#endif
#include <arch/arch.h>
+#include <arch/bitmask.h>
#include <process.h>
#include <atomic.h>
#include <smp.h>
#include <pmap.h>
+#include <trap.h>
#include <schedule.h>
#include <manager.h>
#include <stdio.h>
#include <assert.h>
#include <timing.h>
+#include <hashtable.h>
+#include <slab.h>
#include <sys/queue.h>
/* Process Lists */
-struct proc_list proc_freelist = TAILQ_HEAD_INITIALIZER(proc_freelist);
-spinlock_t freelist_lock = 0;
struct proc_list proc_runnablelist = TAILQ_HEAD_INITIALIZER(proc_runnablelist);
-spinlock_t runnablelist_lock = 0;
+spinlock_t runnablelist_lock = SPINLOCK_INITIALIZER;
+struct kmem_cache *proc_cache;
/* Tracks which cores are idle, similar to the vcoremap. Each value is the
* physical coreid of an unallocated core. */
-spinlock_t idle_lock = 0;
-uint32_t idlecoremap[MAX_NUM_CPUS];
-uint32_t num_idlecores = 0;
+spinlock_t idle_lock = SPINLOCK_INITIALIZER;
+uint32_t LCKD(&idle_lock) (RO idlecoremap)[MAX_NUM_CPUS];
+uint32_t LCKD(&idle_lock) num_idlecores = 0;
-/*
- * While this could be done with just an assignment, this gives us the
+/* Helper function to return a core to the idlemap. It causes some more lock
+ * acquisitions (like in a for loop), but it's a little easier. Plus, one day
+ * we might be able to do this without locks (for the putting). */
+static void put_idle_core(uint32_t coreid)
+{
+ spin_lock(&idle_lock);
+ idlecoremap[num_idlecores++] = coreid;
+ spin_unlock(&idle_lock);
+}
+
+/* Other helpers, implemented later. */
+static uint32_t get_free_vcoreid(struct proc *SAFE p, uint32_t prev);
+static uint32_t get_busy_vcoreid(struct proc *SAFE p, uint32_t prev);
+static int32_t __get_vcoreid(int32_t *corelist, size_t num, int32_t pcoreid);
+static int32_t get_vcoreid(struct proc *SAFE p, int32_t pcoreid);
+static inline void __wait_for_ipi(const char *fnname);
+
+/* PID management. */
+#define PID_MAX 32767 // goes from 0 to 32767, with 0 reserved
+static DECL_BITMASK(pid_bmask, PID_MAX + 1);
+spinlock_t pid_bmask_lock = SPINLOCK_INITIALIZER;
+struct hashtable *pid_hash;
+spinlock_t pid_hash_lock; // initialized in proc_init
+
+/* Finds the next free entry (zero) entry in the pid_bitmask. Set means busy.
+ * PID 0 is reserved (in proc_init). A return value of 0 is a failure (and
+ * you'll also see a warning, for now). Consider doing this with atomics. */
+static pid_t get_free_pid(void)
+{
+ static pid_t next_free_pid = 1;
+ pid_t my_pid = 0;
+
+ spin_lock(&pid_bmask_lock);
+ // atomically (can lock for now, then change to atomic_and_return
+ FOR_CIRC_BUFFER(next_free_pid, PID_MAX + 1, i) {
+ // always points to the next to test
+ next_free_pid = (next_free_pid + 1) % (PID_MAX + 1);
+ if (!GET_BITMASK_BIT(pid_bmask, i)) {
+ SET_BITMASK_BIT(pid_bmask, i);
+ my_pid = i;
+ break;
+ }
+ }
+ spin_unlock(&pid_bmask_lock);
+ if (!my_pid)
+ warn("Shazbot! Unable to find a PID! You need to deal with this!\n");
+ return my_pid;
+}
+
+/* Return a pid to the pid bitmask */
+static void put_free_pid(pid_t pid)
+{
+ spin_lock(&pid_bmask_lock);
+ CLR_BITMASK_BIT(pid_bmask, pid);
+ spin_unlock(&pid_bmask_lock);
+}
+
+/* While this could be done with just an assignment, this gives us the
* opportunity to check for bad transitions. Might compile these out later, so
- * we shouldn't rely on them for sanity checking from userspace.
- */
-int proc_set_state(struct proc *p, uint32_t state)
+ * we shouldn't rely on them for sanity checking from userspace. */
+int __proc_set_state(struct proc *p, uint32_t state)
{
uint32_t curstate = p->state;
/* Valid transitions:
* RBS -> D
* RBM -> D
*
- * This isn't allowed yet, may be later.
+ * This isn't allowed yet, should be later. Is definitely causable.
* C -> D
*/
#if 1 // some sort of correctness flag
return 0;
}
-/* Change this when we aren't using an array */
-struct proc *get_proc(unsigned pid)
+/* Returns a pointer to the proc with the given pid, or 0 if there is none */
+struct proc *pid2proc(pid_t pid)
+{
+ spin_lock(&pid_hash_lock);
+ struct proc *p = hashtable_search(pid_hash, (void*)pid);
+ spin_unlock(&pid_hash_lock);
+ /* if the refcnt was 0, decref and return 0 (we failed). (TODO) */
+ if (p)
+ proc_incref(p, 1); // TODO:(REF) to do this all atomically and not panic
+ return p;
+}
+
+/* Performs any intialization related to processes, such as create the proc
+ * cache, prep the scheduler, etc. When this returns, we should be ready to use
+ * any process related function. */
+void proc_init(void)
+{
+ proc_cache = kmem_cache_create("proc", sizeof(struct proc),
+ MAX(HW_CACHE_ALIGN, __alignof__(struct proc)), 0, 0, 0);
+ /* Init PID mask and hash. pid 0 is reserved. */
+ SET_BITMASK_BIT(pid_bmask, 0);
+ spinlock_init(&pid_hash_lock);
+ spin_lock(&pid_hash_lock);
+ pid_hash = create_hashtable(100, __generic_hash, __generic_eq);
+ spin_unlock(&pid_hash_lock);
+ schedule_init();
+ /* Init idle cores. core 0 is not idle, all others are (for now) */
+ spin_lock(&idle_lock);
+ num_idlecores = num_cpus - 1;
+ for (int i = 0; i < num_idlecores; i++)
+ idlecoremap[i] = i + 1;
+ spin_unlock(&idle_lock);
+ atomic_init(&num_envs, 0);
+}
+
+static void
+proc_init_procinfo(struct proc* p)
+{
+ p->env_procinfo->pid = p->pid;
+ p->env_procinfo->ppid = p->ppid;
+ p->env_procinfo->tsc_freq = system_timing.tsc_freq;
+ // TODO: maybe do something smarter here
+ p->env_procinfo->max_harts = MAX(1,num_cpus); // hack to use all cores
+}
+
+/* Allocates and initializes a process, with the given parent. Currently
+ * writes the *p into **pp, and returns 0 on success, < 0 for an error.
+ * Errors include:
+ * - ENOFREEPID if it can't get a PID
+ * - ENOMEM on memory exhaustion */
+static error_t proc_alloc(struct proc *SAFE*SAFE pp, pid_t parent_id)
+{
+ error_t r;
+ struct proc *p;
+
+ if (!(p = kmem_cache_alloc(proc_cache, 0)))
+ return -ENOMEM;
+
+ { INITSTRUCT(*p)
+
+ // Setup the default map of where to get cache colors from
+ p->cache_colors_map = global_cache_colors_map;
+ p->next_cache_color = 0;
+
+ /* Initialize the address space */
+ if ((r = env_setup_vm(p)) < 0) {
+ kmem_cache_free(proc_cache, p);
+ return r;
+ }
+
+ /* Get a pid, then store a reference in the pid_hash */
+ if (!(p->pid = get_free_pid())) {
+ kmem_cache_free(proc_cache, p);
+ return -ENOFREEPID;
+ }
+ spin_lock(&pid_hash_lock);
+ hashtable_insert(pid_hash, (void*)p->pid, p);
+ spin_unlock(&pid_hash_lock);
+
+ /* Set the basic status variables. */
+ spinlock_init(&p->proc_lock);
+ p->exitcode = 0;
+ p->ppid = parent_id;
+ p->state = PROC_CREATED; // shouldn't go through state machine for init
+ p->env_refcnt = 2; // one for the object, one for the ref we pass back
+ p->env_flags = 0;
+ p->env_entry = 0; // cheating. this really gets set in load_icode
+ p->num_vcores = 0;
+ p->heap_bottom = (void*)UTEXT;
+ p->heap_top = (void*)UTEXT;
+ memset(&p->vcoremap, -1, sizeof(p->vcoremap));
+ memset(&p->resources, 0, sizeof(p->resources));
+ memset(&p->env_ancillary_state, 0, sizeof(p->env_ancillary_state));
+ memset(&p->env_tf, 0, sizeof(p->env_tf));
+ proc_init_trapframe(&p->env_tf,0);
+
+ /* Initialize the contents of the e->env_procinfo structure */
+ proc_init_procinfo(p);
+ /* Initialize the contents of the e->env_procdata structure */
+
+ /* Initialize the generic syscall ring buffer */
+ SHARED_RING_INIT(&p->env_procdata->syscallring);
+ /* Initialize the backend of the syscall ring buffer */
+ BACK_RING_INIT(&p->syscallbackring,
+ &p->env_procdata->syscallring,
+ SYSCALLRINGSIZE);
+
+ /* Initialize the generic sysevent ring buffer */
+ SHARED_RING_INIT(&p->env_procdata->syseventring);
+ /* Initialize the frontend of the sysevent ring buffer */
+ FRONT_RING_INIT(&p->syseventfrontring,
+ &p->env_procdata->syseventring,
+ SYSEVENTRINGSIZE);
+ *pp = p;
+ atomic_inc(&num_envs);
+
+ proc_init_arch(p);
+
+ printd("[%08x] new process %08x\n", current ? current->pid : 0, p->pid);
+ } // INIT_STRUCT
+ return 0;
+}
+
+/* Creates a process from the specified binary, which is of size size.
+ * Currently, the binary must be a contiguous block of memory, which needs to
+ * change. On any failure, it just panics, which ought to be sorted. */
+struct proc *proc_create(uint8_t *binary, size_t size)
+{
+ struct proc *p;
+ error_t r;
+ pid_t curid;
+
+ curid = (current ? current->pid : 0);
+ if ((r = proc_alloc(&p, curid)) < 0)
+ panic("proc_create: %e", r); // one of 3 quaint usages of %e.
+ if(binary != NULL)
+ env_load_icode(p, NULL, binary, size);
+ return p;
+}
+
+/* This is called by proc_decref, once the last reference to the process is
+ * gone. Don't call this otherwise (it will panic). It will clean up the
+ * address space and deallocate any other used memory. */
+static void __proc_free(struct proc *p)
{
- // should have some error checking when we do this for real
- return &envs[ENVX(pid)];
+ physaddr_t pa;
+
+ printd("[PID %d] freeing proc: %d\n", current ? current->pid : 0, p->pid);
+ // All parts of the kernel should have decref'd before __proc_free is called
+ assert(p->env_refcnt == 0);
+
+ proc_free_arch(p);
+
+ // Free any colors allocated to this process
+ if(p->cache_colors_map != global_cache_colors_map) {
+ for(int i=0; i<llc_cache->num_colors; i++)
+ cache_color_free(llc_cache, p->cache_colors_map);
+ cache_colors_map_free(p->cache_colors_map);
+ }
+
+ // Flush all mapped pages in the user portion of the address space
+ env_user_mem_free(p);
+
+ // free the page directory
+ pa = p->env_cr3;
+ p->env_pgdir = 0;
+ p->env_cr3 = 0;
+ page_decref(pa2page(pa));
+
+ /* Remove self from the pid hash, return PID. Note the reversed order. */
+ spin_lock(&pid_hash_lock);
+ if (!hashtable_remove(pid_hash, (void*)p->pid))
+ panic("Proc not in the pid table in %s", __FUNCTION__);
+ spin_unlock(&pid_hash_lock);
+ put_free_pid(p->pid);
+ atomic_dec(&num_envs);
+
+ /* Dealloc the struct proc */
+ kmem_cache_free(proc_cache, p);
}
-/* Whether or not actor can control target */
+/* Whether or not actor can control target. Note we currently don't need
+ * locking for this. TODO: think about that, esp wrt proc's dying. */
bool proc_controls(struct proc *actor, struct proc *target)
{
- return target->env_parent_id == actor->env_id;
+ return ((actor == target) || (target->ppid == actor->pid));
}
-/*
- * Dispatches a process to run, either on the current core in the case of a
- * RUNNABLE_S, or on its partition in the case of a RUNNABLE_M.
