linux-kernel - Re: [PATCH 11/11] mm,sched: conditionally skip lazy TLB mm refcounting

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:   Mon, 6 Aug 2018 19:50:13 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Rik van Riel <riel@...riel.com>
Cc:     linux-kernel@...r.kernel.org, kernel-team@...com, mingo@...nel.org,
        luto@...nel.org, x86@...nel.org, efault@....de,
        dave.hansen@...el.com
Subject: Re: [PATCH 11/11] mm,sched: conditionally skip lazy TLB mm
 refcounting

On Fri, Aug 03, 2018 at 01:39:52PM -0400, Rik van Riel wrote:

> > At the very least the Changelog needs to explain why we cannot do
> > away with the swizzle now and how doing the swizzle without the
> > refcounting is not completely broken (I think I see, but urgh).
> 
> The changelog for patches 9 & 10 explains, I think.

They hint at it :-)

> What is missing from my explanation?
> How would you like to see it explained?

Maybe a few words like:

"Since ->active_mm is still used in a few sites, we must keep the
current tracking, such that we will not hit a kthread's NULL mm. Note
that lazy_tlb_exit_mmap() switches ->active_mm to &init_mm before taking
out the lazy mm."

That said; I'm not entirely sure we'll actually touch active_mm if we're
not a user task. The perf thing for example will only touch active_mm
when user_mode(regs).

But whatever, this was the only hickup.

Acked-by: Peter Zijlstra (Intel) <peterz@...radead.org>

Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)