Page MenuHomeFreeBSD

tuexen (Michael Tuexen)
User

Projects (4)

User Details

User Since
Feb 4 2016, 4:45 PM (562 w, 5 d)

Recent Activity

Yesterday

tuexen closed D61685: tcp: improve consistency.
Tue, Nov 17, 7:55 PM
tuexen committed rGe3483dbc91b5: tcp: improve consistency.
Tue, Nov 17, 7:55 PM
tuexen retitled D61685: tcp: improve consistency from tcp: small refactoring
Tue, Nov 17, 7:26 PM
tuexen requested review of D61685: tcp: improve consistency.
Tue, Nov 17, 7:02 PM

Mon, Nov 16

tuexen committed rGd174085f9f89: tcp: improve sender side SWS avoidance.
Mon, Nov 16, 5:30 PM
tuexen closed D61605: tcp: improve sender side SWS avoidance .
Mon, Nov 16, 5:30 PM
tuexen requested review of D61605: tcp: improve sender side SWS avoidance .
Mon, Nov 16, 12:00 AM

Sun, Nov 15

tuexen closed D61586: tcp bbr: improve handling of retransmissions in LAST-ACK.
Sun, Nov 15, 9:41 PM
tuexen committed rG3c6fd437b182: tcp bbr: improve handling of retransmissions in LAST-ACK.
Sun, Nov 15, 9:41 PM
tuexen closed D61600: tcp bbr: improve persits timeout computation.
Sun, Nov 15, 8:25 PM
tuexen committed rG751882926dc0: tcp bbr: improve persits timeout computation.
Sun, Nov 15, 8:25 PM
tuexen requested review of D61600: tcp bbr: improve persits timeout computation.
Sun, Nov 15, 8:12 PM
tuexen updated the diff for D61586: tcp bbr: improve handling of retransmissions in LAST-ACK.

Provide patch with context.

Sun, Nov 15, 7:11 PM
tuexen reopened D61586: tcp bbr: improve handling of retransmissions in LAST-ACK.
Sun, Nov 15, 7:09 PM
tuexen closed D61581: tcp bbr: improve retransmit timout calculation.
Sun, Nov 15, 7:09 PM
tuexen committed rG2f98793a13fe: tcp bbr: improve retransmit timout calculation.
Sun, Nov 15, 7:09 PM
tuexen added a reverting change for rG47ef696a7696: tcp bbr: improve retransmit timout calculation: rG7659a4f98c6c: Revert "tcp bbr: improve retransmit timout calculation".
Sun, Nov 15, 7:03 PM
tuexen committed rG7659a4f98c6c: Revert "tcp bbr: improve retransmit timout calculation".
Sun, Nov 15, 7:03 PM
tuexen added a reverting change for D61586: tcp bbr: improve handling of retransmissions in LAST-ACK: rG7659a4f98c6c: Revert "tcp bbr: improve retransmit timout calculation".
Sun, Nov 15, 7:03 PM
tuexen closed D61585: tcp bbr: improve invalidating rtt.
Sun, Nov 15, 7:00 PM
tuexen committed rG8018d75601df: tcp bbr: improve invalidating rtt.
Sun, Nov 15, 7:00 PM
tuexen committed rG47ef696a7696: tcp bbr: improve retransmit timout calculation.
Sun, Nov 15, 6:55 PM
tuexen closed D61586: tcp bbr: improve handling of retransmissions in LAST-ACK.
Sun, Nov 15, 6:55 PM

Sat, Nov 14

tuexen requested review of D61586: tcp bbr: improve handling of retransmissions in LAST-ACK.
Sat, Nov 14, 11:45 PM
tuexen updated the test plan for D61585: tcp bbr: improve invalidating rtt.
Sat, Nov 14, 11:08 PM
tuexen requested review of D61585: tcp bbr: improve invalidating rtt.
Sat, Nov 14, 11:04 PM
tuexen committed rG34436bb7cf97: tcp bbr: cleanup.
Sat, Nov 14, 9:27 PM
tuexen requested review of D61581: tcp bbr: improve retransmit timout calculation.
Sat, Nov 14, 9:03 PM

Fri, Nov 13

tuexen accepted D61509: ip6_forward: Free the ICMPv6 copy on all error paths.
Fri, Nov 13, 4:38 PM
tuexen added a reviewer for D61509: ip6_forward: Free the ICMPv6 copy on all error paths: msvoelker.
Fri, Nov 13, 4:30 PM

Thu, Nov 12

tuexen added a comment to D60460: tcp: count only the SACK blocks that fit in tcp_maxseg().

