KVM: SEV: Do cache maintenance on the source VM during intra-host migrationgrafted

Manually perform cache maintenance on the source VM during intra-host
migration to ensure no stale data is left in CPU caches after the VM is
destroyed.  Because the source VM is "converted" to a non-SEV VM, KVM's
memory reclaim flows won't trigger cache maintenance, e.g. when all guest
memory is reclaimed in response to detaching from the mmu_notifier.

Note, relying on the destination VM to do cache maintenance isn't an option
as KVM doesn't require identical guest memory configurations, i.e. the
source VM may have access to memory that the destination VM does not.
Enforcing equivalent memory configurations is infeasible, as it would
require a *deep* comparison of memslots, e.g. to verify that not only are
the memslot identical, but what the memslots point at is also identical.

Fixes: b56639318bb2 ("KVM: SEV: Add support for SEV intra host migration")
Cc: stable@vger.kernel.org
Reported-by: Stefan Teodorescu <fane@google.com>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-ID: <20260923163721.1584779-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 223 insertions, 0 deletions

diff --git a/security/apparmor/capability.c b/security/apparmor/capability.c
new file mode 100644
index 000000000..2f3c6c843
--- /dev/null
+++ b/security/apparmor/capability.c
@@ -0,0 +1,223 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * AppArmor security module
+ *
+ * This file contains AppArmor capability mediation functions
+ *
+ * Copyright (C) 1998-2008 Novell/SUSE
+ * Copyright 2009-2010 Canonical Ltd.
+ */
+
+#include <linux/capability.h>
+#include <linux/errno.h>
+#include <linux/gfp.h>
+#include <linux/security.h>
+#include <linux/timekeeping.h>
+
+#include "include/apparmor.h"
+#include "include/capability.h"
+#include "include/cred.h"
+#include "include/policy.h"
+#include "include/audit.h"
+
+/*
+ * Table of capability names: we generate it from capabilities.h.
+ */
+#include "capability_names.h"
+
+struct aa_sfs_entry aa_sfs_entry_caps[] = {
+	AA_SFS_FILE_STRING("mask", AA_SFS_CAPS_MASK),
+	AA_SFS_FILE_BOOLEAN("extended", 1),
+	{ }
+};
+
+struct audit_cache {
+	const struct cred *ad_subj_cred;
+	/* Capabilities go from 0 to CAP_LAST_CAP */
+	u64 ktime_ns_expiration[CAP_LAST_CAP+1];
+};
+
+static DEFINE_PER_CPU(struct audit_cache, audit_cache);
+
+/**
+ * audit_cb - call back for capability components of audit struct
+ * @ab: audit buffer   (NOT NULL)
+ * @va: audit struct to audit data from  (NOT NULL)
+ */
+static void audit_cb(struct audit_buffer *ab, void *va)
+{
+	struct common_audit_data *sa = va;
+
+	audit_log_format(ab, " capname=");
+	audit_log_untrustedstring(ab, capability_names[sa->u.cap]);
+}
+
+/**
+ * audit_caps - audit a capability
+ * @ad: audit data
+ * @profile: profile being tested for confinement (NOT NULL)
+ * @cap: capability tested
+ * @error: error code returned by test
+ *
+ * Do auditing of capability and handle, audit/complain/kill modes switching
+ * and duplicate message elimination.
+ *
+ * Returns: 0 or ad->error on success,  error code on failure
+ */
+static int audit_caps(struct apparmor_audit_data *ad, struct aa_profile *profile,
+		      int cap, int error)
+{
+	const u64 AUDIT_CACHE_TIMEOUT_NS = 1000*1000*1000; /* 1 second */
+
+	struct aa_ruleset *rules = profile->label.rules[0];
+	struct audit_cache *ent;
+	int type = AUDIT_APPARMOR_AUTO;
+
+	ad->error = error;
+
+	if (likely(!error)) {
+		/* test if auditing is being forced */
+		if (likely((AUDIT_MODE(profile) != AUDIT_ALL) &&
+			   !cap_raised(rules->caps.audit, cap)))
+			return 0;
+		type = AUDIT_APPARMOR_AUDIT;
+	} else if (KILL_MODE(profile) ||
+		   cap_raised(rules->caps.kill, cap)) {
+		type = AUDIT_APPARMOR_KILL;
+	} else if (cap_raised(rules->caps.quiet, cap) &&
+		   AUDIT_MODE(profile) != AUDIT_NOQUIET &&
+		   AUDIT_MODE(profile) != AUDIT_ALL) {
