From: Peter Zijlstra <peterz-wEGCiKHe2LqWVfeAwA7xHQ@public.gmane.org>
To: Nick Piggin <npiggin-l3A5Bk7waGM@public.gmane.org>
Cc: Andrew Morton
<akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>,
shaggy-V7BBcbaFuwjMbYB6QlFGEg@public.gmane.org,
axboe-tSWWG44O7X1aa/9Udqfwiw@public.gmane.org,
linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org,
linux-arch-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org,
Clark Williams <williams-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
Ingo Molnar <mingo-X9Un+BFzKDI@public.gmane.org>
Subject: Re: [patch 2/2]: introduce fast_gup
Date: Thu, 17 Apr 2008 17:03:25 +0200 [thread overview]
Message-ID: <1208444605.7115.2.camel@twins> (raw)
In-Reply-To: <20080328030023.GC8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
On Fri, 2008-03-28 at 04:00 +0100, Nick Piggin wrote:
> +++ linux-2.6/arch/x86/mm/gup.c
> @@ -0,0 +1,198 @@
> +/*
> + * Lockless fast_gup for x86
> + *
> + * Copyright (C) 2007 Nick Piggin
> + * Copyright (C) 2007 Novell Inc.
> + */
> +#include <linux/sched.h>
> +#include <linux/mm.h>
> +#include <linux/vmstat.h>
> +#include <asm/pgtable.h>
> +
> +/*
> + * The performance critical leaf functions are made noinline otherwise gcc
> + * inlines everything into a single function which results in too much
> + * register pressure.
> + */
> +static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
> + unsigned long end, int write, struct page **pages, int *nr)
> +{
> + unsigned long mask, result;
> + pte_t *ptep;
> +
> + result = _PAGE_PRESENT|_PAGE_USER;
> + if (write)
> + result |= _PAGE_RW;
> + mask = result | _PAGE_SPECIAL;
> +
> + ptep = pte_offset_map(&pmd, addr);
> + do {
> + /*
> + * XXX: careful. On 3-level 32-bit, the pte is 64 bits, and
> + * we need to make sure we load the low word first, then the
> + * high. This means _PAGE_PRESENT should be clear if the high
> + * word was not valid. Currently, the C compiler can issue
> + * the loads in any order, and I don't know of a wrapper
> + * function that will do this properly, so it is broken on
> + * 32-bit 3-level for the moment.
> + */
> + pte_t pte = *ptep;
> + struct page *page;
> +
> + if ((pte_val(pte) & mask) != result)
> + return 0;
> + VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
> + page = pte_page(pte);
> + get_page(page);
> + pages[*nr] = page;
> + (*nr)++;
> +
> + } while (ptep++, addr += PAGE_SIZE, addr != end);
> + pte_unmap(ptep - 1);
> +
> + return 1;
> +}
Would this be sufficient to address that comment's conern?
Index: linux-2.6/arch/x86/mm/gup.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/gup.c
+++ linux-2.6/arch/x86/mm/gup.c
@@ -36,8 +36,16 @@ static noinline int gup_pte_range(pmd_t
* function that will do this properly, so it is broken on
* 32-bit 3-level for the moment.
*/
- pte_t pte = *ptep;
struct page *page;
+ pte_t pte;
+
+#ifdef CONFIG_X86_PAE
+ pte.pte_low = ptep->pte_low;
+ barrier();
+ pte.pte_high = ptep->pte_high;
+#else
+ pte = *ptep;
+#endif
if ((pte_val(pte) & mask) != result)
return 0;
WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Nick Piggin <npiggin@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>,
shaggy@austin.ibm.com, axboe@kernel.dk, linux-mm@kvack.org,
linux-arch@vger.kernel.org, torvalds@linux-foundation.org,
Clark Williams <williams@redhat.com>, Ingo Molnar <mingo@elte.hu>
Subject: Re: [patch 2/2]: introduce fast_gup
Date: Thu, 17 Apr 2008 17:03:25 +0200 [thread overview]
Message-ID: <1208444605.7115.2.camel@twins> (raw)
Message-ID: <20080417150325.KXecs9azQsTXFaAcGp3biuYFbCzbJkl9zG-8VcjOobU@z> (raw)
In-Reply-To: <20080328030023.GC8083@wotan.suse.de>
On Fri, 2008-03-28 at 04:00 +0100, Nick Piggin wrote:
> +++ linux-2.6/arch/x86/mm/gup.c
> @@ -0,0 +1,198 @@
> +/*
> + * Lockless fast_gup for x86
> + *
> + * Copyright (C) 2007 Nick Piggin
> + * Copyright (C) 2007 Novell Inc.
