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.

drivers/md/md.c

@@ -1180,7 +1180,7 @@ static int super_90_validate(struct mddev *mddev, struct md_rdev *rdev)
 			mddev->bitmap_info.offset =
 				mddev->bitmap_info.default_offset;
 			mddev->bitmap_info.space =
-				mddev->bitmap_info.space;
+				mddev->bitmap_info.default_space;
 		}
 
 	} else if (mddev->pers == NULL) {
@@ -3429,7 +3429,7 @@ safe_delay_store(struct mddev *mddev, const char *cbuf, size_t len)
 		mddev->safemode_delay = (msec*HZ)/1000;
 		if (mddev->safemode_delay == 0)
 			mddev->safemode_delay = 1;
-		if (mddev->safemode_delay < old_delay)
+		if (mddev->safemode_delay < old_delay || old_delay == 0)
 			md_safemode_timeout((unsigned long)mddev);
 	}
 	return len;
@@ -5144,7 +5144,7 @@ int md_run(struct mddev *mddev)
 	
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	
-	if (mddev->flags)
+	if (mddev->flags & MD_UPDATE_SB_FLAGS)
 		md_update_sb(mddev, 0);
 
 	md_new_event(mddev);
@@ -5289,7 +5289,7 @@ static void __md_stop_writes(struct mddev *mddev)
 	md_super_wait(mddev);
 
 	if (mddev->ro == 0 &&
-	    (!mddev->in_sync || mddev->flags)) {
+	    (!mddev->in_sync || (mddev->flags & MD_UPDATE_SB_FLAGS))) {
 		/* mark array as shutdown cleanly */
 		mddev->in_sync = 1;
 		md_update_sb(mddev, 1);
@@ -5337,8 +5337,14 @@ static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
 		err = -EBUSY;
 		goto out;
 	}
-	if (bdev)
-		sync_blockdev(bdev);
+	if (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags)) {
+		/* Someone opened the device since we flushed it
+		 * so page cache could be dirty and it is too late
+		 * to flush.  So abort
+		 */
+		mutex_unlock(&mddev->open_mutex);
+		return -EBUSY;
+	}
 	if (mddev->pers) {
 		__md_stop_writes(mddev);
 
@@ -5373,14 +5379,14 @@ static int do_md_stop(struct mddev * mddev, int mode,
 		mutex_unlock(&mddev->open_mutex);
 		return -EBUSY;
 	}
-	if (bdev)
-		/* It is possible IO was issued on some other
-		 * open file which was closed before we took ->open_mutex.
-		 * As that was not the last close __blkdev_put will not
-		 * have called sync_blockdev, so we must.
+	if (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags)) {
+		/* Someone opened the device since we flushed it
+		 * so page cache could be dirty and it is too late
+		 * to flush.  So abort
 		 */
-		sync_blockdev(bdev);
-
+		mutex_unlock(&mddev->open_mutex);
+		return -EBUSY;
+	}
 	if (mddev->pers) {
 		if (mddev->ro)
 			set_disk_ro(disk, 0);
@@ -5628,10 +5634,7 @@ static int get_bitmap_file(struct mddev * mddev, void __user * arg)
 	char *ptr, *buf = NULL;
 	int err = -ENOMEM;
 
-	if (md_allow_write(mddev))
 	file = kmalloc(sizeof(*file), GFP_NOIO);
-	else
-		file = kmalloc(sizeof(*file), GFP_KERNEL);
 
 	if (!file)
 		goto out;
@@ -6420,6 +6423,20 @@ static int md_ioctl(struct block_device *bdev, fmode_t mode,
 						 !test_bit(MD_RECOVERY_NEEDED,
 							   &mddev->flags),
 						 msecs_to_jiffies(5000));
+	if (cmd == STOP_ARRAY || cmd == STOP_ARRAY_RO) {
+		/* Need to flush page cache, and ensure no-one else opens
+		 * and writes
+		 */
+		mutex_lock(&mddev->open_mutex);
+		if (atomic_read(&mddev->openers) > 1) {
+			mutex_unlock(&mddev->open_mutex);
+			err = -EBUSY;
+			goto abort;
+		}
+		set_bit(MD_STILL_CLOSED, &mddev->flags);
+		mutex_unlock(&mddev->open_mutex);
+		sync_blockdev(bdev);
+	}
 	err = mddev_lock(mddev);
 	if (err) {
 		printk(KERN_INFO 
@@ -6673,6 +6690,7 @@ static int md_open(struct block_device *bdev, fmode_t mode)
 
 	err = 0;
 	atomic_inc(&mddev->openers);
+	clear_bit(MD_STILL_CLOSED, &mddev->flags);
 	mutex_unlock(&mddev->open_mutex);
 
 	check_disk_change(bdev);
@@ -7817,7 +7835,7 @@ void md_check_recovery(struct mddev *mddev)
 				sysfs_notify_dirent_safe(mddev->sysfs_state);
 		}
 
