* [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions
@ 2026-07-23 15:51 Aleksa Paunovic via B4 Relay
2026-07-27 21:03 ` Jesse Taube
2026-08-10 15:57 ` Conor Dooley
0 siblings, 2 replies; 6+ messages in thread
From: Aleksa Paunovic via B4 Relay @ 2026-07-23 15:51 UTC (permalink / raw)
To: Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti,
Will Deacon, Peter Zijlstra, Mark Rutland, Yury Norov,
Djordje Todorovic, Rasmus Villemoes, Charlie Jenkins,
Conor Dooley, Jessica Clarke, Samuel Holland, Paul Walmsley,
Boqun Feng, Gary Guo, Boqun Feng
Cc: linux-riscv, linux-kernel, Chao-ying Fu, Aleksandar Rikalo,
Aleksa Paunovic
From: Chao-ying Fu <cfu@mips.com>
MIPS P8700 does not natively support Zaamo instructions.
They are emulated with Zalrsc extension instructions instead [1].
Since the emulation is implemented through M-mode traps in the SBI
layer, it is best to avoid using these instructions wherever possible on
the P8700.
Implement kernel atomic operations using LR/SC sequences only.
This is achieved by using the errata mechanism, with minimal
interference on other cores.
Signed-off-by: Chao-ying Fu <cfu@mips.com>
Signed-off-by: Aleksandar Rikalo <arikalo@gmail.com>
Co-developed-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
[1] https://mips.com/wp-content/uploads/2026/03/MIPS_P8700_P8700-F_Programmers_Reference_Guide_Rev1.86_2-17-2026.pdf
---
The patch was tested on QEMU configured to emulate an eight-hart MIPS P8700 CPU.
Testing done since v3: futex kselftests and perf futex tests. These tests caught the issues described below.
The same tests were executed on the Boston board with a single-hart P8700 core.
Since the main issue was with an incorrectly written erratum, it shouldn't affect Vladimir's version [1].
However, since chips supporting only one part of the A extension are rare, we believe it might be
better to address this using the alternative mechanism, instead of demanding that the wider community
relax the A extension requirement.
Changes in v4:
- The amo part of the ALT_TEST_AND_OP_BIT_ORD erratum erroneously hardcoded zero as the destination register.
This is fixed in v4.
- futex.h was missing the ANDN case.
- Link to v3: https://lore.kernel.org/r/20250901-p8700-zalrsc-v3-1-ec64fabbe093@htecgroup.com
Changes in v3:
- Use alternatives to replace AMO instructions with LR/SC
- Rebase on Alexandre Ghiti's "for-next" branch.
- Link to v2: https://lore.kernel.org/linux-riscv/20241225082412.36727-1-arikalo@gmail.com/
Links:
[1] https://lore.kernel.org/linux-riscv/20260120-lrsc-only-v2-0-a522e640d27d@mobileye.com/
Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
---
arch/riscv/Kconfig.errata | 11 ++
arch/riscv/errata/mips/errata.c | 13 +-
arch/riscv/include/asm/atomic.h | 29 ++--
arch/riscv/include/asm/bitops.h | 28 ++--
arch/riscv/include/asm/cmpxchg.h | 9 +-
arch/riscv/include/asm/errata_list.h | 215 +++++++++++++++++++++++++++
arch/riscv/include/asm/errata_list_vendors.h | 3 +-
arch/riscv/include/asm/futex.h | 40 ++---
arch/riscv/kernel/entry.S | 10 +-
9 files changed, 290 insertions(+), 68 deletions(-)
diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
index 3c945d086c7d0266b685f9506d58b0662af071c4..cd5bd5e8eb395418c3ad103dbd836c775b7fd901 100644
--- a/arch/riscv/Kconfig.errata
+++ b/arch/riscv/Kconfig.errata
@@ -44,6 +44,17 @@ config ERRATA_MIPS_P8700_PAUSE_OPCODE
If you are not using the P8700 processor, say n.
+config ERRATA_MIPS_P8700_AMO_ZALRSC
+ bool "Replace AMO instructions with LR/SC on MIPS P8700"
+ depends on ERRATA_MIPS && 64BIT
+ default n
+ help
+ The MIPS P8700 does not implement the full A extension,
+ implementing only Zalrsc. Enabling this will replace
+ all AMO instructions with LR/SC instructions on the P8700.
+
+ If you are not using the P8700 processor, say n.
+
config ERRATA_SIFIVE
bool "SiFive errata"
depends on RISCV_ALTERNATIVE
diff --git a/arch/riscv/errata/mips/errata.c b/arch/riscv/errata/mips/errata.c
index e984a8152208c34690f89d8101571b097485c360..08c5efd58bf242d3831957f91f6338ef49f61238 100644
--- a/arch/riscv/errata/mips/errata.c
+++ b/arch/riscv/errata/mips/errata.c
@@ -23,13 +23,22 @@ static inline bool errata_probe_pause(void)
return true;
}
-static u32 mips_errata_probe(void)
+static inline bool errata_probe_zalrsc(unsigned long archid)
+{
+ return archid == 0x8000000000000201;
+}
+
+static u32 mips_errata_probe(unsigned long archid)
{
u32 cpu_req_errata = 0;
if (errata_probe_pause())
cpu_req_errata |= BIT(ERRATA_MIPS_P8700_PAUSE_OPCODE);
+ if (errata_probe_zalrsc(archid))
+ cpu_req_errata |= BIT(ERRATA_MIPS_P8700_ZALRSC);
+
+
return cpu_req_errata;
}
@@ -38,7 +47,7 @@ void mips_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
unsigned int stage)
{
struct alt_entry *alt;
- u32 cpu_req_errata = mips_errata_probe();
+ u32 cpu_req_errata = mips_errata_probe(archid);
u32 tmp;
BUILD_BUG_ON(ERRATA_MIPS_NUMBER >= RISCV_VENDOR_EXT_ALTERNATIVES_BASE);
diff --git a/arch/riscv/include/asm/atomic.h b/arch/riscv/include/asm/atomic.h
index 3f33dc54f94b2e14cb4270dbd0493625fbb977bd..9f848ec1705ff1afc6342e53965fab82309f9fe2 100644
--- a/arch/riscv/include/asm/atomic.h
+++ b/arch/riscv/include/asm/atomic.h
@@ -54,12 +54,9 @@ static __always_inline void arch_atomic64_set(atomic64_t *v, s64 i)
static __always_inline \
void arch_atomic##prefix##_##op(c_type i, atomic##prefix##_t *v) \
{ \
- __asm__ __volatile__ ( \
- " amo" #asm_op "." #asm_type " zero, %1, %0" \
- : "+A" (v->counter) \
- : "r" (I) \
- : "memory"); \
-} \
+ register __maybe_unused c_type ret, temp; \
+ ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp); \
+}
#ifdef CONFIG_GENERIC_ATOMIC64
#define ATOMIC_OPS(op, asm_op, I) \
@@ -89,24 +86,16 @@ static __always_inline \
c_type arch_atomic##prefix##_fetch_##op##_relaxed(c_type i, \
atomic##prefix##_t *v) \
{ \
- register c_type ret; \
- __asm__ __volatile__ ( \
- " amo" #asm_op "." #asm_type " %1, %2, %0" \
- : "+A" (v->counter), "=r" (ret) \
- : "r" (I) \
- : "memory"); \
+ register __maybe_unused c_type ret, temp; \
+ ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp); \
return ret; \
} \
static __always_inline \
c_type arch_atomic##prefix##_fetch_##op(c_type i, atomic##prefix##_t *v) \
-{ \
- register c_type ret; \
- __asm__ __volatile__ ( \
- " amo" #asm_op "." #asm_type ".aqrl %1, %2, %0" \
- : "+A" (v->counter), "=r" (ret) \
- : "r" (I) \
- : "memory"); \
- return ret; \
+{ \
+ register __maybe_unused c_type ret, temp; \
+ ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp); \
+ return ret; \
}
#define ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_type, c_type, prefix) \
diff --git a/arch/riscv/include/asm/bitops.h b/arch/riscv/include/asm/bitops.h
index 3c1a15be54d804d7f609d8f4033710970f7b215f..caf8048c29f34182ee0740e836a1305c450b68f4 100644
--- a/arch/riscv/include/asm/bitops.h
+++ b/arch/riscv/include/asm/bitops.h
@@ -170,30 +170,27 @@ static __always_inline int variable_fls(unsigned int x)
#if (BITS_PER_LONG == 64)
#define __AMO(op) "amo" #op ".d"
+#define __LR "lr.d"
+#define __SC "sc.d"
#elif (BITS_PER_LONG == 32)
#define __AMO(op) "amo" #op ".w"
+#define __LR "lr.w"
+#define __SC "sc.w"
#else
#error "Unexpected BITS_PER_LONG"
#endif
#define __test_and_op_bit_ord(op, mod, nr, addr, ord) \
({ \
- unsigned long __res, __mask; \
+ __maybe_unused unsigned long __res, __mask, __temp; \
__mask = BIT_MASK(nr); \
- __asm__ __volatile__ ( \
- __AMO(op) #ord " %0, %2, %1" \
- : "=r" (__res), "+A" (addr[BIT_WORD(nr)]) \
- : "r" (mod(__mask)) \
- : "memory"); \
+ ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp); \
((__res & __mask) != 0); \
})
#define __op_bit_ord(op, mod, nr, addr, ord) \
- __asm__ __volatile__ ( \
- __AMO(op) #ord " zero, %1, %0" \
- : "+A" (addr[BIT_WORD(nr)]) \
- : "r" (mod(BIT_MASK(nr))) \
- : "memory");
+ __maybe_unused unsigned long __res, __temp; \
+ ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp);
#define __test_and_op_bit(op, mod, nr, addr) \
__test_and_op_bit_ord(op, mod, nr, addr, .aqrl)
@@ -337,12 +334,9 @@ static __always_inline void arch___clear_bit_unlock(
static __always_inline bool arch_xor_unlock_is_negative_byte(unsigned long mask,
volatile unsigned long *addr)
{
- unsigned long res;
- __asm__ __volatile__ (
- __AMO(xor) ".rl %0, %2, %1"
- : "=r" (res), "+A" (*addr)
- : "r" (__NOP(mask))
- : "memory");
+ __maybe_unused unsigned long res, temp;
+
+ ALT_ARCH_XOR_UNLOCK(mask, addr, res, temp);
return (res & BIT(7)) != 0;
}
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index 8712cf9c69dcb3690058c60e60481fc4c3ffea06..21035ebe0e9193c1793ddc5488346fcb04d3ff3a 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/arch/riscv/include/asm/cmpxchg.h
@@ -56,13 +56,8 @@
#define __arch_xchg(sfx, prepend, append, r, p, n) \
({ \
- __asm__ __volatile__ ( \
- prepend \
- " amoswap" sfx " %0, %2, %1\n" \
- append \
- : "=r" (r), "+A" (*(p)) \
- : "r" (n) \
- : "memory"); \
+ __typeof__(*(__ptr)) __maybe_unused temp; \
+ ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp); \
})
#define _arch_xchg(ptr, new, sc_sfx, swap_sfx, prepend, \
diff --git a/arch/riscv/include/asm/errata_list.h b/arch/riscv/include/asm/errata_list.h
index 6694b5ccdcf85cfe7e767ea4de981b34f2b17b04..76c2f2a0f3b53f5507ecb2680d302679b3f97c69 100644
--- a/arch/riscv/include/asm/errata_list.h
+++ b/arch/riscv/include/asm/errata_list.h
@@ -25,6 +25,7 @@ ALTERNATIVE(__stringify(RISCV_PTR do_page_fault), \
__stringify(RISCV_PTR sifive_cip_453_page_fault_trp), \
SIFIVE_VENDOR_ID, ERRATA_SIFIVE_CIP_453, \
CONFIG_ERRATA_SIFIVE_CIP_453)
+
#else /* !__ASSEMBLER__ */
#define ALT_SFENCE_VMA_ASID(asid) \
@@ -53,6 +54,220 @@ asm(ALTERNATIVE( \
: /* no inputs */ \
: "memory")
+#ifdef CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC
+#define ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp) \
+asm(ALTERNATIVE( \
+ " amo" #asm_op "." #asm_type " zero, %3, %0\n" \
+ __nops(3), \
+ "1: lr." #asm_type " %1, %0\n" \
+ " " #asm_op " %2, %1, %3\n" \
+ " sc." #asm_type " %2, %2, %0\n" \
+ " bnez %2, 1b\n", \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
+ : "r" (I) \
