linux-kernel - [PATCH] vfio: Fix endianness handling for emulated BARs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1403091391-31780-1-git-send-email-aik@ozlabs.ru>
Date:	Wed, 18 Jun 2014 21:36:31 +1000
From:	Alexey Kardashevskiy <aik@...abs.ru>
To:	linuxppc-dev@...ts.ozlabs.org
Cc:	Alexey Kardashevskiy <aik@...abs.ru>,
	Benjamin Herrenschmidt <benh@...nel.crashing.org>,
	Alex Williamson <alex.williamson@...hat.com>,
	kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
	Alexander Graf <agraf@...e.de>,
	Nikunj A Dadhania <nikunj@...ux.vnet.ibm.com>
Subject: [PATCH] vfio: Fix endianness handling for emulated BARs

VFIO exposes BARs to user space as a byte stream so userspace can
read it using pread()/pwrite(). Since this is a byte stream, VFIO should
not do byte swapping and simply return values as it gets them from
PCI device.

Instead, the existing code assumes that byte stream in read/write is
little-endian and it fixes endianness for values which it passes to
ioreadXX/iowriteXX helpers. This works for little-endian as PCI is
little endian and le32_to_cpu/... are stubs.

This also works for big endian but rather by an accident: it reads 4 bytes
from the stream (@val is big endian), converts to CPU format (which should
be big endian) as it was little endian (@val becomes actually little
endian) and calls iowrite32() which does not do swapping on big endian
system.

This removes byte swapping and makes use ioread32be/iowrite32be
(and 16bit versions) on big-endian systems. The "be" helpers take
native endian values and do swapping at the moment of writing to a PCI
register using one of "store byte-reversed" instructions.

Suggested-by: Benjamin Herrenschmidt <benh@...nel.crashing.org>
Signed-off-by: Alexey Kardashevskiy <aik@...abs.ru>
---
 drivers/vfio/pci/vfio_pci_rdwr.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)

diff --git a/drivers/vfio/pci/vfio_pci_rdwr.c b/drivers/vfio/pci/vfio_pci_rdwr.c
index 210db24..f363b5a 100644
--- a/drivers/vfio/pci/vfio_pci_rdwr.c
+++ b/drivers/vfio/pci/vfio_pci_rdwr.c
@@ -21,6 +21,18 @@
 
 #include "vfio_pci_private.h"
 
+#ifdef __BIG_ENDIAN__
+#define ioread16_native		ioread16be
+#define ioread32_native		ioread32be
+#define iowrite16_native	iowrite16be
+#define iowrite32_native	iowrite32be
+#else
+#define ioread16_native		ioread16
+#define ioread32_native		ioread32
+#define iowrite16_native	iowrite16
+#define iowrite32_native	iowrite32
+#endif
+
 /*
  * Read or write from an __iomem region (MMIO or I/O port) with an excluded
  * range which is inaccessible.  The excluded range drops writes and fills
@@ -50,9 +62,9 @@ static ssize_t do_io_rw(void __iomem *io, char __user *buf,
 				if (copy_from_user(&val, buf, 4))
 					return -EFAULT;
 
-				iowrite32(le32_to_cpu(val), io + off);
+				iowrite32_native(val, io + off);
 			} else {
-				val = cpu_to_le32(ioread32(io + off));
+				val = ioread32_native(io + off);
 
 				if (copy_to_user(buf, &val, 4))
 					return -EFAULT;
@@ -66,9 +78,9 @@ static ssize_t do_io_rw(void __iomem *io, char __user *buf,
 				if (copy_from_user(&val, buf, 2))
 					return -EFAULT;
 
-				iowrite16(le16_to_cpu(val), io + off);
+				iowrite16_native(val, io + off);
 			} else {
-				val = cpu_to_le16(ioread16(io + off));
+				val = ioread16_native(io + off);
 
 				if (copy_to_user(buf, &val, 2))
 					return -EFAULT;
-- 
2.0.0

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/