- * This should never be called to "restart" a core.
- */
+/* Dispatches a process to run, either on the current core in the case of a
+ * RUNNABLE_S, or on its partition in the case of a RUNNABLE_M. This should
+ * never be called to "restart" a core. This expects that the "instructions"
+ * for which core(s) to run this on will be in the vcoremap, which needs to be
+ * set externally.
+ *
+ * When a process goes from RUNNABLE_M to RUNNING_M, its vcoremap will be
+ * "packed" (no holes in the vcore->pcore mapping), vcore0 will continue to run
+ * it's old core0 context, and the other cores will come in at the entry point.
+ * Including in the case of preemption.
+ *
+ * This won't return if the current core is going to be one of the processes
+ * cores (either for _S mode or for _M if it's in the vcoremap). proc_run will
+ * eat your reference if it does not return. */
void proc_run(struct proc *p)
{
+ bool self_ipi_pending = FALSE;
spin_lock_irqsave(&p->proc_lock);
switch (p->state) {
case (PROC_DYING):
spin_unlock_irqsave(&p->proc_lock);
- printk("Process %d not starting due to async death\n", p->env_id);
- // There should be no core cleanup to do (like decref).
- assert(current != p);
+ printk("Process %d not starting due to async death\n", p->pid);
// if we're a worker core, smp_idle, o/w return
if (!management_core())
smp_idle(); // this never returns
return;
case (PROC_RUNNABLE_S):
- proc_set_state(p, PROC_RUNNING_S);
- // We will want to know where this process is running, even if it is
- // only in RUNNING_S. can use the vcoremap, which makes death easy.
- // we may need the pcoremap entry to mark it as a RUNNING_S core, or
- // else update it here. (TODO) (PCORE)
+ __proc_set_state(p, PROC_RUNNING_S);
+ /* We will want to know where this process is running, even if it is
+ * only in RUNNING_S. can use the vcoremap, which makes death easy.
+ * Also, this is the signal used in trap.c to know to save the tf in
+ * env_tf.
+ * We may need the pcoremap entry to mark it as a RUNNING_S core, or
+ * else update it here. (TODO) (PCORE) */
p->num_vcores = 0;
p->vcoremap[0] = core_id();
spin_unlock_irqsave(&p->proc_lock);
- // This normally doesn't return, but might error out in the future.
+ /* Transferring our reference to startcore, where p will become
+ * current. If it already is, decref in advance. This is similar
+ * to __startcore(), in that it sorts out the refcnt accounting. */
+ if (current == p)
+ proc_decref(p, 1);
proc_startcore(p, &p->env_tf);
break;
case (PROC_RUNNABLE_M):
- proc_set_state(p, PROC_RUNNING_M);
/* vcoremap[i] holds the coreid of the physical core allocated to
* this process. It is set outside proc_run. For the active
* message, a0 = struct proc*, a1 = struct trapframe*. */
if (p->num_vcores) {
+ __proc_set_state(p, PROC_RUNNING_M);
+ int i = 0;
+ /* Up the refcnt, since num_vcores are going to start using this
+ * process and have it loaded in their 'current'. */
+ p->env_refcnt += p->num_vcores; // TODO: (REF) use incref
+ /* If the core we are running on is in the vcoremap, we will get
+ * an IPI (once we reenable interrupts) and never return. */
+ if (__get_vcoreid(p->vcoremap, p->num_vcores, core_id()) != -1)
+ self_ipi_pending = TRUE;
// TODO: handle silly state (HSS)
- // set virtual core 0 to run the main context
- send_active_msg_sync(p->vcoremap[0], __startcore, p,
- &p->env_tf, 0);
- /* handle the others. note the sync message will spin until
- * there is a free active message slot, which could lock up the
- * system. think about this. (TODO) */
- for (int i = 1; i < p->num_vcores; i++)
- send_active_msg_sync(p->vcoremap[i], __startcore, p, 0, i);
+ // set virtual core 0 to run the main context on transition
+ if (p->env_flags & PROC_TRANSITION_TO_M) {
+ p->env_flags &= !PROC_TRANSITION_TO_M;
+#ifdef __IVY__
+ send_active_message(p->vcoremap[0], __startcore, p,
+ &p->env_tf, (void *SNT)0);
+#else
+ send_active_message(p->vcoremap[0], (void *)__startcore,
+ (void *)p, (void *)&p->env_tf, 0);
+#endif
+ i = 1; // start at vcore1 in the loop below
+ }
+ /* handle the others. */
+ for (/* i set above */; i < p->num_vcores; i++)
+#ifdef __IVY__
+ send_active_message(p->vcoremap[i], __startcore,
+ p, (trapframe_t *CT(1))NULL, (void *SNT)i);
+#else
+ send_active_message(p->vcoremap[i], (void *)__startcore,
+ (void *)p, (void *)0, (void *)i);
+#endif
+ } else {
+ warn("Tried to proc_run() an _M with no vcores!");
}
- /* There a subtle (attempted) race avoidance here. proc_startcore
- * can handle a death message, but we can't have the startcore come
- * after the death message. Otherwise, it would look like a new
- * process. So we hold the lock to make sure our message went out
- * before a possible death message.
+ /* Unlock and decref/wait for the IPI if one is pending. This will
+ * eat the reference if we aren't returning.
