Drivers: hv: vmbus: finally fix hv_need_to_signal_on_read()
authorDexuan Cui <decui@microsoft.com>
Sat, 28 Jan 2017 18:46:02 +0000 (11:46 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 31 Jan 2017 09:59:48 +0000 (10:59 +0100)
commit433e19cf33d34bb6751c874a9c00980552fe508c
treece6547ef2987fbb289fa28f03536328a42781651
parent191e885a2e130e639bb0c8ee350d7047294f2ce6
Drivers: hv: vmbus: finally fix hv_need_to_signal_on_read()

Commit a389fcfd2cb5 ("Drivers: hv: vmbus: Fix signaling logic in
hv_need_to_signal_on_read()")
added the proper mb(), but removed the test "prev_write_sz < pending_sz"
when making the signal decision.

As a result, the guest can signal the host unnecessarily,
and then the host can throttle the guest because the host
thinks the guest is buggy or malicious; finally the user
running stress test can perceive intermittent freeze of
the guest.

This patch brings back the test, and properly handles the
in-place consumption APIs used by NetVSC (see get_next_pkt_raw(),
put_pkt_raw() and commit_rd_index()).

Fixes: a389fcfd2cb5 ("Drivers: hv: vmbus: Fix signaling logic in
hv_need_to_signal_on_read()")

Signed-off-by: Dexuan Cui <decui@microsoft.com>
Reported-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
Tested-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/hv/ring_buffer.c
drivers/net/hyperv/netvsc.c
include/linux/hyperv.h