Align atomic storage

This series adds the __aligned attribute to atomic_t and atomic64_t definitions in include/linux and include/asm-generic (respectively) to get natural alignment of both types on csky, m68k, microblaze, nios2, openrisc and sh.

This series also adds Kconfig options to enable a new run-time warning to help reveal misaligned atomic accesses on platforms which don't trap that.

The performance impact is expected to vary across platforms and workloads. The measurements I made on m68k show that some workloads run faster and others slower.

---

Changed since v4:
- Dropped parisc header file patch as it's been merged already.
- Submitted as PATCH instead of RFC.

Changed since v3:
- Rebased on v6.17.
- New patch to resolve header dependency issue on parisc.
- Dropped documentation patch.

Changed since v2:
- Specify natural alignment for atomic64_t.
- CONFIG_DEBUG_ATOMIC checks for natural alignment again.
- New patch to add weakened alignment check.
- New patch for explicit alignment in BPF header.

---

Finn Thain (3):
bpf: Explicitly align bpf_res_spin_lock
atomic: Specify alignment for atomic_t and atomic64_t
atomic: Add option for weaker alignment check

Peter Zijlstra (1):
atomic: Add alignment check to instrumented atomic operations

 include/asm-generic/atomic64.h   |  2 +-
 include/asm-generic/rqspinlock.h |  2 +-
 include/linux/instrumented.h     | 15 +++++++++++++++
 include/linux/types.h            |  2 +-
 kernel/bpf/rqspinlock.c          |  1 -
 lib/Kconfig.debug                | 18 ++++++++++++++++++
 6 files changed, 36 insertions(+), 4 deletions(-)

2.49.1

Align bpf_res_spin_lock to avoid a BUILD_BUG_ON() when the alignment changes, as it will do on m68k when, in a subsequent patch, the minimum alignment of the atomic_t member of struct rqspinlock gets increased from 2 to 4. Drop the BUILD_BUG_ON() as it becomes redundant.

---
 include/asm-generic/rqspinlock.h | 2 +-
 kernel/bpf/rqspinlock.c          | 1 -
 2 files changed, 1 insertion(+), 2 deletions(-)

index 6d4244d643df..eac2dcd31b83 100644
@@ -28,7 +28,7 @@ struct rqspinlock {
  */
 struct bpf_res_spin_lock {
 	u32 val;
-};
+} __aligned(__alignof__(struct rqspinlock));
 
 struct qspinlock;
 #ifdef CONFIG_QUEUED_SPINLOCKS

index a00561b1d3e5..02f1f671e624 100644
@@ -692,7 +692,6 @@ __bpf_kfunc int bpf_res_spin_lock(struct bpf_res_spin_lock *lock)
 	int ret;
 
 	BUILD_BUG_ON(sizeof(rqspinlock_t) != sizeof(struct bpf_res_spin_lock));
-	BUILD_BUG_ON(__alignof__(rqspinlock_t) != __alignof__(struct bpf_res_spin_lock));
 
 	preempt_disable();
 	ret = res_spin_lock((rqspinlock_t *)lock);

2.49.1

Some recent commits incorrectly assumed 4-byte alignment of locks. That assumption fails on Linux/m68k (and, interestingly, would have failed on Linux/cris also). The jump label implementation makes a similar alignment assumption.

The expectation that atomic_t and atomic64_t variables will be naturally aligned seems reasonable, as indeed they are on 64-bit architectures. But atomic64_t isn't naturally aligned on csky, m68k, microblaze, nios2, openrisc and sh. Neither atomic_t nor atomic64_t are naturally aligned on m68k.

This patch brings a little uniformity by specifying natural alignment for atomic types. One benefit is that atomic64_t variables do not get split across a page boundary. The cost is that some structs grow which leads to cache misses and wasted memory.

---
Changed since v2:
- Specify natural alignment for atomic64_t.
Changed since v1:
- atomic64_t now gets an __aligned attribute too.
- The 'Fixes' tag has been dropped because Lance sent a different fix for commit e711faaafbe5 ("hung_task: replace blocker_mutex with encoded blocker") that's suitable for -stable.

---
 include/asm-generic/atomic64.h | 2 +-
 include/linux/types.h          | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

index 100d24b02e52..f22ccfc0df98 100644
@@ -10,7 +10,7 @@
 #include <linux/types.h>
 
 typedef struct {
-	s64 counter;
+	s64 __aligned(sizeof(s64)) counter;
 } atomic64_t;
 
 #define ATOMIC64_INIT(i)	{ (i) }

index 6dfdb8e8e4c3..a225a518c2c3 100644
@@ -179,7 +179,7 @@ typedef phys_addr_t resource_size_t;
 typedef unsigned long irq_hw_number_t;
 
 typedef struct {
-	int counter;
+	int __aligned(sizeof(int)) counter;
 } atomic_t;
 
 #define ATOMIC_INIT(i) { (i) }

2.49.1

From: Peter Zijlstra <[email protected]>

Add a Kconfig option for debug builds which logs a warning when an instrumented atomic operation takes place that's misaligned. Some platforms don't trap for this.

