* [RFC][PATCH] lkdtm/usercopy: Add tests for other memory types
@ 2022-05-12 18:36 Kees Cook
2022-05-12 18:56 ` Matthew Wilcox
0 siblings, 1 reply; 3+ messages in thread
From: Kees Cook @ 2022-05-12 18:36 UTC (permalink / raw)
To: Matthew Wilcox
Cc: Kees Cook, Greg Kroah-Hartman, Arnd Bergmann, linux-kernel,
linux-hardening
Add coverage for the recently added usercopy checks for vmap, kmap, and
folios.
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Kees Cook <keescook@chromium.org>
---
Hi,
So, this kmap test fails the "good" copy step... Is copy_to_user()
limited somewhere else to not touch kmap'd pages? I can't find it...
Also, is my use of folios correct here?
-Kees
---
drivers/misc/lkdtm/usercopy.c | 104 ++++++++++++++++++++++++++++++++++
1 file changed, 104 insertions(+)
diff --git a/drivers/misc/lkdtm/usercopy.c b/drivers/misc/lkdtm/usercopy.c
index 945806db2a13..d1c585bb7a8d 100644
--- a/drivers/misc/lkdtm/usercopy.c
+++ b/drivers/misc/lkdtm/usercopy.c
@@ -5,6 +5,7 @@
*/
#include "lkdtm.h"
#include <linux/slab.h>
+#include <linux/highmem.h>
#include <linux/vmalloc.h>
#include <linux/sched/task_stack.h>
#include <linux/mman.h>
@@ -341,6 +342,106 @@ static void lkdtm_USERCOPY_KERNEL(void)
vm_munmap(user_addr, PAGE_SIZE);
}
+/*
+ * This expects "kaddr" to point to a PAGE_SIZE allocation, which means
+ * a more complete test that would include copy_from_user() would risk
+ * memory corruption. Just test copy_to_user() here, as that exercises
+ * almost exactly the same code paths.
+ */
+static void do_usercopy_page_span(const char *name, void *kaddr)
+{
+ unsigned long uaddr;
+
+ uaddr = vm_mmap(NULL, 0, PAGE_SIZE, PROT_READ | PROT_WRITE,
+ MAP_ANONYMOUS | MAP_PRIVATE, 0);
+ if (uaddr >= TASK_SIZE) {
+ pr_warn("Failed to allocate user memory\n");
+ return;
+ }
+
+ /* Initialize contents. */
+ memset(kaddr, 0xAA, PAGE_SIZE);
+
+ /* Bump the kaddr forward to detect a page-spanning overflow. */
+ kaddr += PAGE_SIZE / 2;
+
+ pr_info("attempting good copy_to_user() from kernel %s: %px\n",
+ name, kaddr);
+ if (copy_to_user((void __user *)uaddr, kaddr,
+ unconst + (PAGE_SIZE / 2))) {
+ pr_err("copy_to_user() failed unexpectedly?!\n");
+ goto free_user;
+ }
+
+ pr_info("attempting bad copy_to_user() from kernel %s: %px\n",
+ name, kaddr);
+ if (copy_to_user((void __user *)uaddr, kaddr,
+ unconst + PAGE_SIZE)) {
+ pr_warn("Good, copy_to_user() failed, but lacked Oops(?!)\n");
+ goto free_user;
+ }
+
+ pr_err("FAIL: bad copy_to_user() not detected!\n");
+ pr_expected_config_param(CONFIG_HARDENED_USERCOPY, "hardened_usercopy");
+
+free_user:
+ vm_munmap(uaddr, PAGE_SIZE);
+}
+
+static void lkdtm_USERCOPY_VMALLOC(void)
+{
+ void *addr;
+
+ addr = vmalloc(PAGE_SIZE);
+ if (!addr) {
+ pr_err("vmalloc() failed!?\n");
+ return;
+ }
+ do_usercopy_page_span("vmalloc", addr);
+ vfree(addr);
+}
+
+static void lkdtm_USERCOPY_KMAP(void)
+{
+ struct page *page;
+ void *addr;
+
+ page = alloc_pages(GFP_KERNEL | __GFP_ZERO, 0);
+ if (!page) {
+ pr_err("alloc_pages() failed!?\n");
+ return;
+ }
+ addr = kmap_atomic(page);
+ if (addr) {
+ do_usercopy_page_span("kmap", addr);
+ kunmap_atomic(addr);
+ } else {
+ pr_err("kmap_atomic() failed!?\n");
+ }
+ __free_pages(page, 0);
+}
+
+static void lkdtm_USERCOPY_FOLIO(void)
+{
+ struct folio *folio;
+ void *addr;
+
+ /*
+ * FIXME: Folio checking currently misses 0-order allocations, so
+ * allocate and bump forward to the last page.
