Message-ID: <20251106161117.350395-14-imbrenda@linux.ibm.com>
Date: Thu,  6 Nov 2025 17:11:07 +0100
From: Claudio Imbrenda <imbrenda@...ux.ibm.com>
To: kvm@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, linux-s390@...r.kernel.org,
        borntraeger@...ibm.com, frankja@...ux.ibm.com, nsg@...ux.ibm.com,
        nrb@...ux.ibm.com, seiden@...ux.ibm.com, schlameuss@...ux.ibm.com,
        hca@...ux.ibm.com, svens@...ux.ibm.com, agordeev@...ux.ibm.com,
        gor@...ux.ibm.com, david@...hat.com, gerald.schaefer@...ux.ibm.com
Subject: [PATCH v3 13/23] KVM: s390: KVM page table management functions: CMMA

Add page table management functions to be used for KVM guest (gmap)
page tables.

This patch adds functions to handle CMMA and the ESSA instruction.

Signed-off-by: Claudio Imbrenda <imbrenda@...ux.ibm.com>
---
 arch/s390/kvm/dat.c | 262 ++++++++++++++++++++++++++++++++++++++++++++
 arch/s390/kvm/dat.h |  27 +++++
 2 files changed, 289 insertions(+)

diff --git a/arch/s390/kvm/dat.c b/arch/s390/kvm/dat.c
index 666aebec72eb..d2a68cee8ff6 100644
--- a/arch/s390/kvm/dat.c
+++ b/arch/s390/kvm/dat.c
@@ -1098,3 +1098,265 @@ int dat_set_prefix_notif_bit(union asce asce, gfn_t gfn)
 		return -EAGAIN;
 	return 0;
 }
+
+/**
+ * dat_perform_essa() - perform ESSA actions on the PGSTE.
+ * @asce: the asce to operate on.
+ * @gfn: the guest page frame to operate on.
+ * @orc: the specific action to perform, see the ESSA_SET_* macros.
+ * @state: the storage attributes to be returned to the guest.
+ * @dirty: returns whether the function dirtied a previously clean entry.
+ *
+ * Context: Called with kvm->mmu_lock held.
+ *
+ * Return:
+ * * 1 if the page state has been altered and the page is to be added to the CBRL
+ * * 0 if the page state has been altered, but the page is not to be added to the CBRL
+ * * -1 if the page state has not been altered and the page is not to be added to the CBRL
+ */
+int dat_perform_essa(union asce asce, gfn_t gfn, int orc, union essa_state *state, bool *dirty)
+{
+	union crste *crstep;
+	union pgste pgste;
+	union pte *ptep;
+	int res = 0;
+
+	if (dat_entry_walk(NULL, gfn, asce, 0, TABLE_TYPE_PAGE_TABLE, &crstep, &ptep)) {
+		*state = (union essa_state) { .exception = 1 };
+		return -1;
+	}
+
+	pgste = pgste_get_lock(ptep);
+
+	*state = (union essa_state) {
+		.content = (ptep->h.i << 1) + (ptep->h.i && pgste.zero),
+		.nodat = pgste.nodat,
+		.usage = pgste.usage,
+		};
+
+	switch (orc) {
+	case ESSA_GET_STATE:
+		res = -1;
+		break;
+	case ESSA_SET_STABLE:
+		pgste.usage = PGSTE_GPS_USAGE_STABLE;
+		pgste.nodat = 0;
+		break;
+	case ESSA_SET_UNUSED:
+		pgste.usage = PGSTE_GPS_USAGE_UNUSED;
+		if (ptep->h.i)
+			res = 1;
+		break;
+	case ESSA_SET_VOLATILE:
+		pgste.usage = PGSTE_GPS_USAGE_VOLATILE;
+		if (ptep->h.i)
+			res = 1;
+		break;
+	case ESSA_SET_POT_VOLATILE:
+		if (!ptep->h.i) {
+			pgste.usage = PGSTE_GPS_USAGE_POT_VOLATILE;
+		} else if (pgste.zero) {
+			pgste.usage = PGSTE_GPS_USAGE_VOLATILE;
+		} else if (!pgste.gc) {
+			pgste.usage = PGSTE_GPS_USAGE_VOLATILE;
+			res = 1;
+		}
+		break;
+	case ESSA_SET_STABLE_RESIDENT:
+		pgste.usage = PGSTE_GPS_USAGE_STABLE;
+		/*
+		 * Since the resident state can go away any time after this
+		 * call, we will not make this page resident. We can revisit
+		 * this decision if a guest will ever start using this.
+		 */
+		break;
+	case ESSA_SET_STABLE_IF_RESIDENT:
+		if (!ptep->h.i)
+			pgste.usage = PGSTE_GPS_USAGE_STABLE;
+		break;
+	case ESSA_SET_STABLE_NODAT:
+		pgste.usage = PGSTE_GPS_USAGE_STABLE;
+		pgste.nodat = 1;
+		break;
+	default:
+		WARN_ONCE(1, "Invalid ORC!");
+		res = -1;
+		break;
+	}
+	/* If we are discarding a page, set it to logical zero */
+	pgste.zero = res == 1;
+	if (orc > 0) {
+		*dirty = !pgste.cmma_d;
+		pgste.cmma_d = 1;
+	}
+
+	pgste_set_unlock(ptep, pgste);
+
+	return res;
+}
+
+static long dat_reset_cmma_pte(union pte *ptep, gfn_t gfn, gfn_t next, struct dat_walk *walk)
+{
+	union pgste pgste;
+
+	pgste = pgste_get_lock(ptep);
+	pgste.usage = 0;
+	pgste.nodat = 0;
+	pgste.cmma_d = 0;
+	pgste_set_unlock(ptep, pgste);
+	if (need_resched())
+		return next;
+	return 0;
+}
+
+long dat_reset_cmma(union asce asce, gfn_t start)
+{
+	const struct dat_walk_ops dat_reset_cmma_ops = {
+		.pte_entry = dat_reset_cmma_pte,
+	};
+
+	return _dat_walk_gfn_range(start, asce_end(asce), asce, &dat_reset_cmma_ops,
+				   DAT_WALK_IGN_HOLES, NULL);
+}
+
+struct dat_get_cmma_state {
+	gfn_t start;
+	gfn_t end;
+	unsigned int count;
+	u8 *values;
+	atomic64_t *remaining;
+};
+
+static long __dat_peek_cmma_pte(union pte *ptep, gfn_t gfn, gfn_t next, struct dat_walk *walk)
+{
+	struct dat_get_cmma_state *state = walk->priv;
+	union pgste pgste;
+
+	pgste = pgste_get_lock(ptep);
+	state->values[gfn - walk->start] = pgste.usage | (pgste.nodat << 6);
+	pgste_set_unlock(ptep, pgste);
+	state->end = next;
+
+	return 0;
+}
+
+static long __dat_peek_cmma_crste(union crste *crstep, gfn_t gfn, gfn_t next, struct dat_walk *walk)
+{
+	struct dat_get_cmma_state *state = walk->priv;
+
+	if (crstep->h.i)
+		state->end = min(walk->end, next);
+	return 0;
+}
+
+int dat_peek_cmma(gfn_t start, union asce asce, unsigned int *count, u8 *values)
+{
+	const struct dat_walk_ops ops = {
+		.pte_entry = __dat_peek_cmma_pte,
+		.pmd_entry = __dat_peek_cmma_crste,
+		.pud_entry = __dat_peek_cmma_crste,
+		.p4d_entry = __dat_peek_cmma_crste,
+		.pgd_entry = __dat_peek_cmma_crste,
+	};
+	struct dat_get_cmma_state state = { .values = values, };
+	int rc;
+
+	rc = _dat_walk_gfn_range(start, start + *count, asce, &ops, DAT_WALK_DEFAULT, &state);
+	*count = state.end - start;
+	/* Return success if at least one value was saved, otherwise an error. */
+	return (rc == -EFAULT && *count > 0) ? 0 : rc;
+}
+
+static long __dat_get_cmma_pte(union pte *ptep, gfn_t gfn, gfn_t next, struct dat_walk *walk)
+{
+	struct dat_get_cmma_state *state = walk->priv;
+	union pgste pgste;
+
+	if (state->start != -1) {
+		if ((gfn - state->end) > KVM_S390_MAX_BIT_DISTANCE)
+			return 1;
+		if (gfn - state->start >= state->count)
+			return 1;
+	}
+
+	if (!READ_ONCE(*pgste_of(ptep)).cmma_d)
+		return 0;
+
+	pgste = pgste_get_lock(ptep);
+	if (pgste.cmma_d) {
+		if (state->start == -1)
+			state->start = gfn;
+		pgste.cmma_d = 0;
+		atomic64_dec(state->remaining);
+		state->values[gfn - state->start] = pgste.usage | pgste.nodat << 6;
+		state->end = next;
+	}
+	pgste_set_unlock(ptep, pgste);
+	return 0;
+}
+
+int dat_get_cmma(union asce asce, gfn_t *start, unsigned int *count, u8 *values, atomic64_t *rem)
+{
+	const struct dat_walk_ops ops = { .pte_entry = __dat_get_cmma_pte, };
+	struct dat_get_cmma_state state = {
+		.remaining = rem,
+		.values = values,
+		.count = *count,
+		.start = -1,
+	};
+
+	_dat_walk_gfn_range(*start, asce_end(asce), asce, &ops, DAT_WALK_IGN_HOLES, &state);
+
+	if (state.start == -1) {
+		*count = 0;
+	} else {
+		*count = state.end - state.start;
+		*start = state.start;
+	}
+
+	return 0;
+}
+
+struct dat_set_cmma_state {
+	unsigned long mask;
+	const u8 *bits;
+};
+
+static long __dat_set_cmma_pte(union pte *ptep, gfn_t gfn, gfn_t next, struct dat_walk *walk)
+{
+	struct dat_set_cmma_state *state = walk->priv;
+	union pgste pgste, tmp;
+
+	tmp.val = (state->bits[gfn - walk->start] << 24) & state->mask;
+
+	pgste = pgste_get_lock(ptep);
+	pgste.usage = tmp.usage;
+	pgste.nodat = tmp.nodat;
+	pgste_set_unlock(ptep, pgste);
+
+	return 0;
+}
+
+/*
+ * This function sets the CMMA attributes for the given pages. If the input
+ * buffer has zero length, no action is taken, otherwise the attributes are
+ * set and the mm->context.uses_cmm flag is set.
+ */
+int dat_set_cmma_bits(struct kvm_s390_mmu_cache *mc, union asce asce, gfn_t gfn,
+		      unsigned long count, unsigned long mask, const uint8_t *bits)
+{
+	const struct dat_walk_ops ops = { .pte_entry = __dat_set_cmma_pte, };
+	struct dat_set_cmma_state state = { .mask = mask, .bits = bits, };
+	union crste *crstep;
+	union pte *ptep;
+	gfn_t cur;
+	int rc;
+
+	for (cur = ALIGN_DOWN(gfn, _PAGE_ENTRIES); cur < gfn + count; cur += _PAGE_ENTRIES) {
+		rc = dat_entry_walk(mc, cur, asce, DAT_WALK_ALLOC, TABLE_TYPE_PAGE_TABLE,
+				    &crstep, &ptep);
+		if (rc)
+			return rc;
+	}
+	return _dat_walk_gfn_range(gfn, gfn + count, asce, &ops, DAT_WALK_IGN_HOLES, &state);
+}
diff --git a/arch/s390/kvm/dat.h b/arch/s390/kvm/dat.h
index 0dc80893ed7c..098cd454b161 100644
--- a/arch/s390/kvm/dat.h
+++ b/arch/s390/kvm/dat.h
@@ -18,6 +18,15 @@
 #include <asm/tlbflush.h>
 #include <asm/dat-bits.h>
 
+/*
+ * Base address and length must be sent at the start of each block, therefore
+ * it's cheaper to send some clean data, as long as it's less than the size of
+ * two longs.
+ */
+#define KVM_S390_MAX_BIT_DISTANCE (2 * sizeof(void *))
+/* For consistency */
+#define KVM_S390_CMMA_SIZE_MAX ((u32)KVM_S390_SKEYS_MAX)
+
 #define _ASCE(x) ((union asce) { .val = (x), })
 #define NULL_ASCE _ASCE(0)
 
@@ -434,6 +443,17 @@ static inline union crste _crste_fc1(kvm_pfn_t pfn, int tt, bool writable, bool
 	return res;
 }
 
+union essa_state {
+	unsigned char val;
+	struct {
+		unsigned char		: 2;
+		unsigned char nodat	: 1;
+		unsigned char exception	: 1;
+		unsigned char usage	: 2;
+		unsigned char content	: 2;
+	};
+};
+
 /**
  * struct vsie_rmap - reverse mapping for shadow page table entries
  * @next: pointer to next rmap in the list
@@ -523,6 +543,13 @@ bool dat_test_age_gfn(union asce asce, gfn_t start, gfn_t end);
 int dat_link(struct kvm_s390_mmu_cache *mc, union asce asce, int level,
 	     bool uses_skeys, struct guest_fault *f);
 
+int dat_perform_essa(union asce asce, gfn_t gfn, int orc, union essa_state *state, bool *dirty);
+long dat_reset_cmma(union asce asce, gfn_t start_gfn);
+int dat_peek_cmma(gfn_t start, union asce asce, unsigned int *count, u8 *values);
+int dat_get_cmma(union asce asce, gfn_t *start, unsigned int *count, u8 *values, atomic64_t *rem);
+int dat_set_cmma_bits(struct kvm_s390_mmu_cache *mc, union asce asce, gfn_t gfn,
+		      unsigned long count, unsigned long mask, const uint8_t *bits);
+
 int kvm_s390_mmu_cache_topup(struct kvm_s390_mmu_cache *mc);
 
 #define GFP_KVM_S390_MMU_CACHE (GFP_ATOMIC | __GFP_ACCOUNT | __GFP_NOWARN)
-- 
2.51.1

Powered by blists - more mailing lists