On Sun, 09 Apr 2006 14:15:32 -0700 (PDT), "David S. Miller" <[EMAIL PROTECTED]> wrote:
> > I wonder if this will cause other problems by invoked scripts > seeing the state not advanced yet? > I don't know. An other way would be to allow reading 'address' also when device is in NETREG_REGISTERING state. I think the time when show_address() can be called with the device in this state is exactly the spotted race window (beetween creation of the sysfs entry and reg_state being set to NETREG_REGISTERED). The attached patch does that and worked here. It sounds to me less likely to have other side effects, but sure i may be wrong. -- TGL.
--- net/core/net-sysfs.c.orig 2006-04-10 01:03:52.000000000 +0200 +++ net/core/net-sysfs.c 2006-04-10 01:19:48.000000000 +0200 @@ -110,7 +110,7 @@ ssize_t ret = -EINVAL; read_lock(&dev_base_lock); - if (dev_isalive(net)) + if (dev_isalive(net) || (net->reg_state == NETREG_REGISTERING)) ret = format_addr(buf, net->dev_addr, net->addr_len); read_unlock(&dev_base_lock); return ret;
