Skip to contentSkip to conversation

zram: fix slot lock bit position on big-endian 64-bit

PATCH Applied David Carlier 2 participants 2 messages Aug 10 – Aug 11

Subsystems (3)
Stable branch Block layer zram
lore
Share & tools
lore

David Carlier

author
#0 lore raw

PATCH 1/1 zram: fix slot lock bit position on big-endian 64-bit

The slot lock is a bit operation on the whole __lock word, which flags and ac_time alias as two u32s. On little-endian the lock bit lands in the position ZRAM_ENTRY_LOCK reserves in flags, so the aliasing works out. On 64-bit big-endian it lands in ac_time
instead: with ZRAM_TRACK_ENTRY_ACTIME enabled, storing the access time from mark_slot_accessed() or slot_free() wipes out the held lock bit, letting another CPU take the same slot lock; an access time value with that bit set makes the slot look locked forever.

Shift the lock bit into the flags half of the word on big-endian 64-bit.

Fixes 2e8ff2f51dde ("zram: use u32 for entry ac_time tracking")
Cc [email protected]
Signed-off-by David Carlier <[email protected]>
---
 drivers/block/zram/zram_drv.c |  6 +++---
 drivers/block/zram/zram_drv.h | 14 ++++++++++++++
 2 files changed, 17 insertions(+), 3 deletions(-)
drivers/block/zram/zram_drv.c +3 -3
index d09fdca49cbd..a9b3bb1d3bef 100644
@@ -74,7 +74,7 @@ static __must_check bool slot_trylock(struct zram *zram, unsigned long index)
 {
 	unsigned long *lock = &zram->table[index].__lock;
 
-	if (!test_and_set_bit_lock(ZRAM_ENTRY_LOCK, lock)) {
+	if (!test_and_set_bit_lock(ZRAM_ENTRY_LOCK_BIT, lock)) {
 		mutex_acquire(&zram->table_lock_map, 0, 1, _RET_IP_);
 		lock_acquired(&zram->table_lock_map, _RET_IP_);
 		return true;

@@ -88,7 +88,7 @@ static void slot_lock(struct zram *zram, unsigned long index)
 	unsigned long *lock = &zram->table[index].__lock;
 
 	mutex_acquire(&zram->table_lock_map, 0, 0, _RET_IP_);
-	wait_on_bit_lock(lock, ZRAM_ENTRY_LOCK, TASK_UNINTERRUPTIBLE);
+	wait_on_bit_lock(lock, ZRAM_ENTRY_LOCK_BIT, TASK_UNINTERRUPTIBLE);
 	lock_acquired(&zram->table_lock_map, _RET_IP_);
 }
 

@@ -97,7 +97,7 @@ static void slot_unlock(struct zram *zram, unsigned long index)
 	unsigned long *lock = &zram->table[index].__lock;
 
 	mutex_release(&zram->table_lock_map, _RET_IP_);
-	clear_and_wake_up_bit(ZRAM_ENTRY_LOCK, lock);
+	clear_and_wake_up_bit(ZRAM_ENTRY_LOCK_BIT, lock);
 }
 
 static inline bool init_done(struct zram *zram)
drivers/block/zram/zram_drv.h +14 -0
index 4fddc582f3b8..7a55d751417e 100644
@@ -15,6 +15,7 @@
 #ifndef _ZRAM_DRV_H_
 #define _ZRAM_DRV_H_
 
+#include <asm/byteorder.h>
 #include <linux/rwsem.h>
 #include <linux/zsmalloc.h>
 

@@ -57,6 +58,19 @@ enum zram_pageflags {
 	__NR_ZRAM_PAGEFLAGS,
 };
 
+/*
+ * The slot lock is a bit-wait lock on the whole __lock word, while
+ * flags and ac_time alias that word as two u32s.  The lock bit must
+ * land in the slot that ZRAM_ENTRY_LOCK reserves in attr.flags; on
+ * 64-bit big-endian the flags word maps to the upper half of __lock,
+ * so the bit position has to be shifted up.
+ */
+#if defined(CONFIG_64BIT) && defined(__BIG_ENDIAN)
+#define ZRAM_ENTRY_LOCK_BIT  (ZRAM_ENTRY_LOCK + 32)
+#else
+#define ZRAM_ENTRY_LOCK_BIT  ZRAM_ENTRY_LOCK
+#endif
+
 /*
  * Allocated for each disk page.  We use bit-lock (ZRAM_ENTRY_LOCK bit
  * of flags) to save memory.  There can be plenty of entries and standard
Signature · 1 line
2.55.0
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 9 lines · David Carlier

The slot lock is a bit operation on the whole __lock word, which flags and ac_time alias as two u32s. On little-endian the lock bit lands in the position ZRAM_ENTRY_LOCK reserves in flags, so the aliasing works out. On 64-bit big-endian it lands in ac_time
instead: with ZRAM_TRACK_ENTRY_ACTIME enabled, storing the access time from mark_slot_accessed() or slot_free() wipes out the held lock bit, letting another CPU take the same slot lock; an access time value with that bit set makes the slot look locked forever.

Quoted 6 lines

Shift the lock bit into the flags half of the word on big-endian 64-bit.

Fixes 2e8ff2f51dde ("zram: use u32 for entry ac_time tracking")
Cc [email protected]
Signed-off-by David Carlier <[email protected]>
Reviewed-by Sergey Senozhatsky <[email protected]>