Skip to contentSkip to conversation

arm: support get_user_pages_fast

PATCH In review Minchan Kim 3 participants 7 messages Sep 17, 2018 – Sep 28, 2018 Arm (32-bit)

lore
Share & tools
lore
#0 lore raw

cover arm: support get_user_pages_fast

Recently, I got a report get_user_pages_fast helps app's launching time due to reducing uninterruptible sleep time because it could reduce mmap_sem lock contentions when app is launching.

To support gupf in ARM-non-LPAE, first patch reorders memory type table to use 5th bit of the page table.

Second patch introduces L_PTE_SPECIAL for arm so that last patch can support get_user_pags_fast.

I would greatly appreciate if guys review that I screw up something, especially, architecture stuffs.

* from RFC
* Remove CPU_v7M - Catalin
* define DEV_NONSHARED to SHARED - Catalin
* https://lkml.kernel.org/r/20180910165011.zcighyuabshsmxjb at armageddon.cambridge.arm.com

Minchan Kim (3):
arm: mm: reordering memory type table
arm: mm: introduce L_PTE_SPECIAL
arm: mm: support get_user_pages_fast

 arch/arm/Kconfig                      |   2 +-
 arch/arm/include/asm/pgtable-2level.h |  22 ++-
 arch/arm/include/asm/pgtable-3level.h |   6 -
 arch/arm/include/asm/pgtable.h        |  13 ++
 arch/arm/mm/Makefile                  |   6 +
 arch/arm/mm/gup.c                     | 221 ++++++++++++++++++++++++++
 arch/arm/mm/proc-macros.S             |   4 +-
 7 files changed, 259 insertions(+), 15 deletions(-)
 create mode 100644 arch/arm/mm/gup.c
Signature · 1 line
2.19.0.397.gdd90340f6a-goog
#1 lore raw

PATCH 1/3 arm: mm: reordering memory type table

To use bit 5 in page table as L_PTE_SPECIAL, we need a room for that. It seems we don't need 4 bits for the memory type with ARMv6+. If it's true, let's reorder bits to make bit 5 free.

We will use the bit for L_PTE_SPECIAL in next patch.

A note from Catalin
"

Quoted 6 lines

Anyway, on ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to shared device in hardware. Looking through the arm32 code, it seems that MT_DEVICE_NONSHARED is used by arch/arm/mach-shmobile/setup-r8a7779.c and IIUC that's a v7 platform (R-Car H1, Cortex-A9). I think the above should be defined to L_PTE_MT_DEV_SHARED, unless I miss any place where DEV_NONSHARED is relevant on ARMv6 (adding Simon to confirm on shmbile).

"

Cc Russell King <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc Will Deacon <[email protected]>
Cc Steve Capper <[email protected]>
Cc Simon Horman <[email protected]>
---
 arch/arm/include/asm/pgtable-2level.h | 19 +++++++++++++++----
 arch/arm/mm/proc-macros.S             |  4 ++--
 2 files changed, 17 insertions(+), 6 deletions(-)
arch/arm/include/asm/pgtable-2level.h +15 -4
index 92fd2c8a9af0..514b13c27b43 100644
@@ -164,14 +164,25 @@
 #define L_PTE_MT_BUFFERABLE	(_AT(pteval_t, 0x01) << 2)	/* 0001 */
 #define L_PTE_MT_WRITETHROUGH	(_AT(pteval_t, 0x02) << 2)	/* 0010 */
 #define L_PTE_MT_WRITEBACK	(_AT(pteval_t, 0x03) << 2)	/* 0011 */
+#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */
+#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x05) << 2)	/* 0101 */
 #define L_PTE_MT_MINICACHE	(_AT(pteval_t, 0x06) << 2)	/* 0110 (sa1100, xscale) */
 #define L_PTE_MT_WRITEALLOC	(_AT(pteval_t, 0x07) << 2)	/* 0111 */
-#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */
-#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#if defined(CONFIG_CPU_V7) || defined (CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K)
+/*
+ * On ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to
+ * shared device in hardware.
+ */
+#define L_PTE_MT_DEV_NONSHARED	L_PTE_MT_DEV_SHARED
+#define L_PTE_MT_DEV_WC		L_PTE_MT_BUFFERABLE
+#define L_PTE_MT_DEV_CACHED	L_PTE_MT_WRITEBACK
+#define L_PTE_MT_MASK		(_AT(pteval_t, 0x07) << 2)
+#else
 #define L_PTE_MT_DEV_WC		(_AT(pteval_t, 0x09) << 2)	/* 1001 */
 #define L_PTE_MT_DEV_CACHED	(_AT(pteval_t, 0x0b) << 2)	/* 1011 */
-#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x0f) << 2)	/* 1111 */
-#define L_PTE_MT_MASK		(_AT(pteval_t, 0x0f) << 2)
+#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#define L_PTE_MT_MASK           (_AT(pteval_t, 0x0f) << 2)
+#endif
 