> + */
> +#include <linux/sched.h>
> +#include <linux/mm.h>
> +#include <linux/vmstat.h>
> +#include <asm/pgtable.h>
> +
> +/*
> + * The performance critical leaf functions are made noinline otherwise gcc
> + * inlines everything into a single function which results in too much
> + * register pressure.
> + */
> +static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
> + unsigned long end, int write, struct page **pages, int *nr)
> +{
> + unsigned long mask, result;
> + pte_t *ptep;
> +
> + result = _PAGE_PRESENT|_PAGE_USER;
> + if (write)
> + result |= _PAGE_RW;
> + mask = result | _PAGE_SPECIAL;
> +
> + ptep = pte_offset_map(&pmd, addr);
> + do {
> + /*
> + * XXX: careful. On 3-level 32-bit, the pte is 64 bits, and
> + * we need to make sure we load the low word first, then the
> + * high. This means _PAGE_PRESENT should be clear if the high
> + * word was not valid. Currently, the C compiler can issue
> + * the loads in any order, and I don't know of a wrapper
> + * function that will do this properly, so it is broken on
> + * 32-bit 3-level for the moment.
> + */
> + pte_t pte = *ptep;
> + struct page *page;
> +
> + if ((pte_val(pte) & mask) != result)
> + return 0;
> + VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
> + page = pte_page(pte);
> + get_page(page);
> + pages[*nr] = page;
> + (*nr)++;
> +
> + } while (ptep++, addr += PAGE_SIZE, addr != end);
> + pte_unmap(ptep - 1);
> +
> + return 1;
> +}
Would this be sufficient to address that comment's conern?
Index: linux-2.6/arch/x86/mm/gup.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/gup.c
+++ linux-2.6/arch/x86/mm/gup.c
@@ -36,8 +36,16 @@ static noinline int gup_pte_range(pmd_t
* function that will do this properly, so it is broken on
* 32-bit 3-level for the moment.
*/
- pte_t pte = *ptep;
struct page *page;
+ pte_t pte;
+
+#ifdef CONFIG_X86_PAE
+ pte.pte_low = ptep->pte_low;
+ barrier();
+ pte.pte_high = ptep->pte_high;
+#else
+ pte = *ptep;
+#endif
if ((pte_val(pte) & mask) != result)
return 0;
WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Nick Piggin <npiggin@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>,
shaggy@austin.ibm.com, axboe@kernel.dk, linux-mm@kvack.org,
linux-arch@vger.kernel.org, torvalds@linux-foundation.org,
Clark Williams <williams@redhat.com>, Ingo Molnar <mingo@elte.hu>
Subject: Re: [patch 2/2]: introduce fast_gup
Date: Thu, 17 Apr 2008 17:03:25 +0200 [thread overview]
Message-ID: <1208444605.7115.2.camel@twins> (raw)
In-Reply-To: <20080328030023.GC8083@wotan.suse.de>
On Fri, 2008-03-28 at 04:00 +0100, Nick Piggin wrote:
> +++ linux-2.6/arch/x86/mm/gup.c
> @@ -0,0 +1,198 @@
> +/*
> + * Lockless fast_gup for x86
> + *
> + * Copyright (C) 2007 Nick Piggin
> + * Copyright (C) 2007 Novell Inc.
> + */
> +#include <linux/sched.h>
> +#include <linux/mm.h>
> +#include <linux/vmstat.h>
> +#include <asm/pgtable.h>
> +
> +/*
> + * The performance critical leaf functions are made noinline otherwise gcc
> + * inlines everything into a single function which results in too much
> + * register pressure.
> + */
> +static noinline int gup_pte_range(pmd_t pmd, unsigned long addr,
> + unsigned long end, int write, struct page **pages, int *nr)
> +{
> + unsigned long mask, result;
> + pte_t *ptep;
> +
> + result = _PAGE_PRESENT|_PAGE_USER;
> + if (write)
> + result |= _PAGE_RW;
> + mask = result | _PAGE_SPECIAL;
> +
> + ptep = pte_offset_map(&pmd, addr);
> + do {
> + /*
> + * XXX: careful. On 3-level 32-bit, the pte is 64 bits, and
> + * we need to make sure we load the low word first, then the
> + * high. This means _PAGE_PRESENT should be clear if the high
> + * word was not valid. Currently, the C compiler can issue
> + * the loads in any order, and I don't know of a wrapper
> + * function that will do this properly, so it is broken on
> + * 32-bit 3-level for the moment.
