inappropriate IPv4 address offered by DHCP server (IDFGH-8574)

[sergioaiello]

Answers checklist.

• I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
• I have updated my IDF branch (master or release) to the latest version and checked that the issue is present there.
• I have searched the issue tracker for a similar issue and not found a similar issue.

General issue report

Hope this problem has not been previously flagged.

Based on AP networking configuration
IP4_ADDR(&ipInfo.ip, 192,168,200,9);
IP4_ADDR(&ipInfo.gw, 192,168,200,9);
IP4_ADDR(&ipInfo.netmask, 255,255,255,248);
the addresses 192.168.200.8 and 192.168.200.15 should be reserved
Vice versa the HDCP server of the ESP32 AP tried to assign it. The STA does not answer and does not associate.
Note that only three IP addresses has been assigned on 5 available:

• 192.168.200.12
• 13
• 14

I (743059) esp_netif_lwip: DHCP server assigned IP to a station, IP is: 192.168.200.15
D (743059) event: no handlers have been registered for event IP_EVENT:2 posted to loop 0x3ffbdc18
D (744349) esp_netif_lwip: esp_netif_get_ip_info esp_netif:0x3ffbeb44
D (744489) esp_netif_lwip: esp_netif_get_ip_info esp_netif:0x3ffbeb44
I (744489) esp_netif_lwip: DHCP server assigned IP to a station, IP is: 192.168.200.15
D (744499) event: no handlers have been registered for event IP_EVENT:2 posted to loop 0x3ffbdc18
D (745739) esp_netif_lwip: esp_netif_get_ip_info esp_netif:0x3ffbeb44
D (745929) esp_netif_lwip: esp_netif_get_ip_info esp_netif:0x3ffbeb44

I can't clarify the reason why the AP's DHCP svr does not consider the two addresses 192.168.200.10 and .11 (usually they are assigned as expected)

The wireshark trace reports, similarly to lwip's log. a DHCP offer of the 192.168.200.15 address. The STA does not answer to AP's HDCP MSGs

[xueyunfei998]

[xueyunfei998]

I can receive 10 addresses for the test here, and you can modify the address configuration here.

location:esp-idf/components/esp_netif/esp_netif_defaults.c

[sergioaiello]

@xueyunfei998 / "I can receive 10 addresses for the test here"
Are you saying the DHCP server does offer the IP address to 10 STAs?
If so, I think it is an error since only 8 IP addresses can be used in a /29 subnet. And on top of 8 IP addresses, the DHCP server can offer to STAs only 5.
Please, clarify
Thank you

[xueyunfei998]

hi @sergioaiello

I mean: according to your configuration, softap can be allocated from the address of 192.168.200.10, not to 10 stas.

[sergioaiello]

hi xueyunfei998,
ok, clear. Thank you.
Have you checked to IPv4 address range?
I experienced that AP does not stop after .14. It offers also .15. That's a bug to me. Isn't it?

[xueyunfei998]

hi @sergioaiello

Yes, the subnet mask you set has no effect on idf.

[sergioaiello]

@xueyunfei998
is a /29 mask allowed? will a patch be released?
thank you

[xueyunfei998]

0001-lwip-bugfix-softap-excedes-the-range-of-subnet.txt
hi @sergioaiello
Use this patch test.

[sergioaiello]

Hi xueyunfei998,
thank you for the patch. I see my "dhcpserver.c" is different (I'm using IDF release 4.4)
Please, see here below an excerpt with the "dhcps_poll_set" function.
How can apply your patch?
Thank you.

static void dhcps_poll_set(u32_t ip)
{
u32_t softap_ip = 0, local_ip = 0;
u32_t start_ip = 0;
u32_t end_ip = 0;

if (dhcps_poll.enable == true) {
    softap_ip = htonl(ip);
    start_ip = htonl(dhcps_poll.start_ip.addr);
    end_ip = htonl(dhcps_poll.end_ip.addr);

    /*config ip information can't contain local ip*/
    if ((start_ip <= softap_ip) && (softap_ip <= end_ip)) {
        dhcps_poll.enable = false;
    } else {
        /*config ip information must be in the same segment as the local ip*/
        softap_ip >>= 8;

        if (((start_ip >> 8 != softap_ip) || (end_ip >> 8 != softap_ip))
                || (end_ip - start_ip > DHCPS_MAX_LEASE)) {
            dhcps_poll.enable = false;
        }
    }
}

if (dhcps_poll.enable == false) {
    local_ip = softap_ip = htonl(ip);
    softap_ip &= 0xFFFFFF00;
    local_ip &= 0xFF;

    if (local_ip >= 0x80) {
        local_ip -= DHCPS_MAX_LEASE;
    } else {
        local_ip ++;
    }

    bzero(&dhcps_poll, sizeof(dhcps_poll));
    dhcps_poll.start_ip.addr = softap_ip | local_ip;
    dhcps_poll.end_ip.addr = softap_ip | (local_ip + DHCPS_MAX_LEASE - 1);
    dhcps_poll.start_ip.addr = htonl(dhcps_poll.start_ip.addr);
    dhcps_poll.end_ip.addr = htonl(dhcps_poll.end_ip.addr);
}

}

[sergioaiello]

sorry, to enter even though your question was @xueyunfei998, but I'm on hold waiting for a support to test the fix (please, see comment on Nov 19th. Thank you.

[sergioaiello]

Hi, just loaded ESP-IDF 5.0 I'll test the patch soon.
Thank you.

[sergioaiello]

Hi, the issue is still there :(
In the "C:\Users\sergi\esp\esp-idf\components\lwip\apps\dhcpserver" folder I edited the file "dhcpserver.c"/"dhcps_poll_set" as follows:
"
....
static void dhcps_poll_set(dhcps_t *dhcps, u32_t ip)
{
u32_t softap_ip = 0, local_ip = 0, temp_local_ip = 0; // u32_t softap_ip = 0, local_ip = 0;
u32_t start_ip = 0;
u32_t end_ip = 0;
dhcps_lease_t *dhcps_poll = &dhcps->dhcps_poll;
if (dhcps_poll->enable == true) {
softap_ip = htonl(ip);
start_ip = htonl(dhcps_poll->start_ip.addr);
end_ip = htonl(dhcps_poll->end_ip.addr);

    /*config ip information can't contain local ip*/
    if ((start_ip <= softap_ip) && (softap_ip <= end_ip)) {
        dhcps_poll->enable = false;
    } else {
        /*config ip information must be in the same segment as the local ip*/
        softap_ip >>= 8;

        if (((start_ip >> 8 != softap_ip) || (end_ip >> 8 != softap_ip))
                || (end_ip - start_ip > DHCPS_MAX_LEASE)) {
            dhcps_poll->enable = false;
        }
    }
}

if (dhcps_poll->enable == false) {
    local_ip = softap_ip = htonl(ip);
    softap_ip &= 0xFFFFFF00;
    temp_local_ip = local_ip &= 0xFF; // local_ip &= 0xFF;

    if (local_ip >= 0x80) {
        local_ip -= DHCPS_MAX_LEASE;
    } else {
        local_ip ++;
    }

    bzero(dhcps_poll, sizeof(*dhcps_poll));
    dhcps_poll->start_ip.addr = softap_ip | local_ip;
    dhcps_poll->end_ip.addr = softap_ip | (temp_local_ip + 0xFFFFFFFF- htonl(dhcps->dhcps_mask.addr) - 1); // dhcps_poll->end_ip.addr = softap_ip | (local_ip + DHCPS_MAX_LEASE - 1);
    dhcps_poll->start_ip.addr = htonl(dhcps_poll->start_ip.addr);
    dhcps_poll->end_ip.addr = htonl(dhcps_poll->end_ip.addr);
}

}
......
"
Is that correct?
If not, please, provide me a document that explains how-to test patches.
Thank you.

[AxelLin]

@xueyunfei998
Any update? This issue was marked as "Resolution: Done" 3 weeks ago, but I still cannot find the fix in any branch.

[xueyunfei998]

dhcpserver.txt
hi @sergioaiello

Sorry for not replying in time.

The above is a complete pacth based on version 4.4.

[sergioaiello]

Please, let me know what to test in IDF v5.0
My last test (about one month ago) was not successful (the patch did not solve the issue ... on the opposite I found an other problem that I did not investigate since not relevant with this case and limited time)

Is the file dhcpserver.txt also relevant for IDF v5.0?
Should it be tested?
Thank you

[AxelLin]

@xueyunfei998 This bug is also in v4.3: cab17c7, please backport the fix for v4.3.

[AxelLin]

@xueyunfei998 v4.3.5 still has this bug.

[xueyunfei998]

hi @sergioaiello

The dhcp address pool allocation algorithm has been revised again. Please use the master version to test it.

hi @AxelLin

dhcp server's new memory pool allocation algorithm has been backport to 4.3.please test

[AxelLin]

@xueyunfei998 @mhdong

I have a question regarding the change in espressif/esp-lwip@4f24c9b ,
the commit log tells nothing about what was fixed.

Does this commit fix any issue?
If yes, why not mention it in commit log? and what was actually fixed?
If no, what is the reason to lower the retry backoff time?