John Fischer wrote:
> Garrett,
>
> The timeout seems a little quick for a short work week especially
> since there are numerous folks on vacation this week. Is there a
> reason why the timeout wasn't a little longer?
I just set it for the default one week. If someone says "want more
time", or the case doesn't appear to receive review, then it will be
extended. I don't think there is any particular urgency here.
>
> Also what are the interface taxonomy classifications for the imported
> and exported interfaces? I did not see anything mentioned.
Good point. I think the imported SPCR interface in BIOS is specified by
the ACPI specification, and falls outside of PSARC classification.
I'm guessing that $CONSOLE is probably meant to be a newly exported
Committed interface.
The actual contents of $CONSOLE appear to be imported from the previous
kernel boot command arguments. I'm not sure of their existing
stability, but my first guess here is also Committed.
The changes to /etc/ttydefs are not themselves an interface, but simply
a change to the delivered defaults. So no interface changes there.
spcr should probably be Uncommitted (my opinion), but I'll let the
project team (Dan Mick) specify for sure.
Dan, can you comment here?
>
> Should we expect another case to resolve the 2 step process on
> the SPARC platform? Or should that be considered a bug that
> can be fixed at anytime?
I'd consider it a bug or rfe to be fixed anytime after this project
integrates, without a need to come back to PSARC. (Or, if needed, could
be done via a self-review case.)
- Garrett
>
> Thanks,
>
> John
>
>
> Garrett D'Amore - sun microsystems wrote:
>> Template Version: @(#)sac_nextcase 1.68 02/23/09 SMI
>> This information is Copyright 2009 Sun Microsystems
>> 1. Introduction
>> 1.1. Project/Component Working Name:
>> Serial Port Console Redirection
>> 1.2. Name of Document Author/Supplier:
>> Author: Daniel Mick
>> 1.3 Date of This Document:
>> 21 December, 2009
>> 4. Technical Description
>>
>> PROBLEM:
>>
>> x86 platforms have been able for some time to redirect BIOS and early
>> boot output to a serial port, as has the Solaris kernel; however,
>> configuration of both is separate, and error-prone. This has been
>> unavoidable, because, until recently, there's been no interface to
>> query the BIOS settings for serial console.
>>
>> Further, OpenSolaris now tries to enter graphics mode immediately in
>> Grub, and stay there during kernel boot until Xorg takes over
>> (so-called "happy face" boot). This works well on workstations with
>> VGAs and displays, and not so well on remote servers with serial
>> consoles (the graphics output is at least garbage to the serial
>> console, and in worse cases, can wedge the console).
>> SOLUTION:
>>
>> Microsoft has defined an ACPI table to transmit this information called
>> SPCR (Serial Port Console Redirection). Many machines support this
>> table, and its contents are sufficient to allow Grub and the kernel to
>> match their configuration to the BIOS configuration. Using this
>> information, Grub can automatically avoid graphics mode when serial
>> console is configured, and we can remove the requirement for manual
>> configuration of serial parameters for the kernel in Grub's menu.lst.
>>
>> IMPLEMENTATION
>>
>> Grub has been modified to read and use the information in SPCR. A new
>> keyword $CONSOLE is added for use on the kernel$ boot line in menu.lst;
>> it will expand to properties which select either the appropriate serial
>> port and mode (for example,
>> "console=ttya,ttya-mode=19200,8,n,1,-,ttya-ignore-cd=true") or the
>> normal text or graphics console ("console=text" or "console=graphics")
>> as the Solaris release requires. (Currently Nevada uses text,
>> OpenSolaris uses graphics.) A typical use in menu.lst would look
>> like:
>>
>> kernel$ /platform/i86pc/kernel/$ISADIR/unix -B $ZFS-BOOTFS,$CONSOLE
>>
>> In addition, a new "spcr" command has been defined, to display the
>> settings in the SPCR table (if it exists). This command is useful for
>> diagnostic purposes.
>>
>> This requires no change in the kernel to use; however, one file is also
>> changed in the Solaris system image: /etc/ttydefs, which currently
>> configures ttymon(1) to force the console to 9600 baud. It turns out
>> that if the baud rate setting is removed, ttymon simply uses the
>> preexisting baud rate without change. This is exactly what we'd like
>> for the x86 case, as it means that baud rate choice is done in exactly
>> one place: the BIOS. It also allows one fewer configuration step on
>> SPARC systems, where the ttya-mode variable can be set in OBP, and then
>> ttydefs will not need change. (Unfortunately, the SPARC serial drivers
>> take it upon themselves to initialize the baud rate from settings in
>> /kernel/drv/options.conf, ignoring ttya-mode, so on SPARC we haven't
>> yet made it a one-step process...but this does at least remove one step
>> on SPARC).
>>
>> SUMMARY OF CHANGE
>>
>> 1) new keyword for use on Grub menu lines '$CONSOLE'
>> 2) new Grub command 'spcr'
>> 3) removal of baud rate from console: line in /etc/ttydefs
>>
>>
>>
>>
>> 6. Resources and Schedule
>> 6.4. Steering Committee requested information
>> 6.4.1. Consolidation C-team Name:
>> ON
>> 6.5. ARC review type: FastTrack
>> 6.6. ARC Exposure: open
>>
>> _______________________________________________
>> opensolaris-arc mailing list
>> opensolaris-arc at opensolaris.org
