Page MenuHomeFreeBSD

jtl (Jonathan T. Looney)
UserAdministrator

Projects

User Details

User Since
Oct 29 2015, 5:25 PM (519 w, 2 d)
Roles
Administrator

Recent Activity

Thu, Oct 9

jtl committed rG049010b03a98: x86: Defer non-fatal MCA message output from the HW interrupt context (authored by jtl).
x86: Defer non-fatal MCA message output from the HW interrupt context
Thu, Oct 9, 5:21 PM
jtl committed rG8b6bbdb80fc0: x86: Close race condition on MCA task queues at startup (authored by jtl).
x86: Close race condition on MCA task queues at startup
Thu, Oct 9, 5:21 PM
jtl committed rG1b00e2ad86b3: x86: Keep cumulative MCA statistics in the kernel (authored by jtl).
x86: Keep cumulative MCA statistics in the kernel
Thu, Oct 9, 5:20 PM
jtl committed rG3f0df559cbe4: x86: Defer non-fatal MCA message output from the HW interrupt context (authored by jtl).
x86: Defer non-fatal MCA message output from the HW interrupt context
Thu, Oct 9, 5:17 PM
jtl committed rG8d0f269989d8: x86: Close race condition on MCA task queues at startup (authored by jtl).
x86: Close race condition on MCA task queues at startup
Thu, Oct 9, 5:16 PM
jtl committed rGbe85c5d65750: x86: Keep cumulative MCA statistics in the kernel (authored by jtl).
x86: Keep cumulative MCA statistics in the kernel
Thu, Oct 9, 5:15 PM

Wed, Oct 8

jtl committed rG55c28005f544: x86: support sending non-fatal MCA messages to syslog (authored by jtl).
x86: support sending non-fatal MCA messages to syslog
Wed, Oct 8, 7:14 PM
jtl closed D12278: x86: support sending non-fatal MCA messages to syslog.
Wed, Oct 8, 7:14 PM
jtl committed rG0d14826de7d9: x86: Allow MCA messages to be rate-limited (authored by jtl).
x86: Allow MCA messages to be rate-limited
Wed, Oct 8, 7:13 PM
jtl closed D52946: x86: Allow MCA messages to be rate-limited.
Wed, Oct 8, 7:12 PM
jtl added inline comments to D52946: x86: Allow MCA messages to be rate-limited.
Wed, Oct 8, 3:58 PM
jtl updated the diff for D52946: x86: Allow MCA messages to be rate-limited.

updates to incorporate review feedback

Wed, Oct 8, 3:35 PM
jtl added inline comments to D52946: x86: Allow MCA messages to be rate-limited.
Wed, Oct 8, 3:30 PM
jtl added inline comments to D52946: x86: Allow MCA messages to be rate-limited.
Wed, Oct 8, 3:15 PM
jtl updated the diff for D12278: x86: support sending non-fatal MCA messages to syslog.

incorporate review feedback; rebase on top of changes to D52946

Wed, Oct 8, 12:39 PM
jtl updated the diff for D52946: x86: Allow MCA messages to be rate-limited.

Reworked patch to use a large-ish buffer allocated from BSS and protected by a lock if called from an interrupt context. As noted in the commit message, this will result in increased lock contention during an interrupt storm which exceeds the capacity of the free list; however, overall lock contention should still be lower than it was when mca_log() was called with the mca_lock held.

Wed, Oct 8, 11:52 AM

Tue, Oct 7

jtl added a comment to D52946: x86: Allow MCA messages to be rate-limited.

Not sure whether it needs to be addressed or not, but for completeness there is the possibility of some sequence of events occasionally emptying the mca_freelist, and then my understanding is that mca_record_entry() will work in "degraded mode" directly call mca_log() in an interrupt context (next to "MCA: Unable to allocate space for an event.\n"), and as long as that situation persists, the corresponding MCA printf() won't be ratechecked any more.

It could be a very unlikely theoretical case with the mca_postscan() changes, but with previous version of code, a couple of machines were able to trigger the "MCA: Unable to allocate space for an event" situation in netflix fleet.

Tue, Oct 7, 6:52 PM
jtl closed D52942: x86: Add a way to inject artificial MCA events for testing.
Tue, Oct 7, 4:43 PM
jtl committed rG1c2fc62e4a96: x86: Add a way to inject artificial MCA events for testing (authored by lprylli_netflix.com).
x86: Add a way to inject artificial MCA events for testing
Tue, Oct 7, 4:43 PM
jtl updated the diff for D12278: x86: support sending non-fatal MCA messages to syslog.

Updating to apply cleanly on top of D52946, which introduces the use of sbuf to gather the log message.

