Re: p = kmalloc(sizeof(*p), )

[Denis Vlasenko <vda@ilport.com.ua>]

> > >         p = kmalloc(sizeof(*p), ...)
> > > 
> > >    is not grep-friendly, and can not be used to identify potential
> > >    initialisation sites.  However:
> > > 
> > >         p = kmalloc(sizeof(struct foo), ...)
> > > 
> > >    is grep-friendly, and will lead you to inspect each place where
> > >    such a structure is allocated for correct initialisation.
> > 
> > I would disagree on this one. You can still grep all the places where
> > the local variable is declared in. Furthermore, structs are not always
> > initialized where they're kmalloc'd so you need to manually inspect
> > anyway.
> 
> Think about it some more.  You've added a new member to struct foo.
> You want to fix up all the places which allocate struct foo to
> initialise this new member.  Grepping for 'struct foo' returns 100
> files.  Grepping for kmalloc in those 100 files returns 100 files.
> 
> Do you open all 100 in an editor and manually try and locate the five
> kmalloc instances of this structure, and end up missing some.
> 
> Or do you do the sane thing and use kmalloc(sizeof(struct foo), ...)
> and grep for "kmalloc[[:space:]]*(sizeof[[:space:]]*(struct foo)"
> which returns only the five files and fix those up with knowledge
> that you've found all the instances?

Both are inferior to Alans macro

p = typed_kmalloc(struct foo, ...);

which has greppable struct name, saves typing sizeof() and also
gives you typechecking (fails with "pointers to different types"
if p is not struct foo*).
--
vda