From: Will Deacon <will@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>,
linux-m68k@lists.linux-m68k.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 4/5] m68k,mm: Extend table allocator for multiple sizes
Date: Wed, 29 Jan 2020 12:17:53 +0000 [thread overview]
Message-ID: <20200129121752.GB31582@willie-the-truck> (raw)
In-Reply-To: <20200129104345.491163937@infradead.org>
On Wed, Jan 29, 2020 at 11:39:45AM +0100, Peter Zijlstra wrote:
> In addition to the PGD/PMD table size (128*4) add a PTE table size
> (64*4) to the table allocator. This completely removes the pte-table
> overhead compared to the old code, even for dense tables.
>
> Notes:
>
> - the allocator gained __flush_page_to_ram(), since the old
> page-based allocator had that.
>
> - the allocator gained a list_empty() check to deal with there not
> being any pages at all.
>
> - the free mask is extended to cover more than the 8 bits required
> for the (512 byte) PGD/PMD tables.
>
> - NR_PAGETABLE accounting is restored.
>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
> ---
> arch/m68k/include/asm/motorola_pgalloc.h | 24 +++++-----
> arch/m68k/mm/init.c | 6 +-
> arch/m68k/mm/memory.c | 70 ++++++++++++++++++++-----------
> 3 files changed, 61 insertions(+), 39 deletions(-)
>
> --- a/arch/m68k/include/asm/motorola_pgalloc.h
> +++ b/arch/m68k/include/asm/motorola_pgalloc.h
> @@ -5,61 +5,61 @@
> #include <asm/tlb.h>
> #include <asm/tlbflush.h>
>
> -extern pmd_t *get_pointer_table(void);
> -extern int free_pointer_table(pmd_t *);
> +extern void *get_pointer_table(int type);
Could be prettier/obfuscated with an enum type?
> --- a/arch/m68k/mm/memory.c
> +++ b/arch/m68k/mm/memory.c
> @@ -27,24 +27,34 @@
> arch/sparc/mm/srmmu.c ... */
>
> typedef struct list_head ptable_desc;
> -static LIST_HEAD(ptable_list);
> +
> +static struct list_head ptable_list[2] = {
> + LIST_HEAD_INIT(ptable_list[0]),
> + LIST_HEAD_INIT(ptable_list[1]),
> +};
>
> #define PD_PTABLE(page) ((ptable_desc *)&(virt_to_page(page)->lru))
> #define PD_PAGE(ptable) (list_entry(ptable, struct page, lru))
> -#define PD_MARKBITS(dp) (*(unsigned char *)&PD_PAGE(dp)->index)
> +#define PD_MARKBITS(dp) (*(unsigned int *)&PD_PAGE(dp)->index)
> +
> +static const int ptable_shift[2] = {
> + 7+2, /* PGD, PMD */
> + 6+2, /* PTE */
> +};
>
> -#define PTABLE_SIZE (PTRS_PER_PMD * sizeof(pmd_t))
> +#define ptable_size(type) (1U << ptable_shift[type])
> +#define ptable_mask(type) ((1U << (PAGE_SIZE / ptable_size(type))) - 1)
>
> -void __init init_pointer_table(unsigned long ptable)
> +void __init init_pointer_table(unsigned long ptable, int type)
> {
> ptable_desc *dp;
> unsigned long page = ptable & PAGE_MASK;
> - unsigned char mask = 1 << ((ptable - page)/PTABLE_SIZE);
> + unsigned int mask = 1U << ((ptable - page)/ptable_size(type));
>
> dp = PD_PTABLE(page);
> if (!(PD_MARKBITS(dp) & mask)) {
> - PD_MARKBITS(dp) = 0xff;
> - list_add(dp, &ptable_list);
> + PD_MARKBITS(dp) = ptable_mask(type);
> + list_add(dp, &ptable_list[type]);
> }
>
> PD_MARKBITS(dp) &= ~mask;
> @@ -57,12 +67,10 @@ void __init init_pointer_table(unsigned
> return;
> }
>
> -pmd_t *get_pointer_table (void)
> +void *get_pointer_table (int type)
> {
> - ptable_desc *dp = ptable_list.next;
> - unsigned char mask = PD_MARKBITS (dp);
> - unsigned char tmp;
> - unsigned int off;
> + ptable_desc *dp = ptable_list[type].next;
> + unsigned int mask, tmp, off;
nit, but if you do:
unsigned int mask = list_empty(&ptable_list[type]) ? 0 : PD_MARKBITS(dp);
then you can leave the existing mask logic as-is.
Will
next prev parent reply other threads:[~2020-01-29 12:18 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-29 10:39 [PATCH 0/5] Rewrite Motorola MMU page-table layout Peter Zijlstra
2020-01-29 10:39 ` [PATCH 1/5] m68k,mm: Restructure motorola mmu " Peter Zijlstra
2020-01-29 10:39 ` [PATCH 2/5] m68k,mm: Improve kernel_page_table() Peter Zijlstra
2020-01-29 10:39 ` [PATCH 3/5] m68k,mm: Use table allocator for pgtables Peter Zijlstra
2020-01-29 12:11 ` Will Deacon
2020-01-29 12:24 ` Peter Zijlstra
2020-01-29 10:39 ` [PATCH 4/5] m68k,mm: Extend table allocator for multiple sizes Peter Zijlstra
2020-01-29 12:17 ` Will Deacon [this message]
2020-01-29 12:43 ` Peter Zijlstra
2020-01-29 13:15 ` Will Deacon
2020-01-29 10:39 ` [PATCH 5/5] m68k,mm: Fully initialize the page-table allocator Peter Zijlstra
2020-01-29 10:49 ` [PATCH 0/5] Rewrite Motorola MMU page-table layout John Paul Adrian Glaubitz
2020-01-29 11:54 ` Peter Zijlstra
2020-01-29 12:05 ` John Paul Adrian Glaubitz
2020-01-29 12:30 ` Peter Zijlstra
2020-01-29 18:52 ` Michael Schmitz
2020-01-29 19:31 ` Peter Zijlstra
2020-01-30 7:31 ` Michael Schmitz
2020-01-30 8:16 ` Peter Zijlstra
2020-01-30 19:12 ` Michael Schmitz
2020-01-31 6:31 ` Greg Ungerer
2020-01-31 9:38 ` Will Deacon
2020-01-31 10:22 ` Peter Zijlstra
2020-01-31 11:14 ` Peter Zijlstra
2020-01-31 11:18 ` Will Deacon
2020-01-31 11:31 ` Peter Zijlstra
2020-01-31 11:43 ` Will Deacon
2020-01-31 13:04 ` Greg Ungerer
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=20200129121752.GB31582@willie-the-truck \
--to=will@kernel.org \
--cc=geert@linux-m68k.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-m68k@lists.linux-m68k.org \
--cc=peterz@infradead.org \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).