On Fri, Apr 30, 2021 at 11:50 AM Eric W. Biederman
<ebiederm@xxxxxxxxxxxx> wrote:
>
> A printk specific variant could easily be modified to always restart or
> to simply ignore renames and changes to the mount tree.
Exactly. I think a "ignore renames and mount tree changes" version for
printk would be the right thing.
Yeah, you can in theory get inconsistent results, but everything is
RCU-protected, so you'd get the same kind of "its' kind of valid, but
in race situations you might get a mix of two components" that '%pd'
gives for a dentry case.
That would allow people to use '%pD' and get reasonable results,
without having them actually interact with locks (that may or may not
be held by the thread trying to print debug messages).
Linus