 #ifndef __ASSEMBLY__
 
arch/arm/mm/proc-macros.S +2 -2
index 81d0efb055c6..367a89d5aeca 100644
@@ -138,7 +138,7 @@
 	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
 	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
 	.long	PTE_BUFFERABLE					@ L_PTE_MT_DEV_SHARED
-	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
 	.long	0x00						@ L_PTE_MT_MINICACHE (not present)
 	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC
 	.long	0x00						@ unused
@@ -148,7 +148,7 @@
 	.long	PTE_EXT_TEX(2)					@ L_PTE_MT_DEV_NONSHARED
 	.long	0x00						@ unused
 	.long	0x00						@ unused
-	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
+	.long	0x00						@ unused
 	.endm
 
 	.macro	armv6_set_pte_ext pfx
Signature · 1 line
2.19.0.397.gdd90340f6a-goog
#2 lore raw

Re: [PATCH v1 1/3] arm: mm: reordering memory type table

Hi Guys,

Could you have a chance to review this patchset?

Thanks!

Quoted 86 lines · Minchan Kim

To use bit 5 in page table as L_PTE_SPECIAL, we need a room for that. It seems we don't need 4 bits for the memory type with ARMv6+. If it's true, let's reorder bits to make bit 5 free.

We will use the bit for L_PTE_SPECIAL in next patch.

A note from Catalin
"

Quoted 6 lines

Anyway, on ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to shared device in hardware. Looking through the arm32 code, it seems that MT_DEVICE_NONSHARED is used by arch/arm/mach-shmobile/setup-r8a7779.c and IIUC that's a v7 platform (R-Car H1, Cortex-A9). I think the above should be defined to L_PTE_MT_DEV_SHARED, unless I miss any place where DEV_NONSHARED is relevant on ARMv6 (adding Simon to confirm on shmbile).

"

Cc Russell King <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc Will Deacon <[email protected]>
Cc Steve Capper <[email protected]>
Cc Simon Horman <[email protected]>
---
 arch/arm/include/asm/pgtable-2level.h | 19 +++++++++++++++----
 arch/arm/mm/proc-macros.S             |  4 ++--
 2 files changed, 17 insertions(+), 6 deletions(-)
arch/arm/include/asm/pgtable-2level.h +15 -4
index 92fd2c8a9af0..514b13c27b43 100644
@@ -164,14 +164,25 @@
 #define L_PTE_MT_BUFFERABLE	(_AT(pteval_t, 0x01) << 2)	/* 0001 */
 #define L_PTE_MT_WRITETHROUGH	(_AT(pteval_t, 0x02) << 2)	/* 0010 */
 #define L_PTE_MT_WRITEBACK	(_AT(pteval_t, 0x03) << 2)	/* 0011 */
+#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */
+#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x05) << 2)	/* 0101 */
 #define L_PTE_MT_MINICACHE	(_AT(pteval_t, 0x06) << 2)	/* 0110 (sa1100, xscale) */
 #define L_PTE_MT_WRITEALLOC	(_AT(pteval_t, 0x07) << 2)	/* 0111 */
-#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */
-#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#if defined(CONFIG_CPU_V7) || defined (CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K)
+/*
+ * On ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to
+ * shared device in hardware.
+ */
+#define L_PTE_MT_DEV_NONSHARED	L_PTE_MT_DEV_SHARED
+#define L_PTE_MT_DEV_WC		L_PTE_MT_BUFFERABLE
+#define L_PTE_MT_DEV_CACHED	L_PTE_MT_WRITEBACK
+#define L_PTE_MT_MASK		(_AT(pteval_t, 0x07) << 2)
+#else
 #define L_PTE_MT_DEV_WC		(_AT(pteval_t, 0x09) << 2)	/* 1001 */
 #define L_PTE_MT_DEV_CACHED	(_AT(pteval_t, 0x0b) << 2)	/* 1011 */
-#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x0f) << 2)	/* 1111 */
-#define L_PTE_MT_MASK		(_AT(pteval_t, 0x0f) << 2)
+#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#define L_PTE_MT_MASK           (_AT(pteval_t, 0x0f) << 2)
+#endif
 
