Re: New Defects reported by Coverity Scan for XenProject

[Jan Beulich <jbeulich@xxxxxxxx>]

On 24.06.2026 14:33, scan-admin@xxxxxxxxxxxx wrote:
> ** CID 1695360:       Integer handling issues  (BAD_SHIFT)
> /xen/common/numa.c: 483           in extract_lsb_from_nodes()
> 
> 
> _____________________________________________________________________________________________
> *** CID 1695360:         Integer handling issues  (BAD_SHIFT)
> /xen/common/numa.c: 483             in extract_lsb_from_nodes()
> 477     
> 478         if ( nodes_used <= 1 )
> 479             i = min(PADDR_BITS, BITS_PER_LONG - 1);
> 480         else
> 481             i = find_first_bit(&bitfield, sizeof(unsigned long) * 8);
> 482     
>>>>     CID 1695360:         Integer handling issues  (BAD_SHIFT)
>>>>     In expression "memtop - 1UL >> i", right shifting by more than 63 bits 
>>>> has undefined behavior.  The shift amount, "i", is 64.
> 483         memnodemapsize = ((memtop - 1) >> i) + 1;
> 484     
> 485         return i;
> 486     }

Without contextual knowledge I think the complaint is reasonable. However,
for nodes_used to be greater than 1, we will need to have ORed spdx into
bitfield on an iteration later than the 1st one. Yet only on the 1st
iteration can spdx be 0 (or, if nodes[] wasn't sorted, only on one of the
at least two involved iterations). Hence bitfield isn't going to be 0, and
thus find_first_bit() is going to return a value lower than BITS_PER_LONG.

Jan