Re: [LinuxBIOS] Fallback checking normal

[Thomas Ekstrand]

> -----Original Message-----
> From: Stefan Reinauer [*mailto:[e-mail]*]
> Sent: den 23 mars 2007 17:55
> To: Thomas Ekstrand (AL/EAB)
> Cc: [e-mail]
> Subject: Re: [LinuxBIOS] Fallback checking normal
> Thomas Ekstrand [e-mail]
>
> [070323 17:02]:
> > I'm wondering wether the fallback ever checks the normal image to see
> > if it is ok before jumping to it?
>
> LinuxBIOSv3 will do this.

Is there a date set for LinuxBIOSv3?
> > If not, why?
>
> The mechanism works in a real life check. A flag is set
> before the image is jumped to, and that flag is cleared by
> the OS loader (or by the OS, if you wish). Whenever the
> system is booted with the flag set, it uses the fallback
> image instead of the normal image.
>
> So this check does not only check for a correct checksum, but
> whether the normal image was actually able to boot the OS.
> This makes this feature not only usable for flash failures
> but also for developer failures. ;-)
> > How is LinuxBIOS support for:
> > SMBIOS
> > SMI
>
> LinuxBIOS does "support" having those (ie. it does not
> prevent them) but does not implement any system management
> functionality, as it is not needed (and even discouraged) to
> have them. What do you need them for?

Donno, just checking things out! :)
I'm doing a technical report on LinuxBIOS as part of a school project.

Maybe it can be of some use in the documenation or something?
I will say that I'm no expert so it's basically a summary of what you
can find on the internet. but still...
> > Vmware
>
> Some people on this list have been working on vmware support.
> See the web archive.

Is there a search engine for the archive?
It's kinda hard to know which month to look at :)
> Stefan

Thanks for taking your time!
/Thomas

[LinuxBIOS] Fallback checking normal

[Thomas Ekstrand]

> -----Original Message-----
> From: Stefan Reinauer [

*mailto:[e-mail]* ] > Sent: den 23 mars
2007 17:55
> To: Thomas Ekstrand (AL/EAB)
> Cc: [e-mail]
> Subject: Re: [LinuxBIOS] Fallback checking normal
> Thomas Ekstrand [e-mail]
>
> [070323 17:02]:
> > I'm wondering wether the fallback ever checks the normal image to see
> > if it is ok before jumping to it?
>
> LinuxBIOSv3 will do this.

Is there a date set for LinuxBIOSv3?
> > If not, why?
>
> The mechanism works in a real life check. A flag is set
> before the image is jumped to, and that flag is cleared by
> the OS loader (or by the OS, if you wish). Whenever the
> system is booted with the flag set, it uses the fallback
> image instead of the normal image.
>
> So this check does not only check for a correct checksum, but
> whether the normal image was actually able to boot the OS.
> This makes this feature not only usable for flash failures
> but also for developer failures. ;-)
> > How is LinuxBIOS support for:
> > SMBIOS
> > SMI
>
> LinuxBIOS does "support" having those (ie. it does not
> prevent them) but does not implement any system management
> functionality, as it is not needed (and even discouraged) to
> have them. What do you need them for?

Donno, just checking things out! :)
I'm doing a technical report on LinuxBIOS as part of a school project.

Maybe it can be of some use in the documenation or something?
I will say that I'm no expert so it's basically a summary of what you
can find on the internet. but still...
> > Vmware
>
> Some people on this list have been working on vmware support.
> See the web archive.

