In 'mthca_buddy_init()', the 'buddy->bits[n]' bitmap has just been
allocated, so no concurrent accesses can occur.

The other accesses to the 'buddy->bits[n]' bitmap are protected by the
'buddy->lock' spinlock, so no concurrent accesses can occur.

So prefer the non-atomic '__[set|clear]_bit()' functions to save a few
cycles.

Link: https://lore.kernel.org/r/a19b88ccdbc03972fd97306b998731814283041f.1637785902.git.christophe.jaillet@wanadoo.frSigned-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>