On Tue, Jun 20, 2023 at 07:43:35PM +0200, Daniel Wagner wrote:
> On Tue, Jun 20, 2023 at 05:07:43PM +0300, Sagi Grimberg wrote:
> > Hmm... is this hapenning with non-fc as well?
>
> I haven't seen a problem for TCP or RDMA yet but in principle it should also
> exists. I can do some tracing to see if we have also problem thern. Two of the
> udev rule match on the subsystem and the event type.
The only udev rule which gets triggered during blktests execution is this one:
# nvme-fc transport generated events (old-style for compatibility)
ACTION=="change", SUBSYSTEM=="fc", ENV{FC_EVENT}=="nvmediscovery", \
ENV{NVMEFC_HOST_TRADDR}=="*", ENV{NVMEFC_TRADDR}=="*", \
RUN+="@SYSTEMCTL@ --no-block restart nvmf-connect@--device=none\t--transport=fc\t--traddr=$env{NVMEFC_TRADDR}\t--trsvcid=none\t--host-traddr=$env{NVMEFC_HOST_TRADDR}.service"
That explain why blktests didn't get disturbed by the autoconnect rule as this
rule has a match on the fc subsystem.
