blob: ca4b726a75b12c0fc1b948bd91477fbe43186705 [file] [log] [blame]
/* Copyright (c) 2010-13 The Regents of the University of California
* Barret Rhoden <brho@cs.berkeley.edu>
* See LICENSE for details.
*
* Kernel threading. These are for blocking within the kernel for whatever
* reason, usually during blocking IO operations. */
#include <kthread.h>
#include <slab.h>
#include <page_alloc.h>
#include <pmap.h>
#include <smp.h>
#include <schedule.h>
#include <kstack.h>
#include <kmalloc.h>
#include <arch/uaccess.h>
#define KSTACK_NR_GUARD_PGS 1
#define KSTACK_GUARD_SZ (KSTACK_NR_GUARD_PGS * PGSIZE)
static struct kmem_cache *kstack_cache;
/* We allocate KSTKSIZE + PGSIZE vaddrs. So for one-page stacks, we get two
* pages. blob points to the bottom of this space. Our job is to allocate the
* physical pages for the stack and set up the virtual-to-physical mappings. */
int kstack_ctor(void *blob, void *priv, int flags)
{
void *stackbot;
stackbot = kpages_alloc(KSTKSIZE, flags);
if (!stackbot)
return -1;
if (map_vmap_segment((uintptr_t)blob, 0x123456000, KSTACK_NR_GUARD_PGS,
PTE_NONE))
goto error;
if (map_vmap_segment((uintptr_t)blob + KSTACK_GUARD_SZ, PADDR(stackbot),
KSTKSIZE / PGSIZE, PTE_KERN_RW))
goto error;
return 0;
error:
/* On failure, we only need to undo what our dtor would do. The unmaps
* happen in the vmap_arena ffunc. */
kpages_free(stackbot, KSTKSIZE);
return -1;
}
/* The vmap_arena free will unmap the vaddrs on its own. We just need to free
* the physical memory we allocated in ctor. Although we still have mappings
* and TLB entries pointing to the memory after we free it (and thus it can be
* reused), this is no more dangerous than just freeing the stack. Errant
* pointers into an old kstack are still dangerous. */
void kstack_dtor(void *blob, void *priv)
{
void *stackbot;
pte_t pte;
pte = pgdir_walk(boot_pgdir, blob + KSTACK_GUARD_SZ, 0);
assert(pte_walk_okay(pte));
stackbot = KADDR(pte_get_paddr(pte));
kpages_free(stackbot, KSTKSIZE);
}
uintptr_t get_kstack(void)
{
void *blob;
blob = kmem_cache_alloc(kstack_cache, MEM_ATOMIC);
/* TODO: think about MEM_WAIT within kthread/blocking code. */
assert(blob);
return (uintptr_t)blob + KSTKSIZE + KSTACK_GUARD_SZ;
}
void put_kstack(uintptr_t stacktop)
{
kmem_cache_free(kstack_cache, (void*)(stacktop - KSTKSIZE
- KSTACK_GUARD_SZ));
}
uintptr_t *kstack_bottom_addr(uintptr_t stacktop)
{
/* canary at the bottom of the stack */
assert(!PGOFF(stacktop));
return (uintptr_t*)(stacktop - KSTKSIZE);
}
struct kmem_cache *kthread_kcache;
void kthread_init(void)
{
kthread_kcache = kmem_cache_create("kthread", sizeof(struct kthread),
__alignof__(struct kthread), 0,
NULL, 0, 0, NULL);
kstack_cache = kmem_cache_create("kstack", KSTKSIZE + KSTACK_GUARD_SZ,
PGSIZE, 0, vmap_arena, kstack_ctor,
kstack_dtor, NULL);
}
/* Used by early init routines (smp_boot, etc) */
struct kthread *__kthread_zalloc(void)
{
struct kthread *kthread;
kthread = kmem_cache_alloc(kthread_kcache, 0);
assert(kthread);
memset(kthread, 0, sizeof(struct kthread));
return kthread;
}
/* Helper during early boot, where we jump from the bootstack to a real kthread
* stack, then run f(). Note that we don't have a kthread yet (done in smp.c).
*
* After this, our callee (f) can free the bootstack, if we care, by adding it
* to the base arena (use the KERNBASE addr, not the KERN_LOAD_ADDR). */
void __use_real_kstack(void (*f)(void *arg))
{
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
uintptr_t new_stacktop;
new_stacktop = get_kstack();
set_stack_top(new_stacktop);
__reset_stack_pointer(0, new_stacktop, f);
}
/* Starts kthread on the calling core. This does not return, and will handle
* the details of cleaning up whatever is currently running (freeing its stack,
* etc). Pairs with sem_down(). */
void restart_kthread(struct kthread *kthread)
{
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
uintptr_t current_stacktop;
struct kthread *cur_kth;
struct proc *old_proc;
/* Avoid messy complications. The kthread will enable_irqsave() when it
* comes back up. */
disable_irq();
/* Free any spare, since we need the current to become the spare.
