Page MenuHomeFreeBSD

glebius (Gleb Smirnoff)
User

Projects (6)

User Details

User Since
May 28 2014, 2:27 PM (651 w, 3 h)

Recent Activity

Yesterday

glebius committed rGdf20cc34b9f8: tests/ssl_sendfile: fix flakyness under load.
Tue, Nov 17, 8:05 PM
glebius committed rGf5c21693e43e: tests: unix: Check that MSG_PEEK returns -1 for SCM_RIGHTS (authored by Patrick Crandol <patrick@crandol.com>).
Tue, Nov 17, 5:33 PM
glebius committed rG8f926f291143: unix: Do not copy out internalized SCM_RIGHTS on MSG_PEEK (authored by Patrick Crandol <patrick@crandol.com>).
Tue, Nov 17, 5:33 PM

Mon, Nov 16

glebius committed rGea390ec31eb6: tests/netinet/ip_reass: skip when IP reassembly is disabled (authored by nickbanks_netflix.com).
Mon, Nov 16, 8:19 PM

Sat, Nov 14

glebius committed rGb352368bfbe8: snmp_hostres(3): check if sysconf(3) fails.
Sat, Nov 14, 7:53 PM
glebius added inline comments to D61518: mlx5en: only pass TCP packets to sw LRO.
Sat, Nov 14, 7:14 PM

Fri, Nov 13

glebius accepted D61518: mlx5en: only pass TCP packets to sw LRO.

I always assumed that this already happens for any LRO.

Fri, Nov 13, 9:20 PM
glebius committed rG4edbd4e35c4e: snmp_hostres(3): strengthen work on the devd(8) socket.
Fri, Nov 13, 2:38 AM
glebius closed D61146: snmp_hostres(3): strengthen work on the devd(8) socket.
Fri, Nov 13, 2:38 AM
glebius committed rGb1a743405697: devd: expose maximum record size via sysconf(3).
Fri, Nov 13, 2:38 AM
glebius closed D61346: devd: expose maximum record size via sysconf(3).
Fri, Nov 13, 2:38 AM
glebius committed rG343afa2118ba: unistd.h: sort sysconf(3) arguments.
Fri, Nov 13, 2:38 AM
glebius closed D61345: unistd.h: sort sysconf(3) arguments.
Fri, Nov 13, 2:38 AM
glebius committed rG039ec3ca1bec: devd: define _PATH_DEVD_SOCKET in <paths.h>.
Fri, Nov 13, 2:38 AM
glebius closed D61344: devd: define _PATH_DEVD_SOCKET in <paths.h>.
Fri, Nov 13, 2:38 AM

Thu, Nov 12

glebius added a comment to D61361: callout: optionally round-robin default callout CPU, instead of always using 0.

I would suggest to enable this by default. What are our potential pitfalls?

Thu, Nov 12, 6:17 PM
glebius added a comment to D61385: ip_forward: Take the nexthop reference only for ip_output().

Including net/route/route_var.h is wrong and protocol shouldn't manage nhop references itself without fib_algo help.

Thu, Nov 12, 6:12 PM
glebius accepted D61334: aio_mlock.2: Expand document description.

Thanks! I don't think such trivial changes need review process.

Thu, Nov 12, 6:09 PM
glebius added a comment to D61320: if_firewire: Be more robust while checking the frame size.

P.S. AFAIK the ubiquitous use of signed integers for unsigned entities that we inherited from 4.3BSD comes from late Bruce Evans. He always insisted that at machine level everything is signed, that's why. I never fully bought this point. It seems to me that by today standards all software development norms is that use unsigned types for unsigned entities. And even a very smart man can be wrong sometimes. Thus, we should do sweeping changes to get rid of this heritage. Any other opinions?

Thu, Nov 12, 6:05 PM · firewire
glebius updated subscribers of D61320: if_firewire: Be more robust while checking the frame size.
Thu, Nov 12, 6:02 PM · firewire
glebius updated the diff for D61346: devd: expose maximum record size via sysconf(3).

Provide hw.bus.devctl_maxrecord.

Thu, Nov 12, 5:31 PM

Tue, Nov 10

glebius updated the diff for D61146: snmp_hostres(3): strengthen work on the devd(8) socket.

Use MSG_WAITALL. Not really important for this devd client, but just
follow the suggested practice.

Tue, Nov 10, 5:37 PM
glebius added a comment to D61346: devd: expose maximum record size via sysconf(3).
In D61346#1407002, @mjg wrote:

Now that I'm taking another look, even if going with sysconf, looking at the uma zone is a no-go. One day it might change (e.g., it may get removed altogether) and this will break userspace. A dedicated sysctl with the size is needed, but then I don't see any justificatin for adding a sysconf.

