Re: [PATCH] net: nfc: nci: fix for UBSAN: shift-out-of-bounds in nci_activate_target

[Krzysztof Kozlowski]

On 19/04/2023 10:26, Krzysztof Kozlowski wrote:
> On 19/04/2023 03:16, Anup Sharma wrote:
>> syzbot found  UBSAN: shift-out-of-bounds in nci_activate_target [1],
>> when nci_target->supported_protocols is bigger than UNIT_MAX,
> 
> UINT_MAX?
> 
>> where supported_protocols is unsigned 32-bit interger type.
> 
> integer?
> 
>>
>> 32 is the maximum allowed for supported_protocols. Added a check
>> for it. 
>>
>> [1] UBSAN: shift-out-of-bounds in net/nfc/nci/core.c:912:45
>> shift exponent 4294967071 is too large for 32-bit type 'int'
>> Call Trace:
>>  <TASK>
>>  __dump_stack lib/dump_stack.c:88 [inline]
>>  dump_stack_lvl+0x136/0x150 lib/dump_stack.c:106
>>  ubsan_epilogue lib/ubsan.c:217 [inline]
>>  __ubsan_handle_shift_out_of_bounds+0x221/0x5a0 lib/ubsan.c:387
>>  nci_activate_target.cold+0x1a/0x1f net/nfc/nci/core.c:912
>>  nfc_activate_target+0x1f8/0x4c0 net/nfc/core.c:420
>>  nfc_genl_activate_target+0x1f3/0x290 net/nfc/netlink.c:900
>>  genl_family_rcv_msg_doit.isra.0+0x1e6/0x2d0 net/netlink/genetlink.c:968
>>  genl_family_rcv_msg net/netlink/genetlink.c:1048 [inline]
>>
>> Reported-by: syzbot+0839b78e119aae1fec78@xxxxxxxxxxxxxxxxxxxxxxxxx
>> Link: https://syzkaller.appspot.com/bug?id=19cf2724120ef8c51c8d2566df0cc34617188433
>>
>> Signed-off-by: anupsharma <anupnewsmail@xxxxxxxxx>
>> ---
>>  net/nfc/nci/core.c | 5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/net/nfc/nci/core.c b/net/nfc/nci/core.c
>> index fff755dde30d..e9d968bd1cd9 100644
>> --- a/net/nfc/nci/core.c
>> +++ b/net/nfc/nci/core.c
>> @@ -908,6 +908,11 @@ static int nci_activate_target(struct nfc_dev *nfc_dev,
>>  		pr_err("unable to find the selected target\n");
>>  		return -EINVAL;
>>  	}
>> +	
>> +	if (nci_target->supported_protocols >= 32) {
> 
> I don't think it makes any sense. How do you protect from UBSAN reported
> shift? Why supported_protocols cannot be 33? You are not shifting the
> supported_protocols...
> 
>> +		pr_err("Too many supported protocol by the device\n");
>> +		return -EINVAL;
> 
> I am pretty sure that you broke now NFC. Test the patches first and
> share your test scenario.

BTW, ISO15693 is here protocol 128, so definitely more than 32.

Best regards,
Krzysztof