1319673709 Q * dowdle Remote host closed the connection 1319675318 Q * ksn Ping timeout: 480 seconds 1319675598 Q * guerby Ping timeout: 480 seconds 1319678818 M * Bertl off to bed now ... have a good one everyone! 1319678825 N * Bertl Bertl_zZ 1319679383 Q * clopez Ping timeout: 480 seconds 1319680394 J * ksn ~ksn@197.168.189.9 1319680911 J * treaki_ ~treaki@p5B0331CE.dip.t-dialin.net 1319681223 Q * treaki Ping timeout: 480 seconds 1319684378 Q * hparker Ping timeout: 480 seconds 1319684920 J * hparker ~hparker@linux.homershut.net 1319685708 Q * eyck Ping timeout: 480 seconds 1319688294 Q * hparker synthon.oftc.net charm.oftc.net 1319688294 Q * ghislain synthon.oftc.net charm.oftc.net 1319688294 Q * kshannon synthon.oftc.net charm.oftc.net 1319688352 J * hparker ~hparker@linux.homershut.net 1319688352 J * kshannon ~kris@122.252.14.166 1319689601 J * sannes1 ~ace@cm-84.209.106.118.getinternet.no 1319690571 J * eyck ~eyck@77.79.198.68 1319691063 Q * eyck Ping timeout: 480 seconds 1319693493 Q * FireEgl Ping timeout: 480 seconds 1319693853 J * derjohn_mob ~aj@213.238.45.2 1319694109 J * FireEgl FireEgl@2001:470:e056:1:f885:13a6:cafc:86d1 1319694642 J * ncopa ~ncopa@3.203.202.84.customer.cdi.no 1319695519 J * eyck ~eyck@77.79.198.63 1319698272 J * guerby ~guerby@nc10d.tetaneutral.net 1319698397 J * renihs ~arf@83-65-34-34.arsenal.xdsl-line.inode.at 1319701615 J * ghislain ~AQUEOS@adsl2.aqueos.com 1319704254 J * bonbons ~bonbons@2001:960:7ab:0:5851:bc0:e077:d949 1319706941 Q * ksn Read error: Connection reset by peer 1319706958 J * ksn_ ~ksn@197.110.35.138 1319707089 J * clopez ~clopez@155.99.117.91.static.mundo-r.com 1319709011 Q * fisted Read error: Connection reset by peer 1319709097 J * fisted ~fisted@xdsl-87-78-210-20.netcologne.de 1319709602 Q * Hunger Ping timeout: 480 seconds 1319709773 Q * ksn_ Ping timeout: 480 seconds 1319711140 J * Hunger ~Hunger@Hunger.hu 1319712413 Q * hparker Ping timeout: 480 seconds 1319712489 J * hparker ~hparker@linux.homershut.net 1319713102 Q * hparker Quit: Quit 1319714418 Q * Aiken Remote host closed the connection 1319714836 N * Bertl_zZ Bertl 1319714884 M * Bertl morning folks! 1319715071 M * daniel_hozac morning Bertl 1319715220 M * Bertl hey, got a minute for me? 1319715289 M * daniel_hozac sure 1319715319 M * Bertl http://paste.linux-vserver.org/20644 1319715337 M * Bertl this is the final check regarding ptrace access 1319715371 M * Bertl I'm not sure what the purpose of the last check (regarding VXF_STATE_ADMIN) really is about 1319715379 M * Bertl -about 1319715416 M * daniel_hozac isn't that to prevent ptrace on processes in guests without VXF_STATE_ADMIN? 1319715429 M * daniel_hozac i.e. unadministrable guests 1319715460 M * Bertl ah, good point 1319715545 M * Bertl then I probably have to look elsewhere ... the original issue is that the spectator context cannot access /proc//io because we get -EACCES there 1319715590 M * daniel_hozac hmm 1319715593 M * Bertl from a quick glance at the code, the only relevant place was this check, but it cannot possibly trigger on a default system 1319715609 M * daniel_hozac can you reproduce it? 1319715645 M * Bertl chcontext --xid 1 -- cat /proc/1/io 1319715820 M * Bertl does chcontext drop capabilities on xid=1? 1319715939 M * daniel_hozac no, it shouldn't... 1319716031 M * daniel_hozac it looks like any error code from ptrace_may_access will return EACCES though. 1319716079 M * Bertl yep, just saw that, so probably something like the namespace causes this 1319716102 M * Bertl which would mean that we either miss CAP_SYS_PTRACE 1319716132 M * Bertl or the task isn't dumpable then? 1319716158 M * Bertl anyway, will add some debug info to that check and see if something triggers 1319716179 J * Crumbz ~Crumbz@host-89-240-244-133.as13285.net 1319716197 M * daniel_hozac hmm 1319716222 M * daniel_hozac task_ns_capable and the ns_capable do the same thing. 