On Thu, Apr 21, 2022 at 07:40:25AM +0000, David Laight wrote:
> From: Yury Norov
> > Sent: 20 April 2022 23:25
> >
> > Manipulating 64-bit arrays with bitmap functions is potentially dangerous
> > because on 32-bit BE machines the order of halfwords doesn't match. Another
> > issue is that compiler may throw a warning about out-of-boundary access.
> >
> > This patch adds bitmap_{from,to}_arr64 functions in addition to existing
> > bitmap_{from,to}_arr32.
> >
> > Signed-off-by: Yury Norov <yury.norov@xxxxxxxxx>
> > ---
> > include/linux/bitmap.h | 23 +++++++++++++++++----
> > lib/bitmap.c | 47 ++++++++++++++++++++++++++++++++++++++++++
> > 2 files changed, 66 insertions(+), 4 deletions(-)
> >
> > diff --git a/include/linux/bitmap.h b/include/linux/bitmap.h
> > index 10d805c2893c..f78c534fb814 100644
> > --- a/include/linux/bitmap.h
> > +++ b/include/linux/bitmap.h
> > @@ -292,6 +292,24 @@ void bitmap_to_arr32(u32 *buf, const unsigned long *bitmap,
> > (const unsigned long *) (bitmap), (nbits))
> > #endif
> >
> > +/*
> > + * On 64-bit systems bitmaps are represented as u64 arrays internally. On LE32
> > + * machines the order of hi and lo parts of nubmers match the bitmap structure.
> > + * In both cases conversion is not needed when copying data from/to arrays of
> > + * u64.
> > + */
> > +#if (BITS_PER_LONG == 32) && defined(__BIG_ENDIAN)
>
> I think I'd change the condition to (inverting it):
> #if (BITS_PER_LONG == 64) || defined(__LITTLE_ENDIAN)
> since that is the condition when the layout matches.
Sorry, I don't understand about 'layout matches'. Why this way is
better than another?
> > +void bitmap_from_arr64(unsigned long *bitmap, const u64 *buf, unsigned int nbits);
> > +void bitmap_to_arr64(u64 *buf, const unsigned long *bitmap, unsigned int nbits);
> > +#else
> > +#define bitmap_from_arr64(bitmap, buf, nbits) \
> > + bitmap_copy_clear_tail((unsigned long *) (bitmap), \
> > + (const unsigned long *) (buf), (nbits))
> > +#define bitmap_to_arr64(buf, bitmap, nbits) \
> > + bitmap_copy_clear_tail((unsigned long *) (buf), \
> > + (const unsigned long *) (bitmap), (nbits))
> > +#endif
> > +
> > static inline int bitmap_and(unsigned long *dst, const unsigned long *src1,
> > const unsigned long *src2, unsigned int nbits)
> > {
> > @@ -596,10 +614,7 @@ static inline void bitmap_next_set_region(unsigned long *bitmap,
> > */
> > static inline void bitmap_from_u64(unsigned long *dst, u64 mask)
> > {
> > - dst[0] = mask & ULONG_MAX;
> > -
> > - if (sizeof(mask) > sizeof(unsigned long))
> > - dst[1] = mask >> 32;
> > + bitmap_from_arr64(dst, &mask, 64);
> > }
>
> I'd leave this alone.
I'd change it in sake of consistency. Let's see what others say.
> > /**
> > diff --git a/lib/bitmap.c b/lib/bitmap.c
> > index d9a4480af5b9..aea9493f4216 100644
> > --- a/lib/bitmap.c
> > +++ b/lib/bitmap.c
> > @@ -1533,5 +1533,52 @@ void bitmap_to_arr32(u32 *buf, const unsigned long *bitmap, unsigned int nbits)
> > buf[halfwords - 1] &= (u32) (UINT_MAX >> ((-nbits) & 31));
> > }
> > EXPORT_SYMBOL(bitmap_to_arr32);
> > +#endif
> > +
> > +#if (BITS_PER_LONG == 32) && defined(__BIG_ENDIAN)
> > +/**
> > + * bitmap_from_arr64 - copy the contents of u64 array of bits to bitmap
> > + * @bitmap: array of unsigned longs, the destination bitmap
> > + * @buf: array of u64 (in host byte order), the source bitmap
> > + * @nbits: number of bits in @bitmap
> > + */
> > +void bitmap_from_arr64(unsigned long *bitmap, const u64 *buf, unsigned int nbits)
> > +{
> > + while (nbits > 0) {
>
> This looks like a for look to me...
Can you explain why for() is better then while() here? Is generated
code better, or something else?
> > + u64 val = *buf++;
> > +
> > + *bitmap++ = (unsigned long)val;
> > + if (nbits > 32)
> > + *bitmap++ = (unsigned long)(val >> 32);
>
> No need for either cast.
Yep, thanks.
> > + nbits -= 64;
> > + }
> >
> > + /* Clear tail bits in last word beyond nbits. */
> > + if (nbits % BITS_PER_LONG)
> > + bitmap[-1] &= BITMAP_LAST_WORD_MASK(nbits);
> > +}
> > +EXPORT_SYMBOL(bitmap_from_arr64);
> > +
> > +/**
> > + * bitmap_to_arr64 - copy the contents of bitmap to a u64 array of bits
> > + * @buf: array of u64 (in host byte order), the dest bitmap
> > + * @bitmap: array of unsigned longs, the source bitmap
> > + * @nbits: number of bits in @bitmap
> > + */
> > +void bitmap_to_arr64(u64 *buf, const unsigned long *bitmap, unsigned int nbits)
> > +{
> > + unsigned long *end = bitmap + BITS_TO_LONGS(nbits);
This should be const unsigned long.
> > +
> > + while (bitmap < end) {
>
> Another for loop...
>
> > + *buf = *bitmap++;
> > + if (bitmap < end)
> > + *buf |= *bitmap++ << 32;
>
> That is UB.
It's -Wshift-count-overflow. Should be
*buf |= (u64)(*bitmap++) << 32;
> Did you even compile this??
Yes. That's it, my BE32 platform is arm, and I had to disable CONFIG_WERROR
because arm breaks build with that, and forgot about it. :(
I boot-tested it on mips with the fix above with no issues.
> David
>
> > + buf++;
> > + }
> > +
> > + /* Clear tail bits in last element of array beyond nbits. */
> > + if (nbits % 64)
> > + buf[-1] &= GENMASK_ULL(nbits, 0);
> > +}
> > +EXPORT_SYMBOL(bitmap_to_arr64);
> > #endif
> > --
> > 2.32.0
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)
