diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2013-09-10 16:03:41 -0400 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2013-09-10 16:03:41 -0400 |
| commit | 4d7696f1b05f4aeb586c74868fe3da2731daca4b (patch) | |
| tree | dd6cf4d41df2c0a1f52a85a3f8b8af5a9ebdeb5d /lib/raid6 | |
| parent | b05430fc9341fea7a6228a3611c850a476809596 (diff) | |
| parent | bfc90cb0936f5b972706625f38f72c7cb726c20a (diff) | |
Merge tag 'md/3.12' of git://neil.brown.name/md
Pull md update from Neil Brown:
"Headline item is multithreading for RAID5 so that more IO/sec can be
supported on fast (SSD) devices. Also TILE-Gx SIMD suppor for RAID6
calculations and an assortment of bug fixes"
* tag 'md/3.12' of git://neil.brown.name/md:
raid5: only wakeup necessary threads
md/raid5: flush out all pending requests before proceeding with reshape.
md/raid5: use seqcount to protect access to shape in make_request.
raid5: sysfs entry to control worker thread number
raid5: offload stripe handle to workqueue
raid5: fix stripe release order
raid5: make release_stripe lockless
md: avoid deadlock when dirty buffers during md_stop.
md: Don't test all of mddev->flags at once.
md: Fix apparent cut-and-paste error in super_90_validate
raid6/test: replace echo -e with printf
RAID: add tilegx SIMD implementation of raid6
md: fix safe_mode buglet.
md: don't call md_allow_write in get_bitmap_file.
Diffstat (limited to 'lib/raid6')
| -rw-r--r-- | lib/raid6/Makefile | 6 | ||||
| -rw-r--r-- | lib/raid6/algos.c | 3 | ||||
| -rw-r--r-- | lib/raid6/test/Makefile | 9 | ||||
| -rw-r--r-- | lib/raid6/tilegx.uc | 86 |
4 files changed, 103 insertions, 1 deletions
diff --git a/lib/raid6/Makefile b/lib/raid6/Makefile index b4625787c7ee..c7dab0645554 100644 --- a/lib/raid6/Makefile +++ b/lib/raid6/Makefile | |||
| @@ -6,6 +6,7 @@ raid6_pq-y += algos.o recov.o tables.o int1.o int2.o int4.o \ | |||
| 6 | raid6_pq-$(CONFIG_X86) += recov_ssse3.o recov_avx2.o mmx.o sse1.o sse2.o avx2.o | 6 | raid6_pq-$(CONFIG_X86) += recov_ssse3.o recov_avx2.o mmx.o sse1.o sse2.o avx2.o |
| 7 | raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o | 7 | raid6_pq-$(CONFIG_ALTIVEC) += altivec1.o altivec2.o altivec4.o altivec8.o |
| 8 | raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o | 8 | raid6_pq-$(CONFIG_KERNEL_MODE_NEON) += neon.o neon1.o neon2.o neon4.o neon8.o |
| 9 | raid6_pq-$(CONFIG_TILEGX) += tilegx8.o | ||
| 9 | 10 | ||
| 10 | hostprogs-y += mktables | 11 | hostprogs-y += mktables |
| 11 | 12 | ||
| @@ -110,6 +111,11 @@ $(obj)/neon8.c: UNROLL := 8 | |||
| 110 | $(obj)/neon8.c: $(src)/neon.uc $(src)/unroll.awk FORCE | 111 | $(obj)/neon8.c: $(src)/neon.uc $(src)/unroll.awk FORCE |
| 111 | $(call if_changed,unroll) | 112 | $(call if_changed,unroll) |
| 112 | 113 | ||
| 114 | targets += tilegx8.c | ||
| 115 | $(obj)/tilegx8.c: UNROLL := 8 | ||
| 116 | $(obj)/tilegx8.c: $(src)/tilegx.uc $(src)/unroll.awk FORCE | ||
| 117 | $(call if_changed,unroll) | ||
| 118 | |||
| 113 | quiet_cmd_mktable = TABLE $@ | 119 | quiet_cmd_mktable = TABLE $@ |