Tue, Nov 10, 5:33 PM

Mon, Nov 9

glebius updated the diff for D61146: snmp_hostres(3): strengthen work on the devd(8) socket.
  • Rebase it over official #define for the socket path and sysconf() value
Mon, Nov 9, 5:44 PM
glebius requested review of D61346: devd: expose maximum record size via sysconf(3).
Mon, Nov 9, 5:42 PM
glebius requested review of D61345: unistd.h: sort sysconf(3) arguments.
Mon, Nov 9, 5:42 PM
glebius requested review of D61344: devd: define _PATH_DEVD_SOCKET in <paths.h>.
Mon, Nov 9, 5:41 PM

Fri, Nov 6

glebius added inline comments to D61234: uma: reuse VM cache buckets when replacement allocation fails.
Fri, Nov 6, 7:01 PM
glebius added a comment to D61037: if_epair: Hash the IP header, not the Ethernet header, for fanout.

Please check if this fixes flakyness of tests/sys/net/bpf:inject in presence of RSS and if it does, please remove the check there.

Fri, Nov 6, 6:50 PM
glebius accepted D61010: inpcb: treat the ephemeral port range as inclusive.

Thanks! I would suggest to split into two commits, since two changes are related but actually different and they also live in different chunks. No reason to squash them.

Fri, Nov 6, 6:46 PM
glebius added a reviewer for D61146: snmp_hostres(3): strengthen work on the devd(8) socket: kevans.
Fri, Nov 6, 12:00 AM

Mon, Nov 2

glebius closed D60967: udp6: make udp6_ctlinput() same as IPv4 version.
Mon, Nov 2, 9:43 PM
glebius committed rGa08dc822f494: udp6: make udp6_ctlinput() same as IPv4 version.
Mon, Nov 2, 9:43 PM
glebius updated the diff for D60987: mbuf: declare mbuf hash types as distinct types, not flags.
  • Limit size of iflib's if_rxd_info.
  • Remove the test.
  • Hide hash macros/functions from userland.
Mon, Nov 2, 7:55 PM
glebius added a comment to D58768: unix: rename unp_connectat_peer() to unp_resolve_peer().

Thanks! I'm a bit lost in the branches there. What branches do include changes that you post here on reviews.f.o?

Mon, Nov 2, 7:34 PM
glebius added a comment to D58768: unix: rename unp_connectat_peer() to unp_resolve_peer().

@inquire_JohnEricson.me do you have a public git repo with your changes? Checked out https://github.com/Ericson2314 already. Is that yours? But no freebsd clone repos found there.

Mon, Nov 2, 7:23 PM
glebius requested review of D61146: snmp_hostres(3): strengthen work on the devd(8) socket.
Mon, Nov 2, 7:04 PM

Sat, Oct 31

glebius added a comment to D61010: inpcb: treat the ephemeral port range as inclusive.

I agree that of by one needs to be fixed. Suggesting both your fix and fix to sysctl handler, so that invalid range is refused.

Sat, Oct 31, 4:15 PM
glebius added a comment to D58768: unix: rename unp_connectat_peer() to unp_resolve_peer().

I will. I'm waiting for review on D60968 and D60969 (btw, you are reviewer) and then I will push several as batch. There are also github submissions on unix(4) waiting on me.

Sat, Oct 31, 4:13 PM

Fri, Oct 30

glebius added a comment to D61041: netinet6: Send redirects from ip6_tryforward().

Looks correct! Don't feel myself expert enough to approve.

Fri, Oct 30, 8:08 PM
glebius added inline comments to D61038: netinet6: Keep a pfil-chosen next hop in ip6_tryforward().
Fri, Oct 30, 8:07 PM
glebius accepted D61042: netinet6: Do not count a packet a hook consumed as discarded in ip6_tryforward().
Fri, Oct 30, 7:58 PM
glebius added a comment to D61038: netinet6: Keep a pfil-chosen next hop in ip6_tryforward().

Looks correct. Not an expert here, though.

Fri, Oct 30, 7:54 PM
glebius added a comment to D61034: netinet6: Check the hop limit after pfil in ip6_tryforward().

Looks like this review was open for about 4 hours.

Fri, Oct 30, 7:39 PM
glebius added a comment to D61010: inpcb: treat the ephemeral port range as inclusive.

AFAIK, the random port allocation can't be really considered a security feature against connection forging these days. An attacker can just inject 50k packets in a millisecond. However, the random port allocation is still useful to avoid revealing information on how many connections a host makes. For the latter purpose a cheaper random is fine.

Fri, Oct 30, 7:35 PM
glebius added a comment to D61010: inpcb: treat the ephemeral port range as inclusive.

IMHO, it would be better to cover invalid sysctl setting in sysctl_net_ipport_check().

