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;
MIME-Version: 1.0
References: <20210508122530.1971-1-justin.he@arm.com> <20210508122530.1971-2-justin.he@arm.com>
In-Reply-To: <20210508122530.1971-2-justin.he@arm.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sat, 8 May 2021 08:30:43 -0700
Message-ID: <CAHk-=wgSFUUWJKW1DXa67A0DXVzQ+OATwnC3FCwhqfTJZsvj1A@mail.gmail.com>
Subject: Re: [PATCH RFC 1/3] fs: introduce helper d_path_fast()
To:     Jia He <justin.he@arm.com>
Cc:     Petr Mladek <pmladek@suse.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Sergey Senozhatsky <senozhatsky@chromium.org>,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Rasmus Villemoes <linux@rasmusvillemoes.dk>,
        Jonathan Corbet <corbet@lwn.net>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Al Viro <viro@ftp.linux.org.uk>,
        Heiko Carstens <hca@linux.ibm.com>,
        Vasily Gorbik <gor@linux.ibm.com>,
        Christian Borntraeger <borntraeger@de.ibm.com>,
        "Eric W . Biederman" <ebiederm@xmission.com>,
        "Darrick J. Wong" <darrick.wong@oracle.com>,
        "Peter Zijlstra (Intel)" <peterz@infradead.org>,
        Ira Weiny <ira.weiny@intel.com>,
        Eric Biggers <ebiggers@google.com>,
        "Ahmed S. Darwish" <a.darwish@linutronix.de>,
        "open list:DOCUMENTATION" <linux-doc@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        linux-s390 <linux-s390@vger.kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Sat, May 8, 2021 at 5:29 AM Jia He <justin.he@arm.com> wrote:
>
> This helper is similar to d_path except for not taking seqlock/spinlock.

I see why you did it that way, but conditional locking is something we
really really try to avoid in the kernel.

It basically makes a lot of static tools unable to follow the locking
rules, and it makes it hard for people to se what's going on too.

So instead of passing a "bool need_lock" thing down, the way to do
these things is generally to extract the code inside the locked region
into a helper function of its own, and then you have

  __unlocked_version(...)
  {
       .. do the actual work
  }

  locked_version(..)
  {
      take_lock(..)
      retval = __unlocked_version(..);
      release_lock(..);
      return retval;
  }

this prepend_path() case is a bit more complicated because there's two
layers of locking, but I think the pattern should still work fine.

In fact, I think it would clean up prepend_path() and make it more
legible to have the two layers of mount_lock / rename_lock be done in
callers with the restarting being done as a loop in the caller rather
than as "goto restart_*".

              Linus