Back to Home / #uml / 2007 / 11 / Prev Day | Next Day
#uml IRC Logs for 2007-11-09

---Logopened Fri Nov 09 00:00:56 2007
00:46|-|dsoul [darksoul@vice.ii.uj.edu.pl] has joined #uml
00:50|-|Magotari_ [~karol@chello089076066108.chello.pl] has joined #uml
00:50|-|weasel_ [weasel@weasel.chair.oftc.net] has joined #uml
00:53|-|Netsplit synthon.oftc.net <-> kilo.oftc.net quits: Magotari, dsoul_, weasel
06:45|-|Magotari_ [~karol@chello089076066108.chello.pl] has quit [Quit: Lost terminal]
07:28|-|lily [~a@221.209.198.232] has joined #uml
07:28|-|lily [~a@221.209.198.232] has quit []
08:45|-|dang [~dang@nemesis.fprintf.net] has quit [Quit: Leaving.]
09:07|-|dang [~dang@aa-redwall.nexthop.com] has joined #uml
10:12|-|karol [~karol@abhu203.neoplus.adsl.tpnet.pl] has joined #uml
10:12|-|karol changed nick to Magotari
11:04|-|jdike [~jdike@pool-71-248-190-161.bstnma.fios.verizon.net] has joined #uml
11:04<jdike-#uml->>Hi guys
11:11|-|da-x_ [karrde@bzq-79-182-10-21.red.bezeqint.net] has quit [Quit: Client exiting]
11:15<peterz-#uml->>hey jeff
11:16|-|karrde_ [karrde@bzq-79-182-118-83.red.bezeqint.net] has joined #uml
11:21<kokoko1-#uml->>yo
11:30|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has joined #uml
11:40|-|jjkola [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has joined #uml
11:40<Magotari-#uml->>Hey, everyone.
11:40<jjkola-#uml->>hello
11:42<jjkola-#uml->>I think I have found some new information regarding the network problems
11:43<jdike-#uml->>packet loss?
11:43<jjkola-#uml->>I'm not sure about that anymore
11:44<jjkola-#uml->>first things first, if I run commands on guests to make them restart network interfaces very often then I can keep my guests mostly up
11:45<jjkola-#uml->>but if I don't do that then within 24h if I don't keep make connections to those hosts they lock up or go down
11:45<jdike-#uml->>what locks up or goes down?
11:46<jdike-#uml->>connections or UMLs?
11:46<jjkola-#uml->>umls
11:46<jdike-#uml->>lock up as in hang?
11:46<jdike-#uml->>and go down as in crash?
11:46<jjkola-#uml->>yes
11:46<jdike-#uml->>do you have any stack traces?
11:46<jdike-#uml->>and what version?
11:47<jjkola-#uml->>2.6.23 on both side
11:47<jjkola-#uml->>it's always the same back trace
11:47<jjkola-#uml->>I think I already showed one to you a few days ago
11:48<jdike-#uml->>show me again
11:50<jjkola-#uml->>here is one: http://rafb.net/p/sMs0El55.html
11:51<jdike-#uml->>Oh yeah
11:52<jdike-#uml->>can you ask gdb for the sigcontext for the SIGSEGV?
11:52<jjkola-#uml->>one other interesting thing seems to be that if I send a flood of pings from the host to outside then the round trip time seems to be proportional to amount of other connections
11:53<jjkola-#uml->>ping times has varied from 16 ms to about 9000 ms
11:53<jjkola-#uml->>to address and within a few minutes
11:53<jjkola-#uml->>*to same address
11:54<jjkola-#uml->>I don't have hanged umls at the moment so I can't check that
11:54<jdike-#uml->>next time you do, let it sit there if you can so I can have you poke at it
11:55<jjkola-#uml->>ok
11:56<jjkola-#uml->>as both the host and guests seems to be affected I'm starting to think this may have something to do with the bridge code
11:57|-|karrde_ [karrde@bzq-79-182-118-83.red.bezeqint.net] has quit [Quit: Client exiting]
11:58<jjkola-#uml->>all the network connections go through one or the other virtual bridges on the host side (although I have also uml_switch running for guests for internal connections)
12:00<jjkola-#uml->>is there any newer code for bridges than which can be found on uml site?
