feat(procfs): implement /proc/pid/maps and /proc/meminfo

[lianux-mm]

Dependencies

This PR depends on the API changes in ArceOS.
I have submitted a corresponding PR to Starry-OS/arceos to expose the necessary memory interfaces.

Upstream PR:

feat(mm): expose memory interfaces for procfs support (maps & meminfo) #39
Starry-OS/arceos#39

(Note: CI will fail with a fetch error until #39 is merged and the submodule pointer is updated.)

---

Summary

This PR implements the standard Linux interfaces /proc/pid/maps and /proc/meminfo for Starry.

Crucially, it also includes significant improvements to the underlying SeqFile infrastructure to correctly handle multi-page content generation, resolving issues where output would be truncated at 4KB.

Implementation Details

1. /proc/pid/maps

• Standard Compliance: Aligned output format strictly with Linux AArch64 standards (12-digit hex addresses, path padding to column 73).
• Region Identification: Added logic to automatically identify and label special regions like [heap] and [stack].

2. /proc/meminfo

• Implemented global memory statistics reporting (Total, Free, Available).
• Connects to the new global memory statistics API exposed in axmm.

3. SeqFile Infrastructure Improvements (Critical)

• Large Output Support: Updated metadata to report a large virtual file size (1MB) instead of the physical buffer size. This bypasses the VFS layer's EOF check, allowing cat to read files larger than 4KB (e.g., a process with hundreds of VMAs).
• Safety: Added output.fill(0) in the read path to prevent dirty data leakage.
• State Machine: Refined offset tracking to ensure seamless multi-page reads.

---

Test result

1. /proc/pid/maps Validation

• Used a custom C program to allocate a large number of VMAs (via mmap) to stress-test the SeqFile pagination logic.
• Result: The system successfully handled the large output (> 4KB), correctly generating the maps file across multiple reads without truncation or duplication.

2. /proc/meminfo Validation

• Verified that /proc/meminfo is implemented and correctly reports system memory

[lianux-mm]

Test program

#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <unistd.h>
#include <errno.h>
#include <fcntl.h>

#define N_MAPS 4096
#define READ_BUF_SIZE 1024 

int main() {
    printf("=== Step 1: Spawning VMAs ===\n");
    void *addrs[N_MAPS];
    for (int i = 0; i < N_MAPS; i++) {
        int prot = (i % 2 == 0) ? PROT_READ : (PROT_READ | PROT_WRITE);
        addrs[i] = mmap(NULL, 4096, prot, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
        if (addrs[i] == MAP_FAILED) {
            perror("mmap failed");
            return 1;
        }
    }
    printf("Allocated %d VMAs.\n", N_MAPS);

    printf("\n=== Step 2: Reading /proc/self/maps ===\n");
    
    int fd = open("/proc/self/maps", O_RDONLY);
    if (fd < 0) {
        perror("Failed to open /proc/self/maps");
        return 1;
    }

    char buf[READ_BUF_SIZE];
    int total_bytes = 0;
    int n;
    

    while ((n = read(fd, buf, READ_BUF_SIZE - 1)) > 0) {
        buf[n] = '\0'; 
        printf("%s", buf);
        total_bytes += n;
    }

    if (n < 0) {
        perror("\nError reading file");
    }

    close(fd);

    printf("\n=== Step 3: Verification ===\n");
    printf("Total bytes read: %d\n", total_bytes);

    if (total_bytes > 4096) {
        printf("SUCCESS: File size > 4096 bytes. Multi-page reading works!\n");
    } else {
        printf("WARNING: File size <= 4096. Did not trigger multi-page logic.\n");
    }


    for (int i = 0; i < N_MAPS; i++) munmap(addrs[i], 4096);

    return 0;
}