From: "Schumaker, Anna" <Anna.Schumaker@netapp.com>
To: "trondmy@gmail.com" <trondmy@gmail.com>,
"bcodding@redhat.com" <bcodding@redhat.com>
Cc: "linux-nfs@vger.kernel.org" <linux-nfs@vger.kernel.org>,
"jamespharvey20@gmail.com" <jamespharvey20@gmail.com>
Subject: Re: [PATCH v2] SUNRPC: fix race to sk_err after xs_error_report
Date: Wed, 2 Oct 2019 14:12:42 +0000 [thread overview]
Message-ID: <4aad6fda0afd63e6020dfe843004fa73edf665a5.camel@netapp.com> (raw)
In-Reply-To: <d851f5c2e3aa2b1784a0cc7d42858781b6d07ab8.1570014121.git.bcodding@redhat.com>
Hi Ben,
On Wed, 2019-10-02 at 07:03 -0400, Benjamin Coddington wrote:
> Since commit 4f8943f80883 ("SUNRPC: Replace direct task wakeups from
> softirq context") there has been a race to the value of the sk_err if both
> XPRT_SOCK_WAKE_ERROR and XPRT_SOCK_WAKE_DISCONNECT are set. In that case,
> we may end up losing the sk_err value that existed when xs_error_report was
> called.
>
> Fix this by reverting to the previous behavior: instead of using SO_ERROR
> to retrieve the value at a later time (which might also return sk_err_soft),
> copy the sk_err value onto struct sock_xprt, and use that value to wake
> pending tasks.
>
> Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
> Fixes: 4f8943f80883 ("SUNRPC: Replace direct task wakeups from softirq context")
> 8<-----------------------------------------------------------------------------
>
> Changes on V2:
> - move xprt_err into a hole in struct sock_xprt
> - add an memory barrier to ensure the error is visble to the
> error_worker
>
> ---
> include/linux/sunrpc/xprtsock.h | 1 +
> net/sunrpc/xprtsock.c | 16 ++++++++--------
> 2 files changed, 9 insertions(+), 8 deletions(-)
>
> diff --git a/include/linux/sunrpc/xprtsock.h b/include/linux/sunrpc/xprtsock.h
> index 7638dbe7bc50..a940de03808d 100644
> --- a/include/linux/sunrpc/xprtsock.h
> +++ b/include/linux/sunrpc/xprtsock.h
> @@ -61,6 +61,7 @@ struct sock_xprt {
> struct mutex recv_mutex;
> struct sockaddr_storage srcaddr;
> unsigned short srcport;
> + int xprt_err;
>
> /*
> * UDP socket buffer size parameters
> diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
> index e2176c167a57..25ff097d6e21 100644
> --- a/net/sunrpc/xprtsock.c
> +++ b/net/sunrpc/xprtsock.c
> @@ -1250,12 +1250,15 @@ static void xs_error_report(struct sock *sk)
> goto out;
>
> transport = container_of(xprt, struct sock_xprt, xprt);
> - err = -sk->sk_err;
It looks like "err" is unused after this change, so can you remove the variable declaration as
well?
Thanks,
Anna
> - if (err == 0)
> + transport->xprt_err = -sk->sk_err;
> + if (transport->xprt_err == 0)
> goto out;
> dprintk("RPC: xs_error_report client %p, error=%d...\n",
> - xprt, -err);
> - trace_rpc_socket_error(xprt, sk->sk_socket, err);
> + xprt, -transport->xprt_err);
> + trace_rpc_socket_error(xprt, sk->sk_socket, transport->xprt_err);
> +
> + /* barrier ensures xprt_err is set before XPRT_SOCK_WAKE_ERROR */
> + smp_mb__before_atomic();
> xs_run_error_worker(transport, XPRT_SOCK_WAKE_ERROR);
> out:
> read_unlock_bh(&sk->sk_callback_lock);
> @@ -2470,7 +2473,6 @@ static void xs_wake_write(struct sock_xprt *transport)
> static void xs_wake_error(struct sock_xprt *transport)
> {
> int sockerr;
> - int sockerr_len = sizeof(sockerr);
>
> if (!test_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
> return;
> @@ -2479,9 +2481,7 @@ static void xs_wake_error(struct sock_xprt *transport)
> goto out;
> if (!test_and_clear_bit(XPRT_SOCK_WAKE_ERROR, &transport->sock_state))
> goto out;
> - if (kernel_getsockopt(transport->sock, SOL_SOCKET, SO_ERROR,
> - (char *)&sockerr, &sockerr_len) != 0)
> - goto out;
> + sockerr = xchg(&transport->xprt_err, 0);
> if (sockerr < 0)
> xprt_wake_pending_tasks(&transport->xprt, sockerr);
> out:
prev parent reply other threads:[~2019-10-02 14:12 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-02 11:03 [PATCH v2] SUNRPC: fix race to sk_err after xs_error_report Benjamin Coddington
2019-10-02 14:12 ` Schumaker, Anna [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=4aad6fda0afd63e6020dfe843004fa73edf665a5.camel@netapp.com \
--to=anna.schumaker@netapp.com \
--cc=bcodding@redhat.com \
--cc=jamespharvey20@gmail.com \
--cc=linux-nfs@vger.kernel.org \
--cc=trondmy@gmail.com \
/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).