This is a note to let you know that I've just added the patch titled arm64: futex: Bound number of LDXR/STXR loops in FUTEX_WAKE_OP to the 4.9-stable tree which can be found at: http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary The filename of the patch is: arm64-futex-bound-number-of-ldxr-stxr-loops-in-futex_wake_op.patch and it can be found in the queue-4.9 subdirectory. If you, or anyone else, feels it should not be added to the stable tree, please let <stable@xxxxxxxxxxxxxxx> know about it. >From foo@baz Mon Mar 29 07:48:09 AM CEST 2021 From: Ben Hutchings <ben@xxxxxxxxxxxxxxx> Date: Sun, 28 Mar 2021 22:42:51 +0200 Subject: arm64: futex: Bound number of LDXR/STXR loops in FUTEX_WAKE_OP To: stable@xxxxxxxxxxxxxxx Cc: Lee Jones <lee.jones@xxxxxxxxxx>, "Luis Claudio R. Goncalves" <lgoncalv@xxxxxxxxxx>, Florian Fainelli <f.fainelli@xxxxxxxxx> Message-ID: <YGDqS7bKz9n1Jto3@xxxxxxxxxxxxxxx> Content-Disposition: inline From: Will Deacon <will.deacon@xxxxxxx> commit 03110a5cb2161690ae5ac04994d47ed0cd6cef75 upstream. Our futex implementation makes use of LDXR/STXR loops to perform atomic updates to user memory from atomic context. This can lead to latency problems if we end up spinning around the LL/SC sequence at the expense of doing something useful. Rework our futex atomic operations so that we return -EAGAIN if we fail to update the futex word after 128 attempts. The core futex code will reschedule if necessary and we'll try again later. Fixes: 6170a97460db ("arm64: Atomic operations") Signed-off-by: Will Deacon <will.deacon@xxxxxxx> [bwh: Backported to 4.9: adjust context] Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- arch/arm64/include/asm/futex.h | 59 +++++++++++++++++++++++++---------------- 1 file changed, 37 insertions(+), 22 deletions(-) --- a/arch/arm64/include/asm/futex.h +++ b/arch/arm64/include/asm/futex.h @@ -26,7 +26,12 @@ #include <asm/errno.h> #include <asm/sysreg.h> +#define FUTEX_MAX_LOOPS 128 /* What's the largest number you can think of? */ + #define __futex_atomic_op(insn, ret, oldval, uaddr, tmp, oparg) \ +do { \ + unsigned int loops = FUTEX_MAX_LOOPS; \ + \ asm volatile( \ ALTERNATIVE("nop", SET_PSTATE_PAN(0), ARM64_HAS_PAN, \ CONFIG_ARM64_PAN) \ @@ -34,21 +39,26 @@ "1: ldxr %w1, %2\n" \ insn "\n" \ "2: stlxr %w0, %w3, %2\n" \ -" cbnz %w0, 1b\n" \ -" dmb ish\n" \ +" cbz %w0, 3f\n" \ +" sub %w4, %w4, %w0\n" \ +" cbnz %w4, 1b\n" \ +" mov %w0, %w7\n" \ "3:\n" \ +" dmb ish\n" \ " .pushsection .fixup,\"ax\"\n" \ " .align 2\n" \ -"4: mov %w0, %w5\n" \ +"4: mov %w0, %w6\n" \ " b 3b\n" \ " .popsection\n" \ _ASM_EXTABLE(1b, 4b) \ _ASM_EXTABLE(2b, 4b) \ ALTERNATIVE("nop", SET_PSTATE_PAN(1), ARM64_HAS_PAN, \ CONFIG_ARM64_PAN) \ - : "=&r" (ret), "=&r" (oldval), "+Q" (*uaddr), "=&r" (tmp) \ - : "r" (oparg), "Ir" (-EFAULT) \ - : "memory") + : "=&r" (ret), "=&r" (oldval), "+Q" (*uaddr), "=&r" (tmp), \ + "+r" (loops) \ + : "r" (oparg), "Ir" (-EFAULT), "Ir" (-EAGAIN) \ + : "memory"); \ +} while (0) static inline int arch_futex_atomic_op_inuser(int op, int oparg, int *oval, u32 __user *uaddr) @@ -59,23 +69,23 @@ arch_futex_atomic_op_inuser(int op, int switch (op) { case FUTEX_OP_SET: - __futex_atomic_op("mov %w3, %w4", + __futex_atomic_op("mov %w3, %w5", ret, oldval, uaddr, tmp, oparg); break; case FUTEX_OP_ADD: - __futex_atomic_op("add %w3, %w1, %w4", + __futex_atomic_op("add %w3, %w1, %w5", ret, oldval, uaddr, tmp, oparg); break; case FUTEX_OP_OR: - __futex_atomic_op("orr %w3, %w1, %w4", + __futex_atomic_op("orr %w3, %w1, %w5", ret, oldval, uaddr, tmp, oparg); break; case FUTEX_OP_ANDN: - __futex_atomic_op("and %w3, %w1, %w4", + __futex_atomic_op("and %w3, %w1, %w5", ret, oldval, uaddr, tmp, ~oparg); break; case FUTEX_OP_XOR: - __futex_atomic_op("eor %w3, %w1, %w4", + __futex_atomic_op("eor %w3, %w1, %w5", ret, oldval, uaddr, tmp, oparg); break; default: @@ -95,6 +105,7 @@ futex_atomic_cmpxchg_inatomic(u32 *uval, u32 oldval, u32 newval) { int ret = 0; + unsigned int loops = FUTEX_MAX_LOOPS; u32 val, tmp; u32 __user *uaddr; @@ -106,21 +117,25 @@ futex_atomic_cmpxchg_inatomic(u32 *uval, ALTERNATIVE("nop", SET_PSTATE_PAN(0), ARM64_HAS_PAN, CONFIG_ARM64_PAN) " prfm pstl1strm, %2\n" "1: ldxr %w1, %2\n" -" sub %w3, %w1, %w4\n" -" cbnz %w3, 3f\n" -"2: stlxr %w3, %w5, %2\n" -" cbnz %w3, 1b\n" -" dmb ish\n" +" sub %w3, %w1, %w5\n" +" cbnz %w3, 4f\n" +"2: stlxr %w3, %w6, %2\n" +" cbz %w3, 3f\n" +" sub %w4, %w4, %w3\n" +" cbnz %w4, 1b\n" +" mov %w0, %w8\n" "3:\n" +" dmb ish\n" +"4:\n" " .pushsection .fixup,\"ax\"\n" -"4: mov %w0, %w6\n" -" b 3b\n" +"5: mov %w0, %w7\n" +" b 4b\n" " .popsection\n" - _ASM_EXTABLE(1b, 4b) - _ASM_EXTABLE(2b, 4b) + _ASM_EXTABLE(1b, 5b) + _ASM_EXTABLE(2b, 5b) ALTERNATIVE("nop", SET_PSTATE_PAN(1), ARM64_HAS_PAN, CONFIG_ARM64_PAN) - : "+r" (ret), "=&r" (val), "+Q" (*uaddr), "=&r" (tmp) - : "r" (oldval), "r" (newval), "Ir" (-EFAULT) + : "+r" (ret), "=&r" (val), "+Q" (*uaddr), "=&r" (tmp), "+r" (loops) + : "r" (oldval), "r" (newval), "Ir" (-EFAULT), "Ir" (-EAGAIN) : "memory"); *uval = val; Patches currently in stable-queue which might be from ben@xxxxxxxxxxxxxxx are queue-4.9/futex-drop-hb-lock-before-enqueueing-on-the-rtmutex.patch queue-4.9/futex-rework-futex_lock_pi-to-use-rt_mutex_-_proxy_lock.patch queue-4.9/futex-avoid-freeing-an-active-timer.patch queue-4.9/futex-rt_mutex-introduce-rt_mutex_init_waiter.patch queue-4.9/futex-fix-incorrect-should_fail_futex-handling.patch queue-4.9/futex-rt_mutex-fix-rt_mutex_cleanup_proxy_lock.patch queue-4.9/futex-use-smp_store_release-in-mark_wake_futex.patch queue-4.9/futex-fix-possible-missed-wakeup.patch queue-4.9/locking-futex-allow-low-level-atomic-operations-to-return-eagain.patch queue-4.9/futex-prevent-robust-futex-exit-race.patch queue-4.9/futex-handle-transient-ownerless-rtmutex-state-correctly.patch queue-4.9/futex-handle-early-deadlock-return-correctly.patch queue-4.9/arm64-futex-bound-number-of-ldxr-stxr-loops-in-futex_wake_op.patch