I have a FreeBSD 8-STABLE system that's been running stably since I last upgraded and rebooted on May 8. Yesterday, I updated /usr/src to get ZFS v28 and also seem to have gotten rid of my nice, solid re0 network interface: re0: <RealTek 8168/8111 B/C/CP/D/DP/E PCIe Gigabit Ethernet> port 0xb000-0xb0ff mem 0xea210000-0xea210fff,0xea200000-0xea20ffff irq 16 at device 0.0 on pci5 re0: Using 1 MSI-X message re0: Chip rev. 0x3c000000 re0: MAC rev. 0x00400000 miibus0: <MII bus> on re0 I'm too tired from lack of sleep due to getting the system back up and running to remember all the details, but the summary is that it started autodetecting its media as 10baseT/UTP. Almost immediately after boot - sometimes while still playing in single-user mode - I'd start seeing "no buffer space available" error messages all over the place. Forcing media to 1000baseTX/full-duplex fixed the problem for a few minutes, but it wouldn't stay in that state and would shortly start throwing "no buffer space available" errors again. Until I've gotten some sleep and have more mental energy to figure out exactly what's going on, I've found that forcing the media to 100baseTX keeps it solidly chugging along (if a little slowly). Anyway, that's where I'm at now. If your re NIC is giving you fits this morning, try setting it to 100baseTX and see if that'll get you running until a better fix comes along. - Kirk
Kirk Strauser
2011-Jun-22 00:42 UTC
SOLVED (was: re0 died last night; here's how I half-revived it)
I found the problem: sometime between the May 8 kernel I'd been using and the new one (latest build: 15:02:36 CST today), my system decided to devour socket buffers. I set kern.ipc.maxsockbuf=16777216 and have over an hour of stable multi-user uptime, which is a vast improvement! On Jun 9, 2011, at 9:37 AM, Kirk Strauser wrote:> I have a FreeBSD 8-STABLE system that's been running stably since I > last upgraded and rebooted on May 8. Yesterday, I updated /usr/src > to get ZFS v28 and also seem to have gotten rid of my nice, solid > re0 network interface: > > re0: <RealTek 8168/8111 B/C/CP/D/DP/E PCIe Gigabit Ethernet> port > 0xb000-0xb0ff mem 0xea210000-0xea210fff,0xea200000-0xea20ffff irq 16 > at device 0.0 on pci5 > re0: Using 1 MSI-X message > re0: Chip rev. 0x3c000000 > re0: MAC rev. 0x00400000 > miibus0: <MII bus> on re0 > > I'm too tired from lack of sleep due to getting the system back up > and running to remember all the details, but the summary is that it > started autodetecting its media as 10baseT/UTP. Almost immediately > after boot - sometimes while still playing in single-user mode - I'd > start seeing "no buffer space available" error messages all over the > place. > > Forcing media to 1000baseTX/full-duplex fixed the problem for a few > minutes, but it wouldn't stay in that state and would shortly start > throwing "no buffer space available" errors again. Until I've gotten > some sleep and have more mental energy to figure out exactly what's > going on, I've found that forcing the media to 100baseTX keeps it > solidly chugging along (if a little slowly). > > Anyway, that's where I'm at now. If your re NIC is giving you fits > this morning, try setting it to 100baseTX and see if that'll get you > running until a better fix comes along. > > - Kirk > > _______________________________________________ > freebsd-stable@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/freebsd-stable > To unsubscribe, send any mail to "freebsd-stable-unsubscribe@freebsd.org > "