Tue, Oct 7, 2:38 PM
jtl committed rGe770e32aa3a0: x86: Reduce amount of time the MCA lock is held while emitting records (authored by jtl).
x86: Reduce amount of time the MCA lock is held while emitting records
Tue, Oct 7, 2:27 PM
jtl closed D52938: x86: Reduce amount of time the MCA lock is held while emitting records.
Tue, Oct 7, 2:27 PM
jtl updated the diff for D52946: x86: Allow MCA messages to be rate-limited.

Remove unneeded check for mca_startup_done

Tue, Oct 7, 2:16 PM
jtl abandoned D52943: x86: print MCAs to the console before MCA initialization.

Note that I am not 100% sure this can actually happen, since I am not 100% sure we will ever enable interrupts before we run mca_setup().

I don't quite follow, MCEs will be delivered even if interrupts are disabled. _mca_init() enables delivery of MCEs at SI_SUB_CPU time; what's wrong with calling mca_record_entry() after that point?

Tue, Oct 7, 2:13 PM
jtl accepted D52944: sockstat: support reporting of BBLog state.
Tue, Oct 7, 1:14 PM
jtl updated the diff for D52946: x86: Allow MCA messages to be rate-limited.

Fix logic in case (mode != polled && (mca_startup_done || (rec->mr_status & MC_STATUS_UC)))

Tue, Oct 7, 12:56 PM

Mon, Oct 6

jtl requested review of D52946: x86: Allow MCA messages to be rate-limited.
Mon, Oct 6, 9:30 PM
jtl added a comment to D52942: x86: Add a way to inject artificial MCA events for testing.
In D52942#1209601, @jtl wrote:

Updated to hide this code under the DIAGNOSTIC kernel config option.

Mon, Oct 6, 6:49 PM
jtl updated the diff for D52942: x86: Add a way to inject artificial MCA events for testing.

Updated to hide this code under the DIAGNOSTIC kernel config option.

Mon, Oct 6, 6:48 PM
jtl requested review of D52943: x86: print MCAs to the console before MCA initialization.
Mon, Oct 6, 6:40 PM
jtl requested review of D52942: x86: Add a way to inject artificial MCA events for testing.
Mon, Oct 6, 6:13 PM
jtl committed rG1cc96642675b: tcp: close two minor races with debug messages (authored by jtl).
tcp: close two minor races with debug messages
Mon, Oct 6, 4:24 PM
jtl committed rGa1f20ce89403: tcp: close two minor races with debug messages (authored by jtl).
tcp: close two minor races with debug messages
Mon, Oct 6, 4:23 PM
jtl requested review of D52938: x86: Reduce amount of time the MCA lock is held while emitting records.
Mon, Oct 6, 4:07 PM
jtl committed rG5e1611a37a02: x86: Defer non-fatal MCA message output from the HW interrupt context (authored by jtl).
x86: Defer non-fatal MCA message output from the HW interrupt context
Mon, Oct 6, 4:04 PM
jtl closed D12275: x86: Defer non-fatal MCA message output from the hardware interrupt context.
Mon, Oct 6, 4:03 PM
jtl added inline comments to D12275: x86: Defer non-fatal MCA message output from the hardware interrupt context.
Mon, Oct 6, 4:02 PM
jtl committed rG4a5d8670a7d5: x86: Close race condition on MCA task queues at startup (authored by jtl).
x86: Close race condition on MCA task queues at startup
Mon, Oct 6, 4:00 PM
jtl closed D12276: x86: Close race condition on MCA task queues at startup.
Mon, Oct 6, 4:00 PM
jtl accepted D52934: tcp: improve SEG.ACK validation in SYN-RECEIVED.
Mon, Oct 6, 3:25 PM

Fri, Oct 3

jtl abandoned D16471: Some TCP reassembly improvements.
Fri, Oct 3, 6:29 PM
jtl updated the diff for D12275: x86: Defer non-fatal MCA message output from the hardware interrupt context.

update diff to account for 8 years of bitrot

Fri, Oct 3, 6:25 PM
jtl abandoned D12277: Allow the user more control over MCA throttling.

AFAIK, no one has bemoaned the lack of this feature in the 8 years this has been sitting here. I think it is safe to abandon this.

Fri, Oct 3, 4:38 PM
jtl updated the diff for D12276: x86: Close race condition on MCA task queues at startup.

Updating the diff to account for 8 years of bit rot.

Fri, Oct 3, 4:32 PM
jtl committed rG5f1f7d8457d4: x86: Keep cumulative MCA statistics in the kernel (authored by jtl).
x86: Keep cumulative MCA statistics in the kernel
Fri, Oct 3, 4:04 PM
jtl closed D52709: x86: Keep cumulative MCA statistics in the kernel.
Fri, Oct 3, 4:04 PM
jtl updated the diff for D52709: x86: Keep cumulative MCA statistics in the kernel.

incorporating review feedback

Fri, Oct 3, 2:20 PM
jtl added a comment to D52709: x86: Keep cumulative MCA statistics in the kernel.

