feat: pause(2) support
[lunaix-os.git] / lunaix-os / kernel / sched.c
index 8f65087f8253175f0a2dd2f4f26ab34afd718a4a..50f5f67ceaa6cf209d885fd7e9b78ebaa3ef8635 100644 (file)
@@ -1,12 +1,15 @@
 #include <arch/x86/interrupts.h>
 #include <arch/x86/tss.h>
+
 #include <hal/apic.h>
 #include <hal/cpu.h>
+
 #include <lunaix/mm/kalloc.h>
+#include <lunaix/mm/pmm.h>
 #include <lunaix/mm/vmm.h>
 #include <lunaix/process.h>
 #include <lunaix/sched.h>
-
+#include <lunaix/signal.h>
 #include <lunaix/spike.h>
 #include <lunaix/status.h>
 #include <lunaix/syscall.h>
@@ -18,8 +21,6 @@ volatile struct proc_info* __current;
 
 struct proc_info dummy;
 
-extern void __proc_table;
-
 struct scheduler sched_ctx;
 
 LOG_MODULE("SCHED")
@@ -28,11 +29,14 @@ void
 sched_init()
 {
     size_t pg_size = ROUNDUP(sizeof(struct proc_info) * MAX_PROCESS, 0x1000);
-    assert_msg(vmm_alloc_pages(
-                 KERNEL_PID, &__proc_table, pg_size, PG_PREM_RW, PP_FGPERSIST),
-               "Fail to allocate proc table");
 
-    sched_ctx = (struct scheduler){ ._procs = (struct proc_info*)&__proc_table,
+    for (size_t i = 0; i <= pg_size; i += 4096) {
+        uintptr_t pa = pmm_alloc_page(KERNEL_PID, PP_FGPERSIST);
+        vmm_set_mapping(
+          PD_REFERENCED, PROC_START + i, pa, PG_PREM_RW, VMAP_NULL);
+    }
+
+    sched_ctx = (struct scheduler){ ._procs = (struct proc_info*)PROC_START,
                                     .ptable_len = 0,
                                     .procs_index = 0 };
 }
@@ -45,22 +49,20 @@ run(struct proc_info* proc)
     }
     proc->state = PROC_RUNNING;
 
-    // FIXME: 这里还是得再考虑一下。
-    // tss_update_esp(__current->intr_ctx.esp);
-
-    if (__current->page_table != proc->page_table) {
-        __current = proc;
-        cpu_lcr3(__current->page_table);
-        // from now on, the we are in the kstack of another process
-    } else {
-        __current = proc;
-    }
+    /*
+        将tss.esp0设置为上次调度前的esp值。
+        当处理信号时,上下文信息是不会恢复的,而是保存在用户栈中,然后直接跳转进位于用户空间的sig_wrapper进行
+          信号的处理。当用户自定义的信号处理函数返回时,sigreturn的系统调用才开始进行上下文的恢复(或者说是进行
+          另一次调度。
+        由于这中间没有进行地址空间的交换,所以第二次跳转使用的是同一个内核栈,而之前默认tss.esp0的值是永远指向最顶部
+        这样一来就有可能会覆盖更早的上下文信息(比如嵌套的信号捕获函数)
+    */
+    tss_update_esp(proc->intr_ctx.registers.esp);
 
     apic_done_servicing();
 
     asm volatile("pushl %0\n"
-                 "jmp soft_iret\n" ::"r"(&__current->intr_ctx)
-                 : "memory");
+                 "jmp switch_to\n" ::"r"(proc)); // kernel/asm/x86/interrupt.S
 }
 
 void
@@ -70,6 +72,8 @@ schedule()
         return;
     }
 
+    // 上下文切换相当的敏感!我们不希望任何的中断打乱栈的顺序……
+    cpu_disable_interrupt();
     struct proc_info* next;
     int prev_ptr = sched_ctx.procs_index;
     int ptr = prev_ptr;
@@ -97,6 +101,7 @@ __DEFINE_LXSYSCALL1(unsigned int, sleep, unsigned int, seconds)
     if (!seconds) {
         return 0;
     }
+
     if (__current->timer) {
         return __current->timer->counter / timer_context()->running_frequency;
     }
@@ -119,31 +124,60 @@ __DEFINE_LXSYSCALL(void, yield)
     schedule();
 }
 
