linux-kernel - Re: [PATCH 3/5] zram: use zram_slot_lock instead of raw bit_spin

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]

Date:   Mon, 3 Apr 2017 15:34:20 +0900
From:   Minchan Kim <minchan@...nel.org>
To:     Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        linux-kernel@...r.kernel.org,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
        kernel-team@....com
Subject: Re: [PATCH 3/5] zram: use zram_slot_lock instead of raw
 bit_spin_lock op

Hi Sergey,

On Mon, Apr 03, 2017 at 03:08:58PM +0900, Sergey Senozhatsky wrote:
> Hello Minchan,
> 
> On (04/03/17 14:17), Minchan Kim wrote:
> > With this clean-up phase, I want to use zram's wrapper function
> > to lock table access which is more consistent with other zram's
> > functions.
> 
> which reminds me of...
> 
> there was a discussion a long time ago, -rt people absolutely
> hate bit spin_locks and they suggested us to replace it with
> normal spin_locks (and I promised to take a look at it, but
> got interrupted and never really returned back to it).
> 
> for !lockdep builds the impact is somewhat small; for lockdep
> builds we increase the memory usage, but
> 
> a) lockdep builds are debug builds by definition, no one runs lockdep
>    enabled kernels in production
> 
> b) we have lockdep in zram now, which is nice

It's really one I want to have.

> 
> c) spin_locks probably have better fairness guarantees

In fact, it wouldn't be an imporant because zram's slot lock contention
is not heavy.
> 
> 
> what do you think? can we, in this patch set, also replce bit
> spin_locks with a normal spin_lock?

With changing only zram side from bit_spin_lock to spin_lock,
it would be crippled. I mean zsmalloc should be changed, too
and it's really hard. :(