ftrace: Synchronize the initialization of ftrace_opsgrafted

There's some internal state that ftrace_ops needs to have set, but since
it can be declared outside of the ftrace.c code, it calls
ftrace_ops_init() on the ops in every global function. The issue is that
if two tasks call it on the same ops at the same time it is possible to
have the initialization of one corrupt the initialization of the other
call.

Create a ops_mutex to use to synchronize every initialization of the
ftrace_ops. The mutex is taken within checking the ftrace_ops flag that
states it was initializied but the flag is checked again after the mutex
has been taken. Checking first outside the mutex allows it to shortcut
having to take the mutex. But then the check needs to be done again after
the mute is taken in case of races.

Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20260902095501.6b59af20@gandalf.local.home
Fixes: f04f24fb7e48d ("ftrace, kprobes: Fix a deadlock on ftrace_regex_lock")
Reported-by: sashiko-bot@kernel.org
Close: https://lore.kernel.org/all/20260829025528.49A831F000E9@smtp.kernel.org/
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

1 files changed, 134 insertions, 0 deletions

diff --git a/arch/xtensa/lib/mulsi3.S b/arch/xtensa/lib/mulsi3.S
new file mode 100644
index 000000000..c6b4fd46b
--- /dev/null
+++ b/arch/xtensa/lib/mulsi3.S
@@ -0,0 +1,134 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later WITH GCC-exception-2.0 */
+#include <linux/linkage.h>
+#include <asm/asmmacro.h>
+#include <asm/core.h>
+
+	.macro	do_addx2 dst, as, at, tmp
+#if XCHAL_HAVE_ADDX
+	addx2	\dst, \as, \at
+#else
+	slli	\tmp, \as, 1
+	add	\dst, \tmp, \at
+#endif
+	.endm
+
+	.macro	do_addx4 dst, as, at, tmp
+#if XCHAL_HAVE_ADDX
+	addx4	\dst, \as, \at
+#else
+	slli	\tmp, \as, 2
+	add	\dst, \tmp, \at
+#endif
+	.endm
+
+	.macro	do_addx8 dst, as, at, tmp
+#if XCHAL_HAVE_ADDX
+	addx8	\dst, \as, \at
+#else
+	slli	\tmp, \as, 3
+	add	\dst, \tmp, \at
+#endif
+	.endm
+
+ENTRY(__mulsi3)
+
+	abi_entry_default
+
+#if XCHAL_HAVE_MUL32
+	mull	a2, a2, a3
+
+#elif XCHAL_HAVE_MUL16
+	or	a4, a2, a3
+	srai	a4, a4, 16
+	bnez	a4, .LMUL16
+	mul16u	a2, a2, a3
+	abi_ret_default
+.LMUL16:
+	srai	a4, a2, 16
+	srai	a5, a3, 16
+	mul16u	a7, a4, a3
+	mul16u	a6, a5, a2
+	mul16u	a4, a2, a3
+	add	a7, a7, a6
+	slli	a7, a7, 16
+	add	a2, a7, a4
+
+#elif XCHAL_HAVE_MAC16
+	mul.aa.hl a2, a3
+	mula.aa.lh a2, a3
+	rsr	a5, ACCLO
+	umul.aa.ll a2, a3
+	rsr	a4, ACCLO
+	slli	a5, a5, 16
+	add	a2, a4, a5
+
+#else /* !MUL32 && !MUL16 && !MAC16 */
+
+	/* Multiply one bit at a time, but unroll the loop 4x to better
+	   exploit the addx instructions and avoid overhead.
+	   Peel the first iteration to save a cycle on init.  */
+
+	/* Avoid negative numbers.  */
+	xor	a5, a2, a3	/* Top bit is 1 if one input is negative.  */
+	do_abs	a3, a3, a6
+	do_abs	a2, a2, a6
+
+	/* Swap so the second argument is smaller.  */
+	sub	a7, a2, a3
+	mov	a4, a3
+	movgez	a4, a2, a7	/* a4 = max (a2, a3) */
+	movltz	a3, a2, a7	/* a3 = min (a2, a3) */
+
+	movi	a2, 0
+	extui	a6, a3, 0, 1
+	movnez	a2, a4, a6
+
+	do_addx2 a7, a4, a2, a7
+	extui	a6, a3, 1, 1
+	movnez	a2, a7, a6
+
+	do_addx4 a7, a4, a2, a7
+	extui	a6, a3, 2, 1
+	movnez	a2, a7, a6
+
+	do_addx8 a7, a4, a2, a7
+	extui	a6, a3, 3, 1
+	movnez	a2, a7, a6
+
+	bgeui	a3, 16, .Lmult_main_loop
+	neg	a3, a2
+	movltz	a2, a3, a5
+	abi_ret_default
+
+	.align	4
+.Lmult_main_loop:
+	srli	a3, a3, 4
+	slli	a4, a4, 4
+
+	add	a7, a4, a2
+	extui	a6, a3, 0, 1
+	movnez	a2, a7, a6
+
+	do_addx2 a7, a4, a2, a7
+	extui	a6, a3, 1, 1
+	movnez	a2, a7, a6
+
+	do_addx4 a7, a4, a2, a7
+	extui	a6, a3, 2, 1
+	movnez	a2, a7, a6
+
+	do_addx8 a7, a4, a2, a7
+	extui	a6, a3, 3, 1
+	movnez	a2, a7, a6
+
+	bgeui	a3, 16, .Lmult_main_loop
+
+	neg	a3, a2
+	movltz	a2, a3, a5
+
+#endif /* !MUL32 && !MUL16 && !MAC16 */
+
+	abi_ret_default
+
+ENDPROC(__mulsi3)
+EXPORT_SYMBOL(__mulsi3)