1TODO 2==== 3 4There is a potential for deadlock when allocating a struct sk_buff for 5data that needs to be written out to aoe storage. If the data is 6being written from a dirty page in order to free that page, and if 7there are no other pages available, then deadlock may occur when a 8free page is needed for the sk_buff allocation. This situation has 9not been observed, but it would be nice to eliminate any potential for 10deadlock under memory pressure. 11 12Because ATA over Ethernet is not fragmented by the kernel's IP code, 13the destructor member of the struct sk_buff is available to the aoe 14driver. By using a mempool for allocating all but the first few 15sk_buffs, and by registering a destructor, we should be able to 16efficiently allocate sk_buffs without introducing any potential for 17deadlock. 18