Re: [PATCH v3 2/2] riscv: Add Zawrs support for spinlocks

From: Palmer Dabbelt
Date: Wed May 24 2023 - 19:00:52 EST

On Wed, 24 May 2023 10:05:52 PDT (-0700), ajones@xxxxxxxxxxxxxxxx wrote:
On Sun, May 21, 2023 at 01:47:15PM +0200, Heiko Stuebner wrote:
From: Christoph Müllner <christoph.muellner@xxxxxxxx>

The current RISC-V code uses the generic ticket lock implementation,
that calls the macros smp_cond_load_relaxed() and smp_cond_load_acquire().
Currently, RISC-V uses the generic implementation of these macros.
This patch introduces a RISC-V specific implementation, of these
macros, that peels off the first loop iteration

I guess this peeling off of the first iteration is because it's expected
that the load generated by READ_ONCE() is more efficient than lr.w/d? If
we're worried about unnecessary use of lr.w/d, then shouldn't we look
for a solution that doesn't issue those instructions when we don't have
the Zawrs extension?

It's actually just a consequence of how the Linux hooks are described: they're macros that take a C expression to test in the loop, and we can't handle C expressions in LR/SC loops as that'd require compiler support and nobody's figured out how to do that correctly yet (there were some patches, but they had issues). So we need to do this awkward bit of checking without the reservation and then waiting with the reservation.

That's kind of a clunky pattern, so happy to hear if you've got a better idea. It's been a year or so since I looked at this last, and from looking at the spec I'd need to go dig into the "valid reservations in non-constrained loops" bit to see if we can get away with something. The semantics aren't clear from just one read of the spec.

Another option would be to call out the common cases of smp_cond_load_* as explicit hooks, with generic implementations that just do the C expression. There's only a few of these calls in the kernel and they fall into a few patterns, so we might just need ==0, !=0, and !=long-in-register.

and modifies the waiting
loop such, that it is possible to use the WRS.STO instruction of the Zawrs
ISA extension to stall the CPU.

The resulting implementation of smp_cond_load_*() will only work for
32-bit or 64-bit types for RV64 and 32-bit types for RV32.
This is caused by the restrictions of the LR instruction (RISC-V only
has LR.W and LR.D). Compiler assertions guard this new restriction.

This patch uses the existing RISC-V ISA extension framework
to detect the presence of Zawrs at run-time.
If available a NOP instruction will be replaced by WRS.NTO or WRS.STO.

I haven't looked at the patch, but I'd expect we NOP out the whole LR/WRS sequence? I don't remember any reason to have the load reservation without the WRS, but it's been a while so I might be forgetting something.

If we do need the LR, we should really then we replacing the WRS with an SC instead to avoid a dangling reservation as IIRC those have a performance hit on the SiFive cores (and presumably anyone else who locks cache lines).


The whole mechanism is gated by Kconfig setting, which defaults to Y.

The Zawrs specification can be found here:
https://github.com/riscv/riscv-zawrs/blob/main/zawrs.adoc

IIUC it's ratified? HEAD is kind of an odd commit saying it's ratified and changing some text: https://github.com/riscv/riscv-zawrs/commit/861483d99eedc60cbe9fd2ceb18dbf28d0905c9c

Signed-off-by: Christoph Müllner <christoph.muellner@xxxxxxxx>
[rebase, update to review comments]
Signed-off-by: Heiko Stuebner <heiko.stuebner@xxxxxxxx>
---
arch/riscv/Kconfig | 10 +++++
arch/riscv/include/asm/barrier.h | 64 ++++++++++++++++++++++++++++
arch/riscv/include/asm/errata_list.h | 14 ++++++
arch/riscv/include/asm/hwcap.h | 1 +
arch/riscv/kernel/cpu.c | 1 +
arch/riscv/kernel/cpufeature.c | 1 +
6 files changed, 91 insertions(+)

diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index 348c0fa1fc8c..60ff303ff58a 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -460,6 +460,16 @@ config RISCV_ISA_SVPBMT

If you don't know what to do here, say Y.