Is there a search engine for the archive?
It's kinda hard to know which month to look at :)
> Stefan

Thanks for taking your time!
/Thomas

[LinuxBIOS] Fallback checking normal

[Stefan Reinauer]

* Thomas Ekstrand  [070327 10:25]:
> > > I'm wondering wether the fallback ever checks the normal image to see
> > > if it is ok before jumping to it?
> >
> > LinuxBIOSv3 will do this.
>
> Is there a date set for LinuxBIOSv3?

the current version can be downloaded from

svn://linuxbios.org/repository/LinuxBIOSv3
> > > How is LinuxBIOS support for:
> > > SMBIOS
> > > SMI
> >
> > LinuxBIOS does "support" having those (ie. it does not
> > prevent them) but does not implement any system management
> > functionality, as it is not needed (and even discouraged) to
> > have them. What do you need them for?
>
> Donno, just checking things out! :)
> I'm doing a technical report on LinuxBIOS as part of a school project.

Great!
> Maybe it can be of some use in the documenation or something?
> I will say that I'm no expert so it's basically a summary of what you
> can find on the internet. but still...

I think LinuxBIOS can greatly use this. We're still not exactly
successfully conveying what LinuxBIOS is to our potential users. Any
step towards that helps.
> > > Vmware
> >
> > Some people on this list have been working on vmware support.
> > See the web archive.
>
> Is there a search engine for the archive?
> It's kinda hard to know which month to look at :)

google. Add the string "site:linuxbios.org/pipermail" to your search
string and you will only search the linuxbios mailing list archives.

Stefan

[LinuxBIOS] Fallback checking normal

[WarrenHead]

Stefan Reinauer wrote:
> Thomas Ekstrand  [070327 10:25]:
> > Maybe it can be of some use in the documenation or something?
> > I will say that I'm no expert so it's basically a summary of what you
> > can find on the internet. but still...
>
> I think LinuxBIOS can greatly use this. We're still not exactly
> successfully conveying what LinuxBIOS is to our potential users. Any
> step towards that helps.
>
> Stefan

May I be of help here by acting as a complete nono?
I have been reading the emails fly by here for the last two weeks and 
been trying to read through the entire website, but I still lack a few 
reasons for even wanting to begin with linuxbios. As a user mind you, 
because otherwise this project is ofcourse wildly interesting.
So, as an ordinary person who just owns a pc, what would I get out of 
using linuxbios instead of the proprietary one that came with my machine 
already?
I did watch the FOSDEM presentation video of Ron Minnich, so I'm not 
entirely in the blue. I understand the proper support for clusters and 
the point about your bios not contacting its manufacturers without you 
knowing.
But, I find the point (expressed as a big plus) of having a prompt 
waiting for input, just a few seconds after booting, a bit vague.

What exactly am I supposed to be able to do right away at that point?
Is my own special kernel running yet, are my filesystems mounted, am I 
able to start my own favorite movie/music player?

As a normal user, sofar, I see linuxbios simply shaving off a few 
seconds of booting time, and thats it.
In other words, I see its merit as being an open system that can grow 
into something extremely cool, but I don't see its necessity, especially 
right now.

(I do however hope that Luc Verhaegen keeps at it and brings support for 
the via mini-itx boards into play, since these are mediacenters and 
should be instant-on.
And I do hope that the intel 440BX chipsets becomes supported, since 
that might be very cool for the freesco project that uses such ancient 
hardware and could do with a flashable ROM.)

Cheers, Warren

[LinuxBIOS] Fallback checking normal

[Peter Stuge]

Hi,

On Tue, Mar 27, 2007 at 04:41:37PM +0200, WarrenHead wrote:
> May I be of help here by acting as a complete nono?
> I have been reading the emails fly by here for the last two weeks and 
> been trying to read through the entire website, but I still lack a few 
> reasons for even wanting to begin with linuxbios. As a user mind you, 
> because otherwise this project is ofcourse wildly interesting.
> So, as an ordinary person who just owns a pc, what would I get out of 
> using linuxbios instead of the proprietary one that came with my
> machine already?

The warm and fuzzy open source feeling, but not much else right now.
> I did watch the FOSDEM presentation video of Ron Minnich, so I'm not 
> entirely in the blue. I understand the proper support for clusters and 
> the point about your bios not contacting its manufacturers without you 
> knowing.
> But, I find the point (expressed as a big plus) of having a prompt 
> waiting for input, just a few seconds after booting, a bit vague.
>
> What exactly am I supposed to be able to do right away at that point?

