svn commit: r322969 - in head: sbin/mdconfig sys/dev/md sys/sys

[Maxim Sobolev]

John, OK, maybe you are right and the current status quo was just an
accident. I am curious what do you and other people think about expressing
expected structure size and padding more explicitly instead of trying to
accommodate for sometimes intricate play between alignment and type size
with something like char[N]? I.e. along the following lines:

#if __WORDSIZE < 64
#define MD_IOCTL_LEN    436
#else
#define MD_IOCTL_LEN    448
#endif

struct md_ioctl {
    union {
        struct _md_ioctl_payload {
            unsigned        version;     /* Structure layout version */
            unsigned        unit;        /* unit number */
            enum md_types   type ;       /* type of disk */
            char            *file;       /* pathname of file to mount */
            off_t           mediasize;   /* size of disk in bytes */
            unsigned        sectorsize;  /* sectorsize */
            unsigned        options;     /* options */
            u_int64_t       base;        /* base address */
            int             fwheads;     /* firmware heads */
            int             fwsectors;   /* firmware sectors */
            char            *label;      /* label of the device */
        } md;
        char raw[MD_IOCTL_LEN]; /* payload + padding for future ideas */
    };
};
CTASSERT(sizeof(struct md_ioctl) == MD_IOCTL_LEN);

I've tested that it DTRT on i386 and amd64, need to validate on ARM and
MIPS. The code impact should be minimal and easy to apply, i.e. ptr->md_ywz
becomes ptr->md.xyz.

On Tue, Aug 29, 2017 at 7:30 AM, John Baldwin <jhb at freebsd.org> wrote:

> On Monday, August 28, 2017 04:40:49 PM Maxim Sobolev wrote:
> > John, well, this depends on how you look at it. The padding element size
> is
> > "int", which when you account for the alignment has the nice property on
> > both 32 and 64-bit arches that no matter what kind of element you add
> > (char, short, int or void *), you only need to bring down MDNPAD by 1 to
> > keep the structure size the same.
>
> That's only true once.  Add a second pointer and this trick won't work
> again
> on 64-bit platforms.  It only worked this time by accident because there
> was
> unaccounted-for padding at the end of the structure, not on purpose.  In
> other places in the tree where ABI is more carefully maintained we use
> explicit spare pointers or spare longs instead (e.g. padding in
> kinfo_proc).
>
> --
> John Baldwin
>
>


-- 
Maksym Sobolyev
Sippy Software, Inc.
Internet Telephony (VoIP) Experts
Tel (Canada): +1-778-783-0474
Tel (Toll-Free): +1-855-747-7779
Fax: +1-866-857-6942
Web: http://www.sippysoft.com
MSN: sales at sippysoft.com
Skype: SippySoft