MeowOS-For-R3P2/patch_v2/reader_app/app/shelf_index.c

/* app/shelf_index.c — 书架索引落盘/校验（实现见 app/shelf_index.h）。
 *
 * 落盘格式 v2（全部小端、手工序列化，不依赖结构体布局）：
 *   [header 24B]  "MEOWIDX2" | u32 version | u32 header_size | u32 root_len | u32 rsv
 *   [root bytes]
 *   [dir 记录 ×N] u32 rel_len | u32 books_here | u64 mtime_sec | u32 mtime_nsec | u32 rsv | rel
 *                 只含 Books 根（rel=""）与一级目录（rel 不含 '/'）；level-2+ 不落盘。
 *   [book 记录 ×M] u32 rel_len | u32 size | u32 fmt | u32 rsv | rel
 *   [trailer 24B] u32 ndirs | u32 nbooks | u32 flags | u32 depth_skipped | u32 crc32 | u32 magic
 * crc32 覆盖 [0, trailer_off)（含 header/root/全部记录），用来发现截断/坏页。
 * rel 一律相对 Books 根，路径做 ".."/绝对路径/内嵌 NUL 校验，杜绝索引坏档造成越界读。
 *
 * 校验范围只有根 + 一级目录（真机冷缓存 434 次 stat=3.15s，4 次≈20ms；见头文件）。
 * level-2 及更深处的变更不会自动失效，需点「刷新」——这是本方案明确接受的代价。
 *
 * 写入顺序：临时文件 `shelf.idx.tmp` 全量写完（flash 尽力 fsync）后 rename 发布。
 * 扫描中途被杀只会留下 tmp（下次覆盖），正式索引要么是旧的、要么是完整新的。
 */
#include "shelf_index.h"
#include "meow_ui.h"        /* meow_config_dir() */

#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <unistd.h>

#define IDX_HDR_SIZE     24
#define IDX_TRL_SIZE     24
#define IDX_TRL_MAGIC    0x314D5749u   /* "IWM1" */
#define IDX_REL_MAX      1000          /* 单条 rel 上限（真机 <100B，留足余量） */
#define IDX_DIR_REC_FIX  24            /* dir 记录定长部分 */
#define IDX_BOOK_REC_FIX 16            /* book 记录定长部分 */
#define IDX_MAX_SIZE     (8 * 1024 * 1024)

static const unsigned char IDX_MAGIC[8] = { 'M','E','O','W','I','D','X','2' };

/* ---- 小端序列化 ---- */
static void put32(unsigned char *p, uint32_t v)
{
    p[0] = (unsigned char)(v & 0xFF); p[1] = (unsigned char)((v >> 8) & 0xFF);
    p[2] = (unsigned char)((v >> 16) & 0xFF); p[3] = (unsigned char)((v >> 24) & 0xFF);
}
static uint32_t get32(const unsigned char *p)
{
    return (uint32_t)p[0] | ((uint32_t)p[1] << 8) | ((uint32_t)p[2] << 16) | ((uint32_t)p[3] << 24);
}
static void put64(unsigned char *p, uint64_t v)
{
    put32(p, (uint32_t)(v & 0xFFFFFFFFu));
    put32(p + 4, (uint32_t)((v >> 32) & 0xFFFFFFFFu));
}
static uint64_t get64(const unsigned char *p)
{
    return (uint64_t)get32(p) | ((uint64_t)get32(p + 4) << 32);
}

/* ---- CRC32（IEEE 802.3，查表；表首次使用时生成 1KB） ---- */
static uint32_t g_crc_tbl[256];
static int      g_crc_ready;

static void crc_table_init(void)
{
    for (uint32_t i = 0; i < 256; i++) {
        uint32_t c = i;
        for (int k = 0; k < 8; k++)
            c = (c & 1) ? (0xEDB88320u ^ (c >> 1)) : (c >> 1);
        g_crc_tbl[i] = c;
    }
    g_crc_ready = 1;
}
static uint32_t crc_step(uint32_t c, const void *buf, size_t n)
{
    if (!g_crc_ready) crc_table_init();
    const unsigned char *p = (const unsigned char *)buf;
    while (n--) c = g_crc_tbl[(c ^ *p++) & 0xFF] ^ (c >> 8);
    return c;
}

