Still testing.
One place I don't know what to do about is the vfs_hash_insert call. Apart from that all error paths that I see got a vgone call.
Differential D23215
[wip] ufs: add missing vgone calls to accomodate unlocked vnode with i_mode == 0 mjg on Jan 16 2020, 8:45 PM. Authored by Tags None Referenced Files
Details
Still testing. One place I don't know what to do about is the vfs_hash_insert call. Apart from that all error paths that I see got a vgone call.
Diff Detail
Event TimelineComment Actions Well in that spirit I would argue there should be a (debug) flag where the filesystem can denote that the vnode is ready for use (or ready to no longer be used, after vgone). Then the vfs layer can assert on it in a generic manner. Combining vgone + vput into something in the lines of vabort is fine with me.
Comment Actions This got sorted out with r357070 for now, but I think reliance on vput atomicity is an avoidable constraint on what can be done in that primitive and stuff like should go in for all filesystems. Comment Actions I think that this change can/should be committed regardless of other actions in this area. It is correct for the places where you put it. |