Hello,
Thanks for the report, I have commited a fix that is a bit different.
Samuel
Richard, le mar. 13 oct. 2026 19:33:50 +0800, a ecrit:
> If opening the backing store fails during a metadata-only open,
> check_dev() ignores the error and leaves dev->store NULL.
> create_partitions() then passes it to ped_device_new_from_store(),
> triggering Parted's "source != NULL" assertion.
>
> Ensure the backing store is open before probing partitions and
> propagate any error.
>
> To reproduce (with /tmp/no-such-image absent):
> touch /tmp/storeio-node
> settrans -ac /tmp/storeio-node /hurd/storeio -r -T file /tmp/no-such-image
> stat /tmp/storeio-node/1
>
> Before this patch, storeio aborts and stat reports "Computer bought
> the farm". With the patch, stat reports "No such file or directory"
> and storeio remains running.
>
> ---
> storeio/part.c | 7 ++++++-
> 1 file changed, 6 insertions(+), 1 deletion(-)
>
> diff --git a/storeio/part.c b/storeio/part.c
> index 47d37df7..142427db 100644
> --- a/storeio/part.c
> +++ b/storeio/part.c
> @@ -17,6 +17,7 @@
> along with this program; if not, write to the Free Software
> Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA. */
>
> +#include <fcntl.h>
> #include <parted/parted.h>
> #include <hurd/store.h>
>
> @@ -91,6 +92,11 @@ create_partitions (void)
> {
> debug ("create_partitions:\n");
> struct node *dir = netfs_root_node;
> + /* A lookup can reach us before the root has been opened for I/O. */
> + error_t err = check_dev (dir, NULL, O_READ);
> + if (err)
> + return err;
> +
> size_t last_partition_num =
> (size_t) set_last_partition_num (dir->nn->dev->store);
> if (last_partition_num == 0)
> @@ -100,7 +106,6 @@ create_partitions (void)
> return ENOTDIR;
> }
>
> - error_t err;
> dir->nn->entries_num = last_partition_num;
> dir->nn->entries = malloc (last_partition_num * sizeof (struct node *));
> if (!dir->nn->entries)
>
