Each architecture generally implements fine-tuned checksum functions to
leverage the instruction set. This patch adds the main checksum
functions that are used in networking. Tested on QEMU, this series
allows the CHECKSUM_KUNIT tests to complete an average of 50.9% faster.
This patch takes heavy use of the Zbb extension using alternatives
patching.
To test this patch, enable the configs for KUNIT, then CHECKSUM_KUNIT.
I have attempted to make these functions as optimal as possible, but I
have not ran anything on actual riscv hardware. My performance testing
has been limited to inspecting the assembly, running the algorithms on
x86 hardware, and running in QEMU.
ip_fast_csum is a relatively small function so even though it is
possible to read 64 bits at a time on compatible hardware, the
bottleneck becomes the clean up and setup code so loading 32 bits at a
time is actually faster.
Relies on https://lore.kernel.org/lkml/[email protected]/
---
The algorithm proposed to replace the default csum_fold can be seen to
compute the same result by running all 2^32 possible inputs.
static inline unsigned int ror32(unsigned int word, unsigned int shift)
{
return (word >> (shift & 31)) | (word << ((-shift) & 31));
}
unsigned short csum_fold(unsigned int csum)
{
unsigned int sum = csum;
sum = (sum & 0xffff) + (sum >> 16);
sum = (sum & 0xffff) + (sum >> 16);
return ~sum;
}
unsigned short csum_fold_arc(unsigned int csum)
{
return ((~csum - ror32(csum, 16)) >> 16);
}
int main()
{
unsigned int start = 0x0;
do {
if (csum_fold(start) != csum_fold_arc(start)) {
printf("Not the same %u\n", start);
return -1;
}
start += 1;
} while(start != 0x0);
printf("The same\n");
return 0;
}
Cc: Paul Walmsley <[email protected]>
Cc: Albert Ou <[email protected]>
Cc: Arnd Bergmann <[email protected]>
To: Charlie Jenkins <[email protected]>
To: Palmer Dabbelt <[email protected]>
To: Conor Dooley <[email protected]>
To: Samuel Holland <[email protected]>
To: David Laight <[email protected]>
To: Xiao Wang <[email protected]>
To: Evan Green <[email protected]>
To: [email protected]
To: [email protected]
To: [email protected]
Signed-off-by: Charlie Jenkins <[email protected]>
---
Changes in v13:
- Move cast from patch 4 to patch 3
- Link to v12: https://lore.kernel.org/r/[email protected]
Changes in v12:
- Rebase onto 6.7-rc5
- Add performance stats in the cover letter
- Link to v11: https://lore.kernel.org/r/[email protected]
Changes in v11:
- Extensive modifications to comply to sparse
- Organize include statements (Xiao)
- Add csum_ipv6_magic to commit message (Xiao)
- Remove extraneous len statement (Xiao)
- Add kasan_check_read call (Xiao)
- Improve comment field checksum.h (Xiao)
- Consolidate "buff" and "len" into one parameter "end" (Xiao)
- Link to v10: https://lore.kernel.org/r/[email protected]
Changes in v10:
- Move tests that were riscv-specific to be arch agnostic (Arnd)
- Link to v9: https://lore.kernel.org/r/[email protected]
Changes in v9:
- Use ror64 (Xiao)
- Move do_csum and csum_ipv6_magic headers to patch 4 (Xiao)
- Remove word "IP" from checksum headers (Xiao)
- Swap to using ifndef CONFIG_32BIT instead of ifdef CONFIG_64BIT (Xiao)
- Run no alignment code when buff is aligned (Xiao)
- Consolidate two do_csum implementations overlap into do_csum_common
- Link to v8: https://lore.kernel.org/r/[email protected]
Changes in v8:
- Speedups of 12% without Zbb and 21% with Zbb when cpu supports fast
misaligned accesses for do_csum
- Various formatting updates
- Patch now relies on https://lore.kernel.org/lkml/[email protected]/
- Link to v7: https://lore.kernel.org/r/[email protected]
Changes in v7:
- Included linux/bitops.h in asm-generic/checksum.h to use ror (Conor)
- Optimized loop in do_csum (David)
- Used ror instead of shifting (David)
- Unfortunately had to reintroduce ifdefs because gcc is not smart
enough to not throw warnings on code that will never execute
- Use ifdef instead of IS_ENABLED on __LITTLE_ENDIAN because IS_ENABLED
does not work on that
- Only optimize for zbb when alternatives is enabled in do_csum
- Link to v6: https://lore.kernel.org/r/[email protected]
Changes in v6:
- Fix accuracy of commit message for csum_fold
- Fix indentation
- Link to v5: https://lore.kernel.org/r/[email protected]
Changes in v5:
- Drop vector patches
- Check ZBB enabled before doing any ZBB code (Conor)
- Check endianness in IS_ENABLED
- Revert to the simpler non-tree based version of ipv6_csum_magic since
David pointed out that the tree based version is not better.
- Link to v4: https://lore.kernel.org/r/[email protected]
Changes in v4:
- Suggestion by David Laight to use an improved checksum used in
arch/arc.
- Eliminates zero-extension on rv32, but not on rv64.
- Reduces data dependency which should improve execution speed on
rv32 and rv64
- Still passes CHECKSUM_KUNIT and RISCV_CHECKSUM_KUNIT on rv32 and
rv64 with and without zbb.
- Link to v3: https://lore.kernel.org/r/[email protected]
Changes in v3:
- Use riscv_has_extension_likely and has_vector where possible (Conor)
- Reduce ifdefs by using IS_ENABLED where possible (Conor)
- Use kernel_vector_begin in the vector code (Samuel)
- Link to v2: https://lore.kernel.org/r/[email protected]
Changes in v2:
- After more benchmarking, rework functions to improve performance.
- Remove tests that overlapped with the already existing checksum
tests and make tests more extensive.
- Use alternatives to activate code with Zbb and vector extensions
- Link to v1: https://lore.kernel.org/r/[email protected]
---
Charlie Jenkins (5):
asm-generic: Improve csum_fold
riscv: Add static key for misaligned accesses
riscv: Add checksum header
riscv: Add checksum library
kunit: Add tests for csum_ipv6_magic and ip_fast_csum
arch/riscv/include/asm/checksum.h | 93 ++++++++++
arch/riscv/include/asm/cpufeature.h | 2 +
arch/riscv/kernel/cpufeature.c | 30 ++++
arch/riscv/lib/Makefile | 1 +
arch/riscv/lib/csum.c | 326 ++++++++++++++++++++++++++++++++++++
include/asm-generic/checksum.h | 6 +-
lib/checksum_kunit.c | 284 ++++++++++++++++++++++++++++++-
7 files changed, 738 insertions(+), 4 deletions(-)
---
base-commit: a39b6ac3781d46ba18193c9dbb2110f31e9bffe9
change-id: 20230804-optimize_checksum-db145288ac21
--
- Charlie
This csum_fold implementation introduced into arch/arc by Vineet Gupta
is better than the default implementation on at least arc, x86, and
riscv. Using GCC trunk and compiling non-inlined version, this
implementation has 41.6667%, 25% fewer instructions on riscv64, x86-64
respectively with -O3 optimization. Most implmentations override this
default in asm, but this should be more performant than all of those
other implementations except for arm which has barrel shifting and
sparc32 which has a carry flag.
Signed-off-by: Charlie Jenkins <[email protected]>
Reviewed-by: David Laight <[email protected]>
---
include/asm-generic/checksum.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/include/asm-generic/checksum.h b/include/asm-generic/checksum.h
index 43e18db89c14..ad928cce268b 100644
--- a/include/asm-generic/checksum.h
+++ b/include/asm-generic/checksum.h
@@ -2,6 +2,8 @@
#ifndef __ASM_GENERIC_CHECKSUM_H
#define __ASM_GENERIC_CHECKSUM_H
+#include <linux/bitops.h>
+
/*
* computes the checksum of a memory block at buff, length len,
* and adds in "sum" (32-bit)
@@ -31,9 +33,7 @@ extern __sum16 ip_fast_csum(const void *iph, unsigned int ihl);
static inline __sum16 csum_fold(__wsum csum)
{
u32 sum = (__force u32)csum;
- sum = (sum & 0xffff) + (sum >> 16);
- sum = (sum & 0xffff) + (sum >> 16);
- return (__force __sum16)~sum;
+ return (__force __sum16)((~sum - ror32(sum, 16)) >> 16);
}
#endif
--
2.43.0
Support static branches depending on the value of misaligned accesses.
This will be used by a later patch in the series. All cpus must be
considered "fast" for this static branch to be flipped.
Signed-off-by: Charlie Jenkins <[email protected]>
---
arch/riscv/include/asm/cpufeature.h | 2 ++
arch/riscv/kernel/cpufeature.c | 30 ++++++++++++++++++++++++++++++
2 files changed, 32 insertions(+)
diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h
index a418c3112cd6..7b129e5e2f07 100644
--- a/arch/riscv/include/asm/cpufeature.h
+++ b/arch/riscv/include/asm/cpufeature.h
@@ -133,4 +133,6 @@ static __always_inline bool riscv_cpu_has_extension_unlikely(int cpu, const unsi
return __riscv_isa_extension_available(hart_isa[cpu].isa, ext);
}
+DECLARE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
+
#endif
diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
index b3785ffc1570..095eb6ebdcaa 100644
--- a/arch/riscv/kernel/cpufeature.c
+++ b/arch/riscv/kernel/cpufeature.c
@@ -10,6 +10,7 @@
#include <linux/bitmap.h>
#include <linux/cpuhotplug.h>
#include <linux/ctype.h>
+#include <linux/jump_label.h>
#include <linux/log2.h>
#include <linux/memory.h>
#include <linux/module.h>
@@ -728,6 +729,35 @@ void riscv_user_isa_enable(void)
csr_set(CSR_SENVCFG, ENVCFG_CBZE);
}
+DEFINE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
+
+static int set_unaligned_access_static_branches(void)
+{
+ /*
+ * This will be called after check_unaligned_access_all_cpus so the
+ * result of unaligned access speed for all cpus will be available.
