linux-kernel - Re: [PATCH 2.6.18-rc5] PCI: sort device lists breadth-first

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20060912030848.GA18381@lists.us.dell.com>
Date:	Mon, 11 Sep 2006 22:08:48 -0500
From:	Matt Domsch <Matt_Domsch@...l.com>
To:	Bill Davidsen <davidsen@....com>
Cc:	Arjan van de Ven <arjan@...radead.org>,
	Dave Jones <davej@...hat.com>,
	linux-pci@...ey.karlin.mff.cuni.cz, Greg KH <greg@...ah.com>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2.6.18-rc5] PCI: sort device lists breadth-first

On Mon, Sep 11, 2006 at 03:36:06PM -0400, Bill Davidsen wrote:
> Arjan van de Ven wrote:
> >Maybe the kernel's initial ordering should do a numeric sort by mac
> >address or something.. (or userspace should)
> >
> That wouldn't match any existing setup, and would be subject to mid-list 
> insertions if a NIC were added/replaced. And that is fragile.
> 
> I was looking for an easy way to do PCI slot to MAC, and from there MAC 
> to IP, so any NIC plugged into a given slot could be called eth0 (for 
> instance) and given the "right" IP address, but that's not easy. Can be 
> done with some searching in /sys, but it's non-trivial.

So, I did almost this in
userspace. http://linux.dell.com/files/name_eths/.  It uses the PCI
IRQ Routing table to determine if a PCI device is embedded on the
motherboard, or is in an add-in slot, and if so, which slot.  It
orders the list of thereby possible PCI NICs with all the embeddeds
first, in ascending PCI breadth-first order, then orders all the
add-in NICs in PCI slot number ascending order, subsort PCI
breadth-first for those nifty multiport cards.  It rewrites
modprobe.conf to load network drivers 'proper' order and outputs an
/etc/mactab file that can be used by the second half of the script to
write HWADDR lines into Red Hat-style ifcfg-eth* files, and into
openSuSE udev ethernet rules file.

This works great, until you add another NIC into an add-in slot
somewhere in the middle (e.g. you have 2 embedded NICs eth0 and eth1,
and a NIC card in PCI slot 4 eth2, then at some later point you add a
NIC card into PCI slot 2).  You either have to manually configure a
name for the new card, or run name_eths again and expect the NIC in
PCI slot 2 to become eth2, and the one in slot4 to become eth3.

The pure C udev helper I'm working on behaves similarly, though would
negate the need to edit config files, as it assigns a name "on the
fly" at device discovery time.

For the relatively rare cases of adding a NIC, I'm OK with this.  I
don't have a better way to handle it, but am open to ideas.

We could assign names like eth-embedded-1, eth-embedded-2,
eth-slot2-1, eth-slot4-1 if we wanted to change how people think of
ethernet names (and this would be similar to how large network
switches work: blade N, port M.  We've got 15 usable chars in the name
after all...

Thanks,
Matt

-- 
Matt Domsch
Software Architect
Dell Linux Solutions linux.dell.com & www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/