[PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[Paolo Abeni]

It turns out that the issue reported by Geliang on the pending backlog
patches is actually on older one that the new code made more easy
reproducible.

The first patch in the series addresses it[1], the 2nd one introduce new
mibs to hopefully catch this problem sooned.

I have a pktdrill to validate this scenario. I'll share that soon.

@Geliang could you please validate the above in your testbed, on top of
the backlog and splice patches?

[1] at least here, at least with the packet drill reproducer ;)
---
v1 -> v2:
  - drop unneeded delta in patch 1/2

Paolo Abeni (2):
  mptcp: restore window probe
  mptcp: zero window probe mib

 net/mptcp/mib.c      | 1 +
 net/mptcp/mib.h      | 1 +
 net/mptcp/protocol.c | 8 +++++++-
 3 files changed, 9 insertions(+), 1 deletion(-)

-- 
2.51.0

[PATCH v2 mptcp-net 1/2] mptcp: restore window probe

[Paolo Abeni]

Since commit 72377ab2d671 ("mptcp: more conservative check for zero
probes") the MPTCP-level zero window probe check is always disabled, as
the TCP-level write queue always contains at least the newly allocated
skb.

Refine the relevant check tacking in account that the above condition
and that such skb can have zero length.

Fixes: 72377ab2d671 ("mptcp: more conservative check for zero probes")
Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
v1 -> v2:
  - drop unneeded changes (Mat)
---
 net/mptcp/protocol.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index bde76b7311f986..a7fa69cf79f106 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -1339,7 +1339,12 @@ static int mptcp_sendmsg_frag(struct sock *sk, struct sock *ssk,
 	if (copy == 0) {
 		u64 snd_una = READ_ONCE(msk->snd_una);
 
-		if (snd_una != msk->snd_nxt || tcp_write_queue_tail(ssk)) {
+		/* No need for zero probe if there are any data pending
+		 * either at the msk or ssk level; skb is the current write
+		 * queue tail and can be empty at this point.
+		 */
+		if (snd_una != msk->snd_nxt || skb->len ||
+		    skb != tcp_send_head(ssk)) {
 			tcp_remove_empty_skb(ssk);
 			return 0;
 		}
-- 
2.51.0

[PATCH v2 mptcp-net 2/2] mptcp: zero window probe mib

[Paolo Abeni]

Explicitly account for MPTCP-level zero windows probe, to catch
hopefully earlier issues alike the one addressed by the previous
patch.

Reviewed-by: Mat Martineau <martineau@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
 net/mptcp/mib.c      | 1 +
 net/mptcp/mib.h      | 1 +
 net/mptcp/protocol.c | 1 +
 3 files changed, 3 insertions(+)

diff --git a/net/mptcp/mib.c b/net/mptcp/mib.c
index 6003e47c770a7c..eb4645a9c5ac07 100644
--- a/net/mptcp/mib.c
+++ b/net/mptcp/mib.c
@@ -85,6 +85,7 @@ static const struct snmp_mib mptcp_snmp_list[] = {
 	SNMP_MIB_ITEM("DssFallback", MPTCP_MIB_DSSFALLBACK),
 	SNMP_MIB_ITEM("SimultConnectFallback", MPTCP_MIB_SIMULTCONNFALLBACK),
 	SNMP_MIB_ITEM("FallbackFailed", MPTCP_MIB_FALLBACKFAILED),
+	SNMP_MIB_ITEM("MPTcpWinProbe", MPTCP_MIB_MPTCPWINPROBE),
 };
 
