[PATCH V2 0/8] THP fixes for ppc64

[PATCH V2 0/8] THP fixes for ppc64

[Aneesh Kumar K.V]

Hi,

This patch series fixes machine check exception that we observed when using
transparent huge page along with 4k hash pte on power bare metal platform.
Patch "powerpc: mm: Use read barrier when creating real_pte" is not really
related to thp, but was added in the series  because it is fixing a related
issue with 4k pte. That patch was not really needed to fix the MCE.

Changes from V1:
* Rebased to latest upstream

Aneesh Kumar K.V (8):
  powerpc: thp: Add write barrier after updating the valid bit
  powerpc: thp: don't recompute vsid and ssize in loop on invalidate
  powerpc: thp: invalidate old 64K based hash page mapping before insert
    of 4k pte
  powerpc: thp: Handle combo pages in invalidate
  powerpc: thp: inalidate with vpn in loop
  powerpc: thp: use ACCESS_ONCE when loading pmdp
  powerpc: mm: Use read barrier when creating real_pte
  powerpc: thp: Add tracepoints to track hugepage invalidate

 arch/powerpc/include/asm/machdep.h        |  6 +--
 arch/powerpc/include/asm/pgtable-ppc64.h  |  2 +-
 arch/powerpc/include/asm/pte-hash64-64k.h | 30 +++++++++--
 arch/powerpc/mm/hash_native_64.c          | 40 ++++----------
 arch/powerpc/mm/hugepage-hash64.c         | 88 +++++++++++++++++++++++++++----
 arch/powerpc/mm/pgtable_64.c              | 44 ++++++++++------
 arch/powerpc/mm/tlb_hash64.c              |  6 ++-
 arch/powerpc/platforms/pseries/lpar.c     | 20 +++----
 include/trace/events/thp.h                | 88 +++++++++++++++++++++++++++++++
 9 files changed, 245 insertions(+), 79 deletions(-)
 create mode 100644 include/trace/events/thp.h

-- 
1.9.1

[PATCH V2 1/8] powerpc: thp: Add write barrier after updating the valid bit

[Aneesh Kumar K.V]

With hugepages, we store the hpte valid information in the pte page
whose address is stored in the second half of the PMD. Use a
write barrier to make sure clearing pmd busy bit and updating
hpte valid info are ordered properly.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/mm/hugepage-hash64.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/mm/hugepage-hash64.c b/arch/powerpc/mm/hugepage-hash64.c
index 826893fcb3a7..11f9a37ca2c6 100644
--- a/arch/powerpc/mm/hugepage-hash64.c
+++ b/arch/powerpc/mm/hugepage-hash64.c
@@ -172,8 +172,11 @@ repeat:
 		mark_hpte_slot_valid(hpte_slot_array, index, slot);
 	}
 	/*
-	 * No need to use ldarx/stdcx here
+	 * The hpte valid is stored in the pgtable whose address is in the
+	 * second half of the PMD. Order this against clearing of the busy bit in
+	 * huge pmd.
 	 */
+	smp_wmb();
 	*pmdp = __pmd(new_pmd & ~_PAGE_BUSY);
 	return 0;
 }
-- 
1.9.1

[PATCH V2 2/8] powerpc: thp: don't recompute vsid and ssize in loop on invalidate

[Aneesh Kumar K.V]

The segment identifier and segment size will remain the same in
the loop, So we can compute it outside. We also change the
hugepage_invalidate interface so that we can use it the later patch

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/machdep.h    |  6 +++---
 arch/powerpc/mm/hash_native_64.c      | 19 +++++--------------
 arch/powerpc/mm/pgtable_64.c          | 24 ++++++++++++------------
 arch/powerpc/platforms/pseries/lpar.c | 20 ++++++--------------
 4 files changed, 26 insertions(+), 43 deletions(-)

diff --git a/arch/powerpc/include/asm/machdep.h b/arch/powerpc/include/asm/machdep.h
index 44e90516519b..b125ceab149c 100644
--- a/arch/powerpc/include/asm/machdep.h
+++ b/arch/powerpc/include/asm/machdep.h
@@ -57,10 +57,10 @@ struct machdep_calls {
 	void            (*hpte_removebolted)(unsigned long ea,
 					     int psize, int ssize);
 	void		(*flush_hash_range)(unsigned long number, int local);
-	void		(*hugepage_invalidate)(struct mm_struct *mm,
+	void		(*hugepage_invalidate)(unsigned long vsid,
+					       unsigned long addr,
 					       unsigned char *hpte_slot_array,
-					       unsigned long addr, int psize);
-
+					       int psize, int ssize);
 	/* special for kexec, to be called in real mode, linear mapping is
 	 * destroyed as well */
 	void		(*hpte_clear_all)(void);
diff --git a/arch/powerpc/mm/hash_native_64.c b/arch/powerpc/mm/hash_native_64.c
index cf1d325eae8b..fb89d7695a9a 100644
--- a/arch/powerpc/mm/hash_native_64.c
+++ b/arch/powerpc/mm/hash_native_64.c
@@ -412,18 +412,18 @@ static void native_hpte_invalidate(unsigned long slot, unsigned long vpn,
 	local_irq_restore(flags);
 }
 
-static void native_hugepage_invalidate(struct mm_struct *mm,
+static void native_hugepage_invalidate(unsigned long vsid,
+				       unsigned long addr,
 				       unsigned char *hpte_slot_array,
-				       unsigned long addr, int psize)
+				       int psize, int ssize)
 {
-	int ssize = 0, i;
-	int lock_tlbie;
+	int i, lock_tlbie;
 	struct hash_pte *hptep;
 	int actual_psize = MMU_PAGE_16M;
 	unsigned int max_hpte_count, valid;
 	unsigned long flags, s_addr = addr;
 	unsigned long hpte_v, want_v, shift;
-	unsigned long hidx, vpn = 0, vsid, hash, slot;
+	unsigned long hidx, vpn = 0, hash, slot;
 
 	shift = mmu_psize_defs[psize].shift;
 	max_hpte_count = 1U << (PMD_SHIFT - shift);
@@ -437,15 +437,6 @@ static void native_hugepage_invalidate(struct mm_struct *mm,
 
 		/* get the vpn */
 		addr = s_addr + (i * (1ul << shift));
-		if (!is_kernel_addr(addr)) {
-			ssize = user_segment_size(addr);
-			vsid = get_vsid(mm->context.id, addr, ssize);
-			WARN_ON(vsid == 0);
-		} else {
-			vsid = get_kernel_vsid(addr, mmu_kernel_ssize);
-			ssize = mmu_kernel_ssize;
-		}
-
 		vpn = hpt_vpn(addr, vsid, ssize);
 		hash = hpt_hash(vpn, shift, ssize);
 		if (hidx & _PTEIDX_SECONDARY)
diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c
index 3b3c4d34c7a0..5039f3b04d6e 100644
--- a/arch/powerpc/mm/pgtable_64.c
+++ b/arch/powerpc/mm/pgtable_64.c
@@ -745,12 +745,21 @@ void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr,
 	if (!hpte_slot_array)
 		return;
 
-	/* get the base page size */
+	/* get the base page size,vsid and segment size */
 	psize = get_slice_psize(mm, s_addr);
+	if (!is_kernel_addr(s_addr)) {
+		ssize = user_segment_size(s_addr);
+		vsid = get_vsid(mm->context.id, s_addr, ssize);
+		WARN_ON(vsid == 0);
+	} else {
+		vsid = get_kernel_vsid(s_addr, mmu_kernel_ssize);
+		ssize = mmu_kernel_ssize;
+	}
 
 	if (ppc_md.hugepage_invalidate)
-		return ppc_md.hugepage_invalidate(mm, hpte_slot_array,
-						  s_addr, psize);
+		return ppc_md.hugepage_invalidate(vsid, s_addr,
+						  hpte_slot_array,
+						  psize, ssize);
 	/*
 	 * No bluk hpte removal support, invalidate each entry
 	 */
@@ -768,15 +777,6 @@ void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr,
 
 		/* get the vpn */
 		addr = s_addr + (i * (1ul << shift));
-		if (!is_kernel_addr(addr)) {
-			ssize = user_segment_size(addr);
-			vsid = get_vsid(mm->context.id, addr, ssize);
-			WARN_ON(vsid == 0);
-		} else {
-			vsid = get_kernel_vsid(addr, mmu_kernel_ssize);
-			ssize = mmu_kernel_ssize;
-		}
-
 		vpn = hpt_vpn(addr, vsid, ssize);
 		hash = hpt_hash(vpn, shift, ssize);
 		if (hidx & _PTEIDX_SECONDARY)
diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c
index fbfcef514aa7..34e64237fff9 100644
--- a/arch/powerpc/platforms/pseries/lpar.c
+++ b/arch/powerpc/platforms/pseries/lpar.c
@@ -431,16 +431,17 @@ static void __pSeries_lpar_hugepage_invalidate(unsigned long *slot,
 		spin_unlock_irqrestore(&pSeries_lpar_tlbie_lock, flags);
 }
 
-static void pSeries_lpar_hugepage_invalidate(struct mm_struct *mm,
-				       unsigned char *hpte_slot_array,
-				       unsigned long addr, int psize)
+static void pSeries_lpar_hugepage_invalidate(unsigned long vsid,
+					     unsigned long addr,
+					     unsigned char *hpte_slot_array,
+					     int psize, int ssize)
 {
-	int ssize = 0, i, index = 0;
+	int i, index = 0;
 	unsigned long s_addr = addr;
 	unsigned int max_hpte_count, valid;
 	unsigned long vpn_array[PPC64_HUGE_HPTE_BATCH];
 	unsigned long slot_array[PPC64_HUGE_HPTE_BATCH];
-	unsigned long shift, hidx, vpn = 0, vsid, hash, slot;
+	unsigned long shift, hidx, vpn = 0, hash, slot;
 
 	shift = mmu_psize_defs[psize].shift;
 	max_hpte_count = 1U << (PMD_SHIFT - shift);
@@ -453,15 +454,6 @@ static void pSeries_lpar_hugepage_invalidate(struct mm_struct *mm,
 
 		/* get the vpn */
 		addr = s_addr + (i * (1ul << shift));
-		if (!is_kernel_addr(addr)) {
-			ssize = user_segment_size(addr);
-			vsid = get_vsid(mm->context.id, addr, ssize);
-			WARN_ON(vsid == 0);
-		} else {
-			vsid = get_kernel_vsid(addr, mmu_kernel_ssize);
-			ssize = mmu_kernel_ssize;
-		}
-
 		vpn = hpt_vpn(addr, vsid, ssize);
 		hash = hpt_hash(vpn, shift, ssize);
 		if (hidx & _PTEIDX_SECONDARY)
-- 
1.9.1

[PATCH V2 3/8] powerpc: thp: invalidate old 64K based hash page mapping before insert of 4k pte

[Aneesh Kumar K.V]

If we changed base page size of the segment, either via sub_page_protect
or via remap_4k_pfn, we do a demote_segment which doesn't flush the hash
table entries. We do a lazy hash page table flush for all mapped pages
in the demoted segment. This happens when we handle hash page fault
for these pages.

We use _PAGE_COMBO bit along with _PAGE_HASHPTE to indicate whether a
pte is backed by 4K hash pte. If we find _PAGE_COMBO not set on the pte,
that implies that we could possibly have older 64K hash pte entries in
the hash page table and we need to invalidate those entries.

Handle this correctly for 16M pages

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/mm/hugepage-hash64.c | 79 ++++++++++++++++++++++++++++++++++-----
 1 file changed, 70 insertions(+), 9 deletions(-)

diff --git a/arch/powerpc/mm/hugepage-hash64.c b/arch/powerpc/mm/hugepage-hash64.c
index 11f9a37ca2c6..1fb609dcc49b 100644
--- a/arch/powerpc/mm/hugepage-hash64.c
+++ b/arch/powerpc/mm/hugepage-hash64.c
@@ -18,6 +18,57 @@
 #include <linux/mm.h>
 #include <asm/machdep.h>
 
+static void invalidate_old_hpte(unsigned long vsid, unsigned long addr,
+				pmd_t *pmdp, unsigned int psize, int ssize)
+{
+	int i, max_hpte_count, valid;
+	unsigned long s_addr;
+	unsigned char *hpte_slot_array;
+	unsigned long hidx, shift, vpn, hash, slot;
+
+	s_addr = addr & HPAGE_PMD_MASK;
+	hpte_slot_array = get_hpte_slot_array(pmdp);
+	/*
+	 * IF we try to do a HUGE PTE update after a withdraw is done.
+	 * we will find the below NULL. This happens when we do
+	 * split_huge_page_pmd
+	 */
+	if (!hpte_slot_array)
+		return;
+
+	if (ppc_md.hugepage_invalidate)
+		return ppc_md.hugepage_invalidate(vsid, s_addr, hpte_slot_array,
+						  psize, ssize);
+	/*
+	 * No bluk hpte removal support, invalidate each entry
+	 */
+	shift = mmu_psize_defs[psize].shift;
+	max_hpte_count = HPAGE_PMD_SIZE >> shift;
+	for (i = 0; i < max_hpte_count; i++) {
+		/*
+		 * 8 bits per each hpte entries
+		 * 000| [ secondary group (one bit) | hidx (3 bits) | valid bit]
+		 */
+		valid = hpte_valid(hpte_slot_array, i);
+		if (!valid)
+			continue;
+		hidx =  hpte_hash_index(hpte_slot_array, i);
+
+		/* get the vpn */
+		addr = s_addr + (i * (1ul << shift));
+		vpn = hpt_vpn(addr, vsid, ssize);
+		hash = hpt_hash(vpn, shift, ssize);
+		if (hidx & _PTEIDX_SECONDARY)
+			hash = ~hash;
+
+		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
+		slot += hidx & _PTEIDX_GROUP_IX;
+		ppc_md.hpte_invalidate(slot, vpn, psize,
+				       MMU_PAGE_16M, ssize, 0);
+	}
+}
+
+
 int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
 		    pmd_t *pmdp, unsigned long trap, int local, int ssize,
 		    unsigned int psize)
@@ -85,6 +136,15 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
 	vpn = hpt_vpn(ea, vsid, ssize);
 	hash = hpt_hash(vpn, shift, ssize);
 	hpte_slot_array = get_hpte_slot_array(pmdp);
+	if (psize == MMU_PAGE_4K) {
+		/*
+		 * invalidate the old hpte entry if we have that mapped via 64K
+		 * base page size. This is because demote_segment won't flush
+		 * hash page table entries.
+		 */
+		if ((old_pmd & _PAGE_HASHPTE) && !(old_pmd & _PAGE_COMBO))
+			invalidate_old_hpte(vsid, ea, pmdp, MMU_PAGE_64K, ssize);
+	}
 
 	valid = hpte_valid(hpte_slot_array, index);
 	if (valid) {
@@ -107,11 +167,8 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
 			 * safely update this here.
 			 */
 			valid = 0;
-			new_pmd &= ~_PAGE_HPTEFLAGS;
 			hpte_slot_array[index] = 0;
-		} else
-			/* clear the busy bits and set the hash pte bits */
-			new_pmd = (new_pmd & ~_PAGE_HPTEFLAGS) | _PAGE_HASHPTE;
+		}
 	}
 
 	if (!valid) {
@@ -119,11 +176,7 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
 
 		/* insert new entry */
 		pa = pmd_pfn(__pmd(old_pmd)) << PAGE_SHIFT;
-repeat:
-		hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;
-
-		/* clear the busy bits and set the hash pte bits */
-		new_pmd = (new_pmd & ~_PAGE_HPTEFLAGS) | _PAGE_HASHPTE;
+		new_pmd |= _PAGE_HASHPTE;
 
 		/* Add in WIMG bits */
 		rflags |= (new_pmd & (_PAGE_WRITETHRU | _PAGE_NO_CACHE |
@@ -132,6 +185,8 @@ repeat:
 		 * enable the memory coherence always
 		 */
 		rflags |= HPTE_R_M;
+repeat:
+		hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;
 
 		/* Insert into the hash table, primary slot */
 		slot = ppc_md.hpte_insert(hpte_group, vpn, pa, rflags, 0,
@@ -172,6 +227,12 @@ repeat:
 		mark_hpte_slot_valid(hpte_slot_array, index, slot);
 	}
 	/*
+	 * Mark the pte with _PAGE_COMBO, if we are trying to hash it with
+	 * base page size 4k.
+	 */
+	if (psize == MMU_PAGE_4K)
+		new_pmd |= _PAGE_COMBO;
+	/*
 	 * The hpte valid is stored in the pgtable whose address is in the
 	 * second half of the PMD. Order this against clearing of the busy bit in
 	 * huge pmd.
-- 
1.9.1

[PATCH V2 4/8] powerpc: thp: Handle combo pages in invalidate

[Aneesh Kumar K.V]

If we changed base page size of the segment, either via sub_page_protect
or via remap_4k_pfn, we do a demote_segment which doesn't flush the hash
table entries. We do a lazy hash page table flush for all mapped pages
in the demoted segment. This happens when we handle hash page fault for
these pages.

We use _PAGE_COMBO bit along with _PAGE_HASHPTE to indicate whether a
pte is backed by 4K hash pte. If we find _PAGE_COMBO not set on the pte,
that implies that we could possibly have older 64K hash pte entries in
the hash page table and we need to invalidate those entries.

Use _PAGE_COMBO to determine the page size with which we should
invalidate the hash table entries on unmap.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/pgtable-ppc64.h |  2 +-
 arch/powerpc/mm/pgtable_64.c             | 14 +++++++++++---
 arch/powerpc/mm/tlb_hash64.c             |  2 +-
 3 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/arch/powerpc/include/asm/pgtable-ppc64.h b/arch/powerpc/include/asm/pgtable-ppc64.h
index eb9261024f51..7b3d54fae46f 100644
--- a/arch/powerpc/include/asm/pgtable-ppc64.h
+++ b/arch/powerpc/include/asm/pgtable-ppc64.h
@@ -413,7 +413,7 @@ static inline char *get_hpte_slot_array(pmd_t *pmdp)
 }
 
 extern void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr,
-				   pmd_t *pmdp);
+				   pmd_t *pmdp, unsigned long old_pmd);
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 extern pmd_t pfn_pmd(unsigned long pfn, pgprot_t pgprot);
 extern pmd_t mk_pmd(struct page *page, pgprot_t pgprot);
diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c
index 5039f3b04d6e..948a81e02ddb 100644
--- a/arch/powerpc/mm/pgtable_64.c
+++ b/arch/powerpc/mm/pgtable_64.c
@@ -538,7 +538,7 @@ unsigned long pmd_hugepage_update(struct mm_struct *mm, unsigned long addr,
 	*pmdp = __pmd((old & ~clr) | set);
 #endif
 	if (old & _PAGE_HASHPTE)
-		hpte_do_hugepage_flush(mm, addr, pmdp);
+		hpte_do_hugepage_flush(mm, addr, pmdp, old);
 	return old;
 }
 
@@ -645,7 +645,7 @@ void pmdp_splitting_flush(struct vm_area_struct *vma,
 	if (!(old & _PAGE_SPLITTING)) {
 		/* We need to flush the hpte */
 		if (old & _PAGE_HASHPTE)
-			hpte_do_hugepage_flush(vma->vm_mm, address, pmdp);
+			hpte_do_hugepage_flush(vma->vm_mm, address, pmdp, old);
 	}
 	/*
 	 * This ensures that generic code that rely on IRQ disabling
@@ -723,7 +723,7 @@ void pmdp_invalidate(struct vm_area_struct *vma, unsigned long address,
  * neesd to be flushed.
  */
 void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr,
-			    pmd_t *pmdp)
+			    pmd_t *pmdp, unsigned long old_pmd)
 {
 	int ssize, i;
 	unsigned long s_addr;
@@ -746,7 +746,15 @@ void hpte_do_hugepage_flush(struct mm_struct *mm, unsigned long addr,
 		return;
 
 	/* get the base page size,vsid and segment size */
+#ifdef CONFIG_DEBUG_VM
 	psize = get_slice_psize(mm, s_addr);
+	BUG_ON(psize == MMU_PAGE_16M);
+#endif
+	if (old_pmd & _PAGE_COMBO)
+		psize = MMU_PAGE_4K;
+	else
+		psize = MMU_PAGE_64K;
+
 	if (!is_kernel_addr(s_addr)) {
 		ssize = user_segment_size(s_addr);
 		vsid = get_vsid(mm->context.id, s_addr, ssize);
diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c
index c99f6510a0b2..9adda5790463 100644
--- a/arch/powerpc/mm/tlb_hash64.c
+++ b/arch/powerpc/mm/tlb_hash64.c
@@ -216,7 +216,7 @@ void __flush_hash_table_range(struct mm_struct *mm, unsigned long start,
 		if (!(pte & _PAGE_HASHPTE))
 			continue;
 		if (unlikely(hugepage_shift && pmd_trans_huge(*(pmd_t *)pte)))
-			hpte_do_hugepage_flush(mm, start, (pmd_t *)pte);
+			hpte_do_hugepage_flush(mm, start, (pmd_t *)ptep, pte);
 		else
 			hpte_need_flush(mm, start, ptep, pte, 0);
 	}
-- 
1.9.1

[PATCH V2 5/8] powerpc: thp: inalidate with vpn in loop

[Aneesh Kumar K.V]

As per ISA, for 4k base page size we compare 14..65 bits of VA specified
with the entry_VA in tlb. That implies we need to make sure we do a
tlbie with all the possible 4k va we used to access the 16MB hugepage.
With 64k base page size we compare 14..57 bits of VA. Hence we cannot
ignore the lower 24 bits of va while tlbie .We also cannot tlb
invalidate a 16MB entry with just one tlbie instruction because
we don't track which va was used to instantiate the tlb entry.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/mm/hash_native_64.c | 23 +++++++----------------
 1 file changed, 7 insertions(+), 16 deletions(-)

diff --git a/arch/powerpc/mm/hash_native_64.c b/arch/powerpc/mm/hash_native_64.c
index fb89d7695a9a..afc0a8295f84 100644
--- a/arch/powerpc/mm/hash_native_64.c
+++ b/arch/powerpc/mm/hash_native_64.c
@@ -417,7 +417,7 @@ static void native_hugepage_invalidate(unsigned long vsid,
 				       unsigned char *hpte_slot_array,
 				       int psize, int ssize)
 {
-	int i, lock_tlbie;
+	int i;
 	struct hash_pte *hptep;
 	int actual_psize = MMU_PAGE_16M;
 	unsigned int max_hpte_count, valid;
@@ -456,22 +456,13 @@ static void native_hugepage_invalidate(unsigned long vsid,
 		else
 			/* Invalidate the hpte. NOTE: this also unlocks it */
 			hptep->v = 0;
+		/*
+		 * We need to do tlb invalidate for all the address, tlbie
+		 * instruction compares entry_VA in tlb with the VA specified
+		 * here
+		 */
+		tlbie(vpn, psize, actual_psize, ssize, 0);
 	}
-	/*
-	 * Since this is a hugepage, we just need a single tlbie.
-	 * use the last vpn.
-	 */
-	lock_tlbie = !mmu_has_feature(MMU_FTR_LOCKLESS_TLBIE);
-	if (lock_tlbie)
-		raw_spin_lock(&native_tlbie_lock);
-
-	asm volatile("ptesync":::"memory");
-	__tlbie(vpn, psize, actual_psize, ssize);
-	asm volatile("eieio; tlbsync; ptesync":::"memory");
-
-	if (lock_tlbie)
-		raw_spin_unlock(&native_tlbie_lock);
-
 	local_irq_restore(flags);
 }
 
-- 
1.9.1

[PATCH V2 6/8] powerpc: thp: use ACCESS_ONCE when loading pmdp

[Aneesh Kumar K.V]

We would get wrong results in compiler recomputed old_pmd. Avoid
that by using ACCESS_ONCE

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/mm/hugepage-hash64.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/powerpc/mm/hugepage-hash64.c b/arch/powerpc/mm/hugepage-hash64.c
index 1fb609dcc49b..5f5e6328c21c 100644
--- a/arch/powerpc/mm/hugepage-hash64.c
+++ b/arch/powerpc/mm/hugepage-hash64.c
@@ -84,7 +84,9 @@ int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
 	 * atomically mark the linux large page PMD busy and dirty
 	 */
 	do {
-		old_pmd = pmd_val(*pmdp);
+		pmd_t pmd = ACCESS_ONCE(*pmdp);
+
+		old_pmd = pmd_val(pmd);
 		/* If PMD busy, retry the access */
 		if (unlikely(old_pmd & _PAGE_BUSY))
 			return 0;
-- 
1.9.1

[PATCH V2 7/8] powerpc: mm: Use read barrier when creating real_pte

[Aneesh Kumar K.V]

On ppc64 we support 4K hash pte with 64K page size. That requires
us to track the hash pte slot information on a per 4k basis. We do that
by storing the slot details in the second half of pte page. The pte bit
_PAGE_COMBO is used to indicate whether the second half need to be
looked while building real_pte. We need to use read memory barrier while
doing that so that load of hidx is not reordered w.r.t _PAGE_COMBO
check. On the store side we already do a lwsync in __hash_page_4K

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/pte-hash64-64k.h | 30 +++++++++++++++++++++++++-----
 1 file changed, 25 insertions(+), 5 deletions(-)

diff --git a/arch/powerpc/include/asm/pte-hash64-64k.h b/arch/powerpc/include/asm/pte-hash64-64k.h
index b6d2d42f84b5..4f4ec2ab45c9 100644
--- a/arch/powerpc/include/asm/pte-hash64-64k.h
+++ b/arch/powerpc/include/asm/pte-hash64-64k.h
@@ -46,11 +46,31 @@
  * in order to deal with 64K made of 4K HW pages. Thus we override the
  * generic accessors and iterators here
  */
-#define __real_pte(e,p) 	((real_pte_t) { \
-			(e), (pte_val(e) & _PAGE_COMBO) ? \
-				(pte_val(*((p) + PTRS_PER_PTE))) : 0 })
-#define __rpte_to_hidx(r,index)	((pte_val((r).pte) & _PAGE_COMBO) ? \
-        (((r).hidx >> ((index)<<2)) & 0xf) : ((pte_val((r).pte) >> 12) & 0xf))
+#define __real_pte __real_pte
+static inline real_pte_t __real_pte(pte_t pte, pte_t *ptep)
+{
+	real_pte_t rpte;
+
+	rpte.pte = pte;
+	rpte.hidx = 0;
+	if (pte_val(pte) & _PAGE_COMBO) {
+		/*
+		 * Make sure we order the hidx load against the _PAGE_COMBO
+		 * check. The store side ordering is done in __hash_page_4K
+		 */
+		smp_rmb();
+		rpte.hidx = pte_val(*((ptep) + PTRS_PER_PTE));
+	}
+	return rpte;
+}
+
+static inline unsigned long __rpte_to_hidx(real_pte_t rpte, unsigned long index)
+{
+	if ((pte_val(rpte.pte) & _PAGE_COMBO))
+		return (rpte.hidx >> (index<<2)) & 0xf;
+	return (pte_val(rpte.pte) >> 12) & 0xf;
+}
+
 #define __rpte_to_pte(r)	((r).pte)
 #define __rpte_sub_valid(rpte, index) \
 	(pte_val(rpte.pte) & (_PAGE_HPTE_SUB0 >> (index)))
-- 
1.9.1

[PATCH V2 8/8] powerpc: thp: Add tracepoints to track hugepage invalidate

[Aneesh Kumar K.V]

Add tracepoint to track hugepage invalidate. This help us
in debugging difficult to track bugs.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/mm/pgtable_64.c |  6 +++
 arch/powerpc/mm/tlb_hash64.c |  4 ++
 include/trace/events/thp.h   | 88 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 98 insertions(+)
 create mode 100644 include/trace/events/thp.h

diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c
index 948a81e02ddb..c8d709ab489d 100644
--- a/arch/powerpc/mm/pgtable_64.c
+++ b/arch/powerpc/mm/pgtable_64.c
@@ -54,6 +54,9 @@
 
 #include "mmu_decl.h"
 
+#define CREATE_TRACE_POINTS
+#include <trace/events/thp.h>
+
 /* Some sanity checking */
 #if TASK_SIZE_USER64 > PGTABLE_RANGE
 #error TASK_SIZE_USER64 exceeds pagetable range
@@ -537,6 +540,7 @@ unsigned long pmd_hugepage_update(struct mm_struct *mm, unsigned long addr,
 	old = pmd_val(*pmdp);
 	*pmdp = __pmd((old & ~clr) | set);
 #endif
+	trace_hugepage_update(addr, old, clr, set);
 	if (old & _PAGE_HASHPTE)
 		hpte_do_hugepage_flush(mm, addr, pmdp, old);
 	return old;
@@ -642,6 +646,7 @@ void pmdp_splitting_flush(struct vm_area_struct *vma,
 	 * If we didn't had the splitting flag set, go and flush the
 	 * HPTE entries.
 	 */
+	trace_hugepage_splitting(address, old);
 	if (!(old & _PAGE_SPLITTING)) {
 		/* We need to flush the hpte */
 		if (old & _PAGE_HASHPTE)
@@ -709,6 +714,7 @@ void set_pmd_at(struct mm_struct *mm, unsigned long addr,
 	assert_spin_locked(&mm->page_table_lock);
 	WARN_ON(!pmd_trans_huge(pmd));
 #endif
+	trace_hugepage_set_pmd(addr, pmd);
 	return set_pte_at(mm, addr, pmdp_ptep(pmdp), pmd_pte(pmd));
 }
 
diff --git a/arch/powerpc/mm/tlb_hash64.c b/arch/powerpc/mm/tlb_hash64.c
index 9adda5790463..d2a94b85dbc2 100644
--- a/arch/powerpc/mm/tlb_hash64.c
+++ b/arch/powerpc/mm/tlb_hash64.c
@@ -30,6 +30,8 @@
 #include <asm/tlb.h>
 #include <asm/bug.h>
 
