The plan is to commit first series of commits (see my github link), then do drivers in small batches, asking for testing and resorting to committing untested after timeout, if no volunteers found. When all drivers are done, finalize with the last two commits. Looks like we are all positive, so I'm going to proceed.

Wed, Oct 9, 9:59 PM

glebius updated the diff for D21943: Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker..

Address reviewers comments.

Wed, Oct 9, 9:56 PM

glebius added inline comments to D21943: Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker..

Wed, Oct 9, 9:45 PM

glebius committed rS353357: Revert most of the multicast changes from r353292. This needs a more.

Revert most of the multicast changes from r353292. This needs a more

Wed, Oct 9, 5:03 PM

glebius committed rS353356: ip6_output() has a complex set of gotos, and some can jump out of.

ip6_output() has a complex set of gotos, and some can jump out of

Wed, Oct 9, 5:02 PM

glebius committed rS353355: Cleanup unneeded includes that crept in with r353292..

Cleanup unneeded includes that crept in with r353292.

Wed, Oct 9, 5:00 PM

glebius committed rS353350: ifnet_byindex_ref() requires network epoch..

ifnet_byindex_ref() requires network epoch.

Wed, Oct 9, 4:22 PM

glebius committed rS353349: Enter network epoch in domain callouts..

Enter network epoch in domain callouts.

Wed, Oct 9, 4:21 PM

glebius committed rS353336: Revert changes to rip6_bind() from r353292. This function is always.

Revert changes to rip6_bind() from r353292. This function is always

Wed, Oct 9, 5:52 AM

Tue, Oct 8

glebius committed rS353314: Remove epoch assertion from if_setlladdr(). Originally this function was.

Remove epoch assertion from if_setlladdr(). Originally this function was

Tue, Oct 8, 5:56 PM

glebius updated the summary of D21943: Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker..

Tue, Oct 8, 5:34 PM

glebius retitled D21943: Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker. from Correct and final KBI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker. to Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker..

Tue, Oct 8, 5:34 PM

glebius retitled D21943: Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker. from Correct and final ABI to traverse through interface address lists for drivers. Removal of not stack allocated epoch_tracker. to Correct and final KBI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker..

Tue, Oct 8, 5:33 PM

glebius created D21943: Correct and final KPI to traverse through interface address lists for drivers.Removal of not stack allocated epoch_tracker..

Tue, Oct 8, 5:31 PM

glebius committed rS353313: Quickly plug another regression from r353292. Again, multicast locking needs.

Quickly plug another regression from r353292. Again, multicast locking needs

Tue, Oct 8, 4:59 PM

glebius committed rS353312: In DIAGNOSTIC block of if_delmulti_ifma_flags() enter the network epoch..

In DIAGNOSTIC block of if_delmulti_ifma_flags() enter the network epoch.

Tue, Oct 8, 4:46 PM

Mon, Oct 7

glebius committed rS353292: Widen NET_EPOCH coverage..

Widen NET_EPOCH coverage.

Mon, Oct 7, 10:40 PM

Thu, Oct 3

glebius committed rS353030: Fix build failure from r353026. Somehow module build allowed this..

Fix build failure from r353026. Somehow module build allowed this.

Thu, Oct 3, 4:42 AM

glebius committed rS353027: Protect access to seq->xwin[] with the seq mutex..

Protect access to seq->xwin[] with the seq mutex.

Thu, Oct 3, 2:35 AM

glebius closed D21803: netgraph/ng_bridge: Replace NG_BRIDGE_MAX_LINKS with unlimited links.

Thu, Oct 3, 2:33 AM

glebius committed rS353026: - Remove the compile time limit for number of links a ng_bridge node.

- Remove the compile time limit for number of links a ng_bridge node

Thu, Oct 3, 2:33 AM

Mon, Sep 30

glebius added a comment to D19111: Summary: widen net_epoch coverage up to all packet processing.

In D19111#477072, @hselasky wrote:

Hi,
I see wrappers are missing for WLAN:
net80211/ieee80211_input.c: ifp->if_input(ifp, m);
net80211/ieee80211_hostap.c: ifp->if_input(ifp, m);
??

Mon, Sep 30, 8:53 PM

glebius accepted D21803: netgraph/ng_bridge: Replace NG_BRIDGE_MAX_LINKS with unlimited links.

Mon, Sep 30, 8:46 PM

Sun, Sep 29

glebius added inline comments to D21179: Add header definition for RFC4340, Datagram Congestion Control Protocol.

Sun, Sep 29, 5:24 AM

Sat, Sep 28

glebius added a comment to D19111: Summary: widen net_epoch coverage up to all packet processing.

@gallatin yes, I already got a branch that does this. Create it for Olivier Cochard's benchmarking of ipfw. https://github.com/glebius/FreeBSD/tree/net_epoch+IFF_KNOWSEPOCH+pfil

Sat, Sep 28, 12:04 AM

Fri, Sep 27

glebius added a comment to D19111: Summary: widen net_epoch coverage up to all packet processing.

@bz That's fine, no hurry. I'm not sure that enter/exit points are 100% matching for VNETs. For example there is nothing wrong for a packet to travel between VNETs within one epoch.

Fri, Sep 27, 7:16 PM

glebius added a comment to D19111: Summary: widen net_epoch coverage up to all packet processing.

@hselasky Good catch, thanks.

Fri, Sep 27, 7:14 PM

glebius added a reviewer for D19111: Summary: widen net_epoch coverage up to all packet processing: bz.

Fri, Sep 27, 7:12 PM

glebius updated the diff for D19111: Summary: widen net_epoch coverage up to all packet processing.

Enter the net_epoch in uether_rxflush(). While here, rewrite the whole

Fri, Sep 27, 7:12 PM

glebius added inline comments to D19111: Summary: widen net_epoch coverage up to all packet processing.

Fri, Sep 27, 6:56 PM

Thu, Sep 26

glebius added a comment to D19111: Summary: widen net_epoch coverage up to all packet processing.

I'd like to go forward with this change. It will bring a number of new epoch recursions, that don't exist today, but I want to resolve them separately, one per subsystem, otherwise the changeset will get bigger and more messy. I need your review, guys. :)

Thu, Sep 26, 10:13 PM

glebius added reviewers for D19111: Summary: widen net_epoch coverage up to all packet processing: hselasky, kristof, gallatin.

Thu, Sep 26, 10:11 PM

glebius updated the diff for D19111: Summary: widen net_epoch coverage up to all packet processing.

WIP: compiles, boots in VM
Enter epoch in DAD timer.
Revert entering epoch for setsockopt. This needs to be fine grained.
- In pfi_initialize_vnet() rely on IFNET_RLOCK() which is sx(9) for
Revert if_purgeaddrs() change. This function is called in control path.
Enter epoch for duration of ip6_output() just as we do for IPv4.
Enter epoch in route_output().
Revert this branch changes to in6_joingroup_locked(). Function can
Another ND6 timer.
Enter epoch before calling nd6_prefix_onlink() and nd6_lookup().
pfi_kif_update() now requires epoch.
Resolve rebase glitch.
Revert back defrouter_select_fib().
Fix few lock order issues.
Revert changes to ifa_maintain_loopback_route().
Merge remote-tracking branch 'FreeBSD/master' into net_epoch
Merge branch 'epoch-trace' into net_epoch
Merge branch 'epoch-trace' into net_epoch
Merge branch 'master' into net_epoch
Catch up on c6f2373c081e1431319aefa29218d59c1d01ff30.
Fix mismerge.
Merge remote-tracking branch 'FreeBSD/master' into net_epoch
Restore whitespace to what we have in head.
Merge remote-tracking branch 'FreeBSD/master' into net_epoch

Thu, Sep 26, 10:03 PM

glebius committed rS352778: Move EPOCH_TRACE to opt_global.h, so that any external modules that.

Move EPOCH_TRACE to opt_global.h, so that any external modules that

Thu, Sep 26, 9:13 PM

Wed, Sep 25

glebius committed rS352725: style(9): remove extraneous empty lines.

style(9): remove extraneous empty lines

Wed, Sep 25, 8:46 PM

glebius committed rS352707: Add debugging facility EPOCH_TRACE that checks that epochs entered are.

Add debugging facility EPOCH_TRACE that checks that epochs entered are

Wed, Sep 25, 6:26 PM

glebius committed rS352702: Enhance the 'ps' command so that it prints a line per proc and a line.

Enhance the 'ps' command so that it prints a line per proc and a line

Wed, Sep 25, 6:03 PM

Tue, Sep 24

glebius updated the diff for D21610: EPOCH tracing support for debugging..

Merge remote-tracking branch 'FreeBSD/master' into epoch-trace
More from Hans's review.

Tue, Sep 24, 9:19 PM

glebius added inline comments to D21610: EPOCH tracing support for debugging..

Tue, Sep 24, 5:31 PM

glebius added inline comments to D21610: EPOCH tracing support for debugging..

Tue, Sep 24, 12:27 AM

Sep 13 2019

glebius committed rS352311: Drivers may pass runt packets to filter. This is okay..

Drivers may pass runt packets to filter. This is okay.

Sep 13 2019, 10:36 PM

glebius updated the diff for D21610: EPOCH tracing support for debugging..

Improve per Hans's suggestions.

Sep 13 2019, 9:28 PM

glebius added inline comments to D21610: EPOCH tracing support for debugging..

Sep 13 2019, 9:26 PM

Sep 11 2019

glebius created D21610: EPOCH tracing support for debugging..

Sep 11 2019, 9:40 PM

Sep 10 2019

glebius committed rS352116: Fix 'calendar -a' in several ways..

Fix 'calendar -a' in several ways.

Sep 10 2019, 4:22 AM

glebius committed rS352115: Remove pointless playing with LC_TIME, which should have been done in r205821..

Remove pointless playing with LC_TIME, which should have been done in r205821.

Sep 10 2019, 4:01 AM

Aug 30 2019

glebius committed rS351616: Use mbuf queue instead of ifqueue in USB network drivers..

Use mbuf queue instead of ifqueue in USB network drivers.

Aug 30 2019, 12:05 AM

glebius committed rS351615: Allow mbuf queues to be unlimited..

Allow mbuf queues to be unlimited.

Aug 30 2019, 12:04 AM

Aug 16 2019

glebius added inline comments to D21277: Add kernel-side support for in-kernel TLS..

Aug 16 2019, 6:21 PM

Aug 7 2019

glebius committed rS350688: resource.h is sufficient here, no need for resoursevar.h..

resource.h is sufficient here, no need for resoursevar.h.

Aug 7 2019, 6:32 PM

Aug 6 2019

glebius committed rS350636: Since r350426 this KASSERT doesn't serve any useful purpose..

Since r350426 this KASSERT doesn't serve any useful purpose.

Aug 6 2019, 4:11 PM

Jun 25 2019

glebius added a comment to D17996: Implement non-thread dependant network address list locks.

I think that if_maddr_rlock + accessing list head on ifnet is just a plain wrong KPI. The if_maddr_count+if_maddr_array is a bit better, but requires to allocate memory, while usually being used in a context where we can't wait.

Jun 25 2019, 2:06 AM