On Wed, Dec 06, 2023 at 02:56:05PM +0800, Yu Kuai wrote: > > > - invalidate_inode_pages2( > > > - blkif->vbd.bdev_handle->bdev->bd_inode->i_mapping); > > > + invalidate_bdev(blkif->vbd.bdev_handle->bdev); > > > > blkbak is a bdev exported. I don't think it should ever call > > invalidate_inode_pages2, through a wrapper or not. > > I'm not sure about this. I'm not familiar with xen/blkback, but I saw > that xen-blkback will open a bdev from xen_vbd_create(), hence this > looks like a dm/md for me, hence it sounds reasonable to sync + > invalidate the opened bdev while initialization. Please kindly correct > me if I'm wrong. I guess we have enough precedence for this, so the switchover here isn't wrong. But all this invalidating of the bdev cache seems to be asking for trouble.
