2020-06-05 09:14:17

by Bibo Mao

[permalink] [raw]
Subject: [PATCH 1/2] MIPS: set page access bit with pgprot on some MIPS platform

On MIPS system which has rixi hardware bit, page access bit is not
set in pgrot. For memory reading, there will be one page fault to
allocate physical page; however valid bit is not set, there will
be the second fast tlb-miss fault handling to set valid/access bit.

This patch set page access/valid bit with pgrot if there is reading
access privilege. It will reduce one tlb-miss handling for memory
reading access.

The valid/access bit will be cleared in order to track memory
accessing activity. If the page is accessed, tlb-miss fast handling
will set valid/access bit, pte_sw_mkyoung is not necessary in slow
page fault path. This patch removes pte_sw_mkyoung function which
is defined as empty function except MIPS system.

Signed-off-by: Bibo Mao <[email protected]>
---
arch/mips/include/asm/pgtable.h | 11 +++++++++--
arch/mips/mm/cache.c | 34 +++++++++++++++++-----------------
include/asm-generic/pgtable.h | 16 ----------------
mm/memory.c | 3 ---
4 files changed, 26 insertions(+), 38 deletions(-)

diff --git a/arch/mips/include/asm/pgtable.h b/arch/mips/include/asm/pgtable.h
index 85b39c9..e2452ab 100644
--- a/arch/mips/include/asm/pgtable.h
+++ b/arch/mips/include/asm/pgtable.h
@@ -25,6 +25,15 @@
struct mm_struct;
struct vm_area_struct;

+#define __PP _PAGE_PRESENT
+#define __NX _PAGE_NO_EXEC
+#define __NR _PAGE_NO_READ
+#define ___W _PAGE_WRITE
+#define ___A _PAGE_ACCESSED
+#define ___R (_PAGE_SILENT_READ | _PAGE_ACCESSED)
+#define __PC _page_cachable_default
+
+
#define PAGE_NONE __pgprot(_PAGE_PRESENT | _PAGE_NO_READ | \
_page_cachable_default)
#define PAGE_SHARED __pgprot(_PAGE_PRESENT | _PAGE_WRITE | \
@@ -414,8 +423,6 @@ static inline pte_t pte_mkyoung(pte_t pte)
return pte;
}

-#define pte_sw_mkyoung pte_mkyoung
-
#ifdef CONFIG_MIPS_HUGE_TLB_SUPPORT
static inline int pte_huge(pte_t pte) { return pte_val(pte) & _PAGE_HUGE; }

diff --git a/arch/mips/mm/cache.c b/arch/mips/mm/cache.c
index ad6df1c..f814e43 100644
--- a/arch/mips/mm/cache.c
+++ b/arch/mips/mm/cache.c
@@ -158,23 +158,23 @@ void __update_cache(unsigned long address, pte_t pte)
static inline void setup_protection_map(void)
{
if (cpu_has_rixi) {
- protection_map[0] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_NO_READ);
- protection_map[1] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC);
- protection_map[2] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_NO_READ);
- protection_map[3] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC);
- protection_map[4] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
- protection_map[5] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
- protection_map[6] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
- protection_map[7] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
-
- protection_map[8] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_NO_READ);
- protection_map[9] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC);
- protection_map[10] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_WRITE | _PAGE_NO_READ);
- protection_map[11] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_NO_EXEC | _PAGE_WRITE);
- protection_map[12] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
- protection_map[13] = __pgprot(_page_cachable_default | _PAGE_PRESENT);
- protection_map[14] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_WRITE);
- protection_map[15] = __pgprot(_page_cachable_default | _PAGE_PRESENT | _PAGE_WRITE);
+ protection_map[0] = __pgprot(__PC | __PP | __NX | __NR);
+ protection_map[1] = __pgprot(__PC | __PP | __NX | ___R);
+ protection_map[2] = __pgprot(__PC | __PP | __NX | __NR);
+ protection_map[3] = __pgprot(__PC | __PP | __NX | ___R);
+ protection_map[4] = __pgprot(__PC | __PP | ___R);
+ protection_map[5] = __pgprot(__PC | __PP | ___R);
+ protection_map[6] = __pgprot(__PC | __PP | ___R);
+ protection_map[7] = __pgprot(__PC | __PP | ___R);
+
+ protection_map[8] = __pgprot(__PC | __PP | __NX | __NR);
+ protection_map[9] = __pgprot(__PC | __PP | __NX | ___R);
+ protection_map[10] = __pgprot(__PC | __PP | __NX | ___W | __NR);
+ protection_map[11] = __pgprot(__PC | __PP | __NX | ___W | ___R);
+ protection_map[12] = __pgprot(__PC | __PP | ___R);
+ protection_map[13] = __pgprot(__PC | __PP | ___R);
+ protection_map[14] = __pgprot(__PC | __PP | ___W | ___R);
+ protection_map[15] = __pgprot(__PC | __PP | ___W | ___R);

