39a0a891d8
- vdpa: move memory listener to the realize stage - vdpa: implement vdpa device migration - vhost: implement post resume bh - vhost: implement migration state notifier for vdpa device - vhost: implement savevm_hanlder for vdpa device - vhost: implement vhost_vdpa_device_suspend/resume - vhost: implement vhost-vdpa suspend/resume - vhost: add vhost_dev_suspend/resume_op - vhost: introduce bytemap for vhost backend logging - vhost-vdpa: add migration log ops for VhostOps - vhost-vdpa: add VHOST_BACKEND_F_BYTEMAPLOG - vhost: fix null pointer access - ui/gtk: prevent ui lock up when dpy_gl_update called again before current draw event occurs - hw/usb: dev-mtp: Use g_mkdir() - target/ppc/cpu-models: Remove the "default" CPU alias - hw/intc/arm_gicv3: ICC_PMR_EL1 high bits should be RAZ - Fixed a QEMU hang when guest poweroff in COLO mode - migration/colo: More accurate update checkpoint time - scripts/entitlement.sh: Use backward-compatible cp flags - block/nvme: fix infinite loop in nvme_free_req_queue_cb() - hw/net: npcm7xx_emc fix missing queue_flush - Add dummy Aspeed AST2600 Display Port MCU (DPMCU) - hw/arm: ast2600: Fix address mapping of second SPI controller - vhost-user-blk: reconnect on any error during realize - vhost-user-blk: propagate error return from generic vhost - hw/riscv: boot: Reduce FDT address alignment constraints - Revert "hw/virtio/virtio-iommu-pci: Enforce the device is plugged on the root bus" - Fix STM32F2XX USART data register readout - block: use 'unsigned' for in_flight field on driver state - sphinx: change default language to 'en' - tests/qtest: Fix two format strings - trivial typos: namesapce - hw/ide/atapi.c: Correct typos (CD-CDROM -> CD-ROM) - hw/virtio/virtio-iommu-pci: Enforce the device is plugged on the root bus - hw/display/next-fb: Fix comment typo - xen/pass-through: merge emulated bits correctly mainline inclusion commit be9c61da9fc57eb7d293f380d0805ca6f46c2657 category: bugfix - tests/qtest/migration-test.c: spelling fix: bandwith - target/i386/cpu: Improve error message for property "vendor" - balloon: Fix a misleading error message - target/arm: Don't set syndrome ISS for loads and stores with writeback mainline inclusion commit 53ae2fdef1f5661cbaa2ea571c517f98e6041cb8 category: bugfix - disas/hppa: Show hexcode of instruction along with disassembly - tcg/loongarch64: Fix tcg_out_mov() Aborted - ui/qmp-cmds: Improve two error messages - qga: Improve guest-exec-status error message - hmp: Improve sync-profile error message - spapr/pci: Correct "does not support hotplugging error messages - xen/pass-through: don't create needless register group mainline inclusion commit c0e86b7624cb9d6db03e0d48cf82659e5b89a6a6 category: bugfix Signed-off-by: Jiabo Feng <fengjiabo1@huawei.com>
65 lines
2.2 KiB
Diff
65 lines
2.2 KiB
Diff
From ba31baabf9ad582c8a256a58123c036b6a70ba15 Mon Sep 17 00:00:00 2001
|
|
From: Luo Yifan <luoyifan_yewu@cmss.chinamobile.com>
|
|
Date: Fri, 1 Dec 2023 10:00:41 +0800
|
|
Subject: [PATCH] block/nvme: fix infinite loop in nvme_free_req_queue_cb()
|
|
MIME-Version: 1.0
|
|
Content-Type: text/plain; charset=UTF-8
|
|
Content-Transfer-Encoding: 8bit
|
|
|
|
cherry picked from commit cf4fbc3030c974fff726756a7ceef8386cdf500b
|
|
|
|
When the request free list is exhausted the coroutine waits on
|
|
q->free_req_queue for the next free request. Whenever a request is
|
|
completed a BH is scheduled to invoke nvme_free_req_queue_cb() and wake
|
|
up waiting coroutines.
|
|
|
|
1. nvme_get_free_req() waits for a free request:
|
|
|
|
while (q->free_req_head == -1) {
|
|
...
|
|
trace_nvme_free_req_queue_wait(q->s, q->index);
|
|
qemu_co_queue_wait(&q->free_req_queue, &q->lock);
|
|
...
|
|
}
|
|
|
|
2. nvme_free_req_queue_cb() wakes up the coroutine:
|
|
|
|
while (qemu_co_enter_next(&q->free_req_queue, &q->lock)) {
|
|
^--- infinite loop when free_req_head == -1
|
|
}
|
|
|
|
nvme_free_req_queue_cb() and the coroutine form an infinite loop when
|
|
q->free_req_head == -1. Fix this by checking q->free_req_head in
|
|
nvme_free_req_queue_cb(). If the free request list is exhausted, don't
|
|
wake waiting coroutines. Eventually an in-flight request will complete
|
|
and the BH will be scheduled again, guaranteeing forward progress.
|
|
|
|
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
|
|
Message-id: 20211208152246.244585-1-stefanha@redhat.com
|
|
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
|
|
Signed-off-by: Luo Yifan <luoyifan_yewu@cmss.chinamobile.com>
|
|
---
|
|
block/nvme.c | 5 +++--
|
|
1 file changed, 3 insertions(+), 2 deletions(-)
|
|
|
|
diff --git a/block/nvme.c b/block/nvme.c
|
|
index e4f336d79c..fa360b9b3c 100644
|
|
--- a/block/nvme.c
|
|
+++ b/block/nvme.c
|
|
@@ -206,8 +206,9 @@ static void nvme_free_req_queue_cb(void *opaque)
|
|
NVMeQueuePair *q = opaque;
|
|
|
|
qemu_mutex_lock(&q->lock);
|
|
- while (qemu_co_enter_next(&q->free_req_queue, &q->lock)) {
|
|
- /* Retry all pending requests */
|
|
+ while (q->free_req_head != -1 &&
|
|
+ qemu_co_enter_next(&q->free_req_queue, &q->lock)) {
|
|
+ /* Retry waiting requests */
|
|
}
|
|
qemu_mutex_unlock(&q->lock);
|
|
}
|
|
--
|
|
2.27.0
|
|
|