+ */
+
+ int cpu;
+ bool fast_misaligned_access_speed = true;
+
+ for_each_online_cpu(cpu) {
+ int this_perf = per_cpu(misaligned_access_speed, cpu);
+
+ if (this_perf != RISCV_HWPROBE_MISALIGNED_FAST) {
+ fast_misaligned_access_speed = false;
+ break;
+ }
+ }
+
+ if (fast_misaligned_access_speed)
+ static_branch_enable(&fast_misaligned_access_speed_key);
+
+ return 0;
+}
+
+arch_initcall_sync(set_unaligned_access_static_branches);
+
#ifdef CONFIG_RISCV_ALTERNATIVE
/*
* Alternative patch sites consider 48 bits when determining when to patch
--
2.43.0
Provide a 32 and 64 bit version of do_csum. When compiled for 32-bit
will load from the buffer in groups of 32 bits, and when compiled for
64-bit will load in groups of 64 bits.
Additionally provide riscv optimized implementation of csum_ipv6_magic.
Signed-off-by: Charlie Jenkins <[email protected]>
Acked-by: Conor Dooley <[email protected]>
Reviewed-by: Xiao Wang <[email protected]>
---
arch/riscv/include/asm/checksum.h | 11 ++
arch/riscv/lib/Makefile | 1 +
arch/riscv/lib/csum.c | 326 ++++++++++++++++++++++++++++++++++++++
3 files changed, 338 insertions(+)
diff --git a/arch/riscv/include/asm/checksum.h b/arch/riscv/include/asm/checksum.h
index 5a810126aac7..a5b60b54b101 100644
--- a/arch/riscv/include/asm/checksum.h
+++ b/arch/riscv/include/asm/checksum.h
@@ -12,6 +12,17 @@
#define ip_fast_csum ip_fast_csum
+extern unsigned int do_csum(const unsigned char *buff, int len);
+#define do_csum do_csum
+
+/* Default version is sufficient for 32 bit */
+#ifndef CONFIG_32BIT
+#define _HAVE_ARCH_IPV6_CSUM
+__sum16 csum_ipv6_magic(const struct in6_addr *saddr,
+ const struct in6_addr *daddr,
+ __u32 len, __u8 proto, __wsum sum);
+#endif
+
/* Define riscv versions of functions before importing asm-generic/checksum.h */
#include <asm-generic/checksum.h>
diff --git a/arch/riscv/lib/Makefile b/arch/riscv/lib/Makefile
index 26cb2502ecf8..2aa1a4ad361f 100644
--- a/arch/riscv/lib/Makefile
+++ b/arch/riscv/lib/Makefile
@@ -6,6 +6,7 @@ lib-y += memmove.o
lib-y += strcmp.o
lib-y += strlen.o
lib-y += strncmp.o
+lib-y += csum.o
lib-$(CONFIG_MMU) += uaccess.o
lib-$(CONFIG_64BIT) += tishift.o
lib-$(CONFIG_RISCV_ISA_ZICBOZ) += clear_page.o
diff --git a/arch/riscv/lib/csum.c b/arch/riscv/lib/csum.c
new file mode 100644
index 000000000000..06ce8e7250d9
--- /dev/null
+++ b/arch/riscv/lib/csum.c
@@ -0,0 +1,326 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Checksum library
+ *
+ * Influenced by arch/arm64/lib/csum.c
+ * Copyright (C) 2023 Rivos Inc.
+ */
+#include <linux/bitops.h>
+#include <linux/compiler.h>
+#include <linux/jump_label.h>
+#include <linux/kasan-checks.h>
+#include <linux/kernel.h>
+
+#include <asm/cpufeature.h>
+
+#include <net/checksum.h>
+
+/* Default version is sufficient for 32 bit */
+#ifndef CONFIG_32BIT
+__sum16 csum_ipv6_magic(const struct in6_addr *saddr,
+ const struct in6_addr *daddr,
+ __u32 len, __u8 proto, __wsum csum)
+{
+ unsigned int ulen, uproto;
+ unsigned long sum = (__force unsigned long)csum;
+
+ sum += (__force unsigned long)saddr->s6_addr32[0];
+ sum += (__force unsigned long)saddr->s6_addr32[1];
+ sum += (__force unsigned long)saddr->s6_addr32[2];
+ sum += (__force unsigned long)saddr->s6_addr32[3];
+
+ sum += (__force unsigned long)daddr->s6_addr32[0];
+ sum += (__force unsigned long)daddr->s6_addr32[1];
+ sum += (__force unsigned long)daddr->s6_addr32[2];
+ sum += (__force unsigned long)daddr->s6_addr32[3];
+
+ ulen = (__force unsigned int)htonl((unsigned int)len);
+ sum += ulen;
+
+ uproto = (__force unsigned int)htonl(proto);
+ sum += uproto;
+
+ /*
+ * Zbb support saves 4 instructions, so not worth checking without
+ * alternatives if supported
+ */
+ if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
+ IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
+ unsigned long fold_temp;
+
+ /*
+ * Zbb is likely available when the kernel is compiled with Zbb
+ * support, so nop when Zbb is available and jump when Zbb is
+ * not available.
+ */
+ asm_volatile_goto(ALTERNATIVE("j %l[no_zbb]", "nop", 0,
+ RISCV_ISA_EXT_ZBB, 1)
+ :
+ :
+ :
+ : no_zbb);
+ asm(".option push \n\
+ .option arch,+zbb \n\
+ rori %[fold_temp], %[sum], 32 \n\
+ add %[sum], %[fold_temp], %[sum] \n\
+ srli %[sum], %[sum], 32 \n\
+ not %[fold_temp], %[sum] \n\
+ roriw %[sum], %[sum], 16 \n\
+ subw %[sum], %[fold_temp], %[sum] \n\
+ .option pop"
+ : [sum] "+r" (sum), [fold_temp] "=&r" (fold_temp));
+ return (__force __sum16)(sum >> 16);
+ }
+no_zbb:
+ sum += ror64(sum, 32);
+ sum >>= 32;
+ return csum_fold((__force __wsum)sum);
+}
+EXPORT_SYMBOL(csum_ipv6_magic);
+#endif /* !CONFIG_32BIT */
+
+#ifdef CONFIG_32BIT
+#define OFFSET_MASK 3
+#elif CONFIG_64BIT
+#define OFFSET_MASK 7
+#endif
+
+static inline __no_sanitize_address unsigned long
+do_csum_common(const unsigned long *ptr, const unsigned long *end,
+ unsigned long data)
+{
+ unsigned int shift;
+ unsigned long csum = 0, carry = 0;
+
+ /*
+ * Do 32-bit reads on RV32 and 64-bit reads otherwise. This should be
+ * faster than doing 32-bit reads on architectures that support larger
+ * reads.
+ */
+ while (ptr < end) {
+ csum += data;
+ carry += csum < data;
+ data = *(ptr++);
+ }
+
+ /*
+ * Perform alignment (and over-read) bytes on the tail if any bytes
+ * leftover.
+ */
+ shift = ((long)ptr - (long)end) * 8;
+#ifdef __LITTLE_ENDIAN
+ data = (data << shift) >> shift;
+#else
+ data = (data >> shift) << shift;
+#endif
+ csum += data;
+ carry += csum < data;
+ csum += carry;
+ csum += csum < carry;
+
+ return csum;
+}
+
+/*
+ * Algorithm accounts for buff being misaligned.
+ * If buff is not aligned, will over-read bytes but not use the bytes that it
+ * shouldn't. The same thing will occur on the tail-end of the read.
+ */
+static inline __no_sanitize_address unsigned int
+do_csum_with_alignment(const unsigned char *buff, int len)
+{
+ unsigned int offset, shift;
+ unsigned long csum, data;
+ const unsigned long *ptr, *end;
+
+ /*
+ * Align address to closest word (double word on rv64) that comes before
+ * buff. This should always be in the same page and cache line.
+ * Directly call KASAN with the alignment we will be using.
+ */
+ offset = (unsigned long)buff & OFFSET_MASK;
+ kasan_check_read(buff, len);
+ ptr = (const unsigned long *)(buff - offset);
+
+ /*
+ * Clear the most significant bytes that were over-read if buff was not
+ * aligned.
+ */
+ shift = offset * 8;
+ data = *(ptr++);
+#ifdef __LITTLE_ENDIAN
+ data = (data >> shift) << shift;
+#else
+ data = (data << shift) >> shift;
+#endif
+ end = (const unsigned long *)(buff + len);
+ csum = do_csum_common(ptr, end, data);
+
+ /*
+ * Zbb support saves 6 instructions, so not worth checking without
+ * alternatives if supported
+ */
+ if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
+ IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
+ unsigned long fold_temp;
+
+ /*
+ * Zbb is likely available when the kernel is compiled with Zbb
+ * support, so nop when Zbb is available and jump when Zbb is
+ * not available.
+ */
+ asm_volatile_goto(ALTERNATIVE("j %l[no_zbb]", "nop", 0,
+ RISCV_ISA_EXT_ZBB, 1)
+ :
+ :
+ :
+ : no_zbb);
+
+#ifdef CONFIG_32BIT
+ asm_volatile_goto(".option push \n\
+ .option arch,+zbb \n\
+ rori %[fold_temp], %[csum], 16 \n\
+ andi %[offset], %[offset], 1 \n\
+ add %[csum], %[fold_temp], %[csum] \n\
+ beq %[offset], zero, %l[end] \n\
+ rev8 %[csum], %[csum] \n\
+ .option pop"
+ : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
+ : [offset] "r" (offset)
+ :
+ : end);
+
+ return (unsigned short)csum;
+#else /* !CONFIG_32BIT */
+ asm_volatile_goto(".option push \n\
+ .option arch,+zbb \n\
+ rori %[fold_temp], %[csum], 32 \n\
+ add %[csum], %[fold_temp], %[csum] \n\
+ srli %[csum], %[csum], 32 \n\
+ roriw %[fold_temp], %[csum], 16 \n\
+ addw %[csum], %[fold_temp], %[csum] \n\
+ andi %[offset], %[offset], 1 \n\
+ beq %[offset], zero, %l[end] \n\
+ rev8 %[csum], %[csum] \n\
+ .option pop"
+ : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
+ : [offset] "r" (offset)
+ :
+ : end);
+
+ return (csum << 16) >> 48;
+#endif /* !CONFIG_32BIT */
+end:
+ return csum >> 16;
+ }
+no_zbb:
+#ifndef CONFIG_32BIT
+ csum += ror64(csum, 32);
+ csum >>= 32;
+#endif
+ csum = (u32)csum + ror32((u32)csum, 16);
+ if (offset & 1)
+ return (u16)swab32(csum);
+ return csum >> 16;
+}
+
+/*
+ * Does not perform alignment, should only be used if machine has fast
+ * misaligned accesses, or when buff is known to be aligned.
+ */
+static inline __no_sanitize_address unsigned int
+do_csum_no_alignment(const unsigned char *buff, int len)
+{
+ unsigned long csum, data;
+ const unsigned long *ptr, *end;
+
+ ptr = (const unsigned long *)(buff);
+ data = *(ptr++);
+
+ kasan_check_read(buff, len);
+
+ end = (const unsigned long *)(buff + len);
+ csum = do_csum_common(ptr, end, data);
+
+ /*
+ * Zbb support saves 6 instructions, so not worth checking without
+ * alternatives if supported
+ */
+ if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
+ IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
+ unsigned long fold_temp;
+
+ /*
+ * Zbb is likely available when the kernel is compiled with Zbb
+ * support, so nop when Zbb is available and jump when Zbb is
+ * not available.
+ */
+ asm_volatile_goto(ALTERNATIVE("j %l[no_zbb]", "nop", 0,
+ RISCV_ISA_EXT_ZBB, 1)
+ :
+ :
+ :
+ : no_zbb);
+
+#ifdef CONFIG_32BIT
+ asm (".option push \n\
+ .option arch,+zbb \n\
+ rori %[fold_temp], %[csum], 16 \n\
+ add %[csum], %[fold_temp], %[csum] \n\
+ .option pop"
+ : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
+ :
+ : );
+
+#else /* !CONFIG_32BIT */
+ asm (".option push \n\
+ .option arch,+zbb \n\
+ rori %[fold_temp], %[csum], 32 \n\
+ add %[csum], %[fold_temp], %[csum] \n\
+ srli %[csum], %[csum], 32 \n\
+ roriw %[fold_temp], %[csum], 16 \n\
+ addw %[csum], %[fold_temp], %[csum] \n\
+ .option pop"
+ : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp)
+ :
+ : );
+#endif /* !CONFIG_32BIT */
+ return csum >> 16;
+ }
+no_zbb:
+#ifndef CONFIG_32BIT
+ csum += ror64(csum, 32);
+ csum >>= 32;
+#endif
+ csum = (u32)csum + ror32((u32)csum, 16);
+ return csum >> 16;
+}
+
+/*
+ * Perform a checksum on an arbitrary memory address.
+ * Will do a light-weight address alignment if buff is misaligned, unless
+ * cpu supports fast misaligned accesses.
+ */
+unsigned int do_csum(const unsigned char *buff, int len)
+{
+ if (unlikely(len <= 0))
+ return 0;
+
+ /*
+ * Significant performance gains can be seen by not doing alignment
+ * on machines with fast misaligned accesses.
+ *
+ * There is some duplicate code between the "with_alignment" and
+ * "no_alignment" implmentations, but the overlap is too awkward to be
+ * able to fit in one function without introducing multiple static
+ * branches. The largest chunk of overlap was delegated into the
+ * do_csum_common function.
+ */
+ if (static_branch_likely(&fast_misaligned_access_speed_key))
+ return do_csum_no_alignment(buff, len);
+
+ if (((unsigned long)buff & OFFSET_MASK) == 0)
+ return do_csum_no_alignment(buff, len);
+
+ return do_csum_with_alignment(buff, len);
+}
--
2.43.0
Provide checksum algorithms that have been designed to leverage riscv
instructions such as rotate. In 64-bit, can take advantage of the larger
register to avoid some overflow checking.
Signed-off-by: Charlie Jenkins <[email protected]>
Acked-by: Conor Dooley <[email protected]>
Reviewed-by: Xiao Wang <[email protected]>
---
arch/riscv/include/asm/checksum.h | 82 +++++++++++++++++++++++++++++++++++++++
1 file changed, 82 insertions(+)
diff --git a/arch/riscv/include/asm/checksum.h b/arch/riscv/include/asm/checksum.h
new file mode 100644
index 000000000000..5a810126aac7
--- /dev/null
+++ b/arch/riscv/include/asm/checksum.h
@@ -0,0 +1,82 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Checksum routines
+ *
+ * Copyright (C) 2023 Rivos Inc.
+ */
+#ifndef __ASM_RISCV_CHECKSUM_H
+#define __ASM_RISCV_CHECKSUM_H
+
+#include <linux/in6.h>
+#include <linux/uaccess.h>
+
+#define ip_fast_csum ip_fast_csum
+
+/* Define riscv versions of functions before importing asm-generic/checksum.h */
+#include <asm-generic/checksum.h>
+
+/**
+ * Quickly compute an IP checksum with the assumption that IPv4 headers will
+ * always be in multiples of 32-bits, and have an ihl of at least 5.
+ *
+ * @ihl: the number of 32 bit segments and must be greater than or equal to 5.
+ * @iph: assumed to be word aligned given that NET_IP_ALIGN is set to 2 on
+ * riscv, defining IP headers to be aligned.
+ */
+static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
+{
+ unsigned long csum = 0;
+ int pos = 0;
+
+ do {
+ csum += ((const unsigned int *)iph)[pos];
+ if (IS_ENABLED(CONFIG_32BIT))
+ csum += csum < ((const unsigned int *)iph)[pos];
+ } while (++pos < ihl);
+
+ /*
+ * ZBB only saves three instructions on 32-bit and five on 64-bit so not
+ * worth checking if supported without Alternatives.
+ */
+ if (IS_ENABLED(CONFIG_RISCV_ISA_ZBB) &&
+ IS_ENABLED(CONFIG_RISCV_ALTERNATIVE)) {
+ unsigned long fold_temp;
+
+ asm_volatile_goto(ALTERNATIVE("j %l[no_zbb]", "nop", 0,
+ RISCV_ISA_EXT_ZBB, 1)
+ :
+ :
+ :
+ : no_zbb);
+
+ if (IS_ENABLED(CONFIG_32BIT)) {
+ asm(".option push \n\
+ .option arch,+zbb \n\
+ not %[fold_temp], %[csum] \n\
+ rori %[csum], %[csum], 16 \n\
+ sub %[csum], %[fold_temp], %[csum] \n\
+ .option pop"
+ : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp));
+ } else {
+ asm(".option push \n\
+ .option arch,+zbb \n\
+ rori %[fold_temp], %[csum], 32 \n\
+ add %[csum], %[fold_temp], %[csum] \n\
+ srli %[csum], %[csum], 32 \n\
+ not %[fold_temp], %[csum] \n\
+ roriw %[csum], %[csum], 16 \n\
+ subw %[csum], %[fold_temp], %[csum] \n\
+ .option pop"
+ : [csum] "+r" (csum), [fold_temp] "=&r" (fold_temp));
+ }
+ return (__force __sum16)(csum >> 16);
+ }
+no_zbb:
+#ifndef CONFIG_32BIT
+ csum += ror64(csum, 32);
+ csum >>= 32;
+#endif
+ return csum_fold((__force __wsum)csum);
+}
+
+#endif /* __ASM_RISCV_CHECKSUM_H */
--
2.43.0
Supplement existing checksum tests with tests for csum_ipv6_magic and
ip_fast_csum.