+		/* quiet auditing */
+		return error;
+	}
+
+	/* Do simple duplicate message elimination */
+	ent = &get_cpu_var(audit_cache);
+	/* If the capability was never raised the timestamp check would also catch that */
+	if (ad->subj_cred == ent->ad_subj_cred && ktime_get_ns() <= ent->ktime_ns_expiration[cap]) {
+		put_cpu_var(audit_cache);
+		if (COMPLAIN_MODE(profile))
+			return complain_error(error);
+		return error;
+	} else {
+		put_cred(ent->ad_subj_cred);
+		ent->ad_subj_cred = get_cred(ad->subj_cred);
+		ent->ktime_ns_expiration[cap] = ktime_get_ns() + AUDIT_CACHE_TIMEOUT_NS;
+	}
+	put_cpu_var(audit_cache);
+
+	return aa_audit(type, profile, ad, audit_cb);
+}
+
+/**
+ * profile_capable - test if profile allows use of capability @cap
+ * @profile: profile being enforced    (NOT NULL, NOT unconfined)
+ * @cap: capability to test if allowed
+ * @opts: CAP_OPT_NOAUDIT bit determines whether audit record is generated
+ * @ad: audit data (NOT NULL)
+ *
+ * Returns: 0 if allowed else -EPERM
+ */
+static int profile_capable(struct aa_profile *profile, int cap,
+			   unsigned int opts, struct apparmor_audit_data *ad)
+{
+	struct aa_ruleset *rules = profile->label.rules[0];
+	aa_state_t state;
+	int error;
+
+	state = RULE_MEDIATES(rules, ad->class);
+	if (state) {
+		struct aa_perms perms = { };
+		u32 request;
+
+		/* caps broken into 256 x 32 bit permission chunks */
+		state = aa_dfa_next(rules->policy->dfa, state, cap >> 5);
+		request = 1 << (cap & 0x1f);
+		perms = *aa_lookup_perms(rules->policy, state);
+		aa_apply_modes_to_perms(profile, &perms);
+
+		if (opts & CAP_OPT_NOAUDIT) {
+			if (perms.complain & request)
+				ad->info = "optional: no audit";
+			else
+				ad = NULL;
+		}
+		return aa_check_perms(profile, &perms, request, ad,
+				      audit_cb);
+	}
+
+	/* fallback to old caps mediation that doesn't support conditionals */
+	if (cap_raised(rules->caps.allow, cap) &&
+	    !cap_raised(rules->caps.denied, cap))
+		error = 0;
+	else
+		error = -EPERM;
+
+	if (opts & CAP_OPT_NOAUDIT) {
+		if (!COMPLAIN_MODE(profile))
+			return error;
+		/* audit the cap request in complain mode but note that it
+		 * should be optional.
+		 */
+		ad->info = "optional: no audit";
+	}
+
+	return audit_caps(ad, profile, cap, error);
+}
+
+/**
+ * aa_capable - test permission to use capability
+ * @subj_cred: cred we are testing capability against
+ * @label: label being tested for capability (NOT NULL)
+ * @cap: capability to be tested
+ * @opts: CAP_OPT_NOAUDIT bit determines whether audit record is generated
+ *
+ * Look up capability in profile capability set.
+ *
+ * Returns: 0 on success, or else an error code.
+ */
+int aa_capable(const struct cred *subj_cred, struct aa_label *label,
+	       int cap, unsigned int opts)
+{
+	struct aa_profile *profile;
+	int error = 0;
+	DEFINE_AUDIT_DATA(ad, LSM_AUDIT_DATA_CAP, AA_CLASS_CAP, OP_CAPABLE);
+
+	ad.subj_cred = subj_cred;
+	ad.common.u.cap = cap;
+	error = fn_for_each_confined(label, profile,
+			profile_capable(profile, cap, opts, &ad));
+
+	return error;
+}
+
+kernel_cap_t aa_profile_capget(const struct aa_profile *profile)
+{
+	struct aa_ruleset *rules = profile->label.rules[0];
+	aa_state_t state;
+
+	state = RULE_MEDIATES(rules, AA_CLASS_CAP);
+	if (state) {
+		kernel_cap_t caps = CAP_EMPTY_SET;
+		int i;
+
+		/* caps broken into up to 256, 32 bit permission chunks */
+		for (i = 0; i < (CAP_LAST_CAP >> 5); i++) {
+			struct aa_perms perms = { };
+			aa_state_t tmp;
+
+			tmp = aa_dfa_next(rules->policy->dfa, state, i);
+			perms = *aa_lookup_perms(rules->policy, tmp);
+			aa_apply_modes_to_perms(profile, &perms);
+			caps.val |= ((u64)(perms.allow)) << (i * 5);
+			caps.val |= ((u64)(perms.complain)) << (i * 5);
+		}
+		return caps;
+	}
+
+	/* fallback to old caps */
+	if (COMPLAIN_MODE(profile))
+		return CAP_FULL_SET;
+
+	return rules->caps.allow;
+}