* Without the spare, we can't free our current kthread/stack (we could
* free the kthread, but not the stack, since we're still on it). And
* we can't free anything after popping kthread, since we never return.
* */
if (pcpui->spare) {
put_kstack(pcpui->spare->stacktop);
kmem_cache_free(kthread_kcache, pcpui->spare);
}
cur_kth = pcpui->cur_kthread;
current_stacktop = cur_kth->stacktop;
assert(!cur_kth->sysc); /* catch bugs, prev user should clear */
/* Set the spare stuff (current kthread, which includes its stacktop) */
pcpui->spare = cur_kth;
/* When a kthread runs, its stack is the default kernel stack */
set_stack_top(kthread->stacktop);
pcpui->cur_kthread = kthread;
/* Only change current if we need to (the kthread was in process
* context) */
if (kthread->proc) {
if (kthread->proc == pcpui->cur_proc) {
/* We're already loaded, but we do need to drop the
* extra ref stored in kthread->proc. */
proc_decref(kthread->proc);
kthread->proc = 0;
} else {
/* Load our page tables before potentially decreffing
* cur_proc.
*
* We don't need to do an EPT flush here. The EPT is
* flushed and managed in sync with the VMCS. We won't
* run a different VM (and thus *need* a different EPT)
* without first removing the old GPC, which ultimately
* will result in a flushed EPT (on x86, this actually
* happens when we clear_owning_proc()). */
lcr3(kthread->proc->env_cr3);
/* Might have to clear out an existing current. If they
* need to be set later (like in restartcore), it'll be
* done on demand. */
old_proc = pcpui->cur_proc;
/* Transfer our counted ref from kthread->proc to
* cur_proc. */
pcpui->cur_proc = kthread->proc;
kthread->proc = 0;
if (old_proc)
proc_decref(old_proc);
}
}
/* Finally, restart our thread */
longjmp(&kthread->context, 1);
}
/* Kmsg handler to launch/run a kthread. This must be a routine message, since
* it does not return. */
static void __launch_kthread(uint32_t srcid, long a0, long a1, long a2)
{
struct kthread *kthread = (struct kthread*)a0;
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
struct proc *cur_proc = pcpui->cur_proc;
if (pcpui->owning_proc && pcpui->owning_proc != kthread->proc) {
/* Some process should be running here that is not the same as
* the kthread. This means the _M is getting interrupted or
* otherwise delayed. If we want to do something other than run
* it (like send the kmsg to another pcore, or ship the context
* from here to somewhere else/deschedule it (like for an _S)),
* do it here.
*
* If you want to do something here, call out to the ksched,
* then abandon_core(). */
cmb(); /* do nothing/placeholder */
}
/* o/w, just run the kthread. any trapframes that are supposed to run
* or were interrupted will run whenever the kthread smp_idles() or
* otherwise finishes. */
restart_kthread(kthread);
assert(0);
}
/* Call this when a kthread becomes runnable/unblocked. We don't do anything
* particularly smart yet, but when we do, we can put it here. */
void kthread_runnable(struct kthread *kthread)
{
int dst;
/* TODO: KSCHED - this is a scheduling decision. The kthread can be
* woken up by threads from somewhat unrelated processes. Consider
* unlocking a sem or kicking an RV from an MCP's syscall. Where was
* this kthread running before? Did it belong to the MCP? Is the
* kthread from an old MCP that was on this core, but there is now a new
* MCP? (This can happen with alarms, currently).
*
* For ktasks, they tend to sleep on an RV forever. Once they migrate
* to a core other than core 0 due to blocking on a qlock/sem, they will
* tend to stay on that core forever, interfering with an unrelated MCP.
*
* We could consider some sort of core affinity, but for now, we can
* just route all ktasks to core 0. Note this may hide some bugs that
* would otherwise be exposed by running in parallel. */
if (is_ktask(kthread))
dst = 0;
else
dst = core_id();
send_kernel_message(dst, __launch_kthread, (long)kthread, 0, 0,
KMSG_ROUTINE);
}
/* Stop the current kthread. It'll get woken up next time we run routine kmsgs,
* after all existing kmsgs are processed. */
void kthread_yield(void)
{
struct semaphore local_sem, *sem = &local_sem;
sem_init(sem, 0);
run_as_rkm(sem_up, sem);
sem_down(sem);
}
void kthread_usleep(uint64_t usec)
{
ERRSTACK(1);
/* TODO: classic ksched issue: where do we want the wake up to happen?
