[PATCH net-next v1 1/1] net: selftest: provide option to disable generic selftests

Some systems may need to disable selftests to reduce kernel size or for
some policy reasons. This patch provide option to disable generic selftests.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Fixes: 3e1e58d64c3d ("net: add generic selftest support")
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
---
 net/Kconfig | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/net/Kconfig b/net/Kconfig
index f5ee7c65e6b4..dac98c73fcd8 100644
--- a/net/Kconfig
+++ b/net/Kconfig

@@ -431,7 +431,12 @@ config SOCK_VALIDATE_XMIT
 
 config NET_SELFTESTS
 	def_tristate PHYLIB
+	prompt "Support for generic selftests"
 	depends on PHYLIB && INET
+	help
+	  These selftests are build automatically if any driver with generic
+	  selftests support is enabled. This option can be used to disable
+	  selftests to reduce kernel size.
 
 config NET_SOCK_MSG
 	bool

-- 
2.29.2

On 4/30/21 2:53 AM, Oleksij Rempel wrote:

	                      built

+	  selftests support is enabled. This option can be used to disable
+	  selftests to reduce kernel size.
 
 config NET_SOCK_MSG
 	bool

Thanks for the patch/option. But I think it should just default to n,
not PHYLIB.

-- 
~Randy

On Fri, Apr 30, 2021 at 08:13:12AM -0700, Randy Dunlap wrote:

On 4/30/21 2:53 AM, Oleksij Rempel wrote:

quoted

Some systems may need to disable selftests to reduce kernel size or for
some policy reasons. This patch provide option to disable generic selftests.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Fixes: 3e1e58d64c3d ("net: add generic selftest support")
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
---
 net/Kconfig | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/net/Kconfig b/net/Kconfig
index f5ee7c65e6b4..dac98c73fcd8 100644
--- a/net/Kconfig
+++ b/net/Kconfig

@@ -431,7 +431,12 @@ config SOCK_VALIDATE_XMIT
 
 config NET_SELFTESTS
 	def_tristate PHYLIB
+	prompt "Support for generic selftests"
 	depends on PHYLIB && INET
+	help
+	  These selftests are build automatically if any driver with generic

	                      built

quoted

+	  selftests support is enabled. This option can be used to disable
+	  selftests to reduce kernel size.
 
 config NET_SOCK_MSG
 	bool

Thanks for the patch/option. But I think it should just default to n,
not PHYLIB.

It should be enabled by default for every device supporting this kind of
selftests. This tests extend functionality of cable tests, which are not
optional. Disabling it by default makes even less sense, at least for
me.

It depends on PHYLIB, if PHYLIB is build as module, this
this part should be build as module too. And since Geert asking to make
it optional, I provided this patch.

Regards,
Oleksij
-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

quoted

Thanks for the patch/option. But I think it should just default to n,
not PHYLIB.

It should be enabled by default for every device supporting this kind of
selftests.

I agree.

I still wonder if there is confusion about self test here. Maybe
putting ethtool into the description will help people understand it
has nothing to do with the kernel self test infrastructure and kernel
self testing.

	Andrew

Hi Oleksij,

On Fri, Apr 30, 2021 at 5:42 PM Oleksij Rempel [off-list ref] wrote:

On Fri, Apr 30, 2021 at 08:13:12AM -0700, Randy Dunlap wrote:

quoted

On 4/30/21 2:53 AM, Oleksij Rempel wrote:

quoted

Some systems may need to disable selftests to reduce kernel size or for
some policy reasons. This patch provide option to disable generic selftests.

Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Fixes: 3e1e58d64c3d ("net: add generic selftest support")
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
---
 net/Kconfig | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/net/Kconfig b/net/Kconfig
index f5ee7c65e6b4..dac98c73fcd8 100644
--- a/net/Kconfig
+++ b/net/Kconfig

@@ -431,7 +431,12 @@ config SOCK_VALIDATE_XMIT

 config NET_SELFTESTS
    def_tristate PHYLIB
+   prompt "Support for generic selftests"
    depends on PHYLIB && INET
+   help
+     These selftests are build automatically if any driver with generic

                            built

quoted

+     selftests support is enabled. This option can be used to disable
+     selftests to reduce kernel size.

 config NET_SOCK_MSG
    bool

Thanks for the patch/option. But I think it should just default to n,
not PHYLIB.

It should be enabled by default for every device supporting this kind of
selftests. This tests extend functionality of cable tests, which are not
optional. Disabling it by default makes even less sense, at least for
me.

Drivers using it ( AG71XX, FEC, NET_DSA) already select it, right?
Or better, imply it, which assumes it's optional.

It depends on PHYLIB, if PHYLIB is build as module, this
this part should be build as module too. And since Geert asking to make
it optional, I provided this patch.

Gr{oetje,eeting}s,

                        Geert


--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

On 4/30/21 9:23 AM, Andrew Lunn wrote:

quoted

quoted

Thanks for the patch/option. But I think it should just default to n,
not PHYLIB.

It should be enabled by default for every device supporting this kind of
selftests.

I agree.

I still wonder if there is confusion about self test here. Maybe

Probably.

putting ethtool into the description will help people understand it
has nothing to do with the kernel self test infrastructure and kernel
self testing.

So it's a hardware check that is required to be run if it's implemented
in a driver?

Required by who/what?

thanks.
-- 
~Randy

On Fri, Apr 30, 2021 at 09:33:42AM -0700, Randy Dunlap wrote:

On 4/30/21 9:23 AM, Andrew Lunn wrote:

quoted

quoted

quoted

Thanks for the patch/option. But I think it should just default to n,
not PHYLIB.

It should be enabled by default for every device supporting this kind of
selftests.

I agree.

I still wonder if there is confusion about self test here. Maybe

Probably.

quoted

putting ethtool into the description will help people understand it
has nothing to do with the kernel self test infrastructure and kernel
self testing.

So it's a hardware check that is required to be run if it's implemented
in a driver?

Required by who/what?

It is not required, but it is a useful debug tool for the educated
user. Root can run the self tests on the network interface. If the
self test pass, but the interface does not work, you probably have a
cabling or configuration issue. Networking is complex and being able
to eliminate the interface hardware lets you concentrate on some other
part of the problem. You can then maybe next use the cable test option
of ethtool to see if the cable has a problem.

   Andrew