From: Yonglong Li <liyonglong@chinatelecom.cn>
To: Geliang Tang <geliangtang@gmail.com>
Cc: mptcp@lists.linux.dev
Subject: Re: [MPTCP][PATCH mptcp-next] Squash to "mptcp: build ADD_ADDR/echo-ADD_ADDR option according pm.add_signal"
Date: Mon, 12 Jul 2021 17:21:47 +0800 [thread overview]
Message-ID: <8b424320-ae9e-18bb-2011-cb7562497f13@chinatelecom.cn> (raw)
In-Reply-To: <CA+WQbwsd_3AqgwawQu_6socwuviNXMrck-BxjMJ+PU-D9QVX9g@mail.gmail.com>
On 2021/7/12 17:07, Geliang Tang wrote:
> Geliang Tang <geliangtang@gmail.com> 于2021年7月12日周一 下午4:44写道:
>>
>> Yonglong Li <liyonglong@chinatelecom.cn> 于2021年7月12日周一 下午4:07写道:
>>>
>>>
>>>
>>> On 2021/7/12 15:33, Geliang Tang wrote:
>>>> Hi Yonglong,
>>>>
>>>> Yonglong Li <liyonglong@chinatelecom.cn> 于2021年7月12日周一 上午9:34写道:
>>>>>
>>>>>
>>>>>
>>>>> On 2021/7/11 23:15, Geliang Tang wrote:
>>>>>> I think there're still some issues in v8:
>>>>>>
>>>>>> The remaining value is incorrect since "remaining += opt_size;" in the
>>>>>> "drop other suboptions" checks has been called twice in
>>>>>> mptcp_pm_add_addr_signal and mptcp_established_options_add_addr.
>>>>>>
>>>>> I think "remaining" in mptcp_pm_add_addr_signal does not touch "remaining" in
>>>>> mptcp_established_options_add_addr.
>>>>>
>>>>>> opts->local and opts->remote in mptcp_pm_add_addr_signal need be
>>>>>> populate after the length chech, not before the check.]
>>>>>>
>>>>>> The squash-to patch keeped the more orignal code unchanged, and just do
>>>>>> the least, necessary modifications.
>>>>>>
>>>>> Agree opts->local and opts->remote should be asigned after the length check.
>>>>> But if keep the length check out of mptcp_pm_add_addr_signal (out of pm lock )
>>>>> as orignal code, there is a race that:
>>>>>
>>>>> ==> a add addr event (pm.addr_signal == MPTCP_ADD_ADDR_SIGNAL)
>>>>> ==> call mptcp_pm_add_addr_signal
>>>>> ==> a echo add addr event trigger (pm.addr_signal == MPTCP_ADD_ADDR_ECHO|MPTCP_ADD_ADDR_SIGNAL)
>>>>> ==> at this time opts->remote is empty and the length is incorrect.
>>>>>
>>>>
>>>> What will happen in v8 when this race occurs? How dose v8 deal with the
>>>> race?
>>> Hi Geliang, thinks for your patience.
>>>
>>> I think v8 doesn't have this issue:
>>> ==> a add addr event (pm.addr_signal == MPTCP_ADD_ADDR_SIGNAL)
>>> ==> call mptcp_pm_add_addr_signal, save pm.addr_signal to add_addr and save addr in opts under pm.lock
>>> ==> a echo add addr event trigger (pm.addr_signal == MPTCP_ADD_ADDR_ECHO), but add_addr doesn't changed.
>>> ==> use add_addr and opts to check length.
>>> ==> next send ack process will deal with MPTCP_ADD_ADDR_ECHO event.
>>
>> Thanks for your explanation.
>>
>> I think this squash-to patch did the same thing:
>>
>> ==> an add addr event (pm.addr_signal == MPTCP_ADD_ADDR_SIGNAL)
>> ==> call mptcp_pm_add_addr_signal, save echo bit of pm.addr_signal to
>> 'echo' (echo = false), save the port number to 'port', and save addr
>> in opts under pm.lock
>> ==> an echo add addr event trigger (pm.addr_signal ==
>> MPTCP_ADD_ADDR_ECHO), but 'echo' and 'port' don't changed.
>> ==> use 'echo' to get the address family, use 'family', 'echo' and
>> 'port' to check length.
>> ==> next send ack process will deal with MPTCP_ADD_ADDR_ECHO event.
>>
>> Do you think so?
