new file mode 100644
@@ -0,0 +1,388 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/*
+ * Copyright (C) 2022 Enjellic Systems Development, LLC
+ * Author: Dr. Greg Wettstein <greg@enjellic.com>
+ *
+ * Implements updates to an external modeling engine.
+ */
+
+#include <linux/seq_file.h>
+
+#include "tsem.h"
+
+enum export_events_type {
+ AGGREGATE_EVENT = 1,
+ EXPORT_EVENT,
+ LOG_EVENT
+};
+
+struct action_description {
+ enum tsem_event_type type;
+ enum tsem_action_type action;
+ char comm[TASK_COMM_LEN];
+};
+
+struct export_event {
+ struct list_head list;
+ enum export_events_type type;
+ union {
+ u8 *aggregate[WP256_DIGEST_SIZE];
+ struct tsem_event *ep;
+ struct action_description action;
+ } u;
+};
+
+static const char * const tsem_actions[TSEM_ACTION_CNT] = {
+ "LOG",
+ "DENY"
+};
+
+static void trigger_event(struct tsem_TMA_context *ctx)
+{
+ ctx->external->have_event = true;
+ wake_up_interruptible(&ctx->external->wq);
+}
+
+static void show_event_coe(struct tsem_event *ep, struct seq_file *page)
+{
+ seq_printf(page, "export pid{%u} ", ep->pid);
+
+ seq_printf(page, "event{process=%s, filename=%s, ", ep->comm,
+ ep->pathname == NULL ? "none" : ep->pathname);
+ seq_printf(page, "type=%s, task_id=%*phN} ", tsem_names[ep->event],
+ WP256_DIGEST_SIZE, ep->task_id);
+
+ seq_printf(page, "COE{uid=%d, euid=%d, suid=%d, gid=%d, ", ep->COE.uid,
+ ep->COE.euid, ep->COE.suid, ep->COE.gid);
+ seq_printf(page, "egid=%d, sgid=%d, fsuid=%d, fsgid=%d, ",
+ ep->COE.egid, ep->COE.sgid, ep->COE.fsuid, ep->COE.fsgid);
+ seq_printf(page, "cap=0x%llx} ", ep->COE.capability.value);
+}
+
+static void show_file(struct tsem_event *ep, struct seq_file *page)
+{
+ seq_printf(page, "file{flags=%u, uid=%d, gid=%d, mode=0%o, ",
+ ep->file.flags, ep->file.uid, ep->file.gid, ep->file.mode);
+ seq_printf(page, "name_length=%u, name=%*phN, s_magic=0x%0x, ",
+ ep->file.name_length, WP256_DIGEST_SIZE, ep->file.name,
+ ep->file.s_magic);
+ seq_printf(page, "s_id=%s, s_uuid=%*phN, digest=%*phN}\n",
+ ep->file.s_id, (int) sizeof(ep->file.s_uuid),
+ ep->file.s_uuid, WP256_DIGEST_SIZE, ep->file.digest);
+}
+
+static void show_mmap_file(struct tsem_event *ep, struct seq_file *page)
+{
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{type=%u, reqprot=%u, ", tsem_names[ep->event],
+ ep->CELL.mmap_file.anonymous, ep->CELL.mmap_file.reqprot);
+ seq_printf(page, "prot=%u, flags=%u}", ep->CELL.mmap_file.prot,
+ ep->CELL.mmap_file.flags);
+
+ if (!ep->CELL.mmap_file.anonymous) {
+ seq_puts(page, " ");
+ show_file(ep, page);
+ } else
+ seq_puts(page, "\n");
+}
+
+static void show_ipv4_socket(struct tsem_event *ep, struct seq_file *page)
+{
+ struct sockaddr_in *ipv4 = &ep->CELL.socket_connect.u.ipv4;
+
+ show_event_coe(ep, page);
+ seq_printf(page, "%s{family=%u, port=%u, addr=%u}\n",
+ tsem_names[ep->event], ipv4->sin_family, ipv4->sin_port,
+ ipv4->sin_addr.s_addr);
+}
+
+static void show_ipv6_socket(struct tsem_event *ep, struct seq_file *page)
+{
+ struct sockaddr_in6 *ipv6 = &ep->CELL.socket_connect.u.ipv6;
+
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{family=%u, port=%u, flow=%u, ",
+ tsem_names[ep->event], ipv6->sin6_family,
+ ipv6->sin6_port, ipv6->sin6_flowinfo);
+ seq_printf(page, "scope=%u, addr=%*phN}\n", ipv6->sin6_scope_id,
+ (int) sizeof(ipv6->sin6_addr.in6_u.u6_addr8),
+ ipv6->sin6_addr.in6_u.u6_addr8);
+}
+
+static void show_socket(struct tsem_event *ep, struct seq_file *page)
+{
+ struct tsem_socket_connect_args *scp = &ep->CELL.socket_connect;
+
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{family=%u, addr=%*phN}\n", tsem_names[ep->event],
+ scp->family, WP256_DIGEST_SIZE, scp->u.mapping);
+}
+
+static void show_socket_create(struct tsem_event *ep, struct seq_file *page)
+{
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{family=%u, type=%u, ", tsem_names[ep->event],
+ ep->CELL.socket_create.family, ep->CELL.socket_create.type);
+ seq_printf(page, "protocol=%u, kern=%u}\n",
+ ep->CELL.socket_create.protocol,
+ ep->CELL.socket_create.kern);
+}
+
+static void show_socket_accept(struct tsem_event *ep, struct seq_file *page)
+{
