Page MenuHomeFreeBSD

iflib: allow clone detach if not yet init
ClosedPublic

Authored by kevans on Mar 9 2021, 12:17 PM.
Tags
None
Referenced Files
F175592595: D29145.diff
Sun, Oct 11, 9:52 PM
Unknown Object (File)
Tue, Sep 29, 7:47 AM
Unknown Object (File)
Mon, Sep 28, 7:45 PM
Unknown Object (File)
Sat, Sep 26, 7:51 PM
Unknown Object (File)
Sep 5 2026, 5:53 PM
Unknown Object (File)
Sep 5 2026, 5:52 PM
Unknown Object (File)
Sep 5 2026, 5:23 AM
Unknown Object (File)
Sep 4 2026, 6:16 PM
Subscribers

Details

Summary

If we hit an error during init, then we'll unwind our state and attempt
to detach the device -- don't block it.

This was discovered by creating a wg0 with missing parameters; said
failure ended up leaving this orphaned device in place and ended up
panicking the system upon enumeration of the dev.* sysctl space.

MFC after: 3 days

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

This happens after an error from iflib_pseudo_register()?

This happens after an error from iflib_pseudo_register()?

Correct- the error path I was specifically seeing is iflib_clone_create() -> iflib_pseudo_register() -> IFDI_CLONEATTACH(); iflib_clone_create tries to delete the child but gets smacked with an EBUSY from this. I thought about setting IFC_IN_DETACH, but it's all barely-constructed at that point.

This happens after an error from iflib_pseudo_register()?

Correct- the error path I was specifically seeing is iflib_clone_create() -> iflib_pseudo_register() -> IFDI_CLONEATTACH(); iflib_clone_create tries to delete the child but gets smacked with an EBUSY from this. I thought about setting IFC_IN_DETACH, but it's all barely-constructed at that point.

Sure, I think this way is fine. Sorry for the breakage.

This revision is now accepted and ready to land.Mar 9 2021, 3:40 PM
This revision was automatically updated to reflect the committed changes.