Re: New Defects reported by Coverity Scan for XenProject

[David Vrabel <david.vrabel@citrix.com>]

On 13/11/13 13:51, Ian Campbell wrote:
> Hi David,
> 
> Coverity picked up some issues in the kexec patches. At least the
> locking one looks valid to me...

Thanks.  Andy just pointed them out to me as well.

> I didn't investigate the endianness one.

We believe these are coverity being confused by the underlying
atomic_read()/atomic_write() macros and getting the type wrong.

David

> On Wed, 2013-11-13 at 05:34 -0800, scan-admin@coverity.com wrote:
>> ________________________________________________________________________
>> CID 1128573: Missing unlock (LOCK)
>>
>> /xen/common/kexec.c: 788 ( lock)
>>    785    
>>    786        *old = NULL;
>>    787    
>>>>> "_spin_lock(spinlock_t *)" locks "kexec_lock".
>>    788        spin_lock(&kexec_lock);
>>    789    
>>    790        if ( test_bit(KEXEC_FLAG_IN_PROGRESS, &kexec_flags) )
>>    791        {
>>    792            spin_unlock(&kexec_lock);
>>   
>>
>> /xen/common/kexec.c: 797 ( missing_unlock)
>>    794        }
>>    795    
>>    796        if ( kexec_load_get_bits(type, &base, &bit) )
>>>>> CID 1128573: Missing unlock (LOCK)
>>>>> Returning without unlocking "kexec_lock".
>>    797            return -EINVAL;
>>    798    
>>    799        pos = (test_bit(bit, &kexec_flags) != 0);
>>    800        old_slot = base + pos;
>>    801        new_slot = base + !pos;
>>   
>> ________________________________________________________________________
>> CID 1128572: Reliance on integer endianness (INCOMPATIBLE_CAST)
>>
>> /xen/arch/x86/machine_kexec.c: 58 ( incompatible_cast)
>>    55            l3_page = kimage_alloc_control_page(image, 0);
>>    56            if ( !l3_page )
>>    57                goto out;
>>>>> CID 1128572: Reliance on integer endianness (INCOMPATIBLE_CAST)
>>>>> Pointer "&l4->l4" points to an object whose effective type is "unsigned long" (64 bits, unsigned) but is dereferenced as a narrower "unsigned int" (32 bits, unsigned).  This may lead to unexpected results depending on machine endianness.
>>    58            l4e_write(l4, l4e_from_page(l3_page, __PAGE_HYPERVISOR));
>>    59        }
>>    60        else
>>    61            l3_page = l4e_get_page(*l4);
>>    62    
>>   
>> ________________________________________________________________________
>> CID 1128571: Reliance on integer endianness (INCOMPATIBLE_CAST)
>>
>> /xen/arch/x86/machine_kexec.c: 70 ( incompatible_cast)
>>    67            l2_page = kimage_alloc_control_page(image, 0);
>>    68            if ( !l2_page )
>>    69                goto out;
>>>>> CID 1128571: Reliance on integer endianness (INCOMPATIBLE_CAST)
>>>>> Pointer "&l3->l3" points to an object whose effective type is "unsigned long" (64 bits, unsigned) but is dereferenced as a narrower "unsigned int" (32 bits, unsigned).  This may lead to unexpected results depending on machine endianness.
>>    70            l3e_write(l3, l3e_from_page(l2_page, __PAGE_HYPERVISOR));
>>    71        }
>>    72        else
>>    73            l2_page = l3e_get_page(*l3);
>>    74    
>>   
>> ________________________________________________________________________
>> To view the defects in Coverity Scan visit, http://scan.coverity.com