12:02<jjkola-#uml->>ops, my bad, it was the code for tunctl not brctl that was on the uml site
12:03|-|karrde_ [karrde@bzq-79-178-67-136.red.bezeqint.net] has joined #uml
12:29|-|ram [~ram@bi01p1.co.us.ibm.com] has joined #uml
12:56<Magotari-#uml->>Oh great.
12:57<jjkola-#uml->>Magotari: problems?
12:57<Magotari-#uml->>Instead of submitting a patch today, I get to redo all my work. Why? Because I just blasted the tree to splinters.
12:57<Magotari-#uml->>Little happy accidents. *sigh*
12:57<jjkola-#uml->>ouch
12:58<Magotari-#uml->>Under 30 lines, but I am a test maniac. I usually test the patches a lot. So even though I could just redo it in a few hours, I am not sending in something untested.
13:01<jjkola-#uml->>I see
13:13<Magotari-#uml->>End of the world is cancelled. I found the patch, safely put away. "Just in case."
13:16|-|Nem^ [~Nem@dslb-084-056-254-226.pools.arcor-ip.net] has quit [Ping timeout: 480 seconds]
13:20|-|tyler29 [~tyler@ARennes-257-1-10-50.w81-250.abo.wanadoo.fr] has joined #uml
13:22<Magotari-#uml->>jdike: You have been doing this for a while so maybe you know. Any idea how long the -mm release cycle is? Mainline so far has been fairly regular, two to three months and a week or two between the rcs, but I have not seen a -mm for a while.
13:22<Magotari-#uml->>(Downloading a clean rc2 tree.)
13:26|-|Nem^ [~Nem@dslb-084-057-246-223.pools.arcor-ip.net] has joined #uml
13:33|-|mgross [~mgross@jffwprtest.jf.intel.com] has joined #uml
13:34<Magotari-#uml->>Applies cleanly.
13:36<jdike-#uml->>rc2-mm is taking a while
13:36<jdike-#uml->>longer than usual
14:02|-|jjkola1 [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has joined #uml
14:06|-|jjkola [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has quit [Ping timeout: 480 seconds]
14:16|-|Magotari [~karol@abhu203.neoplus.adsl.tpnet.pl] has quit [Ping timeout: 480 seconds]
14:18|-|Magotari [~karol@abgr170.neoplus.adsl.tpnet.pl] has joined #uml
14:28|-|jjkola1 changed nick to jjkola
15:01|-|tyler29 [~tyler@ARennes-257-1-10-50.w81-250.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
15:06<Magotari-#uml->>jdike: Style question. At the end of void functions, do I put in a return; or is the implicit return good enough?
15:06<jdike-#uml->>the implicit return is enough
15:06<Magotari-#uml->>Just at the end, I know the ones in the middle are needed.
15:06<Magotari-#uml->>Ok. Thank you.
15:14|-|tyler29 [~tyler@ARennes-257-1-62-174.w81-53.abo.wanadoo.fr] has joined #uml
15:24|-|waldner [~waldner@host-84-220-175-96.cust-adsl.tiscali.it] has joined #uml
15:26|-|lazblaz [~103623425@AClermont-Ferrand-156-1-95-214.w86-206.abo.wanadoo.fr] has joined #uml
15:26|-|lazblaz [~103623425@AClermont-Ferrand-156-1-95-214.w86-206.abo.wanadoo.fr] has left #uml []
15:26<jdike-#uml->>waldner
15:27<waldner-#uml->>jdike: I'm getting the asm/page.h error even with 2.6.24-rc2 on amd64
15:27<waldner-#uml->>yes
15:27<jdike-#uml->>asm/page.h doesn't exist?
15:28<waldner-#uml->>no, it doesn't
15:28<waldner-#uml->>I'm googling to see if it's a local problem of mine and whether I should upgrade my headers
15:28<waldner-#uml->>I have no asm/page.h under /usr/include
15:29<jdike-#uml->>http://rafb.net/p/sw6obR84.txt
15:29<jdike-#uml->>you can just delete the include
15:29<waldner-#uml->>ah ok
15:29<jdike-#uml->>that patch hasn't made it to mainline yet, but it's on its way
15:30<waldner-#uml->>ok, it seems it's compiling fine now
15:30<waldner-#uml->>hold on
15:31<waldner-#uml->>this asm/page.h has been quite a nasty one, hasn't it?
