Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   "Andrea Parri (Microsoft)" <parri.andrea@gmail.com>
To:     linux-kernel@vger.kernel.org
Cc:     kys@microsoft.com, haiyangz@microsoft.com, sthemmin@microsoft.com,
        wei.liu@kernel.org, linux-hyperv@vger.kernel.org,
        mikelley@microsoft.com,
        "Andrea Parri (Microsoft)" <parri.andrea@gmail.com>
Subject: [PATCH v2 3/3] Drivers: hv: vmbus: Check for pending channel interrupts before taking a CPU offline
Date:   Wed, 14 Apr 2021 17:01:18 +0200
Message-Id: <20210414150118.2843-4-parri.andrea@gmail.com>
In-Reply-To: <20210414150118.2843-1-parri.andrea@gmail.com>
References: <20210414150118.2843-1-parri.andrea@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

Check that enough time has passed such that the modify channel message
has been processed before taking a CPU offline.

Signed-off-by: Andrea Parri (Microsoft) <parri.andrea@gmail.com>
---
 drivers/hv/hv.c | 49 +++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/drivers/hv/hv.c b/drivers/hv/hv.c
index 3e6ff83adff42..dc9aa1130b22f 100644
--- a/drivers/hv/hv.c
+++ b/drivers/hv/hv.c
@@ -15,6 +15,7 @@
 #include <linux/hyperv.h>
 #include <linux/random.h>
 #include <linux/clockchips.h>
+#include <linux/delay.h>
 #include <linux/interrupt.h>
 #include <clocksource/hyperv_timer.h>
 #include <asm/mshyperv.h>
@@ -292,6 +293,41 @@ void hv_synic_disable_regs(unsigned int cpu)
 		disable_percpu_irq(vmbus_irq);
 }
 
+#define HV_MAX_TRIES 3
+/*
+ * Scan the event flags page of 'this' CPU looking for any bit that is set.  If we find one
+ * bit set, then wait for a few milliseconds.  Repeat these steps for a maximum of 3 times.
+ * Return 'true', if there is still any set bit after this operation; 'false', otherwise.
+ *
+ * If a bit is set, that means there is a pending channel interrupt.  The expectation is
+ * that the normal interrupt handling mechanism will find and process the channel interrupt
+ * "very soon", and in the process clear the bit.
+ */
+static bool hv_synic_event_pending(void)
+{
+	struct hv_per_cpu_context *hv_cpu = this_cpu_ptr(hv_context.cpu_context);
+	union hv_synic_event_flags *event =
+		(union hv_synic_event_flags *)hv_cpu->synic_event_page + VMBUS_MESSAGE_SINT;
+	unsigned long *recv_int_page = event->flags; /* assumes VMBus version >= VERSION_WIN8 */
+	bool pending;
+	u32 relid;
+	int tries = 0;
+
+retry:
+	pending = false;
+	for_each_set_bit(relid, recv_int_page, HV_EVENT_FLAGS_COUNT) {
+		/* Special case - VMBus channel protocol messages */
+		if (relid == 0)
+			continue;
+		pending = true;
+		break;
+	}
+	if (pending && tries++ < HV_MAX_TRIES) {
+		usleep_range(10000, 20000);
+		goto retry;
+	}
+	return pending;
+}
 
 int hv_synic_cleanup(unsigned int cpu)
 {
@@ -336,6 +372,19 @@ int hv_synic_cleanup(unsigned int cpu)
 	if (channel_found && vmbus_connection.conn_state == CONNECTED)
 		return -EBUSY;
 
+	if (vmbus_proto_version >= VERSION_WIN10_V4_1) {
+		/*
+		 * channel_found == false means that any channels that were previously
+		 * assigned to the CPU have been reassigned elsewhere with a call of
+		 * vmbus_send_modifychannel().  Scan the event flags page looking for
+		 * bits that are set and waiting with a timeout for vmbus_chan_sched()
+		 * to process such bits.  If bits are still set after this operation
+		 * and VMBus is connected, fail the CPU offlining operation.
+		 */
+		if (hv_synic_event_pending() && vmbus_connection.conn_state == CONNECTED)
+			return -EBUSY;
+	}
+
 	hv_stimer_legacy_cleanup(cpu);
 
 	hv_synic_disable_regs(cpu);
-- 
2.25.1