*/
struct timer_chain *tchain = &per_cpu_info[core_id()].tchain;
struct rendez rv;
int ret_zero(void *ignored)
{
return 0;
}
/* "discard the error" style (we run the conditional code) */
if (!waserror()) {
rendez_init(&rv);
rendez_sleep_timeout(&rv, ret_zero, 0, usec);
}
poperror();
}
static void __ktask_wrapper(uint32_t srcid, long a0, long a1, long a2)
{
ERRSTACK(1);
void (*fn)(void*) = (void (*)(void*))a0;
void *arg = (void*)a1;
char *name = (char*)a2;
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
assert(is_ktask(pcpui->cur_kthread));
pcpui->cur_kthread->name = name;
/* There are some rendezs out there that aren't wrapped. Though no one
* can abort them. Yet. */
if (waserror()) {
printk("Ktask %s threw error %s\n", name, current_errstr());
goto out;
}
enable_irq();
fn(arg);
out:
disable_irq();
pcpui->cur_kthread->name = 0;
poperror();
/* if we blocked, when we return, PRKM will smp_idle() */
}
/* Creates a kernel task, running fn(arg), named "name". This is just a routine
* kernel message that happens to have a name, and is allowed to block. It
* won't be associated with any process. For lack of a better place, we'll just
* start it on the calling core. Caller (and/or fn) need to deal with the
* storage for *name. */
void ktask(char *name, void (*fn)(void*), void *arg)
{
send_kernel_message(core_id(), __ktask_wrapper, (long)fn, (long)arg,
(long)name, KMSG_ROUTINE);
}
/* Semaphores, using kthreads directly */
static void db_blocked_kth(struct kth_db_info *db);
static void db_unblocked_kth(struct kth_db_info *db);
static void db_init(struct kth_db_info *db, int type);
static void sem_init_common(struct semaphore *sem, int signals)
{
TAILQ_INIT(&sem->waiters);
sem->nr_signals = signals;
db_init(&sem->db, KTH_DB_SEM);
}
void sem_init(struct semaphore *sem, int signals)
{
sem_init_common(sem, signals);
spinlock_init(&sem->lock);
}
void sem_init_irqsave(struct semaphore *sem, int signals)
{
sem_init_common(sem, signals);
spinlock_init_irqsave(&sem->lock);
}
bool sem_trydown_bulk(struct semaphore *sem, int nr_signals)
{
bool ret = FALSE;
/* lockless peek */
if (sem->nr_signals - nr_signals < 0)
return ret;
spin_lock(&sem->lock);
if (sem->nr_signals - nr_signals >= 0) {
sem->nr_signals--;
ret = TRUE;
}
spin_unlock(&sem->lock);
return ret;
}
bool sem_trydown(struct semaphore *sem)
{
return sem_trydown_bulk(sem, 1);
}
/* Bottom-half of sem_down. This is called after we jumped to the new stack. */
static void __attribute__((noreturn)) __sem_unlock_and_idle(void *arg)
{
struct semaphore *sem = (struct semaphore*)arg;
spin_unlock(&sem->lock);
smp_idle();
}
static void pre_block_check(int nr_locks)
{
struct per_cpu_info *pcpui = this_pcpui_ptr();
assert(can_block(pcpui));
/* Make sure we aren't holding any locks (only works if SPINLOCK_DEBUG)
*/
if (pcpui->lock_depth > nr_locks)
panic("Kthread tried to sleep, with lockdepth %d\n", pcpui->lock_depth);
}
static struct kthread *save_kthread_ctx(void)
{
struct kthread *kthread, *new_kthread;
register uintptr_t new_stacktop;
struct per_cpu_info *pcpui = this_pcpui_ptr();
assert(pcpui->cur_kthread);
/* We're probably going to sleep, so get ready. We'll check again
* later. */
kthread = pcpui->cur_kthread;
/* We need to have a spare slot for restart, so we also use it when
* sleeping. Right now, we need a new kthread to take over if/when our
* current kthread sleeps. Use the spare, and if not, get a new one.
