From 474fd6e80fe529e9adeeb7ea9d4e5d6c4da0b7fe Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Tue, 23 Aug 2016 13:30:24 +0200 Subject: RAID/s390: add SIMD implementation for raid6 gen/xor Using vector registers is slightly faster: raid6: vx128x8 gen() 19705 MB/s raid6: vx128x8 xor() 11886 MB/s raid6: using algorithm vx128x8 gen() 19705 MB/s raid6: .... xor() 11886 MB/s, rmw enabled vs the software algorithms: raid6: int64x1 gen() 3018 MB/s raid6: int64x1 xor() 1429 MB/s raid6: int64x2 gen() 4661 MB/s raid6: int64x2 xor() 3143 MB/s raid6: int64x4 gen() 5392 MB/s raid6: int64x4 xor() 3509 MB/s raid6: int64x8 gen() 4441 MB/s raid6: int64x8 xor() 3207 MB/s raid6: using algorithm int64x4 gen() 5392 MB/s raid6: .... xor() 3509 MB/s, rmw enabled Signed-off-by: Martin Schwidefsky --- lib/raid6/.gitignore | 1 + lib/raid6/Makefile | 6 ++ lib/raid6/algos.c | 3 + lib/raid6/s390vx.uc | 168 +++++++++++++++++++++++++++++++++++++++++++++++++++ 4 files changed, 178 insertions(+) create mode 100644 lib/raid6/s390vx.uc (limited to 'lib') diff --git a/lib/raid6/.gitignore b/lib/raid6/.gitignore index 0a7e494b2bcd..f01b1cb04f91 100644 --- a/lib/raid6/.gitignore +++ b/lib/raid6/.gitignore @@ -3,3 +3,4 @@ altivec*.c int*.c tables.c neon?.c +s390vx?.c diff --git a/lib/raid6/Makefile b/lib/raid6/Makefile index 3b10a48fa040..667b9607f8c0 100644 --- a/lib/raid6/Makefile +++ b/lib/raid6/Makefile @@ -7,6 +7,7 @@ raid6_pq-$(CONFIG_X86) += recov_ssse3.o recov_avx2.o mmx.o sse1.o sse2.o avx2.o raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o raid6_pq-$(CONFIG_TILEGX) += tilegx8.o +raid6_pq-$(CONFIG_S390) += s390vx8.o hostprogs-y += mktables @@ -116,6 +117,11 @@ $(obj)/tilegx8.c: UNROLL := 8 $(obj)/tilegx8.c: $(src)/tilegx.uc $(src)/unroll.awk FORCE $(call if_changed,unroll) +targets += s390vx8.c +$(obj)/s390vx8.c: UNROLL := 8 +$(obj)/s390vx8.c: $(src)/s390vx.uc $(src)/unroll.awk FORCE + $(call if_changed,unroll) + quiet_cmd_mktable = TABLE $@ cmd_mktable = $(obj)/mktables > $@ || ( rm -f $@ && exit 1 ) diff --git a/lib/raid6/algos.c b/lib/raid6/algos.c index 975c6e0434bd..e1923b602bbc 100644 --- a/lib/raid6/algos.c +++ b/lib/raid6/algos.c @@ -68,6 +68,9 @@ const struct raid6_calls * const raid6_algos[] = { #endif #if defined(CONFIG_TILEGX) &raid6_tilegx8, +#endif +#if defined(CONFIG_S390) + &raid6_s390vx8, #endif &raid6_intx1, &raid6_intx2, diff --git a/lib/raid6/s390vx.uc b/lib/raid6/s390vx.uc new file mode 100644 index 000000000000..7b45191a655f --- /dev/null +++ b/lib/raid6/s390vx.uc @@ -0,0 +1,168 @@ +/* + * raid6_vx$#.c + * + * $#-way unrolled RAID6 gen/xor functions for s390 + * based on the vector facility + * + * Copyright IBM Corp. 2016 + * Author(s): Martin Schwidefsky + * + * This file is postprocessed using unroll.awk. + */ + +#include +#include + +asm(".include \"asm/vx-insn.h\"\n"); + +#define NSIZE 16 + +static inline void LOAD_CONST(void) +{ + asm volatile("VREPIB %v24,7"); + asm volatile("VREPIB %v25,0x1d"); +} + +/* + * The SHLBYTE() operation shifts each of the 16 bytes