Skip to contentSkip to conversation

powerpc/32: Use BATs/LTLBs for STRICT_KERNEL_RWX

PATCH Superseded Christophe Leroy 2 participants 20 messages Feb 20, 2019 – Feb 21, 2019 PowerPC

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

v3 cover powerpc/32: Use BATs/LTLBs for STRICT_KERNEL_RWX

The purpose of this serie is to:
- use BATs with STRICT_KERNEL_RWX on book3s (See patch 13 for details.)
- use LTLBs with STRICT_KERNEL_RWX on 8xx (See patch 15 for a few details.)

v3:
- Reordered to avoid build failure due to setibat() not being used for several steps in the serie. Now the patch using setibat() is next to the one adding setibat().
- Fixed mmu_mapin_ram() in patch 3 to return base in all cases, thanks Jonathan for the test
- Fixed build failure on 8xx when CONFIG_PERF_EVENTS is set due to too many instructions in Exception 0x1200
- Made 8M alignment for data the default on 8xx when STRICT_KERNEL_RWX is selected.
- Added patch 1 to not set additionnal bat on the wii when requesting nobats. The only purpose of this patch is to be backported, as this function is removed later in the series.

v2:
- Fix patch 2 (was patch 3 in v1) based on feedback from Jonathan.
- Added support for 8xx with LTLBs.
- Added systematic population of pagetables for Abatron BDI.

Christophe Leroy (16):
powerpc/wii: properly disable use of BATs when requested. powerpc/mm/32: add base address to mmu_mapin_ram() powerpc/mm/32s: rework mmu_mapin_ram()
powerpc/mm/32s: use generic mmu_mapin_ram() for all blocks. powerpc/32: always populate page tables for Abatron BDI. powerpc/wii: remove wii_mmu_mapin_mem2()
powerpc/mm/32s: use _PAGE_EXEC in setbat()
powerpc/32: add helper to write into segment registers powerpc/mmu: add is_strict_kernel_rwx() helper powerpc/kconfig: define PAGE_SHIFT inside Kconfig powerpc/kconfig: define CONFIG_DATA_SHIFT and CONFIG_ETEXT_SHIFT powerpc/mm/32s: add setibat() clearibat() and update_bats() powerpc/mm/32s: Use BATs for STRICT_KERNEL_RWX powerpc/kconfig: make _etext and data areas alignment configurable on

    Book3s 32
  powerpc/8xx: don't disable large TLBs with CONFIG_STRICT_KERNEL_RWX
  powerpc/kconfig: make _etext and data areas alignment configurable on
    8xx
 arch/powerpc/Kconfig                          |  60 +++++++++
 arch/powerpc/include/asm/book3s/32/mmu-hash.h |   2 +
 arch/powerpc/include/asm/book3s/32/pgtable.h  |  11 ++
 arch/powerpc/include/asm/mmu.h                |  11 ++
 arch/powerpc/include/asm/nohash/32/mmu-8xx.h  |   3 +-
 arch/powerpc/include/asm/page.h               |  13 +-
 arch/powerpc/include/asm/reg.h                |   5 +
 arch/powerpc/kernel/head_32.S                 |  35 +++++
 arch/powerpc/kernel/head_8xx.S                |  54 ++++++--
 arch/powerpc/kernel/vmlinux.lds.S             |   9 +-
 arch/powerpc/mm/40x_mmu.c                     |   2 +-
 arch/powerpc/mm/44x_mmu.c                     |   2 +-
 arch/powerpc/mm/8xx_mmu.c                     |  33 ++++-
 arch/powerpc/mm/fsl_booke_mmu.c               |   2 +-
 arch/powerpc/mm/init_32.c                     |   6 +-
 arch/powerpc/mm/mmu_decl.h                    |  10 +-
 arch/powerpc/mm/pgtable_32.c                  |  38 +++---
 arch/powerpc/mm/ppc_mmu_32.c                  | 180 ++++++++++++++++++++++----
 arch/powerpc/platforms/embedded6xx/wii.c      |  24 ----
 19 files changed, 390 insertions(+), 110 deletions(-)
Signature · 1 line
2.13.3
#1 lore raw

PATCH v3 1/16 powerpc/wii: properly disable use of BATs when requested.

'nobats' kernel parameter or some options like CONFIG_DEBUG_PAGEALLOC deny the use of BATS for mapping memory.

This patch makes sure that the specific wii RAM mapping function takes it into account as well.

Fixes de32400dd26e ("wii: use both mem1 and mem2 as ram")
---
 arch/powerpc/platforms/embedded6xx/wii.c | 4 ++++
 1 file changed, 4 insertions(+)
arch/powerpc/platforms/embedded6xx/wii.c +4 -0
index ecf703ee3a76..ac4ee88efc80 100644
@@ -83,6 +83,10 @@ unsigned long __init wii_mmu_mapin_mem2(unsigned long top)
 	/* MEM2 64MB@0x10000000 */
 	delta = wii_hole_start + wii_hole_size;
 	size = top - delta;
+
+	if (__map_without_bats)
+		return delta;
+
 	for (bl = 128<<10; bl < max_size; bl <<= 1) {
 		if (bl * 2 > size)
 			break;
Signature · 1 line
2.13.3
#2 lore raw

PATCH v3 2/16 powerpc/mm/32: add base address to mmu_mapin_ram()

In reply to Christophe Leroy

At the time being, mmu_mapin_ram() always maps RAM from the beginning. But some platforms like the WII have to map a second block of RAM.

This patch adds to mmu_mapin_ram() the base address of the block. At the moment, only base address 0 is supported.

---
 arch/powerpc/mm/40x_mmu.c       | 2 +-
 arch/powerpc/mm/44x_mmu.c       | 2 +-
 arch/powerpc/mm/8xx_mmu.c       | 2 +-
 arch/powerpc/mm/fsl_booke_mmu.c | 2 +-
 arch/powerpc/mm/mmu_decl.h      | 2 +-
 arch/powerpc/mm/pgtable_32.c    | 6 +++---
 arch/powerpc/mm/ppc_mmu_32.c    | 2 +-
 7 files changed, 9 insertions(+), 9 deletions(-)
arch/powerpc/mm/40x_mmu.c +1 -1
index 61ac468c87c6..b9cf6f8764b0 100644
@@ -93,7 +93,7 @@ void __init MMU_init_hw(void)
 #define LARGE_PAGE_SIZE_16M	(1<<24)
 #define LARGE_PAGE_SIZE_4M	(1<<22)
 
-unsigned long __init mmu_mapin_ram(unsigned long top)
+unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	unsigned long v, s, mapped;
 	phys_addr_t p;
arch/powerpc/mm/44x_mmu.c +1 -1
index ea2b9af08a48..aad127acdbaa 100644
@@ -170,7 +170,7 @@ void __init MMU_init_hw(void)
 	flush_instruction_cache();
 }
 
-unsigned long __init mmu_mapin_ram(unsigned long top)
+unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	unsigned long addr;
 	unsigned long memstart = memstart_addr & ~(PPC_PIN_SIZE - 1);
arch/powerpc/mm/8xx_mmu.c +1 -1
index e2c32bdb6023..46bc26ef71e9 100644
@@ -99,7 +99,7 @@ static void __init mmu_patch_cmp_limit(s32 *site, unsigned long mapped)
 	modify_instruction_site(site, 0xffff, (unsigned long)__va(mapped) >> 16);
 }
 
-unsigned long __init mmu_mapin_ram(unsigned long top)
+unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	unsigned long mapped;
 
arch/powerpc/mm/fsl_booke_mmu.c +1 -1
index 080d49b26c3a..210cbc1faf63 100644
@@ -221,7 +221,7 @@ unsigned long map_mem_in_cams(unsigned long ram, int max_cam_idx, bool dryrun)
 #error "LOWMEM_CAM_NUM must be less than NUM_TLBCAMS"
 #endif
 
