[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 Eitan Adler changed: What|Removed |Added Status|In Progress |Closed

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #93 from m...@sentex.net --- All working great since this commit for this particular bug -- You are receiving this mail because: You are the assignee for the bug. ___

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 Ed Maste changed: What|Removed |Added Status|New |In Progress --

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #92 from commit-h...@freebsd.org --- A commit references this bug: Author: kib Date: Wed Feb 14 00:31:45 UTC 2018 New revision: 329254 URL: https://svnweb.freebsd.org/changeset/base/329254 Log: Ensure memory consistency on

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #91 from m...@sentex.net --- I just confirmed this bug happens on a my Epyc system as well. The compile hung on the first try. A reboot with the patch applied in https://reviews.freebsd.org/D14347 fixed the issue.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #90 from m...@sentex.net --- (In reply to Kurt Jaeger from comment #88) The deadlock I was running into with net/samba47 builds would happen with and without SMT enabled. With the patch at https://reviews.freebsd.org/D14347 I

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #89 from Ed Maste --- (In reply to Kurt Jaeger from comment #88) I'm running with SMT disabled in BIOS still. I configured it like this while testing/debugging this issue. I will make sure that I have the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #88 from Kurt Jaeger --- Can all cores be used to build packages or does it require BIOS settings to not start the hyperthreading cores ? Any significant performance impact ? -- You are receiving this mail

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #87 from Ed Maste --- (In reply to mike from comment #86) The freebsd-hackers patch is now also available in Phabricator at https://reviews.freebsd.org/D14347 With the D14347 patch applied to the -CURRENT

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #86 from m...@sentex.net --- The patch in https://docs.freebsd.org/cgi/getmsg.cgi?fetch=417183+0+archive/2018/freebsd-hackers/20180211.freebsd-hackers seems to have fixed this issue for me in prelim testing. On a RELENG 11

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #85 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #84) (gdb) thread 4 [Switching to thread 4 (LWP 101503 of process 43500)] #0 _umtx_op_err () at /usr/src/lib/libthr/arch/amd64/amd64/_umtx_op_err.S:37

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #84 from Konstantin Belousov --- (In reply to mike from comment #83) It is yet another (different) lock leakage. Please switch to the Thread 4 (LWP 101503) and from the frame 1 do "print *mtx". -- You are

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #83 from m...@sentex.net --- Created attachment 190490 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190490=edit process that is hung in usem state with libc path and python linked to debug libs -- You are

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #82 from Konstantin Belousov --- (In reply to mike from comment #81) I can't say anything without debug info. Start with the backtraces, as usual. -- You are receiving this mail because: You are the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #81 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #79) > BTW, if using the patch from the comment 59, is the hang reproducible ? I got it to hang on RELENG_11 just now. Perhaps its some combo of the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #80 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #79) re: patch 59 I will apply it by itself and try again patch -p1 < p.59 Hmm... Looks like a unified diff to me... The text leading up to this was:

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #79 from Konstantin Belousov --- BTW, if using the patch from the comment 59, is the hang reproducible ? -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #78 from Konstantin Belousov --- (In reply to Ed Maste from comment #77) Ok. It could be a consequence of the discovered issue occuring in other code fragment, it can be something else. So I can wait for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #77 from Ed Maste --- (In reply to Ed Maste from comment #75) I spoke too soon; I encountered a stuck build on my threadripper in the configuration described in comment 75. -- You are receiving this mail

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #76 from m...@sentex.net --- (In reply to Ed Maste from comment #75) Stalled out again. It managed 15 builds. My RELENG_11 box did as well after 23 builds root@amdtestr12:/usr/ports/sysutils/dmidecode # procstat -k 34990

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #75 from Ed Maste --- (In reply to Konstantin Belousov from comment #69) I'm testing net/samba47 builds on my threadripper with a libc including the patch from comment #69 applied to my work tree and it has

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #74 from Konstantin Belousov --- (In reply to mike from comment #72) I admit that it is strange that it hangs. It is quite possible that mfence trick dos not work, but then the 'else' part of the patch must

