Re: [PATCH] xen/xenbus: handle potential dangling pointer issue in xen_pcibk_xenbus_probe

[Yunseong Kim <yskelg@xxxxxxxxx>]

On 5/29/24 9:27 오후, Jürgen Groß wrote:
> On 29.05.24 14:22, yskelg@xxxxxxxxx wrote:
>> From: Yunseong Kim <yskelg@xxxxxxxxx>
>>
>> If 'xen_pcibk_init_devices()' fails. This ensures that 'pdev->xdev' does
>> not point to 'xdev' when 'pdev' is freed.
>>
>> Signed-off-by: Yunseong Kim <yskelg@xxxxxxxxx>
>> ---
>>   drivers/xen/xen-pciback/xenbus.c | 1 +
>>   1 file changed, 1 insertion(+)
>>
>> diff --git a/drivers/xen/xen-pciback/xenbus.c
>> b/drivers/xen/xen-pciback/xenbus.c
>> index b11e401f1b1e..348d6803b8c0 100644
>> --- a/drivers/xen/xen-pciback/xenbus.c
>> +++ b/drivers/xen/xen-pciback/xenbus.c
>> @@ -54,6 +54,7 @@ static struct xen_pcibk_device *alloc_pdev(struct
>> xenbus_device *xdev)
>>       INIT_WORK(&pdev->op_work, xen_pcibk_do_op);
>>         if (xen_pcibk_init_devices(pdev)) {
>> +        pdev->xdev = NULL;
>>           kfree(pdev);
>>           pdev = NULL;
>>       }
> 
> NAK.
> 
> This doesn't make any sense, as pdev is freed.
> 
> 
> Juergen

Thanks for the code review Juergen!

I don't know if it's a big deal, but I think that if pdev->xdev don't
replace it with NULL, 'struct xenbus_device *xdev' address value stays
in the kmalloc cached memory.

And most of the kernel code I've seen does this (even when it's
'free_pdev()'), so I sent a patch.

In the very unlikely event that. I was thinking if someone look at the
value without initializing the kmalloced memory, someone might see the
'xenbus_device *xdev address'.

Warm Regards,

Yunseong Kim