-unsigned long __init mmu_mapin_ram(unsigned long top)
+unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	return tlbcam_addrs[tlbcam_index - 1].limit - PAGE_OFFSET + 1;
 }
arch/powerpc/mm/mmu_decl.h +1 -1
index c4a717da65eb..61730023dde3 100644
@@ -130,7 +130,7 @@ extern void wii_memory_fixups(void);
  */
 #ifdef CONFIG_PPC32
 extern void MMU_init_hw(void);
-extern unsigned long mmu_mapin_ram(unsigned long top);
+unsigned long mmu_mapin_ram(unsigned long base, unsigned long top);
 #endif
 
 #ifdef CONFIG_PPC_FSL_BOOK3E
arch/powerpc/mm/pgtable_32.c +3 -3
index ded71126ce4c..b4858818523f 100644
@@ -258,15 +258,15 @@ void __init mapin_ram(void)
 
 #ifndef CONFIG_WII
 	top = total_lowmem;
-	s = mmu_mapin_ram(top);
+	s = mmu_mapin_ram(0, top);
 	__mapin_ram_chunk(s, top);
 #else
 	if (!wii_hole_size) {
-		s = mmu_mapin_ram(total_lowmem);
+		s = mmu_mapin_ram(0, total_lowmem);
 		__mapin_ram_chunk(s, total_lowmem);
 	} else {
 		top = wii_hole_start;
-		s = mmu_mapin_ram(top);
+		s = mmu_mapin_ram(0, top);
 		__mapin_ram_chunk(s, top);
 
 		top = memblock_end_of_DRAM();
arch/powerpc/mm/ppc_mmu_32.c +1 -1
index 3f4193201ee7..b260ced065b4 100644
@@ -73,7 +73,7 @@ unsigned long p_block_mapped(phys_addr_t pa)
 	return 0;
 }
 
-unsigned long __init mmu_mapin_ram(unsigned long top)
+unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	unsigned long tot, bl, done;
 	unsigned long max_size = (256<<20);
Signature · 1 line
2.13.3
#3 lore raw

PATCH v3 3/16 powerpc/mm/32s: rework mmu_mapin_ram()

In reply to Christophe Leroy

This patch reworks mmu_mapin_ram() to be more generic and map as much blocks as possible. It now supports blocks not starting at address 0.

It scans DBATs array to find free ones instead of forcing the use of BAT2 and BAT3.

---
 arch/powerpc/mm/ppc_mmu_32.c | 63 ++++++++++++++++++++++++++++----------------
 1 file changed, 41 insertions(+), 22 deletions(-)
arch/powerpc/mm/ppc_mmu_32.c +41 -22
index b260ced065b4..5fc59b195fef 100644
@@ -73,39 +73,58 @@ unsigned long p_block_mapped(phys_addr_t pa)
 	return 0;
 }
 
+static int find_free_bat(void)
+{
+	int b;
+
+	if (cpu_has_feature(CPU_FTR_601)) {
+		for (b = 0; b < 4; b++) {
+			struct ppc_bat *bat = BATS[b];
+
+			if (!(bat[0].batl & 0x40))
+				return b;
+		}
+	} else {
+		int n = mmu_has_feature(MMU_FTR_USE_HIGH_BATS) ? 8 : 4;
+
+		for (b = 0; b < n; b++) {
+			struct ppc_bat *bat = BATS[b];
+
+			if (!(bat[1].batu & 3))
+				return b;
+		}
+	}
+	return -1;
+}
+
+static unsigned int block_size(unsigned long base, unsigned long top)
+{
+	unsigned int max_size = (cpu_has_feature(CPU_FTR_601) ? 8 : 256) << 20;
+	unsigned int base_shift = (fls(base) - 1) & 31;
+	unsigned int block_shift = (fls(top - base) - 1) & 31;
+
+	return min3(max_size, 1U << base_shift, 1U << block_shift);
+}
+
 unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
-	unsigned long tot, bl, done;
-	unsigned long max_size = (256<<20);
+	int idx;
 
 	if (__map_without_bats) {
 		printk(KERN_DEBUG "RAM mapped without BATs\n");
-		return 0;
+		return base;
 	}
 
-	/* Set up BAT2 and if necessary BAT3 to cover RAM. */
+	while ((idx = find_free_bat()) != -1 && base != top) {
+		unsigned int size = block_size(base, top);
 
-	/* Make sure we don't map a block larger than the
-	   smallest alignment of the physical address. */
-	tot = top;
-	for (bl = 128<<10; bl < max_size; bl <<= 1) {
-		if (bl * 2 > tot)
+		if (size < 128 << 10)
 			break;
+		setbat(idx, PAGE_OFFSET + base, base, size, PAGE_KERNEL_X);
+		base += size;
 	}
 
-	setbat(2, PAGE_OFFSET, 0, bl, PAGE_KERNEL_X);
-	done = (unsigned long)bat_addrs[2].limit - PAGE_OFFSET + 1;
-	if ((done < tot) && !bat_addrs[3].limit) {
-		/* use BAT3 to cover a bit more */
-		tot -= done;
-		for (bl = 128<<10; bl < max_size; bl <<= 1)
-			if (bl * 2 > tot)
-				break;
-		setbat(3, PAGE_OFFSET+done, done, bl, PAGE_KERNEL_X);
-		done = (unsigned long)bat_addrs[3].limit - PAGE_OFFSET + 1;
-	}
-
-	return done;
+	return base;
 }
 
 /*
Signature · 1 line
2.13.3
#4 lore raw

PATCH v3 4/16 powerpc/mm/32s: use generic mmu_mapin_ram() for all blocks.

In reply to Christophe Leroy

Now that mmu_mapin_ram() is able to handle other blocks than the one starting at 0, the WII can use it for all its blocks.

---
 arch/powerpc/mm/pgtable_32.c | 25 +++++++------------------
 1 file changed, 7 insertions(+), 18 deletions(-)
arch/powerpc/mm/pgtable_32.c +7 -18
index b4858818523f..fd665c32a1f7 100644
@@ -254,26 +254,15 @@ static void __init __mapin_ram_chunk(unsigned long offset, unsigned long top)
 
 void __init mapin_ram(void)
 {
-	unsigned long s, top;
-
-#ifndef CONFIG_WII
-	top = total_lowmem;
-	s = mmu_mapin_ram(0, top);
-	__mapin_ram_chunk(s, top);
-#else
-	if (!wii_hole_size) {
-		s = mmu_mapin_ram(0, total_lowmem);
-		__mapin_ram_chunk(s, total_lowmem);
-	} else {
-		top = wii_hole_start;
-		s = mmu_mapin_ram(0, top);
-		__mapin_ram_chunk(s, top);
+	struct memblock_region *reg;
+
+	for_each_memblock(memory, reg) {
+		unsigned long base = reg->base;
+		unsigned long top = base + reg->size;
 
-		top = memblock_end_of_DRAM();
-		s = wii_mmu_mapin_mem2(top);
-		__mapin_ram_chunk(s, top);
+		base = mmu_mapin_ram(base, top);
+		__mapin_ram_chunk(base, top);
 	}
-#endif
 }
 
 /* Scan the real Linux page tables and return a PTE pointer for
Signature · 1 line
2.13.3
#5 lore raw

PATCH v3 5/16 powerpc/32: always populate page tables for Abatron BDI.

In reply to Christophe Leroy

When CONFIG_BDI_SWITCH is set, the page tables have to be populated allthough large TLBs are used, because the BDI switch knows nothing about those large TLBs which are handled directly in TLB miss logic.

