diff mbox series

[v3,1/2] lockdown: kexec_file: prevent unsigned kernel image when KEXEC_SIG not enabled

Message ID 20221230065850.897967-1-coxu@redhat.com
State New
Headers
Series [v3,1/2] lockdown: kexec_file: prevent unsigned kernel image when KEXEC_SIG not enabled |

Commit Message

Coiby Xu Dec. 30, 2022, 6:58 a.m. UTC 
  A kernel builder may not enable KEXEC_SIG and some architectures like
ppc64 simply don't have KEXEC_SIG. In these cases, unless both
IMA_ARCH_POLICY and secure boot also enabled, lockdown doesn't prevent
unsigned kernel image from being kexec'ed via the kexec_file_load
syscall whereas it could prevent one via the kexec_load syscall. Mandate
signature verification for those cases.

Fixes: 155bdd30af17 ("kexec_file: Restrict at runtime if the kernel is locked down")
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Jiri Bohac <jbohac@suse.cz>
Cc: David Howells <dhowells@redhat.com>
Cc: kexec@lists.infradead.org
Cc: linux-integrity@vger.kernel.org
Signed-off-by: Coiby Xu <coxu@redhat.com>
Reviewed-by: Mimi Zohar <zohar@linux.ibm.com>
---
v3
 - add lockdown tests to explain why kexec_file_load failed [Mimi]

v2
 - collect reviewed-by tag from Mimi
 - s/mandate_signatute_verification/mandate_signature_verification [Mimi]
 - return the status of kexec_image_verify_sig correctly when signature
   verification is not mandated
---

 kernel/kexec_file.c | 26 +++++++++++++++++---------
 1 file changed, 17 insertions(+), 9 deletions(-)

Comments

Mimi Zohar Jan. 3, 2023, 2:42 p.m. UTC | #1 
Hi Coiby,

On Fri, 2022-12-30 at 14:58 +0800, Coiby Xu wrote:
> A kernel builder may not enable KEXEC_SIG and some architectures like
> ppc64 simply don't have KEXEC_SIG. In these cases, unless both
> IMA_ARCH_POLICY and secure boot also enabled, lockdown doesn't prevent
> unsigned kernel image from being kexec'ed via the kexec_file_load
> syscall whereas it could prevent one via the kexec_load syscall. Mandate
> signature verification for those cases.

The phrase "unless both IMA_ARCH_POLICY and secure boot also enabled"
doesn't reflect the code.  IMA could contain a custom policy rule which
requires the kexec kernel image signature verification as well.  Refer
to the comment now in mandate_signature_verification().

thanks,

Mimi
diff mbox series

Patch

diff --git a/kernel/kexec_file.c b/kernel/kexec_file.c
index dd5983010b7b..2c1054ab21ef 100644
--- a/kernel/kexec_file.c
+++ b/kernel/kexec_file.c
@@ -125,6 +125,17 @@  void kimage_file_post_load_cleanup(struct kimage *image)
 	image->image_loader_data = NULL;
 }
 
+static bool mandate_signature_verification(void)
+{
+	/*
+	 * If IMA is guaranteed to appraise a signature on the kexec
+	 * image, permit it even if the kernel is otherwise locked
+	 * down.
+	 */
+	return !ima_appraise_signature(READING_KEXEC_IMAGE) &&
+	       security_locked_down(LOCKDOWN_KEXEC);
+}
+
 #ifdef CONFIG_KEXEC_SIG
 #ifdef CONFIG_SIGNED_PE_FILE_VERIFICATION
 int kexec_kernel_verify_pe_sig(const char *kernel, unsigned long kernel_len)
@@ -168,13 +179,7 @@  kimage_validate_signature(struct kimage *image)
 			return ret;
 		}
 
-		/*
-		 * If IMA is guaranteed to appraise a signature on the kexec
-		 * image, permit it even if the kernel is otherwise locked
-		 * down.
-		 */
-		if (!ima_appraise_signature(READING_KEXEC_IMAGE) &&
-		    security_locked_down(LOCKDOWN_KEXEC))
+		if (mandate_signature_verification())
 			return -EPERM;
 
 		pr_debug("kernel signature verification failed (%d).\n", ret);
@@ -211,10 +216,13 @@  kimage_file_prepare_segments(struct kimage *image, int kernel_fd, int initrd_fd,
 
 #ifdef CONFIG_KEXEC_SIG
 	ret = kimage_validate_signature(image);
-
+#else
+	if (mandate_signature_verification())
+		ret = -EPERM;
+#endif
 	if (ret)
 		goto out;
-#endif
+
 	/* It is possible that there no initramfs is being loaded */
 	if (!(flags & KEXEC_FILE_NO_INITRAMFS)) {
 		ret = kernel_read_file_from_fd(initrd_fd, 0, &image->initrd_buf,