-		if (mddev->flags)
+		if (mddev->flags & MD_UPDATE_SB_FLAGS)
 			md_update_sb(mddev, 0);
 
 		if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&

drivers/md/md.h

@@ -209,7 +209,11 @@ struct mddev {
 #define MD_CHANGE_DEVS	0	/* Some device status has changed */
 #define MD_CHANGE_CLEAN 1	/* transition to or from 'clean' */
 #define MD_CHANGE_PENDING 2	/* switch from 'clean' to 'active' in progress */
+#define MD_UPDATE_SB_FLAGS (1 | 2 | 4)	/* If these are set, md_update_sb needed */
 #define MD_ARRAY_FIRST_USE 3    /* First use of array, needs initialization */
+#define MD_STILL_CLOSED	4	/* If set, then array has not been opened since
+				 * md_ioctl checked on it.
+				 */
 
 	int				suspended;
 	atomic_t			active_io;

drivers/md/raid5.h

@@ -197,6 +197,7 @@ enum reconstruct_states {
 struct stripe_head {
 	struct hlist_node	hash;
 	struct list_head	lru;	      /* inactive_list or handle_list */
+	struct llist_node	release_list;
 	struct r5conf		*raid_conf;
 	short			generation;	/* increments with every
 						 * reshape */
@@ -211,6 +212,8 @@ struct stripe_head {
 	enum check_states	check_state;
 	enum reconstruct_states reconstruct_state;
 	spinlock_t		stripe_lock;
+	int			cpu;
+	struct r5worker_group	*group;
 	/**
 	 * struct stripe_operations
 	 * @target - STRIPE_OP_COMPUTE_BLK target
@@ -321,6 +324,7 @@ enum {
 	STRIPE_OPS_REQ_PENDING,
 	STRIPE_ON_UNPLUG_LIST,
 	STRIPE_DISCARD,
+	STRIPE_ON_RELEASE_LIST,
 };
 
 /*
@@ -363,6 +367,19 @@ struct disk_info {
 	struct md_rdev	*rdev, *replacement;
 };
 
+struct r5worker {
+	struct work_struct work;
+	struct r5worker_group *group;
+	bool working;
+};
+
+struct r5worker_group {
+	struct list_head handle_list;
+	struct r5conf *conf;
+	struct r5worker *workers;
+	int stripes_cnt;
+};
+
 struct r5conf {
 	struct hlist_head	*stripe_hashtbl;
 	struct mddev		*mddev;
@@ -386,6 +403,7 @@ struct r5conf {
 	int			prev_chunk_sectors;
 	int			prev_algo;
 	short			generation; /* increments with every reshape */
+	seqcount_t		gen_lock;	/* lock against generation changes */
 	unsigned long		reshape_checkpoint; /* Time we last updated
 						     * metadata */
 	long long		min_offset_diff; /* minimum difference between
@@ -445,6 +463,7 @@ struct r5conf {
 	 */
 	atomic_t		active_stripes;
 	struct list_head	inactive_list;
+	struct llist_head	released_stripes;
 	wait_queue_head_t	wait_for_stripe;
 	wait_queue_head_t	wait_for_overlap;
 	int			inactive_blocked;	/* release of inactive stripes blocked,
@@ -458,6 +477,9 @@ struct r5conf {
 	 * the new thread here until we fully activate the array.
 	 */
 	struct md_thread	*thread;
+	struct r5worker_group	*worker_groups;
+	int			group_cnt;
+	int			worker_cnt_per_group;
 };
 
 /*

include/linux/raid/pq.h

@@ -101,6 +101,7 @@ extern const struct raid6_calls raid6_altivec8;
 extern const struct raid6_calls raid6_avx2x1;
 extern const struct raid6_calls raid6_avx2x2;
 extern const struct raid6_calls raid6_avx2x4;
+extern const struct raid6_calls raid6_tilegx8;
 
 struct raid6_recov_calls {
 	void (*data2)(int, size_t, int, int, void **);

lib/raid6/Makefile

@@ -6,6 +6,7 @@ raid6_pq-y	+= algos.o recov.o tables.o int1.o int2.o int4.o \
 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
 
 hostprogs-y	+= mktables
 
@@ -110,6 +111,11 @@ $(obj)/neon8.c:   UNROLL := 8
 $(obj)/neon8.c:   $(src)/neon.uc $(src)/unroll.awk FORCE
 	$(call if_changed,unroll)
 
