Yes, but apparently there is no space in the NFSv3 protocol used to encode the information anywhere else. This is based on the words of the article, not knowledge. Even after the problem is known, the extra bits still go somewhere into the readdir() cookie, only now they're stored in some different way. Eww.
Posted Mar 29, 2013 9:48 UTC (Fri) by dvdeug (subscriber, #10998)
[Link]
Sure, I get that. But if you get an opaque id that takes 64 bits, and you don't preserve all 64 bits, you've broken the API, not the kernel. Don't mumble things about offsets and your constraints; it's your fault. At the very least, it never would have got into an infinite loop if they'd checked their assumption on every offset they got from the kernel.