1319716302 M * daniel_hozac the vs_check_bit in ns_capable looks wrong to me. 1319716329 M * daniel_hozac although not related to this. 1319716425 M * Bertl how so? (I'm referring to the wrong part)? 1319716454 M * daniel_hozac it circumvents the actual capability check 1319716462 M * daniel_hozac shouldn't that be a return false? 1319716598 M * Bertl IIRC, we added the VXC_CAP_MASK for exactly that purpose, but it shouldn't be used now/anymore as it is 0 1319716663 M * Bertl so vs_check_bit(VXC_CAP_MASK, cap) should always give false 1319716673 M * daniel_hozac fair enough 1319716698 M * Bertl but I'd opt to remove that contruct as we obviously do not need it 1319716720 M * Bertl (and it just leads to confusion :) 1319716727 M * daniel_hozac yeah : 1319717000 J * ksn ~ksn@197.110.35.138 1319717035 M * daniel_hozac i guess we just need some printk's to see which check is failing. 1319718184 J * thierryp ~thierry@zankai.inria.fr 1319721136 J * hparker ~hparker@2001:470:1f0f:32c:beae:c5ff:fe01:b647 1319721964 Q * treaki_ Ping timeout: 480 seconds 1319723613 Q * ksn Ping timeout: 480 seconds 1319724570 M * Bertl it seems, we actually hit the ident/state_admin check 1319724591 M * Bertl will output a little more details on the flags/check results 1319725293 M * Bertl so, the problem is this, when xid=1 tries to inspect a host process, we end up with both checks returning true and block the access 1319725327 M * Bertl I'm now going to check the case where we try to inspect a guest process 1319725379 M * Bertl the question is if we actually want to change that behaviour, because host processes can be inspected from the host context 1319725536 M * Bertl ser: are you sure you can only see I/O for htop not for guest processes? 1319725552 M * Bertl ser: it is expected that you cannot see I/O for host processes 1319725593 Q * FireEgl Ping timeout: 480 seconds 1319725899 Q * micah Quit: Lost terminal 1319725915 M * Bertl daniel_hozac: 'vurpm urpmi mag1 -- screen' 1319725924 M * Bertl ncontext: execvp(""): Permission denied 1319725932 M * Bertl what am I doing wrong here? 1319726141 J * micah ~micah@micah.riseup.net 1319726759 J * dowdle ~dowdle@scott.coe.montana.edu 1319726822 M * Crumbz Bertl, you seem to be the only person who ever answers question on this server. :P 1319726831 M * Crumbz channel* 1319726840 M * Bertl nah, that is a misconception 1319726873 M * Crumbz Or just a coincidence? ;p 1319727125 Q * Crumbz Quit: Leaving 1319728040 Q * ncopa Quit: Leaving 1319728205 N * [Guy] Guy- 1319729697 Q * nkukard Ping timeout: 480 seconds 1319730439 J * nkukard ~nkukard@41-133-198-167.dsl.mweb.co.za 1319731514 Q * derjohn_mob Ping timeout: 480 seconds 1319732098 Q * nkukard Ping timeout: 480 seconds 1319732347 J * sweil ~stefan@p5086FEB8.dip.t-dialin.net 1319732889 J * nkukard ~nkukard@41-133-198-167.dsl.mweb.co.za 1319733328 J * FireEgl FireEgl@2001:470:e056:1:7052:b613:6257:460f 1319733489 J * treaki_ ~treaki@31-19-232-27-dynip.superkabel.de 1319733517 Q * thierryp Remote host closed the connection 1319733886 M * daniel_hozac Bertl: hmm, it looks fine. i'll have to look at it later. 1319734193 Q * FireEgl Ping timeout: 480 seconds 1319734779 J * FireEgl ~FireEgl@173-16-9-169.client.mchsi.com 1319734937 Q * treaki_ Ping timeout: 480 seconds 1319735953 Q * chrissbx Ping timeout: 480 seconds 1319737178 Q * FireEgl Ping timeout: 480 seconds 1319737811 J * FireEgl ~FireEgl@173-16-9-169.client.mchsi.com 1319738768 Q * renihs Ping timeout: 480 seconds 1319739327 Q * mike Remote host closed the connection 1319739565 J * mike mike@no.phear.eu 1319739922 Q * nicholi Quit: leaving 1319739955 J * nicholi ~nicholi@rrcs-76-79-196-34.west.biz.rr.com 1319741593 M * ser Bertl: I cannot see any of guests IO 1319741593 Q * FireEgl Read error: Connection reset by peer 1319741603 M * ser Bertl: the aonly IO I can is htop itself 1319741696 M * ser which is on the host, btw 1319742038 M * Bertl okay, but you can enter/administrate the guests? 