 #ifndef __ASSEMBLY__
 
arch/arm/mm/proc-macros.S +2 -2
index 81d0efb055c6..367a89d5aeca 100644
@@ -138,7 +138,7 @@
 	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
 	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
 	.long	PTE_BUFFERABLE					@ L_PTE_MT_DEV_SHARED
-	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
 	.long	0x00						@ L_PTE_MT_MINICACHE (not present)
 	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC
 	.long	0x00						@ unused
@@ -148,7 +148,7 @@
 	.long	PTE_EXT_TEX(2)					@ L_PTE_MT_DEV_NONSHARED
 	.long	0x00						@ unused
 	.long	0x00						@ unused
-	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
+	.long	0x00						@ unused
 	.endm
 
 	.macro	armv6_set_pte_ext pfx
Signature · 1 line
2.19.0.397.gdd90340f6a-goog
#3 lore raw

Re: [PATCH v1 1/3] arm: mm: reordering memory type table

In reply to Minchan Kim
Quoted 9 lines · Minchan Kim

To use bit 5 in page table as L_PTE_SPECIAL, we need a room for that. It seems we don't need 4 bits for the memory type with ARMv6+. If it's true, let's reorder bits to make bit 5 free.

We will use the bit for L_PTE_SPECIAL in next patch.

A note from Catalin
"

Quoted 1 line

Anyway, on ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to

I meant 'ARMv7+LPAE' since ARMv6 never had the LPAE feature (please correct the code comment below as well).

I was wrong with the classic ARMv7, only ARMv7+LPAE makes all device memory shareable in hardware (even if not enabled). With classic ARMv7 (that is pre-Cortex-A7/A15), the shareable bit in combination with PRRR allows the Device Non-shareable configuration.

Anyway, it doesn't matter here since the L_PTE_SHARED bit is set separately in the mem_types[] array, the L_PTE_MT_* definitions are just for the actual memory type ignoring shareability. We just need to make sure the comments are correct.

Quoted 5 lines
Quoted 5 lines

shared device in hardware. Looking through the arm32 code, it seems that MT_DEVICE_NONSHARED is used by arch/arm/mach-shmobile/setup-r8a7779.c and IIUC that's a v7 platform (R-Car H1, Cortex-A9). I think the above should be defined to L_PTE_MT_DEV_SHARED, unless I miss any place where DEV_NONSHARED is relevant on ARMv6 (adding Simon to confirm on shmbile).

It would be good to figure out the DEV_NONSHARED on ARMv6 relevance. I don't think we break R-Car H1 since the shareability bit wouldn't be set for DEV_NONSHARED.

Quoted 9 lines
arch/arm/include/asm/pgtable-2level.h +1 -0
index 92fd2c8a9af0..514b13c27b43 100644
@@ -164,14 +164,25 @@
 #define L_PTE_MT_BUFFERABLE	(_AT(pteval_t, 0x01) << 2)	/* 0001 */
 #define L_PTE_MT_WRITETHROUGH	(_AT(pteval_t, 0x02) << 2)	/* 0010 */
 #define L_PTE_MT_WRITEBACK	(_AT(pteval_t, 0x03) << 2)	/* 0011 */
+#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */

arch/arm/include/asm/pgtable-2level.h line 175

Quoted 10 lines
arch/arm/include/asm/pgtable-2level.h +6 -2
+#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x05) << 2)	/* 0101 */
 #define L_PTE_MT_MINICACHE	(_AT(pteval_t, 0x06) << 2)	/* 0110 (sa1100, xscale) */
 #define L_PTE_MT_WRITEALLOC	(_AT(pteval_t, 0x07) << 2)	/* 0111 */
-#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */
-#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#if defined(CONFIG_CPU_V7) || defined (CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K)
+/*
+ * On ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to
+ * shared device in hardware.
+ */

