Page MenuHomeFreeBSD

kristof (Kristof Provost)
User

Projects

User Details

User Since
Sep 28 2014, 7:22 PM (229 w, 3 d)

Recent Activity

Yesterday

kristof added a comment to D19248: tun: VIMAGE fix for if_tun cloner.

Thanks. The original reporter of the bug discovered that too.

Wed, Feb 20, 12:54 PM

Tue, Feb 19

kristof added reviewers for D19248: tun: VIMAGE fix for if_tun cloner: network, bz.
Tue, Feb 19, 6:28 PM
kristof created D19248: tun: VIMAGE fix for if_tun cloner.
Tue, Feb 19, 6:27 PM

Mon, Feb 11

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

This one boots, but panics when I kldload pfsync:

Mon, Feb 11, 9:44 PM

Sun, Feb 10

kristof accepted D19131: Decrease the time the kernel takes to install a new PF config with a large number of queues.
Sun, Feb 10, 1:46 PM

Sat, Feb 9

kristof added a comment to D19124: Fix HFSC configuration bug introduced in r343287.

I have no objections to the patch, but I don't know enough about HFSC to meaningfully review this, I'm afraid.

Sat, Feb 9, 1:58 PM

Fri, Feb 8

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

I'm pretty sure that ipv6_activate_all_interfaces="YES" in /etc/rc.conf is the trigger for the previous panic.
That also matches nicely with the backtrace. Presumably nd6_dad_timer() doesn't enter the NET_EPOCH.

Fri, Feb 8, 3:15 PM
kristof added a comment to D19111: Summary: widen net_epoch coverage up to all packet processing.

This panics my test vm:

Fri, Feb 8, 10:43 AM

Sat, Feb 2

kristof accepted D18924: bridge: Fix spurious warnings about capabilities.

LGTM

Sat, Feb 2, 10:10 AM

Tue, Jan 29

kristof added a comment to D18951: New pfil(9).

Did you happen to do any benchmarking on this? I'd have expected "Sync pfil hooks epoch(9)" to give us a (small) performance improvement, but my initial test shows a small reduction in forwarding performance (with pf loaded).

Tue, Jan 29, 8:34 PM

Mon, Jan 28

kristof accepted D19005: ifconfig: fix endianness bug displaying pfsync interfaces.
Mon, Jan 28, 5:44 PM
kristof added inline comments to D19005: ifconfig: fix endianness bug displaying pfsync interfaces.
Mon, Jan 28, 5:10 PM

Fri, Jan 25

kristof added a comment to D18951: New pfil(9).

This panics during the netipsec and pf regression tests:

Fri, Jan 25, 1:57 AM

Thu, Jan 24

kristof accepted D18919: In ifconfig(8), don't build, sort and search all system addresses when performing a non-status action on a single interface.

To be clear: I don't object to your change. It's just that ifconfig is a bit of a mess and any opportunity for cleanup is tempting. Especially if I don't have to do it.

Thu, Jan 24, 9:04 AM
kristof added inline comments to D18919: In ifconfig(8), don't build, sort and search all system addresses when performing a non-status action on a single interface.
Thu, Jan 24, 3:45 AM
kristof added inline comments to D18919: In ifconfig(8), don't build, sort and search all system addresses when performing a non-status action on a single interface.
Thu, Jan 24, 3:31 AM
kristof added inline comments to D18919: In ifconfig(8), don't build, sort and search all system addresses when performing a non-status action on a single interface.
Thu, Jan 24, 3:13 AM
kristof accepted D18918: Don't re-evaluate ALTQ kernel configuration due to events on non-ALTQ interfaces.
Thu, Jan 24, 2:58 AM

Jan 21 2019

kristof added a reviewer for D18909: pfctl: Point users to net.pf.request_maxcount if large requests are rejected: network.
Jan 21 2019, 3:50 AM
kristof created D18909: pfctl: Point users to net.pf.request_maxcount if large requests are rejected.
Jan 21 2019, 3:50 AM

Jan 18 2019

kristof added a comment to D18882: Fix pfsync breaking carp.

Yeah, looking at my original commit I did manage to lose that flag. I’ve not tested yet, but this patch is almost certainly correct.

Jan 18 2019, 12:41 AM

Jan 15 2019

kristof added a comment to D18759: Improve pf.conf parsing speed for large numbers of queues.

Also, I’d really, really like some basic tests for altq. It should be reasonably straightforward to do something based on the existing pf tests. At least, if altq can work on top of epairs.

Jan 15 2019, 12:12 AM

Jan 12 2019

kristof accepted D18759: Improve pf.conf parsing speed for large numbers of queues.
Jan 12 2019, 5:55 AM

Jan 8 2019

kristof accepted D18779: Don't set if_linkmib for vlan(4)..

LGTM.

Jan 8 2019, 8:14 AM

Jan 7 2019

kristof added a comment to D18759: Improve pf.conf parsing speed for large numbers of queues.

I’ll try to review asap, but I’m on holiday wirh limited internet access right now.

Jan 7 2019, 9:05 PM

Dec 29 2018

kristof created D18679: libxo: Fix XML output if a container name is a number.
Dec 29 2018, 12:52 PM

Dec 8 2018

kristof created D18483: pf: Fix endless loop on NAT exhaustion with sticky-address.
Dec 8 2018, 3:07 PM

Dec 5 2018

kristof added a comment to D18373: pfsync: Performance improvement.
In D18373#392613, @eri wrote:

Not a blocker but:

  • It would also be nice to have measure if the other side can keep up with the blast of state updates now?
  • Even better, provide the same bucket mechanism on reception so it can be distributed on the various cores
Dec 5 2018, 7:35 PM
kristof added a comment to D18373: pfsync: Performance improvement.
In D18373#392610, @eri wrote:

Can you please measure the latency of syncing states with this change against previous latency?

Dec 5 2018, 7:25 PM

Dec 3 2018

kristof updated the diff for D18373: pfsync: Performance improvement.

Fix build with ! VIMAGE

Dec 3 2018, 11:04 AM
kristof updated the diff for D18373: pfsync: Performance improvement.

Address man page remark.

Dec 3 2018, 8:38 AM

Dec 2 2018

kristof updated the diff for D18373: pfsync: Performance improvement.

Minor improvements, style and such.
Small man page addition.

Dec 2 2018, 5:20 PM
kristof abandoned D17994: pfsync: Insert static trace points.

Abandoned in favour of D18373.

Dec 2 2018, 4:53 PM
kristof abandoned D17992: pfsync: Reduce contention on PFSYNC_LOCK().

Abandoned in favour of D18373.

Dec 2 2018, 4:49 PM
kristof abandoned D17993: pfsync: Call pfsyncintr() directly from pfsync_msg_intr() rather than scheduling a swi.

Abandoned in favour of D18373.

Dec 2 2018, 4:49 PM

Dec 1 2018

kristof added inline comments to D17376: IPv6 Fragmentation Regression Tests from OpenBSD.
Dec 1 2018, 5:24 PM

Nov 29 2018

kristof updated the diff for D18373: pfsync: Performance improvement.

Single pfsyncintr swi, rather than one per bucket.

Nov 29 2018, 9:16 PM
kristof added a comment to D18373: pfsync: Performance improvement.
In D18373#390595, @mjg wrote:

once more i don't have a full picture so can't give a proper review.

Well, the remarks you've had have been pretty helpful so far.

Nov 29 2018, 7:08 PM

Nov 28 2018

kristof added a comment to D18373: pfsync: Performance improvement.

And here's my flame graph: https://people.freebsd.org/~kp/pfsync/buckets.svg

Nov 28 2018, 10:44 PM
kristof added a comment to D18373: pfsync: Performance improvement.

This is a slightly rough first version. I'm sure there are some remaining style issues (and there might be cleanup issues too).

Nov 28 2018, 10:40 PM
kristof added reviewers for D18373: pfsync: Performance improvement: network, mjg.
Nov 28 2018, 10:40 PM
kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().

I have an alternative version in https://reviews.freebsd.org/D18373.
It splits pfsync up into buckets, with their own lock. Performance is slightly better than this, and it's in many ways a simpler change.

Nov 28 2018, 10:40 PM
kristof created D18373: pfsync: Performance improvement.
Nov 28 2018, 10:37 PM

Nov 26 2018

kristof added a comment to D17994: pfsync: Insert static trace points.
In D17994#389260, @eri wrote:

Can you add some text to the manual pages for documenting the feature? Possibly linking to some example?

Nov 26 2018, 10:28 AM

Nov 22 2018

kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().
In D17992#387985, @mjg wrote:

Single-threaded processing is definitely a bottle neck. Perhaps you can have more than one thread pushing stuff and/or some of it can be done by submitting threads instead.

Nov 22 2018, 8:19 PM

Nov 21 2018

kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().
In D17992#386171, @mjg wrote:

So both ring and swi kicking code are significant players. I think a simple and probably good enough solution would just add more rings, perhaps based on the number of hardware threads. Assuming the traffic is hashed to distribute among them, the rings could mostly remain unshared with unrelated threads. Sending out of the traffic would just combine data from all rings.

