From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tejun Heo Subject: Re: [PATCH 4/4] memcg: always enable kmemcg on the default hierarchy Date: Tue, 1 Sep 2015 14:51:57 -0400 Message-ID: <20150901185157.GD18956@htj.dyndns.org> References: <1440775530-18630-1-git-send-email-tj@kernel.org> <1440775530-18630-5-git-send-email-tj@kernel.org> <20150828164918.GJ9610@esperanza> <20150828171438.GD21463@dhcp22.suse.cz> <20150828174140.GN26785@mtj.duckdns.org> <20150901124459.GC8810@dhcp22.suse.cz> Mime-Version: 1.0 Return-path: DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=sender:date:from:to:cc:subject:message-id:references:mime-version :content-type:content-disposition:in-reply-to:user-agent; bh=NN+VANBVAas+oL3yC4maMvD2znwijVE05jYKxUC3/EI=; b=t4SH4VgG1RRFockbYW9tnt3T4Ke+FAZitw8k/o+NAok+uFE+1jYWZoDZ8jHxMe73sT ZnbWtYzCo1gkAq7Af5U5neHa6KONWLsF9mzU2Jr1vUr7cUG2rZ8z96jVR5TMaMTJV6zN u8i4h4oS14NeAGQUCfQQTLNceBZAMz8SeTPQ4ZYEoItxiXO3CSgQieHOYs6R9jbGNu6K NyzsTonentSV5/EOYL/CYfjytcWaqddKYP0yL0O6zxFfIYH1oquGOV10winKFPfK4BNC VwycKjoYEJlmh5YeaoPl8en06RCJrdc8fJLj0RVtjBLppsZfANnLZLgYvPou++nRxftd whWw== Content-Disposition: inline In-Reply-To: <20150901124459.GC8810-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org> Sender: cgroups-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Michal Hocko Cc: Vladimir Davydov , hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org, cgroups-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org, kernel-team-b10kYP2dOMg@public.gmane.org Hello, On Tue, Sep 01, 2015 at 02:44:59PM +0200, Michal Hocko wrote: > The runtime overhead is not negligible and I do not see why everybody > should be paying that price by default. I can definitely see the reason why > somebody would want to enable the kmem accounting but many users will > probably never care because the kernel footprint would be in the noise > wrt. user memory. We said the same thing about hierarchy support. Sure, it's not the same but I think it's wiser to keep the architectural decisions at a higher level. I don't think kmem overhead is that high but if this actually is a problem we'd need a per-cgroup knob anyway. > > We wanna put all memory consumptions under the same roof by default. > > But I am not sure we will ever achieve this. E.g. hugetlb memory is way > too different to be under the same charging by default IMO. Also all > the random drivers calling into the page allocator directly in the user > context would need to charge explicitly. Oh I meant the big ones. I don't think we'll achieve 100% coverage either but even just catching the major ones, kmem and tcp socket buffers, should remove most ambiguities around memory consumption. Thanks. -- tejun