Re: [PATCH 03/12] DRBD: bitmap

[Philipp Reisner <philipp.reisner@linbit.com>]

On Wednesday 08 April 2009 12:16:11 Nikanth K wrote:
> On Mon, Mar 23, 2009 at 9:17 PM, Philipp Reisner
>
> <philipp.reisner@linbit.com> wrote:
> > +/* definition of bits in bm_flags */
> > +#define BM_LOCKED 0
> > +#define BM_MD_IO_ERROR (BITS_PER_LONG-1) /* 31? 63? */
>
> Wonder whether this should be made same for 32-bit as well as 64-bit?
> Especially for x86_64 as the machine can become 32-bit to 64-bit or
> vice-versa after reboot.
>

This is only a bitnumber for an in memory flag word (bm_flags). 
I changed that to 

#define BM_LOCKED       0
#define BM_MD_IO_ERROR  1

What was before was just unnecessary complex.

>
> <snip>
>
> > +#if 0
> > +#define catch_oob_access_start() do {  \
> > +       do {                            \
> > +               if ((bm-p_addr) >= PAGE_SIZE/sizeof(long)) { \
> > +                       printk(KERN_ALERT "drbd_bitmap.c:%u %s: p_addr:%p
> > bm:%p %d\n", \ +                                       __LINE__ ,
> > __func__ , p_addr, bm, (bm-p_addr)); \ +                       break;    
> >      \
> > +               }
> > +#define catch_oob_access_end() \
> > +       } while (0); } while (0)
> > +#else
> > +#define catch_oob_access_start() do {
> > +#define catch_oob_access_end() } while (0)
> > +#endif
> > +
>
> Probably should be changed to be based on a config debug option?

I removed the macro definitions as well as these macro invocations.
They where used to find bugs years ago. Unnecessary nowadays.

> <snip>
>
> > +/*
> > + * since (b->bm_bits % BITS_PER_LONG) != 0,
> > + * this masks out the remaining bits.
> > + * Rerturns the number of bits cleared.
> > + */
> > +STATIC int bm_clear_surplus(struct drbd_bitmap *b)
> > +{
> > +       const unsigned long mask = (1UL << (b->bm_bits &
> > (BITS_PER_LONG-1))) - 1;
>
> Should BM_MD_IO_ERROR be used instead of  (BITS_PER_LONG-1)?
> Or at least this deserves a macro as it is used in many places.
>

Well, BITS_PER_LONG is now used in two functions only (after I
removed that from the definition of BM_MD_IO_ERROR). I will leave
it open coded, as it is now.

Thanks for these comments!

-Phil
-- 
: Dipl-Ing Philipp Reisner
: LINBIT | Your Way to High Availability
: Tel: +43-1-8178292-50, Fax: +43-1-8178292-82
: http://www.linbit.com

DRBD(R) and LINBIT(R) are registered trademarks of LINBIT, Austria.