Fri, Oct 30, 7:32 PM
glebius added reviewers for D61010: inpcb: treat the ephemeral port range as inclusive: glebius, network.
Fri, Oct 30, 7:25 PM
glebius added a comment to D60971: move SEC_TO_NSEC() out of pf code.

IMHO, more review is needed here. This is a system header that is installed in /usr/include. We should check if other common APIs exist or not for this operation. Does it match existing operations that we already define there?

Fri, Oct 30, 7:23 PM
glebius added inline comments to D60894: pf: optimize state key comparison on amd4.
Fri, Oct 30, 7:05 PM
glebius committed rG60529b9c1b29: libalias: index fully specified inbound links by remote endpoint (authored by paulo_nlink.com.br).
Fri, Oct 30, 5:26 PM
glebius added a comment to D60987: mbuf: declare mbuf hash types as distinct types, not flags.

tests/sys/sys/mbuf_test.c still refs M_HASHTYPE_INNER so it needs to be updated or removed

Fri, Oct 30, 12:25 AM

Thu, Oct 29

glebius added inline comments to D58290: inpcb: add single peer local port caching.
Thu, Oct 29, 11:43 PM
glebius updated the diff for D60987: mbuf: declare mbuf hash types as distinct types, not flags.
  • Implement M_HASHTYPE_ISHASH(m) macro with M_HASHTYPE_GET() macro
Thu, Oct 29, 11:00 PM
glebius added inline comments to D58290: inpcb: add single peer local port caching.
Thu, Oct 29, 9:21 PM
glebius added a comment to D60987: mbuf: declare mbuf hash types as distinct types, not flags.

In general, I'm not sure I like this change. The old code seems less complex

Thu, Oct 29, 9:06 PM
glebius updated the diff for D60987: mbuf: declare mbuf hash types as distinct types, not flags.
  • Fix critical typo in m_adj_decap(), thanks @kbowling
  • Differentiate between M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH
  • M_HASHTYPE_OPAQUE is not really a hash, thanks @gallatin
Thu, Oct 29, 9:02 PM
glebius added a comment to D60987: mbuf: declare mbuf hash types as distinct types, not flags.

Does m_adj_decap() use rssflags &= M_HASHFLAG_INNER, which retains the inner flag where it should clear it?

Thu, Oct 29, 8:55 PM
glebius updated the diff for D60967: udp6: make udp6_ctlinput() same as IPv4 version.

Fixups from Michael.

Thu, Oct 29, 8:37 PM
glebius added inline comments to D60894: pf: optimize state key comparison on amd4.
Thu, Oct 29, 8:35 PM
glebius accepted D60970: carp: Fix packet size checks in carp6_input().
Thu, Oct 29, 8:30 PM
glebius added a comment to D60969: unix: add Linux-compatible abstract sockets.

It doesn't apply cause it depends on D60968.

Thu, Oct 29, 8:27 PM
glebius added a comment to D60987: mbuf: declare mbuf hash types as distinct types, not flags.

this may go back to flow director stuff way back when.

OPAQUE_HASH is that it's a hash that isn't necessarily RSS but has hash properties. Ie it's not some small numerical space of possible values. I think OPAQUE was used by flow director to be able to direct /very specific flows/ in specific ways, rather than how RSS determines how to distribute traffic based on hash.

Thu, Oct 29, 8:26 PM
glebius added a comment to D58769: unix: let unp_connect_peer() own the listener address buffer.

@inquire_JohnEricson.me can you please review my suggestions and update the change, or if you 100% agree just express your agreement to push the change together with my suggestions?

Thu, Oct 29, 8:21 PM
glebius added a comment to D60987: mbuf: declare mbuf hash types as distinct types, not flags.
In D60987#1398461, @np wrote:

There is no summary or proposed commit message so the motivation for this change is unclear.

Thu, Oct 29, 8:12 PM
glebius reopened D58290: inpcb: add single peer local port caching.

@markj can you please repost the review feedback you posted as reply to commit email?

Thu, Oct 29, 8:00 PM
glebius added a comment to D60987: mbuf: declare mbuf hash types as distinct types, not flags.

The one thing I don't understand is the duality between M_HASHTYPE_OPAQUE and M_HASHTYPE_OPAQUE_HASH. I removed this duality as I don't see any sense in it.

