[Buildroot] [PATCHv5] system: allow/disallow root login, accept encoded passwords

[Arnout Vandecappelle <arnout@mind.be>]

On 10/04/15 23:42, Yann E. MORIN wrote:
> From: Lorenzo Catucci <lorenzo@sancho.ccd.uniroma2.it>
> 
> Currently, there is only two possibilities regarding the root account:
>   - it is enabled with no password (the default)
>   - it is enabled, using a clear-text, user-provided password
> 
> This is deemed insufficient in many cases, especially when the .config
> file has to be published (e.g. for the GPL compliance, or any other
> reason.).
> 
> Fix that in two ways:
> 
>   - add a bolean option that allows/disallows root login altogether,
>     which defaults to 'y' to keep backward compatibility;
> 
>   - accept already-encoded passwords, which we recognise as starting
>     with either of $1$, $5$ or $6$ (resp. for md5, sha256 or sha512).
> 
> Signed-off-by: Lorenzo M. Catucci <lorenzo@sancho.ccd.uniroma2.it>
> [yann.morin.1998 at free.fr:
>   - don't add a choice to select between clear-text/encoded password,
>     use a single prompt;
>   - differentiate in the password hook itself;
>   - rewrite parts of the help entry;
>   - rewrite and expand the commit log
> ]
> Signed-off-by: Yann E. MORIN <yann.morin.1998@free.fr>
> Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
> 
> ---
> Changes v4 -> v5:
>   - use makefile syntax instead of shell  (Thomas)
>   - typoes  (Thomas)
>   - fix up the commit log (it never was possible to disable root login)
> ---
>  system/Config.in | 28 +++++++++++++++++++---------
>  system/system.mk | 22 ++++++++++++++++------
>  2 files changed, 35 insertions(+), 15 deletions(-)
> 
> diff --git a/system/Config.in b/system/Config.in
> index 431524d..6ba34ba 100644
> --- a/system/Config.in
> +++ b/system/Config.in
> @@ -177,26 +177,36 @@ endif
>  
>  if BR2_ROOTFS_SKELETON_DEFAULT
>  
> +config BR2_TARGET_ENABLE_ROOT_LOGIN
> +	bool "Enable root login"
> +	default "y"

 No quotes around bool values.

 However, since the default is y while it is normally n, and since we have to do
something special for the n case, wouldn't it make more sense to call it
BR2_TARGET_DISABLE_ROOT_LOGIN?

> +	help
> +	  Enable root login password
> +
>  config BR2_TARGET_GENERIC_ROOT_PASSWD
>  	string "Root password"
>  	default ""
> +	depends on BR2_TARGET_ENABLE_ROOT_LOGIN
>  	help
> -	  Set the initial root password (in clear). It will be md5-encrypted.
> +	  Set the initial root password.
>  
>  	  If set to empty (the default), then no root password will be set,
>  	  and root will need no password to log in.
>  
> -	  WARNING! WARNING!
> -	  Although pretty strong, MD5 is now an old hash function, and
> -	  suffers from some weaknesses, which makes it susceptible to attacks.
> -	  It is showing its age, so this root password should not be trusted
> -	  to properly secure any product that can be shipped to the wide,
> -	  hostile world.
> +	  If the password starts with any of $1$, $5$ or $6$, it is considered
> +	  to be already crypt-encoded with respectively md5, sha256 or sha512.
> +	  Any other value is taken to be a clear-text value, and is crypt-encoded
> +	  as per the "Passwords encoding" scheme, above.
> +
> +	  Note: "$" signs in the hashed password must be doubled. For example,
> +	  if the hashed password is "$1$longsalt$v35DIIeMo4yUfI23yditq0", then
> +	  you must enter it as "$$1$$longsalt$$v35DIIeMo4yUfI23yditq0".

 Perhaps explain why:

This is necessary because make will interpret the $ as variable expansion.