*
* Note we do this with interrupts disabled (which protects us from
* concurrent modifications). */
if (pcpui->spare) {
new_kthread = pcpui->spare;
new_stacktop = new_kthread->stacktop;
pcpui->spare = 0;
/* The old flags could have KTH_IS_KTASK set. The reason is
* that the launching of blocked kthreads also uses PRKM, and
* that KMSG (__launch_kthread) doesn't return. Thus the
* soon-to-be spare kthread, that is launching another, has
* flags & KTH_IS_KTASK set. */
new_kthread->flags = KTH_DEFAULT_FLAGS;
new_kthread->proc = 0;
new_kthread->name = 0;
} else {
new_kthread = __kthread_zalloc();
new_kthread->flags = KTH_DEFAULT_FLAGS;
new_stacktop = get_kstack();
new_kthread->stacktop = new_stacktop;
}
/* Set the core's new default stack and kthread */
set_stack_top(new_stacktop);
pcpui->cur_kthread = new_kthread;
/* Kthreads that are ktasks are not related to any process, and do not
* need to work in a process's address space. They can operate in any
* address space that has the kernel mapped (like boot_pgdir, or any
* pgdir). Some ktasks may switch_to, at which point they do care about
* the address space and must maintain a reference.
*
* Normal kthreads need to stay in the process context, but we want the
* core (which could be a vcore) to stay in the context too. */
if ((kthread->flags & KTH_SAVE_ADDR_SPACE) && current) {
kthread->proc = current;
/* In the future, we could check owning_proc. If it isn't set,
* we could clear current and transfer the refcnt to
* kthread->proc. If so, we'll need to reset the cr3 to
* something (boot_cr3 or owning_proc's cr3), which might not be
* worth the potentially excessive TLB flush. */
proc_incref(kthread->proc, 1);
} else {
assert(kthread->proc == 0);
}
return kthread;
}
static void unsave_kthread_ctx(struct kthread *kthread)
{
struct per_cpu_info *pcpui = this_pcpui_ptr();
struct kthread *new_kthread = pcpui->cur_kthread;
printd("[kernel] Didn't sleep, unwinding...\n");
/* Restore the core's current and default stacktop */
if (kthread->flags & KTH_SAVE_ADDR_SPACE) {
proc_decref(kthread->proc);
kthread->proc = 0;
}
set_stack_top(kthread->stacktop);
pcpui->cur_kthread = kthread;
/* Save the allocs as the spare */
assert(!pcpui->spare);
pcpui->spare = new_kthread;
}
/* This downs the semaphore and suspends the current kernel context on its
* waitqueue if there are no pending signals. */
void sem_down(struct semaphore *sem)
{
bool irqs_were_on = irq_is_enabled();
struct kthread *kthread;
pre_block_check(0);
/* Try to down the semaphore. If there is a signal there, we can skip
* all of the sleep prep and just return. */
#ifdef CONFIG_SEM_SPINWAIT
for (int i = 0; i < CONFIG_SEM_SPINWAIT_NR_LOOPS; i++) {
if (sem_trydown(sem))
goto block_return_path;
cpu_relax();
}
#else
if (sem_trydown(sem))
goto block_return_path;
#endif
kthread = save_kthread_ctx();
if (setjmp(&kthread->context))
goto block_return_path;
spin_lock(&sem->lock);
sem->nr_signals -= 1;
if (sem->nr_signals < 0) {
TAILQ_INSERT_TAIL(&sem->waiters, kthread, link);
db_blocked_kth(&sem->db);
/* At this point, we know we'll sleep and change stacks. Once
* we unlock the sem, we could have the kthread restarted
* (possibly on another core), so we need to leave the old stack
* before unlocking. If we don't and we stay on the stack, then
* if we take an IRQ or NMI (NMI that doesn't change stacks,
* unlike x86_64), we'll be using the stack at the same time as
* the kthread. We could just disable IRQs, but that wouldn't
* protect us from NMIs that don't change stacks. */
__reset_stack_pointer(sem, current_kthread->stacktop,
__sem_unlock_and_idle);
assert(0);
}
spin_unlock(&sem->lock);
unsave_kthread_ctx(kthread);
block_return_path:
printd("[kernel] Returning from being 'blocked'! at %llu\n", read_tsc());
/* restart_kthread and longjmp did not reenable IRQs. We need to make
* sure irqs are on if they were on when we started to block. If they
* were already on and we short-circuited the block, it's harmless to
* reenable them. */
if (irqs_were_on)
enable_irq();
}
void sem_down_bulk(struct semaphore *sem, int nr_signals)
{
/* This is far from ideal. Our current sem code expects a 1:1 pairing
* of signals to waiters. For instance, if we have 10 waiters of -1
* each or 1 waiter of -10, we can't tell from looking at the overall
* structure. We'd need to track the desired number of signals per
* waiter.
*
* Note that if there are a bunch of signals available, sem_down will
* quickly do a try_down and return, so we won't block repeatedly. But
* if we do block, we could wake up N times. */
for (int i = 0; i < nr_signals; i++)
sem_down(sem);
}
/* Ups the semaphore. If it was < 0, we need to wake up someone, which we do.
