On 7/23/19 4:06 PM, David Miller wrote:

From: Shannon Nelson <redacted>
Date: Tue, 23 Jul 2019 15:50:43 -0700

quoted

On 7/23/19 2:33 PM, David Miller wrote:

quoted

Generally interface address changes are expected to be synchronous.

Yeah, this bothers me a bit as well, but the address change calls come
in under spin_lock_bh(), and I'm reluctant to make an AdminQ call
under the _bh that could block for a few seconds.

So it's not about memory allocation but rather the fact that the device
might take a while to complete?

Memory allocation may or may not be involved, but yes, mainly we're 
doing another spin_lock on a firmware command that waits for an ACK or 
ERROR answer, and in extreme cases could possibly timeout on a dead 
firmware.  I know that i40e and ice do much the same thing, and I 
believe mlx5 as well, for the same reasons.  I suspect others do as well.

Can you start the operation synchronously yet complete it async?

This could be possible, but would likely require a bunch more messy 
logic to track async AdminQ requests, that otherwise is unnecessary.

sln