On Sat, Oct 28, 2017 at 12:18:47AM +0900, Akira Yokosawa wrote:
> On 2017/10/27 18:33, Yubin Ruan wrote:
> > And here are some more modification to some wording in chapter 7, but I am not
> > sure whether you like it or not.
> >
> > Anyway, chapter 7 makes me feel good ;-) It makes me know that home-brewing a
> > lock primitives with atomic instructions(which is what I was doing) is
> > something that is possible and used in production.
> >
> > Thanks,
> > Yubin
> >
> > diff --git a/locking/locking.tex b/locking/locking.tex
> > index 14db27d..a9f46f1 100644
> > --- a/locking/locking.tex
> > +++ b/locking/locking.tex
> > @@ -2166,8 +2166,8 @@ Signal-handler deadlocks can be explicitly avoided as follows:
> > of a signal handler.
> > \item If the application invokes the library function
> > while holding a lock acquired within a given signal
> > - handler, then that signal must be blocked every time that the
> > - library function is called outside of a signal handler.
> > + handler, then that signal must be blocked every time that lock
> > + is to be acquired outside of a signal handler.
>
> The talking point here is library function and signal handler. So something like:
>
> + handler, then that signal must be blocked every time that one
> + of related library functions is called outside of the signal handler.
>
> Thoughts?
I believe that the major problem is that I didn't really specify the
situation and constraints. How about the following?
Thanx, Paul
-----------------------------------------------------------------------
commit 85025eaceff8cf10c9b674ba70d33dbd012e27ce
Author: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
Date: Fri Oct 27 16:21:33 2017 -0700
locking: Clarify explicit avoidance of signal-handler deadlocks
Reported-by: Yubin Ruan <ablacktshirt@xxxxxxxxx>
Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
diff --git a/locking/locking.tex b/locking/locking.tex
index 14db27d1148b..025320b294ed 100644
--- a/locking/locking.tex
+++ b/locking/locking.tex
@@ -2157,7 +2157,17 @@ locks, the strategies in the next section may prove helpful.
\subsubsection{Explicitly Avoid Signal-Handler Deadlocks}
\label{sec:locking:Explicitly Avoid Signal-Handler Deadlocks}
-Signal-handler deadlocks can be explicitly avoided as follows:
+Suppose that a given library function is known to acquire locks,
+but does not block signals.
+Suppose further that it is necessary to invoke that function both from
+within and outside of a signal handler, and that it is not permissible
+to modify this library function.
+Of course, if no special action is taken, then if a signal arrives
+while that library function is holding its lock, deadlock can occur
+when the signal handler invokes that same library function,
+which in turn attempts to re-acquire that same lock.
+
+Such deadlocks can be avoided as follows:
\begin{enumerate}
\item If the application invokes the library function from
--
To unsubscribe from this list: send the line "unsubscribe perfbook" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
