[PATCH v2 0/3] fix SELinux W^X bypass via ptrace

[PATCH v2 0/3] fix SELinux W^X bypass via ptrace

[Jann Horn]

This fixes a bypass of SELinux' W^X protection via ptrace.
For more details, see the commit messages of patches 2/3 and 3/3.

Jann Horn (3):
  fs/exec: don't force writing memory access
  mm: add LSM hook for writes to readonly memory
  selinux: require EXECMEM for forced ptrace poke

 drivers/gpu/drm/etnaviv/etnaviv_gem.c   |  3 +-
 drivers/gpu/drm/i915/i915_gem_userptr.c |  2 +-
 drivers/infiniband/core/umem_odp.c      |  4 +-
 fs/exec.c                               |  4 +-
 fs/proc/base.c                          | 68 +++++++++++++++++++++-------
 fs/proc/internal.h                      |  4 +-
 fs/proc/task_mmu.c                      |  4 +-
 fs/proc/task_nommu.c                    |  2 +-
 include/linux/lsm_hooks.h               |  9 ++++
 include/linux/mm.h                      | 12 ++++-
 include/linux/sched.h                   |  4 +-
 include/linux/security.h                | 10 +++++
 kernel/events/uprobes.c                 |  6 ++-
 kernel/fork.c                           |  6 ++-
 mm/gup.c                                | 80 +++++++++++++++++++++++++--------
 mm/memory.c                             | 22 ++++++---
 mm/nommu.c                              | 22 +++++----
 mm/process_vm_access.c                  |  8 ++--
 security/security.c                     |  8 ++++
 security/selinux/hooks.c                | 15 +++++++
 security/tomoyo/domain.c                |  2 +-
 virt/kvm/async_pf.c                     |  3 +-
 virt/kvm/kvm_main.c                     |  9 ++--
 23 files changed, 230 insertions(+), 77 deletions(-)

-- 
2.1.4

[PATCH v2 1/3] fs/exec: don't force writing memory access

[Jann Horn]

This shouldn't change behavior in any way - at this point, there should be
no non-writable mappings, only the initial stack mapping -, but this change
makes it easier to reason about the correctness of the following commits
that place restrictions on forced memory writes.

Signed-off-by: Jann Horn <jann@thejh.net>
Reviewed-by: Janis Danisevskis <jdanis@android.com>
---
 fs/exec.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/exec.c b/fs/exec.c
index 6fcfb3f..d607da8 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -204,7 +204,7 @@ static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
 	 * doing the exec and bprm->mm is the new process's mm.
 	 */
 	ret = get_user_pages_remote(current, bprm->mm, pos, 1, write,
-			1, &page, NULL);
+			0, &page, NULL);
 	if (ret <= 0)
 		return NULL;
 
-- 
2.1.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH v2 2/3] mm: add LSM hook for writes to readonly memory

[Jann Horn]

SELinux attempts to make it possible to whitelist trustworthy sources of
code that may be mapped into memory, and Android makes use of this feature.
To prevent an attacker from bypassing this by modifying R+X memory through
/proc/$pid/mem or PTRACE_POKETEXT, it is necessary to call a security hook
in check_vma_flags().

The new security hook security_forced_write() takes three arguments:

 - The modified VMA, so the security check can e.g. test for executability.
 - The subject performing the access. For remote accesses, this may be
   different from the target of the access. This can e.g. be used to create
   a security policy that permits a privileged debugger to set software
   breakpoints in the address space of a sandboxed process.
 - The target of the access. This is useful if only a subset of the
   processes on the system should be prevented from executing arbitrary
   code, as is the case on Android.

changed in v2:
 - fix comment (Janis Danisevsk)
 - simplify code a bit (Janis Danisevsk)

Signed-off-by: Jann Horn <jann@thejh.net>
Reviewed-by: Janis Danisevskis <jdanis@android.com>
---
 drivers/gpu/drm/etnaviv/etnaviv_gem.c   |  3 +-
 drivers/gpu/drm/i915/i915_gem_userptr.c |  2 +-
 drivers/infiniband/core/umem_odp.c      |  4 +-
 fs/exec.c                               |  4 +-
 fs/proc/base.c                          | 68 +++++++++++++++++++++-------
 fs/proc/internal.h                      |  4 +-
 fs/proc/task_mmu.c                      |  4 +-
 fs/proc/task_nommu.c                    |  2 +-
 include/linux/lsm_hooks.h               |  9 ++++
 include/linux/mm.h                      | 12 ++++-
 include/linux/sched.h                   |  4 +-
 include/linux/security.h                | 10 +++++
 kernel/events/uprobes.c                 |  6 ++-
 kernel/fork.c                           |  6 ++-
 mm/gup.c                                | 80 +++++++++++++++++++++++++--------
 mm/memory.c                             | 22 ++++++---
 mm/nommu.c                              | 22 +++++----
 mm/process_vm_access.c                  |  8 ++--
 security/security.c                     |  8 ++++
 security/tomoyo/domain.c                |  2 +-
 virt/kvm/async_pf.c                     |  3 +-
 virt/kvm/kvm_main.c                     |  9 ++--
 22 files changed, 215 insertions(+), 77 deletions(-)

diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.c b/drivers/gpu/drm/etnaviv/etnaviv_gem.c
index 5ce3603..873130d 100644
--- a/drivers/gpu/drm/etnaviv/etnaviv_gem.c
+++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.c
@@ -758,7 +758,8 @@ static struct page **etnaviv_gem_userptr_do_get_pages(
 
 	down_read(&mm->mmap_sem);
 	while (pinned < npages) {
-		ret = get_user_pages_remote(task, mm, ptr, npages - pinned,
+		ret = get_user_pages_remote(task, mm, NULL, NULL, ptr,
+					    npages - pinned,
 					    !etnaviv_obj->userptr.ro, 0,
 					    pvec + pinned, NULL);
 		if (ret < 0)
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index 2314c88..1633081 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -544,7 +544,7 @@ __i915_gem_userptr_get_pages_worker(struct work_struct *_work)
 			down_read(&mm->mmap_sem);
 			while (pinned < npages) {
 				ret = get_user_pages_remote
-					(work->task, mm,
+					(work->task, mm, NULL, NULL,
 					 obj->userptr.ptr + pinned * PAGE_SIZE,
 					 npages - pinned,
 					 !obj->userptr.read_only, 0,
diff --git a/drivers/infiniband/core/umem_odp.c b/drivers/infiniband/core/umem_odp.c
index 75077a0..cbebfaa 100644
--- a/drivers/infiniband/core/umem_odp.c
+++ b/drivers/infiniband/core/umem_odp.c
@@ -572,8 +572,8 @@ int ib_umem_odp_map_dma_pages(struct ib_umem *umem, u64 user_virt, u64 bcnt,
 		 * complex (and doesn't gain us much performance in most use
 		 * cases).
 		 */
-		npages = get_user_pages_remote(owning_process, owning_mm,
-				user_virt, gup_num_pages,
+		npages = get_user_pages_remote(owning_process, owning_mm, NULL,
+				NULL, user_virt, gup_num_pages,
 				access_mask & ODP_WRITE_ALLOWED_BIT,
 				0, local_page_list, NULL);
 		up_read(&owning_mm->mmap_sem);
diff --git a/fs/exec.c b/fs/exec.c
index d607da8..c57a8f0 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -203,8 +203,8 @@ static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
 	 * We are doing an exec().  'current' is the process
 	 * doing the exec and bprm->mm is the new process's mm.
 	 */
-	ret = get_user_pages_remote(current, bprm->mm, pos, 1, write,
-			0, &page, NULL);
+	ret = get_user_pages_remote(current, bprm->mm, NULL, NULL, pos, 1,
+			write, 0, &page, NULL);
 	if (ret <= 0)
 		return NULL;
 
diff --git a/fs/proc/base.c b/fs/proc/base.c
index ac0df4d..9575975 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -113,6 +113,11 @@ struct pid_entry {
 	union proc_op op;
 };
 
