Thanks for the review, I've fixed the comments and I would like to push this ASAP, since it will allow me to build FreeBSD install images on the Xen CI loop without any external patches. Long term I think it would be better to add an "unmount" argument to zfsboot, but that can be done later.

Fix Devin's comments.

Use zpool export instead of zfs unmount -a

Thanks you very much! The handbook looks fantastic, only some minor comments.

This is not suitable because distextract requires a TTY in order to run, a new set of patches is coming...

In any case, the patch is incorrect, there's a race. Consider the following scenario:

I've asked several times before requesting the EN that you test stable/10, stable/11 and HEAD to make sure they where working fine, and I was told that everything was working fine and that all issues where fixed, yet here we are, we requested a set of ENs and it's still not working properly.

In D9832#207524, @bhavesh.davda_gmail.com wrote:

The mbuf(9) man page specifically says this:

If a particular network interface just indicates success or failure of TCP or UDP checksum validation without returning the exact value of the checksum to the host CPU, its driver can mark CSUM_DATA_VALID and CSUM_PSEUDO_HDR in csum_flags, and set csum_data to 0xFFFF hexadecimal to indicate a valid checksum. It is a peculiarity of the algorithm used that the Internet checksum calculated over any valid packet will be 0xFFFF as long as the original checksum field is included.

@royger 's proposed patch @ https://bugs.freebsd.org/bugzilla/attachment.cgi?id=180335&action=diff&headers=0#b/sys/dev/xen/netfront/netfront.c_sec1 fixes the Xen netfront driver to conform to this recommendation in the mbuf(9) man page.

However, m->m_pkthdr.csum_data == 0xFFFF only helps udp_input() that I previously cited to "validate" that the received UDP checksum is "correct":

if (uh->uh_sum) {
        u_short uh_sum;
        if ((m->m_pkthdr.csum_flags & CSUM_DATA_VALID) &&
            !cscov_partial) {
                if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
                        uh_sum = m->m_pkthdr.csum_data; // this will be 0xffff with the Xen netfront driver change
                        ...
                uh_sum ^= 0xffff;  // This will make uh_sum '0'
        }
        if (uh_sum) { // This will be false as a result of the above, so packet won't be dropped. But the UDP checksum in the packet payload won't have been modified at all
                UDPSTAT_INC(udps_badsum);
                m_freem(m);
                return (IPPROTO_DONE);
        }

So in other words, my proposed patch in this review won't work.

FWIW, this was fixed not long ago in isc-dhcp with the following commit:

In D9832#207512, @bhavesh.davda_gmail.com wrote:

Can folks please take a quick look at https://bugs.freebsd.org/bugzilla/attachment.cgi?id=180307&action=diff to see if that is a more acceptable patch?

Add fix for XSA-211

Ping?

Fix compile issues with clang 4.0. Now both xen-kernel and tools compile with clang 4.0 from HEAD.

dev.xnb.0.unit_test_results: 53 Tests Passed

Oh, right. This is the actual output, looks like 1 test failed:

This is the output I get:

Sorry to ask, but how should I test it?

LGTM

D9630 is now committed, feel free to push this.

Thanks for doing this (I was planning to do it this weekend), LGTM. Just a minor comment regarding a repeated code chunk. In any case feel free to commit with or without that fixed, it's a general improvement after all.