bk.'s questions

The LSI Megaraid controller has a Battery back up for its write cache, which allows, in the event of an unanticipated power outage, to retain the cached, not yet written data, so it can write it out to disk next time the server is booted up. In the event of a clean shutdown, this back up is presumably not necessary, as part of the clean shutdown will involve the controller writing out the cached data to disk.

How does the raid controller get informed that the shutdown is about to happen? Presumably this flush must be the last thing to occur after other OS shutdown code, which might involve writing out data or logs to the raid.

Is there some PCIe level signalling that informs peripherals of a pending power down? Does the Megaraid_sas kernel module inform the card of the pending shutdown? Or does a MegaCli command to flush the cache need to be run from an init script for the shutdown?

I've manually added a Megaraid card and additional set of drives to a Linux system which was installed without one, and want to be sure that I don't need to manually add something in to guarantee the cache is consistent on shutdown.

I've seen PCAPS recently from three separate companies networks where two devices are configured to transmit traffic to each other over UDP, where one side sends a VLAN header on each packet, and the other side does not.For example, two phones sending RTP to each other. X->Y is on VLAN 1, Y->X is not on a VLAN.

I'd considered this to be a misconfiguration, but since I've now seen it three times, and it appears to work fine, is there an advantage doing this? Why/When should this be done?

Suppose I had two Cisco switches each outputting some network traffic to a SPAN port, and I needed to send the sum of all that traffic to a third device for monitoring that traffic via libpcap. The easiest way to accomplish this would be to have two NIC's in the target device and send one SPAN port to each, but suppose the target device only contained a single NIC and was not expandable.

Obviously I can't route the SPAN traffic through a switch or router since the switch won't passively send all the traffic to my monitoring device disregarding MAC, IP, ARP, etc. However, thats exactly what an old school "Passive" Hub would do.

Will two SPAN ports "play nice" outputting to a single hub or will they "talk over" each other and ignore resource contention,etc. I know a SPAN port operating as a normal ethernet device in that it can't receive any traffic, respond to ARP requests, etc, but does it keep enough of the Ethernet smarts to coexist with other transmitters on the same circuit, or would it need to be the only transmitter on the wire?

I have an old Linux machine running a 2.4.37 kernel. In the machine are two NIC cards both on the same network with different IPs. For some testing, we want each IP's traffic routed through that NIC, but what is actually happening is all the traffic for both IPs is getting routed through one of the ethernet devices and the other is unused. If I take another machine on the network with an empty ARP Cache and ping both IPs, and then check the ARP Cache, I see that it has the MAC for one of the two NICs under both IPs. Basically the Linux box is allowing both NICs to respond to ARP requests for each other- which is the default behaviour. However, I have arp_filter enabled for both NICs which is supposed to eliminate this from happening, but appears to have no effect. I've also tried disabling rp_filter, but that seems to have no effect either. Here are my relevant sysctl entries:

net.ipv4.conf.eth1.arp_filter = 1
net.ipv4.conf.eth1.rp_filter = 0
net.ipv4.conf.eth0.arp_filter = 1
net.ipv4.conf.eth0.rp_filter = 0
net.ipv4.conf.lo.arp_filter = 1
net.ipv4.conf.lo.rp_filter = 0
net.ipv4.conf.default.arp_filter = 1
net.ipv4.conf.default.rp_filter = 0
net.ipv4.conf.all.arp_filter = 1
net.ipv4.conf.all.rp_filter = 0

I find tons of links to people with the same problem on google, but once they set arp_filter=1 the problem goes away for them, but not for me.

Edit: I've also been playing with various arp_ignore and arp_announce settings, but no matter what I do I can't get each interface to only respond to ARPs for its own IP

I have several Linux servers connected to APC UPSs. They are connected to the UPS via USB and configured to shutdown if the UPS sends a signal that mains power has been lost and has not returned after a period of time. This works well. In addition they are configured via jumpers to automatically power on if power is lost and then returns. This also works well.

The scenario that is causing issues is when mains power is lost, and the server shuts down, but then mains power is restored before the UPS runs out. In that case we'd like the Server to start up again, but since power is never actually lost post-UPS, the server never sees a power-return event and therefore doesn't power on.

I'm hoping there exists a device that can plug into mains power pre-UPS, and sense the power return state and send wake up signals perhaps via Wake-on-Lan to preconfigured servers to wake them up once mains power has returned that someone can point me to.

Alternate suggestions are also appreciated, only thing I can't do is replace the server hardware, and I'd prefer to keep the same UPSs if possible.

