On 9/4/19 8:09 PM, Jeff Layton wrote:
It seems much similar to
https://lore.kernel.org/r/20190419084810.63732-1-houtao1@xxxxxxxxxx/
Thanks, yes two same symptoms with different use cases.
But except the fact that here, we see the issue with GPFS quite
frequently. So let's hope that we could have some solution to this
problem in upstream.
Agreed. Looks a lot like the same issue.
From the thread:-
>> We could simply use d_really_is_negative() there, avoiding all that
>> mess. If and when we get around to whiteouts-in-dcache (i.e. if
>> unionfs series gets resurrected), we can revisit that
I didn't get this part. Does it mean, d_really_is_negative can only be
used, once whiteouts-in-dcache series is resurrected?
If yes, meanwhile could we have any other solution in place?
I think Al was saying that you could change this to use
d_really_is_negative now but the whiteouts-in-dcache series would have
to deal with that. That series seems to be stalled for the time being,
so I wouldn't let it stop you from changing this.
It wouldn't hurt to put in some comments with this change too, to make
sure everyone understands why that's being used there.
Thanks Jeff. I did look at the thread again and yes you are right.
I have created the patch and added some comments as per your suggestion.
Once it is tested again internally (although it does looks like it
should work fine), I will post the patch.
Appreciate your help!!
-ritesh