Signed-off-by: Charlie Jenkins <[email protected]>
---
lib/checksum_kunit.c | 284 ++++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 283 insertions(+), 1 deletion(-)
diff --git a/lib/checksum_kunit.c b/lib/checksum_kunit.c
index 0eed92b77ba3..af3e5ca4e170 100644
--- a/lib/checksum_kunit.c
+++ b/lib/checksum_kunit.c
@@ -1,15 +1,21 @@
// SPDX-License-Identifier: GPL-2.0+
/*
- * Test cases csum_partial and csum_fold
+ * Test cases csum_partial, csum_fold, ip_fast_csum, csum_ipv6_magic
*/
#include <kunit/test.h>
#include <asm/checksum.h>
+#include <net/ip6_checksum.h>
#define MAX_LEN 512
#define MAX_ALIGN 64
#define TEST_BUFLEN (MAX_LEN + MAX_ALIGN)
+#define IPv4_MIN_WORDS 5
+#define IPv4_MAX_WORDS 15
+#define NUM_IPv6_TESTS 200
+#define NUM_IP_FAST_CSUM_TESTS 181
+
/* Values for a little endian CPU. Byte swap each half on big endian CPU. */
static const u32 random_init_sum = 0x2847aab;
static const u8 random_buf[] = {
@@ -209,6 +215,237 @@ static const u32 init_sums_no_overflow[] = {
0xffff0000, 0xfffffffb,
};
+static const __sum16 expected_csum_ipv6_magic[] = {
+ 0x18d4, 0x3085, 0x2e4b, 0xd9f4, 0xbdc8, 0x78f, 0x1034, 0x8422, 0x6fc0,
+ 0xd2f6, 0xbeb5, 0x9d3, 0x7e2a, 0x312e, 0x778e, 0xc1bb, 0x7cf2, 0x9d1e,
+ 0xca21, 0xf3ff, 0x7569, 0xb02e, 0xca86, 0x7e76, 0x4539, 0x45e3, 0xf28d,
+ 0xdf81, 0x8fd5, 0x3b5d, 0x8324, 0xf471, 0x83be, 0x1daf, 0x8c46, 0xe682,
+ 0xd1fb, 0x6b2e, 0xe687, 0x2a33, 0x4833, 0x2d67, 0x660f, 0x2e79, 0xd65e,
+ 0x6b62, 0x6672, 0x5dbd, 0x8680, 0xbaa5, 0x2229, 0x2125, 0x2d01, 0x1cc0,
+ 0x6d36, 0x33c0, 0xee36, 0xd832, 0x9820, 0x8a31, 0x53c5, 0x2e2, 0xdb0e,
+ 0x49ed, 0x17a7, 0x77a0, 0xd72e, 0x3d72, 0x7dc8, 0x5b17, 0xf55d, 0xa4d9,
+ 0x1446, 0x5d56, 0x6b2e, 0x69a5, 0xadb6, 0xff2a, 0x92e, 0xe044, 0x3402,
+ 0xbb60, 0xec7f, 0xe7e6, 0x1986, 0x32f4, 0x8f8, 0x5e00, 0x47c6, 0x3059,
+ 0x3969, 0xe957, 0x4388, 0x2854, 0x3334, 0xea71, 0xa6de, 0x33f9, 0x83fc,
+ 0x37b4, 0x5531, 0x3404, 0x1010, 0xed30, 0x610a, 0xc95, 0x9aed, 0x6ff,
+ 0x5136, 0x2741, 0x660e, 0x8b80, 0xf71, 0xa263, 0x88af, 0x7a73, 0x3c37,
+ 0x1908, 0x6db5, 0x2e92, 0x1cd2, 0x70c8, 0xee16, 0xe80, 0xcd55, 0x6e6,
+ 0x6434, 0x127, 0x655d, 0x2ea0, 0xb4f4, 0xdc20, 0x5671, 0xe462, 0xe52b,
+ 0xdb44, 0x3589, 0xc48f, 0xe60b, 0xd2d2, 0x66ad, 0x498, 0x436, 0xb917,
+ 0xf0ca, 0x1a6e, 0x1cb7, 0xbf61, 0x2870, 0xc7e8, 0x5b30, 0xe4a5, 0x168,
+ 0xadfc, 0xd035, 0xe690, 0xe283, 0xfb27, 0xe4ad, 0xb1a5, 0xf2d5, 0xc4b6,
+ 0x8a30, 0xd7d5, 0x7df9, 0x91d5, 0x63ed, 0x2d21, 0x312b, 0xab19, 0xa632,
+ 0x8d2e, 0xef06, 0x57b9, 0xc373, 0xbd1f, 0xa41f, 0x8444, 0x9975, 0x90cb,
+ 0xc49c, 0xe965, 0x4eff, 0x5a, 0xef6d, 0xe81a, 0xe260, 0x853a, 0xff7a,
+ 0x99aa, 0xb06b, 0xee19, 0xcc2c, 0xf34c, 0x7c49, 0xdac3, 0xa71e, 0xc988,
+ 0x3845, 0x1014
+};
+
+static const __sum16 expected_fast_csum[] = {
+ 0xda83, 0x45da, 0x4f46, 0x4e4f, 0x34e, 0xe902, 0xa5e9, 0x87a5, 0x7187,
+ 0x5671, 0xf556, 0x6df5, 0x816d, 0x8f81, 0xbb8f, 0xfbba, 0x5afb, 0xbe5a,
+ 0xedbe, 0xabee, 0x6aac, 0xe6b, 0xea0d, 0x67ea, 0x7e68, 0x8a7e, 0x6f8a,
+ 0x3a70, 0x9f3a, 0xe89e, 0x75e8, 0x7976, 0xfa79, 0x2cfa, 0x3c2c, 0x463c,
+ 0x7146, 0x7a71, 0x547a, 0xfd53, 0x99fc, 0xb699, 0x92b6, 0xdb91, 0xe8da,
+ 0x5fe9, 0x1e60, 0xae1d, 0x39ae, 0xf439, 0xa1f4, 0xdda1, 0xede, 0x790f,
+ 0x579, 0x1206, 0x9012, 0x2490, 0xd224, 0x5cd2, 0xa65d, 0xca7, 0x220d,
+ 0xf922, 0xbf9, 0x920b, 0x1b92, 0x361c, 0x2e36, 0x4d2e, 0x24d, 0x2,
+ 0xcfff, 0x90cf, 0xa591, 0x93a5, 0x7993, 0x9579, 0xc894, 0x50c8, 0x5f50,
+ 0xd55e, 0xcad5, 0xf3c9, 0x8f4, 0x4409, 0x5043, 0x5b50, 0x55b, 0x2205,
+ 0x1e22, 0x801e, 0x3780, 0xe137, 0x7ee0, 0xf67d, 0x3cf6, 0xa53c, 0x2ea5,
+ 0x472e, 0x5147, 0xcf51, 0x1bcf, 0x951c, 0x1e95, 0xc71e, 0xe4c7, 0xc3e4,
+ 0x3dc3, 0xee3d, 0xa4ed, 0xf9a4, 0xcbf8, 0x75cb, 0xb375, 0x50b4, 0x3551,
+ 0xf835, 0x19f8, 0x8c1a, 0x538c, 0xad52, 0xa3ac, 0xb0a3, 0x5cb0, 0x6c5c,
+ 0x5b6c, 0xc05a, 0x92c0, 0x4792, 0xbe47, 0x53be, 0x1554, 0x5715, 0x4b57,
+ 0xe54a, 0x20e5, 0x21, 0xd500, 0xa1d4, 0xa8a1, 0x57a9, 0xca57, 0x5ca,
+ 0x1c06, 0x4f1c, 0xe24e, 0xd9e2, 0xf0d9, 0x4af1, 0x474b, 0x8146, 0xe81,
+ 0xfd0e, 0x84fd, 0x7c85, 0xba7c, 0x17ba, 0x4a17, 0x964a, 0xf595, 0xff5,
+ 0x5310, 0x3253, 0x6432, 0x4263, 0x2242, 0xe121, 0x32e1, 0xf632, 0xc5f5,
+ 0x21c6, 0x7d22, 0x8e7c, 0x418e, 0x5641, 0x3156, 0x7c31, 0x737c, 0x373,
+ 0x2503, 0xc22a, 0x3c2, 0x4a04, 0x8549, 0x5285, 0xa352, 0xe8a3, 0x6fe8,
+ 0x1a6f, 0x211a, 0xe021, 0x38e0, 0x7638, 0xf575, 0x9df5, 0x169e, 0xf116,
+ 0x23f1, 0xcd23, 0xece, 0x660f, 0x4866, 0x6a48, 0x716a, 0xee71, 0xa2ee,
+ 0xb8a2, 0x61b9, 0xa361, 0xf7a2, 0x26f7, 0x1127, 0x6611, 0xe065, 0x36e0,
+ 0x1837, 0x3018, 0x1c30, 0x721b, 0x3e71, 0xe43d, 0x99e4, 0x9e9a, 0xb79d,
+ 0xa9b7, 0xcaa, 0xeb0c, 0x4eb, 0x1305, 0x8813, 0xb687, 0xa9b6, 0xfba9,
+ 0xd7fb, 0xccd8, 0x2ecd, 0x652f, 0xae65, 0x3fae, 0x3a40, 0x563a, 0x7556,
+ 0x2776, 0x1228, 0xef12, 0xf9ee, 0xcef9, 0x56cf, 0xa956, 0x24a9, 0xba24,
+ 0x5fba, 0x665f, 0xf465, 0x8ff4, 0x6d8f, 0x346d, 0x5f34, 0x385f, 0xd137,
+ 0xb8d0, 0xacb8, 0x55ac, 0x7455, 0xe874, 0x89e8, 0xd189, 0xa0d1, 0xb2a0,
+ 0xb8b2, 0x36b8, 0x5636, 0xd355, 0x8d3, 0x1908, 0x2118, 0xc21, 0x990c,
+ 0x8b99, 0x158c, 0x7815, 0x9e78, 0x6f9e, 0x4470, 0x1d44, 0x341d, 0x2634,
+ 0x3f26, 0x793e, 0xc79, 0xcc0b, 0x26cc, 0xd126, 0x1fd1, 0xb41f, 0xb6b4,
+ 0x22b7, 0xa122, 0xa1, 0x7f01, 0x837e, 0x3b83, 0xaf3b, 0x6fae, 0x916f,
+ 0xb490, 0xffb3, 0xceff, 0x50cf, 0x7550, 0x7275, 0x1272, 0x2613, 0xaa26,
+ 0xd5aa, 0x7d5, 0x9607, 0x96, 0xb100, 0xf8b0, 0x4bf8, 0xdd4c, 0xeddd,
