1434609 - nsHalfOpenSocket::SetFastOpenConnected MOZ_ASSERT((mFastOpenStatus... )) faiied

Status: RESOLVED FIXED

(Core :: Networking, defect, P2)

Description

[Marco Bonardo [:mak]]

Attachment: mach_run_log.txt

+++ This bug was initially created as a clone of Bug #1432254 +++

I'm hitting the assertion with TFO_DISABLED, the output of mach run is attached.

winver reports build 1709 (16299.192)
As requested the output of netsh int tcp show global (it's in italian, sorry)

Parametri globali TCP
----------------------------------------------
Stato Receive-Side Scaling          : enabled
Stato Chimney Offload               : disabled
Livello regolazione automatica finestra ricezione    : normal
Provider controllo congestione componente aggiuntivo  : default
Funzionalità ECN                      : disabled
Timestamp RFC 1323                 : disabled
RTO iniziale                         :  3000
Stato unione segmenti ricezione    : enabled
Resilienza RTT non SACK             : disabled
Max. ritrasmissioni SYN             : 2
Apertura rapida                           : enabled
Fallback apertura rapida                  : enabled
Profilo trasmissione cadenzata                      : off

Comment 1

[Marco Bonardo [:mak]]

I suppose "Apertura rapida" (Fast Open) is what you're looking for.

Comment 2

[Honza Bambas (:mayhemer)]

(In reply to Marco Bonardo [::mak] from comment #1)
> I suppose "Apertura rapida" (Fast Open) is what you're looking for.

yes, thanks.

Comment 3

[Honza Bambas (:mayhemer)]

I can also see this with TFO_DATA_COOKIE_NOT_ACCEPTED in xpcshell tests (./mach test toolkit/components/extensions/test/xpcshell/test_ext_proxy_auth.js)

Comment 4

[Honza Bambas (:mayhemer)]

ah.. that's the original bug, my tree is just not updated...

Comment 5

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

Consistent failure for Windows 10 debug (opt) builds for me. Believe this started when pulling on the 2nd (thought hadn't pulled in awhile before that), continues to today. Debug shows TFO_DISABLED is reason for assert.

`netsh int tcp show global` result:

TCP Global Parameters
----------------------------------------------
Receive-Side Scaling State          : enabled
Chimney Offload State               : disabled
Receive Window Auto-Tuning Level    : normal
Add-On Congestion Control Provider  : default
ECN Capability                      : disabled
RFC 1323 Timestamps                 : disabled
Initial RTO                         : 3000
Receive Segment Coalescing State    : enabled
Non Sack Rtt Resiliency             : disabled
Max SYN Retransmissions             : 2
Fast Open                           : enabled
Fast Open Fallback                  : enabled
Pacing Profile                      : off

Comment 6

[Milan Sreckovic [:milan] (needinfo for best results)]

I hit the same assertion as well; haven't been able to run debug since.  Windows 10.

Comment 7

[Bryce Seager van Dyk [:bryce] (he/him) - Not reading bugmail]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=0edee4e364cbb9d33b2c9bd87d9c81f991a7313f

Comment 8

[Kerem Kat]

Commenting out the assertion allows running firefox.

Comment 9

[Honza Bambas (:mayhemer)]

can any of you please provide a stack trace?  I was to see where from it's called so it's easier to chace for me, thanks.

Comment 10

[Honza Bambas (:mayhemer)]

OK, I think based on e.g. [1] we can safely add TFO_DISABLED into the list.  It's a valid TFO at this stage.

[1] https://searchfox.org/mozilla-central/rev/0c0ddaa7e859a2b76a56a0e2e9c0de88af166812/netwerk/base/TCPFastOpenLayer.cpp#435

Comment 11

[Honza Bambas (:mayhemer)]

Attachment: v1

Comment 12

[Honza Bambas (:mayhemer)]

Comment on attachment 8952353 [details] [diff] [review]
v1

Valentin, in case you are willing to do an r here and get to it first, thanks.