I would change this to something like:

/*
* On ARMv7 or ARMv7+LPAE, the non-shared and shared device types get
* mapped to the same TEX remapping index. On classic ARMv7, the
* shareability is controlled by the PRRR[17:16] field, indexed by
* L_PTE_SHARED. On ARMv7+LPAE the device mapping is always shareable. */

Quoted 29 lines
arch/arm/include/asm/pgtable-2level.h +8 -2
+#define L_PTE_MT_DEV_NONSHARED	L_PTE_MT_DEV_SHARED
+#define L_PTE_MT_DEV_WC		L_PTE_MT_BUFFERABLE
+#define L_PTE_MT_DEV_CACHED	L_PTE_MT_WRITEBACK
+#define L_PTE_MT_MASK		(_AT(pteval_t, 0x07) << 2)
+#else
 #define L_PTE_MT_DEV_WC		(_AT(pteval_t, 0x09) << 2)	/* 1001 */
 #define L_PTE_MT_DEV_CACHED	(_AT(pteval_t, 0x0b) << 2)	/* 1011 */
-#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x0f) << 2)	/* 1111 */
-#define L_PTE_MT_MASK		(_AT(pteval_t, 0x0f) << 2)
+#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#define L_PTE_MT_MASK           (_AT(pteval_t, 0x0f) << 2)
+#endif
 
 #ifndef __ASSEMBLY__
 
arch/arm/mm/proc-macros.S +1 -1
index 81d0efb055c6..367a89d5aeca 100644
@@ -138,7 +138,7 @@
 	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
 	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
 	.long	PTE_BUFFERABLE					@ L_PTE_MT_DEV_SHARED
-	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
 	.long	0x00						@ L_PTE_MT_MINICACHE (not present)
 	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC
 	.long	0x00						@ unused

arch/arm/mm/proc-macros.S line 152

Quoted 6 lines
arch/arm/mm/proc-macros.S +1 -1
@@ -148,7 +148,7 @@
 	.long	PTE_EXT_TEX(2)					@ L_PTE_MT_DEV_NONSHARED
 	.long	0x00						@ unused
 	.long	0x00						@ unused
-	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
+	.long	0x00						@ unused
 	.endm

Looking at the L_PTE_MT_VECTORS uses, I don't think this gives you what you intended. vecs_pgprot in build_mem_type_table() actually combines the cache policy bits with L_PTE_MT_VECTORS and this might have been the reason why it was on the last position (all bits 1). So the default cachepolicy of L_PTE_MT_WRITEBACK or'ed with the new L_PTE_MT_VECTORS gives you 0b0111 which is position 7 instead of 5. This would map onto L_PTE_MT_WRITEALLOC (which is not that bad) but misses the APX bit which marks the vectors page r/w for kernel and ro for user.

I don't think this matters since the kernel no longer writes to the vectors page at run-time but it needs cleaning up a bit (and testing in case I missed something). IOW, do we still need a dedicated mapping type for the vectors or we can simply use the read-only user page attributes?

Signature · 1 line
Catalin
#4 lore raw

Re: [PATCH v1 1/3] arm: mm: reordering memory type table

Hi Catalin,

Sorry for the late response. It was big holiday here.
I will correct what you pointed out and resubmit patch next week.

Thanks for the review!

Quoted 115 lines · Catalin Marinas
Quoted 9 lines · Minchan Kim

To use bit 5 in page table as L_PTE_SPECIAL, we need a room for that. It seems we don't need 4 bits for the memory type with ARMv6+. If it's true, let's reorder bits to make bit 5 free.

We will use the bit for L_PTE_SPECIAL in next patch.

A note from Catalin
"

Quoted 1 line

Anyway, on ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to

I meant 'ARMv7+LPAE' since ARMv6 never had the LPAE feature (please correct the code comment below as well).

I was wrong with the classic ARMv7, only ARMv7+LPAE makes all device memory shareable in hardware (even if not enabled). With classic ARMv7 (that is pre-Cortex-A7/A15), the shareable bit in combination with PRRR allows the Device Non-shareable configuration.

Anyway, it doesn't matter here since the L_PTE_SHARED bit is set separately in the mem_types[] array, the L_PTE_MT_* definitions are just for the actual memory type ignoring shareability. We just need to make sure the comments are correct.