/* ---- 路径推导：<MeowData>/<pkg>/data → <MeowData>/<pkg>/index/shelf.idx ---- */
static void strip_data_suffix(const char *d, char *out, int outsz)
{
    if (!d || !d[0]) { strncpy(out, "/tmp", (size_t)outsz - 1); out[outsz - 1] = 0; return; }
    size_t n = strlen(d);
    const char *suffix = "/data";
    size_t sl = strlen(suffix);
    size_t keep = (n >= sl && strcmp(d + n - sl, suffix) == 0) ? n - sl : n;
    if (keep >= (size_t)outsz) keep = (size_t)outsz - 1;
    memcpy(out, d, keep);
    out[keep] = 0;
}

/* mkdir -p（逐分量，EEXIST 视为成功）。 */
static int mkdir_p(const char *path)
{
    char tmp[520];
    if (!path || !path[0] || strlen(path) >= sizeof tmp) return -1;
    memcpy(tmp, path, strlen(path) + 1);
    for (char *p = tmp + 1; *p; p++) {
        if (*p != '/') continue;
        *p = 0; mkdir(tmp, 0755); *p = '/';
    }
    mkdir(tmp, 0755);
    return 0;
}

/* 索引目录 <MeowData>/<pkg>/index（只算路径；mkdir 只在写索引时做）。 */
static int index_dir(char *out, int outsz)
{
    const char *data = meow_config_dir();
    if (!data) return -2;
    char base[400];
    strip_data_suffix(data, base, sizeof base);
    if (snprintf(out, (size_t)outsz, "%s/index", base) >= outsz) return -3;
    return 0;
}

int shelf_index_path(const char *root, char *out, int outsz)
{
    (void)root;   /* 索引跟 App 数据目录走，不跟 Books 路径走 */
    if (!out || outsz <= 0) return -1;
    out[0] = 0;
    char dir[440];
    if (index_dir(dir, sizeof dir) != 0) return -2;
    if (snprintf(out, (size_t)outsz, "%s/shelf.idx", dir) >= outsz) return -5;
    return 0;
}

/* ---- rel 校验：相对路径、无 ".." 分量、无内嵌 NUL、长度受控 ---- */
static int rel_valid(const char *rel, int len, int allow_empty)
{
    if (len < 0 || len > IDX_REL_MAX) return 0;
    if (len == 0) return allow_empty;
    if (rel[0] == '/') return 0;
    if (memchr(rel, 0, (size_t)len) != NULL) return 0;
    int comp = 0;
    for (int i = 0; i <= len; i++) {
        char ch = (i == len) ? '/' : rel[i];
        if (ch == '/') {
            if (i - comp == 2 && rel[comp] == '.' && rel[comp + 1] == '.') return 0;
            comp = i + 1;
        }
    }
    return 1;
}

/* 目录 rel（""/a/b）拼成绝对路径；返回 0=成功。 */
static int join_path(const char *root, const char *rel, int rel_len, char *out, int outsz)
{
    if (rel_len == 0) {
        if ((int)strlen(root) >= outsz) return -1;
        memcpy(out, root, strlen(root) + 1);
        return 0;
    }
    int n = snprintf(out, (size_t)outsz, "%s/%s", root, rel);
    return (n < 0 || n >= outsz) ? -1 : 0;
}

/* book rel → title/group/top。与 shelf_scan_books 的写法保持逐字一致
 * （path 必须是同一个绝对路径，进度键 rdprog_hash(path) 才能对上）。 */
static void fill_from_rel(shelf_book *b, const char *rel, int rel_len)
{
    const char *base = rel;
    for (int i = 0; i < rel_len; i++) if (rel[i] == '/') base = rel + i + 1;
    char name[256];
    int nl = rel_len - (int)(base - rel);
    if (nl >= (int)sizeof name) nl = (int)sizeof name - 1;
    memcpy(name, base, (size_t)nl);
    name[nl] = 0;
    shelf_copy_utf8(b->title, (int)sizeof b->title, name);
    char *dot = strrchr(b->title, '.');
    if (dot) *dot = 0;

