netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "Guilherme G. Piccoli" <gpiccoli@canonical.com>
To: netanel@amazon.com, akiyano@amazon.com, netdev@vger.kernel.org
Cc: gtzalik@amazon.com, saeedb@amazon.com, zorik@amazon.com,
	gpiccoli@canonical.com, kernel@gpiccoli.net, gshan@redhat.com,
	gavin.guo@canonical.com, jay.vosburgh@canonical.com,
	pedro.principeza@canonical.com
Subject: [PATCH] net: ena: Add PCI shutdown handler to allow safe kexec
Date: Fri, 20 Mar 2020 09:55:34 -0300	[thread overview]
Message-ID: <20200320125534.28966-1-gpiccoli@canonical.com> (raw)

Currently ENA only provides the PCI remove() handler, used during rmmod
for example. This is not called on shutdown/kexec path; we are potentially
creating a failure scenario on kexec:

(a) Kexec is triggered, no shutdown() / remove() handler is called for ENA;
instead pci_device_shutdown() clears the master bit of the PCI device,
stopping all DMA transactions;

(b) Kexec reboot happens and the device gets enabled again, likely having
its FW with that DMA transaction buffered; then it may trigger the (now
invalid) memory operation in the new kernel, corrupting kernel memory area.

This patch aims to prevent this, by implementing a shutdown() handler
quite similar to the remove() one - the difference being the handling
of the netdev, which is unregistered on remove(), but following the
convention observed in other drivers, it's only detached on shutdown().

This prevents an odd issue in AWS Nitro instances, in which after the 2nd
kexec the next one will fail with an initrd corruption, caused by a wild
DMA write to invalid kernel memory. The lspci output for the adapter
present in my instance is:

00:05.0 Ethernet controller [0200]: Amazon.com, Inc. Elastic Network
Adapter (ENA) [1d0f:ec20]

Suggested-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: Guilherme G. Piccoli <gpiccoli@canonical.com>
---


The idea for this patch came from an informal conversation with my
friend Gavin Shan, based on his past experience with similar issues.
I'd like to thank him for the great suggestion!

As a test metric, I've performed 1000 kexecs with this patch, whereas
without this one, the 3rd kexec failed with initrd corruption. Also,
one test that I've done before writing the patch was just to rmmod
the driver before the kexecs, and it worked fine too.

I suggest we add this patch in stable releases as well.
Thanks in advance for reviews,

Guilherme


 drivers/net/ethernet/amazon/ena/ena_netdev.c | 51 ++++++++++++++++----
 1 file changed, 41 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ethernet/amazon/ena/ena_netdev.c b/drivers/net/ethernet/amazon/ena/ena_netdev.c
index 0b2fd96b93d7..7a5c01ff2ee8 100644
--- a/drivers/net/ethernet/amazon/ena/ena_netdev.c
+++ b/drivers/net/ethernet/amazon/ena/ena_netdev.c
@@ -4325,13 +4325,15 @@ static int ena_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 
 /*****************************************************************************/
 
-/* ena_remove - Device Removal Routine
+/* __ena_shutoff - Helper used in both PCI remove/shutdown routines
  * @pdev: PCI device information struct
+ * @shutdown: Is it a shutdown operation? If false, means it is a removal
  *
- * ena_remove is called by the PCI subsystem to alert the driver
- * that it should release a PCI device.
+ * __ena_shutoff is a helper routine that does the real work on shutdown and
+ * removal paths; the difference between those paths is with regards to whether
+ * dettach or unregister the netdevice.
  */
-static void ena_remove(struct pci_dev *pdev)
+static void __ena_shutoff(struct pci_dev *pdev, bool shutdown)
 {
 	struct ena_adapter *adapter = pci_get_drvdata(pdev);
 	struct ena_com_dev *ena_dev;
@@ -4350,13 +4352,17 @@ static void ena_remove(struct pci_dev *pdev)
 
 	cancel_work_sync(&adapter->reset_task);
 
-	rtnl_lock();
+	rtnl_lock(); /* lock released inside the below if-else block */
 	ena_destroy_device(adapter, true);
-	rtnl_unlock();
-
-	unregister_netdev(netdev);
-
-	free_netdev(netdev);
+	if (shutdown) {
+		netif_device_detach(netdev);
+		dev_close(netdev);
+		rtnl_unlock();
+	} else {
+		rtnl_unlock();
+		unregister_netdev(netdev);
+		free_netdev(netdev);
+	}
 
 	ena_com_rss_destroy(ena_dev);
 
@@ -4371,6 +4377,30 @@ static void ena_remove(struct pci_dev *pdev)
 	vfree(ena_dev);
 }
 
+/* ena_remove - Device Removal Routine
+ * @pdev: PCI device information struct
+ *
+ * ena_remove is called by the PCI subsystem to alert the driver
+ * that it should release a PCI device.
+ */
+
+static void ena_remove(struct pci_dev *pdev)
+{
+	__ena_shutoff(pdev, false);
+}
+
+/* ena_shutdown - Device Shutdown Routine
+ * @pdev: PCI device information struct
+ *
+ * ena_shutdown is called by the PCI subsystem to alert the driver that
+ * a shutdown/reboot (or kexec) is happening and device must be disabled.
+ */
+
+static void ena_shutdown(struct pci_dev *pdev)
+{
+	__ena_shutoff(pdev, true);
+}
+
 #ifdef CONFIG_PM
 /* ena_suspend - PM suspend callback
  * @pdev: PCI device information struct
@@ -4420,6 +4450,7 @@ static struct pci_driver ena_pci_driver = {
 	.id_table	= ena_pci_tbl,
 	.probe		= ena_probe,
 	.remove		= ena_remove,
+	.shutdown	= ena_shutdown,
 #ifdef CONFIG_PM
 	.suspend    = ena_suspend,
 	.resume     = ena_resume,
-- 
2.25.1


             reply	other threads:[~2020-03-20 12:55 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-20 12:55 Guilherme G. Piccoli [this message]
2020-03-24  4:04 ` [PATCH] net: ena: Add PCI shutdown handler to allow safe kexec David Miller
2020-03-24 20:04 ` Bhupesh Sharma
2020-03-25 19:04 ` David Miller
2020-03-25 19:06   ` Guilherme G. Piccoli
2020-03-23  9:05 Jubran, Samih
2020-03-23 14:09 ` Guilherme G. Piccoli

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200320125534.28966-1-gpiccoli@canonical.com \
    --to=gpiccoli@canonical.com \
    --cc=akiyano@amazon.com \
    --cc=gavin.guo@canonical.com \
    --cc=gshan@redhat.com \
    --cc=gtzalik@amazon.com \
    --cc=jay.vosburgh@canonical.com \
    --cc=kernel@gpiccoli.net \
    --cc=netanel@amazon.com \
    --cc=netdev@vger.kernel.org \
    --cc=pedro.principeza@canonical.com \
    --cc=saeedb@amazon.com \
    --cc=zorik@amazon.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).