patch-author-guide.md

kpatch Patch Author Guide

Because kpatch-build is relatively easy to use, it can be easy to assume that a successful patch module build means that the patch is safe to apply. But in fact that's a very dangerous assumption.

There are many pitfalls that can be encountered when creating a live patch. This document attempts to guide the patch creation process. It's a work in progress. If you find it useful, please contribute!

Table of contents

• Patch analysis
• kpatch vs livepatch vs kGraft
• Patch upgrades
• Data structure changes
• Data semantic changes
• Init code changes
• Header file changes
• Dealing with unexpected changed functions
• Removing references to static local variables
• Code removal
• Once macros
• inline implies notrace
• Jump labels and static calls
• Sibling calls
• Exported symbol versioning
• System calls

Patch analysis

kpatch provides some guarantees, but it does not guarantee that all patches are safe to apply. Every patch must also be analyzed in-depth by a human.

The most important point here cannot be stressed enough. Here comes the bold:

Do not blindly apply patches. There is no substitute for human analysis and reasoning on a per-patch basis. All patches must be thoroughly analyzed by a human kernel expert who completely understands the patch and the affected code and how they relate to the live patching environment.

kpatch vs livepatch vs kGraft

This document assumes that the kpatch-build tool is being used to create livepatch kernel modules. Other live patching systems may have different consistency models, their own guarantees, and other subtle differences. The guidance in this document applies only to kpatch-build generated livepatches.

Patch upgrades

Due to potential unexpected interactions between patches, it's highly recommended that when patching a system which has already been patched, the second patch should be a cumulative upgrade which is a superset of the first patch.

Since upstream kernel 5.1, livepatch supports a "replace" flag to help the management of cumulative patches. With the flag set, the kernel will load the cumulative patch and unload all existing patches in one transition. kpatch-build enables the replace flag by default. If replace behavior is not desired, the user can disable it with -R|--non-replace.

Data structure changes

kpatch patches functions, not data. If the original patch involves a change to a data structure, the patch will require some rework, as changes to data structures are not allowed by default.

Usually you have to get creative. There are several possible ways to handle this:

Change the code which uses the data structure

Sometimes, instead of changing the data structure itself, you can change the code which uses it.

For example, consider this patch. which has the following hunk:

@@ -3270,6 +3277,7 @@ static int (*const svm_exit_handlers[])(struct vcpu_svm *svm) = {
 	[SVM_EXIT_EXCP_BASE + PF_VECTOR]	= pf_interception,
 	[SVM_EXIT_EXCP_BASE + NM_VECTOR]	= nm_interception,
 	[SVM_EXIT_EXCP_BASE + MC_VECTOR]	= mc_interception,
+	[SVM_EXIT_EXCP_BASE + AC_VECTOR]	= ac_interception,
 	[SVM_EXIT_INTR]				= intr_interception,
 	[SVM_EXIT_NMI]				= nmi_interception,
 	[SVM_EXIT_SMI]				= nop_on_interception,

svm_exit_handlers[] is an array of function pointers. The patch adds a ac_interception function pointer to the array at index [SVM_EXIT_EXCP_BASE + AC_VECTOR]. That change is incompatible with kpatch.

Looking at the source file, we can see that this function pointer is only accessed by a single function, handle_exit():

        if (exit_code >= ARRAY_SIZE(svm_exit_handlers)
            || !svm_exit_handlers[exit_code]) {
                WARN_ONCE(1, "svm: unexpected exit reason 0x%x\n", exit_code);
                kvm_queue_exception(vcpu, UD_VECTOR);
                return 1;
        }

        return svm_exit_handlers[exit_code](svm);

So an easy solution here is to just change the code to manually check for the new case before looking in the data structure:

@@ -3580,6 +3580,9 @@ static int handle_exit(struct kvm_vcpu *vcpu)
                return 1;
        }

+       if (exit_code == SVM_EXIT_EXCP_BASE + AC_VECTOR)
+               return ac_interception(svm);
+
        return svm_exit_handlers[exit_code](svm);
 }

Not only is this an easy solution, it's also safer than touching data since svm_exit_handlers[] may be in use by tasks that haven't been patched yet.

