Altix I/O code re-org

Altix I/O code re-org

[Pat Gefre]

We have redone the I/O layer in the Altix code.

We are posting this code for review before submitting for
inclusion in the 2.5 tree.

The code can be seen at:
ftp://oss.sgi.com/projects/sn2/sn2-update/

The general changes are:
o added new hardware support
o ran all code thru Lindent
o code cleanup (typedefs, include files, etc.)
o simplified the directory structure (all files under arch/ia64/sn/io/
  are deleted, new files are under arch/ia64/sn/ioif/)
o code size reduced by >50%
o major reorg of the code itself
o copyright updates



The code is organized in the following manner:

diffs_of_common_c_files - this has the diffs of 'C' files that were changed
                          but not renamed.
diffs_of_common_h_files - this has the diffs of include files that were changed
                          but not renamed.
diffs_of_renamed_files  - this has the diffs of all files that were renamed.


source_code.tar.gz      - tarball of new source files.
Source_code_tree/       - directory of new source files.

The diffstats are:

=> diffstat diffs_of_common_c_files
 irq.c   |  438 ++++++++++++++++++++++++++++++----------------------------------
 setup.c |  296 ++++++++++++++++++++-----------------------
 2 files changed, 349 insertions(+), 385 deletions(-)

=> diffstat diffs_of_common_h_files
 addrs.h           |   23 
 arch.h            |    5 
 geo.h             |   31 -
 hcl.h             |   52 -
 intr.h            |    2 
 io.h              |    2 
 iograph.h         |   26 
 klconfig.h        |   34 -
 ksys/l1.h         |    4 
 module.h          |   55 -
 nodepda.h         |   60 -
 pci/pic.h         |  725 ++++++++++++++---------
 pda.h             |    3 
 router.h          |    7 
 sgi.h             |   23 
 sn2/addrs.h       |   58 +
 sn2/arch.h        |    3 
 sn2/geo.h         |    7 
 sn2/intr.h        |   26 
 sn2/shub.h        |    3 
 sn2/shubio.h      | 1669 ++++++++++++++++++++++++------------------------------
 sn_cpuid.h        |    1 
 sn_sal.h          |   59 -
 sndrv.h           |    1 
 types.h           |    2 
 xtalk/xbow.h      |  456 +-------------
 xtalk/xbow_info.h |    2 
 27 files changed, 1479 insertions(+), 1860 deletions(-)

=> diffstat diffs_of_renamed_files
 bte_error.c                                         |   53 
 iomv.c                                              |    2 
 klconflib.c                                         |  545 +----
 local.tree/arch/ia64/sn/ioif/pci/pcibr_err.c        | 1870 --------------------
 local.tree/arch/ia64/sn/ioif/topology/shubmon_dvr.c |  171 -
 pci_dma.c                                           |  375 +---
 pcibr_ate.c                                         |  339 ++-
 pcibr_intr.c                                        |  710 -------
 8 files changed, 586 insertions(+), 3479 deletions(-)

-- 

Patrick Gefre
Silicon Graphics, Inc.                     (E-Mail)  pfg@sgi.com
2750 Blue Water Rd                         (Voice)   (651) 683-3127
Eagan, MN 55121-1400                       (FAX)     (651) 683-3054

Re: Altix I/O code re-org

[Christoph Hellwig]

On Thu, Jul 22, 2004 at 10:14:31AM -0500, Pat Gefre wrote:
> The code is organized in the following manner:
> 
> diffs_of_common_c_files - this has the diffs of 'C' files that were changed
>                           but not renamed.
> diffs_of_common_h_files - this has the diffs of include files that were changed
>                           but not renamed.
> diffs_of_renamed_files  - this has the diffs of all files that were renamed.
> 
> 
> source_code.tar.gz      - tarball of new source files.
> Source_code_tree/       - directory of new source files.

I haven't looked at the code yet (and won't until I'm back from OLS), but
for sure thats not how we will merge the code.  Please split it into small,
self-contained patches that do one thing each. 

Re: Altix I/O code re-org

[Jesse Barnes]

On Thursday, July 22, 2004 11:14 am, Pat Gefre wrote:
> We have redone the I/O layer in the Altix code.
>
> We are posting this code for review before submitting for
> inclusion in the 2.5 tree.
>
> The code can be seen at:
> ftp://oss.sgi.com/projects/sn2/sn2-update/
>
> The general changes are:
> o added new hardware support
> o ran all code thru Lindent
> o code cleanup (typedefs, include files, etc.)
> o simplified the directory structure (all files under arch/ia64/sn/io/
>   are deleted, new files are under arch/ia64/sn/ioif/)
> o code size reduced by >50%
> o major reorg of the code itself
> o copyright updates

One of the most important changes this patch makes is to rip out all of the 
SGI PCI probing code.  Our PROM now probes for I/O devices and tells the 
kernel where they are (similar to the ACPI model, which we may get to 
eventually).  The result is much more readable code with less duplication.  
Please take a look and let us know if you have any feedback, since we'd like 
to get this in as soon as we release a PROM that supports probing.