* Returns TRUE if we woke someone, FALSE o/w (used for debugging in some
* places). If we need more control, we can implement a version of the old
* __up_sem() again. */
bool sem_up(struct semaphore *sem)
{
struct kthread *kthread = 0;
spin_lock(&sem->lock);
if (sem->nr_signals++ < 0) {
assert(!TAILQ_EMPTY(&sem->waiters));
/* could do something with 'priority' here */
kthread = TAILQ_FIRST(&sem->waiters);
TAILQ_REMOVE(&sem->waiters, kthread, link);
db_unblocked_kth(&sem->db);
} else {
assert(TAILQ_EMPTY(&sem->waiters));
}
spin_unlock(&sem->lock);
/* Note that once we call kthread_runnable(), we cannot touch the sem
* again. Some sems are on stacks. The caller can touch sem, if it
* knows about the memory/usage of the sem. Likewise, we can't touch
* the kthread either. */
if (kthread) {
kthread_runnable(kthread);
return TRUE;
}
return FALSE;
}
bool sem_trydown_bulk_irqsave(struct semaphore *sem, int nr_signals)
{
bool ret;
int8_t irq_state = 0;
disable_irqsave(&irq_state);
ret = sem_trydown_bulk(sem, nr_signals);
enable_irqsave(&irq_state);
return ret;
}
bool sem_trydown_irqsave(struct semaphore *sem)
{
return sem_trydown_bulk_irqsave(sem, 1);
}
void sem_down_bulk_irqsave(struct semaphore *sem, int nr_signals)
{
int8_t irq_state = 0;
disable_irqsave(&irq_state);
sem_down_bulk(sem, nr_signals);
enable_irqsave(&irq_state);
}
void sem_down_irqsave(struct semaphore *sem)
{
sem_down_bulk_irqsave(sem, 1);
}
bool sem_up_irqsave(struct semaphore *sem)
{
bool retval;
int8_t irq_state = 0;
disable_irqsave(&irq_state);
retval = sem_up(sem);
enable_irqsave(&irq_state);
return retval;
}
/* Sem debugging */
#ifdef CONFIG_SEMAPHORE_DEBUG
static struct kth_db_tailq objs_with_waiters =
TAILQ_HEAD_INITIALIZER(objs_with_waiters);
static spinlock_t objs_with_waiters_lock = SPINLOCK_INITIALIZER_IRQSAVE;
static struct kthread_tailq *db_get_waiters(struct kth_db_info *db)
{
struct semaphore *sem;
struct cond_var *cv;
switch (db->type) {
case KTH_DB_SEM:
return &container_of(db, struct semaphore, db)->waiters;
case KTH_DB_CV:
return &container_of(db, struct cond_var, db)->waiters;
}
panic("Bad type %d in db %p\n", db->type, db);
}
static spinlock_t *db_get_spinlock(struct kth_db_info *db)
{
struct semaphore *sem;
struct cond_var *cv;
switch (db->type) {
case KTH_DB_SEM:
return &container_of(db, struct semaphore, db)->lock;
case KTH_DB_CV:
return container_of(db, struct cond_var, db)->lock;
}
panic("Bad type %d in db %p\n", db->type, db);
}
static void db_blocked_kth(struct kth_db_info *db)
{
spin_lock_irqsave(&objs_with_waiters_lock);
if (!db->on_list) {
TAILQ_INSERT_HEAD(&objs_with_waiters, db, link);
db->on_list = true;
}
spin_unlock_irqsave(&objs_with_waiters_lock);
}
static void db_unblocked_kth(struct kth_db_info *db)
{
spin_lock_irqsave(&objs_with_waiters_lock);
if (TAILQ_EMPTY(db_get_waiters(db))) {
TAILQ_REMOVE(&objs_with_waiters, db, link);
db->on_list = false;
}
spin_unlock_irqsave(&objs_with_waiters_lock);
}
static void db_init(struct kth_db_info *db, int type)
{
db->type = type;
db->on_list = false;
}
static bool __obj_has_pid(struct kth_db_info *db, pid_t pid)
{
struct kthread *kth_i;
if (pid == -1)
return true;
TAILQ_FOREACH(kth_i, db_get_waiters(db), link) {
if (kth_i->proc) {
if (kth_i->proc->pid == pid)
return true;
} else {
if (pid == 0)
return true;
}
}
return false;
}
static void db_print_obj(struct kth_db_info *db, pid_t pid)
{
struct kthread *kth_i;
/* Always safe to irqsave. We trylock, since the lock ordering is
* obj_lock
* -> list_lock. */
if (!spin_trylock_irqsave(db_get_spinlock(db)))
return;
if (!__obj_has_pid(db, pid)) {
spin_unlock_irqsave(db_get_spinlock(db));
return;
}
printk("Object %p (%3s):\n", db, db->type == KTH_DB_SEM ? "sem" :
db->type == KTH_DB_CV ? "cv" : "unk");
TAILQ_FOREACH(kth_i, db_get_waiters(db), link)
printk("\tKthread %p (%s), proc %d, sysc %p, pc/frame %p %p\n",
