>> * 215-220: I don't understand this comment. What is its purpose? What
>> is the impact of sh_cnt not being accurate?
>>
> I think it might not be needed any more after we move dls_destroy() as part
> of softmac_destroy(). I will test it out.
>
I tried this out, and the special handling for GLDv3 is still needed. I
changed the comments to:
/*
* Note that for GLDv3 devices, we create devfs minor nodes
* for VLANs as well. Assuming a GLDv3 driver on which only
* a VLAN is created. During the detachment of this device
* instance, the following would happen:
* a. the pre-detach callback softmac_destroy() succeeds.
* Because the physical link itself is not in use,
* softmac_destroy() succeeds and destroys softmac_head_t;
* b. the device detachment fails in mac_unregister() because
* this MAC is still used by a VLAN.
* c. the post-attachment callback is then called which leads
* us here. Note that ddi_minor_node_count() returns 3
* (including the minior node of the VLAN). In that case,
* we must correct the minor node count to 2 as that is
* the count of minor nodes that go through post-attachment.
*/
Thanks
- Cathy