+struct mem_private {
+	struct mm_struct *mm;
+	const struct cred *object_cred;
+};
+
 #define NOD(NAME, MODE, IOP, FOP, OP) {			\
 	.name = (NAME),					\
 	.len  = sizeof(NAME) - 1,			\
@@ -252,7 +257,7 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 	 * Inherently racy -- command line shares address space
 	 * with code and data.
 	 */
-	rv = access_remote_vm(mm, arg_end - 1, &c, 1, 0);
+	rv = access_remote_vm(mm, NULL, NULL, arg_end - 1, &c, 1, 0);
 	if (rv <= 0)
 		goto out_free_page;
 
@@ -270,7 +275,8 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 			int nr_read;
 
 			_count = min3(count, len, PAGE_SIZE);
-			nr_read = access_remote_vm(mm, p, page, _count, 0);
+			nr_read = access_remote_vm(mm, NULL, NULL, p, page,
+						   _count, 0);
 			if (nr_read < 0)
 				rv = nr_read;
 			if (nr_read <= 0)
@@ -305,7 +311,8 @@ static ssize_t proc_pid_cmdline_read(struct file *file, char __user *buf,
 			bool final;
 
 			_count = min3(count, len, PAGE_SIZE);
-			nr_read = access_remote_vm(mm, p, page, _count, 0);
+			nr_read = access_remote_vm(mm, NULL, NULL, p, page,
+						   _count, 0);
 			if (nr_read < 0)
 				rv = nr_read;
 			if (nr_read <= 0)
@@ -354,7 +361,8 @@ skip_argv:
 			bool final;
 
 			_count = min3(count, len, PAGE_SIZE);
-			nr_read = access_remote_vm(mm, p, page, _count, 0);
+			nr_read = access_remote_vm(mm, NULL, NULL, p, page,
+						   _count, 0);
 			if (nr_read < 0)
 				rv = nr_read;
 			if (nr_read <= 0)
@@ -403,7 +411,7 @@ static const struct file_operations proc_pid_cmdline_ops = {
 static int proc_pid_auxv(struct seq_file *m, struct pid_namespace *ns,
 			 struct pid *pid, struct task_struct *task)
 {
-	struct mm_struct *mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
+	struct mm_struct *mm = mm_access(task, NULL, PTRACE_MODE_READ_FSCREDS);
 	if (mm && !IS_ERR(mm)) {
 		unsigned int nwords = 0;
 		do {
@@ -801,13 +809,15 @@ static const struct file_operations proc_single_file_operations = {
 };
 
 
-struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
+struct mm_struct *proc_mem_open(struct inode *inode,
+				const struct cred **object_cred,
+				unsigned int mode)
 {
 	struct task_struct *task = get_proc_task(inode);
 	struct mm_struct *mm = ERR_PTR(-ESRCH);
 
 	if (task) {
-		mm = mm_access(task, mode | PTRACE_MODE_FSCREDS);
+		mm = mm_access(task, object_cred, mode | PTRACE_MODE_FSCREDS);
 		put_task_struct(task);
 
 		if (!IS_ERR_OR_NULL(mm)) {
@@ -823,7 +833,7 @@ struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
 
 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
 {
-	struct mm_struct *mm = proc_mem_open(inode, mode);
+	struct mm_struct *mm = proc_mem_open(inode, NULL, mode);
 
 	if (IS_ERR(mm))
 		return PTR_ERR(mm);
@@ -834,18 +844,36 @@ static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
 
 static int mem_open(struct inode *inode, struct file *file)
 {
-	int ret = __mem_open(inode, file, PTRACE_MODE_ATTACH);
+	struct mem_private *private = kmalloc(sizeof(*private), GFP_KERNEL);
+	struct mm_struct *mm;
+
+	if (!private)
+		return -ENOMEM;
+
+	mm = proc_mem_open(inode, &private->object_cred, PTRACE_MODE_ATTACH);
+
+	if (!mm)
+		private->object_cred = NULL;
+
+	if (IS_ERR(mm)) {
+		kfree(private);
+		return PTR_ERR(mm);
+	}
+
+	private->mm = mm;
+	file->private_data = private;
 
 	/* OK to pass negative loff_t, we can catch out-of-range */
 	file->f_mode |= FMODE_UNSIGNED_OFFSET;
 
-	return ret;
+	return 0;
 }
 
 static ssize_t mem_rw(struct file *file, char __user *buf,
 			size_t count, loff_t *ppos, int write)
 {
-	struct mm_struct *mm = file->private_data;
+	struct mem_private *private = file->private_data;
+	struct mm_struct *mm = private->mm;
 	unsigned long addr = *ppos;
 	ssize_t copied;
 	char *page;
@@ -869,7 +897,9 @@ static ssize_t mem_rw(struct file *file, char __user *buf,
 			break;
 		}
 
-		this_len = access_remote_vm(mm, addr, page, this_len, write);
+		this_len = access_remote_vm(mm, file->f_cred,
+					    private->object_cred, addr,
+					    page, this_len, write);
 		if (!this_len) {
 			if (!copied)
 				copied = -EIO;
@@ -924,9 +954,13 @@ loff_t mem_lseek(struct file *file, loff_t offset, int orig)
 
 static int mem_release(struct inode *inode, struct file *file)
 {
-	struct mm_struct *mm = file->private_data;
-	if (mm)
-		mmdrop(mm);
+	struct mem_private *private = file->private_data;
+
+	if (private->mm) {
+		mmdrop(private->mm);
+		put_cred(private->object_cred);
+	}
+	kfree(private);
 	return 0;
 }
 
@@ -981,7 +1015,7 @@ static ssize_t environ_read(struct file *file, char __user *buf,
 		max_len = min_t(size_t, PAGE_SIZE, count);
 		this_len = min(max_len, this_len);
 
-		retval = access_remote_vm(mm, (env_start + src),
+		retval = access_remote_vm(mm, NULL, NULL, (env_start + src),
 			page, this_len, 0);
 
 		if (retval <= 0) {
@@ -1873,7 +1907,7 @@ static int map_files_d_revalidate(struct dentry *dentry, unsigned int flags)
 	if (!task)
 		goto out_notask;
 
