linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	stable@vger.kernel.org,
	Stephen Brennan <stephen.s.brennan@oracle.com>,
	Douglas Anderson <dianders@chromium.org>,
	Daniel Thompson <daniel.thompson@linaro.org>,
	Linus Torvalds <torvalds@linux-foundation.org>
Subject: [PATCH 5.18 01/47] lockdown: also lock down previous kgdb use
Date: Fri, 27 May 2022 10:49:41 +0200	[thread overview]
Message-ID: <20220527084801.415044459@linuxfoundation.org> (raw)
In-Reply-To: <20220527084801.223648383@linuxfoundation.org>

From: Daniel Thompson <daniel.thompson@linaro.org>

commit eadb2f47a3ced5c64b23b90fd2a3463f63726066 upstream.

KGDB and KDB allow read and write access to kernel memory, and thus
should be restricted during lockdown.  An attacker with access to a
serial port (for example, via a hypervisor console, which some cloud
vendors provide over the network) could trigger the debugger so it is
important that the debugger respect the lockdown mode when/if it is
triggered.

Fix this by integrating lockdown into kdb's existing permissions
mechanism.  Unfortunately kgdb does not have any permissions mechanism
(although it certainly could be added later) so, for now, kgdb is simply
and brutally disabled by immediately exiting the gdb stub without taking
any action.

For lockdowns established early in the boot (e.g. the normal case) then
this should be fine but on systems where kgdb has set breakpoints before
the lockdown is enacted than "bad things" will happen.

CVE: CVE-2022-21499
Co-developed-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Signed-off-by: Stephen Brennan <stephen.s.brennan@oracle.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 include/linux/security.h    |    2 +
 kernel/debug/debug_core.c   |   24 +++++++++++++++++
 kernel/debug/kdb/kdb_main.c |   62 +++++++++++++++++++++++++++++++++++++++++---
 security/security.c         |    2 +
 4 files changed, 87 insertions(+), 3 deletions(-)

--- a/include/linux/security.h
+++ b/include/linux/security.h
@@ -121,10 +121,12 @@ enum lockdown_reason {
 	LOCKDOWN_DEBUGFS,
 	LOCKDOWN_XMON_WR,
 	LOCKDOWN_BPF_WRITE_USER,
+	LOCKDOWN_DBG_WRITE_KERNEL,
 	LOCKDOWN_INTEGRITY_MAX,
 	LOCKDOWN_KCORE,
 	LOCKDOWN_KPROBES,
 	LOCKDOWN_BPF_READ_KERNEL,
+	LOCKDOWN_DBG_READ_KERNEL,
 	LOCKDOWN_PERF,
 	LOCKDOWN_TRACEFS,
 	LOCKDOWN_XMON_RW,
--- a/kernel/debug/debug_core.c
+++ b/kernel/debug/debug_core.c
@@ -53,6 +53,7 @@
 #include <linux/vmacache.h>
 #include <linux/rcupdate.h>
 #include <linux/irq.h>
+#include <linux/security.h>
 
 #include <asm/cacheflush.h>
 #include <asm/byteorder.h>
@@ -752,6 +753,29 @@ cpu_master_loop:
 				continue;
 			kgdb_connected = 0;
 		} else {
+			/*
+			 * This is a brutal way to interfere with the debugger
+			 * and prevent gdb being used to poke at kernel memory.
+			 * This could cause trouble if lockdown is applied when
+			 * there is already an active gdb session. For now the
+			 * answer is simply "don't do that". Typically lockdown
+			 * *will* be applied before the debug core gets started
+			 * so only developers using kgdb for fairly advanced
+			 * early kernel debug can be biten by this. Hopefully
+			 * they are sophisticated enough to take care of
+			 * themselves, especially with help from the lockdown
+			 * message printed on the console!
+			 */
+			if (security_locked_down(LOCKDOWN_DBG_WRITE_KERNEL)) {
+				if (IS_ENABLED(CONFIG_KGDB_KDB)) {
+					/* Switch back to kdb if possible... */
+					dbg_kdb_mode = 1;
+					continue;
+				} else {
+					/* ... otherwise just bail */
+					break;
+				}
+			}
 			error = gdb_serial_stub(ks);
 		}
 
