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: <200710172249.13877.nickpiggin@yahoo.com.au>
Date:	Wed, 17 Oct 2007 22:49:13 +1000
From:	Nick Piggin <nickpiggin@...oo.com.au>
To:	"Eric W. Biederman" <ebiederm@...ssion.com>
Cc:	Andrew Morton <akpm@...ux-foundation.org>,
	Christian Borntraeger <borntraeger@...ibm.com>,
	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	Martin Schwidefsky <schwidefsky@...ibm.com>,
	"Theodore Ts'o" <tytso@....edu>
Subject: Re: [patch][rfc] rewrite ramdisk

On Wednesday 17 October 2007 20:30, Eric W. Biederman wrote:
> Nick Piggin <nickpiggin@...oo.com.au> writes:
> > On Tuesday 16 October 2007 18:08, Nick Piggin wrote:
> >> On Tuesday 16 October 2007 14:57, Eric W. Biederman wrote:
> >> > > What magic restrictions on page allocations? Actually we have
> >> > > fewer restrictions on page allocations because we can use
> >> > > highmem!
> >> >
> >> > With the proposed rewrite yes.
> >
> > Here's a quick first hack...
> >
> > Comments?
>
> I have beaten my version of this into working shape, and things
> seem ok.
>
> However I'm beginning to think that the real solution is to remove
> the dependence on buffer heads for caching the disk mapping for
> data pages, and move the metadata buffer heads off of the block
> device page cache pages.  Although I am just a touch concerned
> there may be an issue with using filesystem tools while the
> filesystem is mounted if I move the metadata buffer heads.
>
> If we were to move the metadata buffer heads (assuming I haven't
> missed some weird dependency) then I think there a bunch of
> weird corner cases that would be simplified.

I'd prefer just to go along the lines of what I posted. It's
a pure block device driver which knows absolutely nothing about
vm or vfs.

What are you guys using rd for, and is it really important to
have this supposed buffercache optimisation?


> I guess that is where I look next.
>
> Oh for what it is worth I took a quick look at fsblock and I don't think
> struct fsblock makes much sense as a block mapping translation layer for
> the data path where the current page caches works well.

Except the page cache doesn't store any such translation. fsblock
works very nicely as a buffer_head and nobh-mode replacement there,
but we could probably go one better (for many filesystems) by using
a tree.


> For less 
> then the cost of 1 fsblock I can cache all of the translations for a
> 1K filesystem on a 4K page.

I'm not exactly sure what you mean, unless you're talking about an
extent based data structure. fsblock is fairly slim as far as a
generic solution goes. You could probably save 4 or 8 bytes in it,
but I don't know if it's worth the trouble.


> I haven't looked to see if fsblock makes sense to use as a buffer head
> replacement yet.

Well I don't want to get too far off topic on the subject of fsblock,
and block mappings (because I think the ramdisk driver simply should
have nothing to do with any of that)... but fsblock is exactly a
buffer head replacement (so if it doesn't make sense, then I've
screwed something up badly! ;))
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