| 114 | cmd_mktable = $(obj)/mktables > $@ || ( rm -f $@ && exit 1 ) | 120 | cmd_mktable = $(obj)/mktables > $@ || ( rm -f $@ && exit 1 ) |
| 115 | 121 | ||
diff --git a/lib/raid6/algos.c b/lib/raid6/algos.c index 74e6f5629dbc..f0b1aa3586d1 100644 --- a/lib/raid6/algos.c +++ b/lib/raid6/algos.c | |||
| @@ -66,6 +66,9 @@ const struct raid6_calls * const raid6_algos[] = { | |||
| 66 | &raid6_altivec4, | 66 | &raid6_altivec4, |
| 67 | &raid6_altivec8, | 67 | &raid6_altivec8, |
| 68 | #endif | 68 | #endif |
| 69 | #if defined(CONFIG_TILEGX) | ||
| 70 | &raid6_tilegx8, | ||
| 71 | #endif | ||
| 69 | &raid6_intx1, | 72 | &raid6_intx1, |
| 70 | &raid6_intx2, | 73 | &raid6_intx2, |
| 71 | &raid6_intx4, | 74 | &raid6_intx4, |
diff --git a/lib/raid6/test/Makefile b/lib/raid6/test/Makefile index 28afa1a06e03..29090f3db677 100644 --- a/lib/raid6/test/Makefile +++ b/lib/raid6/test/Makefile | |||
| @@ -40,13 +40,16 @@ else ifeq ($(HAS_NEON),yes) | |||
| 40 | OBJS += neon.o neon1.o neon2.o neon4.o neon8.o | 40 | OBJS += neon.o neon1.o neon2.o neon4.o neon8.o |
| 41 | CFLAGS += -DCONFIG_KERNEL_MODE_NEON=1 | 41 | CFLAGS += -DCONFIG_KERNEL_MODE_NEON=1 |
| 42 | else | 42 | else |
| 43 | HAS_ALTIVEC := $(shell echo -e '\#include <altivec.h>\nvector int a;' |\ | 43 | HAS_ALTIVEC := $(shell printf '\#include <altivec.h>\nvector int a;\n' |\ |
| 44 | gcc -c -x c - >&/dev/null && \ | 44 | gcc -c -x c - >&/dev/null && \ |
| 45 | rm ./-.o && echo yes) | 45 | rm ./-.o && echo yes) |
| 46 | ifeq ($(HAS_ALTIVEC),yes) | 46 | ifeq ($(HAS_ALTIVEC),yes) |
| 47 | OBJS += altivec1.o altivec2.o altivec4.o altivec8.o | 47 | OBJS += altivec1.o altivec2.o altivec4.o altivec8.o |
| 48 | endif | 48 | endif |
| 49 | endif | 49 | endif |
| 50 | ifeq ($(ARCH),tilegx) | ||
| 51 | OBJS += tilegx8.o | ||
| 52 | endif | ||
| 50 | 53 | ||
| 51 | .c.o: | 54 | .c.o: |
| 52 | $(CC) $(CFLAGS) -c -o $@ $< | 55 | $(CC) $(CFLAGS) -c -o $@ $< |
| @@ -109,11 +112,15 @@ int16.c: int.uc ../unroll.awk | |||
| 109 | int32.c: int.uc ../unroll.awk | 112 | int32.c: int.uc ../unroll.awk |
| 110 | $(AWK) ../unroll.awk -vN=32 < int.uc > $@ | 113 | $(AWK) ../unroll.awk -vN=32 < int.uc > $@ |
| 111 | 114 | ||
| 115 | tilegx8.c: tilegx.uc ../unroll.awk | ||
| 116 | $(AWK) ../unroll.awk -vN=8 < tilegx.uc > $@ | ||
| 117 | |||
| 112 | tables.c: mktables | 118 | tables.c: mktables |
| 113 | ./mktables > tables.c | 119 | ./mktables > tables.c |
| 114 | 120 | ||
| 115 | clean: | 121 | clean: |
| 116 | rm -f *.o *.a mktables mktables.c *.uc int*.c altivec*.c neon*.c tables.c raid6test | 122 | rm -f *.o *.a mktables mktables.c *.uc int*.c altivec*.c neon*.c tables.c raid6test |
| 123 | rm -f tilegx*.c | ||
| 117 | 124 | ||
| 118 | spotless: clean | 125 | spotless: clean |
| 119 | rm -f *~ | 126 | rm -f *~ |
diff --git a/lib/raid6/tilegx.uc b/lib/raid6/tilegx.uc new file mode 100644 index 000000000000..e7c29459cbcd --- /dev/null +++ b/lib/raid6/tilegx.uc | |||
| @@ -0,0 +1,86 @@ | |||
| 1 | /* -*- linux-c -*- ------------------------------------------------------- * | ||
