82574L hangs (with r233708 e1000 driver).

[Jason Wolfe]

On Sat, Apr 7, 2012 at 6:37 AM, Konstantin Belousov <kostikbel at gmail.com> wrote:
> I bought Intel Atom motherboard DN2800MT, which has integrated if_em
> LOM, reported by pciconf as
> em0 at pci0:1:0:0: class=0x020000 card=0x20128086 chip=0x10d38086 rev=0x00 hdr=0x00
> 82574L Gigabit Network Connection
> It seems that any non-trivial network activity on the interface causes
> reliable interface hang, which can be temporary cured by ifconfig down/up.
> This happens both with outdated stable/9 sys/dev/e1000 and with driver
> from HEAD merged into stable/9. I currently use the copy of dev/e1000
> at rev. r233708. Disabling MSI-X makes the hand to occur slightly less
> often.
>
> I can reproduce the hang in approximately a minute by scp'ing large
> file from other machine to /dev/null on the DN2800MT. This makes the
> board completely unusable for me.
>
> The driver reports itself as
> em0: <Intel(R) PRO/1000 Network Connection 7.3.2> port 0x2000-0x201f mem 0xc0400000-0xc041ffff,0xc0000000-0xc03fffff,0xc0420000-0xc0423fff irq 16 at device 0.0 on pci1
> em0: attempting to allocate 3 MSI-X vectors (5 supported)
> msi: routing MSI-X IRQ 258 to local APIC 0 vector 60
> msi: routing MSI-X IRQ 259 to local APIC 0 vector 61
> msi: routing MSI-X IRQ 260 to local APIC 0 vector 62
> em0: using IRQs 258-260 for MSI-X
> em0: Using MSIX interrupts with 3 vectors
> em0: bpf attached
> em0: Ethernet address: 00:22:4d:7a:47:f6
> em0: Link is up 1000 Mbps Full Duplex
>
> The board is connected to ProCurve switch, there is no link flaps.
>
> When hang occur, dmesg output of # sysctl dev.em.0.debug=1
> dev.em.0.debug: Interface is RUNNING and ACTIVE
> em0: hw tdh = 357, hw tdt = 357
> em0: hw rdh = 323, hw rdt = 273
> em0: Tx Queue Status = 0
> em0: TX descriptors avail = 1024
> em0: Tx Descriptors avail failure = 0
> em0: RX discarded packets = 0
> em0: RX Next to Check = 274
> em0: RX Next to Refresh = 273
>
> [..snip..]
> Any help ?

Konstantin,

It doesn't appear to be the exact issue I was seeing, but here is a
patch that did resolve issues with my 82574L buffers filling and
requiring a down/up to recover.  It was the result of the 'Intel
82574L interface wedging - em7.3.2/8.2-STABLE' if you'd like to peruse
that.

Jason
-------------- next part --------------
A non-text attachment was scrubbed...
Name: if_em7.2.3_fix.diff
Type: application/octet-stream
Size: 11756 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-net/attachments/20120407/829bd92e/if_em7.2.3_fix.obj