@nprice ping again. I would really like to have this in tree. If you don't want to commit this, let me know and I can do it.

Thu, Nov 12, 8:33 PM
tuexen committed rGd40890ffbd8b: TCP RACK: fix RSM computation for TCP fast open.
Thu, Nov 12, 6:11 PM

Sat, Nov 7

tuexen committed rGff066579fd77: TCP RACK: fix RSM computation for TCP fast open.
Sat, Nov 7, 1:38 PM
tuexen closed D61149: TCP RACK: fix RSM computation for TCP fast open.
Sat, Nov 7, 1:38 PM

Fri, Nov 6

tuexen accepted D61159: Mark Johnsons Claude code found a duplicate call to rack_check_reorder_ack() in rack.c.
Fri, Nov 6, 7:23 PM
tuexen added a reviewer for D61149: TCP RACK: fix RSM computation for TCP fast open: nickbanks_netflix.com.
Fri, Nov 6, 9:44 AM

Wed, Nov 4

tuexen added a comment to D60460: tcp: count only the SACK blocks that fit in tcp_maxseg().

@nprice Can you commit this?

Wed, Nov 4, 4:05 PM
tuexen added a comment to D61149: TCP RACK: fix RSM computation for TCP fast open.

{F177033943}Hi Michael,

While your patch fixes 297006 panic, it still panics with Robert's code from 297273. From what I see the code path in 297273 is different.

I agree. There was a problem on my side. I added packetdrill reproducers to the PRs.

I attached kyua tests for both cases generated with Claude Code.

Maxim

Wed, Nov 4, 12:55 PM
tuexen updated the summary of D61149: TCP RACK: fix RSM computation for TCP fast open.
Wed, Nov 4, 12:53 PM

Tue, Nov 3

tuexen added a comment to D61128: tpm: crb: Log cmd failures, correct Pluton mailbox handshake.

I tested this patch on a Framework AMD laptop and it resolves the problem for me. No more messages observed.

Tue, Nov 3, 9:18 PM

Mon, Nov 2

tuexen requested review of D61149: TCP RACK: fix RSM computation for TCP fast open.
Mon, Nov 2, 10:04 PM
tuexen accepted D61136: Broadcom tester's were able to trigger a KASSERT in rack due to an accounting issue..

This fixes PR 296474.
It would be great if this fix can go in soon, such we can get it into 15.2. Please refer to the PR.

Mon, Nov 2, 9:16 PM

Fri, Oct 30

tuexen accepted D60967: udp6: make udp6_ctlinput() same as IPv4 version.
Fri, Oct 30, 7:02 PM

Thu, Oct 29

tuexen accepted D60980: vtnet: honor the advertised MTU limit on attach.
Thu, Oct 29, 7:58 PM
tuexen committed rG7b8151b0cd1b: sctp: improve handling of SO_SETFIB socket option (authored by albin.hellqvist_ericsson.com).
Thu, Oct 29, 6:46 PM
tuexen committed rGfafba5672b1a: sysctl: don't loop when EAGAIN is returned.
Thu, Oct 29, 8:45 AM
tuexen committed rG5ad9c6b31504: sctp: improve handling of SO_SETFIB socket option (authored by albin.hellqvist_ericsson.com).
Thu, Oct 29, 8:42 AM
tuexen committed rG434994d73c5e: sctp: improve handling of SO_SETFIB socket option (authored by albin.hellqvist_ericsson.com).
Thu, Oct 29, 8:42 AM
tuexen committed rG335bee05d914: sockstat: provide more BBLog information.
Thu, Oct 29, 8:40 AM
tuexen committed rGc31cc4376dac: sockstat: improve consistency.
Thu, Oct 29, 8:40 AM
tuexen committed rG2001681da179: sockstat: improve consistency.
Thu, Oct 29, 8:40 AM
tuexen added a reverting change for rG2001681da179: sockstat: improve consistency: rGa949836043f8: Revert "sockstat: improve consistency".
Thu, Oct 29, 8:40 AM
tuexen committed rG0b3f85d53e29: sctp: hold tcp lock when calling sctp_calc_rwnd() (authored by Lukas Book <lkbook@outlook.de>).
Thu, Oct 29, 8:40 AM
tuexen committed rGa949836043f8: Revert "sockstat: improve consistency".
Thu, Oct 29, 8:40 AM
tuexen committed rGbbd9741f0b84: sctp: improve locking validation (authored by Lukas Book <lkbook@outlook.de>).
Thu, Oct 29, 8:40 AM
tuexen committed rG3b09a9346691: Revert "sockstat: improve consistency".
Thu, Oct 29, 8:40 AM
tuexen added a reverting change for rG71b1f005a56e: sockstat: improve consistency: rG3b09a9346691: Revert "sockstat: improve consistency".
Thu, Oct 29, 8:40 AM
tuexen committed rGa492f3101668: epair: allow disabling receive checksum offloading (authored by msvoelker).
Thu, Oct 29, 8:40 AM
tuexen committed rGf9b8c1530cf6: tcpdrop: improve handling of -C and -S.
Thu, Oct 29, 8:40 AM
tuexen committed rG0ab8bd1eee93: sctp: initialize inp to avoid uninitialized use (authored by bz).
Thu, Oct 29, 8:40 AM
tuexen committed rGae93abe7e9a0: tcp md5: improve consistency.
Thu, Oct 29, 8:40 AM
tuexen committed rGf23fa56df5df: tcp md5: fix accounting of ACK with bad signatures.
Thu, Oct 29, 8:40 AM
tuexen committed rG6fa319d8c670: tcp md5: fix accounting for SYN segments with unexpected signature.
Thu, Oct 29, 8:40 AM
tuexen committed rG3d5546429b99: tcp md5: improve handling of tcps_sig_err_sigopt counter.
Thu, Oct 29, 8:40 AM
tuexen committed rG6873fec25b54: tcp: minor cleanup.
Thu, Oct 29, 8:40 AM
tuexen committed rG71b1f005a56e: sockstat: improve consistency.
Thu, Oct 29, 8:40 AM
tuexen committed rGa991de89495b: tcp: cleanup whitespaces.
Thu, Oct 29, 8:40 AM

Thu, Oct 22

tuexen added inline comments to D60967: udp6: make udp6_ctlinput() same as IPv4 version.
Thu, Oct 22, 5:47 AM

Wed, Oct 21

tuexen accepted D60938: sctp: Allow to Ignore iface type when determining address scope.
Wed, Oct 21, 6:03 PM
tuexen added inline comments to D60938: sctp: Allow to Ignore iface type when determining address scope.
Wed, Oct 21, 5:47 PM
tuexen accepted D60938: sctp: Allow to Ignore iface type when determining address scope.
Wed, Oct 21, 5:36 PM

Oct 16 2026

tuexen added inline comments to D60759: tcp/rack: reject wma_divisor values <= 0.
Oct 16 2026, 8:58 PM
tuexen accepted D60759: tcp/rack: reject wma_divisor values <= 0.
Oct 16 2026, 8:46 PM
tuexen added inline comments to D60759: tcp/rack: reject wma_divisor values <= 0.
Oct 16 2026, 8:45 PM
tuexen added inline comments to D60759: tcp/rack: reject wma_divisor values <= 0.
Oct 16 2026, 6:59 PM
tuexen committed rG38479a483278: sctp: improve locking validation (authored by Lukas Book <lkbook@outlook.de>).
Oct 16 2026, 7:50 AM
tuexen added inline comments to D60759: tcp/rack: reject wma_divisor values <= 0.
Oct 16 2026, 6:45 AM
tuexen accepted D60759: tcp/rack: reject wma_divisor values <= 0.
Oct 16 2026, 6:45 AM

Oct 15 2026

tuexen accepted D60460: tcp: count only the SACK blocks that fit in tcp_maxseg().
Oct 15 2026, 5:29 PM

Oct 14 2026

tuexen accepted D60733: tcp: update the send/recv pipes in hostcache during connection lifetime.
Oct 14 2026, 7:52 PM
tuexen accepted D60736: mbuf: remove remnants of caddr_t.
Oct 14 2026, 7:50 PM
tuexen accepted D60732: tcp: refactor tcp_autosndbuf() and associated sysctls.
Oct 14 2026, 7:48 PM
tuexen accepted D60731: tcp: refactor tcp_autorcvbuf() and associated sysctls.
Oct 14 2026, 7:44 PM

Oct 11 2026

tuexen added a comment to D60460: tcp: count only the SACK blocks that fit in tcp_maxseg().