Use a kpatch callback macro

Kpatch supports the kernel's livepatch (Un)patching callbacks. The kernel API requires callback registration through struct klp_callbacks, but to do so through kpatch-build, kpatch-macros.h defines the following:

• KPATCH_PRE_PATCH_CALLBACK - executed before patching
• KPATCH_POST_PATCH_CALLBACK - executed after patching
• KPATCH_PRE_UNPATCH_CALLBACK - executed before unpatching, complements the post-patch callback.
• KPATCH_POST_UNPATCH_CALLBACK - executed after unpatching, complements the pre-patch callback.

A pre-patch callback routine has the following signature:

static int callback(patch_object *obj) { }
KPATCH_PRE_PATCH_CALLBACK(callback);

and any non-zero return status indicates failure to the kernel. For more information on pre-patch callback failure, see the Pre-patch return status section below.

Post-patch, pre-unpatch, and post-unpatch callback routines all share the following signature:

static void callback(patch_object *obj) { }
KPATCH_POST_PATCH_CALLBACK(callback);            /* or */
KPATCH_PRE_UNPATCH_CALLBACK(callback);           /* or */
KPATCH_POST_UNPATCH_CALLBACK(callback);

Generally pre-patch callbacks are paired with post-unpatch callbacks, meaning that anything the former allocates or sets up should be torn down by the former callback. Likewise for post-patch and pre-unpatch callbacks.

Pre-patch return status

If kpatch is currently patching already loaded objects (vmlinux always by definition as well as any currently loaded kernel modules), a non-zero pre-patch callback status stops the current patch in progress. The kpatch-module is rejected, completely reverted, and unloaded.

If an already loaded kpatch is patching an incoming kernel module, then a failing pre-patch callback will result in the kernel module loader rejecting the new module.

In both cases, if a pre-patch callback fails, none of its other associated callbacks will be executed.

Callback context

• For patches to vmlinux or already loaded kernel modules, callback functions will be run around the livepatch transitions in the klp_enable_patch() callchain. This is executed automatically on kpatch module init.

• For patches to kernel modules which haven't been loaded yet, a module-notifier will execute callbacks when the module is loaded into the MODULE_STATE_COMING state. The pre and post-patch callbacks are called before any module_init code.

Example: a kpatch fix for CVE-2016-5389 could utilize the KPATCH_PRE_PATCH_CALLBACK and KPATCH_POST_UNPATCH_CALLBACK macros to modify variable sysctl_tcp_challenge_ack_limit in-place:

+#include "kpatch-macros.h"
+
+static bool kpatch_write = false;
+static int kpatch_pre_patch_tcp_send_challenge_ack(patch_object *obj)
+{
+	if (sysctl_tcp_challenge_ack_limit == 100) {
+		sysctl_tcp_challenge_ack_limit = 1000;
+		kpatch_write = true;
+	}
+	return 0;
+}
static void kpatch_post_unpatch_tcp_send_challenge_ack(patch_object *obj)
+{
+	if (kpatch_write && sysctl_tcp_challenge_ack_limit == 1000)
+		sysctl_tcp_challenge_ack_limit = 100;
+}
+KPATCH_PRE_PATCH_CALLBACK(kpatch_pre_patch_tcp_send_challenge_ack);
+KPATCH_POST_UNPATCH_CALLBACK(kpatch_post_unpatch_tcp_send_challenge_ack);

Don't forget to protect access to data as needed. Spinlocks and mutexes / sleeping locks may be used (this is a change of behavior from when kpatch relied on the kpatch.ko support module and stop_machine() context.)

Be careful when upgrading. If patch A has a pre/post-patch callback which writes to X, and then you load patch B which is a superset of A, in some cases you may want to prevent patch B from writing to X, if A is already loaded.

Use a shadow variable

If you need to add a field to an existing data structure, or even many existing data structures, you can use the kernel's Shadow Variable API.

Example: The shadow-newpid.patch integration test employs shadow variables to add a rolling counter to the new struct task_struct instances. A simplified version is presented here.

