Re: [PATCH v2 2/2] misc: dice: Add driver to forward secrets to userspace

On Thu, Dec 09, 2021 at 04:31:53PM +0100, Greg Kroah-Hartman wrote:

On Thu, Dec 09, 2021 at 03:11:23PM +0000, David Brazdil wrote:

quoted

Open Profile for DICE is a protocol for deriving unique secrets at boot
used by some Android devices. The firmware/bootloader hands over secrets
in a reserved memory region, which this driver takes ownership of and
exposes it to userspace via a misc device.

Userspace obtains the region's size using an ioctl and mmaps the memory
to its address space. This mapping cannot be write+shared, giving
userspace a guarantee that the secrets have not been overwritten by
another process.

Userspace can also issue an ioctl requesting that the memory be wiped by
the driver. Because both the kernel and userspace mappings use
write-combine semantics, all clients will observe the memory as zeroed
after the ioctl has returned.

Cc: Andrew Scull <redacted>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: David Brazdil <redacted>
---
 .../userspace-api/ioctl/ioctl-number.rst      |   1 +
 drivers/misc/Kconfig                          |   8 +
 drivers/misc/Makefile                         |   1 +
 drivers/misc/dice.c                           | 161 ++++++++++++++++++

Nice, almost 100 lines shorter than before!

Much better, thanks for the changes, but it can be made simpler, see
comments below:

Yep, thanks for all the tips!

quoted

 include/uapi/linux/dice.h                     |  14 ++
 5 files changed, 185 insertions(+)
 create mode 100644 drivers/misc/dice.c
 create mode 100644 include/uapi/linux/dice.h

diff --git a/Documentation/userspace-api/ioctl/ioctl-number.rst b/Documentation/userspace-api/ioctl/ioctl-number.rst
index cfe6cccf0f44..4b8bee2ffd1e 100644
--- a/Documentation/userspace-api/ioctl/ioctl-number.rst
+++ b/Documentation/userspace-api/ioctl/ioctl-number.rst

@@ -341,6 +341,7 @@ Code  Seq#    Include File                                           Comments
 0xAE  40-FF  linux/kvm.h                                             Kernel-based Virtual Machine
                                                                      <mailto:kvm@vger.kernel.org>
 0xAE  20-3F  linux/nitro_enclaves.h                                  Nitro Enclaves
+0xAE  40-5F  uapi/linux/dice.h                                       Open Profile for DICE driver

Why the huge range?  You are only really using 40 and 41.  Stick to that
please.

I understood the comments at the top of this file as encouraging devs to
err on the side of caution and be more liberal with the reservations in
case more ioctls are needed in the future. But I agree that it is highly
unlikely this driver will ever need 32 of them.

quoted

 0xAF  00-1F  linux/fsl_hypervisor.h                                  Freescale hypervisor
 0xB0  all                                                            RATIO devices in development:
                                                                      <mailto:vgo@ratio.de>

diff --git a/drivers/misc/Kconfig b/drivers/misc/Kconfig
index 0f5a49fc7c9e..7165f4b6c41b 100644
--- a/drivers/misc/Kconfig
+++ b/drivers/misc/Kconfig

@@ -470,6 +470,14 @@ config HISI_HIKEY_USB
 	  switching between the dual-role USB-C port and the USB-A host ports
 	  using only one USB controller.

+config DICE
+	tristate "Open Profile for DICE driver"
+	depends on OF_RESERVED_MEM
+	help
+	  This driver allows to ownership of a reserved memory region
+	  containing DICE secrets and expose them to userspace via
+	  a character device.

What is the module name, please add that here.

And "dice" is a very generic name.  I don't mind, but if you want to
name it a bit more specific, that might be better.

Does "open-dice" sound good? I think that's the shorthand used on the
official website.

quoted

+
 source "drivers/misc/c2port/Kconfig"
 source "drivers/misc/eeprom/Kconfig"
 source "drivers/misc/cb710/Kconfig"

diff --git a/drivers/misc/Makefile b/drivers/misc/Makefile
index a086197af544..f73c6bb23ccd 100644
--- a/drivers/misc/Makefile
+++ b/drivers/misc/Makefile

@@ -59,3 +59,4 @@ obj-$(CONFIG_UACCE)		+= uacce/
 obj-$(CONFIG_XILINX_SDFEC)	+= xilinx_sdfec.o
 obj-$(CONFIG_HISI_HIKEY_USB)	+= hisi_hikey_usb.o
 obj-$(CONFIG_HI6421V600_IRQ)	+= hi6421v600-irq.o
+obj-$(CONFIG_DICE)		+= dice.o

diff --git a/drivers/misc/dice.c b/drivers/misc/dice.c
new file mode 100644
index 000000000000..06f3754feb71
--- /dev/null
+++ b/drivers/misc/dice.c

@@ -0,0 +1,161 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2021 - Google LLC
+ * Author: David Brazdil <dbrazdil@google.com>
+ *
+ * Driver for Open Profile for DICE.
+ *
+ * This driver takes ownership of a reserved memory region containing secrets
+ * derived following the Open Profile for DICE. The contents of the memory
+ * region are not interpreted by the kernel but can be mapped into a userspace
+ * process via a misc device. The memory region can also be wiped, removing
+ * the secrets from memory.
+ *
+ * Userspace can access the data by (w/o error handling):
+ *
+ *     int fd = open("/dev/dice", O_RDONLY | O_CLOEXEC);
+ *     size_t size = ioctl(fd, DICE_GET_SIZE);
+ *     void *data = mmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
+ *     ioctl(fd, DICE_WIPE);
+ *     close(fd);
+ */
+
+#include <linux/dice.h>
+#include <linux/io.h>
+#include <linux/miscdevice.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/of_reserved_mem.h>
+#include <linux/platform_device.h>
+
+static int dice_mmap(struct file *filp, struct vm_area_struct *vma);
+static long dice_ioctl(struct file *filp, unsigned int cmd, unsigned long arg);
+
+static const struct file_operations dice_fops = {
+	.mmap = dice_mmap,
+	.unlocked_ioctl = dice_ioctl,
+};
+
+static struct miscdevice dice_misc = {
+	.name	= "dice",
+	.minor	= MISC_DYNAMIC_MINOR,
+	.fops	= &dice_fops,
+	.mode	= 0400,
+};
+
+static struct reserved_mem *dice_rmem;
+static DEFINE_SPINLOCK(dice_lock);

These should be per-device, not global, right?  Please put them in a
local structure that lives off of the platform device.  Then everything
is dynamic and nothing is static making it all much simpler again.

Do that and you can get rid of the pre-definitions of dice_mmap() and
dice_ioctl() above as well.0

Ack

quoted

+
+static int dice_mmap(struct file *filp, struct vm_area_struct *vma)
+{
+	/* Do not allow userspace to modify the underlying data. */
+	if ((vma->vm_flags & VM_WRITE) && (vma->vm_flags & VM_SHARED))
+		return -EPERM;
+
+	/* Create write-combine mapping so all clients observe a wipe. */
+	vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
+	vma->vm_flags |= VM_DONTCOPY | VM_DONTDUMP;
+	return vm_iomap_memory(vma, dice_rmem->base, dice_rmem->size);
+}
+
+static int dice_wipe(void)
+{
+	void *kaddr;
+
+	spin_lock(&dice_lock);
+	kaddr = devm_memremap(dice_misc.this_device, dice_rmem->base,
+			      dice_rmem->size, MEMREMAP_WC);
+	if (IS_ERR(kaddr)) {
+		spin_unlock(&dice_lock);
+		return PTR_ERR(kaddr);
+	}
+
+	memzero_explicit(kaddr, dice_rmem->size);
+	devm_memunmap(dice_misc.this_device, kaddr);

Do you really need to call memzero_explicit()?  This isn't "local"
memory, if the compiler "optimizes away" a normal call, it would be
_VERY_ broken.

I agree that it's not needed here. Seemed in line with other code that
handles secrets but i'm happy to just use memset.

quoted

+	spin_unlock(&dice_lock);
+	return 0;
+}
+
+static long dice_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
+{
+	switch (cmd) {
+	case DICE_GET_SIZE:
+		/* Checked against INT_MAX in dice_probe(). */
+		return dice_rmem->size;
+	case DICE_WIPE:
+		return dice_wipe();
+	}
+
+	return -ENOIOCTLCMD;

-ENOTTY please.

I have no personal attachment to ENOIOCTLCMD, but it is documented as
"no ioctl command" and converted to ENOTTY before returning to userspace.
That made me think this was the right thing to do.

As you only have 2 ioctls, why not just use read/write for this?  Write
would cause dice_wipe() to happen, and read would return the size in the
buffer provided.  Then no ioctl is needed at all.

Fine by me but does feel like a bit of a hack. Is that a common pattern?

quoted

+}
+
+static int __init dice_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct device_node *rmem_np;
+	struct reserved_mem *rmem;
+	int ret;
+
+	if (dice_rmem) {
+		dev_err(dev, "only one instance of device allowed\n");

Why?  That's a device-specific thing, not a driver-specific thing.  Make
this all dynamic and you don't need to care about it at all.

Again, simplification :)

thanks,

greg k-h