+config RISCV_ISA_ZAWRS
+ bool "Zawrs extension support"

This should be "Zawrs extension support for ..."

+ depends on RISCV_ALTERNATIVE
+ default y
+ help
+ Adds support to dynamically detect the presence of the Zawrs extension
+ (wait for reservation set) and enable its usage.
+
+ If you don't know what to do here, say Y.
+
config TOOLCHAIN_HAS_ZBB
bool
default y
diff --git a/arch/riscv/include/asm/barrier.h b/arch/riscv/include/asm/barrier.h
index 110752594228..bd64cfe5ae10 100644
--- a/arch/riscv/include/asm/barrier.h
+++ b/arch/riscv/include/asm/barrier.h
@@ -12,6 +12,8 @@

#ifndef __ASSEMBLY__

+#include <asm/errata_list.h>
+
#define nop() __asm__ __volatile__ ("nop")
#define __nops(n) ".rept " #n "\nnop\n.endr\n"
#define nops(n) __asm__ __volatile__ (__nops(n))
@@ -44,6 +46,36 @@ do { \
___p1; \
})

+#define ___smp_load_reservedN(pfx, ptr) \
+({ \
+ typeof(*ptr) ___p1; \
+ __asm__ __volatile__ ("lr." pfx " %[p], %[c]\n" \
+ : [p]"=&r" (___p1), [c]"+A"(*ptr)); \
+ ___p1; \
+})
+
+#define __smp_load_reserved_relaxed(ptr) \
+({ \
+ typeof(*ptr) ___p1; \
+ if (sizeof(*ptr) == sizeof(int)) \
+ ___p1 = ___smp_load_reservedN("w", ptr); \
+ else if (sizeof(*ptr) == sizeof(long)) \
+ ___p1 = ___smp_load_reservedN("d", ptr); \
+ else \
+ compiletime_assert(0, \
+ "Need type compatible with LR/SC instructions " \
+ "for " __stringify(ptr)); \
+ ___p1; \

It's more common to use a switch for these things, embedding the
lr.w/d asm in each case, something like the macros in
arch/riscv/include/asm/cmpxchg.h. Can we stick with that pattern?

+})
+
+#define __smp_load_reserved_acquire(ptr) \
+({ \
+ typeof(*ptr) ___p1; \
+ ___p1 = __smp_load_reserved_relaxed(ptr); \
+ RISCV_FENCE(r, rw); \
+ ___p1; \
+})
+
/*
* This is a very specific barrier: it's currently only used in two places in
* the kernel, both in the scheduler. See include/linux/spinlock.h for the two
@@ -71,6 +103,38 @@ do { \
*/
#define smp_mb__after_spinlock() RISCV_FENCE(iorw,iorw)

+#define smp_cond_load_relaxed(ptr, cond_expr) \
+({ \
+ typeof(ptr) __PTR = (ptr); \
+ __unqual_scalar_typeof(*ptr) VAL; \
+ VAL = READ_ONCE(*__PTR); \
+ if (!cond_expr) { \
+ for (;;) { \
+ VAL = __smp_load_reserved_relaxed(__PTR); \
+ if (cond_expr) \
+ break; \
+ ALT_WRS_STO(); \
+ } \
+ } \
+ (typeof(*ptr))VAL; \
+})
+
+#define smp_cond_load_acquire(ptr, cond_expr) \
+({ \
+ typeof(ptr) __PTR = (ptr); \
+ __unqual_scalar_typeof(*ptr) VAL; \
+ VAL = smp_load_acquire(__PTR); \
+ if (!cond_expr) { \
+ for (;;) { \
+ VAL = __smp_load_reserved_acquire(__PTR); \
+ if (cond_expr) \
+ break; \
+ ALT_WRS_STO(); \
+ } \
+ } \
+ (typeof(*ptr))VAL; \
+})
+
#include <asm-generic/barrier.h>

