Does struct qstr->name should be terminated ?

Does struct qstr->name should be terminated ?

[Jerome de Vivie]

I've work with "struct qstr" for a while without bothering with the last 
caracter: qstr->name[qstr->len]. I get oops from my module but after 
investigation, i found lines in fs/namei.c that are using this caracter:

In do_rename:
         /* unless the source is a directory trailing slashes give 
-ENOTDIR */
         if (!S_ISDIR(old_dentry->d_inode->i_mode)) {
                 error = -ENOTDIR;
                 if (oldnd.last.name[oldnd.last.len])
                         goto exit4;
                 if (newnd.last.name[newnd.last.len])
                         goto exit4;
         }

So, do i need to initialize this caracter when allocating new qstr ? Is 
this caracter is used in other place in the VFS ?

regards,

j.

Re: Does struct qstr->name should be terminated ?

[Jan Hudec]

[-- Attachment #1: Type: text/plain, Size: 1091 bytes --]

On Tue, Mar 23, 2004 at 11:56:01 +0100, Jerome de Vivie wrote:
> 
> I've work with "struct qstr" for a while without bothering with the last 
> caracter: qstr->name[qstr->len]. I get oops from my module but after 
> investigation, i found lines in fs/namei.c that are using this caracter:
> 
> In do_rename:
>         /* unless the source is a directory trailing slashes give 
> -ENOTDIR */
>         if (!S_ISDIR(old_dentry->d_inode->i_mode)) {
>                 error = -ENOTDIR;
>                 if (oldnd.last.name[oldnd.last.len])
>                         goto exit4;
>                 if (newnd.last.name[newnd.last.len])
>                         goto exit4;
>         }
> 
> So, do i need to initialize this caracter when allocating new qstr ? Is 
> this caracter is used in other place in the VFS ?

I'd think that it's treated as a nomral string in quite a few places, so
it needs to be NUL terminated. (But not searched the sources.)

-------------------------------------------------------------------------------
						 Jan 'Bulb' Hudec <bulb@ucw.cz>

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Does struct qstr->name should be terminated ?

[Matthew Wilcox]

On Wed, Mar 24, 2004 at 10:09:26AM +0100, Jan Hudec wrote:
> I'd think that it's treated as a nomral string in quite a few places, so
> it needs to be NUL terminated. (But not searched the sources.)

Why not?  It isn't hard to find.

fs/dcache.c:703:        str[name->len] = 0;

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain

Re: Does struct qstr->name should be terminated ?

[Bryan Henderson]

>> I'd think that it's treated as a nomral string in quite a few places, 
so
>> it needs to be NUL terminated. (But not searched the sources.)
>
>Why not?  It isn't hard to find.
>
>fs/dcache.c:703:        str[name->len] = 0;

That's technically an area where it's treated like a normal string, but 
not really relevant to the question (which is "what if I don't 
nul-terminate the thing"?  Do you have an example handy of a strcmp() or 
such that would fail if the NUL terminator were not there?

Re: Does struct qstr->name should be terminated ?

[Jerome de Vivie]

Bryan Henderson wrote:
>>>I'd think that it's treated as a nomral string in quite a few places, 
> 
> so
> 
>>>it needs to be NUL terminated. (But not searched the sources.)
>>
>>Why not?  It isn't hard to find.
>>
>>fs/dcache.c:703:        str[name->len] = 0;
> 
> 
> That's technically an area where it's treated like a normal string, but 

I thougth that qstr->name was not treated as a normal string and has
been design to speedup d_compare test with strncmp instead of strcmp.
That's why i never bother with the null termination caracter.

> not really relevant to the question (which is "what if I don't 
> nul-terminate the thing"?  Do you have an example handy of a strcmp() or 
> such that would fail if the NUL terminator were not there?

In  yourfs->d_lookup, you can reallocate qstr and pad d_name with random
caracters. After that, unexpected behavior could happen after a mv(1):
the VFS could increment qstr->len. The result is one more caracter for
some entries after a ls(1).