    /* group = rel 去掉最后一段；top = 第一段。*/
    int glen = (int)(base - rel);
    if (glen > 0) glen--;                     /* 去掉结尾 '/' */
    char g[IDX_REL_MAX + 1];
    if (glen < 0) glen = 0;
    memcpy(g, rel, (size_t)glen);
    g[glen] = 0;
    shelf_copy_utf8(b->group, (int)sizeof b->group, g);
    const char *sl = strchr(g, '/');
    int tlen = sl ? (int)(sl - g) : glen;
    char t[IDX_REL_MAX + 1];
    memcpy(t, g, (size_t)tlen);
    t[tlen] = 0;
    shelf_copy_utf8(b->top, (int)sizeof b->top, t);
}

/* ================= 载入 ================= */

int shelf_index_load(const char *root, shelf_book *books, int cap,
                     shelf_ts_fn ts_fn, shelf_scan_report *rep)
{
    if (rep) { rep->nbooks = 0; rep->truncated = 0; rep->depth_skipped = 0; rep->root_missing = 0; }
    if (!root || !books || cap <= 0) return 0;

    char ipath[500];
    if (shelf_index_path(root, ipath, sizeof ipath) != 0) return 0;
    /* O_NOFOLLOW/O_NONBLOCK + S_ISREG：索引是 App 自己的普通文件；卡上若有同名 FIFO/符号链接，
     * 直接拒绝重扫，避免 fopen 阻塞或读到卡外内容。 */
    int fd = open(ipath, O_RDONLY | O_NONBLOCK | O_NOFOLLOW);
    if (fd < 0) return 0;
    struct stat fst;
    if (fstat(fd, &fst) != 0 || !S_ISREG(fst.st_mode) || fst.st_size <= 0) {
        close(fd);
        return 0;
    }
    FILE *f = fdopen(fd, "rb");
    if (!f) { close(fd); return 0; }

    int ok = 0;
    long fsz = (long)fst.st_size;
    if (fsz < IDX_HDR_SIZE + IDX_TRL_SIZE || fsz > IDX_MAX_SIZE) goto done;

    /* trailer 先读：拿 ndirs/nbooks/crc，再从头顺序校验。 */
    unsigned char tr[IDX_TRL_SIZE];
    if (fseek(f, fsz - IDX_TRL_SIZE, SEEK_SET) != 0) goto done;
    if (fread(tr, 1, sizeof tr, f) != sizeof tr) goto done;
    if (get32(tr + 20) != IDX_TRL_MAGIC) goto done;
    uint32_t ndirs = get32(tr + 0);
    uint32_t nbooks = get32(tr + 4);
    uint32_t flags = get32(tr + 8);
    uint32_t depth_skipped = get32(tr + 12);
    uint32_t crc_expect = get32(tr + 16);
    if (ndirs < 1 || ndirs > 65535) goto done;
    if (nbooks > (uint32_t)cap || nbooks > (uint32_t)SHELF_MAX_BOOKS) goto done;
    if (depth_skipped > 0x7FFFFFFFu) goto done;

    long trailer_off = fsz - IDX_TRL_SIZE;
    if (fseek(f, 0, SEEK_SET) != 0) goto done;
    uint32_t crc = 0xFFFFFFFFu;

    /* header + root */
    unsigned char h[IDX_HDR_SIZE];
    if (fread(h, 1, sizeof h, f) != sizeof h) goto done;
    crc = crc_step(crc, h, sizeof h);
    if (memcmp(h, IDX_MAGIC, 8) != 0) goto done;
    if (get32(h + 8) != SHELF_INDEX_VERSION) goto done;
    if (get32(h + 12) != IDX_HDR_SIZE) goto done;
    uint32_t root_len = get32(h + 16);
    if (root_len > IDX_REL_MAX || (int)root_len != (int)strlen(root)) goto done;
    char rootbuf[IDX_REL_MAX + 1];
    if (root_len > 0 && fread(rootbuf, 1, root_len, f) != root_len) goto done;
    if (root_len > 0) crc = crc_step(crc, rootbuf, root_len);
    rootbuf[root_len] = 0;
    if (strcmp(rootbuf, root) != 0) goto done;   /* 换了卡/挂载点 → 重扫 */

