Re: [RFC PATCH 2/2] ansible.cfg: generate an ansible.cfg file in TOPDIR

[Chuck Lever <cel@kernel.org>]

On 9/1/25 8:00 AM, Daniel Gomez wrote:
> On 28/08/2025 22.28, Chuck Lever wrote:
>> From: Chuck Lever <chuck.lever@oracle.com>
>>
>> I need an ansible.cfg that is generated by the Kconfig menu, but
>> whose pathname is selected dynamically based on where kdevops is
>> being run rather than having that path baked into the .config.
> 
> I'm trying to review the previous conversation before these patches as well as
> this series. Please, let me know if I'm understanding something wrong:
> 
> I think what we need is not to have individual configuration paths (as we
> have now) for ansible.cfg, inventory, etc. What we want is to have one sandbox
> folder, our TOPDIR, and have our kdevops sandbox directory there. All files
> generated and consumed by kdevops should then use that path.
> 
> Would this work?

Well I think that would work for my particular usage, however: If I'm
reading between the lines of the Ansible pathname help texts, it appears
that these settings can specify search paths, not a single file.

In particular, the ANSIBLE_CFG_INVENTORY_FILE string is a comma-
separated list. And, if ANSIBLE_CFG_FILE is an empty string, then there
is a default set of locations /outside of/ the CWD that Ansible searches
for a configuration.

I'm not against your suggestion of specifying a sandbox pathname at all;
that would be a good UX simplification. There are these details, though,
and I don't have any suggestions about that at the moment.


> See below:
> 
> kdevops-tree $ make defconfig-selftest-xarray O=path/to/sandbox0
> # Now path/to/sandbox0 contains the generated .config, extra_vars.yml, etc
> kdevops-tree $ make menuconfig O=path/to/sandbox0
> kdevops-tree $ make O=path/to/sandbox0
> kdevops-tree $ make bringup O=path/to/sandbox0
> ...
> kdevops-tree $ make defconfig-tmpfs O=path/to/sandbox1
> ...
> path/to/sandbox0 $ make linux
> path/to/sandbox0 $ make selftests
> path/to/sandbox0 $ make selftests-xarray
> ...
> path/to/sandbox1 $ make
> path/to/sandbox1 $ make bringup
> path/to/sandbox1 $ make linux
> path/to/sandbox1 $ make fstests
> path/to/sandbox1 $ make fstests-baseline
> 
> To clarify: instead of letting users override individual file paths, we allow
> customization of the topdir directory (with O=) where all files are created.
> If O= is not passed, then PWD is assumed. I think it would still be useful to
> support customizing individual files, but only relative to TOPDIR. That would
> keep the structure cleaner, for example, using subdirectories for Ansible files,
> kconfigs, SSH, virsh, and so on.
> 
>> Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
>> ---
>>  Makefile                     |  8 ++++++++
>>  kconfigs/Kconfig.ansible_cfg | 12 ++++++++++++
>>  2 files changed, 20 insertions(+)
>>
>> diff --git a/Makefile b/Makefile
>> index b6fdbe67fc07..b014faa5632e 100644
>> --- a/Makefile
>> +++ b/Makefile
>> @@ -27,10 +27,15 @@ KDEVOPS_NODES_ROLE_TEMPLATE_DIR :=		$(KDEVOPS_PLAYBOOKS_DIR)/roles/gen_nodes/tem
>>  export KDEVOPS_NODES_TEMPLATE :=
>>  export KDEVOPS_MRPROPER :=
>>  
>> +ifeq (y,$(CONFIG_ANSIBLE_CFG_FILE_DEFAULT))
>> +ANSIBLE_CFG_FILE := $(TOPDIR_PATH)/ansible.cfg
>> +export ANSIBLE_CONFIG := $(ANSIBLE_CFG_FILE)
>> +else
> 
> What I'm suggesting above is to always use this new option. And perhaps
> keep ANSIBLE_CFG_FILE configurable but inside TOPDIR. For example:
> ANSIBLE_CFG_FILE=ansible/ansible.cfg.
> 
>>  ifneq ($(strip $(CONFIG_ANSIBLE_CFG_FILE)),)
>>  ANSIBLE_CFG_FILE := $(shell echo $(CONFIG_ANSIBLE_CFG_FILE) | tr --delete '"')
>>  export ANSIBLE_CONFIG := $(ANSIBLE_CFG_FILE)
>>  endif
>> +endif
>>  ANSIBLE_INVENTORY_FILE := $(shell echo $(CONFIG_ANSIBLE_CFG_INVENTORY) | tr --delete '"')
>>  
>>  KDEVOPS_INSTALL_TARGETS :=
>> @@ -86,6 +91,9 @@ CFLAGS += $(INCLUDES)
>>  ANSIBLE_EXTRA_ARGS += kdevops_version='$(PROJECTRELEASE)'
>>  ANSIBLE_EXTRA_ARGS += topdir_path_has_sha256sum='$(TOPDIR_PATH_HAS_SHA256SUM)'
>>  ANSIBLE_EXTRA_ARGS += topdir_path_sha256sum='$(TOPDIR_PATH_SHA256SUM)'
>> +ifeq (y,$(CONFIG_ANSIBLE_CFG_FILE_DEFAULT))
>> +ANSIBLE_EXTRA_ARGS += ansible_cfg_file='$(ANSIBLE_CFG_FILE)'
>> +endif
>>  
>>  export KDEVOPS_HOSTS_TEMPLATE := hosts.j2
>>  
>> diff --git a/kconfigs/Kconfig.ansible_cfg b/kconfigs/Kconfig.ansible_cfg
>> index dbd189f5d19e..a44f760912cf 100644
>> --- a/kconfigs/Kconfig.ansible_cfg
>> +++ b/kconfigs/Kconfig.ansible_cfg
>> @@ -18,6 +18,16 @@ config ANSIBLE_CFG_INVENTORY_SET_BY_CLI
>>  	bool
>>  	default $(shell, scripts/check-cli-set-var.sh KDEVOPS_HOSTS)
>>  
>> +config ANSIBLE_CFG_FILE_DEFAULT
>> +	bool "Use a sensible default location for the ansible.cfg file"
>> +	default N
>> +	help
>> +	  When this option is set to Y, kdevops creates an ansible.cfg
>> +	  in a sensible default location, using the settings defined
>> +	  in this Kconfig menu.
>> +
>> +if !ANSIBLE_CFG_FILE_DEFAULT
>> +
> 
> I think we don't need this if you agree with the approach above.
> 
>>  config ANSIBLE_CFG_FILE
>>  	string "Ansible configuration file"
>>  	output yaml
>> @@ -41,6 +51,8 @@ config ANSIBLE_CFG_FILE
>>  	  For more details, refer to the Ansible documentation:
>>  	  https://docs.ansible.com/ansible/latest/reference_appendices/config.html#the-configuration-file
>>  
>> +endif # ANSIBLE_CFG_FILE_DEFAULT
>> +
>>  menu "Ansible Callback Plugin Configuration"
>>  choice
>>  	prompt "Ansible Callback Plugin"


-- 
Chuck Lever