--- a/kernel/debug/kdb/kdb_main.c
+++ b/kernel/debug/kdb/kdb_main.c
@@ -45,6 +45,7 @@
 #include <linux/proc_fs.h>
 #include <linux/uaccess.h>
 #include <linux/slab.h>
+#include <linux/security.h>
 #include "kdb_private.h"
 
 #undef	MODULE_PARAM_PREFIX
@@ -166,10 +167,62 @@ struct task_struct *kdb_curr_task(int cp
 }
 
 /*
- * Check whether the flags of the current command and the permissions
- * of the kdb console has allow a command to be run.
+ * Update the permissions flags (kdb_cmd_enabled) to match the
+ * current lockdown state.
+ *
+ * Within this function the calls to security_locked_down() are "lazy". We
+ * avoid calling them if the current value of kdb_cmd_enabled already excludes
+ * flags that might be subject to lockdown. Additionally we deliberately check
+ * the lockdown flags independently (even though read lockdown implies write
+ * lockdown) since that results in both simpler code and clearer messages to
+ * the user on first-time debugger entry.
+ *
+ * The permission masks during a read+write lockdown permits the following
+ * flags: INSPECT, SIGNAL, REBOOT (and ALWAYS_SAFE).
+ *
+ * The INSPECT commands are not blocked during lockdown because they are
+ * not arbitrary memory reads. INSPECT covers the backtrace family (sometimes
+ * forcing them to have no arguments) and lsmod. These commands do expose
+ * some kernel state but do not allow the developer seated at the console to
+ * choose what state is reported. SIGNAL and REBOOT should not be controversial,
+ * given these are allowed for root during lockdown already.
+ */
+static void kdb_check_for_lockdown(void)
+{
+	const int write_flags = KDB_ENABLE_MEM_WRITE |
+				KDB_ENABLE_REG_WRITE |
+				KDB_ENABLE_FLOW_CTRL;
+	const int read_flags = KDB_ENABLE_MEM_READ |
+			       KDB_ENABLE_REG_READ;
+
+	bool need_to_lockdown_write = false;
+	bool need_to_lockdown_read = false;
+
+	if (kdb_cmd_enabled & (KDB_ENABLE_ALL | write_flags))
+		need_to_lockdown_write =
+			security_locked_down(LOCKDOWN_DBG_WRITE_KERNEL);
+
+	if (kdb_cmd_enabled & (KDB_ENABLE_ALL | read_flags))
+		need_to_lockdown_read =
+			security_locked_down(LOCKDOWN_DBG_READ_KERNEL);
+
+	/* De-compose KDB_ENABLE_ALL if required */
+	if (need_to_lockdown_write || need_to_lockdown_read)
+		if (kdb_cmd_enabled & KDB_ENABLE_ALL)
+			kdb_cmd_enabled = KDB_ENABLE_MASK & ~KDB_ENABLE_ALL;
+
+	if (need_to_lockdown_write)
+		kdb_cmd_enabled &= ~write_flags;
+
+	if (need_to_lockdown_read)
+		kdb_cmd_enabled &= ~read_flags;
+}
+
+/*
+ * Check whether the flags of the current command, the permissions of the kdb
+ * console and the lockdown state allow a command to be run.
  */
-static inline bool kdb_check_flags(kdb_cmdflags_t flags, int permissions,
+static bool kdb_check_flags(kdb_cmdflags_t flags, int permissions,
 				   bool no_args)
 {
 	/* permissions comes from userspace so needs massaging slightly */
@@ -1180,6 +1233,9 @@ static int kdb_local(kdb_reason_t reason
 		kdb_curr_task(raw_smp_processor_id());
 
 	KDB_DEBUG_STATE("kdb_local 1", reason);
+
+	kdb_check_for_lockdown();
+
 	kdb_go_count = 0;
 	if (reason == KDB_REASON_DEBUG) {
 		/* special case below */
--- a/security/security.c
+++ b/security/security.c
@@ -59,10 +59,12 @@ const char *const lockdown_reasons[LOCKD
 	[LOCKDOWN_DEBUGFS] = "debugfs access",
 	[LOCKDOWN_XMON_WR] = "xmon write access",
 	[LOCKDOWN_BPF_WRITE_USER] = "use of bpf to write user RAM",
+	[LOCKDOWN_DBG_WRITE_KERNEL] = "use of kgdb/kdb to write kernel RAM",
 	[LOCKDOWN_INTEGRITY_MAX] = "integrity",
 	[LOCKDOWN_KCORE] = "/proc/kcore access",
 	[LOCKDOWN_KPROBES] = "use of kprobes",
 	[LOCKDOWN_BPF_READ_KERNEL] = "use of bpf to read kernel RAM",
+	[LOCKDOWN_DBG_READ_KERNEL] = "use of kgdb/kdb to read kernel RAM",
 	[LOCKDOWN_PERF] = "unsafe use of perf",
 	[LOCKDOWN_TRACEFS] = "use of tracefs",
 	[LOCKDOWN_XMON_RW] = "xmon read and write access",



