23891 - deadlock or thread contention displaying IMAP messages

Status: VERIFIED FIXED

(MailNews Core :: Networking, defect, P3)

Description

[Alec Flett]

I've been getting what looks like either deadlock or just lock thrashing
displaying some IMAP messages. My CPU hits 100%, but about 50% of that time is
in the kernel, so I think it's lock contention.

Here are the two threads that are contending:
(gdb) where 15
#0  pthread_handle_sigrestart (sig=-1073754820, ctx={gs = 27260,
      __gsh = 16412, fs = 0, __fsh = 0, es = 11124, __esh = 2053, ds = 52692,
      __dsh = 49151, edi = 134556552, esi = 134556548, ebp = 0, esp = 0,
      ebx = 134555896, edx = 16382, ecx = 134555916, eax = 134555912,
      trapno = 0, err = 0, eip = 1077146764, cs = 52608, __csh = 49151,
      eflags = 1077146764, esp_at_signal = 1075604092, ss = 18112,
      __ssh = 16412, fpstate = 0xbfffcda4, oldmask = 1075569153, cr2 = 0})
    at pthread.c:561
#1  0x401be211 in __pthread_lock (lock=0x8052b84, self=0x401c46c0)
    at restart.h:32
#2  0x401bb99a in __pthread_mutex_lock (mutex=0x8052b74) at mutex.c:84
#3  0x4019c6fb in PR_Lock (lock=0x8052b74) at ptsynch.c:167
#4  0x4019d4ec in PR_EnterMonitor (mon=0x8052b70) at ptsynch.c:492
#5  0x40186dbd in PR_CEnterMonitor (address=0x87f4990) at prcmon.c:308
#6  0x4135726a in nsImapProtocol::PseudoInterruptMsgLoad (this=0x87f4990,
    aImapUrl=0x940d048, interrupted=0xbfffcf0c) at nsImapProtocol.cpp:831
#7  0x4134126f in nsImapIncomingServer::PseudoInterruptMsgLoad (
    this=0x8454ef0, aImapUrl=0x940d048, interrupted=0xbfffcf0c)
    at nsImapIncomingServer.cpp:1374
#8  0x4137029c in nsImapService::FetchMessage (this=0x86e6ed0,
    aImapUrl=0x940d048, aImapAction=<incomplete type>,
    aImapMailFolder=0x82b9ba0, aImapMessage=0x82b9c30, aURL=0x0,
    aDisplayConsumer=0x85c9930, messageIdentifierList=0x8cdaed8 "4699",
    messageIdsAreUID=1) at nsImapService.cpp:516
#9  0x4136e4c3 in nsImapService::DisplayMessage (this=0x86e6ed0,
    aMessageURI=0x940cb38 "imap_message://[e-mail]/Outside#4699",
    aDisplayConsumer=0x85c9930, aMsgWindow=0x84660e0, aUrlListener=0x0,
    aURL=0x0) at nsImapService.cpp:273
#10 0x40af58c8 in ?? ()
   from /home1/alecf/seamonkey/mozilla/dist/bin/components/libmailnews.so
#11 0x40129095 in XPTC_InvokeByIndex (that=0x8452930, methodIndex=11,
    paramCount=1, params=0xbfffd284) at xptcinvoke_unixish_x86.cpp:159
#12 0x40964a63 in ?? ()
   from /home1/alecf/seamonkey/mozilla/dist/bin/components/libxpconnect.so
#13 0x409661d2 in ?? ()
   from /home1/alecf/seamonkey/mozilla/dist/bin/components/libxpconnect.so
#14 0x40069669 in ?? ()
   from /home1/alecf/seamonkey/mozilla/dist/bin/libmozjs.so


and:

(gdb) where
#0  pthread_handle_sigrestart (sig=-1088423240, ctx={gs = 27260,
      __gsh = 16412, fs = 31976, __fsh = 2286, es = 10488, __esh = 2053,
      ds = 27260, __dsh = 16412, edi = 134555916, esi = 134555912, ebp = 0,
      esp = 0, ebx = 134555912, edx = 0, ecx = 0, eax = 0, trapno = 0,
      err = 0, eip = 0, cs = 0, __csh = 0, eflags = 0, esp_at_signal = 0,
      ss = 0, __ssh = 0, fpstate = 0x0, oldmask = 0, cr2 = 0}) at pthread.c:561
#1  0x401be211 in __pthread_lock (lock=0x8052908, self=0xbf1ffe60)
    at restart.h:32
#2  0x401bb99a in __pthread_mutex_lock (mutex=0x80528f8) at mutex.c:84
#3  0x4019c6fb in PR_Lock (lock=0x80528f8) at ptsynch.c:167
#4  0x40186df4 in PR_CExitMonitor (address=0x8ee7ce8) at prcmon.c:317
#5  0x41356f87 in nsImapProtocol::ImapThreadIsRunning (this=0x8ee7ce8)
    at nsImapProtocol.cpp:729
#6  0x4135749e in nsImapProtocol::ImapThreadMainLoop (this=0x8ee7ce8)
    at nsImapProtocol.cpp:905
#7  0x41355e53 in nsImapProtocol::Run (this=0x8ee7ce8)
    at nsImapProtocol.cpp:685
#8  0x4011b345 in nsThread::Main (arg=0x8ee8950) at nsThread.cpp:83
#9  0x401a58fb in _pt_root (arg=0x8ee8968) at ptthread.c:157
#10 0x401baeca in pthread_start_thread (arg=0xbf1ffe60) at manager.c:213


I've been seeing this on and off for a week or so... this is on Linux 2.2.12smp,
which means I'm running an SMP kernel, though I'm on a single-processor box, so
it's possible that thread behavior or timing is different than a stock linux
kernel.

The machine is "nefarious" and it's my primary machine.... davidmc's
"dragonballz" has the same configuration.

Comment 1

[Alec Flett]

adding bienvenu & sspitzer to the CC

Comment 2

[Alec Flett]

just noticed I'm missing a frame there in libmailnews.so:
#10 0x40af58c8 in nsMessenger::OpenURL (this=0x8452930,
    url=0x940cb00 "imap_message://[e-mail]/Outside#4699")
    at nsMessenger.cpp:318

Comment 3

[Scott MacGregor]

Alec, do you actually get past the contention? Or are you just stuck once this
occurrs?

Comment 4

[Scott MacGregor]

I've sent Alec a patch which may help alleviate this problem. We'll see.

Comment 5

[Scott MacGregor]

Hey Alec, have you still been running into this problem since you've been
running with that patch I sent you?

Comment 6

[Scott MacGregor]

I checked in my patch. Feel free to re-open alec if you start seeing the
contention problems again.

Comment 7

[lchiang]

I think this is ok for alec now or else he would have reopened the bug :-)