Thu, Oct 29, 7:56 PM
glebius requested review of D60987: mbuf: declare mbuf hash types as distinct types, not flags.
Thu, Oct 29, 7:51 PM
glebius closed D60732: tcp: refactor tcp_autosndbuf() and associated sysctls.
Thu, Oct 29, 5:21 PM
glebius closed D60731: tcp: refactor tcp_autorcvbuf() and associated sysctls.
Thu, Oct 29, 5:21 PM
glebius committed rG00c5cf7b6457: ng_pppoe: fix m_pullup() logic.
Thu, Oct 29, 8:46 AM
glebius committed rG930be41aa757: ipfw: declare the ipfw_nat64 module.
Thu, Oct 29, 8:43 AM
glebius added a reverting change for rG289082428c76: inpcb: add single peer local port caching: rG91bff7d7ed46: Revert "inpcb: add single peer local port caching".
Thu, Oct 29, 8:43 AM
glebius committed rG91bff7d7ed46: Revert "inpcb: add single peer local port caching".
Thu, Oct 29, 8:43 AM
glebius committed rG289082428c76: inpcb: add single peer local port caching.
Thu, Oct 29, 8:42 AM
glebius added a reverting change for D58290: inpcb: add single peer local port caching: rG91bff7d7ed46: Revert "inpcb: add single peer local port caching".
Thu, Oct 29, 8:42 AM
glebius closed D58290: inpcb: add single peer local port caching.
Thu, Oct 29, 8:42 AM
glebius committed rG267e4a1a8e9e: pfsync: use result of m_pullup().
Thu, Oct 29, 8:41 AM

Thu, Oct 22

glebius added inline comments to D60894: pf: optimize state key comparison on amd4.
Thu, Oct 22, 5:29 AM
glebius added a comment to D60953: unix: do not copy un-externalized control on MSG_PEEK.

Ok, migrating to github, given there is a history there and today PHP had a bad day :) In general, using reviews.freebsd.org is preferred over github.

Thu, Oct 22, 4:52 AM
glebius accepted D60949: pf: Store the rounded source node hash size.

Oops, screwed up. Thanks for the fix!

Thu, Oct 22, 4:48 AM
glebius added a comment to D60894: pf: optimize state key comparison on amd4.
In D60894#1397813, @mjg wrote:

I did not see any point in benchmarking this, it is strictly less work in the fast path.

Thu, Oct 22, 4:44 AM
glebius added inline comments to D60894: pf: optimize state key comparison on amd4.
Thu, Oct 22, 4:42 AM
glebius requested changes to D58770: unix: set and clear UNP_CONNECTING in the same function.

I have alternative D60968. It completes the same goal of set and clear UNP_CONNECTING in the same function and makes all operation linear and thus more readable. I also need it for abstract sockets in D60969. Can you please review?

Thu, Oct 22, 4:11 AM
glebius requested review of D60969: unix: add Linux-compatible abstract sockets.
Thu, Oct 22, 4:09 AM
glebius requested review of D60968: unix: inline unp_connect_peer() into unp_connectat().
Thu, Oct 22, 4:08 AM
glebius requested review of D60967: udp6: make udp6_ctlinput() same as IPv4 version.
Thu, Oct 22, 4:04 AM
glebius committed rG9c67d6ed9632: inpcb: inp_next() never returns a freed inpcb.
Thu, Oct 22, 3:32 AM
glebius committed rG49548ea75c91: udp: in_pcblookup() never returns a freed inpcb.
Thu, Oct 22, 3:32 AM
glebius committed rGabb8b02550c3: icmp6: make icmp6_notify_error() more readable.
Thu, Oct 22, 3:32 AM
glebius committed rG331333ba0186: ipfw: hold sleepable "upper half lock" for periodic state expiry.
Thu, Oct 22, 3:32 AM
glebius committed rG7146ad04f18e: tcp: refactor tcp_autosndbuf() and associated sysctls.
Thu, Oct 22, 3:32 AM
glebius committed rG04cd9bca4dab: tcp: refactor tcp_autorcvbuf() and associated sysctls.
Thu, Oct 22, 3:32 AM
glebius closed D60869: ipfw: hold sleepable "upper half lock" for periodic state expiry.
Thu, Oct 22, 3:32 AM

Tue, Oct 20

glebius closed D60689: net*: use inlined m_pullup().
Tue, Oct 20, 12:50 AM

Mon, Oct 19

glebius committed rG4488a4f0486c: inpcb: make ipport_randomized static.
Mon, Oct 19, 7:43 PM
glebius added a comment to D60830: tcp: build tcp_sack.c and tcp_reass.c under bounds-safety.

Is the VNET change related to the TCP changes? Are TCP changes related at all?

Mon, Oct 19, 6:14 PM
glebius added a comment to D58290: inpcb: add single peer local port caching.

We are in the late testing stages of this patch. I'd like to push it before stabweek Monday. Tag @markj if you want to take an extra look.

Mon, Oct 19, 6:07 PM

Oct 19 2026

glebius committed rGf2bba0192166: ipfw: keep flag(s) definition next to member that holds them.
Oct 19 2026, 5:33 PM
glebius committed rGbc9e66886a25: ipfw: gc disabled old hashing code.
Oct 19 2026, 5:33 PM
glebius committed rG671babc79497: mbuf: make m_pullup() an inline over most expected case.
Oct 19 2026, 5:33 PM