[LinuxBIOS] LinuxBIOS Solaris Dev

[Kaufman, Adam]

Hey guys. My company has been looking at using LinuxBIOS for some of our
products. Since a fair amount of our customers are using Solaris, we are
interested in getting involved with that effort. I've seen a few
references to porting the flashrom utility to Solaris, which would be a
good priority for us. I poked around the LB source a bit, and it seems
like it will be some work to get that running in Solaris. I wondered if
you guys have researched this effort at all, and if so, what ideas you
had for implementation.

Thanks,

Adam

[ron minnich]

I don't see a solaris port as a huge problem. flashrom just uses mmap
and solaris has that.

If you want to try a make and see what happens, I'd like to hear how it goes.

ron

[Uwe Hermann]

On Wed, Jan 31, 2007 at 02:44:05PM -0700, ron minnich wrote:
> I don't see a solaris port as a huge problem. flashrom just uses mmap
> and solaris has that.
>
> If you want to try a make and see what happens, I'd like to hear how it goes.

Just a short note: if you build flashrom without having the whole
LinuxBIOS source tree around, you'll need to copy the file
src/include/boot/linuxbios_tables.h into the flashrom source directory
and apply the attached patch to make it use the file.

Btw, should we apply this in svn anyways? flashrom cannot be used
indepently otherwise...

Uwe.

[Uwe Hermann]

On Wed, Jan 31, 2007 at 11:42:43PM +0100, Stefan Reinauer wrote:
> Uwe Hermann wrote:
> > On Wed, Jan 31, 2007 at 02:44:05PM -0700, ron minnich wrote:
> > > I don't see a solaris port as a huge problem. flashrom just uses mmap
> > > and solaris has that.
> > >
> > > If you want to try a make and see what happens, I'd like to hear how it goes.
> >
> > Just a short note: if you build flashrom without having the whole
> > LinuxBIOS source tree around, you'll need to copy the file
> > src/include/boot/linuxbios_tables.h into the flashrom source directory
> > and apply the attached patch to make it use the file.
> >
> > Btw, should we apply this in svn anyways? flashrom cannot be used
> > indepently otherwise...
>
> Yes. Should we have a "make update_includes" target or something
> similar, or mention this in the README at least?

Hm, why not just put a copy of the file in util/flashrom? Will there be
(backwards-)compatibility problems when the file changes in LinuxBIOS?
I don't think so, but maybe I'm wrong.

Patch attached.

Uwe.

[Uwe Hermann]

On Thu, Feb 01, 2007 at 02:26:45AM +0100, Uwe Hermann wrote:
> Patch attached.

Forgot the sign-off. New patch.

Uwe.

[Stefan Reinauer]

* Uwe Hermann  [070201 02:33]:
> On Thu, Feb 01, 2007 at 02:26:45AM +0100, Uwe Hermann wrote:
> > Patch attached.
>
> Forgot the sign-off. New patch.

Acked-by: Stefan Reinauer

and applied.

[Kaufman, Adam]

Well, I had to do a bit of messing around, but I got both pciutils and
flashrom to compile on my Solaris 10 box. My biggest concern was the
/dev/mem wouldn't allow the same access we get on linux...  here's what
I get:

bash-3.00# ./flashrom
Calibrating delay loop... ok
Can not mmap /dev/mem at 00000000 errno(6):No such device or address

Ideas?

Thanks,
-ak

-----Original Message-----
From: ron minnich [mailto:[e-mail]] 
Sent: Wednesday, January 31, 2007 4:44 PM
To: Kaufman, Adam
Cc: [e-mail]
Subject: Re: [LinuxBIOS] LinuxBIOS Solaris Dev

I don't see a solaris port as a huge problem. flashrom just uses mmap
and solaris has that.

If you want to try a make and see what happens, I'd like to hear how it
goes.

ron

[Peter Stuge]

On Wed, Jan 31, 2007 at 07:16:32PM -0500, Kaufman,  Adam wrote:
> Well, I had to do a bit of messing around, but I got both pciutils and
> flashrom to compile on my Solaris 10 box. My biggest concern was the
> /dev/mem wouldn't allow the same access we get on linux...  here's what
> I get:
>
> bash-3.00# ./flashrom
> Calibrating delay loop... ok
> Can not mmap /dev/mem at 00000000 errno(6):No such device or address
>
> Ideas?

Dunno about userspace IO in Solaris.

Here are some semi-useful links about device drivers at least:

Writing Device Drivers
http://docs.sun.com/app/docs/doc/816-4854?q=816-4854
(Grab the PDF, I find docs.sun.com to be terribly slow.)

Writing 64-bit Clean Device Drivers for Solaris 10 for x64 Platforms:
A 32-bit and 64-bit Compatible Guide
http://partneradvantage.sun.com/protected/solaris10/adoptionkit/x86/Porting_...

//Peter

[Stefan Reinauer]

Kaufman, Adam wrote:
> Well, I had to do a bit of messing around, but I got both pciutils and
> flashrom to compile on my Solaris 10 box. My biggest concern was the
> /dev/mem wouldn't allow the same access we get on linux...  here's what
> I get:
>
> bash-3.00# ./flashrom
> Calibrating delay loop... ok
> Can not mmap /dev/mem at 00000000 errno(6):No such device or address
>
> Ideas?

It looks /dev/mem only contains ram on solaris, not the complete address
space.

There's a workaround, try using /dev/xsvc instead. It is part of X (or
XFree?) as it seems.

http://blogs.sun.com/dmick/entry/patch_for_dmidecode_2_6

[Kaufman, Adam]

Awesome! I think that did it. I'm not at the office right now, so I don't want to reboot my test system, but it doesn't look like anything blew up. Thanks for the help! If everything is good, I'll clean it up and make a patch. Should be fairly minimal.