> + */
> + pte_t pte = *ptep;
> + struct page *page;
> +
> + if ((pte_val(pte) & mask) != result)
> + return 0;
> + VM_BUG_ON(!pfn_valid(pte_pfn(pte)));
> + page = pte_page(pte);
> + get_page(page);
> + pages[*nr] = page;
> + (*nr)++;
> +
> + } while (ptep++, addr += PAGE_SIZE, addr != end);
> + pte_unmap(ptep - 1);
> +
> + return 1;
> +}
Would this be sufficient to address that comment's conern?
Index: linux-2.6/arch/x86/mm/gup.c
===================================================================
--- linux-2.6.orig/arch/x86/mm/gup.c
+++ linux-2.6/arch/x86/mm/gup.c
@@ -36,8 +36,16 @@ static noinline int gup_pte_range(pmd_t
* function that will do this properly, so it is broken on
* 32-bit 3-level for the moment.
*/
- pte_t pte = *ptep;
struct page *page;
+ pte_t pte;
+
+#ifdef CONFIG_X86_PAE
+ pte.pte_low = ptep->pte_low;
+ barrier();
+ pte.pte_high = ptep->pte_high;
+#else
+ pte = *ptep;
+#endif
if ((pte_val(pte) & mask) != result)
return 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>
next prev parent reply other threads:[~2008-04-17 15:03 UTC|newest]
Thread overview: 106+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-03-28 2:54 [patch 0/2]: lockless get_user_pages patchset Nick Piggin
2008-03-28 2:54 ` Nick Piggin
2008-03-28 2:54 ` Nick Piggin
[not found] ` <20080328025455.GA8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-03-28 2:55 ` [patch 1/2]: x86: implement pte_special Nick Piggin
2008-03-28 2:55 ` Nick Piggin
2008-03-28 2:55 ` Nick Piggin
[not found] ` <20080328025541.GB8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-03-28 3:23 ` David Miller
2008-03-28 3:23 ` David Miller, Nick Piggin
2008-03-28 3:23 ` David Miller
[not found] ` <20080327.202334.250213398.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2008-03-28 3:31 ` Nick Piggin
2008-03-28 3:31 ` Nick Piggin
2008-03-28 3:31 ` Nick Piggin
[not found] ` <20080328033149.GD8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-03-28 3:44 ` David Miller
2008-03-28 3:44 ` David Miller, Nick Piggin
2008-03-28 3:44 ` David Miller
[not found] ` <20080327.204431.201380891.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2008-03-28 4:04 ` Nick Piggin
2008-03-28 4:04 ` Nick Piggin
2008-03-28 4:04 ` Nick Piggin
[not found] ` <20080328040442.GE8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-03-28 4:09 ` David Miller
2008-03-28 4:09 ` David Miller, Nick Piggin
2008-03-28 4:09 ` David Miller
[not found] ` <20080327.210910.101408473.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2008-03-28 4:15 ` Nick Piggin
2008-03-28 4:15 ` Nick Piggin
2008-03-28 4:15 ` Nick Piggin
[not found] ` <20080328041519.GF8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-03-28 4:16 ` David Miller
2008-03-28 4:16 ` David Miller, Nick Piggin
2008-03-28 4:16 ` David Miller
[not found] ` <20080327.211632.02770342.davem-fT/PcQaiUtIeIZ0/mPfg9Q@public.gmane.org>
2008-03-28 4:19 ` Nick Piggin
2008-03-28 4:19 ` Nick Piggin
2008-03-28 4:19 ` Nick Piggin
2008-03-28 4:17 ` Nick Piggin
2008-03-28 4:17 ` Nick Piggin
2008-03-28 4:17 ` Nick Piggin
2008-03-28 3:00 ` [patch 2/2]: introduce fast_gup Nick Piggin
2008-03-28 3:00 ` Nick Piggin
2008-03-28 3:00 ` Nick Piggin
[not found] ` <20080328030023.GC8083-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-03-28 10:01 ` Jens Axboe
2008-03-28 10:01 ` Jens Axboe
2008-03-28 10:01 ` Jens Axboe
2008-04-17 15:03 ` Peter Zijlstra [this message]
2008-04-17 15:03 ` Peter Zijlstra
2008-04-17 15:03 ` Peter Zijlstra
2008-04-17 15:25 ` Linus Torvalds
2008-04-17 15:25 ` Linus Torvalds
