All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
To: Anshuman Khandual <khandual@linux.vnet.ibm.com>,
	linux-mm@kvack.org, linux-kernel@vger.kernel.org,
	linuxppc-dev@lists.ozlabs.org
Cc: hughd@google.com, kirill@shutemov.name,
	n-horiguchi@ah.jp.nec.com, akpm@linux-foundation.org,
	mgorman@techsingularity.net, mpe@ellerman.id.au
Subject: Re: [RFC 4/9] powerpc/mm: Split huge_pte_alloc function for BOOK3S 64K
Date: Thu, 10 Mar 2016 01:25:53 +0530	[thread overview]
Message-ID: <878u1r1l4m.fsf@linux.vnet.ibm.com> (raw)
In-Reply-To: <1457525450-4262-4-git-send-email-khandual@linux.vnet.ibm.com>

Anshuman Khandual <khandual@linux.vnet.ibm.com> writes:

> [ text/plain ]
> From: root <root@ltcalpine2-lp8.aus.stglabs.ibm.com>
>
> Currently the 'huge_pte_alloc' function has two versions, one for the
> BOOK3S and the other one for the BOOK3E platforms. This change splits
> the BOOK3S version into two parts, one for the 4K page size based
> implementation and the other one for the 64K page sized implementation.
> This change is one of the prerequisites towards enabling GENERAL_HUGETLB
> implementation for BOOK3S 64K based huge pages.

I really wish we reduce #ifdefs in C code and start splitting hash
and nonhash code out where ever we can. 

What we really want here is a book3s version and in book3s version use
powerpc specific huge_pte_alloc only if GENERAL_HUGETLB was not defined.
Don't limit it to 64k linux page size. We should select between powerpc
specific implementation and generic code using GENERAL_HUGETLB define.


>
> Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
> ---
>  arch/powerpc/mm/hugetlbpage.c | 67 +++++++++++++++++++++++++++----------------
>  1 file changed, 43 insertions(+), 24 deletions(-)
>
> diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
> index 744e24b..a49c6ae 100644
> --- a/arch/powerpc/mm/hugetlbpage.c
> +++ b/arch/powerpc/mm/hugetlbpage.c
> @@ -59,6 +59,7 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
>  	return __find_linux_pte_or_hugepte(mm->pgd, addr, NULL, NULL);
>  }
>
> +#if !defined(CONFIG_PPC_64K_PAGES) || !defined(CONFIG_PPC_BOOK3S_64)
>  static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
>  			   unsigned long address, unsigned pdshift, unsigned pshift)
>  {
> @@ -117,6 +118,7 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
>  	spin_unlock(&mm->page_table_lock);
>  	return 0;
>  }
> +#endif /* !defined(CONFIG_PPC_64K_PAGES) || !defined(CONFIG_PPC_BOOK3S_64) */
>
>  /*
>   * These macros define how to determine which level of the page table holds
> @@ -131,6 +133,7 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
>  #endif
>
>  #ifdef CONFIG_PPC_BOOK3S_64
> +#ifdef CONFIG_PPC_4K_PAGES
>  /*
>   * At this point we do the placement change only for BOOK3S 64. This would
>   * possibly work on other subarchs.
> @@ -146,32 +149,23 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz
>
>  	addr &= ~(sz-1);
>  	pg = pgd_offset(mm, addr);
> -
> -	if (pshift == PGDIR_SHIFT)
> -		/* 16GB huge page */
> -		return (pte_t *) pg;
> -	else if (pshift > PUD_SHIFT)
> -		/*
> -		 * We need to use hugepd table
> -		 */
> +	if (pshift > PUD_SHIFT) {
>  		hpdp = (hugepd_t *)pg;
> -	else {
> -		pdshift = PUD_SHIFT;
> -		pu = pud_alloc(mm, pg, addr);
> -		if (pshift == PUD_SHIFT)
> -			return (pte_t *)pu;
> -		else if (pshift > PMD_SHIFT)
> -			hpdp = (hugepd_t *)pu;
> -		else {
> -			pdshift = PMD_SHIFT;
> -			pm = pmd_alloc(mm, pu, addr);
> -			if (pshift == PMD_SHIFT)
> -				/* 16MB hugepage */
> -				return (pte_t *)pm;
> -			else
> -				hpdp = (hugepd_t *)pm;
> -		}
> +		goto hugepd_search;
> +	}
> +
> +	pdshift = PUD_SHIFT;
> +	pu = pud_alloc(mm, pg, addr);
> +	if (pshift > PMD_SHIFT) {
> +		hpdp = (hugepd_t *)pu;
> +		goto hugepd_search;
>  	}
> +
> +	pdshift = PMD_SHIFT;
> +	pm = pmd_alloc(mm, pu, addr);
> +	hpdp = (hugepd_t *)pm;
> +
> +hugepd_search:
>  	if (!hpdp)
>  		return NULL;
>
> @@ -184,6 +178,31 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz
>  }
>
>  #else
> +pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
> +{
> +	pgd_t *pg;
> +	pud_t *pu;
> +	pmd_t *pm;
> +	unsigned pshift = __ffs(sz);
> +
> +	addr &= ~(sz-1);
> +	pg = pgd_offset(mm, addr);
> +
> +	if (pshift == PGDIR_SHIFT)	/* 16GB Huge Page */
> +		return (pte_t *)pg;
> +
> +	pu = pud_alloc(mm, pg, addr);	/* NA, skipped */
> +	if (pshift == PUD_SHIFT)
> +		return (pte_t *)pu;
> +
> +	pm = pmd_alloc(mm, pu, addr);	/* 16MB Huge Page */
> +	if (pshift == PMD_SHIFT)
> +		return (pte_t *)pm;
> +
> +	return NULL;
> +}
> +#endif
> +#else
>
>  pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
>  {
> -- 
> 2.1.0