-ak

whodini# ./flashrom -w /mnt/adamk/tyan/2882v309.rom
Calibrating delay loop... ok
No LinuxBIOS table found.
Enabling flash write on AMD8111...OK
Pm49FL004 found at physical address: 0xfff80000
Flash part is Pm49FL004 (512 KB)
Flash image seems to be a legacy BIOS. Disabling checks.
Programming Page: 0007 at address: 0x00070000

-----Original Message-----
From: Stefan Reinauer [mailto:[e-mail]]
Sent: Wed 31/01/2007 20:41
To: Kaufman,  Adam
Cc: [e-mail]
Subject: Re: [LinuxBIOS] LinuxBIOS Solaris Dev

Kaufman, Adam wrote:
> Well, I had to do a bit of messing around, but I got both pciutils and
> flashrom to compile on my Solaris 10 box. My biggest concern was the
> /dev/mem wouldn't allow the same access we get on linux...  here's what
> I get:
>
> bash-3.00# ./flashrom
> Calibrating delay loop... ok
> Can not mmap /dev/mem at 00000000 errno(6):No such device or address
>
> Ideas?

It looks /dev/mem only contains ram on solaris, not the complete address
space.

There's a workaround, try using /dev/xsvc instead. It is part of X (or
XFree?) as it seems.

http://blogs.sun.com/dmick/entry/patch_for_dmidecode_2_6

[Kaufman, Adam]

Ok, here's what I came up with for a Solaris 10 patch. I don't know that
it's the best way, but I tried to make somewhat clean. Feel free to
rewrite/suggest anything. Note that Solaris 10 removed the ability to
compile libc static, and I don't know if we think it's worth the effort,
so I just made checks to disable it for SunOS.

Thanks,

Adam

________________________________

From: [e-mail]
[mailto:[e-mail]] On Behalf Of Kaufman, Adam
Sent: Wednesday, January 31, 2007 9:00 PM
To: [e-mail]
Subject: Re: [LinuxBIOS] LinuxBIOS Solaris Dev

Awesome! I think that did it. I'm not at the office right now, so I
don't want to reboot my test system, but it doesn't look like anything
blew up. Thanks for the help! If everything is good, I'll clean it up
and make a patch. Should be fairly minimal.

-ak

whodini# ./flashrom -w /mnt/adamk/tyan/2882v309.rom
Calibrating delay loop... ok
No LinuxBIOS table found.
Enabling flash write on AMD8111...OK
Pm49FL004 found at physical address: 0xfff80000
Flash part is Pm49FL004 (512 KB)
Flash image seems to be a legacy BIOS. Disabling checks.
Programming Page: 0007 at address: 0x00070000

-----Original Message-----
From: Stefan Reinauer [mailto:[e-mail]]
Sent: Wed 31/01/2007 20:41
To: Kaufman,  Adam
Cc: [e-mail]
Subject: Re: [LinuxBIOS] LinuxBIOS Solaris Dev

Kaufman, Adam wrote:
> Well, I had to do a bit of messing around, but I got both pciutils and
> flashrom to compile on my Solaris 10 box. My biggest concern was the
> /dev/mem wouldn't allow the same access we get on linux...  here's

what
> I get:
>
> bash-3.00# ./flashrom
> Calibrating delay loop... ok
> Can not mmap /dev/mem at 00000000 errno(6):No such device or address
>
> Ideas?

It looks /dev/mem only contains ram on solaris, not the complete address
space.

There's a workaround, try using /dev/xsvc instead. It is part of X (or
XFree?) as it seems.

http://blogs.sun.com/dmick/entry/patch_for_dmidecode_2_6

[Stefan Reinauer]

Kaufman, Adam wrote:
> Ok, here’s what I came up with for a Solaris 10 patch. I don’t know that
> it’s the best way, but I tried to make somewhat clean. Feel free to
> rewrite/suggest anything. Note that Solaris 10 removed the ability to
> compile libc static, and I don’t know if we think it’s worth the effort,
> so I just made checks to disable it for SunOS.

mem.h does not look right:
> #if defined(sunos)
> #define MEM_DEV "/dev/mem"
> #else
> #define MEM_DEV "/dev/xsvc"
> #endif

should it be

#if !defined(__sun)

maybe?

Please post a Signed-off-by:
http://linuxbios.org/Development_Guidelines#How_to_contribute

[Peter Stuge]

On Fri, Feb 02, 2007 at 08:45:53AM -0500, Kaufman,  Adam wrote:
> Ok, here's what I came up with for a Solaris 10 patch.

Could all of the defines be compressed into one place? Maybe one
mem.c per system and defines controlling the build process?
Am I being overzealous about portability?

Also, it would be nice if the error messages printed the device that
was being used.

//Peter

[Stefan Reinauer]

Peter Stuge wrote:
> On Fri, Feb 02, 2007 at 08:45:53AM -0500, Kaufman,  Adam wrote:
> > Ok, here's what I came up with for a Solaris 10 patch.
>
> Could all of the defines be compressed into one place? Maybe one
> mem.c per system and defines controlling the build process?
> Am I being overzealous about portability?

Check the attached patch. The utility is, as many parts of the linuxbios
source, a little bit overpopulated with includes.

There's no seperate mem.[ch] anymore, as the code is pretty much
identical except it is using a define instead of a hardcode now.

With 2 lines in flash.h you could make it work on BeOS as well. :-)
> Also, it would be nice if the error messages printed the device that
> was being used.

done.

If you're ok, please Ack this patch

Stefan