[v2,1/2] ras: fix an off-by-one error in __find_elem()

From: Cong Wang <xiyou.wangcong@gmail.com>
To: Borislav Petkov <bp@alien8.de>
Cc: LKML <linux-kernel@vger.kernel.org>,
	linux-edac@vger.kernel.org, Tony Luck <tony.luck@intel.com>,
	Thomas Gleixner <tglx@linutronix.de>
Subject: [v2,1/2] ras: fix an off-by-one error in __find_elem()
Date: Tue, 16 Apr 2019 16:16:08 -0700	[thread overview]
Message-ID: <CAM_iQpVd02zkVJ846cj-Fg1yUNuz6tY5q1Vpj4LrXmE06dPYYg@mail.gmail.com> (raw)

On Tue, Apr 16, 2019 at 2:46 PM Borislav Petkov <bp@alien8.de> wrote:
>
> On Tue, Apr 16, 2019 at 02:33:50PM -0700, Cong Wang wrote:
> > ce_arr.array[] is always within the range [0, ce_arr.n-1].
> > However, the binary search code in __find_elem() uses ce_arr.n
> > as the maximum index, which could lead to an off-by-one
> > out-of-bound access right after the while loop. In this case,
> > we should not even read it, just return -ENOKEY instead.
> >
> > Note, this could cause a kernel crash if ce_arr.n is exactly
> > MAX_ELEMS.
>
> "Could cause"?
>
> I'm still waiting for a demonstration. You can build a case through
> writing values in the debugfs nodes I pointed you at or even with a
> patch ontop preparing the exact conditions for it to crash. And then
> give me that "recipe" to trigger it here in a VM.

It is actually fairly easy:

1) Fill the whole page with PFN's:
for i in `seq 0 511`; do echo $i >> /sys/kernel/debug/ras/cec/pfn; done

2) Set thresh to 1 in order to trigger the deletion:
echo 1 > /sys/kernel/debug/ras/cec/count_threshold

3) Repeatedly add and remove the last element:
echo 512 >> /sys/kernel/debug/ras/cec/pfn
(until you get a crash.)

In case you still don't get it, here it is:

[   57.732593] BUG: unable to handle kernel paging request at ffff9c667bca0000
[   57.734994] #PF error: [PROT] [WRITE]
[   57.735891] PGD 75601067 P4D 75601067 PUD 75605067 PMD 7bca1063 PTE
800000007bca0061
[   57.737702] Oops: 0003 [#1] SMP PTI
[   57.738533] CPU: 0 PID: 649 Comm: bash Not tainted 5.1.0-rc5+ #561
[   57.739965] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
BIOS ?-20180724_192412-buildhw-07.phx2.fedoraproject.org-1.fc29
04/01/2014
[   57.742892] RIP: 0010:__memmove+0x57/0x1a0
[   57.743853] Code: 00 72 05 40 38 fe 74 3b 48 83 ea 20 48 83 ea 20
4c 8b 1e 4c 8b 56 08 4c 8b 4e 10 4c 8b 46 18 48 8d 76 20 4c 89 1f 4c
89 57 08 <4c> 89 4f 10 4c 89 47 18 48 8d 7f 20 73 d4 48 83 c2 20 e9 a2
00 00
[   57.748150] RSP: 0018:ffffbe2ec0c8bdf8 EFLAGS: 00010206
[   57.749371] RAX: ffff9c667a5c1ff0 RBX: 0000000000000001 RCX: 0000000000000ff8
[   57.751018] RDX: 00000007fe921fb8 RSI: ffff9c667bca0018 RDI: ffff9c667bc9fff0
[   57.752674] RBP: 0000000000000200 R08: 0000000000000000 R09: 0000015c00000000
[   57.754325] R10: 0000000000040001 R11: 5a5a5a5a5a5a5a5a R12: 0000000000000004
[   57.755976] R13: ffff9c6671787778 R14: ffff9c6671787728 R15: ffff9c6671787750
[   57.757631] FS:  00007f33ca294740(0000) GS:ffff9c667d800000(0000)
knlGS:0000000000000000
[   57.759689] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   57.761023] CR2: ffff9c667bca0000 CR3: 000000007061e000 CR4: 00000000000406f0
[   57.762681] Call Trace:
[   57.763275]  del_elem.constprop.1+0x39/0x40
[   57.764260]  cec_add_elem+0x1e4/0x211
[   57.765129]  simple_attr_write+0xa2/0xc3
[   57.766057]  debugfs_attr_write+0x45/0x5c
[   57.767005]  full_proxy_write+0x4b/0x65
[   57.767911]  ? full_proxy_poll+0x50/0x50
[   57.768844]  vfs_write+0xb8/0xf5
[   57.769613]  ksys_write+0x6b/0xb8
[   57.770407]  do_syscall_64+0x57/0x65
[   57.771249]  entry_SYSCALL_64_after_hwframe+0x49/0xbe

I will leave it as a homework for explaining why the crash is inside
memmove(). ;)

Thanks.