Re: [PATCH 4/6] vhost_vdpa: support doorbell mapping via mmap

From: Michael S. Tsirkin
Date: Wed Jun 03 2020 - 02:35:08 EST


On Wed, Jun 03, 2020 at 12:18:44PM +0800, Jason Wang wrote:
>
> On 2020/6/2 äå9:31, Michael S. Tsirkin wrote:
> > On Tue, Jun 02, 2020 at 02:49:38PM +0800, Jason Wang wrote:
> > > On 2020/6/2 äå12:56, Michael S. Tsirkin wrote:
> > > > On Tue, Jun 02, 2020 at 03:22:49AM +0800, kbuild test robot wrote:
> > > > > Hi Jason,
> > > > >
> > > > > I love your patch! Yet something to improve:
> > > > >
> > > > > [auto build test ERROR on vhost/linux-next]
> > > > > [also build test ERROR on linus/master v5.7 next-20200529]
> > > > > [if your patch is applied to the wrong git tree, please drop us a note to help
> > > > > improve the system. BTW, we also suggest to use '--base' option to specify the
> > > > > base tree in git format-patch, please seehttps://stackoverflow.com/a/37406982]
> > > > >
> > > > > url:https://github.com/0day-ci/linux/commits/Jason-Wang/vDPA-doorbell-mapping/20200531-070834
> > > > > base:https://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost.git linux-next
> > > > > config: m68k-randconfig-r011-20200601 (attached as .config)
> > > > > compiler: m68k-linux-gcc (GCC) 9.3.0
> > > > > reproduce (this is a W=1 build):
> > > > > wgethttps://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
> > > > > chmod +x ~/bin/make.cross
> > > > > # save the attached .config to linux build tree
> > > > > COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=m68k
> > > > >
> > > > > If you fix the issue, kindly add following tag as appropriate
> > > > > Reported-by: kbuild test robot<lkp@xxxxxxxxx>
> > > > >
> > > > > All errors (new ones prefixed by >>, old ones prefixed by <<):
> > > > >
> > > > > drivers/vhost/vdpa.c: In function 'vhost_vdpa_fault':
> > > > > > > drivers/vhost/vdpa.c:754:22: error: implicit declaration of function 'pgprot_noncached' [-Werror=implicit-function-declaration]
> > > > > 754 | vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
> > > > > | ^~~~~~~~~~~~~~~~
> > > > > > > drivers/vhost/vdpa.c:754:22: error: incompatible types when assigning to type 'pgprot_t' {aka 'struct <anonymous>'} from type 'int'
> > > > > cc1: some warnings being treated as errors
> > > > >
> > > > > vim +/pgprot_noncached +754 drivers/vhost/vdpa.c
> > > > >
> > > > > 742
> > > > > 743 static vm_fault_t vhost_vdpa_fault(struct vm_fault *vmf)
> > > > > 744 {
> > > > > 745 struct vhost_vdpa *v = vmf->vma->vm_file->private_data;
> > > > > 746 struct vdpa_device *vdpa = v->vdpa;
> > > > > 747 const struct vdpa_config_ops *ops = vdpa->config;
> > > > > 748 struct vdpa_notification_area notify;
> > > > > 749 struct vm_area_struct *vma = vmf->vma;
> > > > > 750 u16 index = vma->vm_pgoff;
> > > > > 751
> > > > > 752 notify = ops->get_vq_notification(vdpa, index);
> > > > > 753
> > > > > > 754 vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
> > > > > 755 if (remap_pfn_range(vma, vmf->address & PAGE_MASK,
> > > > > 756 notify.addr >> PAGE_SHIFT, PAGE_SIZE,
> > > > > 757 vma->vm_page_prot))
> > > > > 758 return VM_FAULT_SIGBUS;
> > > > > 759
> > > > > 760 return VM_FAULT_NOPAGE;
> > > > > 761 }
> > > > > 762
> > > > Yes well, all this remapping clearly has no chance to work
> > > > on systems without CONFIG_MMU.
> > >
> > > It looks to me mmap can work according to Documentation/nommu-mmap.txt. But
> > > I'm not sure it's worth to bother.
> > >
> > > Thanks
> >
> > Well
> >
> > int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
> > unsigned long pfn, unsigned long size, pgprot_t prot)
> > {
> > if (addr != (pfn << PAGE_SHIFT))
> > return -EINVAL;
> >
> > vma->vm_flags |= VM_IO | VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP;
> > return 0;
> > }
> > EXPORT_SYMBOL(remap_pfn_range);
> >
> >
> > So things aren't going to work if you have a fixed PFN
> > which is the case of the hardware device.
>
>
> Looking at the implementation of some drivers e.g mtd_char. If I read the
> code correctly, we can do this by providing get_unmapped_area method and use
> physical address directly.
>
> But start form CONFIG_MMU should be fine. Do you prefer making vhost_vdpa
> depends on CONFIG_MMU or just fail mmap when CONFIG_MMU is not configured?
>
> Thanks

I'd just not specify the mmap callback at all.

>
> >
> >
> > > >
> > > >