Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':
u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'
Function setup_environment() in board/sunxi/board.c calls eth_env_set_enetaddr() to setup stable mac address for ethernet interfaces.
This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See fdt_fixup_ethernet().
This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)
cmd/nvedit.c +12 -0
@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);
include/env_internal.h +11 -0
@@ -273,6 +273,17 @@ struct env_driver {extern struct hsearch_data env_htab; +/** +* eth_parse_enetaddr() - Parse a MAC address +* +* Convert a string MAC address +* +* @addr: MAC address in aa:bb:cc:dd:ee:ff format, where each part is a 2-digit +* hex value +* @enetaddr: Place to put MAC address (6 bytes) +*/ +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr); +#endif /* DO_DEPS_ONLY */ #endif /* _ENV_INTERNAL_H_ */
include/net.h +0 -11
@@ -875,15 +875,4 @@ int update_tftp(ulong addr, char *interface, char *devstring);/**********************************************************************/ -/** -* eth_parse_enetaddr() - Parse a MAC address -* -* Convert a string MAC address -* -* @addr: MAC address in aa:bb:cc:dd:ee:ff format, where each part is a 2-digit -* hex value -* @enetaddr: Place to put MAC address (6 bytes) -*/ -void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr); -#endif /* __NET_H__ */
net/net.c +0 -12
@@ -1628,15 +1628,3 @@ ushort env_get_vlan(char *var) {return string_to_vlan (env_get (var)); } - -void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) -{ -char *end; -int i; - -for (i =0 ; i <6 ; ++i) { - enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; -if (addr) - addr = (*end) ? end +1 : end; - } -}
Signature · 1 line
2.23.0
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0
Please, don't move the definition to env_internal.h but to env.h as board/renesas/sh7753evb/sh7753evb.c and others are using eth_parse_enetaddr().
env_internal.h explicitly states "It should not be included by board files".
Please, execute Travis CI tests to ensure you do not break any other board.
Best regards
Heinrich
Quoted 59 lines
include/env_internal.h +11
-0@@ -273,6 +273,17 @@ struct env_driver {extern struct hsearch_data env_htab; +/** +* eth_parse_enetaddr() - Parse a MAC address +* +* Convert a string MAC address +* +* @addr: MAC address in aa:bb:cc:dd:ee:ff format, where each part is a 2-digit +* hex value +* @enetaddr: Place to put MAC address (6 bytes) +*/ +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr); +#endif /* DO_DEPS_ONLY */ #endif /* _ENV_INTERNAL_H_ */ include/net.h +0
-11@@ -875,15 +875,4 @@ int update_tftp(ulong addr, char *interface, char *devstring);/**********************************************************************/ -/** -* eth_parse_enetaddr() - Parse a MAC address -* -* Convert a string MAC address -* -* @addr: MAC address in aa:bb:cc:dd:ee:ff format, where each part is a 2-digit -* hex value -* @enetaddr: Place to put MAC address (6 bytes) -*/ -void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr); -#endif /* __NET_H__ */ net/net.c +0
-12@@ -1628,15 +1628,3 @@ ushort env_get_vlan(char *var) {return string_to_vlan (env_get (var)); } - -void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) -{ -char *end; -int i; - -for (i =0 ; i <6 ; ++i) { - enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; -if (addr) - addr = (*end) ? end +1 : end; - } -}
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Quoted 60 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
Quoted 3 lines
env_internal.hexplicitly states "It should not be included by board files".Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.
Quoted 64 lines
Best regards
Heinrich
Quoted 59 lines
include/env_internal.h +11
-0@@ -273,6 +273,17 @@ struct env_driver {extern struct hsearch_data env_htab; +/** +* eth_parse_enetaddr() - Parse a MAC address +* +* Convert a string MAC address +* +* @addr: MAC address in aa:bb:cc:dd:ee:ff format, where each part is a 2-digit +* hex value +* @enetaddr: Place to put MAC address (6 bytes) +*/ +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr); +#endif /* DO_DEPS_ONLY */ #endif /* _ENV_INTERNAL_H_ */ include/net.h +0
-11@@ -875,15 +875,4 @@ int update_tftp(ulong addr, char *interface, char *devstring);/**********************************************************************/ -/** -* eth_parse_enetaddr() - Parse a MAC address -* -* Convert a string MAC address -* -* @addr: MAC address in aa:bb:cc:dd:ee:ff format, where each part is a 2-digit -* hex value -* @enetaddr: Place to put MAC address (6 bytes) -*/ -void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr); -#endif /* __NET_H__ */ net/net.c +0
-12@@ -1628,15 +1628,3 @@ ushort env_get_vlan(char *var) {return string_to_vlan (env_get (var)); } - -void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) -{ -char *end; -int i; - -for (i =0 ; i <6 ; ++i) { - enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; -if (addr) - addr = (*end) ? end +1 : end; - } -}
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Quoted 63 lines · OndÅ™ej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".
Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.
Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).
The file telling Travis what to do is in .travis.yml in U-Boot.
Best regards
Heinrich
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Hi,
Quoted 73 lines · Heinrich Schuchardt
Quoted 63 lines · OndÅ™ej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.
Quoted 8 lines
Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).The file telling Travis what to do is in .travis.yml in U-Boot.
I think it would be better to do a patch to move this function into a common/ file, and add a new config to test for this case.
I understand the desire for a revert, but no build was broken due to the original patch, and the expected behaviour was not obvious.
Regards,
Simon
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Quoted 82 lines · Simon Glass
Hi,
Quoted 79 lines · Heinrich Schuchardt
Quoted 63 lines · OndÅ™ej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).
Quoted 5 lines
Quoted 2 lines
The file telling Travis what to do is in .travis.yml in U-Boot.
I think it would be better to do a patch to move this function into a common/ file, and add a new config to test for this case.
Do we really need a CONFIG? Doesn't the linker take care of eliminating unused global functions?
Regards
Heinrich
Quoted 7 lines
I understand the desire for a revert, but no build was broken due to the original patch, and the expected behaviour was not obvious.
Regards,
Simon
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
In reply to Simon GlassQuoted 82 lines · Simon Glass
Hi,
Quoted 79 lines · Heinrich Schuchardt
Quoted 63 lines · OndÅ™ej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).
Quoted 8 lines
Quoted 2 lines
The file telling Travis what to do is in .travis.yml in U-Boot.
I think it would be better to do a patch to move this function into a common/ file, and add a new config to test for this case.
I understand the desire for a revert, but no build was broken due to the original patch, and the expected behaviour was not obvious.
My build was broken, where previously it worked with the same config on v2019.07. Otherwise I wouldn't complain. Or are u-boot builds not meant to be user configurable, at least at the high level, like disabling usb/net, or other big subystems that may be unnecessary and take a lot of space and boot time?
I'm more used to the Linux kernel, where it's sort of expected that random configurations at least build, if not boot.
I don't mind having local patches for my specific configurations, if expectations for straying from defconfig are different for u-boot.
regards,
o.
Quoted 2 lines
Regards,
Simon
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Hi Ondřej,
Quoted 90 lines · Ondřej Jirman
Quoted 88 lines · Simon Glass
Hi,
Quoted 73 lines · Heinrich Schuchardt
Quoted 63 lines · Ondřej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.Quoted 8 lines
Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).The file telling Travis what to do is in .travis.yml in U-Boot.
I think it would be better to do a patch to move this function into a common/ file, and add a new config to test for this case.
Quoted 8 lines
Quoted 2 lines
I understand the desire for a revert, but no build was broken due to the original patch, and the expected behaviour was not obvious.
My build was broken, where previously it worked with the same config on v2019.07. Otherwise I wouldn't complain. Or are u-boot builds not meant to be user configurable, at least at the high level, like disabling usb/net, or other big subystems that may be unnecessary and take a lot of space and boot time?
It sounds like your board / build config is not in the mainline tree, so there is no way Simon could have known it would break you, and it didn't break the existing boards, hence his comment. I strongly encourage you to send a series adding your config so that it has an opportunity to be build tested.
Quoted 2 lines
I'm more used to the Linux kernel, where it's sort of expected that random configurations at least build, if not boot.
No, that is the expectation here too, and generally we accomplish that.
Quoted 12 lines
I don't mind having local patches for my specific configurations, if expectations for straying from defconfig are different for u-boot.
regards,
o.Quoted 2 lines
Regards,
Simon_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Hi Joe,
Quoted 102 lines · Joe Hershberger
Hi Ondřej,
Quoted 90 lines · Ondřej Jirman
Quoted 88 lines · Simon Glass
Hi,
Quoted 73 lines · Heinrich Schuchardt
Quoted 63 lines · Ondřej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.Quoted 8 lines
Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).The file telling Travis what to do is in .travis.yml in U-Boot.
I think it would be better to do a patch to move this function into a common/ file, and add a new config to test for this case.
Quoted 8 lines
Quoted 2 lines
I understand the desire for a revert, but no build was broken due to the original patch, and the expected behaviour was not obvious.
My build was broken, where previously it worked with the same config on v2019.07. Otherwise I wouldn't complain. Or are u-boot builds not meant to be user configurable, at least at the high level, like disabling usb/net, or other big subystems that may be unnecessary and take a lot of space and boot time?
Quoted 5 lines
It sounds like your board / build config is not in the mainline tree, so there is no way Simon could have known it would break you, and it didn't break the existing boards, hence his comment. I strongly encourage you to send a series adding your config so that it has an opportunity to be build tested.
I'm using orangepi_pc_defconfig. It's mainline.
I just disable a few things, like USB and NET. That's enough for it to break the build.
I don't think my minimalistic config would be proper as a defconfig for that particular board. Anyway, the kernel has feature that generates random configs for revealing these kinds of issues.
regards,
o.
Quoted 17 lines
Quoted 2 lines
I'm more used to the Linux kernel, where it's sort of expected that random configurations at least build, if not boot.
No, that is the expectation here too, and generally we accomplish that.
Quoted 12 lines
I don't mind having local patches for my specific configurations, if expectations for straying from defconfig are different for u-boot.
regards,
o.Quoted 2 lines
Regards,
Simon_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Quoted 112 lines · Ondřej Jirman
Hi Joe,
Quoted 102 lines · Joe Hershberger
Hi Ondřej,
Quoted 90 lines · Ondřej Jirman
Quoted 88 lines · Simon Glass
Hi,
Quoted 73 lines · Heinrich Schuchardt
Quoted 63 lines · Ondřej Jirman
Quoted 62 lines · Heinrich Schuchardt
include/env_internal.h
Quoted 54 lines · megous at megous.com
From: Ondrej Jirman <[email protected]>
The reverted patch causes linking error with disabled
CONFIG_NET:
cmd/built-in.o: In function `eth_env_get_enetaddr':u-boot-v2019.10/cmd/nvedit.c:363: undefined reference to `eth_parse_enetaddr'Function
setup_environment()inboard/sunxi/board.ccallseth_env_set_enetaddr()to setup stable mac address for ethernet interfaces.This needs to be implemented and succeed even if net is disabled in u-boot, as it ensures Linux will not generate random MAC addresses, and picks the ones provided by u-boot via DT. See
fdt_fixup_ethernet().This feature is independent of the whole network stack and network drivers in u-boot.
This revert fixes the linking error.
Signed-off-by Ondrej Jirman <[email protected]>--- cmd/nvedit.c | 12 ++++++++++++ include/env_internal.h | 11 +++++++++++ include/net.h | 11 ----------- net/net.c | 12 ------------ 4 files changed, 23 insertions(+), 23 deletions(-)cmd/nvedit.c +12
-0@@ -358,6 +358,18 @@ ulong env_get_hex(const char *varname, ulong default_val)return value; } +void eth_parse_enetaddr (const char *addr,uint8_t *enetaddr) +{ +char *end; +int i; + +for (i =0 ; i <6 ; ++i) { + enetaddr[i] = addr ?simple_strtoul (addr, &end,16 ) :0 ; +if (addr) + addr = (*end) ? end +1 : end; + } +} +int eth_env_get_enetaddr (const char *name,uint8_t *enetaddr) {eth_parse_enetaddr (env_get (name), enetaddr);include/env_internal.h +0
-0Please, don't move the definition to
env_internal.hbut toenv.hasboard/renesas/sh7753evb/sh7753evb.cand others are usingeth_parse_enetaddr().
env_internal.hexplicitly states "It should not be included by board files".Quoted 7 lines
Quoted 1 line
Please, execute Travis CI tests to ensure you do not break any other board.
I haven't found any documentation in the tree or on the u-boot website on how to do that.
regards,
o.Quoted 8 lines
Create a repository with U-Boot on Github (e.g. fork
https://github.com/trini/u-boot). Logon in https://travis-ci.org/ with your Github account. Allow Travis to access the U-Boot repository. In settings select "Build pushed requests". Push your commit to Github. Now Travis should start building (takes about 4 hours).The file telling Travis what to do is in .travis.yml in U-Boot.
I think it would be better to do a patch to move this function into a common/ file, and add a new config to test for this case.
Quoted 8 lines
Quoted 2 lines
I understand the desire for a revert, but no build was broken due to the original patch, and the expected behaviour was not obvious.
My build was broken, where previously it worked with the same config on v2019.07. Otherwise I wouldn't complain. Or are u-boot builds not meant to be user configurable, at least at the high level, like disabling usb/net, or other big subystems that may be unnecessary and take a lot of space and boot time?
Quoted 5 lines
It sounds like your board / build config is not in the mainline tree, so there is no way Simon could have known it would break you, and it didn't break the existing boards, hence his comment. I strongly encourage you to send a series adding your config so that it has an opportunity to be build tested.
Quoted 4 lines
I'm using
orangepi_pc_defconfig. It's mainline.I just disable a few things, like USB and NET. That's enough for it to break the build.
Clearly the point is that the actual problematic config is not mainline.
Quoted 2 lines
I don't think my minimalistic config would be proper as a defconfig for that particular board.
I was not suggesting to replace it, simply to add a minimal one. There are plenty of examples of boards with several defconfigs.
Quoted 2 lines
Anyway, the kernel has feature that generates random configs for revealing these kinds of issues.
Are you suggesting that you can port this to U-Boot so we can test in a similar way?
Quoted 24 lines
regards,
o.Quoted 17 lines
Quoted 2 lines
I'm more used to the Linux kernel, where it's sort of expected that random configurations at least build, if not boot.
No, that is the expectation here too, and generally we accomplish that.
Quoted 12 lines
I don't mind having local patches for my specific configurations, if expectations for straying from defconfig are different for u-boot.
regards,
o.Quoted 2 lines
Regards,
Simon_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Quoted 12 lines · Joe Hershberger
Quoted 10 lines · OndÅ™ej Jirman
Quoted 5 lines
It sounds like your board / build config is not in the mainline tree, so there is no way Simon could have known it would break you, and it didn't break the existing boards, hence his comment. I strongly encourage you to send a series adding your config so that it has an opportunity to be build tested.
I'm using
orangepi_pc_defconfig. It's mainline.I just disable a few things, like USB and NET. That's enough for it to break the build.
Quoted 7 lines
Clearly the point is that the actual problematic config is not mainline.
Quoted 2 lines
I don't think my minimalistic config would be proper as a defconfig for that particular board.
I was not suggesting to replace it, simply to add a minimal one. There are plenty of examples of boards with several defconfigs.
Interesting, I may add one then. Not sure what sunxi maintainer will think of that, but if it has value for testing, why not. Probably just one minimal config would have caught this, so I guess it has some value.
Thanks for suggestion.
Quoted 5 lines
Quoted 2 lines
Anyway, the kernel has feature that generates random configs for revealing these kinds of issues.
Are you suggesting that you can port this to U-Boot so we can test in a similar way?
It's a Kconfig feature, you can already use it. Try make randconfig inside u-boot.
regards,
o.
Quoted 24 lines
Quoted 24 lines
regards,
o.Quoted 17 lines
Quoted 2 lines
I'm more used to the Linux kernel, where it's sort of expected that random configurations at least build, if not boot.
No, that is the expectation here too, and generally we accomplish that.
Quoted 12 lines
I don't mind having local patches for my specific configurations, if expectations for straying from defconfig are different for u-boot.
regards,
o.Quoted 2 lines
Regards,
Simon_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot_______________________________________________
U-Boot mailing list
U-Boot at lists.denx.de
https://lists.denx.de/listinfo/u-boot
PATCH 1/1 Revert "env: net: Move eth_parse_enetaddr() to net.c/h"
Hi Ondřej,
Quoted 28 lines · Ondřej Jirman
Quoted 12 lines · Joe Hershberger
Quoted 10 lines · Ondřej Jirman
Quoted 5 lines
It sounds like your board / build config is not in the mainline tree, so there is no way Simon could have known it would break you, and it didn't break the existing boards, hence his comment. I strongly encourage you to send a series adding your config so that it has an opportunity to be build tested.
I'm using
orangepi_pc_defconfig. It's mainline.I just disable a few things, like USB and NET. That's enough for it to break the build.
Quoted 7 lines
Clearly the point is that the actual problematic config is not mainline.
Quoted 2 lines
I don't think my minimalistic config would be proper as a defconfig for that particular board.
I was not suggesting to replace it, simply to add a minimal one. There are plenty of examples of boards with several defconfigs.
Interesting, I may add one then. Not sure what sunxi maintainer will think of that, but if it has value for testing, why not. Probably just one minimal config would have caught this, so I guess it has some value.
Thanks for suggestion.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Anyway, the kernel has feature that generates random configs for revealing these kinds of issues.
Are you suggesting that you can port this to U-Boot so we can test in a similar way?
It's a
Kconfigfeature, you can already use it. Try make randconfig inside u-boot.
Another suggestion that might be better: add a new sandbox_nonet build. I suspect this would throw up quite a few issues. Also since it is more generic the build coverage would likely be better.
Regards,
Simon