From: Tariq Saeed <tariq.x.saeed@oracle.com>
To: ocfs2-devel@oss.oracle.com
Subject: [Ocfs2-devel] [patch 03/11] ocfs2/o2net: incorrect to terminate accepting connections loop upon rejecting an invalid one
Date: Mon, 07 Apr 2014 20:07:41 -0700 [thread overview]
Message-ID: <534367FD.8060403@oracle.com> (raw)
In-Reply-To: <20140319140143.b3a0fa14b49284fb79c1bd61@linux-foundation.org>
On 03/19/2014 02:01 PM, Andrew Morton wrote:
> On Fri, 24 Jan 2014 14:22:57 -0800 tariq saeed <tariq.x.saeed@oracle.com> wrote:
>
>> On 1/24/2014 1:55 PM, Mark Fasheh wrote:
>>> On Fri, Jan 24, 2014 at 12:47:02PM -0800, akpm at linux-foundation.org wrote:
>>>> From: Tariq Saeed <tariq.x.saeed@oracle.com>
>>>> Subject: ocfs2/o2net: incorrect to terminate accepting connections loop upon rejecting an invalid one
>>>>
>>>> When o2net-accept-one() rejects an illegal connection, it terminates the
>>>> loop picking up the remaining queued connections. This fix will continue
>>>> accepting connections till the queue is emtpy.
>>>>
>>>> Addresses Orabug 17489469.
>>> Thanks for sending this, review comments below.
>>>
>>>
>>>> diff -puN fs/ocfs2/cluster/tcp.c~ocfs2-o2net-incorrect-to-terminate-accepting-connections-loop-upon-rejecting-an-invalid-one fs/ocfs2/cluster/tcp.c
>>>> --- a/fs/ocfs2/cluster/tcp.c~ocfs2-o2net-incorrect-to-terminate-accepting-connections-loop-upon-rejecting-an-invalid-one
>>>> +++ a/fs/ocfs2/cluster/tcp.c
>>>> @@ -1826,7 +1826,7 @@ int o2net_register_hb_callbacks(void)
>>>>
>>>> /* ------------------------------------------------------------ */
>>>>
>>>> -static int o2net_accept_one(struct socket *sock)
>>>> +static int o2net_accept_one(struct socket *sock, int *more)
>>>> {
>>>> int ret, slen;
>>>> struct sockaddr_in sin;
>>>> @@ -1837,6 +1837,7 @@ static int o2net_accept_one(struct socke
>>>> struct o2net_node *nn;
>>>>
>>>> BUG_ON(sock == NULL);
>>>> + *more = 0;
>>>> ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type,
>>>> sock->sk->sk_protocol, &new_sock);
>>>> if (ret)
>>>> @@ -1848,6 +1849,7 @@ static int o2net_accept_one(struct socke
>>>> if (ret < 0)
>>>> goto out;
>>>>
>>>> + *more = 1;
>>>> new_sock->sk->sk_allocation = GFP_ATOMIC;
>>>>
>>>> ret = o2net_set_nodelay(new_sock);
>>>> @@ -1949,8 +1951,15 @@ out:
>>>> static void o2net_accept_many(struct work_struct *work)
>>>> {
>>>> struct socket *sock = o2net_listen_sock;
>>>> - while (o2net_accept_one(sock) == 0)
>>>> + int more;
>>>> + int err;
>>>> +
>>>> + for (;;) {
>>>> + err = o2net_accept_one(sock, &more);
>>>> + if (!more)
>>>> + break;
>>> We're throwing out 'err' here and trusting the variable 'more'. However, err
>>> could be set and more would be 0 regardless of whether there actually are
>>> more connections to be had. This makes more sense given when 'more' is set:
>>
>> Thanks for the comments.
>> To understand the consequences of ignoring the err, we need to look at
>> what is going on.
>> We get a softIRQ when a connection packet (tcp SYN). It is critical to
>> note that we may not
>> get a softIRQ_for every connection s_ince connection packets can arrive
>> back-to-back (as happened in this bug). So, one softIRQ could be
>> delivered for > 1 pending accept.
