]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/kern/coalition.c
xnu-3247.1.106.tar.gz
[apple/xnu.git] / osfmk / kern / coalition.c
index 95babd63280f2211f645ba2ac4c114e4e5889126..a5cf30c939af6506fb6b5be5cec707330c0c83bb 100644 (file)
 
 #include <kern/coalition.h>
 #include <kern/host.h>
-#include <kern/ledger.h>
 #include <kern/kalloc.h>
+#include <kern/ledger.h>
 #include <kern/mach_param.h> /* for TASK_CHUNK */
 #include <kern/task.h>
 #include <kern/zalloc.h>
-#include <kern/sfi.h>
 
 #include <libkern/OSAtomic.h>
 
 
 #include <sys/errno.h>
 
+/*
+ * BSD interface functions
+ */
+int coalitions_get_list(int type, struct procinfo_coalinfo *coal_list, int list_sz);
+boolean_t coalition_is_leader(task_t task, int coal_type, coalition_t *coal);
+int coalition_get_task_count(coalition_t coal);
+uint64_t coalition_get_page_count(coalition_t coal, int *ntasks);
+int coalition_get_pid_list(coalition_t coal, uint32_t rolemask, int sort_order,
+                                 int *pid_list, int list_sz);
+
 /* defined in task.c */
 extern ledger_template_t task_ledger_template;
 
@@ -67,20 +76,103 @@ lck_grp_attr_t coalitions_lck_grp_attr;
 decl_lck_mtx_data(static,coalitions_list_lock);
 static uint64_t coalition_count;
 static uint64_t coalition_next_id = 1;
-static queue_head_t coalitions;
+static queue_head_t coalitions_q;
 
-coalition_t default_coalition;
+coalition_t init_coalition[COALITION_NUM_TYPES];
 
 zone_t coalition_zone;
 
-struct coalition {
-       uint64_t id;                    /* monotonically increasing */
+static const char *coal_type_str(int type)
+{
+       switch(type) {
+       case COALITION_TYPE_RESOURCE:
+               return "RESOURCE";
+       case COALITION_TYPE_JETSAM:
+               return "JETSAM";
+       default:
+               return "<unknown>";
+       }
+}
+
+struct coalition_type {
+       int type;
+       int has_default;
+       /*
+        * init
+        * pre-condition: coalition just allocated (unlocked), unreferenced,
+        *                type field set
+        */
+       kern_return_t (*init)(coalition_t coal, boolean_t privileged);
+
+       /*
+        * dealloc
+        * pre-condition: coalition unlocked
+        * pre-condition: coalition refcount=0, active_count=0,
+        *                termrequested=1, terminated=1, reaped=1
+        */
+       void          (*dealloc)(coalition_t coal);
+
+       /*
+        * adopt_task
+        * pre-condition: coalition locked
+        * pre-condition: coalition !repead and !terminated
+        */
+       kern_return_t (*adopt_task)(coalition_t coal, task_t task);
+
+       /*
+        * remove_task
+        * pre-condition: coalition locked
+        * pre-condition: task has been removed from coalition's task list
+        */
+       kern_return_t (*remove_task)(coalition_t coal, task_t task);
+
+       /*
+        * set_taskrole
+        * pre-condition: coalition locked
+        * pre-condition: task added to coalition's task list,
+        *                active_count >= 1 (at least the given task is active)
+        */
+       kern_return_t (*set_taskrole)(coalition_t coal, task_t task, int role);
+
+       /*
+        * get_taskrole
+        * pre-condition: coalition locked
+        * pre-condition: task added to coalition's task list,
+        *                active_count >= 1 (at least the given task is active)
+        */
+       int (*get_taskrole)(coalition_t coal, task_t task);
+
+       /*
+        * iterate_tasks
+        * pre-condition: coalition locked
+        */
+       void (*iterate_tasks)(coalition_t coal, void *ctx, void (*callback)(coalition_t, void *, task_t));
+};
+
+/*
+ * COALITION_TYPE_RESOURCE
+ */
 
+static kern_return_t i_coal_resource_init(coalition_t coal, boolean_t privileged);
+static void          i_coal_resource_dealloc(coalition_t coal);
+static kern_return_t i_coal_resource_adopt_task(coalition_t coal, task_t task);
+static kern_return_t i_coal_resource_remove_task(coalition_t coal, task_t task);
+static kern_return_t i_coal_resource_set_taskrole(coalition_t coal,
+                                                task_t task, int role);
+static int           i_coal_resource_get_taskrole(coalition_t coal, task_t task);
+static void          i_coal_resource_iterate_tasks(coalition_t coal, void *ctx,
+                                                  void (*callback)(coalition_t, void *, task_t));
+
+struct i_resource_coalition {
        ledger_t ledger;
        uint64_t bytesread;
        uint64_t byteswritten;
        uint64_t gpu_time;
 
+       uint64_t task_count;      /* tasks that have started in this coalition */
+       uint64_t dead_task_count; /* tasks that have exited in this coalition;
+                                    subtract from task_count to get count
+                                    of "active" tasks */
        /*
         * Count the length of time this coalition had at least one active task.
         * This can be a 'denominator' to turn e.g. cpu_time to %cpu.
@@ -88,33 +180,103 @@ struct coalition {
        uint64_t last_became_nonempty_time;
        uint64_t time_nonempty;
 
-       uint64_t task_count;            /* Count of tasks that have started in this coalition */
-       uint64_t dead_task_count;       /* Count of tasks that have exited in this coalition; subtract from task_count to get count of "active" */
-       queue_head_t tasks;             /* List of active tasks in the coalition */
-
-       queue_chain_t coalitions;       /* global list of coalitions */
+       queue_head_t tasks;         /* List of active tasks in the coalition */
+};
 
-       decl_lck_mtx_data(,lock)        /* Coalition lock. */
+/*
+ * COALITION_TYPE_JETSAM
+ */
 
-       uint32_t ref_count;             /* Number of references to the memory containing this struct */
-       uint32_t active_count;          /* Number of members of (tasks in) the coalition, plus vouchers referring to the coalition */
+static kern_return_t i_coal_jetsam_init(coalition_t coal, boolean_t privileged);
+static void          i_coal_jetsam_dealloc(coalition_t coal);
+static kern_return_t i_coal_jetsam_adopt_task(coalition_t coal, task_t task);
+static kern_return_t i_coal_jetsam_remove_task(coalition_t coal, task_t task);
+static kern_return_t i_coal_jetsam_set_taskrole(coalition_t coal,
+                                              task_t task, int role);
+static int           i_coal_jetsam_get_taskrole(coalition_t coal, task_t task);
+static void          i_coal_jetsam_iterate_tasks(coalition_t coal, void *ctx,
+                                                void (*callback)(coalition_t, void *, task_t));
+
+struct i_jetsam_coalition {
+       task_t       leader;
+       queue_head_t extensions;
+       queue_head_t services;
+       queue_head_t other;
+};
 
-       unsigned int privileged : 1;    /* Members of this coalition may create and manage coalitions and may posix_spawn processes into selected coalitions */
 
+/*
+ * main coalition structure
+ */
+struct coalition {
+       uint64_t id;                /* monotonically increasing */
+       uint32_t type;
+       uint32_t ref_count;         /* Number of references to the memory containing this struct */
+       uint32_t active_count;      /* Number of members of (tasks in) the
+                                      coalition, plus vouchers referring
+                                      to the coalition */
+       uint32_t focal_task_count;   /* Number of TASK_FOREGROUND_APPLICATION tasks in the coalition */
+       uint32_t nonfocal_task_count; /* Number of TASK_BACKGROUND_APPLICATION tasks in the coalition */
+
+       /* coalition flags */
+       uint32_t privileged : 1;    /* Members of this coalition may create
+                                      and manage coalitions and may posix_spawn
+                                      processes into selected coalitions */
        /* ast? */
-
        /* voucher */
+       uint32_t termrequested : 1; /* launchd has requested termination when coalition becomes empty */
+       uint32_t terminated : 1;    /* coalition became empty and spawns are now forbidden */
+       uint32_t reaped : 1;        /* reaped, invisible to userspace, but waiting for ref_count to go to zero */
+       uint32_t notified : 1;      /* no-more-processes notification was sent via special port */
+#if defined(DEVELOPMENT) || defined(DEBUG)
+       uint32_t should_notify : 1; /* should this coalition send notifications (default: yes) */
+#endif
 