Thanks for the review! I agree on the documentation comment.

Fri, Oct 3, 1:52 PM

Thu, Oct 2

jtl added a comment to D52872: tcp: bump max snd buffer size for autoscaling.
In D52872#1207881, @jtl wrote:

Is this worthy of a release note?

Yes, the combination of D52871, D52872, and D52873.

Thu, Oct 2, 9:20 PM
jtl accepted D52873: sockets: bump socket buffer limit.

My main concern about these three revisions is that they are somewhat susceptible of remote manipulation, and that could make it easier to DoS a server. However, I view that as a tradeoff that the user needs to make, and think an appropriate release note should suffice to warn about these issues.

Thu, Oct 2, 9:19 PM
jtl accepted D52871: tcp: bump max rcv buffer size for autoscaling.

My main concern about these three revisions is that they are somewhat susceptible of remote manipulation, and that could make it easier to DoS a server. However, I view that as a tradeoff that the user needs to make, and think an appropriate release note should suffice to warn about these issues.

Thu, Oct 2, 9:18 PM
jtl accepted D52872: tcp: bump max snd buffer size for autoscaling.

Is this worthy of a release note?

Thu, Oct 2, 8:31 PM
jtl added a comment to D52709: x86: Keep cumulative MCA statistics in the kernel.

If anyone is planning to review this (or if I should ask more people), let me know. Otherwise, I will probably just commit this soon. AFAICT, this is fairly innocuous (new feature with low risk of breaking existing things). But, there is a reason we get these things reviewed...

Thu, Oct 2, 8:23 PM
jtl committed rGad38f6a0b466: tcp: close two minor races with debug messages (authored by jtl).
tcp: close two minor races with debug messages
Thu, Oct 2, 8:10 PM
jtl closed D52868: tcp: close two minor races with debug messages.
Thu, Oct 2, 8:10 PM
jtl added inline comments to D52849: tcp: improve segment validation in SYN-RECEIVED.
Thu, Oct 2, 5:34 PM
jtl requested review of D52868: tcp: close two minor races with debug messages.
Thu, Oct 2, 5:33 PM
jtl accepted D52731: sys/netinet6: Use atomic(9) for dad_failures counter.

Thanks for doing this! LGTM other than one nit.

Thu, Oct 2, 5:07 PM
jtl added inline comments to D52849: tcp: improve segment validation in SYN-RECEIVED.
Thu, Oct 2, 4:33 PM

Wed, Oct 1

jtl accepted D52840: tcp: keep SYN-cache entry when sending of challenge ACK fails.
Wed, Oct 1, 9:00 PM

Mon, Sep 29

jtl committed rGf801346706ea: ip6: add SO_BINTIME support (authored by jtl).
ip6: add SO_BINTIME support
Mon, Sep 29, 3:14 PM
jtl committed rG5b59ff1e1607: ip6: add SO_BINTIME support (authored by jtl).
ip6: add SO_BINTIME support
Mon, Sep 29, 3:08 PM

Wed, Sep 24

jtl requested review of D52709: x86: Keep cumulative MCA statistics in the kernel.
Wed, Sep 24, 2:47 PM

Mon, Sep 15

jtl committed rGcd02a8a9f8be: ip6: add SO_BINTIME support (authored by jtl).
ip6: add SO_BINTIME support
Mon, Sep 15, 4:50 PM
jtl closed D52504: ip6: add SO_BINTIME support.
Mon, Sep 15, 4:50 PM

Fri, Sep 12

jtl added inline comments to D52504: ip6: add SO_BINTIME support.
Fri, Sep 12, 6:25 PM
jtl requested review of D52504: ip6: add SO_BINTIME support.
Fri, Sep 12, 6:24 PM

May 29 2025

jtl added inline comments to D50581: sendfile: don't hack sb_lowat for sockets that manage the watermark.
May 29 2025, 3:23 PM

Jun 27 2024

jtl added inline comments to D45746: callout: provide CALLOUT_TRYLOCK flag.
Jun 27 2024, 5:31 PM

Jun 26 2024

jtl accepted D45411: tcp: simplify endpoint creation at the passive side.
Jun 26 2024, 7:29 PM
jtl accepted D45749: tcp: fix cleanup when in_pcbconnect() fails in syncache_socket().
Jun 26 2024, 7:24 PM
jtl accepted D45748: tcp: get rid of TDP_INTCPCALLOUT.
Jun 26 2024, 5:27 PM
jtl accepted D45747: tcp: use CALLOUT_TRYLOCK for the TCP callout.
Jun 26 2024, 5:27 PM
jtl accepted D45746: callout: provide CALLOUT_TRYLOCK flag.

