guile-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: RFC: Foreign objects facility


From: Andy Wingo
Subject: Re: RFC: Foreign objects facility
Date: Mon, 28 Apr 2014 18:08:44 +0200
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/24.3 (gnu/linux)

On Sun 27 Apr 2014 18:00, Mark H Weaver <address@hidden> writes:

> Andy Wingo <address@hidden> writes:
>
>> I propose to provide a new interface that will eventually make SMOBs
>> obsolete.  This new interface is based on structs with raw fields -- the
>> 'u' fields.  (See
>> http://www.gnu.org/software/guile/docs/master/guile.html/Vtables.html#Vtables
>> for description of 'u' fields.  Note that the documentation is wrong --
>> these fields are indeed traced by the GC.)
>
> Sounds like a good idea, in general.

Thanks; with this note, Ludo's approbation, and a doc update I feel OK
with pushing to stable-2.0.  Let's hammer out the details there.

>>     SCM scm_make_foreign_object_1 (SCM type, scm_t_bits val0);
>>     SCM scm_make_foreign_object_2 (SCM type, scm_t_bits val0,
>>                                     scm_t_bits val1);
>>     SCM scm_make_foreign_object_3 (SCM type, scm_t_bits val0,
>>                                     scm_t_bits val1, scm_t_bits val2);
>
> If we include an interface like this, I think we should have more of
> these, but see below.

I changed this interface to take void* pointers, with the idea that
that's the common thing.  Same with scm_foreign_object_ref.  But I added
scm_foreign_object_signed_ref, unsigned_set_x, etc, with notes in the
manual that you can indeed provide fewer initializers than fields, and
then initialize the fields yourself later.

> #ifndef __GNUC__
>   if (bits > SCM_T_SIGNED_BITS_MAX)
>     return -1 - (scm_t_signed_bits) ~bits;
> #endif
>   return (scm_t_signed_bits) bits;
> }

I think I forgot to actually implement this bit though :P

Tutorial-style documentation (in the Programming in C chapter, replacing
the SMOB tutorial) attached.  There is also reference-style
documentation in the API chapter.

Andy

1 Defining New Foreign Object Types
===================================

The "foreign object type" facility is Guile's mechanism for importing
object and types from C or other languages into Guile's system.  If you
have a C 'struct foo' type, for example, you can define a corresponding
Guile foreign object type that allows Scheme code to handle 'struct foo
*' objects.

   To define a new foreign object type, the programmer provides Guile
with some essential information about the type -- what its name is, how
many fields it has, and its finalizer (if any) -- and Guile allocates a
fresh type for it.  Foreign objects can be accessed from Scheme or from
C.

1.1 Defining Foreign Object Types
---------------------------------

To create a new foreign object type from C, call
'scm_make_foreign_object_type'.  It returns a value of type 'SCM' which
identifies the new type.

   Here is how one might declare a new type representing eight-bit
gray-scale images:

     #include <libguile.h>

     struct image {
       int width, height;
       char *pixels;

       /* The name of this image */
       SCM name;

       /* A function to call when this image is
          modified, e.g., to update the screen,
          or SCM_BOOL_F if no action necessary */
       SCM update_func;
     };

     static SCM image_type image_type;

     void
     init_image_type (void)
     {
       SCM name, slots;
       scm_t_struct_finalize finalizer;

       name = scm_from_utf8_symbol ("image");
       slots = scm_list_1 (scm_from_utf8_symbol ("data"));
       finalizer = NULL;

       image_type =
         scm_make_foreign_object_type (name, slots, finalizer);
     }

   The result is an initialized 'image_type' value that identifies the
new foreign object type.  The next section describes how to create
foreign objects and how to access their slots.

1.2 Creating Foreign Objects
----------------------------

Foreign objects contain zero or more "slots" of data.  A slot can hold a
pointer, an integer that fits into a 'size_t' or 'ssize_t', or a 'SCM'
value.

   All objects of a given foreign type have the same number of slots.
In the example from the previous section, the 'image' type has one slot,
because the slots list passed to 'scm_make_foreign_object_type' is of
length one.  (The actual names given to slots are unimportant for most
users of the C interface, but can be used on the Scheme side to
introspect on the foreign object.)

   To construct a foreign object and initialize its first slot, call
'scm_make_foreign_object_1 (TYPE, FIRST_SLOT_VALUE)'.  There are
similarly named constructors for initializing 0, 1, 2, or 3 slots, or
initializing N slots via an array.  *Note Foreign Objects::, for full
details.  Any fields that are not explicitly initialized are set to 0.