Quoted 5 lines
Quoted 5 lines

shared device in hardware. Looking through the arm32 code, it seems that MT_DEVICE_NONSHARED is used by arch/arm/mach-shmobile/setup-r8a7779.c and IIUC that's a v7 platform (R-Car H1, Cortex-A9). I think the above should be defined to L_PTE_MT_DEV_SHARED, unless I miss any place where DEV_NONSHARED is relevant on ARMv6 (adding Simon to confirm on shmbile).

It would be good to figure out the DEV_NONSHARED on ARMv6 relevance. I don't think we break R-Car H1 since the shareability bit wouldn't be set for DEV_NONSHARED.

Quoted 9 lines
arch/arm/include/asm/pgtable-2level.h +1 -0
index 92fd2c8a9af0..514b13c27b43 100644
@@ -164,14 +164,25 @@
 #define L_PTE_MT_BUFFERABLE	(_AT(pteval_t, 0x01) << 2)	/* 0001 */
 #define L_PTE_MT_WRITETHROUGH	(_AT(pteval_t, 0x02) << 2)	/* 0010 */
 #define L_PTE_MT_WRITEBACK	(_AT(pteval_t, 0x03) << 2)	/* 0011 */
+#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */

arch/arm/include/asm/pgtable-2level.h line 175

Quoted 10 lines
arch/arm/include/asm/pgtable-2level.h +6 -2
+#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x05) << 2)	/* 0101 */
 #define L_PTE_MT_MINICACHE	(_AT(pteval_t, 0x06) << 2)	/* 0110 (sa1100, xscale) */
 #define L_PTE_MT_WRITEALLOC	(_AT(pteval_t, 0x07) << 2)	/* 0111 */
-#define L_PTE_MT_DEV_SHARED	(_AT(pteval_t, 0x04) << 2)	/* 0100 */
-#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#if defined(CONFIG_CPU_V7) || defined (CONFIG_CPU_V6) || defined(CONFIG_CPU_V6K)
+/*
+ * On ARMv7 or ARMv6+LPAE, the non-shared device gets mapped to
+ * shared device in hardware.
+ */

I would change this to something like:

/*
* On ARMv7 or ARMv7+LPAE, the non-shared and shared device types get
* mapped to the same TEX remapping index. On classic ARMv7, the
* shareability is controlled by the PRRR[17:16] field, indexed by
* L_PTE_SHARED. On ARMv7+LPAE the device mapping is always shareable. */

Quoted 29 lines
arch/arm/include/asm/pgtable-2level.h +8 -2
+#define L_PTE_MT_DEV_NONSHARED	L_PTE_MT_DEV_SHARED
+#define L_PTE_MT_DEV_WC		L_PTE_MT_BUFFERABLE
+#define L_PTE_MT_DEV_CACHED	L_PTE_MT_WRITEBACK
+#define L_PTE_MT_MASK		(_AT(pteval_t, 0x07) << 2)
+#else
 #define L_PTE_MT_DEV_WC		(_AT(pteval_t, 0x09) << 2)	/* 1001 */
 #define L_PTE_MT_DEV_CACHED	(_AT(pteval_t, 0x0b) << 2)	/* 1011 */
-#define L_PTE_MT_VECTORS	(_AT(pteval_t, 0x0f) << 2)	/* 1111 */
-#define L_PTE_MT_MASK		(_AT(pteval_t, 0x0f) << 2)
+#define L_PTE_MT_DEV_NONSHARED	(_AT(pteval_t, 0x0c) << 2)	/* 1100 */
+#define L_PTE_MT_MASK           (_AT(pteval_t, 0x0f) << 2)
+#endif
 