---
Changed since v2:
- Always check for natural alignment.
---
Checkpatch.pl says...
ERROR: Missing Signed-off-by: line by nominal patch author 'Peter Zijlstra <[email protected]>'

---
 include/linux/instrumented.h |  4 ++++
 lib/Kconfig.debug            | 10 ++++++++++
 2 files changed, 14 insertions(+)

index 711a1f0d1a73..402a999a0d6b 100644
@@ -7,6 +7,7 @@
 #ifndef _LINUX_INSTRUMENTED_H
 #define _LINUX_INSTRUMENTED_H
 
+#include <linux/bug.h>
 #include <linux/compiler.h>
 #include <linux/kasan-checks.h>
 #include <linux/kcsan-checks.h>

@@ -67,6 +68,7 @@ static __always_inline void instrument_atomic_read(const volatile void *v, size_
 {
 	kasan_check_read(v, size);
 	kcsan_check_atomic_read(v, size);
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_DEBUG_ATOMIC) && ((unsigned long)v & (size - 1)));
 }
 
 /**

@@ -81,6 +83,7 @@ static __always_inline void instrument_atomic_write(const volatile void *v, size
 {
 	kasan_check_write(v, size);
 	kcsan_check_atomic_write(v, size);
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_DEBUG_ATOMIC) && ((unsigned long)v & (size - 1)));
 }
 
 /**

@@ -95,6 +98,7 @@ static __always_inline void instrument_atomic_read_write(const volatile void *v,
 {
 	kasan_check_write(v, size);
 	kcsan_check_atomic_read_write(v, size);
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_DEBUG_ATOMIC) && ((unsigned long)v & (size - 1)));
 }
 
 /**

index 713cc94caa02..d080d23d9a87 100644
@@ -1356,6 +1356,16 @@ config DEBUG_PREEMPT
 	  depending on workload as it triggers debugging routines for each
 	  this_cpu operation. It should only be used for debugging purposes.
 
+config DEBUG_ATOMIC
+	bool "Debug atomic variables"
+	depends on DEBUG_KERNEL
+	help
+	  If you say Y here then the kernel will add a runtime alignment check
+	  to atomic accesses. Useful for architectures that do not have trap on
+	  mis-aligned access.
+
+	  This option has potentially significant overhead.
+
 menu "Lock Debugging (spinlocks, mutexes, etc...)"
 
 config LOCK_DEBUGGING_SUPPORT

2.49.1

Add a new Kconfig symbol to make CONFIG_DEBUG_ATOMIC more useful on those architectures which do not align dynamic allocations to 8-byte boundaries. Without this, CONFIG_DEBUG_ATOMIC produces excessive WARN splats.

---
Changed since v3:
- Dropped #include <linux/cache.h> to avoid a build failure on arm64.
- Rewrote Kconfig help text to better describe preferred alignment.
- Refactored to avoid line-wrapping and duplication.

---
 include/linux/instrumented.h | 17 ++++++++++++++---
 lib/Kconfig.debug            |  8 ++++++++
 2 files changed, 22 insertions(+), 3 deletions(-)

index 402a999a0d6b..ca946c5860be 100644
@@ -56,6 +56,17 @@ static __always_inline void instrument_read_write(const volatile void *v, size_t
 	kcsan_check_read_write(v, size);
 }
 
+static __always_inline void instrument_atomic_check_alignment(const volatile void *v, size_t size)
+{
+	if (IS_ENABLED(CONFIG_DEBUG_ATOMIC)) {
+		unsigned int mask = size - 1;
+
+		if (IS_ENABLED(CONFIG_DEBUG_ATOMIC_LARGEST_ALIGN))
+			mask &= sizeof(struct { long x; } __aligned_largest) - 1;
+		WARN_ON_ONCE((unsigned long)v & mask);
+	}
+}
+
 /**
  * instrument_atomic_read - instrument atomic read access
  * @v: address of access

@@ -68,7 +79,7 @@ static __always_inline void instrument_atomic_read(const volatile void *v, size_
 {
 	kasan_check_read(v, size);
 	kcsan_check_atomic_read(v, size);
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_DEBUG_ATOMIC) && ((unsigned long)v & (size - 1)));
+	instrument_atomic_check_alignment(v, size);
 }
 
 /**

@@ -83,7 +94,7 @@ static __always_inline void instrument_atomic_write(const volatile void *v, size
 {
 	kasan_check_write(v, size);
 	kcsan_check_atomic_write(v, size);
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_DEBUG_ATOMIC) && ((unsigned long)v & (size - 1)));
+	instrument_atomic_check_alignment(v, size);
 }
 
 /**

@@ -98,7 +109,7 @@ static __always_inline void instrument_atomic_read_write(const volatile void *v,
 {
 	kasan_check_write(v, size);
 	kcsan_check_atomic_read_write(v, size);
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_DEBUG_ATOMIC) && ((unsigned long)v & (size - 1)));
+	instrument_atomic_check_alignment(v, size);
 }
 
 /**

index d080d23d9a87..60f9deef00fd 100644
@@ -1366,6 +1366,14 @@ config DEBUG_ATOMIC
 
 	  This option has potentially significant overhead.
 
+config DEBUG_ATOMIC_LARGEST_ALIGN
+	bool "Check alignment only up to __aligned_largest"
+	depends on DEBUG_ATOMIC
+	help
+	  If you say Y here then the check for natural alignment of
+	  atomic accesses will be constrained to the compiler's largest
+	  alignment for scalar types.
+
 menu "Lock Debugging (spinlocks, mutexes, etc...)"
 
 config LOCK_DEBUGGING_SUPPORT

2.49.1

Looks nice, thanks.

I don't know which tree this is aimed at. I grabbed a copy for mm.git's mm-nonmm-unstable tree, can drop it again if this pops up in linux-next via another route.