svn commit: r257184 - in head/sys/dev: mii usb/net xl

[Adrian Chadd]

Cool.

Just make sure you check the odd ones like USB, where ejecting may
occur during or just before the callout runs, and the callout tries
accessing registers that aren't there.

Sure the previous was racy, but it may have been "not racy enough" to
cause most people to never see a panic.



-adrian


On 27 October 2013 00:19, Gleb Smirnoff <glebius at freebsd.org> wrote:
> On Sat, Oct 26, 2013 at 06:57:04PM -0700, Adrian Chadd wrote:
> A> hm, so none of the modified PHYs nor their consumers will get upset?
> A> Eg, if the NIC is down, is touching the PHY registers going to be a
> A> problem?
>
> The IFF_UP was always an administrative flag. I believe drivers initialize
> all their resources before being set an address with SIOCSIFADDR.
>
> If a driver doesn't, then the previous code still was racy.
>
> Anyway, I am going to handle any fallouts.
>
>
> --
> Totus tuus, Glebius.