[Bugs] [Bug 1235582] snapd crashed due to stack overflow

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=1235582



--- Comment #2 from krishnan parthasarathi <kparthas at redhat.com> ---
RCA
----

The stack overflow was seen when older snapshots were being deleted while new
ones were being created concurently. In the setup detailed above, snapshot
scheduler creates snapshots periodically and auto-delete of snapshots is
enabled. When no. of snapshots in the system (of the volume) exceeds the
soft-limit configured, snapshots are (auto-)deleted. The crash happened when a
scheduled snapshot-create coincided with the auto-delete triggered
snapshot-delete operation.

Implementation detail
----------------------
Snapshot daemon uses gfapi interface to serve user-serviceable snapshots. gfapi
interface creates a new glfs object for every snapshot (volume) serviced. This
object is 'linked' with a global xlator object until the time glfs object is
fully initialized (i.e, set-volume operation is complete). The global xlator
object's ctx (glusterfs_ctx_t) object is being modified in a thread-unsafe
manner and could refer to a destroyed ctx (which belonged to glfs representing
a deleted snapshot).

Fix outline
------------
All initialisation managment operations (e.g, RPCs like DUMP_VERSION,
SET_VOLUME, etc.) must refer to the corresponding translator objects in the
glfs' graph.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=gfSlGEm2qM&a=cc_unsubscribe