linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* multiplexing /dev/console to all console= entries
@ 2018-11-05  1:45 Dusty Mabe
  2018-11-07  9:03 ` Jiri Slaby
  0 siblings, 1 reply; 3+ messages in thread
From: Dusty Mabe @ 2018-11-05  1:45 UTC (permalink / raw)
  To: Greg Kroah-Hartman, Jiri Slaby; +Cc: linux-kernel


Hi Greg, Jiri,

I'd like to bring up a question about multiplexing /dev/console to
all specified console= devices on the kernel command line rather than just
the last entry. In my personal experience using linux systems and also in
my experience dealing with users of various server images (Fedora Cloud,
Fedora Atomic Host, Fedora Server, CentOS, RHEL, etc..) it is often hard
to debug properly when something goes wrong because of a number of reasons:

- the user doesn't know the difference between a serial console and 
  a VGA console
- the user doesn't know what the settings were on the image they booted
- the user doesn't know how to access the various console devices using
  the cloud environments they are on (i.e. they know of one but not the other)
- in an attempt to share configs for disk image creation the creators add
  extra console= entries to cover various console devices and various
  architectures (see [1] where we are covering VGA/serial, hvc0, AMA0. etc)

I believe it would simplify debugging/troubleshooting with users if we
were able to multiplex /dev/console to all console= entries. This
conversation actually started as a PR to dracut [2] to get all of the
systemd messages to be printed to all console= devices, but it was 
pointed out there [3] that rather than change the the settings in dracut
it would be better to attempt to change the kernel behavior.

If a patch were submitted to enable this type of behavior would such a
patch be acceptable? Maybe it could be a compile time option and also
have a kernel arg option to enable/disable it.


Note: I apologize if this functionality already exists and I missed it
      somehow.

Note: I grabbed info from [4] about who to contact for tty subsystem 
      questions. I apologize in advance if you are the wrong people to
      contact.

Thanks!
Dusty

[1] https://pagure.io/fedora-kickstarts/blob/ee84a1650b8188cbd00d54a688838207c34906f4/f/fedora-atomic.ks#_36
[2] https://github.com/dracutdevs/dracut/pull/445
[3] https://github.com/dracutdevs/dracut/pull/445#issuecomment-424755057
[4] https://github.com/torvalds/linux/blob/82aa4671516a3203261c835e98c3eecab10c994d/MAINTAINERS#L15149-L15151


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: multiplexing /dev/console to all console= entries
  2018-11-05  1:45 multiplexing /dev/console to all console= entries Dusty Mabe
@ 2018-11-07  9:03 ` Jiri Slaby
  2018-11-07 21:34   ` Dusty Mabe
  0 siblings, 1 reply; 3+ messages in thread
From: Jiri Slaby @ 2018-11-07  9:03 UTC (permalink / raw)
  To: Dusty Mabe, Greg Kroah-Hartman; +Cc: linux-kernel

On 05. 11. 18, 2:45, Dusty Mabe wrote:
> If a patch were submitted to enable this type of behavior would such a
> patch be acceptable? Maybe it could be a compile time option and also
> have a kernel arg option to enable/disable it.

Hi,

it might kill the system on some setups where people specify more serial
(or similarly slow consoles). Output to some console is so slow that the
system does nothing else than writing to the consoles. What might work
as the first step to check it out is adding a flag to the console=
option which would mark a console as a /dev/console device too and
multiplex only those specified by the user. Then let's see how it goes
with userspace logging out in the wild...

thanks,
-- 
js
suse labs

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: multiplexing /dev/console to all console= entries
  2018-11-07  9:03 ` Jiri Slaby
@ 2018-11-07 21:34   ` Dusty Mabe
  0 siblings, 0 replies; 3+ messages in thread
From: Dusty Mabe @ 2018-11-07 21:34 UTC (permalink / raw)
  To: Jiri Slaby, Greg Kroah-Hartman; +Cc: linux-kernel



On 11/7/18 4:03 AM, Jiri Slaby wrote:
> On 05. 11. 18, 2:45, Dusty Mabe wrote:
>> If a patch were submitted to enable this type of behavior would such a
>> patch be acceptable? Maybe it could be a compile time option and also
>> have a kernel arg option to enable/disable it.
> 
> Hi,
> 
> it might kill the system on some setups where people specify more serial
> (or similarly slow consoles). Output to some console is so slow that the
> system does nothing else than writing to the consoles. What might work
> as the first step to check it out is adding a flag to the console=
> option which would mark a console as a /dev/console device too and
> multiplex only those specified by the user. Then let's see how it goes
> with userspace logging out in the wild...
> 

Thanks Jiri!

I think that seems reasonable. Do you have a good idea of what you'd like it
to look like? Some options

- console=ttyS0,primary
- console=ttyS0,mux
- console=ttyS0+
- consolex=ttyS0

These are ideas. Bad ones, probably. But just throwing out ideas. 

Dusty

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2018-11-07 21:34 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-11-05  1:45 multiplexing /dev/console to all console= entries Dusty Mabe
2018-11-07  9:03 ` Jiri Slaby
2018-11-07 21:34   ` Dusty Mabe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).