Hi Jeff,
On 12/24/2012 09:07 AM, Jeff Liu wrote:
> On 12/22/2012 01:36 AM, Goffredo Baroncelli wrote:
>> On 12/21/2012 07:42 AM, Jeff Liu wrote:
[...]
>>> I don't think we should replace strlen() with strnlen() since it's
>>> totally wrong if the length of label is more than BTRFS_LABEL_SIZE -1,
>>> we can not just truncating the label and return it in this case.
>>
>> This for me is sufficient, or we could copy all the label buffer,
>> without further check:
>>
>> copy_to_user(arg, label, BTRFS_LABEL_SIZE)
> That sounds ok to me, but it's better to limit the length to be
> 'BTRFS_LABEL_SIZE - 1' bytes, or else, it would copy 256 bytes back to
> the user space if the label length is beyond or equal to the array limits.
Sorry I don't understand your reply. The ioctl has as argument an array
of BTRFS_LABEL_SIZE characters, so it should not be any problem of page
fault (with the exception of an user who passes a shorter array). So it
wouldn't be any problem to copy
BTRFS_LABEL_SIZE character, or I am missing something ?
The question is another. Is a kernel responsibility to assure that the
returned string is zero terminated ? If yes (and I think so) we should
truncate the string before the copy:
label[BTRFS_LABEL_SIZE-1] = 0;
copy_to_user(arg, label, BTRFS_LABEL_SIZE);
>>
>>
>>> Add BUG_ON(strlen(label) > BTRFS_LABEL_SIZE - 1) is reasonable instead.
>>
>> I agree with Stefan, this is not a correct use of BUG_ON; a warning is
>> sufficient (there is un-correct data read from disk).
> Maybe like following?
>
> size_t len = strlen(label);
If label (who is an array read from the disk) is not zero terminated,
this would lead to a possible page fault. I suggest to use strnlen() or
similar.
>
> if (len > BTRFS_LABEL_SIZE - 1) {
> WARN(1, "btrfs: device label has weird length %zu bytes, "
> "it will be truncated to %d bytes.\n",
> len, BTRFS_LABEL_SIZE - 1);
>
> len = BTRFS_LABEL_SIZE - 1;
> }
As message I suggest:
WARN(1, "btrfs: device label is not zero terminated, it will be
truncated to %d bytes.\n",
BTRFS_LABEL-1 );
NOTE: it is suggested to not span the string on more lines, to help
searching in the source a messages with grep. It is an exception of the
rule of the max 80 columns text width.
Happy Christmas you too
GB
>
> I'll post a new patch with those changes if no other objections.
>
> Happy Christmas.
> -Jeff
>>
>>>
>>> Thanks,
>>> -Jeff
>>>>
>>>>> + mutex_unlock(&root->fs_info->volume_mutex);
>>>>> +
>>>>> + return ret ? -EFAULT : 0;
>>>>> +}
>>>>> +
>>>>> long btrfs_ioctl(struct file *file, unsigned int
>>>>> cmd, unsigned long arg)
>>>>> {
>>>>> @@ -3797,6 +3810,8 @@ long btrfs_ioctl(struct file *file, unsigned int
>>>>> return btrfs_ioctl_qgroup_create(root, argp);
>>>>> case BTRFS_IOC_QGROUP_LIMIT:
>>>>> return btrfs_ioctl_qgroup_limit(root, argp);
>>>>> + case BTRFS_IOC_GET_FSLABEL:
>>>>> + return btrfs_ioctl_get_fslabel(file, argp);
>>>>> }
>>>>>
>>>>> return -ENOTTY;
>>>>> diff --git a/fs/btrfs/ioctl.h b/fs/btrfs/ioctl.h
>>>>> index 731e287..5b2cbef 100644
>>>>> --- a/fs/btrfs/ioctl.h
>>>>> +++ b/fs/btrfs/ioctl.h
>>>>> @@ -451,6 +451,8 @@ struct btrfs_ioctl_send_args {
>>>>> struct btrfs_ioctl_qgroup_create_args)
>>>>> #define BTRFS_IOC_QGROUP_LIMIT _IOR(BTRFS_IOCTL_MAGIC, 43, \
>>>>> struct btrfs_ioctl_qgroup_limit_args)
>>>>> +#define BTRFS_IOC_GET_FSLABEL _IOR(BTRFS_IOCTL_MAGIC, 49, \
>>>>> + char[BTRFS_LABEL_SIZE])
>>>>> #define BTRFS_IOC_GET_DEV_STATS _IOWR(BTRFS_IOCTL_MAGIC, 52, \
>>>>> struct btrfs_ioctl_get_dev_stats)
>>>>> #endif
>>>>
>>>>
>>>
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>>> the body of a message to majord...@vger.kernel.org
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>>
>>
>>
>
>
--
gpg @keyserver.linux.it: Goffredo Baroncelli (kreijackATinwind.it>
Key fingerprint BBF5 1610 0B64 DAC6 5F7D 17B2 0EDA 9B37 8B82 E0B5
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
