On Wed, Jun 09, 2021 at 09:50:39AM -0700, Tom Rix wrote:
>
> On 6/9/21 9:38 AM, Greg KH wrote:
> > On Wed, Jun 09, 2021 at 08:08:06AM -0700, Tom Rix wrote:
> > > On 6/9/21 7:53 AM, Greg KH wrote:
> > > > On Wed, Jun 09, 2021 at 07:22:03AM -0700, [email protected] wrote:
> > > > > From: Tom Rix <[email protected]>
> > > > >
> > > > > The incoming xrt patchset has a toplevel subdir xrt/
> > > > > The current fpga/ uses a single dir with filename prefixes to subdivide owners
> > > > > For consistency, there should be only one way to organize the fpga/ dir.
> > > > > Because the subdir model scales better, refactor to use it.
> > > > > The discussion wrt xrt is here:
> > > > > https://lore.kernel.org/linux-fpga/[email protected]/
> > > > >
> > > > > Follow drivers/net/ethernet/ which has control configs
> > > > > NET_VENDOR_BLA that map to drivers/net/ethernet/bla
> > > > > Since fpgas do not have many vendors, drop the 'VENDOR' and use
> > > > > FPGA_BLA.
> > > > >
> > > > > There are several new subdirs
> > > > > altera/
> > > > > dfl/
> > > > > lattice/
> > > > > xilinx/
> > > > >
> > > > > Each subdir has a Kconfig that has a new/reused
> > > > >
> > > > > if FPGA_BLA
> > > > > ... existing configs ...
> > > > > endif FPGA_BLA
> > > > >
> > > > > Which is sourced into the main fpga/Kconfig
> > > > >
> > > > > Each subdir has a Makefile whose transversal is controlled in the
> > > > > fpga/Makefile by
> > > > >
> > > > > obj-$(CONFIG_FPGA_BLA) += bla/
> > > > >
> > > > > Some cleanup to arrange thing alphabetically and make fpga/Makefile's
> > > > > whitespace look more like net/'s
> > > > >
> > > > > Changes from
> > > > > v1
> > > > > Drop renaming files
> > > > > Cleanup makefiles
> > > > You can rename the files, you just can not rename the .ko objects
> > > > without everyone knowing what you are doing and you trying to bury it in
> > > > the middle of a differently described patch.
> > > >
> > > > If you want to do that, do you? I don't really understand why you want
> > > > to move things around right now other than "we have 40 files in one
> > > > directory, ick!".
> > > I am trying to resolve the layout inconsistency between what we have and
> > > what the xrt patchset does.
> > Why does it matter? New stuff can be added to a new dir, why worry
> > about old stuff? What does it hurt?
> >
> > > The big issue is the files vs dirs.
> > >
> > > Over specified filenames is secondary, so I dropped them.
> > >
> > > 40 files in one dir is itself not a problem.
> > >
> > > having 40 files and an xrt/ is.
> > Why is that a "problem"?
> >
> > > fpga/ layout should be consistent so the Makefile and Kconfig are easier to
> > > maintain.
> > Is it somehow hard to maintain today? Seems pretty trivial to me...
>
> This change was to help move xrt along.
>
> If you are fine with xrt/, I will drop this patchset.
Who has objected to xrt/ being the only new subdirectory?
My main complaints here are:
- these patches were not tested
- you renamed kernel modules "accidentally"
- you forgot SPDX lines
- lack of description of why these files being moved was
necessary in the changelog where you moved the files
Remember, patch 0/X never shows up in changelogs...
You can do better :)
greg k-h
On 6/9/21 10:13 AM, Greg KH wrote:
> On Wed, Jun 09, 2021 at 09:50:39AM -0700, Tom Rix wrote:
>> On 6/9/21 9:38 AM, Greg KH wrote:
>>> On Wed, Jun 09, 2021 at 08:08:06AM -0700, Tom Rix wrote:
>>>> On 6/9/21 7:53 AM, Greg KH wrote:
>>>>> On Wed, Jun 09, 2021 at 07:22:03AM -0700, [email protected] wrote:
>>>>>> From: Tom Rix <[email protected]>
>>>>>>
>>>>>> The incoming xrt patchset has a toplevel subdir xrt/
>>>>>> The current fpga/ uses a single dir with filename prefixes to subdivide owners
>>>>>> For consistency, there should be only one way to organize the fpga/ dir.