+ : "memory")
+
+#define ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp) \
+asm(ALTERNATIVE( \
+ " amo" #asm_op "." #asm_type " %1, %3, %0\n" \
+ __nops(3), \
+ "1: lr." #asm_type " %1, %0\n" \
+ " " #asm_op " %2, %1, %3\n" \
+ " sc." #asm_type " %2, %2, %0\n" \
+ " bnez %2, 1b\n", \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
+ : "r" (I) \
+ : "memory")
+
+#define ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp) \
+asm(ALTERNATIVE( \
+ " amo" #asm_op "." #asm_type ".aqrl %1, %3, %0\n"\
+ __nops(3), \
+ "1: lr." #asm_type ".aqrl %1, %0\n" \
+ " " #asm_op " %2, %1, %3\n" \
+ " sc." #asm_type ".aqrl %2, %2, %0\n" \
+ " bnez %2, 1b\n", \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
+ : "r" (I) \
+ : "memory")
+/* BITOPS.h */
+#define ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp) \
+asm(ALTERNATIVE( \
+ __AMO(op) #ord " %0, %3, %1\n" \
+ __nops(3), \
+ "1: " __LR #ord " %0, %1\n" \
+ #op " %2, %0, %3\n" \
+ __SC #ord " %2, %2, %1\n" \
+ "bnez %2, 1b\n", \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "=&r" (__res), "+A" (addr[BIT_WORD(nr)]), "=&r" (__temp) \
+ : "r" (mod(__mask)) \
+ : "memory")
+
+#define ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp) \
+asm(ALTERNATIVE( \
+ __AMO(op) #ord " zero, %3, %1\n" \
+ __nops(3), \
+ "1: " __LR #ord " %0, %1\n" \
+ #op " %2, %0, %3\n" \
+ __SC #ord " %2, %2, %1\n" \
+ "bnez %2, 1b\n", \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "=&r" (__res), "+A" (addr[BIT_WORD(nr)]), "=&r" (__temp) \
+ : "r" (mod(BIT_MASK(nr))) \
+ : "memory")
+
+#define ALT_ARCH_XOR_UNLOCK(mask, addr, __res, __temp) \
+asm(ALTERNATIVE( \
+ __AMO(xor) ".rl %0, %3, %1\n" \
+ __nops(3), \
+ "1: " __LR ".rl %0, %1\n" \
+ "xor %2, %0, %3\n" \
+ __SC ".rl %2, %2, %1\n" \
+ "bnez %2, 1b\n", \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "=&r" (__res), "+A" (*addr), "=&r" (__temp) \
+ : "r" (__NOP(mask)) \
+ : "memory")
+
+#define ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp) \
+asm(ALTERNATIVE( \
+ prepend \
+ " amoswap" sfx " %0, %3, %1\n" \
+ __nops(2) \
+ append, \
+ prepend \
+ "1: lr" sfx " %0, %1\n" \
+ " sc" sfx " %2, %3, %1\n" \
+ " bnez %2, 1b\n" \
+ append, \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : "=&r" (r), "+A" (*(p)), "=&r" (temp) \
+ : "r" (n) \
+ : "memory")
+
+/* FUTEX.H */
+#define ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp) \
+asm(ALTERNATIVE( \
+ "1: amo" #insn ".w.aqrl %[ov],%z[op],%[u]\n" \
+ __nops(3) \
+ "2:\n" \
+ _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
+ "1: lr.w.aqrl %[ov], %[u]\n" \
+ " " #insn" %[t], %[ov], %z[op]\n" \
+ " sc.w.aqrl %[t], %[t], %[u]\n" \
+ " bnez %[t], 1b\n" \
+ "2:\n" \
+ _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : [r] "+r" (ret), [ov] "=&r" (oldval), \
+ [t] "=&r" (temp), [u] "+m" (*uaddr) \
+ : [op] "Jr" (oparg) \
+ : "memory")
+
+#define ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp) \
+asm(ALTERNATIVE( \
+ "1: amoswap.w.aqrl %[ov],%z[op],%[u]\n" \
+ __nops(3) \
+ "2:\n" \
+ _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
+ "1: lr.w.aqrl %[ov], %[u]\n" \
+ " mv %[t], %z[op]\n" \
+ " sc.w.aqrl %[t], %[t], %[u]\n" \
+ " bnez %[t], 1b\n" \
+ "2:\n" \
+ _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
+ MIPS_VENDOR_ID, \
+ ERRATA_MIPS_P8700_ZALRSC, \
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
+ : [r] "+r" (ret), [ov] "=&r" (oldval), \
+ [t] "=&r" (temp), [u] "+m" (*uaddr) \
+ : [op] "Jr" (oparg) \
+ : "memory")
+
+#else
+#define ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp) \
+asm("amo" #asm_op "." #asm_type " zero, %1, %0" \
+ : "+A" (v->counter) \
+ : "r" (I) \
+ : "memory")
+
+#define ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp) \
+asm("amo" #asm_op "." #asm_type " %1, %2, %0" \
+ : "+A" (v->counter), "=r" (ret) \
+ : "r" (I) \
+ : "memory")
+
+#define ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp) \
+asm("amo" #asm_op "." #asm_type ".aqrl %1, %2, %0" \
+ : "+A" (v->counter), "=r" (ret) \
+ : "r" (I) \
+ : "memory")
+
+#define ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp) \
+asm(__AMO(op) #ord " %0, %2, %1" \
+ : "=r" (__res), "+A" (addr[BIT_WORD(nr)]) \
+ : "r" (mod(__mask)) \
+ : "memory")
+
+#define ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp) \
+asm(__AMO(op) #ord " zero, %1, %0" \
+ : "+A" (addr[BIT_WORD(nr)]) \
+ : "r" (mod(BIT_MASK(nr))) \
+ : "memory")
+
+#define ALT_ARCH_XOR_UNLOCK(mask, addr, __res, __temp) \
+asm(__AMO(xor) ".rl %0, %2, %1" \
+ : "=r" (res), "+A" (*addr) \
+ : "r" (__NOP(mask)) \
+ : "memory")
+
+#define ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp) \
+asm(prepend \
+ " amoswap" sfx " %0, %2, %1\n" \
+ append \
+ : "=r" (r), "+A" (*(p)) \
+ : "r" (n) \
+ : "memory")
+
+#define ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp) \
+asm("1: amo" #insn ".w.aqrl %[ov],%z[op],%[u]\n" \
+ "2:\n" \
+ _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
+ : [r] "+r" (ret), [ov] "=&r" (oldval), \
+ [u] "+m" (*uaddr) \
+ : [op] "Jr" (oparg) \
+ : "memory")
+
+#define ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp) \
+asm("1: amoswap.w.aqrl %[ov],%z[op],%[u]\n" \
+ "2:\n" \
+ _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
+ : [r] "+r" (ret), [ov] "=&r" (oldval), \
+ [u] "+m" (*uaddr) \
+ : [op] "Jr" (oparg) \
+ : "memory")
+#endif
+
/*
* _val is marked as "will be overwritten", so need to set it to 0
* in the default case.
diff --git a/arch/riscv/include/asm/errata_list_vendors.h b/arch/riscv/include/asm/errata_list_vendors.h
index ec7eba3734371a2d8b68fbd4cbd88a8e7135a413..d47a98ab93d12ce1f6ae44449f80bb703987e985 100644
--- a/arch/riscv/include/asm/errata_list_vendors.h
+++ b/arch/riscv/include/asm/errata_list_vendors.h
@@ -23,7 +23,8 @@
#ifdef CONFIG_ERRATA_MIPS
#define ERRATA_MIPS_P8700_PAUSE_OPCODE 0
-#define ERRATA_MIPS_NUMBER 1
+#define ERRATA_MIPS_P8700_ZALRSC 1
+#define ERRATA_MIPS_NUMBER 2
#endif
#endif /* ASM_ERRATA_LIST_VENDORS_H */
diff --git a/arch/riscv/include/asm/futex.h b/arch/riscv/include/asm/futex.h
index 90c86b115e008a1fb08f3da64382fb4a64d9cc2f..bc3e8c0586410d1fc3a1a7d35b7775f9c339719f 100644
--- a/arch/riscv/include/asm/futex.h
+++ b/arch/riscv/include/asm/futex.h
@@ -12,6 +12,7 @@
#include <linux/errno.h>
#include <asm/asm.h>
#include <asm/asm-extable.h>
+#include <asm/errata_list.h>
/* We don't even really need the extable code, but for now keep it simple */
#ifndef CONFIG_MMU
@@ -19,48 +20,47 @@
#define __disable_user_access() do { } while (0)
#endif
-#define __futex_atomic_op(insn, ret, oldval, uaddr, oparg) \
+#define __futex_atomic_op(insn, ret, oldval, uaddr, oparg, temp) \
{ \
__enable_user_access(); \
- __asm__ __volatile__ ( \
- "1: " insn " \n" \
- "2: \n" \
- _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
- : [r] "+r" (ret), [ov] "=&r" (oldval), \
- [u] "+m" (*uaddr) \
- : [op] "Jr" (oparg) \
- : "memory"); \
+ ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp); \
+ __disable_user_access(); \
+}
+
+#define __futex_atomic_swap(ret, oldval, uaddr, oparg, temp) \
+{ \
+ __enable_user_access(); \
+ ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp); \
__disable_user_access(); \
}
static inline int
arch_futex_atomic_op_inuser(int op, int oparg, int *oval, u32 __user *uaddr)
{
- int oldval = 0, ret = 0;
+ int __maybe_unused oldval = 0, ret = 0, temp = 0;
if (!access_ok(uaddr, sizeof(u32)))
return -EFAULT;
switch (op) {
case FUTEX_OP_SET:
- __futex_atomic_op("amoswap.w.aqrl %[ov],%z[op],%[u]",
- ret, oldval, uaddr, oparg);
+ __futex_atomic_swap(ret, oldval, uaddr, oparg, temp);
break;
case FUTEX_OP_ADD:
- __futex_atomic_op("amoadd.w.aqrl %[ov],%z[op],%[u]",
- ret, oldval, uaddr, oparg);
+ __futex_atomic_op(add,
+ ret, oldval, uaddr, oparg, temp);
break;
case FUTEX_OP_OR:
- __futex_atomic_op("amoor.w.aqrl %[ov],%z[op],%[u]",
- ret, oldval, uaddr, oparg);
+ __futex_atomic_op(or,
+ ret, oldval, uaddr, oparg, temp);
break;
case FUTEX_OP_ANDN:
- __futex_atomic_op("amoand.w.aqrl %[ov],%z[op],%[u]",
- ret, oldval, uaddr, ~oparg);
+ __futex_atomic_op(and,
+ ret, oldval, uaddr, ~oparg, temp);
break;
case FUTEX_OP_XOR:
- __futex_atomic_op("amoxor.w.aqrl %[ov],%z[op],%[u]",
- ret, oldval, uaddr, oparg);
+ __futex_atomic_op(xor,
+ ret, oldval, uaddr, oparg, temp);
break;
default:
ret = -ENOSYS;
diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index 60eb221296a604694c5f936d0c23c637dc44298e..8a7f9cb57f8450fd04107e2b57096c41bcefb722 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -73,7 +73,15 @@
beq a2, zero, .Lnew_vmalloc_restore_context
/* Atomically reset the current cpu bit in new_vmalloc */
- amoxor.d a0, a1, (a0)
+ ALTERNATIVE("amoxor.d a0, a1, (a0); \
+ .rept 3; nop; .endr;",
+ "1: lr.d a2, (a0); \
+ xor a2, a2, a1; \
+ sc.d a2, a2, (a0); \
+ bnez a2, 1b;",
+ MIPS_VENDOR_ID,
+ ERRATA_MIPS_P8700_ZALRSC,
+ CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC)
/* Only emit a sfence.vma if the uarch caches invalid entries */
ALTERNATIVE("sfence.vma", "nop", 0, RISCV_ISA_EXT_SVVPTC, 1)
---
base-commit: c369299895a591d96745d6492d4888259b004a9e
change-id: 20250714-p8700-zalrsc-f3894be40d06
Best regards,
--
Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions
2026-07-23 15:51 [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions Aleksa Paunovic via B4 Relay
@ 2026-07-27 21:03 ` Jesse Taube
2026-07-27 22:37 ` Jesse Taube
2026-08-10 15:57 ` Conor Dooley
1 sibling, 1 reply; 6+ messages in thread
From: Jesse Taube @ 2026-07-27 21:03 UTC (permalink / raw)
To: aleksa.paunovic
Cc: Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti,
Will Deacon, Peter Zijlstra, Mark Rutland, Yury Norov,
Djordje Todorovic, Rasmus Villemoes, Charlie Jenkins,
Conor Dooley, Jessica Clarke, Samuel Holland, Paul Walmsley,
Boqun Feng, Gary Guo, linux-riscv, linux-kernel, Chao-ying Fu,
Aleksandar Rikalo, Charles Mirabile
On Thu, Jul 23, 2026 at 11:53 AM Aleksa Paunovic via B4 Relay
<devnull+aleksa.paunovic.htecgroup.com@kernel.org> wrote:
>
> From: Chao-ying Fu <cfu@mips.com>
>
> MIPS P8700 does not natively support Zaamo instructions.
