virtio_mem_init_kdump() marks the device DRIVER_OK before it checks for
CONFIG_PROC_VMCORE. Without vmcore support it then fails with -EBUSY,
and virtio_mem_probe() deletes the virtqueue on that error path without
resetting the device first. The device is left with DRIVER_OK set and a
queue that points at freed memory. Nothing has been queued at that
point, so the device has nothing to process, but the driver should not
leave a device it is giving up on marked ready.
Before commit a9403425b3cf ("virtio-mem: mark device ready before
registering callbacks in kdump mode"), virtio_mem_probe() called
virtio_device_ready() only after virtio_mem_init() had succeeded, so this
path never exposed a ready device.
Move virtio_device_ready() into the CONFIG_PROC_VMCORE branch, right
before the vmcore callbacks that need it are registered, so the -EBUSY
path never sets DRIVER_OK.
Fixes: a9403425b3cf ("virtio-mem: mark device ready before registering
callbacks in kdump mode")
Signed-off-by: Yuho Choi <[email protected]>
---
v2:
- Drop Cc: stable, as no practical kdump kernel disables PROC_VMCORE
(David).
- Say in the changelog that nothing is queued on this path.
- Rebased on v7.4-rc1.
v1: https://lore.kernel.org/r/[email protected]
Compile-tested with W=1, with and without CONFIG_PROC_VMCORE.
drivers/virtio/virtio_mem.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/virtio/virtio_mem.c b/drivers/virtio/virtio_mem.c
index e18dd736f2ec..b28d2da7d5b6 100644
--- a/drivers/virtio/virtio_mem.c
+++ b/drivers/virtio/virtio_mem.c
@@ -2817,9 +2817,9 @@ static int virtio_mem_vmcore_get_device_ram(struct
vmcore_cb *cb,
static int virtio_mem_init_kdump(struct virtio_mem *vm)
{
+#ifdef CONFIG_PROC_VMCORE
/* We must be prepared to receive a callback immediately. */
virtio_device_ready(vm->vdev);
-#ifdef CONFIG_PROC_VMCORE
dev_info(&vm->vdev->dev, "memory hot(un)plug disabled in kdump
kernel\n");
vm->vmcore_cb.pfn_is_ram = virtio_mem_vmcore_pfn_is_ram;
#ifdef CONFIG_PROC_VMCORE_DEVICE_RAM
--
2.43.0