---
 arch/powerpc/mm/pgtable_32.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
arch/powerpc/mm/pgtable_32.c +4 -1
index fd665c32a1f7..94bd7d013557 100644
@@ -261,7 +261,10 @@ void __init mapin_ram(void)
 		unsigned long top = base + reg->size;
 
 		base = mmu_mapin_ram(base, top);
-		__mapin_ram_chunk(base, top);
+		if (IS_ENABLED(CONFIG_BDI_SWITCH))
+			__mapin_ram_chunk(reg->base, top);
+		else
+			__mapin_ram_chunk(base, top);
 	}
 }
 
Signature · 1 line
2.13.3
#6 lore raw

PATCH v3 6/16 powerpc/wii: remove wii_mmu_mapin_mem2()

In reply to Christophe Leroy

wii_mmu_mapin_mem2() is not used anymore, remove it.

---
 arch/powerpc/platforms/embedded6xx/wii.c | 28 ----------------------------
 1 file changed, 28 deletions(-)
arch/powerpc/platforms/embedded6xx/wii.c +0 -28
index ac4ee88efc80..235fe81aa2b1 100644
@@ -54,10 +54,6 @@
 static void __iomem *hw_ctrl;
 static void __iomem *hw_gpio;
 
-unsigned long wii_hole_start;
-unsigned long wii_hole_size;
-
-
 static int __init page_aligned(unsigned long x)
 {
 	return !(x & (PAGE_SIZE-1));
@@ -69,30 +65,6 @@ void __init wii_memory_fixups(void)
 
 	BUG_ON(memblock.memory.cnt != 2);
 	BUG_ON(!page_aligned(p[0].base) || !page_aligned(p[1].base));
-
-	/* determine hole */
-	wii_hole_start = ALIGN(p[0].base + p[0].size, PAGE_SIZE);
-	wii_hole_size = p[1].base - wii_hole_start;
-}
-
-unsigned long __init wii_mmu_mapin_mem2(unsigned long top)
-{
-	unsigned long delta, size, bl;
-	unsigned long max_size = (256<<20);
-
-	/* MEM2 64MB@0x10000000 */
-	delta = wii_hole_start + wii_hole_size;
-	size = top - delta;
-
-	if (__map_without_bats)
-		return delta;
-
-	for (bl = 128<<10; bl < max_size; bl <<= 1) {
-		if (bl * 2 > size)
-			break;
-	}
-	setbat(4, PAGE_OFFSET+delta, delta, bl, PAGE_KERNEL_X);
-	return delta + bl;
 }
 
 static void __noreturn wii_spin(void)
Signature · 1 line
2.13.3
#7 lore raw

PATCH v3 7/16 powerpc/mm/32s: use _PAGE_EXEC in setbat()

In reply to Christophe Leroy

Do not set IBAT when setbat() is called without _PAGE_EXEC

---
 arch/powerpc/mm/ppc_mmu_32.c | 10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)
arch/powerpc/mm/ppc_mmu_32.c +6 -4
index 5fc59b195fef..ff8580c6ab11 100644
@@ -131,6 +131,7 @@ unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
  * Set up one of the I/D BAT (block address translation) register pairs.
  * The parameters are not checked; in particular size must be a power
  * of 2 between 128k and 256M.
+ * On 603+, only set IBAT when _PAGE_EXEC is set
  */
 void __init setbat(int index, unsigned long virt, phys_addr_t phys,
 		   unsigned int size, pgprot_t prot)
@@ -157,11 +158,12 @@ void __init setbat(int index, unsigned long virt, phys_addr_t phys,
 			bat[1].batu |= 1; 	/* Vp = 1 */
 		if (flags & _PAGE_GUARDED) {
 			/* G bit must be zero in IBATs */
-			bat[0].batu = bat[0].batl = 0;
-		} else {
-			/* make IBAT same as DBAT */
-			bat[0] = bat[1];
+			flags &= ~_PAGE_EXEC;
 		}
+		if (flags & _PAGE_EXEC)
+			bat[0] = bat[1];
+		else
+			bat[0].batu = bat[0].batl = 0;
 	} else {
 		/* 601 cpu */
 		if (bl > BL_8M)
Signature · 1 line
2.13.3
#8 lore raw

PATCH v3 8/16 powerpc/32: add helper to write into segment registers

In reply to Christophe Leroy

This patch add an helper which wraps 'mtsrin' instruction to write into segment registers.

---
 arch/powerpc/include/asm/reg.h | 5 +++++
 1 file changed, 5 insertions(+)
arch/powerpc/include/asm/reg.h +5 -0
index 1c98ef1f2d5b..a70cbaf5c26f 100644
@@ -1425,6 +1425,11 @@ static inline void msr_check_and_clear(unsigned long bits)
 #define mfsrin(v)	({unsigned int rval; \
 			asm volatile("mfsrin %0,%1" : "=r" (rval) : "r" (v)); \
 					rval;})
+
+static inline void mtsrin(u32 val, u32 idx)
+{
+	asm volatile("mtsrin %0, %1" : : "r" (val), "r" (idx));
+}
 #endif
 
 #define proc_trap()	asm volatile("trap")
Signature · 1 line
2.13.3
#9 lore raw

PATCH v3 9/16 powerpc/mmu: add is_strict_kernel_rwx() helper

In reply to Christophe Leroy

Add a helper to know whether STRICT_KERNEL_RWX is enabled.

This is based on rodata_enabled flag which is defined only when CONFIG_STRICT_KERNEL_RWX is selected.

---
 arch/powerpc/include/asm/mmu.h | 11 +++++++++++
 arch/powerpc/mm/init_32.c      |  4 +---
 2 files changed, 12 insertions(+), 3 deletions(-)
arch/powerpc/include/asm/mmu.h +11 -0
index 6d22a8e78fe2..d34ad1657d7b 100644
@@ -289,6 +289,17 @@ static inline u16 get_mm_addr_key(struct mm_struct *mm, unsigned long address)
 }
 #endif /* CONFIG_PPC_MEM_KEYS */
 
+#ifdef CONFIG_STRICT_KERNEL_RWX
+static inline bool strict_kernel_rwx_enabled(void)
+{
+	return rodata_enabled;
+}
+#else
+static inline bool strict_kernel_rwx_enabled(void)
+{
+	return false;
+}
+#endif
 #endif /* !__ASSEMBLY__ */
 
 /* The kernel use the constants below to index in the page sizes array.
arch/powerpc/mm/init_32.c +1 -3
index 3e59e5d64b01..ee5a430b9a18 100644
@@ -108,12 +108,10 @@ static void __init MMU_setup(void)
 		__map_without_bats = 1;
 		__map_without_ltlbs = 1;
 	}
-#ifdef CONFIG_STRICT_KERNEL_RWX
-	if (rodata_enabled) {
+	if (strict_kernel_rwx_enabled()) {
 		__map_without_bats = 1;
 		__map_without_ltlbs = 1;
 	}
-#endif
 }
 
 /*
Signature · 1 line
2.13.3
#10 lore raw

PATCH v3 10/16 powerpc/kconfig: define PAGE_SHIFT inside Kconfig

In reply to Christophe Leroy

This patch defined CONFIG_PPC_PAGE_SHIFT in order to be able to use PAGE_SHIFT value inside Kconfig.

