On 7/13/16, 9:28 PM, "Martin K. Petersen" <martin.peter...@oracle.com> wrote:
>>>>>> "Johannes" == Johannes Thumshirn <jthumsh...@suse.de> writes:
>
>Johannes> qla2xxx first calls request_irq() and then does the setup of
>Johannes> the queue entry data needed in the interrupt handlers in when
>Johannes> using MSI-X. This could lead to a NULL pointer dereference
>Johannes> when an IRQ fires between the request_irq() call and the
>Johannes> assignment of the qentry data structure to the
>rsp-> msix field. A possible case for such a race would be in the kdump
>Johannes> case when the HBA's IRQs are still enabled but the driver is
>Johannes> undergoing a new initialisation and thus is not aware of
>Johannes> already activated IRQs in the HBA.
>
>Qlogic folks: Please review!
Looks Good.
Reviewed-by: Himanshu Madhani <himanshu.madh...@qlogic.com>
>
>--
>Martin K. Petersen Oracle Linux Engineering
N�����r��y����b�X��ǧv�^�){.n�+����{���"�{ay� ʇڙ�,j��f���h���z� �w���
���j:+v���w�j�m��������zZ+�����ݢj"��!�i