 /* mptcp_mib_alloc - allocate percpu mib counters
diff --git a/net/mptcp/mib.h b/net/mptcp/mib.h
index 309bac6fea3252..f83a113700522e 100644
--- a/net/mptcp/mib.h
+++ b/net/mptcp/mib.h
@@ -88,6 +88,7 @@ enum linux_mptcp_mib_field {
 	MPTCP_MIB_DSSFALLBACK,		/* Bad or missing DSS */
 	MPTCP_MIB_SIMULTCONNFALLBACK,	/* Simultaneous connect */
 	MPTCP_MIB_FALLBACKFAILED,	/* Can't fallback due to msk status */
+	MPTCP_MIB_MPTCPWINPROBE,	/* MPTCP-level zero window probe */
 	__MPTCP_MIB_MAX
 };
 
diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
index a7fa69cf79f106..7f2690f8d92756 100644
--- a/net/mptcp/protocol.c
+++ b/net/mptcp/protocol.c
@@ -1395,6 +1395,7 @@ static int mptcp_sendmsg_frag(struct sock *sk, struct sock *ssk,
 		 mpext->dsn64);
 
 	if (zero_window_probe) {
+		MPTCP_INC_STATS(sock_net(ssk), MPTCP_MIB_MPTCPWINPROBE);
 		mptcp_subflow_ctx(ssk)->rel_write_seq += copy;
 		mpext->frozen = 1;
 		if (READ_ONCE(msk->csum_enabled))
-- 
2.51.0

Re: [PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[MPTCP CI]

Hi Paolo,

Thank you for your modifications, that's great!

Our CI did some validations and here is its report:

- KVM Validation: normal (except selftest_mptcp_join): Success! ✅
- KVM Validation: normal (only selftest_mptcp_join): Success! ✅
- KVM Validation: debug (except selftest_mptcp_join): Success! ✅
- KVM Validation: debug (only selftest_mptcp_join): Success! ✅
- KVM Validation: btf-normal (only bpftest_all): Success! ✅
- KVM Validation: btf-debug (only bpftest_all): Success! ✅
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/18709417366

Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/5d3c078227a8
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1014382


If there are some issues, you can reproduce them using the same environment as
the one used by the CI thanks to a docker image, e.g.:

    $ cd [kernel source code]
    $ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
        --pull always mptcp/mptcp-upstream-virtme-docker:latest \
        auto-normal

For more details:

    https://github.com/multipath-tcp/mptcp-upstream-virtme-docker


Please note that despite all the efforts that have been already done to have a
stable tests suite when executed on a public CI like here, it is possible some
reported issues are not due to your modifications. Still, do not hesitate to
help us improve that ;-)

Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)

Re: [PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[Geliang Tang]

Hi Paolo,

On Wed, 2025-10-22 at 09:39 +0200, Paolo Abeni wrote:
> It turns out that the issue reported by Geliang on the pending
> backlog
> patches is actually on older one that the new code made more easy
> reproducible.
> 
> The first patch in the series addresses it[1], the 2nd one introduce
> new
> mibs to hopefully catch this problem sooned.
> 
> I have a pktdrill to validate this scenario. I'll share that soon.
> 
> @Geliang could you please validate the above in your testbed, on top
> of
> the backlog and splice patches?

Thanks very much. I've already started testing, and so far everything
is running smoothly. I'll let it run loop tests overnight and will have
the results by tomorrow morning. I'll provide you with feedback then.

Thanks,
-Geliang

> 
> [1] at least here, at least with the packet drill reproducer ;)
> ---
> v1 -> v2:
>   - drop unneeded delta in patch 1/2
> 
> Paolo Abeni (2):
>   mptcp: restore window probe
>   mptcp: zero window probe mib
> 
>  net/mptcp/mib.c      | 1 +
>  net/mptcp/mib.h      | 1 +
>  net/mptcp/protocol.c | 8 +++++++-
>  3 files changed, 9 insertions(+), 1 deletion(-)
> 

Re: [PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[Geliang Tang]

