On Fri, 05 Jan 2024 11:57:03 -0600, Dave Hansen <dave.hansen@xxxxxxxxx>
wrote:
On 10/30/23 11:20, Haitao Huang wrote:
@@ -527,16 +530,13 @@ void sgx_mark_page_reclaimable(struct
sgx_epc_page *page)
int sgx_unmark_page_reclaimable(struct sgx_epc_page *page)
{
spin_lock(&sgx_global_lru.lock);
- if (page->flags & SGX_EPC_PAGE_RECLAIMER_TRACKED) {
- /* The page is being reclaimed. */
- if (list_empty(&page->list)) {
- spin_unlock(&sgx_global_lru.lock);
- return -EBUSY;
- }
-
- list_del(&page->list);
- page->flags &= ~SGX_EPC_PAGE_RECLAIMER_TRACKED;
+ if (sgx_epc_page_reclaim_in_progress(page->flags)) {
+ spin_unlock(&sgx_global_lru.lock);
+ return -EBUSY;
}
+
+ list_del(&page->list);
+ sgx_epc_page_reset_state(page);
I want to know how much if this series is basically line-for-line
abstraction shifting like:
- page->flags &= ~SGX_EPC_PAGE_RECLAIMER_TRACKED;
+ sgx_epc_page_reset_state(page);
versus actually adding complexity. That way, I might be able to offer
some advice on where this can be pared down. That's really hard to do
with the current series.
Please don't just "introduce new page states". This should have first
abstracted out the sgx_epc_page_reclaim_in_progress() operation, using
the list_empty() check as the implementation.
Then, in a separate patch, introduce the concept of the "reclaim in
progress" flag and finally flip the implementation over.
Ditto for the sgx_epc_page_reset_state() abstraction. It should have
been introduced separately as a concept and then have the implementation
changed.
On in to patch 10 (which is much too big) which introduces the
sgx_lru_list() abstraction.
Sure. I'll try to refactor according to this plan.
Thanks
Haitao