kth_i, kth_i->name, kth_i->proc ? kth_i->proc->pid : 0,
kth_i->sysc, jmpbuf_get_pc(&kth_i->context),
jmpbuf_get_fp(&kth_i->context));
printk("\n");
spin_unlock_irqsave(db_get_spinlock(db));
}
void print_db_blk_info(pid_t pid)
{
struct kth_db_info *db_i;
print_lock();
printk("All objects with waiters:\n");
spin_lock_irqsave(&objs_with_waiters_lock);
TAILQ_FOREACH(db_i, &objs_with_waiters, link)
db_print_obj(db_i, pid);
spin_unlock_irqsave(&objs_with_waiters_lock);
print_unlock();
}
#else
static void db_blocked_kth(struct kth_db_info *db)
{
}
static void db_unblocked_kth(struct kth_db_info *db)
{
}
static void db_init(struct kth_db_info *db, int type)
{
}
void print_db_blk_info(pid_t pid)
{
printk("Failed to print all sems: build with CONFIG_SEMAPHORE_DEBUG\n");
}
#endif /* CONFIG_SEMAPHORE_DEBUG */
static void __cv_raw_init(struct cond_var *cv)
{
TAILQ_INIT(&cv->waiters);
cv->nr_waiters = 0;
db_init(&cv->db, KTH_DB_CV);
}
/* Condition variables, using semaphores and kthreads */
void cv_init(struct cond_var *cv)
{
__cv_raw_init(cv);
cv->lock = &cv->internal_lock;
spinlock_init(cv->lock);
}
void cv_init_irqsave(struct cond_var *cv)
{
__cv_raw_init(cv);
cv->lock = &cv->internal_lock;
spinlock_init_irqsave(cv->lock);
}
void cv_init_with_lock(struct cond_var *cv, spinlock_t *lock)
{
__cv_raw_init(cv);
cv->lock = lock;
}
void cv_init_irqsave_with_lock(struct cond_var *cv, spinlock_t *lock)
{
cv_init_with_lock(cv, lock);
}
void cv_lock(struct cond_var *cv)
{
spin_lock(cv->lock);
}
void cv_unlock(struct cond_var *cv)
{
spin_unlock(cv->lock);
}
void cv_lock_irqsave(struct cond_var *cv, int8_t *irq_state)
{
disable_irqsave(irq_state);
cv_lock(cv);
}
void cv_unlock_irqsave(struct cond_var *cv, int8_t *irq_state)
{
cv_unlock(cv);
enable_irqsave(irq_state);
}
static void __attribute__((noreturn)) __cv_unlock_and_idle(void *arg)
{
struct cond_var *cv = arg;
cv_unlock(cv);
smp_idle();
}
/* Comes in locked. Regarding IRQs, the initial cv_lock_irqsave would have
* disabled irqs. When this returns, IRQs would still be disabled. If it was a
* regular cv_lock(), IRQs will be enabled when we return. */
void cv_wait_and_unlock(struct cond_var *cv)
{
bool irqs_were_on = irq_is_enabled();
struct kthread *kthread;
pre_block_check(1);
kthread = save_kthread_ctx();
if (setjmp(&kthread->context)) {
/* When the kthread restarts, IRQs are off. */
if (irqs_were_on)
enable_irq();
return;
}
TAILQ_INSERT_TAIL(&cv->waiters, kthread, link);
cv->nr_waiters++;
db_blocked_kth(&cv->db);
__reset_stack_pointer(cv, current_kthread->stacktop,
__cv_unlock_and_idle);
assert(0);
}
/* Comes in locked. Note cv_lock does not disable irqs. They should still be
* disabled from the initial cv_lock_irqsave(), which cv_wait_and_unlock()
* maintained. */
void cv_wait(struct cond_var *cv)
{
cv_wait_and_unlock(cv);
cv_lock(cv);
}
/* Helper, wakes exactly one, and there should have been at least one waiter. */
static void __cv_wake_one(struct cond_var *cv)
{
struct kthread *kthread;
kthread = TAILQ_FIRST(&cv->waiters);
TAILQ_REMOVE(&cv->waiters, kthread, link);
db_unblocked_kth(&cv->db);
kthread_runnable(kthread);
}
void __cv_signal(struct cond_var *cv)
{
if (cv->nr_waiters) {
cv->nr_waiters--;
__cv_wake_one(cv);
}
}
void __cv_broadcast(struct cond_var *cv)
{
while (cv->nr_waiters) {
cv->nr_waiters--;
__cv_wake_one(cv);
}
}
void cv_signal(struct cond_var *cv)
{
spin_lock(cv->lock);
__cv_signal(cv);
spin_unlock(cv->lock);
}
void cv_broadcast(struct cond_var *cv)
{
spin_lock(cv->lock);
__cv_broadcast(cv);
spin_unlock(cv->lock);
}
void cv_signal_irqsave(struct cond_var *cv, int8_t *irq_state)
{
disable_irqsave(irq_state);
cv_signal(cv);
enable_irqsave(irq_state);
}
void cv_broadcast_irqsave(struct cond_var *cv, int8_t *irq_state)
{
disable_irqsave(irq_state);
cv_broadcast(cv);
enable_irqsave(irq_state);
}
/* Helper, aborts and releases a CLE. dereg_ spinwaits on abort_in_progress.