+ u8 *p;
+ int size;
+ struct tsem_socket_accept_args *sap = &ep->CELL.socket_accept;
+
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{family=%u, type=%u, port=%u, addr=",
+ tsem_names[ep->event], sap->family, sap->type, sap->port);
+
+ if (sap->family == AF_INET) {
+ seq_printf(page, "%u}\n", sap->ipv4);
+ return;
+ }
+
+ if (sap->family == AF_INET6) {
+ p = sap->ipv6.in6_u.u6_addr8;
+ size = sizeof(sap->ipv6.in6_u.u6_addr8);
+ } else {
+ p = sap->tsip->digest;
+ size = sizeof(sap->tsip->digest);
+ }
+ seq_printf(page, "%*phN}\n", size, p);
+}
+
+static void show_socket_events(struct tsem_event *ep, struct seq_file *page)
+{
+ switch (ep->event) {
+ case TSEM_SOCKET_CREATE:
+ show_socket_create(ep, page);
+ break;
+
+ case TSEM_SOCKET_CONNECT:
+ case TSEM_SOCKET_BIND:
+ switch (ep->CELL.socket_connect.family) {
+ case AF_INET:
+ show_ipv4_socket(ep, page);
+ break;
+ case AF_INET6:
+ show_ipv6_socket(ep, page);
+ break;
+ default:
+ show_socket(ep, page);
+ break;
+ }
+ break;
+
+ case TSEM_SOCKET_ACCEPT:
+ show_socket_accept(ep, page);
+ break;
+
+ default:
+ break;
+ }
+}
+
+static void show_task_kill(struct tsem_event *ep, struct seq_file *page)
+{
+ struct tsem_task_kill_args *args = &ep->CELL.task_kill;
+
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{cross=%u, signal=%u, target=%*phN}\n",
+ tsem_names[ep->event], args->cross_model,
+ args->signal, WP256_DIGEST_SIZE, args->target);
+}
+
+static void show_event_generic(struct tsem_event *ep, struct seq_file *page)
+{
+ show_event_coe(ep, page);
+
+ seq_printf(page, "%s{type=%s}\n", tsem_names[ep->event],
+ tsem_names[ep->CELL.event_type]);
+}
+
+int tsem_export_show(struct seq_file *page)
+{
+ ssize_t retn = -ENODATA;
+ struct export_event *mp;
+ struct tsem_event *ep;
+ struct tsem_TMA_context *ctx = tsem_context(current);
+
+ if (!ctx->id)
+ return -EPERM;
+
+ mutex_lock(&ctx->external->measurement_mutex);
+ if (list_empty(&ctx->external->measurement_list))
+ goto done;
+ mp = list_first_entry(&ctx->external->measurement_list,
+ struct export_event, list);
+
+ switch (mp->type) {
+ case AGGREGATE_EVENT:
+ seq_printf(page, "aggregate %*phN\n", WP256_DIGEST_SIZE,
+ mp->u.aggregate);
+ break;
+
+ case EXPORT_EVENT:
+ ep = mp->u.ep;
+ switch (ep->event) {
+ case TSEM_FILE_OPEN:
+ show_event_coe(ep, page);
+ show_file(ep, page);
+ break;
+
+ case TSEM_MMAP_FILE:
+ show_mmap_file(ep, page);
+ break;
+
+ case TSEM_SOCKET_CREATE:
+ case TSEM_SOCKET_CONNECT:
+ case TSEM_SOCKET_BIND:
+ case TSEM_SOCKET_ACCEPT:
+ show_socket_events(ep, page);
+ break;
+
+ case TSEM_TASK_KILL:
+ show_task_kill(ep, page);
+ break;
+
+ case TSEM_GENERIC_EVENT:
+ show_event_generic(ep, page);
+ break;
+
+ default:
+ break;
+ }
+ tsem_event_put(ep);
+ break;
+
+ case LOG_EVENT:
+ seq_printf(page, "log process{%s} event{%s} action{%s}\n",
+ mp->u.action.comm, tsem_names[mp->u.action.type],
+ tsem_actions[mp->u.action.action]);
+ break;
+ }
+
+ list_del(&mp->list);
+ kfree(mp);
+ retn = 0;
+
+ done:
+ mutex_unlock(&ctx->external->measurement_mutex);
+ return retn;
+}
+
+int tsem_export_event(struct tsem_event *ep)
+{
+ int retn = 0;
+ struct tsem_task *task = tsem_task(current);
+ struct tsem_TMA_context *ctx = task->context;
+ struct export_event *mp;
+
+ if (!ctx->external)
+ return 0;
+
+ mp = kzalloc(sizeof(struct export_event), GFP_KERNEL);
+ if (!mp) {
+ retn = -ENOMEM;
+ goto done;
+ }
+ mp->type = EXPORT_EVENT;
+ mp->u.ep = ep;
+ tsem_event_get(ep);
+
+ mutex_lock(&ctx->external->measurement_mutex);
+ list_add_tail(&mp->list, &ctx->external->measurement_list);
+ mutex_unlock(&ctx->external->measurement_mutex);
+
+ task->trust_status |= TSEM_TASK_TRUST_PENDING;
+ trigger_event(ctx);
+
+ while (task->trust_status & TSEM_TASK_TRUST_PENDING) {
+ set_current_state(TASK_INTERRUPTIBLE);
+ schedule();
+ if (signal_pending(current)) {
+ if (sigismember(¤t->pending.signal, SIGKILL) ||
+ sigismember(¤t->signal->shared_pending.signal,
+ SIGKILL))
+ task->trust_status = TSEM_TASK_UNTRUSTED;
+ }
+ }
+
+ done:
+ return retn;
+}
+
+/**
+ * tsem_export_action() - Exports the action taken to a security violation.
+ * @event: The TSEM event type number for which the log event is being
+ * generated.
+ *
+ * This function queues for export a description of an event that
+ * was being disciplined.
+ *
+ * Return: This function returns 0 if the export was successful or
+ * an error value if it was not.
+ */
+int tsem_export_action(enum tsem_event_type event)
+{
+ struct tsem_TMA_context *ctx = tsem_context(current);
+ struct export_event *exp;
+
+ exp = kzalloc(sizeof(struct export_event), GFP_KERNEL);
+ if (!exp)
+ return -ENOMEM;
+
+ exp->type = LOG_EVENT;
+ exp->u.action.type = event;
+ exp->u.action.action = ctx->actions[event];
+ strcpy(exp->u.action.comm, current->comm);
+
+ mutex_lock(&ctx->external->measurement_mutex);
+ list_add_tail(&exp->list, &ctx->external->measurement_list);
+ mutex_unlock(&ctx->external->measurement_mutex);
+
+ trigger_event(ctx);
+
+ return 0;
+}
+
+/**
+ * tsem_export_aggregate() - Exports the hardware aggregate value.
+ *
+ * This function exports the hardware aggregate measurement for
+ * the platform on which the TSEM LSM is being run on.
+ *
+ * Return: This function returns a value of 0 if the export was
+ * successful or a non-zero return value if the export was
+ * not successful.
+ */
+int tsem_export_aggregate(void)
+{
+ struct tsem_TMA_context *ctx = tsem_context(current);
+ struct export_event *exp;
+
+ exp = kzalloc(sizeof(struct export_event), GFP_KERNEL);
+ if (!exp)
+ return -ENOMEM;
+
+ exp->type = AGGREGATE_EVENT;
+ memcpy(exp->u.aggregate, tsem_trust_aggregate(),
+ sizeof(exp->u.aggregate));
+
+ mutex_lock(&ctx->external->measurement_mutex);
+ list_add_tail(&exp->list, &ctx->external->measurement_list);
+ mutex_unlock(&ctx->external->measurement_mutex);
+
+ trigger_event(ctx);
+
+ return 0;
+}
The functionality for surfacing security model events to an external modeling domain is implemented in the export.c file. ASCII descriptions of the events are presented to a userspace trust orchestrator through the following pseudo-files: /sys/fs/tsem/ExternalTMA/N Where N is replaced with security model domain identifier. The following event types are exported: AGGREGATE_EVENT EXPORT_EVENT LOG_EVENT The AGGREGATE_EVENT is used to inject the hardware platform aggregate that was computed over TPM Platform Configuration Registers 0 through 7 at the time the LSM was initialized. In TSEM modeling this is the first security state point committed to a model. An EXPORT_EVENT is used to surface the description of either an explicitly or generically modeled security state event for injection into a security model run by an external orchestrator and its associated Trusted Modeling Agent (TMA). A LOG_EVENT is used to export descriptions of security events that are invoked by untrusted processes. The modeling and logging by external orchestrators allow the implementation of out-of-band notifications of security forensics events that occur. The /sys/fs/tsem/ExternalTMA/N pseudo-files implement a pollable interface that the trust orchestrators can use to wait on events. After placing the event description into the device queue the process is placed in an interruptible sleep state. After the TMA completes modeling of the event, the trust orchestrator is responsible for using the tsemfs control plane to wake the process that exported the event and set its status to either trusted or untrusted. Signed-off-by: Greg Wettstein <greg@enjellic.com> --- security/tsem/export.c | 388 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 388 insertions(+) create mode 100644 security/tsem/export.c