>> _This is the KEY point. _
>>
>> If we terminate the loop calling o2net_accept_one() upon seeing an
>> error, what happens
>> to the rest of the connections in the queue. If no new connection
>> arrives for hours, no new softIRQ
>> will be delivered, and the connections will just sit in the queue.
>
> Please note that I had to edit your email to undo the top-posting so I
> could reply to it. Please don't top-post.
>
> Mark, are you now OK with the patch as-is?
Mark, do you have further questios?
>
>
> From: Tariq Saeed <tariq.x.saeed@oracle.com>
> Subject: ocfs2/o2net: incorrect to terminate accepting connections loop upon rejecting an invalid one
>
> When o2net-accept-one() rejects an illegal connection, it terminates the
> loop picking up the remaining queued connections. This fix will continue
> accepting connections till the queue is emtpy.
>
> Addresses Orabug 17489469.
>
> Signed-off-by: Tariq Saseed <tariq.x.saeed@oracle.com>
> Cc: Mark Fasheh <mfasheh@suse.com>
> Cc: Joel Becker <jlbec@evilplan.org>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
>
> fs/ocfs2/cluster/tcp.c | 13 +++++++++++--
> 1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff -puN fs/ocfs2/cluster/tcp.c~ocfs2-o2net-incorrect-to-terminate-accepting-connections-loop-upon-rejecting-an-invalid-one fs/ocfs2/cluster/tcp.c
> --- a/fs/ocfs2/cluster/tcp.c~ocfs2-o2net-incorrect-to-terminate-accepting-connections-loop-upon-rejecting-an-invalid-one
> +++ a/fs/ocfs2/cluster/tcp.c
> @@ -1826,7 +1826,7 @@ int o2net_register_hb_callbacks(void)
>
> /* ------------------------------------------------------------ */
>
> -static int o2net_accept_one(struct socket *sock)
> +static int o2net_accept_one(struct socket *sock, int *more)
> {
> int ret, slen;
> struct sockaddr_in sin;
> @@ -1837,6 +1837,7 @@ static int o2net_accept_one(struct socke
> struct o2net_node *nn;
>
> BUG_ON(sock == NULL);
> + *more = 0;
> ret = sock_create_lite(sock->sk->sk_family, sock->sk->sk_type,
> sock->sk->sk_protocol, &new_sock);
> if (ret)
> @@ -1848,6 +1849,7 @@ static int o2net_accept_one(struct socke
> if (ret < 0)
> goto out;
>
> + *more = 1;
> new_sock->sk->sk_allocation = GFP_ATOMIC;
>
> ret = o2net_set_nodelay(new_sock);
> @@ -1949,8 +1951,15 @@ out:
> static void o2net_accept_many(struct work_struct *work)
> {
> struct socket *sock = o2net_listen_sock;
> - while (o2net_accept_one(sock) == 0)
> + int more;
> + int err;
> +
> + for (;;) {
> + err = o2net_accept_one(sock, &more);
> + if (!more)
> + break;
> cond_resched();
> + }
> }
>
> static void o2net_listen_data_ready(struct sock *sk, int bytes)
> _
>
next prev parent reply other threads:[~2014-04-08 3:07 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-01-24 20:47 [Ocfs2-devel] [patch 03/11] ocfs2/o2net: incorrect to terminate accepting connections loop upon rejecting an invalid one akpm at linux-foundation.org
2014-01-24 21:55 ` Mark Fasheh
2014-01-24 22:15 ` Srinivas Eeda
2014-01-24 22:22 ` tariq saeed
2014-03-19 21:01 ` Andrew Morton
2014-04-08 3:07 ` Tariq Saeed [this message]
2014-04-08 19:42 ` Mark Fasheh
2014-04-08 22:39 ` Tariq Saeed
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=534367FD.8060403@oracle.com \
--to=tariq.x.saeed@oracle.com \
--cc=ocfs2-devel@oss.oracle.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).