+ *
+ * There a subtle race avoidance here. proc_startcore can handle a
+ * death message, but we can't have the startcore come after the
+ * death message. Otherwise, it would look like a new process. So
+ * we hold the lock til after we send our message, which prevents a
+ * possible death message.
* - Likewise, we need interrupts to be disabled, in case one of the
* messages was for us, and reenable them after letting go of the
* lock. This is done by spin_lock_irqsave, so be careful if you
* change this.
- * - This can also be done far more intelligently / efficiently,
- * like skipping in case it's busy and coming back later.
* - Note there is no guarantee this core's interrupts were on, so
* it may not get the message for a while... */
- spin_unlock_irqsave(&p->proc_lock);
+ __proc_unlock_ipi_pending(p, self_ipi_pending);
break;
default:
spin_unlock_irqsave(&p->proc_lock);
- panic("Invalid process state in proc_run()!!");
+ panic("Invalid process state %p in proc_run()!!", p->state);
}
}
-/*
- * Runs the given context (trapframe) of process p on the core this code
- * executes on. The refcnt tracks how many cores have "an interest" in this
- * process, which so far just means it uses the process's page table. See the
- * massive comments around the incref function
+/* Runs the given context (trapframe) of process p on the core this code
+ * executes on.
*
* Given we are RUNNING_*, an IPI for death or preemption could come in:
* 1. death attempt (IPI to kill whatever is on your core):
* I think we need to make it such that the kernel in "process context" never
* gets removed from the core (displaced from its stack) without going through
* some "bundling" code.
- */
+ *
+ * A note on refcnting: this function will not return, and your proc reference
+ * will end up stored in current. This will make no changes to p's refcnt, so
+ * do your accounting such that there is only the +1 for current. This means if
+ * it is already in current (like in the trap return path), don't up it. If
+ * it's already in current and you have another reference (like pid2proc or from
+ * an IPI), then down it (which is what happens in __startcore()). If it's not
+ * in current and you have one reference, like proc_run(non_current_p), then
+ * also do nothing. The refcnt for your *p will count for the reference stored
+ * in current. */
void proc_startcore(struct proc *p, trapframe_t *tf) {
// it's possible to be DYING, but it's a rare race.
//if (p->state & (PROC_RUNNING_S | PROC_RUNNING_M))
// printk("dying before (re)startcore on core %d\n", core_id());
-
// sucks to have ints disabled when doing env_decref and possibly freeing
disable_irq();
if (per_cpu_info[core_id()].preempt_pending) {
}
/* If the process wasn't here, then we need to load its address space. */
if (p != current) {
- if (proc_incref(p)) {
- // getting here would mean someone tried killing this while we tried
- // to start one of it's contexts (from scratch, o/w we had it's CR3
- // loaded already)
- // if this happens, a no-op death-IPI ought to be on its way... we can
- // just smp_idle()
- smp_idle();
- }
+ /* Do not incref here. We were given the reference to current,
+ * pre-upped. */
lcr3(p->env_cr3);
- // we unloaded the old cr3, so decref it (if it exists)
- // TODO: Consider moving this to wherever we really "mean to leave the
- // process's context".
+ /* This is "leaving the process context" of the previous proc. The
+ * previous lcr3 unloaded the previous proc's context. This should
+ * rarely happen, since we usually proactively leave process context,
+ * but is the fallback. */
if (current)
- proc_decref(current);
- current = p;
+ proc_decref(current, 1);
+ set_current_proc(p);
}
/* need to load our silly state, preferably somewhere other than here so we
* can avoid the case where the context was just running here. it's not
* (Last core/kernel thread to decref cleans up and deallocates resources.)
*
* Note that some cores can be processing async calls, but will eventually
- * decref. Should think about this more.
- */
+ * decref. Should think about this more, like some sort of callback/revocation.
+ *
+ * This will eat your reference if it won't return. Note that this function
+ * needs to change anyways when we make __death more like __preempt. (TODO) */
void proc_destroy(struct proc *p)
{
- // Note this code relies on this lock disabling interrupts, similar to
- // proc_run.
+ /* TODO: this corelist is taking up a lot of space on the stack */
+ uint32_t corelist[MAX_NUM_CPUS];
+ size_t num_cores_freed;
+ bool self_ipi_pending = FALSE;
spin_lock_irqsave(&p->proc_lock);
+
+ /* TODO: (DEATH) look at this again when we sort the __death IPI */
+ if (current == p)
+ self_ipi_pending = TRUE;
+
switch (p->state) {
- case PROC_DYING:
- return; // someone else killed this already.
- case PROC_RUNNABLE_S:
+ case PROC_DYING: // someone else killed this already.
+ __proc_unlock_ipi_pending(p, self_ipi_pending);
+ return;
case PROC_RUNNABLE_M:
+ /* Need to reclaim any cores this proc might have, even though it's
+ * not running yet. */
+ __proc_take_allcores(p, NULL, NULL, NULL, NULL);
+ // fallthrough
+ case PROC_RUNNABLE_S:
// Think about other lists, like WAITING, or better ways to do this
deschedule_proc(p);
break;
// here's how to do it manually
if (current == p) {
lcr3(boot_cr3);
- proc_decref(p); // this decref is for the cr3
+ proc_decref(p, 1); // this decref is for the cr3
current = NULL;
}
#endif
- send_active_msg_sync(p->vcoremap[0], __death, 0, 0, 0);
+ send_active_message(p->vcoremap[0], __death, (void *SNT)0,
+ (void *SNT)0, (void *SNT)0);
#if 0
/* right now, RUNNING_S only runs on a mgmt core (0), not cores
* managed by the idlecoremap. so don't do this yet. */
- spin_lock(&idle_lock);
- idlecoremap[num_idlecores++] = p->vcoremap[0];
- spin_unlock(&idle_lock);
+ put_idle_core(p->vcoremap[0]);
#endif
break;
case PROC_RUNNING_M:
* deallocate the cores.
