Re: [RFC PATCH 1/7 v1]powerpc: Free up four PTE bits to accommodate memory keys

From: Ram Pai <linuxram@us.ibm.com>
To: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org,
	benh@kernel.crashing.org, paulus@samba.org, mpe@ellerman.id.au,
	khandual@linux.vnet.ibm.com, bsingharora@gmail.com,
	dave.hansen@intel.com, hbabu@us.ibm.com
Subject: Re: [RFC PATCH 1/7 v1]powerpc: Free up four PTE bits to accommodate memory keys
Date: Tue, 13 Jun 2017 14:52:24 -0700	[thread overview]
Message-ID: <20170613215224.GA5590@ram.oc3035372033.ibm.com> (raw)
In-Reply-To: <87tw3k1obo.fsf@skywalker.in.ibm.com>

On Tue, Jun 13, 2017 at 10:22:43AM +0530, Aneesh Kumar K.V wrote:
> Ram Pai <linuxram@us.ibm.com> writes:
> 
> > Rearrange  PTE   bits to  free  up  bits 3, 4, 5  and  6  for
> > memory keys. Bit 3, 4, 5, 6 and 57  shall  be used for memory
> > keys.
> >
> > The patch does the following change to the 64K PTE format
> >
> > H_PAGE_BUSY moves from bit 3 to bit 7
> > H_PAGE_F_SECOND which occupied bit 4 moves to the second part
> > 	of the pte.
> > H_PAGE_F_GIX which  occupied bit 5, 6 and 7 also moves to the
> > 	second part of the pte.
> >
> > The second part of the PTE will hold
> >    a (H_PAGE_F_SECOND|H_PAGE_F_GIX)  for  64K page backed pte,
> >    and sixteen (H_PAGE_F_SECOND|H_PAGE_F_GIX)  for 4k  backed
> > 	pte.
> >
> > the four  bits((H_PAGE_F_SECOND|H_PAGE_F_GIX) that represent a slot
> > is initialized to 0xF indicating a invalid slot. if a hashpage does
> > get  allocated  to  the  0xF  slot, it is released and not used. In
> > other words, even  though  0xF  is  a valid slot we discard it  and
> > consider it as invalid slot(HPTE_SOFT_INVALID). This  gives  us  an
> > opportunity to  not  depend on a bit in the primary PTE in order to
> > determine the validity of a slot.
> >
> > When  we  release  a  0xF slot we also release a legitimate primary
> > slot  and  unmap  that  entry. This  is  to  ensure  that we do get
> > a legimate non-0xF slot the next time we retry for a slot.
> >
> > Though treating 0xF slot as invalid reduces the number of available
> > slots and make have a effect on the performance, the probabilty
> > of hitting a 0xF is extermely low.
> >
> > Compared  to the current scheme, the above described scheme reduces
> > the number of false hash table updates  significantly  and  has the
> > added  advantage  of  releasing  four  valuable  PTE bits for other
> > purpose.
> >
> > This idea was jointly developed by Paul Mackerras, Aneesh, Michael
> > Ellermen and myself.
> >
> > 4K PTE format remain unchanged currently.
> >
> 
> Can you also split this patch into two. One which changes
> __hash_page_4k() ie, linux pte format w.r.t 4k hash pte. Second patch
> with changes w.r.t __hash_page_64k() ie, pte format w.r.t 64k hash pte.

ok. A v2 version of the patch series will be out in a day or two.
RP