   To get or set the value of a slot by index, you can use the
'scm_foreign_object_ref' and 'scm_foreign_object_set_x' functions.
These functions take and return values as 'void *' pointers; there are
corresponding convenience procedures like '_signed_ref',
'_unsigned_set_x' and so on for dealing with slots as signed or unsigned
integers.

   Foreign objects fields that are pointers can be tricky to manage.  If
possible, it is best that all memory that is referenced by a foreign
object be managed by the garbage collector.  That way, the GC can
automatically ensure that memory is accessible when it is needed, and
freed when it becomes inaccessible.  If this is not the case for your
program - for example, if you are exposing an object to Scheme that was
allocated by some other, Guile-unaware part of your program - then you
will probably need to implement a finalizer.  *Note Foreign Object
Memory Management::, for more.

   Continuing the example from the previous section, if the global
variable 'image_type' contains the type returned by
'scm_make_foreign_object_type', here is how we could construct a foreign
object whose "data" field contains a pointer to a freshly allocated
'struct image':

     SCM
     make_image (SCM name, SCM s_width, SCM s_height)
     {
       struct image *image;
       int width = scm_to_int (s_width);
       int height = scm_to_int (s_height);

       /* Allocate the `struct image'.  Because we
          use scm_gc_malloc, this memory block will
          be automatically reclaimed when it becomes
          inaccessible, and its members will be traced
          by the garbage collector.  */
       image = (struct image *)
         scm_gc_malloc (sizeof (struct image), "image");

       image->width = width;
       image->height = height;

       /* Allocating the pixels with
          scm_gc_malloc_pointerless means that the
          pixels data is collectable by GC, but
          that GC shouldn't spend time tracing its
          contents for nested pointers because there
          aren't any.  */
       image->pixels =
         scm_gc_malloc_pointerless (width * height, "image pixels");

       image->name = name;
       image->update_func = SCM_BOOL_F;

       /* Now wrap the struct image* in a new foreign
          object, and return that object.  */
       return scm_make_foreign_object_1 (image_type, image);
     }

   We use 'scm_gc_malloc_pointerless' for the pixel buffer to tell the
garbage collector not to scan it for pointers.  Calls to
'scm_gc_malloc', 'scm_make_foreign_object_1', and
'scm_gc_malloc_pointerless' raise an exception in out-of-memory
conditions; the garbage collector is able to reclaim previously
allocated memory if that happens.

1.3 Type Checking of Foreign Objects
------------------------------------

Functions that operate on foreign objects should check that the passed
'SCM' value indeed is of the correct type before accessing its data.
They can do this with 'scm_assert_foreign_object_type'.

   For example, here is a simple function that operates on an image
object, and checks the type of its argument.

     SCM
     clear_image (SCM image_obj)
     {
       int area;
       struct image *image;

       scm_assert_foreign_object_type (image_type, image_obj);

       image = scm_foreign_object_ref (image_obj, 0);
       area = image->width * image->height;
       memset (image->pixels, 0, area);

       /* Invoke the image's update function.  */
       if (scm_is_true (image->update_func))
         scm_call_0 (image->update_func);

       return SCM_UNSPECIFIED;
     }

1.4 Foreign Object Memory Management
------------------------------------

Once a foreign object has been released to the tender mercies of the
Scheme system, it must be prepared to survive garbage collection.  In
the example above, all the memory associated with the foreign object is
managed by the garbage collector because we used the 'scm_gc_'
allocation functions.  Thus, no special care must be taken: the garbage
collector automatically scans them and reclaims any unused memory.

   However, when data associated with a foreign object is managed in
some other way--e.g., 'malloc''d memory or file descriptors--it is
possible to specify a "finalizer" function to release those resources
when the foreign object is reclaimed.

   As discussed in *note Garbage Collection::, Guile's garbage collector
will reclaim inaccessible memory as needed.  This reclamation process
runs concurrently with the main program.  When Guile analyzes the heap
and determines that an object's memory can be reclaimed, that memory is
put on a "free list" of objects that can be reclaimed.  Usually that's
the end of it--the object is available for immediate re-use.  However
some objects can have "finalizers" associated with them--functions that
are called on reclaimable objects to effect any external cleanup
actions.

   Finalizers are tricky business and it is best to avoid them.  They
can be invoked at unexpected times, or not at all--for example, they are
not invoked on process exit.  They don't help the garbage collector do
its job; in fact, they are a hindrance.  Furthermore, they perturb the
garbage collector's internal accounting.  The GC decides to scan the
heap when it thinks that it is necessary, after some amount of
allocation.  Finalizable objects almost always represent an amount of
allocation that is invisible to the garbage collector.  The effect can
be that the actual resource usage of a system with finalizable objects
is higher than what the GC thinks it should be.