> They are emulated with Zalrsc extension instructions instead [1].
> Since the emulation is implemented through M-mode traps in the SBI
> layer, it is best to avoid using these instructions wherever possible on
> the P8700.
>
> Implement kernel atomic operations using LR/SC sequences only.
> This is achieved by using the errata mechanism, with minimal
> interference on other cores.
>
> Signed-off-by: Chao-ying Fu <cfu@mips.com>
> Signed-off-by: Aleksandar Rikalo <arikalo@gmail.com>
> Co-developed-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
>
> [1] https://mips.com/wp-content/uploads/2026/03/MIPS_P8700_P8700-F_Programmers_Reference_Guide_Rev1.86_2-17-2026.pdf
>
> ---
> The patch was tested on QEMU configured to emulate an eight-hart MIPS P8700 CPU.
Can you share the tests. I made my own tests a while back here:
https://github.com/Mr-Bossman/zalrsc-buildroot/tree/master
Though they don't seem to boot without the zaamo extension
Thanks,
Jesse Taube
> Testing done since v3: futex kselftests and perf futex tests. These tests caught the issues described below.
> The same tests were executed on the Boston board with a single-hart P8700 core.
>
> Since the main issue was with an incorrectly written erratum, it shouldn't affect Vladimir's version [1].
> However, since chips supporting only one part of the A extension are rare, we believe it might be
> better to address this using the alternative mechanism, instead of demanding that the wider community
> relax the A extension requirement.
>
> Changes in v4:
> - The amo part of the ALT_TEST_AND_OP_BIT_ORD erratum erroneously hardcoded zero as the destination register.
> This is fixed in v4.
> - futex.h was missing the ANDN case.
> - Link to v3: https://lore.kernel.org/r/20250901-p8700-zalrsc-v3-1-ec64fabbe093@htecgroup.com
>
> Changes in v3:
> - Use alternatives to replace AMO instructions with LR/SC
> - Rebase on Alexandre Ghiti's "for-next" branch.
> - Link to v2: https://lore.kernel.org/linux-riscv/20241225082412.36727-1-arikalo@gmail.com/
>
> Links:
> [1] https://lore.kernel.org/linux-riscv/20260120-lrsc-only-v2-0-a522e640d27d@mobileye.com/
>
> Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> ---
> arch/riscv/Kconfig.errata | 11 ++
> arch/riscv/errata/mips/errata.c | 13 +-
> arch/riscv/include/asm/atomic.h | 29 ++--
> arch/riscv/include/asm/bitops.h | 28 ++--
> arch/riscv/include/asm/cmpxchg.h | 9 +-
> arch/riscv/include/asm/errata_list.h | 215 +++++++++++++++++++++++++++
> arch/riscv/include/asm/errata_list_vendors.h | 3 +-
> arch/riscv/include/asm/futex.h | 40 ++---
> arch/riscv/kernel/entry.S | 10 +-
> 9 files changed, 290 insertions(+), 68 deletions(-)
>
> diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
> index 3c945d086c7d0266b685f9506d58b0662af071c4..cd5bd5e8eb395418c3ad103dbd836c775b7fd901 100644
> --- a/arch/riscv/Kconfig.errata
> +++ b/arch/riscv/Kconfig.errata
> @@ -44,6 +44,17 @@ config ERRATA_MIPS_P8700_PAUSE_OPCODE
>
> If you are not using the P8700 processor, say n.
>
> +config ERRATA_MIPS_P8700_AMO_ZALRSC
> + bool "Replace AMO instructions with LR/SC on MIPS P8700"
> + depends on ERRATA_MIPS && 64BIT
> + default n
> + help
> + The MIPS P8700 does not implement the full A extension,
> + implementing only Zalrsc. Enabling this will replace
> + all AMO instructions with LR/SC instructions on the P8700.
> +
> + If you are not using the P8700 processor, say n.
> +
> config ERRATA_SIFIVE
> bool "SiFive errata"
> depends on RISCV_ALTERNATIVE
> diff --git a/arch/riscv/errata/mips/errata.c b/arch/riscv/errata/mips/errata.c
> index e984a8152208c34690f89d8101571b097485c360..08c5efd58bf242d3831957f91f6338ef49f61238 100644
> --- a/arch/riscv/errata/mips/errata.c
> +++ b/arch/riscv/errata/mips/errata.c
> @@ -23,13 +23,22 @@ static inline bool errata_probe_pause(void)
> return true;
> }
>
> -static u32 mips_errata_probe(void)
> +static inline bool errata_probe_zalrsc(unsigned long archid)
> +{
> + return archid == 0x8000000000000201;
> +}
> +
> +static u32 mips_errata_probe(unsigned long archid)
> {
> u32 cpu_req_errata = 0;
>
> if (errata_probe_pause())
> cpu_req_errata |= BIT(ERRATA_MIPS_P8700_PAUSE_OPCODE);
>
> + if (errata_probe_zalrsc(archid))
> + cpu_req_errata |= BIT(ERRATA_MIPS_P8700_ZALRSC);
> +
> +
> return cpu_req_errata;
> }
>
> @@ -38,7 +47,7 @@ void mips_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
> unsigned int stage)
> {
> struct alt_entry *alt;
> - u32 cpu_req_errata = mips_errata_probe();
> + u32 cpu_req_errata = mips_errata_probe(archid);
> u32 tmp;
>
> BUILD_BUG_ON(ERRATA_MIPS_NUMBER >= RISCV_VENDOR_EXT_ALTERNATIVES_BASE);
> diff --git a/arch/riscv/include/asm/atomic.h b/arch/riscv/include/asm/atomic.h
> index 3f33dc54f94b2e14cb4270dbd0493625fbb977bd..9f848ec1705ff1afc6342e53965fab82309f9fe2 100644
> --- a/arch/riscv/include/asm/atomic.h
> +++ b/arch/riscv/include/asm/atomic.h
> @@ -54,12 +54,9 @@ static __always_inline void arch_atomic64_set(atomic64_t *v, s64 i)
> static __always_inline \
> void arch_atomic##prefix##_##op(c_type i, atomic##prefix##_t *v) \
> { \
> - __asm__ __volatile__ ( \
> - " amo" #asm_op "." #asm_type " zero, %1, %0" \
> - : "+A" (v->counter) \
> - : "r" (I) \
> - : "memory"); \
> -} \
> + register __maybe_unused c_type ret, temp; \
> + ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp); \
> +}
>
> #ifdef CONFIG_GENERIC_ATOMIC64
> #define ATOMIC_OPS(op, asm_op, I) \
> @@ -89,24 +86,16 @@ static __always_inline \
> c_type arch_atomic##prefix##_fetch_##op##_relaxed(c_type i, \
> atomic##prefix##_t *v) \
> { \
> - register c_type ret; \
> - __asm__ __volatile__ ( \
> - " amo" #asm_op "." #asm_type " %1, %2, %0" \
> - : "+A" (v->counter), "=r" (ret) \
> - : "r" (I) \
> - : "memory"); \
> + register __maybe_unused c_type ret, temp; \
> + ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp); \
> return ret; \
> } \
> static __always_inline \
> c_type arch_atomic##prefix##_fetch_##op(c_type i, atomic##prefix##_t *v) \
> -{ \
> - register c_type ret; \
> - __asm__ __volatile__ ( \
> - " amo" #asm_op "." #asm_type ".aqrl %1, %2, %0" \
> - : "+A" (v->counter), "=r" (ret) \
> - : "r" (I) \
> - : "memory"); \
> - return ret; \
> +{ \
> + register __maybe_unused c_type ret, temp; \
> + ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp); \
> + return ret; \
> }
>
> #define ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_type, c_type, prefix) \
> diff --git a/arch/riscv/include/asm/bitops.h b/arch/riscv/include/asm/bitops.h
> index 3c1a15be54d804d7f609d8f4033710970f7b215f..caf8048c29f34182ee0740e836a1305c450b68f4 100644
> --- a/arch/riscv/include/asm/bitops.h
> +++ b/arch/riscv/include/asm/bitops.h
> @@ -170,30 +170,27 @@ static __always_inline int variable_fls(unsigned int x)
>
> #if (BITS_PER_LONG == 64)
> #define __AMO(op) "amo" #op ".d"
> +#define __LR "lr.d"
> +#define __SC "sc.d"
> #elif (BITS_PER_LONG == 32)
> #define __AMO(op) "amo" #op ".w"
> +#define __LR "lr.w"
> +#define __SC "sc.w"
> #else
> #error "Unexpected BITS_PER_LONG"
> #endif
>
> #define __test_and_op_bit_ord(op, mod, nr, addr, ord) \
> ({ \
> - unsigned long __res, __mask; \
> + __maybe_unused unsigned long __res, __mask, __temp; \
> __mask = BIT_MASK(nr); \
> - __asm__ __volatile__ ( \
> - __AMO(op) #ord " %0, %2, %1" \
> - : "=r" (__res), "+A" (addr[BIT_WORD(nr)]) \
> - : "r" (mod(__mask)) \
> - : "memory"); \
> + ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp); \
> ((__res & __mask) != 0); \
> })
>
> #define __op_bit_ord(op, mod, nr, addr, ord) \
> - __asm__ __volatile__ ( \
> - __AMO(op) #ord " zero, %1, %0" \
> - : "+A" (addr[BIT_WORD(nr)]) \
> - : "r" (mod(BIT_MASK(nr))) \
> - : "memory");
> + __maybe_unused unsigned long __res, __temp; \
> + ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp);
>
> #define __test_and_op_bit(op, mod, nr, addr) \
> __test_and_op_bit_ord(op, mod, nr, addr, .aqrl)
