Bug 1306054

Summary: Enable netconsole in Fedora kernel builds
Product: [Fedora] Fedora Reporter: Shawn Starr <shawn.starr>
Component: kernelAssignee: Neil Horman <nhorman>
Status: CLOSED NOTABUG QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: rawhideCC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda, mchehab, nhorman, shawn.starr
Target Milestone: ---   
Target Release: ---   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2016-02-13 04:17:05 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:

Description Shawn Starr 2016-02-10 00:01:19 UTC
Description of problem:
It's difficult to debug issues with kernels when netconsole is built as a module only. 

Version-Release number of selected component (if applicable):
Any

How reproducible:
Always

Is there a specific reason this is not enabled? I can compile the kernel from SRPM but it would be nice if this was built-in.

Comment 1 Josh Boyer 2016-02-10 00:54:35 UTC
99% of Fedora installs are not going to use this.  Why would we require every machine to load the code unconditionally and sit there unused?

I'm not in favor of building this into the kernel.  Perhaps in the -debug variants at best.

Comment 2 Shawn Starr 2016-02-10 15:53:18 UTC
How about this, could we enable this for the -debug and the rawhide-kernel-nodebug ones?

Comment 3 Neil Horman 2016-02-10 19:22:19 UTC
I'm not sure I understand the difficulty.  netconsole can be loaded from the initramfs fairly easily, which is relatively early in the boot process.  The window of time between when it would be loaded as a component of a monolithic build vs a module is extremely small.  I think everyones time would be better served understanding your specific use case so that we can find a good way to address it without having to build the bits into the kernel

Comment 4 Shawn Starr 2016-02-10 22:49:43 UTC
well, in this use case, Linux 4.4 boots my W500 laptop, Linux 4.5 fails to boot with any display... and i think the i915 driver broke, but I can't debug since I can't get it to show me anything.

I have a USB-to-serial but this is bound too late on bootup by the time it locks up.

Comment 5 Neil Horman 2016-02-11 11:44:03 UTC
Serial ports over usb should be something that you enable in the bios, they should be available for use in earlyprintk, and general console access, well before netconsole would be activated, even as a builtin.  Lets start by trying to figure out whats going on there.  Do you have serial over usb support enabled in bios?

Comment 6 Shawn Starr 2016-02-11 17:55:01 UTC
this laptop is pre-UEFI I don't recall an option in the BIOS for Serial over USB support at all. It just has options for USB support/legacy.

Comment 7 Neil Horman 2016-02-11 18:13:37 UTC
many bioses have the option to redirect serial output over a usb device. That should get you access to bios over serial, and make that usb port your console device.  Even if you don't have it though, you can redirect the console to a usb port from the kernel command line.  Just stop the grub2 boot process, and manually edit the kernel command line to append this:
console=ttyUSB0,9600n8

Adjusting the usb port, baud, parity and data bits accordingly.  that will make the usb tty device your console instead of the monitor, and you can connect a usb to serial converter, and monitor boot up from another laptop using minicom or other serial terminal program.

Comment 8 Shawn Starr 2016-02-13 04:17:05 UTC
I may have a workaround, this laptop has vPro on it, so I can use SOL which is even better than any other option.

Let's close this, since I have HW can handle this.