Sep 26 2017
Shouldn't we also apply something like/* Map our computed MSS into the 3-bit index. */ - mss = min(tcp_mssopt(&sc->sc_inc), max(sc->sc_peer_mss, V_tcp_minmss)); + mss = sc->sc_peer_mss; for (i = nitems(tcp_sc_msstab) - 1; tcp_sc_msstab[i] > mss && i > 0; i--) ;
since we will use the computed index to restore the sc->sc_peer_mss later.
Sep 23 2017
This seems generally sensible.
I'm not quite sure about keeping the "Non-standard RSS hash types" comment, as M_HASHTYPE_OPAQUE is basically intended to be a catch-all for any other model, and is not specifically intended to be RSS (hence not having RSS in its name). So I'd be tempted to drop that comment.
Did microsoft update the referenced link? I don't see the removal notice in it.
Sep 22 2017
Sep 21 2017
Sep 20 2017
Sep 14 2017
Sep 13 2017
Sep 12 2017
Sep 7 2017
Address Hans' comment
Sep 5 2017
Sep 1 2017
Aug 31 2017
Aug 11 2017
Aug 10 2017
Aug 1 2017
Jul 25 2017
Add comment about the network VF and hn(4) work together
Sorry for jumping out of the blue, but I don't understand what this is all about. You seem to use lagg together with VFs (I assume PCI Virtual Functions), yet there's no commit message at all explaining what's going on, I think you should add a proper commit message describing how this is supposed to work.
Jul 24 2017
Jul 21 2017
In D11003#242017, @kevin.bowling_kev009.com wrote:
@sepherosa_gmail.com does this look good to you now?
Jul 19 2017
Jul 18 2017
Consistent temp variable naming
Fix eventhandler deregister on detach path
Jul 17 2017
Jun 30 2017
Jun 21 2017
And your implementation actually suffers the same issue as Linux: if any of the listen socket is closed, the sockets on the closed listen socket's accept queue are dropped; but well, that's may be acceptable, since I don't think Linux even fixed that.
Would implementing https://github.com/DragonFlyBSD/DragonFlyBSD/commit/02ad2f0b874fb0a45eb69750219f79f5e8982272 solve this?
Jun 20 2017
Jun 16 2017
I think you took a pretty old version of the code. I fixed the hash stability issue in a later version; you probably want to check that. And your implementation actually suffers the same issue as Linux: if any of the listen socket is closed, the sockets on the closed listen socket's accept queue are dropped; but well, that's may be acceptable, since I don't think Linux even fixed that.
May 3 2017
I'd kind of prefer we fix the drivers, though I think this patch is fine for now. I have an open item to refactor even the ACPI code to not use pci_cfgreg* directly but instead punt that up to the parent of acpi0 (nexus0) to support arm64 (and to support systems with multiple domains).
May 2 2017
minor style fixes; remove unnecessary comment.
Apr 20 2017
It nullifies Hyper-V hot channel effect, thus actually hurts performance. Drop it.
Apr 7 2017
Here's an idea from gnn that I agree with:
The fundamental issue here is that the syncookie is good enough to pass validation, but there is sufficient information in the syncache entry to reject it. The patch proposed here is a workaround for deleting the syncache entry.
I think the original problem is that we deleted the syncache entry. Why not set a flag or change the state of the syncache entry to prevent future packets from being accepted for that connection, and let the syncache expiry mechanism deal with deleting the syncache entry eventually.
Apr 1 2017
Mar 30 2017
Looks pretty good to me. Some minor code style suggestion.
Mar 6 2017
Feb 27 2017
We plan to commit this near the end of this week, if no objection comes. It detects possible errors, and obviously will not make the situation worse than before.
Feb 22 2017
Feb 21 2017
Jan 27 2017
Jan 19 2017
OK, I see. Please fix the commit message when committing then. BTW, in addition to include/dev/hyperv, there are usr.sbin/hyperv and several manpages that you probably want to manage w/ MK_HYPERV.
I don't see why it is needed. If hyperv is not in kernel or the ref rdtsc is not the current timecounter, this path will never be hit. Or do you have other concerns?
Jan 18 2017
If no objection comes, we plan to commit this earlier next week.
Jan 9 2017
Jan 6 2017
Jan 5 2017
I think the change is a step in the right direction. Certainly, "ifconfig xxN down" followed by an implicit UP should not cause any change to the routing table. Does anyone know why the "down" is removing the route? That seems wrong to me.
Jan 4 2017
storvsc parts looks good to me.
Dec 30 2016
See the comment, rest looks good.
I would suggest rewording the message a bit to something like:Defer if_up() until after the interface's if_ioctl method is called. This ensures the interface is initialized by the interface driver before it can be used by the rest of the system.
Dec 29 2016
Dec 28 2016
Dec 26 2016
The cause is that the prefix route was removed by in_scrubprefix() in the PRC_IFDOWN handler and never reinstalled upon PRC_IFUP because the reinstallation is done only for ifa passed to SIOCAIFADDR. Just calling if_up(ifp) looks too heavy to me because it causes extra pr_ctlinput() calls of each protocol, not only inetdomain, in order to recover the routes. What do you think about adding reinstalltion of the prefix routes in the protocol specific PRC_IFUP handler instead?