> @@ -337,12 +334,9 @@ static __always_inline void arch___clear_bit_unlock(
> static __always_inline bool arch_xor_unlock_is_negative_byte(unsigned long mask,
> volatile unsigned long *addr)
> {
> - unsigned long res;
> - __asm__ __volatile__ (
> - __AMO(xor) ".rl %0, %2, %1"
> - : "=r" (res), "+A" (*addr)
> - : "r" (__NOP(mask))
> - : "memory");
> + __maybe_unused unsigned long res, temp;
> +
> + ALT_ARCH_XOR_UNLOCK(mask, addr, res, temp);
> return (res & BIT(7)) != 0;
> }
>
> diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
> index 8712cf9c69dcb3690058c60e60481fc4c3ffea06..21035ebe0e9193c1793ddc5488346fcb04d3ff3a 100644
> --- a/arch/riscv/include/asm/cmpxchg.h
> +++ b/arch/riscv/include/asm/cmpxchg.h
> @@ -56,13 +56,8 @@
>
> #define __arch_xchg(sfx, prepend, append, r, p, n) \
> ({ \
> - __asm__ __volatile__ ( \
> - prepend \
> - " amoswap" sfx " %0, %2, %1\n" \
> - append \
> - : "=r" (r), "+A" (*(p)) \
> - : "r" (n) \
> - : "memory"); \
> + __typeof__(*(__ptr)) __maybe_unused temp; \
> + ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp); \
> })
>
> #define _arch_xchg(ptr, new, sc_sfx, swap_sfx, prepend, \
> diff --git a/arch/riscv/include/asm/errata_list.h b/arch/riscv/include/asm/errata_list.h
> index 6694b5ccdcf85cfe7e767ea4de981b34f2b17b04..76c2f2a0f3b53f5507ecb2680d302679b3f97c69 100644
> --- a/arch/riscv/include/asm/errata_list.h
> +++ b/arch/riscv/include/asm/errata_list.h
> @@ -25,6 +25,7 @@ ALTERNATIVE(__stringify(RISCV_PTR do_page_fault), \
> __stringify(RISCV_PTR sifive_cip_453_page_fault_trp), \
> SIFIVE_VENDOR_ID, ERRATA_SIFIVE_CIP_453, \
> CONFIG_ERRATA_SIFIVE_CIP_453)
> +
> #else /* !__ASSEMBLER__ */
>
> #define ALT_SFENCE_VMA_ASID(asid) \
> @@ -53,6 +54,220 @@ asm(ALTERNATIVE( \
> : /* no inputs */ \
> : "memory")
>
> +#ifdef CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC
> +#define ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp) \
> +asm(ALTERNATIVE( \
> + " amo" #asm_op "." #asm_type " zero, %3, %0\n" \
> + __nops(3), \
> + "1: lr." #asm_type " %1, %0\n" \
> + " " #asm_op " %2, %1, %3\n" \
> + " sc." #asm_type " %2, %2, %0\n" \
> + " bnez %2, 1b\n", \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
> + : "r" (I) \
> + : "memory")
> +
> +#define ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp) \
> +asm(ALTERNATIVE( \
> + " amo" #asm_op "." #asm_type " %1, %3, %0\n" \
> + __nops(3), \
> + "1: lr." #asm_type " %1, %0\n" \
> + " " #asm_op " %2, %1, %3\n" \
> + " sc." #asm_type " %2, %2, %0\n" \
> + " bnez %2, 1b\n", \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
> + : "r" (I) \
> + : "memory")
> +
> +#define ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp) \
> +asm(ALTERNATIVE( \
> + " amo" #asm_op "." #asm_type ".aqrl %1, %3, %0\n"\
> + __nops(3), \
> + "1: lr." #asm_type ".aqrl %1, %0\n" \
> + " " #asm_op " %2, %1, %3\n" \
> + " sc." #asm_type ".aqrl %2, %2, %0\n" \
> + " bnez %2, 1b\n", \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
> + : "r" (I) \
> + : "memory")
> +/* BITOPS.h */
> +#define ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp) \
> +asm(ALTERNATIVE( \
> + __AMO(op) #ord " %0, %3, %1\n" \
> + __nops(3), \
> + "1: " __LR #ord " %0, %1\n" \
> + #op " %2, %0, %3\n" \
> + __SC #ord " %2, %2, %1\n" \
> + "bnez %2, 1b\n", \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "=&r" (__res), "+A" (addr[BIT_WORD(nr)]), "=&r" (__temp) \
> + : "r" (mod(__mask)) \
> + : "memory")
> +
> +#define ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp) \
> +asm(ALTERNATIVE( \
> + __AMO(op) #ord " zero, %3, %1\n" \
> + __nops(3), \
> + "1: " __LR #ord " %0, %1\n" \
> + #op " %2, %0, %3\n" \
> + __SC #ord " %2, %2, %1\n" \
> + "bnez %2, 1b\n", \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "=&r" (__res), "+A" (addr[BIT_WORD(nr)]), "=&r" (__temp) \
> + : "r" (mod(BIT_MASK(nr))) \
> + : "memory")
> +
> +#define ALT_ARCH_XOR_UNLOCK(mask, addr, __res, __temp) \
> +asm(ALTERNATIVE( \
> + __AMO(xor) ".rl %0, %3, %1\n" \
> + __nops(3), \
> + "1: " __LR ".rl %0, %1\n" \
> + "xor %2, %0, %3\n" \
> + __SC ".rl %2, %2, %1\n" \
> + "bnez %2, 1b\n", \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "=&r" (__res), "+A" (*addr), "=&r" (__temp) \
> + : "r" (__NOP(mask)) \
> + : "memory")
> +
> +#define ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp) \
> +asm(ALTERNATIVE( \
> + prepend \
> + " amoswap" sfx " %0, %3, %1\n" \
> + __nops(2) \
> + append, \
> + prepend \
> + "1: lr" sfx " %0, %1\n" \
> + " sc" sfx " %2, %3, %1\n" \
> + " bnez %2, 1b\n" \
> + append, \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : "=&r" (r), "+A" (*(p)), "=&r" (temp) \
> + : "r" (n) \
> + : "memory")
> +
> +/* FUTEX.H */
> +#define ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp) \
> +asm(ALTERNATIVE( \
> + "1: amo" #insn ".w.aqrl %[ov],%z[op],%[u]\n" \
> + __nops(3) \
> + "2:\n" \
> + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> + "1: lr.w.aqrl %[ov], %[u]\n" \
> + " " #insn" %[t], %[ov], %z[op]\n" \
> + " sc.w.aqrl %[t], %[t], %[u]\n" \
> + " bnez %[t], 1b\n" \
> + "2:\n" \
> + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> + [t] "=&r" (temp), [u] "+m" (*uaddr) \
> + : [op] "Jr" (oparg) \
> + : "memory")
> +
> +#define ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp) \
> +asm(ALTERNATIVE( \
> + "1: amoswap.w.aqrl %[ov],%z[op],%[u]\n" \
> + __nops(3) \
> + "2:\n" \
> + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> + "1: lr.w.aqrl %[ov], %[u]\n" \
> + " mv %[t], %z[op]\n" \
> + " sc.w.aqrl %[t], %[t], %[u]\n" \
> + " bnez %[t], 1b\n" \
> + "2:\n" \
> + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> + MIPS_VENDOR_ID, \
> + ERRATA_MIPS_P8700_ZALRSC, \
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> + [t] "=&r" (temp), [u] "+m" (*uaddr) \
> + : [op] "Jr" (oparg) \
> + : "memory")
> +
> +#else
> +#define ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp) \
> +asm("amo" #asm_op "." #asm_type " zero, %1, %0" \
> + : "+A" (v->counter) \
> + : "r" (I) \
> + : "memory")
> +
> +#define ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp) \
> +asm("amo" #asm_op "." #asm_type " %1, %2, %0" \
> + : "+A" (v->counter), "=r" (ret) \
> + : "r" (I) \
> + : "memory")
> +
> +#define ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp) \
> +asm("amo" #asm_op "." #asm_type ".aqrl %1, %2, %0" \
> + : "+A" (v->counter), "=r" (ret) \
> + : "r" (I) \
> + : "memory")
> +
> +#define ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp) \
> +asm(__AMO(op) #ord " %0, %2, %1" \
> + : "=r" (__res), "+A" (addr[BIT_WORD(nr)]) \
> + : "r" (mod(__mask)) \
> + : "memory")
> +
> +#define ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp) \
> +asm(__AMO(op) #ord " zero, %1, %0" \
> + : "+A" (addr[BIT_WORD(nr)]) \
> + : "r" (mod(BIT_MASK(nr))) \
> + : "memory")
> +
> +#define ALT_ARCH_XOR_UNLOCK(mask, addr, __res, __temp) \
> +asm(__AMO(xor) ".rl %0, %2, %1" \
> + : "=r" (res), "+A" (*addr) \
> + : "r" (__NOP(mask)) \
> + : "memory")
> +
> +#define ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp) \
> +asm(prepend \
> + " amoswap" sfx " %0, %2, %1\n" \
> + append \
> + : "=r" (r), "+A" (*(p)) \
> + : "r" (n) \
> + : "memory")
> +
> +#define ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp) \
> +asm("1: amo" #insn ".w.aqrl %[ov],%z[op],%[u]\n" \
> + "2:\n" \
> + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
> + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> + [u] "+m" (*uaddr) \
> + : [op] "Jr" (oparg) \
> + : "memory")
> +
> +#define ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp) \
> +asm("1: amoswap.w.aqrl %[ov],%z[op],%[u]\n" \
> + "2:\n" \
> + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
> + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> + [u] "+m" (*uaddr) \
> + : [op] "Jr" (oparg) \
> + : "memory")
> +#endif
> +
> /*
> * _val is marked as "will be overwritten", so need to set it to 0
> * in the default case.
