Hi, On 26-12-17 19:11, Shaun Assam wrote: Thank you for tracking this down. I've just looked at the source-code implementing this and by default enabling this does not change anything except for a couple of (Dell) systems in a DMI quirk list or when someone explicitly asks for this on the kernel cmdline, so turning this on seems safe to me. But the ultimate decision on this is up to the Fedora kernel team (when they get back from their holidays). Regards, Hans

Hi Laura, Thanks for the clarification regarding the ACPI specs. If setting the acpi_rev to "yes" causes any issues, could the option be enabled but set to "no"? This would allow users the ability to turn it "on" with kernel parameters in GRUB. Regards, Shaun Assam From: Laura Abbott Sent: Wednesday, January 3, 11:24 AM Subject: Re: Fix for rhbz #1409801 To: Shaun Assam, kernel(a)lists.fedoraproject.org On 12/26/2017 10:11 AM, Shaun Assam wrote: > Hello Kernel Maintainers, > > I would like to submit a fix for bug #1409801 regarding the ethernet > adapter not working on Dell Latitude 3350 (https://bugzilla.redhat.com/ > show_bug.cgi?id=1409801). This issue is only related to the Fedora > kernels from 4.4+, and does not affect the upstream kernels used by > other distributions like Ubuntu or Arch. > > This issue can be resolved by turning on the ACPI_REV_OVERRIDE option > in the kernel config files for i686 and x86_64. Currently the > configuration is set to: > > # CONFIG_ACPI_REV_OVERRIDE_POSSIBLE is not set > > By changing it to: > > CONFIG_ACPI_REV_OVERRIDE_POSSIBLE=y > > The following boot errors no longer occur: > > [ 1.972457] pci 0000:00:1c.2: Error enabling bridge (-16), > continuing > [ 1.973304] r8169 0000:03:00.0 (unnamed net_device) (uninitialized): > enable failure > > In addition to fixing the ethernet issue, it also fixes an issue with > the Bluetooth device on the laptop automatically switching on after > shutdown/reboot (not when resuming from sleep/hibernate). The Bluetooth > icon would appear in the GNOME System Menu and would have to manually > be turned off. Since applying the ACPI_REV fix, this is no longer a > problem. I have been testing this change for the last three days and > have not encountered any issues resulting from this fix. > > The issue regarding this bug has been patched in the upstream kernel > for sometime and is already included in the code for current versions > of the Fedora kernel (see drivers/acpi/blacklist.c for details -- the > code applies to Dell laptops/computers). > > I would like to request this fix be included in future Fedora kernel > releases so we may close this bug and possibly fix other, opened, bugs > related to this configuration. I have attached the config files to this > email for your convenience (around line 75). So for the longest time we didn't want to turn this option on. While it certainly fixes some problems, overriding the ACPI table like this is questionable from an ACPI spec perspective. When the option was first introduced, it was only supposed to be for the Dell XPS 13 for a sound issue and only until more of the sound framework was updated. Clearly other machines have been added since then and it doesn't seem to be going away. The benefits to users outweigh my personal opinions at this point (it's still a complete hack) so I'll turn it on in rawhide. We're close enough to a new kernel that we can pick it up for F26/F27 when the rebase happens. Thanks, Laura