JakeHillion/scx-upstream
1
0
Fork 0
mirror of https://github.com/sched-ext/scx.git synced 2024-11-25 04:00:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

Revert "scx_flatcg: Keep cgroup rb nodes stashed"

Browse Source 
This reverts commit 3b7f33ea1b.

I haven't root caused it yet but it's easy to reproduce stall and trigger
the watchdog after the commit - just running stress in multiple cgroups
easily triggers stalls after a couple tens of seconds. Let's revert it for
now.
This commit is contained in:
Tejun Heo 2024-06-19 14:43:04 -10:00
parent 92ca7f385c
commit 7a40059b55
1 changed files with 18 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
scheds/c/scx_flatcg.bpf.c
View File

@ -99,7 +99,6 @@ struct cgv_node {
struct bpf_rb_node rb_node;
__u64 cvtime;
__u64 cgid;
struct bpf_refcount refcount;
};
private(CGV_TREE) struct bpf_spin_lock cgv_tree_lock;
@ -289,17 +288,14 @@ static void cgrp_enqueued(struct cgroup *cgrp, struct fcg_cgrp_ctx *cgc)
}
stash = bpf_map_lookup_elem(&cgv_node_stash, &cgid);
if (!stash || !stash->node) {
if (!stash) {
scx_bpf_error("cgv_node lookup failed for cgid %llu", cgid);
return;
}
cgv_node = bpf_refcount_acquire(stash->node);
/* NULL if the node is already on the rbtree */
cgv_node = bpf_kptr_xchg(&stash->node, NULL);
if (!cgv_node) {
/*
* Node never leaves cgv_node_stash, this should only happen if
* fcg_cgroup_exit deletes the stashed node
*/
stat_inc(FCG_STAT_ENQ_RACE);
return;
}
@ -612,6 +608,7 @@ void BPF_STRUCT_OPS(fcg_cgroup_set_weight, struct cgroup *cgrp, u32 weight)
static bool try_pick_next_cgroup(u64 *cgidp)
{
struct bpf_rb_node *rb_node;
struct cgv_node_stash *stash;
struct cgv_node *cgv_node;
struct fcg_cgrp_ctx *cgc;
struct cgroup *cgrp;
@ -695,6 +692,12 @@ static bool try_pick_next_cgroup(u64 *cgidp)
return true;
out_stash:
stash = bpf_map_lookup_elem(&cgv_node_stash, &cgid);
if (!stash) {
stat_inc(FCG_STAT_PNC_GONE);
goto out_free;
}
/*
* Paired with cmpxchg in cgrp_enqueued(). If they see the following
* transition, they'll enqueue the cgroup. If they are earlier, we'll
@ -707,9 +710,16 @@ out_stash:
bpf_rbtree_add(&cgv_tree, &cgv_node->rb_node, cgv_node_less);
bpf_spin_unlock(&cgv_tree_lock);
stat_inc(FCG_STAT_PNC_RACE);
return false;
} else {
cgv_node = bpf_kptr_xchg(&stash->node, cgv_node);
if (cgv_node) {
scx_bpf_error("unexpected !NULL cgv_node stash");
goto out_free;
}
}
return false;
out_free:
bpf_obj_drop(cgv_node);
return false;