* The rule is that the vcoremap is set before proc_run, and reset
* within proc_destroy */
- spin_lock(&idle_lock);
- for (int i = 0; i < p->num_vcores; i++) {
- send_active_msg_sync(p->vcoremap[i], __death, 0, 0, 0);
- // give the pcore back to the idlecoremap
- assert(num_idlecores < num_cpus); // sanity
- idlecoremap[num_idlecores++] = p->vcoremap[i];
- p->vcoremap[i] = 0; // TODO: might need a better signal
- }
- spin_unlock(&idle_lock);
+ __proc_take_allcores(p, __death, (void *SNT)0, (void *SNT)0,
+ (void *SNT)0);
break;
default:
- panic("Weird state in proc_destroy");
+ panic("Weird state(0x%08x) in proc_destroy", p->state);
}
- proc_set_state(p, PROC_DYING);
+ __proc_set_state(p, PROC_DYING);
+ /* this decref is for the process in general */
+ p->env_refcnt--; // TODO (REF)
+ //proc_decref(p, 1);
- atomic_dec(&num_envs);
- /* TODO: (REF) dirty hack. decref currently uses a lock, but needs to use
- * CAS instead (another lock would be slightly less ghetto). but we need to
- * decref before releasing the lock, since that could enable interrupts,
- * which would have us receive the DEATH IPI if this was called locally by
- * the target process. */
- //proc_decref(p); // this decref is for the process in general
- p->env_refcnt--;
- size_t refcnt = p->env_refcnt; // need to copy this in so it's not reloaded
+ /* Unlock and possible decref and wait. A death IPI should be on its way,
+ * either from the RUNNING_S one, or from proc_take_cores with a __death.
+ * in general, interrupts should be on when you call proc_destroy locally,
+ * but currently aren't for all things (like traphandlers). */
+ __proc_unlock_ipi_pending(p, self_ipi_pending);
+ return;
+}
+
+/* Helper function. Starting from prev, it will find the next free vcoreid,
+ * which is the next slot with a -1 in it.
+ * You better hold the lock before calling this. */
+static uint32_t get_free_vcoreid(struct proc *SAFE p, uint32_t prev)
+{
+ uint32_t i;
+ for (i = prev; i < MAX_NUM_CPUS; i++)
+ if (p->vcoremap[i] == -1)
+ break;
+ if (i + 1 >= MAX_NUM_CPUS)
+ warn("At the end of the vcorelist. Might want to check that out.");
+ return i;
+}
+
+/* Helper function. Starting from prev, it will find the next busy vcoreid,
+ * which is the next slot with something other than a -1 in it.
+ * You better hold the lock before calling this. */
+static uint32_t get_busy_vcoreid(struct proc *SAFE p, uint32_t prev)
+{
+ uint32_t i;
+ for (i = prev; i < MAX_NUM_CPUS; i++)
+ if (p->vcoremap[i] != -1)
+ break;
+ if (i + 1 >= MAX_NUM_CPUS)
+ warn("At the end of the vcorelist. Might want to check that out.");
+ return i;
+}
+
+/* Helper function. Find the vcoreid for a given physical core id. If we use
+ * some sort of pcoremap, we can avoid this linear search. You better hold the
+ * lock before calling this. Returns -1 on failure. */
+static int32_t __get_vcoreid(int32_t *corelist, size_t num, int32_t pcoreid)
+{
+ int32_t i;
+ bool found = FALSE;
+ for (i = 0; i < num; i++)
+ if (corelist[i] == pcoreid) {
+ found = TRUE;
+ break;
+ }
+ if (found)
+ return i;
+ else
+ return -1;
+}
- /* After unlocking, we can receive a DEATH IPI and clean up */
+/* Helper function. Just like the one above, but this one panics on failure.
+ * You better hold the lock before calling this. */
+static int32_t get_vcoreid(struct proc *SAFE p, int32_t pcoreid)
+{
+ int32_t vcoreid = __get_vcoreid(p->vcoremap, p->num_vcores, pcoreid);
+ assert(vcoreid != -1);
+ return vcoreid;
+}
+
+/* Use this when you are waiting for an IPI that you sent yourself. In most
+ * cases, interrupts should already be on (like after a spin_unlock_irqsave from
+ * process context), but aren't always, like in proc_destroy(). We might be
+ * able to remove the enable_irq in the future. Think about this (TODO).
+ *
+ * Note this means all non-proc management interrupt handlers must return (which
+ * they need to do anyway), so that we get back to this point. */
+static inline void __wait_for_ipi(const char *fnname)
+{
+ enable_irq();
+ udelay(1000000);
+ panic("Waiting too long on core %d for an IPI in %s()!", core_id(), fnname);
+}
+
+/* Yields the calling core. Must be called locally (not async) for now.
+ * - If RUNNING_S, you just give up your time slice and will eventually return.
+ * - If RUNNING_M, you give up the current vcore (which never returns), and
+ * adjust the amount of cores wanted/granted.
+ * - If you have only one vcore, you switch to RUNNABLE_M. When you run again,
+ * you'll have one guaranteed core, starting from the entry point.
+ *
+ * - RES_CORES amt_wanted will be the amount running after taking away the
+ * yielder, unless there are none left, in which case it will be 1.
+ *
+ * This does not return (abandon_core()), so it will eat your reference. */
+void proc_yield(struct proc *SAFE p)
+{
+ spin_lock_irqsave(&p->proc_lock);
+ switch (p->state) {
+ case (PROC_RUNNING_S):
+ p->env_tf= *current_tf;
+ env_push_ancillary_state(p);
+ __proc_set_state(p, PROC_RUNNABLE_S);
+ schedule_proc(p);
+ break;
+ case (PROC_RUNNING_M):
+ p->resources[RES_CORES].amt_granted = --(p->num_vcores);
+ p->resources[RES_CORES].amt_wanted = p->num_vcores;
+ // give up core
+ p->vcoremap[get_vcoreid(p, core_id())] = -1;
+ // add to idle list
+ put_idle_core(core_id());
+ // last vcore? then we really want 1, and to yield the gang
+ if (p->num_vcores == 0) {
+ // might replace this with m_yield, if we have it directly
+ p->resources[RES_CORES].amt_wanted = 1;
+ __proc_set_state(p, PROC_RUNNABLE_M);
+ schedule_proc(p);
+ }
+ break;
+ default:
+ // there are races that can lead to this (async death, preempt, etc)
+ panic("Weird state(0x%08x) in proc_yield", p->state);
+ }
spin_unlock_irqsave(&p->proc_lock);
+ proc_decref(p, 1);
+ /* Clean up the core and idle. For mgmt cores, they will ultimately call
+ * manager, which will call schedule() and will repick the yielding proc. */
+ abandon_core();
+}
- // coupled with the refcnt-- above, from decref. if this happened,
- // proc_destroy was called "remotely", and with no one else refcnting
- if (!refcnt)
- env_free(p);
+/* Gives process p the additional num cores listed in corelist. You must be
+ * RUNNABLE_M or RUNNING_M before calling this. If you're RUNNING_M, this will
+ * startup your new cores at the entry point with their virtual IDs. If you're
+ * RUNNABLE_M, you should call proc_run after this so that the process can start
+ * to use its cores.