Comment 13

[Honza Bambas (:mayhemer)]

Comment on attachment 8952353 [details] [diff] [review]
v1

thanks

Comment 14

[Milan Sreckovic [:milan] (needinfo for best results)]

Assertion failure: (mFastOpenStatus == TFO_NOT_TRIED) || (mFastOpenStatus == TFO_DATA_SENT) || (mFastOpenStatus == TFO_TRIED) || (mFastOpenStatus == TFO_DATA_COOKIE_NOT_ACCEPTED), at c:/Users/Milan/Repos/mozilla-central/netwerk/protocol/http/nsHttpConnectionMgr.cpp:4597
[Parent 16708, Gecko_IOThread] WARNING: pipe error: 109: file c:/Users/Milan/Repos/mozilla-central/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[Parent 16708, Main Thread] WARNING: attempt to modify an immutable nsStandardURL: file c:/Users/Milan/Repos/mozilla-central/netwerk/base/nsStandardURL.cpp, line 1698
#01: mozilla::net::nsHttpConnectionMgr::nsHalfOpenSocket::SetFastOpenConnected (c:\users\milan\repos\mozilla-central\netwerk\protocol\http\nshttpconnectionmgr.cpp:4597)
#02: mozilla::net::nsSocketTransport::OnSocketConnected (c:\users\milan\repos\mozilla-central\netwerk\base\nssockettransport2.cpp:1907)
#03: mozilla::net::nsSocketTransport::OnSocketReady (c:\users\milan\repos\mozilla-central\netwerk\base\nssockettransport2.cpp:2255)
#04: mozilla::net::nsSocketTransportService::DoPollIteration (c:\users\milan\repos\mozilla-central\netwerk\base\nssockettransportservice2.cpp:1156)
#05: mozilla::net::nsSocketTransportService::Run (c:\users\milan\repos\mozilla-central\netwerk\base\nssockettransportservice2.cpp:927)
#06: nsThread::ProcessNextEvent (c:\users\milan\repos\mozilla-central\xpcom\threads\nsthread.cpp:1041)
#07: NS_ProcessNextEvent (c:\users\milan\repos\mozilla-central\xpcom\threads\nsthreadutils.cpp:517)
#08: mozilla::ipc::MessagePumpForNonMainThreads::Run (c:\users\milan\repos\mozilla-central\ipc\glue\messagepump.cpp:364)
#09: MessageLoop::RunInternal (c:\users\milan\repos\mozilla-central\ipc\chromium\src\base\message_loop.cc:327)
#10: MessageLoop::RunHandler (c:\users\milan\repos\mozilla-central\ipc\chromium\src\base\message_loop.cc:320)
#11: MessageLoop::Run (c:\users\milan\repos\mozilla-central\ipc\chromium\src\base\message_loop.cc:300)
#12: nsThread::ThreadFunc (c:\users\milan\repos\mozilla-central\xpcom\threads\nsthread.cpp:425)
#13: _PR_NativeRunThread (c:\users\milan\repos\mozilla-central\nsprpub\pr\src\threads\combined\pruthr.c:397)
#14: pr_root (c:\users\milan\repos\mozilla-central\nsprpub\pr\src\md\windows\w95thred.c:137)
#15: o_strcpy_s[C:\WINDOWS\System32\ucrtbase.dll +0x3ab7d]
#16: BaseThreadInitThunk[C:\WINDOWS\System32\KERNEL32.DLL +0x18654]
#17: patched_BaseThreadInitThunk (c:\users\milan\repos\mozilla-central\mozglue\build\windowsdllblocklist.cpp:839)
#18: RtlGetAppContainerNamedObjectPath[C:\WINDOWS\SYSTEM32\ntdll.dll +0x64a77]
#19: RtlGetAppContainerNamedObjectPath[C:\WINDOWS\SYSTEM32\ntdll.dll +0x64a47]

Comment 15

[Kerem Kat]

I had the same stack trace.

Comment 16

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/760d011572df
Add TFO_DISABLED as one of possible states after finishing TCP Fast Open (TFO) connection, r=valentin

Comment 17

[Raul Gurzau (:RaulG)]

https://hg.mozilla.org/mozilla-central/rev/760d011572df