filesystem encryption problem.

filesystem encryption problem.

[Amir Hezarkhani]

hello
I am working on adding a simple encryption to file contents in ext4 driver
(for learning purposes) I added simple XOR encryption to aio_read and
aio_write functions and it worked until I faced this problem:

when I open a file in encrypted filesystem using VIM text editor and when I
try to save it it gives me this error:

>>pointer block id wrong
>>can not find line 1

and it just corrupts the entire file!

this is my aio_write function:

aio_write_enc(struct kiocb *iocb, const struct iovec *iov,
        unsigned long nr_segs, loff_t pos)
{
    size_t i;
    ssize_t ret;
    char *data=vmalloc(sizeof(char)*iov->iov_len);
    copy_from_user(data,iov->iov_base,iov->iov_len);

    for(i=0;i<iov->iov_len;i++)
    {
        data[i]^=5;
    }
    struct iovec iov_enc= { .iov_base = iov->iov_base, .iov_len =
iov->iov_len };

    copy_to_user(iov_enc.iov_base,data,iov->iov_len);
    ret=ext4_file_write(iocb,&iov_enc,nr_segs,pos);
    vfree(data);
    return ret;
}

this just changes the data and then calls original function.

is there anything wrong with this function? can anyone help me?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20150712/66e6c67a/attachment.html 

filesystem encryption problem.

[Freeman Zhang]

-------- Original Message --------
> hello
> I am working on adding a simple encryption to file contents in ext4 driver
> (for learning purposes) I added simple XOR encryption to aio_read and
> aio_write functions and it worked until I faced this problem:
> 
> when I open a file in encrypted filesystem using VIM text editor and when I
> try to save it it gives me this error:
> 
>>> pointer block id wrong
>>> can not find line 1
> 
> and it just corrupts the entire file!
> 
> this is my aio_write function:
> 
> aio_write_enc(struct kiocb *iocb, const struct iovec *iov,
>         unsigned long nr_segs, loff_t pos)
> {
>     size_t i;
>     ssize_t ret;
>     char *data=vmalloc(sizeof(char)*iov->iov_len);
>     copy_from_user(data,iov->iov_base,iov->iov_len);
> 
>     for(i=0;i<iov->iov_len;i++)
>     {
>         data[i]^=5;
>     }
>     struct iovec iov_enc= { .iov_base = iov->iov_base, .iov_len =
> iov->iov_len };
> 
>     copy_to_user(iov_enc.iov_base,data,iov->iov_len);
>     ret=ext4_file_write(iocb,&iov_enc,nr_segs,pos);
>     vfree(data);
>     return ret;
> }
> 
> this just changes the data and then calls original function.
> 
> is there anything wrong with this function? can anyone help me?
> 
> 
> 
Hi Amir,

I'm not quite sure about what's wrong with your function, but here are
two suggestions I got from the list when I did similar things:

1. wrapfs
2. ecryptfs

I think you should check these two stackable filesystems if you haven't.

Hope this can help a little bit!

Freeman

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
Url : http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20150712/a9929b1d/attachment.bin 

filesystem encryption problem.

[Rohan Puri]

On Sun, Jul 12, 2015 at 8:08 PM, Freeman Zhang <freeman.zhang1992@gmail.com>
wrote:

> -------- Original Message --------
> > hello
> > I am working on adding a simple encryption to file contents in ext4
> driver
> > (for learning purposes) I added simple XOR encryption to aio_read and
> > aio_write functions and it worked until I faced this problem:
> >
> > when I open a file in encrypted filesystem using VIM text editor and
> when I
> > try to save it it gives me this error:
> >
> >>> pointer block id wrong
> >>> can not find line 1
> >
> > and it just corrupts the entire file!
> >
> > this is my aio_write function:
> >
> > aio_write_enc(struct kiocb *iocb, const struct iovec *iov,
> >         unsigned long nr_segs, loff_t pos)
> > {
> >     size_t i;
> >     ssize_t ret;
> >     char *data=vmalloc(sizeof(char)*iov->iov_len);
> >     copy_from_user(data,iov->iov_base,iov->iov_len);
> >
> >     for(i=0;i<iov->iov_len;i++)
> >     {
> >         data[i]^=5;
> >     }
> >     struct iovec iov_enc= { .iov_base = iov->iov_base, .iov_len =
> > iov->iov_len };
> >
> >     copy_to_user(iov_enc.iov_base,data,iov->iov_len);
> >     ret=ext4_file_write(iocb,&iov_enc,nr_segs,pos);
> >     vfree(data);
> >     return ret;
> > }
> >
> > this just changes the data and then calls original function.
> >
> > is there anything wrong with this function? can anyone help me?
> >
> >
> >
> Hi Amir,
>
> I'm not quite sure about what's wrong with your function, but here are
> two suggestions I got from the list when I did similar things:
>
> 1. wrapfs
> 2. ecryptfs
>
> I think you should check these two stackable filesystems if you haven't.
>
> Hope this can help a little bit!
>
> Freeman
>
>
> _______________________________________________
> Kernelnewbies mailing list
> Kernelnewbies at kernelnewbies.org
> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>
>
Hi Amir,

I agree with Freeman Zhang over here. The way you are doing it is not
right. There is a mechanism to create stacks of file system and you should
go down that path.

Having said this, you should definitely debug the issue that you are
facing. Some pointers : -
1. As you have already mentioned that this is happening only for vim and
not while regular read(using cat, etc), you need to check what vim does
special to read a file. I would suggest make use of strace and do reading
with and without vim, maybe you will get something of interest.
2. re-read code to check, you might be messing up while write or read.

Apart from these some basic practices you need to follow is : -

1. check for error conditions, like you missed checking error from
vmalloc() and the below code will execute even if it failed, this should be
avoided.
2. copy_from_user & again copying back to user is in-efficient.


Enjoy life,
Rohan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20150712/b26a2e91/attachment.html 

filesystem encryption problem.

[Amir Hezarkhani]

Thank for replies. About copy_to_user and copy_from_user, whats the better
way?
I dont have much experience in kernel development but I'm trying to learn.
Can you recommend me some books, documents, etc so I can learn more about
filesystems in kernel. I am also interested to learn how mmap works because
I have problems with execution of binary files in my encrypted filesystem.
On Jul 12, 2015 8:30 PM, <kernelnewbies-request@kernelnewbies.org> wrote:

>
>
> On Sun, Jul 12, 2015 at 8:08 PM, Freeman Zhang <
> freeman.zhang1992 at gmail.com> wrote:
>
>> -------- Original Message --------
>> > hello
>> > I am working on adding a simple encryption to file contents in ext4
>> driver
>> > (for learning purposes) I added simple XOR encryption to aio_read and
>> > aio_write functions and it worked until I faced this problem:
>> >
>> > when I open a file in encrypted filesystem using VIM text editor and
>> when I
>> > try to save it it gives me this error:
>> >
>> >>> pointer block id wrong
>> >>> can not find line 1
>> >
>> > and it just corrupts the entire file!
>> >
>> > this is my aio_write function:
>> >
>> > aio_write_enc(struct kiocb *iocb, const struct iovec *iov,
>> >         unsigned long nr_segs, loff_t pos)
>> > {
>> >     size_t i;
>> >     ssize_t ret;
>> >     char *data=vmalloc(sizeof(char)*iov->iov_len);
>> >     copy_from_user(data,iov->iov_base,iov->iov_len);
>> >
>> >     for(i=0;i<iov->iov_len;i++)
>> >     {
>> >         data[i]^=5;
>> >     }
>> >     struct iovec iov_enc= { .iov_base = iov->iov_base, .iov_len =
>> > iov->iov_len };
>> >
>> >     copy_to_user(iov_enc.iov_base,data,iov->iov_len);
>> >     ret=ext4_file_write(iocb,&iov_enc,nr_segs,pos);
>> >     vfree(data);
>> >     return ret;
>> > }
>> >
>> > this just changes the data and then calls original function.
>> >
>> > is there anything wrong with this function? can anyone help me?
>> >
>> >
>> >
>> Hi Amir,
>>
>> I'm not quite sure about what's wrong with your function, but here are
>> two suggestions I got from the list when I did similar things:
>>
>> 1. wrapfs
>> 2. ecryptfs
>>
>> I think you should check these two stackable filesystems if you haven't.
>>
>> Hope this can help a little bit!
>>
>> Freeman
>>
>>
>> _______________________________________________
>> Kernelnewbies mailing list
>> Kernelnewbies at kernelnewbies.org
>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>>
>>
> Hi Amir,
>
> I agree with Freeman Zhang over here. The way you are doing it is not
> right. There is a mechanism to create stacks of file system and you should
> go down that path.
>
> Having said this, you should definitely debug the issue that you are
> facing. Some pointers : -
> 1. As you have already mentioned that this is happening only for vim and
> not while regular read(using cat, etc), you need to check what vim does
> special to read a file. I would suggest make use of strace and do reading
> with and without vim, maybe you will get something of interest.
> 2. re-read code to check, you might be messing up while write or read.
>
> Apart from these some basic practices you need to follow is : -
>
> 1. check for error conditions, like you missed checking error from
> vmalloc() and the below code will execute even if it failed, this should be
> avoided.
> 2. copy_from_user & again copying back to user is in-efficient.
>
>
> Enjoy life,
> Rohan
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20150712/111dc5bb/attachment.html 