* This can throw a PF */
static void __abort_and_release_cle(struct cv_lookup_elm *cle)
{
int8_t irq_state = 0;
/* At this point, we have a handle on the syscall that we want to abort
* (via the cle), and we know none of the memory will disappear on us
* (deregers wait on the flag). So we'll signal ABORT, which rendez
* will pick up next time it is awake. Then we make sure it is awake
* with a broadcast. */
atomic_or(&cle->sysc->flags, SC_ABORT);
/* flags write before signal; atomic op provided CPU mb */
cmb();
cv_broadcast_irqsave(cle->cv, &irq_state);
/* broadcast writes before abort flag; atomic op provided CPU mb */
cmb();
atomic_dec(&cle->abort_in_progress);
}
/* Attempts to abort p's sysc. It will only do so if the sysc lookup succeeds,
* so we can handle "guesses" for syscalls that might not be sleeping. This
* style of "do it if you know you can" is the best way here - anything else
* runs into situations where you don't know if the memory is safe to touch or
* not (we're doing a lookup via pointer address, and only dereferencing if that
* succeeds). Even something simple like letting userspace write SC_ABORT is
* very hard for them, since they don't know a sysc's state for sure (under the
* current system).
*
* Here are the rules:
* - if you're flagged SC_ABORT, you don't sleep
* - if you sleep, you're on the list
* - if you are on the list or abort_in_progress is set, CV is signallable, and
* all the memory for CLE is safe */
bool abort_sysc(struct proc *p, uintptr_t sysc)
{
ERRSTACK(1);
struct cv_lookup_elm *cle;
int8_t irq_state = 0;
spin_lock_irqsave(&p->abort_list_lock);
TAILQ_FOREACH(cle, &p->abortable_sleepers, link) {
if ((uintptr_t)cle->sysc == sysc) {
/* Note: we could have multiple aborters, so we need to
* use a numeric refcnt instead of a flag. */
atomic_inc(&cle->abort_in_progress);
break;
}
}
spin_unlock_irqsave(&p->abort_list_lock);
if (!cle)
return FALSE;
if (!waserror()) /* discard error */
__abort_and_release_cle(cle);
poperror();
return TRUE;
}
/* This will abort any abortables at the time the call was started for which
* should_abort(cle, arg) returns true. New abortables could be registered
* concurrently.
*
* One caller for this is proc_destroy(), in which case DYING_ABORT will be set,
* and new abortables will quickly abort and dereg when they see their proc is
* DYING_ABORT. */
static int __abort_all_sysc(struct proc *p,
bool (*should_abort)(struct cv_lookup_elm*, void*),
void *arg)
{
ERRSTACK(1);
struct cv_lookup_elm *cle;
int8_t irq_state = 0;
struct cv_lookup_tailq abortall_list;
uintptr_t old_proc = switch_to(p);
int ret = 0;
/* Concerns: we need to not remove them from their original list, since
* concurrent wake ups will cause a dereg, which will remove from the
* list. We also can't touch freed memory, so we need a refcnt to keep
* cles around. */
TAILQ_INIT(&abortall_list);
spin_lock_irqsave(&p->abort_list_lock);
TAILQ_FOREACH(cle, &p->abortable_sleepers, link) {
if (!should_abort(cle, arg))
continue;
atomic_inc(&cle->abort_in_progress);
TAILQ_INSERT_HEAD(&abortall_list, cle, abortall_link);
ret++;
}
spin_unlock_irqsave(&p->abort_list_lock);
if (!waserror()) { /* discard error */
TAILQ_FOREACH(cle, &abortall_list, abortall_link)
__abort_and_release_cle(cle);
}
poperror();
switch_back(p, old_proc);
return ret;
}
static bool always_abort(struct cv_lookup_elm *cle, void *arg)
{
return TRUE;
}
void abort_all_sysc(struct proc *p)
{
__abort_all_sysc(p, always_abort, 0);
}
/* cle->sysc could be a bad pointer. we can either use copy_from_user (btw,
* we're already in their addr space) or we can use a waserror in
* __abort_all_sysc(). Both options are fine. I went with it here for a couple
* reasons. It is only this abort function pointer that accesses sysc, though