Due to a proprietary kernel module that I don't have the source to and is very picky about what kernel versions it will load into (even with modprobe --f), I find myself in need of running a 2.6.16.XX kernel on my Debian5 machine. The machine boots fine with the 2.6.26-2 stock kernel, and I have successfully build and booted 2.6.26 and 2.6.31 based kernels by making a .deb and the ndoing dpkg -i.

However, when I do the same approach for 2.6.16, the kernel hangs at boot. I'm testing this in a VMWare image, so I don't think its an issue of newer hardware not supported by the older kernel.

For a working kernel, at boot I get:

Uncompressing Linux.. OK booting the kernel
Loading, please wait...
mdadm: No devices listed in the conf file were found
kinit name_to_dev_t /dev/hda5 (dev5,3)
...

With 2.6.16.60, I never get the kinit message. It hangs after the mdadm line. There are no mdadm arrays on this machine, so I doubt its an issue inside the mdadm stuff, which is supposed to just error out as it does in the 2.6.26 case above, but for some reason I'm getting stuck getting into kinit.

I've been banging my head against this wall so I'm very open to suggestions on how to go about troubleshooting this.

I Have a Windows XP Machine Running wireshark, connected to a Mirror port on a network. I'm capturing with no filtering, and it can only see half of some two way TCP conversations. I had thought it was a mirror port issue on the switch, but I can take the same ethernet cable, plug it into a laptop running the same version of wireshark and see both sides of the conversation. I also see a lot more random network activity such as NBNS queries, LLDP multicasts, and Dynamic Truning protocol packets.

This does not appear to be a promiscuous mode issue, because I do see half of a TCP conversation from point A to point B, and I am point C. I've tried replacing the network card but that isn't it. Its not random packet loss, because I see every packet for one side of the one conversation (based on sequence numbers)

I'm looking for any Windows configurations or other programs or clues that could be preventing Wireshark from capturing all packets.

I am trying to run Firefox inside of a Linux chroot. The chroot has a copy of Debian Etch installed via debootstrap. If the chroot is running on a machine w/ a 2.6 kernel, I can run Firefox 2.0, 3.0 and 3.5 all fine. My problem occurs when I move the chroot to a machine running a 2.4.37 kernel, then I can run 2.0 alright, but 3.0 and 3.5 both hang silently with no display and no errors output.

With Firefox 3.5 on the Kernel 2.4 machine, strace -f ./firefox outputs: (only the last few lines before the hang shown)

[pid  8220] set_thread_area({entry_number:-1 -> -1, base_addr:0x41bef140, limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, seg_not_present:0, useable:1}) = -1 ENOSYS (Function not implemented)
[pid  8220] modify_ldt(1, {entry_number:0, base_addr:0x41bef140, limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, seg_not_present:0, useable:1}, 16) = 0
[pid  8220] munmap(0x40019000, 11144)   = 0
[pid  8220] getpid()                    = 8220
[pid  8220] rt_sigaction(SIGRTMIN, {0x400241e0, [], SA_RESTORER, 0x416f58f8}, NULL, 8) =     0
[pid  8220] rt_sigaction(SIGRT_1, {0x40023cc0, [RTMIN], SA_RESTORER, 0x416f58f8}, NULL, 8) = 0
[pid  8220] rt_sigaction(SIGRT_2, {0x40023500, [], SA_RESTORER, 0x416f58f8}, NULL, 8) = 0
[pid  8220] rt_sigprocmask(SIG_BLOCK, [RTMIN], NULL, 8) = 0
[pid  8220] rt_sigprocmask(SIG_UNBLOCK, [RT_1], NULL, 8) = 0
[pid  8220] open("/proc/cpuinfo", O_RDONLY) = 3
[pid  8220] read(3, "processor\t: 0\nvendor_id\t: Genuin"..., 1024) = 422
[pid  8220] read(3, "", 1024)           = 0
[pid  8220] close(3)                    = 0
[pid  8220] readlink("/etc/malloc.conf", 0xbfffe47b, 4096) = -1 ENOENT (No such file or     directory)
[pid  8220] rt_sigprocmask(SIG_SETMASK, NULL, [RTMIN], 8) = 0
[pid  8220] rt_sigsuspend([]

The set_thread_area call failing with "Function Not Implemented" along with the fact that since I'm runninng in a chroot, only the kernel is different between my two running environments, leads me to believe that Firefox does not run on the 2.4 kernel. However, Mozilla lists a long list of requirements for running Firefox, and 2.6 kernel is not one of them. Can anyone suggest a workaround for how I can get this running?