On Sun, Oct 31, 2004 at 10:44:33PM -0600, Matt Domsch wrote:
> I think -EOVERFLOW would be appropriate return for dev->addr_len >
> sizeof sa_data, yes?
> I'd prefer though, that an "obsolete" function, be marked as such
> somehow (perhaps print a net_ratelimit()ed KERN_DEBUG message when
> it's called telling apps to move to rtnetlink), and that the behaviour
> for everything except overflow be consistent with the prior
> implementation, at least until such a time that all the apps in the
> distros are converted to rtnetlink.
How's this look?
This makes ioctl(SIOCIFHWADDR) behavior in 2.6 consistent with that of
previous kernels wherever possible. It returns -EOVERFLOW if it can't
represent the address, zeros the data if dev->addr_len is zero, and
prints a KERN_DEBUG message telling people to fix their applications
to use rtnetlink.
Compiles on x86 and x86-64 cleanly.
Signed-off-by: Matt Domsch <Matt_Domsch@xxxxxxxx>
Sr. Software Engineer, Lead Engineer
Dell Linux Solutions linux.dell.com & www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
===== net/core/dev.c 1.169 vs edited =====
--- 1.169/net/core/dev.c 2004-10-26 11:09:33 -05:00
+++ edited/net/core/dev.c 2004-11-01 11:20:25 -06:00
@@ -2375,8 +2375,16 @@
return dev_set_mtu(dev, ifr->ifr_mtu);
- memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr,
- min(sizeof ifr->ifr_hwaddr.sa_data, (size_t)
+ if (net_ratelimit())
+ printk(KERN_DEBUG "Warning: %s uses obsolete
ioctl(SIOCGIFHWADDR), please convert it to rtnetlink(3,7)\n", current->comm);
+ if ((size_t) dev->addr_len > sizeof
+ return -EOVERFLOW;
+ else if (!dev->addr_len)
+ memset(ifr->ifr_hwaddr.sa_data, 0, sizeof
+ memcpy(ifr->ifr_hwaddr.sa_data, dev->dev_addr,
+ min(sizeof ifr->ifr_hwaddr.sa_data,
ifr->ifr_hwaddr.sa_family = dev->type;