WARNING: multiple messages have this Message-ID (diff)
From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
To: Anshuman Khandual <khandual@linux.vnet.ibm.com>,
	linux-mm@kvack.org, linux-kernel@vger.kernel.org,
	linuxppc-dev@lists.ozlabs.org
Cc: hughd@google.com, kirill@shutemov.name,
	n-horiguchi@ah.jp.nec.com, akpm@linux-foundation.org,
	mgorman@techsingularity.net, mpe@ellerman.id.au
Subject: Re: [RFC 4/9] powerpc/mm: Split huge_pte_alloc function for BOOK3S 64K
Date: Thu, 10 Mar 2016 01:25:53 +0530	[thread overview]
Message-ID: <878u1r1l4m.fsf@linux.vnet.ibm.com> (raw)
In-Reply-To: <1457525450-4262-4-git-send-email-khandual@linux.vnet.ibm.com>

Anshuman Khandual <khandual@linux.vnet.ibm.com> writes:

> [ text/plain ]
> From: root <root@ltcalpine2-lp8.aus.stglabs.ibm.com>
>
> Currently the 'huge_pte_alloc' function has two versions, one for the
> BOOK3S and the other one for the BOOK3E platforms. This change splits
> the BOOK3S version into two parts, one for the 4K page size based
> implementation and the other one for the 64K page sized implementation.
> This change is one of the prerequisites towards enabling GENERAL_HUGETLB
> implementation for BOOK3S 64K based huge pages.

I really wish we reduce #ifdefs in C code and start splitting hash
and nonhash code out where ever we can. 

What we really want here is a book3s version and in book3s version use
powerpc specific huge_pte_alloc only if GENERAL_HUGETLB was not defined.
Don't limit it to 64k linux page size. We should select between powerpc
specific implementation and generic code using GENERAL_HUGETLB define.


>
> Signed-off-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
> ---
>  arch/powerpc/mm/hugetlbpage.c | 67 +++++++++++++++++++++++++++----------------
>  1 file changed, 43 insertions(+), 24 deletions(-)
>
> diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
> index 744e24b..a49c6ae 100644
> --- a/arch/powerpc/mm/hugetlbpage.c
> +++ b/arch/powerpc/mm/hugetlbpage.c
> @@ -59,6 +59,7 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
>  	return __find_linux_pte_or_hugepte(mm->pgd, addr, NULL, NULL);
>  }
>
> +#if !defined(CONFIG_PPC_64K_PAGES) || !defined(CONFIG_PPC_BOOK3S_64)
>  static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
>  			   unsigned long address, unsigned pdshift, unsigned pshift)
>  {
> @@ -117,6 +118,7 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
>  	spin_unlock(&mm->page_table_lock);
>  	return 0;
>  }
> +#endif /* !defined(CONFIG_PPC_64K_PAGES) || !defined(CONFIG_PPC_BOOK3S_64) */
>
>  /*
>   * These macros define how to determine which level of the page table holds
> @@ -131,6 +133,7 @@ static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
>  #endif
>
>  #ifdef CONFIG_PPC_BOOK3S_64
> +#ifdef CONFIG_PPC_4K_PAGES
>  /*
>   * At this point we do the placement change only for BOOK3S 64. This would
>   * possibly work on other subarchs.
> @@ -146,32 +149,23 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz
>
>  	addr &= ~(sz-1);
>  	pg = pgd_offset(mm, addr);
> -
> -	if (pshift == PGDIR_SHIFT)
> -		/* 16GB huge page */
> -		return (pte_t *) pg;
> -	else if (pshift > PUD_SHIFT)
> -		/*
> -		 * We need to use hugepd table
> -		 */
> +	if (pshift > PUD_SHIFT) {
>  		hpdp = (hugepd_t *)pg;
> -	else {
> -		pdshift = PUD_SHIFT;
> -		pu = pud_alloc(mm, pg, addr);
> -		if (pshift == PUD_SHIFT)
> -			return (pte_t *)pu;
> -		else if (pshift > PMD_SHIFT)
> -			hpdp = (hugepd_t *)pu;
> -		else {
> -			pdshift = PMD_SHIFT;
> -			pm = pmd_alloc(mm, pu, addr);
> -			if (pshift == PMD_SHIFT)
> -				/* 16MB hugepage */
> -				return (pte_t *)pm;
> -			else
> -				hpdp = (hugepd_t *)pm;
> -		}
> +		goto hugepd_search;
> +	}
> +
> +	pdshift = PUD_SHIFT;
> +	pu = pud_alloc(mm, pg, addr);
> +	if (pshift > PMD_SHIFT) {
> +		hpdp = (hugepd_t *)pu;
> +		goto hugepd_search;
>  	}
> +
> +	pdshift = PMD_SHIFT;
> +	pm = pmd_alloc(mm, pu, addr);
> +	hpdp = (hugepd_t *)pm;
> +
> +hugepd_search:
>  	if (!hpdp)
>  		return NULL;
>
> @@ -184,6 +178,31 @@ pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz
>  }
>
>  #else
> +pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
> +{
> +	pgd_t *pg;
> +	pud_t *pu;
> +	pmd_t *pm;
> +	unsigned pshift = __ffs(sz);
> +
> +	addr &= ~(sz-1);
> +	pg = pgd_offset(mm, addr);
> +
> +	if (pshift == PGDIR_SHIFT)	/* 16GB Huge Page */
> +		return (pte_t *)pg;
> +
> +	pu = pud_alloc(mm, pg, addr);	/* NA, skipped */
> +	if (pshift == PUD_SHIFT)
> +		return (pte_t *)pu;
> +
> +	pm = pmd_alloc(mm, pu, addr);	/* 16MB Huge Page */
> +	if (pshift == PMD_SHIFT)
> +		return (pte_t *)pm;
> +
> +	return NULL;
> +}
> +#endif
> +#else
>
>  pte_t *huge_pte_alloc(struct mm_struct *mm, unsigned long addr, unsigned long sz)
>  {
> -- 
> 2.1.0