Hi Paolo,

On Wed, 2025-10-22 at 18:07 +0800, Geliang Tang wrote:
> Hi Paolo,
> 
> On Wed, 2025-10-22 at 09:39 +0200, Paolo Abeni wrote:
> > It turns out that the issue reported by Geliang on the pending
> > backlog
> > patches is actually on older one that the new code made more easy
> > reproducible.
> > 
> > The first patch in the series addresses it[1], the 2nd one
> > introduce
> > new
> > mibs to hopefully catch this problem sooned.
> > 
> > I have a pktdrill to validate this scenario. I'll share that soon.
> > 
> > @Geliang could you please validate the above in your testbed, on
> > top
> > of
> > the backlog and splice patches?
> 
> Thanks very much. I've already started testing, and so far everything
> is running smoothly. I'll let it run loop tests overnight and will
> have
> the results by tomorrow morning. I'll provide you with feedback then.

Thank you so much! This fix really works. I've run hundreds of splice
test loops with the following three patchsets applied, and no test
failures occurred:

	mptcp: restore zero window probe, v2
	mptcp: introduce backlog processing, v6
	implement mptcp read_sock, v13

Please add my tag for this set:

	Reported-and-tested-by: Geliang Tang <geliang@kernel.org>

> 
> Thanks,
> -Geliang
> 
> > 
> > [1] at least here, at least with the packet drill reproducer ;)
> > ---
> > v1 -> v2:
> >   - drop unneeded delta in patch 1/2

This change in v1 is valid. Later I will send a dedicated cleanup patch
for -next to use tcp_sk() in mptcp_sendmsg_frag() with your "Suggested-
by" tag.

Thanks,
-Geliang

> > 
> > Paolo Abeni (2):
> >   mptcp: restore window probe
> >   mptcp: zero window probe mib
> > 
> >  net/mptcp/mib.c      | 1 +
> >  net/mptcp/mib.h      | 1 +
> >  net/mptcp/protocol.c | 8 +++++++-
> >  3 files changed, 9 insertions(+), 1 deletion(-)
> > 
> 
> 

Re: [PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[Matthieu Baerts]

Hi Geliang,

On 23/10/2025 08:24, Geliang Tang wrote:
> Hi Paolo,
> 
> On Wed, 2025-10-22 at 18:07 +0800, Geliang Tang wrote:
>> Hi Paolo,
>>
>> On Wed, 2025-10-22 at 09:39 +0200, Paolo Abeni wrote:
>>> It turns out that the issue reported by Geliang on the pending
>>> backlog
>>> patches is actually on older one that the new code made more easy
>>> reproducible.
>>>
>>> The first patch in the series addresses it[1], the 2nd one
>>> introduce
>>> new
>>> mibs to hopefully catch this problem sooned.
>>>
>>> I have a pktdrill to validate this scenario. I'll share that soon.
>>>
>>> @Geliang could you please validate the above in your testbed, on
>>> top
>>> of
>>> the backlog and splice patches?
>>
>> Thanks very much. I've already started testing, and so far everything
>> is running smoothly. I'll let it run loop tests overnight and will
>> have
>> the results by tomorrow morning. I'll provide you with feedback then.
> 
> Thank you so much! This fix really works. I've run hundreds of splice
> test loops with the following three patchsets applied, and no test
> failures occurred:
> 
> 	mptcp: restore zero window probe, v2
> 	mptcp: introduce backlog processing, v6
> 	implement mptcp read_sock, v13
> 
> Please add my tag for this set:
> 
> 	Reported-and-tested-by: Geliang Tang <geliang@kernel.org>

Thank you for the test!

To avoid issues with checkpatch and to be consistent, I'm going to add
this on the first patch:

  Reported-by: (...)
  Closes:
https://lore.kernel.org/d0a814c364e744ca6b836ccd5b6e9146882e8d42.camel@kernel.org
  (...)
  Tested-by: (...)

And only the Tested-by on the second one. I hope that's OK!

