linux-kernel - Re: Tagged pointers in the XArray

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]

Message-ID: <20180829161756.GB30396@bombadil.infradead.org>
Date:   Wed, 29 Aug 2018 09:17:56 -0700
From:   Matthew Wilcox <willy@...radead.org>
To:     linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        linux-fsdevel@...r.kernel.org
Cc:     Gao Xiang <gaoxiang25@...wei.com>,
        zhong jiang <zhongjiang@...wei.com>,
        Chao Yu <yuchao0@...wei.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Subject: Re: Tagged pointers in the XArray

On Tue, Aug 28, 2018 at 03:27:27PM -0700, Matthew Wilcox wrote:
> I find myself caught between two traditions.
> 
> On the one hand, the radix tree has been calling the page cache dirty &
> writeback bits "tags" for over a decade.
> 
> On the other hand, using some of the bits _in a pointer_ as a tag has been
> common practice since at least the 1960s.
> https://en.wikipedia.org/wiki/Tagged_pointer and
> https://en.wikipedia.org/wiki/31-bit
> 
> EROFS wants to use tagged pointers in the radix tree / xarray.  Right now,
> they're building them by hand, which is predictably grotty-looking.
> I think it's reasonable to provide this functionality as part of the
> XArray API, _but_ it's confusing to have two different things called tags.
> 
> I've done my best to document my way around this, but if we want to rename
> the things that the radix tree called tags to avoid the problem entirely,
> now is the time to do it.  Anybody got a Good Idea?

I have two ideas now.

First, we could rename radix tree tags to xarray marks.  That is,

xa_mark_t
xa_set_mark()
xas_clear_mark()
xas_for_each_marked() { }
xa_marked()
etc

Second, we could call the tagged pointers typed pointers.  That is,

void *xa_mk_type(void *p, unsigned int type);
void *xa_to_ptr(void *entry);
int xa_ptr_type(void *entry);

Any better ideas, or violent revulsion to either of the above ideas?