On Mon, Nov 25, 2019 at 6:46 PM J . Bruce Fields bfields@fieldses.org wrote:
On Sun, Nov 24, 2019 at 09:31:45PM +0200, Amir Goldstein wrote:
Push clamping timestamps down the call stack into notify_change(), so in-kernel callers like nfsd and overlayfs will get similar timestamp set behavior as utimes.
So, nfsd has always bypassed timestamp_truncate() and we've never noticed till now? What are the symptoms? (Do timestamps go backwards after cache eviction on filesystems with large time granularity?)
Clamping seems to be new behavior since v5.4-rc1. Before that clamping was done implicitly when hitting the disk IIUC, so it was observed mostly after cache eviction.
Looks like generic/402 has never run in my tests:
generic/402 [not run] no kernel support for y2038 sysfs switch
The test in its current form is quite recent as well or at the _require has changed recently. See acb2ba78 - overlay: support timestamp range check
You'd probably need something similar for nfs (?)
Thanks, Amir.