| 2 | * | ||
| 3 | * Copyright 2002 H. Peter Anvin - All Rights Reserved | ||
| 4 | * Copyright 2012 Tilera Corporation - All Rights Reserved | ||
| 5 | * | ||
| 6 | * This program is free software; you can redistribute it and/or modify | ||
| 7 | * it under the terms of the GNU General Public License as published by | ||
| 8 | * the Free Software Foundation, Inc., 53 Temple Place Ste 330, | ||
| 9 | * Boston MA 02111-1307, USA; either version 2 of the License, or | ||
| 10 | * (at your option) any later version; incorporated herein by reference. | ||
| 11 | * | ||
| 12 | * ----------------------------------------------------------------------- */ | ||
| 13 | |||
| 14 | /* | ||
| 15 | * tilegx$#.c | ||
| 16 | * | ||
| 17 | * $#-way unrolled TILE-Gx SIMD for RAID-6 math. | ||
| 18 | * | ||
| 19 | * This file is postprocessed using unroll.awk. | ||
| 20 | * | ||
| 21 | */ | ||
| 22 | |||
| 23 | #include <linux/raid/pq.h> | ||
| 24 | |||
| 25 | /* Create 8 byte copies of constant byte */ | ||
| 26 | # define NBYTES(x) (__insn_v1addi(0, x)) | ||
| 27 | # define NSIZE 8 | ||
| 28 | |||
| 29 | /* | ||
| 30 | * The SHLBYTE() operation shifts each byte left by 1, *not* | ||
| 31 | * rolling over into the next byte | ||
| 32 | */ | ||
| 33 | static inline __attribute_const__ u64 SHLBYTE(u64 v) | ||
| 34 | { | ||
| 35 | /* Vector One Byte Shift Left Immediate. */ | ||
| 36 | return __insn_v1shli(v, 1); | ||
| 37 | } | ||
| 38 | |||
| 39 | /* | ||
| 40 | * The MASK() operation returns 0xFF in any byte for which the high | ||
| 41 | * bit is 1, 0x00 for any byte for which the high bit is 0. | ||
| 42 | */ | ||
| 43 | static inline __attribute_const__ u64 MASK(u64 v) | ||
| 44 | { | ||
| 45 | /* Vector One Byte Shift Right Signed Immediate. */ | ||
| 46 | return __insn_v1shrsi(v, 7); | ||
| 47 | } | ||
| 48 | |||
| 49 | |||
| 50 | void raid6_tilegx$#_gen_syndrome(int disks, size_t bytes, void **ptrs) | ||
| 51 | { | ||
| 52 | u8 **dptr = (u8 **)ptrs; | ||
| 53 | u64 *p, *q; | ||
| 54 | int d, z, z0; | ||
| 55 | |||
| 56 | u64 wd$$, wq$$, wp$$, w1$$, w2$$; | ||
| 57 | u64 x1d = NBYTES(0x1d); | ||
| 58 | u64 * z0ptr; | ||
| 59 | |||
| 60 | z0 = disks - 3; /* Highest data disk */ | ||
| 61 | p = (u64 *)dptr[z0+1]; /* XOR parity */ | ||
| 62 | q = (u64 *)dptr[z0+2]; /* RS syndrome */ | ||
| 63 | |||
| 64 | z0ptr = (u64 *)&dptr[z0][0]; | ||
| 65 | for ( d = 0 ; d < bytes ; d += NSIZE*$# ) { | ||
| 66 | wq$$ = wp$$ = *z0ptr++; | ||
| 67 | for ( z = z0-1 ; z >= 0 ; z-- ) { | ||
| 68 | wd$$ = *(u64 *)&dptr[z][d+$$*NSIZE]; | ||
| 69 | wp$$ = wp$$ ^ wd$$; | ||
| 70 | w2$$ = MASK(wq$$); | ||
| 71 | w1$$ = SHLBYTE(wq$$); | ||
| 72 | w2$$ = w2$$ & x1d; | ||
| 73 | w1$$ = w1$$ ^ w2$$; | ||
| 74 | wq$$ = w1$$ ^ wd$$; | ||
| 75 | } | ||
| 76 | *p++ = wp$$; | ||
| 77 | *q++ = wq$$; | ||
| 78 | } | ||
| 79 | } | ||
| 80 | |||
| 81 | const struct raid6_calls raid6_tilegx$# = { | ||
| 82 | raid6_tilegx$#_gen_syndrome, | ||
| 83 | NULL, | ||
| 84 | "tilegx$#", | ||
| 85 | 0 | ||
| 86 | }; | ||
