撤下 free service 机制以修复多核下发生保护性异常的问题.

vfs 增加文件系统元数据结构体, 将init进程调用从 shell 移动到 kmain.
添加 openfs umount2 readlink sysinfo getgid getppid 系统调用.

Co-authored-by: suhuajun-github <115517663+suhuajun-github@users.noreply.github.com>

Author: xiaoyi1212

assets/initramfs.img

@@ -120,12 +120,15 @@ _Noreturn void apu_entry() {
     apu_idle->cpu_timer                             = nano_time();
     apu_idle->cpu_id                                = current_cpu->id;
     apu_idle->status                                = RUNNING;
+    apu_idle->fs = apu_idle->gs = 0;
+    apu_idle->fs_base           = read_fsbase();
+    apu_idle->gs_base           = read_gsbase();
     char name[50];
     sprintf(name, "CP_IDLE_CPU%u", current_cpu->id);
     memcpy(apu_idle->name, name, strlen(name));
-    apu_idle->name[strlen(name)]      = '\0';
-    current_cpu->idle_pcb             = apu_idle;
-    current_cpu->ready                = true;
+    apu_idle->name[strlen(name)] = '\0';
+    current_cpu->idle_pcb        = apu_idle;
+    current_cpu->ready           = true;
     change_current_tcb(apu_idle);
     apu_idle->parent_group = kernel_group;
     apu_idle->group_index  = queue_enqueue(kernel_group->pcb_queue, apu_idle);
@@ -145,8 +148,6 @@ _Noreturn void apu_entry() {
 
     setup_syscall(false);
 
-    create_kernel_thread((void *)halt_service, NULL, "free service", kernel_group, lapic_id());
-
     open_interrupt;
     loop __asm__ volatile("hlt");
 }
@@ -180,7 +181,7 @@ void smp_setup() {
         __asm__ volatile("pause" ::: "memory");
     }
 
-    current_cpu->idle_pcb             = kernel_head_task;
+    current_cpu->idle_pcb = kernel_head_task;
     kernel_head_task->queue_index =
         queue_enqueue(((smp_cpu_t *)read_kgsbase())->scheduler_queue, kernel_head_task);
     if (kernel_head_task->queue_index == (size_t)-1) {
@@ -203,7 +204,5 @@ void smp_setup() {
     insert_sched_entity(((struct eevdf_t *)current_cpu->sched_handle)->root, idle_entity);
     eevdf_sched->current = idle_entity;
     open_interrupt;
-
-    create_kernel_thread((void *)halt_service, NULL, "free_service", kernel_group, lapic_id());
     kinfo("%d processors have been enabled.", cpu_count);
 }

src/x86_64/cpu/smp.c

@@ -170,13 +170,37 @@ errno_t cpfs_symlink(void *parent, const char *name, vfs_node_t node) {
     return EOK;
 }
 
+static void free_child_node(cpfs_file_t *parent, cpfs_file_t *dir) {
+    if (dir->is_dir) {
+        do {
+            cpfs_file_t *tmp = NULL;
+            cpfs_file_t *pos, *n;
+            llist_for_each(pos, n, &dir->child_node, curr_node) {
+                tmp = pos;
+            }
+            if (tmp == NULL) break;
+            free_child_node(dir, tmp);
+            free(dir);
+        } while (true);
+        return;
+    }
+    llist_delete(&dir->curr_node);
+    free(dir->data);
+    free(dir);
+}
+
+void cpfs_unmount(void *root) {
+    cpfs_file_t *handle = root;
+    free_child_node(NULL, handle);
+}
+
 static int dummy() {
     return -ENOSYS;
 }
 
 static struct vfs_callback cpfs_callbacks = {
     .mount    = cpfs_mount,
-    .unmount  = (void *)empty,
+    .unmount  = cpfs_unmount,
     .mkdir    = cpfs_mkdir,
     .close    = cpfs_close,
     .stat     = cpfs_stat,

src/x86_64/fs/cpio/cpfs.c

@@ -10,18 +10,23 @@
 #include "kprint.h"
 #include "krlibc.h"
 #include "list.h"
+#include "llist.h"
 #include "pcb.h"
 #include "pipefs.h"
 
 vfs_node_t rootdir = NULL;
 
 static void empty_func() {}
 
-struct vfs_callback vfs_empty_callback;
+struct vfs_callback   vfs_empty_callback;
+struct vfs_filesystem vfs_empty_filesystem;
 
 vfs_callback_t fs_callbacks[256] = {
     [0] = &vfs_empty_callback,
 };
+
+struct llist_header fs_metadata_list;
+
 static int fs_nextid = 1;
 
 #define callbackof(node, _name_) (fs_callbacks[(node)->fsid]->_name_)
@@ -256,9 +261,14 @@ int vfs_regist(const char *name, vfs_callback_t callback) {
     for (size_t i = 0; i < sizeof(struct vfs_callback) / sizeof(void *); i++) {
         if (((void **)callback)[i] == NULL) return VFS_STATUS_FAILED;
     }
-    int id = fs_nextid++;
-
+    int id           = fs_nextid++;
     fs_callbacks[id] = callback;
+
+    vfs_filesystem_t filesystem = malloc(sizeof(struct vfs_filesystem));
+    filesystem->callback        = callback;
+    strcpy(filesystem->name, name);
+    llist_init_head(&filesystem->node);
+    llist_append(&fs_metadata_list, &filesystem->node);
     return id;
 }
 
@@ -477,6 +487,11 @@ errno_t vfs_unmount(const char *path) {
     return VFS_STATUS_FAILED;
 }
 
+size_t vfs_readlink(vfs_node_t node, char *buf, size_t bufsize) {
+    size_t ret = callbackof(node, readlink)(node, buf, 0, bufsize);
+    return ret;
+}
+
 void *general_map(vfs_read_t read_callback, void *file, uint64_t addr, uint64_t len, uint64_t prot,
                   uint64_t flags, uint64_t offset) {
     pcb_t current_task        = get_current_task()->parent_group;
@@ -580,7 +595,7 @@ bool vfs_init() {
     for (size_t i = 0; i < sizeof(struct vfs_callback) / sizeof(void *); i++) {
         ((void **)&vfs_empty_callback)[i] = (void *)empty_func;
     }
-
+    llist_init_head(&fs_metadata_list);
     rootdir       = vfs_node_alloc(NULL, "/");
     rootdir->type = file_dir;
     kinfo("Virtual File System initialize.");

src/x86_64/fs/vfs.c

@@ -3,6 +3,5 @@
 #include "pcb.h"
 
 void add_death_proc(pcb_t task);
-void halt_service();
 
 void killer_setup();

src/x86_64/include/killer.h

@@ -3,4 +3,7 @@
 #define MAX_COMMAND_LEN 100
 #define MAX_ARG_NR      50
 
+bool exec(int argc, char **argv);
+void mount(int argc, char **argv);
+
 _Noreturn void shell_setup();

src/x86_64/include/shell.h

@@ -11,7 +11,45 @@
 #define SYSCALL_FAULT_(name) ((uint64_t)-(name))
 #define FD_SETSIZE           1024
 
