From mboxrd@z Thu Jan  1 00:00:00 1970
From: hch@lst.de (Christoph Hellwig)
Date: Fri, 17 May 2019 11:05:21 +0200
Subject: [PATCHv2 6/6] nvme-pci: Use host managed power state for suspend
In-Reply-To: <CAJZ5v0j0V10BYrME=KU1AJXGDMRUFFHiZEHQnsXhNJGPZKBSUw@mail.gmail.com>
References: <20190515163625.21776-6-keith.busch@intel.com>
 <CAJZ5v0iC44O_gLMHnLui+tH+BuHKJzh_WC7PV3JFJwhj041=1A@mail.gmail.com>
 <20190516142657.GD23333@localhost.localdomain>
 <70235CA3-0FBB-4A06-996F-647A0D95C6D0@canonical.com>
 <64e8e0252a4042b99dd3d0def15b1780@AUSX13MPC105.AMER.DELL.COM>
 <20190516193822.GA23853@localhost.localdomain>
 <CAJZ5v0gBQVjy70CbA6CzkkjsGQYGeO6fDjJimvadM5_oq=3qeA@mail.gmail.com>
 <20190516203950.GB23853@localhost.localdomain>
 <CAJZ5v0hZSiQuuq2+P+uLd3uE=ruuTW+5DmAdjqcr39=7poUe-g@mail.gmail.com>
 <CAJZ5v0j0V10BYrME=KU1AJXGDMRUFFHiZEHQnsXhNJGPZKBSUw@mail.gmail.com>
Message-ID: <20190517090521.GA15509@lst.de>

On Fri, May 17, 2019@10:39:19AM +0200, Rafael J. Wysocki wrote:
> I forgot about one thing which is relevant here, sorry about that.

While we got your attention, let me repeat two questions / requests
that seem to have got lost:

 (1) in what system power states are the devices not allowed to
     every use DMA to access host memory?  Disabling the host memory
     buffer for NVMe devices can be somewhat expensive, so we'd prefer
     to only do that if we really have to
 (2) can we get some good kerneldoc comments for
     pm_suspend_via_firmware and pm_suspend_via_s2idle explaining
     when exactly they will return true, and how a driver can make
     use of these facts?  Right now these appear a little too much
     black magic to be useful