c - 我更改fs/myext2/file.c读写操作后，我的 'cat'、 'cp'、 'echo'都被kill掉了

Question

我将 fs/ext2/复制为 fs/myext2/，并修改了我需要修改的所有内容，并且在我更改 fs/myext2/file.c 之前一切正常。当我这样改变时，效果很好 (

在file.c中添加如下代码，new_sync_write和new_sync_read函数是从fs/read_write.c中复制过来的。

我还添加了另一个头文件 linux/uio.h,

将 .read = new_sync_read_crypt, .write = new_sync_write_crypt 添加到 const struct file_operations myext2_file_operations)。

ssize_t new_sync_write_crypt(struct file *filp, const char __user *buf, size_t len, loff_t *ppos){
    char *mybuf = buf;
    int i;
    /*for(i=0;i<len;i++){
        mybuf[i] += 25;
    }*/
    printk("haha write encrypt %ld\n",len);
    return new_sync_write(filp,mybuf,len,ppos);
}

ssize_t new_sync_read_crypt(struct file *filp, char __user *buf, size_t len, loff_t *ppos){

    ssize_t ret = new_sync_read(filp,buf,len,ppos);
    int i;
    /*for(i = 0; i < len; i++)
        buf[i] -= 25;*/
    printk("haha read encrypt %ld\n",len);
    return ret;
}

“做得很好”意味着，在/mnt 下，我 echo "1234567"> test.txt，并且在 log(dmesg) 中它确实有正确的输出, 我可以 'cp', 'cat'。

但是在我移动/**/之后，一切都出错了。我不能 'cat test.txt'，它说 killed。

root@ubuntu:/mnt# cat test.txt
killed
root@ubuntu:/mnt#

在日志中:

[ 5640.036210] BUG: unable to handle kernel paging request at b733f000
[ 5640.036215] IP: [<f9072aa0>] new_sync_read_crypt+0xc0/0x100 [myext2]
[ 5640.036220] *pdpt = 0000000011da5001 *pde = 0000000004708067 *pte =     0000000000000000 
[ 5640.036222] Oops: 0002 [#21] SMP 
[ 5640.036225] Modules linked in: myext2(OE) rfcomm bnep crc32_pclmul vmw_balloon aesni_intel aes_i586 xts lrw gf128mul ablk_helper cryptd joydev input_leds serio_raw snd_ens1371 snd_ac97_codec gameport ac97_bus vmw_vsock_vmci_transport vsock snd_pcm snd_seq_midi uvcvideo videobuf2_vmalloc videobuf2_memops snd_seq_midi_event videobuf2_v4l2 snd_rawmidi btusb btrtl btbcm videobuf2_core btintel videodev bluetooth media snd_seq snd_seq_device snd_timer snd soundcore nfit 8250_fintek mac_hid vmw_vmci i2c_piix4 shpchp parport_pc ppdev lp parport autofs4 hid_generic usbhid hid psmouse vmwgfx ttm drm_kms_helper ahci libahci mptspi mptscsih pcnet32 mii syscopyarea sysfillrect sysimgblt fb_sys_fops drm mptbase scsi_transport_spi pata_acpi fjes [last unloaded: myext2]
[ 5640.036249] CPU: 0 PID: 9011 Comm: cat Tainted: G      D    OE   4.6.0 #1
[ 5640.036250] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 05/19/2017
[ 5640.036252] task: f549dc40 ti: d1da6000 task.ti: d1da6000
[ 5640.036253] EIP: 0060:[<f9072aa0>] EFLAGS: 00010206 CPU: 0
[ 5640.036255] EIP is at new_sync_read_crypt+0xc0/0x100 [myext2]
[ 5640.036256] EAX: b735f000 EBX: b733f000 ECX: d1da7f60 EDX: 00000000
[ 5640.036258] ESI: 00020000 EDI: 00000000 EBP: d1da7ed8 ESP: d1da7e88
[ 5640.036259]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[ 5640.036260] CR0: 80050033 CR2: b733f000 CR3: 30ef0120 CR4: 003406f0
[ 5640.036265] Stack:
[ 5640.036265]  00000001 00020000 d1da7f60 b733f000 00020000 00000000 00000000 00020000
[ 5640.036268]  d1da7e94 00000001 f2bdac00 00000000 00000000 00000000 00000000 00000000
[ 5640.036271]  ffa41c7b f2bdac00 f90729e0 d1da7f60 d1da7f2c c11ea28f d1da7f60 c6e50cf8
[ 5640.036274] Call Trace:
[ 5640.036277]  [<f90729e0>] ? myext2_empty_dir+0x170/0x170 [myext2]
[ 5640.036280]  [<c11ea28f>] __vfs_read+0x2f/0x100
[ 5640.036282]  [<c11ea76e>] ? rw_verify_area+0x5e/0x140
[ 5640.036284]  [<c11ea8bf>] vfs_read+0x6f/0x140
[ 5640.036286]  [<c11eb9e1>] SyS_read+0x51/0xb0
[ 5640.036288]  [<c100394d>] do_fast_syscall_32+0x8d/0x140
[ 5640.036291]  [<c17baa2e>] sysenter_past_esp+0x47/0x75
[ 5640.036292] Code: 8d 45 d8 ff 51 10 3d ef fd ff ff 89 c7 74 54 8b 45 dc 8b 4d b8 85 f6 8b 55 e0 89 01 8d 04 33 89 51 04 74 11 8d b4 26 00 00 00 00 <80> 2b 19 83 c3 01 39 d8 75 f6 89 74 24 04 c7 04 24 b5 b9 07 f9
[ 5640.036311] EIP: [<f9072aa0>] new_sync_read_crypt+0xc0/0x100     [myext2] SS:ESP 0068:d1da7e88
[ 5640.036320] CR2: 00000000b733f000
[ 5640.036322] ---[ end trace d357a556f62edff8 ]---

Answer 1

我会尝试回答。

<强>1。阅读部分:

您依赖于传递的长度，而不是返回的长度ret:

ssize_t ret = new_sync_read(filp,buf,len,ppos);

所以在缓冲区的末尾，当 len 高于要读取的其余部分时，new_sync_read 返回 ret，它低于 len 并且您的循环写入边界之外。

我的修复建议:

ssize_t new_sync_read_crypt(struct file *filp, char __user *buf, size_t len, loff_t *ppos){

    ssize_t ret = new_sync_read(filp,buf,len,ppos);        
    size_t i;
    for(i = 0; i < ret; i++)
        buf[i] -= 25;
    return ret;
}

我还会添加 if (buf) 来保护循环，以防传递 buf=NULL 和 len=-1 是合法的

<强>2。写部分:

这里的问题比较棘手。

ssize_t new_sync_write_crypt(struct file *filp, const char __user *buf, size_t len, loff_t *ppos){
    char *mybuf = buf;

在这里，您将常量强制转换为非常量，因此您可以加密，这违反了 const 规则。你不能那样做，例如，如果常量缓冲区指向一个字符串文字。

您必须使用 malloc 创建一个副本

size_t new_sync_write_crypt(struct file *filp, const char __user *buf, size_t len, loff_t *ppos){
    char *mybuf = malloc(len);
    size_t i;
    for(i=0;i<len;i++){
        mybuf[i] = buf[i]+25;
    }

    size_t r = new_sync_write(filp,mybuf,len,ppos);
    free(mybuf);
    return r;
}