| 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
| ||
|
Message-ID: <158662721802.1893045.12301414116114602646.stgit@dwillia2-desk3.amr.corp.intel.com>
Date: Sat, 11 Apr 2020 10:51:35 -0700
From: Dan Williams <dan.j.williams@...el.com>
To: gregkh@...uxfoundation.org
Cc: Arnd Bergmann <arnd@...db.de>, Ingo Molnar <mingo@...hat.com>,
Kees Cook <keescook@...omium.org>,
Russell King <linux@....linux.org.uk>,
Andrew Morton <akpm@...ux-foundation.org>,
linux-kernel@...r.kernel.org, linux-nvdimm@...ts.01.org
Subject: [RFC PATCH] /dev/mem: Revoke mappings when a driver claims the
region
Commit 90a545e98126 ("restrict /dev/mem to idle io memory ranges")
introduced CONFIG_IO_STRICT_DEVMEM with the goal of protecting the
kernel against scenarios where a /dev/mem user tramples memory that a
kernel driver owns. However, this protection only prevents *new* read(),
write() and mmap() requests. Established mappings prior to the driver
calling request_mem_region() are left alone.
Especially with persistent memory, and the core kernel metadata that is
stored there, there are plentiful scenarios for a /dev/mem user to
violate the expectations of the driver and cause amplified damage.
Teach request_mem_region() to find and shoot down active /dev/mem
mappings that it believes it has successfully claimed for the exclusive
use of the driver.
Cc: Arnd Bergmann <arnd@...db.de>
Cc: Ingo Molnar <mingo@...hat.com>
Cc: Kees Cook <keescook@...omium.org>
Cc: Russell King <linux@....linux.org.uk>
Cc: Andrew Morton <akpm@...ux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
Fixes: 90a545e98126 ("restrict /dev/mem to idle io memory ranges")
Signed-off-by: Dan Williams <dan.j.williams@...el.com>
---
I marked this RFC because after writing it I realized we potentially
have the same problem with /dev/port, and many mmap drivers in general.
Is there a wider solution I'm missing?
drivers/char/mem.c | 104 +++++++++++++++++++++++++++++++++++++++++++-
include/linux/ioport.h | 6 +++
include/uapi/linux/magic.h | 1
kernel/resource.c | 5 ++
4 files changed, 114 insertions(+), 2 deletions(-)
diff --git a/drivers/char/mem.c b/drivers/char/mem.c
index 43dd0891ca1e..3a3b7b37c5d2 100644
--- a/drivers/char/mem.c
+++ b/drivers/char/mem.c
@@ -31,11 +31,15 @@
#include <linux/uio.h>
#include <linux/uaccess.h>
#include <linux/security.h>
+#include <linux/pseudo_fs.h>
+#include <uapi/linux/magic.h>
+#include <linux/mount.h>
#ifdef CONFIG_IA64
# include <linux/efi.h>
#endif
+#define DEVMEM_MINOR 1
#define DEVPORT_MINOR 4
static inline unsigned long size_inside_page(unsigned long start,
@@ -805,12 +809,66 @@ static loff_t memory_lseek(struct file *file, loff_t offset, int orig)
return ret;
}
+static struct inode *devmem_inode;
+
+#ifdef CONFIG_IO_STRICT_DEVMEM
+void revoke_devmem(struct resource *res)
+{
+ struct inode *inode = READ_ONCE(devmem_inode);
+
+ /*
+ * Check that the initialization has completed. Losing the race
+ * is ok because it means drivers are claiming resources before
+ * the fs_initcall level of init and prevent /dev/mem from
+ * establishing mappings.
+ */
+ smp_rmb();
+ if (!inode)
+ return;
+
+ /*
+ * The expectation is that the driver has successfully marked
+ * the resource busy by this point, so devmem_is_allowed()
+ * should start returning false, however for performance this
+ * does not iterate the entire resource range.
+ */
+ if (devmem_is_allowed(PHYS_PFN(res->start)) &&
+ devmem_is_allowed(PHYS_PFN(res->end))) {
+ /*
+ * *cringe* iomem=relaxed says "go ahead, what's the
+ * worst that can happen?"
+ */
+ return;
+ }
+
+ unmap_mapping_range(inode->i_mapping, res->start, resource_size(res), 1);
+}
+#endif
+
static int open_port(struct inode *inode, struct file *filp)
{
+ int rc;
+
if (!capable(CAP_SYS_RAWIO))
return -EPERM;
- return security_locked_down(LOCKDOWN_DEV_MEM);
+ rc = security_locked_down(LOCKDOWN_DEV_MEM);
+ if (rc)
+ return rc;
+
+ if (iminor(inode) != DEVMEM_MINOR)
+ return 0;
+
+ /*
+ * Use a unified address space to have a single point to manage
+ * revocations when drivers want to take over a /dev/mem mapped
+ * range.
