Also, any meaningful change to a manual page requires bumping the .Dd value at the beginning of the page to today's date.

I agree that scripts shouldn't use -h.

In D48205#1101785, @markj wrote:

Isn't it used by svc_vc.c and clnt_vc.c? Any consumers of those interfaces would use this, it looks like, but there is no use of XDR_GET/SETPOS there that is affected by the bug, I think.

• use ptrdiff_t
• put back break

In D48054#1101622, @krzysztof.galazka_intel.com wrote:

LGTM but let me ask @bartosz.sobczak_intel.com to take a look at irdma changes.

In D48254#1100422, @markj wrote:

Are you sure it was exactly the same? I fixed another problem recently which has similar symptoms: b64c5a0ace59.

Don't have expertise to review, but thanks a lot for working on that. Was following your discussion with Dmitry. I hit this problem many many times, when running a different mix of modules on FreeBSD CURRENT and never bothered to understand it. Always resorted to recompile a static kernel.

btw, I got this script. Named wrap74, it would wrap lines to 74 chars. And you can have several symlinks/hardlinks for it, with different names, e.g. wrap79, wrap64.

In the sack_rxmit block there is another cwin calculation that uses tcp_compute_pipe() without capping with off. Later this cwin is used to calculate len. However, I have never caught such condition, so this revision leaves the above tcp_compute_pipe() use untouched.

Oh, pardon my Chirstmas hangover :)

Sorry for slightly diverting the topic. Once regular use of 240/4 is enabled by default, does it make any sense to maintain the IPv4 multicast support at all?

On second thought this probably needs to be mentioned in Relnotes and not MFC-ed. Cause it might be that somebody was relying on the stack to drop those packets before firewall.

Is this for stable/13 only? If you care about closing the bug, just go forward and commit it. The driver was removed in recent versions of FreeBSD.

uint8_t code

The change is a matter of taste, of course. So I decided to add all contributors to the script to create a kind of a vote. Anybody else using git-arc are also very welcome to join the review and vote their (dis)agreement.

Later I'd probably bring the input and output to same KBI, so that both return same signedness error for unlocked return. But this requires a sweeping change over stacks, so later.

Negate result of ctf_do_queued_segments() and compare error to >= 0.

Thanks! Please apply same changes to netinet6/icmp6.c

In D47935#1094165, @markj wrote:

I see, thanks. One other question: does the system in question have more than one NUMA domain? If so, we will use a large import quantum, KVA_NUMA_IMPORT_QUANTUM, and we need to be able to allocate more than two PTPs in order to grow the map by that much. That is, when vm_ndomains > 1, we should set vmd_interrupt_min to a larger value.

In D47935#1093253, @markj wrote:

Also, in your stack, why are we using kmem_malloc() to allocate a slab for inpcbs? That should be going through uma_small_alloc(), which doesn't need KVA. Was the kernel compiled with KASAN enabled?

In D47935#1093091, @markj wrote:

What's the motivation for this change?

Do you have any idea on the inp->inp_socket->so_type != SOCK_STREAM check?