#endif /* __ASSEMBLY__ */
diff --git a/arch/riscv/include/asm/errata_list.h b/arch/riscv/include/asm/errata_list.h
index fb1a810f3d8c..36a72a07d62c 100644
--- a/arch/riscv/include/asm/errata_list.h
+++ b/arch/riscv/include/asm/errata_list.h
@@ -44,6 +44,20 @@ asm(ALTERNATIVE("sfence.vma %0", "sfence.vma", SIFIVE_VENDOR_ID, \
ERRATA_SIFIVE_CIP_1200, CONFIG_ERRATA_SIFIVE_CIP_1200) \
: : "r" (addr) : "memory")

+#define ZAWRS_WRS_NTO ".long 0x00d00073"
+#define ALT_WRS_NTO() \
+asm volatile(ALTERNATIVE( \
+ "nop\n\t", \
+ ZAWRS_WRS_NTO "\n\t", \
+ 0, RISCV_ISA_EXT_ZAWRS, CONFIG_RISCV_ISA_ZAWRS))

We don't use this. Do we need to define it now?

+
+#define ZAWRS_WRS_STO ".long 0x01d00073"

I'd prefer we use insn-def.h to define instructions, rather than scatter
.long's around, but since this instruction doesn't have any inputs, then
I guess it's not so important to use insn-def.h.

+#define ALT_WRS_STO() \
+asm volatile(ALTERNATIVE( \
+ "nop\n\t", \
+ ZAWRS_WRS_STO "\n\t", \
+ 0, RISCV_ISA_EXT_ZAWRS, CONFIG_RISCV_ISA_ZAWRS))
+

These alternatives aren't supporting errata, so putting them in
errata_list.h doesn't seem appropriate. I think cpufeature.h
would be better.

/*
* _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/hwcap.h b/arch/riscv/include/asm/hwcap.h
index e0c40a4c63d5..4233d87539ab 100644
--- a/arch/riscv/include/asm/hwcap.h
+++ b/arch/riscv/include/asm/hwcap.h
@@ -46,6 +46,7 @@
#define RISCV_ISA_EXT_ZICBOZ 34
#define RISCV_ISA_EXT_SMAIA 35
#define RISCV_ISA_EXT_SSAIA 36
+#define RISCV_ISA_EXT_ZAWRS 37

#define RISCV_ISA_EXT_MAX 64
#define RISCV_ISA_EXT_NAME_LEN_MAX 32
diff --git a/arch/riscv/kernel/cpu.c b/arch/riscv/kernel/cpu.c
index c96aa56cf1c7..ac9c604d8936 100644
--- a/arch/riscv/kernel/cpu.c
+++ b/arch/riscv/kernel/cpu.c
@@ -184,6 +184,7 @@ static struct riscv_isa_ext_data isa_ext_arr[] = {
__RISCV_ISA_EXT_DATA(zicbom, RISCV_ISA_EXT_ZICBOM),
__RISCV_ISA_EXT_DATA(zicboz, RISCV_ISA_EXT_ZICBOZ),
__RISCV_ISA_EXT_DATA(zihintpause, RISCV_ISA_EXT_ZIHINTPAUSE),
+ __RISCV_ISA_EXT_DATA(zawrs, RISCV_ISA_EXT_ZAWRS),
__RISCV_ISA_EXT_DATA(zbb, RISCV_ISA_EXT_ZBB),
__RISCV_ISA_EXT_DATA(smaia, RISCV_ISA_EXT_SMAIA),
__RISCV_ISA_EXT_DATA(ssaia, RISCV_ISA_EXT_SSAIA),
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index b1d6b7e4b829..d4a22a7aed99 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -236,6 +236,7 @@ void __init riscv_fill_hwcap(void)
SET_ISA_EXT_MAP("svinval", RISCV_ISA_EXT_SVINVAL);
SET_ISA_EXT_MAP("svnapot", RISCV_ISA_EXT_SVNAPOT);
SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT);
+ SET_ISA_EXT_MAP("zawrs", RISCV_ISA_EXT_ZAWRS);
SET_ISA_EXT_MAP("zbb", RISCV_ISA_EXT_ZBB);
SET_ISA_EXT_MAP("zicbom", RISCV_ISA_EXT_ZICBOM);
SET_ISA_EXT_MAP("zicboz", RISCV_ISA_EXT_ZICBOZ);
--
2.39.0


Thanks,
drew