[PATCH] remove GPL restriction from set_dumpable()

[PATCH] remove GPL restriction from set_dumpable()

[Jiri Kosina]

From: Jiri Kosina <jkosina@suse.cz>

remove GPL restriction from set_dumpable()

Commit 6c5d5238 introduced a set_dumpable() function that replaced the 
direct access to mm_struct->dumpable. I don't think there is any reason to 
restrict this function to EXPORT_SYMBOL_GPL() -- previously any module 
could modify current->mm->dumpable without any resitrictions, so it makes 
a little sense to turn this into 'internal interface' at once.

There in fact are 3rd party modules that modify the dumpable flag, and 
this patch should fix the situation for them once again (for example 
vmware).

Signed-off-by: Jiri Kosina <jkosina@suse.cz>

diff --git a/fs/exec.c b/fs/exec.c
index 073b0b8..a204720 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1699,7 +1699,7 @@ void set_dumpable(struct mm_struct *mm, int value)
 		break;
 	}
 }
-EXPORT_SYMBOL_GPL(set_dumpable);
+EXPORT_SYMBOL(set_dumpable);
 
 int get_dumpable(struct mm_struct *mm)
 {

Re: [PATCH] remove GPL restriction from set_dumpable()

[Arjan van de Ven]

On Sun, 14 Oct 2007 13:04:31 +0200 (CEST)
Jiri Kosina <jkosina@suse.cz> wrote:

> From: Jiri Kosina <jkosina@suse.cz>
> 
> remove GPL restriction from set_dumpable()
> 
> Commit 6c5d5238 introduced a set_dumpable() function that replaced
> the direct access to mm_struct->dumpable. I don't think there is any
> reason to restrict this function to EXPORT_SYMBOL_GPL() -- previously
> any module could modify current->mm->dumpable without any
> resitrictions, so it makes a little sense to turn this into 'internal
> interface' at once.
> 

but modules really have no business setting process's dumpable flags...
other than very internal modules.

Re: [PATCH] remove GPL restriction from set_dumpable()

[Christoph Hellwig]

On Sun, Oct 14, 2007 at 01:04:31PM +0200, Jiri Kosina wrote:
> From: Jiri Kosina <jkosina@suse.cz>
> 
> remove GPL restriction from set_dumpable()
> 
> Commit 6c5d5238 introduced a set_dumpable() function that replaced the 
> direct access to mm_struct->dumpable. I don't think there is any reason to 
> restrict this function to EXPORT_SYMBOL_GPL() -- previously any module 
> could modify current->mm->dumpable without any resitrictions, so it makes 
> a little sense to turn this into 'internal interface' at once.
> 
> There in fact are 3rd party modules that modify the dumpable flag, and 
> this patch should fix the situation for them once again (for example 
> vmware).

Nack, the just shouldn't do such things at all.  I start to get really
sick of patches adding random exports everywhere.

Re: [PATCH] remove GPL restriction from set_dumpable()

[Jiri Kosina]

On Sun, 14 Oct 2007, Christoph Hellwig wrote:

> > Commit 6c5d5238 introduced a set_dumpable() function that replaced the 
> > direct access to mm_struct->dumpable. I don't think there is any 
> > reason to restrict this function to EXPORT_SYMBOL_GPL() -- previously 
> > any module could modify current->mm->dumpable without any 
> > resitrictions, so it makes a little sense to turn this into 'internal 
> > interface' at once.
> Nack, the just shouldn't do such things at all.  I start to get really 
> sick of patches adding random exports everywhere.

I actually don't care that much whether this is merged or not. In fact the 
function itself is pretty trivial and standalone, so any 3rd party module 
willing to modify current->mm->flags can just reimplement it line-by-line 
themselves without breaking the license anyway ...

My main point here was that we should probably better document somewhere 
what are the intended usage scenarios for EXPORT_SYMBOL() vs. 
EXPORT_SYMBOL_GPL(). "Really internal interface" seems a little bit vague 
to me.

-- 
Jiri Kosina
SUSE Labs