-#define syscall_(name)                                                                             \
+// 一个非常取巧的宏魔法, 可以简化 syscall 函数的定义
+#define __EXPAND_PARAMS(...) __VA_ARGS__
+#define __CONCAT_IMPL(a, b)  a##b
+#define __CONCAT(a, b)       __CONCAT_IMPL(a, b)
+
+#define __ARGS_COUNT_IMPL(_0, _1, _2, _3, _4, _5, _6, N, ...) N
+
+#define __ARGS_COUNT(...) __EXPAND_PARAMS(__ARGS_COUNT_IMPL(__VA_ARGS__, 6, 5, 4, 3, 2, 1, 0))
+
+#define __SYSCALL_IMPL_0(NAME)                                                                     \
+    uint64_t syscall_##NAME(uint64_t arg0, uint64_t arg1, uint64_t arg2, uint64_t arg3,            \
+                            uint64_t arg4, uint64_t arg5, struct syscall_regs *regs)
+
+#define __SYSCALL_IMPL_1(NAME, P1)                                                                 \
+    uint64_t syscall_##NAME(P1, uint64_t arg1, uint64_t arg2, uint64_t arg3, uint64_t arg4,        \
+                            uint64_t arg5, struct syscall_regs *regs)
+
+#define __SYSCALL_IMPL_2(NAME, P1, P2)                                                             \
+    uint64_t syscall_##NAME(P1, P2, uint64_t arg2, uint64_t arg3, uint64_t arg4, uint64_t arg5,    \
+                            struct syscall_regs *regs)
+
+#define __SYSCALL_IMPL_3(NAME, P1, P2, P3)                                                         \
+    uint64_t syscall_##NAME(P1, P2, P3, uint64_t arg3, uint64_t arg4, uint64_t arg5,               \
+                            struct syscall_regs *regs)
+
+#define __SYSCALL_IMPL_4(NAME, P1, P2, P3, P4)                                                     \
+    uint64_t syscall_##NAME(P1, P2, P3, P4, uint64_t arg4, uint64_t arg5, struct syscall_regs *regs)
+
+#define __SYSCALL_IMPL_5(NAME, P1, P2, P3, P4, P5)                                                 \
+    uint64_t syscall_##NAME(P1, P2, P3, P4, P5, uint64_t arg5, struct syscall_regs *regs)
+
+#define __SYSCALL_IMPL_6(NAME, P1, P2, P3, P4, P5, P6)                                             \
+    uint64_t syscall_##NAME(P1, P2, P3, P4, P5, P6, struct syscall_regs *regs)
+
+#define __SYSCALL_DISPATCH(N, NAME, ...) __CONCAT(__SYSCALL_IMPL_, N)(NAME, ##__VA_ARGS__)
+
+#define syscall_(NAME, ...) __SYSCALL_DISPATCH(__ARGS_COUNT(0, ##__VA_ARGS__), NAME, ##__VA_ARGS__)
+
+#define syscall_def_(name)                                                                         \
     uint64_t syscall_##name(                                                                       \
         uint64_t arg0 __attribute__((unused)), uint64_t arg1 __attribute__((unused)),              \
         uint64_t arg2 __attribute__((unused)), uint64_t arg3 __attribute__((unused)),              \
@@ -114,12 +152,16 @@
 #define SYSCALL_LINK        86
 #define SYSCALL_UNLINK      87
 #define SYSCALL_SYMLINK     88
+#define SYSCALL_READLINK    89
+#define SYSCALL_SYSINFO     99
 #define SYSCALL_GETUID      102
+#define SYSCALL_GETGID      104
 #define SYSCALL_SETUID      105
 #define SYSCALL_SETGID      106
 #define SYSCALL_GETEUID     107
 #define SYSCALL_GETEGID     108
 #define SYSCALL_SETPGID     109
+#define SYSCALL_GETPPID     110
 #define SYSCALL_GETGROUPS   115
 #define SYScall_GETPGID     121
 #define SYSCALL_SIGSUSPEND  130
@@ -128,6 +170,7 @@
 #define SYSCALL_ARCH_PRCTL  158
 #define SYSCALL_G_AFFINITY  160
 #define SYSCALL_MOUNT       165
+#define SYSCALL_UMOUNT2     166
 #define SYSCALL_REBOOT      169
 #define SYSCALL_GET_TID     186
 #define SYSCALL_FUTEX       202
@@ -146,6 +189,7 @@
 #define SYSCALL_PIPE2       293
 #define SYSCALL_CP_F_RANGE  326
 #define SYSCALL_STATX       332
+#define SYSCALL_FSOPEN      430
 #define SYSCALL_FACCESSAT2  439
 
 // CoolPotOS 平台特有系统调用号定义
@@ -287,6 +331,23 @@ struct statx {
                                          /* 0x100 */
 };
 
+struct sysinfo {
+    int64_t  uptime;    /* Seconds since boot */
+    uint64_t loads[3];  /* 1, 5, and 15 minute load averages */
+    uint64_t totalram;  /* Total usable main memory size */
+    uint64_t freeram;   /* Available memory size */
+    uint64_t sharedram; /* Amount of shared memory */
+    uint64_t bufferram; /* Memory used by buffers */
+    uint64_t totalswap; /* Total swap space size */
+    uint64_t freeswap;  /* swap space still available */
+    uint16_t procs;     /* Number of current processes */
+    uint16_t pad;       /* Explicit padding for m68k */
+    uint64_t totalhigh; /* Total high memory size */
+    uint64_t freehigh;  /* Available high memory size */
+    uint32_t mem_unit;  /* Memory unit size in bytes */
+    char     _f[20 - 2 * sizeof(uint64_t) - sizeof(uint32_t)]; /* Padding: libc5 uses this.. */
+};
+
 struct cpos_meminfo {
     uint64_t used;      // 已用内存
     uint64_t available; // 未使用内存

src/x86_64/include/syscall.h

@@ -64,6 +64,7 @@
 
 #include "ctype.h"
 #include "list.h"
+#include "llist.h"
 
 typedef struct vfs_node *vfs_node_t;
 
