[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20180813081120.h47tvkg6vhxc42jg@localhost.localdomain>
Date: Mon, 13 Aug 2018 10:11:20 +0200
From: Lukas Czerner <lczerner@...hat.com>
To: "Theodore Y. Ts'o" <tytso@....edu>
Cc: linux-ext4@...r.kernel.org
Subject: Re: [2/2] e2fsck: remove unnecessary reserve_stdio_fds()
On Sat, Aug 11, 2018 at 08:47:43PM -0400, Theodore Y. Ts'o wrote:
> On Thu, Aug 09, 2018 at 10:35:46AM +0200, Lukas Czerner wrote:
> > Standard stream are always open, we do not need to "check" it
> > specifically. Remove reserve_stdio_fds().
>
> I don't remember the exact circumstances, since it was twenty years
> ago, but that's not *always* true. Remember that e2fsck is run by
> init or some other early boot sequence, and not all such programs
> are.... sane. In fact, this was added because some buggy init called
> e2fsck with fd 2 closed, so the file system was opened using fd 2.
> And then an error message ended up corrupting the file system....
Uff, even though we do not really want to optimize for a broken code, it
does make sense to me to protect user from possible data loss in thi
crazy case.
>
> This function was added in response to that failure, because
> sometimes, the world *is* out to get you. It is fair to fix the file
> descriptor leak, so how about this?
Fair enough.
Reviewed-by: Lukas Czerner <lczerner@...hat.com>
Thanks!
-Lukas
>
> - Ted
>
>
> commit 352701d9e2258299fe3bffb1d112566c0e4a7cdf
> Author: Theodore Ts'o <tytso@....edu>
> Date: Sat Aug 11 20:47:08 2018 -0400
>
> e2fsck: fix fd leak in reserve_stdio_fds
>
> Signed-off-by: Theodore Ts'o <tytso@....edu>
>
> diff --git a/e2fsck/unix.c b/e2fsck/unix.c
> index 90065b395..2df22b171 100644
> --- a/e2fsck/unix.c
> +++ b/e2fsck/unix.c
> @@ -617,9 +617,10 @@ static void reserve_stdio_fds(void)
> fprintf(stderr, _("ERROR: Couldn't open "
> "/dev/null (%s)\n"),
> strerror(errno));
> - break;
> + return;
> }
> }
> + (void) close(fd);
> }
>
> #ifdef HAVE_SIGNAL_H
Powered by blists - more mailing lists