>> Thanks,
>> -Geliang
>>
>>>
>>> [1] at least here, at least with the packet drill reproducer ;)
>>> ---
>>> v1 -> v2:
>>>   - drop unneeded delta in patch 1/2
> 
> This change in v1 is valid. Later I will send a dedicated cleanup patch
> for -next to use tcp_sk() in mptcp_sendmsg_frag() with your "Suggested-
> by" tag.

Please only add such patch when we modify code around in net-next.

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[Matthieu Baerts]

Hi Paolo, Geliang, Mat,

On 22/10/2025 09:39, Paolo Abeni wrote:
> It turns out that the issue reported by Geliang on the pending backlog
> patches is actually on older one that the new code made more easy
> reproducible.
> 
> The first patch in the series addresses it[1], the 2nd one introduce new
> mibs to hopefully catch this problem sooned.
> 
> I have a pktdrill to validate this scenario. I'll share that soon.

Thank you for the new pktdrill. I don't know if you saw my question:

  https://github.com/multipath-tcp/packetdrill/pull/179

I also just did a quick review and test while at it


> @Geliang could you please validate the above in your testbed, on top of
> the backlog and splice patches?

Thank you for the patches and the reviews! I already applied the
patches, I guess the issues with Packetdrill are in the test, not the
kernel modifications.


New patches for t/upstream-net and t/upstream:
- b7749bb6ab95: mptcp: restore window probe
- 5d8d442266f9: mptcp: zero window probe mib
- Results: 94b2567634ae..affaafb5552e (export-net)
- Results: 5d6d50a980ff..ed6f82a81575 (export)

Tests are now in progress:

- export-net:
https://github.com/multipath-tcp/mptcp_net-next/commit/5db161c8d173fc05a4d26fd5df5e8c6e90ac8c19/checks
- export:
https://github.com/multipath-tcp/mptcp_net-next/commit/785e3842513ff5993f32c918f90dbee62244ed63/checks

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH v2 mptcp-net 2/2] mptcp: zero window probe mib

[Matthieu Baerts]

Hi Paolo,

On 22/10/2025 09:39, Paolo Abeni wrote:
> Explicitly account for MPTCP-level zero windows probe, to catch
> hopefully earlier issues alike the one addressed by the previous
> patch.
> 
> Reviewed-by: Mat Martineau <martineau@kernel.org>
> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
> ---
>  net/mptcp/mib.c      | 1 +
>  net/mptcp/mib.h      | 1 +
>  net/mptcp/protocol.c | 1 +
>  3 files changed, 3 insertions(+)
> 
> diff --git a/net/mptcp/mib.c b/net/mptcp/mib.c
> index 6003e47c770a7c..eb4645a9c5ac07 100644
> --- a/net/mptcp/mib.c
> +++ b/net/mptcp/mib.c
> @@ -85,6 +85,7 @@ static const struct snmp_mib mptcp_snmp_list[] = {
>  	SNMP_MIB_ITEM("DssFallback", MPTCP_MIB_DSSFALLBACK),
>  	SNMP_MIB_ITEM("SimultConnectFallback", MPTCP_MIB_SIMULTCONNFALLBACK),
>  	SNMP_MIB_ITEM("FallbackFailed", MPTCP_MIB_FALLBACKFAILED),
> +	SNMP_MIB_ITEM("MPTcpWinProbe", MPTCP_MIB_MPTCPWINPROBE),

Detail: I just noticed that after having applied the patch: Do we need
to double the 'MPTCP' keyword here? The full MIB counter name is:

  MPTcpExtMPTcpWinProbe

Maybe just WinProbe is enough? → MPTcpExtWinProbe

Or the second MPTCP using capital letters, similar to "MPTCPRetrans"?

(I can fix that directly in the tree.)

