Re: [PATCH v13 11/14] x86/sgx: Abstract check for global reclaimable pages

From: Huang, Kai
Date: Thu May 02 2024 - 19:23:34 EST




On 1/05/2024 7:51 am, Haitao Huang wrote:
From: Kristen Carlson Accardi <kristen@xxxxxxxxxxxxxxx>

For the global reclaimer to determine if any page available for
reclamation at the global level, it currently only checks for emptiness
of the global LRU. That will be inadequate when pages are tracked in
multiple LRUs, one per cgroup. For this purpose, create a new helper,
sgx_can_reclaim_global(), to abstract this check. Currently it only
checks the global LRU, later will check emptiness of LRUs of all cgroups
when per-cgroup tracking is turned on.

Replace all the checks for emptiness of the global LRU,
list_empty(&sgx_global_lru.reclaimable), with calls to
sgx_can_reclaim_global().

Rename sgx_should_reclaim() to sgx_should_reclaim_global() as it is used
to check if a global reclamation should be performed.

Co-developed-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
Signed-off-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
Signed-off-by: Kristen Carlson Accardi <kristen@xxxxxxxxxxxxxxx>
Co-developed-by: Haitao Huang <haitao.huang@xxxxxxxxxxxxxxx>
Signed-off-by: Haitao Huang <haitao.huang@xxxxxxxxxxxxxxx>
Tested-by: Jarkko Sakkinen <jarkko@xxxxxxxxxx>
---

Free free to add:

Reviewed-by: Kai Huang <kai.huang@xxxxxxxxx>

One thing below:

[...]

-static bool sgx_should_reclaim(unsigned long watermark)
+static bool sgx_should_reclaim_global(unsigned long watermark)
{
return atomic_long_read(&sgx_nr_free_pages) < watermark &&
- !list_empty(&sgx_global_lru.reclaimable);
+ sgx_can_reclaim_global();
}
static void sgx_reclaim_pages_global(struct mm_struct *charge_mm)
@@ -405,7 +413,7 @@ static void sgx_reclaim_pages_global(struct mm_struct *charge_mm)
*/
void sgx_reclaim_direct(void)
{
- if (sgx_should_reclaim(SGX_NR_LOW_PAGES))
+ if (sgx_should_reclaim_global(SGX_NR_LOW_PAGES))
sgx_reclaim_pages_global(current->mm);
}

Hmm.. Sorry for not pointing out in the previous version.

Perhaps it makes more sense to do the rename in the patch ...

x86/sgx: Add basic EPC reclamation flow for cgroup

.. where we have actually introduced the concept of per-cgroup reclaim, and we literally have below change in that patch:

void sgx_reclaim_direct(void)
{
if (sgx_should_reclaim(SGX_NR_LOW_PAGES))
- sgx_reclaim_pages();
+ sgx_reclaim_pages_global();
}

So in that patch, the sgx_should_reclaim() literally just means we should do gloabl reclaim, but not the per-cgruop reclaim. Thus, perhaps we just do the renaming here together with the new sgx_reclaim_pages_global().

If there's a new version needed, please move the renaming to that patch?