On Mon, 2010-11-29 at 19:51 +0300, Alexey Zaytsev wrote:

[rewriting history!]
quoted text
> struct fanotify_event_metadata {
>        __u32 event_len; /* Including the options */
>        __u8 vers;
>        __u8 options_offset; /* Aka header length */
>        __u16 reserved;
>        __aligned_u64 mask;
>        __s32 fd;
>        __s32 pid;
>        /* Options go here. */
> };

quoted text
> and let's make both vers and options_offset u8, just in case we need the
> other 2 bytes in the future:

So the last discussion is around vers and options_offset.

Alexy:				Tvrtko:
__u8 vers;			__u16 vers;
__u8 options_offset;		__u16 options_offset;
__u16 unused;

The only type of long option that first comes to mind is a filename.   A
filename could easily blow out the __u8 options_offset.  I probably
shouldn't put that in the event_metadata since it wouldn't be fixed
length and it wouldn't allow fixed offset extention of the
event_metadata, so maybe I shouldn't worry about it.  I'm trying to
think of reasons why __u8 isn't adequate other than my usual "just make
it bigger than we ever need"

You'll notice I'm using __u64 for the mask, even though we don't come
close to filling up an __u32 at this point.

Even though I can't think of a likely reason __u8 is bad I think I like
the 'Tvrtko' option better.  I think we should do a compromise:

Eric:
__u8 vers;
__u8 unused;
__u16 options_offset;

If we ever overload vers we can expand into another field.  Would could
change unused into vers2 and define the version as vers + vers2;  vers2
could even exist somewhere else in the metadata.  That can be done once
we get to version 254 while maintaining backwards compat beyond 255.  If
we ever overflow options_offset we are screwed since old userspace
wouldn't know how to handle things.

So, if you want to send me a patch that implements the above (along with
the obvious version bump to 3, I'll queue it up for this merge window
even though we have an ABI compatible solution.

-Eric

--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/