From: Johannes Berg johannes.berg@intel.com
[ Upstream commit 85e73968a040c642fd38f6cba5b73b61f5d0f052 ]
When creating an external event, the current time needs to be propagated to other participants of a simulation. This is done in the places here where we kick a virtq etc.
However, it must be done for _all_ external events, and that includes making the initial socket connection and later closing it. Call time_travel_propagate_time() to do this before making or closing the socket connection.
Apparently, at least for the initial connection creation, due to the remote side in my use cases using microseconds (rather than nanoseconds), this wasn't a problem yet; only started failing between 5.14-rc1 and 5.15-rc1 (didn't test others much), or possibly depending on the configuration, where more delays happen before the virtio devices are initialized.
Fixes: 88ce64249233 ("um: Implement time-travel=ext") Signed-off-by: Johannes Berg johannes.berg@intel.com Signed-off-by: Richard Weinberger richard@nod.at Signed-off-by: Sasha Levin sashal@kernel.org --- arch/um/drivers/virtio_uml.c | 4 ++++ 1 file changed, 4 insertions(+)
diff --git a/arch/um/drivers/virtio_uml.c b/arch/um/drivers/virtio_uml.c index d11b3d41c3785..d5d768188b3ba 100644 --- a/arch/um/drivers/virtio_uml.c +++ b/arch/um/drivers/virtio_uml.c @@ -1076,6 +1076,8 @@ static void virtio_uml_release_dev(struct device *d) container_of(d, struct virtio_device, dev); struct virtio_uml_device *vu_dev = to_virtio_uml_device(vdev);
+ time_travel_propagate_time(); + /* might not have been opened due to not negotiating the feature */ if (vu_dev->req_fd >= 0) { um_free_irq(VIRTIO_IRQ, vu_dev); @@ -1109,6 +1111,8 @@ static int virtio_uml_probe(struct platform_device *pdev) vu_dev->pdev = pdev; vu_dev->req_fd = -1;
+ time_travel_propagate_time(); + do { rc = os_connect_socket(pdata->socket_path); } while (rc == -EINTR);