> diff --git a/arch/riscv/include/asm/errata_list_vendors.h b/arch/riscv/include/asm/errata_list_vendors.h
> index ec7eba3734371a2d8b68fbd4cbd88a8e7135a413..d47a98ab93d12ce1f6ae44449f80bb703987e985 100644
> --- a/arch/riscv/include/asm/errata_list_vendors.h
> +++ b/arch/riscv/include/asm/errata_list_vendors.h
> @@ -23,7 +23,8 @@
>
> #ifdef CONFIG_ERRATA_MIPS
> #define ERRATA_MIPS_P8700_PAUSE_OPCODE 0
> -#define ERRATA_MIPS_NUMBER 1
> +#define ERRATA_MIPS_P8700_ZALRSC 1
> +#define ERRATA_MIPS_NUMBER 2
> #endif
>
> #endif /* ASM_ERRATA_LIST_VENDORS_H */
> diff --git a/arch/riscv/include/asm/futex.h b/arch/riscv/include/asm/futex.h
> index 90c86b115e008a1fb08f3da64382fb4a64d9cc2f..bc3e8c0586410d1fc3a1a7d35b7775f9c339719f 100644
> --- a/arch/riscv/include/asm/futex.h
> +++ b/arch/riscv/include/asm/futex.h
> @@ -12,6 +12,7 @@
> #include <linux/errno.h>
> #include <asm/asm.h>
> #include <asm/asm-extable.h>
> +#include <asm/errata_list.h>
>
> /* We don't even really need the extable code, but for now keep it simple */
> #ifndef CONFIG_MMU
> @@ -19,48 +20,47 @@
> #define __disable_user_access() do { } while (0)
> #endif
>
> -#define __futex_atomic_op(insn, ret, oldval, uaddr, oparg) \
> +#define __futex_atomic_op(insn, ret, oldval, uaddr, oparg, temp) \
> { \
> __enable_user_access(); \
> - __asm__ __volatile__ ( \
> - "1: " insn " \n" \
> - "2: \n" \
> - _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
> - : [r] "+r" (ret), [ov] "=&r" (oldval), \
> - [u] "+m" (*uaddr) \
> - : [op] "Jr" (oparg) \
> - : "memory"); \
> + ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp); \
> + __disable_user_access(); \
> +}
> +
> +#define __futex_atomic_swap(ret, oldval, uaddr, oparg, temp) \
> +{ \
> + __enable_user_access(); \
> + ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp); \
> __disable_user_access(); \
> }
>
> static inline int
> arch_futex_atomic_op_inuser(int op, int oparg, int *oval, u32 __user *uaddr)
> {
> - int oldval = 0, ret = 0;
> + int __maybe_unused oldval = 0, ret = 0, temp = 0;
>
> if (!access_ok(uaddr, sizeof(u32)))
> return -EFAULT;
>
> switch (op) {
> case FUTEX_OP_SET:
> - __futex_atomic_op("amoswap.w.aqrl %[ov],%z[op],%[u]",
> - ret, oldval, uaddr, oparg);
> + __futex_atomic_swap(ret, oldval, uaddr, oparg, temp);
> break;
> case FUTEX_OP_ADD:
> - __futex_atomic_op("amoadd.w.aqrl %[ov],%z[op],%[u]",
> - ret, oldval, uaddr, oparg);
> + __futex_atomic_op(add,
> + ret, oldval, uaddr, oparg, temp);
> break;
> case FUTEX_OP_OR:
> - __futex_atomic_op("amoor.w.aqrl %[ov],%z[op],%[u]",
> - ret, oldval, uaddr, oparg);
> + __futex_atomic_op(or,
> + ret, oldval, uaddr, oparg, temp);
> break;
> case FUTEX_OP_ANDN:
> - __futex_atomic_op("amoand.w.aqrl %[ov],%z[op],%[u]",
> - ret, oldval, uaddr, ~oparg);
> + __futex_atomic_op(and,
> + ret, oldval, uaddr, ~oparg, temp);
> break;
> case FUTEX_OP_XOR:
> - __futex_atomic_op("amoxor.w.aqrl %[ov],%z[op],%[u]",
> - ret, oldval, uaddr, oparg);
> + __futex_atomic_op(xor,
> + ret, oldval, uaddr, oparg, temp);
> break;
> default:
> ret = -ENOSYS;
> diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> index 60eb221296a604694c5f936d0c23c637dc44298e..8a7f9cb57f8450fd04107e2b57096c41bcefb722 100644
> --- a/arch/riscv/kernel/entry.S
> +++ b/arch/riscv/kernel/entry.S
> @@ -73,7 +73,15 @@
> beq a2, zero, .Lnew_vmalloc_restore_context
>
> /* Atomically reset the current cpu bit in new_vmalloc */
> - amoxor.d a0, a1, (a0)
> + ALTERNATIVE("amoxor.d a0, a1, (a0); \
> + .rept 3; nop; .endr;",
> + "1: lr.d a2, (a0); \
> + xor a2, a2, a1; \
> + sc.d a2, a2, (a0); \
> + bnez a2, 1b;",
> + MIPS_VENDOR_ID,
> + ERRATA_MIPS_P8700_ZALRSC,
> + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC)
>
> /* Only emit a sfence.vma if the uarch caches invalid entries */
> ALTERNATIVE("sfence.vma", "nop", 0, RISCV_ISA_EXT_SVVPTC, 1)
>
> ---
> base-commit: c369299895a591d96745d6492d4888259b004a9e
> change-id: 20250714-p8700-zalrsc-f3894be40d06
>
> Best regards,
> --
> Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
>
>
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions
2026-07-27 21:03 ` Jesse Taube
@ 2026-07-27 22:37 ` Jesse Taube
2026-08-10 15:20 ` Aleksa Paunovic
0 siblings, 1 reply; 6+ messages in thread
From: Jesse Taube @ 2026-07-27 22:37 UTC (permalink / raw)
To: aleksa.paunovic
Cc: Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti,
Will Deacon, Peter Zijlstra, Mark Rutland, Yury Norov,
Djordje Todorovic, Rasmus Villemoes, Conor Dooley, Jessica Clarke,
Samuel Holland, Paul Walmsley, Boqun Feng, Gary Guo, linux-riscv,
linux-kernel, Chao-ying Fu, Aleksandar Rikalo, Charles Mirabile,
Charles Jenkins
On Mon, Jul 27, 2026 at 5:03 PM Jesse Taube <jtaubepe@redhat.com> wrote:
>
> On Thu, Jul 23, 2026 at 11:53 AM Aleksa Paunovic via B4 Relay
> <devnull+aleksa.paunovic.htecgroup.com@kernel.org> wrote:
> >
> > From: Chao-ying Fu <cfu@mips.com>
> >
> > MIPS P8700 does not natively support Zaamo instructions.
> > They are emulated with Zalrsc extension instructions instead [1].
> > Since the emulation is implemented through M-mode traps in the SBI
> > layer, it is best to avoid using these instructions wherever possible on
> > the P8700.
> >
> > Implement kernel atomic operations using LR/SC sequences only.
> > This is achieved by using the errata mechanism, with minimal
> > interference on other cores.
> >
> > Signed-off-by: Chao-ying Fu <cfu@mips.com>
> > Signed-off-by: Aleksandar Rikalo <arikalo@gmail.com>
> > Co-developed-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> > Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
Tested-by: Jesse Taube <jtaubepe@redhat.com>
> >
> > [1] https://mips.com/wp-content/uploads/2026/03/MIPS_P8700_P8700-F_Programmers_Reference_Guide_Rev1.86_2-17-2026.pdf
> >
> > ---
> > The patch was tested on QEMU configured to emulate an eight-hart MIPS P8700 CPU.
>
> Can you share the tests. I made my own tests a while back here:
> https://github.com/Mr-Bossman/zalrsc-buildroot/tree/master
> Though they don't seem to boot without the zaamo extension
I found out that the devicetree needs to have the amo extention to work.
Somewhere there is a patch to split support into Zalrsc and Zaamo, but
here is one on my tree
https://github.com/Mr-Bossman/linux/commit/2bde8c4382a55cb82e769b0c8d9d8bf1b9c9164d
> Thanks,
> Jesse Taube
>
> > Testing done since v3: futex kselftests and perf futex tests. These tests caught the issues described below.
> > The same tests were executed on the Boston board with a single-hart P8700 core.
> >
> > Since the main issue was with an incorrectly written erratum, it shouldn't affect Vladimir's version [1].
> > However, since chips supporting only one part of the A extension are rare, we believe it might be
> > better to address this using the alternative mechanism, instead of demanding that the wider community
> > relax the A extension requirement.
> >
> > Changes in v4:
> > - The amo part of the ALT_TEST_AND_OP_BIT_ORD erratum erroneously hardcoded zero as the destination register.
> > This is fixed in v4.
> > - futex.h was missing the ANDN case.
> > - Link to v3: https://lore.kernel.org/r/20250901-p8700-zalrsc-v3-1-ec64fabbe093@htecgroup.com
> >
> > Changes in v3:
> > - Use alternatives to replace AMO instructions with LR/SC
> > - Rebase on Alexandre Ghiti's "for-next" branch.
> > - Link to v2: https://lore.kernel.org/linux-riscv/20241225082412.36727-1-arikalo@gmail.com/
> >
> > Links:
> > [1] https://lore.kernel.org/linux-riscv/20260120-lrsc-only-v2-0-a522e640d27d@mobileye.com/
> >
> > Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> > ---
> > arch/riscv/Kconfig.errata | 11 ++
> > arch/riscv/errata/mips/errata.c | 13 +-
> > arch/riscv/include/asm/atomic.h | 29 ++--
> > arch/riscv/include/asm/bitops.h | 28 ++--
> > arch/riscv/include/asm/cmpxchg.h | 9 +-
> > arch/riscv/include/asm/errata_list.h | 215 +++++++++++++++++++++++++++
> > arch/riscv/include/asm/errata_list_vendors.h | 3 +-
> > arch/riscv/include/asm/futex.h | 40 ++---
> > arch/riscv/kernel/entry.S | 10 +-
> > 9 files changed, 290 insertions(+), 68 deletions(-)
> >
> > diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
> > index 3c945d086c7d0266b685f9506d58b0662af071c4..cd5bd5e8eb395418c3ad103dbd836c775b7fd901 100644
> > --- a/arch/riscv/Kconfig.errata
> > +++ b/arch/riscv/Kconfig.errata
> > @@ -44,6 +44,17 @@ config ERRATA_MIPS_P8700_PAUSE_OPCODE
> >
> > If you are not using the P8700 processor, say n.
> >
> > +config ERRATA_MIPS_P8700_AMO_ZALRSC
> > + bool "Replace AMO instructions with LR/SC on MIPS P8700"
> > + depends on ERRATA_MIPS && 64BIT
> > + default n
> > + help
> > + The MIPS P8700 does not implement the full A extension,
> > + implementing only Zalrsc. Enabling this will replace
> > + all AMO instructions with LR/SC instructions on the P8700.
> > +
> > + If you are not using the P8700 processor, say n.
> > +
> > config ERRATA_SIFIVE
> > bool "SiFive errata"
> > depends on RISCV_ALTERNATIVE
> > diff --git a/arch/riscv/errata/mips/errata.c b/arch/riscv/errata/mips/errata.c
> > index e984a8152208c34690f89d8101571b097485c360..08c5efd58bf242d3831957f91f6338ef49f61238 100644
> > --- a/arch/riscv/errata/mips/errata.c
> > +++ b/arch/riscv/errata/mips/errata.c
> > @@ -23,13 +23,22 @@ static inline bool errata_probe_pause(void)
> > return true;
> > }
> >
> > -static u32 mips_errata_probe(void)
> > +static inline bool errata_probe_zalrsc(unsigned long archid)
> > +{
> > + return archid == 0x8000000000000201;
> > +}
> > +
> > +static u32 mips_errata_probe(unsigned long archid)
> > {
> > u32 cpu_req_errata = 0;
> >
> > if (errata_probe_pause())
> > cpu_req_errata |= BIT(ERRATA_MIPS_P8700_PAUSE_OPCODE);
> >
> > + if (errata_probe_zalrsc(archid))
> > + cpu_req_errata |= BIT(ERRATA_MIPS_P8700_ZALRSC);
> > +
> > +
> > return cpu_req_errata;
> > }
> >
> > @@ -38,7 +47,7 @@ void mips_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
> > unsigned int stage)
> > {
We should be patching this in at RISCV_ALTERNATIVES_EARLY_BOOT
This gives as few instructions as possible that aren't patched. Please
do something
similar to how thead does it.
