Very Computer

by **Carl Appello** » Sun, 06 Oct 2002 05:15:01

I'm running into a process that's catching an exception
STATUS_POSSIBLE_DEADLOCK. This seems to be raised by ntdll when an attempt
to acquire a Critical Section (EnterCriticalSection) times out. On the
system I'm looking at, there are a couple of interesting internal settings
in ntdll. These are:

RtlpTimoutDisable (set to 0)
RtlpTimeout (set to a large negative number representing 2 minutes worth of
time.)

My process is designed to wait forever to acquire a critical section, and
croaks after 2 minutes with the above settings.

On another system, I see these parameters set to:

RtlpTimoutDisable (set to 1)
RtlpTimeout (set to a large negative number representing 30 days.)

I vaguely recall a setup to debug CriticalSection deadlocks that modifies
the "normal" values for these parameters, setting a fairly short (2 minute)
timeout. But I can't remember for the life of me seeing a reference to the
info. Can anyone jog my memory?

Thanks,
Carl Appellof

by **Pavel Lebedinsk** » Sun, 06 Oct 2002 07:18:28

The timeout is set here:

Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager]
"CriticalSectionTimeout"=dword:00278d00

The value is in seconds.

> I'm running into a process that's catching an exception
> STATUS_POSSIBLE_DEADLOCK. This seems to be raised by ntdll when an
attempt
> to acquire a Critical Section (EnterCriticalSection) times out. On the
> system I'm looking at, there are a couple of interesting internal settings
> in ntdll. These are:

> RtlpTimoutDisable (set to 0)
> RtlpTimeout (set to a large negative number representing 2 minutes worth
of
> time.)

> My process is designed to wait forever to acquire a critical section, and
> croaks after 2 minutes with the above settings.

> On another system, I see these parameters set to:

> RtlpTimoutDisable (set to 1)
> RtlpTimeout (set to a large negative number representing 30 days.)

> I vaguely recall a setup to debug CriticalSection deadlocks that modifies
> the "normal" values for these parameters, setting a fairly short (2
minute)
> timeout. But I can't remember for the life of me seeing a reference to
the
> info. Can anyone jog my memory?

> Thanks,
> Carl Appellof

by **arkady** » Mon, 07 Oct 2002 19:53:55

Are you do TryEnterCriticalSection() before ?
Arkady

Quote:> I'm running into a process that's catching an exception
> STATUS_POSSIBLE_DEADLOCK. This seems to be raised by ntdll when an
attempt
> to acquire a Critical Section (EnterCriticalSection) times out. On the
> system I'm looking at, there are a couple of interesting internal settings
> in ntdll. These are:

> RtlpTimoutDisable (set to 0)
> RtlpTimeout (set to a large negative number representing 2 minutes worth
of
> time.)

> My process is designed to wait forever to acquire a critical section, and
> croaks after 2 minutes with the above settings.

> On another system, I see these parameters set to:

> RtlpTimoutDisable (set to 1)
> RtlpTimeout (set to a large negative number representing 30 days.)

> I vaguely recall a setup to debug CriticalSection deadlocks that modifies
> the "normal" values for these parameters, setting a fairly short (2
minute)
> timeout. But I can't remember for the life of me seeing a reference to
the
> info. Can anyone jog my memory?

> Thanks,
> Carl Appellof

by **Carl Appello** » Wed, 09 Oct 2002 03:33:51

Thanks to Pavel for the pointer to the registry setting.

My problems started after installing Windows 2000 SP3, which seems to have
changed my CriticalSectionTimeout from 30 days to 2 minutes! Has anyone
else seen this change after installing W2K SP3 ?

Carl

> The timeout is set here:

> Windows Registry Editor Version 5.00
> [HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager]
> "CriticalSectionTimeout"=dword:00278d00

> The value is in seconds.

> > I'm running into a process that's catching an exception
> > STATUS_POSSIBLE_DEADLOCK. This seems to be raised by ntdll when an
> attempt
> > to acquire a Critical Section (EnterCriticalSection) times out. On the
> > system I'm looking at, there are a couple of interesting internal
settings
> > in ntdll. These are:

> > RtlpTimoutDisable (set to 0)
> > RtlpTimeout (set to a large negative number representing 2 minutes worth
> of
> > time.)

> > My process is designed to wait forever to acquire a critical section,
and
> > croaks after 2 minutes with the above settings.

> > On another system, I see these parameters set to:

> > RtlpTimoutDisable (set to 1)
> > RtlpTimeout (set to a large negative number representing 30 days.)

> > I vaguely recall a setup to debug CriticalSection deadlocks that
modifies
> > the "normal" values for these parameters, setting a fairly short (2
> minute)
> > timeout. But I can't remember for the life of me seeing a reference to
> the
> > info. Can anyone jog my memory?

> > Thanks,
> > Carl Appellof

Very Computer

Critical Section timeout

Critical Section timeout

Critical Section timeout

Critical Section timeout

Critical Section timeout