Re: [PATCH v2 4/8] metaperf: Include linux/param.h explicitly for HZ macro

[Eryu Guan <eguan@redhat.com>]

On Thu, Jan 25, 2018 at 11:02:15AM +1100, Dave Chinner wrote:
> On Wed, Jan 24, 2018 at 11:58:59AM +0800, Eryu Guan wrote:
> > On Mon, Jan 22, 2018 at 12:24:06PM -0800, Christoph Hellwig wrote:
> > > On Mon, Jan 22, 2018 at 03:57:15PM +0000, Rostislav Skudnov wrote:
> > > > Glibc includes linux/param.h when we include sys/param.h, whereas musl
> > > > libc does not do that. HZ is a Linux-specific macro, therefore include
> > > > the header file that defines it explicitly.
> > > > 
> > > > Signed-off-by: Rostislav Skudnov <rostislav@tuxera.com>
> > > 
> > > Using HZ in userspace doesn't make any sense at all.  The kernel HZ
> > > (as in the granulairy for jiffies) can vary between architectures
> > > and even configurations.  I guess it wants a hard coded 100 here, but
> > > someone will have to do a detailed analysis.
> > 
> > No one is using metaperf now, all the old benchmark infrastructure was
> > removed by commit b020416d51ff ("xfstests: remove bench infrastructure")
> > 
> > I think we can remove src/metaperf.c too.
> 
> I use metaperf (and dirperf) quite regularly when testing my
> patches, so I'd really like them to remin available on all my test
> machines that have fstests installed....

Thanks for the heads-up! Let's keep dmiperf.c and metaperf.c then.

I tend to take this patch as is for now to fix the build problem with
musl libc first. I think the "using HZ in userspace problem" is another
issue which can be fixed by a separate patch (if necessary and someone
has interest).

Thanks,
Eryu