Index | Thread | Search

From:
"Theo de Raadt" <deraadt@openbsd.org>
Subject:
Re: em(4): detect BMC network reachable
To:
Jan Klemkow <j.klemkow@wemelug.de>
Cc:
tech@openbsd.org
Date:
Fri, 06 Nov 2026 16:07:44 -0600

Download raw body.

Thread
Wordy but complete.  OK deraadt

Jan Klemkow <j.klemkow@wemelug.de> wrote:

> On Fri, Nov 06, 2026 at 09:33:35AM -0600, Theo de Raadt wrote:
> > +Traffic to and from the BMC is potentially handled by the controller's
> > +management firmware and is not visible to the operating system.
> > 
> > Well I think this does not address the real concern. It means a
> > typcally buggy BMC is on this network.  It means some catagory of
> > packets is stolen.  It can also mean the port carries two MAC
> > addresses. There are a bunch of problems, and I don't think this
> > sentencies conveys it enough.
> 
> ok?
> 
> Index: sys/dev/pci/if_em.c
> ===================================================================
> RCS file: /cvs/src/sys/dev/pci/if_em.c,v
> diff -u -p -r1.384 if_em.c
> --- sys/dev/pci/if_em.c	14 Aug 2026 06:40:24 -0000	1.384
> +++ sys/dev/pci/if_em.c	6 Nov 2026 12:23:51 -0000
> @@ -619,6 +619,16 @@ em_attach(struct device *parent, struct 
>  		printf("%s: PHY reset is blocked due to SOL/IDER session.\n",
>  		    DEVNAME(sc));
>  
> +	/* Indicate whether the BMC is reachable via this port */
> +	if (sc->hw.mac_type == em_i210 || sc->hw.mac_type == em_i350) {
> +		uint32_t manc = E1000_READ_REG(&sc->hw, MANC);
> +
> +		if (ISSET(manc, E1000_MANC_RCV_TCO_EN) &&
> +		    ISSET(manc, E1000_MANC_EN_BMC2NET))
> +			printf("%s: warning - BMC is enabled on this interface\n",
> +			    DEVNAME(sc));
> +	}
> +
>  	/* Identify 82544 on PCI-X */
>  	if (sc->hw.bus_type == em_bus_type_pcix &&
>  	    sc->hw.mac_type == em_82544)
> Index: sys/dev/pci/if_em_hw.h
> ===================================================================
> RCS file: /cvs/src/sys/dev/pci/if_em_hw.h,v
> diff -u -p -r1.103 if_em_hw.h
> --- sys/dev/pci/if_em_hw.h	14 Aug 2026 07:21:33 -0000	1.103
> +++ sys/dev/pci/if_em_hw.h	6 Nov 2026 12:22:13 -0000
> @@ -2245,6 +2245,7 @@ struct e1000_adv_tx_context_desc {
>  #define E1000_MANC_SMB_DATA_IN   0x08000000 /* SMBus Data In */
>  #define E1000_MANC_SMB_DATA_OUT  0x10000000 /* SMBus Data Out */
>  #define E1000_MANC_SMB_CLK_OUT   0x20000000 /* SMBus Clock Out */
> +#define E1000_MANC_EN_BMC2NET    0x20000000 /* MC <-> network traffic */
>  
>  #define E1000_MANC_SMB_DATA_OUT_SHIFT  28 /* SMBus Data Out Shift */
>  #define E1000_MANC_SMB_CLK_OUT_SHIFT   29 /* SMBus Clock Out Shift */
> Index: share/man/man4/em.4
> ===================================================================
> RCS file: /cvs/src/share/man/man4/em.4,v
> diff -u -p -r1.55 em.4
> --- share/man/man4/em.4	8 Sep 2021 20:29:21 -0000	1.55
> +++ share/man/man4/em.4	6 Nov 2026 13:42:22 -0000
> @@ -193,6 +193,23 @@ Only full-duplex mode is supported at th
>  .Pp
>  For more information on configuring this device, see
>  .Xr ifconfig 8 .
> +.Sh DIAGNOSTICS
> +.Bl -tag -width Ds
> +.It "warning - BMC is enabled on this interface"
> +.Pp
> +A baseboard management controller (BMC) shares this interface with the host
> +and is directly reachable from the connected network.
> +BMC firmware is often buggy and has a history of security vulnerabilities,
> +so exposing it to an untrusted network is a risk.
> +The controller diverts certain categories of received packets to the BMC,
> +for example ARP, RMCP, or traffic addressed to the BMC.
> +These packets are consumed by the BMC and are not visible to the operating
> +system.
> +The BMC may also use its own MAC address,
> +so the port can appear on the network with two MAC addresses.
> +Depending on the system, the BMC's network access may be disabled in the BIOS
> +or BMC configuration.
> +.El
>  .Sh SEE ALSO
>  .Xr arp 4 ,
>  .Xr gcu 4 ,