---
 arch/powerpc/Kconfig            |  7 +++++++
 arch/powerpc/include/asm/page.h | 13 ++-----------
 2 files changed, 9 insertions(+), 11 deletions(-)
arch/powerpc/Kconfig +7 -0
index 849b0d5ac3d1..417e52a27f63 100644
@@ -708,6 +708,13 @@ config PPC_256K_PAGES
 
 endchoice
 
+config PPC_PAGE_SHIFT
+	int
+	default 18 if PPC_256K_PAGES
+	default 16 if PPC_64K_PAGES
+	default 14 if PPC_16K_PAGES
+	default 12
+
 config THREAD_SHIFT
 	int "Thread shift" if EXPERT
 	range 13 15
arch/powerpc/include/asm/page.h +2 -11
index aa4497175bd3..ed870468ef6f 100644
@@ -20,20 +20,11 @@
 
 /*
  * On regular PPC32 page size is 4K (but we support 4K/16K/64K/256K pages
- * on PPC44x). For PPC64 we support either 4K or 64K software
+ * on PPC44x and 4K/16K on 8xx). For PPC64 we support either 4K or 64K software
  * page size. When using 64K pages however, whether we are really supporting
  * 64K pages in HW or not is irrelevant to those definitions.
  */
-#if defined(CONFIG_PPC_256K_PAGES)
-#define PAGE_SHIFT		18
-#elif defined(CONFIG_PPC_64K_PAGES)
-#define PAGE_SHIFT		16
-#elif defined(CONFIG_PPC_16K_PAGES)
-#define PAGE_SHIFT		14
-#else
-#define PAGE_SHIFT		12
-#endif
-
+#define PAGE_SHIFT		CONFIG_PPC_PAGE_SHIFT
 #define PAGE_SIZE		(ASM_CONST(1) << PAGE_SHIFT)
 
 #ifndef __ASSEMBLY__
Signature · 1 line
2.13.3
#11 lore raw

PATCH v3 11/16 powerpc/kconfig: define CONFIG_DATA_SHIFT and CONFIG_ETEXT_SHIFT

In reply to Christophe Leroy

CONFIG_STRICT_KERNEL_RWX requires a special alignment for DATA for some subarches. Today it is just defined as an #ifdef in vmlinux.lds.S

In order to get more flexibility, this patch moves the definition of this alignment in Kconfig

On some subarches, CONFIG_STRICT_KERNEL_RWX will require a special alignment of _etext.

This patch also adds a configuration item for it in Kconfig

---
 arch/powerpc/Kconfig              | 9 +++++++++
 arch/powerpc/kernel/vmlinux.lds.S | 9 +++------
 2 files changed, 12 insertions(+), 6 deletions(-)
arch/powerpc/Kconfig +9 -0
index 417e52a27f63..edef40a2b446 100644
@@ -725,6 +725,15 @@ config THREAD_SHIFT
 	  Used to define the stack size. The default is almost always what you
 	  want. Only change this if you know what you are doing.
 
+config ETEXT_SHIFT
+	int
+	default PPC_PAGE_SHIFT
+
+config DATA_SHIFT
+	int
+	default 24 if STRICT_KERNEL_RWX && PPC64
+	default PPC_PAGE_SHIFT
+
 config FORCE_MAX_ZONEORDER
 	int "Maximum zone order"
 	range 8 9 if PPC64 && PPC_64K_PAGES
arch/powerpc/kernel/vmlinux.lds.S +3 -6
index c3efb972c8c1..060a1acd7c6d 100644
@@ -12,11 +12,8 @@
 #include <asm/cache.h>
 #include <asm/thread_info.h>
 
-#if defined(CONFIG_STRICT_KERNEL_RWX) && !defined(CONFIG_PPC32)
-#define STRICT_ALIGN_SIZE	(1 << 24)
-#else
-#define STRICT_ALIGN_SIZE	PAGE_SIZE
-#endif
+#define STRICT_ALIGN_SIZE	(1 << CONFIG_DATA_SHIFT)
+#define ETEXT_ALIGN_SIZE	(1 << CONFIG_ETEXT_SHIFT)
 
 ENTRY(_stext)
 
@@ -131,7 +128,7 @@ SECTIONS
 
 	} :kernel
 
-	. = ALIGN(PAGE_SIZE);
+	. = ALIGN(ETEXT_ALIGN_SIZE);
 	_etext = .;
 	PROVIDE32 (etext = .);
 
Signature · 1 line
2.13.3
#12 lore raw

PATCH v3 12/16 powerpc/mm/32s: add setibat() clearibat() and update_bats()

In reply to Christophe Leroy

setibat() and clearibat() allows to manipulate IBATs independently of DBATs.

update_bats() allows to update bats after init. This is done with MMU off.

---
 arch/powerpc/include/asm/book3s/32/mmu-hash.h |  2 ++
 arch/powerpc/kernel/head_32.S                 | 35 +++++++++++++++++++++++++++
 arch/powerpc/mm/ppc_mmu_32.c                  | 32 ++++++++++++++++++++++++
 3 files changed, 69 insertions(+)
arch/powerpc/include/asm/book3s/32/mmu-hash.h +2 -0
index 0c261ba2c826..5cb588395fdc 100644
@@ -92,6 +92,8 @@ typedef struct {
 	unsigned long vdso_base;
 } mm_context_t;
 
+void update_bats(void);
+
 /* patch sites */
 extern s32 patch__hash_page_A0, patch__hash_page_A1, patch__hash_page_A2;
 extern s32 patch__hash_page_B, patch__hash_page_C;
arch/powerpc/kernel/head_32.S +35 -0
index c2f564690778..91b302b0797f 100644
@@ -1104,6 +1104,41 @@ BEGIN_MMU_FTR_SECTION
 END_MMU_FTR_SECTION_IFSET(MMU_FTR_USE_HIGH_BATS)
 	blr
 
+_ENTRY(update_bats)
+	lis	r4, 1f@h
+	ori	r4, r4, 1f@l
+	tophys(r4, r4)
+	mfmsr	r6
+	mflr	r7
+	li	r3, MSR_KERNEL & ~(MSR_IR | MSR_DR)
+	rlwinm	r0, r6, 0, ~MSR_RI
+	rlwinm	r0, r0, 0, ~MSR_EE
+	mtmsr	r0
+	mtspr	SPRN_SRR0, r4
+	mtspr	SPRN_SRR1, r3
+	SYNC
+	RFI
+1:	bl	clear_bats
+	lis	r3, BATS@ha
+	addi	r3, r3, BATS@l
+	tophys(r3, r3)
+	LOAD_BAT(0, r3, r4, r5)
+	LOAD_BAT(1, r3, r4, r5)
+	LOAD_BAT(2, r3, r4, r5)
+	LOAD_BAT(3, r3, r4, r5)
+BEGIN_MMU_FTR_SECTION
+	LOAD_BAT(4, r3, r4, r5)
+	LOAD_BAT(5, r3, r4, r5)
+	LOAD_BAT(6, r3, r4, r5)
+	LOAD_BAT(7, r3, r4, r5)
+END_MMU_FTR_SECTION_IFSET(MMU_FTR_USE_HIGH_BATS)
+	li	r3, MSR_KERNEL & ~(MSR_IR | MSR_DR | MSR_RI)
+	mtmsr	r3
+	mtspr	SPRN_SRR0, r7
+	mtspr	SPRN_SRR1, r6
+	SYNC
+	RFI
+
 flush_tlbs:
 	lis	r10, 0x40
 1:	addic.	r10, r10, -0x1000
arch/powerpc/mm/ppc_mmu_32.c +32 -0
index ff8580c6ab11..66f1319e8e20 100644
@@ -106,6 +106,38 @@ static unsigned int block_size(unsigned long base, unsigned long top)
 	return min3(max_size, 1U << base_shift, 1U << block_shift);
 }
 
+/*
+ * Set up one of the IBAT (block address translation) register pairs.
+ * The parameters are not checked; in particular size must be a power
+ * of 2 between 128k and 256M.
+ * Only for 603+ ...
+ */
+static void setibat(int index, unsigned long virt, phys_addr_t phys,
+		    unsigned int size, pgprot_t prot)
+{
+	unsigned int bl = (size >> 17) - 1;
+	int wimgxpp;
+	struct ppc_bat *bat = BATS[index];
+	unsigned long flags = pgprot_val(prot);
+
+	if (!cpu_has_feature(CPU_FTR_NEED_COHERENT))
+		flags &= ~_PAGE_COHERENT;
+
+	wimgxpp = (flags & _PAGE_COHERENT) | (_PAGE_EXEC ? BPP_RX : BPP_XX);
+	bat[0].batu = virt | (bl << 2) | 2; /* Vs=1, Vp=0 */
+	bat[0].batl = BAT_PHYS_ADDR(phys) | wimgxpp;
+	if (flags & _PAGE_USER)
+		bat[0].batu |= 1;	/* Vp = 1 */
+}
+
+static void clearibat(int index)
+{
+	struct ppc_bat *bat = BATS[index];
+
+	bat[0].batu = 0;
+	bat[0].batl = 0;
+}
+
 unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	int idx;
Signature · 1 line
2.13.3
#13 lore raw

PATCH v3 13/16 powerpc/mm/32s: Use BATs for STRICT_KERNEL_RWX

In reply to Christophe Leroy

Today, STRICT_KERNEL_RWX is based on the use of regular pages to map kernel pages.