Excellent question.

The metric of boot-to-shell is used to describe what kind of boot
times can be expected for a full system, whatever that may be.

System integrators will build their own stuff on top of the "boot
shell state" while home users may just want a distribution to start
on their system as usual.

The metric is the least common denominator, but only integrators will
be familiar with it. :(

While LinuxBIOS doesn't really care what you run on your system it's
still a problem for us because the boundary is not immediately clear
to users.
> Is my own special kernel running yet, are my filesystems mounted, am
> I able to start my own favorite movie/music player?

Yes, your own kernel is running at that time. Anything else depends
on the startup scripts that have been run at boot.

In my car PC I made some custom start scripts to fire up xmms2 and
start playing at boot.

I used the same board for a theatre lighting project, where I used a
kernel with the viafb driver, and could run mplayer right after boot.
> As a normal user, sofar, I see linuxbios simply shaving off a few
> seconds of booting time, and thats it.

I would say it's more than a few, perhaps more like 30 on an EPIA.
> In other words, I see its merit as being an open system that can
> grow into something extremely cool, but I don't see its necessity,
> especially right now.

With EFI and data retention not currently here I agree there is
little pressing necessity yet, but there will be soon enough.
> (I do however hope that Luc Verhaegen keeps at it and brings
> support for the via mini-itx boards into play, since these are
> mediacenters and should be instant-on.

Three mini-ITX boards (EPIA, -M, -MII) have been supported in
LinuxBIOS v2 for quite some time, and work pretty well. :)

(While not perfect still usable for some purposes. I've had problems
with the PCI slot.)
> And I do hope that the intel 440BX chipsets becomes supported,
> since that might be very cool for the freesco project that uses
> such ancient hardware and could do with a flashable ROM.)

There is indeed a lot of 430/440 hardware around. Once the code works
I think we can expect even more attention.

In summary I'd say that v1 and v2, while good enough for controlled
deployment, have proven to be great learning experiences for
understanding the problems of a fully general BIOS and that v3 will
be able to focus more on end users. One example is using Kconfig for
configuring the BIOS. (This is the same configuration system as used
by the Linux kernel, so you'd say make menuconfig or make xconfig to
configure your LinuxBIOS too.)

//Peter

[LinuxBIOS] Fallback checking normal

[Alex Mauer]

Peter Stuge wrote:
> Three mini-ITX boards (EPIA, -M, -MII) have been supported in
> LinuxBIOS v2 for quite some time, and work pretty well. :)

I'd like to mention that the EPIA is in fact not supported.

The -M and -MII are, but the plain EPIA is not.

-Alex Mauer "hawke"

[LinuxBIOS] Fallback checking normal

[Peter Stuge]

On Thu, Mar 29, 2007 at 10:41:40AM -0500, Alex Mauer wrote:
> Peter Stuge wrote:
> > Three mini-ITX boards (EPIA, -M, -MII) have been supported in
> > LinuxBIOS v2 for quite some time, and work pretty well. :)
>
> I'd like to mention that the EPIA is in fact not supported.
>
> The -M and -MII are, but the plain EPIA is not.

What about targets/via/epia ?

//Peter

[LinuxBIOS] Fallback checking normal

[Alex Mauer]

Peter Stuge wrote:
> On Thu, Mar 29, 2007 at 10:41:40AM -0500, Alex Mauer wrote:
> > The -M and -MII are, but the plain EPIA is not.
>
> What about targets/via/epia ?

As far as I know that is the plain Epia.

-Alex Mauer "hawke"

[LinuxBIOS] Fallback checking normal

[Carl-Daniel Hailfinger]

On 27.03.2007 16:41, WarrenHead wrote:
> So, as an ordinary person who just owns a pc, what would I get out of 
> using linuxbios instead of the proprietary one that came with my machine 
> already?

