Re: [Xen-devel] [PATCH V2 3/6] [RFC] xen/common: Introduce _xrealloc function

[Jan Beulich <jbeulich@xxxxxxxx>]

On 06.08.2019 21:51, Volodymyr Babchuk wrote:
> Hi Jan,
>
> Jan Beulich writes:
>
> > On 02.08.2019 18:39, Oleksandr Tyshchenko wrote:
> > > --- a/xen/common/xmalloc_tlsf.c
> > > +++ b/xen/common/xmalloc_tlsf.c
> > > @@ -610,6 +610,27 @@ void *_xzalloc(unsigned long size, unsigned long align)
> > >        return p ? memset(p, 0, size) : p;
> > >    }
> > >    
> > > +void *_xrealloc(void *p, unsigned long new_size, unsigned long align)
> > > +{
> > > +    void *new_p;
> > > +
> > > +    if ( !new_size )
> > > +    {
> > > +        xfree(p);
> > > +        return NULL;
> > > +    }
> > > +
> > > +    new_p = _xmalloc(new_size, align);
> > > +
> > > +    if ( new_p && p )
> > > +    {
> > > +        memcpy(new_p, p, new_size);
> > > +        xfree(p);
> > > +    }
> > > +
> > > +    return new_p;
> > > +}
> > While I can see why having a re-allocation function may be handy,
> > explicit / direct use of _xmalloc() and _xzalloc() are discouraged,
> > in favor of the more type-safe underscore-less variants. I can't
> > see though how a type-safe "realloc" could look like, except for
> > arrays. If resizing arrays is all you're after, I'd like to
> > recommend to go that route rather then the suggested one here. If
> > resizing arbitrary objects is the goal, then what you suggest may
> > be the only route, but I'd still be not overly happy to see such
> > added.
> I can see 3 uses for realloc:
>
>   a. re-allocate generic data buffer
>   b. re-allocate array
>   c. re-allocate struct with flexible buffer.
>
> option c. is about structures like this:
>
> struct arrlen
> {
>          size_t len;
>          int data[1];
> };
>
> This is Oleksandr's case.
>
> So for option a. we can use _xreallocate(ptr, size, align)
> For option b. we can use xrealloc_array(_ptr, _type, _num)
> And for option c. I propose to implement the following macro:
>
> #define realloc_flex_struct(_ptr, _type, _field, _len)                        \
>   ((_type *)_xrealloc(_ptr, offsetof(_type, _field[_len]) , __alignof__(_type)))
>
> It can be used in the following way:
>
> newptr = realloc_flex_struct(ptr, struct arrlen, newsize);
>
> As you can see, this approach is type-safe and covers Oleksanrd's case.
This looks fine to me, but then wants to be accompanied by a
similar xmalloc_flex_struct(), which could be used right away
to replace a number of open-coded instances of the above.

There's one more thing for the re-alloc case though (besides
cosmetic aspects): The incoming pointer should also be verified
to be of correct type.

Jan

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel