Skip to contentSkip to conversation

KVM: arm64: pKVM memory transitions cleanup

PATCH In review Oliver Upton 3 participants 7 messages Oct 28, 2022 – Nov 10, 2022

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

cover KVM: arm64: pKVM memory transitions cleanup

In order to help resolve my own bikeshedding on the outstanding pKVM patches [1], small deck of patches to polish up the existing memory transitions. Mainly:

- Rejig the layout of pkvm_mem_transition
- Stop using out pointers to get at the 'completer' addr
- Use better-fitting terminology (source/target) to describe the addresses involved in a memory transition

Applies to 6.1-rc2. Politely compile tested, and that's just about it.

Oliver Upton (2):
KVM: arm64: Clean out the odd handling of completer_addr
KVM: arm64: Redefine pKVM memory transitions in terms of source/target

 arch/arm64/kvm/hyp/nvhe/mem_protect.c | 124 +++++++++++---------------
 1 file changed, 52 insertions(+), 72 deletions(-)

base-commit: 247f34f7b80357943234f93f247a1ae6b6c3a740

Signature · 1 line
2.38.1.273.g43a17bfeac-goog
#1 lore raw

PATCH 1/2 KVM: arm64: Clean out the odd handling of completer_addr

The layout of struct pkvm_mem_transition is a bit weird; the destination address for the transition is actually stashed in the initiator address context. Even weirder so, that address is thrown inside a union and return from helpers by use of an out pointer.

Rip out the whole mess and move the destination address into the destination context sub-struct. No functional change intended.

---
 arch/arm64/kvm/hyp/nvhe/mem_protect.c | 70 ++++++++++-----------------
 1 file changed, 25 insertions(+), 45 deletions(-)
arch/arm64/kvm/hyp/nvhe/mem_protect.c +25 -45
index 1e78acf9662e..3636a24e1b34 100644
@@ -393,17 +393,12 @@ struct pkvm_mem_transition {
 		enum pkvm_component_id	id;
 		/* Address in the initiator's address space */
 		u64			addr;
-
-		union {
-			struct {
-				/* Address in the completer's address space */
-				u64	completer_addr;
-			} host;
-		};
 	} initiator;
 
 	struct {
 		enum pkvm_component_id	id;
+		/* Address in the completer's address space */
+		u64			addr;
 	} completer;
 };
 
@@ -471,43 +466,35 @@ static int __host_set_page_state_range(u64 addr, u64 size,
 	return host_stage2_idmap_locked(addr, size, prot);
 }
 
-static int host_request_owned_transition(u64 *completer_addr,
-					 const struct pkvm_mem_transition *tx)
+static int host_request_owned_transition(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
 	u64 addr = tx->initiator.addr;
 
-	*completer_addr = tx->initiator.host.completer_addr;
 	return __host_check_page_state_range(addr, size, PKVM_PAGE_OWNED);
 }
 
-static int host_request_unshare(u64 *completer_addr,
-				const struct pkvm_mem_transition *tx)
+static int host_request_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
 	u64 addr = tx->initiator.addr;
 
-	*completer_addr = tx->initiator.host.completer_addr;
 	return __host_check_page_state_range(addr, size, PKVM_PAGE_SHARED_OWNED);
 }
 
-static int host_initiate_share(u64 *completer_addr,
-			       const struct pkvm_mem_transition *tx)
+static int host_initiate_share(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
 	u64 addr = tx->initiator.addr;
 
-	*completer_addr = tx->initiator.host.completer_addr;
 	return __host_set_page_state_range(addr, size, PKVM_PAGE_SHARED_OWNED);
 }
 
-static int host_initiate_unshare(u64 *completer_addr,
-				 const struct pkvm_mem_transition *tx)
+static int host_initiate_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
 	u64 addr = tx->initiator.addr;
 
-	*completer_addr = tx->initiator.host.completer_addr;
 	return __host_set_page_state_range(addr, size, PKVM_PAGE_OWNED);
 }
 
@@ -537,7 +524,7 @@ static bool __hyp_ack_skip_pgtable_check(const struct pkvm_mem_transition *tx)
 		 tx->initiator.id != PKVM_ID_HOST);
 }
 
-static int hyp_ack_share(u64 addr, const struct pkvm_mem_transition *tx,
+static int hyp_ack_share(const struct pkvm_mem_transition *tx,
 			 enum kvm_pgtable_prot perms)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
@@ -548,34 +535,35 @@ static int hyp_ack_share(u64 addr, const struct pkvm_mem_transition *tx,
 	if (__hyp_ack_skip_pgtable_check(tx))
 		return 0;
 
-	return __hyp_check_page_state_range(addr, size, PKVM_NOPAGE);
+	return __hyp_check_page_state_range(tx->completer.addr, size, PKVM_NOPAGE);
 }
 
-static int hyp_ack_unshare(u64 addr, const struct pkvm_mem_transition *tx)
+static int hyp_ack_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
 
 	if (__hyp_ack_skip_pgtable_check(tx))
 		return 0;
 
-	return __hyp_check_page_state_range(addr, size,
+	return __hyp_check_page_state_range(tx->completer.addr, size,
 					    PKVM_PAGE_SHARED_BORROWED);
 }
 
-static int hyp_complete_share(u64 addr, const struct pkvm_mem_transition *tx,
+static int hyp_complete_share(const struct pkvm_mem_transition *tx,
 			      enum kvm_pgtable_prot perms)
 {
-	void *start = (void *)addr, *end = start + (tx->nr_pages * PAGE_SIZE);
+	void *start = (void *)tx->completer.addr;
+	void *end = start + (tx->nr_pages * PAGE_SIZE);
 	enum kvm_pgtable_prot prot;
 
 	prot = pkvm_mkstate(perms, PKVM_PAGE_SHARED_BORROWED);
 	return pkvm_create_mappings_locked(start, end, prot);
 }
 
-static int hyp_complete_unshare(u64 addr, const struct pkvm_mem_transition *tx)
+static int hyp_complete_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
-	int ret = kvm_pgtable_hyp_unmap(&pkvm_pgtable, addr, size);
+	int ret = kvm_pgtable_hyp_unmap(&pkvm_pgtable, tx->completer.addr, size);
 
 	return (ret != size) ? -EFAULT : 0;
 }
@@ -583,12 +571,11 @@ static int hyp_complete_unshare(u64 addr, const struct pkvm_mem_transition *tx)
 static int check_share(struct pkvm_mem_share *share)
 {
 	const struct pkvm_mem_transition *tx = &share->tx;
-	u64 completer_addr;
 	int ret;
 
 	switch (tx->initiator.id) {
 	case PKVM_ID_HOST:
-		ret = host_request_owned_transition(&completer_addr, tx);
+		ret = host_request_owned_transition(tx);
 		break;
 	default:
 		ret = -EINVAL;
@@ -599,7 +586,7 @@ static int check_share(struct pkvm_mem_share *share)
 
 	switch (tx->completer.id) {
 	case PKVM_ID_HYP:
-		ret = hyp_ack_share(completer_addr, tx, share->completer_prot);
+		ret = hyp_ack_share(tx, share->completer_prot);
 		break;
 	default:
 		ret = -EINVAL;
@@ -611,12 +598,11 @@ static int check_share(struct pkvm_mem_share *share)
 static int __do_share(struct pkvm_mem_share *share)
 {
 	const struct pkvm_mem_transition *tx = &share->tx;
-	u64 completer_addr;
 	int ret;
 
 	switch (tx->initiator.id) {
 	case PKVM_ID_HOST:
-		ret = host_initiate_share(&completer_addr, tx);
+		ret = host_initiate_share(tx);
 		break;
 	default:
 		ret = -EINVAL;
@@ -627,7 +613,7 @@ static int __do_share(struct pkvm_mem_share *share)
 
 	switch (tx->completer.id) {
 	case PKVM_ID_HYP:
-		ret = hyp_complete_share(completer_addr, tx, share->completer_prot);
+		ret = hyp_complete_share(tx, share->completer_prot);
 		break;
 	default:
 		ret = -EINVAL;
@@ -659,12 +645,11 @@ static int do_share(struct pkvm_mem_share *share)
 static int check_unshare(struct pkvm_mem_share *share)
 {
 	const struct pkvm_mem_transition *tx = &share->tx;
-	u64 completer_addr;
 	int ret;
 
 	switch (tx->initiator.id) {
 	case PKVM_ID_HOST:
-		ret = host_request_unshare(&completer_addr, tx);
+		ret = host_request_unshare(tx);
 		break;
 	default:
 		ret = -EINVAL;
@@ -675,7 +660,7 @@ static int check_unshare(struct pkvm_mem_share *share)
 
 	switch (tx->completer.id) {
 	case PKVM_ID_HYP:
-		ret = hyp_ack_unshare(completer_addr, tx);
+		ret = hyp_ack_unshare(tx);
 		break;
 	default:
 		ret = -EINVAL;
@@ -687,12 +672,11 @@ static int check_unshare(struct pkvm_mem_share *share)
 static int __do_unshare(struct pkvm_mem_share *share)
 {
 	const struct pkvm_mem_transition *tx = &share->tx;
-	u64 completer_addr;
 	int ret;
 
 	switch (tx->initiator.id) {
 	case PKVM_ID_HOST:
-		ret = host_initiate_unshare(&completer_addr, tx);
+		ret = host_initiate_unshare(tx);
 		break;
 	default:
 		ret = -EINVAL;
@@ -703,7 +687,7 @@ static int __do_unshare(struct pkvm_mem_share *share)
 
 	switch (tx->completer.id) {
 	case PKVM_ID_HYP:
-		ret = hyp_complete_unshare(completer_addr, tx);
+		ret = hyp_complete_unshare(tx);
 		break;
 	default:
 		ret = -EINVAL;
@@ -743,12 +727,10 @@ int __pkvm_host_share_hyp(u64 pfn)
 			.initiator	= {
 				.id	= PKVM_ID_HOST,
 				.addr	= host_addr,
-				.host	= {
-					.completer_addr = hyp_addr,
-				},
 			},
 			.completer	= {
 				.id	= PKVM_ID_HYP,
+				.addr	= hyp_addr,
 			},
 		},
 		.completer_prot	= PAGE_HYP,
@@ -776,12 +758,10 @@ int __pkvm_host_unshare_hyp(u64 pfn)
 			.initiator	= {
 				.id	= PKVM_ID_HOST,
 				.addr	= host_addr,
-				.host	= {
-					.completer_addr = hyp_addr,
-				},
 			},
 			.completer	= {
 				.id	= PKVM_ID_HYP,
+				.addr	= hyp_addr,
 			},
 		},
 		.completer_prot	= PAGE_HYP,
Signature · 1 line
2.38.1.273.g43a17bfeac-goog
#3 lore raw

PATCH 2/2 KVM: arm64: Redefine pKVM memory transitions in terms of source/target

In reply to Oliver Upton

Perhaps it is just me, but the 'initiator' and 'completer' terms are slightly confusing descriptors for the addresses involved in a memory transition. Apply a rename to instead describe memory transitions in terms of a source and target address.

No functional change intended.

---
 arch/arm64/kvm/hyp/nvhe/mem_protect.c | 68 +++++++++++++--------------
 1 file changed, 34 insertions(+), 34 deletions(-)
arch/arm64/kvm/hyp/nvhe/mem_protect.c +34 -34
index 3636a24e1b34..3ea389a8166f 100644
@@ -391,20 +391,20 @@ struct pkvm_mem_transition {
 
 	struct {
 		enum pkvm_component_id	id;
-		/* Address in the initiator's address space */
+		/* Address in the source's address space */
 		u64			addr;
-	} initiator;
+	} source;
 
 	struct {
 		enum pkvm_component_id	id;
-		/* Address in the completer's address space */
+		/* Address in the target's address space */
 		u64			addr;
-	} completer;
+	} target;
 };
 
 struct pkvm_mem_share {
 	const struct pkvm_mem_transition	tx;
-	const enum kvm_pgtable_prot		completer_prot;
+	const enum kvm_pgtable_prot		target_prot;
 };
 
 struct check_walk_data {
@@ -469,7 +469,7 @@ static int __host_set_page_state_range(u64 addr, u64 size,
 static int host_request_owned_transition(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
-	u64 addr = tx->initiator.addr;
+	u64 addr = tx->source.addr;
 
 	return __host_check_page_state_range(addr, size, PKVM_PAGE_OWNED);
 }
@@ -477,7 +477,7 @@ static int host_request_owned_transition(const struct pkvm_mem_transition *tx)
 static int host_request_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
-	u64 addr = tx->initiator.addr;
+	u64 addr = tx->source.addr;
 
 	return __host_check_page_state_range(addr, size, PKVM_PAGE_SHARED_OWNED);
 }
@@ -485,7 +485,7 @@ static int host_request_unshare(const struct pkvm_mem_transition *tx)
 static int host_initiate_share(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
-	u64 addr = tx->initiator.addr;
+	u64 addr = tx->source.addr;
 
 	return __host_set_page_state_range(addr, size, PKVM_PAGE_SHARED_OWNED);
 }
@@ -493,7 +493,7 @@ static int host_initiate_share(const struct pkvm_mem_transition *tx)
 static int host_initiate_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
-	u64 addr = tx->initiator.addr;
+	u64 addr = tx->source.addr;
 
 	return __host_set_page_state_range(addr, size, PKVM_PAGE_OWNED);
 }
@@ -521,7 +521,7 @@ static int __hyp_check_page_state_range(u64 addr, u64 size,
 static bool __hyp_ack_skip_pgtable_check(const struct pkvm_mem_transition *tx)
 {
 	return !(IS_ENABLED(CONFIG_NVHE_EL2_DEBUG) ||
-		 tx->initiator.id != PKVM_ID_HOST);
+		 tx->source.id != PKVM_ID_HOST);
 }
 
 static int hyp_ack_share(const struct pkvm_mem_transition *tx,
@@ -535,7 +535,7 @@ static int hyp_ack_share(const struct pkvm_mem_transition *tx,
 	if (__hyp_ack_skip_pgtable_check(tx))
 		return 0;
 
-	return __hyp_check_page_state_range(tx->completer.addr, size, PKVM_NOPAGE);
+	return __hyp_check_page_state_range(tx->target.addr, size, PKVM_NOPAGE);
 }
 
 static int hyp_ack_unshare(const struct pkvm_mem_transition *tx)
@@ -545,14 +545,14 @@ static int hyp_ack_unshare(const struct pkvm_mem_transition *tx)
 	if (__hyp_ack_skip_pgtable_check(tx))
 		return 0;
 
-	return __hyp_check_page_state_range(tx->completer.addr, size,
+	return __hyp_check_page_state_range(tx->target.addr, size,
 					    PKVM_PAGE_SHARED_BORROWED);
 }
 
 static int hyp_complete_share(const struct pkvm_mem_transition *tx,
 			      enum kvm_pgtable_prot perms)
 {
-	void *start = (void *)tx->completer.addr;
+	void *start = (void *)tx->target.addr;
 	void *end = start + (tx->nr_pages * PAGE_SIZE);
 	enum kvm_pgtable_prot prot;
 
@@ -563,7 +563,7 @@ static int hyp_complete_share(const struct pkvm_mem_transition *tx,
 static int hyp_complete_unshare(const struct pkvm_mem_transition *tx)
 {
 	u64 size = tx->nr_pages * PAGE_SIZE;
-	int ret = kvm_pgtable_hyp_unmap(&pkvm_pgtable, tx->completer.addr, size);
+	int ret = kvm_pgtable_hyp_unmap(&pkvm_pgtable, tx->target.addr, size);
 
 	return (ret != size) ? -EFAULT : 0;
 }
@@ -573,7 +573,7 @@ static int check_share(struct pkvm_mem_share *share)
 	const struct pkvm_mem_transition *tx = &share->tx;
 	int ret;
 
-	switch (tx->initiator.id) {
+	switch (tx->source.id) {
 	case PKVM_ID_HOST:
 		ret = host_request_owned_transition(tx);
 		break;
@@ -584,9 +584,9 @@ static int check_share(struct pkvm_mem_share *share)
 	if (ret)
 		return ret;
 
-	switch (tx->completer.id) {
+	switch (tx->target.id) {
 	case PKVM_ID_HYP:
-		ret = hyp_ack_share(tx, share->completer_prot);
+		ret = hyp_ack_share(tx, share->target_prot);
 		break;
 	default:
 		ret = -EINVAL;
@@ -600,7 +600,7 @@ static int __do_share(struct pkvm_mem_share *share)
 	const struct pkvm_mem_transition *tx = &share->tx;
 	int ret;
 
-	switch (tx->initiator.id) {
+	switch (tx->source.id) {
 	case PKVM_ID_HOST:
 		ret = host_initiate_share(tx);
 		break;
@@ -611,9 +611,9 @@ static int __do_share(struct pkvm_mem_share *share)
 	if (ret)
 		return ret;
 
-	switch (tx->completer.id) {
+	switch (tx->target.id) {
 	case PKVM_ID_HYP:
-		ret = hyp_complete_share(tx, share->completer_prot);
+		ret = hyp_complete_share(tx, share->target_prot);
 		break;
 	default:
 		ret = -EINVAL;
@@ -628,8 +628,8 @@ static int __do_share(struct pkvm_mem_share *share)
  * The page owner grants access to another component with a given set
  * of permissions.
  *
- * Initiator: OWNED	=> SHARED_OWNED
- * Completer: NOPAGE	=> SHARED_BORROWED
+ * Source: OWNED	=> SHARED_OWNED
+ * Target: NOPAGE	=> SHARED_BORROWED
  */
 static int do_share(struct pkvm_mem_share *share)
 {
@@ -647,7 +647,7 @@ static int check_unshare(struct pkvm_mem_share *share)
 	const struct pkvm_mem_transition *tx = &share->tx;
 	int ret;
 
-	switch (tx->initiator.id) {
+	switch (tx->source.id) {
 	case PKVM_ID_HOST:
 		ret = host_request_unshare(tx);
 		break;
@@ -658,7 +658,7 @@ static int check_unshare(struct pkvm_mem_share *share)
 	if (ret)
 		return ret;
 
-	switch (tx->completer.id) {
+	switch (tx->target.id) {
 	case PKVM_ID_HYP:
 		ret = hyp_ack_unshare(tx);
 		break;
@@ -674,7 +674,7 @@ static int __do_unshare(struct pkvm_mem_share *share)
 	const struct pkvm_mem_transition *tx = &share->tx;
 	int ret;
 
-	switch (tx->initiator.id) {
+	switch (tx->source.id) {
 	case PKVM_ID_HOST:
 		ret = host_initiate_unshare(tx);
 		break;
@@ -685,7 +685,7 @@ static int __do_unshare(struct pkvm_mem_share *share)
 	if (ret)
 		return ret;
 
-	switch (tx->completer.id) {
+	switch (tx->target.id) {
 	case PKVM_ID_HYP:
 		ret = hyp_complete_unshare(tx);
 		break;
@@ -702,8 +702,8 @@ static int __do_unshare(struct pkvm_mem_share *share)
  * The page owner revokes access from another component for a range of
  * pages which were previously shared using do_share().
  *
- * Initiator: SHARED_OWNED	=> OWNED
- * Completer: SHARED_BORROWED	=> NOPAGE
+ * Source: SHARED_OWNED	=> OWNED
+ * Target: SHARED_BORROWED	=> NOPAGE
  */
 static int do_unshare(struct pkvm_mem_share *share)
 {
@@ -724,16 +724,16 @@ int __pkvm_host_share_hyp(u64 pfn)
 	struct pkvm_mem_share share = {
 		.tx	= {
 			.nr_pages	= 1,
-			.initiator	= {
+			.source	= {
 				.id	= PKVM_ID_HOST,
 				.addr	= host_addr,
 			},
-			.completer	= {
+			.target	= {
 				.id	= PKVM_ID_HYP,
 				.addr	= hyp_addr,
 			},
 		},
-		.completer_prot	= PAGE_HYP,
+		.target_prot	= PAGE_HYP,
 	};
 
 	host_lock_component();
@@ -755,16 +755,16 @@ int __pkvm_host_unshare_hyp(u64 pfn)
 	struct pkvm_mem_share share = {
 		.tx	= {
 			.nr_pages	= 1,
-			.initiator	= {
+			.source	= {
 				.id	= PKVM_ID_HOST,
 				.addr	= host_addr,
 			},
-			.completer	= {
+			.target	= {
 				.id	= PKVM_ID_HYP,
 				.addr	= hyp_addr,
 			},
 		},
-		.completer_prot	= PAGE_HYP,
+		.target_prot	= PAGE_HYP,
 	};
 
 	host_lock_component();
Signature · 1 line
2.38.1.273.g43a17bfeac-goog
#4 lore raw

Re: [PATCH 2/2] KVM: arm64: Redefine pKVM memory transitions in terms of source/target

Hey Oliver,

Quoted 4 lines · Oliver Upton

Perhaps it is just me, but the 'initiator' and 'completer' terms are slightly confusing descriptors for the addresses involved in a memory transition. Apply a rename to instead describe memory transitions in terms of a source and target address.

Just to provide some rationale for the initiator/completer terminology, the very first implementation we did of this used 'sender/recipient (or something along those lines I think), and we ended up confusing ourselves massively. The main issue is that memory doesn't necessarily 'flow' in the same direction as the transition. It's all fine for a donation or a share, but reclaim and unshare become funny. 'The recipient of an unshare' can be easily misunderstood, I think.

So yeah, we ended up with initiator/completer, which may not be the prettiest terminology, but it was useful to disambiguate things at least.

Thanks for the review!
Quentin

#5 lore raw

Re: [PATCH 2/2] KVM: arm64: Redefine pKVM memory transitions in terms of source/target

Quentin,

Quoted 16 lines · Quentin Perret

Hey Oliver,

Quoted 4 lines · Oliver Upton

Perhaps it is just me, but the 'initiator' and 'completer' terms are slightly confusing descriptors for the addresses involved in a memory transition. Apply a rename to instead describe memory transitions in terms of a source and target address.

Just to provide some rationale for the initiator/completer terminology, the very first implementation we did of this used 'sender/recipient (or something along those lines I think), and we ended up confusing ourselves massively. The main issue is that memory doesn't necessarily 'flow' in the same direction as the transition. It's all fine for a donation or a share, but reclaim and unshare become funny. 'The recipient of an unshare' can be easily misunderstood, I think.

Quoted 3 lines

So yeah, we ended up with initiator/completer, which may not be the prettiest terminology, but it was useful to disambiguate things at least.

I see, thanks for the background :) If I've managed to re-ambiguate the language here then LMK. Frankly, I'm more strongly motivated on the first patch anyway.

--
Thanks,
Oliver

maintainer
#2 lore raw

Re: [PATCH 1/2] KVM: arm64: Clean out the odd handling of completer_addr

In reply to Oliver Upton

Hi Oliver,

Quoted 26 lines · Oliver Upton

The layout of struct pkvm_mem_transition is a bit weird; the destination address for the transition is actually stashed in the initiator address context. Even weirder so, that address is thrown inside a union and return from helpers by use of an out pointer.

Rip out the whole mess and move the destination address into the destination context sub-struct. No functional change intended.

---
 arch/arm64/kvm/hyp/nvhe/mem_protect.c | 70 ++++++++++-----------------
 1 file changed, 25 insertions(+), 45 deletions(-)
arch/arm64/kvm/hyp/nvhe/mem_protect.c +0 -5
index 1e78acf9662e..3636a24e1b34 100644
@@ -393,17 +393,12 @@ struct pkvm_mem_transition {
 		enum pkvm_component_id	id;
 		/* Address in the initiator's address space */
 		u64			addr;
-
-		union {
-			struct {
-				/* Address in the completer's address space */
-				u64	completer_addr;

arch/arm64/kvm/hyp/nvhe/mem_protect.c struct pkvm_mem_transition { · line 403

Quoted 10 lines
arch/arm64/kvm/hyp/nvhe/mem_protect.c +2 -2
-			} host;
-		};
 	} initiator;
 
 	struct {
 		enum pkvm_component_id	id;
+		/* Address in the completer's address space */
+		u64			addr;
 	} completer;
 };

I'm reasonably sure we'll end up putting this back like we had it as we gain support for guest-initiatied transitions, where we have to walk the guest stage-2 page-table to figure out the physical address of the memory being shared, which is the host (completer's) IPA thanks to the identity mapping there.

So here's what I'll do: I'll post a v6 of the EL2 state series, and I'll include this at the end (before the RFC patch) and let Marc decide whether to go ahead with it. I do agree that it cleans things up for now but, as above, I think that's likely to be a short-lived change.

Sound reasonable?

Cheers,

Will

maintainer
#6 lore raw

Re: [PATCH 2/2] KVM: arm64: Redefine pKVM memory transitions in terms of source/target

In reply to Oliver Upton
Quoted 15 lines · Oliver Upton
Quoted 14 lines · Quentin Perret
Quoted 4 lines · Oliver Upton

Perhaps it is just me, but the 'initiator' and 'completer' terms are slightly confusing descriptors for the addresses involved in a memory transition. Apply a rename to instead describe memory transitions in terms of a source and target address.

Just to provide some rationale for the initiator/completer terminology, the very first implementation we did of this used 'sender/recipient (or something along those lines I think), and we ended up confusing ourselves massively. The main issue is that memory doesn't necessarily 'flow' in the same direction as the transition. It's all fine for a donation or a share, but reclaim and unshare become funny. 'The recipient of an unshare' can be easily misunderstood, I think.

Quoted 7 lines
Quoted 3 lines

So yeah, we ended up with initiator/completer, which may not be the prettiest terminology, but it was useful to disambiguate things at least.

I see, thanks for the background :) If I've managed to re-ambiguate the language here then LMK. Frankly, I'm more strongly motivated on the first patch anyway.

Having been previously tangled up in the confusion mentioned by Quentin, I'm also strongly in favour of leaving the terminology as-is for the time being. Once we have some of the more interesting memory transitions (i.e. approaching the cross-product of host/guest/hyp/trustzone doing share/unshare/donate) then I think we'll be in a much better position to improve the naming, but whatever we change now is very unlikely to stick and the patches as we have them now are at least consistent.

I replied separately on the first patch, as I don't really have a strong opinion on that one.

Will