Re: [RFT v5h printk: allow increasing the ring buffer depending on the number of CPUs

["Petr Mládek" <pmladek@suse.cz>]

On Wed 2014-06-18 12:59:26, Luis R. Rodriguez wrote:
> On Wed, Jun 18, 2014 at 10:31:02AM +0200, Petr Mládek wrote:
> > On Wed 2014-06-18 02:18:16, Luis R. Rodriguez wrote:
> > I am happy with this solution. And I agree that it is better to split
> > log_buf_len_align() in a separate patch as you suggested in the other
> > mail.
> 
> OK just to be on safe side I noticed memblock_virt_alloc() and
> memblock_virt_alloc_nopanic() allow passing an explicit alignment
> requirement, traced back the orignal code with no good reason to
> not use the LOG_ALIGN, so I think using that would be the safest
> thing to do. Will roll that into the first patch, curious if the
> folks that ran into the alignment issues on ARM could reproduce
> an align barf without this on some situations, perhaps not because
> of the power of 2 thing and since the min value for LOG_BUF_SHIFT
> is 12.

Great catch. It makes sense to me. There is no reason to have aligned
stores when the buffer itself is not properly aligned.

IMHO, it would make sense to have separate patch for this change. It might be
candidate for stable releases.

Best Regards,
Petr