Drivers: hv: kvp: switch to using the hvutil_device_state state machine
authorVitaly Kuznetsov <vkuznets@redhat.com>
Sun, 12 Apr 2015 01:07:47 +0000 (18:07 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 24 May 2015 19:17:41 +0000 (12:17 -0700)
commit97bf16cd309805ebf82ffcc4063a65e06169651f
tree443d0899e71a29df4bcf58da958c714944fdb45f
parent636c88da6df3bb2f978b48d3a7ed55423da84d19
Drivers: hv: kvp: switch to using the hvutil_device_state state machine

Switch to using the hvutil_device_state state machine from using 2 different state variables: kvp_transaction.active and
in_hand_shake.

State transitions are:
-> HVUTIL_DEVICE_INIT when driver loads or on device release
-> HVUTIL_READY if the handshake was successful
-> HVUTIL_HOSTMSG_RECEIVED when there is a non-negotiation message from the host
-> HVUTIL_USERSPACE_REQ after we sent the message to the userspace daemon
   -> HVUTIL_USERSPACE_RECV after/if the userspace daemon has replied
-> HVUTIL_READY after we respond to the host
-> HVUTIL_DEVICE_DYING on driver unload

In hv_kvp_onchannelcallback() process ICMSGTYPE_NEGOTIATE messages even when
the userspace daemon is disconnected, otherwise we can make the host think
we don't support KVP and disable the service completely.

Unfortunately there is no good way we can figure out that the userspace daemon
has died (unless we start treating all timeouts as such). In case the daemon
restarts we skip the negotiation procedure (so the daemon is supposed to has
the same version). This behavior is unchanged from in_handshake approach.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Tested-by: Alex Ng <alexng@microsoft.com>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/hv/hv_kvp.c