   All those caveats aside, some foreign object types will need
finalizers.  For example, if we had a foreign object type that wrapped
file descriptors--and we aren't suggesting this, as Guile already has
ports --then you might define the type like this:

     static SCM file_type;

     static void
     finalize_file (SCM file)
     {
       int fd = scm_foreign_object_signed_ref (file, 0);
       if (fd >= 0)
         {
           scm_foreign_object_signed_set_x (file, 0, -1);
           close (fd);
         }
     }

     static void
     init_file_type (void)
     {
       SCM name, slots;
       scm_t_struct_finalize finalizer;

       name = scm_from_utf8_symbol ("file");
       slots = scm_list_1 (scm_from_utf8_symbol ("fd"));
       finalizer = finalize_file;

       image_type =
         scm_make_foreign_object_type (name, slots, finalizer);
     }

     static SCM
     make_file (int fd)
     {
       return scm_make_foreign_object_1 (file_type, (void *) fd);
     }

   Note that the finalizer may be invoked in ways and at times you might
not expect.  In particular, if the user's Guile is built with support
for threads, the finalizer may be called from any thread that is running
Guile.  In Guile 2.0, finalizers are invoked via "asyncs", which
interleaves them with running Scheme code; *note System asyncs::.  In
Guile 2.2 there will be a dedicated finalization thread, to ensure that
the finalization doesn't run within the critical section of any other
thread known to Guile.

   In either case, finalizers run concurrently with the main program,
and so they need to be async-safe and thread-safe.  If for some reason
this is impossible, perhaps because you are embedding Guile in some
application that is not itself thread-safe, you have a few options.  One
is to use guardians instead of finalizers, and arrange to pump the
guardians for finalizable objects.  *Note Guardians::, for more
information.  The other option is to disable automatic finalization
entirely, and arrange to call 'scm_run_finalizers ()' at appropriate
points.  *Note Foreign Objects::, for more on these interfaces.

   Finalizers are allowed to allocate memory, access GC-managed memory,
and in general can do anything any Guile user code can do.  This was not
the case in Guile 1.8, where finalizers were much more restricted.  In
particular, in Guile 2.0, finalizers can resuscitate objects.  We do not
recommend that users avail themselves of this possibility, however, as a
resuscitated object can re-expose other finalizable objects that have
been already finalized back to Scheme.  These objects will not be
finalized again, but they could cause use-after-free problems to code
that handles objects of that particular foreign object type.  To guard
against this possibility, robust finalization routines should clear
state from the foreign object, as in the above 'free_file' example.

   One final caveat.  Foreign object finalizers are associated with the
lifetime of a foreign object, not of its fields.  If you access a field
of a finalizable foreign object, and do not arrange to keep a reference
on the foreign object itself, it could be that the outer foreign object
gets finalized while you are working with its field.

   For example, consider a procedure to read some data from a file, from
our example above.

     SCM
     read_bytes (SCM file, SCM n)
     {
       int fd;
       SCM buf;
       size_t len, pos;

       scm_assert_foreign_object_type (file_type, file);

       fd = scm_foreign_object_signed_ref (file, 0);
       if (fd < 0)
         scm_wrong_type_arg_msg ("read-bytes", SCM_ARG1,
                                 file, "open file");

       len = scm_to_size_t (n);
       SCM buf = scm_c_make_bytevector (scm_to_size_t (n));

       pos = 0;
       while (pos < len)
         {
           char *bytes = SCM_BYTEVECTOR_CONTENTS (buf);
           ssize_t count = read (fd, bytes + pos, len - pos);
           if (count < 0)
             scm_syserror ("read-bytes");
           if (count == 0)
             break;
           pos += count;
         }

       scm_remember_upto_here_1 (file);

       return scm_values (scm_list_2 (buf, scm_from_size_t (pos)));
     }

   After the prelude, only the 'fd' value is used and the C compiler has
no reason to keep the 'file' object around.  If 'scm_c_make_bytevector'
results in a garbage collection, 'file' might not be on the stack or
anywhere else and could be finalized, leaving 'read' to read a closed
(or, in a multi-threaded program, possibly re-used) file descriptor.
The use of 'scm_remember_upto_here_1' prevents this, by creating a
reference to 'file' after all data accesses.  *Note Garbage Collection
Functions::.