-       /* state of the coalition */
-       unsigned int termrequested : 1;         /* launchd has requested termination when coalition becomes empty */
-       unsigned int terminated : 1;            /* coalition became empty and spawns are now forbidden */
-       unsigned int reaped : 1;                /* reaped, invisible to userspace, but waiting for ref_count to go to zero */
-       unsigned int notified : 1;              /* no-more-processes notification was sent via special port */
+       queue_chain_t coalitions;   /* global list of coalitions */
 
-       uint32_t focal_tasks_count;     /* count of TASK_FOREGROUND_APPLICATION tasks in the coalition */
-       uint32_t non_focal_tasks_count; /* count of TASK_BACKGROUND_APPLICATION tasks in the coalition */
+       decl_lck_mtx_data(,lock)    /* Coalition lock. */
+
+       /* put coalition type-specific structures here */
+       union {
+               struct i_resource_coalition  r;
+               struct i_jetsam_coalition    j;
+       };
+};
+
+/*
+ * register different coalition types:
+ * these must be kept in the order specified in coalition.h
+ */
+static const struct coalition_type
+s_coalition_types[COALITION_NUM_TYPES] = {
+       {
+               COALITION_TYPE_RESOURCE,
+               1,
+               i_coal_resource_init,
+               i_coal_resource_dealloc,
+               i_coal_resource_adopt_task,
+               i_coal_resource_remove_task,
+               i_coal_resource_set_taskrole,
+               i_coal_resource_get_taskrole,
+               i_coal_resource_iterate_tasks,
+       },
+       {
+               COALITION_TYPE_JETSAM,
+               1,
+               i_coal_jetsam_init,
+               i_coal_jetsam_dealloc,
+               i_coal_jetsam_adopt_task,
+               i_coal_jetsam_remove_task,
+               i_coal_jetsam_set_taskrole,
+               i_coal_jetsam_get_taskrole,
+               i_coal_jetsam_iterate_tasks,
+       },
 };
 
+#define coal_call(coal, func, ...) \
+       (s_coalition_types[(coal)->type].func)(coal, ## __VA_ARGS__)
+
+
 #define coalition_lock(c) do{ lck_mtx_lock(&c->lock); }while(0)
 #define coalition_unlock(c) do{ lck_mtx_unlock(&c->lock); }while(0)
 
@@ -133,26 +295,137 @@ coalition_notify_user(uint64_t id, uint32_t flags)
 }
 
 /*
- * coalition_find_by_id_internal
- * Returns: Coalition object with specified id, NOT referenced.
- *          If not found, returns COALITION_NULL.
- * Condition: coalitions_list_lock must be LOCKED.
+ *
+ * COALITION_TYPE_RESOURCE
+ *
  */
-static coalition_t
-coalition_find_by_id_internal(uint64_t coal_id)
+static kern_return_t
+i_coal_resource_init(coalition_t coal, boolean_t privileged)
 {
-       if (coal_id == 0) {
-               return COALITION_NULL;
+       (void)privileged;
+       assert(coal && coal->type == COALITION_TYPE_RESOURCE);
+       coal->r.ledger = ledger_instantiate(task_ledger_template,
+                                           LEDGER_CREATE_ACTIVE_ENTRIES);
+       if (coal->r.ledger == NULL)
+               return KERN_RESOURCE_SHORTAGE;
+
+       queue_init(&coal->r.tasks);
+
+       return KERN_SUCCESS;
+}
+
+static void
+i_coal_resource_dealloc(coalition_t coal)
+{
+       assert(coal && coal->type == COALITION_TYPE_RESOURCE);
+       ledger_dereference(coal->r.ledger);
+}
+
+static kern_return_t
+i_coal_resource_adopt_task(coalition_t coal, task_t task)
+{
+       struct i_resource_coalition *cr;
+
+       assert(coal && coal->type == COALITION_TYPE_RESOURCE);
+       assert(queue_empty(&task->task_coalition[COALITION_TYPE_RESOURCE]));
+
+       cr = &coal->r;
+       cr->task_count++;
+
+       if (cr->task_count < cr->dead_task_count) {
+               panic("%s: coalition %p id:%llu type:%s task_count(%llu) < dead_task_count(%llu)",
+                     __func__, coal, coal->id, coal_type_str(coal->type),
+                     cr->task_count, cr->dead_task_count);
        }
 
-       lck_mtx_assert(&coalitions_list_lock, LCK_MTX_ASSERT_OWNED);
-       coalition_t coal;
-       queue_iterate(&coalitions, coal, coalition_t, coalitions) {
-               if (coal->id == coal_id) {
-                       return coal;
-               }
+       /* If moving from 0->1 active tasks */
+       if (cr->task_count - cr->dead_task_count == 1) {
+               cr->last_became_nonempty_time = mach_absolute_time();
        }
-       return COALITION_NULL;
+
+       /* put the task on the coalition's list of tasks */
+       enqueue_tail(&cr->tasks, &task->task_coalition[COALITION_TYPE_RESOURCE]);
+
+       coal_dbg("Added PID:%d to id:%llu, task_count:%llu, dead_count:%llu, nonempty_time:%llu",
+                task_pid(task), coal->id, cr->task_count, cr->dead_task_count,
+                cr->last_became_nonempty_time);
+
+       return KERN_SUCCESS;
+}
+
+static kern_return_t
+i_coal_resource_remove_task(coalition_t coal, task_t task)
+{
+       struct i_resource_coalition *cr;
+
+       assert(coal && coal->type == COALITION_TYPE_RESOURCE);
+       assert(task->coalition[COALITION_TYPE_RESOURCE] == coal);
+       assert(!queue_empty(&task->task_coalition[COALITION_TYPE_RESOURCE]));
+
+       /*
+        * handle resource coalition accounting rollup for dead tasks
+        */
+       cr = &coal->r;
+
+       cr->dead_task_count++;
+
+       if (cr->task_count < cr->dead_task_count) {
+               panic("%s: coalition %p id:%llu type:%s task_count(%llu) < dead_task_count(%llu)",
+                     __func__, coal, coal->id, coal_type_str(coal->type), cr->task_count, cr->dead_task_count);
+       }
+
+       /* If moving from 1->0 active tasks */
+       if (cr->task_count - cr->dead_task_count == 0) {
+               uint64_t last_time_nonempty = mach_absolute_time() - cr->last_became_nonempty_time;
+               cr->last_became_nonempty_time = 0;
+               cr->time_nonempty += last_time_nonempty;
+       }
+
+       ledger_rollup(cr->ledger, task->ledger);
+       cr->bytesread += task->task_io_stats->disk_reads.size;
+       cr->byteswritten += task->task_io_stats->total_io.size - task->task_io_stats->disk_reads.size;
+       cr->gpu_time += task_gpu_utilisation(task);
+
+       /* remove the task from the coalition's list */
+       remqueue(&task->task_coalition[COALITION_TYPE_RESOURCE]);
+       queue_chain_init(task->task_coalition[COALITION_TYPE_RESOURCE]);
+
+       coal_dbg("removed PID:%d from id:%llu, task_count:%llu, dead_count:%llu",
+                task_pid(task), coal->id, cr->task_count, cr->dead_task_count);
+
+       return KERN_SUCCESS;
+}
+
+static kern_return_t
+i_coal_resource_set_taskrole(__unused coalition_t coal,
+                           __unused task_t task, __unused int role)
+{
+       return KERN_SUCCESS;
+}
+
+static int
+i_coal_resource_get_taskrole(__unused coalition_t coal, __unused task_t task)
+{
+       task_t t;
+
+       assert(coal && coal->type == COALITION_TYPE_RESOURCE);
+
+       qe_foreach_element(t, &coal->r.tasks, task_coalition[COALITION_TYPE_RESOURCE]) {
+               if (t == task)
+                       return COALITION_TASKROLE_UNDEF;
+       }
+
+       return -1;
+}
+
+static void
+i_coal_resource_iterate_tasks(coalition_t coal, void *ctx, void (*callback)(coalition_t, void *, task_t))
+{
+       task_t t;
+       assert(coal && coal->type == COALITION_TYPE_RESOURCE);
+
+       qe_foreach_element(t, &coal->r.tasks, task_coalition[COALITION_TYPE_RESOURCE])
+               callback(coal, ctx, t);
 }
 
 kern_return_t
@@ -161,10 +434,12 @@ coalition_resource_usage_internal(coalition_t coal, struct coalition_resource_us
        kern_return_t kr;
        ledger_amount_t credit, debit;
 
+       if (coal->type != COALITION_TYPE_RESOURCE)
+               return KERN_INVALID_ARGUMENT;
+
        ledger_t sum_ledger = ledger_instantiate(task_ledger_template, LEDGER_CREATE_ACTIVE_ENTRIES);
-       if (sum_ledger == LEDGER_NULL) {
+       if (sum_ledger == LEDGER_NULL)
                return KERN_RESOURCE_SHORTAGE;
-       }
 
        coalition_lock(coal);
 
@@ -172,29 +447,51 @@ coalition_resource_usage_internal(coalition_t coal, struct coalition_resource_us
         * Start with the coalition's ledger, which holds the totals from all
         * the dead tasks.
         */
-       ledger_rollup(sum_ledger, coal->ledger);
-       uint64_t bytesread = coal->bytesread;
-       uint64_t byteswritten = coal->byteswritten;
-       uint64_t gpu_time = coal->gpu_time;
+       ledger_rollup(sum_ledger, coal->r.ledger);
+       uint64_t bytesread = coal->r.bytesread;
+       uint64_t byteswritten = coal->r.byteswritten;
+       uint64_t gpu_time = coal->r.gpu_time;
+       int64_t cpu_time_billed_to_me = 0;
+       int64_t cpu_time_billed_to_others = 0;
+
+       kr = ledger_get_balance(sum_ledger, task_ledgers.cpu_time_billed_to_me, (int64_t *)&cpu_time_billed_to_me);
+       if (kr != KERN_SUCCESS || cpu_time_billed_to_me < 0) {
+#if DEVELOPMENT || DEBUG
+               printf("ledger_get_balance failed or ledger negative in coalition_resource_usage_internal: %lld\n", cpu_time_billed_to_me);
+#endif /* DEVELOPMENT || DEBUG */
+               cpu_time_billed_to_me = 0;
+       }
+
+       kr = ledger_get_balance(sum_ledger, task_ledgers.cpu_time_billed_to_others, (int64_t *)&cpu_time_billed_to_others);
+       if (kr != KERN_SUCCESS || cpu_time_billed_to_others < 0) {
+#if DEVELOPMENT || DEBUG
+               printf("ledger_get_balance failed or ledger negative in coalition_resource_usage_internal: %lld\n", cpu_time_billed_to_others);
+#endif /* DEVELOPMENT || DEBUG */
+               cpu_time_billed_to_others = 0;
+       }
 
        /*
         * Add to that all the active tasks' ledgers. Tasks cannot deallocate
         * out from under us, since we hold the coalition lock.
+        * Do not use the on-behalf of cpu time from ledger for live tasks, since
+        * it will not have cpu time for active linkages between tasks.
         */
        task_t task;
-       queue_iterate(&coal->tasks, task, task_t, coalition_tasks) {
+       qe_foreach_element(task, &coal->r.tasks, task_coalition[COALITION_TYPE_RESOURCE]) {
                ledger_rollup(sum_ledger, task->ledger);
                bytesread += task->task_io_stats->disk_reads.size;
                byteswritten += task->task_io_stats->total_io.size - task->task_io_stats->disk_reads.size;
                gpu_time += task_gpu_utilisation(task);
+               cpu_time_billed_to_me += (int64_t)bank_billed_time(task->bank_context);
+               cpu_time_billed_to_others += (int64_t)bank_serviced_time(task->bank_context);
        }
 
        /* collect information from the coalition itself */
-       cru_out->tasks_started = coal->task_count;
-       cru_out->tasks_exited = coal->dead_task_count;
+       cru_out->tasks_started = coal->r.task_count;
+       cru_out->tasks_exited = coal->r.dead_task_count;
 
-       uint64_t time_nonempty = coal->time_nonempty;
-       uint64_t last_became_nonempty_time = coal->last_became_nonempty_time;
+       uint64_t time_nonempty = coal->r.time_nonempty;
+       uint64_t last_became_nonempty_time = coal->r.last_became_nonempty_time;
 
        coalition_unlock(coal);
 
@@ -205,6 +502,8 @@ coalition_resource_usage_internal(coalition_t coal, struct coalition_resource_us
                credit = 0;
        }
        cru_out->cpu_time = credit;
+       cru_out->cpu_time_billed_to_me = (uint64_t)cpu_time_billed_to_me;
+       cru_out->cpu_time_billed_to_others = (uint64_t)cpu_time_billed_to_others;
 
        kr = ledger_get_entries(sum_ledger, task_ledgers.interrupt_wakeups,
                        &credit, &debit);
@@ -235,43 +534,242 @@ coalition_resource_usage_internal(coalition_t coal, struct coalition_resource_us
        return KERN_SUCCESS;
 }
 
+/*
+ *
+ * COALITION_TYPE_JETSAM
+ *
+ */
+static kern_return_t
+i_coal_jetsam_init(coalition_t coal, boolean_t privileged)
+{
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+       (void)privileged;
+
+       coal->j.leader= TASK_NULL;
+       queue_head_init(coal->j.extensions);
+       queue_head_init(coal->j.services);
+       queue_head_init(coal->j.other);
+
+       return KERN_SUCCESS;
+}
+
+static void
+i_coal_jetsam_dealloc(__unused coalition_t coal)
+{
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+
+       /* the coalition should be completely clear at this point */
+       assert(queue_empty(&coal->j.extensions));
+       assert(queue_empty(&coal->j.services));
+       assert(queue_empty(&coal->j.other));
+       assert(coal->j.leader == TASK_NULL);
+}
+
+static kern_return_t
+i_coal_jetsam_adopt_task(coalition_t coal, task_t task)
+{
+       struct i_jetsam_coalition *cj;
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+
+       cj = &coal->j;
+
+       assert(queue_empty(&task->task_coalition[COALITION_TYPE_JETSAM]));
+
+       /* put each task initially in the "other" list */
+       enqueue_tail(&cj->other, &task->task_coalition[COALITION_TYPE_JETSAM]);
+       coal_dbg("coalition %lld adopted PID:%d as UNDEF",
+                coal->id, task_pid(task));
+
+       return KERN_SUCCESS;
+}
+
+static kern_return_t
+i_coal_jetsam_remove_task(coalition_t coal, task_t task)
+{
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+       assert(task->coalition[COALITION_TYPE_JETSAM] == coal);
+
+       coal_dbg("removing PID:%d from coalition id:%lld",
+                task_pid(task), coal->id);
+
+       if (task == coal->j.leader) {
+               coal->j.leader = NULL;
+               coal_dbg("    PID:%d was the leader!", task_pid(task));
+       } else {
+               assert(!queue_empty(&task->task_coalition[COALITION_TYPE_JETSAM]));
+       }
+
+       /* remove the task from the specific coalition role queue */
+       remqueue(&task->task_coalition[COALITION_TYPE_JETSAM]);
+       queue_chain_init(task->task_coalition[COALITION_TYPE_RESOURCE]);
+
+       return KERN_SUCCESS;
+}
+
+static kern_return_t
+i_coal_jetsam_set_taskrole(coalition_t coal, task_t task, int role)
+{
+       struct i_jetsam_coalition *cj;
+       queue_t q = NULL;
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+       assert(task->coalition[COALITION_TYPE_JETSAM] == coal);
+
+       cj = &coal->j;
+
+       switch (role) {
+       case COALITION_TASKROLE_LEADER:
+               coal_dbg("setting PID:%d as LEADER of %lld",
+                        task_pid(task), coal->id);
+               if (cj->leader != TASK_NULL) {
+                       /* re-queue the exiting leader onto the "other" list */
+                       coal_dbg("    re-queue existing leader (%d) as OTHER",
+                                task_pid(cj->leader));
+                       re_queue_tail(&cj->other, &cj->leader->task_coalition[COALITION_TYPE_JETSAM]);
+               }
+               /*
+                * remove the task from the "other" list
+                * (where it was put by default)
+                */
+               remqueue(&task->task_coalition[COALITION_TYPE_JETSAM]);
+               queue_chain_init(task->task_coalition[COALITION_TYPE_JETSAM]);
+
+               /* set the coalition leader */
+               cj->leader = task;
+               break;
+       case COALITION_TASKROLE_UNDEF:
+               coal_dbg("setting PID:%d as UNDEF in %lld",
+                        task_pid(task), coal->id);
+               q = (queue_t)&cj->other;
+               break;
+       case COALITION_TASKROLE_XPC:
+               coal_dbg("setting PID:%d as XPC in %lld",
+                        task_pid(task), coal->id);
+               q = (queue_t)&cj->services;
+               break;
+       case COALITION_TASKROLE_EXT:
+               coal_dbg("setting PID:%d as EXT in %lld",
+                        task_pid(task), coal->id);
+               q = (queue_t)&cj->extensions;
+               break;
+       default:
+               panic("%s: invalid role(%d) for task", __func__, role);
+               return KERN_INVALID_ARGUMENT;
+       }
+
+       if (q != NULL)
+               re_queue_tail(q, &task->task_coalition[COALITION_TYPE_JETSAM]);
+
+       return KERN_SUCCESS;
+}
+
+static int
+i_coal_jetsam_get_taskrole(coalition_t coal, task_t task)
+{
+       struct i_jetsam_coalition *cj;
+       task_t t;
+
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+       assert(task->coalition[COALITION_TYPE_JETSAM] == coal);
+
+       cj = &coal->j;
+
+       if (task == cj->leader)
+               return COALITION_TASKROLE_LEADER;
+
+       qe_foreach_element(t, &cj->services, task_coalition[COALITION_TYPE_JETSAM]) {
+               if (t == task)
+                       return COALITION_TASKROLE_XPC;
+       }
+
+       qe_foreach_element(t, &cj->extensions, task_coalition[COALITION_TYPE_JETSAM]) {
+               if (t == task)
+                       return COALITION_TASKROLE_EXT;
+       }
+
+       qe_foreach_element(t, &cj->other, task_coalition[COALITION_TYPE_JETSAM]) {
+               if (t == task)
+                       return COALITION_TASKROLE_UNDEF;
+       }
+
+       /* task not in the coalition?! */
+       return -1;
+}
+
+static void
+i_coal_jetsam_iterate_tasks(coalition_t coal, void *ctx, void (*callback)(coalition_t, void *, task_t))
+{
+       struct i_jetsam_coalition *cj;
+       task_t t;
+
+       assert(coal && coal->type == COALITION_TYPE_JETSAM);
+
+       cj = &coal->j;
+
+       if (cj->leader)
+               callback(coal, ctx, cj->leader);
+
+       qe_foreach_element(t, &cj->services, task_coalition[COALITION_TYPE_JETSAM])
+               callback(coal, ctx, t);
+
+       qe_foreach_element(t, &cj->extensions, task_coalition[COALITION_TYPE_JETSAM])
+               callback(coal, ctx, t);
+
+       qe_foreach_element(t, &cj->other, task_coalition[COALITION_TYPE_JETSAM])
+               callback(coal, ctx, t);
+}
+
+
+/*
+ *
+ * Main Coalition implementation
+ *
+ */
+
 /*
  * coalition_create_internal
  * Returns: New coalition object, referenced for the caller and unlocked.
  * Condition: coalitions_list_lock must be UNLOCKED.
  */
 kern_return_t
-coalition_create_internal(coalition_t *out, boolean_t privileged)
+coalition_create_internal(int type, boolean_t privileged, coalition_t *out)
 {
-       struct coalition *new_coal = (struct coalition *)zalloc(coalition_zone);
-       if (new_coal == COALITION_NULL) {
+       kern_return_t kr;
+       struct coalition *new_coal;
+
+       if (type < 0 || type > COALITION_TYPE_MAX)
+               return KERN_INVALID_ARGUMENT;
+
+       new_coal = (struct coalition *)zalloc(coalition_zone);
+       if (new_coal == COALITION_NULL)
                return KERN_RESOURCE_SHORTAGE;
-       }
        bzero(new_coal, sizeof(*new_coal));
 
-       new_coal->ledger = ledger_instantiate(task_ledger_template, LEDGER_CREATE_ACTIVE_ENTRIES);
-       if (new_coal->ledger == NULL) {
+       new_coal->type = type;
+
+       /* initialize type-specific resources */
+       kr = coal_call(new_coal, init, privileged);
+       if (kr != KERN_SUCCESS) {
                zfree(coalition_zone, new_coal);
-               return KERN_RESOURCE_SHORTAGE;
+               return kr;
        }
 
        /* One for caller, one for coalitions list */
        new_coal->ref_count = 2;
 
        new_coal->privileged = privileged ? TRUE : FALSE;
+#if defined(DEVELOPMENT) || defined(DEBUG)
+       new_coal->should_notify = 1;
+#endif
 
        lck_mtx_init(&new_coal->lock, &coalitions_lck_grp, &coalitions_lck_attr);
-       queue_init(&new_coal->tasks);
 
        lck_mtx_lock(&coalitions_list_lock);
        new_coal->id = coalition_next_id++;
        coalition_count++;
-       queue_enter(&coalitions, new_coal, coalition_t, coalitions);
+       enqueue_tail(&coalitions_q, &new_coal->coalitions);
        lck_mtx_unlock(&coalitions_list_lock);
 
-#if COALITION_DEBUG
-       printf("%s: new coal id %llu\n", __func__, new_coal->id);
-#endif
+       coal_dbg("id:%llu, type:%s", new_coal->id, coal_type_str(new_coal->type));
 
        *out = new_coal;
        return KERN_SUCCESS;
@@ -284,37 +782,61 @@ coalition_create_internal(coalition_t *out, boolean_t privileged)
 void
 coalition_release(coalition_t coal)
 {
-       boolean_t do_dealloc = FALSE;
-
        /* TODO: This can be done with atomics. */
        coalition_lock(coal);
        coal->ref_count--;
-       if (coal->ref_count == 0) {
-               do_dealloc = TRUE;
-       }
+
 #if COALITION_DEBUG
        uint32_t rc = coal->ref_count;
+       uint32_t ac = coal->active_count;
 #endif /* COALITION_DEBUG */
 
+       coal_dbg("id:%llu type:%s ref_count:%u active_count:%u%s",
+                coal->id, coal_type_str(coal->type), rc, ac,
+                rc <= 0 ? ", will deallocate now" : "");
+
+       if (coal->ref_count > 0) {
+               coalition_unlock(coal);
+               return;
+       }
+
+       assert(coal->termrequested);
+       assert(coal->terminated);
+       assert(coal->active_count == 0);
+       assert(coal->reaped);
+       assert(coal->focal_task_count == 0);
+       assert(coal->nonfocal_task_count == 0);
+
+       coal_call(coal, dealloc);
+
        coalition_unlock(coal);
 
-#if COALITION_DEBUG
-       printf("%s: coal %llu ref_count-- -> %u%s\n", __func__, coal->id, rc,
-                       do_dealloc ? ", will deallocate now" : "");
-#endif /* COALITION_DEBUG */
+       lck_mtx_destroy(&coal->lock, &coalitions_lck_grp);
 
-       if (do_dealloc) {
-               assert(coal->termrequested);
-               assert(coal->terminated);
-               assert(coal->active_count == 0);
-               assert(coal->reaped);
-               assert(coal->focal_tasks_count == 0);
-               assert(coal->non_focal_tasks_count == 0);
+       zfree(coalition_zone, coal);
+}
 
-               ledger_dereference(coal->ledger);
-               lck_mtx_destroy(&coal->lock, &coalitions_lck_grp);
-               zfree(coalition_zone, coal);
+/*
+ * coalition_find_by_id_internal
+ * Returns: Coalition object with specified id, NOT referenced.
+ *          If not found, returns COALITION_NULL.
+ * Condition: coalitions_list_lock must be LOCKED.
+ */
+static coalition_t
+coalition_find_by_id_internal(uint64_t coal_id)
+{
+       if (coal_id == 0) {
+               return COALITION_NULL;
        }
+
+       lck_mtx_assert(&coalitions_list_lock, LCK_MTX_ASSERT_OWNED);
+       coalition_t coal;
+       qe_foreach_element(coal, &coalitions_q, coalitions) {
+               if (coal->id == coal_id) {
+                       return coal;
+               }
+       }
+       return COALITION_NULL;
 }
 
 /*
@@ -346,8 +868,8 @@ coalition_find_by_id(uint64_t cid)
        }
 
        if (coal->ref_count == 0) {
-               panic("resurrecting coalition %p id %llu, active_count = %u\n",
-                               coal, coal->id, coal->active_count);
+               panic("resurrecting coalition %p id:%llu type:%s, active_count:%u\n",
+                               coal, coal->id, coal_type_str(coal->type), coal->active_count);
        }
        coal->ref_count++;
 #if COALITION_DEBUG
@@ -357,9 +879,9 @@ coalition_find_by_id(uint64_t cid)
        coalition_unlock(coal);
        lck_mtx_unlock(&coalitions_list_lock);
 
-#if COALITION_DEBUG
-       printf("%s: coal %llu ref_count++ -> %u\n", __func__, coal->id, rc);
-#endif
+       coal_dbg("id:%llu type:%s ref_count:%u",
+                coal->id, coal_type_str(coal->type), rc);
+
        return coal;
 }
 
@@ -397,8 +919,8 @@ coalition_find_and_activate_by_id(uint64_t cid)
        }
 
        if (coal->ref_count == 0) {
-               panic("resurrecting coalition %p id %llu, active_count = %u\n",
-                               coal, coal->id, coal->active_count);
+               panic("resurrecting coalition %p id:%llu type:%s, active_count:%u\n",
+                               coal, coal->id, coal_type_str(coal->type), coal->active_count);
        }
 
        coal->ref_count++;
@@ -412,10 +934,9 @@ coalition_find_and_activate_by_id(uint64_t cid)
        coalition_unlock(coal);
        lck_mtx_unlock(&coalitions_list_lock);
 
-#if COALITION_DEBUG
-       printf("%s: coal %llu ref_count++ -> %u, active_count++ -> %u\n",
-                       __func__, coal->id, rc, ac);
-#endif
+       coal_dbg("id:%llu type:%s ref_count:%u, active_count:%u",
+                coal->id, coal_type_str(coal->type), rc, ac);
+
        return coal;
 }
 
@@ -425,10 +946,41 @@ coalition_id(coalition_t coal)
        return coal->id;
 }
 
-uint64_t
-task_coalition_id(task_t task)
+void
+task_coalition_ids(task_t task, uint64_t ids[COALITION_NUM_TYPES])
+{
+       int i;
+       for (i = 0; i < COALITION_NUM_TYPES; i++) {
+               if (task->coalition[i])
+                       ids[i] = task->coalition[i]->id;
+               else
+                       ids[i] = 0;
+       }
+}
+
+void
+task_coalition_roles(task_t task, int roles[COALITION_NUM_TYPES])
 {
-       return task->coalition->id;
+       int i;
+       memset(roles, 0, COALITION_NUM_TYPES * sizeof(roles[0]));
+
+       for (i = 0; i < COALITION_NUM_TYPES; i++) {
+               if (task->coalition[i]) {
+                       coalition_lock(task->coalition[i]);
+                       roles[i] = coal_call(task->coalition[i],
+                                            get_taskrole, task);
+                       coalition_unlock(task->coalition[i]);
+               } else {
+                       roles[i] = -1;
+               }
+       }
+}
+
+
+int
+coalition_type(coalition_t coal)
+{
+       return coal->type;
 }
 
 boolean_t
@@ -438,50 +990,107 @@ coalition_is_privileged(coalition_t coal)
 }
 
 boolean_t
-task_is_in_privileged_coalition(task_t task)
+task_is_in_privileged_coalition(task_t task, int type)
 {
-       return task->coalition->privileged || unrestrict_coalition_syscalls;
+       if (type < 0 || type > COALITION_TYPE_MAX)
+               return FALSE;
+       if (unrestrict_coalition_syscalls)
+               return TRUE;
+       if (!task->coalition[type])
+               return FALSE;
+       return task->coalition[type]->privileged;
 }
 
-/*
- * coalition_get_ledger
- * Returns: Coalition's ledger, NOT referenced.
- * Condition: Caller must have a coalition reference.
- */
-ledger_t
-coalition_get_ledger(coalition_t coal)
+void task_coalition_update_gpu_stats(task_t task, uint64_t gpu_ns_delta)
 {
-       return coal->ledger;
+       coalition_t coal;
+
+       assert(task != TASK_NULL);
+       if (gpu_ns_delta == 0)
+               return;
+
+       coal = task->coalition[COALITION_TYPE_RESOURCE];
+       assert(coal != COALITION_NULL);
+
+       coalition_lock(coal);
+       coal->r.gpu_time += gpu_ns_delta;
+       coalition_unlock(coal);
 }
 
-/*
- * This is the function to use when you already hold an activation on the
- * coalition, and want to extend it to a second activation owned by a new
- * object, like when a task in the coalition calls fork(). This is analogous
- * to taking a second reference when you already hold one.
- * See also coalition_find_and_activate_by_id.
- */
-kern_return_t
-coalition_extend_active(coalition_t coal)
+uint32_t task_coalition_adjust_focal_count(task_t task, int count)
 {
-       coalition_lock(coal);
+       coalition_t coal;
+       uint32_t ret;
 
-       if (coal->reaped) {
-               panic("cannot make a reaped coalition active again");
-       }
+       /*
+        * For now: only use the resource coalition. Perhaps in the
+        * future we may combine all coalition types, or even make
+        * a special coalition type just for this.
+        */
+       coal = task->coalition[COALITION_TYPE_RESOURCE];
+       assert(coal != COALITION_NULL);
 
-       if (coal->terminated) {
-               coalition_unlock(coal);
-               return KERN_TERMINATED;
-       }
+       ret = hw_atomic_add(&coal->focal_task_count, count);
 
-       assert(coal->active_count > 0);
-       coal->active_count++;
+       /* catch underflow */
+       assert(ret != UINT32_MAX);
+       return ret;
+}
+
+uint32_t task_coalition_focal_count(task_t task)
+{
+       coalition_t coal;
+       coal = task->coalition[COALITION_TYPE_RESOURCE];
+       assert(coal != COALITION_NULL);
+
+       return coal->focal_task_count;
+}
+
+uint32_t task_coalition_adjust_nonfocal_count(task_t task, int count)
+{
+       coalition_t coal;
+       uint32_t ret;
+
+       /*
+        * For now: only use the resource coalition. Perhaps in the
+        * future we may combine all coalition types, or even make
+        * a special coalition type just for this.
+        */
+       coal = task->coalition[COALITION_TYPE_RESOURCE];
+       assert(coal != COALITION_NULL);
+
+       ret = hw_atomic_add(&coal->nonfocal_task_count, count);
+
+       /* catch underflow */
+       assert(ret != UINT32_MAX);
+       return ret;
+}
+
+uint32_t task_coalition_nonfocal_count(task_t task)
+{
+       coalition_t coal;
+       coal = task->coalition[COALITION_TYPE_RESOURCE];
+       assert(coal != COALITION_NULL);
+
+       return coal->nonfocal_task_count;
+}
+
+void coalition_for_each_task(coalition_t coal, void *ctx,
+                            void (*callback)(coalition_t, void *, task_t))
+{
+       assert(coal != COALITION_NULL);
+
+       coal_dbg("iterating tasks in coalition %p id:%llu type:%s, active_count:%u",
+                coal, coal->id, coal_type_str(coal->type), coal->active_count);
+
+       coalition_lock(coal);
+
+       coal_call(coal, iterate_tasks, ctx, callback);
 
        coalition_unlock(coal);
-       return KERN_SUCCESS;
 }
 
+
 void
 coalition_remove_active(coalition_t coal)
 {
@@ -506,13 +1115,26 @@ coalition_remove_active(coalition_t coal)
                assert(!coal->notified);
 
                coal->notified = TRUE;
+#if defined(DEVELOPMENT) || defined(DEBUG)
+               do_notify = coal->should_notify;
+#else
                do_notify = TRUE;
+#endif
                notify_id = coal->id;
                notify_flags = 0;
        }
 
+#if COALITION_DEBUG
+       uint64_t cid = coal->id;
+       uint32_t rc = coal->ref_count;
+       int      ac = coal->active_count;
+       int      ct = coal->type;
+#endif
        coalition_unlock(coal);
 
+       coal_dbg("id:%llu type:%s ref_count:%u, active_count:%u,%s",
+                cid, coal_type_str(ct), rc, ac, do_notify ? " NOTIFY" : " ");
+
        if (do_notify) {
                coalition_notify_user(notify_id, notify_flags);
        }
@@ -520,10 +1142,10 @@ coalition_remove_active(coalition_t coal)
 
 /* Used for kernel_task, launchd, launchd's early boot tasks... */
 kern_return_t
-coalition_default_adopt_task(task_t task)
+coalitions_adopt_init_task(task_t task)
 {
        kern_return_t kr;
-       kr = coalition_adopt_task(default_coalition, task);
+       kr = coalitions_adopt_task(init_coalition, task);
        if (kr != KERN_SUCCESS) {
                panic("failed to adopt task %p into default coalition: %d", task, kr);
        }
@@ -531,14 +1153,16 @@ coalition_default_adopt_task(task_t task)
 }
 
 /*
- * coalition_adopt_task
+ * coalition_adopt_task_internal
  * Condition: Coalition must be referenced and unlocked. Will fail if coalition
  * is already terminated.
  */
-kern_return_t
-coalition_adopt_task(coalition_t coal, task_t task)
+static kern_return_t
+coalition_adopt_task_internal(coalition_t coal, task_t task)
 {
-       if (task->coalition) {
+       kern_return_t kr;
+
+       if (task->coalition[coal->type]) {
                return KERN_ALREADY_IN_SET;
        }
 
@@ -549,72 +1173,152 @@ coalition_adopt_task(coalition_t coal, task_t task)
                return KERN_TERMINATED;
        }
 
+       kr = coal_call(coal, adopt_task, task);
+       if (kr != KERN_SUCCESS)
+               goto out_unlock;
+
        coal->active_count++;
 
        coal->ref_count++;
-       task->coalition = coal;
-
-       queue_enter(&coal->tasks, task, task_t, coalition_tasks);
-       coal->task_count++;
-
-       if(coal->task_count < coal->dead_task_count) {
-               panic("%s: coalition %p id %llu task_count < dead_task_count", __func__, coal, coal->id);
-       }
 
-       /* If moving from 0->1 active tasks */
-       if (coal->task_count - coal->dead_task_count == 1) {
-               coal->last_became_nonempty_time = mach_absolute_time();
-       }
+       task->coalition[coal->type] = coal;
 
+out_unlock:
 #if COALITION_DEBUG
+       (void)coal; /* need expression after label */
+       uint64_t cid = coal->id;
        uint32_t rc = coal->ref_count;
+       uint32_t ct = coal->type;
 #endif
-
        coalition_unlock(coal);
 
+       coal_dbg("task:%d, id:%llu type:%s ref_count:%u, kr=%d",
+                task_pid(task), cid, coal_type_str(ct), rc, kr);
+       return kr;
+}
+
+static kern_return_t
+coalition_remove_task_internal(task_t task, int type)
+{
+       kern_return_t kr;
+
+       coalition_t coal = task->coalition[type];
+
+       if (!coal)
+               return KERN_SUCCESS;
+
+       assert(coal->type == (uint32_t)type);
+
+       coalition_lock(coal);
+
+       kr = coal_call(coal, remove_task, task);
+
 #if COALITION_DEBUG
-       if (rc) {
-               printf("%s: coal %llu ref_count++ -> %u\n", __func__, coal->id, rc);
-       }
+       uint64_t cid = coal->id;
+       uint32_t rc = coal->ref_count;
+       int      ac = coal->active_count;
+       int      ct = coal->type;
 #endif
-       return KERN_SUCCESS;
+       coalition_unlock(coal);
+
+       coal_dbg("id:%llu type:%s ref_count:%u, active_count:%u, kr=%d",
+                cid, coal_type_str(ct), rc, ac, kr);
+
+       coalition_remove_active(coal);
+
+       return kr;
 }
 
 /*
- * coalition_remove_task
- * Condition: task must be referenced and UNLOCKED; task's coalition must be UNLOCKED
+ * coalitions_adopt_task
+ * Condition: All coalitions must be referenced and unlocked.
+ * Will fail if any coalition is already terminated.
  */
 kern_return_t
-coalition_remove_task(task_t task)
+coalitions_adopt_task(coalition_t *coals, task_t task)
 {
-       coalition_t coal = task->coalition;
-       assert(coal);
+       int i;
+       kern_return_t kr;
 
-       coalition_lock(coal);
+       if (!coals || coals[COALITION_TYPE_RESOURCE] == COALITION_NULL)
+               return KERN_INVALID_ARGUMENT;
+
+       /* verify that the incoming coalitions are what they say they are */
+       for (i = 0; i < COALITION_NUM_TYPES; i++)
+               if (coals[i] && coals[i]->type != (uint32_t)i)
+                       return KERN_INVALID_ARGUMENT;
+
+       for (i = 0; i < COALITION_NUM_TYPES; i++) {
+               kr = KERN_SUCCESS;
+               if (coals[i])
+                       kr = coalition_adopt_task_internal(coals[i], task);
+               if (kr != KERN_SUCCESS) {
+                       /* dis-associate any coalitions that just adopted this task */
+                       while (--i >= 0) {
+                               if (task->coalition[i])
+                                       coalition_remove_task_internal(task, i);
+                       }
+                       break;
+               }
+       }
+       return kr;
+}
 
-       queue_remove(&coal->tasks, task, task_t, coalition_tasks);
-       coal->dead_task_count++;
+/*
+ * coalitions_remove_task
+ * Condition: task must be referenced and UNLOCKED; all task's coalitions must be UNLOCKED
+ */
+kern_return_t
+coalitions_remove_task(task_t task)
+{
+       kern_return_t kr;
+       int i;
 
-       if(coal->task_count < coal->dead_task_count) {
-               panic("%s: coalition %p id %llu task_count < dead_task_count", __func__, coal, coal->id);
+       for (i = 0; i < COALITION_NUM_TYPES; i++) {
+               kr = coalition_remove_task_internal(task, i);
+               assert(kr == KERN_SUCCESS);
        }
 
-       /* If moving from 1->0 active tasks */
-       if (coal->task_count - coal->dead_task_count == 0) {
-               uint64_t last_time_nonempty = mach_absolute_time() - coal->last_became_nonempty_time;
-               coal->last_became_nonempty_time = 0;
-               coal->time_nonempty += last_time_nonempty;
+       return kr;
+}
+
+/*
+ * task_release_coalitions
+ * helper function to release references to all coalitions in which
+ * 'task' is a member.
+ */
+void
+task_release_coalitions(task_t task)
+{
+       int i;
+       for (i = 0; i < COALITION_NUM_TYPES; i++) {
+               if (task->coalition[i])
+                       coalition_release(task->coalition[i]);
        }
+}
 
-       ledger_rollup(coal->ledger, task->ledger);
-       coal->bytesread += task->task_io_stats->disk_reads.size;
-       coal->byteswritten += task->task_io_stats->total_io.size - task->task_io_stats->disk_reads.size;
-       coal->gpu_time += task_gpu_utilisation(task);
+/*
+ * coalitions_set_roles
+ * for each type of coalition, if the task is a member of a coalition of
+ * that type (given in the coalitions parameter) then set the role of
+ * the task within that that coalition.
+ */
+kern_return_t coalitions_set_roles(coalition_t coalitions[COALITION_NUM_TYPES],
+                                  task_t task, int roles[COALITION_NUM_TYPES])
+{
+       kern_return_t kr = KERN_SUCCESS;
+       int i;
 
-       coalition_unlock(coal);
+       for (i = 0; i < COALITION_NUM_TYPES; i++) {
+               if (!coalitions[i])
+                       continue;
+               coalition_lock(coalitions[i]);
+               kr = coal_call(coalitions[i], set_taskrole, task, roles[i]);
+               coalition_unlock(coalitions[i]);
+               assert(kr == KERN_SUCCESS);
+       }
 
-       coalition_remove_active(coal);
-       return KERN_SUCCESS;
+       return kr;
 }
 
 /*
@@ -624,7 +1328,9 @@ coalition_remove_task(task_t task)
 kern_return_t
 coalition_request_terminate_internal(coalition_t coal)
 {
-       if (coal == default_coalition) {
+       assert(coal->type >= 0 && coal->type <= COALITION_TYPE_MAX);
+
+       if (coal == init_coalition[coal->type]) {
                return KERN_DEFAULT_SET;
        }
 
@@ -649,8 +1355,9 @@ coalition_request_terminate_internal(coalition_t coal)
        if (coal->active_count == 0) {
                /*
                 * We only notify once, when active_count reaches zero.
-                * We just decremented, so if it reached zero, we mustn't have
-                * notified already.
+                * We just set termrequested to zero. If the active count
+                * was already at zero (tasks died before we could request
+                * a termination notification), we should notify.
                 */
                assert(!coal->terminated);
                coal->terminated = TRUE;
@@ -658,7 +1365,11 @@ coalition_request_terminate_internal(coalition_t coal)
                assert(!coal->notified);
 
                coal->notified = TRUE;
+#if defined(DEVELOPMENT) || defined(DEBUG)
+               do_notify = coal->should_notify;
+#else
                do_notify = TRUE;
+#endif
                note_id = coal->id;
                note_flags = 0;
        }
@@ -679,7 +1390,9 @@ coalition_request_terminate_internal(coalition_t coal)
 kern_return_t
 coalition_reap_internal(coalition_t coal)
 {
-       if (coal == default_coalition) {
+       assert(coal->type <= COALITION_TYPE_MAX);
+
+       if (coal == init_coalition[coal->type]) {
                return KERN_DEFAULT_SET;
        }
 
@@ -707,7 +1420,7 @@ coalition_reap_internal(coalition_t coal)
 
        lck_mtx_lock(&coalitions_list_lock);
        coalition_count--;
-       queue_remove(&coalitions, coal, coalition_t, coalitions);
+       remqueue(&coal->coalitions);
        lck_mtx_unlock(&coalitions_list_lock);
 
        /* Release the list's reference and launchd's reference. */
@@ -717,16 +1430,43 @@ coalition_reap_internal(coalition_t coal)
        return KERN_SUCCESS;
 }
 
+#if defined(DEVELOPMENT) || defined(DEBUG)
+int coalition_should_notify(coalition_t coal)
+{
+       int should;
+       if (!coal)
+               return -1;
+       coalition_lock(coal);
+       should = coal->should_notify;
+       coalition_unlock(coal);
+
+       return should;
+}
+
+void coalition_set_notify(coalition_t coal, int notify)
+{
+       if (!coal)
+               return;
+       coalition_lock(coal);
+       coal->should_notify = !!notify;
+       coalition_unlock(coal);
+}
+#endif
+
 void
-coalition_init(void)
+coalitions_init(void)
 {
+       kern_return_t kr;
+       int i;
+       const struct coalition_type *ctype;
+
        coalition_zone = zinit(
                        sizeof(struct coalition),
                        CONFIG_COALITION_MAX * sizeof(struct coalition),
                        COALITION_CHUNK * sizeof(struct coalition),
                        "coalitions");
        zone_change(coalition_zone, Z_NOENCRYPT, TRUE);
-       queue_init(&coalitions);
+       queue_head_init(coalitions_q);
 
        if (!PE_parse_boot_argn("unrestrict_coalition_syscalls", &unrestrict_coalition_syscalls,
                sizeof (unrestrict_coalition_syscalls))) {
@@ -740,54 +1480,402 @@ coalition_init(void)
 
        init_task_ledgers();
 
-       kern_return_t kr = coalition_create_internal(&default_coalition, TRUE);
-       if (kr != KERN_SUCCESS) {
-               panic("%s: could not create default coalition: %d", __func__, kr);
+       for (i = 0, ctype = &s_coalition_types[0]; i < COALITION_NUM_TYPES; ctype++, i++) {
+               /* verify the entry in the global coalition types array */
+               if (ctype->type != i ||
+                   !ctype->init ||
+                   !ctype->dealloc ||
+                   !ctype->adopt_task ||
+                   !ctype->remove_task) {
+                       panic("%s: Malformed coalition type %s(%d) in slot for type:%s(%d)",
+                             __func__, coal_type_str(ctype->type), ctype->type, coal_type_str(i), i);
+               }
+               if (!ctype->has_default)
+                       continue;
+               kr = coalition_create_internal(ctype->type, TRUE, &init_coalition[ctype->type]);
+               if (kr != KERN_SUCCESS)
+                       panic("%s: could not create init %s coalition: kr:%d",
+                             __func__, coal_type_str(i), kr);
        }
+
        /* "Leak" our reference to the global object */
 }
 
-/* coalition focal tasks */
-uint32_t coalition_adjust_focal_task_count(coalition_t coal, int count)
+/*
+ * BSD Kernel interface functions
+ *
+ */
+static void coalition_fill_procinfo(struct coalition *coal,
+                                   struct procinfo_coalinfo *coalinfo)
 {
-       return hw_atomic_add(&coal->focal_tasks_count, count);
+       coalinfo->coalition_id = coal->id;
+       coalinfo->coalition_type = coal->type;
+       coalinfo->coalition_tasks = coalition_get_task_count(coal);
 }
 
-uint32_t coalition_focal_task_count(coalition_t coal)
+
+int coalitions_get_list(int type, struct procinfo_coalinfo *coal_list, int list_sz)
 {
-       return coal->focal_tasks_count;
+       int ncoals = 0;
+       struct coalition *coal;
+
+       lck_mtx_lock(&coalitions_list_lock);
+       qe_foreach_element(coal, &coalitions_q, coalitions) {
+               if (!coal->reaped && (type < 0 || type == (int)coal->type)) {
+                       if (coal_list && ncoals < list_sz)
+                               coalition_fill_procinfo(coal, &coal_list[ncoals]);
+                       ++ncoals;
+               }
+       }
+       lck_mtx_unlock(&coalitions_list_lock);
+
+       return ncoals;
 }
 
-uint32_t coalition_adjust_non_focal_task_count(coalition_t coal, int count)
+/*
+ * Jetsam coalition interface
+ *
+ */
+boolean_t coalition_is_leader(task_t task, int coal_type, coalition_t *coal)
 {
-       return hw_atomic_add(&coal->non_focal_tasks_count, count);
+       coalition_t c;
+       boolean_t ret;
+
+       if (coal) /* handle the error cases gracefully */
+               *coal = COALITION_NULL;
+
+       if (!task)
+               return FALSE;
+
+       if (coal_type > COALITION_TYPE_MAX)
+               return FALSE;
+
+       c = task->coalition[coal_type];
+       if (!c)
+               return FALSE;
+
+       assert((int)c->type == coal_type);
+
+       coalition_lock(c);
+
+       if (coal)
+               *coal = c;
+
+       ret = FALSE;
+       if (c->type == COALITION_TYPE_JETSAM && c->j.leader == task)
+               ret = TRUE;
+
+       coalition_unlock(c);
+
+       return ret;
 }
 
-uint32_t coalition_non_focal_task_count(coalition_t coal)
+
+int coalition_get_task_count(coalition_t coal)
 {
-       return coal->non_focal_tasks_count;
+       int ntasks = 0;
+       struct queue_entry *qe;
+       if (!coal)
+               return 0;
+
+       coalition_lock(coal);
+       switch (coal->type) {
+       case COALITION_TYPE_RESOURCE:
+               qe_foreach(qe, &coal->r.tasks)
+                       ntasks++;
+               break;
+       case COALITION_TYPE_JETSAM:
+               if (coal->j.leader)
+                       ntasks++;
+               qe_foreach(qe, &coal->j.other)
+                       ntasks++;
+               qe_foreach(qe, &coal->j.extensions)
+                       ntasks++;
+               qe_foreach(qe, &coal->j.services)
+                       ntasks++;
+               break;
+       default:
+               break;
+       }
+       coalition_unlock(coal);
+
+       return ntasks;
 }
 
-/* Call sfi_reevaluate() for every thread in the coalition */
-void coalition_sfi_reevaluate(coalition_t coal, task_t updated_task) {
+
+static uint64_t i_get_list_footprint(queue_t list, int type, int *ntasks)
+{
        task_t task;
-       thread_t thread;
+       uint64_t bytes = 0;
 
-       coalition_lock(coal);
+       qe_foreach_element(task, list, task_coalition[type]) {
+               bytes += get_task_phys_footprint(task);
+               coal_dbg("    [%d] task_pid:%d, type:%d, footprint:%lld",
+                        *ntasks, task_pid(task), type, bytes);
+               *ntasks += 1;
+       }
 
-       queue_iterate(&coal->tasks, task, task_t, coalition_tasks) {
+       return bytes;
+}
 
-               /* Skip the task we're doing this on behalf of - it's already updated */
-               if (task == updated_task)
-                       continue;
+uint64_t coalition_get_page_count(coalition_t coal, int *ntasks)
+{
+       uint64_t bytes = 0;
+       int num_tasks = 0;
+
+       if (ntasks)
+               *ntasks = 0;
+       if (!coal)
+               return bytes;
 
-               task_lock(task);
+       coalition_lock(coal);
 
-               queue_iterate(&task->threads, thread, thread_t, task_threads) {
-                               sfi_reevaluate(thread);
+       switch (coal->type) {
+       case COALITION_TYPE_RESOURCE:
+               bytes += i_get_list_footprint(&coal->r.tasks, COALITION_TYPE_RESOURCE, &num_tasks);
+               break;
+       case COALITION_TYPE_JETSAM:
+               if (coal->j.leader) {
+                       bytes += get_task_phys_footprint(coal->j.leader);
+                       num_tasks = 1;
                }
-               task_unlock(task);
+               bytes += i_get_list_footprint(&coal->j.extensions, COALITION_TYPE_JETSAM, &num_tasks);
+               bytes += i_get_list_footprint(&coal->j.services, COALITION_TYPE_JETSAM, &num_tasks);
+               bytes += i_get_list_footprint(&coal->j.other, COALITION_TYPE_JETSAM, &num_tasks);
+               break;
+       default:
+               break;
        }
+
        coalition_unlock(coal);
+
+       if (ntasks)
+               *ntasks = num_tasks;
+
+       return bytes / PAGE_SIZE_64;
 }
 
