On 04/05/2020 19:57, Clay McClure wrote:

My recent commit b6d49cab44b5 ("net: Make PTP-specific drivers depend on PTP_1588_CLOCK")
exposes a missing dependency in defconfigs that select TI_CPTS without
selecting PTP_1588_CLOCK, leading to linker errors of the form:

drivers/net/ethernet/ti/cpsw.o: in function `cpsw_ndo_stop':
cpsw.c:(.text+0x680): undefined reference to `cpts_unregister'
drivers/net/ethernet/ti/cpsw.o: in function `cpsw_remove':
cpsw.c:(.text+0x81c): undefined reference to `cpts_release'
drivers/net/ethernet/ti/cpsw.o: in function `cpsw_rx_handler':
cpsw.c:(.text+0x1324): undefined reference to `cpts_rx_timestamp'
drivers/net/ethernet/ti/cpsw.o: in function `cpsw_ndo_open':
cpsw.c:(.text+0x15ec): undefined reference to `cpts_register'
drivers/net/ethernet/ti/cpsw.o: in function `cpsw_probe':
cpsw.c:(.text+0x2468): undefined reference to `cpts_release'

That's because TI_CPTS_MOD (which is the symbol gating the _compilation_
of cpts.c) now depends on PTP_1588_CLOCK, and so is not enabled in these
configurations, but TI_CPTS (which is the symbol gating _calls_ to the
cpts functions) _is_ enabled. So we end up compiling calls to functions
that don't exist, resulting in the linker errors.

The reason we have two symbols (TI_CPTS and TI_CPTS_MOD) for the same
driver is due to commit be9ca0d33c85 ("cpsw/netcp: work around reverse
cpts dependency"), which introduced TI_CPTS_MOD because (quoting the
commit message):

quoted

The dependency is reversed: cpsw and netcp call into cpts,
but cpts depends on the other two in Kconfig. This can lead
to cpts being a loadable module and its callers built-in:

drivers/net/ethernet/ti/cpsw.o: In function `cpsw_remove':
cpsw.c:(.text.cpsw_remove+0xd0): undefined reference to `cpts_release'
drivers/net/ethernet/ti/cpsw.o: In function `cpsw_rx_handler':
cpsw.c:(.text.cpsw_rx_handler+0x2dc): undefined reference to `cpts_rx_timestamp'
drivers/net/ethernet/ti/cpsw.o: In function `cpsw_tx_handler':
cpsw.c:(.text.cpsw_tx_handler+0x7c): undefined reference to `cpts_tx_timestamp'
drivers/net/ethernet/ti/cpsw.o: In function `cpsw_ndo_stop':

Both forms of linker error -- those caused by defconfigs that select
TI_CPTS without PTP_1588_CLOCK and those caused by configuring TI_CPSW
as a built-in and TI_CPTS as a module -- can be avoided by using the
IS_REACHABLE() macro to gate calls to cpts functions, and using the
TI_CPTS symbol to gate compilation of cpts.c. cpts.h already provides
the no-op stub implementations of the cpts functions required to make
this work, we just need to change the existing IS_ENABLED(TI_CPTS)
guards to IS_REACHABLE(TI_CPTS).

With this change there is no longer any need for the TI_CPTS_MOD symbol,
so we can remove it.

To preserve the existing behavior of defconfigs that select TI_CPTS, we
must also select PTP_1588_CLOCK so that the dependency is satisfied.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Fixes: b6d49cab44b5 ("net: Make PTP-specific drivers depend on PTP_1588_CLOCK")

It's better if you send v2 not as reply to v1.

just to clarify. After these two patches
  - the PTP_1588_CLOCK can still be set to "M"
  - which will cause TI_CPTS to be "M",
  - but TI_CPSW will still be "Y".

and all above will build and produce built-in CPSW without CPTS support
and cpts.ko which is loadable, but not functional.

Sorry, I'm a little bit lost regarding the target you'are trying to achieve.
At least previously "imply PTP_1588_CLOCK" allowed to select properly PTP_1588_CLOCK
without modifying every defconfig.

+	default y if TI_CPSW=y || TI_KEYSTONE_NETCP=y || TI_CPSW_SWITCHDEV=y

even with above statement it's possible to force TI_CPTS="M" while CPSW/NETCP="Y"

+	default m

I could be mistaken by above 2 lines seems can be 'imply TI_CPTS'
in TI_CPSW, TI_KEYSTONE_NETCP, TI_CPSW_SWITCHDEV

  	---help---
  	  This driver supports the Common Platform Time Sync unit of
  	  the CPSW Ethernet Switch and Keystone 2 1g/10g Switch Subsystem.
  	  The unit can time stamp PTP UDP/IPv4 and Layer 2 packets, and the
  	  driver offers a PTP Hardware Clock.
  
-config TI_CPTS_MOD
-	tristate
-	depends on TI_CPTS
-	depends on PTP_1588_CLOCK
-	default y if TI_CPSW=y || TI_KEYSTONE_NETCP=y || TI_CPSW_SWITCHDEV=y
-	default m

and this prevented user from forcing TI_CPTS="M" while CPSW/NETCP="Y"

Below small diff should fix build fail:

diff --git a/drivers/net/ethernet/ti/Kconfig b/drivers/net/ethernet/ti/Kconfig
index 8e348780efb6..eeaee47598aa 100644
--- a/drivers/net/ethernet/ti/Kconfig
+++ b/drivers/net/ethernet/ti/Kconfig

@@ -81,6 +81,7 @@ config TI_CPTS
         depends on TI_CPSW || TI_KEYSTONE_NETCP || TI_CPSW_SWITCHDEV || COMPILE_TEST
         depends on COMMON_CLK
         depends on POSIX_TIMERS
+       depends on PTP_1588_CLOCK
         ---help---
           This driver supports the Common Platform Time Sync unit of
           the CPSW Ethernet Switch and Keystone 2 1g/10g Switch Subsystem.

@@ -90,7 +91,6 @@ config TI_CPTS
  config TI_CPTS_MOD
         tristate
         depends on TI_CPTS
-       depends on PTP_1588_CLOCK
         default y if TI_CPSW=y || TI_KEYSTONE_NETCP=y || TI_CPSW_SWITCHDEV=y
         default m

Then separate patch can be used to enable PTP_1588_CLOCK in defconfigs.

My personal opinion - it might be better to revert TI CPTS part from
b6d49cab44b5 ("net: Make PTP-specific drivers depend on PTP_1588_CLOCK")
at all.

-- 
Best regards,
grygorii

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel