On Mon, May 30, 2016 at 06:22:35PM +0300, Dmitry Fleytman wrote: > > > On 30 May 2016, at 18:19 PM, Michael S. Tsirkin <m...@redhat.com> wrote: > > > > On Mon, May 30, 2016 at 06:14:56PM +0300, Dmitry Fleytman wrote: > >> Does DSN generation function pass unaligned offsets? > >> It does not look like it does… > >> > >> > >> It does according to clang sanitiser. > > > > > > Oh so it's a clang false positive? > > I think not. > The capability itself is 8-bytes aligned but 64-bit serial number inside of > it is not because of 32 bit header in front of it.
Oh right. Things like this should really go into commit log in the future. For now a code comment in pci set/get that explains that alignment in capabilities is generally at dword not qword boundary would be enough. > > > > -- > > MST