diff options
author | Max Laier <mlaier@FreeBSD.org> | 2021-02-24 23:56:16 +0000 |
---|---|---|
committer | Ryan Libby <rlibby@FreeBSD.org> | 2021-02-24 23:56:16 +0000 |
commit | 14b5a3c7d5c034c2a5a487b5e2d0de79c2801a65 (patch) | |
tree | e87e97d334f51a2fa5c6a6b219363a122b4bb350 | |
parent | 9a227a2fd642ec057a0ec70d67d5699d65553294 (diff) | |
download | src-14b5a3c7d5c034c2a5a487b5e2d0de79c2801a65.tar.gz src-14b5a3c7d5c034c2a5a487b5e2d0de79c2801a65.zip |
vm pqbatch: move unmanaged page assert under pagequeue lock
This KASSERT is overzealous because of the following race condition:
1) A managed page which is currently in PQ_LAUNDRY is freed.
vm_page_free_prep calls vm_page_dequeue_deferred()
The page state is:
PQ_LAUNDRY, PGA_DEQUEUE|PGA_ENQUEUED
2) The laundry worker comes around and pick up the page and calls
vm_pageout_defer(m, PQ_LAUNDRY, true) to check if page is still in the
queue. We do a vm_page_astate_load and get
PQ_LAUNDRY, PGA_DEQUEUE|PGA_ENQUEUED
as per above.
3) The laundry worker is pre-empted and another thread allocates our page
from the free pool. For example vm_page_alloc_domain_after calls
vm_page_dequeue() and sets VPO_UNMANAGED because we are allocating for
an OBJT_UNMANAGED object.
The page state is:
PQ_NONE, 0 - VPO_UNMANAGED
4) The laundry worker resumes, and processes vm_pageout_defer based on the
stale astate which leads to a call to vm_page_pqbatch_submit, which will
trip on the KASSERT.
Submitted by: mlaier
Reviewed by: markj, rlibby
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D28563
-rw-r--r-- | sys/vm/vm_page.c | 7 |
1 files changed, 2 insertions, 5 deletions
diff --git a/sys/vm/vm_page.c b/sys/vm/vm_page.c index c36b8cdc5762..20fbbc304490 100644 --- a/sys/vm/vm_page.c +++ b/sys/vm/vm_page.c @@ -3545,9 +3545,8 @@ vm_pqbatch_process_page(struct vm_pagequeue *pq, vm_page_t m, uint8_t queue) counter_u64_add(queue_nops, 1); break; } - KASSERT(old.queue != PQ_NONE || - (old.flags & PGA_QUEUE_STATE_MASK) == 0, - ("%s: page %p has unexpected queue state", __func__, m)); + KASSERT((m->oflags & VPO_UNMANAGED) == 0, + ("%s: page %p is unmanaged", __func__, m)); new = old; if ((old.flags & PGA_DEQUEUE) != 0) { @@ -3594,8 +3593,6 @@ vm_page_pqbatch_submit(vm_page_t m, uint8_t queue) struct vm_pagequeue *pq; int domain; - KASSERT((m->oflags & VPO_UNMANAGED) == 0, - ("page %p is unmanaged", m)); KASSERT(queue < PQ_COUNT, ("invalid queue %d", queue)); domain = vm_page_domain(m); |