@@ -37,13 +37,12 @@ The supervisor joins the workload's **user namespace only** to preserve UID/GID
3737mapping for shared-volume access. PID, mount, and network namespaces remain
3838separate. The channel volume uses shared SELinux relabeling (` :z ` ).
3939
40- Before starting either container, the driver uploads volume-relative archives
41- directly to the channel and workspace volume destinations. A rootfs upload on a
42- stopped Podman container does not populate nested named volumes. Restart restores
43- only the channel bootstrap into the existing channel volume, preserving the
44- workspace. The workload starts before the supervisor so its user namespace exists
45- when the supervisor joins it; a stopped supervisor resolves that namespace again
46- on its next start.
40+ Before starting either container, the driver uploads bootstrap files to the
41+ channel volume. For managed ` /sandbox ` , it also prepares the workspace volume;
42+ custom image workspaces need no upload. Restart restores only the channel
43+ bootstrap, preserving the workspace. The workload starts before the supervisor
44+ so its user namespace exists when the supervisor joins it; a stopped supervisor
45+ resolves that namespace again on its next start.
4746
4847The runtime must pass the sandbox's unprivileged enforcement probe, including
4948nested seccomp notification and Landlock. Unsupported runtime defaults fail
@@ -100,16 +99,15 @@ mounts. For a custom path, image volumes and driver mounts cannot cover the
10099workspace or one of its parents; mounts nested below it remain valid. Podman
101100creates a private volume for each image-declared path nested below it.
102101
103- For a custom path, Podman mounts a persistent workspace volume there. When the
104- volume is first created, Podman copies any files already in that image directory
105- into it. OpenShell keeps their ownership and permissions, starts as the final
106- non-root user, and rejects the image if that user cannot reach and write the
107- directory. Agent commands use the path as their working directory;
102+ For a custom path, Podman uses the image's container filesystem directly.
103+ OpenShell keeps the image directory's ownership and permissions, starts as the
104+ final non-root user, and rejects the image if that user cannot reach and write
105+ the directory. Agent commands use the path as their working directory;
108106` filesystem.include_workdir ` grants access to it when enabled.
109107
110- For ` /sandbox ` , OpenShell prepares the managed workspace before switching to
111- the non-root user. The separate supervisor receives the path but does not mount
112- the workspace. Test custom images with the Podman configuration you will use .
108+ For ` /sandbox ` , OpenShell creates a workspace volume and prepares it before
109+ switching to the non-root user. The separate supervisor receives the path but
110+ does not mount the workspace .
113111
114112## Lifecycle and readiness
115113
0 commit comments