https://elixir.bootlin.com/linux/v7.2-rc4/source/arch/riscv/errata/thead/errata.c#L222
Thanks,
Jesse Taube
> > struct alt_entry *alt;
> > - u32 cpu_req_errata = mips_errata_probe();
> > + u32 cpu_req_errata = mips_errata_probe(archid);
> > u32 tmp;
> >
> > BUILD_BUG_ON(ERRATA_MIPS_NUMBER >= RISCV_VENDOR_EXT_ALTERNATIVES_BASE);
> > diff --git a/arch/riscv/include/asm/atomic.h b/arch/riscv/include/asm/atomic.h
> > index 3f33dc54f94b2e14cb4270dbd0493625fbb977bd..9f848ec1705ff1afc6342e53965fab82309f9fe2 100644
> > --- a/arch/riscv/include/asm/atomic.h
> > +++ b/arch/riscv/include/asm/atomic.h
> > @@ -54,12 +54,9 @@ static __always_inline void arch_atomic64_set(atomic64_t *v, s64 i)
> > static __always_inline \
> > void arch_atomic##prefix##_##op(c_type i, atomic##prefix##_t *v) \
> > { \
> > - __asm__ __volatile__ ( \
> > - " amo" #asm_op "." #asm_type " zero, %1, %0" \
> > - : "+A" (v->counter) \
> > - : "r" (I) \
> > - : "memory"); \
> > -} \
> > + register __maybe_unused c_type ret, temp; \
> > + ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp); \
> > +}
> >
> > #ifdef CONFIG_GENERIC_ATOMIC64
> > #define ATOMIC_OPS(op, asm_op, I) \
> > @@ -89,24 +86,16 @@ static __always_inline \
> > c_type arch_atomic##prefix##_fetch_##op##_relaxed(c_type i, \
> > atomic##prefix##_t *v) \
> > { \
> > - register c_type ret; \
> > - __asm__ __volatile__ ( \
> > - " amo" #asm_op "." #asm_type " %1, %2, %0" \
> > - : "+A" (v->counter), "=r" (ret) \
> > - : "r" (I) \
> > - : "memory"); \
> > + register __maybe_unused c_type ret, temp; \
> > + ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp); \
> > return ret; \
> > } \
> > static __always_inline \
> > c_type arch_atomic##prefix##_fetch_##op(c_type i, atomic##prefix##_t *v) \
> > -{ \
> > - register c_type ret; \
> > - __asm__ __volatile__ ( \
> > - " amo" #asm_op "." #asm_type ".aqrl %1, %2, %0" \
> > - : "+A" (v->counter), "=r" (ret) \
> > - : "r" (I) \
> > - : "memory"); \
> > - return ret; \
> > +{ \
> > + register __maybe_unused c_type ret, temp; \
> > + ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp); \
> > + return ret; \
> > }
> >
> > #define ATOMIC_OP_RETURN(op, asm_op, c_op, I, asm_type, c_type, prefix) \
> > diff --git a/arch/riscv/include/asm/bitops.h b/arch/riscv/include/asm/bitops.h
> > index 3c1a15be54d804d7f609d8f4033710970f7b215f..caf8048c29f34182ee0740e836a1305c450b68f4 100644
> > --- a/arch/riscv/include/asm/bitops.h
> > +++ b/arch/riscv/include/asm/bitops.h
> > @@ -170,30 +170,27 @@ static __always_inline int variable_fls(unsigned int x)
> >
> > #if (BITS_PER_LONG == 64)
> > #define __AMO(op) "amo" #op ".d"
> > +#define __LR "lr.d"
> > +#define __SC "sc.d"
> > #elif (BITS_PER_LONG == 32)
> > #define __AMO(op) "amo" #op ".w"
> > +#define __LR "lr.w"
> > +#define __SC "sc.w"
> > #else
> > #error "Unexpected BITS_PER_LONG"
> > #endif
> >
> > #define __test_and_op_bit_ord(op, mod, nr, addr, ord) \
> > ({ \
> > - unsigned long __res, __mask; \
> > + __maybe_unused unsigned long __res, __mask, __temp; \
> > __mask = BIT_MASK(nr); \
> > - __asm__ __volatile__ ( \
> > - __AMO(op) #ord " %0, %2, %1" \
> > - : "=r" (__res), "+A" (addr[BIT_WORD(nr)]) \
> > - : "r" (mod(__mask)) \
> > - : "memory"); \
> > + ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp); \
> > ((__res & __mask) != 0); \
> > })
> >
> > #define __op_bit_ord(op, mod, nr, addr, ord) \
> > - __asm__ __volatile__ ( \
> > - __AMO(op) #ord " zero, %1, %0" \
> > - : "+A" (addr[BIT_WORD(nr)]) \
> > - : "r" (mod(BIT_MASK(nr))) \
> > - : "memory");
> > + __maybe_unused unsigned long __res, __temp; \
> > + ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp);
> >
> > #define __test_and_op_bit(op, mod, nr, addr) \
> > __test_and_op_bit_ord(op, mod, nr, addr, .aqrl)
> > @@ -337,12 +334,9 @@ static __always_inline void arch___clear_bit_unlock(
> > static __always_inline bool arch_xor_unlock_is_negative_byte(unsigned long mask,
> > volatile unsigned long *addr)
> > {
> > - unsigned long res;
> > - __asm__ __volatile__ (
> > - __AMO(xor) ".rl %0, %2, %1"
> > - : "=r" (res), "+A" (*addr)
> > - : "r" (__NOP(mask))
> > - : "memory");
> > + __maybe_unused unsigned long res, temp;
> > +
> > + ALT_ARCH_XOR_UNLOCK(mask, addr, res, temp);
> > return (res & BIT(7)) != 0;
> > }
> >
> > diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
> > index 8712cf9c69dcb3690058c60e60481fc4c3ffea06..21035ebe0e9193c1793ddc5488346fcb04d3ff3a 100644
> > --- a/arch/riscv/include/asm/cmpxchg.h
> > +++ b/arch/riscv/include/asm/cmpxchg.h
> > @@ -56,13 +56,8 @@
> >
> > #define __arch_xchg(sfx, prepend, append, r, p, n) \
> > ({ \
> > - __asm__ __volatile__ ( \
> > - prepend \
> > - " amoswap" sfx " %0, %2, %1\n" \
> > - append \
> > - : "=r" (r), "+A" (*(p)) \
> > - : "r" (n) \
> > - : "memory"); \
> > + __typeof__(*(__ptr)) __maybe_unused temp; \
> > + ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp); \
> > })
> >
> > #define _arch_xchg(ptr, new, sc_sfx, swap_sfx, prepend, \
> > diff --git a/arch/riscv/include/asm/errata_list.h b/arch/riscv/include/asm/errata_list.h
> > index 6694b5ccdcf85cfe7e767ea4de981b34f2b17b04..76c2f2a0f3b53f5507ecb2680d302679b3f97c69 100644
> > --- a/arch/riscv/include/asm/errata_list.h
> > +++ b/arch/riscv/include/asm/errata_list.h
> > @@ -25,6 +25,7 @@ ALTERNATIVE(__stringify(RISCV_PTR do_page_fault), \
> > __stringify(RISCV_PTR sifive_cip_453_page_fault_trp), \
> > SIFIVE_VENDOR_ID, ERRATA_SIFIVE_CIP_453, \
> > CONFIG_ERRATA_SIFIVE_CIP_453)
> > +
> > #else /* !__ASSEMBLER__ */
> >
> > #define ALT_SFENCE_VMA_ASID(asid) \
> > @@ -53,6 +54,220 @@ asm(ALTERNATIVE( \
> > : /* no inputs */ \
> > : "memory")
> >
> > +#ifdef CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC
> > +#define ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp) \
> > +asm(ALTERNATIVE( \
> > + " amo" #asm_op "." #asm_type " zero, %3, %0\n" \
> > + __nops(3), \
> > + "1: lr." #asm_type " %1, %0\n" \
> > + " " #asm_op " %2, %1, %3\n" \
> > + " sc." #asm_type " %2, %2, %0\n" \
> > + " bnez %2, 1b\n", \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
> > + : "r" (I) \
> > + : "memory")
> > +
> > +#define ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp) \
> > +asm(ALTERNATIVE( \
> > + " amo" #asm_op "." #asm_type " %1, %3, %0\n" \
> > + __nops(3), \
> > + "1: lr." #asm_type " %1, %0\n" \
> > + " " #asm_op " %2, %1, %3\n" \
> > + " sc." #asm_type " %2, %2, %0\n" \
> > + " bnez %2, 1b\n", \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
> > + : "r" (I) \
> > + : "memory")
> > +
> > +#define ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp) \
> > +asm(ALTERNATIVE( \
> > + " amo" #asm_op "." #asm_type ".aqrl %1, %3, %0\n"\
> > + __nops(3), \
> > + "1: lr." #asm_type ".aqrl %1, %0\n" \
> > + " " #asm_op " %2, %1, %3\n" \
> > + " sc." #asm_type ".aqrl %2, %2, %0\n" \
> > + " bnez %2, 1b\n", \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "+A" (v->counter), "=&r" (ret), "=&r" (temp) \
> > + : "r" (I) \
> > + : "memory")
> > +/* BITOPS.h */
> > +#define ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp) \
> > +asm(ALTERNATIVE( \
> > + __AMO(op) #ord " %0, %3, %1\n" \
> > + __nops(3), \
> > + "1: " __LR #ord " %0, %1\n" \
> > + #op " %2, %0, %3\n" \
> > + __SC #ord " %2, %2, %1\n" \
> > + "bnez %2, 1b\n", \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "=&r" (__res), "+A" (addr[BIT_WORD(nr)]), "=&r" (__temp) \
> > + : "r" (mod(__mask)) \
> > + : "memory")
> > +
> > +#define ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp) \
> > +asm(ALTERNATIVE( \
> > + __AMO(op) #ord " zero, %3, %1\n" \
> > + __nops(3), \
> > + "1: " __LR #ord " %0, %1\n" \
> > + #op " %2, %0, %3\n" \
> > + __SC #ord " %2, %2, %1\n" \
> > + "bnez %2, 1b\n", \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "=&r" (__res), "+A" (addr[BIT_WORD(nr)]), "=&r" (__temp) \
> > + : "r" (mod(BIT_MASK(nr))) \
> > + : "memory")
> > +
> > +#define ALT_ARCH_XOR_UNLOCK(mask, addr, __res, __temp) \
> > +asm(ALTERNATIVE( \
> > + __AMO(xor) ".rl %0, %3, %1\n" \
> > + __nops(3), \
> > + "1: " __LR ".rl %0, %1\n" \
> > + "xor %2, %0, %3\n" \
> > + __SC ".rl %2, %2, %1\n" \
> > + "bnez %2, 1b\n", \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "=&r" (__res), "+A" (*addr), "=&r" (__temp) \
> > + : "r" (__NOP(mask)) \
> > + : "memory")
> > +
> > +#define ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp) \
> > +asm(ALTERNATIVE( \
> > + prepend \
> > + " amoswap" sfx " %0, %3, %1\n" \
> > + __nops(2) \
> > + append, \
> > + prepend \
> > + "1: lr" sfx " %0, %1\n" \
> > + " sc" sfx " %2, %3, %1\n" \
> > + " bnez %2, 1b\n" \
> > + append, \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : "=&r" (r), "+A" (*(p)), "=&r" (temp) \
> > + : "r" (n) \
> > + : "memory")
> > +
> > +/* FUTEX.H */
> > +#define ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp) \
> > +asm(ALTERNATIVE( \
> > + "1: amo" #insn ".w.aqrl %[ov],%z[op],%[u]\n" \
> > + __nops(3) \
> > + "2:\n" \
> > + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> > + "1: lr.w.aqrl %[ov], %[u]\n" \
> > + " " #insn" %[t], %[ov], %z[op]\n" \
> > + " sc.w.aqrl %[t], %[t], %[u]\n" \
> > + " bnez %[t], 1b\n" \
> > + "2:\n" \
> > + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> > + [t] "=&r" (temp), [u] "+m" (*uaddr) \
> > + : [op] "Jr" (oparg) \
> > + : "memory")
> > +
> > +#define ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp) \
> > +asm(ALTERNATIVE( \
> > + "1: amoswap.w.aqrl %[ov],%z[op],%[u]\n" \
> > + __nops(3) \
> > + "2:\n" \
> > + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> > + "1: lr.w.aqrl %[ov], %[u]\n" \
> > + " mv %[t], %z[op]\n" \
> > + " sc.w.aqrl %[t], %[t], %[u]\n" \
> > + " bnez %[t], 1b\n" \
> > + "2:\n" \
> > + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]), \
> > + MIPS_VENDOR_ID, \
> > + ERRATA_MIPS_P8700_ZALRSC, \
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC) \
> > + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> > + [t] "=&r" (temp), [u] "+m" (*uaddr) \
> > + : [op] "Jr" (oparg) \
> > + : "memory")
> > +
> > +#else
> > +#define ALT_ATOMIC_OP(asm_op, I, asm_type, v, ret, temp) \
> > +asm("amo" #asm_op "." #asm_type " zero, %1, %0" \
> > + : "+A" (v->counter) \
> > + : "r" (I) \
> > + : "memory")
> > +
> > +#define ALT_ATOMIC_FETCH_OP_RELAXED(asm_op, I, asm_type, v, ret, temp) \
> > +asm("amo" #asm_op "." #asm_type " %1, %2, %0" \
> > + : "+A" (v->counter), "=r" (ret) \
> > + : "r" (I) \
> > + : "memory")
> > +
> > +#define ALT_ATOMIC_FETCH_OP(asm_op, I, asm_type, v, ret, temp) \
> > +asm("amo" #asm_op "." #asm_type ".aqrl %1, %2, %0" \
> > + : "+A" (v->counter), "=r" (ret) \
> > + : "r" (I) \
> > + : "memory")
> > +
> > +#define ALT_TEST_AND_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __mask, __temp) \
> > +asm(__AMO(op) #ord " %0, %2, %1" \
> > + : "=r" (__res), "+A" (addr[BIT_WORD(nr)]) \
> > + : "r" (mod(__mask)) \
> > + : "memory")
> > +
> > +#define ALT_OP_BIT_ORD(op, mod, nr, addr, ord, __res, __temp) \
> > +asm(__AMO(op) #ord " zero, %1, %0" \
> > + : "+A" (addr[BIT_WORD(nr)]) \
> > + : "r" (mod(BIT_MASK(nr))) \
> > + : "memory")
> > +
> > +#define ALT_ARCH_XOR_UNLOCK(mask, addr, __res, __temp) \
> > +asm(__AMO(xor) ".rl %0, %2, %1" \
> > + : "=r" (res), "+A" (*addr) \
> > + : "r" (__NOP(mask)) \
> > + : "memory")
> > +
> > +#define ALT_ARCH_XCHG(sfx, prepend, append, r, p, n, temp) \
> > +asm(prepend \
> > + " amoswap" sfx " %0, %2, %1\n" \
> > + append \
> > + : "=r" (r), "+A" (*(p)) \
> > + : "r" (n) \
> > + : "memory")
> > +
> > +#define ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp) \
> > +asm("1: amo" #insn ".w.aqrl %[ov],%z[op],%[u]\n" \
> > + "2:\n" \
> > + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
> > + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> > + [u] "+m" (*uaddr) \
> > + : [op] "Jr" (oparg) \
> > + : "memory")
> > +
> > +#define ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp) \
> > +asm("1: amoswap.w.aqrl %[ov],%z[op],%[u]\n" \
> > + "2:\n" \
> > + _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
> > + : [r] "+r" (ret), [ov] "=&r" (oldval), \
> > + [u] "+m" (*uaddr) \
> > + : [op] "Jr" (oparg) \
> > + : "memory")
> > +#endif
> > +
> > /*
> > * _val is marked as "will be overwritten", so need to set it to 0
> > * in the default case.