>>>>>> Because the subdir model scales better, refactor to use it.
>>>>>> The discussion wrt xrt is here:
>>>>>> https://lore.kernel.org/linux-fpga/[email protected]/
>>>>>>
>>>>>> Follow drivers/net/ethernet/ which has control configs
>>>>>> NET_VENDOR_BLA that map to drivers/net/ethernet/bla
>>>>>> Since fpgas do not have many vendors, drop the 'VENDOR' and use
>>>>>> FPGA_BLA.
>>>>>>
>>>>>> There are several new subdirs
>>>>>> altera/
>>>>>> dfl/
>>>>>> lattice/
>>>>>> xilinx/
>>>>>>
>>>>>> Each subdir has a Kconfig that has a new/reused
>>>>>>
>>>>>> if FPGA_BLA
>>>>>> ... existing configs ...
>>>>>> endif FPGA_BLA
>>>>>>
>>>>>> Which is sourced into the main fpga/Kconfig
>>>>>>
>>>>>> Each subdir has a Makefile whose transversal is controlled in the
>>>>>> fpga/Makefile by
>>>>>>
>>>>>> obj-$(CONFIG_FPGA_BLA) += bla/
>>>>>>
>>>>>> Some cleanup to arrange thing alphabetically and make fpga/Makefile's
>>>>>> whitespace look more like net/'s
>>>>>>
>>>>>> Changes from
>>>>>> v1
>>>>>> Drop renaming files
>>>>>> Cleanup makefiles
>>>>> You can rename the files, you just can not rename the .ko objects
>>>>> without everyone knowing what you are doing and you trying to bury it in
>>>>> the middle of a differently described patch.
>>>>>
>>>>> If you want to do that, do you? I don't really understand why you want
>>>>> to move things around right now other than "we have 40 files in one
>>>>> directory, ick!".
>>>> I am trying to resolve the layout inconsistency between what we have and
>>>> what the xrt patchset does.
>>> Why does it matter? New stuff can be added to a new dir, why worry
>>> about old stuff? What does it hurt?
>>>
>>>> The big issue is the files vs dirs.
>>>>
>>>> Over specified filenames is secondary, so I dropped them.
>>>>
>>>> 40 files in one dir is itself not a problem.
>>>>
>>>> having 40 files and an xrt/ is.
>>> Why is that a "problem"?
>>>
>>>> fpga/ layout should be consistent so the Makefile and Kconfig are easier to
>>>> maintain.
>>> Is it somehow hard to maintain today? Seems pretty trivial to me...
>> This change was to help move xrt along.
>>
>> If you are fine with xrt/, I will drop this patchset.
> Who has objected to xrt/ being the only new subdirectory?
Maybe just me, but it has been mostly me doing the review.
all of my easy comments have been nearly resolved.
now I am looking at bigger issues like this, should subdev's move out of
fpga/ etc.
>
> My main complaints here are:
> - these patches were not tested
> - you renamed kernel modules "accidentally"
> - you forgot SPDX lines
> - lack of description of why these files being moved was
> necessary in the changelog where you moved the files
>
> Remember, patch 0/X never shows up in changelogs...
I will respin and collapse the patches to a single patch with a better
commit log.
They aren't really useful except as a full change.
Testing will be done for dfl.
Tom
>
> You can do better :)
>
> greg k-h
>
On Wed, Jun 09, 2021 at 11:52:44AM -0700, Tom Rix wrote:
> > My main complaints here are:
> > - these patches were not tested
> > - you renamed kernel modules "accidentally"
> > - you forgot SPDX lines
> > - lack of description of why these files being moved was
> > necessary in the changelog where you moved the files
> >
> > Remember, patch 0/X never shows up in changelogs...
>
> I will respin and collapse the patches to a single patch with a better
> commit log.
They should not be a single patch, I never said that at all :(
Please read what I wrote above, did I ever mention there was too many
patches in the series here?
{sigh}
greg k-h