[PATCH 5.15.y] mm: Fix copy_from_user_nofault().

From: Alexei Starovoitov <[email protected]>

[ Upstream commit d319f344561de23e810515d109c7278919bff7b0 ]

There are several issues with copy_from_user_nofault():

- access_ok() is designed for user context only and for that reason
it has WARN_ON_IN_IRQ() which triggers when bpf, kprobe, eprobe
and perf on ppc are calling it from irq.

- it's missing nmi_uaccess_okay() which is a nop on all architectures
except x86 where it's required.
The comment in arch/x86/mm/tlb.c explains the details why it's necessary.
Calling copy_from_user_nofault() from bpf, [ke]probe without this check is not safe.

- __copy_from_user_inatomic() under CONFIG_HARDENED_USERCOPY is calling
check_object_size()->__check_object_size()->check_heap_object()->find_vmap_area()->spin_lock()
which is not safe to do from bpf, [ke]probe and perf due to potential deadlock.

Fix all three issues. At the end the copy_from_user_nofault() becomes
equivalent to copy_from_user_nmi() from safety point of view with
a difference in the return value.

Reported-by: Hsin-Wei Hung <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: Florian Lehner <[email protected]>
Tested-by: Hsin-Wei Hung <[email protected]>
Tested-by: Florian Lehner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
[ merge conflicts 

  copy_from_user_nofault: caused by force_uaccess_begin and
  force_uaccess_end functions. moved the code around to call those
  functions in the same way after the additional checks.

  check_heap_object: completely different implementation in 5.15.y.
  removed any changes from this commit as the current implementation
  doesn't call the same lock which caused problems. ]
Signed-off-by: Jakov Novak <[email protected]>
---
Note: I have checked all syzbot issues linked with this backport and
verified that they work locally (the ones with repros) and on the
syzbot servers and I ran KUnit tests to make sure this doesn't break anything.
This is my first time sending a patch to the stable repo, so I am a bit
unsure if there is more to be done for testing, apart from inspecting the
code and running those mentioned tests.

 mm/maccess.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/mm/maccess.c b/mm/maccess.c
index ded4bfaba7f3..2b186ce822c9 100644
--- a/mm/maccess.c
+++ b/mm/maccess.c
@@ -5,6 +5,7 @@
 #include <linux/export.h>
 #include <linux/mm.h>
 #include <linux/uaccess.h>
+#include <asm/tlb.h>
 
 bool __weak copy_from_kernel_nofault_allowed(const void *unsafe_src,
 		size_t size)
@@ -221,13 +222,18 @@ long strncpy_from_kernel_nofault(char *dst, const void *unsafe_addr, long count)
 long copy_from_user_nofault(void *dst, const void __user *src, size_t size)
 {
 	long ret = -EFAULT;
-	mm_segment_t old_fs = force_uaccess_begin();
+	mm_segment_t old_fs;
 
-	if (access_ok(src, size)) {
-		pagefault_disable();
-		ret = __copy_from_user_inatomic(dst, src, size);
-		pagefault_enable();
-	}
+	if (!__access_ok(src, size))
+		return ret;
+
+	if (!nmi_uaccess_okay())
+		return ret;
+
+	old_fs = force_uaccess_begin();
+	pagefault_disable();
+	ret = __copy_from_user_inatomic(dst, src, size);
+	pagefault_enable();
 	force_uaccess_end(old_fs);
 
 	if (ret)
-- 
2.54.0