Re: [RFC v4 05/11] vdpa: Support transferring virtual addressing during DMA mapping

[Dan Carpenter <dan.carpenter@oracle.com>]

[-- Attachment #1: Type: text/plain, Size: 4395 bytes --]

Hi Xie,

url:    https://github.com/0day-ci/linux/commits/Xie-Yongji/Introduce-VDUSE-vDPA-Device-in-Userspace/20210223-200222
base:   https://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost.git linux-next
config: i386-randconfig-m021-20210223 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0

If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>

New smatch warnings:
drivers/vhost/vdpa.c:660 vhost_vdpa_va_map() warn: inconsistent returns '&dev->mm->mmap_lock'.

vim +660 drivers/vhost/vdpa.c

d8433d0f11798a Xie Yongji 2021-02-23  614  static int vhost_vdpa_va_map(struct vhost_vdpa *v,
d8433d0f11798a Xie Yongji 2021-02-23  615  			     u64 iova, u64 size, u64 uaddr, u32 perm)
d8433d0f11798a Xie Yongji 2021-02-23  616  {
d8433d0f11798a Xie Yongji 2021-02-23  617  	struct vhost_dev *dev = &v->vdev;
d8433d0f11798a Xie Yongji 2021-02-23  618  	u64 offset, map_size, map_iova = iova;
d8433d0f11798a Xie Yongji 2021-02-23  619  	struct vdpa_map_file *map_file;
d8433d0f11798a Xie Yongji 2021-02-23  620  	struct vm_area_struct *vma;
d8433d0f11798a Xie Yongji 2021-02-23  621  	int ret;
d8433d0f11798a Xie Yongji 2021-02-23  622  
d8433d0f11798a Xie Yongji 2021-02-23  623  	mmap_read_lock(dev->mm);
d8433d0f11798a Xie Yongji 2021-02-23  624  
d8433d0f11798a Xie Yongji 2021-02-23  625  	while (size) {
d8433d0f11798a Xie Yongji 2021-02-23  626  		vma = find_vma(dev->mm, uaddr);
d8433d0f11798a Xie Yongji 2021-02-23  627  		if (!vma) {
d8433d0f11798a Xie Yongji 2021-02-23  628  			ret = -EINVAL;
d8433d0f11798a Xie Yongji 2021-02-23  629  			goto err;
d8433d0f11798a Xie Yongji 2021-02-23  630  		}
d8433d0f11798a Xie Yongji 2021-02-23  631  		map_size = min(size, vma->vm_end - uaddr);
d8433d0f11798a Xie Yongji 2021-02-23  632  		offset = (vma->vm_pgoff << PAGE_SHIFT) + uaddr - vma->vm_start;
d8433d0f11798a Xie Yongji 2021-02-23  633  		map_file = kzalloc(sizeof(*map_file), GFP_KERNEL);
d8433d0f11798a Xie Yongji 2021-02-23  634  		if (!map_file) {
d8433d0f11798a Xie Yongji 2021-02-23  635  			ret = -ENOMEM;
d8433d0f11798a Xie Yongji 2021-02-23  636  			goto err;
d8433d0f11798a Xie Yongji 2021-02-23  637  		}
d8433d0f11798a Xie Yongji 2021-02-23  638  		if (vma->vm_file && (vma->vm_flags & VM_SHARED) &&
d8433d0f11798a Xie Yongji 2021-02-23  639  			!(vma->vm_flags & (VM_IO | VM_PFNMAP))) {
d8433d0f11798a Xie Yongji 2021-02-23  640  			map_file->file = get_file(vma->vm_file);
d8433d0f11798a Xie Yongji 2021-02-23  641  			map_file->offset = offset;
d8433d0f11798a Xie Yongji 2021-02-23  642  		}
d8433d0f11798a Xie Yongji 2021-02-23  643  		ret = vhost_vdpa_map(v, map_iova, map_size, uaddr,
d8433d0f11798a Xie Yongji 2021-02-23  644  				     perm, map_file);
d8433d0f11798a Xie Yongji 2021-02-23  645  		if (ret) {
d8433d0f11798a Xie Yongji 2021-02-23  646  			if (map_file->file)
d8433d0f11798a Xie Yongji 2021-02-23  647  				fput(map_file->file);
d8433d0f11798a Xie Yongji 2021-02-23  648  			kfree(map_file);
d8433d0f11798a Xie Yongji 2021-02-23  649  			goto err;
d8433d0f11798a Xie Yongji 2021-02-23  650  		}
d8433d0f11798a Xie Yongji 2021-02-23  651  		size -= map_size;
d8433d0f11798a Xie Yongji 2021-02-23  652  		uaddr += map_size;
d8433d0f11798a Xie Yongji 2021-02-23  653  		map_iova += map_size;
d8433d0f11798a Xie Yongji 2021-02-23  654  	}
d8433d0f11798a Xie Yongji 2021-02-23  655  	mmap_read_unlock(dev->mm);
                                                ^^^^^^^^^^^^^^^^^^^^^^^^^
The success path unlocks.

d8433d0f11798a Xie Yongji 2021-02-23  656  
d8433d0f11798a Xie Yongji 2021-02-23  657  	return 0;
d8433d0f11798a Xie Yongji 2021-02-23  658  err:
d8433d0f11798a Xie Yongji 2021-02-23  659  	vhost_vdpa_unmap(v, iova, map_iova - iova);
d8433d0f11798a Xie Yongji 2021-02-23 @660  	return ret;

The error path doesn't seem to unlock.  I did look at vhost_vdpa_unmap()
briefly, but I didn't see any reason that we don't need to unlock but I
may have missed it.  The point of sending these bug reports as soon as
possible is that hopefully you know the answer off the top of your
head.  :)

d8433d0f11798a Xie Yongji 2021-02-23  661  }

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

[-- Attachment #2: config.gz --]
[-- Type: application/gzip, Size: 37926 bytes --]