Re: When will the new BCE driver in HEAD be incorporated into RELENG_6?
Bill Moran wrote: [snip] >> The driver from HEAD (if_bce.c 1.17) is looking much better - thanks >> Scott! >> >> We couldn't trigger this bug using UDP NFS mounts. >> >> Neither could we trigger it with multiple simultaneous TCP connections. > > I'm seeing similar improvement. I've been testing the new version since > I came in this morning (About 2.5 hours so far), with no failures. [snip] Adding to the excitement, our band of attack squirrels haven't been able to bring down our 2950 yet with this one. We are using the if_bce.c from the HEAD w/ a small chunk about vlans commented out (for any wondering). We're trying outbound UDP_STREAM tests w/ netperf of big and small payloads to a machine on the same 100Mb switch. Nothing fancy, but what we could once crash in seconds now hangs tough till the end of the test. Sam -- Lumeta Corporation www.lumeta.com The world's most security conscious organizations rely on Lumeta to find and prioritize what's in IT, so they can manage and secure IT better. Lumeta Corporation 220 Davidson Ave, 4th Floor Somerset, NJ 08873 ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: Diskless PXEboot crashes at kernel loading.
Hi all, I can't emphasise enough how useful running a TCPDUMP on the server in question is to find out what subtle thing it is that I missed. Generally speaking a quick-crash like that is a file not being found like the modules directory or something silly like that. If you dump the NFS traffic you will probably see the client asking for a file handle for file X and getting denied by the server. If you're not familiar w/ TCPDUMP, here are some handy commands: # Avoid SSH and dump everything to the screen... tcpdump -s 1500 -X udp # Write it to a file so you can open it in ethereal later... tcpdump -s 1500 -X -w nfs.dump udp # Read in the dump file and read it through less, should you not want to us # ethereal after the last step. :) tcpdump -X -r nfs.dump |less Note that if you have a slow or missing DNS entry, give these calls a -n to just get the data and not worry about the ND entries. Hope this helps! PXE boots are never kind. :) Sam Baskinger Software Engineer Lumeta - Securing the Network in the Face of Change Lumeta Corporation > > > And the crash happens. > > Note that the crash occurs for whatever option > > 1 to 6 I choose from the FreeBSD boot menu. > > > Does someone understand the crash messages? > > Not that I understand those messages but some time ago I've had a > similar case. It took me some hours to realise that I try to load > amd64 kernel to i386 diskless station... ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: Problems installing JDK 1.5
I'm assuming that you installed linux compatibility and then the linux-jdk 1.4? I failed to do that and saw a very similar error a while back. :) Sam Baskinger Lumeta - Securing the Network in the Face of Change www.lumeta.com Juergen Nickelsen wrote: Hello, on a more or less newly installed machine with 5.5-STABLE (updated recently) I tried to install JDK 1.5 via /usr/ports/. After retrieving all necessary files, the build failed quite mysteriously, even more so as I had succeeded with another machine fine last year. In the build protocol the problems seem to begin like this: Java HotSpot(TM) Client VM warning: Can't detect initial thread stack location /usr/ports/java/jdk15/work/control/build/bsd-i586/gensrc/sun/nio/cs/StandardChar\ sets.java:226: expected and then a series of other errors in the Java code; as if it is compiled using the wrong compiler version. It looks weird to me, as I think I fetched the correct files and did everything as carefully as last year. Is this a known problem? Does anyone know a way around it? Is the port currently broken, perhaps? Details can be found in <http://www.schlabun.org/screenlog.2>; and I'll gladly provide more details if necessary. Regards, Juergen. ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]" ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: Problems installing JDK 1.5
Juergen Nickelsen wrote: On Feb 27 2007, Chris Timmons wrote: Don't forget that the linux JDK will want you to have linprocfs mounted. I recently had a similar build experience of jdk-1.5.0p4 which cleared up fine after I remembered about linprocfs. linprocfs /compat/linux/proc linprocfs rw 0 0 Hm. I cannot imagine how this could lead to the problems I saw, but another thing to try. Thanks! Linux systems store tons of need-to-know data in the /proc file system. If you've ever installed a chrooted Linux environment you must 'mount -o bind' the /proc file system from the running system to the chroot for many OS functions to work correctly. I suspect that Java was looking for some process information in linprocfs and since it wasn't mounted, it blew up. :) Hope this clears up the "why" question. :) Sam Baskinger Lumeta - Securing the Network in the Face of Change www.lumeta.com ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: SMP doesn't work without ACPI?
Adding a datapoint: Dell 1950s exhibit similar behaviour but have 2 cores in a single physical CPU. Hope this helps the discussion along. Sam Baskinger Software Engineer Lumeta - Securing the Network in the Face of Change Ivan Voras wrote: Continuing my problems with the IBM blade: Booting with ACPI module enabled (btw. live boot CD with sysinstall doesn't load ACPI, but the installed system does?) hangs the system somewhere after first USB bus is found (booting verbose doesn't show any new lines before or after this step). It appears to be a real hang instead of a timeout because I left it 30 minutes and it didn't continue. Booting without ACPI on the other hand doesn't find all the CPU's :( Here's sysctl output: sysctl -a | grep smp kern.timecounter.smp_tsc: 0 kern.smp.forward_roundrobin_enabled: 1 kern.smp.forward_signal_enabled: 1 kern.smp.cpus: 1 kern.smp.disabled: 0 kern.smp.active: 0 kern.smp.maxcpus: 16 Here's the mptable -dmesg output: === MPTable --- MP Floating Pointer Structure: location: EBDA physical address: 0x0009d140 signature:'_MP_' length: 16 bytes version: 1.4 checksum: 0xfe mode: Virtual Wire --- MP Config Table Header: physical address: 0x0009e9b0 signature:'PCMP' base table length:388 version: 1.4 checksum: 0x77 OEM ID: 'IBM ENSW' Product ID: 'LEWIS SMP ' OEM table pointer:0x OEM table size: 0 entry count: 37 local APIC address: 0xfee0 extended table length:408 extended table checksum: 159 --- MP Config Base Table Entries: -- Processors: APIC ID Version State Family Model Step Flags 0 0x10BSP, usable 15 1 2 0x0301 2 0x10AP, usable 15 1 2 0x0301 1 0x10AP, usable 15 1 2 0x0301 3 0x10AP, usable 15 1 2 0x0301 -- Bus:Bus ID Type 0 PCI 1 PCI 2 PCI 3 PCI 4 PCI 5 PCI 6 PCI 7 PCI 8 PCI 9 ISA -- I/O APICs: APIC ID Version State Address 14 0x11usable 0xfec0 13 0x11usable 0xfec02000 -- I/O Ints: TypePolarityTrigger Bus ID IRQAPIC ID PIN# INT conformsconforms9 1 141 INT conformsconforms9 0 142 INT conformsconforms9 4 144 INT conformsconforms9 6 146 INT active-hiedge9 8 148 INT conformsconforms911 14 11 INT conformsconforms912 14 12 INT conformsconforms913 14 13 INT conformsconforms914 14 14 INT conformsconforms915 14 15 INT conformsconforms0 3:A 143 INT conformsconforms0 3:A 143 INT conformsconforms0 3:A 143 INT conformsconforms0 5:A 130 INT conformsconforms2 4:A 131 INT conformsconforms2 5:A 132 INT conformsconforms3 4:A 133 INT conformsconforms3 5:A 134 INT conformsconforms3 5:B 135 -- Local Ints: TypePolarityTrigger Bus ID IRQAPIC ID PIN# NMI conformsconforms9 02551 ExtINT conformsconforms9 02550 ---
Re: SMP doesn't work without ACPI?
The 1950s that I have (IIRC as I installed them a few months ago) hang at about the same location when ACPI is enabled. I'll see if I can't pull one down and recreate the behavior. I should note that I'm running something after 6.2-RELEASE. Again, I'll try to recreate and get some data from the machines. Sam Scott Long wrote: I've had no problem getting FreeBSD 6 to boot on Dell 1950 and 2950 machines. Where does it hang for you, and what changes have you made to your kernel config? Scott Sam Baskinger wrote: Adding a datapoint: Dell 1950s exhibit similar behaviour but have 2 cores in a single physical CPU. Hope this helps the discussion along. Sam Baskinger Software Engineer Lumeta - Securing the Network in the Face of Change Ivan Voras wrote: Continuing my problems with the IBM blade: Booting with ACPI module enabled (btw. live boot CD with sysinstall doesn't load ACPI, but the installed system does?) hangs the system somewhere after first USB bus is found (booting verbose doesn't show any new lines before or after this step). It appears to be a real hang instead of a timeout because I left it 30 minutes and it didn't continue. Booting without ACPI on the other hand doesn't find all the CPU's :( Here's sysctl output: sysctl -a | grep smp kern.timecounter.smp_tsc: 0 kern.smp.forward_roundrobin_enabled: 1 kern.smp.forward_signal_enabled: 1 kern.smp.cpus: 1 kern.smp.disabled: 0 kern.smp.active: 0 kern.smp.maxcpus: 16 Here's the mptable -dmesg output: === MPTable --- MP Floating Pointer Structure: location: EBDA physical address: 0x0009d140 signature:'_MP_' length: 16 bytes version: 1.4 checksum: 0xfe mode: Virtual Wire --- MP Config Table Header: physical address: 0x0009e9b0 signature:'PCMP' base table length:388 version: 1.4 checksum: 0x77 OEM ID: 'IBM ENSW' Product ID: 'LEWIS SMP ' OEM table pointer:0x OEM table size: 0 entry count: 37 local APIC address: 0xfee0 extended table length:408 extended table checksum: 159 --- MP Config Base Table Entries: -- Processors: APIC ID Version State Family Model Step Flags 0 0x10BSP, usable 15 1 2 0x0301 2 0x10AP, usable 15 1 2 0x0301 1 0x10AP, usable 15 1 2 0x0301 3 0x10AP, usable 15 1 2 0x0301 -- Bus:Bus ID Type 0 PCI 1 PCI 2 PCI 3 PCI 4 PCI 5 PCI 6 PCI 7 PCI 8 PCI 9 ISA -- I/O APICs: APIC ID Version State Address 14 0x11usable 0xfec0 13 0x11usable 0xfec02000 -- I/O Ints: TypePolarityTrigger Bus ID IRQAPIC ID PIN# INT conformsconforms9 1 141 INT conformsconforms9 0 142 INT conformsconforms9 4 144 INT conformsconforms9 6 146 INT active-hiedge9 8 148 INT conformsconforms911 14 11 INT conformsconforms912 14 12 INT conformsconforms913 14 13 INT conformsconforms914 14 14 INT conformsconforms915 14 15 INT conformsconforms0 3:A 143 INT conformsconforms0 3:A 143 INT conformsconforms0 3:A 143 INT conformsconforms0 5:A 130 INT conformsconforms2 4:A 131 INT conformsc
Re: Problems installing JDK 1.5
We've seen JVM lockups during high contention Threading operations with the jdk15 port. The same code succeeded when running under diablo-jdk. I have precious little information on the details of the failure we saw so I'm hesitant to toss the jdk15 port squarely under the bus without more poking at the issue, but time just isn't on my side lately. :) Perhaps this will jog the memories of others who can elaborate. Juergen Nickelsen wrote: Stefan Walter <[EMAIL PROTECTED]> writes: It might be easier for you to just use the java/diablo-jdk15 port, which installs precompiled binaries. Although I have, with the much appreciated help from people here, successfully installed the jdk15 port in between, I thought about diablo-jdk15, too. Would there be any difference between both ports in terms of compatibility, features, and performance from the viewpoint of the Java developer and user? -- Sam Baskinger Software Engineer Lumeta - Securing the Network in the Face of Change www.lumeta.com ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: GELI versus GBDE?
I've been working on a ruby script to manage some geli file systems and have had some good experience using "-k -" to make it read from standard in. It's mixed with popen calls instead of a more bash-y version, but it works. :) I have not tried running it w/o a terminal allocated, but I suspect that won't make much of a difference. (If the script wasn't in such sorry shape at the moment I would copy it along, but I don't think anyone wants to see it now. ;) ) Sam Lumeta - Securing the Network in the Face of Change www.lumeta.com Nikolay Mirin wrote: Anyway, the other reasons that GBDE suck are: 1) Lots of annoying ENOMEM messages, since the memory allocation calls gbde makes are somewhat specific as I understand. One can ignore those messages. 2) GELI provides a onetime key feature, which makes it incredibly convenient for swap and /tmp encryption. 3) The secret key in GELI can be split between the keyfile and the passphrase. The only inconvenience I had with GELI is that if one wants to read a passphrase in a script once and then open a bunch of volumes, than one has to use "expect" to feed the passphrase to geli. It requires the terminal input and won't accept the stdin. GBDE does not have such issue. P.S. One can actually have both in kernel. Christian Brueffer said the following on 16.04.2007 11:21: On Sun, Apr 15, 2007 at 08:56:07AM -0500, Nikolay Mirin wrote: Definitely GELI. GBDE will become obsolete very soon as some other things like vinum and such. It was there just as a test of concept as I understand. Many those different disk subsystems are incompatible in fact, the case of GBDE and Vinum is mentioned as an example in the handbook. Read more about GEOM, as this system will unite all possible disk techniqies. Also, GELI takes advantage of crypto-hardware, but I believe that one gets a benefit out of it only if the main CPU is very slow. There are currently no plans to remove GBDE. The problems with Vinum you mention stemmed from the fact, that the original Vinum was not GEOM aware, thus, GELI couldn't have been used with it as well. gvinum has been in existance for some time now and it's fully compatible to both GBDE and GELI. - Christian ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]" ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "[EMAIL PROTECTED]"