+ 0x98ed, 0x2599, 0x9325, 0xeb92, 0x8feb, 0xcc8f, 0x2acd, 0x392b, 0x3b39,
+ 0xcb3b, 0x6acb, 0xd46a, 0xb8d4, 0x6ab8, 0x106a, 0x2f10, 0x892f, 0x789,
+ 0xc806, 0x45c8, 0x7445, 0x3c74, 0x3a3c, 0xcf39, 0xd7ce, 0x58d8, 0x6e58,
+ 0x336e, 0x1034, 0xee10, 0xe9ed, 0xc2e9, 0x3fc2, 0xd53e, 0xd2d4, 0xead2,
+ 0x8fea, 0x2190, 0x1162, 0xbe11, 0x8cbe, 0x6d8c, 0xfb6c, 0x6dfb, 0xd36e,
+ 0x3ad3, 0xf3a, 0x870e, 0xc287, 0x53c3, 0xc54, 0x5b0c, 0x7d5a, 0x797d,
+ 0xec79, 0x5dec, 0x4d5e, 0x184e, 0xd618, 0x60d6, 0xb360, 0x98b3, 0xf298,
+ 0xb1f2, 0x69b1, 0xf969, 0xef9, 0xab0e, 0x21ab, 0xe321, 0x24e3, 0x8224,
+ 0x5481, 0x5954, 0x7a59, 0xff7a, 0x7dff, 0x1a7d, 0xa51a, 0x46a5, 0x6b47,
+ 0xe6b, 0x830e, 0xa083, 0xff9f, 0xd0ff, 0xffd0, 0xe6ff, 0x7de7, 0xc67d,
+ 0xd0c6, 0x61d1, 0x3a62, 0xc3b, 0x150c, 0x1715, 0x4517, 0x5345, 0x3954,
+ 0xdd39, 0xdadd, 0x32db, 0x6a33, 0xd169, 0x86d1, 0xb687, 0x3fb6, 0x883f,
+ 0xa487, 0x39a4, 0x2139, 0xbe20, 0xffbe, 0xedfe, 0x8ded, 0x368e, 0xc335,
+ 0x51c3, 0x9851, 0xf297, 0xd6f2, 0xb9d6, 0x95ba, 0x2096, 0xea1f, 0x76e9,
+ 0x4e76, 0xe04d, 0xd0df, 0x80d0, 0xa280, 0xfca2, 0x75fc, 0xef75, 0x32ef,
+ 0x6833, 0xdf68, 0xc4df, 0x76c4, 0xb77, 0xb10a, 0xbfb1, 0x58bf, 0x5258,
+ 0x4d52, 0x6c4d, 0x7e6c, 0xb67e, 0xccb5, 0x8ccc, 0xbe8c, 0xc8bd, 0x9ac8,
+ 0xa99b, 0x52a9, 0x2f53, 0xc30, 0x3e0c, 0xb83d, 0x83b7, 0x5383, 0x7e53,
+ 0x4f7e, 0xe24e, 0xb3e1, 0x8db3, 0x618e, 0xc861, 0xfcc8, 0x34fc, 0x9b35,
+ 0xaa9b, 0xb1aa, 0x5eb1, 0x395e, 0x8639, 0xd486, 0x8bd4, 0x558b, 0x2156,
+ 0xf721, 0x4ef6, 0x14f, 0x7301, 0xdd72, 0x49de, 0x894a, 0x9889, 0x8898,
+ 0x7788, 0x7b77, 0x637b, 0xb963, 0xabb9, 0x7cab, 0xc87b, 0x21c8, 0xcb21,
+ 0xdfca, 0xbfdf, 0xf2bf, 0x6af2, 0x626b, 0xb261, 0x3cb2, 0xc63c, 0xc9c6,
+ 0xc9c9, 0xb4c9, 0xf9b4, 0x91f9, 0x4091, 0x3a40, 0xcc39, 0xd1cb, 0x7ed1,
+ 0x537f, 0x6753, 0xa167, 0xba49, 0x88ba, 0x7789, 0x3877, 0xf037, 0xd3ef,
+ 0xb5d4, 0x55b6, 0xa555, 0xeca4, 0xa1ec, 0xb6a2, 0x7b7, 0x9507, 0xfd94,
+ 0x82fd, 0x5c83, 0x765c, 0x9676, 0x3f97, 0xda3f, 0x6fda, 0x646f, 0x3064,
+ 0x5e30, 0x655e, 0x6465, 0xcb64, 0xcdca, 0x4ccd, 0x3f4c, 0x243f, 0x6f24,
+ 0x656f, 0x6065, 0x3560, 0x3b36, 0xac3b, 0x4aac, 0x714a, 0x7e71, 0xda7e,
+ 0x7fda, 0xda7f, 0x6fda, 0xff6f, 0xc6ff, 0xedc6, 0xd4ed, 0x70d5, 0xeb70,
+ 0xa3eb, 0x80a3, 0xca80, 0x3fcb, 0x2540, 0xf825, 0x7ef8, 0xf87e, 0x73f8,
+ 0xb474, 0xb4b4, 0x92b5, 0x9293, 0x93, 0x3500, 0x7134, 0x9071, 0xfa8f,
+ 0x51fa, 0x1452, 0xba13, 0x7ab9, 0x957a, 0x8a95, 0x6e8a, 0x6d6e, 0x7c6d,
+ 0x447c, 0x9744, 0x4597, 0x8945, 0xef88, 0x8fee, 0x3190, 0x4831, 0x8447,
+ 0xa183, 0x1da1, 0xd41d, 0x2dd4, 0x4f2e, 0xc94e, 0xcbc9, 0xc9cb, 0x9ec9,
+ 0x319e, 0xd531, 0x20d5, 0x4021, 0xb23f, 0x29b2, 0xd828, 0xecd8, 0x5ded,
+ 0xfc5d, 0x4dfc, 0xd24d, 0x6bd2, 0x5f6b, 0xb35e, 0x7fb3, 0xee7e, 0x56ee,
+ 0xa657, 0x68a6, 0x8768, 0x7787, 0xb077, 0x4cb1, 0x764c, 0xb175, 0x7b1,
+ 0x3d07, 0x603d, 0x3560, 0x3e35, 0xb03d, 0xd6b0, 0xc8d6, 0xd8c8, 0x8bd8,
+ 0x3e8c, 0x303f, 0xd530, 0xf1d4, 0x42f1, 0xca42, 0xddca, 0x41dd, 0x3141,
+ 0x132, 0xe901, 0x8e9, 0xbe09, 0xe0bd, 0x2ce0, 0x862d, 0x3986, 0x9139,
+ 0x6d91, 0x6a6d, 0x8d6a, 0x1b8d, 0xac1b, 0xedab, 0x54ed, 0xc054, 0xcebf,
+ 0xc1ce, 0x5c2, 0x3805, 0x6038, 0x5960, 0xd359, 0xdd3, 0xbe0d, 0xafbd,
+ 0x6daf, 0x206d, 0x2c20, 0x862c, 0x8e86, 0xec8d, 0xa2ec, 0xa3a2, 0x51a3,
+ 0x8051, 0xfd7f, 0x91fd, 0xa292, 0xaf14, 0xeeae, 0x59ef, 0x535a, 0x8653,
+ 0x3986, 0x9539, 0xb895, 0xa0b8, 0x26a0, 0x2227, 0xc022, 0x77c0, 0xad77,
+ 0x46ad, 0xaa46, 0x60aa, 0x8560, 0x4785, 0xd747, 0x45d7, 0x2346, 0x5f23,
+ 0x25f, 0x1d02, 0x71d, 0x8206, 0xc82, 0x180c, 0x3018, 0x4b30, 0x4b,
+ 0x3001, 0x1230, 0x2d12, 0x8c2d, 0x148d, 0x4015, 0x5f3f, 0x3d5f, 0x6b3d,
+ 0x396b, 0x473a, 0xf746, 0x44f7, 0x8945, 0x3489, 0xcb34, 0x84ca, 0xd984,
+ 0xf0d9, 0xbcf0, 0x63bd, 0x3264, 0xf332, 0x45f3, 0x7346, 0x5673, 0xb056,
+ 0xd3b0, 0x4ad4, 0x184b, 0x7d18, 0x6c7d, 0xbb6c, 0xfeba, 0xe0fe, 0x10e1,
+ 0x5410, 0x2954, 0x9f28, 0x3a9f, 0x5a3a, 0xdb59, 0xbdc, 0xb40b, 0x1ab4,
+ 0x131b, 0x5d12, 0x6d5c, 0xe16c, 0xb0e0, 0x89b0, 0xba88, 0xbb, 0x3c01,
+ 0xe13b, 0x6fe1, 0x446f, 0xa344, 0x81a3, 0xfe81, 0xc7fd, 0x38c8, 0xb38,
+ 0x1a0b, 0x6d19, 0xf36c, 0x47f3, 0x6d48, 0xb76d, 0xd3b7, 0xd8d2, 0x52d9,
+ 0x4b53, 0xa54a, 0x34a5, 0xc534, 0x9bc4, 0xed9b, 0xbeed, 0x3ebe, 0x233e,
+ 0x9f22, 0x4a9f, 0x774b, 0x4577, 0xa545, 0x64a5, 0xb65, 0x870b, 0x487,
+ 0x9204, 0x5f91, 0xd55f, 0x35d5, 0x1a35, 0x71a, 0x7a07, 0x4e7a, 0xfc4e,
+ 0x1efc, 0x481f, 0x7448, 0xde74, 0xa7dd, 0x1ea7, 0xaa1e, 0xcfaa, 0xfbcf,
+ 0xedfb, 0x6eee, 0x386f, 0x4538, 0x6e45, 0xd96d, 0x11d9, 0x7912, 0x4b79,
+ 0x494b, 0x6049, 0xac5f, 0x65ac, 0x1366, 0x5913, 0xe458, 0x7ae4, 0x387a,
+ 0x3c38, 0xb03c, 0x76b0, 0x9376, 0xe193, 0x42e1, 0x7742, 0x6476, 0x3564,
+ 0x3c35, 0x6a3c, 0xcc69, 0x94cc, 0x5d95, 0xe5e, 0xee0d, 0x4ced, 0xce4c,
+ 0x52ce, 0xaa52, 0xdaaa, 0xe4da, 0x1de5, 0x4530, 0x5445, 0x3954, 0xb639,
+ 0x81b6, 0x7381, 0x1574, 0xc215, 0x10c2, 0x3f10, 0x6b3f, 0xe76b, 0x7be7,
+ 0xbc7b, 0xf7bb, 0x41f7, 0xcc41, 0x38cc, 0x4239, 0xa942, 0x4a9, 0xc504,
+ 0x7cc4, 0x437c, 0x6743, 0xea67, 0x8dea, 0xe88d, 0xd8e8, 0xdcd8, 0x17dd,
+ 0x5718, 0x958, 0xa609, 0x41a5, 0x5842, 0x159, 0x9f01, 0x269f, 0x5a26,
+ 0x405a, 0xc340, 0xb4c3, 0xd4b4, 0xf4d3, 0xf1f4, 0x39f2, 0xe439, 0x67e4,
+ 0x4168, 0xa441, 0xdda3, 0xdedd, 0x9df, 0xab0a, 0xa5ab, 0x9a6, 0xba09,
+ 0x9ab9, 0xad9a, 0x5ae, 0xe205, 0xece2, 0xecec, 0x14ed, 0xd614, 0x6bd5,
+ 0x916c, 0x3391, 0x6f33, 0x206f, 0x8020, 0x780, 0x7207, 0x2472, 0x8a23,
