[PATCH] SYS_futex FUTEX_WAIT requires a valid timeout argument

[PATCH] SYS_futex FUTEX_WAIT requires a valid timeout argument

[patrick.oppenlander]

From: Patrick Oppenlander <patrick.oppenlander@gmail.com>

---
 src/internal/pthread_impl.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
index fc2def63..97522bfe 100644
--- a/src/internal/pthread_impl.h
+++ b/src/internal/pthread_impl.h
@@ -155,8 +155,8 @@ static inline void __wake(volatile void *addr, int cnt, int priv)
 static inline void __futexwait(volatile void *addr, int val, int priv)
 {
 	if (priv) priv = FUTEX_PRIVATE;
-	__syscall(SYS_futex, addr, FUTEX_WAIT|priv, val) != -ENOSYS ||
-	__syscall(SYS_futex, addr, FUTEX_WAIT, val);
+	__syscall(SYS_futex, addr, FUTEX_WAIT|priv, val, 0) != -ENOSYS ||
+	__syscall(SYS_futex, addr, FUTEX_WAIT, val, 0);
 }
 
 void __acquire_ptc(void);
-- 
2.17.0

Re: [PATCH] SYS_futex FUTEX_WAIT requires a valid timeout argument

[Rich Felker]

On Fri, Jun 01, 2018 at 10:49:20AM +1000, patrick.oppenlander@gmail.com wrote:
> From: Patrick Oppenlander <patrick.oppenlander@gmail.com>
> 
> ---
>  src/internal/pthread_impl.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
> index fc2def63..97522bfe 100644
> --- a/src/internal/pthread_impl.h
> +++ b/src/internal/pthread_impl.h
> @@ -155,8 +155,8 @@ static inline void __wake(volatile void *addr, int cnt, int priv)
>  static inline void __futexwait(volatile void *addr, int val, int priv)
>  {
>  	if (priv) priv = FUTEX_PRIVATE;
> -	__syscall(SYS_futex, addr, FUTEX_WAIT|priv, val) != -ENOSYS ||
> -	__syscall(SYS_futex, addr, FUTEX_WAIT, val);
> +	__syscall(SYS_futex, addr, FUTEX_WAIT|priv, val, 0) != -ENOSYS ||
> +	__syscall(SYS_futex, addr, FUTEX_WAIT, val, 0);
>  }
>  
>  void __acquire_ptc(void);
> -- 
> 2.17.0

This looks correct. Did you hit things that broke because it was
missing?

Rich

Re: [PATCH] SYS_futex FUTEX_WAIT requires a valid timeout argument

[Patrick Oppenlander]

[-- Attachment #1: Type: text/plain, Size: 1309 bytes --]

On Fri., 1 Jun. 2018, 11:01 Rich Felker, <dalias@libc.org> wrote:

> On Fri, Jun 01, 2018 at 10:49:20AM +1000, patrick.oppenlander@gmail.com
> wrote:
> > From: Patrick Oppenlander <patrick.oppenlander@gmail.com>
> >
> > ---
> >  src/internal/pthread_impl.h | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
> > index fc2def63..97522bfe 100644
> > --- a/src/internal/pthread_impl.h
> > +++ b/src/internal/pthread_impl.h
> > @@ -155,8 +155,8 @@ static inline void __wake(volatile void *addr, int
> cnt, int priv)
> >  static inline void __futexwait(volatile void *addr, int val, int priv)
> >  {
> >       if (priv) priv = FUTEX_PRIVATE;
> > -     __syscall(SYS_futex, addr, FUTEX_WAIT|priv, val) != -ENOSYS ||
> > -     __syscall(SYS_futex, addr, FUTEX_WAIT, val);
> > +     __syscall(SYS_futex, addr, FUTEX_WAIT|priv, val, 0) != -ENOSYS ||
> > +     __syscall(SYS_futex, addr, FUTEX_WAIT, val, 0);
> >  }
> >
> >  void __acquire_ptc(void);
> > --
> > 2.17.0
>
> This looks correct. Did you hit things that broke because it was
> missing?
>
> Rich
>

Syscall argument validation in my nommu kernel caught it. I'm sure Linux
would catch it too.

I didn't prove that it causes observable issues in userspace.

Patrick

>

[-- Attachment #2: Type: text/html, Size: 2229 bytes --]

Re: [PATCH] SYS_futex FUTEX_WAIT requires a valid timeout argument

[Rich Felker]

On Fri, Jun 01, 2018 at 11:23:11AM +1000, Patrick Oppenlander wrote:
> On Fri., 1 Jun. 2018, 11:01 Rich Felker, <dalias@libc.org> wrote:
> 
> > On Fri, Jun 01, 2018 at 10:49:20AM +1000, patrick.oppenlander@gmail.com
> > wrote:
> > > From: Patrick Oppenlander <patrick.oppenlander@gmail.com>
> > >
> > > ---
> > >  src/internal/pthread_impl.h | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
> > > index fc2def63..97522bfe 100644
> > > --- a/src/internal/pthread_impl.h
> > > +++ b/src/internal/pthread_impl.h
> > > @@ -155,8 +155,8 @@ static inline void __wake(volatile void *addr, int
> > cnt, int priv)
> > >  static inline void __futexwait(volatile void *addr, int val, int priv)
> > >  {
> > >       if (priv) priv = FUTEX_PRIVATE;
> > > -     __syscall(SYS_futex, addr, FUTEX_WAIT|priv, val) != -ENOSYS ||
> > > -     __syscall(SYS_futex, addr, FUTEX_WAIT, val);
> > > +     __syscall(SYS_futex, addr, FUTEX_WAIT|priv, val, 0) != -ENOSYS ||
> > > +     __syscall(SYS_futex, addr, FUTEX_WAIT, val, 0);
> > >  }
> > >
> > >  void __acquire_ptc(void);
> > > --
> > > 2.17.0
> >
> > This looks correct. Did you hit things that broke because it was
> > missing?
> >
> > Rich
> >
> 
> Syscall argument validation in my nommu kernel caught it. I'm sure Linux
> would catch it too.
> 
> I didn't prove that it causes observable issues in userspace.

OK. I just like to know if there is a known visible failure users have
seen so I can document it in the commit message. But if nothing was
seen, no problem.

Rich

Re: [PATCH] SYS_futex FUTEX_WAIT requires a valid timeout argument

[Patrick Oppenlander]

On Fri, Jun 1, 2018 at 11:48 AM, Rich Felker <dalias@libc.org> wrote:
> OK. I just like to know if there is a known visible failure users have
> seen so I can document it in the commit message. But if nothing was
> seen, no problem.

Makes sense.

The test I was running (a pthread_barrier test from the open posix
test suite) passed even though the kernel was returning a EFAULT in
that case.

I can reproduce it and give you a backtrace of the failure path if it helps.

Patrick