> > diff --git a/arch/riscv/include/asm/errata_list_vendors.h b/arch/riscv/include/asm/errata_list_vendors.h
> > index ec7eba3734371a2d8b68fbd4cbd88a8e7135a413..d47a98ab93d12ce1f6ae44449f80bb703987e985 100644
> > --- a/arch/riscv/include/asm/errata_list_vendors.h
> > +++ b/arch/riscv/include/asm/errata_list_vendors.h
> > @@ -23,7 +23,8 @@
> >
> > #ifdef CONFIG_ERRATA_MIPS
> > #define ERRATA_MIPS_P8700_PAUSE_OPCODE 0
> > -#define ERRATA_MIPS_NUMBER 1
> > +#define ERRATA_MIPS_P8700_ZALRSC 1
> > +#define ERRATA_MIPS_NUMBER 2
> > #endif
> >
> > #endif /* ASM_ERRATA_LIST_VENDORS_H */
> > diff --git a/arch/riscv/include/asm/futex.h b/arch/riscv/include/asm/futex.h
> > index 90c86b115e008a1fb08f3da64382fb4a64d9cc2f..bc3e8c0586410d1fc3a1a7d35b7775f9c339719f 100644
> > --- a/arch/riscv/include/asm/futex.h
> > +++ b/arch/riscv/include/asm/futex.h
> > @@ -12,6 +12,7 @@
> > #include <linux/errno.h>
> > #include <asm/asm.h>
> > #include <asm/asm-extable.h>
> > +#include <asm/errata_list.h>
> >
> > /* We don't even really need the extable code, but for now keep it simple */
> > #ifndef CONFIG_MMU
> > @@ -19,48 +20,47 @@
> > #define __disable_user_access() do { } while (0)
> > #endif
> >
> > -#define __futex_atomic_op(insn, ret, oldval, uaddr, oparg) \
> > +#define __futex_atomic_op(insn, ret, oldval, uaddr, oparg, temp) \
> > { \
> > __enable_user_access(); \
> > - __asm__ __volatile__ ( \
> > - "1: " insn " \n" \
> > - "2: \n" \
> > - _ASM_EXTABLE_UACCESS_ERR(1b, 2b, %[r]) \
> > - : [r] "+r" (ret), [ov] "=&r" (oldval), \
> > - [u] "+m" (*uaddr) \
> > - : [op] "Jr" (oparg) \
> > - : "memory"); \
> > + ALT_FUTEX_ATOMIC_OP(insn, ret, oldval, uaddr, oparg, temp); \
> > + __disable_user_access(); \
> > +}
> > +
> > +#define __futex_atomic_swap(ret, oldval, uaddr, oparg, temp) \
> > +{ \
> > + __enable_user_access(); \
> > + ALT_FUTEX_ATOMIC_SWAP(ret, oldval, uaddr, oparg, temp); \
> > __disable_user_access(); \
> > }
> >
> > static inline int
> > arch_futex_atomic_op_inuser(int op, int oparg, int *oval, u32 __user *uaddr)
> > {
> > - int oldval = 0, ret = 0;
> > + int __maybe_unused oldval = 0, ret = 0, temp = 0;
> >
> > if (!access_ok(uaddr, sizeof(u32)))
> > return -EFAULT;
> >
> > switch (op) {
> > case FUTEX_OP_SET:
> > - __futex_atomic_op("amoswap.w.aqrl %[ov],%z[op],%[u]",
> > - ret, oldval, uaddr, oparg);
> > + __futex_atomic_swap(ret, oldval, uaddr, oparg, temp);
> > break;
> > case FUTEX_OP_ADD:
> > - __futex_atomic_op("amoadd.w.aqrl %[ov],%z[op],%[u]",
> > - ret, oldval, uaddr, oparg);
> > + __futex_atomic_op(add,
> > + ret, oldval, uaddr, oparg, temp);
> > break;
> > case FUTEX_OP_OR:
> > - __futex_atomic_op("amoor.w.aqrl %[ov],%z[op],%[u]",
> > - ret, oldval, uaddr, oparg);
> > + __futex_atomic_op(or,
> > + ret, oldval, uaddr, oparg, temp);
> > break;
> > case FUTEX_OP_ANDN:
> > - __futex_atomic_op("amoand.w.aqrl %[ov],%z[op],%[u]",
> > - ret, oldval, uaddr, ~oparg);
> > + __futex_atomic_op(and,
> > + ret, oldval, uaddr, ~oparg, temp);
> > break;
> > case FUTEX_OP_XOR:
> > - __futex_atomic_op("amoxor.w.aqrl %[ov],%z[op],%[u]",
> > - ret, oldval, uaddr, oparg);
> > + __futex_atomic_op(xor,
> > + ret, oldval, uaddr, oparg, temp);
> > break;
> > default:
> > ret = -ENOSYS;
> > diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
> > index 60eb221296a604694c5f936d0c23c637dc44298e..8a7f9cb57f8450fd04107e2b57096c41bcefb722 100644
> > --- a/arch/riscv/kernel/entry.S
> > +++ b/arch/riscv/kernel/entry.S
> > @@ -73,7 +73,15 @@
> > beq a2, zero, .Lnew_vmalloc_restore_context
> >
> > /* Atomically reset the current cpu bit in new_vmalloc */
> > - amoxor.d a0, a1, (a0)
> > + ALTERNATIVE("amoxor.d a0, a1, (a0); \
> > + .rept 3; nop; .endr;",
> > + "1: lr.d a2, (a0); \
> > + xor a2, a2, a1; \
> > + sc.d a2, a2, (a0); \
> > + bnez a2, 1b;",
> > + MIPS_VENDOR_ID,
> > + ERRATA_MIPS_P8700_ZALRSC,
> > + CONFIG_ERRATA_MIPS_P8700_AMO_ZALRSC)
> >
> > /* Only emit a sfence.vma if the uarch caches invalid entries */
> > ALTERNATIVE("sfence.vma", "nop", 0, RISCV_ISA_EXT_SVVPTC, 1)
> >
> > ---
> > base-commit: c369299895a591d96745d6492d4888259b004a9e
> > change-id: 20250714-p8700-zalrsc-f3894be40d06
> >
> > Best regards,
> > --
> > Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> >
> >
> >
> > _______________________________________________
> > linux-riscv mailing list
> > linux-riscv@lists.infradead.org
> > http://lists.infradead.org/mailman/listinfo/linux-riscv
> >
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions
2026-07-27 22:37 ` Jesse Taube
@ 2026-08-10 15:20 ` Aleksa Paunovic
0 siblings, 0 replies; 6+ messages in thread
From: Aleksa Paunovic @ 2026-08-10 15:20 UTC (permalink / raw)
To: jtaubepe@redhat.com
Cc: Aleksa Paunovic, alex@ghiti.fr, aou@eecs.berkeley.edu,
arikalo@gmail.com, boqun@kernel.org, cfu@mips.com,
cmirabil@redhat.com, conor@kernel.org, Djordje Todorovic,
gary@garyguo.net, jrtc27@jrtc27.com, linux-kernel@vger.kernel.org,
linux-riscv@lists.infradead.org, linux@rasmusvillemoes.dk,
mark.rutland@arm.com, palmer@dabbelt.com,
paul.walmsley@sifive.com, peterz@infradead.org, pjw@kernel.org,
samuel.holland@sifive.com, thecharlesjenkins@gmail.com,
will@kernel.org, yury.norov@gmail.com
Hi Jesse,
On 7/28/26 00:37, Jesse Taube wrote:
> On Mon, Jul 27, 2026 at 5:03 PM Jesse Taube <jtaubepe@redhat.com> wrote:
>> On Thu, Jul 23, 2026 at 11:53 AM Aleksa Paunovic via B4 Relay
>> <devnull+aleksa.paunovic.htecgroup.com@kernel.org> wrote:
>>> From: Chao-ying Fu <cfu@mips.com>
>>>
>>> MIPS P8700 does not natively support Zaamo instructions.
>>> They are emulated with Zalrsc extension instructions instead [1].
>>> Since the emulation is implemented through M-mode traps in the SBI
>>> layer, it is best to avoid using these instructions wherever possible on
>>> the P8700.
>>>
>>> Implement kernel atomic operations using LR/SC sequences only.
>>> This is achieved by using the errata mechanism, with minimal
>>> interference on other cores.
>>>
>>> Signed-off-by: Chao-ying Fu <cfu@mips.com>
>>> Signed-off-by: Aleksandar Rikalo <arikalo@gmail.com>
>>> Co-developed-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
>>> Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> Tested-by: Jesse Taube <jtaubepe@redhat.com>
Thank you for taking the time to test the patch!
>
>>> [1] https://mips.com/wp-content/uploads/2026/03/MIPS_P8700_P8700-F_Programmers_Reference_Guide_Rev1.86_2-17-2026.pdf
>>>
>>> ---
>>> The patch was tested on QEMU configured to emulate an eight-hart MIPS P8700 CPU.
>> Can you share the tests. I made my own tests a while back here:
>> https://github.com/Mr-Bossman/zalrsc-buildroot/tree/master
>> Though they don't seem to boot without the zaamo extension
> I found out that the devicetree needs to have the amo extention to work.
> Somewhere there is a patch to split support into Zalrsc and Zaamo, but
> here is one on my tree
> https://github.com/Mr-Bossman/linux/commit/2bde8c4382a55cb82e769b0c8d9d8bf1b9c9164d
That's correct. We still pass 'a' to the riscv,isa-extensions list in our private dts.
I think [1] both addresses this issue and overlaps with this patch.
If it's not a major issue though, we are content with leaving things as they are.
The tests I ran were generic kselftests (and kernel modules). Compiled with GCC 15.1.0.
>> Thanks,
>> Jesse Taube
>>
>>> Testing done since v3: futex kselftests and perf futex tests. These tests caught the issues described below.
>>> The same tests were executed on the Boston board with a single-hart P8700 core.
>>>
>>> Since the main issue was with an incorrectly written erratum, it shouldn't affect Vladimir's version [1].
>>> However, since chips supporting only one part of the A extension are rare, we believe it might be
>>> better to address this using the alternative mechanism, instead of demanding that the wider community
>>> relax the A extension requirement.
>>>
>>> Changes in v4:
>>> - The amo part of the ALT_TEST_AND_OP_BIT_ORD erratum erroneously hardcoded zero as the destination register.
>>> This is fixed in v4.