>  
>  	  WARNING! WARNING!
> -	  The password appears in clear in the .config file, and may appear
> +	  The password appears as-is in the .config file, and may appear
>  	  in the build log! Avoid using a valuable password if either the
> -	  .config file or the build log may be distributed!
> +	  .config file or the build log may be distributed, or at the
> +	  very least use a strong cryptographic hash for your password!
>  
>  choice
>  	bool "/bin/sh"
> diff --git a/system/system.mk b/system/system.mk
> index 4a1eb4a..b500a01 100644
> --- a/system/system.mk
> +++ b/system/system.mk
> @@ -34,7 +34,7 @@ endef
>  TARGET_FINALIZE_HOOKS += SYSTEM_ISSUE
>  endif
>  
> -ifneq ($(TARGET_GENERIC_ROOT_PASSWD),)
> +ifeq ($(BR2_TARGET_ENABLE_ROOT_LOGIN),y)
>  TARGETS += host-mkpasswd
>  endif
>  
> @@ -69,12 +69,22 @@ TARGET_FINALIZE_HOOKS += SET_NETWORK
>  
>  ifeq ($(BR2_ROOTFS_SKELETON_DEFAULT),y)
>  
> -define SYSTEM_ROOT_PASSWD
> -	[ -n "$(TARGET_GENERIC_ROOT_PASSWD)" ] && \
> -		TARGET_GENERIC_ROOT_PASSWD_HASH=$$($(MKPASSWD) -m "$(TARGET_GENERIC_PASSWD_METHOD)" "$(TARGET_GENERIC_ROOT_PASSWD)"); \
> -	$(SED) "s,^root:[^:]*:,root:$$TARGET_GENERIC_ROOT_PASSWD_HASH:," $(TARGET_DIR)/etc/shadow
> +ifeq ($(BR2_TARGET_ENABLE_ROOT_LOGIN),y)
> +ifeq ($(TARGET_GENERIC_ROOT_PASSWD),)
> +SYSTEM_ROOT_PASSWORD =
> +else ifneq ($(or $(filter $$1$$%,$(TARGET_GENERIC_ROOT_PASSWD)),$(filter $$5$$%,$(TARGET_GENERIC_ROOT_PASSWD)),$(filter $$6$$%,$(TARGET_GENERIC_ROOT_PASSWD))),)

 filter allows multiple patterns, so:

else ifneq ($(filter $$1$$% $$5$$% $$6$$%,$(TARGET_GENERIC_ROOT_PASSWD)),)

> +SYSTEM_ROOT_PASSWORD = $(TARGET_GENERIC_ROOT_PASSWD)
> +else
> +SYSTEM_ROOT_PASSWORD = $(shell $(MKPASSWD) -m "$(TARGET_GENERIC_PASSWD_METHOD)" "$(TARGET_GENERIC_ROOT_PASSWD)")
> +endif
> +else # !BR2_TARGET_ENABLE_ROOT_LOGIN
> +SYSTEM_ROOT_PASSWORD = *

 Even though Peter prefers positive logic, I think in this case it is more
important to keep the logic close to the condition, i.e.:

ifeq ($(BR2_TARGET_ENABLE_ROOT_LOGIN),)
SYSTEM_ROOT_PASSWORD = *
else ifeq ($(TARGET_GENERIC_ROOT_PASSWD),)
...

 Of course, if it becomes _DISABLE_ then it will be positive logic after all :-)

 Regards,
 Arnout


> +endif
> +
> +define SYSTEM_SET_ROOT_PASSWD
> +	$(SED) 's,^root:[^:]*:,root:$(SYSTEM_ROOT_PASSWORD):,' $(TARGET_DIR)/etc/shadow
>  endef
> -TARGET_FINALIZE_HOOKS += SYSTEM_ROOT_PASSWD
> +TARGET_FINALIZE_HOOKS += SYSTEM_SET_ROOT_PASSWD
>  
>  ifeq ($(BR2_SYSTEM_BIN_SH_NONE),y)
>  define SYSTEM_BIN_SH
> 


-- 
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F