/* scan2.c - find qwords whose value points into a /dev/nvidia* mapping of
   the target process, and report the u32 stored there (semaphore hunt).
   usage: scan2 PID [max_print] */
#define _GNU_SOURCE
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdint.h>
#include <fcntl.h>
#include <unistd.h>

#define MAXR 4096
static struct { uint64_t lo, hi; char name[64]; } nv[MAXR], wr[MAXR];
static int nnv, nwr;

static int in_nv(uint64_t a, int *idx)
{
    for (int i = 0; i < nnv; i++)
        if (a >= nv[i].lo && a < nv[i].hi) { if (idx) *idx = i; return 1; }
    return 0;
}

static void load_maps(int pid)
{
    char path[64], line[1024];
    snprintf(path, sizeof path, "/proc/%d/maps", pid);
    FILE *f = fopen(path, "r");
    if (!f) { perror("maps"); exit(1); }
    while (fgets(line, sizeof line, f)) {
        uint64_t lo, hi;
        char perms[16], name[128] = "";
        if (sscanf(line, "%lx-%lx %15s %*s %*s %*s %127[^\n]", &lo, &hi, perms, name) < 3)
            continue;
        if (strchr(perms, 'w') && nwr < MAXR) { wr[nwr].lo = lo; wr[nwr].hi = hi; nwr++; }
        if (strstr(line, "nvidia") && nnv < MAXR) {
            nv[nnv].lo = lo; nv[nnv].hi = hi;
            snprintf(nv[nnv].name, sizeof nv[nnv].name, "%s", name);
            nnv++;
        }
    }
    fclose(f);
}

int main(int argc, char **argv)
{
    if (argc < 2) { fprintf(stderr, "usage: scan2 PID [max_print]\n"); return 1; }
    int pid = atoi(argv[1]);
    int maxp = argc > 2 ? atoi(argv[2]) : 200;
    load_maps(pid);
    printf("writable=%d nvidia=%d\n", nwr, nnv);

    char mp[64];
    snprintf(mp, sizeof mp, "/proc/%d/mem", pid);
    int fd = open(mp, O_RDONLY);
    if (fd < 0) { perror("mem"); return 1; }

    size_t CH = 4 << 20;
    uint8_t *buf = malloc(CH + 64);
    int printed = 0, scanned = 0;
    for (int r = 0; r < nwr; r++) {
        uint64_t a = wr[r].lo;
        while (a < wr[r].hi) {
            size_t n = (size_t)((wr[r].hi - a) < CH ? (wr[r].hi - a) : CH);
            ssize_t got = pread(fd, buf, n, (off_t)a);
            if (got <= 0) {
                if (n > (256u << 10)) { n = 256u << 10; continue; }
                a += 4096; continue;
            }
            if ((size_t)got < n) n = got;
            n -= n % 8;
            uint64_t *w = (uint64_t *)buf;
            for (size_t i = 0; i < n / 8; i++, scanned++) {
                uint64_t v = w[i];
                int idx;
                if (!in_nv(v, &idx)) continue;
                uint32_t val = 0;
                if (pread(fd, &val, 4, (off_t)v) != 4) continue;
                if (val > 0x100000) continue;
                if (printed < maxp)
                    printf("at 0x%lx -> 0x%lx (%s) val=%u\n", a + i * 8, v, nv[idx].name, val);
                printed++;
            }
            a += n;
        }
    }
    printf("scanned=%d printed=%d\n", scanned, printed);
    return 0;
}