I looked only at the 'kernel' bits, except where I wandered off by accident.
Overall, LGTM (tee-hee). General (these stray into design, sorry): - Various components (e.g. aggr and vlan) base decisions on both the apparent and native media types of the mac. How appropriate and extensible is this? For example, both aggr and vlan disallow the creation of their respective entities if the native media type of the underlying mac is not DL_ETHER. This may be an appropriate decision if the underlying type is DL_WIFI, but what if there is some other underlying type? usr/src/uts/common/io/mac/plugins/mac_wifi.c: - IEEE80211_HDRSIZE: why "+ 1"? - mac_wifi_header_info() makes no attempt to deal with chained mblks. Is there a clear requirement on drivers that they must present the headers in a single block? - mac_wifi_header_uncook(), line 404: mhi_hdrsize >= sizeof (struct ether_header) ? usr/src/uts/common/sys/dld.h: - how will we increase the size of DLD_SECOBJ_VAL_MAX when that becomes necessary? usr/src/uts/common/io/dld/dld_drv.c: - drv_secobj_init(): is the kmem_cache actually worthwhile? The constructor does nothing but bzero, the kmem_cache_alloc() caller sets all of the fields and the kmem_cache_free() caller doesn't 're-init' the bufer. - line 707: prefer "sizeof (*sgp)", given line 710. - line 727: spurious blank. dme. -- David Edmondson, Sun Microsystems, http://www.dme.org _______________________________________________ networking-discuss mailing list [email protected]
