Skip to contentSkip to conversation

powerpc/pseries: Don't opencode HPTE_V_BOLTED

PATCH In review Aneesh Kumar K.V 2 participants 4 messages Oct 24, 2019 – Oct 30, 2019 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/3 powerpc/pseries: Don't opencode HPTE_V_BOLTED

No functional change in this patch.

---
 arch/powerpc/platforms/pseries/lpar.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/platforms/pseries/lpar.c +1 -1
index f87a5c64e24d..3126fc02e50b 100644
@@ -774,7 +774,7 @@ static long pSeries_lpar_hpte_remove(unsigned long hpte_group)
 
 		/* don't remove a bolted entry */
 		lpar_rc = plpar_pte_remove(H_ANDCOND, hpte_group + slot_offset,
-					   (0x1UL << 4), &dummy1, &dummy2);
+					   HPTE_V_BOLTED, &dummy1, &dummy2);
 		if (lpar_rc == H_SUCCESS)
 			return i;
 
Signature · 1 line
2.21.0
#1 lore raw

PATCH v2 2/3 powerpc/pseries: Don't fail hash page table insert for bolted mapping

If the hypervisor returned H_PTEG_FULL for H_ENTER hcall, retry a hash page table insert by removing a random entry from the group.

After some runtime, it is very well possible to find all the 8 hash page table entry slot in the hpte group used for mapping. Don't fail a bolted entry insert in that case. With Storage class memory a user can find this error easily since a namespace enable/disable is equivalent to memory add/remove.

This results in failures as reported below:

$ ndctl create-namespace -r region1 -t pmem -m devdax -a 65536 -s 100M
libndctl: ndctl_dax_enable: dax1.3: failed to enable
  Error: namespace1.2: failed to enable

failed to create namespace: No such device or address

In kernel log we find the details as below:

Unable to create mapping for hot added memory 0xc000042006000000..0xc00004200d000000: -1
dax_pmem: probe of dax1.3 failed with error -14

This indicates that we failed to create a bolted hash table entry for direct-map address backing the namespace.

We also observe failures such that not all namespaces will be enabled with ndctl enable-namespace all command.

---
Changes from v1:
* update commit message

 arch/powerpc/mm/book3s64/hash_utils.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
arch/powerpc/mm/book3s64/hash_utils.c +8 -1
index 6c123760164e..6e5a769ebcb8 100644
@@ -294,7 +294,14 @@ int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
 		ret = mmu_hash_ops.hpte_insert(hpteg, vpn, paddr, tprot,
 					       HPTE_V_BOLTED, psize, psize,
 					       ssize);
-
+		if (ret == -1) {
+			/* Try to remove a non bolted entry */
+			ret = mmu_hash_ops.hpte_remove(hpteg);
+			if (ret != -1)
+				ret = mmu_hash_ops.hpte_insert(hpteg, vpn, paddr, tprot,
+							       HPTE_V_BOLTED, psize, psize,
+							       ssize);
+		}
 		if (ret < 0)
 			break;
 
Signature · 1 line
2.21.0
#2 lore raw

PATCH v2 3/3 powerpc/book3s64/hash: Use secondary hash for bolted mapping if the primary is full

In reply to Aneesh Kumar K.V

With bolted hash page table entry, kernel currently only use primary hash group when inserting the hash page table entry. In the rare case where kernel find all the 8 primary hash slot occupied by bolted entries, this can result in hash page table insert failure for bolted entries. Avoid this by using the secondary hash group.

This is different from what kernel does for the non-bolted mapping. With non-bolted entries kernel will try secondary before removing an existing entry from hash page table group. With bolted prefer primary hash group and hence try to insert the page table entry by removing a slot from primary before trying the secondary hash group.

---
Changes from v1:
* update commit message

 arch/powerpc/mm/book3s64/hash_native.c | 38 ++++++++++++++++++++------
 arch/powerpc/mm/book3s64/hash_utils.c  | 13 ++++++++-
 arch/powerpc/platforms/pseries/lpar.c  | 14 ++++++++--
 3 files changed, 52 insertions(+), 13 deletions(-)
arch/powerpc/mm/book3s64/hash_native.c +29 -9
index 523e42eb11da..d2d8237ea9d5 100644
@@ -482,19 +482,12 @@ static long native_hpte_updatepp(unsigned long slot, unsigned long newpp,
 	return ret;
 }
 
-static long native_hpte_find(unsigned long vpn, int psize, int ssize)
+static long __native_hpte_find(unsigned long want_v, unsigned long slot)
 {
 	struct hash_pte *hptep;
-	unsigned long hash;
+	unsigned long hpte_v;
 	unsigned long i;
-	long slot;
-	unsigned long want_v, hpte_v;
 
-	hash = hpt_hash(vpn, mmu_psize_defs[psize].shift, ssize);
-	want_v = hpte_encode_avpn(vpn, psize, ssize);
-
-	/* Bolted mappings are only ever in the primary group */
-	slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
 	for (i = 0; i < HPTES_PER_GROUP; i++) {
 
 		hptep = htab_address + slot;
@@ -508,6 +501,33 @@ static long native_hpte_find(unsigned long vpn, int psize, int ssize)
 	return -1;
 }
 
+static long native_hpte_find(unsigned long vpn, int psize, int ssize)
+{
+	unsigned long hpte_group;
+	unsigned long want_v;
+	unsigned long hash;
+	long slot;
+
+	hash = hpt_hash(vpn, mmu_psize_defs[psize].shift, ssize);
+	want_v = hpte_encode_avpn(vpn, psize, ssize);
+
+	/*
+	 * We try to keep bolted entries always in primary hash
+	 * But in some case we can find them in secondary too.
+	 */
+	hpte_group = (hash & htab_hash_mask) * HPTES_PER_GROUP;
+	slot = __native_hpte_find(want_v, hpte_group);
+	if (slot < 0) {
+		/* Try in secondary */
+		hpte_group = (~hash & htab_hash_mask) * HPTES_PER_GROUP;
+		slot = __native_hpte_find(want_v, hpte_group);
+		if (slot < 0)
+			return -1;
+	}
+
+	return slot;
+}
+
 /*
  * Update the page protection bits. Intended to be used to create
  * guard pages for kernel data structures on pages which are bolted
arch/powerpc/mm/book3s64/hash_utils.c +12 -1
index 6e5a769ebcb8..a9d1f72de848 100644
@@ -263,6 +263,7 @@ int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
 		unsigned long vsid = get_kernel_vsid(vaddr, ssize);
 		unsigned long vpn  = hpt_vpn(vaddr, vsid, ssize);
 		unsigned long tprot = prot;
+		bool secondary_hash = false;
 
 		/*
 		 * If we hit a bad address return error.
@@ -291,17 +292,27 @@ int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
 		hpteg = ((hash & htab_hash_mask) * HPTES_PER_GROUP);
 
 		BUG_ON(!mmu_hash_ops.hpte_insert);
+repeat:
 		ret = mmu_hash_ops.hpte_insert(hpteg, vpn, paddr, tprot,
 					       HPTE_V_BOLTED, psize, psize,
 					       ssize);
 		if (ret == -1) {
-			/* Try to remove a non bolted entry */
+			/*
+			 * Try to to keep bolted entries in primary.
+			 * Remove non bolted entries and try insert again
+			 */
 			ret = mmu_hash_ops.hpte_remove(hpteg);
 			if (ret != -1)
 				ret = mmu_hash_ops.hpte_insert(hpteg, vpn, paddr, tprot,
 							       HPTE_V_BOLTED, psize, psize,
 							       ssize);
+			if (ret == -1 && !secondary_hash) {
+				secondary_hash = true;
+				hpteg = ((~hash & htab_hash_mask) * HPTES_PER_GROUP);
+				goto repeat;
+			}
 		}
+
 		if (ret < 0)
 			break;
 
arch/powerpc/platforms/pseries/lpar.c +11 -3
index 3126fc02e50b..74c59a1e9627 100644
@@ -938,11 +938,19 @@ static long pSeries_lpar_hpte_find(unsigned long vpn, int psize, int ssize)
 	hash = hpt_hash(vpn, mmu_psize_defs[psize].shift, ssize);
 	want_v = hpte_encode_avpn(vpn, psize, ssize);
 
-	/* Bolted entries are always in the primary group */
+	/*
+	 * We try to keep bolted entries always in primary hash
+	 * But in some case we can find them in secondary too.
+	 */
 	hpte_group = (hash & htab_hash_mask) * HPTES_PER_GROUP;
 	slot = __pSeries_lpar_hpte_find(want_v, hpte_group);
-	if (slot < 0)
-		return -1;
+	if (slot < 0) {
+		/* Try in secondary */
+		hpte_group = (~hash & htab_hash_mask) * HPTES_PER_GROUP;
+		slot = __pSeries_lpar_hpte_find(want_v, hpte_group);
+		if (slot < 0)
+			return -1;
+	}
 	return hpte_group + slot;
 }
 
Signature · 1 line
2.21.0