in + * vector register y left by 1 bit and stores the result in + * vector register x. + */ +static inline void SHLBYTE(int x, int y) +{ + asm volatile ("VAB %0,%1,%1" : : "i" (x), "i" (y)); +} + +/* + * For each of the 16 bytes in the vector register y the MASK() + * operation returns 0xFF if the high bit of the byte is 1, + * or 0x00 if the high bit is 0. The result is stored in vector + * register x. + */ +static inline void MASK(int x, int y) +{ + asm volatile ("VESRAVB %0,%1,24" : : "i" (x), "i" (y)); +} + +static inline void AND(int x, int y, int z) +{ + asm volatile ("VN %0,%1,%2" : : "i" (x), "i" (y), "i" (z)); +} + +static inline void XOR(int x, int y, int z) +{ + asm volatile ("VX %0,%1,%2" : : "i" (x), "i" (y), "i" (z)); +} + +static inline void LOAD_DATA(int x, int n, u8 *ptr) +{ + typedef struct { u8 _[16*n]; } addrtype; + register addrtype *__ptr asm("1") = (addrtype *) ptr; + + asm volatile ("VLM %2,%3,0,%r1" + : : "m" (*__ptr), "a" (__ptr), "i" (x), "i" (x + n - 1)); +} + +static inline void STORE_DATA(int x, int n, u8 *ptr) +{ + typedef struct { u8 _[16*n]; } addrtype; + register addrtype *__ptr asm("1") = (addrtype *) ptr; + + asm volatile ("VSTM %2,%3,0,1" + : "=m" (*__ptr) : "a" (__ptr), "i" (x), "i" (x + n - 1)); +} + +static inline void COPY_VEC(int x, int y) +{ + asm volatile ("VLR %0,%1" : : "i" (x), "i" (y)); +} + +static void raid6_s390vx$#_gen_syndrome(int disks, size_t bytes, void **ptrs) +{ + struct kernel_fpu vxstate; + u8 **dptr, *p, *q; + int d, z, z0; + + kernel_fpu_begin(&vxstate, KERNEL_VXR); + LOAD_CONST(); + + dptr = (u8 **) ptrs; + z0 = disks - 3; /* Highest data disk */ + p = dptr[z0 + 1]; /* XOR parity */ + q = dptr[z0 + 2]; /* RS syndrome */ + + for (d = 0; d < bytes; d += $#*NSIZE) { + LOAD_DATA(0,$#,&dptr[z0][d]); + COPY_VEC(8+$$,0+$$); + for (z = z0 - 1; z >= 0; z--) { + MASK(16+$$,8+$$); + AND(16+$$,16+$$,25); + SHLBYTE(8+$$,8+$$); + XOR(8+$$,8+$$,16+$$); + LOAD_DATA(16,$#,&dptr[z][d]); + XOR(0+$$,0+$$,16+$$); + XOR(8+$$,8+$$,16+$$); + } + STORE_DATA(0,$#,&p[d]); + STORE_DATA(8,$#,&q[d]); + } + kernel_fpu_end(&vxstate, KERNEL_VXR); +} + +static void raid6_s390vx$#_xor_syndrome(int disks, int start, int stop, + size_t bytes, void **ptrs) +{ + struct kernel_fpu vxstate; + u8 **dptr, *p, *q; + int d, z, z0; + + dptr = (u8 **) ptrs; + z0 = stop; /* P/Q right side optimization */ + p = dptr[disks - 2]; /* XOR parity */ + q = dptr[disks - 1]; /* RS syndrome */ + + kernel_fpu_begin(&vxstate, KERNEL_VXR); + LOAD_CONST(); + + for (d = 0; d < bytes; d += $#*NSIZE) { + /* P/Q data pages */ + LOAD_DATA(0,$#,&dptr[z0][d]); + COPY_VEC(8+$$,0+$$); + for (z = z0 - 1; z >= start; z--) { + MASK(16+$$,8+$$); + AND(16+$$,16+$$,25); + SHLBYTE(8+$$,8+$$); + XOR(8+$$,8+$$,16+$$); + LOAD_DATA(16,$#,&dptr[z][d]); + XOR(0+$$,0+$$,16+$$); + XOR(8+$$,8+$$,16+$$); + } + /* P/Q left side optimization */ + for (z = start - 1; z >= 0; z--) { + MASK(16+$$,8+$$); + AND(16+$$,16+$$,25); + SHLBYTE(8+$$,8+$$); + XOR(8+$$,8+$$,16+$$); + } + LOAD_DATA(16,$#,&p[d]); + XOR(16+$$,16+$$,0+$$); + STORE_DATA(16,$#,&p[d]); + LOAD_DATA(16,$#,&q[d]); + XOR(16+$$,16+$$,8+$$); + STORE_DATA(16,$#,&q[d]); + } + kernel_fpu_end(&vxstate, KERNEL_VXR); +} + +static int raid6_s390vx$#_valid(void) +{ + return MACHINE_HAS_VX; +} + +const struct raid6_calls raid6_s390vx$# = { + raid6_s390vx$#_gen_syndrome, + raid6_s390vx$#_xor_syndrome, + raid6_s390vx$#_valid, + "vx128x$#", + 1 +}; -- cgit v1.2.3 From f5b55fa1f81d518925d68b50d2316850c525d1ad Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Wed, 31 Aug 2016 09:27:35 +0200 Subject: RAID/s390: provide raid6 recovery optimization The XC instruction can be used to improve the speed of the raid6 recovery. The loops now operate on blocks of 256 bytes. Signed-off-by: Martin Schwidefsky --- lib/raid6/Makefile | 2 +- lib/raid6/algos.c | 3 ++ lib/raid6/recov_s390xc.c | 116 +++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 120 insertions(+), 1 deletion(-) create mode 100644 lib/raid6/recov_s390xc.c (limited to 'lib') diff --git a/lib/raid6/Makefile b/lib/raid6/Makefile index 667b9607f8c0..29f503ebfd60 100644 --- a/lib/raid6/Makefile +++ b/lib/raid6/Makefile @@ -7,7 +7,7 @@ raid6_pq-$(CONFIG_X86) += recov_ssse3.o recov_avx2.o mmx.o sse1.o sse2.o avx2.o raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o raid6_pq-$(CONFIG_TILEGX) += tilegx8.o -raid6_pq-$(CONFIG_S390) += s390vx8.o +raid6_pq-$(CONFIG_S390) += s390vx8.o recov_s390xc.o hostprogs-y += mktables diff --git a/lib/raid6/algos.c b/lib/raid6/algos.c index e1923b602bbc..592ff49df47d 100644 --- a/lib/raid6/algos.c +++ b/lib/raid6/algos.c @@ -97,6 +97,9 @@ const struct raid6_recov_calls *const raid6_recov_algos[] = { #endif #ifdef CONFIG_AS_SSSE3 &raid6_recov_ssse3, +#endif +#ifdef CONFIG_S390 + &raid6_recov_s390xc, #endif &raid6_recov_intx1, NULL diff --git a/lib/raid6/recov_s390xc.c b/lib/raid6/recov_s390xc.c new file mode 100644 index 000000000000..b042dac826cc --- /dev/null +++ b/lib/raid6/recov_s390xc.c @@ -0,0 +1,116 @@ +/* + * RAID-6 data recovery in dual failure mode based on the XC instruction. + * + * Copyright IBM Corp. 2016 + * Author(s): Martin Schwidefsky + */ + +#include +#include + +static inline void xor_block(u8 *p1, u8 *p2) +{ + typedef struct { u8 _[256]; } addrtype; + + asm volatile( + " xc 0(256,%[p1]),0(%[p2])\n" + : "+m" (*(addrtype *) p1) : "m" (*(addrtype *) p2), + [p1] "a" (p1), [p2] "a" (p2) : "cc"); +} + +/* Recover two failed data blocks. */ +static void raid6_2data_recov_s390xc(int disks, size_t bytes, int faila, + int failb, void **ptrs) +{ + u8 *p, *q, *dp, *dq; + const u8 *pbmul; /* P multiplier table for B data */ + const u8 *qmul; /* Q multiplier table (for both) */ + int i; + + p = (u8 *)ptrs[disks-2]; + q = (u8 *)ptrs[disks-1]; + + /* Compute syndrome with zero for the missing data pages + Use the dead data pages as temporary storage for + delta p and delta q */ + dp = (u8 *)ptrs[faila]; + ptrs[faila] = (void *)raid6_empty_zero_page; + ptrs[disks-2] = dp; + dq = (u8 *)ptrs[failb]; + ptrs[failb] = (void *)raid6_empty_zero_page; + ptrs[disks-1] = dq; + + raid6_call.gen_syndrome(disks, bytes, ptrs); + + /* Restore pointer table */ + ptrs[faila] = dp; + ptrs[failb] = dq; + ptrs[disks-2] = p; + ptrs[disks-1] = q; + + /* Now, pick the proper data tables */ + pbmul = raid6_gfmul[raid6_gfexi[failb-faila]]; + qmul = raid6_gfmul[raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]]; + + /* Now do it... */ + while (bytes) { + xor_block(dp, p); + xor_block(dq, q); + for (i = 0; i < 256; i++) + dq[i] = pbmul[dp[i]] ^ qmul[dq[i]]; + xor_block(dp, dq); + p += 256; + q += 256; + dp += 256; + dq += 256; + bytes -= 256; + } +} + +/* Recover failure of one data block plus the P block */ +static void raid6_datap_recov_s390xc(int disks, size_t bytes, int faila, + void **ptrs) +{ + u8 *p, *q, *dq; + const u8 *qmul; /* Q multiplier table */ + int i; + + p = (u8 *)ptrs[disks-2]; + q = (u8 *)ptrs[disks-1]; + + /* Compute syndrome with zero for the missing data page + Use the dead data page as temporary storage for delta q */ + dq = (u8 *)ptrs[faila]; + ptrs[faila] = (void *)raid6_empty_zero_page; + ptrs[disks-1] = dq; + + raid6_call.gen_syndrome(disks, bytes, ptrs); + + /* Restore pointer table */ + ptrs[faila] = dq; + ptrs[disks-1] = q; + + /* Now, pick the proper data tables */ + qmul = raid6_gfmul[raid6_gfinv[raid6_gfexp[faila]]]; + + /* Now do it... */ + while (bytes) { + xor_block(dq, q); + for (i = 0; i < 256; i++) + dq[i] = qmul[dq[i]]; + xor_block(p, dq); + p += 256; + q += 256; + dq += 256; + bytes -= 256; + } +} + + +const struct raid6_recov_calls raid6_recov_s390xc = { + .data2 = raid6_2data_recov_s390xc, + .datap = raid6_datap_recov_s390xc, + .valid = NULL, + .name = "s390xc", + .priority = 1, +}; -- cgit v1.2.3 From 725c4d22bbc4fcac5779963e0ff9cdf232afbb90 Mon Sep 17 00:00:00 2001 From: Christian Borntraeger Date: Mon, 12 Sep 2016 14:37:19 +0200 Subject: ubsan: allow to disable the null sanitizer Some architectures use a hardware defined structure at address zero. Checking for a null pointer will result in many ubsan reports. Allow users to disable the null sanitizer. Signed-off-by: Christian Borntraeger Acked-by: Andrey Ryabinin Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky --- lib/Kconfig.ubsan | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'lib') diff --git a/lib/Kconfig.ubsan b/lib/Kconfig.ubsan index 39494af9a84a..bc6e651df68c 100644 --- a/lib/Kconfig.ubsan +++ b/lib/Kconfig.ubsan @@ -1,6 +1,9 @@ config ARCH_HAS_UBSAN_SANITIZE_ALL bool +config ARCH_WANTS_UBSAN_NO_NULL + def_bool n + config UBSAN bool "Undefined behaviour sanity checker" help @@ -34,3 +37,11 @@ config UBSAN_ALIGNMENT This option enables detection of unaligned memory accesses. Enabling this option on architectures that support unaligned accesses may produce a lot of false positives. + +config UBSAN_NULL + bool "Enable checking of null pointers" + depends on UBSAN + default y if !ARCH_WANTS_UBSAN_NO_NULL + help + This option enables detection of memory accesses via a + null pointer. -- cgit v1.2.3