+ */
+ inode->i_mapping = devmem_inode->i_mapping;
+ inode->i_mapping->host = devmem_inode;
+ filp->f_mapping = inode->i_mapping;
+
+ return 0;
}
#define zero_lseek null_lseek
@@ -885,7 +943,7 @@ static const struct memdev {
fmode_t fmode;
} devlist[] = {
#ifdef CONFIG_DEVMEM
- [1] = { "mem", 0, &mem_fops, FMODE_UNSIGNED_OFFSET },
+ [DEVMEM_MINOR] = { "mem", 0, &mem_fops, FMODE_UNSIGNED_OFFSET },
#endif
#ifdef CONFIG_DEVKMEM
[2] = { "kmem", 0, &kmem_fops, FMODE_UNSIGNED_OFFSET },
@@ -939,6 +997,46 @@ static char *mem_devnode(struct device *dev, umode_t *mode)
static struct class *mem_class;
+static int devmem_fs_init_fs_context(struct fs_context *fc)
+{
+ return init_pseudo(fc, DEVMEM_MAGIC) ? 0 : -ENOMEM;
+}
+
+static struct file_system_type devmem_fs_type = {
+ .name = "devmem",
+ .owner = THIS_MODULE,
+ .init_fs_context = devmem_fs_init_fs_context,
+ .kill_sb = kill_anon_super,
+};
+
+static int devmem_init_inode(void)
+{
+ static struct vfsmount *devmem_vfs_mount;
+ static int devmem_fs_cnt;
+ struct inode *inode;
+ int rc;
+
+ rc = simple_pin_fs(&devmem_fs_type, &devmem_vfs_mount, &devmem_fs_cnt);
+ if (rc < 0) {
+ pr_err("Cannot mount /dev/mem pseudo filesystem: %d\n", rc);
+ return rc;
+ }
+
+ inode = alloc_anon_inode(devmem_vfs_mount->mnt_sb);
+ if (IS_ERR(inode)) {
+ rc = PTR_ERR(inode);
+ pr_err("Cannot allocate inode for /dev/mem: %d\n", rc);
+ simple_release_fs(&devmem_vfs_mount, &devmem_fs_cnt);
+ return rc;
+ }
+
+ /* publish /dev/mem initialized */
+ WRITE_ONCE(devmem_inode, inode);
+ smp_wmb();
+
+ return 0;
+}
+
static int __init chr_dev_init(void)
{
int minor;
@@ -960,6 +1058,8 @@ static int __init chr_dev_init(void)
*/
if ((minor == DEVPORT_MINOR) && !arch_has_dev_port())
continue;
+ if ((minor == DEVMEM_MINOR) && devmem_init_inode() != 0)
+ continue;
device_create(mem_class, NULL, MKDEV(MEM_MAJOR, minor),
NULL, devlist[minor].name);
diff --git a/include/linux/ioport.h b/include/linux/ioport.h
index a9b9170b5dd2..6c3eca90cbc4 100644
--- a/include/linux/ioport.h
+++ b/include/linux/ioport.h
@@ -301,5 +301,11 @@ struct resource *devm_request_free_mem_region(struct device *dev,
struct resource *request_free_mem_region(struct resource *base,
unsigned long size, const char *name);
+#ifdef CONFIG_IO_STRICT_DEVMEM
+void revoke_devmem(struct resource *res);
+#else
+static inline void revoke_devmem(struct resource *res) { };
+#endif
+
#endif /* __ASSEMBLY__ */
#endif /* _LINUX_IOPORT_H */
diff --git a/include/uapi/linux/magic.h b/include/uapi/linux/magic.h
index d78064007b17..f3956fc11de6 100644
--- a/include/uapi/linux/magic.h
+++ b/include/uapi/linux/magic.h
@@ -94,6 +94,7 @@
#define BALLOON_KVM_MAGIC 0x13661366
#define ZSMALLOC_MAGIC 0x58295829
#define DMA_BUF_MAGIC 0x444d4142 /* "DMAB" */
+#define DEVMEM_MAGIC 0x454d444d /* "DMEM" */
#define Z3FOLD_MAGIC 0x33
#define PPC_CMM_MAGIC 0xc7571590
diff --git a/kernel/resource.c b/kernel/resource.c
index 76036a41143b..841737bbda9e 100644
--- a/kernel/resource.c
+++ b/kernel/resource.c
@@ -1126,6 +1126,7 @@ struct resource * __request_region(struct resource *parent,
{
DECLARE_WAITQUEUE(wait, current);
struct resource *res = alloc_resource(GFP_KERNEL);
+ struct resource *orig_parent = parent;
if (!res)
return NULL;
@@ -1176,6 +1177,10 @@ struct resource * __request_region(struct resource *parent,
break;
}
write_unlock(&resource_lock);
+
+ if (res && orig_parent == &iomem_resource)
+ revoke_devmem(res);
+
return res;
}
EXPORT_SYMBOL(__request_region);
Powered by blists - more mailing lists