On Sun, 23 Mar 2008, Robert P. J. Day wrote:
> On Sun, 23 Mar 2008, Julia Lawall wrote:
>
> > On Sun, 23 Mar 2008, Robert P. J. Day wrote:
> >
> > >
> > > as i'm currently perusing the implementation of the kernel linked
> > > list, it occurred to me to wonder if there are still parts of the
> > > kernel that insist on implementing their own design unnecessarily
> > > (which they're not supposed to do). anyone know of any such examples,
> > > out of curiosity? thanks.
> >
> > There are a lot of occurrences of structures that have a field of type
> > pointer to the structure, often with the suggestive name "next". But at
> > least some of them have only a next field and no prev field. Are you
> > interested in them anyway?
> >
> > An example is:
> >
> > arch/um/include/aio.h
> >
> > struct aio_context {
> > int reply_fd;
> > struct aio_context *next;
> > };
>
> based on my own grepping, i noticed a zillion of those things -- i'm
> just thinking that some of them are probably local implementations of
> a linked list structure that could be rewritten using the in-kernel
> linked list structure defined in <linux/list.h>.
>
> i doubt that could be done quickly -- it's undoubtedly a long-term
> project that could be transformed little by little as people run
> across examples of it. just something to keep an eye out for if
> you're feeling ambitious.
The question still stands, though. list_head has both next and prev
fields. Is it desirable to convert a singly linked list to a doubly
linked list just for the benefit of using the functions in list.h, at a
cost of more memory?
I found 89 files that declare a structure with multiple self-pointer-typed
fields, and over 400 more that declare a structure with only one
self-pointer field. Perhaps there is a need for an API for singly-linked
lists as well? Or perhaps they are too simple to be worth it?
Another common pattern is a next field and a parent field. Those should
probably not be converted either, for readability, even if doing so would
be correct.
julia
--
To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