-	mm = mm_access(task, PTRACE_MODE_READ_FSCREDS);
+	mm = mm_access(task, NULL, PTRACE_MODE_READ_FSCREDS);
 	if (IS_ERR_OR_NULL(mm))
 		goto out;
 
diff --git a/fs/proc/internal.h b/fs/proc/internal.h
index 7931c55..53e4ad6 100644
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -288,7 +288,9 @@ struct proc_maps_private {
 #endif
 };
 
-struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode);
+struct mm_struct *proc_mem_open(struct inode *inode,
+				const struct cred **object_cred,
+				unsigned int mode);
 
 extern const struct file_operations proc_pid_maps_operations;
 extern const struct file_operations proc_tid_maps_operations;
diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index f6fa99e..f7ab4aa 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -230,7 +230,7 @@ static int proc_maps_open(struct inode *inode, struct file *file,
 		return -ENOMEM;
 
 	priv->inode = inode;
-	priv->mm = proc_mem_open(inode, PTRACE_MODE_READ);
+	priv->mm = proc_mem_open(inode, NULL, PTRACE_MODE_READ);
 	if (IS_ERR(priv->mm)) {
 		int err = PTR_ERR(priv->mm);
 
@@ -1443,7 +1443,7 @@ static int pagemap_open(struct inode *inode, struct file *file)
 {
 	struct mm_struct *mm;
 
-	mm = proc_mem_open(inode, PTRACE_MODE_READ);
+	mm = proc_mem_open(inode, NULL, PTRACE_MODE_READ);
 	if (IS_ERR(mm))
 		return PTR_ERR(mm);
 	file->private_data = mm;
diff --git a/fs/proc/task_nommu.c b/fs/proc/task_nommu.c
index faacb0c..6243696 100644
--- a/fs/proc/task_nommu.c
+++ b/fs/proc/task_nommu.c
@@ -287,7 +287,7 @@ static int maps_open(struct inode *inode, struct file *file,
 		return -ENOMEM;
 
 	priv->inode = inode;
-	priv->mm = proc_mem_open(inode, PTRACE_MODE_READ);
+	priv->mm = proc_mem_open(inode, NULL, PTRACE_MODE_READ);
 	if (IS_ERR(priv->mm)) {
 		int err = PTR_ERR(priv->mm);
 
diff --git a/include/linux/lsm_hooks.h b/include/linux/lsm_hooks.h
index 101bf19..b62d989 100644
--- a/include/linux/lsm_hooks.h
+++ b/include/linux/lsm_hooks.h
@@ -1154,6 +1154,11 @@
  *	to the @parent process for tracing.
  *	@parent contains the task_struct structure for debugger process.
  *	Return 0 if permission is granted.
+ * @forced_write:
+ *	Check whether @subject_cred is permitted to forcibly write to the
+ *	non-writable mapping @vma that belongs to a process with objective
+ *	credentials @object_cred.
+ *	Return 0 if permission is granted.
  * @capget:
  *	Get the @effective, @inheritable, and @permitted capability sets for
  *	the @target process.  The hook may also perform permission checking to
@@ -1317,6 +1322,9 @@ union security_list_options {
 	int (*ptrace_access_check)(struct task_struct *child,
 					unsigned int mode);
 	int (*ptrace_traceme)(struct task_struct *parent);
+	int (*forced_write)(struct vm_area_struct *vma,
+				const struct cred *subject_cred,
+				const struct cred *object_cred);
 	int (*capget)(struct task_struct *target, kernel_cap_t *effective,
 			kernel_cap_t *inheritable, kernel_cap_t *permitted);
 	int (*capset)(struct cred *new, const struct cred *old,
@@ -1629,6 +1637,7 @@ struct security_hook_heads {
 	struct list_head binder_transfer_file;
 	struct list_head ptrace_access_check;
 	struct list_head ptrace_traceme;
+	struct list_head forced_write;
 	struct list_head capget;
 	struct list_head capset;
 	struct list_head capable;
diff --git a/include/linux/mm.h b/include/linux/mm.h
index ef815b9..ef92319 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -23,6 +23,7 @@
 #include <linux/page_ext.h>
 #include <linux/err.h>
 #include <linux/page_ref.h>
+#include <linux/cred.h>
 
 struct mempolicy;
 struct anon_vma;
@@ -1279,14 +1280,19 @@ static inline int fixup_user_fault(struct task_struct *tsk,
 #endif
 
 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
-extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
-		void *buf, int len, int write);
+extern int access_remote_vm(struct mm_struct *mm,
+		const struct cred *subject_cred, const struct cred *object_cred,
+		unsigned long addr, void *buf, int len, int write);
 
 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
+		      const struct cred *subject_cred,
+		      const struct cred *object_cred,
 		      unsigned long start, unsigned long nr_pages,
 		      unsigned int foll_flags, struct page **pages,
 		      struct vm_area_struct **vmas, int *nonblocking);
 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
+			    const struct cred *subject_cred,
+			    const struct cred *object_cred,
 			    unsigned long start, unsigned long nr_pages,
 			    int write, int force, struct page **pages,
 			    struct vm_area_struct **vmas);
@@ -1296,6 +1302,8 @@ long get_user_pages(unsigned long start, unsigned long nr_pages,
 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
 		    int write, int force, struct page **pages, int *locked);
 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
+			       const struct cred *subject_cred,
+			       const struct cred *object_cred,
 			       unsigned long start, unsigned long nr_pages,
 			       int write, int force, struct page **pages,
 			       unsigned int gup_flags);
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 62c68e5..c792341 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2853,7 +2853,9 @@ extern struct mm_struct *get_task_mm(struct task_struct *task);
  * and ptrace_may_access with the mode parameter passed to it
  * succeeds.
  */
-extern struct mm_struct *mm_access(struct task_struct *task, unsigned int mode);
+extern struct mm_struct *mm_access(struct task_struct *task,
+				   const struct cred **object_cred,
+				   unsigned int mode);
 /* Remove the current tasks stale references to the old mm_struct */
 extern void mm_release(struct task_struct *, struct mm_struct *);
 
diff --git a/include/linux/security.h b/include/linux/security.h
index 7831cd5..8f0dbce 100644
--- a/include/linux/security.h
+++ b/include/linux/security.h
@@ -193,6 +193,9 @@ int security_binder_transfer_file(struct task_struct *from,
 				  struct task_struct *to, struct file *file);
 int security_ptrace_access_check(struct task_struct *child, unsigned int mode);
 int security_ptrace_traceme(struct task_struct *parent);
+int security_forced_write(struct vm_area_struct *vma,
+			  const struct cred *subject_cred,
+			  const struct cred *object_cred);
 int security_capget(struct task_struct *target,
 		    kernel_cap_t *effective,
 		    kernel_cap_t *inheritable,
@@ -424,6 +427,13 @@ static inline int security_ptrace_traceme(struct task_struct *parent)
 	return cap_ptrace_traceme(parent);
 }
 