2008-04-17 15:25 ` Linus Torvalds
[not found] ` <alpine.LFD.1.00.0804170814090.2879-5CScLwifNT1QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org>
2008-04-17 16:12 ` Peter Zijlstra
2008-04-17 16:12 ` Peter Zijlstra
2008-04-17 16:12 ` Peter Zijlstra
2008-04-17 16:18 ` Linus Torvalds
2008-04-17 16:18 ` Linus Torvalds
2008-04-17 16:18 ` Linus Torvalds
[not found] ` <alpine.LFD.1.00.0804170916470.2879-5CScLwifNT1QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org>
2008-04-17 16:35 ` Peter Zijlstra
2008-04-17 16:35 ` Peter Zijlstra
2008-04-17 16:35 ` Peter Zijlstra
2008-04-17 16:40 ` Linus Torvalds
2008-04-17 16:40 ` Linus Torvalds
2008-04-17 16:40 ` Linus Torvalds
[not found] ` <alpine.LFD.1.00.0804170940270.2879-5CScLwifNT1QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org>
2008-04-17 17:23 ` Peter Zijlstra
2008-04-17 17:23 ` Peter Zijlstra
2008-04-17 17:23 ` Peter Zijlstra
2008-04-17 18:28 ` Linus Torvalds
2008-04-17 18:28 ` Linus Torvalds
2008-04-17 18:28 ` Linus Torvalds
[not found] ` <alpine.LFD.1.00.0804171127310.2879-5CScLwifNT1QetFLy7KEm3xJsTq8ys+cHZ5vskTnxNA@public.gmane.org>
2008-04-22 3:14 ` Nick Piggin
2008-04-22 3:14 ` Nick Piggin
2008-04-22 3:14 ` Nick Piggin
2008-04-18 6:31 ` Geert Uytterhoeven
2008-04-18 6:31 ` Geert Uytterhoeven
2008-04-18 6:31 ` Geert Uytterhoeven
2008-04-18 14:40 ` Linus Torvalds
2008-04-18 14:40 ` Linus Torvalds
2008-04-18 14:40 ` Linus Torvalds
2008-04-18 9:58 ` Jeremy Fitzhardinge
2008-04-18 9:58 ` Jeremy Fitzhardinge
2008-04-18 9:58 ` Jeremy Fitzhardinge
2008-04-21 12:00 ` Avi Kivity
2008-04-21 12:00 ` Avi Kivity
2008-04-21 12:00 ` Avi Kivity
[not found] ` <480C81C4.8030200-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2008-04-21 12:30 ` Peter Zijlstra
2008-04-21 12:30 ` Peter Zijlstra
2008-04-21 12:30 ` Peter Zijlstra
2008-04-21 13:26 ` Avi Kivity
2008-04-21 13:26 ` Avi Kivity
2008-04-21 13:26 ` Avi Kivity
[not found] ` <480C9619.2050201-atKUWr5tajBWk0Htik3J/w@public.gmane.org>
2008-04-21 14:35 ` Peter Zijlstra
2008-04-21 14:35 ` Peter Zijlstra
2008-04-21 14:35 ` Peter Zijlstra
2008-04-22 3:23 ` Nick Piggin
2008-04-22 3:23 ` Nick Piggin
2008-04-22 3:23 ` Nick Piggin
[not found] ` <20080422032319.GB21993-B4tOwbsTzaBolqkO4TVVkw@public.gmane.org>
2008-04-22 7:19 ` Avi Kivity
2008-04-22 7:19 ` Avi Kivity
2008-04-22 7:19 ` Avi Kivity
2008-04-22 8:07 ` Ingo Molnar
2008-04-22 8:07 ` Ingo Molnar
2008-04-22 8:07 ` Ingo Molnar
2008-04-22 9:42 ` Peter Zijlstra
2008-04-22 9:42 ` Peter Zijlstra
2008-04-22 9:42 ` Peter Zijlstra
2008-04-22 9:46 ` Nick Piggin
2008-04-22 9:46 ` Nick Piggin
2008-04-22 9:46 ` Nick Piggin
2008-05-14 18:33 ` Dave Kleikamp
2008-05-14 18:33 ` Dave Kleikamp
2008-05-15 1:13 ` Nick Piggin
2008-05-15 1:13 ` Nick Piggin
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=1208444605.7115.2.camel@twins \
--to=peterz-wegcikhe2lqwvfeawa7xhq@public.gmane.org \
--cc=akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org \
--cc=axboe-tSWWG44O7X1aa/9Udqfwiw@public.gmane.org \
--cc=linux-arch-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=linux-mm-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org \
--cc=mingo-X9Un+BFzKDI@public.gmane.org \
--cc=npiggin-l3A5Bk7waGM@public.gmane.org \
--cc=shaggy-V7BBcbaFuwjMbYB6QlFGEg@public.gmane.org \
--cc=torvalds-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org \
--cc=williams-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.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 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.