linux-kernel - Re: [PATCH 15/17] fs/ext4: use cached rbtrees

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

Message-ID: <20170719225039.GC14395@linux-80c1.suse>
Date:   Wed, 19 Jul 2017 15:50:39 -0700
From:   Davidlohr Bueso <dave@...olabs.net>
To:     Jan Kara <jack@...e.cz>
Cc:     akpm@...ux-foundation.org, mingo@...nel.org, peterz@...radead.org,
        torvalds@...ux-foundation.org, kirill.shutemov@...ux.intel.com,
        hch@...radead.org, ldufour@...ux.vnet.ibm.com, mhocko@...e.com,
        mgorman@...hsingularity.net, linux-kernel@...r.kernel.org,
        tytso@....edu, Davidlohr Bueso <dbueso@...e.de>
Subject: Re: [PATCH 15/17] fs/ext4: use cached rbtrees

On Wed, 19 Jul 2017, Jan Kara wrote:

>Hum, so I'm somewhat undecided whether this is worth the churn. For free
>blocks rb_tree we use rb_first() only in ext4_mb_generate_from_freelist()
>which gets called only when generating new buddy bitmap from on-disk bitmap
>and we traverse the whole tree after that - thus the extra cost of
>rb_first() is a) well hidden in the total cost of iteration, b) rather rare
>anyway.

Yes, during some of the testing I did for this patch, I could not find trees
with more than just a handful of nodes, ie: while removing file(s). 

>
>Similarly for the H-tree directory code, we call rb_first() in
>ext4_dx_readdir() only to start an iteration over the whole B-tree and in
>such case I don't think optimizing rb_first() makes a big difference
>(maintaining cached value is going to have about the same cost as we save
>by using it).

Your reaction is rather expected, actually. Lets just leave this patch out.

Thanks,
Davidlohr