I am trying to trigger a "sent too much" KASSERT but was not successful up to now. Can you provide a BBLog file of such a situation?

Here's a bblog file and a packetdrill repro


Oct 11 2026, 9:23 PM
tuexen committed rG429f2fb48b92: sctp: hold tcp lock when calling sctp_calc_rwnd() (authored by Lukas Book <lkbook@outlook.de>).
Oct 11 2026, 8:05 PM
tuexen added a comment to D60460: tcp: count only the SACK blocks that fit in tcp_maxseg().

I am trying to trigger a "sent too much" KASSERT but was not successful up to now. Can you provide a BBLog file of such a situation?

Oct 11 2026, 7:45 PM

Oct 9 2026

tuexen added a comment to D60460: tcp: count only the SACK blocks that fit in tcp_maxseg().

Let me write a packetdrill-test to make sure we catch an KASSERT-issue.

Oct 9 2026, 5:03 AM

Oct 8 2026

tuexen abandoned D60281: tcp: improve stack switching.

D60444 solves the problem in a better way.

Oct 8 2026, 1:13 PM
tuexen committed rGd39b12e9030b: sysctl: don't loop when EAGAIN is returned.
Oct 8 2026, 1:09 PM
tuexen closed D60444: sysctl: don't loop when EAGAIN is returned.
Oct 8 2026, 1:09 PM
tuexen added a comment to D60416: iflib: Ring transmit doorbells once per receive batch.

I need to think about this more. For your workload, I see the benefit. For mine (busy CDN server), where we're running TCP, and where our TX rings are constantly > 80% full, I think this just adds overhead (due to full rings) and risk (due to TCP).

TCP is also interesting, because one RX packet can generate megabytes of TX. I'm not sure its appropriate at all for TCP. I'm going to add some folks from the transport side.

Oct 8 2026, 5:52 AM

Oct 7 2026

tuexen added a comment to D60281: tcp: improve stack switching.
In D60281#1386309, @phk wrote:

I have no objection to removing the sysctl EAGAIN loop, if anything else uses it, that needs to be fixed properly.

Oct 7 2026, 3:13 PM
tuexen requested review of D60444: sysctl: don't loop when EAGAIN is returned.
Oct 7 2026, 3:13 PM
tuexen updated the diff for D60281: tcp: improve stack switching.

Improve comment.

Oct 7 2026, 10:33 AM
tuexen added a comment to D60281: tcp: improve stack switching.

So what happens here is that the KPI of TCP stack switching is being changed under pressure of one of its consumers that has a special handling of EAGAIN. And we are substituting this EAGAIN to factually incorrect EINVAL. No, the argument is valid, the connection exists it is just its state that doesn't allow the switch (supposedly temporarily).

Oct 7 2026, 10:31 AM
tuexen updated the summary of D60281: tcp: improve stack switching.
Oct 7 2026, 10:21 AM
tuexen updated the diff for D60281: tcp: improve stack switching.

As suggested by Gleb: lie about error in sysctl_setsockopt() to avoid a loop in userland_sysctl().

Oct 7 2026, 10:20 AM

Oct 6 2026

tuexen updated the summary of D60281: tcp: improve stack switching.
Oct 6 2026, 5:57 PM
tuexen updated the diff for D60281: tcp: improve stack switching.

Let tfb_tcp_handoff_ok() return bool. That is simpler and out of tree stacks a caught by the compiler.

Oct 6 2026, 5:54 PM

Oct 5 2026

tuexen added a comment to D60281: tcp: improve stack switching.

An alternative approach would be to change int (*tfb_tcp_handoff_ok)(struct tcpcb *) to bool (*tfb_tcp_handoff_ok)(struct tcpcb *) and let the caller map false to EINVAL. Would that be clearer? That would also allow the compiler to catch the change in TCP stacks other than BBR and RACK.

Oct 5 2026, 6:54 PM