On Book3s 32, it has three consequences:
- Using pages instead of BAT for mapping kernel linear memory severely impacts performance.
- Exec protection is not effective because no-execute cannot be set at page level (except on 603 which doesn't have hash tables)
- Write protection is not effective because PP bits do not provide RO mode for kernel-only pages (except on 603 which handles it in software via PAGE_DIRTY)

On the 603+, we have:
- Independent IBAT and DBAT allowing limitation of exec parts.
- NX bit can be set in segment registers to forbit execution on memory mapped by pages.
- RO mode on DBATs even for kernel-only blocks.

On the 601, there is nothing much we can do other than warn the user about it, because:
- BATs are common to instructions and data.
- BAT do not provide RO mode for kernel-only blocks.
- segment registers don't have the NX bit.

In order to use IBAT for exec protection, this patch:
- Aligns _etext to BAT block sizes (128kb)
- Set NX bit in kernel segment register (Except on vmalloc area when CONFIG_MODULES is selected)
- Maps kernel text with IBATs.

In order to use DBAT for exec protection, this patch:
- Aligns RW DATA to BAT block sizes (4M)
- Maps kernel RO area with write prohibited DBATs
- Maps remaining memory with remaining DBATs

Here is what we get with this patch on a 832x when activating STRICT_KERNEL_RWX:

Symbols:
c0000000 T _stext
c0680000 R __start_rodata
c0680000 R _etext
c0800000 T __init_begin
c0800000 T _sinittext

~# cat /sys/kernel/debug/block_address_translation ---[ Instruction Block Address Translation ]--- 0: 0xc0000000-0xc03fffff 0x00000000 Kernel EXEC coherent 1: 0xc0400000-0xc05fffff 0x00400000 Kernel EXEC coherent 2: 0xc0600000-0xc067ffff 0x00600000 Kernel EXEC coherent

3:         -
4:         -
5:         -
6:         -
7:         -

---[ Data Block Address Translation ]---
0: 0xc0000000-0xc07fffff 0x00000000 Kernel RO coherent 1: 0xc0800000-0xc0ffffff 0x00800000 Kernel RW coherent 2: 0xc1000000-0xc1ffffff 0x01000000 Kernel RW coherent 3: 0xc2000000-0xc3ffffff 0x02000000 Kernel RW coherent 4: 0xc4000000-0xc7ffffff 0x04000000 Kernel RW coherent 5: 0xc8000000-0xcfffffff 0x08000000 Kernel RW coherent 6: 0xd0000000-0xdfffffff 0x10000000 Kernel RW coherent
7: -

~# cat /sys/kernel/debug/segment_registers
---[ User Segments ]---
0x00000000-0x0fffffff Kern key 1 User key 1 VSID 0xa085d0 0x10000000-0x1fffffff Kern key 1 User key 1 VSID 0xa086e1 0x20000000-0x2fffffff Kern key 1 User key 1 VSID 0xa087f2 0x30000000-0x3fffffff Kern key 1 User key 1 VSID 0xa08903 0x40000000-0x4fffffff Kern key 1 User key 1 VSID 0xa08a14 0x50000000-0x5fffffff Kern key 1 User key 1 VSID 0xa08b25 0x60000000-0x6fffffff Kern key 1 User key 1 VSID 0xa08c36 0x70000000-0x7fffffff Kern key 1 User key 1 VSID 0xa08d47 0x80000000-0x8fffffff Kern key 1 User key 1 VSID 0xa08e58 0x90000000-0x9fffffff Kern key 1 User key 1 VSID 0xa08f69 0xa0000000-0xafffffff Kern key 1 User key 1 VSID 0xa0907a 0xb0000000-0xbfffffff Kern key 1 User key 1 VSID 0xa0918b

---[ Kernel Segments ]---
0xc0000000-0xcfffffff Kern key 0 User key 1 No Exec VSID 0x000ccc 0xd0000000-0xdfffffff Kern key 0 User key 1 No Exec VSID 0x000ddd 0xe0000000-0xefffffff Kern key 0 User key 1 No Exec VSID 0x000eee 0xf0000000-0xffffffff Kern key 0 User key 1 No Exec VSID 0x000fff

Aligning _etext to 128kb allows to map up to 32Mb text with 8 IBATs:
16Mb + 8Mb + 4Mb + 2Mb + 1Mb + 512kb + 256kb + 128kb (+ 128kb) = 32Mb (A 9th IBAT is unneeded as 32Mb would need only a single 32Mb block)

Aligning data to 4M allows to map up to 512Mb data with 8 DBATs:
16Mb + 8Mb + 4Mb + 4Mb + 32Mb + 64Mb + 128Mb + 256Mb = 512Mb

Because some processors only have 4 BATs and because some targets need DBATs for mapping other areas, the following patch will allow to modify _etext and data alignment.

---
 arch/powerpc/Kconfig                         |  2 +
 arch/powerpc/include/asm/book3s/32/pgtable.h | 11 ++++
 arch/powerpc/mm/init_32.c                    |  4 +-
 arch/powerpc/mm/mmu_decl.h                   |  8 +++
 arch/powerpc/mm/pgtable_32.c                 | 10 +++-
 arch/powerpc/mm/ppc_mmu_32.c                 | 87 ++++++++++++++++++++++++++--
 6 files changed, 112 insertions(+), 10 deletions(-)
arch/powerpc/Kconfig +2 -0
index edef40a2b446..640a7cfba9d0 100644
@@ -727,11 +727,13 @@ config THREAD_SHIFT
 
 config ETEXT_SHIFT
 	int
+	default 17 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default PPC_PAGE_SHIFT
 
 config DATA_SHIFT
 	int
 	default 24 if STRICT_KERNEL_RWX && PPC64
+	default 22 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default PPC_PAGE_SHIFT
 
 config FORCE_MAX_ZONEORDER
arch/powerpc/include/asm/book3s/32/pgtable.h +11 -0
index 49d76adb9bc5..aa8406b8f7ba 100644
@@ -174,7 +174,18 @@ static inline bool pte_user(pte_t pte)
  * of RAM.  -- Cort
  */
 #define VMALLOC_OFFSET (0x1000000) /* 16M */
+
+/*
+ * With CONFIG_STRICT_KERNEL_RWX, kernel segments are set NX. But when modules
+ * are used, NX cannot be set on VMALLOC space. So vmalloc VM space and linear
+ * memory shall not share segments.
+ */
+#if defined(CONFIG_STRICT_KERNEL_RWX) && defined(CONFIG_MODULES)
+#define VMALLOC_START ((_ALIGN((long)high_memory, 256L << 20) + VMALLOC_OFFSET) & \
+		       ~(VMALLOC_OFFSET - 1))
+#else
 #define VMALLOC_START ((((long)high_memory + VMALLOC_OFFSET) & ~(VMALLOC_OFFSET-1)))
+#endif
 #define VMALLOC_END	ioremap_bot
 