+ 0xb689, 0x3ab6, 0xf739, 0x97f6, 0xb097, 0xa4b0, 0xe6a4, 0x88e6, 0x2789,
+ 0xb28, 0x350b, 0x1f35, 0x431e, 0x1043, 0xc30f, 0x79c3, 0x379, 0x5703,
+ 0x3256, 0x4732, 0x7247, 0x9d72, 0x489d, 0xd348, 0xa4d3, 0x7ca4, 0xbf7b,
+ 0x45c0, 0x7b45, 0x337b, 0x4034, 0x843f, 0xd083, 0x35d0, 0x6335, 0x4d63,
+ 0xe14c, 0xcce0, 0xfecc, 0x35ff, 0x5636, 0xf856, 0xeef8, 0x2def, 0xfc2d,
+ 0x4fc, 0x6e04, 0xb66d, 0x78b6, 0xbb78, 0x3dbb, 0x9a3d, 0x839a, 0x9283,
+ 0x593, 0xd504, 0x23d5, 0x5424, 0xd054, 0x61d0, 0xdb61, 0x17db, 0x1f18,
+ 0x381f, 0x9e37, 0x679e, 0x1d68, 0x381d, 0x8038, 0x917f, 0x491, 0xbb04,
+ 0x23bb, 0x4124, 0xd41, 0xa30c, 0x8ba3, 0x8b8b, 0xc68b, 0xd2c6, 0xebd2,
+ 0x93eb, 0xbd93, 0x99bd, 0x1a99, 0xea19, 0x58ea, 0xcf58, 0x73cf, 0x1073,
+ 0x9e10, 0x139e, 0xea13, 0xcde9, 0x3ecd, 0x883f, 0xf89, 0x180f, 0x2a18,
+ 0x212a, 0xce20, 0x73ce, 0xf373, 0x60f3, 0xad60, 0x4093, 0x8e40, 0xb98e,
+ 0xbfb9, 0xf1bf, 0x8bf1, 0x5e8c, 0xe95e, 0x14e9, 0x4e14, 0x1c4e, 0x7f1c,
+ 0xe77e, 0x6fe7, 0xf26f, 0x13f2, 0x8b13, 0xda8a, 0x5fda, 0xea5f, 0x4eea,
+ 0xa84f, 0x88a8, 0x1f88, 0x2820, 0x9728, 0x5a97, 0x3f5b, 0xb23f, 0x70b2,
+ 0x2c70, 0x232d, 0xf623, 0x4f6, 0x905, 0x7509, 0xd675, 0x28d7, 0x9428,
+ 0x3794, 0xf036, 0x2bf0, 0xba2c, 0xedb9, 0xd7ed, 0x59d8, 0xed59, 0x4ed,
+ 0xe304, 0x18e3, 0x5c19, 0x3d5c, 0x753d, 0x6d75, 0x956d, 0x7f95, 0xc47f,
+ 0x83c4, 0xa84, 0x2e0a, 0x5f2e, 0xb95f, 0x77b9, 0x6d78, 0xf46d, 0x1bf4,
+ 0xed1b, 0xd6ed, 0xe0d6, 0x5e1, 0x3905, 0x5638, 0xa355, 0x99a2, 0xbe99,
+ 0xb4bd, 0x85b4, 0x2e86, 0x542e, 0x6654, 0xd765, 0x73d7, 0x3a74, 0x383a,
+ 0x2638, 0x7826, 0x7677, 0x9a76, 0x7e99, 0x2e7e, 0xea2d, 0xa6ea, 0x8a7,
+ 0x109, 0x3300, 0xad32, 0x5fad, 0x465f, 0x2f46, 0xc62f, 0xd4c5, 0xad5,
+ 0xcb0a, 0x4cb, 0xb004, 0x7baf, 0xe47b, 0x92e4, 0x8e92, 0x638e, 0x1763,
+ 0xc17, 0xf20b, 0x1ff2, 0x8920, 0x5889, 0xcb58, 0xf8cb, 0xcaf8, 0x84cb,
+ 0x9f84, 0x8a9f, 0x918a, 0x4991, 0x8249, 0xff81, 0x46ff, 0x5046, 0x5f50,
+ 0x725f, 0xf772, 0x8ef7, 0xe08f, 0xc1e0, 0x1fc2, 0x9e1f, 0x8b9d, 0x108b,
+ 0x411, 0x2b04, 0xb02a, 0x1fb0, 0x1020, 0x7a0f, 0x587a, 0x8958, 0xb188,
+ 0xb1b1, 0x49b2, 0xb949, 0x7ab9, 0x917a, 0xfc91, 0xe6fc, 0x47e7, 0xbc47,
+ 0x8fbb, 0xea8e, 0x34ea, 0x2635, 0x1726, 0x9616, 0xc196, 0xa6c1, 0xf3a6,
+ 0x11f3, 0x4811, 0x3e48, 0xeb3e, 0xf7ea, 0x1bf8, 0xdb1c, 0x8adb, 0xe18a,
+ 0x42e1, 0x9d42, 0x5d9c, 0x6e5d, 0x286e, 0x4928, 0x9a49, 0xb09c, 0xa6b0,
+ 0x2a7, 0xe702, 0xf5e6, 0x9af5, 0xf9b, 0x810f, 0x8080, 0x180, 0x1702,
+ 0x5117, 0xa650, 0x11a6, 0x1011, 0x550f, 0xd554, 0xbdd5, 0x6bbe, 0xc66b,
+ 0xfc7, 0x5510, 0x5555, 0x7655, 0x177, 0x2b02, 0x6f2a, 0xb70, 0x9f0b,
+ 0xcf9e, 0xf3cf, 0x3ff4, 0xcb40, 0x8ecb, 0x768e, 0x5277, 0x8652, 0x9186,
+ 0x9991, 0x5099, 0xd350, 0x93d3, 0x6d94, 0xe6d, 0x530e, 0x3153, 0xa531,
+ 0x64a5, 0x7964, 0x7c79, 0x467c, 0x1746, 0x3017, 0x3730, 0x538, 0x5,
+ 0x1e00, 0x5b1e, 0x955a, 0xae95, 0x3eaf, 0xff3e, 0xf8ff, 0xb2f9, 0xa1b3,
+ 0xb2a1, 0x5b2, 0xad05, 0x7cac, 0x2d7c, 0xd32c, 0x80d2, 0x7280, 0x8d72,
+ 0x1b8e, 0x831b, 0xac82, 0xfdac, 0xa7fd, 0x15a8, 0xd614, 0xe0d5, 0x7be0,
+ 0xb37b, 0x61b3, 0x9661, 0x9d95, 0xc79d, 0x83c7, 0xd883, 0xead7, 0xceb,
+ 0xf60c, 0xa9f5, 0x19a9, 0xa019, 0x8f9f, 0xd48f, 0x3ad5, 0x853a, 0x985,
+ 0x5309, 0x6f52, 0x1370, 0x6e13, 0xa96d, 0x98a9, 0x5198, 0x9f51, 0xb69f,
+ 0xa1b6, 0x2ea1, 0x672e, 0x2067, 0x6520, 0xaf65, 0x6eaf, 0x7e6f, 0xee7e,
+ 0x17ef, 0xa917, 0xcea8, 0x9ace, 0xff99, 0x5dff, 0xdf5d, 0x38df, 0xa39,
+ 0x1c0b, 0xe01b, 0x46e0, 0xcb46, 0x90cb, 0xba90, 0x4bb, 0x9104, 0x9d90,
+ 0xc89c, 0xf6c8, 0x6cf6, 0x886c, 0x1789, 0xbd17, 0x70bc, 0x7e71, 0x17e,
+ 0x1f01, 0xa01f, 0xbaa0, 0x14bb, 0xfc14, 0x7afb, 0xa07a, 0x3da0, 0xbf3d,
+ 0x48bf, 0x8c48, 0x968b, 0x9d96, 0xfd9d, 0x96fd, 0x9796, 0x6b97, 0xd16b,
+ 0xf4d1, 0x3bf4, 0x253c, 0x9125, 0x6691, 0xc166, 0x34c1, 0x5735, 0x1a57,
+ 0xdc19, 0x77db, 0x8577, 0x4a85, 0x824a, 0x9182, 0x7f91, 0xfd7f, 0xb4c3,
+ 0xb5b4, 0xb3b5, 0x7eb3, 0x617e, 0x4e61, 0xa4f, 0x530a, 0x3f52, 0xa33e,
+ 0x34a3, 0x9234, 0xf091, 0xf4f0, 0x1bf5, 0x311b, 0x9631, 0x6a96, 0x386b,
+ 0x1d39, 0xe91d, 0xe8e9, 0x69e8, 0x426a, 0xee42, 0x89ee, 0x368a, 0x2837,
+ 0x7428, 0x5974, 0x6159, 0x1d62, 0x7b1d, 0xf77a, 0x7bf7, 0x6b7c, 0x696c,
+ 0xf969, 0x4cf9, 0x714c, 0x4e71, 0x6b4e, 0x256c, 0x6e25, 0xe96d, 0x94e9,
+ 0x8f94, 0x3e8f, 0x343e, 0x4634, 0xb646, 0x97b5, 0x8997, 0xe8a, 0x900e,
+ 0x8090, 0xfd80, 0xa0fd, 0x16a1, 0xf416, 0xebf4, 0x95ec, 0x1196, 0x8911,
+ 0x3d89, 0xda3c, 0x9fd9, 0xd79f, 0x4bd7, 0x214c, 0x3021, 0x4f30, 0x994e,
+ 0x5c99, 0x6f5d, 0x326f, 0xab31, 0x6aab, 0xe969, 0x90e9, 0x1190, 0xff10,
+ 0xa2fe, 0xe0a2, 0x66e1, 0x4067, 0x9e3f, 0x2d9e, 0x712d, 0x8170, 0xd180,
+ 0xffd1, 0x25ff, 0x3826, 0x2538, 0x5f24, 0xc45e, 0x1cc4, 0xdf1c, 0x93df,
+ 0xc793, 0x80c7, 0x2380, 0xd223, 0x7ed2, 0xfc7e, 0x22fd, 0x7422, 0x1474,
+ 0xb714, 0x7db6, 0x857d, 0xa85, 0xa60a, 0x88a6, 0x4289, 0x7842, 0xc278,
+ 0xf7c2, 0xcdf7, 0x84cd, 0xae84, 0x8cae, 0xb98c, 0x1aba, 0x4d1a, 0x884c,
+ 0x4688, 0xcc46, 0xd8cb, 0x2bd9, 0xbe2b, 0xa2be, 0x72a2, 0xf772, 0xd2f6,
+ 0x75d2, 0xc075, 0xa3c0, 0x63a3, 0xae63, 0x8fae, 0x2a90, 0x5f2a, 0xef5f,
+ 0x5cef, 0xa05c, 0x89a0, 0x5e89, 0x6b5e, 0x736b, 0x773, 0x9d07, 0xe99c,
+ 0x27ea, 0x2028, 0xc20, 0x980b, 0x4797, 0x2848, 0x9828, 0xc197, 0x48c2,
+ 0x2449, 0x7024, 0x570, 0x3e05, 0xd3e, 0xf60c, 0xbbf5, 0x69bb, 0x3f6a,
+ 0x740, 0xf006, 0xe0ef, 0xbbe0, 0xadbb, 0x56ad, 0xcf56, 0xbfce, 0xa9bf,
+ 0x205b, 0x6920, 0xae69, 0x50ae, 0x2050, 0xf01f, 0x27f0, 0x9427, 0x8993,