A shadow PID variable is allocated in do_fork(): it is associated with the current struct task_struct *p value, given an ID of KPATCH_SHADOW_NEWPID, sized accordingly, and allocated as per GFP_KERNEL flag rules. Note that the shadow variable <obj, id> association is global -- hence it is best to provide unique ID enumerations per kpatch as needed.

klp_shadow_alloc() returns a pointer to the shadow variable, so we can dereference and make assignments as usual. In this patch chunk, the shadow newpid is allocated then assigned to a rolling ctr counter value:

diff --git a/kernel/fork.c b/kernel/fork.c
index 9bff3b28c357..18374fd35bd9 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -1751,6 +1751,8 @@ struct task_struct *fork_idle(int cpu)
 	return task;
 }
 
+#include <linux/livepatch.h>
+#define KPATCH_SHADOW_NEWPID 0
 /*
  *  Ok, this is the main fork-routine.
  *
@@ -1794,6 +1796,14 @@ long do_fork(unsigned long clone_flags,
 	if (!IS_ERR(p)) {
 		struct completion vfork;
 		struct pid *pid;
+		int *newpid;
+		static int ctr = 0;
+
+		newpid = klp_shadow_get_or_alloc(p, KPATCH_SHADOW_NEWPID,
+						 sizeof(*newpid), GFP_KERNEL,
+						 NULL, NULL);
+		if (newpid)
+			*newpid = ctr++;
 
 		trace_sched_process_fork(current, p);

A shadow variable may be accessed via klp_shadow_get(). Here the patch modifies task_context_switch_counts() to fetch the shadow variable associated with the current struct task_struct *p object and a KPATCH_SHADOW_NEWPID ID. As in the previous patch chunk, the shadow variable pointer may be accessed as an ordinary pointer type:

diff --git a/fs/proc/array.c b/fs/proc/array.c
index 39684c79e8e2..fe0259d057a3 100644
--- a/fs/proc/array.c
+++ b/fs/proc/array.c
@@ -394,13 +394,19 @@ static inline void task_seccomp(struct seq_file *m, struct task_struct *p)
 	seq_putc(m, '\n');
 }
 
+#include <linux/livepatch.h>
+#define KPATCH_SHADOW_NEWPID 0
 static inline void task_context_switch_counts(struct seq_file *m,
 						struct task_struct *p)
 {
+	int *newpid;
 	seq_printf(m,	"voluntary_ctxt_switches:\t%lu\n"
 			"nonvoluntary_ctxt_switches:\t%lu\n",
 			p->nvcsw,
 			p->nivcsw);
+	newpid = klp_shadow_get(p, KPATCH_SHADOW_NEWPID);
+	if (newpid)
+		seq_printf(m, "newpid:\t%d\n", *newpid);
 }
 
 static void task_cpus_allowed(struct seq_file *m, struct task_struct *task)

A shadow variable is freed by calling klp_shadow_free() and providing the object / enum ID combination. Once freed, the shadow variable is no longer safe to access:

diff --git a/kernel/exit.c b/kernel/exit.c
index 148a7842928d..44b6fe61e912 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -791,6 +791,8 @@ static void check_stack_usage(void)
 static inline void check_stack_usage(void) {}
 #endif
 
+#include <linux/livepatch.h>
+#define KPATCH_SHADOW_NEWPID 0
 void do_exit(long code)
 {
 	struct task_struct *tsk = current;
@@ -888,6 +890,8 @@ void do_exit(long code)
 	check_stack_usage();
 	exit_thread();
 
+	klp_shadow_free(tsk, KPATCH_SHADOW_NEWPID, NULL);
+
 	/*
 	 * Flush inherited counters to the parent - before the parent
 	 * gets woken up by child-exit notifications.

Notes:

• klp_shadow_alloc() and klp_shadow_get_or_alloc() initialize only shadow variable metadata. They allocate variable storage via kmalloc with the gfp_t flags given, but otherwise leave the area untouched. Initialization of a shadow variable is the responsibility of the caller.
• As soon as klp_shadow_alloc() or klp_shadow_get_or_alloc() create a shadow variable, its presence will be reported by klp_shadow_get(). Care should be taken to avoid any potential race conditions between a kernel thread that allocates a shadow variable and concurrent threads that may attempt to use it.
• Patches may need to call klp_shadow_free_all() from a post-unpatch handler to safely cleanup any shadow variables of a particular ID. From post-unpatch context, unloading kpatch module code (aside from .exit) should be completely inactive. As long as these shadow variables were only accessed by the unloaded kpatch, they are be safe to release.

Data semantic changes

Part of the stable-tree backport to fix CVE-2014-0206 changed the reference count semantic of struct kioctx.reqs_active. Associating a shadow variable to new instances of this structure can be used by patched code to handle both new (post-patch) and existing (pre-patch) instances.

