linux-kernel - Re: [PATCH] lib/genalloc.c: include vmalloc.h

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:   Thu, 10 Jan 2019 15:51:08 -0800
From:   Andrew Morton <akpm@...ux-foundation.org>
To:     peter enderborg <peter.enderborg@...y.com>
Cc:     Linus Torvalds <torvalds@...ux-foundation.org>,
        Olof Johansson <olof@...om.net>,
        Linux List Kernel Mailing <linux-kernel@...r.kernel.org>,
        Huang Shijie <sjhuang@...vatar.ai>,
        Alexey Skidanov <alexey.skidanov@...el.com>
Subject: Re: [PATCH] lib/genalloc.c: include vmalloc.h

On Thu, 10 Jan 2019 14:13:44 +0100 peter enderborg <peter.enderborg@...y.com> wrote:

> On 1/7/19 11:56 PM, Andrew Morton wrote:
> > On Sat, 5 Jan 2019 13:35:33 -0800 Linus Torvalds <torvalds@...ux-foundation.org> wrote:
> >
> >> On Sat, Jan 5, 2019 at 1:21 PM Olof Johansson <olof@...om.net> wrote:
> >>> Fixes build break on most ARM/ARM64 defconfigs:
> >> Interesting.
> >>
> >> Andrew, I thought ARM was one of the platforms that your tree compiled
> >> against? Or was some other change just hiding this?
> >>
> > I've become a bit lazy with the cross-compiling because linux-next's
> > build coverage is pretty broad.  But Stephen's holiday interrupted
> > things and this one fell in that window.
> >
> Is it now OK to use vmalloc in fast path's on all platforms?

I'd say not.  It is slower than kmalloc and these changes did nothing
to speed it up.

That being said, vmalloc is for "large" memory allocations and large
amounts of memory take large amount of time to utilize, so unless the
code is allocating then freeing memory without using it (don't do that)
then vmalloc() can never be a "fast path" thing?