On Thu 25-10-18 06:56:37, Vlastimil Babka wrote: > On 10/25/18 1:17 AM, Andrew Morton wrote: > > On Mon, 22 Oct 2018 15:27:54 +0200 Vlastimil Babka <vbabka@xxxxxxx> wrote: > > > >>> : Moreover the oriinal code allowed to trigger > >>> : WARN_ON_ONCE(policy->mode == MPOL_BIND && (gfp & __GFP_THISNODE)); > >>> : in policy_node if the requested node (e.g. cpu local one) was outside of > >>> : the mbind nodemask. This is not possible now. We haven't heard about any > >>> : such warning yet so it is unlikely that it happens but still a signal of > >>> : a wrong code layering. > >> > >> Ah, as I said in the other mail, I think it's inaccurate, the warning > >> was not possible to hit. > >> > >> There's also a slight difference wrt MPOL_BIND. The previous code would > >> avoid using __GFP_THISNODE if the local node was outside of > >> policy_nodemask(). After your patch __GFP_THISNODE is avoided for all > >> MPOL_BIND policies. So there's a difference that if local node is > >> actually allowed by the bind policy's nodemask, previously > >> __GFP_THISNODE would be added, but now it won't be. I don't think it > >> matters that much though, but maybe the changelog could say that > >> (instead of the inaccurate note about warning). Note the other policy > >> where nodemask is relevant is MPOL_INTERLEAVE, and that's unchanged by > >> this patch. > > > > So the above could go into the changelog, yes? > > Yeah. Andrew. Do you want me to repost the patch or you plan to update the changelog yourself? -- Michal Hocko SUSE Labs