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH v2 mptcp-net 2/2] mptcp: zero window probe mib

[Paolo Abeni]

On 10/23/25 6:26 PM, Matthieu Baerts wrote:
> On 22/10/2025 09:39, Paolo Abeni wrote:
>> Explicitly account for MPTCP-level zero windows probe, to catch
>> hopefully earlier issues alike the one addressed by the previous
>> patch.
>>
>> Reviewed-by: Mat Martineau <martineau@kernel.org>
>> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
>> ---
>>  net/mptcp/mib.c      | 1 +
>>  net/mptcp/mib.h      | 1 +
>>  net/mptcp/protocol.c | 1 +
>>  3 files changed, 3 insertions(+)
>>
>> diff --git a/net/mptcp/mib.c b/net/mptcp/mib.c
>> index 6003e47c770a7c..eb4645a9c5ac07 100644
>> --- a/net/mptcp/mib.c
>> +++ b/net/mptcp/mib.c
>> @@ -85,6 +85,7 @@ static const struct snmp_mib mptcp_snmp_list[] = {
>>  	SNMP_MIB_ITEM("DssFallback", MPTCP_MIB_DSSFALLBACK),
>>  	SNMP_MIB_ITEM("SimultConnectFallback", MPTCP_MIB_SIMULTCONNFALLBACK),
>>  	SNMP_MIB_ITEM("FallbackFailed", MPTCP_MIB_FALLBACKFAILED),
>> +	SNMP_MIB_ITEM("MPTcpWinProbe", MPTCP_MIB_MPTCPWINPROBE),
> 
> Detail: I just noticed that after having applied the patch: Do we need
> to double the 'MPTCP' keyword here? The full MIB counter name is:
> 
>   MPTcpExtMPTcpWinProbe
> 
> Maybe just WinProbe is enough? → MPTcpExtWinProbe

I'm sorry for the late reply. Yes WinProbe -> MPTcpExtWinProbe is the
correct/intended name, if you could fix that in the tree would be great,
thanks!

Paolo

Re: [PATCH v2 mptcp-net 2/2] mptcp: zero window probe mib

[Matthieu Baerts]

Hi Paolo,

On 24/10/2025 09:30, Paolo Abeni wrote:
> On 10/23/25 6:26 PM, Matthieu Baerts wrote:
>> On 22/10/2025 09:39, Paolo Abeni wrote:
>>> Explicitly account for MPTCP-level zero windows probe, to catch
>>> hopefully earlier issues alike the one addressed by the previous
>>> patch.
>>>
>>> Reviewed-by: Mat Martineau <martineau@kernel.org>
>>> Signed-off-by: Paolo Abeni <pabeni@redhat.com>
>>> ---
>>>  net/mptcp/mib.c      | 1 +
>>>  net/mptcp/mib.h      | 1 +
>>>  net/mptcp/protocol.c | 1 +
>>>  3 files changed, 3 insertions(+)
>>>
>>> diff --git a/net/mptcp/mib.c b/net/mptcp/mib.c
>>> index 6003e47c770a7c..eb4645a9c5ac07 100644
>>> --- a/net/mptcp/mib.c
>>> +++ b/net/mptcp/mib.c
>>> @@ -85,6 +85,7 @@ static const struct snmp_mib mptcp_snmp_list[] = {
>>>  	SNMP_MIB_ITEM("DssFallback", MPTCP_MIB_DSSFALLBACK),
>>>  	SNMP_MIB_ITEM("SimultConnectFallback", MPTCP_MIB_SIMULTCONNFALLBACK),
>>>  	SNMP_MIB_ITEM("FallbackFailed", MPTCP_MIB_FALLBACKFAILED),
>>> +	SNMP_MIB_ITEM("MPTcpWinProbe", MPTCP_MIB_MPTCPWINPROBE),
>>
>> Detail: I just noticed that after having applied the patch: Do we need
>> to double the 'MPTCP' keyword here? The full MIB counter name is:
>>
>>   MPTcpExtMPTcpWinProbe
>>
>> Maybe just WinProbe is enough? → MPTcpExtWinProbe
> 
> I'm sorry for the late reply. Yes WinProbe -> MPTcpExtWinProbe is the
> correct/intended name, if you could fix that in the tree would be great,
> thanks!