1319742203 M * ser sure i can 1319742219 M * ser iotop also works 1319742311 M * ser which saves my ass for now, as i can use htop and iotop to achieve mostly the same results as htop alone before on 2.6.36 1319742416 M * ser i think your investigation is right, it looks it just cannot read properly values from /proc// - these entries were on my strace 1319742443 Q * hparker Ping timeout: 480 seconds 1319742450 M * ser it works in guests, it works on host, it does not work in context 1 1319742929 J * matti matti@acrux.romke.net 1319742944 P * matti 8-X 1319743210 J * hparker ~hparker@2001:470:1f0f:32c:beae:c5ff:fe01:b647 1319743846 J * Aiken ~Aiken@2001:44b8:2168:1000:21f:d0ff:fed6:d63f 1319744716 Q * Alex[fob] Ping timeout: 480 seconds 1319745342 J * Alex[fob] ~alex@2001:6f8:1c3c:f76::152:1 1319745872 Q * sannes1 Remote host closed the connection 1319747568 J * FireEgl ~FireEgl@173-16-9-169.client.mchsi.com 1319747782 Q * clopez Ping timeout: 480 seconds 1319749193 Q * FireEgl Ping timeout: 480 seconds 1319749648 Q * sweil Remote host closed the connection 1319749804 J * FireEgl ~FireEgl@173-16-9-169.client.mchsi.com 1319750370 Q * FireEgl Remote host closed the connection 1319751239 J * FireEgl FireEgl@2001:470:e056:1:c027:4e85:ce0c:23a6 1319751769 Q * FireEgl Remote host closed the connection 1319752469 M * daniel_hozac Bertl: vurpm should be fixed now. 1319752488 M * daniel_hozac Bertl: do you have some debugging output already from ptrace_may_access? 1319752660 M * daniel_hozac i see how that would trigger for host processes 1319752666 M * daniel_hozac but guests? 1319752788 M * daniel_hozac well, hmm, i can see the guest's io file fine... 1319753721 M * ser daniel_hozac: apt-get install htop; chcontext --xid 1 -- htop 1319753770 J * FireEgl FireEgl@2001:470:e056:1:7d6f:4d19:5b4d:305c 1319754527 M * Bertl yeah, no idea why htop fails 1319754547 M * Bertl but I can see that htop has problems with guest files as well 1319754625 M * Bertl regarding vurpm: can I get a tar (pretty please :)? 1319754848 M * ser Bertl: what do you mean by guests? i run it in a guest and it shows all IO properly 1319754899 J * treaki_ ~treaki@31-19-232-27-dynip.superkabel.de 1319754994 M * Bertl yeah, I can reproduce that 1319755012 M * Bertl I just haven't figured out what keeps htop from doing the same in xid 1 1319755035 M * Bertl because with e.g. 'cat' I can view the I/O stats quite fine from xid=1 1319755045 M * ser OK, sorry, I was sure there is some misunderstanding :) 1319755057 M * ser yes, iotop works well either 1319755064 M * ser it is magic :) 1319755080 M * Bertl so either something in htop is going wrong, or htop is kind of dropping permissions early or something completely different 1319755096 M * ser I can consult htop list if you suggest to 1319755119 M * Bertl yeah, it would definitely bring some outside ideas ... 1319755134 M * ser sorry for my screaming, it is just the basic app in my everyday work :) 1319755137 M * daniel_hozac do we get any errors from htop? 1319755147 M * daniel_hozac (with strace, i mean) 1319755156 M * ser i can upload a strace for you 1319755165 M * Bertl strace shows a bunch of access denied, but not on the pids in question 1319755204 M * daniel_hozac Bertl: new tar uploaded :) 1319755212 M * ser daniel_hozac: if you want me of course to do that :) 1319755222 M * Bertl daniel_hozac: thanks a bunch! \o/ 1319755400 M * daniel_hozac the error is that it shows zeroes? 1319755406 M * ser yes 1319755406 M * Bertl yup 1319755420 M * Bertl but the io stats update when using cat for example 1319755440 M * Bertl easy to test with chcontext --xid 42 -- dd if=/dev/zero of=/dev/null bs=1k & 1319755462 M * daniel_hozac hmm 1319755468 M * daniel_hozac i see numbers. 