15:31<jdike-#uml->>a few headers have caused trouble
15:31<jdike-#uml->>asm/user.h too
15:31<waldner-#uml->>yes, that one too
15:31<jdike-#uml->>libc is getting more careful about what it exports to userspace
15:32<jdike-#uml->>and UML has used headers which it really shouldn't have
15:32<waldner-#uml->>so in the end it's been a good thing
15:32<jdike-#uml->>yup
15:32<jdike-#uml->>UML is getting cleaner
15:33<waldner-#uml->>still compiling (finger crossed)
15:34<jdike-#uml->>hehe
15:34<waldner-#uml->>ok, finished
15:34<waldner-#uml->>btw, is it normal that I get lots of these at the end
15:35<waldner-#uml->>(.text+0xebc1): warning: Using 'getaddrinfo' in statically linked applications requires at runtime the shared libraries from the glibc version used for linking
15:35<jdike-#uml->>I haven't seen that
15:35<jdike-#uml->>seems harmless though
15:36<jdike-#uml->>why are you statically linking?
15:36<jdike-#uml->>ooh
15:36<jdike-#uml->>the port-helper thing happens here
15:36<waldner-#uml->>oh good
15:36<waldner-#uml->>so you can see it for yourself
15:37<waldner-#uml->>I was just trying to see whether it still happened
15:38<waldner-#uml->>...and still happens here too
15:38<jdike-#uml->>one per xterm
15:38<waldner-#uml->>exactly
15:41|-|Blissex [~Blissex@82-69-39-138.dsl.in-addr.zen.co.uk] has joined #uml
15:43<waldner-#uml->>I link statically because sometimes I want to redistribute the binary kernel
15:44<waldner-#uml->>but doesn't matter now
15:50<Magotari-#uml->>jdike: Finally, the patch is sent.
15:51<jdike-#uml->>OK
15:51<Magotari-#uml->>I wanted to change some stuff around path handling, but in the end I decided to research more, so I will put it off till later.
15:52<Magotari-#uml->>The patch I just sent was compile and run tested, so it should be fine.
16:03<jdike-#uml->>Oh man
16:03<jdike-#uml->>that only happens on x86_64
16:04<jdike-#uml->>if(n != sizeof(iov.iov_len))
16:04<jdike-#uml->>should be if(n != iov.iov_len)
16:04<waldner-#uml->>what file?
16:05<jdike-#uml->>arch/um/os-Linux/file.c
16:05<jdike-#uml->>os_rcv_fd
16:06<waldner-#uml->>ok
16:06<jdike-#uml->>someone really needs to shoot me before I hurt someone
16:06<waldner-#uml->>lol
16:07<waldner-#uml->>I'm trying it
16:08<waldner-#uml->>I guess that's because only on amd64 iov.iov_len != sizeof(iov.iov_len)
16:08<waldner-#uml->>correct?
16:12<jdike-#uml->>yup
16:12<jdike-#uml->>one is 8 bytes, one is 4
16:12<waldner-#uml->>but on x86 they are both 4
16:12<jdike-#uml->>yup
16:13<waldner-#uml->>it works!
16:13<jdike-#uml->>here too
16:13<waldner-#uml->>many thanks
16:13<jdike-#uml->>thanks for telling me
16:13<waldner-#uml->>as usual, you were very helpful
16:14<waldner-#uml->>I can finally remove that "killall port-helper" I had at the end of every script :-)
16:14<jdike-#uml->>hehe
16:15<waldner-#uml->>I saw you included vde into UML
16:15<waldner-#uml->>what are the main advantages over uml_switch?
16:16<jdike-#uml->>I included it because the VDE people sent it to me, and it works
16:16<jdike-#uml->>however, it seems to be more general
16:16<jdike-#uml->>i.e. I think it works between hosts
16:17<waldner-#uml->>yes, you can connect vde switches together via virtual wires that pass through ssh
16:17<jdike-#uml->>yeah, that's nice
16:17<waldner-#uml->>well
16:17<waldner-#uml->>gtg now
16:17<waldner-#uml->>thanks again
16:18<Magotari-#uml->>jdike: Have you seen how qemu handles the network by default? You don't have to set up anything. My biggest woe with uml now is a lack of such a thing available without any configuration.
