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