On Tue, 2016-04-19 at 19:53 -0600, David Ahern wrote: > > The kernel can set a flag in the response that it acknowledges the > new attribute/flag. I did that for filtering neigh dumps -- > 21fdd092acc7. >
Hm, that works, but I think it requires writing extra code, which I was kinda trying to avoid. With the patch that Emmanuel wrote, we can restrict the changes to just nla_parse(). Anyway, I think we just have to document the behaviour very precisely, and userspace can make its own decisions. Essentially, apps will have a number of choices: 1) Use the new attribute flag only with commands known to have been added after the kernel support was added. 2) Use the new attribute flag with some required attribute for existing commands, so that older kernel will not find the required attribute and will reject the operation entirely. May or may not fall back to trying the operation again without the flag. 3) Simply use the new flag and do unexpected things on kernels not supporting the rejection mechanism - not much worse than today in many cases. I guess we'll write a proper commit message and send the patch. johannes
