On Thu, 5 Jul 2018 17:12:36 +0200 Manfred Spraul <manf...@colorfullife.com> wrote:
> Hi Dmitry, > > On 07/05/2018 10:36 AM, Dmitry Vyukov wrote: > > [...] > > Hi Manfred, > > > > The series looks like a significant improvement to me. Thanks! > > > > I feel that this code can be further simplified (unless I am missing > > something here). Please take a look at this version: > > > > https://github.com/dvyukov/linux/commit/f77aeaf80f3c4ab524db92184d874b03063fea3a?diff=split > > > > This is on top of your patches. It basically does the same as your > > code, but consolidates all id/seq assignment and dealing with next_id, > > and deduplicates code re CONFIG_CHECKPOINT_RESTORE. Currently it's a > > bit tricky to follow e.g. where exactly next_id is consumed and where > > it needs to be left intact. > > The only difference is that my code assigns new->id earlier. Not sure > > if it can lead to anything bad. But if yes, then it seems that > > currently uninitialized new->id is exposed. If necessary (?) we could > > reset new->id in the same place where we set new->deleted. > Everything looks correct for me, it is better than the current code. > Except that you didn't sign off your last patch. > > As next step: Who can merge the patches towards linux-next? Me. But it's unclear which patchset we're talking about. What's the plan here? To combine both efforts? > The only open point that I see are stress tests of the error codepaths. > > And: > I don't think that the patches are relevant for linux-stable, correct?
