Re: [GIT PULL] Please pull RDMA subsystem changes

[Leon Romanovsky]

On Mon, Apr 29, 2019 at 11:00:40AM +0200, Michal Kubecek wrote:
> On Mon, Apr 29, 2019 at 08:40:37AM +0000, Jason Gunthorpe wrote:
> > On Mon, Apr 29, 2019 at 08:09:47AM +0200, Heiko Carstens wrote:
> > > On Sun, Apr 28, 2019 at 11:52:12AM +0000, Jason Gunthorpe wrote:
> > > > Hi Linus,
> > > >
> > > > Third rc pull request
> > > >
> > > > Nothing particularly special here. There is a small merge conflict
> > > > with Adrea's mm_still_valid patches which is resolved as below:
> > > ...
> > > > Jason Gunthorpe (3):
> > > >       RDMA/mlx5: Do not allow the user to write to the clock page
> > > >       RDMA/mlx5: Use rdma_user_map_io for mapping BAR pages
> > > >       RDMA/ucontext: Fix regression with disassociate
> > >
> > > This doesn't compile. The patch below would fix it, but not sure if
> > > this is what is intended:
> > >
> > > drivers/infiniband/core/uverbs_main.c: In function 'rdma_umap_fault':
> > > drivers/infiniband/core/uverbs_main.c:898:28: error: 'struct vm_fault' has no member named 'vm_start'
> > >    vmf->page = ZERO_PAGE(vmf->vm_start);
> > >                             ^~
> > > diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c
> > > index 7843e89235c3..65fe89b3fa2d 100644
> > > +++ b/drivers/infiniband/core/uverbs_main.c
> > > @@ -895,7 +895,7 @@ static vm_fault_t rdma_umap_fault(struct vm_fault *vmf)
> > >
> > >  	/* Read only pages can just use the system zero page. */
> > >  	if (!(vmf->vma->vm_flags & (VM_WRITE | VM_MAYWRITE))) {
> > > -		vmf->page = ZERO_PAGE(vmf->vm_start);
> > > +		vmf->page = ZERO_PAGE(vmf->vma->vm_start);
> > >  		get_page(vmf->page);
> > >  		return 0;
> > >  	}
> > >
> >
> > Thanks Heiko, this looks right to me.
> >
> > I'm surprised to be seeing this at this point, these patches should
> > have been seen by 0 day for several days now, and they were in
> > linux-next already too..
>
> Most architectures have versions of ZERO_PAGE() which ignore the
> argument so that the code builds anyway. I'm not sure if 0-day also
> tests s390x builds (which is where I ran into this).

According to 0-build results for this patch, the answer is yes, it builds.
s390                        default_defconfig

And it compiles uverbs_main.c (CONFIG_INFINIBAND_USER_ACCESS)
  kernel git:(rdma-next) grep INFIN arch/s390/configs/debug_defconfig
CONFIG_INFINIBAND=m
CONFIG_INFINIBAND_USER_ACCESS=m
CONFIG_MLX4_INFINIBAND=m
CONFIG_MLX5_INFINIBAND=m

Thanks

>
> Michal Kubecek