  reply	other threads:[~2022-05-27  8:51 UTC|newest]

Thread overview: 55+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-27  8:49 [PATCH 5.18 00/47] 5.18.1-rc1 review Greg Kroah-Hartman
2022-05-27  8:49 ` Greg Kroah-Hartman [this message]
2022-05-27  8:49 ` [PATCH 5.18 02/47] HID: amd_sfh: Add support for sensor discovery Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 03/47] random: fix sysctl documentation nits Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 04/47] init: call time_init() before rand_initialize() Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 05/47] ia64: define get_cycles macro for arch-override Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 06/47] s390: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 07/47] parisc: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 08/47] alpha: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 09/47] powerpc: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 10/47] timekeeping: Add raw clock fallback for random_get_entropy() Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 11/47] m68k: use fallback for random_get_entropy() instead of zero Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 12/47] riscv: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 13/47] mips: use fallback for random_get_entropy() instead of just c0 random Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 14/47] arm: use fallback for random_get_entropy() instead of zero Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 15/47] nios2: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 16/47] x86/tsc: Use " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 17/47] um: use " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 18/47] sparc: " Greg Kroah-Hartman
2022-05-27  8:49 ` [PATCH 5.18 19/47] xtensa: " Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 20/47] random: insist on random_get_entropy() existing in order to simplify Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 21/47] random: do not use batches when !crng_ready() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 22/47] random: use first 128 bits of input as fast init Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 23/47] random: do not pretend to handle premature next security model Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 24/47] random: order timer entropy functions below interrupt functions Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 25/47] random: do not use input pool from hard IRQs Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 26/47] random: help compiler out with fast_mix() by using simpler arguments Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 27/47] siphash: use one source of truth for siphash permutations Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 28/47] random: use symbolic constants for crng_init states Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 29/47] random: avoid initializing twice in credit race Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 30/47] random: move initialization out of reseeding hot path Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 31/47] random: remove ratelimiting for in-kernel unseeded randomness Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 32/47] random: use proper jiffies comparison macro Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 33/47] random: handle latent entropy and command line from random_init() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 34/47] random: credit architectural init the exact amount Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 35/47] random: use static branch for crng_ready() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 36/47] random: remove extern from functions in header Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 37/47] random: use proper return types on get_random_{int,long}_wait() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 38/47] random: make consistent use of buf and len Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 39/47] random: move initialization functions out of hot pages Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 40/47] random: move randomize_page() into mm where it belongs Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 41/47] random: unify batched entropy implementations Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 42/47] random: convert to using fops->read_iter() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 43/47] random: convert to using fops->write_iter() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 44/47] random: wire up fops->splice_{read,write}_iter() Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 45/47] random: check for signals after page of pool writes Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 46/47] ACPI: sysfs: Fix BERT error region memory mapping Greg Kroah-Hartman
2022-05-27  8:50 ` [PATCH 5.18 47/47] ALSA: ctxfi: Add SB046x PCI ID Greg Kroah-Hartman
2022-05-27 22:40 ` [PATCH 5.18 00/47] 5.18.1-rc1 review Guenter Roeck
2022-05-28  1:53 ` Justin Forbes
2022-05-28  2:24 ` Ron Economos
2022-05-28  2:50 ` Bagas Sanjaya
2022-05-28 11:16 ` Naresh Kamboju
2022-05-28 13:39 ` Rudi Heitbaum
2022-05-29  2:48 ` Fox Chen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220527084801.415044459@linuxfoundation.org \
    --to=gregkh@linuxfoundation.org \
    --cc=daniel.thompson@linaro.org \
    --cc=dianders@chromium.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=stable@vger.kernel.org \
    --cc=stephen.s.brennan@oracle.com \
    --cc=torvalds@linux-foundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).