* [email protected] ([email protected]) wrote:
> Hi,
Hi Misono,
> I'm testing virtiofs on ARM enviroment.
> Since current ARM does not support DAX, I use cache-size=0 for
> vhost-user-fs-pci parameter.
>
> This results in the following error message during unmount:
> ```
> vhost_user_fs_slave_unmap: unmap when DAX cache not present
> lo_destroy: unmap during destroy failed
> ```
Thanks for reporting this,
> This is because lo_destroy() always calls fuse_virtio_unmap().
> So, I want to fix this by not calling fuse_virtio_unmap() when cache is not
> used,
> but is there any good way for virtiofsd to know vhost-user's cahce-size?
Unfortunately not; that's why I used the special ~0 as a 'unmap all'.
It feels easier to fix the qemu side to be happy unmapping everything
when everything is actually nothing :-)
Does the following (build tested only) patch fix it:
diff --git a/hw/virtio/vhost-user-fs.c b/hw/virtio/vhost-user-fs.c
index 6cacdb24b0..07a04d9b21 100644
--- a/hw/virtio/vhost-user-fs.c
+++ b/hw/virtio/vhost-user-fs.c
@@ -92,10 +92,6 @@ uint64_t vhost_user_fs_slave_unmap(struct vhost_dev *dev,
VhostUserFSSlaveMsg *s
return -1;
}
size_t cache_size = fs->conf.cache_size;
- if (!cache_size) {
- fprintf(stderr, "%s: unmap when DAX cache not present\n", __func__);
- return (uint64_t)-1;
- }
void *cache_host = memory_region_get_ram_ptr(&fs->cache);
unsigned int i;
@@ -106,13 +102,19 @@ uint64_t vhost_user_fs_slave_unmap(struct vhost_dev *dev,
VhostUserFSSlaveMsg *s
*/
for (i = 0; i < VHOST_USER_FS_SLAVE_ENTRIES; i++) {
void *ptr;
+ if (sm->len[i] == ~(uint64_t)0) {
+ /* Special case meaning the whole arena */
+ sm->len[i] = cache_size;
+ }
+
if (sm->len[i] == 0) {
continue;
}
- if (sm->len[i] == ~(uint64_t)0) {
- /* Special case meaning the whole arena */
- sm->len[i] = cache_size;
+ if (!cache_size) {
+ fprintf(stderr, "%s: unmap when DAX cache not present\n",
__func__);
+ res = -1;
+ break;
}
if ((sm->c_offset[i] + sm->len[i]) < sm->len[i] ||
> Thanks.
> Misono
>
> _______________________________________________
> Virtio-fs mailing list
> [email protected]
> https://www.redhat.com/mailman/listinfo/virtio-fs
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
_______________________________________________
Virtio-fs mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/virtio-fs
