qemu-kvm/kvm-block-linux-aio-bound-ioq_submit-recursion-depth.patch

From 9a6c4bad7f575826796a4c690a0fa6bbcda1f5ad Mon Sep 17 00:00:00 2001
From: "Denis V. Lunev" <den@openvz.org>
Date: Sat, 13 Jun 2026 23:03:08 +0300
Subject: [PATCH 32/52] block/linux-aio: bound ioq_submit() recursion depth

RH-Author: Kevin Wolf <kwolf@redhat.com>
RH-MergeRequest: 504: virt-storage: Backport stable branch fixes
RH-Jira: RHEL-186384
RH-Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
RH-Acked-by: Miroslav Rezanina <mrezanin@redhat.com>
RH-Commit: [25/45] 04545c714dacb2571f169b3da3c7cf493dea31bd (kmwolf/centos-qemu-kvm)

qemu_laio_process_completions() wraps its body in defer_call_begin /
defer_call_end. Inside the section, completion callbacks wake coroutines
that queue new aiocbs; laio_do_submit() defers laio_deferred_fn. At the
bottom of qemu_laio_process_completions() the defer_call_end() fires
laio_deferred_fn, which calls ioq_submit(), closing the cycle:

  ioq_submit
    -> io_submit(2)                           // some sync completions
    -> qemu_laio_process_completions          // defer_call_begin
         -> aio_co_wake                       // resumes coroutine
              -> laio_do_submit
                   -> defer_call(laio_deferred_fn, s)   // enqueued
         -> defer_call_end                    // nesting drops to 0
              -> laio_deferred_fn
                   -> ioq_submit              // +1 stack frame, loop

When io_submit(2) returns asynchronously (O_DIRECT) the cycle
terminates in one extra frame: the fresh aiocb is still in flight, no
completion is drained, no coroutine wakes, no new submission queues.
When submissions complete synchronously (non-O_DIRECT, or per-descriptor
drivers such as vmdk) each level enqueues more work for the next
defer_call_end() to drain, so recursion grows without bound and QEMU
crashes with SIGSEGV on the thread guard page.

The cycle was closed by two performance commits, each correct in
isolation:

  076682885d ("block/linux-aio: convert to blk_io_plug_call() API")
    -- introduced laio_deferred_fn and wired
       laio_do_submit -> defer_call(laio_deferred_fn, s).

  84d61e5f36 ("virtio: use defer_call() in virtio_irqfd_notify()")
    -- added defer_call_begin/end around qemu_laio_process_completions
       so virtio-irqfd notifications batch across a completion pass.

The supported aio=native + cache=none pairing keeps submissions
asynchronous, so the cycle stays bounded; nothing in the code enforces
that contract. Observed in production as a SIGSEGV during a backup job
configured with --cached + aio=native; reproducible on upstream with
qemu-io against vmdk.

Cap ioq_submit() recursion with a counter on LaioQueue, which is only
accessed from the AioContext home thread. On overflow, return without
submitting. The pending work is drained by s->completion_bh, which
qemu_laio_process_completions() has already scheduled on entry -- no
work is lost; one event-loop round-trip of latency is paid only when
the bound is hit, which cannot happen on a supported configuration.

Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Kevin Wolf <kwolf@redhat.com>
CC: Hanna Reitz <hreitz@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20260520142503.251959-2-den@openvz.org>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit 6864bec553b2e37699739615e604fc3c7bae0e1d)
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>

Message-ID: <20260613200411.1808021-25-mjt@tls.msk.ru>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
---
 block/linux-aio.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/block/linux-aio.c b/block/linux-aio.c
index 84397de54c..37de9b564b 100644
--- a/block/linux-aio.c
+++ b/block/linux-aio.c
@@ -36,6 +36,19 @@
 /* Maximum number of requests in a batch. (default value) */
 #define DEFAULT_MAX_BATCH 32
 
+/*
+ * Bound on how deep ioq_submit() may recurse on a single LaioQueue via the
+ * ioq_submit -> qemu_laio_process_completions -> defer_call_end ->
+ * laio_deferred_fn -> ioq_submit cycle. The cycle terminates naturally
+ * when io_submit(2) returns asynchronously (O_DIRECT), but can grow
+ * without bound when submissions complete synchronously. On overflow
+ * the caller returns without submitting; the outermost
+ * qemu_laio_process_completions() has already scheduled s->completion_bh
+ * (via qemu_bh_schedule() at the top of that function), which resumes
+ * submission from the next event-loop dispatch.
+ */
+#define IOQ_SUBMIT_MAX_DEPTH 8
+
 struct qemu_laiocb {
     Coroutine *co;
     LinuxAioState *ctx;
@@ -61,6 +74,7 @@ typedef struct {
     unsigned int in_queue;
     unsigned int in_flight;
     bool blocked;
+    unsigned int submit_depth;
     QSIMPLEQ_HEAD(, qemu_laiocb) pending;
 } LaioQueue;
 
@@ -331,6 +345,7 @@ static void ioq_init(LaioQueue *io_q)
     io_q->in_queue = 0;
     io_q->in_flight = 0;
     io_q->blocked = false;
+    io_q->submit_depth = 0;
 }
 
 static void ioq_submit(LinuxAioState *s)
@@ -340,6 +355,11 @@ static void ioq_submit(LinuxAioState *s)
     QEMU_UNINITIALIZED struct iocb *iocbs[MAX_EVENTS];
     QSIMPLEQ_HEAD(, qemu_laiocb) completed;
 
+    if (s->io_q.submit_depth >= IOQ_SUBMIT_MAX_DEPTH) {
+        return;
+    }
+    s->io_q.submit_depth++;
+
     do {
         if (s->io_q.in_flight >= MAX_EVENTS) {
             break;
@@ -385,6 +405,8 @@ static void ioq_submit(LinuxAioState *s)
          * pended requests will be submitted from there.
          */
     }
+
+    s->io_q.submit_depth--;
 }
 
 static uint64_t laio_max_batch(LinuxAioState *s, uint64_t dev_max_batch)
-- 
2.52.0