+ 0x8689, 0x4087, 0x6e40, 0xb16e, 0xa1b1, 0xe8a1, 0x87e8, 0x6f88, 0xfe6f,
+ 0x4cfe, 0xe94d, 0xd5e9, 0x47d6, 0x3148, 0x5f31, 0xc35f, 0x13c4, 0xa413,
+ 0x5a5, 0x2405, 0xc223, 0x66c2, 0x3667, 0x5e37, 0x5f5e, 0x2f5f, 0x8c2f,
+ 0xe48c, 0xd0e4, 0x4d1, 0xd104, 0xe4d0, 0xcee4, 0xfcf, 0x480f, 0xa447,
+ 0x5ea4, 0xff5e, 0xbefe, 0x8dbe, 0x1d8e, 0x411d, 0x1841, 0x6918, 0x5469,
+ 0x1155, 0xc611, 0xaac6, 0x37ab, 0x2f37, 0xca2e, 0x87ca, 0xbd87, 0xabbd,
+ 0xb3ab, 0xcb4, 0xce0c, 0xfccd, 0xa5fd, 0x72a5, 0xf072, 0x83f0, 0xfe83,
+ 0x97fd, 0xc997, 0xb0c9, 0xadb0, 0xe6ac, 0x88e6, 0x1088, 0xbe10, 0x16be,
+ 0xa916, 0xa3a8, 0x46a3, 0x5447, 0xe953, 0x84e8, 0x2085, 0xa11f, 0xfa1,
+ 0xdd0f, 0xbedc, 0x5abe, 0x805a, 0xc97f, 0x6dc9, 0x826d, 0x4a82, 0x934a,
+ 0x5293, 0xd852, 0xd3d8, 0xadd3, 0xf4ad, 0xf3f4, 0xfcf3, 0xfefc, 0xcafe,
+ 0xb7ca, 0x3cb8, 0xa13c, 0x18a1, 0x1418, 0xea13, 0x91ea, 0xf891, 0x53f8,
+ 0xa254, 0xe9a2, 0x87ea, 0x4188, 0x1c41, 0xdc1b, 0xf5db, 0xcaf5, 0x45ca,
+ 0x6d45, 0x396d, 0xde39, 0x90dd, 0x1e91, 0x1e, 0x7b00, 0x6a7b, 0xa46a,
+ 0xc9a3, 0x9bc9, 0x389b, 0x1139, 0x5211, 0x1f52, 0xeb1f, 0xabeb, 0x48ab,
+ 0x9348, 0xb392, 0x17b3, 0x1618, 0x5b16, 0x175b, 0xdc17, 0xdedb, 0x1cdf,
+ 0xeb1c, 0xd1ea, 0x4ad2, 0xd4b, 0xc20c, 0x24c2, 0x7b25, 0x137b, 0x8b13,
+ 0x618b, 0xa061, 0xff9f, 0xfffe, 0x72ff, 0xf572, 0xe2f5, 0xcfe2, 0xd2cf,
+ 0x75d3, 0x6a76, 0xc469, 0x1ec4, 0xfc1d, 0x59fb, 0x455a, 0x7a45, 0xa479,
+ 0xb7a4
+};
+
static u8 tmp_buf[TEST_BUFLEN];
#define full_csum(buff, len, sum) csum_fold(csum_partial(buff, len, sum))
@@ -338,10 +575,55 @@ static void test_csum_no_carry_inputs(struct kunit *test)
}
}
+static void test_ip_fast_csum(struct kunit *test)
+{
+ __sum16 csum_result, expected;
+
+ for (int len = IPv4_MIN_WORDS; len < IPv4_MAX_WORDS; len++) {
+ for (int index = 0; index < NUM_IP_FAST_CSUM_TESTS; index++) {
+ csum_result = ip_fast_csum(random_buf + index, len);
+ expected =
+ expected_fast_csum[(len - IPv4_MIN_WORDS) *
+ NUM_IP_FAST_CSUM_TESTS +
+ index];
+ CHECK_EQ(expected, csum_result);
+ }
+ }
+}
+
+static void test_csum_ipv6_magic(struct kunit *test)
+{
+ const struct in6_addr *saddr;
+ const struct in6_addr *daddr;
+ unsigned int len;
+ unsigned char proto;
+ unsigned int csum;
+
+ const int daddr_offset = sizeof(struct in6_addr);
+ const int len_offset = sizeof(struct in6_addr) + sizeof(struct in6_addr);
+ const int proto_offset = sizeof(struct in6_addr) + sizeof(struct in6_addr) +
+ sizeof(int);
+ const int csum_offset = sizeof(struct in6_addr) + sizeof(struct in6_addr) +
+ sizeof(int) + sizeof(char);
+
+ for (int i = 0; i < NUM_IPv6_TESTS; i++) {
+ saddr = (const struct in6_addr *)(random_buf + i);
+ daddr = (const struct in6_addr *)(random_buf + i +
+ daddr_offset);
+ len = *(unsigned int *)(random_buf + i + len_offset);
+ proto = *(random_buf + i + proto_offset);
+ csum = *(unsigned int *)(random_buf + i + csum_offset);
+ CHECK_EQ(expected_csum_ipv6_magic[i],
+ csum_ipv6_magic(saddr, daddr, len, proto, csum));
+ }
+}
+
static struct kunit_case __refdata checksum_test_cases[] = {
KUNIT_CASE(test_csum_fixed_random_inputs),
KUNIT_CASE(test_csum_all_carry_inputs),
KUNIT_CASE(test_csum_no_carry_inputs),
+ KUNIT_CASE(test_ip_fast_csum),
+ KUNIT_CASE(test_csum_ipv6_magic),
{}
};
--
2.43.0
On Wed, Dec 20, 2023 at 3:37 PM Charlie Jenkins <[email protected]> wrote:
>
> Support static branches depending on the value of misaligned accesses.
> This will be used by a later patch in the series. All cpus must be
> considered "fast" for this static branch to be flipped.
>
> Signed-off-by: Charlie Jenkins <[email protected]>
You didn't pick up my tag from the last spin, so here it is again:
Reviewed-by: Evan Green <[email protected]>
On Thu, Dec 21, 2023 at 7:38 AM Charlie Jenkins <[email protected]> wrote:
>
> Support static branches depending on the value of misaligned accesses.
> This will be used by a later patch in the series. All cpus must be
> considered "fast" for this static branch to be flipped.
>
> Signed-off-by: Charlie Jenkins <[email protected]>
> ---
> arch/riscv/include/asm/cpufeature.h | 2 ++
> arch/riscv/kernel/cpufeature.c | 30 ++++++++++++++++++++++++++++++
> 2 files changed, 32 insertions(+)
>
> diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h
> index a418c3112cd6..7b129e5e2f07 100644
> --- a/arch/riscv/include/asm/cpufeature.h
> +++ b/arch/riscv/include/asm/cpufeature.h
> @@ -133,4 +133,6 @@ static __always_inline bool riscv_cpu_has_extension_unlikely(int cpu, const unsi
> return __riscv_isa_extension_available(hart_isa[cpu].isa, ext);
> }
>
> +DECLARE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
> +
> #endif
> diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> index b3785ffc1570..095eb6ebdcaa 100644
> --- a/arch/riscv/kernel/cpufeature.c
> +++ b/arch/riscv/kernel/cpufeature.c
> @@ -10,6 +10,7 @@
> #include <linux/bitmap.h>
> #include <linux/cpuhotplug.h>
> #include <linux/ctype.h>
> +#include <linux/jump_label.h>
> #include <linux/log2.h>
> #include <linux/memory.h>
> #include <linux/module.h>
> @@ -728,6 +729,35 @@ void riscv_user_isa_enable(void)
> csr_set(CSR_SENVCFG, ENVCFG_CBZE);
> }
>
> +DEFINE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
> +
> +static int set_unaligned_access_static_branches(void)
> +{
> + /*
> + * This will be called after check_unaligned_access_all_cpus so the
> + * result of unaligned access speed for all cpus will be available.