Nov 21 2018, 7:22 PM

Nov 17 2018

kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().
In D17992#385087, @eri wrote:

What needs to be considered here is the assumption of pfsync is that a state created in pf will be synched at shortest possible cycle to the cluster member.
By defering that assumption is relaxed so figuring out baselines of before and after this change would make this more easy to reason about.

Nov 17 2018, 7:49 PM
kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().

Here are flame graphs for my test setup:

  • plain forwarding
  • pf only
  • unpatched pfsync
  • pfsync with this (and the next) patch
Nov 17 2018, 1:51 PM

Nov 16 2018

kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().

I've been thinking about what we could do to split up the lock, rather than take this approach, and I'm not sure how.

Nov 16 2018, 6:08 PM

Nov 15 2018

kristof accepted D2266: I can find no reason to allow packets with both SYN and FIN bits set past this point in the code. The packet should be dropped and not massaged as it is here..

For what it's worth, OpenBSD don't drop here but henning did add a /* XXX why clear instead of drop? */ comment a few years ago.

Nov 15 2018, 8:54 PM
kristof added a comment to D17992: pfsync: Reduce contention on PFSYNC_LOCK().
In D17992#384518, @mjg wrote:

I think the approach taken here is iffy. Basic problem with this is that even if there is no lock contention anymore, you are still suffering from bouncing cache lines. Also swi_sched probably does not appreciate being called very often.

My performance tests show an improvement of 80-90% with this change. The swi_sched is costly, yes, but it's a pretty significant improvement in throughput.

Nov 15 2018, 6:53 AM

Nov 14 2018

kristof created D17994: pfsync: Insert static trace points.
Nov 14 2018, 9:26 PM
kristof created D17993: pfsync: Call pfsyncintr() directly from pfsync_msg_intr() rather than scheduling a swi.
Nov 14 2018, 9:25 PM
kristof created D17992: pfsync: Reduce contention on PFSYNC_LOCK().
Nov 14 2018, 9:25 PM

Nov 1 2018

kristof updated the diff for D17734: pf: Limit the fragment entry queue length to 64 per bucket..

Formatting change.

Nov 1 2018, 10:58 AM
kristof added a comment to D17505: pfsync: Allow module to be unloaded.

I'm not absolutely sure that all possible races are fixed. There still could be dangling ifnet pointers. But that's up to your justification. If you are sure everything is covered, feel free to remove.

I'm not currently aware of any remaining issues here. It's certainly possible that some still exist, but I've seen no reports and the tests don't provoke them.
I think of this change mostly as a 'I will support this now.' statement.

Nov 1 2018, 10:54 AM

Oct 28 2018

kristof added a reviewer for D17734: pf: Limit the fragment entry queue length to 64 per bucket.: network.
Oct 28 2018, 6:31 AM
kristof updated the summary of D17733: pf: Split the fragment reassembly queue into smaller parts.
Oct 28 2018, 6:31 AM
kristof updated the summary of D17732: pf: Count holes rather than fragments for reassembly.
Oct 28 2018, 6:30 AM
kristof created D17734: pf: Limit the fragment entry queue length to 64 per bucket..
Oct 28 2018, 6:30 AM
kristof created D17733: pf: Split the fragment reassembly queue into smaller parts.
Oct 28 2018, 6:30 AM
kristof created D17732: pf: Count holes rather than fragments for reassembly.
Oct 28 2018, 6:29 AM
kristof abandoned D17634: pf tests: Test ':0' ignoring link-local addresses.

Committed as r339836

Oct 28 2018, 6:00 AM
kristof closed D1309: VIMAGE PF fixes #1.

Assorted pf VIMAGE fixes have been done, and pf is now usable inside VIMAGE jails.

Oct 28 2018, 5:59 AM
kristof abandoned D11401: Kernel pf tests.

An alternative approach (VIMAGE based) was committed instead.

Oct 28 2018, 5:58 AM
kristof commandeered D11401: Kernel pf tests.
Oct 28 2018, 5:57 AM

Oct 21 2018

kristof updated the diff for D17633: pf: Make ':0' ignore link-local v6 addresses too.

Make :0 work without ()

Oct 21 2018, 7:46 PM
kristof added a reviewer for D17634: pf tests: Test ':0' ignoring link-local addresses: network.
Oct 21 2018, 12:38 AM
kristof added a reviewer for D17633: pf: Make ':0' ignore link-local v6 addresses too: network.
Oct 21 2018, 12:38 AM
kristof created D17634: pf tests: Test ':0' ignoring link-local addresses.
Oct 21 2018, 12:37 AM
kristof created D17633: pf: Make ':0' ignore link-local v6 addresses too.
Oct 21 2018, 12:37 AM

Oct 20 2018

kristof added a reviewer for D17625: tcpdump: Log uid on pflog interfaces: network.
Oct 20 2018, 6:32 PM
kristof created D17625: tcpdump: Log uid on pflog interfaces.
Oct 20 2018, 6:31 PM
kristof updated the summary of D17504: pf tests: Basic pfsync test.
Oct 20 2018, 6:21 PM
kristof added a reviewer for D17505: pfsync: Allow module to be unloaded: network.
Oct 20 2018, 6:20 PM

Oct 13 2018

kristof added a comment to D17500: Notify that the ifnet will go away, even on vnet shutdown.

We also don't want to not get notified and have an interface hanging around for ever given pf never releases a reference?

Oct 13 2018, 5:10 PM
kristof added inline comments to D17522: libifconfig: Add initial support for if_bridge.
Oct 13 2018, 3:46 PM

Oct 12 2018

kristof added inline comments to D17522: libifconfig: Add initial support for if_bridge.
Oct 12 2018, 9:17 AM
kristof added a comment to D17499: pfsync: Make pfsync callbacks per-vnet.

Yes. Especially because pfsync currently does a lot of work (including taking a single lock, which really kills throughput) even if it won't actually end up sending the sync packets out.
It should be possible to change pfsync so we don't virtualise the callback pointers and instead immediately check if it's enabled and return if not. We would have to be careful about locking around the up/down (and defer) flags, so it'd be a larger change than this.

Oct 12 2018, 7:55 AM
kristof updated the diff for D17502: pfsync: Handle syncdev going away.

Fix comment style

Oct 12 2018, 7:52 AM
kristof added a comment to D17500: Notify that the ifnet will go away, even on vnet shutdown.

The following also resolves the use-after-free issue.

Oct 12 2018, 7:47 AM

Oct 11 2018

kristof added a comment to D17522: libifconfig: Add initial support for if_bridge.

I’ll try to review in the next day or two, but it needs tests ;) (in a separte commit)
I think there are currently no functional tests for if_bridge, but it should be pretty straightforward to spin up a vnet jail, create a bridge in it and then use libifconfig to add/remove interfaces.
It’d have the very nice side-effect of getting us some basic bridge tests too.

Oct 11 2018, 5:52 PM

Oct 10 2018

kristof added a comment to D17499: pfsync: Make pfsync callbacks per-vnet.

Each vnet may choose to set pfsync0 down, which (see pfsyncioctl() / SIOCSIFFLAGS) set or clear these callbacks.
And while it's okay for a single vnet to decide it doesn't want pfsync to be enabled, it's not okay for that decision to affect all vnets.

Oct 10 2018, 5:33 PM
kristof updated the summary of D17508: pfctl tests: Basic test case for PR 231323.
Oct 10 2018, 4:36 PM
kristof added a reviewer for D17507: pfctl: Dup strings: network.
Oct 10 2018, 4:36 PM
kristof created D17508: pfctl tests: Basic test case for PR 231323.
Oct 10 2018, 4:35 PM
kristof created D17507: pfctl: Dup strings.
Oct 10 2018, 4:35 PM
kristof added a comment to D17500: Notify that the ifnet will go away, even on vnet shutdown.

Test plan:
kldload pfsync
cd /usr/tests/sys/netpfil/pf
kyua test

Oct 10 2018, 3:38 PM
kristof added a reviewer for D17506: pfsync: Add missing unlock: network.
Oct 10 2018, 3:21 PM
kristof created D17506: pfsync: Add missing unlock.
Oct 10 2018, 3:20 PM
kristof added a reviewer for D17502: pfsync: Handle syncdev going away: network.
Oct 10 2018, 3:20 PM
kristof added a reviewer for D17501: pfsync: Ensure uninit is done before pf: network.
Oct 10 2018, 3:20 PM
kristof added a reviewer for D17500: Notify that the ifnet will go away, even on vnet shutdown: network.
Oct 10 2018, 3:19 PM
kristof created D17505: pfsync: Allow module to be unloaded.
Oct 10 2018, 3:19 PM
kristof created D17504: pf tests: Basic pfsync test.
Oct 10 2018, 3:19 PM
kristof created D17502: pfsync: Handle syncdev going away.
Oct 10 2018, 3:18 PM
kristof created D17501: pfsync: Ensure uninit is done before pf.
Oct 10 2018, 3:18 PM
kristof created D17500: Notify that the ifnet will go away, even on vnet shutdown.
Oct 10 2018, 3:18 PM
kristof added a reviewer for D17499: pfsync: Make pfsync callbacks per-vnet: network.
Oct 10 2018, 3:17 PM
kristof created D17499: pfsync: Make pfsync callbacks per-vnet.
Oct 10 2018, 3:16 PM