[Pvfs2-developers] patches: permission/acl bug fixes

[Phil Carns]

Sam Lang wrote:
> 
> On Mar 21, 2007, at 8:35 AM, Phil Carns wrote:
> 
>> Sam Lang wrote:
>>
>>>> acl-check-assert.patch:
>>>> ------------------------
>>>
>>> It seems like it should be possible to do that format checking of  
>>> the  acl when the system.posix_acl_access extended attribute is  
>>> set.  Does  it make sense to add a callouts framework to set-eattr  
>>> to do format  checking for specific xattrs?
>>
>>
>> I'm not sure- maybe?  I don't actually know how the file system  that 
>> triggered this problem got bad acls in the first place.
> 
> 
> I thought it would be easy to add some framework code to do the  
> callouts I was thinking of, but it ended up being a little more  
> complicated, and I started feeling like I was painting over the shed,  
> but not even with a different color.  But at that point though I was  
> already too far in to go back.  So I've attached a patch, which I'm  not 
> sure was worth it, but might be in the long run if we get more  extended 
> attributes in pvfs and want to do proper checking of their  keys and 
> values.
> 
> As for the bad acls showing up, it probably didn't happen on the set- 
> eattr anyway, so this patch isn't going to help that.  Most likely  
> there's a bug somewhere else in the code that we haven't found yet.

I like the structure of the patch in that it cleans the xattr namespace 
translation goop out of the state machines.

I think that this may break something, though.  Could you confirm with 
the test/automated/tacl_xattr.sh script on your end?  I may have done 
something wrong, but this test appears to work fine without the attached 
patch, but then has some failures if I apply it.

As a side note, it looks like there are still some noisy messages coming 
from the kernel related to ACLs regardless of whether the patches are 
applied or not.  The tacl_xattr.sh script generates several of these in 
dmesg for me:

pvfs2_acl_chmod: get acl (access) failed with 0

They also pop out during LTP test runs.

-Phil