linux-kernel - Re: [PATCH] maple_tree: do not preallocate nodes for slot stores

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: <20231212170048.7c0afab9c0522c7de6067cd7@linux-foundation.org>
Date:   Tue, 12 Dec 2023 17:00:48 -0800
From:   Andrew Morton <akpm@...ux-foundation.org>
To:     Matthew Wilcox <willy@...radead.org>
Cc:     Sidhartha Kumar <sidhartha.kumar@...cle.com>,
        linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        maple-tree@...ts.infradead.org, liam.howlett@...cle.com,
        zhangpeng.00@...edance.com
Subject: Re: [PATCH] maple_tree: do not preallocate nodes for slot stores

On Tue, 12 Dec 2023 20:57:48 +0000 Matthew Wilcox <willy@...radead.org> wrote:

> On Tue, Dec 12, 2023 at 11:46:40AM -0800, Sidhartha Kumar wrote:
> > +	/* Slot store, does not require additional nodes */
> > +	if ((node_size == mas->end) && ((!mt_in_rcu(mas->tree))
> > +		|| (wr_mas.offset_end - mas->offset == 1)))
> > +		return 0;
> 
> Should we refactor this into a mas_is_slot_store() predicate?
> 
> A few coding-style problems with it as it's currently written:
> 
> 1. The indentation on the second line is wrong.  It makes the
> continuation of the condition look like part of the statement.  Use
> extra whitespace to indent.  eg:
> 
> 	if ((node_size == mas->end) && ((!mt_in_rcu(mas->tree))
> 			|| (wr_mas.offset_end - mas->offset == 1)))
> 		return 0;
> 
> 2. The operator goes last on the line, not at the beginning of the
> continuation line.  ie:
> 
> 	if ((node_size == mas->end) && ((!mt_in_rcu(mas->tree)) ||
> 			(wr_mas.offset_end - mas->offset == 1)))
> 		return 0;
> 
> 3. You don't need parens around the !mt_in_rcu(mas->tree).  There's
> no ambiguity to solve here:
> 
> 	if ((node_size == mas->end) && (!mt_in_rcu(mas->tree) ||
> 			(wr_mas.offset_end - mas->offset == 1)))
> 		return 0;
> 
> But I'd write it as:
> 
> 	if ((node_size == mas->end) &&
> 	    (!mt_in_rcu(mas->tree) || (wr_mas.offset_end - mas->offset == 1)))
> 		return 0;
> 
> because then the whitespace matches how you're supposed to parse the
> condition, and so the next person to read this code will have an easier
> time of it.

Yup.  But I'd suggest going further:

	/* Slot store, does not require additional nodes */
	if (node_size == mas->end) {
		/* comment goes here */
		if (!mt_in_rcu(mas->tree))
			return 0;
		/* and here too */
		if (wr_mas.offset_end - mas->offset == 1)
			return 0;
	}

ie: create space to add those comments explaining the reason for each test.