From: "Rich" 

This is a multi-part message in MIME format.

------=_NextPart_000_03A6_01C5C122.5341FED0
Content-Type: text/plain;
        charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

   Zombies are not what I meant.  Zombies as they exist in unix do not =
exist in Windows.

   The behavior I am describing is that you can't terminate a process or =
thread at an arbitrary location in kernel mode.  This is because the =
kernel mode state must remain in a consistent state.  In user mode this =
doesn't matter because if you are killing a process with prejudice you =
explicitly do not care about user mode state.  Anyway, in order to keep =
kernel mode state consistent threads only terminate at known good = points.
 If the thread is blocked in kernel mode at a point that it can = not be
woken in order to be terminated cleanly, the termination needs to = be
defered until such a point is reached.  This is not a zombie at least = as
the reference to which you linked uses the term..  This is also not =
specific to Windows.

Rich

  "Tony Williams"  wrote in message =
news:4335ccdd{at}w3.nls.net...
  Frank Haber wrote:
  >> The process is killed and will terminate when it is unblocked.
  >=20
  >=20
  > What is this, some sort of Zen Kill, then?  Do I have to redefine =
"kill"=20
  > as a concept?  It's not dead, Jim?
  >=20
  >=20
  > -zombie

  Exactly. They're called zombie processes and they can be a real =
nuisance.

  http://www.losurs.org/docs/zombies has a good explanation.

  --=20
  Tony
------=_NextPart_000_03A6_01C5C122.5341FED0
Content-Type: text/html;
        charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable








   Zombies
are not what I =
meant. =20
Zombies as they exist in unix do not exist in Windows.
 
   The behavior I am =
describing is that=20
you can't terminate a process or thread at an arbitrary location in = kernel=20
mode.  This is because the kernel mode state must remain in a =
consistent=20
state.  In user mode this doesn't matter because if you are killing = a=20
process with prejudice you explicitly do not care about user mode =
state. =20
Anyway, in order to keep kernel mode state consistent threads only = terminate at=20
known good points.  If the thread is blocked in kernel mode at a =
point that=20
it can not be woken in order to be terminated cleanly, the termination = needs to=20
be defered until such a point is reached.  This is not a zombie at
= least as=20
the reference to which you linked uses the term..  This is =
also not=20
specific to Windows.
 
Rich
 

  "Tony Williams" <tonyw{at}blarg.net>">mailto:tonyw{at}blarg.net">tonyw{at}blarg.net>
wrote in =
message news:4335ccdd{at}w3.nls.net...Fr=
ank=20
  Haber wrote:>> The process is killed and will
terminate when =
it is=20
  unblocked.> > > What
is this, some sort of Zen =
Kill,=20
  then?  Do I have to redefine "kill" >
as a concept?  =
It's not=20
  dead, Jim?> > >
-zombieExactly. They're =
called=20
  zombie processes and they can be a real nuisance.http://www.losurs.org/docs/zo" target="new">http://www.losurs.org/docs/zo=">http://www.losurs.org/docs/zombies">http://www.losurs.org/docs/zo=
mbies=20
  has a good explanation.--
Tony

------=_NextPart_000_03A6_01C5C122.5341FED0--

--- BBBS/NT v4.01 Flag-5