>>> - futex.h was missing the ANDN case.
>>> - Link to v3: https://lore.kernel.org/r/20250901-p8700-zalrsc-v3-1-ec64fabbe093@htecgroup.com
>>>
>>> Changes in v3:
>>> - Use alternatives to replace AMO instructions with LR/SC
>>> - Rebase on Alexandre Ghiti's "for-next" branch.
>>> - Link to v2: https://lore.kernel.org/linux-riscv/20241225082412.36727-1-arikalo@gmail.com/
>>>
>>> Links:
>>> [1] https://lore.kernel.org/linux-riscv/20260120-lrsc-only-v2-0-a522e640d27d@mobileye.com/
>>>
>>> Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
>>> ---
>>> arch/riscv/Kconfig.errata | 11 ++
>>> arch/riscv/errata/mips/errata.c | 13 +-
>>> arch/riscv/include/asm/atomic.h | 29 ++--
>>> arch/riscv/include/asm/bitops.h | 28 ++--
>>> arch/riscv/include/asm/cmpxchg.h | 9 +-
>>> arch/riscv/include/asm/errata_list.h | 215 +++++++++++++++++++++++++++
>>> arch/riscv/include/asm/errata_list_vendors.h | 3 +-
>>> arch/riscv/include/asm/futex.h | 40 ++---
>>> arch/riscv/kernel/entry.S | 10 +-
>>> 9 files changed, 290 insertions(+), 68 deletions(-)
>>>
>>> diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
>>> index 3c945d086c7d0266b685f9506d58b0662af071c4..cd5bd5e8eb395418c3ad103dbd836c775b7fd901 100644
>>> --- a/arch/riscv/Kconfig.errata
>>> +++ b/arch/riscv/Kconfig.errata
>>> @@ -44,6 +44,17 @@ config ERRATA_MIPS_P8700_PAUSE_OPCODE
>>>
>>> If you are not using the P8700 processor, say n.
>>>
>>> +config ERRATA_MIPS_P8700_AMO_ZALRSC
>>> + bool "Replace AMO instructions with LR/SC on MIPS P8700"
>>> + depends on ERRATA_MIPS && 64BIT
>>> + default n
>>> + help
>>> + The MIPS P8700 does not implement the full A extension,
>>> + implementing only Zalrsc. Enabling this will replace
>>> + all AMO instructions with LR/SC instructions on the P8700.
>>> +
>>> + If you are not using the P8700 processor, say n.
>>> +
>>> config ERRATA_SIFIVE
>>> bool "SiFive errata"
>>> depends on RISCV_ALTERNATIVE
>>> diff --git a/arch/riscv/errata/mips/errata.c b/arch/riscv/errata/mips/errata.c
>>> index e984a8152208c34690f89d8101571b097485c360..08c5efd58bf242d3831957f91f6338ef49f61238 100644
>>> --- a/arch/riscv/errata/mips/errata.c
>>> +++ b/arch/riscv/errata/mips/errata.c
>>> @@ -23,13 +23,22 @@ static inline bool errata_probe_pause(void)
>>> return true;
>>> }
>>>
>>> -static u32 mips_errata_probe(void)
>>> +static inline bool errata_probe_zalrsc(unsigned long archid)
>>> +{
>>> + return archid == 0x8000000000000201;
>>> +}
>>> +
>>> +static u32 mips_errata_probe(unsigned long archid)
>>> {
>>> u32 cpu_req_errata = 0;
>>>
>>> if (errata_probe_pause())
>>> cpu_req_errata |= BIT(ERRATA_MIPS_P8700_PAUSE_OPCODE);
>>>
>>> + if (errata_probe_zalrsc(archid))
>>> + cpu_req_errata |= BIT(ERRATA_MIPS_P8700_ZALRSC);
>>> +
>>> +
>>> return cpu_req_errata;
>>> }
>>>
>>> @@ -38,7 +47,7 @@ void mips_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
>>> unsigned int stage)
>>> {
> We should be patching this in at RISCV_ALTERNATIVES_EARLY_BOOT
> This gives as few instructions as possible that aren't patched. Please
> do something
> similar to how thead does it.
> https://elixir.bootlin.com/linux/v7.2-rc4/source/arch/riscv/errata/thead/errata.c#L222
Thank you for the suggestion! I just sent out a patch to address this.
Best regards,
Aleksa
Links:
[1] https://lore.kernel.org/linux-riscv/20260120-lrsc-only-v2-1-a522e640d27d@mobileye.com/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions
2026-07-23 15:51 [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions Aleksa Paunovic via B4 Relay
2026-07-27 21:03 ` Jesse Taube
@ 2026-08-10 15:57 ` Conor Dooley
2026-08-10 18:53 ` Conor Dooley
1 sibling, 1 reply; 6+ messages in thread
From: Conor Dooley @ 2026-08-10 15:57 UTC (permalink / raw)
To: Aleksa Paunovic
Cc: Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti,
Will Deacon, Peter Zijlstra, Mark Rutland, Yury Norov,
Djordje Todorovic, Rasmus Villemoes, Charlie Jenkins,
Jessica Clarke, Samuel Holland, Paul Walmsley, Boqun Feng,
Gary Guo, linux-riscv, linux-kernel, Chao-ying Fu,
Aleksandar Rikalo
[-- Attachment #1: Type: text/plain, Size: 4571 bytes --]
On Thu, Jul 23, 2026 at 05:51:53PM +0200, Aleksa Paunovic wrote:
> From: Chao-ying Fu <cfu@mips.com>
>
> MIPS P8700 does not natively support Zaamo instructions.
> They are emulated with Zalrsc extension instructions instead [1].
> Since the emulation is implemented through M-mode traps in the SBI
> layer, it is best to avoid using these instructions wherever possible on
> the P8700.
>
> Implement kernel atomic operations using LR/SC sequences only.
> This is achieved by using the errata mechanism, with minimal
> interference on other cores.
This does not seem appropriate, you don't have an actual erratum.
Rather, you don't meet the minimum extension requirements that linux
currently has and want to avoid emulation.
>
> Signed-off-by: Chao-ying Fu <cfu@mips.com>
> Signed-off-by: Aleksandar Rikalo <arikalo@gmail.com>
> Co-developed-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
>
> [1] https://mips.com/wp-content/uploads/2026/03/MIPS_P8700_P8700-F_Programmers_Reference_Guide_Rev1.86_2-17-2026.pdf
This should be a Link: tag, and be above the signoffs.
>
> ---
> The patch was tested on QEMU configured to emulate an eight-hart MIPS P8700 CPU.
> Testing done since v3: futex kselftests and perf futex tests. These tests caught the issues described below.
> The same tests were executed on the Boston board with a single-hart P8700 core.
>
> Since the main issue was with an incorrectly written erratum, it shouldn't affect Vladimir's version [1].
> However, since chips supporting only one part of the A extension are rare, we believe it might be
> better to address this using the alternative mechanism, instead of demanding that the wider community
> relax the A extension requirement.
Alternatives, sure. Relaxing the A extension requirement on the other
hand basically makes no difference, if we have to insert an alternative into
anything using AMO instructions, may as well do it properly and support Zalrsc
only systems rather than abuse the errata mechanisms. Any multiplatform kernel
(so anything provided by a distro) is going to turn on any errata that are not
listed as being non-portable.
The minimum requirements have been known for a long time too, since the
port was merged, so I personally don't feel charitable here. Your
firmware already can deal with the emulation, so I don't really see why
we should be adding alternatives because people selling (or buying) IP
cannot be bothered to meet the extremely basic minimum requirements of the
software they want to support.
>
> Changes in v4:
> - The amo part of the ALT_TEST_AND_OP_BIT_ORD erratum erroneously hardcoded zero as the destination register.
> This is fixed in v4.
> - futex.h was missing the ANDN case.
> - Link to v3: https://lore.kernel.org/r/20250901-p8700-zalrsc-v3-1-ec64fabbe093@htecgroup.com
>
> Changes in v3:
> - Use alternatives to replace AMO instructions with LR/SC
> - Rebase on Alexandre Ghiti's "for-next" branch.
> - Link to v2: https://lore.kernel.org/linux-riscv/20241225082412.36727-1-arikalo@gmail.com/
>
> Links:
> [1] https://lore.kernel.org/linux-riscv/20260120-lrsc-only-v2-0-a522e640d27d@mobileye.com/
>
> Signed-off-by: Aleksa Paunovic <aleksa.paunovic@htecgroup.com>
> ---
> arch/riscv/Kconfig.errata | 11 ++
> arch/riscv/errata/mips/errata.c | 13 +-
> arch/riscv/include/asm/atomic.h | 29 ++--
> arch/riscv/include/asm/bitops.h | 28 ++--
> arch/riscv/include/asm/cmpxchg.h | 9 +-
> arch/riscv/include/asm/errata_list.h | 215 +++++++++++++++++++++++++++
> arch/riscv/include/asm/errata_list_vendors.h | 3 +-
> arch/riscv/include/asm/futex.h | 40 ++---
> arch/riscv/kernel/entry.S | 10 +-
> 9 files changed, 290 insertions(+), 68 deletions(-)
>
> diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
> index 3c945d086c7d0266b685f9506d58b0662af071c4..cd5bd5e8eb395418c3ad103dbd836c775b7fd901 100644
> --- a/arch/riscv/Kconfig.errata
> +++ b/arch/riscv/Kconfig.errata
> @@ -44,6 +44,17 @@ config ERRATA_MIPS_P8700_PAUSE_OPCODE
>
> If you are not using the P8700 processor, say n.
>
> +config ERRATA_MIPS_P8700_AMO_ZALRSC
> + bool "Replace AMO instructions with LR/SC on MIPS P8700"
> + depends on ERRATA_MIPS && 64BIT
> + default n
Why is this default n if your platform does not work without it?
Cheers,
Conor.
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions
2026-08-10 15:57 ` Conor Dooley
@ 2026-08-10 18:53 ` Conor Dooley
0 siblings, 0 replies; 6+ messages in thread
From: Conor Dooley @ 2026-08-10 18:53 UTC (permalink / raw)
To: Aleksa Paunovic
Cc: Paul Walmsley, Palmer Dabbelt, Albert Ou, Alexandre Ghiti,
Will Deacon, Peter Zijlstra, Mark Rutland, Yury Norov,
Djordje Todorovic, Rasmus Villemoes, Charlie Jenkins,
Jessica Clarke, Samuel Holland, Paul Walmsley, Boqun Feng,
Gary Guo, linux-riscv, linux-kernel, Chao-ying Fu,
Aleksandar Rikalo
[-- Attachment #1: Type: text/plain, Size: 756 bytes --]
On Mon, Aug 10, 2026 at 04:57:00PM +0100, Conor Dooley wrote:
> >
> > diff --git a/arch/riscv/Kconfig.errata b/arch/riscv/Kconfig.errata
> > index 3c945d086c7d0266b685f9506d58b0662af071c4..cd5bd5e8eb395418c3ad103dbd836c775b7fd901 100644
> > --- a/arch/riscv/Kconfig.errata
> > +++ b/arch/riscv/Kconfig.errata
> > @@ -44,6 +44,17 @@ config ERRATA_MIPS_P8700_PAUSE_OPCODE
> >
> > If you are not using the P8700 processor, say n.
> >
> > +config ERRATA_MIPS_P8700_AMO_ZALRSC
> > + bool "Replace AMO instructions with LR/SC on MIPS P8700"
> > + depends on ERRATA_MIPS && 64BIT
> > + default n
>
> Why is this default n if your platform does not work without it?
Disregard this, I said it before I noticed that there was emulation
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2026-08-10 18:53 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-07-23 15:51 [PATCH v4] riscv: Use Zalrsc extension to implement atomic functions Aleksa Paunovic via B4 Relay
2026-07-27 21:03 ` Jesse Taube
2026-07-27 22:37 ` Jesse Taube
2026-08-10 15:20 ` Aleksa Paunovic
2026-08-10 15:57 ` Conor Dooley
2026-08-10 18:53 ` Conor Dooley
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox