by posting code you mean I send it to you, on email right? or tech-kern
mailing list?
And in ext2fs_lookup.c
in function ext2fs_dirbadentry()
else if (reclen % 4 != 0)
error_msg = "rec_len % 4 != 0";
why does this part of code necessary ?
On Sun, Jul 24, 2016 at 10:56 PM, Christos Zoulas <[email protected]>
wrote:
> On Jul 24, 5:52pm, [email protected] (HRISHIKESH GOYAL) wrote:
> -- Subject: Re: GSoC-2016
>
> | Hello,
> |
> | sir, I am still struggling with htree index directory write support. The
> | problem that I am facing currently is in reading a directory file before
> | writing a new entry. During reading a directory, for validating purpose
> | kernel calls ext2fs_dirbadentry(). Issues were following:
> | 1. By giving certain printfs I got to know that actual length of filename
> | and e2d_namlen doesnot match.
> | 2. For some files their partial names is only saved ( like a entry has
> file
> | name "newf" which is supposed to be "newfile10") .
> | 3. And at the end ext2fs_dirbadentry() fail to validate an entry ( file
> | Name- "yyyy" and inode no - some big number. other details is in the
> | attached pic)
> |
> |
> | might be problem is due to bad writing of entries by my htree directory
> | code which has corrupted the entire directory structure. I am trying
> | creating a new file-system disc image during each testing phase.
>
> I think thst the best way is to write code that verifies that the directory
> structure is valid after every index write. This could be done by a
> specialized
> routine, or by calling the readdir routine after each write making sure
> that it works.
>
> | Another thing is I might consider building kernel by commenting
> DIAGNOSTIC
> | in kernel configuration file which may avoid full validation check. But
> | again its not the solution.
>
> This would be moving in the wrong direction.
>
> | Any suggestion will be helpful
>
> Also you can post your code and tests with instructions how to reproduce
> the
> problem and we can help you debug.
>
> christos
>