 #ifndef __ASSEMBLY__
arch/powerpc/mm/init_32.c +1 -3
index ee5a430b9a18..bc28995a37ea 100644
@@ -108,10 +108,8 @@ static void __init MMU_setup(void)
 		__map_without_bats = 1;
 		__map_without_ltlbs = 1;
 	}
-	if (strict_kernel_rwx_enabled()) {
-		__map_without_bats = 1;
+	if (strict_kernel_rwx_enabled())
 		__map_without_ltlbs = 1;
-	}
 }
 
 /*
arch/powerpc/mm/mmu_decl.h +8 -0
index 61730023dde3..98fc94affc29 100644
@@ -165,3 +165,11 @@ unsigned long p_block_mapped(phys_addr_t pa);
 static inline phys_addr_t v_block_mapped(unsigned long va) { return 0; }
 static inline unsigned long p_block_mapped(phys_addr_t pa) { return 0; }
 #endif
+
+#if defined(CONFIG_PPC_BOOK3S_32)
+void mmu_mark_initmem_nx(void);
+void mmu_mark_rodata_ro(void);
+#else
+static inline void mmu_mark_initmem_nx(void) { }
+static inline void mmu_mark_rodata_ro(void) { }
+#endif
arch/powerpc/mm/pgtable_32.c +9 -1
index 94bd7d013557..a3ad09c6e277 100644
@@ -351,7 +351,10 @@ void mark_initmem_nx(void)
 	unsigned long numpages = PFN_UP((unsigned long)_einittext) -
 				 PFN_DOWN((unsigned long)_sinittext);
 
-	change_page_attr(page, numpages, PAGE_KERNEL);
+	if (v_block_mapped((unsigned long)_stext) + 1)
+		mmu_mark_initmem_nx();
+	else
+		change_page_attr(page, numpages, PAGE_KERNEL);
 }
 
 #ifdef CONFIG_STRICT_KERNEL_RWX
@@ -360,6 +363,11 @@ void mark_rodata_ro(void)
 	struct page *page;
 	unsigned long numpages;
 
+	if (v_block_mapped((unsigned long)_sinittext)) {
+		mmu_mark_rodata_ro();
+		return;
+	}
+
 	page = virt_to_page(_stext);
 	numpages = PFN_UP((unsigned long)_etext) -
 		   PFN_DOWN((unsigned long)_stext);
arch/powerpc/mm/ppc_mmu_32.c +81 -6

108 lines not on this page · raw message

Signature · 1 line
2.13.3
#14 lore raw

PATCH v3 14/16 powerpc/kconfig: make _etext and data areas alignment configurable on Book3s 32

In reply to Christophe Leroy

Depending on the number of available BATs for mapping the different kernel areas, it might be needed to increase the alignment of _etext and/or of data areas.

This patchs allows the user to do it via Kconfig.

---
 arch/powerpc/Kconfig | 32 ++++++++++++++++++++++++++++++--
 1 file changed, 30 insertions(+), 2 deletions(-)
arch/powerpc/Kconfig +30 -2
index 640a7cfba9d0..20c4e3a62b90 100644
@@ -725,16 +725,44 @@ config THREAD_SHIFT
 	  Used to define the stack size. The default is almost always what you
 	  want. Only change this if you know what you are doing.
 
+config ETEXT_SHIFT_BOOL
+	bool "Set custom etext alignment" if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	depends on ADVANCED_OPTIONS
+	help
+	  This option allows you to set the kernel end of text alignment. When
+	  RAM is mapped by blocks, the alignment needs to fit the size and
+	  number of possible blocks. The default should be OK for most configs.
+
+	  Say N here unless you know what you are doing.
+
 config ETEXT_SHIFT
-	int
+	int "_etext shift" if ETEXT_SHIFT_BOOL
+	range 17 28 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default 17 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default PPC_PAGE_SHIFT
+	help
+	  On Book3S 32 (603+), IBATs are used to map kernel text.
+	  Smaller is the alignment, greater is the number of necessary IBATs.
+
+config DATA_SHIFT_BOOL
+	bool "Set custom data alignment" if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	depends on ADVANCED_OPTIONS
+	help
+	  This option allows you to set the kernel data alignment. When
+	  RAM is mapped by blocks, the alignment needs to fit the size and
+	  number of possible blocks. The default should be OK for most configs.
+
+	  Say N here unless you know what you are doing.
 
 config DATA_SHIFT
-	int
+	int "Data shift" if DATA_SHIFT_BOOL
 	default 24 if STRICT_KERNEL_RWX && PPC64
+	range 17 28 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default 22 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default PPC_PAGE_SHIFT
+	help
+	  On Book3S 32 (603+), DBATs are used to map kernel text and rodata RO.
+	  Smaller is the alignment, greater is the number of necessary DBATs.
 
 config FORCE_MAX_ZONEORDER
 	int "Maximum zone order"
Signature · 1 line
2.13.3
#15 lore raw

PATCH v3 15/16 powerpc/8xx: don't disable large TLBs with CONFIG_STRICT_KERNEL_RWX

In reply to Christophe Leroy

This patch implements handling of STRICT_KERNEL_RWX with large TLBs directly in the TLB miss handlers.

To do so, etext and sinittext are aligned on 512kB boundaries and the miss handlers use 512kB pages instead of 8Mb pages for addresses close to the boundaries.

It sets RO PP flags for addresses under sinittext.

---
 arch/powerpc/Kconfig                         |  2 ++
 arch/powerpc/include/asm/nohash/32/mmu-8xx.h |  3 +-
 arch/powerpc/kernel/head_8xx.S               | 54 +++++++++++++++++++++-------
 arch/powerpc/mm/8xx_mmu.c                    | 31 +++++++++++++++-
 arch/powerpc/mm/init_32.c                    |  2 +-
 arch/powerpc/mm/mmu_decl.h                   |  2 +-
 6 files changed, 78 insertions(+), 16 deletions(-)
arch/powerpc/Kconfig +2 -0
index 20c4e3a62b90..c4d6c97d7699 100644
@@ -739,6 +739,7 @@ config ETEXT_SHIFT
 	int "_etext shift" if ETEXT_SHIFT_BOOL
 	range 17 28 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default 17 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	default 19 if STRICT_KERNEL_RWX && PPC_8xx
 	default PPC_PAGE_SHIFT
 	help
 	  On Book3S 32 (603+), IBATs are used to map kernel text.
@@ -759,6 +760,7 @@ config DATA_SHIFT
 	default 24 if STRICT_KERNEL_RWX && PPC64
 	range 17 28 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default 22 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	default 19 if STRICT_KERNEL_RWX && PPC_8xx
 	default PPC_PAGE_SHIFT
 	help
 	  On Book3S 32 (603+), DBATs are used to map kernel text and rodata RO.
arch/powerpc/include/asm/nohash/32/mmu-8xx.h +2 -1
index b0f764c827c0..0a1a3fc54e54 100644
@@ -231,9 +231,10 @@ static inline unsigned int mmu_psize_to_shift(unsigned int mmu_psize)
 }
 
 /* patch sites */
-extern s32 patch__itlbmiss_linmem_top;
+extern s32 patch__itlbmiss_linmem_top, patch__itlbmiss_linmem_top8;
 extern s32 patch__dtlbmiss_linmem_top, patch__dtlbmiss_immr_jmp;
 extern s32 patch__fixupdar_linmem_top;
+extern s32 patch__dtlbmiss_romem_top, patch__dtlbmiss_romem_top8;
 
 extern s32 patch__itlbmiss_exit_1, patch__itlbmiss_exit_2;
 extern s32 patch__dtlbmiss_exit_1, patch__dtlbmiss_exit_2, patch__dtlbmiss_exit_3;
arch/powerpc/kernel/head_8xx.S +42 -12

