Bug 102841

Please attach your acpidump:
# acpidump > acpidump.txt
Thanks. There may be a chance we can disable this feature in kernel to avoid this problem perhaps.

(In reply to dblack from comment #0)
> Dell XPS 15 2014 model wakes from sleep when intel SRT(Smart Response
> Technology) is enabled.

do you mean the intel SRT is enabled in BIOS, or there is a driver in Linux kernel?

> Please let me know what information would be helpful
> to fix this issue. The kernel I am running has intel smart connect enabled.
> The issue also appears using the current ubuntu 15.04 kernel.

please attach the output of "cat /proc/acpi/wakeup"

> do you mean the intel SRT is enabled in BIOS, or there is a driver in Linux
> kernel?

I mean intel SRT is enabled in bios. I have CONFIG_INTEL_SMARTCONNECT compiled as a module.

> please attach the output of "cat /proc/acpi/wakeup"

Will do.

Created attachment 186251 [details]
/proc/acpi/

acpidump and dmesg please.

ping

Created attachment 188711 [details]
acpidump

I have attached an acpidump output. However, I have not attached dmesg output yet because I will want to reproduce the bug (the issue may have been fixed in newer kernels than the ubuntu 15.04 kernels that I am using) first.

s/attached an/attached/

I cannot reproduce this issue with SRT enabled on a 4.2.3 kernel.

Created attachment 190581 [details]
dmesg

I managed to reproduce the issue with a 4.2.3 kernel. I have attached dmesg output.

There are some suspend/resume messages in the dmesg, is it that the resume is not intended but by the machine itself(probably due to SRT)?

Since there are multiple suspend/resume cycle, is it that you start the suspend process after the machine resumed itself?

The first suspend was intended but the resume that woke the system up following it was *not* intended. The later suspends and resumes were intended.
When the system failed to suspend properly ->
> [25895.878099] pciehp 0000:00:1c.0:pcie04: link training error: status 0x1001
[25895.878101] pciehp 0000:00:1c.0:pcie04: Failed to check link status

appear on the screen.

s/appear on/appeared on/

As it takes days to reproduce the problem with Intel SRT enabled, I'd propose to spend another a couple of days to reproduce the problem with SRT disabled. Because I don't see why Intel SRT is related...

It doesn't take days to reproduce this problem. It is as simple as booting with SRT enabled. If I boot with SRT disabled then I cannot reproduce this issue. If you have some kernel/acpi debug options I can/should enable to make diagnosing this issue easier please let me know.

please attach the acpidump both with Intel SRT enabled&disabled in the BIOS.

(In reply to dblack from comment #3)
> > do you mean the intel SRT is enabled in BIOS, or there is a driver in Linux
> kernel?
> 
> I mean intel SRT is enabled in bios. I have CONFIG_INTEL_SMARTCONNECT
> compiled as a module.
> 
BTW, I don't know if there is kernel driver for this SRT feature, but at least CONFIG_INTEL_SMARTCONNECT is not the one.

Created attachment 209681 [details]
acpidump_srt_disabled_in_bios

Created attachment 209691 [details]
acpidump_srt_enabled_in_bios

Having the same laptop, I wonder whether this is related to my bug: https://bugzilla.kernel.org/show_bug.cgi?id=116761

Have you tried upgrading you BIOS? You’re apparently at A03, while the current one is A09. You can upgrade from a FreeDOS USB stick by booting in the simplest mode of it (don’t remember the name, but it’s quite obvious AFAIR).

No I have not tried updating the bios as disabling Intel SRT fixes the problem for me.

I've checked the acpidump file, and there is nothing suspicious.

And I don't think there is a driver in Linux kernel that uses this SRT feature.
And I agree that the driver below may be related.
config INTEL_SMARTCONNECT
        tristate "Intel Smart Connect disabling driver"
        depends on ACPI
        ---help---
          Intel Smart Connect is a technology intended to permit devices to
          update state by resuming for a short period of time at regular
          intervals. If a user enables this functionality under Windows and
          then reboots into Linux, the system may remain configured to resume
          on suspend. In the absence of any userspace to support it, the system
          will then remain awake until something triggers another suspend.

          This driver checks to determine whether the device has Intel Smart
          Connect enabled, and if so disables it.

So please build in this intel-smartconnect driver and check if the problem still exists.
If yes, I think this is a BIOS issue that we should raise to Dell instead, and this bug report should be closed.
If no, then this seems to be a configuration issue, and this bug report still should be closed.

Thus close the bug report for now. Please feel free to reopen it if you still have any questions.