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2016-03-09 19:56 UTC|newest]

Thread overview: 44+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-03-09 12:10 [RFC 1/9] mm/hugetlb: Make GENERAL_HUGETLB functions PGD implementation aware Anshuman Khandual
2016-03-09 12:10 ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 2/9] mm/hugetlb: Add follow_huge_pgd function Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-11  3:02   ` Anshuman Khandual
2016-03-11  3:02     ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 3/9] mm/gup: Make follow_page_mask function PGD implementation aware Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-11  3:03   ` Anshuman Khandual
2016-03-11  3:03     ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 4/9] powerpc/mm: Split huge_pte_alloc function for BOOK3S 64K Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-09 19:55   ` Aneesh Kumar K.V [this message]
2016-03-09 19:55     ` Aneesh Kumar K.V
2016-03-10  5:33     ` Anshuman Khandual
2016-03-10  5:33       ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 5/9] powerpc/mm: Split huge_pte_offset " Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-09 22:57   ` Dave Hansen
2016-03-09 22:57     ` Dave Hansen
2016-03-10  3:37     ` Anshuman Khandual
2016-03-10  3:37       ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 6/9] powerpc/hugetlb: Enable ARCH_WANT_GENERAL_HUGETLB " Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-09 19:58   ` Aneesh Kumar K.V
2016-03-09 19:58     ` Aneesh Kumar K.V
2016-03-10  5:12     ` Anshuman Khandual
2016-03-10  5:12       ` Anshuman Khandual
2016-03-21  9:55   ` Rui Teng
2016-03-21  9:55     ` Rui Teng
2016-03-09 12:10 ` [RFC 7/9] powerpc/hugetlb: Change follow_huge_* routines " Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 8/9] powerpc/mm: Enable HugeTLB page migration Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-09 12:10 ` [RFC 9/9] selfttest/powerpc: Add memory page migration tests Anshuman Khandual
2016-03-09 12:10   ` Anshuman Khandual
2016-03-09 20:01   ` Aneesh Kumar K.V
2016-03-09 20:01     ` Aneesh Kumar K.V
2016-03-10  5:05     ` Anshuman Khandual
2016-03-10  5:05       ` Anshuman Khandual
2016-03-11  3:01 ` [RFC 1/9] mm/hugetlb: Make GENERAL_HUGETLB functions PGD implementation aware Anshuman Khandual
2016-03-11  3:01   ` Anshuman Khandual
2016-03-14 20:29   ` Andrew Morton
2016-03-14 20:29     ` Andrew Morton

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=878u1r1l4m.fsf@linux.vnet.ibm.com \
    --to=aneesh.kumar@linux.vnet.ibm.com \
    --cc=akpm@linux-foundation.org \
    --cc=hughd@google.com \
    --cc=khandual@linux.vnet.ibm.com \
    --cc=kirill@shutemov.name \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=linuxppc-dev@lists.ozlabs.org \
    --cc=mgorman@techsingularity.net \
    --cc=mpe@ellerman.id.au \
    --cc=n-horiguchi@ah.jp.nec.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.