virtio_blk_remove() resets the virtio device first and unregisters the block device afterwards, but that second step is what writes barebox' block cache back: blockdevice_unregister() calls writebuffer_flush(), which issues ->write() for every dirty chunk and then ->flush(). Both go to a device that no longer has a driver-owned virtqueue, so nothing answers them.
Booting an OS makes this visible, as ExitBootServices() shuts the devices down: EFI stub: Exiting boot services... ERROR: virtio_blk virtio0: I/O error: type 4 sector 0 count 0 status 1 ret Connection timed out Here it is only the flush that is lost, but a dirty chunk - a variable file written just before booting, say - would be dropped just as quietly. Reset after unregistering, like Linux' virtblk_remove() does around del_gendisk(). Assisted-by: Claude:opus-5 Signed-off-by: Ahmad Fatoum <[email protected]> --- drivers/block/virtio_blk.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c index f1ee76da9750..a82d6774c385 100644 --- a/drivers/block/virtio_blk.c +++ b/drivers/block/virtio_blk.c @@ -115,8 +115,8 @@ static void virtio_blk_remove(struct virtio_device *vdev) { struct virtio_blk_priv *priv = vdev->priv; - vdev->config->reset(vdev); blockdevice_unregister(&priv->blk); + vdev->config->reset(vdev); vdev->config->del_vqs(vdev); free(priv); -- 2.47.3
