On Sat, Sep 21, 2024 at 07:03:18PM +0200, Dag-Erling Smørgrav wrote: > Konstantin Belousov <kostik...@gmail.com> writes: > > Dag-Erling Smørgrav <d...@freebsd.org> writes: > > > Konstantin Belousov <k...@freebsd.org> writes: > > > > commit a52b30ff98cdab82af140285fa7fcdf1036fef27 > > > > > > > > sys_pipe: consistently use cr_ruidinfo for accounting of pipebuf > > > > > > > > Tested by: yasu > > > > Sponsored by: The FreeBSD Foundation > >> > MFC after: 1 week > > > This appears to be the opposite of the patch which you posted on > > > -current and which yasu@ tested [...] > > Before committing anything, I did a self-review and remembered that I > > have did a lot of considerations when implementing swap accounting and > > decided that ruid is the right target for charge. > > > > Besides stating the obvious fact above, what do you expect me to answer/ > > react to your mail? > > My point is that the commit message claims the patch was tested by yasu@ > when in fact it wasn't. If you're convinced that this is the correct > solution then that's fine, and it does appear to work, but don't claim > that it's been tested by others when it hasn't.
The main part of the patch was to ensure consistency in updates: it must be either always uidinfo, or always ruidinfo. Which one specifically would affect the limit's semantic, but not the buggy behavior you reported.