} else {
protection_map[0] = PAGE_NONE;
diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h
index b5278ec..fa5c73f 100644
--- a/include/asm-generic/pgtable.h
+++ b/include/asm-generic/pgtable.h
@@ -244,22 +244,6 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addres
}
#endif

-/*
- * On some architectures hardware does not set page access bit when accessing
- * memory page, it is responsibilty of software setting this bit. It brings
- * out extra page fault penalty to track page access bit. For optimization page
- * access bit can be set during all page fault flow on these arches.
- * To be differentiate with macro pte_mkyoung, this macro is used on platforms
- * where software maintains page access bit.
- */
-#ifndef pte_sw_mkyoung
-static inline pte_t pte_sw_mkyoung(pte_t pte)
-{
- return pte;
-}
-#define pte_sw_mkyoung pte_sw_mkyoung
-#endif
-
#ifndef pte_savedwrite
#define pte_savedwrite pte_write
#endif
diff --git a/mm/memory.c b/mm/memory.c
index c7c8960..8bb31c4 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -2704,7 +2704,6 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
}
flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
entry = mk_pte(new_page, vma->vm_page_prot);
- entry = pte_sw_mkyoung(entry);
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
/*
* Clear the pte entry and flush it first, before updating the
@@ -3379,7 +3378,6 @@ static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
__SetPageUptodate(page);

entry = mk_pte(page, vma->vm_page_prot);
- entry = pte_sw_mkyoung(entry);
if (vma->vm_flags & VM_WRITE)
entry = pte_mkwrite(pte_mkdirty(entry));

@@ -3662,7 +3660,6 @@ vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,

flush_icache_page(vma, page);
entry = mk_pte(page, vma->vm_page_prot);
- entry = pte_sw_mkyoung(entry);
if (write)
entry = maybe_mkwrite(pte_mkdirty(entry), vma);
/* copy-on-write page */
--
1.8.3.1


2020-06-05 09:14:19

by Bibo Mao

[permalink] [raw]
Subject: [PATCH 2/2] MIPS: Add writable-applies-readable policy with pgrot

On Linux system, writable applies readable privilege in most
architectures, this patch adds this policy on MIPS platform
where hardware rixi is supported.

Signed-off-by: Bibo Mao <[email protected]>
---
arch/mips/mm/cache.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/mips/mm/cache.c b/arch/mips/mm/cache.c
index f814e43..dae0617 100644
--- a/arch/mips/mm/cache.c
+++ b/arch/mips/mm/cache.c
@@ -160,7 +160,7 @@ static inline void setup_protection_map(void)
if (cpu_has_rixi) {
protection_map[0] = __pgprot(__PC | __PP | __NX | __NR);
protection_map[1] = __pgprot(__PC | __PP | __NX | ___R);
- protection_map[2] = __pgprot(__PC | __PP | __NX | __NR);
+ protection_map[2] = __pgprot(__PC | __PP | __NX | ___R);
protection_map[3] = __pgprot(__PC | __PP | __NX | ___R);
protection_map[4] = __pgprot(__PC | __PP | ___R);
protection_map[5] = __pgprot(__PC | __PP | ___R);
@@ -169,7 +169,7 @@ static inline void setup_protection_map(void)

protection_map[8] = __pgprot(__PC | __PP | __NX | __NR);
protection_map[9] = __pgprot(__PC | __PP | __NX | ___R);
- protection_map[10] = __pgprot(__PC | __PP | __NX | ___W | __NR);
+ protection_map[10] = __pgprot(__PC | __PP | __NX | ___W | ___R);
protection_map[11] = __pgprot(__PC | __PP | __NX | ___W | ___R);
protection_map[12] = __pgprot(__PC | __PP | ___R);
protection_map[13] = __pgprot(__PC | __PP | ___R);
--
1.8.3.1

2020-06-05 09:41:51

by Jiaxun Yang

[permalink] [raw]
Subject: Re: [PATCH 1/2] MIPS: set page access bit with pgprot on some MIPS platform

On Fri, 5 Jun 2020 17:11:05 +0800
Bibo Mao <[email protected]> wrote:

> On MIPS system which has rixi hardware bit, page access bit is not
> set in pgrot. For memory reading, there will be one page fault to
> allocate physical page; however valid bit is not set, there will
> be the second fast tlb-miss fault handling to set valid/access bit.
>
> This patch set page access/valid bit with pgrot if there is reading
> access privilege. It will reduce one tlb-miss handling for memory
> reading access.
>
> The valid/access bit will be cleared in order to track memory
> accessing activity. If the page is accessed, tlb-miss fast handling
> will set valid/access bit, pte_sw_mkyoung is not necessary in slow
> page fault path. This patch removes pte_sw_mkyoung function which
> is defined as empty function except MIPS system.
>
> Signed-off-by: Bibo Mao <[email protected]>
> ---

Thanks for tracking it down.

Could you please make the patch tittle more clear?
"Some" looks confuse to me, "systems with RIXI" would be better.

- Jiaxun

2020-06-05 09:56:05

by Bibo Mao

[permalink] [raw]
Subject: Re: [PATCH 1/2] MIPS: set page access bit with pgprot on some MIPS platform



On 06/05/2020 05:39 PM, Jiaxun Yang wrote:
> On Fri, 5 Jun 2020 17:11:05 +0800
> Bibo Mao <[email protected]> wrote:
>
>> On MIPS system which has rixi hardware bit, page access bit is not
>> set in pgrot. For memory reading, there will be one page fault to
>> allocate physical page; however valid bit is not set, there will
>> be the second fast tlb-miss fault handling to set valid/access bit.
>>
>> This patch set page access/valid bit with pgrot if there is reading
>> access privilege. It will reduce one tlb-miss handling for memory
>> reading access.
>>
>> The valid/access bit will be cleared in order to track memory
>> accessing activity. If the page is accessed, tlb-miss fast handling
>> will set valid/access bit, pte_sw_mkyoung is not necessary in slow
>> page fault path. This patch removes pte_sw_mkyoung function which
>> is defined as empty function except MIPS system.
>>
>> Signed-off-by: Bibo Mao <[email protected]>
>> ---
>
> Thanks for tracking it down.
>
> Could you please make the patch tittle more clear?
> "Some" looks confuse to me, "systems with RIXI" would be better.

Sure, will add it.

>
> - Jiaxun
>

2020-06-06 01:10:54

by Andrew Morton

[permalink] [raw]
Subject: Re: [PATCH 1/2] MIPS: set page access bit with pgprot on some MIPS platform

On Fri, 5 Jun 2020 17:11:05 +0800 Bibo Mao <[email protected]> wrote:

> On MIPS system which has rixi hardware bit, page access bit is not
> set in pgrot. For memory reading, there will be one page fault to
> allocate physical page; however valid bit is not set, there will
> be the second fast tlb-miss fault handling to set valid/access bit.
>
> This patch set page access/valid bit with pgrot if there is reading
> access privilege. It will reduce one tlb-miss handling for memory
> reading access.
>
> The valid/access bit will be cleared in order to track memory
> accessing activity. If the page is accessed, tlb-miss fast handling
> will set valid/access bit, pte_sw_mkyoung is not necessary in slow
> page fault path. This patch removes pte_sw_mkyoung function which
> is defined as empty function except MIPS system.
>
> ...
>
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -2704,7 +2704,6 @@ static vm_fault_t wp_page_copy(struct vm_fault *vmf)
> }
> flush_cache_page(vma, vmf->address, pte_pfn(vmf->orig_pte));
> entry = mk_pte(new_page, vma->vm_page_prot);
> - entry = pte_sw_mkyoung(entry);
> entry = maybe_mkwrite(pte_mkdirty(entry), vma);
> /*
> * Clear the pte entry and flush it first, before updating the
> @@ -3379,7 +3378,6 @@ static vm_fault_t do_anonymous_page(struct vm_fault *vmf)
> __SetPageUptodate(page);
>
> entry = mk_pte(page, vma->vm_page_prot);
> - entry = pte_sw_mkyoung(entry);
> if (vma->vm_flags & VM_WRITE)
> entry = pte_mkwrite(pte_mkdirty(entry));
>
> @@ -3662,7 +3660,6 @@ vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
>
> flush_icache_page(vma, page);
> entry = mk_pte(page, vma->vm_page_prot);
> - entry = pte_sw_mkyoung(entry);
> if (write)
> entry = maybe_mkwrite(pte_mkdirty(entry), vma);
> /* copy-on-write page */

Only affects mips, so cheerily

Acked-by: Andrew Morton <[email protected]>