+ */
+ folio = folio_alloc(GFP_KERNEL | __GFP_ZERO, 1);
+ if (!folio) {
+ pr_err("folio_alloc() failed!?\n");
+ return;
+ }
+ addr = page_address(&folio->page);
+ if (addr) {
+ do_usercopy_page_span("folio", addr + PAGE_SIZE);
+ }
+ folio_put(folio);
+}
+
void __init lkdtm_usercopy_init(void)
{
/* Prepare cache that lacks SLAB_USERCOPY flag. */
@@ -365,6 +466,9 @@ static struct crashtype crashtypes[] = {
CRASHTYPE(USERCOPY_STACK_FRAME_TO),
CRASHTYPE(USERCOPY_STACK_FRAME_FROM),
CRASHTYPE(USERCOPY_STACK_BEYOND),
+ CRASHTYPE(USERCOPY_VMALLOC),
+ CRASHTYPE(USERCOPY_KMAP),
+ CRASHTYPE(USERCOPY_FOLIO),
CRASHTYPE(USERCOPY_KERNEL),
};
--
2.32.0
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [RFC][PATCH] lkdtm/usercopy: Add tests for other memory types
2022-05-12 18:36 [RFC][PATCH] lkdtm/usercopy: Add tests for other memory types Kees Cook
@ 2022-05-12 18:56 ` Matthew Wilcox
2022-05-12 19:12 ` Kees Cook
0 siblings, 1 reply; 3+ messages in thread
From: Matthew Wilcox @ 2022-05-12 18:56 UTC (permalink / raw)
To: Kees Cook
Cc: Greg Kroah-Hartman, Arnd Bergmann, linux-kernel, linux-hardening
On Thu, May 12, 2022 at 11:36:13AM -0700, Kees Cook wrote:
> +static void lkdtm_USERCOPY_FOLIO(void)
> +{
> + struct folio *folio;
> + void *addr;
> +
> + /*
> + * FIXME: Folio checking currently misses 0-order allocations, so
> + * allocate and bump forward to the last page.
> + */
> + folio = folio_alloc(GFP_KERNEL | __GFP_ZERO, 1);
> + if (!folio) {
> + pr_err("folio_alloc() failed!?\n");
> + return;
> + }
> + addr = page_address(&folio->page);
Ideally, code shouldn't be using &folio->page. If it is, we have a
gap in the folio API. Fortunately, we have folio_address().
> + if (addr) {
> + do_usercopy_page_span("folio", addr + PAGE_SIZE);
> + }
> + folio_put(folio);
> +}
Other than that, this looks sane to me.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [RFC][PATCH] lkdtm/usercopy: Add tests for other memory types
2022-05-12 18:56 ` Matthew Wilcox
@ 2022-05-12 19:12 ` Kees Cook
0 siblings, 0 replies; 3+ messages in thread
From: Kees Cook @ 2022-05-12 19:12 UTC (permalink / raw)
To: Matthew Wilcox
Cc: Greg Kroah-Hartman, Arnd Bergmann, linux-kernel, linux-hardening
On Thu, May 12, 2022 at 07:56:13PM +0100, Matthew Wilcox wrote:
> On Thu, May 12, 2022 at 11:36:13AM -0700, Kees Cook wrote:
> > +static void lkdtm_USERCOPY_FOLIO(void)
> > +{
> > + struct folio *folio;
> > + void *addr;
> > +
> > + /*
> > + * FIXME: Folio checking currently misses 0-order allocations, so
> > + * allocate and bump forward to the last page.
> > + */
> > + folio = folio_alloc(GFP_KERNEL | __GFP_ZERO, 1);
> > + if (!folio) {
> > + pr_err("folio_alloc() failed!?\n");
> > + return;
> > + }
> > + addr = page_address(&folio->page);
>
> Ideally, code shouldn't be using &folio->page. If it is, we have a
> gap in the folio API. Fortunately, we have folio_address().
Ah! Perfect, thanks. In trying to find the right alloc/free pair I
missed folio_address() :)
--
Kees Cook
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-05-12 19:12 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-05-12 18:36 [RFC][PATCH] lkdtm/usercopy: Add tests for other memory types Kees Cook
2022-05-12 18:56 ` Matthew Wilcox
2022-05-12 19:12 ` Kees Cook
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.