    /* 校验记录：只有 Books 根 + 一级目录（v2）。逐个 stat 比对 mtime（唯一允许的校验 IO，
     * 不做 readdir）。level-2+ 的 rel 含 '/'，出现在 v2 文件里即视为坏档。 */
    uint32_t sum_here = 0;
    for (uint32_t i = 0; i < ndirs; i++) {
        unsigned char rh[IDX_DIR_REC_FIX];
        if (fread(rh, 1, sizeof rh, f) != sizeof rh) goto done;
        crc = crc_step(crc, rh, sizeof rh);
        uint32_t rel_len = get32(rh + 0);
        uint32_t books_here = get32(rh + 4);
        uint64_t mt_sec = get64(rh + 8);
        uint32_t mt_nsec = get32(rh + 16);
        if (rel_len > IDX_REL_MAX || books_here > (uint32_t)cap) goto done;
        char rel[IDX_REL_MAX + 1];
        if (rel_len > 0 && fread(rel, 1, rel_len, f) != rel_len) goto done;
        if (rel_len > 0) crc = crc_step(crc, rel, rel_len);
        rel[rel_len] = 0;
        if (!rel_valid(rel, (int)rel_len, 1)) goto done;
        if (rel_len > 0 && strchr(rel, '/')) goto done;   /* v2 只允许一级 */

        char path[400];
        if (join_path(root, rel, (int)rel_len, path, sizeof path) != 0) goto done;
        struct stat st;
        if (stat(path, &st) != 0) goto done;
        if (!S_ISDIR(st.st_mode)) goto done;
        if ((uint64_t)(int64_t)st.st_mtime != mt_sec) goto done;
        if ((uint32_t)(long)st.st_mtim.tv_nsec != mt_nsec) goto done;
        sum_here += books_here;   /* 只作合理性上界：根+一级里直接放的书写数 ≤ nbooks */
    }

    /* 书记录 */
    for (uint32_t i = 0; i < nbooks; i++) {
        unsigned char bh[IDX_BOOK_REC_FIX];
        if (fread(bh, 1, sizeof bh, f) != sizeof bh) goto done;
        crc = crc_step(crc, bh, sizeof bh);
        uint32_t rel_len = get32(bh + 0);
        uint32_t size = get32(bh + 4);
        uint32_t fmt = get32(bh + 8);
        if (rel_len < 1 || rel_len > IDX_REL_MAX || fmt > 4) goto done;
        char rel[IDX_REL_MAX + 1];
        if (fread(rel, 1, rel_len, f) != rel_len) goto done;
        crc = crc_step(crc, rel, rel_len);
        rel[rel_len] = 0;
        if (!rel_valid(rel, (int)rel_len, 0)) goto done;

        shelf_book *b = &books[i];
        memset(b, 0, sizeof *b);
        if (join_path(root, rel, (int)rel_len, b->path, sizeof b->path) != 0) goto done;
        if ((int)strlen(b->path) != (int)(strlen(root) + 1 + rel_len)) goto done;
        fill_from_rel(b, rel, (int)rel_len);
        b->fmt = (int)fmt;
        b->size = (long)size;
        b->ts = 0;   /* 进度不落盘：下面用 ts_fn 现读 */
    }

    /* 必须正好读到 trailer 起点（多/少字节都视为坏档） */
    if (ftell(f) != trailer_off) goto done;
    if ((crc ^ 0xFFFFFFFFu) != crc_expect) goto done;
    /* v2 只校验根+一级：books_here 之和是「根/一级里直接放的书写数」，不可能超过总书数。 */
    if (sum_here > nbooks) goto done;

    /* 校验全过：现读进度 ts（与全量扫描同一口径），重建副题并按最近阅读排序。 */
    for (uint32_t i = 0; i < nbooks; i++) {
        books[i].ts = ts_fn ? ts_fn(books[i].path) : 0;
        shelf_fmt_sub(&books[i]);
    }
    shelf_sort_recent(books, (int)nbooks);

    if (rep) {
        rep->nbooks = (int)nbooks;
        rep->truncated = (flags & 1u) ? 1 : 0;
        rep->depth_skipped = (int)depth_skipped;
        rep->root_missing = 0;
    }
    ok = 1;

done:
    fclose(f);
    return ok;
}

/* ================= 写入 ================= */

struct shelf_idx_writer {
    FILE    *f;
    char     root[400];
    char     dst[500];
    char     tmp[520];
    uint32_t crc;
    int      ndirs;
    int      ok;
};

static int wraw(shelf_idx_writer *w, const void *buf, size_t n)
{
    if (!w->ok) return -1;
    if (fwrite(buf, 1, n, w->f) != n) { w->ok = 0; return -1; }
    w->crc = crc_step(w->crc, buf, n);
    return 0;
}

shelf_idx_writer *shelf_index_writer_open(const char *root)
{
    if (!root) return NULL;
    size_t rl = strlen(root);
    if (rl == 0 || rl >= sizeof(((shelf_idx_writer *)0)->root)) return NULL;

    shelf_idx_writer *w = (shelf_idx_writer *)calloc(1, sizeof *w);
    if (!w) return NULL;
    memcpy(w->root, root, rl + 1);

    if (shelf_index_path(root, w->dst, sizeof w->dst) != 0) { free(w); return NULL; }
    char idir[500];
    snprintf(idir, sizeof idir, "%s", w->dst);
    char *slash = strrchr(idir, '/');
    if (!slash) { free(w); return NULL; }
    *slash = 0;
    mkdir_p(idir);   /* 失败也让 fopen 去失败（只读卡 → 不写索引，书架照常） */
    if (snprintf(w->tmp, sizeof w->tmp, "%s.tmp", w->dst) >= (int)sizeof w->tmp) { free(w); return NULL; }

    w->f = NULL;
    int fd = open(w->tmp, O_WRONLY | O_CREAT | O_TRUNC | O_NOFOLLOW, 0644);
    if (fd < 0) { free(w); return NULL; }
    w->f = fdopen(fd, "wb");
    if (!w->f) { close(fd); unlink(w->tmp); free(w); return NULL; }
    w->ok = 1;
    w->crc = 0xFFFFFFFFu;

    unsigned char h[IDX_HDR_SIZE];
    memset(h, 0, sizeof h);
    memcpy(h, IDX_MAGIC, 8);
    put32(h + 8, SHELF_INDEX_VERSION);
    put32(h + 12, IDX_HDR_SIZE);
    put32(h + 16, (uint32_t)rl);
    if (wraw(w, h, sizeof h) != 0 || wraw(w, root, rl) != 0) {
        shelf_index_writer_abort(w);
        return NULL;
    }
    return w;
}

void shelf_index_writer_dir(void *ctx, const char *rel, int64_t mtime_sec,
                            long mtime_nsec, int books_here)
{
    shelf_idx_writer *w = (shelf_idx_writer *)ctx;
    if (!w || !w->ok || !rel || books_here < 0) return;
    /* v2：只落 Books 根（""）与一级目录（rel 不含 '/'）——真机冷缓存下 434 次 stat=3.15s，
     * 4 次≈20ms；level-2+ 的 mtime 不参与校验（深层变更靠「刷新」，见头文件）。 */
    if (rel[0] && strchr(rel, '/')) return;
    size_t rl = strlen(rel);
    if (rl > IDX_REL_MAX) { w->ok = 0; return; }