Overall, I think this approach has value. See my in-line comments for suggestions on things to review further.

Jun 26 2024, 5:25 PM
jtl accepted D45745: locks: augment lock_class with lc_trylock method.
Jun 26 2024, 5:09 PM

Jun 21 2024

jtl added a comment to D45411: tcp: simplify endpoint creation at the passive side.

Thanks! I like this approach. I've added a few comments about potential enhancements.

Jun 21 2024, 11:35 PM

May 31 2024

jtl accepted D45425: tcp: in tcp_close() call tcp_timer_stop() after pcb has been dropped.
May 31 2024, 6:24 PM
jtl added a comment to D45411: tcp: simplify endpoint creation at the passive side.
In D45411#1035980, @jtl wrote:

Should we check for TCP_FUNC_BEING_REMOVED?

I thought about this and did again after you raised the question. The original code didn't.
...
However, since we don't hold the tcp_function_lock, the tfb can be removed any time before or after we set the TCP function block in tcp_newtcpcb. It cannot go away, since we hold a reference count. I guess that is the reason why there was no check...

May 31 2024, 3:51 PM
jtl added a comment to D45419: tcp: mark TCP stacks which can serve as a default stack.

Can you give a brief explanation of the problem this is supposed to solve?

May 31 2024, 3:18 PM

May 30 2024

jtl added a comment to D45411: tcp: simplify endpoint creation at the passive side.

Should we check for TCP_FUNC_BEING_REMOVED?

May 30 2024, 7:52 PM

Dec 20 2023

jtl added inline comments to D43122: inpcb: reoder inpcb destruction.
Dec 20 2023, 8:23 PM
jtl added inline comments to D43122: inpcb: reoder inpcb destruction.
Dec 20 2023, 3:58 PM

Dec 19 2023

jtl added inline comments to D43122: inpcb: reoder inpcb destruction.
Dec 19 2023, 7:02 PM

Nov 17 2023

jtl added a comment to D42664: tcphpts: enable per default.

I actually planned to make it a loadable module. Given that it creates extra threads that aren't going to be used by default makes its presence questionable.

Nov 17 2023, 11:14 PM

Nov 16 2023

jtl committed rG884eeff20ce9: genoffset.sh: fix build break on MacOS (authored by jtl).
genoffset.sh: fix build break on MacOS
Nov 16 2023, 5:57 PM
jtl closed D42407: Stop using a temporary file for genoffset.sh.
Nov 16 2023, 3:08 PM
jtl committed rGaccfb4cc9346: genoffset.sh: stop using a temporary file (authored by jtl).
genoffset.sh: stop using a temporary file
Nov 16 2023, 3:08 PM

Nov 7 2023

jtl added a comment to D42407: Stop using a temporary file for genoffset.sh.
In D42407#967685, @imp wrote:

I might also note in the commit message you used a here document with the shell output expanded in the commit message... it is a bit of a surprising change based on the commit message.

Nov 7 2023, 8:17 PM
jtl updated the summary of D42407: Stop using a temporary file for genoffset.sh.
Nov 7 2023, 8:13 PM

Oct 30 2023

jtl requested review of D42407: Stop using a temporary file for genoffset.sh.
Oct 30 2023, 5:56 PM

Sep 15 2023

jtl accepted D41877: nvme: Give up when we've failed.
Sep 15 2023, 4:56 PM
jtl added a reviewer for D41867: Deprecate fpu_kern_thread(): kib.
Sep 15 2023, 11:07 AM
jtl requested review of D41867: Deprecate fpu_kern_thread().
Sep 15 2023, 9:51 AM
jtl updated subscribers of D41865: Don't clear FPU registers when using FPU_KERN_NOCTX.
Sep 15 2023, 9:14 AM
jtl requested review of D41865: Don't clear FPU registers when using FPU_KERN_NOCTX.
Sep 15 2023, 8:43 AM

Jun 1 2023

jtl requested review of D40380: Modularize some IPv4/IPv6 protocol code.
Jun 1 2023, 7:06 PM

May 31 2023

jtl committed rG4f2cc73f34eb: tcp: Refactor tcp_get_srtt() (authored by jtl).
tcp: Refactor tcp_get_srtt()
May 31 2023, 7:24 PM
jtl closed D40304: Refactor tcp_get_srtt().
May 31 2023, 7:24 PM

May 27 2023

jtl added a comment to D40267: tcp: fix a bug where unshifting should be put last in tcp_get_srtt().

For what its worth...

May 27 2023, 3:30 PM
jtl requested review of D40304: Refactor tcp_get_srtt().
May 27 2023, 3:06 PM