On Tue, May 24, 2022 at 03:18:53PM -0400, Ryan Stone wrote: > I believe that if we never enter that if statement, then we will leak > the entries that are supposed to be cleaned up. That's better than > the use-after-free that I feared but still not good. I don't know, we are running enhanced IPoIB all the time and I never got any reports about kmemleaks/KASAN in that area. Thanks > > On Tue, May 24, 2022 at 3:08 PM Leon Romanovsky <leon@xxxxxxxxxx> wrote: > > > > On Tue, May 24, 2022 at 09:33:52AM -0400, Ryan Stone wrote: > > > On Tue, May 24, 2022 at 3:07 AM Leon Romanovsky <leon@xxxxxxxxxx> wrote: > > > > IPoIB in mlx5 is HW offloaded version of ulp/ipoib one. AFAIK, it doesn't > > > > change "tx_tail" and we won't enter into this if (...). > > > > > > > > Thanks > > > > > > I don't quite follow this response. Is this the if statement that you > > > mean that we won't fall into? > > > > > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/infiniband/ulp/ipoib/ipoib_ib.c#n682 > > > > Yes, I think so, maybe wrong here.