+#include <trace/events/thp.h>
+
 DEFINE_PER_CPU(struct ppc64_tlb_batch, ppc64_tlb_batch);
 
 /*
@@ -213,6 +215,8 @@ void __flush_hash_table_range(struct mm_struct *mm, unsigned long start,
 		if (ptep == NULL)
 			continue;
 		pte = pte_val(*ptep);
+		if (hugepage_shift)
+			trace_hugepage_invalidate(start, pte_val(pte));
 		if (!(pte & _PAGE_HASHPTE))
 			continue;
 		if (unlikely(hugepage_shift && pmd_trans_huge(*(pmd_t *)pte)))
diff --git a/include/trace/events/thp.h b/include/trace/events/thp.h
new file mode 100644
index 000000000000..b59b065e9e5d
--- /dev/null
+++ b/include/trace/events/thp.h
@@ -0,0 +1,88 @@
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM thp
+
+#if !defined(_TRACE_THP_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_THP_H
+
+#include <linux/types.h>
+#include <linux/tracepoint.h>
+
+TRACE_EVENT(hugepage_invalidate,
+
+	    TP_PROTO(unsigned long addr, unsigned long pte),
+	    TP_ARGS(addr, pte),
+	    TP_STRUCT__entry(
+		    __field(unsigned long, addr)
+		    __field(unsigned long, pte)
+		    ),
+
+	    TP_fast_assign(
+		    __entry->addr = addr;
+		    __entry->pte = pte;
+		    ),
+
+	    TP_printk("hugepage invalidate at addr 0x%lx and pte = 0x%lx",
+		      __entry->addr, __entry->pte)
+);
+
+TRACE_EVENT(hugepage_set_pmd,
+
+	    TP_PROTO(unsigned long addr, unsigned long pmd),
+	    TP_ARGS(addr, pmd),
+	    TP_STRUCT__entry(
+		    __field(unsigned long, addr)
+		    __field(unsigned long, pmd)
+		    ),
+
+	    TP_fast_assign(
+		    __entry->addr = addr;
+		    __entry->pmd = pmd;
+		    ),
+
+	    TP_printk("Set pmd with 0x%lx with 0x%lx", __entry->addr, __entry->pmd)
+);
+
+
+TRACE_EVENT(hugepage_update,
+
+	    TP_PROTO(unsigned long addr, unsigned long pte, unsigned long clr, unsigned long set),
+	    TP_ARGS(addr, pte, clr, set),
+	    TP_STRUCT__entry(
+		    __field(unsigned long, addr)
+		    __field(unsigned long, pte)
+		    __field(unsigned long, clr)
+		    __field(unsigned long, set)
+		    ),
+
+	    TP_fast_assign(
+		    __entry->addr = addr;
+		    __entry->pte = pte;
+		    __entry->clr = clr;
+		    __entry->set = set;
+
+		    ),
+
+	    TP_printk("hugepage update at addr 0x%lx and pte = 0x%lx clr = 0x%lx, set = 0x%lx", __entry->addr, __entry->pte, __entry->clr, __entry->set)
+);
+TRACE_EVENT(hugepage_splitting,
+
+	    TP_PROTO(unsigned long addr, unsigned long pte),
+	    TP_ARGS(addr, pte),
+	    TP_STRUCT__entry(
+		    __field(unsigned long, addr)
+		    __field(unsigned long, pte)
+		    ),
+
+	    TP_fast_assign(
+		    __entry->addr = addr;
+		    __entry->pte = pte;
+		    ),
+
+	    TP_printk("hugepage splitting at addr 0x%lx and pte = 0x%lx",
+		      __entry->addr, __entry->pte)
+);
+
+#endif /* _TRACE_THP_H */
+
+/* This part must be outside protection */
+#include <trace/define_trace.h>
-- 
1.9.1