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #73 from m...@sentex.net --- (In reply to mike from comment #72) To be safe, I am going to do a complete buildworld/kernel and repeat the tests -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #72 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #71) Unfortunately, its back to the original behaviour where it just hangs in the usem state -- You are receiving this mail because: You are the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #71 from Konstantin Belousov --- (In reply to mike from comment #70) Yes, just that libc patch. -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #70 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #69) OK, just this patch right ? Remove the kernel patches ? -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #69 from Konstantin Belousov --- (In reply to mike from comment #68) Please try this, it should work and then it would be the last thing I am asking to test. diff --git a/lib/libc/stdio/_flock_stub.c

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #68 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #67) yes and yes -- You are receiving this mail because: You are the assignee for the bug. ___

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #66 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #65) OK, I am still rebuilding world. While it rebuilds, I tried applying this patch to a stock RELENG_11 box, and it didnt seem to help. Should I try

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #65 from Konstantin Belousov --- (In reply to mike from comment #64) Yes, libc must be pristine, otherwise it already hides the bug. -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #64 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #63) Thanks! I am running with a new kernel. Should I run python linked to the unmodified libs ? -- You are receiving this mail because: You are the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #63 from Konstantin Belousov --- (In reply to Konstantin Belousov from comment #62) Try this. diff --git a/sys/amd64/amd64/cpu_switch.S b/sys/amd64/amd64/cpu_switch.S index 25e3a592171..ef70eefdf37 100644 ---

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #62 from Konstantin Belousov --- (In reply to mike from comment #61) Do not consider it a fix. I do not have any explanation what is going on there. -- You are receiving this mail because: You are the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #61 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #59) Its managed 22 builds now without issue. Prior, I could only do 10 at most. I have a separate RELENG_11 box as well. I take it the same patch

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #60 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #59) I have to confess, I might have borked the previous patch and left the while loop commented out :( I have made sure its right this time. Five

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #59 from Konstantin Belousov --- (In reply to mike from comment #58) Lets try a slight variation. diff --git a/lib/libc/stdio/_flock_stub.c b/lib/libc/stdio/_flock_stub.c index 069ed64c4e9..935bf18b529 100644

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #58 from m...@sentex.net --- Created attachment 190431 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190431=edit with while loop patch funlockfile 0x80146caa0 0x80c83b800 0x0 Bus error (core dumped) -- You are

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #57 from Konstantin Belousov --- (In reply to mike from comment #56) Yes, that was already handled. And, the hang there is the whole system hang. Please apply this debugging patch and try reproduce the issue

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #56 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #55) The DragonFly commit mentioned here http://gitweb.dragonflybsd.org/dragonfly.git/commitdiff/b48dd28447fc8ef62fbc963accd301557fd9ac20 But I FreeBSD

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #55 from Konstantin Belousov --- (In reply to mike from comment #53) The article you linked does not provide any useful information. Nobody knows why this knob helped. That said, I am both disappointed and

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #54 from m...@sentex.net --- Created attachment 190430 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190430=edit latest backtrace post kernel patch -- You are receiving this mail because: You are the assignee for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #53 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #52) No luck. It survived a lot of builds, but still dumped core funlockfile 0x80146ce48 0x80bcd1500 0x0 Bus error (core dumped) BTW, DragonFly and

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #52 from Konstantin Belousov --- Created attachment 190428 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190428=edit Disable RDFSBASE on zens. This is a pure guess based on the _impression_ I get

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #51 from Konstantin Belousov --- (In reply to mike from comment #50) It is not the abort line that was printed for the core analyzed. I expect that the first address printed is 0x80146cd10. -- You are

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #50 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #48) forgot to paste funlockfile 0x80146caa0 0x80d75ff00 0x0 Bus error (core dumped) -- You are receiving this mail because: You are the assignee for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #49 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #48) (gdb) thread 1 [Switching to thread 1 (LWP 101411)] #0 thr_kill () at thr_kill.S:3 3 RSYSCALL(thr_kill) (gdb) list 1 #include

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #48 from Konstantin Belousov --- (In reply to mike from comment #46) What was printed on crash ? >From thread 1 frame 3 please do "p *fp", "p *(fp->_fl_mutex)". -- You are receiving this mail because: You

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #47 from Ed Maste --- Do you have the message it printed before the SIGBUS as well? (If you run it in tmux I'll attach and take a look.) -- You are receiving this mail because: You are the assignee for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #46 from m...@sentex.net --- Created attachment 190408 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190408=edit thread apply all bt on the core dump file post debug patch sig 10 Path worked! -- You are receiving

