Re: [PATCH RFC] random: fix syzkaller fuzzer test int overflow

[Chen Feng]

On 2017/10/24 17:09, Greg KH wrote:
> On Tue, Oct 24, 2017 at 03:44:17PM +0800, Chen Feng wrote:
>> [pid:11940,cpu6,syz-executor][flp_ioctl]cmd[0x1]
>> Restart is not permit
>> =================================================================
>> UBSAN: Undefined behaviour in
>> kernel/linux-4.4/drivers/char/random.c:676:19
>> signed integer overflow:
>> 2147483645 + 268435455 cannot be represented in type 'int'
>> CPU: 4 PID: 11941 Comm: syz-executor Not tainted 4.4.76+ #2
> 
> Does this also happen on 4.14-rc6?

No, mainline also has this issue.
> 
>> TGID: 11928 Comm: syz-executor
>> Hardware name: hi3660 (DT)
>> Call trace:
>> [<ffffffa661090378>] dump_backtrace+0x0/0x314
>> [<ffffffa66109077c>] show_stack+0x1c/0x24
>> [<ffffffa66180cc54>] dump_stack+0xdc/0x130
>> [<ffffffa6618833d0>] ubsan_epilogue+0x18/0x6c
>> [<ffffffa661883c5c>] handle_overflow+0x180/0x1d4
>> [<ffffffa661883cdc>] __ubsan_handle_add_overflow+0x2c/0x34
>> [<ffffffa661ab75a0>] credit_entropy_bits+0x358/0x9a8
>> [<ffffffa661ab85bc>] random_ioctl+0x338/0x384
>> [<ffffffa661399c74>] do_vfs_ioctl+0x60c/0xa4c
>> [<ffffffa66139a150>] SyS_ioctl+0x9c/0xc0
>> [<ffffffa6610838b0>] el0_svc_naked+0x24/0x28
>> =================================================================
>>
>> Signed-off-by: Chen Feng <puck.chen@xxxxxxxxxxxxx>
>> Signed-off-by: Yukun Zhao <zhaoyukun@xxxxxxxxxx>
>> ---
>>  drivers/char/random.c | 5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/char/random.c b/drivers/char/random.c
>> index 1ef2640..6f2bd6a 100644
>> --- a/drivers/char/random.c
>> +++ b/drivers/char/random.c
>> @@ -699,6 +699,11 @@ static void credit_entropy_bits(struct entropy_store *r, int nbits)
>>  	if (cmpxchg(&r->entropy_count, orig, entropy_count) != orig)
>>  		goto retry;
>>  
>> +	if (INT_MAX - nbits < r->entropy_total) {
>> +		WARN_ON(1);
> 
> Why WARN_ON()?  What is that going to help with?
Actually, I am not familiar with the random module....

This patch is RFC to see if some one has better idea.

-feng

> 
> thanks,
> 
> greg k-h
> 
> .
>