>
> And one more thing. How do you test this "mptcp: fix conflicts when using
> pm.add_signal in ADD_ADDR/echo and RM_ADDR process" series? Do you use our
> mptcp_join.sh to test it, or you did some special tests? Does this race
> scenario mentioned above easy to reproduce?
>
> I just did the mptcp_join.sh tests for my squash-to patches, and everything
> looks fine. If you have some special tests, could you please help me to test
> these squash-to patches too? Hope it works in the race scenario.
>
> Thanks.
> -Geliang
>
OK. I will try to test the squash-to patches.
The race scenario is not easy to reproduce. :(
A loop script will try many times to reproduce.
>
>
>>
>>>
>>>>
>>>>> So I think the orignal code is incorrect. WDYT?
>>>>>
>>>>>> Drop the "drop other suboptions" check in mptcp_pm_add_addr_signal.
>>>>>>
>>>>>> Change arguments of mptcp_pm_add_addr_signal.
>>>>>>
>>>>>> Keep mptcp_add_addr_len unchanged.
>>>>>>
>>>>>> Signed-off-by: Geliang Tang <geliangtang@gmail.com>
>>>>>> ---
>>>>>> net/mptcp/options.c | 35 +++++++++++++++++------------------
>>>>>> net/mptcp/pm.c | 23 +++++++++--------------
>>>>>> net/mptcp/protocol.h | 27 +++++++++------------------
>>>>>> 3 files changed, 35 insertions(+), 50 deletions(-)
>>>>>>
>>>>>> diff --git a/net/mptcp/options.c b/net/mptcp/options.c
>>>>>> index 5c0ad9b90866..93ad7b134f74 100644
>>>>>> --- a/net/mptcp/options.c
>>>>>> +++ b/net/mptcp/options.c
>>>>>> @@ -663,16 +663,14 @@ static bool mptcp_established_options_add_addr(struct sock *sk, struct sk_buff *
>>>>>> struct mptcp_sock *msk = mptcp_sk(subflow->conn);
>>>>>> bool drop_other_suboptions = false;
>>>>>> unsigned int opt_size = *size;
>>>>>> - u8 add_addr;
>>>>>> + bool echo;
>>>>>> + bool port;
>>>>>> + u8 family;
>>>>>> int len;
>>>>>>
>>>>>> - if (!mptcp_pm_should_add_signal(msk) ||
>>>>>> - !mptcp_pm_add_addr_signal(msk, skb, opt_size, remaining, opts, &add_addr))
>>>>>> - return false;
>>>>>> -
>>>>>> - if (((add_addr & BIT(MPTCP_ADD_ADDR_ECHO)) ||
>>>>>> - ((add_addr & BIT(MPTCP_ADD_ADDR_SIGNAL)) &&
>>>>>> - (opts->local.family == AF_INET6 || opts->local.port))) &&
>>>>>> + if ((mptcp_pm_should_add_signal_echo(msk) ||
>>>>>> + (mptcp_pm_should_add_signal_addr(msk) &&
>>>>>> + (msk->pm.local.family == AF_INET6 || msk->pm.local.port))) &&
>>>>>> skb && skb_is_tcp_pure_ack(skb)) {
>>>>>> pr_debug("drop other suboptions");
>>>>>> opts->suboptions = 0;
>>>>>> @@ -682,7 +680,12 @@ static bool mptcp_established_options_add_addr(struct sock *sk, struct sk_buff *
>>>>>> drop_other_suboptions = true;
>>>>>> }
>>>>>>
>>>>>> - len = mptcp_add_addr_len(opts, add_addr);
>>>>>> + if (!mptcp_pm_should_add_signal(msk) ||
>>>>>> + !mptcp_pm_add_addr_signal(msk, remaining, &opts->local, &opts->remote, &echo, &port))
>>>>>> + return false;
>>>>>> +
>>>>>> + family = echo ? opts->remote.family : opts->local.family;
>>>>>> + len = mptcp_add_addr_len(family, echo, port);
>>>>>> if (remaining < len)
>>>>>> return false;
>>>>>>
>>>>>> @@ -690,15 +693,14 @@ static bool mptcp_established_options_add_addr(struct sock *sk, struct sk_buff *
>>>>>> if (drop_other_suboptions)
>>>>>> *size -= opt_size;
>>>>>> opts->suboptions |= OPTION_MPTCP_ADD_ADDR;
>>>>>> - if (!(add_addr & BIT(MPTCP_ADD_ADDR_ECHO)) &&
>>>>>> - (add_addr & BIT(MPTCP_ADD_ADDR_SIGNAL))) {
>>>>>> + if (!echo) {