> + */
> +
> + int cpu;
> + bool fast_misaligned_access_speed = true;
> +
> + for_each_online_cpu(cpu) {
Each online_cpu? Is there any offline_cpu that is no
fast_misaligned_access_speed?
Move into your riscv_online_cpu for each CPU, and use stop_machine for
synchronization.
> + int this_perf = per_cpu(misaligned_access_speed, cpu);
> +
> + if (this_perf != RISCV_HWPROBE_MISALIGNED_FAST) {
> + fast_misaligned_access_speed = false;
> + break;
> + }
> + }
> +
> + if (fast_misaligned_access_speed)
> + static_branch_enable(&fast_misaligned_access_speed_key);
> +
> + return 0;
> +}
> +
> +arch_initcall_sync(set_unaligned_access_static_branches);
> +
> #ifdef CONFIG_RISCV_ALTERNATIVE
> /*
> * Alternative patch sites consider 48 bits when determining when to patch
>
> --
> 2.43.0
>
>
--
Best Regards
Guo Ren
On Fri, Dec 22, 2023 at 08:33:18AM +0800, Guo Ren wrote:
> On Thu, Dec 21, 2023 at 7:38 AM Charlie Jenkins <[email protected]> wrote:
> >
> > Support static branches depending on the value of misaligned accesses.
> > This will be used by a later patch in the series. All cpus must be
> > considered "fast" for this static branch to be flipped.
> >
> > Signed-off-by: Charlie Jenkins <[email protected]>
> > ---
> > arch/riscv/include/asm/cpufeature.h | 2 ++
> > arch/riscv/kernel/cpufeature.c | 30 ++++++++++++++++++++++++++++++
> > 2 files changed, 32 insertions(+)
> >
> > diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h
> > index a418c3112cd6..7b129e5e2f07 100644
> > --- a/arch/riscv/include/asm/cpufeature.h
> > +++ b/arch/riscv/include/asm/cpufeature.h
> > @@ -133,4 +133,6 @@ static __always_inline bool riscv_cpu_has_extension_unlikely(int cpu, const unsi
> > return __riscv_isa_extension_available(hart_isa[cpu].isa, ext);
> > }
> >
> > +DECLARE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
> > +
> > #endif
> > diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> > index b3785ffc1570..095eb6ebdcaa 100644
> > --- a/arch/riscv/kernel/cpufeature.c
> > +++ b/arch/riscv/kernel/cpufeature.c
> > @@ -10,6 +10,7 @@
> > #include <linux/bitmap.h>
> > #include <linux/cpuhotplug.h>
> > #include <linux/ctype.h>
> > +#include <linux/jump_label.h>
> > #include <linux/log2.h>
> > #include <linux/memory.h>
> > #include <linux/module.h>
> > @@ -728,6 +729,35 @@ void riscv_user_isa_enable(void)
> > csr_set(CSR_SENVCFG, ENVCFG_CBZE);
> > }
> >
> > +DEFINE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
> > +
> > +static int set_unaligned_access_static_branches(void)
> > +{
> > + /*
> > + * This will be called after check_unaligned_access_all_cpus so the
> > + * result of unaligned access speed for all cpus will be available.
> > + */
> > +
> > + int cpu;
> > + bool fast_misaligned_access_speed = true;
> > +
> > + for_each_online_cpu(cpu) {
> Each online_cpu? Is there any offline_cpu that is no
> fast_misaligned_access_speed?
I think instead of checking offline cpus, it would make more sense to
adjust the static branch when offline cpus come online. Since
riscv_online_cpu is called when a new CPU comes online, I can update the
static branch inside of that function.
>
> Move into your riscv_online_cpu for each CPU, and use stop_machine for
> synchronization.
>
I do not understand what you mean by "Move into your riscv_online_cpu
for each CPU", but I am assuming you are referring to updating the
static branch inside of riscv_online_cpu.
I believe any race condition that could be solved by stop_machine will
become irrelevent by ensuring that the static branch is updated when a
new cpu comes online.
- Charlie
> > + int this_perf = per_cpu(misaligned_access_speed, cpu);
> > +
> > + if (this_perf != RISCV_HWPROBE_MISALIGNED_FAST) {
> > + fast_misaligned_access_speed = false;
> > + break;
> > + }
> > + }
> > +
> > + if (fast_misaligned_access_speed)
> > + static_branch_enable(&fast_misaligned_access_speed_key);
> > +
> > + return 0;
> > +}
> > +
> > +arch_initcall_sync(set_unaligned_access_static_branches);
> > +
> > #ifdef CONFIG_RISCV_ALTERNATIVE
> > /*
> > * Alternative patch sites consider 48 bits when determining when to patch
> >
> > --
> > 2.43.0
> >
> >
>
>
> --
> Best Regards
> Guo Ren
On Fri, Dec 22, 2023 at 9:37 AM Charlie Jenkins <[email protected]> wrote:
>
> On Fri, Dec 22, 2023 at 08:33:18AM +0800, Guo Ren wrote:
> > On Thu, Dec 21, 2023 at 7:38 AM Charlie Jenkins <[email protected]> wrote:
> > >
> > > Support static branches depending on the value of misaligned accesses.
> > > This will be used by a later patch in the series. All cpus must be
> > > considered "fast" for this static branch to be flipped.
> > >
> > > Signed-off-by: Charlie Jenkins <[email protected]>
> > > ---
> > > arch/riscv/include/asm/cpufeature.h | 2 ++
> > > arch/riscv/kernel/cpufeature.c | 30 ++++++++++++++++++++++++++++++
> > > 2 files changed, 32 insertions(+)
> > >
> > > diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h
> > > index a418c3112cd6..7b129e5e2f07 100644
> > > --- a/arch/riscv/include/asm/cpufeature.h
> > > +++ b/arch/riscv/include/asm/cpufeature.h
> > > @@ -133,4 +133,6 @@ static __always_inline bool riscv_cpu_has_extension_unlikely(int cpu, const unsi
> > > return __riscv_isa_extension_available(hart_isa[cpu].isa, ext);
> > > }
> > >
> > > +DECLARE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
> > > +
> > > #endif
> > > diff --git a/arch/riscv/kernel/cpufeature.c b/arch/riscv/kernel/cpufeature.c
> > > index b3785ffc1570..095eb6ebdcaa 100644
> > > --- a/arch/riscv/kernel/cpufeature.c
> > > +++ b/arch/riscv/kernel/cpufeature.c
> > > @@ -10,6 +10,7 @@
> > > #include <linux/bitmap.h>
> > > #include <linux/cpuhotplug.h>
> > > #include <linux/ctype.h>
> > > +#include <linux/jump_label.h>
> > > #include <linux/log2.h>
> > > #include <linux/memory.h>
> > > #include <linux/module.h>
> > > @@ -728,6 +729,35 @@ void riscv_user_isa_enable(void)
> > > csr_set(CSR_SENVCFG, ENVCFG_CBZE);
> > > }
> > >
> > > +DEFINE_STATIC_KEY_FALSE(fast_misaligned_access_speed_key);
> > > +
> > > +static int set_unaligned_access_static_branches(void)
> > > +{
> > > + /*
> > > + * This will be called after check_unaligned_access_all_cpus so the
> > > + * result of unaligned access speed for all cpus will be available.
> > > + */
> > > +
> > > + int cpu;
> > > + bool fast_misaligned_access_speed = true;
> > > +
> > > + for_each_online_cpu(cpu) {
> > Each online_cpu? Is there any offline_cpu that is no
> > fast_misaligned_access_speed?
>
> I think instead of checking offline cpus, it would make more sense to
> adjust the static branch when offline cpus come online. Since
> riscv_online_cpu is called when a new CPU comes online, I can update the
> static branch inside of that function.
>
> >
> > Move into your riscv_online_cpu for each CPU, and use stop_machine for
> > synchronization.
> >
>
> I do not understand what you mean by "Move into your riscv_online_cpu
> for each CPU", but I am assuming you are referring to updating the
> static branch inside of riscv_online_cpu.
I mean in:
arch/riscv/kernel/cpufeature.c: riscv_online_cpu()
Yes,"adjust the static branch when offline cpus come online ..."
>
> I believe any race condition that could be solved by stop_machine will
> become irrelevent by ensuring that the static branch is updated when a
> new cpu comes online.
Em... stop_machine may be not necessary.
>
> - Charlie
>
> > > + int this_perf = per_cpu(misaligned_access_speed, cpu);
> > > +
> > > + if (this_perf != RISCV_HWPROBE_MISALIGNED_FAST) {
> > > + fast_misaligned_access_speed = false;
> > > + break;
> > > + }
> > > + }
> > > +
> > > + if (fast_misaligned_access_speed)
> > > + static_branch_enable(&fast_misaligned_access_speed_key);
> > > +
> > > + return 0;
> > > +}
> > > +
> > > +arch_initcall_sync(set_unaligned_access_static_branches);
> > > +
> > > #ifdef CONFIG_RISCV_ALTERNATIVE
> > > /*
> > > * Alternative patch sites consider 48 bits when determining when to patch
> > >
> > > --
> > > 2.43.0
> > >
> > >
> >
> >
> > --
> > Best Regards
> > Guo Ren
--
Best Regards
Guo Ren