On 2021/04/01 15:18, Greg Kroah-Hartman wrote:
> On Thu, Apr 01, 2021 at 02:57:47PM +0900, Tetsuo Handa wrote:
>> KMSAN complains that the vmci_use_ppn64() == false path in
>> vmci_dbell_register_notification_bitmap() left upper 32bits of
>> bitmap_set_msg.bitmap_ppn64 member uninitialized.
>>
>> KMSAN also complains that vmci_check_host_caps() left the payload part
>> of check_msg uninitialized.
>>
>
> What commit does this "fix"? Can you resend with a proper "Fixes:" tag
> so we know where it needs to be backported to?
It seems that this problem exists since the introduction.
commit 83e2ec765be03e8a8a07619e65df70b48a1db023
Author: George Zhang <georgezh...@vmware.com>
Date: Tue Jan 8 15:53:51 2013 -0800
VMCI: doorbell implementation.
commit 1f166439917b69d3046e2e49fe923579d9181212
Author: George Zhang <georgezh...@vmware.com>
Date: Tue Jan 8 15:55:32 2013 -0800
VMCI: guest side driver implementation.
But this patch should be safe to backport to as old as possible, for
this patch is merely explicitly initializing variables.