+static inline int security_forced_write(struct vm_area_struct *vma,
+					const struct cred *subject_cred,
+					const struct cred *object_cred)
+{
+	return 0;
+}
+
 static inline int security_capget(struct task_struct *target,
 				   kernel_cap_t *effective,
 				   kernel_cap_t *inheritable,
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index 8c50276..26021e1 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -300,7 +300,8 @@ int uprobe_write_opcode(struct mm_struct *mm, unsigned long vaddr,
 
 retry:
 	/* Read the page with vaddr into memory */
-	ret = get_user_pages_remote(NULL, mm, vaddr, 1, 0, 1, &old_page, &vma);
+	ret = get_user_pages_remote(NULL, mm, NULL, NULL, vaddr, 1, 0, 1,
+				    &old_page, &vma);
 	if (ret <= 0)
 		return ret;
 
@@ -1710,7 +1711,8 @@ static int is_trap_at_addr(struct mm_struct *mm, unsigned long vaddr)
 	 * but we treat this as a 'remote' access since it is
 	 * essentially a kernel access to the memory.
 	 */
-	result = get_user_pages_remote(NULL, mm, vaddr, 1, 0, 1, &page, NULL);
+	result = get_user_pages_remote(NULL, mm, NULL, NULL, vaddr, 1, 0, 1,
+				       &page, NULL);
 	if (result < 0)
 		return result;
 
diff --git a/kernel/fork.c b/kernel/fork.c
index beb3172..e6e1fd7 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -847,7 +847,8 @@ struct mm_struct *get_task_mm(struct task_struct *task)
 }
 EXPORT_SYMBOL_GPL(get_task_mm);
 
-struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
+struct mm_struct *mm_access(struct task_struct *task,
+			    const struct cred **object_cred, unsigned int mode)
 {
 	struct mm_struct *mm;
 	int err;
@@ -862,6 +863,9 @@ struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
 		mmput(mm);
 		mm = ERR_PTR(-EACCES);
 	}
+	if (!IS_ERR_OR_NULL(mm) && object_cred)
+		*object_cred = get_task_cred(task);
+
 	mutex_unlock(&task->signal->cred_guard_mutex);
 
 	return mm;
diff --git a/mm/gup.c b/mm/gup.c
index 96b2b2f..481ab81 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -2,6 +2,7 @@
 #include <linux/errno.h>
 #include <linux/err.h>
 #include <linux/spinlock.h>
+#include <linux/security.h>
 
 #include <linux/mm.h>
 #include <linux/memremap.h>
@@ -416,7 +417,17 @@ static int faultin_page(struct task_struct *tsk, struct vm_area_struct *vma,
 	return 0;
 }
 
-static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
+/*
+ * subject_cred must be the subjective credentials using which access is
+ * requested.
+ * object_cred must be the objective credentials of the target task at the time
+ * the mm_struct was acquired.
+ * Both of these may be NULL if FOLL_FORCE is unset or FOLL_WRITE is unset.
+ */
+static int check_vma_flags(struct vm_area_struct *vma,
+			   const struct cred *subject_cred,
+			   const struct cred *object_cred,
+			   unsigned long gup_flags)
 {
 	vm_flags_t vm_flags = vma->vm_flags;
 	int write = (gup_flags & FOLL_WRITE);
@@ -426,9 +437,19 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
 		return -EFAULT;
 
 	if (write) {
+		/* If one of the cred parameters is missing and the WRITE and
+		 * FORCE flags are set, that's a kernel bug.
+		 */
+		if (WARN_ON((gup_flags & FOLL_FORCE) &&
+		    (subject_cred == NULL || object_cred == NULL)))
+			return -EFAULT;
+
 		if (!(vm_flags & VM_WRITE)) {
 			if (!(gup_flags & FOLL_FORCE))
 				return -EFAULT;
+			if (security_forced_write(vma, subject_cred,
+						  object_cred))
+				return -EFAULT;
 			/*
 			 * We used to let the write,force case do COW in a
 			 * VM_MAYWRITE VM_SHARED !VM_WRITE vma, so ptrace could
@@ -517,6 +538,8 @@ static int check_vma_flags(struct vm_area_struct *vma, unsigned long gup_flags)
  * you need some special @gup_flags.
  */
 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
+		const struct cred *subject_cred,
+		const struct cred *object_cred,
 		unsigned long start, unsigned long nr_pages,
 		unsigned int gup_flags, struct page **pages,
 		struct vm_area_struct **vmas, int *nonblocking)
@@ -557,7 +580,8 @@ long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
 				goto next_page;
 			}
 
-			if (!vma || check_vma_flags(vma, gup_flags))
+			if (!vma || check_vma_flags(vma, subject_cred,
+						    object_cred, gup_flags))
 				return i ? : -EFAULT;
 			if (is_vm_hugetlb_page(vma)) {
 				i = follow_hugetlb_page(mm, vma, pages, vmas,
@@ -727,6 +751,8 @@ EXPORT_SYMBOL_GPL(fixup_user_fault);
 
 static __always_inline long __get_user_pages_locked(struct task_struct *tsk,
 						struct mm_struct *mm,
+						const struct cred *subject_cred,
+						const struct cred *object_cred,
 						unsigned long start,
 						unsigned long nr_pages,
 						int write, int force,
@@ -755,8 +781,9 @@ static __always_inline long __get_user_pages_locked(struct task_struct *tsk,
 	pages_done = 0;
 	lock_dropped = false;
 	for (;;) {
-		ret = __get_user_pages(tsk, mm, start, nr_pages, flags, pages,
-				       vmas, locked);
+		ret = __get_user_pages(tsk, mm, subject_cred, object_cred,
+				       start, nr_pages, flags, pages, vmas,
+				       locked);
 		if (!locked)
 			/* VM_FAULT_RETRY couldn't trigger, bypass */
 			return ret;
@@ -795,8 +822,9 @@ static __always_inline long __get_user_pages_locked(struct task_struct *tsk,
 		*locked = 1;
 		lock_dropped = true;
 		down_read(&mm->mmap_sem);
-		ret = __get_user_pages(tsk, mm, start, 1, flags | FOLL_TRIED,
-				       pages, NULL, NULL);
+		ret = __get_user_pages(tsk, mm, subject_cred, object_cred,
+				       start, 1, flags | FOLL_TRIED, pages,
+				       NULL, NULL);
 		if (ret != 1) {
 			BUG_ON(ret > 1);
 			if (!pages_done)
@@ -846,9 +874,10 @@ long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
 			   int write, int force, struct page **pages,
 			   int *locked)
 {
-	return __get_user_pages_locked(current, current->mm, start, nr_pages,
-				       write, force, pages, NULL, locked, true,
-				       FOLL_TOUCH);
+	return __get_user_pages_locked(current, current->mm, current_cred(),
+				       current_real_cred(), start,
+				       nr_pages, write, force, pages, NULL,
+				       locked, true, FOLL_TOUCH);
 }
 EXPORT_SYMBOL(get_user_pages_locked);
 