Thanks,
Jesse

Re: Altix I/O code re-org

[Robin Holt]

On Thu, Jul 22, 2004 at 01:57:53PM -0400, Jesse Barnes wrote:
> On Thursday, July 22, 2004 11:14 am, Pat Gefre wrote:
> > We have redone the I/O layer in the Altix code.
> >
> > We are posting this code for review before submitting for
> > inclusion in the 2.5 tree.
> >
> > The code can be seen at:
> > ftp://oss.sgi.com/projects/sn2/sn2-update/
> >
> > The general changes are:
> > o added new hardware support
> > o ran all code thru Lindent
> > o code cleanup (typedefs, include files, etc.)
> > o simplified the directory structure (all files under arch/ia64/sn/io/
> >   are deleted, new files are under arch/ia64/sn/ioif/)
> > o code size reduced by >50%
> > o major reorg of the code itself
> > o copyright updates
> 
> One of the most important changes this patch makes is to rip out all of the 
> SGI PCI probing code.  Our PROM now probes for I/O devices and tells the 
> kernel where they are (similar to the ACPI model, which we may get to 
> eventually).  The result is much more readable code with less duplication.  
> Please take a look and let us know if you have any feedback, since we'd like 
> to get this in as soon as we release a PROM that supports probing.

I am in the process of making this into four patches that will leave me with
a bootable system.  I need to test out the bte error handling code anyway.

The first patch will handle renaming the files, second and third are the
diffs_common* and then the fourth replaces the is the big patch.  I realize
this doesn't satisfy Christoph's request, but it is a first step.

Thanks,
Robin

Re: Altix I/O code re-org

[Robin Holt]

Pat,

I have a set of patches that cleanly applies against the
http://linux.bkbits.net/linux-2.5 bitkeeper tree using quilt.  They are
not available for the rest of the world yet.  Could you move them from
~holt/ioif to the project's ftp directory on oss.sgi.com?

Thanks,
Robin

PS:  I found a few small problems with the bte code and will soon have
another patch that fixes that up.  Specifically, there were changes
made to bte_error.c and pda.h that are undone by your patch.

Re: Altix I/O code re-org

[Christoph Hellwig]

> PS:  I found a few small problems with the bte code and will soon have
> another patch that fixes that up.  Specifically, there were changes
> made to bte_error.c and pda.h that are undone by your patch.

apropos bte, could you please merge bte_error.c into bte.c - there's no
external callers of functions in bte_error.c except in bte.c

Re: Altix I/O code re-org

[Robin Holt]

On Thu, Jul 22, 2004 at 08:42:31PM +0100, Christoph Hellwig wrote:
> > PS:  I found a few small problems with the bte code and will soon have
> > another patch that fixes that up.  Specifically, there were changes
> > made to bte_error.c and pda.h that are undone by your patch.
> 
> apropos bte, could you please merge bte_error.c into bte.c - there's no
> external callers of functions in bte_error.c except in bte.c

bte.c doesn't call bte_error.  That used to get called from shub_iio_error()
or something like that.  Would it be more appropriate to merge bte_error.c
into the callers file?  I would lean towards putting it into bte.c, but
your previous add implied it should go into the callers file.

Thanks,
Robin

Re: Altix I/O code re-org

[Robin Holt]

On Thu, Jul 22, 2004 at 02:40:50PM -0500, Robin Holt wrote:
> Pat,
> 
> I have a set of patches that cleanly applies against the
> http://linux.bkbits.net/linux-2.5 bitkeeper tree using quilt.  They are
> not available for the rest of the world yet.  Could you move them from
> ~holt/ioif to the project's ftp directory on oss.sgi.com?

Pat, you can ignore moving those patches.  I cannot get anywhere close to
a kernel that builds with this.  The first problem I ran into was
a missing arch/ia64/sn/ioif/agp subdir.  I made a patch to ignore that
subdir and am now running into missing .h files.  I am going exploring
for a workarea to see what your tree looks like.

Thanks,
Robin

Re: Altix I/O code re-org

[Robin Holt]

On Thu, Jul 22, 2004 at 03:29:25PM -0500, Robin Holt wrote:
> On Thu, Jul 22, 2004 at 02:40:50PM -0500, Robin Holt wrote:
> > Pat,
> > 
> > I have a set of patches that cleanly applies against the
> > http://linux.bkbits.net/linux-2.5 bitkeeper tree using quilt.  They are
> > not available for the rest of the world yet.  Could you move them from
> > ~holt/ioif to the project's ftp directory on oss.sgi.com?
> 
> Pat, you can ignore moving those patches.  I cannot get anywhere close to
> a kernel that builds with this.  The first problem I ran into was
> a missing arch/ia64/sn/ioif/agp subdir.  I made a patch to ignore that
> subdir and am now running into missing .h files.  I am going exploring
> for a workarea to see what your tree looks like.
> 

I am out.  I need to get going for a wedding and will be gone tomorrow.

I have updated oss.sgi.com:~holt/ioif/ with the latest set of patches
I was working with.  I also fixed the permissions on those.

Thanks,
Robin