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #45 from Konstantin Belousov --- (In reply to mike from comment #44) Use this instead. diff --git a/lib/libc/stdio/_flock_stub.c b/lib/libc/stdio/_flock_stub.c index 069ed64c4e9..0180f8f285f 100644 ---

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #44 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #43) its unlimited, the defaults. I changed the core pattern and tested it with another shell script. sysctl -w sysctl

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #43 from Konstantin Belousov --- (In reply to mike from comment #42) Check ulimit. -- You are receiving this mail because: You are the assignee for the bug. ___

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #42 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #41) It doesnt seem to be dumping core for some reason ? root@amdtestr12:/ # cd / root@amdtestr12:/ # find . -mtime -2h -type f -name "*.core"

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #41 from Konstantin Belousov --- (In reply to mike from comment #40) There should be a core file from the python process somewhere. Please load it into gdb and gather the same information (at least backtrace

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #40 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #38) python2.7(pid 71466 uid 0) aborted: funlockfile(0x80146caa0, 0x80a718600, 0) -- You are receiving this mail because: You are the assignee for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #39 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #38) # patch -p1 < p Hmm... Looks like a unified diff to me... The text leading up to this was: -- |---

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #38 from Konstantin Belousov --- (In reply to mike from comment #37) Try to apply the following debugging patch: diff --git a/lib/libc/stdio/_flock_stub.c b/lib/libc/stdio/_flock_stub.c index

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #37 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #36) (gdb) print local_close $5 = (int (*)(FILE *)) 0x801237880 (gdb) list 452Otherwise this close() will crash when flushing

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #36 from Konstantin Belousov --- (In reply to mike from comment #35) Thank you. And please, from the same frame 5, do "print local_close". -- You are receiving this mail because: You are the assignee for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #35 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #33) frame 5 has that var ? (gdb) thread 9 [Switching to thread 9 (LWP 101309 of process 48133)] #5 0x000800a82fb7 in close_the_file

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #34 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #33) gdb attach 48133 GNU gdb (GDB) 8.0.1 [GDB v8.0.1 for FreeBSD] Copyright (C) 2017 Free Software Foundation, Inc. License GPLv3+: GNU GPL version 3 or

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #33 from Konstantin Belousov --- (In reply to mike from comment #32) Perhaps you did not switched to the thread 9 before select frame ? -- You are receiving this mail because: You are the assignee for the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #32 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #31) Its not there ? (gdb) frame 7 #7 0x000800b55b5b in call_function (pp_stack=0x7fffde3f3148, oparg=0) at Python/ceval.c:4341 4341

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #31 from Konstantin Belousov --- (In reply to mike from comment #30) Not useful, indeed. Please from the same thread 9, change to the frame 7 and do "print local_fp", "print *local_fp". -- You are receiving

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #30 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #29) (gdb) thread 9 [Switching to thread 9 (LWP 101309 of process 48133)] #0 _umtx_op () at _umtx_op.S:3 3 RSYSCALL(_umtx_op) (gdb) frame 6 #6

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #29 from Konstantin Belousov --- (In reply to mike from comment #28) Now please switch to the thread 9, and from the frame 6, do "print *f" and "print *(f->f_fp)". -- You are receiving this mail because: You

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #28 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #27) (gdb) thread 2 [Switching to thread 2 (LWP 101302 of process 48133)] #0 _umtx_op_err () at /usr/src/lib/libthr/arch/amd64/amd64/_umtx_op_err.S:37

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #27 from Konstantin Belousov --- (In reply to mike from comment #26) Please switch to the Thread 2 (LWP 101302) and do: - on frame 2, "print *curthread" and "print *m" - on frame 4, "print *mutex" - on frame

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #26 from m...@sentex.net --- Created attachment 190385 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190385=edit thread apply all bt info thread at the end -- You are receiving this mail because: You are the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #25 from Konstantin Belousov --- It seems that gdb can do 'thread apply all bt'. At least my copy of gdb 8.1 can. -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #24 from Mark Millard --- (In reply to mike from comment #23) Try: (gdb) thread 11 [Switching to thread 11 (LWP 101275 of process 38415)] #0 _umtx_op_err () at

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #23 from m...@sentex.net --- (In reply to Mark Millard from comment #22) Hi, I installed gdb from the ports as it doesnt seem to be installed by default in HEAD anymore ? bt thread 11 doesnt seem to work in 8.x (gdb) bt thread

