fix receicing of DHCP vendor info (IDFGH-10591)

[ljakob]

Test Setup

DHCP-Server sends forced vendor option

host devbrd-dhcptest    { 
 hardware ethernet cc:db:a7:1d:XX:XX; fixed-address 10.X.X.X;
 option dhcp-parameter-request-list 1,3,28,6,43; # default + vendor
 option vendor-encapsulated-options "hallo test 1234";
}

with current master

this source

 char vendor_buf[128];
 bzero(&vendor_buf[0], sizeof(vendor_buf));
 esp_netif_dhcpc_option(wifi_device, ESP_NETIF_OP_GET, ESP_NETIF_VENDOR_SPECIFIC_INFO, &vendor_buf[0], sizeof(vendor_buf)-1);
 ESP_LOGI(TAG, "vendor string:%s:", &vendor_buf[0]);

logs

vendor string:llahllahllahllah:

what is obviously wrong

Bugs in lwip_default_hooks.c

• the vendor option of the DHCP response is treated as multiple of 32-bit (what is an invalid assumption, see RFC 2132)
• the response is converted from network to host byte order (for no reason)
• the loop will always duplicate the first 4 bytes because offset is not adjusted
• not checked: the 32-bit handling will produce garbage when the last read is shorter than 4 bytes

Hints for everyone who would like to use vendor stuff in DHCP

• query for vendor response in DHCP request: CMakeLists.txt -> add_definitions(-DLWIP_HOOK_DHCP_EXTRA_REQUEST_OPTIONS=,43)

• send custom vendor with request: after WIFI_EVENT_STA_START -> esp_netif_dhcpc_option(wifi_device, ESP_NETIF_OP_SET, ESP_NETIF_VENDOR_CLASS_IDENTIFIER, vendor_string, strlen(vendor_string))

[CLAassistant]

All committers have signed the CLA.