Re: [PATCH 09/11] xfs: rework kmem_alloc_{io,large} to use GFP_* flags

From: Carlos Maiolino <cmaiolino@redhat.com>
To: Dave Chinner <david@fromorbit.com>,
	"Darrick J. Wong" <darrick.wong@oracle.com>,
	linux-xfs@vger.kernel.org
Subject: Re: [PATCH 09/11] xfs: rework kmem_alloc_{io,large} to use GFP_* flags
Date: Thu, 14 Nov 2019 11:31:17 +0100	[thread overview]
Message-ID: <20191114103117.pb6cpyqgnuug3epm@orion> (raw)
In-Reply-To: <20191114094022.u267tdwiistd6blb@orion>

> Ok, I'll leave it as-is (with converted flags).
> 
> > > TBH this series would be better sequenced as (1) get rid of pointless
> > > wrappers, (2) convert kmem* callers to use GFP flags, and (3) save
> > > whatever logic changes for the end.  If nothing else I could pull the
> > > first two parts and leave the third for later.
> > 
> 
> I need to fix the line >80 characters and some other nit picks on the patches,
> so I can split this into this 3-series guideline and submit everything again, no
> worries.
> 
> I do believe though the kmem_zone_alloc() removal falls into the 1st series here
> right? Giving the fact it removes the loop but doesn't really alter the logic at
> all, since __GFP_NOFAIL now implies the 'infinite loop' and the
> congestion_wait()?

NVM, it should be on set (2), not one...

/me goes grab more coffee.

> 
> > 
> > > >  static inline void *
> > > > -kmem_zalloc_large(size_t size, xfs_km_flags_t flags)
> > > > +kmem_zalloc_large(size_t size, gfp_t flags)
> > > >  {
> > > > -	return kmem_alloc_large(size, flags | KM_ZERO);
> > > > +	return kmem_alloc_large(size, flags | __GFP_ZERO);
> > > >  }
> > 
> > I'd also kill these zalloc wrappers and just pass __GFP_ZERO
> > directly.
> 
> Thanks for the review guys.
> 
> Cheers
> 
> -- 
> Carlos

-- 
Carlos