mm/damon/core: reset invalid quota->charge_target_from
Subsystems (3)
SJ Park
maintainerRFC PATCH v2 1/1 mm/damon/core: reset invalid quota->charge_target_from
DAMOS can suddenly stop working if a target process that the quota is just fully charged on is terminated. Fix by catching and processing the corner case.
When DAMOS quota is fully charged, the target and the region to continue applying the action in the next round is saved in damos_quota->charge_{target,addr}_from. In the next round, DAMOS iterates targets and regions from the beginning. It skips applying the action to the regions until it visits and skips the saved target/region.
Virtual address space targets become invalid if the process is terminated. Trying to apply the scheme to invalid target is just a waste of time. Hence commit 6e4930e33329 ("mm/damon/core: fix wasteful CPU calls by skipping non-existent targets") made the logic to skip invalid targets. However, it does skip before the charged target/region skipping/updating.
Let's suppose the user runs DAMOS for multiple virtual address spaces with a quota. The quota exceeded in the middle of a virtual address space. And the process of the address space is terminated. Then the charge_target_from points to the invalid target. The pointer update logic is skipped for the invalid target, so the charge_target_from is never updated. DAMOS action to every target/region is skipped. From the user's perspective, it would look like suddenly DAMOS has stopped working.
No critical leak or crash can happen. The user could reinstall the scheme. But this makes use of DAMOS under certain setups quite unreliable.
When the invalid target is found, further check the corner case and reset the pointer.
This issue was discovered [1] by Sashiko.
[1] https://lore.kernel.org/[email protected]
---
Changes from v1
- v1: https://lore.kernel.org/[email protected]
- Move the corner case check to invalid target found case, to handle dynamic quota uninstall case.
- Add RFC tag again.
Changes from RFC v1
- RFC v1: https://lore.kernel.org/[email protected]
- Drop RFC tag.
mm/damon/core.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
mm/damon/core.c +8 -1
@@ -3389,8 +3389,15 @@ static void kdamond_apply_schemes(struct damon_ctx *c) max_region_sz =damon_region_sz_limit (c);mutex_lock (&c->walk_control_lock);damon_for_each_target (t, c) { -if (c->ops.target_valid && c->ops.target_valid (t) == false) +if (c->ops.target_valid && c->ops.target_valid (t) == false) { +damon_for_each_scheme (s, c) { +if (s->quota.charge_target_from != t) +continue ; + s->quota.charge_target_from =NULL ; + s->quota.charge_addr_from =0 ; + }continue ; + }damos_apply_target (c, t, max_region_sz); }
base-commit: 4c152875952628604696dbba38cd487fdfde28ea
Signature · 1 line
2.47.3
[email protected]
AI reviewSashiko has reviewed this patch and found no issues. It looks great!
Quoted 18 lines
DAMOS can suddenly stop working if a target process that the quota is just fully charged on is terminated. Fix by catching and processing the corner case.
When DAMOS quota is fully charged, the target and the region to continue applying the action in the next round is saved in damos_quota->charge_{target,addr}_from. In the next round, DAMOS iterates targets and regions from the beginning. It skips applying the action to the regions until it visits and skips the saved target/region.
Virtual address space targets become invalid if the process is terminated. Trying to apply the scheme to invalid target is just a waste of time. Hence commit
6e4930e33329("mm/damon/core: fix wasteful CPU calls by skipping non-existent targets") made the logic to skip invalid targets. However, it does skip before the charged target/region [ ... ]
[1] https://lore.kernel.org/[email protected]
Quoted 3 lines
Fixes 6e4930e33329 ("mm/damon/core: fix wasteful CPU calls by skipping non-existent targets")Cc <[email protected]> # 7.0.xSigned-off-by SJ Park <[email protected]>
Sashiko has reviewed this patch and found no issues. It looks great!
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1