Re: Robust futexes: lost wakeups and design flaws in the glibc/kernel synchronization scheme

[Torvald Riegel]

On Thu, 2017-01-12 at 10:39 +0100, Torvald Riegel wrote:
> On Sat, 2016-12-24 at 17:01 +0100, Torvald Riegel wrote:
> > === Robust mutexes have bugs, in both glibc and the kernel
> > 
> > I've been reviewing the implementation of robust mutexes in both glibc
> > and the kernel support code recently because there are several bug
> > reports, for example:
> > https://bugzilla.redhat.com/show_bug.cgi?id=1401665
> > https://sourceware.org/bugzilla/show_bug.cgi?id=19402
> > https://sourceware.org/bugzilla/show_bug.cgi?id=14485
> > 
> > This review revealed a bunch of bugs.  I have committed/proposed patches
> > that fix all glibc-only bugs that I am aware of:
> > https://sourceware.org/ml/libc-alpha/2016-12/msg00587.html
> > https://sourceware.org/ml/libc-alpha/2016-12/msg00862.html
> > https://sourceware.org/ml/libc-alpha/2016-12/msg00863.html
> > https://sourceware.org/ml/libc-alpha/2016-12/msg00947.html

These patches are now all committed to glibc master.  Thus, you can
start fixing/testing now and have a high probability that you won't run
into glibc bugs.