linux-kernel - Re: question about page tables in DAX/FS/PMEM case

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <6e4a5296-0ad0-ab1e-40a0-c1f69d11300a@intel.com>
Date:   Thu, 21 Feb 2019 15:51:05 -0800
From:   Dave Hansen <dave.hansen@...el.com>
To:     Larry Bassel <larry.bassel@...cle.com>,
        Jerome Glisse <jglisse@...hat.com>
Cc:     linux-nvdimm@...ts.01.org, linux-kernel@...r.kernel.org,
        linux-mm@...ck.org
Subject: Re: question about page tables in DAX/FS/PMEM case

On 2/21/19 2:58 PM, Larry Bassel wrote:
> AFAIK there is no hardware benefit from sharing the page table
> directory within different page table. So the only benefit is the
> amount of memory we save.

The hardware benefit from schemes like this is that the CPU caches are
better utilized.  If two processes share page tables, they don't share
TLB entries, but they *do* share the contents of the CPU's caches.  That
will make TLB misses faster.

It probably doesn't matter *that* much in practice because the page
walker doing TLB fills does a pretty good job of hiding all the latency,
but it might matter in extreme cases.