Re: sem_init() fails (when used in a certain way)

[Corinna Vinschen <corinna-cygwin at cygwin dot com>]

On Mar 29 13:24, Christopher Faylor wrote:
> On Tue, Mar 29, 2011 at 05:53:31PM +0200, Corinna Vinschen wrote:
> >On Mar 29 14:41, Jon TURNEY wrote:
> >> $ ./sem_init_malloc_testcase
> >> sem_init: Device or resource busy
> >> [...]
> >> I'm not sure how to fix this:
> >> 
> >> Changing sem_t from a pointer to an instance of class semaphore is not a good
> >> idea as it would change a lot of code, and a non-starter as it breaks ABI by
> >> changing sizeof(sem_t), and I have to assume there is a reason it was
> >> implemented using a pointer in the first place.
> >> 
> >> Removing the is_good_object() check from semaphore::init() (and thus changing
> >> the undefined behaviour when a sem_init() is used twice from 'return EBUSY' to
> >> 'leak some memory') would work.  Perhaps downgrading the error to strace
> >> output "potential repeated semaphore initialization"?
> >
> >This sounds like a good idea to me.  Given that the test can accidentally
> >identify the content of the semaphore as valid, the test is somewhat
> >dangerous.
> >
> >> I hope someone has some better ideas?
> >
> >I don't think there's any other way.  Glibc does not check the semaphore
> >storage at all when calling sem_init and SUSv4 states
> >
> >  "Attempting to initialize an already initialized semaphore results in
> >   undefined behavior."
> >
> >I'd say, just go ahead.
> 
> I think we should put a
> 
>  myfault efault;
>  if (efault.faulted ())
>     ...
> 
> in place of the is_good_object() test and sprinkle those throughout the
> other sem_* functions, if they're not already there.

You can't just replace all is_good_object tests with myfault handlers,
afaics.  The only case where the is_good_object test doesn't make sense
for the reason outlined in Jon's mail are the init methods of the
various object types.  In all other methods the is_good_object test is
still necessary to check the object pointer and to generate the EINVAL
error code reliably.  So the myfault handler could (and probably should)
be added to these methods while keeping the is_good_object test.


Corinna

-- 
Corinna Vinschen                  Please, send mails regarding Cygwin to
Cygwin Project Co-Leader          cygwin AT cygwin DOT com
Red Hat