From mboxrd@z Thu Jan 1 00:00:00 1970 From: Michal Hocko Subject: Re: MADV_DONTNEED semantics? Was: [RFC PATCH] mm: madvise: Ignore repeated MADV_DONTNEED hints Date: Fri, 6 Feb 2015 21:45:36 +0100 Message-ID: <20150206204536.GA24245@dhcp22.suse.cz> References: <54D08483.40209@suse.cz> <20150203105301.GC14259@node.dhcp.inet.fi> <54D0B43D.8000209@suse.cz> <54D0F56A.9050003@gmail.com> <54D22298.3040504@suse.cz> <54D2508A.9030804@suse.cz> <20150205154102.GA20607@dhcp22.suse.cz> <54D4E47E.4020509@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Content-Disposition: inline In-Reply-To: <54D4E47E.4020509-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> Sender: linux-api-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: "Michael Kerrisk (man-pages)" Cc: Vlastimil Babka , "Kirill A. Shutemov" , Dave Hansen , Mel Gorman , "linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org" , Minchan Kim , Andrew Morton , lkml , Linux API , linux-man , Hugh Dickins List-Id: linux-api@vger.kernel.org On Fri 06-02-15 16:57:50, Michael Kerrisk wrote: [...] > > Yes, this wording is better because many users are not aware of > > MAP_ANON|MAP_SHARED being file backed in fact and mmap man page doe= sn't > > mention that. >=20 > (Michal, would you have a text to propose to add to the mmap(2) page? > Maybe it would be useful to add something there.) I am half way on vacation, but I can cook a patch after I am back after week. =20 > > I am just wondering whether it makes sense to mention that MADV_DON= TNEED > > for shared mappings might be surprising and not freeing the backing > > pages thus not really freeing memory until there is a memory > > pressure. But maybe this is too implementation specific for a man > > page. What about the following wording on top of yours? > > " > > Please note that the MADV_DONTNEED hint on shared mappings might no= t > > lead to immediate freeing of pages in the range. The kernel is free= to > > delay this until an appropriate moment. RSS of the calling process = will > > be reduced however. > > " >=20 > Thanks! I added this, but dropped in the word "immediately" in the la= st=20 > sentence, since I assume that was implied. So now we have: >=20 > After a successful MADV_DONTNEED operation, the seman= =E2=80=90 > tics of memory access in the specified region ar= e > changed: subsequent accesses of pages in the range wil= l > succeed, but will result in either repopulating the mem= =E2=80=90 > ory contents from the up-to-date contents of the under= =E2=80=90 > lying mapped file (for shared file mappings, share= d > anonymous mappings, and shmem-based techniques such a= s > System V shared memory segments) or zero-fill-on-deman= d > pages for anonymous private mappings. >=20 > Note that, when applied to shared mappings, MADV_DONT= =E2=80=90 > NEED might not lead to immediate freeing of the pages i= n > the range. The kernel is free to delay freeing th= e > pages until an appropriate moment. The resident se= t > size (RSS) of the calling process will be immediatel= y > reduced however. This sounds good to me and it is definitely much better than the curren= t state. Thanks! > The current draft of the page can be found in a branch, > http://git.kernel.org/cgit/docs/man-pages/man-pages.git/log/?h=3Ddraf= t_madvise >=20 > Thanks, >=20 > Michael >=20 >=20 >=20 > --=20 > Michael Kerrisk > Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/ > Linux/UNIX System Programming Training: http://man7.org/training/ --=20 Michal Hocko SUSE Labs