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #22 from Mark Millard --- (In reply to mike from comment #21) It looks like after: (gdb) thread 11 [Switching to thread 11 (LWP 101275 of process 38415)] #0 _umtx_op_err () at

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #21 from m...@sentex.net --- Created attachment 190376 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190376=edit bt and listing of threads Note, this box is attached off zoo if you want to log in and look around

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #20 from m...@sentex.net --- (In reply to mike from comment #19) It took a few more runs, but after 10 builds, the build hung as usual. attaching the output -- You are receiving this mail because: You are the assignee for the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #19 from m...@sentex.net --- (In reply to Konstantin Belousov from comment #18) Interesting. With -g -O0, I dont get the build hangs. So far in a loop, the port has built 9 times. Normally, it hangs 100% of the time Could be

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #18 from Konstantin Belousov --- (In reply to mike from comment #17) Rebuild libthr with flags "-g -O0" and reinstall it. I need: backtrace of all threads in the process. Plain "bt" for each thread is enough,

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #17 from m...@sentex.net --- Created attachment 190365 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=190365=edit gdb output of stuck python process -- You are receiving this mail because: You are the assignee for

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #16 from m...@sentex.net --- (In reply to Eugene Grosbein from comment #15) I built world with /etc/make.conf KERNCONF=DEBUG DEBUG_FLAGS=-g WITH_DEBUG=YES and then rebuilt python. I am not sure what needs to be done in the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #15 from Eugene Grosbein --- (In reply to Konstantin Belousov from comment #14) Thanks. I personally do not have such problem nor AMD hardware to reproduce that but hope Mike will benefit from this hint. --

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #14 from Konstantin Belousov --- (In reply to Eugene Grosbein from comment #12) Compile everything with debugging symbols, i.e. rtld/libc/libthr, and the application (python interpreter). Than take the

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #13 from m...@sentex.net --- (In reply to Eugene Grosbein from comment #11) Re: timing issues, I did try it on a number of different Intel CPUs ( eg Xeon(R) CPU E5-1650) of various core densities and speeds and was not able to

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 Eugene Grosbein changed: What|Removed |Added CC|

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #11 from Eugene Grosbein --- (In reply to mike from comment #9) These threads locked in "umtxn" state remind me of https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=186114 that revealed long-standing bugs in

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #10 from Ed Maste --- (In reply to mike from comment #9) > Hi, apart from > KDB_UNATTENDED > these options are all part of a standard GENERIC kernel on HEAD, no ? They are, yes. I suspect we're going to

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #9 from m...@sentex.net --- (In reply to Eugene Grosbein from comment #7) Hi, apart from KDB_UNATTENDED these options are all part of a standard GENERIC kernel on HEAD, no ? I added that as well as BREAK_TO_DEBUGGER. The kernel

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #8 from Mark Millard --- I got a different example: Threadripper 1950X context, head -r327485 doing a adm64 -> aarch64 cross build of ports: hung up in uwait. The context is also: Running under

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #7 from Eugene Grosbein --- Could you please re-do your test using kernel with following options added? Be ready for panic, though and prepare to collect crashdump: options KDB options

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #6 from Andriy Gapon --- (In reply to Ed Maste from comment #3) What's nice about this problem is it is debuggable. The build hangs, but the system doesn't. So, it should be possible to examine the system

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 Mark Millard changed: What|Removed |Added CC|

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #4 from m...@sentex.net --- I tried pretty well every combo on my Ryzen board as well-- SMT on / off, No HT, cores only, etc with no change. -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #3 from Ed Maste --- I have reproduced this on my Threadripper, both with and without SMT disabled in BIOS. -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #2 from m...@sentex.net --- Tried with # sysctl -a kern.sched.name kern.sched.name: 4BSD but not obvious difference. -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 --- Comment #1 from m...@sentex.net --- Tried also with kern.eventtimer.periodic=1 but it made no difference that I could tell. Process hung as usual -- You are receiving this mail because: You are the assignee for the bug.

[Bug 225584] Various compile process hang on Ryzen, but not on Intel

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=225584 Bug ID: 225584 Summary: Various compile process hang on Ryzen, but not on Intel Product: Base System Version: CURRENT Hardware: amd64 OS: Any