Oh it was a feeble joke. My experience is that they do not really serve a
useful purpose in most cases, are prone to the extra release problem, and
are easily replaced by other mechanisms. That doesn’t mean that there aren’t
good uses for semaphores, just that most of the usage I’ve seen has been
wrong, buggy, and not needed.
What exactly was the bugcheck? I still think that raising a software
exception, in and of itself, need not cause a bugcheck at dispatch level.
-----Original Message-----
From: Steve Whitman [mailto:xxxxx@cognex.com]
Sent: Thursday, December 20, 2001 2:34 PM
To: NT Developers Interest List
Subject: [ntdev] RE: KeReleaseSemaphore called in a DPC
I orignally thought that exception handling would work at DISPATCH_LEVEL but
I recently tracked a BSOD to a call to KeReleaseSemaphore(). The exception
code was STATUS_SEMAPHORE_COUNT_EXCEEDED. The KeReleaseSemaphore() was
surronded by a try/except handler. The stack trace showed that
ExRaiseStatus() was called prior to the blue screen.
KeReleaseSemaphore() calls ExRaiseStatus() but that routine is not supposed
to be called at an elevated IRQL. Is NT breaking it’s own rules?
On a slightly different note, why do you consider semaphores in NT “BAD
EVIL-DOERS”?
-----Original Message-----
From: Roddy, Mark [mailto:xxxxx@stratus.com]
Sent: Thursday, December 20, 2001 1:03 PM
To: NT Developers Interest List
Subject: [ntdev] RE: KeReleaseSemaphore called in a DPC
Well just to muddy the waters I think I’ll make the claim
that exception
handling works at DISPATCH_LEVEL and in a dpc routine (both
or either.) So,
while semaphores in nt are BAD EVIL-DOERS, and while one
should actually
never exceed the release count barrier, one could put an
exception handler
around the release and actually manage to not bugcheck in
one’s dpc routine.
Anyone think otherwise?
-----Original Message-----
From: Steve Whitman [mailto:xxxxx@cognex.com]
Sent: Thursday, December 20, 2001 11:59 AM
To: NT Developers Interest List
Subject: [ntdev] RE: KeReleaseSemaphore called in a DPC
I believe that exceptions will cause a bugcheck.
ExRaiseStatus() is called
to generate an exception and the DDK states that you can only
call this
function if running at IRQL PASSIVE_LEVEL I would conclude
that exceptions
will cause a bugcheck.
> -----Original Message-----
> From: Jason LaVoie [mailto:lavoie@us.ibm.com]
> Sent: Thursday, December 20, 2001 11:22 AM
> To: NT Developers Interest List
> Subject: [ntdev] RE: KeReleaseSemaphore called in a DPC
>
>
>
> I’ll rephrase the question. Will exceptions cause a Bugcheck at
> Dispatch if they are handled correctly? If not, then the answer to
> this is, catch
> the exception.
>
>
> -----Original Message-----
> From: Steve Whitman [mailto:xxxxx@cognex.com]
> Sent: Thursday, December 20, 2001 10:27 AM
> To: NT Developers Interest List
> Subject: [ntdev] KeReleaseSemaphore called in a DPC
>
>
> Am I correct that this call is potentially dangerous to make within
> a DPC? The DDK documentation states that the function can be called
> at IRQL <=
> DISPATCH_LEVEL provided that the Wait parameter is set to
> FALSE. However
> it
> also states that if the resulting value is greater than the
> limit of the
> semaphore then an exception, STATUS_SEMAPHORE_COUNT_EXCEEDED,
> is raised.
> If
> this occurs within a DPC (or when IRQL == DISPATCH_LEVEL) I
> would expect
> that result to be a BSOD.
>
> - Steve -
>
>
>
>
>
> —
> You are currently subscribed to ntdev as: xxxxx@cognex.com To
> unsubscribe send a blank email to leave-ntdev-$subst(‘Recip.MemberIDChar’)@lists.osr.com
>
You are currently subscribed to ntdev as: xxxxx@stratus.com To
unsubscribe send a blank email to leave-ntdev-$subst(‘Recip.MemberIDChar’)@lists.osr.com
You are currently subscribed to ntdev as: xxxxx@cognex.com To
unsubscribe send a blank email to leave-ntdev-$subst(‘Recip.MemberIDChar’)@lists.osr.com
You are currently subscribed to ntdev as: xxxxx@stratus.com To
unsubscribe send a blank email to leave-ntdev-$subst(‘Recip.MemberIDChar’)@lists.osr.com
You are currently subscribed to ntdev as: $subst(‘Recip.EmailAddr’)
To unsubscribe send a blank email to leave-ntdev-$subst(‘Recip.MemberIDChar’)@lists.osr.com
Of course the count ought never