From: Cyrill Gorcunov <gorcunov@gmail.com> To: Vladimir Davydov <vdavydov.dev@gmail.com> Cc: tml <tarantool-patches@freelists.org>, Alexander Turenko <alexander.turenko@tarantool.org> Subject: Re: [PATCH 2/3] slab_arena: Provide slab_arena_madvise_create to madvice slabs Date: Mon, 6 May 2019 14:03:26 +0300 [thread overview] Message-ID: <20190506110326.GQ2488@uranus.lan> (raw) In-Reply-To: <20190506103836.gxvrre4upkga7tej@esperanza> On Mon, May 06, 2019 at 01:38:36PM +0300, Vladimir Davydov wrote: > > > > if (arena->prealloc) { > > arena->arena = mmap_checked(arena->prealloc, > > @@ -205,6 +208,12 @@ slab_map(struct slab_arena *arena) > > __sync_sub_and_fetch(&arena->used, arena->slab_size); > > quota_release(arena->quota, arena->slab_size); > > } > > + > > +#ifdef TARANTOOL_SMALL_HAS_MADVISE > > + if (ptr && pm_atomic_load(&arena->advice) != MADV_NORMAL) > > + madvise(ptr, arena->slab_size, arena->advice); > > +#endif > > + > > Why use atomic? The arena cannot be used until it's fully initialized. Because otherwise we have a race: with current api madvise may be set up in any moment, say one thread setting up madvise and second thread calls for slab_map. Not a big deal probably but to pair with atomic write into @advice variable. ... > > +int > > +slab_arena_madvise_create(struct slab_arena *arena, int advice) > > I don't see much point in changing madvise settings on the fly. All this to not break backward compatibility. If we're allowed to change it, this won't be a problem to adjust the code. > Let's instead allow to pass DONTDUMP flag at creation time. > This means we have to combine MAP_ and MADV_ flags. I guess it > was a bad idea to pass MAP_ flags directly to small_create() in > the first place. What about introducing small-specific flags like > > SLAB_ARENA_SHARED > SLAB_ARENA_DONTDUMP > > Come to think of it, I assume we don't even need to introduce > SLAB_ARENA_SHARED - we always pass MAP_PRIVATE to small_create(). > Let's drop it, perhaps? > > It will break backward compatibility, but I don't think it really > matters, as tarantool/small is only used by tarantool - it isn't > a big deal if we change its API. We only need to bump the version > number, I guess. > > Alexander, please amend me if I'm wrong. OK, I'll rework in a separate branch, so we will jump back if needed. Cyrill
next prev parent reply other threads:[~2019-05-06 11:03 UTC|newest] Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top 2019-05-01 15:50 [PATCH 0/3] small: Prepare ground for madvise Cyrill Gorcunov 2019-05-01 15:50 ` [PATCH 1/3] build: Check for madvise syscall Cyrill Gorcunov 2019-05-06 10:25 ` Vladimir Davydov 2019-05-06 10:39 ` Cyrill Gorcunov 2019-05-01 15:50 ` [PATCH 2/3] slab_arena: Provide slab_arena_madvise_create to madvice slabs Cyrill Gorcunov 2019-05-06 10:38 ` Vladimir Davydov 2019-05-06 11:03 ` Cyrill Gorcunov [this message] 2019-05-06 13:46 ` Alexander Turenko 2019-05-01 15:50 ` [PATCH 3/3] test: slab_arena -- Verify madvise Cyrill Gorcunov 2019-05-06 10:45 ` Vladimir Davydov 2019-05-06 10:48 ` Cyrill Gorcunov 2019-05-06 10:55 ` Vladimir Davydov 2019-05-06 10:57 ` Cyrill Gorcunov 2019-05-06 11:04 ` Alexander Turenko
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20190506110326.GQ2488@uranus.lan \ --to=gorcunov@gmail.com \ --cc=alexander.turenko@tarantool.org \ --cc=tarantool-patches@freelists.org \ --cc=vdavydov.dev@gmail.com \ --subject='Re: [PATCH 2/3] slab_arena: Provide slab_arena_madvise_create to madvice slabs' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox