[PATCH 3.16 057/114] IB/security: Restrict use of the write() interface

[Ben Hutchings]

3.16.36-rc1 review patch.  If anyone has any objections, please let me know.

------------------

From: Jason Gunthorpe <jgunthorpe@xxxxxxxxxxxxxxxxxxxx>

commit e6bd18f57aad1a2d1ef40e646d03ed0f2515c9e3 upstream.

The drivers/infiniband stack uses write() as a replacement for
bi-directional ioctl().  This is not safe. There are ways to
trigger write calls that result in the return structure that
is normally written to user space being shunted off to user
specified kernel memory instead.

For the immediate repair, detect and deny suspicious accesses to
the write API.

For long term, update the user space libraries and the kernel API
to something that doesn't present the same security vulnerabilities
(likely a structured ioctl() interface).

The impacted uAPI interfaces are generally only available if
hardware from drivers/infiniband is installed in the system.

Reported-by: Jann Horn <jann@xxxxxxxxx>
Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Signed-off-by: Jason Gunthorpe <jgunthorpe@xxxxxxxxxxxxxxxxxxxx>
[ Expanded check to all known write() entry points ]
Signed-off-by: Doug Ledford <dledford@xxxxxxxxxx>
[bwh: Backported to 3.2:
 - Drop changes to hfi1
 - ipath_write() has the same problem, so add the same restriction there]
Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx>
---
--- a/drivers/infiniband/core/ucm.c
+++ b/drivers/infiniband/core/ucm.c
@@ -48,6 +48,7 @@
 
 #include <asm/uaccess.h>
 
+#include <rdma/ib.h>
 #include <rdma/ib_cm.h>
 #include <rdma/ib_user_cm.h>
 #include <rdma/ib_marshall.h>
@@ -1104,6 +1105,9 @@ static ssize_t ib_ucm_write(struct file
 	struct ib_ucm_cmd_hdr hdr;
 	ssize_t result;
 
+	if (WARN_ON_ONCE(!ib_safe_file_access(filp)))
+		return -EACCES;
+
 	if (len < sizeof(hdr))
 		return -EINVAL;
 
--- a/drivers/infiniband/core/ucma.c
+++ b/drivers/infiniband/core/ucma.c
@@ -1487,6 +1487,9 @@ static ssize_t ucma_write(struct file *f
 	struct rdma_ucm_cmd_hdr hdr;
 	ssize_t ret;
 
+	if (WARN_ON_ONCE(!ib_safe_file_access(filp)))
+		return -EACCES;
+
 	if (len < sizeof(hdr))
 		return -EINVAL;
 
--- a/drivers/infiniband/core/uverbs_main.c
+++ b/drivers/infiniband/core/uverbs_main.c
@@ -48,6 +48,8 @@
 
 #include <asm/uaccess.h>
 
+#include <rdma/ib.h>
+
 #include "uverbs.h"
 
 MODULE_AUTHOR("Roland Dreier");
@@ -605,6 +607,9 @@ static ssize_t ib_uverbs_write(struct fi
 	struct ib_uverbs_cmd_hdr hdr;
 	__u32 flags;
 
+	if (WARN_ON_ONCE(!ib_safe_file_access(filp)))
+		return -EACCES;
+
 	if (count < sizeof hdr)
 		return -EINVAL;
 
--- a/drivers/infiniband/hw/ipath/ipath_file_ops.c
+++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c
@@ -45,6 +45,8 @@
 #include <linux/cpu.h>
 #include <asm/pgtable.h>
 
+#include <rdma/ib.h>
+
 #include "ipath_kernel.h"
 #include "ipath_common.h"
 #include "ipath_user_sdma.h"
@@ -2240,6 +2242,9 @@ static ssize_t ipath_write(struct file *
 	ssize_t ret = 0;
 	void *dest;
 
+	if (WARN_ON_ONCE(!ib_safe_file_access(fp)))
+		return -EACCES;
+
 	if (count < sizeof(cmd.type)) {
 		ret = -EINVAL;
 		goto bail;
--- a/drivers/infiniband/hw/qib/qib_file_ops.c
+++ b/drivers/infiniband/hw/qib/qib_file_ops.c
@@ -45,6 +45,8 @@
 #include <linux/delay.h>
 #include <linux/export.h>
 
+#include <rdma/ib.h>
+
 #include "qib.h"
 #include "qib_common.h"
 #include "qib_user_sdma.h"
@@ -2058,6 +2060,9 @@ static ssize_t qib_write(struct file *fp
 	ssize_t ret = 0;
 	void *dest;
 
+	if (WARN_ON_ONCE(!ib_safe_file_access(fp)))
+		return -EACCES;
+
 	if (count < sizeof(cmd.type)) {
 		ret = -EINVAL;
 		goto bail;
--- a/include/rdma/ib.h
+++ b/include/rdma/ib.h
@@ -34,6 +34,7 @@
 #define _RDMA_IB_H
 
 #include <linux/types.h>
+#include <linux/sched.h>
 
 struct ib_addr {
 	union {
@@ -86,4 +87,19 @@ struct sockaddr_ib {
 	__u64			sib_scope_id;
 };
 
+/*
+ * The IB interfaces that use write() as bi-directional ioctl() are
+ * fundamentally unsafe, since there are lots of ways to trigger "write()"
+ * calls from various contexts with elevated privileges. That includes the
+ * traditional suid executable error message writes, but also various kernel
+ * interfaces that can write to file descriptors.
+ *
+ * This function provides protection for the legacy API by restricting the
+ * calling context.
+ */
+static inline bool ib_safe_file_access(struct file *filp)
+{
+	return filp->f_cred == current_cred() && segment_eq(get_fs(), USER_DS);
+}
+
 #endif /* _RDMA_IB_H */