>>>>>> opts->ahmac = add_addr_generate_hmac(msk->local_key,
>>>>>> msk->remote_key,
>>>>>> &opts->local);
>>>>>> }
>>>>>> - pr_debug("addr_signal:%x, echo=%d, local_addr_id=%d, ahmac=%llu, local_port=%d, remote_addr_id=%d, remote_port=%d",
>>>>>> - add_addr, (opts->ahmac == 0), opts->local.id, opts->ahmac,
>>>>>> - ntohs(opts->local.port), opts->remote.id, ntohs(opts->remote.port));
>>>>>> + pr_debug("local_id=%d, local_port=%d, remote_id=%d, remote_port=%d, ahmac=%llu, echo=%d",
>>>>>> + opts->local.id, ntohs(opts->local.port), opts->remote.id,
>>>>>> + ntohs(opts->remote.port), opts->ahmac, echo);
>>>>>>
>>>>>> return true;
>>>>>> }
>>>>>> @@ -1253,13 +1255,10 @@ void mptcp_write_options(__be32 *ptr, const struct tcp_sock *tp,
>>>>>>
>>>>>> mp_capable_done:
>>>>>> if (OPTION_MPTCP_ADD_ADDR & opts->suboptions) {
>>>>>> - struct mptcp_addr_info *addr = &opts->remote;
>>>>>> + struct mptcp_addr_info *addr = opts->ahmac ? &opts->local : &opts->remote;
>>>>>> u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
>>>>>> u8 echo = MPTCP_ADDR_ECHO;
>>>>>>
>>>>>> - if (opts->ahmac)
>>>>>> - addr = &opts->local;
>>>>>> -
>>>>>> #if IS_ENABLED(CONFIG_MPTCP_IPV6)
>>>>>> if (addr->family == AF_INET6)
>>>>>> len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
>>>>>> diff --git a/net/mptcp/pm.c b/net/mptcp/pm.c
>>>>>> index 264f522af530..399b59cb7563 100644
>>>>>> --- a/net/mptcp/pm.c
>>>>>> +++ b/net/mptcp/pm.c
>>>>>> @@ -253,12 +253,13 @@ void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup)
>>>>>>
>>>>>> /* path manager helpers */
>>>>>>
>>>>>> -bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
>>>>>> - unsigned int opt_size, unsigned int remaining,
>>>>>> - struct mptcp_out_options *opts, u8 *add_addr)
>>>>>> +bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
>>>>>> + struct mptcp_addr_info *saddr, struct mptcp_addr_info *daddr,
>>>>>> + bool *echo, bool *port)
>>>>>> {
>>>>>> int ret = false;
>>>>>> u8 add_addr;
>>>>>> + u8 family;
>>>>>>
>>>>>> spin_lock_bh(&msk->pm.lock);
>>>>>>
>>>>>> @@ -266,21 +267,15 @@ bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
>>>>>> if (!mptcp_pm_should_add_signal(msk))
>>>>>> goto out_unlock;
>>>>>>
>>>>>> - opts->local = msk->pm.local;
>>>>>> - opts->remote = msk->pm.remote;
>>>>>> - *add_addr = msk->pm.addr_signal;
>>>>>> + *echo = mptcp_pm_should_add_signal_echo(msk);
>>>>>> + *port = !!(*echo ? msk->pm.remote.port : msk->pm.local.port);
>>>>>>
>>>>>> - if (((msk->pm.addr_signal & BIT(MPTCP_ADD_ADDR_ECHO)) ||
>>>>>> - ((msk->pm.addr_signal & BIT(MPTCP_ADD_ADDR_SIGNAL)) &&
>>>>>> - (msk->pm.local.family == AF_INET6 || msk->pm.local.port))) &&
>>>>>> - skb && skb_is_tcp_pure_ack(skb)) {
>>>>>> - remaining += opt_size;
>>>>>> - }
>>>>>> -
>>>>>> - if (remaining < mptcp_add_addr_len(opts, *add_addr))
>>>>>> + family = *echo ? msk->pm.remote.family : msk->pm.local.family;
>>>>>> + if (remaining < mptcp_add_addr_len(family, *echo, *port))
>>>>>> goto out_unlock;
>>>>>>
>>>>>> *saddr = msk->pm.local;
>>>>>> + *daddr = msk->pm.remote;
>>>>>> add_addr = READ_ONCE(msk->pm.addr_signal);
>>>>>> if (mptcp_pm_should_add_signal_echo(msk))
>>>>>> add_addr &= ~BIT(MPTCP_ADD_ADDR_ECHO);
>>>>>> diff --git a/net/mptcp/protocol.h b/net/mptcp/protocol.h
>>>>>> index 937e0309e340..4b63cc6079fa 100644