16:19<Magotari-#uml->>I think slirp is pretty close to it though.
16:19<jdike-#uml->>slirp sucks though
16:19<jdike-#uml->>what does qemu do?
16:19<Magotari-#uml->>I could not get the monster to run, so I don't know.
16:19<Magotari-#uml->>Well... Let me think how to describe it.
16:20<Magotari-#uml->>First, it does not change anything on the host.
16:20<Magotari-#uml->>Second, it creates a subnet, which is only visible from inside qemu.
16:20<Magotari-#uml->>It runs a dhcp server inside that subnet. You can also contact the host machine via one of the numbers of the subnet.
16:21<Magotari-#uml->>And most importantly, no changes on the host at all. Just like surfing the web, except instead of a browser it is an imaginary computer.
16:21<Magotari-#uml->>That is my understanding of it.
16:21<Magotari-#uml->>This is by default, btw.
16:22<jdike-#uml->>how does contact with the outside world work?
16:22<Magotari-#uml->>I am sure it does not allow a lot of functionality, but it is simple.
16:22<Magotari-#uml->>It just works. :) I have no idea as far as internals go.
16:22<jdike-#uml->>if there's no host setup, it must be something like slirp
16:22<jdike-#uml->>or exactly slirp would be my guess
16:22<Magotari-#uml->>Yeah, the subnets look pretty similar.
16:23<jdike-#uml->>and the magic IP numbers on the subnet sounds like slirp
16:23<Magotari-#uml->>Hmm... Except you don't need slirp for it to work. Maybe built-in?
16:23<jdike-#uml->>maybe
16:24<Magotari-#uml->>"A virtual network in QEMU uses source codes known as SLiRP."
16:24<Magotari-#uml->>Well, that explains a bit.
16:25<jdike-#uml->>yup
16:28|-|tyler29 [~tyler@ARennes-257-1-62-174.w81-53.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
16:33<Magotari-#uml->>One more thing I would like to talk about. In the code I saw you put /usr/lib/uml into the path. And yet, you sometimes use /usr/lib/uml/port-helper. Won't just using port-helper be better in case someone has the utilities in a different place in their path?
16:33|-|Blissex [~Blissex@82-69-39-138.dsl.in-addr.zen.co.uk] has quit [Read error: Connection reset by peer]
16:35<Magotari-#uml->>In one place you first check if the thing is available under the "/usr/lib/uml/port-helper" and if it is not there you just try to use "port-helper".
16:35<Magotari-#uml->>Since the directory is already in our path... Is this needed?
16:36<jdike-#uml->>it might not be though
16:36<jdike-#uml->>although the cases where it wouldn't are
16:36<jdike-#uml->>the distro installs it in a different place
16:37<jdike-#uml->>the user builds it and installs it in a different place
16:37|-|ram [~ram@bi01p1.co.us.ibm.com] has quit [Quit: Leaving]
16:37<jdike-#uml->>in both cases, you'd expect it to be in the path
16:38<jdike-#uml->>but if you take the tarball off the UML site and build and install it
16:38<jdike-#uml->>you'll get stuff in /usr/lib/uml which isn't likely to be in your path
16:39<Magotari-#uml->>Yes. So won't trying to always use the /usr/lib/uml/ directory derail some things? In case the distro did not put it there, or the user did not put it there?
16:39<Magotari-#uml->>I see that happening in drivers/port_user.c, line 173 in rc2.
16:41<Magotari-#uml->>Since we always have the /usr/lib/uml/ directory in the path anyway, if we use just the filename we won't miss a program, but we might find it where we would not otherwise.
16:41<jdike-#uml->>possibly
16:42<jdike-#uml->>we don't have /usr/lib/uml in our path anyway
16:42<Magotari-#uml->>UML inserts that directory into the PATH enviroment variable at start.
16:42<jdike-#uml->>it does?
16:42<Magotari-#uml->>Yes.
16:42<Magotari-#uml->>putenv, if I remember right.
16:42<jdike-#uml->>ooh
16:42<jdike-#uml->>forgot about that
16:42<jdike-#uml->>so, yes, you're right
16:43<Magotari-#uml->>I have already worked on it before, but wanted to confirm if I am right.
16:43<Magotari-#uml->>So I can sort things out there with paths, I think.
16:44<Magotari-#uml->>Gentoo worked around that bit in port_user.c by putting port-helper inside /usr/lib/uml/. All the other utilities are somewhere else.
16:47|-|tyler29 [~tyler@ARennes-257-1-23-159.w81-53.abo.wanadoo.fr] has joined #uml
17:03<Magotari-#uml->>I'm off for today. My fiancee is undergoing a major surgery right now, and I cannot focus too well because of that.
17:03|-|Magotari [~karol@abgr170.neoplus.adsl.tpnet.pl] has quit [Quit: leaving]
17:15|-|dang [~dang@aa-redwall.nexthop.com] has quit [Quit: Leaving.]
17:21|-|tyler29 [~tyler@ARennes-257-1-23-159.w81-53.abo.wanadoo.fr] has quit [Ping timeout: 480 seconds]
17:31|-|waldner [~waldner@host-84-220-175-96.cust-adsl.tiscali.it] has quit [Quit: using sirc version 2.211+KSIRC/1.3.12]
17:47|-|dang [~dang@nemesis.fprintf.net] has joined #uml
18:03|-|jjkola1 [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has joined #uml
18:03|-|jjkola changed nick to Guest132
18:03|-|jjkola1 changed nick to jjkola
18:10|-|Guest132 [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has quit [Ping timeout: 480 seconds]
19:01|-|alexg [alexg@sverige.freeshell.ORG] has joined #uml
19:02<alexg-#uml->>do you need to apply a UML patch for 2.6.21 ?
19:02<alexg-#uml->>all of the 2.6 patches are for earlier versions
19:08<jdike-#uml->>no
19:08<jdike-#uml->>but you'll probably be happier with something later than 2.6.21
19:10|-|jjkola [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has quit [Ping timeout: 480 seconds]
19:10|-|jdike [~jdike@pool-71-248-190-161.bstnma.fios.verizon.net] has quit [Quit: Leaving]
19:14|-|jjkola [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has joined #uml
19:14<alexg-#uml->>thx
19:15<jjkola-#uml->>aagh, just when I got a hanged uml to show for jdike it seems he has left : /
19:15<alexg-#uml->>the uml homepage has instructions on how to run uml they provide a file system and a kernel but never mention setting up a host kernel so I was a bit confused
19:15<alexg-#uml->>I follow the instructions and it just hangs
19:16<alexg-#uml->>and spawns a buch more
19:19<alexg-#uml->>udevd-event[1292]: udev_node_mknod: mknod(/dev/pty/s172, 020660, 3, 172) failed: No space left on device
19:19<alexg-#uml->>Testing root filesystem status: read-only filesystem
19:19<alexg-#uml->>Checking root filesystem:
19:19<alexg-#uml->>Starting sysklogd daemons: /usr/sbin/syslogd syslogd: cannot create /dev/log: No space left on device
19:19<alexg-#uml->>that's where it hangs
19:23|-|mgross [~mgross@jffwprtest.jf.intel.com] has quit [Quit: Leaving]
19:28|-|jjkola1 [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has joined #uml
19:29|-|jjkola changed nick to Guest139
19:29|-|jjkola1 changed nick to jjkola
19:29|-|Guest139 [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has quit [Ping timeout: 480 seconds]
19:40|-|jjkola [~jjkola@dsl-olubrasgw1-fec8de00-184.dhcp.inet.fi] has quit [Quit: *pop*]
19:55<alexg-#uml->>no but what is ? :)
20:02|-|hfb [~hfb@pool-71-106-219-180.lsanca.dsl-w.verizon.net] has quit [Quit: Leaving]
20:40|-|flatronf700B [~flatronf7@ns1.clipsalportal.com] has quit [Ping timeout: 480 seconds]
21:56|-|remus [~remus@76.231.178.131] has quit [Remote host closed the connection]
22:08|-|Nem^ [~Nem@dslb-084-057-246-223.pools.arcor-ip.net] has quit [Ping timeout: 480 seconds]
22:18|-|Nem^ [~Nem@dslb-084-056-243-102.pools.arcor-ip.net] has joined #uml
23:58|-|VS_ChanLog [~stats@ns.theshore.net] has left #uml [Rotating Logs]
23:59|-|VS_ChanLog [~stats@ns.theshore.net] has joined #uml
---Logclosed Sat Nov 10 00:00:16 2007