 #ifndef __ASSEMBLY__
 
arch/arm/mm/proc-macros.S +1 -1
index 81d0efb055c6..367a89d5aeca 100644
@@ -138,7 +138,7 @@
 	.long	PTE_CACHEABLE					@ L_PTE_MT_WRITETHROUGH
 	.long	PTE_CACHEABLE | PTE_BUFFERABLE			@ L_PTE_MT_WRITEBACK
 	.long	PTE_BUFFERABLE					@ L_PTE_MT_DEV_SHARED
-	.long	0x00						@ unused
+	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
 	.long	0x00						@ L_PTE_MT_MINICACHE (not present)
 	.long	PTE_EXT_TEX(1) | PTE_CACHEABLE | PTE_BUFFERABLE	@ L_PTE_MT_WRITEALLOC
 	.long	0x00						@ unused

arch/arm/mm/proc-macros.S line 152

Quoted 6 lines
arch/arm/mm/proc-macros.S +1 -1
@@ -148,7 +148,7 @@
 	.long	PTE_EXT_TEX(2)					@ L_PTE_MT_DEV_NONSHARED
 	.long	0x00						@ unused
 	.long	0x00						@ unused
-	.long	PTE_CACHEABLE | PTE_BUFFERABLE | PTE_EXT_APX	@ L_PTE_MT_VECTORS
+	.long	0x00						@ unused
 	.endm

Looking at the L_PTE_MT_VECTORS uses, I don't think this gives you what you intended. vecs_pgprot in build_mem_type_table() actually combines the cache policy bits with L_PTE_MT_VECTORS and this might have been the reason why it was on the last position (all bits 1). So the default cachepolicy of L_PTE_MT_WRITEBACK or'ed with the new L_PTE_MT_VECTORS gives you 0b0111 which is position 7 instead of 5. This would map onto L_PTE_MT_WRITEALLOC (which is not that bad) but misses the APX bit which marks the vectors page r/w for kernel and ro for user.

I don't think this matters since the kernel no longer writes to the vectors page at run-time but it needs cleaning up a bit (and testing in case I missed something). IOW, do we still need a dedicated mapping type for the vectors or we can simply use the read-only user page attributes?

Signature · 1 line
Catalin
#5 lore raw

PATCH 2/3 arm: mm: introduce L_PTE_SPECIAL

In reply to Minchan Kim

This patch introduces L_PTE_SPECIAL and pte functions for supporting get_user_pages_fast.

Cc Russell King <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc Will Deacon <[email protected]>
Cc Steve Capper <[email protected]>
---
 arch/arm/Kconfig                      |  2 +-
 arch/arm/include/asm/pgtable-2level.h |  3 +--
 arch/arm/include/asm/pgtable-3level.h |  6 ------
 arch/arm/include/asm/pgtable.h        | 13 +++++++++++++
 4 files changed, 15 insertions(+), 9 deletions(-)
arch/arm/Kconfig +1 -1
index e8cd55a5b04c..278cf22b5aa1 100644
@@ -10,7 +10,7 @@ config ARM
 	select ARCH_HAS_FORTIFY_SOURCE
 	select ARCH_HAS_KCOV
 	select ARCH_HAS_MEMBARRIER_SYNC_CORE
-	select ARCH_HAS_PTE_SPECIAL if ARM_LPAE
+	select ARCH_HAS_PTE_SPECIAL if (ARM_LPAE || CPU_V7 || CPU_V6 || CPUV6K)
 	select ARCH_HAS_PHYS_TO_DMA
 	select ARCH_HAS_SET_MEMORY
 	select ARCH_HAS_STRICT_KERNEL_RWX if MMU && !XIP_KERNEL
arch/arm/include/asm/pgtable-2level.h +1 -2
index 514b13c27b43..378668420642 100644
@@ -120,6 +120,7 @@
 #define L_PTE_VALID		(_AT(pteval_t, 1) << 0)		/* Valid */
 #define L_PTE_PRESENT		(_AT(pteval_t, 1) << 0)
 #define L_PTE_YOUNG		(_AT(pteval_t, 1) << 1)
+#define L_PTE_SPECIAL		(_AT(pteval_t, 1) << 5)
 #define L_PTE_DIRTY		(_AT(pteval_t, 1) << 6)
 #define L_PTE_RDONLY		(_AT(pteval_t, 1) << 7)
 #define L_PTE_USER		(_AT(pteval_t, 1) << 8)
@@ -224,8 +225,6 @@ static inline pmd_t *pmd_offset(pud_t *pud, unsigned long addr)
 #define pmd_addr_end(addr,end) (end)
 
 #define set_pte_ext(ptep,pte,ext) cpu_set_pte_ext(ptep,pte,ext)
-#define pte_special(pte)	(0)
-static inline pte_t pte_mkspecial(pte_t pte) { return pte; }
 