There's one instant benefit: Better realtime behaviour for Professional
Audio/Video applications. Let me explain: All recent proprietary BIOS
releases use SMM (maybe not all, but I haven't seen one without).
SMM routines are not under the control of the operating system and will
be executed at arbitrary points in time with unknown (although mostly
noticeable) duration and the operating system will FREEZE during that
time. If you depend on low latency for audio applications, even "hard
realtime" kernels will never be able to guarantee you any latency,
simply because they don't know when the BIOS decides to block the CPU
with SMM.

Regards,
Carl-Daniel

[LinuxBIOS] Fallback checking / SMM disadvantages

[Quux]

SMM is killing hard "real time capability" and also killing hard 
security, unless it's code is public.
Those TCP/IP stacks within SMM-BIOS do not make everyone happy.   --Q

Carl-Daniel Hailfinger schrieb:
> On 27.03.2007 16:41, WarrenHead wrote:
> > So, as an ordinary person who just owns a pc, what would I get out of 
> > using linuxbios instead of the proprietary one that came with my machine 
> > already?
>
> There's one instant benefit: Better realtime behaviour for Professional
> Audio/Video applications. Let me explain: All recent proprietary BIOS
> releases use SMM (maybe not all, but I haven't seen one without).
> SMM routines are not under the control of the operating system and will
> be executed at arbitrary points in time with unknown (although mostly
> noticeable) duration and the operating system will FREEZE during that
> time. If you depend on low latency for audio applications, even "hard
> realtime" kernels will never be able to guarantee you any latency,
> simply because they don't know when the BIOS decides to block the CPU
> with SMM.
>
> Regards,
> Carl-Daniel

[LinuxBIOS] linuxbios selling points - Was: Fallback checking / SMM disadvantages

[WarrenHead]

Quux schreef:
> SMM is killing hard "real time capability" and also killing hard 
> security, unless it's code is public.
> Those TCP/IP stacks within SMM-BIOS do not make everyone happy.   --Q
>
> Carl-Daniel Hailfinger schrieb:
> > On 27.03.2007 16:41, WarrenHead wrote:
> > > So, as an ordinary person who just owns a pc, what would I get out of 
> > > using linuxbios instead of the proprietary one that came with my machine 
> > > already?
> >
> > There's one instant benefit: Better realtime behaviour for Professional
> > Audio/Video applications. Let me explain: All recent proprietary BIOS
> > releases use SMM (maybe not all, but I haven't seen one without).
> > SMM routines are not under the control of the operating system and will
> > be executed at arbitrary points in time with unknown (although mostly
> > noticeable) duration and the operating system will FREEZE during that
> > time. If you depend on low latency for audio applications, even "hard
> > realtime" kernels will never be able to guarantee you any latency,
> > simply because they don't know when the BIOS decides to block the CPU
> > with SMM.
> >
> > Regards,
> > Carl-Daniel

Thanks for your answers.
I for one see the avoidance of bios code that interrupts the careful 
planning of realtime os'ses as something rather straightforward; in the 
way of a sellable benefit. I bet this could really be swept up as a 
selling point, barring you get audio folks to state they see/notice a 
difference.
Are people onto this? Creating a measurable before/after situation?

The security item is interesting, but not as much. Even if 'evil' bios 
manufacturers were daily snooping around on my machine, I can't see it, 
so how would I feel better without it? I haven't experienced anything 
bad, so I can't compare. This feels a bit like reading my horoscope to 
see whether I should even bother to get up, or taking precautionary 
extra vitamins because the label says I need them. Too much fear.
Besides, I can see people saying that this is a feature they want. They 
want some big brother to care about their machine. It makes them feel 
cared for. It's just the secrecy that is bad.

But still, what about the boot time savings? How substantial is this?
Now that hibernation is 'here', this point feels moot. Is it?

Cheers, Warren