1319755476 M * Bertl hmm, kernel? 1319755493 M * daniel_hozac 3.0.4-vs2.3.1-pre10.1 1319755494 M * Bertl and we are talking about htop running in xid=1 1319755498 M * daniel_hozac right 1319755507 M * Bertl interesting 1319755510 M * daniel_hozac vcontext --xid 1 --migrate -- htop in one window 1319755519 M * daniel_hozac chcontext --xid 42 -- dd if=/dev/zero of=/tmp/blah bs=1M count=3000 in another. 1319755543 Q * fisted Ping timeout: 480 seconds 1319755545 M * ser for me it does not show numbers 1319755554 M * ser 3.0.7-vs2.3.1-beng 1319755575 M * Bertl for me neither, tested with 3.0.4-vs2.3.1-pre10.1 here 1319755588 M * daniel_hozac which fields are you monitoring? 1319755602 M * ser IOWR, IO_WR 1319755619 M * Bertl the 5 IO stat one I could find 1319755627 J * fisted ~fisted@xdsl-87-78-208-43.netcologne.de 1319755650 M * Bertl i.e. IO_R/WBYTES, and IO,IORR,IORW 1319755699 M * daniel_hozac /dev/null doesn't work for me. 1319755710 M * daniel_hozac but /tmp/blah works fine. 1319755713 M * Bertl interesting ... 1319755754 M * ser i am testing on real services, they are shown when i enter a particular guest and run htop 1319755795 M * ser is it possible you have different setattr? 1319755824 M * daniel_hozac for /proc? i have stock everything . 1319755861 M * daniel_hozac setattr doesn't apply to per-process entries anyway. 1319755866 M * Bertl I'm inclined to consider this some kind of htop sanity check going wrong maybe? 1319755887 M * ser Bertl: but why daniel's htop works? 1319755909 M * Bertl haven't tested the 'bla' part yet 1319756040 M * ser i have tested :( 1319756174 M * Bertl adaniel_hozac: vurpm urpmi mag1 -- screen \n secure-mount: chdir("/.rpmdb"): No such file or directory 1319756208 M * daniel_hozac hmm 1319756209 M * Bertl I presume I have to create something somewhere or somehow tell it that the database is inside? 1319756232 M * daniel_hozac none of that should be needed for internal databases, i don't know. 1319756330 M * Bertl hmm, maybe the guest got messed up, no problem I try with a new one 1319756395 M * daniel_hozac try just making sure its set as internal 1319756465 M * daniel_hozac i.e. touch /etc/vservers//apps/pkgmgmt/internal 1319756670 Q * ghislain Quit: Leaving. 1319756862 J * andip ~packet@67.213.81.183 1319757027 M * daniel_hozac ser: if you could upload an strace of an htop session along with some pid that is supposed to be doing IO, that would probably be helpful. 1319757042 M * daniel_hozac but i need to get some sleep for now. 1319757049 Q * andip 1319757063 M * daniel_hozac good night! 1319757213 M * Bertl have a good one! 1319757360 M * ser daniel_hozac: i will try, unfortunately next week only, i am due to leave for a conference. but i will do it 1319757395 M * Bertl no hurries, we can partially recreate and reproduce the issue (which is a good thing :) 1319757417 M * ser Bertl, daniel_hozac: thanks a lot, guys 1319757426 M * Bertl but it won't hurt to drop an email to the htop folks, maybe they have a good idea 1319757438 M * ser i will do it now 1319757498 M * Bertl thanks! 1319757529 M * ser i am reading htop archives to find out who was an author of vserver patch 1319757548 M * ser and i will contact him directly to achieve beter results :) 1319757589 M * ser it was Mr Jonathan Sambrook 1319757609 M * Bertl yeah, but I don't think it is related to that patch 1319757625 M * Bertl (which adds support for the xid/nid fields) 1319757635 M * ser but at least he will know what we are chatting about 1319757644 M * Bertl yep, definitely 1319757645 M * ser the rest of htop list does not use vserver 1319757888 Q * treaki_ Ping timeout: 480 seconds 1319757898 J * treaki_ ~treaki@hd679.rmz.uni-lueneburg.de 1319758154 M * ser Ok, I have written him, thanks 1319758166 M * Bertl thank you! 1319758643 Q * bonbons Quit: Leaving 1319759899 Q * treaki_ Ping timeout: 480 seconds