On Mon, 2017-03-13 at 12:28 +0800, Freeman Zhang wrote:
On 3/13/17 12:03 PM, Chang, Cunyin wrote:
> Actually, the error message caused by the function
> spdk_pci_device_detach(), after we detach the device, we try to
> call the function
> rte_eal_pci_probe_one(), but we do not have callback function for
> the dpdk nvme driver at this point, so the error message show up,
> but this function will add the device back to dpdk pci device
> list, then we could probe the device next time when
> we want to use it again. So it has no adverse effects but just one
> error message, but we still trying to fix it now.
Thanks for your reply!
I've noticed it and that is why I think it's kind of misleading. So I
guess we should not focus only on reusing codes but also making it
clearer for both users and developers.
The problem is that DPDK is not exposing a public API to do the
operation we need to do (unload a driver but not remove it entirely
from the list of PCI devices). We're working with the DPDK team to add
a public API to do that while also investigating better ways to work
around the issue that don't leave us with this confusing error message.
All the best!
SPDK mailing list