Re: New Defects reported by Coverity Scan for XenProject

[Jan Beulich <jbeulich@xxxxxxxx>]

On 24.01.2021 11:35, scan-admin@xxxxxxxxxxxx wrote:
> *** CID 1472394:  Concurrent data access violations  (MISSING_LOCK)
> /xen/drivers/passthrough/x86/hvm.c: 1054 in pci_clean_dpci_irq()
> 1048         list_for_each_entry_safe ( digl, tmp, &pirq_dpci->digl_list, 
> list )
> 1049         {
> 1050             list_del(&digl->list);
> 1051             xfree(digl);
> 1052         }
> 1053         /* Note the pirq is now unbound. */
>>>>     CID 1472394:  Concurrent data access violations  (MISSING_LOCK)
>>>>     Accessing "pirq_dpci->flags" without holding lock "domain.event_lock". 
>>>> Elsewhere, "hvm_pirq_dpci.flags" is accessed with "domain.event_lock" held 
>>>> 10 out of 11 times.
> 1054         pirq_dpci->flags = 0;
> 1055     
> 1056         return pt_pirq_softirq_active(pirq_dpci) ? -ERESTART : 0;
> 1057     }

The only (indirect) caller of this function is ...

> 1059     int arch_pci_clean_pirqs(struct domain *d)

... this one, which very clearly acquires the lock in question.
Does anyone have any idea what misleads Coverity here in its
conclusion, and hence possibly what may silence this?

Jan