Thanks! Just did! I also renamed the enum

  MPTCP_MIB_MPTCPWINPROBE → MPTCP_MIB_WINPROBE:

New patches for t/upstream-net and t/upstream:
- 9525948cb5cc: Squash to "mptcp: zero window probe mib"
- Results: f6c0967d41d1..628a66ff3d17 (export-net)
- Results: 8416614a1e27..6354d0b32825 (export)

Tests are now in progress:

- export-net:
https://github.com/multipath-tcp/mptcp_net-next/commit/60ce52cf453b5182b993d2420d01422355ebbd67/checks
- export:
https://github.com/multipath-tcp/mptcp_net-next/commit/a76c4725014b6c51385d301f1027a7891714a6bb/checks

Cheers,
Matt
-- 
Sponsored by the NGI0 Core fund.

Re: [PATCH v2 mptcp-net 0/2] mptcp: restore zero window probe

[Geliang Tang]

Hi Matt,

On Thu, 2025-10-23 at 17:54 +0200, Matthieu Baerts wrote:
> Hi Geliang,
> 
> On 23/10/2025 08:24, Geliang Tang wrote:
> > Hi Paolo,
> > 
> > On Wed, 2025-10-22 at 18:07 +0800, Geliang Tang wrote:
> > > Hi Paolo,
> > > 
> > > On Wed, 2025-10-22 at 09:39 +0200, Paolo Abeni wrote:
> > > > It turns out that the issue reported by Geliang on the pending
> > > > backlog
> > > > patches is actually on older one that the new code made more
> > > > easy
> > > > reproducible.
> > > > 
> > > > The first patch in the series addresses it[1], the 2nd one
> > > > introduce
> > > > new
> > > > mibs to hopefully catch this problem sooned.
> > > > 
> > > > I have a pktdrill to validate this scenario. I'll share that
> > > > soon.
> > > > 
> > > > @Geliang could you please validate the above in your testbed,
> > > > on
> > > > top
> > > > of
> > > > the backlog and splice patches?
> > > 
> > > Thanks very much. I've already started testing, and so far
> > > everything
> > > is running smoothly. I'll let it run loop tests overnight and
> > > will
> > > have
> > > the results by tomorrow morning. I'll provide you with feedback
> > > then.
> > 
> > Thank you so much! This fix really works. I've run hundreds of
> > splice
> > test loops with the following three patchsets applied, and no test
> > failures occurred:
> > 
> > 	mptcp: restore zero window probe, v2
> > 	mptcp: introduce backlog processing, v6
> > 	implement mptcp read_sock, v13
> > 
> > Please add my tag for this set:
> > 
> > 	Reported-and-tested-by: Geliang Tang <geliang@kernel.org>
> 
> Thank you for the test!
> 
> To avoid issues with checkpatch and to be consistent, I'm going to
> add
> this on the first patch:
> 
>   Reported-by: (...)
>   Closes:
> https://lore.kernel.org/d0a814c364e744ca6b836ccd5b6e9146882e8d42.camel@kernel.org
>   (...)
>   Tested-by: (...)
> 
> And only the Tested-by on the second one. I hope that's OK!
> 
> > > Thanks,
> > > -Geliang
> > > 
> > > > 
> > > > [1] at least here, at least with the packet drill reproducer ;)
> > > > ---
> > > > v1 -> v2:
> > > >   - drop unneeded delta in patch 1/2
> > 
> > This change in v1 is valid. Later I will send a dedicated cleanup
> > patch
> > for -next to use tcp_sk() in mptcp_sendmsg_frag() with your
> > "Suggested-
> > by" tag.
> 
> Please only add such patch when we modify code around in net-next.

Sure, I'll send it out together once I modify mptcp_sendmsg_frag in the
future.

Thanks,
-Geliang

> 
> Cheers,
> Matt