Re: pxeboot, machine dependent kernel
On 7 September 2012 14:08, russell russ...@dotplan.dyndns.org wrote: I have doing quite a lot of netbooting lately. However I can not figure out how to configure a specific machine to use a specific kernel. Is there a way for pxeboot to load a kernel based on something machine dependent, for example, mac address? If not, I have been digging around in sys/stand/boot/boot.c while I have not found where to get the mac address yet would it be preferable to a. look for a boot.conf.macaddress before an unadorned boot.conf b. if not otherwise specified fall back to /bsd.macadress c. macro expansion in boot.conf(somthing in the manner of machine $macaddress) I like option a as that seems like it would be easy to put in and provide configuration power where needed while not complicating the setup in the common case of only ever needing one kernel. Have you checked man 8 diskless ? -- Ville
Re: pxeboot, machine dependent kernel
On 09/08/12 03:34, Ville Valkonen wrote: On 7 September 2012 14:08, russell russ...@dotplan.dyndns.org wrote: I have doing quite a lot of netbooting lately. However I can not figure out how to configure a specific machine to use a specific kernel. Is there a way for pxeboot to load a kernel based on something machine dependent, for example, mac address? If not, I have been digging around in sys/stand/boot/boot.c while I have not found where to get the mac address yet would it be preferable to a. look for a boot.conf.macaddress before an unadorned boot.conf b. if not otherwise specified fall back to /bsd.macadress c. macro expansion in boot.conf(somthing in the manner of machine $macaddress) I like option a as that seems like it would be easy to put in and provide configuration power where needed while not complicating the setup in the common case of only ever needing one kernel. Have you checked man 8 diskless ? -- Ville heh, diskless(8), thats my bible. but my problem is. dhcp: filename directive can be per machine but it does not point to a kernel. it points to a pxeboot. pxeboot: can be configured via boot.conf but there is no way to specify a kernel based on the machine actually booting, can only hard code the kernel image in. and even if I kept different pxeboot binarys they would still use the same boot.conf when different machines (say one is amd64 and the other is i386) need different kernels one boot.conf will not work. I was hoping there was something obvious I missed when setting it up. cause right now I am typing in the kernel name by hand when booting, which sucks and kind of defeats the purpose of netbooting. my intention is to hack boot.c(my guess, at this point I am still just looking at source) to check for and use some sort of global kernel macaddress var pxeboot claims to set. It may seem I have no idea what I am doing, this is true. However I figure this is a good chance to learn.
Re: ext2 max filesystem size
Ok, so I'm trying it out. I got the disk ready with: mkfs.ext2 -L cmol-storage -m 0 -I 128 /dev/sdb1 , to get the inodes right according to this: http://efreedom.com/Question/H-33983/Flash-Drive-OpenBSD-Specified-Device-Mat ch-Mounted-Device The disk shows up in dmesg as: umass0 at uhub0 port 1 configuration 1 interface 0 Seagate Backup+ Desk rev 2.10/1.00 addr 2 umass0: using SCSI over Bulk-Only scsibus1 at umass0: 2 targets, initiator 0 sd0 at scsibus1 targ 1 lun 0: Seagate, Backup+ Desk, 0503 SCSI4 0/direct fixed sd0: 2861588MB, 4096 bytes/sec, 732566645 sec total And disklabel shows: # disklabel sd0 # /dev/rsd0c: type: SCSI disk: SCSI disk label: Backup+ Desk duid: flags: bytes/sector: 4096 sectors/track: 63 tracks/cylinder: 255 sectors/cylinder: 16065 cylinders: 45600 total sectors: 732566645 boundstart: 0 boundend: 732566645 drivedata: 0 16 partitions: #size offset fstype [fsize bsize cpg] c:7325666450 unused i:732566272 256 ext2fs However, when I'm trying to mount I get: # mount -t ext2fs /dev/sd0i /mnt mount_ext2fs: /dev/sd0i on /mnt: specified device does not match mounted device Any ideas? On Fri, Sep 7, 2012 at 6:07 PM, Ted Unangst t...@tedunangst.com wrote: On Fri, Sep 07, 2012 at 15:11, Claus Lensbøl wrote: Hi I'm trying to figure out the maximum file system size for ext2 on openbsd, but I can't seem to find it documented anywhere. Does anybody know if the limitation is the old 2TB (as in linux pre 2.6.17 ish), or 32TB as newer implementations? As far as I know, thats was a limit of the linux kernel, not the filesystem, so it doesn't apply, but I'm not that familiar with ext2fs.
Re: pxeboot, machine dependent kernel
On 7 September 2012 23:14, russell russ...@dotplan.dyndns.org wrote: On 09/08/12 03:34, Ville Valkonen wrote: On 7 September 2012 14:08, russell russ...@dotplan.dyndns.org wrote: I have doing quite a lot of netbooting lately. However I can not figure out how to configure a specific machine to use a specific kernel. Is there a way for pxeboot to load a kernel based on something machine dependent, for example, mac address? If not, I have been digging around in sys/stand/boot/boot.c while I have not found where to get the mac address yet would it be preferable to a. look for a boot.conf.macaddress before an unadorned boot.conf b. if not otherwise specified fall back to /bsd.macadress c. macro expansion in boot.conf(somthing in the manner of machine $macaddress) I like option a as that seems like it would be easy to put in and provide configuration power where needed while not complicating the setup in the common case of only ever needing one kernel. Have you checked man 8 diskless ? -- Ville heh, diskless(8), thats my bible. but my problem is. dhcp: filename directive can be per machine but it does not point to a kernel. it points to a pxeboot. pxeboot: can be configured via boot.conf but there is no way to specify a kernel based on the machine actually booting, can only hard code the kernel image in. and even if I kept different pxeboot binarys they would still use the same boot.conf when different machines (say one is amd64 and the other is i386) need different kernels one boot.conf will not work. I was hoping there was something obvious I missed when setting it up. cause right now I am typing in the kernel name by hand when booting, which sucks and kind of defeats the purpose of netbooting. my intention is to hack boot.c(my guess, at this point I am still just looking at source) to check for and use some sort of global kernel macaddress var pxeboot claims to set. It may seem I have no idea what I am doing, this is true. However I figure this is a good chance to learn. Apparently I remembered the contents of the man and its pointers wrong, so for sorry for the noise. Should have checked. -- Ville
Re: pxeboot, machine dependent kernel
On Saturday 08 September 2012 15:11:07 Ville Valkonen wrote: On 7 September 2012 23:14, russell russ...@dotplan.dyndns.org wrote: On 09/08/12 03:34, Ville Valkonen wrote: On 7 September 2012 14:08, russell russ...@dotplan.dyndns.org wrote: I have doing quite a lot of netbooting lately. However I can not figure out how to configure a specific machine to use a specific kernel. Is there a way for pxeboot to load a kernel based on something machine dependent, for example, mac address? If not, I have been digging around in sys/stand/boot/boot.c while I have not found where to get the mac address yet would it be preferable to a. look for a boot.conf.macaddress before an unadorned boot.conf b. if not otherwise specified fall back to /bsd.macadress c. macro expansion in boot.conf(somthing in the manner of machine $macaddress) I like option a as that seems like it would be easy to put in and provide configuration power where needed while not complicating the setup in the common case of only ever needing one kernel. Have you checked man 8 diskless ? -- Ville heh, diskless(8), thats my bible. but my problem is. dhcp: filename directive can be per machine but it does not point to a kernel. it points to a pxeboot. pxeboot: can be configured via boot.conf but there is no way to specify a kernel based on the machine actually booting, can only hard code the kernel image in. and even if I kept different pxeboot binarys they would still use the same boot.conf when different machines (say one is amd64 and the other is i386) need different kernels one boot.conf will not work. I was hoping there was something obvious I missed when setting it up. cause right now I am typing in the kernel name by hand when booting, which sucks and kind of defeats the purpose of netbooting. my intention is to hack boot.c(my guess, at this point I am still just looking at source) to check for and use some sort of global kernel macaddress var pxeboot claims to set. It may seem I have no idea what I am doing, this is true. However I figure this is a good chance to learn. Apparently I remembered the contents of the man and its pointers wrong, so for sorry for the noise. Should have checked. -- Ville How about man dhcpd.conf?
Re: ext2 max filesystem size
On Sat, Sep 08, 2012 at 01:50:26PM +0200, Claus Lensb?l wrote: Ok, so I'm trying it out. I got the disk ready with: mkfs.ext2 -L cmol-storage -m 0 -I 128 /dev/sdb1 , to get the inodes right according to this: http://efreedom.com/Question/H-33983/Flash-Drive-OpenBSD-Specified-Device-Mat ch-Mounted-Device The disk shows up in dmesg as: umass0 at uhub0 port 1 configuration 1 interface 0 Seagate Backup+ Desk rev 2.10/1.00 addr 2 umass0: using SCSI over Bulk-Only scsibus1 at umass0: 2 targets, initiator 0 sd0 at scsibus1 targ 1 lun 0: Seagate, Backup+ Desk, 0503 SCSI4 0/direct fixed sd0: 2861588MB, 4096 bytes/sec, 732566645 sec total And disklabel shows: # disklabel sd0 # /dev/rsd0c: type: SCSI disk: SCSI disk label: Backup+ Desk duid: flags: bytes/sector: 4096 sectors/track: 63 tracks/cylinder: 255 sectors/cylinder: 16065 cylinders: 45600 total sectors: 732566645 boundstart: 0 boundend: 732566645 drivedata: 0 16 partitions: #size offset fstype [fsize bsize cpg] c:7325666450 unused i:732566272 256 ext2fs However, when I'm trying to mount I get: # mount -t ext2fs /dev/sd0i /mnt mount_ext2fs: /dev/sd0i on /mnt: specified device does not match mounted device Any ideas? On Fri, Sep 7, 2012 at 6:07 PM, Ted Unangst t...@tedunangst.com wrote: On Fri, Sep 07, 2012 at 15:11, Claus Lensb?l wrote: Hi I'm trying to figure out the maximum file system size for ext2 on openbsd, but I can't seem to find it documented anywhere. Does anybody know if the limitation is the old 2TB (as in linux pre 2.6.17 ish), or 32TB as newer implementations? As far as I know, thats was a limit of the linux kernel, not the filesystem, so it doesn't apply, but I'm not that familiar with ext2fs. I'm suspicious that the problem lies with: bytes/sector: 4096 since there should be no problem handling inodes larger than 128 bytes since a commit in 2008. Ken
Re: pxeboot, machine dependent kernel
On 8/09/2012 6:14 AM, russell wrote: ... my intention is to hack boot.c(my guess, at this point I am still just looking at source) to check for and use some sort of global kernel macaddress var pxeboot claims to set. ... I played with a similar patch from here many years ago: http://nbender.com/install.netboot/install.html These days I use iPXE instead, it is fairly painless to set up.
problems on ldpd with multiple links between 2 hosts
Hi, I've made some tests on ldpd and found some problems/strange things with the following configuration: 2 hosts (v 5.0) - core3 (loopback 10.0.0.7) - core1 (loopback 10.0.0.9) 2 links between these 2 hosts: - vlan211 (on em9) core3 IP: 10.0.0.125 core1 IP: 10.0.0.126 - vlan212 (on em5) core3 IP: 10.0.0.121 core1 IP: 10.0.0.122 I've keeped only interesting part in command results (there some other ospf neighbors and interfaces). Sorry for this long post, I wanted to reproduce a problem I've got one time but failed to get it again (see a the end of this email). ospf is running on both hosts: core3: ID Pri StateDeadTime Address Iface Uptime 10.0.0.91 FULL/DR 00:00:34 10.0.0.122 vlan212 00:24:38 10.0.0.91 FULL/DR 00:00:34 10.0.0.126 vlan211 00:23:27 core1: ID Pri StateDeadTime Address Iface Uptime 10.0.0.71 FULL/BCKUP 00:00:31 10.0.0.121 vlan212 00:24:50 10.0.0.71 FULL/BCKUP 00:00:31 10.0.0.125 vlan211 00:23:40 ldpd.conf on core3: router-id 10.0.0.7 distribution independent retention liberal advertisement unsolicited interface vlan212 interface vlan211 same on core1 except that router-id is 10.0.0.9 I start ldpd on 2 hosts: core3 logs: Sep 8 14:04:59 core3 ldpd[5529]: startup Sep 8 14:05:00 core3 ldpd[5529]: mpath route not found Sep 8 14:05:05 core3 ldpd[30289]: Connection attempt from unknown neighbor 10.0.0.126: NO HELLO core1 logs: Sep 8 14:04:54 core1 ldpd[8601]: startup Sep 8 14:04:55 core1 ldpd[17310]: received notification from neighbor 10.0.0.7: Session Rejected, No Hello ldpd seems confused by notification received from same host on different links. root@core3:etc$ ldpctl sh nei ID State Address Iface Uptime 10.0.0.9OPERATIONAL/ACTIVE 10.0.0.126 vlan211 00:00:09 root@core1:etc$ ldpctl sh nei ID State Address Iface Uptime 10.0.0.7OPERATIONAL/ACTIVE 10.0.0.125 vlan211 00:00:13 === First problem: only routes from one interface are tagged: root@core1:etc$ ldpctl sh l Destination Nexthop Local LabelRemote Label In Use 10.200.0.176/29 10.0.0.12156 33 yes 10.200.0.176/29 10.0.0.12556 Untagged yes 10.100.0.0/24 10.0.0.12162 39 yes 10.100.0.0/24 10.0.0.12562 Untagged yes Tagged interface correspond to interface in fib: root@core1:etc$ route -n get 10.100.0.1 route to: 10.100.0.1 destination: 10.100.0.0 mask: 255.255.255.0 gateway: 10.0.0.121 interface: vlan212 if address: 10.0.0.122 mpls label: PUSH 39 priority: 32 (ospf) flags: UP,GATEWAY,DONE,MPATH,MPLS use mtuexpire 72 0 0 Now I shutdown vlan211 on core1 with ifconfig vlan211 down No logs on core3 on core1: Sep 8 14:15:03 core1 ldpd[17310]: interface vlan211 down Sep 8 14:15:03 core1 ospfd[23127]: send_packet: error sending packet on interface vlan211: Network is down Sep 8 14:15:03 core1 ospfd[23127]: interface vlan211 down Sep 8 14:15:07 core1 ldpd[17310]: send_packet: error sending packet on interface vlan211: Network is down root@core3:etc$ ospfctl sh nei ID Pri StateDeadTime Address Iface Uptime 10.0.0.91 FULL/DR 00:00:36 10.0.0.122 vlan212 00:37:17 10.0.0.91 DOWN/OTHER 00:00:54 10.0.0.126 vlan211 - root@core1:etc$ ospfctl sh nei ID Pri StateDeadTime Address Iface Uptime 10.0.0.71 FULL/BCKUP 00:00:33 10.0.0.121 vlan212 00:37:08 10.0.0.71 DOWN/DOWN00:01:18 10.0.0.125 vlan211 - root@core3:etc$ ldpctl sh nei ID State Address Iface Uptime 10.0.0.9OPERATIONAL/ACTIVE 10.0.0.126 vlan211 00:11:40 root@core1:etc$ ldpctl sh nei ID State Address Iface Uptime 10.0.0.7OPERATIONAL/DOWN 10.0.0.125 vlan211 00:11:35 Next I restore vlan211 with ifconfig vlan211 up no logs on core3 on core1: Sep 8 14:17:45 core1 ldpd[17310]: interface vlan211 up Sep 8 14:17:45 core1 ospfd[23127]: interface vlan211 up root@core3:etc$ ospfctl sh nei ID Pri StateDeadTime Address Iface Uptime 10.0.0.91 FULL/DR 00:00:37 10.0.0.122 vlan212 00:39:06 10.0.0.91 FULL/DR 00:00:37 10.0.0.126 vlan211 00:00:23 root@core1:etc$ ospfctl sh nei ID Pri StateDeadTime Address Iface Uptime 10.0.0.71 FULL/BCKUP 00:00:34 10.0.0.121 vlan212 00:39:28 10.0.0.71 FULL/BCKUP 00:00:34 10.0.0.125 vlan211 00:00:45 and shut down vlan212 on core1 with ifconfig vlan212 down Sep 8
Re: pxeboot, machine dependent kernel
On Fri, Sep 07, 2012 at 13:14, russell wrote: can be configured via boot.conf but there is no way to specify a kernel based on the machine actually booting, can only hard code the kernel image in. and even if I kept different pxeboot binarys they would still use the same boot.conf when different machines (say one is amd64 and the other is i386) need different kernels one boot.conf will not work. You can also run two tftp servers on two different IPs.
Re: pxeboot, machine dependent kernel
Ted Unangst t...@tedunangst.com wrote: You can also run two tftp servers on two different IPs. The new tftpd also has this: -r socket Issue filename rewrite requests to the specified UNIX domain socket. tftpd will write lines in the format IP OP filename, terminated by a newline, where IP is the client's IP address, and OP is one of read or write. tftpd expects replies in the format filename terminated by a newline. All rewrite requests from the daemon must be answered (even if it is with the original filename) before the TFTP request will continue. By default tftpd does not use filename rewriting. -- Christian naddy Weisgerber na...@mips.inka.de
Re: xenocara not building on amd64-current
On Tue, Sep 04, 2012 at 03:44:35AM -0400, Ted Unangst wrote: A lot of effort is expended trying to get snapshots out quickly after toolchain changes, precisely to make things easy for people. Even if you think you can figure out building from the source, the polite thing to do is to use the snapshots anyway. :) I am very grateful for the effort the developers put into the snapshots, so I don't mean this as criticism. But it is possible for somebody reading the thread to believe that the latest snapshot would allow xenocara to build. As far as I can tell, it does not (yet) on amd64. Since I'm sure it will be fixed by another snapshot soon, this is no big deal.
Re: Hi - Packages Requests For Current
No one is stopping you from porting them yourself. http://www.openbsd.org/faq/faq1.html#HowAbout On Sun, Sep 9, 2012, at 12:43 AM, Das wrote: Hi, I was looking over Current and I noticed these programs were not in OpenBSD; *WinFF*; (This is a really great FFmpeg frontend for converting video formats, I find this works much better at converting then Avidemux) http://code.google.com/p/winff/ *Bleacbit* (Really great for cleaning up and the privacy conscious) http://bleachbit.sourceforge.net/ *LXAppearance *(This works really great in OpenBox for changing GTK Themes, I'd imagine it would work just fine in Fluxbox and Blackbox too... The latest version is actually 0.5.2, just released a few months ago. http://wiki.lxde.org/en/LXAppearance I personally use all these applications and they all work great for me in Slackware 13.37 x86... Thank you very much for your time and consideration, I look forward to seeing these hopefully soon in Current... Cheers Das