   'scm_remember_upto_here_1' is only needed on finalizable objects,
because garbage collection of other values is invisible to the program -
it happens when needed, and is not observable.  But if you can, save
yourself the headache and build your program in such a way that it
doesn't need finalization.

1.5 Foreign Objects and Scheme
------------------------------

It is also possible to create foreign objects and object types from
Scheme, and to access fields of foreign objects from Scheme.  For
example, the file example from the last section could be equivalently
expressed as:

     (define-module (my-file)
       #:use-module (system foreign-object)
       #:use-module ((oop goops) #:select (make))
       #:export (make-file))

     (define (finalize-file file)
       (let ((fd (struct-ref file 0)))
         (unless (< fd 0)
           (struct-set! file 0 -1)
           (close-fdes fd))))

     (define <file>
       (make-foreign-object-type '<file> '(fd)
                                 #:finalizer finalize-file))

     (define (make-file fd)
       (make <file> #:fd fd))

   Here we see that the result of 'make-foreign-object-type', which is
the equivalent of 'scm_make_foreign_object_type', is a struct vtable.
*Note Vtables::, for more information.  To instantiate the foreign
object, which is really a Guile struct, we use 'make'.  (We could have
used 'make-struct/no-tail', but as an implementation detail, finalizers
are attached in the 'initialize' method called by 'make').  To access
the fields, we use 'struct-ref' and 'struct-set!'.  *Note Structure
Basics::.

   There is a convenience syntax, 'define-foreign-object-type', that
defines a type along with a constructor, and getters for the fields.  An
appropriate invocation of 'define-foreign-object-type' for the file
object type could look like this:

     (use-modules (system foreign-object))

     (define-foreign-object-type <file>
       make-file
       (fd)
       #:finalizer finalize-file)

   This defines the '<file>' type with one field, a 'make-file'
constructor, and a getter for the 'fd' field, bound to 'fd'.

   Foreign object types are not only vtables but are actually GOOPS
classes, as hinted at above.  *Note GOOPS::, for more on Guile's
object-oriented programming system.  Thus one can define print and
equality methods using GOOPS:

     (use-modules (oop goops))

     (define-method (write (file <file>) port)
       ;; Assuming existence of the `fd' getter
       (format port "#<<file> ~a>" (fd file)))

     (define-method (equal? (a <file>) (b <file>))
       (eqv? (fd a) (fd b)))

   One can even sub-class foreign types.

     (define-class <named-file> (<file>)
       (name #:init-keyword #:name #:init-value #f #:accessor name))

   The question arises of how to construct these values, given that
'make-file' returns a plain old '<file>' object.  It turns out that you
can use the GOOPS construction interface, where every field of the
foreign object has an associated initialization keyword argument.

     (define* (my-open-file name #:optional (flags O_RDONLY))
       (make <named-file> #:fd (open-fdes name flags) #:name name))

     (define-method (write (file <named-file>) port)
       (format port "#<<file> ~s ~a>" (name file) (fd file)))

   *Note Foreign Objects::, for full documentation on the Scheme
interface to foreign objects.  *Note GOOPS::, for more on GOOPS.

   As a final note, you might wonder how this system supports
encapsulation of sensitive values.  First, we have to recognize that
some facilities are essentially unsafe and have global scope.  For
example, in C, the integrity and confidentiality of a part of a program
is at the mercy of every other part of that program - because any part
of the program can read and write anything in its address space.  At the
same time, principled access to structured data is organized in C on
lexical boundaries; if you don't expose accessors for your object, you
trust other parts of the program not to work around that barrier.

   The situation is not dissimilar in Scheme.  Although Scheme's unsafe
constructs are fewer in number than in C, they do exist.  The '(system
foreign)' module can be used to violate confidentiality and integrity,
and shouldn't be exposed to untrusted code.  Although 'struct-ref' and
'struct-set!' are less unsafe, they still have a cross-cutting
capability of drilling through abstractions.  Performing a 'struct-set!'
on a foreign object slot could cause unsafe foreign code to crash.
Ultimately, structures in Scheme are capabilities for abstraction, and
not abstractions themselves.

   That leaves us with the lexical capabilities, like constructors and
accessors.  Here is where encapsulation lies: the practical degree to
which the innards of your foreign objects are exposed is the degree to
which their accessors are lexically available in user code.  If you want
to allow users to reference fields of your foreign object, provide them
with a getter.  Otherwise you should assume that the only access to your
object may come from your code, which has the relevant authority, or via
code with access to cross-cutting 'struct-ref' and such, which also has
the cross-cutting authority.
-- 
http://wingolog.org/

reply via email to

[Prev in Thread] Current Thread [Next in Thread]