On Thu, Nov 14, 2024 at 12:03:16PM +0100, Amir Goldstein wrote: > On Thu, Nov 14, 2024 at 11:12 AM Christian Brauner <brauner@xxxxxxxxxx> wrote: > > > > On Thu, Nov 14, 2024 at 11:05:36AM +0100, Amir Goldstein wrote: > > > ovl_setup_cred_for_create() decrements one refcount of new creds and > > > ovl_revert_creds() in callers decrements the last refcount. > > > > > > In preparation to revert_creds_light() back to caller creds, pass an > > > explicit reference of the creators creds to the callers and drop the > > > refcount explicitly in the callers after ovl_revert_creds(). > > > > > > Signed-off-by: Amir Goldstein <amir73il@xxxxxxxxx> > > > --- > > > > > > Miklos, Christian, > > > > > > I was chasing a suspect memleak in revert_creds_light() patches. > > > This fix is unrelated to memleak but I think it is needed for > > > correctness anyway. > > > > > > This applies in the middle of the series after adding the > > > ovl_revert_creds() helper. > > > > I'm going to try and reproduce the kmemleak with your ovl_creds branch > > as is and then retry with the series applied as is plus one small fix > > you correctly pointed out. > > Don't bother. The fix that Vinicius sent me was correct. > I still want to use this patch, so would appreciate a review. Ah, thanks for letting me know!
