There is an off-by-one bug here, which can cause host-to-guest write to stall.
When cur_write_sz == pending_sz, we shouldn't signal the host because it's
meaningless: the ring mustn't be 100% full.
But when cached_write_sz == pending_sz, we must signal the host.
Fixes: 433e19cf33d3 ("Drivers: hv: vmbus: finally fix
hv_need_to_signal_on_read()")
Signed-off-by: John Starks <[email protected]>
Signed-off-by: Dexuan Cui <[email protected]>
Cc: Haiyang Zhang <[email protected]>
Cc: Stephen Hemminger <[email protected]>
Cc: "K. Y. Srinivasan" <[email protected]>
Cc: <[email protected]>
---
include/linux/hyperv.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/include/linux/hyperv.h b/include/linux/hyperv.h
index ba93b7e..246bc6b 100644
--- a/include/linux/hyperv.h
+++ b/include/linux/hyperv.h
@@ -1501,11 +1501,11 @@ static inline void hv_signal_on_read(struct vmbus_channel *channel)
cur_write_sz = hv_get_bytes_to_write(rbi);
- if (cur_write_sz < pending_sz)
+ if (cur_write_sz <= pending_sz)
return;
cached_write_sz = hv_get_cached_bytes_to_write(rbi);
- if (cached_write_sz < pending_sz)
+ if (cached_write_sz <= pending_sz)
vmbus_setevent(channel);
return;
--
2.7.4