* that could change. Our syscall aborting isn't plugged into a broader error()
* handler yet, which means we'd want to poperror instead of nexterror in
* __abort_all_sysc, and that would required int ret getting a volatile flag. */
static bool sysc_uses_fd(struct cv_lookup_elm *cle, void *fd)
{
struct syscall local_sysc;
int err;
err = copy_from_user(&local_sysc, cle->sysc, sizeof(struct syscall));
/* Trigger an abort on error */
if (err)
return TRUE;
return syscall_uses_fd(&local_sysc, (int)(long)fd);
}
int abort_all_sysc_fd(struct proc *p, int fd)
{
return __abort_all_sysc(p, sysc_uses_fd, (void*)(long)fd);
}
/* Being on the abortable list means that the CLE, KTH, SYSC, and CV are valid
* memory. The lock ordering is {CV lock, list_lock}. Callers to this *will*
* have CV held. This is done to avoid excessive locking in places like
* rendez_sleep, which want to check the condition before registering. */
void __reg_abortable_cv(struct cv_lookup_elm *cle, struct cond_var *cv)
{
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
cle->cv = cv;
cle->kthread = pcpui->cur_kthread;
/* Could be a ktask. Can build in support for aborting these later */
if (is_ktask(cle->kthread)) {
cle->sysc = 0;
return;
}
cle->sysc = cle->kthread->sysc;
cle->proc = pcpui->cur_proc;
atomic_init(&cle->abort_in_progress, 0);
spin_lock_irqsave(&cle->proc->abort_list_lock);
TAILQ_INSERT_HEAD(&cle->proc->abortable_sleepers, cle, link);
spin_unlock_irqsave(&cle->proc->abort_list_lock);
}
/* We're racing with the aborter too, who will hold the flag in cle to protect
* its ref on our cle. While the lock ordering is CV, list, callers to this
* must *not* have the cv lock held. The reason is this waits on a successful
* abort_sysc, which is trying to cv_{signal,broadcast}, which could wait on the
* CV lock. So if we hold the CV lock, we can deadlock (circular dependency).*/
void dereg_abortable_cv(struct cv_lookup_elm *cle)
{
if (is_ktask(cle->kthread))
return;
assert(cle->proc);
spin_lock_irqsave(&cle->proc->abort_list_lock);
TAILQ_REMOVE(&cle->proc->abortable_sleepers, cle, link);
spin_unlock_irqsave(&cle->proc->abort_list_lock);
/* If we won the race and yanked it out of the list before abort claimed
* it, this will already be FALSE. */
while (atomic_read(&cle->abort_in_progress))
cpu_relax();
}
/* Helper to sleepers to know if they should abort or not. I'll probably extend
* this with things for ktasks in the future. */
bool should_abort(struct cv_lookup_elm *cle)
{
struct syscall local_sysc;
int err;
if (is_ktask(cle->kthread))
return FALSE;
if (cle->proc && (cle->proc->state == PROC_DYING_ABORT))
return TRUE;
if (cle->sysc) {
assert(cle->proc && (cle->proc == current));
err = copy_from_user(&local_sysc, cle->sysc,
offsetof(struct syscall, flags) +
sizeof(cle->sysc->flags));
/* just go ahead and abort if there was an error */
if (err || (atomic_read(&local_sysc.flags) & SC_ABORT))
return TRUE;
}
return FALSE;
}
/* Sometimes the kernel needs to switch out of process context and into a
* 'process-less' kernel thread. This is basically a ktask. We use this mostly
* when performing file ops as the kernel. It's nasty, and all uses of this
* probably should be removed. (TODO: KFOP). */
uintptr_t switch_to_ktask(void)
{
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
struct kthread *kth = pcpui->cur_kthread;
if (is_ktask(kth))
return 0;
/* We leave the SAVE_ADDR_SPACE flag on. Now we're basically a ktask
* that cares about its addr space, since we need to return to it (not
* that we're leaving). */
kth->flags |= KTH_IS_KTASK;
return 1;
}
void switch_back_from_ktask(uintptr_t old_ret)
{
struct per_cpu_info *pcpui = &per_cpu_info[core_id()];
struct kthread *kth = pcpui->cur_kthread;
if (old_ret)
kth->flags &= ~KTH_IS_KTASK;
}