@@ -863,6 +892,8 @@ EXPORT_SYMBOL(get_user_pages_locked);
  * respectively.
  */
 __always_inline long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
+					       const struct cred *subject_cred,
+					       const struct cred *object_cred,
 					       unsigned long start, unsigned long nr_pages,
 					       int write, int force, struct page **pages,
 					       unsigned int gup_flags)
@@ -870,8 +901,9 @@ __always_inline long __get_user_pages_unlocked(struct task_struct *tsk, struct m
 	long ret;
 	int locked = 1;
 	down_read(&mm->mmap_sem);
-	ret = __get_user_pages_locked(tsk, mm, start, nr_pages, write, force,
-				      pages, NULL, &locked, false, gup_flags);
+	ret = __get_user_pages_locked(tsk, mm, subject_cred, object_cred, start,
+				      nr_pages, write, force, pages, NULL,
+				      &locked, false, gup_flags);
 	if (locked)
 		up_read(&mm->mmap_sem);
 	return ret;
@@ -898,7 +930,12 @@ EXPORT_SYMBOL(__get_user_pages_unlocked);
 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
 			     int write, int force, struct page **pages)
 {
-	return __get_user_pages_unlocked(current, current->mm, start, nr_pages,
+	/* None of the current callers actually pass write=1 together with
+	 * force=1, but pass in current_cred() and current_read_cred() in case
+	 * that changes in the future.
+	 */
+	return __get_user_pages_unlocked(current, current->mm, current_cred(),
+					 current_real_cred(), start, nr_pages,
 					 write, force, pages, FOLL_TOUCH);
 }
 EXPORT_SYMBOL(get_user_pages_unlocked);
@@ -959,12 +996,15 @@ EXPORT_SYMBOL(get_user_pages_unlocked);
  * FAULT_FLAG_ALLOW_RETRY to handle_mm_fault.
  */
 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
+		const struct cred *subject_cred,
+		const struct cred *object_cred,
 		unsigned long start, unsigned long nr_pages,
 		int write, int force, struct page **pages,
 		struct vm_area_struct **vmas)
 {
-	return __get_user_pages_locked(tsk, mm, start, nr_pages, write, force,
-				       pages, vmas, NULL, false,
+	return __get_user_pages_locked(tsk, mm, subject_cred, object_cred,
+				       start, nr_pages, write, force, pages,
+				       vmas, NULL, false,
 				       FOLL_TOUCH | FOLL_REMOTE);
 }
 EXPORT_SYMBOL(get_user_pages_remote);
@@ -979,9 +1019,10 @@ long get_user_pages(unsigned long start, unsigned long nr_pages,
 		int write, int force, struct page **pages,
 		struct vm_area_struct **vmas)
 {
-	return __get_user_pages_locked(current, current->mm, start, nr_pages,
-				       write, force, pages, vmas, NULL, false,
-				       FOLL_TOUCH);
+	return __get_user_pages_locked(current, current->mm, current_cred(),
+				       current_real_cred(), start,
+				       nr_pages, write, force, pages, vmas,
+				       NULL, false, FOLL_TOUCH);
 }
 EXPORT_SYMBOL(get_user_pages);
 
@@ -1039,7 +1080,8 @@ long populate_vma_page_range(struct vm_area_struct *vma,
 	 * We made sure addr is within a VMA, so the following will
 	 * not result in a stack expansion that recurses back here.
 	 */
-	return __get_user_pages(current, mm, start, nr_pages, gup_flags,
+	return __get_user_pages(current, mm, current_cred(),
+				current_real_cred(), start, nr_pages, gup_flags,
 				NULL, NULL, nonblocking);
 }
 
@@ -1125,7 +1167,7 @@ struct page *get_dump_page(unsigned long addr)
 	struct vm_area_struct *vma;
 	struct page *page;
 
-	if (__get_user_pages(current, current->mm, addr, 1,
+	if (__get_user_pages(current, current->mm, NULL, NULL, addr, 1,
 			     FOLL_FORCE | FOLL_DUMP | FOLL_GET, &page, &vma,
 			     NULL) < 1)
 		return NULL;
diff --git a/mm/memory.c b/mm/memory.c
index 83be99d..8209d4f 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -3850,6 +3850,7 @@ EXPORT_SYMBOL_GPL(generic_access_phys);
  * given task for page fault accounting.
  */
 static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
+		const struct cred *subject_cred, const struct cred *object_cred,
 		unsigned long addr, void *buf, int len, int write)
 {
 	struct vm_area_struct *vma;
@@ -3862,8 +3863,8 @@ static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
 		void *maddr;
 		struct page *page = NULL;
 
-		ret = get_user_pages_remote(tsk, mm, addr, 1,
-				write, 1, &page, &vma);
+		ret = get_user_pages_remote(tsk, mm, subject_cred, object_cred,
+				addr, 1, write, 1, &page, &vma);
 		if (ret <= 0) {
 #ifndef CONFIG_HAVE_IOREMAP_PROT
 			break;
@@ -3919,28 +3920,35 @@ static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
  *
  * The caller must hold a reference on @mm.
  */
-int access_remote_vm(struct mm_struct *mm, unsigned long addr,
-		void *buf, int len, int write)
+int access_remote_vm(struct mm_struct *mm, const struct cred *subject_cred,
+		     const struct cred *object_cred, unsigned long addr,
+		     void *buf, int len, int write)
 {
-	return __access_remote_vm(NULL, mm, addr, buf, len, write);
+	return __access_remote_vm(NULL, mm, subject_cred, object_cred, addr,
+				  buf, len, write);
 }
 
 /*
  * Access another process' address space.
  * Source/target buffer must be kernel space,
- * Do not walk the page table directly, use get_user_pages
+ * Do not walk the page table directly, use get_user_pages.
+ * @tsk must be ptrace-stopped by current.
  */
 int access_process_vm(struct task_struct *tsk, unsigned long addr,
 		void *buf, int len, int write)
 {
 	struct mm_struct *mm;
 	int ret;
+	const struct cred *object_cred;
 
 	mm = get_task_mm(tsk);
 	if (!mm)
 		return 0;
+	object_cred = get_task_cred(tsk);
 
-	ret = __access_remote_vm(tsk, mm, addr, buf, len, write);
+	ret = __access_remote_vm(tsk, mm, current_cred(), object_cred, addr,
+				 buf, len, write);
+	put_cred(object_cred);
 	mmput(mm);
 
 	return ret;
diff --git a/mm/nommu.c b/mm/nommu.c
index 95daf81..ebb03ba 100644
--- a/mm/nommu.c
+++ b/mm/nommu.c
@@ -110,6 +110,8 @@ unsigned int kobjsize(const void *objp)
 }
 
 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
+		      const struct cred *subject_cred,
+		      const struct cred *object_cred,
 		      unsigned long start, unsigned long nr_pages,
 		      unsigned int foll_flags, struct page **pages,
 		      struct vm_area_struct **vmas, int *nonblocking)
@@ -170,8 +172,8 @@ long get_user_pages(unsigned long start, unsigned long nr_pages,
 	if (force)
 		flags |= FOLL_FORCE;
 
-	return __get_user_pages(current, current->mm, start, nr_pages, flags,
-				pages, vmas, NULL);
+	return __get_user_pages(current, current->mm, NULL, NULL, start,
+				nr_pages, flags, pages, vmas, NULL);
 }
 EXPORT_SYMBOL(get_user_pages);
 
@@ -184,14 +186,16 @@ long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
 EXPORT_SYMBOL(get_user_pages_locked);
 
 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
+			       const struct cred *subject_cred,
+			       const struct cred *object_cred,
 			       unsigned long start, unsigned long nr_pages,
 			       int write, int force, struct page **pages,
 			       unsigned int gup_flags)
 {
 	long ret;
 	down_read(&mm->mmap_sem);
-	ret = __get_user_pages(tsk, mm, start, nr_pages, gup_flags, pages,
-				NULL, NULL);
+	ret = __get_user_pages(tsk, mm, subject_cred, object_cred, start,
+				nr_pages, gup_flags, pages, NULL, NULL);
 	up_read(&mm->mmap_sem);
 	return ret;
 }
@@ -200,8 +204,9 @@ EXPORT_SYMBOL(__get_user_pages_unlocked);
 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
 			     int write, int force, struct page **pages)
 {
-	return __get_user_pages_unlocked(current, current->mm, start, nr_pages,
-					 write, force, pages, 0);
+	return __get_user_pages_unlocked(current, current->mm, NULL, NULL,
+					 start, nr_pages, write, force, pages,
+					 0);
 }
 EXPORT_SYMBOL(get_user_pages_unlocked);
 
@@ -1858,8 +1863,9 @@ static int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
  *
  * The caller must hold a reference on @mm.
  */
-int access_remote_vm(struct mm_struct *mm, unsigned long addr,
-		void *buf, int len, int write)
+int access_remote_vm(struct mm_struct *mm, const struct cred *subject_cred,
+		const struct cred *object_cred, unsigned long addr, void *buf,
+		int len, int write)
 {
 	return __access_remote_vm(NULL, mm, addr, buf, len, write);
 }
diff --git a/mm/process_vm_access.c b/mm/process_vm_access.c
index 07514d4..9f2b6a1 100644
--- a/mm/process_vm_access.c
+++ b/mm/process_vm_access.c
@@ -103,9 +103,9 @@ static int process_vm_rw_single_vec(unsigned long addr,
 		 * add FOLL_REMOTE because task/mm might not
 		 * current/current->mm
 		 */
-		pages = __get_user_pages_unlocked(task, mm, pa, pages,
-						  vm_write, 0, process_pages,
-						  FOLL_REMOTE);
+		pages = __get_user_pages_unlocked(task, mm, NULL, NULL, pa,
+						  pages, vm_write, 0,
+						  process_pages, FOLL_REMOTE);
 		if (pages <= 0)
 			return -EFAULT;
 
@@ -199,7 +199,7 @@ static ssize_t process_vm_rw_core(pid_t pid, struct iov_iter *iter,
 		goto free_proc_pages;
 	}
 
-	mm = mm_access(task, PTRACE_MODE_ATTACH_REALCREDS);
+	mm = mm_access(task, NULL, PTRACE_MODE_ATTACH_REALCREDS);
 	if (!mm || IS_ERR(mm)) {
 		rc = IS_ERR(mm) ? PTR_ERR(mm) : -ESRCH;
 		/*
diff --git a/security/security.c b/security/security.c
index 4838e7f..0c16a6c 100644
--- a/security/security.c
+++ b/security/security.c
@@ -164,6 +164,13 @@ int security_ptrace_traceme(struct task_struct *parent)
 	return call_int_hook(ptrace_traceme, 0, parent);
 }
 
+int security_forced_write(struct vm_area_struct *vma,
+			  const struct cred *subject_cred,
+			  const struct cred *object_cred)
+{
+	return call_int_hook(forced_write, 0, vma, subject_cred, object_cred);
+}
+
 int security_capget(struct task_struct *target,
 		     kernel_cap_t *effective,
 		     kernel_cap_t *inheritable,
@@ -1582,6 +1589,7 @@ struct security_hook_heads security_hook_heads = {
 		LIST_HEAD_INIT(security_hook_heads.ptrace_access_check),
 	.ptrace_traceme =
 		LIST_HEAD_INIT(security_hook_heads.ptrace_traceme),
+	.forced_write =	LIST_HEAD_INIT(security_hook_heads.forced_write),
 	.capget =	LIST_HEAD_INIT(security_hook_heads.capget),
 	.capset =	LIST_HEAD_INIT(security_hook_heads.capset),
 	.capable =	LIST_HEAD_INIT(security_hook_heads.capable),
diff --git a/security/tomoyo/domain.c b/security/tomoyo/domain.c
index ade7c6c..f5acc47 100644
--- a/security/tomoyo/domain.c
+++ b/security/tomoyo/domain.c
@@ -880,7 +880,7 @@ bool tomoyo_dump_page(struct linux_binprm *bprm, unsigned long pos,
 	 * (represented by bprm).  'current' is the process doing
 	 * the execve().
 	 */
-	if (get_user_pages_remote(current, bprm->mm, pos, 1,
+	if (get_user_pages_remote(current, bprm->mm, NULL, NULL, pos, 1,
 				0, 1, &page, NULL) <= 0)
 		return false;
 #else
diff --git a/virt/kvm/async_pf.c b/virt/kvm/async_pf.c
index db96688..e49acb7 100644
--- a/virt/kvm/async_pf.c
+++ b/virt/kvm/async_pf.c
@@ -84,7 +84,8 @@ static void async_pf_execute(struct work_struct *work)
 	 * mm and might be done in another context, so we must
 	 * use FOLL_REMOTE.
 	 */
-	__get_user_pages_unlocked(NULL, mm, addr, 1, 1, 0, NULL, FOLL_REMOTE);
+	__get_user_pages_unlocked(NULL, mm, NULL, NULL, addr, 1, 1, 0, NULL,
+				  FOLL_REMOTE);
 
 	kvm_async_page_present_sync(vcpu, apf);
 
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 1950782..5bedffe 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -1349,15 +1349,15 @@ static int get_user_page_nowait(unsigned long start, int write,
 	if (write)
 		flags |= FOLL_WRITE;
 
-	return __get_user_pages(current, current->mm, start, 1, flags, page,
-			NULL, NULL);
+	return __get_user_pages(current, current->mm, NULL, NULL, start, 1,
+			flags, page, NULL, NULL);
 }
 
 static inline int check_user_page_hwpoison(unsigned long addr)
 {
 	int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE;
 
-	rc = __get_user_pages(current, current->mm, addr, 1,
+	rc = __get_user_pages(current, current->mm, NULL, NULL, addr, 1,
 			      flags, NULL, NULL, NULL);
 	return rc == -EHWPOISON;
 }
@@ -1415,7 +1415,8 @@ static int hva_to_pfn_slow(unsigned long addr, bool *async, bool write_fault,
 		npages = get_user_page_nowait(addr, write_fault, page);
 		up_read(&current->mm->mmap_sem);
 	} else
-		npages = __get_user_pages_unlocked(current, current->mm, addr, 1,
+		npages = __get_user_pages_unlocked(current, current->mm, NULL,
+						   NULL, addr, 1,
 						   write_fault, 0, page,
 						   FOLL_TOUCH|FOLL_HWPOISON);
 	if (npages != 1)
-- 
2.1.4

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH v2 3/3] selinux: require EXECMEM for forced ptrace poke

[Jann Horn]

This is a breaking change for SELinux users that restrict EXECMEM: It might
break gdb if gdb is executed in a domain that does not have EXECMEM
privilege over the debuggee domain.

Unlike most other SELinux hooks, this one takes the subject credentials as
an argument instead of looking up current_cred(). This is done because the
security_forced_write() LSM hook can be invoked from within the write
handler of /proc/$pid/mem, where current_cred() is pretty useless.

Signed-off-by: Jann Horn <jann@thejh.net>
Reviewed-by: Janis Danisevskis <jdanis@android.com>
---
 security/selinux/hooks.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
index 13185a6..e36682a 100644
--- a/security/selinux/hooks.c
+++ b/security/selinux/hooks.c
@@ -2149,6 +2149,20 @@ static int selinux_ptrace_traceme(struct task_struct *parent)
 	return task_has_perm(parent, current, PROCESS__PTRACE);
 }
 
+static int selinux_forced_write(struct vm_area_struct *vma,
+				const struct cred *subject_cred,
+				const struct cred *object_cred)
+{
+	/* Permitting a write to readonly memory is fine - making the readonly
+	 * memory executable afterwards would require EXECMOD permission because
+	 * anon_vma would be non-NULL.
+	 */
+	if ((vma->vm_flags & VM_EXEC) == 0)
+		return 0;
+
+	return cred_has_perm(subject_cred, object_cred, PROCESS__EXECMEM);
+}
+
 static int selinux_capget(struct task_struct *target, kernel_cap_t *effective,
 			  kernel_cap_t *inheritable, kernel_cap_t *permitted)
 {
@@ -6033,6 +6047,7 @@ static struct security_hook_list selinux_hooks[] = {
 
 	LSM_HOOK_INIT(ptrace_access_check, selinux_ptrace_access_check),
 	LSM_HOOK_INIT(ptrace_traceme, selinux_ptrace_traceme),
+	LSM_HOOK_INIT(forced_write, selinux_forced_write),
 	LSM_HOOK_INIT(capget, selinux_capget),
 	LSM_HOOK_INIT(capset, selinux_capset),
 	LSM_HOOK_INIT(capable, selinux_capable),
-- 
2.1.4

Re: [PATCH v2 2/3] mm: add LSM hook for writes to readonly memory

[Andy Lutomirski]

On Wed, Sep 28, 2016 at 3:54 PM, Jann Horn <jann@thejh.net> wrote:
> SELinux attempts to make it possible to whitelist trustworthy sources of
> code that may be mapped into memory, and Android makes use of this feature.
> To prevent an attacker from bypassing this by modifying R+X memory through
> /proc/$pid/mem or PTRACE_POKETEXT, it is necessary to call a security hook
> in check_vma_flags().

If selinux policy allows PTRACE_POKETEXT, is it really so bad for that
to result in code execution?


> -struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
> +struct mm_struct *proc_mem_open(struct inode *inode,
> +                               const struct cred **object_cred,
> +                               unsigned int mode)
>  {

Why are you passing object_cred all over the place like this?  You
have an inode, and an inode implies a task.

For that matter, would it possibly make sense to use MEMCG's mm->owner
and get rid of object_cred entirely?  I can see this causing issues in
strange threading cases, e.g. accessing your own /proc/$$/mem vs
another thread in your process's.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH v2 2/3] mm: add LSM hook for writes to readonly memory

[Jann Horn]

[-- Attachment #1: Type: text/plain, Size: 2046 bytes --]

On Wed, Sep 28, 2016 at 04:22:53PM -0700, Andy Lutomirski wrote:
> On Wed, Sep 28, 2016 at 3:54 PM, Jann Horn <jann@thejh.net> wrote:
> > SELinux attempts to make it possible to whitelist trustworthy sources of
> > code that may be mapped into memory, and Android makes use of this feature.
> > To prevent an attacker from bypassing this by modifying R+X memory through
> > /proc/$pid/mem or PTRACE_POKETEXT, it is necessary to call a security hook
> > in check_vma_flags().
> 
> If selinux policy allows PTRACE_POKETEXT, is it really so bad for that
> to result in code execution?

Have a look at __ptrace_may_access():

	/* Don't let security modules deny introspection */
	if (same_thread_group(task, current))
		return 0;

This means thread A can attach to thread B and poke its memory, and SELinux
can't do anything about it.

I guess another perspective on this would be that it's a problem that
interfaces usable for poking user memory are subject to introspection rules
(as opposed to e.g. /proc/self/maps, where it is actually useful).

> > -struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
> > +struct mm_struct *proc_mem_open(struct inode *inode,
> > +                               const struct cred **object_cred,
> > +                               unsigned int mode)
> >  {
> 
> Why are you passing object_cred all over the place like this?  You
> have an inode, and an inode implies a task.

But the task's mm and objective credentials can change, and only mm_access()
holds the cred_guard_mutex during the mm lookup. Although, if the objective
credentials change because of a setuid execution, being able to poke in the
old mm would be pretty harmless...


> For that matter, would it possibly make sense to use MEMCG's mm->owner
> and get rid of object_cred entirely?

I guess it might.


> I can see this causing issues in
> strange threading cases, e.g. accessing your own /proc/$$/mem vs
> another thread in your process's.

Can you elaborate on that?

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH v2 2/3] mm: add LSM hook for writes to readonly memory

[Jann Horn]

[-- Attachment #1: Type: text/plain, Size: 1360 bytes --]

On Thu, Sep 29, 2016 at 01:32:56AM +0200, Jann Horn wrote:
> On Wed, Sep 28, 2016 at 04:22:53PM -0700, Andy Lutomirski wrote:
> > On Wed, Sep 28, 2016 at 3:54 PM, Jann Horn <jann@thejh.net> wrote:
> > > SELinux attempts to make it possible to whitelist trustworthy sources of
> > > code that may be mapped into memory, and Android makes use of this feature.
> > > To prevent an attacker from bypassing this by modifying R+X memory through
> > > /proc/$pid/mem or PTRACE_POKETEXT, it is necessary to call a security hook
> > > in check_vma_flags().
> > 
> > If selinux policy allows PTRACE_POKETEXT, is it really so bad for that
> > to result in code execution?
> 
> Have a look at __ptrace_may_access():
> 
> 	/* Don't let security modules deny introspection */
> 	if (same_thread_group(task, current))
> 		return 0;
> 
> This means thread A can attach to thread B and poke its memory, and SELinux
> can't do anything about it.
> 
> I guess another perspective on this would be that it's a problem that
> interfaces usable for poking user memory are subject to introspection rules
> (as opposed to e.g. /proc/self/maps, where it is actually useful).

