@@ -117,11 +117,67 @@ __visible bool ex_handler_fprestore(const struct exception_table_entry *fixup,
}
EXPORT_SYMBOL_GPL(ex_handler_fprestore);
+/* Helper to check whether a uaccess fault indicates a kernel bug. */
+static bool bogus_uaccess(struct pt_regs *regs, int trapnr,
+ unsigned long fault_addr)
+{
+ /* This is the normal case: #PF with a fault address in userspace. */
+ if (trapnr == X86_TRAP_PF && fault_addr < TASK_SIZE_MAX)
+ return false;
+
+ /*
+ * This code can be reached for machine checks, but only if the #MC
+ * handler has already decided that it looks like a candidate for fixup.
+ * This e.g. happens when attempting to access userspace memory which
+ * the CPU can't access because of uncorrectable bad memory.
+ */
+ if (trapnr == X86_TRAP_MC)
+ return false;
+
+ /*
+ * There are two remaining exception types we might encounter here:
+ * - #PF for faulting accesses to kernel addresses
+ * - #GP for faulting accesses to noncanonical addresses
+ * Complain about anything else.
+ */
+ if (trapnr != X86_TRAP_PF && trapnr != X86_TRAP_GP) {
+ WARN(1, "unexpected trap %d in uaccess\n", trapnr);
+ return false;
+ }
+
+ /*
+ * This is a faulting memory access in kernel space, on a kernel
+ * address, in a usercopy function. This can e.g. be caused by improper
+ * use of helpers like __put_user and by improper attempts to access
+ * userspace addresses in KERNEL_DS regions.
+ * The one (semi-)legitimate exception are probe_kernel_{read,write}(),
+ * which can be invoked from places like kgdb, /dev/mem (for reading)
+ * and privileged BPF code (for reading).
+ * The probe_kernel_*() functions set the kernel_uaccess_faults_ok flag
+ * to tell us that faulting on kernel addresses, and even noncanonical
+ * addresses, in a userspace accessor does not necessarily imply a
+ * kernel bug, root might just be doing weird stuff.
+ */
+ if (current->kernel_uaccess_faults_ok)
+ return false;
+
+ /* This is bad. Refuse the fixup so that we go into die(). */
+ if (trapnr == X86_TRAP_PF) {
+ pr_emerg("BUG: pagefault on kernel address 0x%lx in non-whitelisted uaccess\n",
+ fault_addr);
+ } else {
+ pr_emerg("BUG: GPF in non-whitelisted uaccess (non-canonical address?)\n");
+ }
+ return true;
+}
+
__visible bool ex_handler_uaccess(const struct exception_table_entry *fixup,
struct pt_regs *regs, int trapnr,
unsigned long error_code,
unsigned long fault_addr)
{
+ if (bogus_uaccess(regs, trapnr, fault_addr))
+ return false;
regs->ip = ex_fixup_addr(fixup);
return true;
}
@@ -132,6 +188,8 @@ __visible bool ex_handler_ext(const struct exception_table_entry *fixup,
unsigned long error_code,
unsigned long fault_addr)
{
+ if (bogus_uaccess(regs, trapnr, fault_addr))
+ return false;
/* Special hack for uaccess_err */
current->thread.uaccess_err = 1;
regs->ip = ex_fixup_addr(fixup);
@@ -2642,6 +2642,7 @@ static long exact_copy_from_user(void *to, const void __user * from,
if (!access_ok(VERIFY_READ, from, n))
return n;
+ current->kernel_uaccess_faults_ok++;
while (n) {
if (__get_user(c, f)) {
memset(t, 0, n);
@@ -2651,6 +2652,7 @@ static long exact_copy_from_user(void *to, const void __user * from,
f++;
n--;
}
+ current->kernel_uaccess_faults_ok--;
return n;
}
@@ -739,6 +739,12 @@ struct task_struct {
unsigned use_memdelay:1;
#endif
+ /*
+ * May usercopy functions fault on kernel addresses?
+ * This is not just a single bit because this can potentially nest.
+ */
+ unsigned int kernel_uaccess_faults_ok;
+
unsigned long atomic_flags; /* Flags requiring atomic access. */
struct restart_block restart_block;
@@ -30,8 +30,10 @@ long __probe_kernel_read(void *dst, const void *src, size_t size)
set_fs(KERNEL_DS);
pagefault_disable();
+ current->kernel_uaccess_faults_ok++;
ret = __copy_from_user_inatomic(dst,
(__force const void __user *)src, size);
+ current->kernel_uaccess_faults_ok--;
pagefault_enable();
set_fs(old_fs);
@@ -58,7 +60,9 @@ long __probe_kernel_write(void *dst, const void *src, size_t size)
set_fs(KERNEL_DS);
pagefault_disable();
+ current->kernel_uaccess_faults_ok++;
ret = __copy_to_user_inatomic((__force void __user *)dst, src, size);
+ current->kernel_uaccess_faults_ok--;
pagefault_enable();
set_fs(old_fs);
@@ -94,11 +98,13 @@ long strncpy_from_unsafe(char *dst, const void *unsafe_addr, long count)
set_fs(KERNEL_DS);
pagefault_disable();
+ current->kernel_uaccess_faults_ok++;
do {
ret = __get_user(*dst++, (const char __user __force *)src++);
} while (dst[-1] && ret == 0 && src - unsafe_addr < count);
+ current->kernel_uaccess_faults_ok--;
dst[-1] = '\0';
pagefault_enable();
set_fs(old_fs);
There have been multiple kernel vulnerabilities that permitted userspace to pass completely unchecked pointers through to userspace accessors: - the waitid() bug - commit 96ca579a1ecc ("waitid(): Add missing access_ok() checks") - the sg/bsg read/write APIs - the infiniband read/write APIs These don't happen all that often, but when they do happen, it is hard to test for them properly; and it is probably also hard to discover them with fuzzing. Even when an unmapped kernel address is supplied to such buggy code, it just returns -EFAULT instead of doing a proper BUG() or at least WARN(). This patch attempts to make such misbehaving code a bit more visible by refusing to do a fixup in the pagefault handler code when a userspace accessor causes #PF on a kernel address and the current context isn't whitelisted. Signed-off-by: Jann Horn <jannh@google.com> --- v3: - whitelist exact_copy_from_user(), at least for now - the alternative would be a somewhat complicated refactor (Kees Cook) arch/x86/mm/extable.c | 58 +++++++++++++++++++++++++++++++++++++++++++ fs/namespace.c | 2 ++ include/linux/sched.h | 6 +++++ mm/maccess.c | 6 +++++ 4 files changed, 72 insertions(+)