    unsigned char rh[IDX_DIR_REC_FIX];
    put32(rh + 0, (uint32_t)rl);
    put32(rh + 4, (uint32_t)books_here);
    put64(rh + 8, (uint64_t)mtime_sec);
    put32(rh + 16, (uint32_t)mtime_nsec);
    put32(rh + 20, 0);
    if (wraw(w, rh, sizeof rh) != 0) return;
    if (rl > 0 && wraw(w, rel, rl) != 0) return;
    w->ndirs++;
}

int shelf_index_writer_commit(shelf_idx_writer *w, const shelf_book *books,
                              int nbooks, const shelf_scan_report *rep)
{
    if (!w) return 0;
    if (!w->ok || nbooks < 0 || (nbooks > 0 && !books) ||
        w->ndirs <= 0 || w->ndirs > 65535) {
        shelf_index_writer_abort(w);
        return 0;
    }

    size_t root_len = strlen(w->root);
    for (int i = 0; i < nbooks; i++) {
        const shelf_book *b = &books[i];
        size_t pl = strlen(b->path);
        if (pl <= root_len + 1 || strncmp(b->path, w->root, root_len) != 0 ||
            b->path[root_len] != '/') {
            /* 路径不是扫描 root 下的相对路径（被截断/异常）→ 宁可不写索引 */
            shelf_index_writer_abort(w);
            return 0;
        }
        const char *rel = b->path + root_len + 1;
        size_t rl = strlen(rel);
        if (rl < 1 || rl > IDX_REL_MAX || b->fmt < 0 || b->fmt > 4 ||
            b->size < 0 || (uint64_t)b->size > 0xFFFFFFFFu) {
            shelf_index_writer_abort(w);
            return 0;
        }
        unsigned char bh[IDX_BOOK_REC_FIX];
        put32(bh + 0, (uint32_t)rl);
        put32(bh + 4, (uint32_t)b->size);
        put32(bh + 8, (uint32_t)b->fmt);
        put32(bh + 12, 0);
        if (wraw(w, bh, sizeof bh) != 0 || wraw(w, rel, rl) != 0) {
            shelf_index_writer_abort(w);
            return 0;
        }
    }

    uint32_t flags = 0;
    int depth = 0;
    if (rep) {
        if (rep->truncated) flags |= 1u;
        if (rep->depth_skipped > 0) { flags |= 2u; depth = rep->depth_skipped; }
    }
    unsigned char tr[IDX_TRL_SIZE];
    put32(tr + 0, (uint32_t)w->ndirs);
    put32(tr + 4, (uint32_t)nbooks);
    put32(tr + 8, flags);
    put32(tr + 12, (uint32_t)depth);
    put32(tr + 16, w->crc ^ 0xFFFFFFFFu);
    put32(tr + 20, IDX_TRL_MAGIC);
    if (wraw(w, tr, sizeof tr) != 0) {
        shelf_index_writer_abort(w);
        return 0;
    }

    /* 刷盘后再 rename：断电最多丢 tmp，正式索引保持旧版本（仍受 mtime 校验保护）。
     * fsync 失败不算致命（FUSE/部分卡返回 EINVAL）：rename 后若内容不完整，CRC 会拒绝。 */
    if (fflush(w->f) != 0) {
        shelf_index_writer_abort(w);
        return 0;
    }
    (void)fsync(fileno(w->f));
    if (fclose(w->f) != 0) {
        w->f = NULL;
        unlink(w->tmp);
        free(w);
        return 0;
    }
    w->f = NULL;
    if (rename(w->tmp, w->dst) != 0) {
        /* exFAT 某些实现不支持「rename 覆盖已存在文件」：先删旧再改名。 */
        unlink(w->dst);
        if (rename(w->tmp, w->dst) != 0) {
            unlink(w->tmp);
            free(w);
            return 0;
        }
    }
    free(w);
    return 1;
}

void shelf_index_writer_abort(shelf_idx_writer *w)
{
    if (!w) return;
    if (w->f) fclose(w->f);
    if (w->tmp[0]) unlink(w->tmp);
    free(w);
}

void shelf_index_remove(const char *root)
{
    char ipath[500];
    if (shelf_index_path(root, ipath, sizeof ipath) == 0) unlink(ipath);
}