+pid_t
+_wait(pid_t wpid, int* status, int options);
+
 __DEFINE_LXSYSCALL1(pid_t, wait, int*, status)
+{
+    return _wait(-1, status, 0);
+}
+
+__DEFINE_LXSYSCALL3(pid_t, waitpid, pid_t, pid, int*, status, int, options)
+{
+    return _wait(pid, status, options);
+}
+
+pid_t
+_wait(pid_t wpid, int* status, int options)
 {
     pid_t cur = __current->pid;
+    int status_flags = 0;
     struct proc_info *proc, *n;
     if (llist_empty(&__current->children)) {
         return -1;
     }
+
+    wpid = wpid ? wpid : -__current->pgid;
+    cpu_enable_interrupt();
 repeat:
     llist_for_each(proc, n, &__current->children, siblings)
     {
-        if (proc->state == PROC_TERMNAT) {
-            goto done;
+        if (!~wpid || proc->pid == wpid || proc->pgid == -wpid) {
+            if (proc->state == PROC_TERMNAT && !options) {
+                status_flags |= PROCTERM;
+                goto done;
+            }
+            if (proc->state == PROC_STOPPED && (options & WUNTRACED)) {
+                status_flags |= PROCSTOP;
+                goto done;
+            }
         }
     }
-    // FIXME: 除了循环,也许有更高效的办法……
-    // (在这里进行schedule,需要重写context switch!)
+    if ((options & WNOHANG)) {
+        return 0;
+    }
+    // 放弃当前的运行机会
+    sched_yield();
     goto repeat;
 
 done:
-    *status = proc->exit_code;
+    cpu_disable_interrupt();
+    *status = (proc->exit_code & 0xffff) | status_flags;
     return destroy_process(proc->pid);
 }
 
-pid_t
-alloc_pid()
+struct proc_info*
+alloc_process()
 {
     pid_t i = 0;
     for (;
@@ -154,28 +188,36 @@ alloc_pid()
     if (i == MAX_PROCESS) {
         panick("Panic in Ponyville shimmer!");
     }
-    return i;
+
+    if (i == sched_ctx.ptable_len) {
+        sched_ctx.ptable_len++;
+    }
+
+    struct proc_info* proc = &sched_ctx._procs[i];
+    memset(proc, 0, sizeof(*proc));
+
+    proc->state = PROC_CREATED;
+    proc->pid = i;
+    proc->created = clock_systime();
+    proc->pgid = proc->pid;
+
+    llist_init_head(&proc->mm.regions);
+    llist_init_head(&proc->children);
+    llist_init_head(&proc->grp_member);
+
+    return proc;
 }
 
 void
-push_process(struct proc_info* process)
+commit_process(struct proc_info* process)
 {
-    int index = process->pid;
-    if (index < 0 || index > sched_ctx.ptable_len) {
-        __current->k_status = LXINVLDPID;
-        return;
-    }
+    assert(process == &sched_ctx._procs[process->pid]);
 
-    if (index == sched_ctx.ptable_len) {
-        sched_ctx.ptable_len++;
+    if (process->state != PROC_CREATED) {
+        __current->k_status = LXINVL;
+        return;
     }
 
-    sched_ctx._procs[index] = *process;
-
-    process = &sched_ctx._procs[index];
-
-    // make sure the address is in the range of process table
-    llist_init_head(&process->children);
     // every process is the child of first process (pid=1)
     if (process->parent) {
         llist_append(&process->parent->children, &process->siblings);
@@ -202,19 +244,17 @@ destroy_process(pid_t pid)
     proc->state = PROC_DESTROY;
     llist_delete(&proc->siblings);
 
-    if (proc->mm.regions) {
-        struct mm_region *pos, *n;
-        llist_for_each(pos, n, &proc->mm.regions->head, head)
-        {
-            lxfree(pos);
-        }
+    struct mm_region *pos, *n;
+    llist_for_each(pos, n, &proc->mm.regions.head, head)
+    {
+        lxfree(pos);
     }
 
-    vmm_mount_pd(PD_MOUNT_2, proc->page_table);
+    vmm_mount_pd(PD_MOUNT_1, proc->page_table);
 
-    __del_pagetable(pid, PD_MOUNT_2);
+    __del_pagetable(pid, PD_MOUNT_1);
 
-    vmm_unmount_pd(PD_MOUNT_2);
+    vmm_unmount_pd(PD_MOUNT_1);
 
     return pid;
 }
@@ -225,6 +265,8 @@ terminate_proc(int exit_code)
     __current->state = PROC_TERMNAT;
     __current->exit_code = exit_code;
 
+    __SET_SIGNAL(__current->parent->sig_pending, _SIGCHLD);
+
     schedule();
 }