Merge tag 'md-3.5' of git://neil.brown.name/md

Pull md updates from NeilBrown:
 "It's been a busy cycle for md - lots of fun stuff here..  if you like
  this kind of thing :-)

  Main features:
   - RAID10 arrays can be reshaped - adding and removing devices and
     changing chunks (not 'far' array though)
   - allow RAID5 arrays to be reshaped with a backup file (not tested
     yet, but the priciple works fine for RAID10).
   - arrays can be reshaped while a bitmap is present - you no longer
     need to remove it first
   - SSSE3 support for RAID6 syndrome calculations

  and of course a number of minor fixes etc."

* tag 'md-3.5' of git://neil.brown.name/md: (56 commits)
  md/bitmap: record the space available for the bitmap in the superblock.
  md/raid10: Remove extras after reshape to smaller number of devices.
  md/raid5: improve removal of extra devices after reshape.
  md: check the return of mddev_find()
  MD RAID1: Further conditionalize 'fullsync'
  DM RAID: Use md_error() in place of simply setting Faulty bit
  DM RAID: Record and handle missing devices
  DM RAID: Set recovery flags on resume
  md/raid5: Allow reshape while a bitmap is present.
  md/raid10: resize bitmap when required during reshape.
  md: allow array to be resized while bitmap is present.
  md/bitmap: make sure reshape request are reflected in superblock.
  md/bitmap: add bitmap_resize function to allow bitmap resizing.
  md/bitmap: use DIV_ROUND_UP instead of open-code
  md/bitmap: create a 'struct bitmap_counts' substructure of 'struct bitmap'
  md/bitmap: make bitmap bitops atomic.
  md/bitmap: make _page_attr bitops atomic.
  md/bitmap: merge bitmap_file_unmap and bitmap_file_put.
  md/bitmap: remove async freeing of bitmap file.
  md/bitmap: convert some spin_lock_irqsave to spin_lock_irq
  ...

8 files changed, 491 insertions, 62 deletions

diff --git a/lib/raid6/Makefile b/lib/raid6/Makefile
index 8a38102770f3..de06dfe165b8 100644
--- a/lib/raid6/Makefile
+++ b/lib/raid6/Makefile
@@ -1,6 +1,6 @@
 obj-$(CONFIG_RAID6_PQ)  += raid6_pq.o
 
-raid6_pq-y      += algos.o recov.o tables.o int1.o int2.o int4.o \
+raid6_pq-y      += algos.o recov.o recov_ssse3.o tables.o int1.o int2.o int4.o \
                    int8.o int16.o int32.o altivec1.o altivec2.o altivec4.o \
                    altivec8.o mmx.o sse1.o sse2.o
 hostprogs-y     += mktables
diff --git a/lib/raid6/algos.c b/lib/raid6/algos.c
index 8b02f60ffc86..589f5f50ad2e 100644
--- a/lib/raid6/algos.c
+++ b/lib/raid6/algos.c
@@ -17,11 +17,11 @@
  */
 
 #include <linux/raid/pq.h>
-#include <linux/module.h>
 #ifndef __KERNEL__
 #include <sys/mman.h>
 #include <stdio.h>
 #else
+#include <linux/module.h>
 #include <linux/gfp.h>
 #if !RAID6_USE_EMPTY_ZERO_PAGE
 /* In .bss so it's zeroed */
@@ -34,10 +34,6 @@ struct raid6_calls raid6_call;
 EXPORT_SYMBOL_GPL(raid6_call);
 
 const struct raid6_calls * const raid6_algos[] = {
-        &raid6_intx1,
-        &raid6_intx2,
-        &raid6_intx4,
-        &raid6_intx8,
 #if defined(__ia64__)
         &raid6_intx16,
         &raid6_intx32,
@@ -61,6 +57,24 @@ const struct raid6_calls * const raid6_algos[] = {
         &raid6_altivec4,
         &raid6_altivec8,
 #endif
+        &raid6_intx1,
+        &raid6_intx2,
+        &raid6_intx4,
+        &raid6_intx8,
+        NULL
+};
+
+void (*raid6_2data_recov)(int, size_t, int, int, void **);
+EXPORT_SYMBOL_GPL(raid6_2data_recov);
+
+void (*raid6_datap_recov)(int, size_t, int, void **);
+EXPORT_SYMBOL_GPL(raid6_datap_recov);
+
+const struct raid6_recov_calls *const raid6_recov_algos[] = {
+#if (defined(__i386__) || defined(__x86_64__)) && !defined(__arch_um__)
+        &raid6_recov_ssse3,
+#endif
+        &raid6_recov_intx1,
         NULL
 };
 