Ugh, I'm talking nonsense, ptrace() doesn't work on threads. (/proc/$pid/mem
works though). And then, ptrace-ish APIs aside, there are those weird
devices that do DMA with force=1.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

Re: [PATCH v2 3/3] selinux: require EXECMEM for forced ptrace poke

[Ingo Molnar]

* Jann Horn <jann@thejh.net> wrote:

> +{
> +	/* Permitting a write to readonly memory is fine - making the readonly
> +	 * memory executable afterwards would require EXECMOD permission because
> +	 * anon_vma would be non-NULL.
> +	 */

Minor stylistic nit: please fix this comment to be visually symmetric. (There's 
another one introduced by one of your other patches.)

Thanks,

	Ingo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH v2 2/3] mm: add LSM hook for writes to readonly memory

[Ingo Molnar]

* Jann Horn <jann@thejh.net> wrote:

> +/*
> + * subject_cred must be the subjective credentials using which access is
> + * requested.
> + * object_cred must be the objective credentials of the target task at the time
> + * the mm_struct was acquired.
> + * Both of these may be NULL if FOLL_FORCE is unset or FOLL_WRITE is unset.

Hm, I have trouble parsing the first sentence.

> -	return __get_user_pages_locked(current, current->mm, start, nr_pages,
> -				       write, force, pages, vmas, NULL, false,
> -				       FOLL_TOUCH);
> +	return __get_user_pages_locked(current, current->mm, current_cred(),
> +				       current_real_cred(), start,
> +				       nr_pages, write, force, pages, vmas,
> +				       NULL, false, FOLL_TOUCH);

So the parameter passing was disgustig before, and now it became super disgusing! 

Would it improve the code if we added a friendly helper structure (or two if 
that's better) to clean up all the interactions within these various functions?

Thanks,

	Ingo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH v2 1/3] fs/exec: don't force writing memory access

[Oleg Nesterov]

On 09/29, Jann Horn wrote:
>
> @@ -204,7 +204,7 @@ static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
>  	 * doing the exec and bprm->mm is the new process's mm.
>  	 */
>  	ret = get_user_pages_remote(current, bprm->mm, pos, 1, write,
> -			1, &page, NULL);
> +			0, &page, NULL);

To me this looks like a reasonable cleanup regardless, FOLL_FORCE
just adds the unnecessary confusion here.

Oleg.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH v2 3/3] selinux: require EXECMEM for forced ptrace poke

[Stephen Smalley]

On 09/28/2016 06:54 PM, Jann Horn wrote:
> This is a breaking change for SELinux users that restrict EXECMEM: It might
> break gdb if gdb is executed in a domain that does not have EXECMEM
> privilege over the debuggee domain.

Since this would break compatibility with existing SELinux policies, you
have to wrap it with a conditional on a policy capability that you can
then enable in newer policies.  See commit
2be4d74f2fd45460d70d4fe65cc1972ef45bf849 for an example.  This requires
a corresponding update to libsepol, and then adding the new policy
capability to your policy (in the policy_capabilities file).

> 
> Unlike most other SELinux hooks, this one takes the subject credentials as
> an argument instead of looking up current_cred(). This is done because the
> security_forced_write() LSM hook can be invoked from within the write
> handler of /proc/$pid/mem, where current_cred() is pretty useless.
> 
> Signed-off-by: Jann Horn <jann@thejh.net>
> Reviewed-by: Janis Danisevskis <jdanis@android.com>
> ---
>  security/selinux/hooks.c | 15 +++++++++++++++
>  1 file changed, 15 insertions(+)
> 
> diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
> index 13185a6..e36682a 100644
> --- a/security/selinux/hooks.c
> +++ b/security/selinux/hooks.c
> @@ -2149,6 +2149,20 @@ static int selinux_ptrace_traceme(struct task_struct *parent)
>  	return task_has_perm(parent, current, PROCESS__PTRACE);
>  }
>  
> +static int selinux_forced_write(struct vm_area_struct *vma,
> +				const struct cred *subject_cred,
> +				const struct cred *object_cred)
> +{
> +	/* Permitting a write to readonly memory is fine - making the readonly
> +	 * memory executable afterwards would require EXECMOD permission because
> +	 * anon_vma would be non-NULL.
> +	 */
> +	if ((vma->vm_flags & VM_EXEC) == 0)
> +		return 0;
> +
> +	return cred_has_perm(subject_cred, object_cred, PROCESS__EXECMEM);
> +}
> +
>  static int selinux_capget(struct task_struct *target, kernel_cap_t *effective,
>  			  kernel_cap_t *inheritable, kernel_cap_t *permitted)
>  {
> @@ -6033,6 +6047,7 @@ static struct security_hook_list selinux_hooks[] = {
>  
>  	LSM_HOOK_INIT(ptrace_access_check, selinux_ptrace_access_check),
>  	LSM_HOOK_INIT(ptrace_traceme, selinux_ptrace_traceme),
> +	LSM_HOOK_INIT(forced_write, selinux_forced_write),
>  	LSM_HOOK_INIT(capget, selinux_capget),
>  	LSM_HOOK_INIT(capset, selinux_capset),
>  	LSM_HOOK_INIT(capable, selinux_capable),
> 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH v2 2/3] mm: add LSM hook for writes to readonly memory

[Jann Horn]

[-- Attachment #1: Type: text/plain, Size: 1789 bytes --]

On Thu, Sep 29, 2016 at 01:32:56AM +0200, Jann Horn wrote:
> On Wed, Sep 28, 2016 at 04:22:53PM -0700, Andy Lutomirski wrote:
> > On Wed, Sep 28, 2016 at 3:54 PM, Jann Horn <jann@thejh.net> wrote:
> > > -struct mm_struct *proc_mem_open(struct inode *inode, unsigned int mode)
> > > +struct mm_struct *proc_mem_open(struct inode *inode,
> > > +                               const struct cred **object_cred,
> > > +                               unsigned int mode)
> > >  {
> > 
> > Why are you passing object_cred all over the place like this?  You
> > have an inode, and an inode implies a task.
> 
> But the task's mm and objective credentials can change, and only mm_access()
> holds the cred_guard_mutex during the mm lookup. Although, if the objective
> credentials change because of a setuid execution, being able to poke in the
> old mm would be pretty harmless...

Actually, no. If you can poke in the pre-execve memory, but are checked
against the (possibly more permissive) objective creds of the post-execve
process, you can affect another process that shares the pre-execve memory
(the case where task B, which calls execve(), was clone()d from task A
with CLONE_VM). So I'm keeping this code the way I wrote it.


> > For that matter, would it possibly make sense to use MEMCG's mm->owner
> > and get rid of object_cred entirely?
> 
> I guess it might.

Actually, I'd prefer not to do that - I think it would be unnecessarily
unintuitive to check against the objective creds of task A when accessing
task B if task B was clone()d from A with clone(CLONE_VM).

> > I can see this causing issues in
> > strange threading cases, e.g. accessing your own /proc/$$/mem vs
> > another thread in your process's.
> 
> Can you elaborate on that?


[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 819 bytes --]