[REGRESSION] DELL XPS 9550 with SM951 NVME SSD quits the bus in battery mode (new in 4.18/4.19)

[REGRESSION] DELL XPS 9550 with SM951 NVME SSD quits the bus in battery mode (new in 4.18/4.19)

[Bockholdt Arne]

Hi all,

I've noticed a regression in the 4.18/4.19 kernel lines with my Dell
XPS 9550 laptop and the SM951 NVME SSD. Only in battery mode the SSD
seems to quit from the PCIe bus and I need to reboot the device. The
laptop is updated to the latest BIOS 1.80. When the device is connected
to AC power the error seems to vanish. It seems to be easily
reproducible by starting a VM via Qemu/KVM, after some seconds the
device and the file system vanishes.
I've tried to disable APST completely by kernel command line parameter
"nvme_core.default_ps_max_latency_us=0" but it doesn't seem to help. 

4.17 doesn't seem to be affected, tested with 4.18.6 and 4.19-rc5.

Is this a known regression? If you need more information, please let me
know.

Kind regards,

	Arne Bockholdt

[REGRESSION] DELL XPS 9550 with SM951 NVME SSD quits the bus in battery mode (new in 4.18/4.19)

[Keith Busch]

On Tue, Sep 25, 2018@07:31:40AM +0000, Bockholdt Arne wrote:
> Hi all,
> 
> I've noticed a regression in the 4.18/4.19 kernel lines with my Dell
> XPS 9550 laptop and the SM951 NVME SSD. Only in battery mode the SSD
> seems to quit from the PCIe bus and I need to reboot the device. The
> laptop is updated to the latest BIOS 1.80. When the device is connected
> to AC power the error seems to vanish. It seems to be easily
> reproducible by starting a VM via Qemu/KVM, after some seconds the
> device and the file system vanishes.
> I've tried to disable APST completely by kernel command line parameter
> "nvme_core.default_ps_max_latency_us=0" but it doesn't seem to help. 
> 
> 4.17 doesn't seem to be affected, tested with 4.18.6 and 4.19-rc5.
> 
> Is this a known regression? If you need more information, please let me
> know.

This regression sounds external to the nvme driver. Would it be possible
to bisect between 4.17 and 4.18?

[REGRESSION] DELL XPS 9550 with SM951 NVME SSD quits the bus in battery mode (new in 4.18/4.19)

[Bockholdt Arne]

On Tue, 2018-09-25@08:04 -0600, Keith Busch wrote:
> On Tue, Sep 25, 2018@07:31:40AM +0000, Bockholdt Arne wrote:
> 
> Hi all,
> 
> I've noticed a regression in the 4.18/4.19 kernel lines with my Dell
> XPS 9550 laptop and the SM951 NVME SSD. Only in battery mode the SSD
> seems to quit from the PCIe bus and I need to reboot the device. The
> laptop is updated to the latest BIOS 1.80. When the device is
> connected
> to AC power the error seems to vanish. It seems to be easily
> reproducible by starting a VM via Qemu/KVM, after some seconds the
> device and the file system vanishes.
> I've tried to disable APST completely by kernel command line
> parameter
> "nvme_core.default_ps_max_latency_us=0" but it doesn't seem to help. 
> 
> 4.17 doesn't seem to be affected, tested with 4.18.6 and 4.19-rc5.
> 
> Is this a known regression? If you need more information, please let
> me
> know.
> 
> This regression sounds external to the nvme driver. Would it be
> possible
> to bisect between 4.17 and 4.18?


This regression happens on my day to day working laptop only and it
would be to time intensive and dangerous to bisect it, sorry. Maybe I
could provide log files if you point me to the right one and tell me if
I should enable some kind of debug flags.

Thanks,

   Arne