@@ -72,59 +86,55 @@ const struct raid6_calls * const raid6_algos[] = {
 #define time_before(x, y) ((x) < (y))
 #endif
 
-/* Try to pick the best algorithm */
-/* This code uses the gfmul table as convenient data set to abuse */
-
-int __init raid6_select_algo(void)
+static inline const struct raid6_recov_calls *raid6_choose_recov(void)
 {
-        const struct raid6_calls * const * algo;
-        const struct raid6_calls * best;
-        char *syndromes;
-        void *dptrs[(65536/PAGE_SIZE)+2];
-        int i, disks;
-        unsigned long perf, bestperf;
-        int bestprefer;
-        unsigned long j0, j1;
+        const struct raid6_recov_calls *const *algo;
+        const struct raid6_recov_calls *best;
 
-        disks = (65536/PAGE_SIZE)+2;
-        for ( i = 0 ; i < disks-2 ; i++ ) {
-                dptrs[i] = ((char *)raid6_gfmul) + PAGE_SIZE*i;
-        }
+        for (best = NULL, algo = raid6_recov_algos; *algo; algo++)
+                if (!best || (*algo)->priority > best->priority)
+                        if (!(*algo)->valid || (*algo)->valid())
+                                best = *algo;
 
-        /* Normal code - use a 2-page allocation to avoid D$ conflict */
-        syndromes = (void *) __get_free_pages(GFP_KERNEL, 1);
+        if (best) {
+                raid6_2data_recov = best->data2;
+                raid6_datap_recov = best->datap;
 
-        if ( !syndromes ) {
-                printk("raid6: Yikes!  No memory available.\n");
-                return -ENOMEM;
-        }
+                printk("raid6: using %s recovery algorithm\n", best->name);
+        } else
+                printk("raid6: Yikes! No recovery algorithm found!\n");
 
-        dptrs[disks-2] = syndromes;
-        dptrs[disks-1] = syndromes + PAGE_SIZE;
+        return best;
+}
+
+static inline const struct raid6_calls *raid6_choose_gen(
+        void *(*const dptrs)[(65536/PAGE_SIZE)+2], const int disks)
+{
+        unsigned long perf, bestperf, j0, j1;
+        const struct raid6_calls *const *algo;
+        const struct raid6_calls *best;
 
-        bestperf = 0;  bestprefer = 0;  best = NULL;
+        for (bestperf = 0, best = NULL, algo = raid6_algos; *algo; algo++) {
+                if (!best || (*algo)->prefer >= best->prefer) {
+                        if ((*algo)->valid && !(*algo)->valid())
+                                continue;
 
-        for ( algo = raid6_algos ; *algo ; algo++ ) {
-                if ( !(*algo)->valid || (*algo)->valid() ) {
                         perf = 0;
 
                         preempt_disable();
                         j0 = jiffies;
-                        while ( (j1 = jiffies) == j0 )
+                        while ((j1 = jiffies) == j0)
                                 cpu_relax();
                         while (time_before(jiffies,
                                             j1 + (1<<RAID6_TIME_JIFFIES_LG2))) {
-                                (*algo)->gen_syndrome(disks, PAGE_SIZE, dptrs);
+                                (*algo)->gen_syndrome(disks, PAGE_SIZE, *dptrs);
                                 perf++;
                         }
                         preempt_enable();
 
-                        if ( (*algo)->prefer > bestprefer ||
-                             ((*algo)->prefer == bestprefer &&
-                              perf > bestperf) ) {
-                                best = *algo;
-                                bestprefer = best->prefer;
+                        if (perf > bestperf) {
                                 bestperf = perf;
+                                best = *algo;
                         }
                         printk("raid6: %-8s %5ld MB/s\n", (*algo)->name,
                                (perf*HZ) >> (20-16+RAID6_TIME_JIFFIES_LG2));
@@ -139,9 +149,46 @@ int __init raid6_select_algo(void)
         } else
                 printk("raid6: Yikes!  No algorithm found!\n");
 
+        return best;
+}
+
+
+/* Try to pick the best algorithm */
+/* This code uses the gfmul table as convenient data set to abuse */
+