filesystem encryption problem.

[Rohan Puri]

On 12 Jul 2015 22:20, "Amir Hezarkhani" <amir6723@gmail.com> wrote:
>
> Thank for replies. About copy_to_user and copy_from_user, whats the
better way?
> I dont have much experience in kernel development but I'm trying to
learn. Can you recommend me some books, documents, etc so I can learn more
about filesystems in kernel. I am also interested to learn how mmap works
because I have problems with execution of binary files in my encrypted
filesystem.
>
> On Jul 12, 2015 8:30 PM, <kernelnewbies-request@kernelnewbies.org> wrote:
>>
>>
>>
>> On Sun, Jul 12, 2015 at 8:08 PM, Freeman Zhang <
freeman.zhang1992@gmail.com> wrote:
>>>
>>> -------- Original Message --------
>>> > hello
>>> > I am working on adding a simple encryption to file contents in ext4
driver
>>> > (for learning purposes) I added simple XOR encryption to aio_read and
>>> > aio_write functions and it worked until I faced this problem:
>>> >
>>> > when I open a file in encrypted filesystem using VIM text editor and
when I
>>> > try to save it it gives me this error:
>>> >
>>> >>> pointer block id wrong
>>> >>> can not find line 1
>>> >
>>> > and it just corrupts the entire file!
>>> >
>>> > this is my aio_write function:
>>> >
>>> > aio_write_enc(struct kiocb *iocb, const struct iovec *iov,
>>> >         unsigned long nr_segs, loff_t pos)
>>> > {
>>> >     size_t i;
>>> >     ssize_t ret;
>>> >     char *data=vmalloc(sizeof(char)*iov->iov_len);
>>> >     copy_from_user(data,iov->iov_base,iov->iov_len);
>>> >
>>> >     for(i=0;i<iov->iov_len;i++)
>>> >     {
>>> >         data[i]^=5;
>>> >     }
>>> >     struct iovec iov_enc= { .iov_base = iov->iov_base, .iov_len =
>>> > iov->iov_len };
>>> >
>>> >     copy_to_user(iov_enc.iov_base,data,iov->iov_len);
>>> >     ret=ext4_file_write(iocb,&iov_enc,nr_segs,pos);
>>> >     vfree(data);
>>> >     return ret;
>>> > }
>>> >
>>> > this just changes the data and then calls original function.
>>> >
>>> > is there anything wrong with this function? can anyone help me?
>>> >
>>> >
>>> >
>>> Hi Amir,
>>>
>>> I'm not quite sure about what's wrong with your function, but here are
>>> two suggestions I got from the list when I did similar things:
>>>
>>> 1. wrapfs
>>> 2. ecryptfs
>>>
>>> I think you should check these two stackable filesystems if you haven't.
>>>
>>> Hope this can help a little bit!
>>>
>>> Freeman
>>>
>>>
>>> _______________________________________________
>>> Kernelnewbies mailing list
>>> Kernelnewbies at kernelnewbies.org
>>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>>>
>>
>> Hi Amir,
>>
>> I agree with Freeman Zhang over here. The way you are doing it is not
right. There is a mechanism to create stacks of file system and you should
go down that path.
>>
>> Having said this, you should definitely debug the issue that you are
facing. Some pointers : -
>> 1. As you have already mentioned that this is happening only for vim and
not while regular read(using cat, etc), you need to check what vim does
special to read a file. I would suggest make use of strace and do reading
with and without vim, maybe you will get something of interest.
>> 2. re-read code to check, you might be messing up while write or read.
>>
>> Apart from these some basic practices you need to follow is : -
>>
>> 1. check for error conditions, like you missed checking error from
vmalloc() and the below code will execute even if it failed, this should be
avoided.
>> 2. copy_from_user & again copying back to user is in-efficient.
>>
>>
>> Enjoy life,
>> Rohan
>
>
> _______________________________________________
> Kernelnewbies mailing list
> Kernelnewbies at kernelnewbies.org
> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
>

Hi Amir,

Please reply at the bottom. Regarding what's the better way would depend on
how you design stuff.

Following is my recommendation :-

For conceptual knowledge of general file systems the best would be OS book
by Prof Remzi Arpaci-Dusseau.
Excellently explained.

For linux kernel conceptual stuff get hold of Robert love Linux kernel
development.

Read lots of kernel generic filesystem code in FS dir. Lots of basic
functionality is implemented in helper functions present in this dir.

Enjoy life,
Rohan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20150712/3014d68e/attachment.html 

filesystem encryption problem.

[Rohan Puri]

On 13 Jul 2015 22:08, "Amir Hezarkhani" <amir6723@gmail.com> wrote:
>
>
> On Jul 13, 2015 1:18 AM, "Rohan Puri" <rohan.puri15@gmail.com> wrote:
> >
> > No issues, you are welcome.
> >
> > Enjoy life,
> > Rohan
> >
> > On 13 Jul 2015 01:19, "Amir Hezarkhani" <amir6723@gmail.com> wrote:
> >>
> >>
> >> On Jul 12, 2015 10:00 PM, "Rohan Puri" <rohan.puri15@gmail.com> wrote:
> >> >
> >> >
> >> > On 12 Jul 2015 22:20, "Amir Hezarkhani" <amir6723@gmail.com> wrote:
> >> > >
> >> > > Thank for replies. About copy_to_user and copy_from_user, whats
the better way?
> >> > > I dont have much experience in kernel development but I'm trying
to learn. Can you recommend me some books, documents, etc so I can learn
more about filesystems in kernel. I am also interested to learn how mmap
works because I have problems with execution of binary files in my
encrypted filesystem.
> >> > >
> >> > > On Jul 12, 2015 8:30 PM, <kernelnewbies-request@kernelnewbies.org>
wrote:
> >> > >>
> >> > >>
> >> > >>
> >> > >> On Sun, Jul 12, 2015 at 8:08 PM, Freeman Zhang <
freeman.zhang1992@gmail.com> wrote:
> >> > >>>
> >> > >>> -------- Original Message --------
> >> > >>> > hello
> >> > >>> > I am working on adding a simple encryption to file contents in
ext4 driver
> >> > >>> > (for learning purposes) I added simple XOR encryption to
aio_read and
> >> > >>> > aio_write functions and it worked until I faced this problem:
> >> > >>> >
> >> > >>> > when I open a file in encrypted filesystem using VIM text
editor and when I
> >> > >>> > try to save it it gives me this error:
> >> > >>> >
> >> > >>> >>> pointer block id wrong
> >> > >>> >>> can not find line 1
> >> > >>> >
> >> > >>> > and it just corrupts the entire file!
> >> > >>> >
> >> > >>> > this is my aio_write function:
> >> > >>> >
> >> > >>> > aio_write_enc(struct kiocb *iocb, const struct iovec *iov,
> >> > >>> >         unsigned long nr_segs, loff_t pos)
> >> > >>> > {
> >> > >>> >     size_t i;
> >> > >>> >     ssize_t ret;
> >> > >>> >     char *data=vmalloc(sizeof(char)*iov->iov_len);
> >> > >>> >     copy_from_user(data,iov->iov_base,iov->iov_len);
> >> > >>> >
> >> > >>> >     for(i=0;i<iov->iov_len;i++)
> >> > >>> >     {
> >> > >>> >         data[i]^=5;
> >> > >>> >     }
> >> > >>> >     struct iovec iov_enc= { .iov_base = iov->iov_base,
.iov_len =
> >> > >>> > iov->iov_len };
> >> > >>> >
> >> > >>> >     copy_to_user(iov_enc.iov_base,data,iov->iov_len);
> >> > >>> >     ret=ext4_file_write(iocb,&iov_enc,nr_segs,pos);
> >> > >>> >     vfree(data);
> >> > >>> >     return ret;
> >> > >>> > }
> >> > >>> >
> >> > >>> > this just changes the data and then calls original function.
> >> > >>> >
> >> > >>> > is there anything wrong with this function? can anyone help me?
> >> > >>> >
> >> > >>> >
> >> > >>> >
> >> > >>> Hi Amir,
> >> > >>>
> >> > >>> I'm not quite sure about what's wrong with your function, but
here are
> >> > >>> two suggestions I got from the list when I did similar things:
> >> > >>>
> >> > >>> 1. wrapfs
> >> > >>> 2. ecryptfs
> >> > >>>
> >> > >>> I think you should check these two stackable filesystems if you
haven't.
> >> > >>>
> >> > >>> Hope this can help a little bit!
> >> > >>>
> >> > >>> Freeman
> >> > >>>
> >> > >>>
> >> > >>> _______________________________________________
> >> > >>> Kernelnewbies mailing list
> >> > >>> Kernelnewbies at kernelnewbies.org
> >> > >>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
> >> > >>>
> >> > >>
> >> > >> Hi Amir,
> >> > >>
> >> > >> I agree with Freeman Zhang over here. The way you are doing it is
not right. There is a mechanism to create stacks of file system and you
should go down that path.
> >> > >>
> >> > >> Having said this, you should definitely debug the issue that you
are facing. Some pointers : -
> >> > >> 1. As you have already mentioned that this is happening only for
vim and not while regular read(using cat, etc), you need to check what vim
does special to read a file. I would suggest make use of strace and do
reading with and without vim, maybe you will get something of interest.
> >> > >> 2. re-read code to check, you might be messing up while write or
read.
> >> > >>
> >> > >> Apart from these some basic practices you need to follow is : -
> >> > >>
> >> > >> 1. check for error conditions, like you missed checking error
from vmalloc() and the below code will execute even if it failed, this
should be avoided.
> >> > >> 2. copy_from_user & again copying back to user is in-efficient.
>
> Yes you are right. This was the problem.
>
> >> > >>
> >> > >>
> >> > >> Enjoy life,
> >> > >> Rohan
> >> > >
> >> > >
> >> > > _______________________________________________
> >> > > Kernelnewbies mailing list
> >> > > Kernelnewbies at kernelnewbies.org
> >> > > http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
> >> > >
> >> >
> >> > Hi Amir,
> >> >
> >> > Please reply at the bottom. Regarding what's the better way would
depend on how you design stuff.
> >> >
> >> > Following is my recommendation :-
> >> >
> >> > For conceptual knowledge of general file systems the best would be
OS book by Prof Remzi Arpaci-Dusseau.
> >> > Excellently explained.
> >> >
> >> > For linux kernel conceptual stuff get hold of Robert love Linux
kernel development.
> >> >
> >> > Read lots of kernel generic filesystem code in FS dir. Lots of basic
functionality is implemented in helper functions present in this dir.
> >> >
> >> > Enjoy life,
> >> > Rohan
> >>
> >> Thanks a lot Rohan.  And sorry about the bad reply.
>
> Ok I solved the problem. As Rohan said, copy_from_user and then
copy_to_user was inefficient. So I removed 'copy to user' part and assigned
'data' to iov_enc.base . thanks guys and Happy codding.

Hi Amir,

Good to know.

Enjoy life,
Rohan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20150713/7232de5a/attachment.html