Re: segfault in ffi_data_to_code_pointer

[Anthony Green <green@redhat.com>]

DJ Delorie <dj@redhat.com> writes:

> Anthony Green <green@redhat.com> writes:
>> No, they should be using ffi_closure_alloc().
>>
>> Can you point me at the ruby code that does this?
>
> ruby 2.6.3
> ext/fiddle/closure.c

Thanks.  This looks right to me.  Perhaps they wrote this before
ffi_closure_alloc() existed?  I don't remember when every bit was
introduced...   Looks like you need to follow up with the ruby people.

Thanks DJ,

AG


> In the failing case, USE_FFI_CLOSURE_ALLOC is not set
>
> static VALUE
> allocate(VALUE klass)
> {
>     fiddle_closure * closure;
>
>     VALUE i = TypedData_Make_Struct(klass, fiddle_closure,
>             &closure_data_type, closure);
>
> fprintf (stderr, "DJ: allocate\n");
> #if USE_FFI_CLOSURE_ALLOC
>     closure->pcl = ffi_closure_alloc(sizeof(ffi_closure), &closure->code);
> #else
>     closure->pcl = mmap(NULL, sizeof(ffi_closure), PROT_READ | PROT_WRITE,
>         MAP_ANON | MAP_PRIVATE, -1, 0);
> #endif
>
>     return i;
> }
>
> initialize()
> {
>  . . .
> #if USE_FFI_CLOSURE_ALLOC
>     result = ffi_prep_closure_loc(pcl, cif, callback,
>                 (void *)self, cl->code);
> #else
>     result = ffi_prep_closure(pcl, cif, callback, (void *)self);
>     cl->code = (void *)pcl;
>     i = mprotect(pcl, sizeof(*pcl), PROT_READ | PROT_EXEC);
>     if (i) {
>         rb_sys_fail("mprotect");
>     }
> #endif
>  . . .
> }