From: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
To: Doug Anderson <dianders@chromium.org>
Cc: David Miller <davem@davemloft.net>,
Jakub Kicinski <kuba@kernel.org>,
Bjorn Andersson <bjorn.andersson@linaro.org>,
netdev <netdev@vger.kernel.org>,
LKML <linux-kernel@vger.kernel.org>,
Alex Elder <elder@linaro.org>
Subject: Re: [PATCH] net: qrtr: ns: Fix the incorrect usage of rcu_read_lock()
Date: Fri, 2 Oct 2020 21:34:02 +0530 [thread overview]
Message-ID: <20201002160401.GB7178@Mani-XPS-13-9360> (raw)
In-Reply-To: <CAD=FV=V3WHgE5hqgRtPayGB1PTcdJge-32wJOgs84=4h3owtsA@mail.gmail.com>
Hi Doug,
On Fri, Oct 02, 2020 at 08:28:51AM -0700, Doug Anderson wrote:
> Hi,
>
> On Fri, Oct 2, 2020 at 7:15 AM Manivannan Sadhasivam
> <manivannan.sadhasivam@linaro.org> wrote:
> >
> > The rcu_read_lock() is not supposed to lock the kernel_sendmsg() API
> > since it has the lock_sock() in qrtr_sendmsg() which will sleep. Hence,
> > fix it by excluding the locking for kernel_sendmsg().
> >
> > Fixes: a7809ff90ce6 ("net: qrtr: ns: Protect radix_tree_deref_slot() using rcu read locks")
> > Reported-by: Doug Anderson <dianders@chromium.org>
> > Tested-by: Alex Elder <elder@linaro.org>
> > Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
> > ---
> > net/qrtr/ns.c | 20 ++++++++++++++------
> > 1 file changed, 14 insertions(+), 6 deletions(-)
> >
> > diff --git a/net/qrtr/ns.c b/net/qrtr/ns.c
> > index 934999b56d60..0515433de922 100644
> > --- a/net/qrtr/ns.c
> > +++ b/net/qrtr/ns.c
> > @@ -203,15 +203,17 @@ static int announce_servers(struct sockaddr_qrtr *sq)
> > /* Announce the list of servers registered in this node */
> > radix_tree_for_each_slot(slot, &node->servers, &iter, 0) {
> > srv = radix_tree_deref_slot(slot);
> > + rcu_read_unlock();
>
> My RCU-fu is mediocre at best and my radix-tree knowledge is
> non-existent. However:
>
> => Reading through radix_tree_deref_slot() it says that if you are
> only holding the read lock that you need to be calling
> radix_tree_deref_retry(). Why don't I see that here?
>
Well, I drew inspiration from peer drivers and didn't look into the API
documentation properly, my bad :(
> => Without any real knowledge, it seems super sketchy to drop the lock
> while iterating over the tree. Somehow that feels unsafe. Hrm, there
> seems to be a function radix_tree_iter_resume() that might be exactly
> what you want, but I'm not totally sure. The only user I can see
> in-tree (other than radix tree regression testing) is btrfs-tests.c
> but it's using it together with radix_tree_deref_slot_protected().
>
> In any case, my totally untested and totally knowedge-free proposal
> would look something like this:
>
> rcu_read_lock();
> /* Announce the list of servers registered in this node */
> radix_tree_for_each_slot(slot, &node->servers, &iter, 0) {
> srv = radix_tree_deref_slot(slot);
> if (!srv)
> continue;
> if (radix_tree_deref_retry(srv)) {
> slot = radix_tree_iter_retry(&iter);
> continue;
> }
> slot = radix_tree_iter_resume(slot, &iter);
> rcu_read_unlock();
>
> ret = service_announce_new(sq, srv);
> if (ret < 0) {
> pr_err("failed to announce new service\n");
> return ret;
> }
>
> rcu_read_lock();
> }
>
> rcu_read_unlock();
>
> What a beast! Given that this doesn't seem to be what anyone else in
> the kernel is doing exactly, it makes me suspect that there's a more
> fundamental design issue here, though...
>
That's how it is supposed to be. So I'm going to roll out next revision with
your suggestion for the rest of the deref_slot() calls also.
Thanks for your time looking into this.
Regards,
Mani
> -Doug
prev parent reply other threads:[~2020-10-02 16:04 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-10-02 14:15 [PATCH] net: qrtr: ns: Fix the incorrect usage of rcu_read_lock() Manivannan Sadhasivam
2020-10-02 15:28 ` Doug Anderson
2020-10-02 16:04 ` Manivannan Sadhasivam [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20201002160401.GB7178@Mani-XPS-13-9360 \
--to=manivannan.sadhasivam@linaro.org \
--cc=bjorn.andersson@linaro.org \
--cc=davem@davemloft.net \
--cc=dianders@chromium.org \
--cc=elder@linaro.org \
--cc=kuba@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).