84 lines not on this page · raw message

arch/powerpc/mm/8xx_mmu.c +30 -1
index 46bc26ef71e9..62d4e6d76cd7 100644
@@ -94,11 +94,20 @@ static void __init mmu_mapin_immr(void)
 		map_kernel_page(v + offset, p + offset, PAGE_KERNEL_NCG);
 }
 
-static void __init mmu_patch_cmp_limit(s32 *site, unsigned long mapped)
+static void mmu_patch_cmp_limit(s32 *site, unsigned long mapped)
 {
 	modify_instruction_site(site, 0xffff, (unsigned long)__va(mapped) >> 16);
 }
 
+static void mmu_patch_addis(s32 *site, long simm)
+{
+	unsigned int instr = *(unsigned int *)patch_site_addr(site);
+
+	instr &= 0xffff0000;
+	instr |= ((unsigned long)simm) >> 16;
+	patch_instruction_site(site, instr);
+}
+
 unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 {
 	unsigned long mapped;
@@ -135,6 +144,26 @@ unsigned long __init mmu_mapin_ram(unsigned long base, unsigned long top)
 	return mapped;
 }
 
+void mmu_mark_initmem_nx(void)
+{
+	if (IS_ENABLED(CONFIG_STRICT_KERNEL_RWX) && CONFIG_ETEXT_SHIFT < 23)
+		mmu_patch_addis(&patch__itlbmiss_linmem_top8,
+				-((long)_etext & ~(LARGE_PAGE_SIZE_8M - 1)));
+	if (!IS_ENABLED(CONFIG_PIN_TLB_TEXT))
+		mmu_patch_cmp_limit(&patch__itlbmiss_linmem_top, __pa(_etext));
+}
+
+#ifdef CONFIG_STRICT_KERNEL_RWX
+void mmu_mark_rodata_ro(void)
+{
+	if (CONFIG_DATA_SHIFT < 23)
+		mmu_patch_addis(&patch__dtlbmiss_romem_top8,
+				-__pa(((unsigned long)_sinittext) &
+				      ~(LARGE_PAGE_SIZE_8M - 1)));
+	mmu_patch_addis(&patch__dtlbmiss_romem_top, -__pa(_sinittext));
+}
+#endif
+
 void __init setup_initial_memory_limit(phys_addr_t first_memblock_base,
 				       phys_addr_t first_memblock_size)
 {
arch/powerpc/mm/init_32.c +1 -1
index bc28995a37ea..41a3513cadc9 100644
@@ -108,7 +108,7 @@ static void __init MMU_setup(void)
 		__map_without_bats = 1;
 		__map_without_ltlbs = 1;
 	}
-	if (strict_kernel_rwx_enabled())
+	if (strict_kernel_rwx_enabled() && !IS_ENABLED(CONFIG_PPC_8xx))
 		__map_without_ltlbs = 1;
 }
 
arch/powerpc/mm/mmu_decl.h +1 -1
index 98fc94affc29..74ff61dabcb1 100644
@@ -166,7 +166,7 @@ static inline phys_addr_t v_block_mapped(unsigned long va) { return 0; }
 static inline unsigned long p_block_mapped(phys_addr_t pa) { return 0; }
 #endif
 
-#if defined(CONFIG_PPC_BOOK3S_32)
+#if defined(CONFIG_PPC_BOOK3S_32) || defined(CONFIG_PPC_8xx)
 void mmu_mark_initmem_nx(void);
 void mmu_mark_rodata_ro(void);
 #else
Signature · 1 line
2.13.3
#16 lore raw

PATCH v3 16/16 powerpc/kconfig: make _etext and data areas alignment configurable on 8xx

In reply to Christophe Leroy

On 8xx, large pages (512kb or 8M) are used to map kernel linear memory. Aligning to 8M reduces TLB misses as only 8M pages are used in that case. We make 8M the default for data.

This patchs allows the user to do it via Kconfig.

---
 arch/powerpc/Kconfig           | 18 +++++++++++++++---
 arch/powerpc/kernel/head_8xx.S |  4 ++--
 2 files changed, 17 insertions(+), 5 deletions(-)
arch/powerpc/Kconfig +15 -3
index c4d6c97d7699..cf30a8f522b9 100644
@@ -726,7 +726,8 @@ config THREAD_SHIFT
 	  want. Only change this if you know what you are doing.
 
 config ETEXT_SHIFT_BOOL
-	bool "Set custom etext alignment" if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	bool "Set custom etext alignment" if STRICT_KERNEL_RWX && \
+					     (PPC_BOOK3S_32 || PPC_8xx)
 	depends on ADVANCED_OPTIONS
 	help
 	  This option allows you to set the kernel end of text alignment. When
@@ -738,6 +739,7 @@ config ETEXT_SHIFT_BOOL
 config ETEXT_SHIFT
 	int "_etext shift" if ETEXT_SHIFT_BOOL
 	range 17 28 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	range 19 23 if STRICT_KERNEL_RWX && PPC_8xx
 	default 17 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
 	default 19 if STRICT_KERNEL_RWX && PPC_8xx
 	default PPC_PAGE_SHIFT
@@ -745,8 +747,13 @@ config ETEXT_SHIFT
 	  On Book3S 32 (603+), IBATs are used to map kernel text.
 	  Smaller is the alignment, greater is the number of necessary IBATs.
 
+	  On 8xx, large pages (512kb or 8M) are used to map kernel linear
+	  memory. Aligning to 8M reduces TLB misses as only 8M pages are used
+	  in that case.
+
 config DATA_SHIFT_BOOL
-	bool "Set custom data alignment" if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	bool "Set custom data alignment" if STRICT_KERNEL_RWX && \
+					    (PPC_BOOK3S_32 || PPC_8xx)
 	depends on ADVANCED_OPTIONS
 	help
 	  This option allows you to set the kernel data alignment. When
@@ -759,13 +766,18 @@ config DATA_SHIFT
 	int "Data shift" if DATA_SHIFT_BOOL
 	default 24 if STRICT_KERNEL_RWX && PPC64
 	range 17 28 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
+	range 19 23 if STRICT_KERNEL_RWX && PPC_8xx
 	default 22 if STRICT_KERNEL_RWX && PPC_BOOK3S_32
-	default 19 if STRICT_KERNEL_RWX && PPC_8xx
+	default 23 if STRICT_KERNEL_RWX && PPC_8xx
 	default PPC_PAGE_SHIFT
 	help
 	  On Book3S 32 (603+), DBATs are used to map kernel text and rodata RO.
 	  Smaller is the alignment, greater is the number of necessary DBATs.
 
+	  On 8xx, large pages (512kb or 8M) are used to map kernel linear
+	  memory. Aligning to 8M reduces TLB misses as only 8M pages are used
+	  in that case.
+
 config FORCE_MAX_ZONEORDER
 	int "Maximum zone order"
 	range 8 9 if PPC64 && PPC_64K_PAGES
arch/powerpc/kernel/head_8xx.S +2 -2
index 01ed8f3c95c8..63f1b7eec3f0 100644
@@ -416,7 +416,7 @@ InstructionTLBMiss:
 #ifndef CONFIG_PIN_TLB_TEXT
 ITLBMissLinear:
 	mtcr	r11
-#ifdef CONFIG_STRICT_KERNEL_RWX
+#if defined(CONFIG_STRICT_KERNEL_RWX) && CONFIG_ETEXT_SHIFT < 23
 	patch_site	0f, patch__itlbmiss_linmem_top8
 
 	mfspr	r10, SPRN_SRR0
@@ -537,7 +537,7 @@ DTLBMissIMMR:
 DTLBMissLinear:
 	mtcr	r11
 	rlwinm	r10, r10, 20, 0x0f800000	/* 8xx supports max 256Mb RAM */
-#ifdef CONFIG_STRICT_KERNEL_RWX
+#if defined(CONFIG_STRICT_KERNEL_RWX) && CONFIG_DATA_SHIFT < 23
 	patch_site	0f, patch__dtlbmiss_romem_top8
 
 0:	subis	r11, r10, (PAGE_OFFSET - 0x80000000)@ha
Signature · 1 line
2.13.3
#17 lore raw
In reply to Christophe Leroy
Quoted 3 lines · Christophe Leroy

The purpose of this serie is to:
- use BATs with STRICT_KERNEL_RWX on book3s (See patch 13 for details.)
- use LTLBs with STRICT_KERNEL_RWX on 8xx (See patch 15 for a few details.)

This doesn't boot qemu-mac99 for me:

spawn ~/src/qemu/ppc-softmmu/qemu-system-ppc -nographic -vga none -M mac99 -m 1G -kernel build/vmlinux -initrd ppc32-initrd.gz -append console=ttyPZ0 init=/bin/sh

Quoted 7 lines
Quoted 7 lines

=============================================================
OpenBIOS 1.1 [Feb 15 2019 10:05] Configuration device id QEMU version 1 machine id 1
CPUs: 1
Memory: 1024M
UUID: 00000000-0000-0000-0000-000000000000 CPU type PowerPC,G4

milliseconds isn't unique.
Welcome to OpenBIOS v1.1 built on Feb 15 2019 10:05

Quoted 3 lines
Quoted 3 lines

[ppc] Kernel already loaded (0x01000000 + 0x00c2c338) (initrd 0x01d2d000 + 0x007e72f0) [ppc] Kernel command line: console=ttyPZ0 init=/bin/sh
switching to new context:

OF stdout device is: /pci@f2000000/mac-io@c/escc@13000/ch-a@13020
Preparing to boot Linux version 5.0.0-rc2-gcc-8.2.0-00125-g4fcb83ca7936 (michael@ka4) (gcc version 8.2.0 (Buildroot 2018.11-rc2-00003-ga0787e9)) #724 Thu Feb 21 12:03:14 AEDT 2019 Detected machine type: 00000400
command line:
memory layout at init:

    memory_limit : 00000000 (16 MB aligned)
    alloc_bottom : 02515000
    alloc_top    : 30000000
    alloc_top_hi : 40000000
    rmo_top      : 30000000
    ram_top      : 40000000

copying OF device tree...
Building dt strings...
Building dt structure...
Device tree strings 0x02516000 -> 0x025150a4
Device tree struct 0x02517000 -> 0x3fde7eb0
Quiescing Open Firmware ...
Booting Linux via __start() @ 0x01000000 ...
FAIL! Booting BE pmac32

That's pmac32 defconfig ish.
I haven't had time to debug it further sorry.

cheers

#18 lore raw

Le 21/02/2019 à 02:47, Michael Ellerman a écrit :

Quoted 42 lines
Quoted 3 lines · Christophe Leroy

The purpose of this serie is to:
- use BATs with STRICT_KERNEL_RWX on book3s (See patch 13 for details.)
- use LTLBs with STRICT_KERNEL_RWX on 8xx (See patch 15 for a few details.)

This doesn't boot qemu-mac99 for me:

   spawn ~/src/qemu/ppc-softmmu/qemu-system-ppc -nographic -vga none -M mac99 -m 1G -kernel build/vmlinux -initrd ppc32-initrd.gz -append console=ttyPZ0 init=/bin/sh
   >> =============================================================
   >> OpenBIOS 1.1 [Feb 15 2019 10:05]
   >> Configuration device id QEMU version 1 machine id 1
   >> CPUs: 1
   >> Memory: 1024M
   >> UUID: 00000000-0000-0000-0000-000000000000
   >> CPU type PowerPC,G4
   milliseconds isn't unique.
   Welcome to OpenBIOS v1.1 built on Feb 15 2019 10:05
   >> [ppc] Kernel already loaded (0x01000000 + 0x00c2c338) (initrd 0x01d2d000 + 0x007e72f0)
   >> [ppc] Kernel command line: console=ttyPZ0 init=/bin/sh
   >> switching to new context:
   OF stdout device is: /pci@f2000000/mac-io@c/escc@13000/ch-a@13020
   Preparing to boot Linux version 5.0.0-rc2-gcc-8.2.0-00125-g4fcb83ca7936 (michael@ka4) (gcc version 8.2.0 (Buildroot 2018.11-rc2-00003-ga0787e9)) #724 Thu Feb 21 12:03:14 AEDT 2019
   Detected machine type: 00000400
   command line:
   memory layout at init:
     memory_limit : 00000000 (16 MB aligned)
     alloc_bottom : 02515000
     alloc_top    : 30000000
     alloc_top_hi : 40000000
     rmo_top      : 30000000
     ram_top      : 40000000
   copying OF device tree...
   Building dt strings...
   Building dt structure...
   Device tree strings 0x02516000 -> 0x025150a4
   Device tree struct  0x02517000 -> 0x3fde7eb0
   Quiescing Open Firmware ...
   Booting Linux via __start() @ 0x01000000 ...
   FAIL! Booting BE pmac32
Quoted 3 lines

That's pmac32 defconfig ish.
I haven't had time to debug it further sorry.

Ok. I boots fine without the '-m 1G'.

I'll find out why.

Christophe

#19 lore raw

Le 21/02/2019 à 07:31, Christophe Leroy a écrit :

Quoted 48 lines

Le 21/02/2019 à 02:47, Michael Ellerman a écrit :

Quoted 45 lines
Quoted 4 lines · Christophe Leroy

The purpose of this serie is to:
- use BATs with STRICT_KERNEL_RWX on book3s (See patch 13 for details.)
- use LTLBs with STRICT_KERNEL_RWX on 8xx (See patch 15 for a few details.)

This doesn't boot qemu-mac99 for me:

   spawn ~/src/qemu/ppc-softmmu/qemu-system-ppc -nographic -vga none
-M mac99 -m 1G -kernel build/vmlinux -initrd ppc32-initrd.gz -append
console=ttyPZ0 init=/bin/sh
   >> =============================================================
   >> OpenBIOS 1.1 [Feb 15 2019 10:05]
   >> Configuration device id QEMU version 1 machine id 1
   >> CPUs: 1
   >> Memory: 1024M
   >> UUID: 00000000-0000-0000-0000-000000000000
   >> CPU type PowerPC,G4
   milliseconds isn't unique.
   Welcome to OpenBIOS v1.1 built on Feb 15 2019 10:05
   >> [ppc] Kernel already loaded (0x01000000 + 0x00c2c338) (initrd 0x01d2d000 + 0x007e72f0)
   >> [ppc] Kernel command line: console=ttyPZ0 init=/bin/sh
   >> switching to new context:
   OF stdout device is: /pci@f2000000/mac-io@c/escc@13000/ch-a@13020
   Preparing to boot Linux version 5.0.0-rc2-gcc-8.2.0-00125-g4fcb83ca7936 (michael@ka4) (gcc version 8.2.0 (Buildroot 2018.11-rc2-00003-ga0787e9)) #724 Thu Feb 21 12:03:14 AEDT 2019 Detected machine type: 00000400
   command line:
   memory layout at init:
     memory_limit : 00000000 (16 MB aligned)
     alloc_bottom : 02515000
     alloc_top    : 30000000
     alloc_top_hi : 40000000
     rmo_top      : 30000000
     ram_top      : 40000000
   copying OF device tree...
   Building dt strings...
   Building dt structure...
   Device tree strings 0x02516000 -> 0x025150a4
   Device tree struct  0x02517000 -> 0x3fde7eb0
   Quiescing Open Firmware ...
   Booting Linux via __start() @ 0x01000000 ...
   FAIL! Booting BE pmac32

Quoted 8 lines
Quoted 3 lines

That's pmac32 defconfig ish.
I haven't had time to debug it further sorry.

Ok. I boots fine without the '-m 1G'.

I'll find out why.

Doesn't boot because it maps memory above total_lowmem.

I'll fix it.

Christophe