+ *
+ * If you're *_S, make sure your core0's TF is set (which is done when coming in
+ * via arch/trap.c and we are RUNNING_S), change your state, then call this.
+ * Then call proc_run().
+ *
+ * The reason I didn't bring the _S cases from core_request over here is so we
+ * can keep this family of calls dealing with only *_Ms, to avoiding caring if
+ * this is called from another core, and to avoid the need_to_idle business.
+ * The other way would be to have this function have the side effect of changing
+ * state, and finding another way to do the need_to_idle.
+ *
+ * The returned bool signals whether or not a stack-crushing IPI will come in
+ * once you unlock after this function.
+ *
+ * WARNING: You must hold the proc_lock before calling this! */
+bool __proc_give_cores(struct proc *SAFE p, int32_t *corelist, size_t num)
+{ TRUSTEDBLOCK
+ bool self_ipi_pending = FALSE;
+ uint32_t free_vcoreid = 0;
+ switch (p->state) {
+ case (PROC_RUNNABLE_S):
+ case (PROC_RUNNING_S):
+ panic("Don't give cores to a process in a *_S state!\n");
+ break;
+ case (PROC_DYING):
+ panic("Attempted to give cores to a DYING process.\n");
+ break;
+ case (PROC_RUNNABLE_M):
+ // set up vcoremap. list should be empty, but could be called
+ // multiple times before proc_running (someone changed their mind?)
+ if (p->num_vcores) {
+ printk("[kernel] Yaaaaaarrrrr! Giving extra cores, are we?\n");
+ // debugging: if we aren't packed, then there's a problem
+ // somewhere, like someone forgot to take vcores after
+ // preempting.
+ for (int i = 0; i < p->num_vcores; i++)
+ assert(p->vcoremap[i]);
+ }
+ // add new items to the vcoremap
+ for (int i = 0; i < num; i++) {
+ // find the next free slot, which should be the next one
+ free_vcoreid = get_free_vcoreid(p, free_vcoreid);
+ printd("setting vcore %d to pcore %d\n", free_vcoreid, corelist[i]);
+ p->vcoremap[free_vcoreid] = corelist[i];
+ p->num_vcores++;
+ }
+ break;
+ case (PROC_RUNNING_M):
+ /* Up the refcnt, since num cores are going to start using this
+ * process and have it loaded in their 'current'. */
+ // TODO: (REF) use proc_incref once we have atomics
+ p->env_refcnt += num;
+ if (__get_vcoreid(corelist, num, core_id()) != -1)
+ self_ipi_pending = TRUE;
+ for (int i = 0; i < num; i++) {
+ free_vcoreid = get_free_vcoreid(p, free_vcoreid);
+ printd("setting vcore %d to pcore %d\n", free_vcoreid, corelist[i]);
+ p->vcoremap[free_vcoreid] = corelist[i];
+ p->num_vcores++;
+ send_active_message(corelist[i], __startcore, p,
+ (struct Trapframe *)0,
+ (void*SNT)free_vcoreid);
+ }
+ break;
+ default:
+ panic("Weird proc state %d in proc_give_cores()!\n", p->state);
+ }
+ return self_ipi_pending;
+}
- /* If we were running the process, we should have received an IPI by now.
- * If not, odds are interrupts are disabled, which shouldn't happen while
- * servicing syscalls. */
- assert(current != p);
- return;
+/* Makes process p's coremap look like corelist (add, remove, etc). Caller
+ * needs to know what cores are free after this call (removed, failed, etc).
+ * This info will be returned via corelist and *num. This will send message to
+ * any cores that are getting removed.
+ *
+ * Before implementing this, we should probably think about when this will be
+ * used. Implies preempting for the message. The more that I think about this,
+ * the less I like it. For now, don't use this, and think hard before
+ * implementing it.
+ *
+ * WARNING: You must hold the proc_lock before calling this! */
+bool __proc_set_allcores(struct proc *SAFE p, int32_t *corelist,
+ size_t *num, amr_t message,TV(a0t) arg0,
+ TV(a1t) arg1, TV(a2t) arg2)
+{
+ panic("Set all cores not implemented.\n");
}
-/*
- * The process refcnt is the number of places the process 'exists' in the
- * system. Creation counts as 1. Having your page tables loaded somewhere
- * (lcr3) counts as another 1. A non-RUNNING_* process should have refcnt at
- * least 1. If the kernel is on another core and in a processes address space
- * (like processing its backring), that counts as another 1.
+/* Takes from process p the num cores listed in corelist, using the given
+ * message for the active message (__death, __preempt, etc). Like the others
+ * in this function group, bool signals whether or not an IPI is pending.
*
- * Note that the actual loading and unloading of cr3 is up to the caller, since
- * that's not the only use for this (and decoupling is more flexible).
