Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   "Andrea Parri (Microsoft)" <parri.andrea@gmail.com>
To:     KY Srinivasan <kys@microsoft.com>,
        Haiyang Zhang <haiyangz@microsoft.com>,
        Stephen Hemminger <sthemmin@microsoft.com>,
        Wei Liu <wei.liu@kernel.org>, Dexuan Cui <decui@microsoft.com>,
        Michael Kelley <mikelley@microsoft.com>,
        Wei Hu <weh@microsoft.com>,
        Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
        Rob Herring <robh@kernel.org>,
        Krzysztof Wilczynski <kw@linux.com>,
        Bjorn Helgaas <bhelgaas@google.com>
Cc:     linux-hyperv@vger.kernel.org, linux-pci@vger.kernel.org,
        linux-kernel@vger.kernel.org,
        "Andrea Parri (Microsoft)" <parri.andrea@gmail.com>
Subject: [PATCH v2 1/6] Drivers: hv: vmbus: Fix handling of messages with transaction ID of zero
Date:   Tue, 19 Apr 2022 14:23:20 +0200
Message-Id: <20220419122325.10078-2-parri.andrea@gmail.com>
In-Reply-To: <20220419122325.10078-1-parri.andrea@gmail.com>
References: <20220419122325.10078-1-parri.andrea@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

vmbus_request_addr() returns 0 (zero) if the transaction ID passed
to as argument is 0.  This is unfortunate for two reasons: first,
netvsc_send_completion() does not check for a NULL cmd_rqst (before
dereferencing the corresponding NVSP message); second, 0 is a *valid*
value of cmd_rqst in netvsc_send_tx_complete(), cf. the call of
vmbus_sendpacket() in netvsc_send_pkt().

vmbus_request_addr() has included the code in question since its
introduction with commit e8b7db38449ac ("Drivers: hv: vmbus: Add
vmbus_requestor data structure for VMBus hardening"); such code was
motivated by the early use of vmbus_requestor by hv_storvsc.  Since
hv_storvsc moved to a tag-based mechanism to generate and retrieve
transaction IDs with commit bf5fd8cae3c8f ("scsi: storvsc: Use
blk_mq_unique_tag() to generate requestIDs"), vmbus_request_addr()
can be modified to return VMBUS_RQST_ERROR if the ID is 0.  This
change solves the issues in hv_netvsc (and makes the handling of
messages with transaction ID of 0 consistent with the semantics
"the ID is not contained in the requestor/invalid ID").

vmbus_next_request_id(), vmbus_request_addr() should still reserve
the ID of 0 for Hyper-V, because Hyper-V will "ignore" (not respond
to) VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED packets/requests with
transaction ID of 0 from the guest.

Fixes: bf5fd8cae3c8f ("scsi: storvsc: Use blk_mq_unique_tag() to generate requestIDs")
Signed-off-by: Andrea Parri (Microsoft) <parri.andrea@gmail.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
---
The above hv_netvsc issues precede bf5fd8cae3c8f; however, these
changes should not be backported to earlier commits since such a
back-port would 'break' hv_storvsc.

 drivers/hv/channel.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/hv/channel.c b/drivers/hv/channel.c
index dc5c35210c16a..20fc8d50a0398 100644
--- a/drivers/hv/channel.c
+++ b/drivers/hv/channel.c
@@ -1245,7 +1245,9 @@ u64 vmbus_next_request_id(struct vmbus_channel *channel, u64 rqst_addr)
 
 	/*
 	 * Cannot return an ID of 0, which is reserved for an unsolicited
-	 * message from Hyper-V.
+	 * message from Hyper-V; Hyper-V does not acknowledge (respond to)
+	 * VMBUS_DATA_PACKET_FLAG_COMPLETION_REQUESTED requests with ID of
+	 * 0 sent by the guest.
 	 */
 	return current_id + 1;
 }
@@ -1270,7 +1272,7 @@ u64 vmbus_request_addr(struct vmbus_channel *channel, u64 trans_id)
 
 	/* Hyper-V can send an unsolicited message with ID of 0 */
 	if (!trans_id)
-		return trans_id;
+		return VMBUS_RQST_ERROR;
 
 	spin_lock_irqsave(&rqstor->req_lock, flags);
 
-- 
2.25.1