Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220214210708.GA2167841@xavier-xps> <CAJfpegvVKWHhhXwOi9jDUOJi2BnYSDxZQrp1_RRrpVjjZ3Rs2w@mail.gmail.com>
 <YguspMvu6M6NJ1hL@zeniv-ca.linux.org.uk> <YgvPbljmJXsR7ESt@zeniv-ca.linux.org.uk>
 <YgvSB6CKAhF5IXFj@casper.infradead.org> <YgvS1XOJMn5CjQyw@zeniv-ca.linux.org.uk>
 <CAJfpegv03YpTPiDnLwbaewQX_KZws5nutays+vso2BVJ1v1+TA@mail.gmail.com>
 <YgzRwhavapo69CAn@miu.piliscsaba.redhat.com> <20220216131814.GA2463301@xavier-xps>
 <CAJfpegsQO-35p6uoG2ZfuCOLPFwnkbTcLc3K8r+HiS2un9au_w@mail.gmail.com>
In-Reply-To: <CAJfpegsQO-35p6uoG2ZfuCOLPFwnkbTcLc3K8r+HiS2un9au_w@mail.gmail.com>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Fri, 18 Feb 2022 16:37:48 +0100
Message-ID: <CAJfpeguFUB_eqdE=CETE+mBhp0ZmeouoFeEs7=O5UibCmPi=CQ@mail.gmail.com>
Subject: Re: race between vfs_rename and do_linkat (mv and link)
To:     Xavier Roche <xavier.roche@algolia.com>
Cc:     Al Viro <viro@zeniv.linux.org.uk>,
        Matthew Wilcox <willy@infradead.org>,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, 16 Feb 2022 at 14:37, Miklos Szeredi <miklos@szeredi.hu> wrote:

> One issue with the patch is nesting of lock_rename() calls in stacked
> fs (rwsem is not allowed to recurse even for read locks).

Scratch that, there's no recursion since do_linkat() is not called
from stacked fs.  And taking link_rwsem is optional for the link
operation, so this is fine.  For stacked fs the race is hopefully
taken care by the top layer locking, no need to repeat it in lower
layers.

I've now sent this patch with a proper header comment to Al.

Thanks,
Miklos