+targets += tilegx8.c
+$(obj)/tilegx8.c:   UNROLL := 8
+$(obj)/tilegx8.c:   $(src)/tilegx.uc $(src)/unroll.awk FORCE
+	$(call if_changed,unroll)
+
 quiet_cmd_mktable = TABLE   $@
       cmd_mktable = $(obj)/mktables > $@ || ( rm -f $@ && exit 1 )
 

lib/raid6/algos.c

@@ -65,6 +65,9 @@ const struct raid6_calls * const raid6_algos[] = {
 	&raid6_altivec2,
 	&raid6_altivec4,
 	&raid6_altivec8,
+#endif
+#if defined(CONFIG_TILEGX)
+	&raid6_tilegx8,
 #endif
 	&raid6_intx1,
 	&raid6_intx2,

lib/raid6/test/Makefile

@@ -40,13 +40,16 @@ else ifeq ($(HAS_NEON),yes)
         OBJS   += neon.o neon1.o neon2.o neon4.o neon8.o
         CFLAGS += -DCONFIG_KERNEL_MODE_NEON=1
 else
-        HAS_ALTIVEC := $(shell echo -e '\#include <altivec.h>\nvector int a;' |\
+        HAS_ALTIVEC := $(shell printf '\#include <altivec.h>\nvector int a;\n' |\
                          gcc -c -x c - >&/dev/null && \
                          rm ./-.o && echo yes)
         ifeq ($(HAS_ALTIVEC),yes)
                 OBJS += altivec1.o altivec2.o altivec4.o altivec8.o
         endif
 endif
+ifeq ($(ARCH),tilegx)
+OBJS += tilegx8.o
+endif
 
 .c.o:
 	$(CC) $(CFLAGS) -c -o $@ $<
@@ -109,11 +112,15 @@ int16.c: int.uc ../unroll.awk
 int32.c: int.uc ../unroll.awk
 	$(AWK) ../unroll.awk -vN=32 < int.uc > $@
 
+tilegx8.c: tilegx.uc ../unroll.awk
+	$(AWK) ../unroll.awk -vN=8 < tilegx.uc > $@
+
 tables.c: mktables
 	./mktables > tables.c
 
 clean:
 	rm -f *.o *.a mktables mktables.c *.uc int*.c altivec*.c neon*.c tables.c raid6test
+	rm -f tilegx*.c
 
 spotless: clean
 	rm -f *~

lib/raid6/tilegx.uc

+/* -*- linux-c -*- ------------------------------------------------------- *
+ *
+ *   Copyright 2002 H. Peter Anvin - All Rights Reserved
+ *   Copyright 2012 Tilera Corporation - All Rights Reserved
+ *
+ *   This program is free software; you can redistribute it and/or modify
+ *   it under the terms of the GNU General Public License as published by
+ *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
+ *   Boston MA 02111-1307, USA; either version 2 of the License, or
+ *   (at your option) any later version; incorporated herein by reference.
+ *
+ * ----------------------------------------------------------------------- */
+
+/*
+ * tilegx$#.c
+ *
+ * $#-way unrolled TILE-Gx SIMD for RAID-6 math.
+ *
+ * This file is postprocessed using unroll.awk.
+ *
+ */
+
+#include <linux/raid/pq.h>
+
+/* Create 8 byte copies of constant byte */
+# define NBYTES(x) (__insn_v1addi(0, x))
+# define NSIZE  8
+
+/*
+ * The SHLBYTE() operation shifts each byte left by 1, *not*
+ * rolling over into the next byte
+ */
+static inline __attribute_const__ u64 SHLBYTE(u64 v)
+{
+	/* Vector One Byte Shift Left Immediate. */
+	return __insn_v1shli(v, 1);
+}
+
+/*
+ * The MASK() operation returns 0xFF in any byte for which the high
+ * bit is 1, 0x00 for any byte for which the high bit is 0.
+ */
+static inline __attribute_const__ u64 MASK(u64 v)
+{
+	/* Vector One Byte Shift Right Signed Immediate. */
+	return __insn_v1shrsi(v, 7);
+}
+
+
+void raid6_tilegx$#_gen_syndrome(int disks, size_t bytes, void **ptrs)
+{
+	u8 **dptr = (u8 **)ptrs;
+	u64 *p, *q;
+	int d, z, z0;
+
+	u64 wd$$, wq$$, wp$$, w1$$, w2$$;
+	u64 x1d = NBYTES(0x1d);
+	u64 * z0ptr;
+
+	z0 = disks - 3;			/* Highest data disk */
+	p = (u64 *)dptr[z0+1];	/* XOR parity */
+	q = (u64 *)dptr[z0+2];	/* RS syndrome */
+
+	z0ptr = (u64 *)&dptr[z0][0];
+	for ( d = 0 ; d < bytes ; d += NSIZE*$# ) {
+		wq$$ = wp$$ = *z0ptr++;
+		for ( z = z0-1 ; z >= 0 ; z-- ) {
+			wd$$ = *(u64 *)&dptr[z][d+$$*NSIZE];
+			wp$$ = wp$$ ^ wd$$;
+			w2$$ = MASK(wq$$);
+			w1$$ = SHLBYTE(wq$$);
+			w2$$ = w2$$ & x1d;
+			w1$$ = w1$$ ^ w2$$;
+			wq$$ = w1$$ ^ wd$$;
+		}
+		*p++ = wp$$;
+		*q++ = wq$$;
+	}
+}
+
+const struct raid6_calls raid6_tilegx$# = {
+	raid6_tilegx$#_gen_syndrome,
+	NULL,
+	"tilegx$#",
+	0
+};