>>>>>> --- a/net/mptcp/protocol.h
>>>>>> +++ b/net/mptcp/protocol.h
>>>>>> @@ -766,25 +766,16 @@ static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
>>>>>> return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
>>>>>> }
>>>>>>
>>>>>> -static inline unsigned int mptcp_add_addr_len(struct mptcp_out_options *opts,
>>>>>> - u8 add_addr)
>>>>>> +static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
>>>>>> {
>>>>>> - struct mptcp_addr_info *addr = &opts->remote;
>>>>>> - u8 len = 0;
>>>>>> + u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
>>>>>>
>>>>>> - if (!(add_addr & BIT(MPTCP_ADD_ADDR_ECHO)) &&
>>>>>> - (add_addr & BIT(MPTCP_ADD_ADDR_SIGNAL))) {
>>>>>> - addr = &opts->local;
>>>>>> + if (family == AF_INET6)
>>>>>> + len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
>>>>>> + if (!echo)
>>>>>> len += MPTCPOPT_THMAC_LEN;
>>>>>> - }
>>>>>> -
>>>>>> - if (addr->family == AF_INET6)
>>>>>> - len += TCPOLEN_MPTCP_ADD_ADDR6_BASE;
>>>>>> - else
>>>>>> - len += TCPOLEN_MPTCP_ADD_ADDR_BASE;
>>>>>> -
>>>>>> /* account for 2 trailing 'nop' options */
>>>>>> - if (addr->port)
>>>>>> + if (port)
>>>>>> len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
>>>>>>
>>>>>> return len;
>>>>>> @@ -798,9 +789,9 @@ static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
>>>>>> return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
>>>>>> }
>>>>>>
>>>>>> -bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, struct sk_buff *skb,
>>>>>> - unsigned int opt_size, unsigned int remaining,
>>>>>> - struct mptcp_out_options *opts, u8 *add_addr);
>>>>>> +bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
>>>>>> + struct mptcp_addr_info *saddr, struct mptcp_addr_info *daddr,
>>>>>> + bool *echo, bool *port);
>>>>>> bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
>>>>>> struct mptcp_rm_list *rm_list);
>>>>>> int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
>>>>>>
>>>>>
>>>>> --
>>>>> Li YongLong
>>>>
>>>
>>> --
>>> Li YongLong
>
--
Li YongLong
next prev parent reply other threads:[~2021-07-12 9:21 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-11 15:15 [MPTCP][PATCH mptcp-next] Squash to "mptcp: make MPTCP_ADD_ADDR_SIGNAL and MPTCP_ADD_ADDR_ECHO separate" Geliang Tang
2021-07-11 15:15 ` [MPTCP][PATCH mptcp-next] Squash to "mptcp: fix ADD_ADDR and RM_ADDR maybe flush addr_signal each other" Geliang Tang
2021-07-12 9:55 ` Yonglong Li
2021-07-12 10:34 ` Geliang Tang
2021-07-12 22:27 ` Mat Martineau
2021-07-11 15:15 ` [MPTCP][PATCH mptcp-next] Squash to "mptcp: build ADD_ADDR/echo-ADD_ADDR option according pm.add_signal" Geliang Tang
2021-07-12 1:34 ` Yonglong Li
2021-07-12 7:33 ` Geliang Tang
2021-07-12 8:06 ` Yonglong Li
2021-07-12 8:44 ` Geliang Tang
2021-07-12 9:07 ` Geliang Tang
2021-07-12 9:21 ` Yonglong Li [this message]
2021-07-12 9:14 ` Yonglong Li
2021-07-12 9:29 ` Geliang Tang
2021-07-12 9:44 ` Yonglong Li
2021-07-12 10:34 ` Geliang Tang
2021-07-12 22:10 ` [MPTCP][PATCH mptcp-next] Squash to "mptcp: make MPTCP_ADD_ADDR_SIGNAL and MPTCP_ADD_ADDR_ECHO separate" Mat Martineau
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=8b424320-ae9e-18bb-2011-cb7562497f13@chinatelecom.cn \
--to=liyonglong@chinatelecom.cn \
--cc=geliangtang@gmail.com \
--cc=mptcp@lists.linux.dev \
/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).