(Note: this example is trimmed to highlight this use-case. Boilerplate code is also required to allocate/free a shadow variable with enum ID KPATCH_SHADOW_REQS_ACTIVE_V2 whenever a new struct kioctx is created/released. No values are ever assigned to the shadow variable.)

diff --git a/fs/aio.c b/fs/aio.c
index ebd06fd0de89..6a33b73c9107 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -280,6 +280,8 @@ static void free_ioctx_rcu(struct rcu_head *head)
  * and ctx->users has dropped to 0, so we know no more kiocbs can be submitted -
  * now it's safe to cancel any that need to be.
  */
+#include <linux/livepatch.h>
+#define KPATCH_SHADOW_REQS_ACTIVE_V2 1
 static void free_ioctx(struct kioctx *ctx)
 {
        struct aio_ring *ring;

Shadow variable existence can be verified before applying the new data semantic of the associated object:

@@ -678,6 +681,8 @@ void aio_complete(struct kiocb *iocb, long res, long res2)
 put_rq:
        /* everything turned out well, dispose of the aiocb. */
        aio_put_req(iocb);
+       if (klp_shadow_get(ctx, KPATCH_SHADOW_REQS_ACTIVE_V2))
+               atomic_dec(&ctx->reqs_active);
 
        /*
         * We have to order our ring_info tail store above and test

Likewise, shadow variable non-existence can be tested to continue applying the old data semantic:

@@ -310,7 +312,8 @@ static void free_ioctx(struct kioctx *ctx)
 
                avail = (head <= ctx->tail ? ctx->tail : ctx->nr_events) - head;
 
-               atomic_sub(avail, &ctx->reqs_active);
+               if (!klp_shadow_get(ctx, KPATCH_SHADOW_REQS_ACTIVE_V2))
+                       atomic_sub(avail, &ctx->reqs_active);
                head += avail;
                head %= ctx->nr_events;
        }
@@ -757,6 +762,8 @@ static long aio_read_events_ring(struct kioctx *ctx,
        pr_debug("%li  h%u t%u\n", ret, head, ctx->tail);
 
        atomic_sub(ret, &ctx->reqs_active);
+       if (!klp_shadow_get(ctx, KPATCH_SHADOW_REQS_ACTIVE_V2))
+               atomic_sub(ret, &ctx->reqs_active);
 out:
        mutex_unlock(&ctx->ring_lock);

The previous example can be extended to use shadow variable storage to handle locking semantic changes. Consider the upstream fix for CVE-2014-2706, which added a ps_lock to struct sta_info to protect critical sections throughout net/mac80211/sta_info.c.

When allocating a new struct sta_info, allocate a corresponding shadow variable large enough to hold a spinlock_t instance, then initialize the spinlock:

diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
index decd30c1e290..758533dda4d8 100644
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -287,6 +287,8 @@ static int sta_prepare_rate_control(struct ieee80211_local *local,
 	return 0;
 }
 
+#include <linux/livepatch.h>
+#define KPATCH_SHADOW_PS_LOCK 2
 struct sta_info *sta_info_alloc(struct ieee80211_sub_if_data *sdata,
 				const u8 *addr, gfp_t gfp)
 {
@@ -295,6 +297,7 @@ struct sta_info *sta_info_alloc(struct ieee80211_sub_if_data *sdata,
 	struct timespec uptime;
 	struct ieee80211_tx_latency_bin_ranges *tx_latency;
 	int i;
+	spinlock_t *ps_lock;
 
 	sta = kzalloc(sizeof(*sta) + local->hw.sta_data_size, gfp);
 	if (!sta)
@@ -330,6 +333,10 @@ struct sta_info *sta_info_alloc(struct ieee80211_sub_if_data *sdata,
 	rcu_read_unlock();
 
 	spin_lock_init(&sta->lock);
+	ps_lock = klp_shadow_alloc(sta, KPATCH_SHADOW_PS_LOCK,
+				   sizeof(*ps_lock), gfp, NULL, NULL);
+	if (ps_lock)
+		spin_lock_init(ps_lock);
 	INIT_WORK(&sta->drv_unblock_wk, sta_unblock);
 	INIT_WORK(&sta->ampdu_mlme.work, ieee80211_ba_session_work);
 	mutex_init(&sta->ampdu_mlme.mtx);

Patched code can reference the shadow variable associated with a given struct sta_info to determine and apply the correct locking semantic for that instance:

diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 97a02d3f7d87..0edb0ed8dc60 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -459,12 +459,15 @@ static int ieee80211_use_mfp(__le16 fc, struct sta_info *sta,
 	return 1;
 }
 
+#include <linux/livepatch.h>
+#define KPATCH_SHADOW_PS_LOCK 2
 static ieee80211_tx_result
 ieee80211_tx_h_unicast_ps_buf(struct ieee80211_tx_data *tx)
 {
 	struct sta_info *sta = tx->sta;
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(tx->skb);
 	struct ieee80211_local *local = tx->local;
+	spinlock_t *ps_lock;
 
 	if (unlikely(!sta))
 		return TX_CONTINUE;
@@ -478,6 +481,23 @@ ieee80211_tx_h_unicast_ps_buf(struct ieee80211_tx_data *tx)
 		       sta->sta.addr, sta->sta.aid, ac);
 		if (tx->local->total_ps_buffered >= TOTAL_MAX_TX_BUFFER)
 			purge_old_ps_buffers(tx->local);
+
+		/* sync with ieee80211_sta_ps_deliver_wakeup */
+		ps_lock = klp_shadow_get(sta, KPATCH_SHADOW_PS_LOCK);
+		if (ps_lock) {
+			spin_lock(ps_lock);
+			/*
+			 * STA woke up the meantime and all the frames on ps_tx_buf have
+			 * been queued to pending queue. No reordering can happen, go
+			 * ahead and Tx the packet.
+			 */
+			if (!test_sta_flag(sta, WLAN_STA_PS_STA) &&
+			    !test_sta_flag(sta, WLAN_STA_PS_DRIVER)) {
+				spin_unlock(ps_lock);
+				return TX_CONTINUE;
+			}
+		}
+
 		if (skb_queue_len(&sta->ps_tx_buf[ac]) >= STA_MAX_TX_BUFFER) {
 			struct sk_buff *old = skb_dequeue(&sta->ps_tx_buf[ac]);
 			ps_dbg(tx->sdata,

Init code changes

Any code which runs in an __init function or during module or device initialization is problematic, as it may have already run before the patch was applied. The patch may require a pre-patch callback which detects whether such init code has run, and which rewrites or changes the original initialization to force it into the desired state. Some changes involving hardware init are inherently incompatible with live patching.

Header file changes

When changing header files, be extra careful. If data is being changed, you probably need to modify the patch. See "Data struct changes" above.

If a function prototype is being changed, make sure it's not an exported function. Otherwise it could break out-of-tree modules. One way to workaround this is to define an entirely new copy of the function (with updated code) and patch in-tree callers to invoke it rather than the deprecated version.

Many header file changes result in a complete rebuild of the kernel tree, which makes kpatch-build have to compare every .o file in the kernel. It slows the build down a lot, and can even fail to build if kpatch-build has any bugs lurking. If it's a trivial header file change, like adding a macro, it's advisable to just move that macro into the .c file where it's needed to avoid changing the header file at all.

Dealing with unexpected changed functions

In general, it's best to patch as minimally as possible. If kpatch-build is reporting some unexpected function changes, it's always a good idea to try to figure out why it thinks they changed. In many cases you can change the source patch so that they no longer change.

Some examples:

• If a changed function was inlined, then the callers which inlined the function will also change. In this case there's nothing you can do to prevent the extra changes.

• If a changed function was originally inlined, but turned into a callable function after patching, consider adding __always_inline to the function definition. Likewise, if a function is only inlined after patching, consider using noinline to prevent the compiler from doing so.

• If your patch adds a call to a function where the original version of the function's ELF symbol has a .constprop or .isra suffix, and the corresponding patched function doesn't, that means the patch caused gcc to no longer perform an interprocedural optimization, which affects the function and all its callers. If you want to prevent this from happening, copy/paste the function with a new name and call the new function from your patch.

• Moving around source code lines can introduce unique instructions if any __LINE__ preprocessor macros are in use. This can be mitigated by adding any new functions to the bottom of source files, using newline whitespace to maintain original line counts, etc. A more exact fix can be employed by modifying the source code that invokes __LINE__ and hard-coding the original line number in place. This occurred in issue #1124 for example.

Removing references to static local variables

Removing references to static locals will fail to patch unless extra steps are taken. Static locals are basically global variables because they outlive the function's scope. They need to be correlated so that the new function will use the old static local. That way patching the function doesn't inadvertently reset the variable to zero; instead the variable keeps its old value.

To work around this limitation one needs to retain the reference to the static local. This might be as simple as adding the variable back in the patched function in a non-functional way and ensuring the compiler doesn't optimize it away.

Code removal

Some fixes may replace or completely remove functions and references to them. Remember that kpatch modules can only add new functions and redirect existing functions, so "removed" functions will continue to exist in kernel address space as effectively dead code.

That means this patch (source code removal of cmdline_proc_show):

diff -Nupr src.orig/fs/proc/cmdline.c src/fs/proc/cmdline.c
--- src.orig/fs/proc/cmdline.c	2016-11-30 19:39:49.317737234 +0000
+++ src/fs/proc/cmdline.c	2016-11-30 19:39:52.696737234 +0000
@@ -3,15 +3,15 @@
 #include <linux/proc_fs.h>
 #include <linux/seq_file.h>
 
-static int cmdline_proc_show(struct seq_file *m, void *v)
-{
-	seq_printf(m, "%s\n", saved_command_line);
-	return 0;
-}
+static int cmdline_proc_show_v2(struct seq_file *m, void *v)
+{
+	seq_printf(m, "%s kpatch\n", saved_command_line);
+	return 0;
+}
 
 static int cmdline_proc_open(struct inode *inode, struct file *file)
 {
-	return single_open(file, cmdline_proc_show, NULL);
+	return single_open(file, cmdline_proc_show_v2, NULL);
 }
 
 static const struct file_operations cmdline_proc_fops = {

will generate an equivalent kpatch module to this patch (dead cmdline_proc_show left in source):

diff -Nupr src.orig/fs/proc/cmdline.c src/fs/proc/cmdline.c
--- src.orig/fs/proc/cmdline.c	2016-11-30 19:39:49.317737234 +0000
+++ src/fs/proc/cmdline.c	2016-11-30 19:39:52.696737234 +0000
@@ -9,9 +9,15 @@ static int cmdline_proc_show(struct seq_
 	return 0;
 }
 
+static int cmdline_proc_show_v2(struct seq_file *m, void *v)
+{
+	seq_printf(m, "%s kpatch\n", saved_command_line);
+	return 0;
+}
+
 static int cmdline_proc_open(struct inode *inode, struct file *file)
 {
-	return single_open(file, cmdline_proc_show, NULL);
+	return single_open(file, cmdline_proc_show_v2, NULL);
 }
 
 static const struct file_operations cmdline_proc_fops = {

In both versions, kpatch-build will determine that only cmdline_proc_open has changed and that cmdline_proc_show_v2 is a new function.

In some patching cases it might be necessary to completely remove the original function to avoid the compiler complaining about a defined, but unused function. This will depend on symbol scope and kernel build options.

"Once" macros

When adding a call to printk_once(), pr_warn_once(), or any other "once" variation of printk(), you'll get the following eror:

ERROR: vmx.o: 1 unsupported section change(s)
vmx.o: WARNING: unable to correlate static local variable __print_once.60588 used by vmx_update_pi_irte, assuming variable is new
vmx.o: changed function: vmx_update_pi_irte
vmx.o: data section .data..read_mostly selected for inclusion
/usr/lib/kpatch/create-diff-object: unreconcilable difference

This error occurs because the printk_once() adds a static local variable to the .data..read_mostly section. kpatch-build strict disallows any changes to that section, because in some cases a change to this section indicates a bug.

To work around this issue, you'll need to manually implement your own "once" logic which doesn't store the static variable in the .data..read_mostly section.

For example, a pr_warn_once() can be replaced with:

	static bool print_once;
	...
	if (!print_once) {
		print_once = true;
		pr_warn("...");
	}

inline implies notrace

The linux kernel defines its own version of "inline" in include/linux/compiler_types.h which includes "notrace" as well:

#if !defined(CONFIG_OPTIMIZE_INLINING)
#define inline inline __attribute__((__always_inline__)) __gnu_inline \
        __inline_maybe_unused notrace
#else
#define inline inline                                    __gnu_inline \
        __inline_maybe_unused notrace
#endif

With the implicit "notrace", use of "inline" in patch sources may lead to kpatch-build errors like the following:

1. __tcp_mtu_to_mss() is marked as inline:

net/ipv4/tcp_output.c:

/* Calculate MSS not accounting any TCP options.  */
static inline int __tcp_mtu_to_mss(struct sock *sk, int pmtu)
{

2. the compiler decides not to inline it and keeps it in its own function-section. Then kpatch-build notices that it doesn't have an fentry/mcount call:

% kpatch-build ...

tcp_output.o: function __tcp_mtu_to_mss has no fentry/mcount call, unable to patch

3. a peek at the generated code:

Disassembly of section .text.__tcp_mtu_to_mss:

0000000000000000 <__tcp_mtu_to_mss>:
   0:   48 8b 87 60 05 00 00    mov    0x560(%rdi),%rax
   7:   0f b7 50 30             movzwl 0x30(%rax),%edx
   b:   0f b7 40 32             movzwl 0x32(%rax),%eax
   f:   29 d6                   sub    %edx,%esi
  11:   83 ee 14                sub    $0x14,%esi
  ...

This could be a little confusing since one might have expected to see changes to all of __tcp_mtu_to_mss() callers (ie, it was inlined as requested). In this case, a simple workaround is to specify __tcp_mtu_to_mss() as __always_inline to force the compiler to do so.

Jump labels and static calls

Late module patching vs special section relocations

Jump labels and static calls can be problematic due to "late module patching", which is a feature (design flaw?) in upstream livepatch. When a livepatch module patches another module, unfortunately the livepatch module doesn't have an official module dependency on the patched module. That means the patched module doesn't even have to be loaded when the livepatch module gets loaded. In that case the patched module gets patched on demand whenever it might get loaded in the future. It also gets unpatched on demand whenever it gets unloaded.

Loading (and patching) the module at some point after loading the livepatch module is called "late module patching". In order to support this (mis?)feature, all relocations in the livepatch module which reference module symbols must be converted to "klp relocations", which get resolved at patching time.

In all modules (livepatch and otherwise), jump labels and static calls rely on special sections which trigger jump-label/static-call code patching when a module gets loaded. But unfortunately those special sections have relocations which need to get resolved, so there's an ordering issue.

When a (livepatch) module gets loaded, first its relocations are resolved, then its special section handling (and code patching) is done. The problem is, for klp relocations, if they reference another module's symbols, and that module isn't loaded, they're not yet defined. So if a .static_call_sites entry tries to reference its corresponding struct static_call_key, but that key lives in another module which is not yet loaded, the key reference won't be resolved, and so mod->static_call_sites will be corrupted when static_call_module_notify() runs when the livepatch module first loads.

Jump labels

With pre-5.8 kernels, kpatch-build will error out if it encounters any jump labels:

oom_kill.o: Found a jump label at out_of_memory()+0x10a, using key cpusets_enabled_key.  Jump labels aren't supported with this kernel.  Use static_key_enabled() instead.

With Linux 5.8+, klp relocation handling is integrated with the module relocation code, so jump labels in patched functions are supported when the static key was originally defined in the kernel proper (vmlinux).

However, if the static key lives in a module, jump labels are not supported in patched code, due to the ordering issue described above. If the jump label is a tracepoint, kpatch-build will silently remove the tracepoint. Otherwise, there will be an error:

vmx.o: Found a jump label at vmx_hardware_enable.cold()+0x23, using key enable_evmcs, which is defined in a module.  Use static_key_enabled() instead.

When you get one of the above errors, the fix is to remove the jump label usage in the patched function, replacing it with a regular C conditional.

This can be done by replacing any usages of static_branch_likely(), static_branch_unlikely(), static_key_true(), and static_key_false() with static_key_enabled() in the patch file.

Static calls

Similarly, static calls are not supported when the corresponding static call key was originally defined in a module. If such a static call is part of a tracepoint, kpatch-build will silently remove it. Otherwise, there will be an error:

cpuid.o: Found a static call at kvm_set_cpuid.cold()+0x32c, using key __SCK__kvm_x86_vcpu_after_set_cpuid, which is defined in a module.  Use KPATCH_STATIC_CALL() instead.

To fix this error, simply replace such static calls with regular indirect branches (or retpolines, if applicable) by adding #include "kpatch-macros.h" to the patch source and replacing usages of static_call() with KPATCH_STATIC_CALL().