 /*
  * We don't have huge page support for short descriptors, for the moment
arch/arm/include/asm/pgtable-3level.h +0 -6
index 6d50a11d7793..b6f52e16b478 100644
@@ -213,12 +213,6 @@ static inline pmd_t *pmd_offset(pud_t *pud, unsigned long addr)
 
 #define pmd_present(pmd)	(pmd_isset((pmd), L_PMD_SECT_VALID))
 #define pmd_young(pmd)		(pmd_isset((pmd), PMD_SECT_AF))
-#define pte_special(pte)	(pte_isset((pte), L_PTE_SPECIAL))
-static inline pte_t pte_mkspecial(pte_t pte)
-{
-	pte_val(pte) |= L_PTE_SPECIAL;
-	return pte;
-}
 
 #define pmd_write(pmd)		(pmd_isclear((pmd), L_PMD_SECT_RDONLY))
 #define pmd_dirty(pmd)		(pmd_isset((pmd), L_PMD_SECT_DIRTY))
arch/arm/include/asm/pgtable.h +13 -0
index a757401129f9..6cc7ce0e423e 100644
@@ -228,6 +228,11 @@ static inline pte_t *pmd_page_vaddr(pmd_t pmd)
 #define pte_dirty(pte)		(pte_isset((pte), L_PTE_DIRTY))
 #define pte_young(pte)		(pte_isset((pte), L_PTE_YOUNG))
 #define pte_exec(pte)		(pte_isclear((pte), L_PTE_XN))
+#ifdef CONFIG_ARCH_HAS_PTE_SPECIAL
+#define pte_special(pte)	(pte_isset((pte), L_PTE_SPECIAL))
+#else
+#define pte_special(pte)	(0)
+#endif
 
 #define pte_valid_user(pte)	\
 	(pte_valid(pte) && pte_isset((pte), L_PTE_USER) && pte_young(pte))
@@ -318,6 +323,14 @@ static inline pte_t pte_mknexec(pte_t pte)
 	return set_pte_bit(pte, __pgprot(L_PTE_XN));
 }
 
+#ifdef CONFIG_ARCH_HAS_PTE_SPECIAL
+static inline pte_t pte_mkspecial(pte_t pte)
+{
+	return set_pte_bit(pte, __pgprot(L_PTE_SPECIAL));
+}
+#else
+static inline pte_t pte_mkspecial(pte_t pte) { return pte; }
+#endif
 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 {
 	const pteval_t mask = L_PTE_XN | L_PTE_RDONLY | L_PTE_USER |
Signature · 1 line
2.19.0.397.gdd90340f6a-goog
#6 lore raw

PATCH 3/3 arm: mm: support get_user_pages_fast

In reply to Minchan Kim

Recently, there was a report get_user_pages_fast helps app launching speed due to reducing uninterruptible sleep time because we don't need to contend for mmap_sem, I believe.

With get_user_pages_fast, that uniterruptible sleep time is reduced about 5~10% by testing.

Cc Russell King <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc Will Deacon <[email protected]>
Cc Steve Capper <[email protected]>
---
 arch/arm/mm/Makefile |   6 ++
 arch/arm/mm/gup.c    | 221 +++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 227 insertions(+)
 create mode 100644 arch/arm/mm/gup.c
arch/arm/mm/Makefile +6 -0
index 7cb1699fbfc4..f55f96d56843 100644
@@ -13,6 +13,12 @@ obj-y				+= nommu.o
 obj-$(CONFIG_ARM_MPU)		+= pmsa-v7.o pmsa-v8.o
 endif
 
+ifneq ($(CONFIG_ARM_LPAE),y)
+ifeq ($(CONFIG_ARCH_HAS_PTE_SPECIAL),y)
+obj-$(CONFIG_MMU)		+= gup.o
+endif
+endif
+
 obj-$(CONFIG_ARM_PTDUMP_CORE)	+= dump.o
 obj-$(CONFIG_ARM_PTDUMP_DEBUGFS)	+= ptdump_debugfs.o
 obj-$(CONFIG_MODULES)		+= proc-syms.o
arch/arm/mm/gup.c new file +221 -0

221 lines not on this page · raw message

Signature · 1 line
2.19.0.397.gdd90340f6a-goog