1527641241 N * obeardly obeardly_zZ 1527642199 J * guerby_ ~guerby@ip165.tetaneutral.net 1527642215 Q * guerby Read error: Connection reset by peer 1527655265 J * _pa ~pav@15-46.dsl.iskon.hr 1527656943 Q * clopez Ping timeout: 480 seconds 1527657015 J * clopez ~tau@neutrino.es 1527659887 Q * _pa Quit: Leaving 1527661580 J * guerby__ ~guerby@ip165.tetaneutral.net 1527661623 J * _pa ~pav@15-46.dsl.iskon.hr 1527661752 J * clopez_ ~tau@neutrino.es 1527661865 Q * clopez magnet.oftc.net kinetic.oftc.net 1527661865 Q * guerby_ magnet.oftc.net kinetic.oftc.net 1527661865 Q * CcxWrk magnet.oftc.net kinetic.oftc.net 1527662148 J * CcxWrk ~ccx@asterix.te2000.cz 1527664113 Q * _pa Quit: Leaving 1527665215 M * Guy- obeardly_zZ: I think Carpoon may have 1527665622 M * Ghislain *hi, i just runned a stress-ng on a stretch guest on 4.9.103 and the guest crashed 1527665632 M * Ghislain stress-ng: error: [30108] stress-ng-userfaultfd: userfaultfd failed, errno = 38 (Function not implemented) 1527665632 M * Ghislain => guest down 1527665878 M * Ghislain https://pastebin.com/raw/vsuJvG32 1527665940 M * Guy- Ghislain: I think an strace log would be useful, to see what function returned errno 38 1527665991 N * Bertl_zZ Bertl 1527665993 M * Bertl morning folks! 1527666084 M * Ghislain well strace do not gives a lot of info 1527666090 M * Ghislain hi bertl 1527666467 M * Guy- Ghislain: surely there must be a syscall that fails due to errno 38 1527666470 J * nikolay ~nikolay@external.oldum.net 1527666540 M * Ghislain trying on the host it does not crash 1527666553 M * Ghislain well perhaps it takes longer 1527666595 M * Ghislain https://pastebin.com/raw/CVQ4CCUF 1527666608 M * Guy- Ghislain: I meant, make it crash while stracing it -- so, presumably, run it in a guest 1527666634 M * Ghislain strace gives nothign i tried it but it show nothing for 10s then crash so the crash is not shown 1527666641 M * Guy- ah 1527666642 M * Ghislain i mean 1527666644 M * Guy- it's this syscall: http://man7.org/linux/man-pages/man2/userfaultfd.2.html 1527666649 M * Guy- says so right in the log 1527666657 M * Ghislain it shwo all stuff then stop, wait 10s+ then crash 1527666679 M * Ghislain but it hsould crash the process not the guest 1527666695 M * Guy- are there other processes in the guest? 1527666736 M * Ghislain oh yes plenty this is a lamp terst guest so you go apache php haproxy spamd dovecot etc. 1527666813 M * Ghislain perhaps it hits a specter protection 1527666937 M * Guy- Ghislain: so you could strace one of the *other* processes when the guest crashes, and see what happens to it 1527667029 M * Ghislain it just stops, strace say nothing (tried on apache) 1527667161 M * Ghislain i will try runnning them one by one to see which test crash the guest 1527669065 M * Ghislain there is so many tests..running 20s each one by one takes....time :) 1527670114 M * Ghislain well one usefaultd test did not crash it 1527670349 M * Bertl userfaultfs is pretty new and 4.11 changed a lot there 1527670388 M * Bertl so that could be some kind of lock/semaphore which is not properly released or blocks 1527670524 M * Ghislain oh ok, hum now i seems to have crashed the host from inside the guest completly :( 1527670600 M * Ghislain oh! it came back 1527670641 M * Ghislain got the same fault error on userfaultd but the guest still runs this time 1527670667 M * Ghislain so the solution is 4.14 then ;p 1527671695 M * Ghislain oh, the dentry number is still going wild :) 1527671709 M * Ghislain fun fact 1527671747 M * Ghislain if i launch the stress in a bash spawned process and not in the vserver enter it crash the bash but not the guest 1527671776 M * Bertl nice, so it is process related 1527671788 M * Ghislain let me try that again 1527671791 M * Bertl please collect all the 'crash' dumps 1527671813 M * Bertl and make sure to have a kernel which allows converting addresses to line numbers :) 1527671954 M * Ghislain so: when i launch it in a root process it crash the guest 1527671967 M * Ghislain when i launch as "nobody" it crash the shell 1527672082 M * Ghislain yes 3 times in a row i confirm. run as rott=> crash the guest, run as nobody => crash the shell 1527672439 M * Ghislain i must recompile the kernel i guess 1527672593 M * Ghislain must i put the 2 delte netlink on it or just the latest ? seems the latest only 1527673084 M * Ghislain anyone for a 4.9.104 compilation with debug symbols ? :D 1527673165 M * Bertl yeah, only the latest 1527673185 M * AlexanderS obeardly_zZ: I have a gitlab inside a vserver guest. But I installed it from source, not using the "omnibus" package. 1527673500 M * Ghislain Bertl: running the stress-ng in the host do not crash it or bash 1527673853 M * Ghislain so it seems to be a vserver only crash. I am compiling 4.9.104 with debug to test it 1527674197 M * Bertl excellent! tx! 1527684292 M * Bertl off for now ... bbl 1527684293 N * Bertl Bertl_oO 1527684523 M * Ghislain hum 4.9.104 did not crashed on stress-ng (unr twice allready) 1527684529 M * Ghislain s/urn/run/ 1527684593 M * Ghislain OH! third time was the good one 1527684597 M * Ghislain just crashed all 1527686763 Q * AlexanderS Ping timeout: 480 seconds 1527687337 J * AlexanderS ~Alexander@home.zedat.fu-berlin.de 1527690437 M * Ghislain ouch 1527690452 M * Ghislain if i ssh into the guest directly 1527690458 M * Ghislain run stress-ng 1527690467 M * Ghislain it crash and i become root on the host 1527691053 M * Ghislain it does not create any crash dump i could see 1527691514 Q * Aiken Remote host closed the connection 1527692093 M * Ghislain 4.9.92 do not seems affected, perhaps you should try the kernel you use 1527692150 M * Ghislain of course the 4.9.104 use a non clean patch that have 2 rejects, one is corrected manueally and the ext2 one i did not touched it and ignored as i dont use the ext2 code 1527693027 M * obeardly_zZ Thanks 1527693034 N * obeardly_zZ obeardly 1527698098 Q * ensc|w Ping timeout: 480 seconds 1527699766 Q * nikolay Remote host closed the connection 1527705272 J * hijacker ~hijacker@external.oldum.net 1527707349 M * obeardly AlexanderS: You still here? 1527707580 Q * hijacker Quit: Leaving 1527707616 M * AlexanderS Yes 1527709540 M * obeardly Did you have any issues with redis? 1527710391 M * obeardly I actually ran into a problem, but it ended up being a corrupt pkg cache. 1527710399 M * obeardly I figured it out finally. 1527712704 J * Aiken ~Aiken@2001:44b8:2168:1000:b26e:bfff:fe2a:b951 1527719099 M * Bertl_oO off to bed now ... have a good one everyone! 1527719101 N * Bertl_oO Bertl_zZ