Sibling calls

GCC may generate sibling calls that are incompatible with kpatch, resulting in an error like: ERROR("Found an unsupported sibling call at foo()+0x123. Add __attribute__((optimize("-fno-optimize-sibling-calls"))) to foo() definition."

For example, if function A() calls function B() at the end of A() and both return similar data-types, GCC may deem them "sibling calls" and apply a tail call optimization in which A() restores the stack to is callee state before setting up B()'s arguments and jumping to B().

This may be an issue for kpatches on PowerPC which modify only A() or B() and the function call crosses a kernel module boundary: the sibling call optimization has changed expected calling conventions and (un)patched code may not be similarly modified.

Commit 8b952bd77130 ("create-diff-object/ppc64le: Don't allow sibling calls") contains an excellent example and description of this problem with annotated disassembly.

Adding __attribute__((optimize("-fno-optimize-sibling-calls"))) instructs GCC to turn off the optimization for the given function.

Exported symbol versioning

Background

CONFIG_MODVERSIONS enables an ABI check between exported kernel symbols and modules referencing those symbols, enforced on module load. When building the kernel, preprocessor output from gcc -E for each source file is passed to scripts/genksyms. The genksyms script recursively expands each exported symbol to its basic types. A hash is generated for each symbol as it traverses back up the symbol tree. The end result is a CRC for each exported function in the Module.symvers file and embedded in the vmlinux kernel object itself.

A similar checksumming is performed when building modules: referenced exported symbol CRCs are stored in the module’s __versions section (you can also find these in plain-text intermediate *.mod.c files.)

When the kernel loads a module, the symbol CRCs found in its __versions are compared to those of the kernel, if the two do not match, the kernel will refuse to load it:

<module>: disagrees about version of symbol <symbol>
<module>: Unknown symbol <symbol> (err -22)

Kpatch detection

After building the original and patched sources, kpatch-build compares the newly calculated Module.symvers against the original. Discrepancies are reported:

ERROR: Version disagreement for symbol <symbol>

These reports should be addressed to ensure that the resulting kpatch module can be loaded.

False positives

It is rare, but possible for a kpatch to introduce inadvertent symbol CRC changes that are not true ABI changes. The following conditions must occur:

1. The kpatch must modify the definition of an exported symbol. For example, introducing a new header file may further define an opaque data type: Before the kpatch, compilation unit U from the original kernel build only knew about a struct S declaration (not its complete type). At the same time, U contains function F, which has an interface that references S. If the kpatch adds a header file to U that now fully defines struct S { int a, b, c; }, its symbol type graph changes, CRCs generated for F are updated, but its ABI remains consistent.

2. The kpatch must introduce either a change or reference to F such that it is included in the resulting kpatch module. This will force a __version entry based on the new CRC.

   Note: if a kpatch doesn't change or reference F such that it is not included in the resulting kpatch module, the new CRC value won't be added to the module's __version table. However, if a future accumulative patch does add a new change or reference to F, the new CRC will become a problem.

Avoidance

Kpatches should introduce new #include directives sparingly. Whenever possible, extract the required definitions from header filers into kpatched compilation units directly.

If additional header files or symbol definitions cannot be avoided, consider surrounding the offending include/definitions in an #ifndef __GENKSYMS__ macro. The genksyms script will skip over those blocks when performing its CRC calculations.

But what about a real ABI change?

If a kpatch introduces a true ABI change, each of calling functions would consequently need to be updated in the kpatch module. For unexported functions, this may be handled safely if the kpatch does indeed update all callers. However, since motivation behind CONFIG_MODVERSIONS is to provide basic ABI verification between the kernel and modules for exported functions, kpatch cannot safely change this ABI without worrying about breaking other out-of-tree drivers. Those drivers have been built against the reference kernel's original set of CRCs and expect the original ABI.

To track down specifically what caused a symbol CRC change, tools like kabi-dw can be employed to produce a detailed symbol definition report. For a kpatch-build, kabi-dw can be modified to operate on .o object files (not just .ko and vmlinux files) and the $CACHEDIR/tmp/{orig, patched} directories compared.

System calls

Attempting to patch a syscall typically results in an error, due to a missing fentry hook in the inner __do_sys##name() function. The fentry hook is missing because of the 'inline' annotation, which invokes 'notrace'.

This problem can be worked around by adding #include "kpatch-syscall.h" and replacing the use of the SYSCALL_DEFINE1 (or similar) macro with the KPATCH_ prefixed version.