Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH 1/2] fs, close_range: add flag CLOSE_RANGE_CLOEXEC
To:     Al Viro <viro@zeniv.linux.org.uk>,
        Giuseppe Scrivano <gscrivan@redhat.com>
Cc:     linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        christian.brauner@ubuntu.com, containers@lists.linux-foundation.org
References: <20201013140609.2269319-1-gscrivan@redhat.com>
 <20201013140609.2269319-2-gscrivan@redhat.com>
 <20201013210925.GJ3576660@ZenIV.linux.org.uk>
From:   Rasmus Villemoes <linux@rasmusvillemoes.dk>
Message-ID: <70fa4f70-38cc-7e18-8156-65a3e50c641e@rasmusvillemoes.dk>
Date:   Tue, 13 Oct 2020 23:49:34 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.10.0
MIME-Version: 1.0
In-Reply-To: <20201013210925.GJ3576660@ZenIV.linux.org.uk>
Content-Type: text/plain; charset=windows-1252
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 13/10/2020 23.09, Al Viro wrote:
> On Tue, Oct 13, 2020 at 04:06:08PM +0200, Giuseppe Scrivano wrote:
>> +		spin_lock(&cur_fds->file_lock);
>> +		fdt = files_fdtable(cur_fds);
>> +		cur_max = fdt->max_fds - 1;
>> +		max_fd = min(max_fd, cur_max);
>> +		while (fd <= max_fd)
>> +			__set_close_on_exec(fd++, fdt);
>> +		spin_unlock(&cur_fds->file_lock);
> 
> 	First of all, this is an atrocious way to set all bits
> in a range.  What's more, you don't want to set it for *all*
> bits - only for the ones present in open bitmap.  It's probably
> harmless at the moment, but let's not create interesting surprises
> for the future.

Eh, why not? They can already be set for unallocated slots:

commit 5297908270549b734c7c2556745e2385b6d4941d
Author: Mateusz Guzik <mguzik@redhat.com>
Date:   Tue Oct 3 12:58:14 2017 +0200

    vfs: stop clearing close on exec when closing a fd

    Codepaths allocating a fd always make sure the bit is set/unset
    depending on flags, thus clearing on close is redundant.

And while we're on that subject, yours truly suggested exactly that two
years prior [1], with a follow-up [2] in 2018 to do what wasn't done in
5297908, but (still) seems like obvious micro-optimizations, given that
the close_on_exec bitmap is not maintained as a subset of the open
bitmap. Mind taking a look at [2]?

[1]
https://lore.kernel.org/lkml/1446543679-28849-1-git-send-email-linux@rasmusvillemoes.dk/t/#u
[2]
https://lore.kernel.org/lkml/20181024160159.25884-1-linux@rasmusvillemoes.dk/

Rasmus