+ * WARNING: You must hold the proc_lock before calling this! */
+bool __proc_take_cores(struct proc *SAFE p, int32_t *corelist,
+ size_t num, amr_t message, TV(a0t) arg0,
+ TV(a1t) arg1, TV(a2t) arg2)
+{ TRUSTEDBLOCK
+ uint32_t vcoreid;
+ bool self_ipi_pending = FALSE;
+ switch (p->state) {
+ case (PROC_RUNNABLE_M):
+ assert(!message);
+ break;
+ case (PROC_RUNNING_M):
+ assert(message);
+ break;
+ default:
+ panic("Weird state %d in proc_take_cores()!\n", p->state);
+ }
+ spin_lock(&idle_lock);
+ assert((num <= p->num_vcores) && (num_idlecores + num <= num_cpus));
+ spin_unlock(&idle_lock);
+ for (int i = 0; i < num; i++) {
+ vcoreid = get_vcoreid(p, corelist[i]);
+ assert(p->vcoremap[vcoreid] == corelist[i]);
+ if (message) {
+ if (p->vcoremap[vcoreid] == core_id())
+ self_ipi_pending = TRUE;
+ send_active_message(corelist[i], message, arg0, arg1, arg2);
+ }
+ // give the pcore back to the idlecoremap
+ put_idle_core(corelist[i]);
+ p->vcoremap[vcoreid] = -1;
+ }
+ p->num_vcores -= num;
+ p->resources[RES_CORES].amt_granted -= num;
+ return self_ipi_pending;
+}
+
+/* Takes all cores from a process, which must be in an _M state. Cores are
+ * placed back in the idlecoremap. If there's a message, such as __death or
+ * __preempt, it will be sent to the cores. The bool signals whether or not an
+ * IPI is coming in once you unlock.
*
- * The refcnt should always be greater than 0 for processes that aren't dying.
- * When refcnt is 0, the process is dying and should not allow any more increfs.
- * A process can be dying with a refcnt greater than 0, since it could be
- * waiting for other cores to "get the message" to die, or a kernel core can be
- * finishing work in the processes's address space.
+ * WARNING: You must hold the proc_lock before calling this! */
+bool __proc_take_allcores(struct proc *SAFE p, amr_t message,
+ TV(a0t) arg0, TV(a1t) arg1, TV(a2t) arg2)
+{
+ uint32_t active_vcoreid = 0;
+ bool self_ipi_pending = FALSE;
+ switch (p->state) {
+ case (PROC_RUNNABLE_M):
+ assert(!message);
+ break;
+ case (PROC_RUNNING_M):
+ assert(message);
+ break;
+ default:
+ panic("Weird state %d in proc_take_allcores()!\n", p->state);
+ }
+ spin_lock(&idle_lock);
+ assert(num_idlecores + p->num_vcores <= num_cpus); // sanity
+ spin_unlock(&idle_lock);
+ for (int i = 0; i < p->num_vcores; i++) {
+ // find next active vcore
+ active_vcoreid = get_busy_vcoreid(p, active_vcoreid);
+ if (message) {
+ if (p->vcoremap[active_vcoreid] == core_id())
+ self_ipi_pending = TRUE;
+ send_active_message(p->vcoremap[active_vcoreid], message,
+ arg0, arg1, arg2);
+ }
+ // give the pcore back to the idlecoremap
+ put_idle_core(p->vcoremap[active_vcoreid]);
+ p->vcoremap[active_vcoreid] = -1;
+ }
+ p->num_vcores = 0;
+ p->resources[RES_CORES].amt_granted = 0;
+ return self_ipi_pending;
+}
+
+/* Helper, to be used when unlocking after calling the above functions that
+ * might cause an IPI to be sent. TODO inline this, so the __FUNCTION__ works.
+ * Will require an overhaul of core_request (break it up, etc) */
+void __proc_unlock_ipi_pending(struct proc *p, bool ipi_pending)
+{
+ if (ipi_pending) {
+ p->env_refcnt--; // TODO: (REF) (atomics)
+ spin_unlock_irqsave(&p->proc_lock);
+ __wait_for_ipi(__FUNCTION__);
+ } else {
+ spin_unlock_irqsave(&p->proc_lock);
+ }
+}
+
+
+/* This takes a referenced process and ups the refcnt by count. If the refcnt
+ * was already 0, then someone has a bug, so panic. Check out the Documentation
+ * for brutal details about refcnting.
*
* Implementation aside, the important thing is that we atomically increment
- * only if it wasn't already 0. If it was 0, then we shouldn't be attaching to
- * the process, so we return an error, which should be handled however is
- * appropriate. We currently use spinlocks, but some sort of clever atomics
- * would work too.
- *
- * Also, no one should ever update the refcnt outside of these functions.
- * Eventually, we'll have Ivy support for this. (TODO)
+ * only if it wasn't already 0. If it was 0, panic.
*
- * TODO: (REF) change to use CAS.
- */
-error_t proc_incref(struct proc *p)
+ * TODO: (REF) change to use CAS / atomics. */
+void proc_incref(struct proc *p, size_t count)
{
- error_t retval = 0;
spin_lock_irqsave(&p->proc_lock);
if (p->env_refcnt)
- p->env_refcnt++;
+ p->env_refcnt += count;
else
- retval = -EBADENV;
+ panic("Tried to incref a proc with no existing refernces!");
spin_unlock_irqsave(&p->proc_lock);
- return retval;
}
-/*
- * When the kernel is done with a process, it decrements its reference count.
- * When the count hits 0, no one is using it and it should be freed.
- * "Last one out" actually finalizes the death of the process. This is tightly
- * coupled with the previous function (incref)
- * Be sure to load a different cr3 before calling this!
+/* When the kernel is done with a process, it decrements its reference count.
+ * When the count hits 0, no one is using it and it should be freed. "Last one
+ * out" actually finalizes the death of the process. This is tightly coupled
+ * with the previous function (incref)
*
* TODO: (REF) change to use CAS. Note that when we do so, we may be holding
- * the process lock when calling env_free().
