Re: Re: [PATCH] tty: vt: Add checks after calling kzalloc

Re: Re: [PATCH] tty: vt: Add checks after calling kzalloc

[Jiasheng Jiang]

On Wed, Aug 31, 2022 at 03:57:42PM +0800, Jiasheng Jiang wrote:
>>  	for (currcons = 0; currcons < MIN_NR_CONSOLES; currcons++) {
>>  		vc_cons[currcons].d = vc = kzalloc(sizeof(struct vc_data), GFP_NOWAIT);
>> +		if (!vc) {
>> +			console_unlock();
>> +			return -ENOMEM;
>> +		}
>>  		INIT_WORK(&vc_cons[currcons].SAK_work, vc_SAK);
>>  		tty_port_init(&vc->port);
>>  		visual_init(vc, currcons, 1);
>>  		/* Assuming vc->vc_{cols,rows,screenbuf_size} are sane here. */
>>  		vc->vc_screenbuf = kzalloc(vc->vc_screenbuf_size, GFP_NOWAIT);
>> +		if (!vc->vc_screenbuf) {
>> +			console_unlock();
>> +			return -ENOMEM;
>> +		}
> 
> This has been attempted many times in the past, sorry.  Unless you can
> prove that this can actually happen in real life, we are going to leave
> these as-is.
> 
> Please do not just do random changes like this without actually testing
> to see if it is possible to happen.

As the harm of vulnerabilities is much higher than the cost of fixing them,
it is acceptable to add harmless security checks that guarantee the
vulnerabilities will never be triggered.

Thanks,
Jiang

Re: Re: [PATCH] tty: vt: Add checks after calling kzalloc

[Greg KH]

On Wed, Aug 31, 2022 at 05:08:10PM +0800, Jiasheng Jiang wrote:
> On Wed, Aug 31, 2022 at 03:57:42PM +0800, Jiasheng Jiang wrote:
> >>  	for (currcons = 0; currcons < MIN_NR_CONSOLES; currcons++) {
> >>  		vc_cons[currcons].d = vc = kzalloc(sizeof(struct vc_data), GFP_NOWAIT);
> >> +		if (!vc) {
> >> +			console_unlock();
> >> +			return -ENOMEM;
> >> +		}
> >>  		INIT_WORK(&vc_cons[currcons].SAK_work, vc_SAK);
> >>  		tty_port_init(&vc->port);
> >>  		visual_init(vc, currcons, 1);
> >>  		/* Assuming vc->vc_{cols,rows,screenbuf_size} are sane here. */
> >>  		vc->vc_screenbuf = kzalloc(vc->vc_screenbuf_size, GFP_NOWAIT);
> >> +		if (!vc->vc_screenbuf) {
> >> +			console_unlock();
> >> +			return -ENOMEM;
> >> +		}
> > 
> > This has been attempted many times in the past, sorry.  Unless you can
> > prove that this can actually happen in real life, we are going to leave
> > these as-is.
> > 
> > Please do not just do random changes like this without actually testing
> > to see if it is possible to happen.
> 
> As the harm of vulnerabilities is much higher than the cost of fixing them,
> it is acceptable to add harmless security checks that guarantee the
> vulnerabilities will never be triggered.

No, not always, many times you are adding new bugs by doing this type of
"unneeded fixes".  We have had this happen in the vt code many times in
the past, let's learn from our mistakes please.

And where is the "vulnerability" here exactly?

thanks,

greg k-h

Re: [PATCH] tty: vt: Add checks after calling kzalloc

[Tetsuo Handa]

Please check if your system can survive, by doing

On 2022/08/31 18:08, Jiasheng Jiang wrote:
> On Wed, Aug 31, 2022 at 03:57:42PM +0800, Jiasheng Jiang wrote:
>>>  	for (currcons = 0; currcons < MIN_NR_CONSOLES; currcons++) {
>>>  		vc_cons[currcons].d = vc = kzalloc(sizeof(struct vc_data), GFP_NOWAIT);

-		vc_cons[currcons].d = vc = kzalloc(sizeof(struct vc_data), GFP_NOWAIT);
+		vc_cons[currcons].d = vc = NULL;

>>> +		if (!vc) {
>>> +			console_unlock();
>>> +			return -ENOMEM;
>>> +		}
>>>  		INIT_WORK(&vc_cons[currcons].SAK_work, vc_SAK);
>>>  		tty_port_init(&vc->port);
>>>  		visual_init(vc, currcons, 1);
>>>  		/* Assuming vc->vc_{cols,rows,screenbuf_size} are sane here. */
>>>  		vc->vc_screenbuf = kzalloc(vc->vc_screenbuf_size, GFP_NOWAIT);

and/or

-		vc->vc_screenbuf = kzalloc(vc->vc_screenbuf_size, GFP_NOWAIT);
+		vc->vc_screenbuf = NULL;

>>> +		if (!vc->vc_screenbuf) {
>>> +			console_unlock();
>>> +			return -ENOMEM;
>>> +		}

in addition to your patch.

This is __init function of built-in module, isn't it?
Who can recover from this allocation failure? ;-)