@@ -132,6 +133,12 @@ typedef struct vfs_callback { // VFS回调函数
     vfs_rename_t rename;      // 重命名文件或文件夹
 } *vfs_callback_t;
 
+typedef struct vfs_filesystem {
+    vfs_callback_t      callback;
+    char                name[10];
+    struct llist_header node;
+} *vfs_filesystem_t;
+
 struct vfs_node {           // vfs节点
     vfs_node_t parent;      // 父目录
     vfs_node_t linkto;      // 符号链接指向的节点
@@ -167,6 +174,7 @@ struct fd {
 
 extern struct vfs_callback vfs_empty_callback;
 extern vfs_node_t          rootdir;
+extern struct llist_header fs_metadata_list;
 
 /**
  * 创建目录节点
@@ -253,6 +261,15 @@ vfs_node_t vfs_open(const char *str);
  */
 errno_t vfs_ioctl(vfs_node_t device, size_t options, void *arg);
 
+/**
+ * 读取一个符号链接文件
+ * @param node 文件节点
+ * @param buf 缓冲区
+ * @param bufsize
+ * @return
+ */
+size_t vfs_readlink(vfs_node_t node, char *buf, size_t bufsize);
+
 bool is_virtual_fs(const char *src);
 
 vfs_node_t vfs_do_search(vfs_node_t dir, const char *name);

src/x86_64/include/vfs.h

@@ -60,6 +60,8 @@
 #    define GIT_VERSION "unknown"
 #endif
 
+#define INITRAMFS_START 1
+
 void kmain();
 
 extern void  error_setup();    // error_handle.c
@@ -82,6 +84,20 @@ _Noreturn void cp_reset() {
     loop;
 }
 
+static void kernel_watchdog() {
+    kwarn("PID 1 process has exited, user session is missing.");
+    kwarn("Should the kernel be restarted?");
+    printk("Reboot kernel(Y/n): ");
+    _terminal_flush();
+    do {
+        char c = kernel_getch();
+        printk("%c\n", c);
+        _terminal_flush();
+        if (c == '\n' || c == 'y') { cp_reset(); }
+        if (c == 'n') break;
+    } while (true);
+}
+
 void kmain() {
     gdt_setup();
     idt_setup();
@@ -144,15 +160,37 @@ void kmain() {
     netfs_setup();
     load_all_kernel_module();
     partition_init();
+#ifdef INITRAMFS_START
     cpio_init();
+#endif
     kinfo("Kernel load Done!");
 
-    pcb_t shell_group = create_process_group("Shell Service", NULL, NULL, "", NULL, NULL, 0);
-    create_kernel_thread((void *)shell_setup, NULL, "KernelShell", shell_group, SIZE_MAX);
+    // pcb_t shell_group = create_process_group("Shell Service", NULL, NULL, "", NULL, NULL, 0);
+    // create_kernel_thread((void *)shell_setup, NULL, "KernelShell", shell_group, SIZE_MAX);
 
     open_interrupt;
     enable_scheduler();
-    change_bsp_weight();
 
+#ifdef INITRAMFS_START
+#    if 0
+    int   argc   = 3;
+    char *argv[] = {"/init", NULL};
+    exec(argc, argv);
+#    else
+    {
+        int   argc         = 3;
+        char *argv_init[2] = {"exec", "/sbin/init"};
+        char *argv_sh[3]   = {"exec", "/bin/busybox", "sh"};
+        char *argv_bash[2] = {"exec", "/bin/bash"};
+        //exec(argc, argv_init);
+        exec(argc, argv_bash);
+        exec(argc, argv_sh);
+    }
+#    endif
+
+#endif
+    kernel_watchdog();
+
+    change_bsp_weight();
     loop __asm__ volatile("hlt");
 }

src/x86_64/main.c

@@ -236,5 +236,5 @@ void init_frame_bitmap() {
     frame_allocator.origin_frames = origin_frames;
     frame_allocator.usable_frames = origin_frames - bitmap_frame_count;
 
-    logkf("Available memory: %lld MiB\n", (origin_frames / 256));
+    logkf("Available memory: %lld MiB bitmap_sz: %llu\n", (origin_frames / 256), bitmap_size);
 }