- */
-void proc_decref(struct proc *p)
+ * the process lock when calling __proc_free(). */
+void proc_decref(struct proc *p, size_t count)
{
spin_lock_irqsave(&p->proc_lock);
- p->env_refcnt--;
+ p->env_refcnt -= count;
size_t refcnt = p->env_refcnt; // need to copy this in so it's not reloaded
spin_unlock_irqsave(&p->proc_lock);
// if we hit 0, no one else will increment and we can check outside the lock
if (!refcnt)
- env_free(p);
+ __proc_free(p);
+ if (refcnt < 0)
+ panic("Too many decrefs!");
}
/* Active message handler to start a process's context on this core. Tightly
* coupled with proc_run() */
-void __startcore(trapframe_t *tf, uint32_t srcid, uint32_t a0, uint32_t a1,
- uint32_t a2)
+#ifdef __IVY__
+void __startcore(trapframe_t *tf, uint32_t srcid, struct proc *CT(1) a0,
+ trapframe_t *CT(1) a1, void *SNT a2)
+#else
+void __startcore(trapframe_t *tf, uint32_t srcid, void * a0, void * a1,
+ void * a2)
+#endif
{
uint32_t coreid = core_id();
- struct proc *p_to_run = (struct proc *SAFE) TC(a0);
+ struct proc *p_to_run = (struct proc *CT(1))a0;
trapframe_t local_tf;
- trapframe_t *tf_to_pop = (trapframe_t *SAFE) TC(a1);
+ trapframe_t *tf_to_pop = (trapframe_t *CT(1))a1;
- printk("Startcore on core %d\n", coreid);
+ printk("[kernel] Startcore on physical core %d for Process %d\n",
+ coreid, p_to_run->pid);
assert(p_to_run);
// TODO: handle silly state (HSS)
if (!tf_to_pop) {
tf_to_pop = &local_tf;
memset(tf_to_pop, 0, sizeof(*tf_to_pop));
- proc_init_trapframe(tf_to_pop);
+ proc_init_trapframe(tf_to_pop,(uint32_t)a2);
// Note the init_tf sets tf_to_pop->tf_esp = USTACKTOP;
- proc_set_tfcoreid(tf_to_pop, a2);
proc_set_program_counter(tf_to_pop, p_to_run->env_entry);
}
+ /* the sender of the amsg increfed, thinking we weren't running current. */
+ if (p_to_run == current)
+ proc_decref(p_to_run, 1);
proc_startcore(p_to_run, tf_to_pop);
}
-/* Active message handler to stop running whatever context is on this core and
- * to idle. Note this leaves no trace of what was running.
- * It's okay if death comes to a core that's already idling and has no current.
- * It could happen if a process decref'd before proc_startcore could incref. */
-void __death(trapframe_t *tf, uint32_t srcid, uint32_t a0, uint32_t a1,
- uint32_t a2)
+/* Stop running whatever context is on this core, load a known-good cr3, and
+ * 'idle'. Note this leaves no trace of what was running. This "leaves the
+ * process's context. */
+void abandon_core(void)
{
/* If we are currently running an address space on our core, we need a known
- * good pgdir before releasing the old one. This is currently the major
- * practical implication of the kernel caring about a processes existence
- * (the inc and decref). This decref corresponds to the incref in
- * proc_startcore (though it's not the only one). */
+ * good pgdir before releasing the old one. We decref, since current no
+ * longer tracks the proc (and current no longer protects the cr3). */
if (current) {
lcr3(boot_cr3);
- proc_decref(current);
- current = NULL;
+ proc_decref(current, 1);
+ set_current_proc(NULL);
}
smp_idle();
}
+/* Active message handler to clean up the core when a process is dying.
+ * Note this leaves no trace of what was running.
+ * It's okay if death comes to a core that's already idling and has no current.
+ * It could happen if a process decref'd before proc_startcore could incref. */
+void __death(trapframe_t *tf, uint32_t srcid, void *SNT a0, void *SNT a1,
+ void *SNT a2)
+{
+ abandon_core();
+}
+
void print_idlecoremap(void)
{
spin_lock(&idle_lock);
+ printk("There are %d idle cores.\n", num_idlecores);
for (int i = 0; i < num_idlecores; i++)
printk("idlecoremap[%d] = %d\n", i, idlecoremap[i]);
spin_unlock(&idle_lock);
}
+
+void print_allpids(void)
+{
+ spin_lock(&pid_hash_lock);
+ if (hashtable_count(pid_hash)) {
+ hashtable_itr_t *phtable_i = hashtable_iterator(pid_hash);
+ printk("PID STATE \n");
+ printk("------------------\n");
+ do {
+ struct proc* p = hashtable_iterator_value(phtable_i);
+ printk("%8d %s\n", hashtable_iterator_key(phtable_i),p ? procstate2str(p->state) : "(null)");
+ } while (hashtable_iterator_advance(phtable_i));
+ }
+ spin_unlock(&pid_hash_lock);
+}
+
+void print_proc_info(pid_t pid)
+{
+ int j = 0;
+ struct proc *p = pid2proc(pid);
+ // not concerned with a race on the state...
+ if (!p) {
+ printk("Bad PID.\n");
+ return;
+ }
+ spinlock_debug(&p->proc_lock);
+ spin_lock_irqsave(&p->proc_lock);
+ printk("struct proc: %p\n", p);
+ printk("PID: %d\n", p->pid);
+ printk("PPID: %d\n", p->ppid);
+ printk("State: 0x%08x\n", p->state);
+ printk("Refcnt: %d\n", p->env_refcnt - 1); // don't report our ref
+ printk("Flags: 0x%08x\n", p->env_flags);
+ printk("CR3(phys): 0x%08x\n", p->env_cr3);
+ printk("Num Vcores: %d\n", p->num_vcores);
+ printk("Vcoremap:\n");
+ for (int i = 0; i < p->num_vcores; i++) {
+ j = get_busy_vcoreid(p, j);
+ printk("\tVcore %d: Pcore %d\n", j, p->vcoremap[j]);
+ j++;
+ }
+ printk("Resources:\n");
+ for (int i = 0; i < MAX_NUM_RESOURCES; i++)
+ printk("\tRes type: %02d, amt wanted: %08d, amt granted: %08d\n", i,
+ p->resources[i].amt_wanted, p->resources[i].amt_granted);
+ printk("Vcore 0's Last Trapframe:\n");
+ print_trapframe(&p->env_tf);
+ spin_unlock_irqsave(&p->proc_lock);
+ proc_decref(p, 1); /* decref for the pid2proc reference */
+}
projects / akaros.git / blobdiff