Re: [PATCH] nvme-pci: Avoid to go into d3cold if device can't use npss. - Kai-Heng Feng

From: Kai-Heng Feng <kai.heng.feng@canonical.com>
To: Christoph Hellwig <hch@lst.de>
Cc: Keith Busch <kbusch@kernel.org>, Koba Ko <koba.ko@canonical.com>,
	Jens Axboe <axboe@fb.com>, Sagi Grimberg <sagi@grimberg.me>,
	linux-nvme <linux-nvme@lists.infradead.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Henrik Juul Hansen <hjhansen2020@gmail.com>,
	Bjorn Helgaas <bhelgaas@google.com>,
	Linux PCI <linux-pci@vger.kernel.org>
Subject: Re: [PATCH] nvme-pci: Avoid to go into d3cold if device can't use npss.
Date: Wed, 26 May 2021 22:21:59 +0800	[thread overview]
Message-ID: <CAAd53p4f2ZFsVRv-Q9maPBSD_uGjj7FoYKYy9MGjBPc6chk_1Q@mail.gmail.com> (raw)
In-Reply-To: <20210526125942.GA25080@lst.de>

On Wed, May 26, 2021 at 8:59 PM Christoph Hellwig <hch@lst.de> wrote:
>
> On Wed, May 26, 2021 at 08:11:41PM +0800, Kai-Heng Feng wrote:
> > On Wed, May 26, 2021 at 10:49 AM Keith Busch <kbusch@kernel.org> wrote:
> > >
> > > On Wed, May 26, 2021 at 10:02:27AM +0800, Koba Ko wrote:
> > > > On Tue, May 25, 2021 at 3:44 PM Christoph Hellwig <hch@lst.de> wrote:
> > > > >
> > > > > On Thu, May 20, 2021 at 11:33:15AM +0800, Koba Ko wrote:
> > > > > > After resume, host can't change power state of the closed controller
> > > > > > from D3cold to D0.
> > > > >
> > > > > Why?
> > > > As per Kai-Heng said, it's a regression introduced by commit
> > > > b97120b15ebd ("nvme-pci:
> > > > use simple suspend when a HMB is enabled"). The affected NVMe is using HMB.
> > >
> > > That really doesn't add up. The mentioned commit restores the driver
> > > behavior for HMB drives that existed prior to d916b1be94b6d from kernel
> > > 5.3. Is that NVMe device broken in pre-5.3 kernels, too?
> >
> > Quite likely. The system in question is a late 2020 Ice Lake laptop,
> > so it was released after 5.3 kernel.
>
> This is just a mess.  We had to disable the sensible power state based
> suspend on these systems because Intel broke it by just cutting the power
> off.  And now the shutdown based one doesn't work either because it can't
> handle d3cold.  Someone we need to stop Intel and the integrators from
> doing stupid things, and I'm not sure how.

To be fair, resuming the NVMe from D3hot is much slower than keep it
at D0, which gives us a faster s2idle resume time. And now AMD also
requires s2idle on their latest laptops.

And it's more like NVMe controllers don't respect PCI D3hot.

>
> But degrading all systems even more is just a bad idea, so I fear we'll
> need a quirk again.  Can you figure out by switching the cards if this
> is the fault of the platform or the nvme device?

Here's the original bug report:
https://bugs.launchpad.net/bugs/1912057

Because the NVMe continues to work after s2idle and the symbol is
rather subtle, so I suspect this is not platform or vendor specific.
Is it possible to disable DMA for HMB NVMe on suspend?

Kai-Heng

>
> >
> > Kai-Heng
> ---end quoted text---