI: ffffffff86678ea0
[ 125.601800] RBP: ffffffff854340c0 R08: ffffed003d873ed5 R09: ffffed003d873ed5
[ 125.603935] R10: 0000000000000001 R11: ffffed003d873ed4 R12: 0000000000000001
[ 125.606113] R13: 0000000000000000 R14: 0000000000000015 R15: ffff8801cb3b7f58
[ 125.608250] FS: 00007fe30d518700(0000) GS:ffff8801ec380000(0000) knlGS:0000000000000000
[ 125.610608] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 125.612331] CR2: ffffffff854340c0 CR3: 00000001d563e001 CR4: 00000000003606e0
[ 125.614470] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 125.616607] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 125.618736] Call Trace:
[ 125.619475] __do_page_fault+0x133/0x780
[ 125.620646] ? mm_fault_error+0x1b0/0x1b0
[ 125.622236] ? async_page_fault+0x8/0x30
[ 125.623388] async_page_fault+0x1e/0x30
[ 125.624526] RIP: 0033:core_pattern+0x0/0x880
[ 125.625786] Code: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 <63> 6f 72 65 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 125.631208] RSP: 002b:00007ffd4cc7a568 EFLAGS: 00010202
[ 125.632737] RAX: ffffffff854340c0 RBX: 0000000000000000 RCX: 0000000000000000
[ 125.635039] RDX: 00007ffd4cc7a678 RSI: 00007ffd4cc7a668 RDI: 0000000000000001
[ 125.637088] RBP: 00007ffd4cc7a580 R08: 0000562d395106f0 R09: 00007fe30d323cb0
[ 125.639153] R10: 0000000000000000 R11: 00007fe30d0d23c0 R12: 0000562d39510530
[ 125.641183] R13: 00007ffd4cc7a660 R14: 0000000000000000 R15: 0000000000000000
[ 125.643221] ---[ end trace fb20716f9d6369bd ]---
=========================
This bug is subject to a 90 day disclosure deadline. After 90 days elapse
or a patch has been made broadly available (whichever is earlier), the bug
report will become visible to the public.
Found by: jannh
