In message m2lkcjvfmw.fsf@ohwell.denx.de you wrote:
Which problem is this patch supposed to fix? Doesn't the "nextserver" option in your DHCP server config file do exactly what you want?
I would not say it tries to "fix a problem" but rather "add an option". In my opinion next-server really serves the same purpose, but of course out of reach of U-Boot. So we gain more freedom in regard to misconfigured DHCP servers.
Hm...
+++ b/include/cmd_confdefs.h @@ -169,6 +169,7 @@ #define CONFIG_BOOTP_SEND_HOSTNAME 0x00000100 #define CONFIG_BOOTP_NTPSERVER 0x00000200 #define CONFIG_BOOTP_TIMEOFFSET 0x00000400 +#define CONFIG_BOOTP_SERVER 0x00000800 #define CONFIG_BOOTP_VENDOREX 0x80000000 @@ -178,7 +179,8 @@ #define CONFIG_BOOTP_DEFAULT (CONFIG_BOOTP_SUBNETMASK | \ CONFIG_BOOTP_GATEWAY | \ CONFIG_BOOTP_HOSTNAME | \
CONFIG_BOOTP_BOOTPATH)
CONFIG_BOOTP_BOOTPATH | \CONFIG_BOOTP_SERVER)#ifndef CONFIG_BOOTP_MASK #define CONFIG_BOOTP_MASK CONFIG_BOOTP_DEFAULT
We don't even have include/cmd_confdefs.h any more...
If I understand corrrectly, this patch changes the behaviour for all boards that currently don't define CONFIG_BOOTP_SERVER.
Is this intentional?
If I understand this correctly, because CONFIG_BOOTP_MASK is tested and I don't see anyone defining this, it will *not* touch any configuration right now. Subsequently for consistency the code in bootp.c _may_ move to also using this MASK variable but then many configs will have to be touched. Many configs potentially can simply move to CONFIG_BOOTP_DEFAULT.
Hmmm... I have to admit that I never liked this CONFIG_BOOTP_MASK stuff so I'm happy it's gone.
So in any case the patch must be cleaned up and resubmitted.
Best regards,
Wolfgang Denk