+struct coal_sort_s {
+       int pid;
+       int usr_order;
+       uint64_t bytes;
+};
+
+/*
+ * return < 0 for a < b
+ *          0 for a == b
+ *        > 0 for a > b
+ */
+typedef int (*cmpfunc_t)(const void *a, const void *b);
+
+extern void
+qsort(void *a, size_t n, size_t es, cmpfunc_t cmp);
+
+static int dflt_cmp(const void *a, const void *b)
+{
+       const struct coal_sort_s *csA = (const struct coal_sort_s *)a;
+       const struct coal_sort_s *csB = (const struct coal_sort_s *)b;
+
+       /*
+        * if both A and B are equal, use a memory descending sort
+        */
+       if (csA->usr_order == csB->usr_order)
+               return (int)((int64_t)csB->bytes - (int64_t)csA->bytes);
+
+       /* otherwise, return the relationship between user specified orders */
+       return (csA->usr_order - csB->usr_order);
+}
+
+static int mem_asc_cmp(const void *a, const void *b)
+{
+       const struct coal_sort_s *csA = (const struct coal_sort_s *)a;
+       const struct coal_sort_s *csB = (const struct coal_sort_s *)b;
+
+       return (int)((int64_t)csA->bytes - (int64_t)csB->bytes);
+}
+
+static int mem_dec_cmp(const void *a, const void *b)
+{
+       const struct coal_sort_s *csA = (const struct coal_sort_s *)a;
+       const struct coal_sort_s *csB = (const struct coal_sort_s *)b;
+
+       return (int)((int64_t)csB->bytes - (int64_t)csA->bytes);
+}
+
+static int usr_asc_cmp(const void *a, const void *b)
+{
+       const struct coal_sort_s *csA = (const struct coal_sort_s *)a;
+       const struct coal_sort_s *csB = (const struct coal_sort_s *)b;
+
+       return (csA->usr_order - csB->usr_order);
+}
+
+static int usr_dec_cmp(const void *a, const void *b)
+{
+       const struct coal_sort_s *csA = (const struct coal_sort_s *)a;
+       const struct coal_sort_s *csB = (const struct coal_sort_s *)b;
+
+       return (csB->usr_order - csA->usr_order);
+}
+
+/* avoid dynamic allocation in this path */
+#define MAX_SORTED_PIDS  80
+
+static int coalition_get_sort_list(coalition_t coal, int sort_order, queue_t list,
+                                  struct coal_sort_s *sort_array, int array_sz)
+{
+       int ntasks = 0;
+       task_t task;
+
+       assert(sort_array != NULL);
+
+       if (array_sz <= 0)
+               return 0;
+
+       if (!list) {
+               /*
+                * this function will only be called with a NULL
+                * list for JETSAM-type coalitions, and is intended
+                * to investigate the leader process
+                */
+               if (coal->type != COALITION_TYPE_JETSAM ||
+                   coal->j.leader == TASK_NULL)
+                       return 0;
+               sort_array[0].pid = task_pid(coal->j.leader);
+               switch (sort_order) {
+               case COALITION_SORT_DEFAULT:
+                       sort_array[0].usr_order = 0;
+                       /* fall-through */
+               case COALITION_SORT_MEM_ASC:
+               case COALITION_SORT_MEM_DEC:
+                       sort_array[0].bytes = get_task_phys_footprint(coal->j.leader);
+                       break;
+               case COALITION_SORT_USER_ASC:
+               case COALITION_SORT_USER_DEC:
+                       sort_array[0].usr_order = 0;
+                       break;
+               default:
+                       break;
+               }
+               return 1;
+       }
+
+       qe_foreach_element(task, list, task_coalition[coal->type]) {
+               if (ntasks >= array_sz) {
+                       printf("WARNING: more than %d pids in coalition %llu\n",
+                              MAX_SORTED_PIDS, coal->id);
+                       break;
+               }
+
+               sort_array[ntasks].pid = task_pid(task);
+
+               switch (sort_order) {
+               case COALITION_SORT_DEFAULT:
+                       sort_array[ntasks].usr_order = 0;
+                       /* fall-through */
+               case COALITION_SORT_MEM_ASC:
+               case COALITION_SORT_MEM_DEC:
+                       sort_array[ntasks].bytes = get_task_phys_footprint(task);
+                       break;
+               case COALITION_SORT_USER_ASC:
+               case COALITION_SORT_USER_DEC:
+                       sort_array[ntasks].usr_order = 0;
+                       break;
+               default:
+                       break;
+               }
+
+               ntasks++;
+       }
+
+       return ntasks;
+}
+
+int coalition_get_pid_list(coalition_t coal, uint32_t rolemask, int sort_order,
+                          int *pid_list, int list_sz)
+{
+       struct i_jetsam_coalition *cj;
+       int ntasks = 0;
+       cmpfunc_t cmp_func = NULL;
+       struct coal_sort_s sort_array[MAX_SORTED_PIDS] = { {0,0,0} }; /* keep to < 2k */
+
+       if (!coal ||
+           !(rolemask & COALITION_ROLEMASK_ALLROLES) ||
+           !pid_list || list_sz < 1) {
+               coal_dbg("Invalid parameters: coal:%p, type:%d, rolemask:0x%x, "
+                        "pid_list:%p, list_sz:%d", coal, coal ? coal->type : -1,
+                        rolemask, pid_list, list_sz);
+               return -EINVAL;
+       }
+
+       switch (sort_order) {
+       case COALITION_SORT_NOSORT:
+               cmp_func = NULL;
+               break;
+       case COALITION_SORT_DEFAULT:
+               cmp_func = dflt_cmp;
+               break;
+       case COALITION_SORT_MEM_ASC:
+               cmp_func = mem_asc_cmp;
+               break;
+       case COALITION_SORT_MEM_DEC:
+               cmp_func = mem_dec_cmp;
+               break;
+       case COALITION_SORT_USER_ASC:
+               cmp_func = usr_asc_cmp;
+               break;
+       case COALITION_SORT_USER_DEC:
+               cmp_func = usr_dec_cmp;
+               break;
+       default:
+               return -ENOTSUP;
+       }
+
+       coalition_lock(coal);
+
+       if (coal->type == COALITION_TYPE_RESOURCE) {
+               ntasks += coalition_get_sort_list(coal, sort_order, &coal->r.tasks,
+                                                 sort_array, MAX_SORTED_PIDS);
+               goto unlock_coal;
+       }
+
+       cj = &coal->j;
+
+       if (rolemask & COALITION_ROLEMASK_UNDEF)
+               ntasks += coalition_get_sort_list(coal, sort_order, &cj->other,
+                                                 sort_array + ntasks,
+                                                 MAX_SORTED_PIDS - ntasks);
+
+       if (rolemask & COALITION_ROLEMASK_XPC)
+               ntasks += coalition_get_sort_list(coal, sort_order, &cj->services,
+                                                 sort_array + ntasks,
+                                                 MAX_SORTED_PIDS - ntasks);
+
+       if (rolemask & COALITION_ROLEMASK_EXT)
+               ntasks += coalition_get_sort_list(coal, sort_order, &cj->extensions,
+                                                 sort_array + ntasks,
+                                                 MAX_SORTED_PIDS - ntasks);
+
+       if (rolemask & COALITION_ROLEMASK_LEADER)
+               ntasks += coalition_get_sort_list(coal, sort_order, NULL,
+                                                 sort_array + ntasks,
+                                                 MAX_SORTED_PIDS - ntasks);
+
+unlock_coal:
+       coalition_unlock(coal);
+
+       /* sort based on the chosen criterion (no sense sorting 1 item) */
+       if (cmp_func && ntasks > 1)
+               qsort(sort_array, ntasks, sizeof(struct coal_sort_s), cmp_func);
+
+       for (int i = 0; i < ntasks; i++) {
+               if (i >= list_sz)
+                       break;
+               coal_dbg(" [%d] PID:%d, footprint:%lld, usr_order:%d",
+                        i, sort_array[i].pid, sort_array[i].bytes,
+                        sort_array[i].usr_order);
+               pid_list[i] = sort_array[i].pid;
+       }
+
+       return ntasks;
+}