Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / ufs / balloc.c
diff --git a/kernel/fs/ufs/balloc.c b/kernel/fs/ufs/balloc.c
new file mode 100644 (file)
index 0000000..a7106ed
--- /dev/null
@@ -0,0 +1,938 @@
+/*
+ *  linux/fs/ufs/balloc.c
+ *
+ * Copyright (C) 1998
+ * Daniel Pirkl <daniel.pirkl@email.cz>
+ * Charles University, Faculty of Mathematics and Physics
+ *
+ * UFS2 write support Evgeniy Dushistov <dushistov@mail.ru>, 2007
+ */
+
+#include <linux/fs.h>
+#include <linux/stat.h>
+#include <linux/time.h>
+#include <linux/string.h>
+#include <linux/buffer_head.h>
+#include <linux/capability.h>
+#include <linux/bitops.h>
+#include <asm/byteorder.h>
+
+#include "ufs_fs.h"
+#include "ufs.h"
+#include "swab.h"
+#include "util.h"
+
+#define INVBLOCK ((u64)-1L)
+
+static u64 ufs_add_fragments(struct inode *, u64, unsigned, unsigned);
+static u64 ufs_alloc_fragments(struct inode *, unsigned, u64, unsigned, int *);
+static u64 ufs_alloccg_block(struct inode *, struct ufs_cg_private_info *, u64, int *);
+static u64 ufs_bitmap_search (struct super_block *, struct ufs_cg_private_info *, u64, unsigned);
+static unsigned char ufs_fragtable_8fpb[], ufs_fragtable_other[];
+static void ufs_clusteracct(struct super_block *, struct ufs_cg_private_info *, unsigned, int);
+
+/*
+ * Free 'count' fragments from fragment number 'fragment'
+ */
+void ufs_free_fragments(struct inode *inode, u64 fragment, unsigned count)
+{
+       struct super_block * sb;
+       struct ufs_sb_private_info * uspi;
+       struct ufs_cg_private_info * ucpi;
+       struct ufs_cylinder_group * ucg;
+       unsigned cgno, bit, end_bit, bbase, blkmap, i;
+       u64 blkno;
+       
+       sb = inode->i_sb;
+       uspi = UFS_SB(sb)->s_uspi;
+       
+       UFSD("ENTER, fragment %llu, count %u\n",
+            (unsigned long long)fragment, count);
+       
+       if (ufs_fragnum(fragment) + count > uspi->s_fpg)
+               ufs_error (sb, "ufs_free_fragments", "internal error");
+
+       mutex_lock(&UFS_SB(sb)->s_lock);
+       
+       cgno = ufs_dtog(uspi, fragment);
+       bit = ufs_dtogd(uspi, fragment);
+       if (cgno >= uspi->s_ncg) {
+               ufs_panic (sb, "ufs_free_fragments", "freeing blocks are outside device");
+               goto failed;
+       }
+               
+       ucpi = ufs_load_cylinder (sb, cgno);
+       if (!ucpi) 
+               goto failed;
+       ucg = ubh_get_ucg (UCPI_UBH(ucpi));
+       if (!ufs_cg_chkmagic(sb, ucg)) {
+               ufs_panic (sb, "ufs_free_fragments", "internal error, bad magic number on cg %u", cgno);
+               goto failed;
+       }
+
+       end_bit = bit + count;
+       bbase = ufs_blknum (bit);
+       blkmap = ubh_blkmap (UCPI_UBH(ucpi), ucpi->c_freeoff, bbase);
+       ufs_fragacct (sb, blkmap, ucg->cg_frsum, -1);
+       for (i = bit; i < end_bit; i++) {
+               if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, i))
+                       ubh_setbit (UCPI_UBH(ucpi), ucpi->c_freeoff, i);
+               else 
+                       ufs_error (sb, "ufs_free_fragments",
+                                  "bit already cleared for fragment %u", i);
+       }
+       
+       fs32_add(sb, &ucg->cg_cs.cs_nffree, count);
+       uspi->cs_total.cs_nffree += count;
+       fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
+       blkmap = ubh_blkmap (UCPI_UBH(ucpi), ucpi->c_freeoff, bbase);
+       ufs_fragacct(sb, blkmap, ucg->cg_frsum, 1);
+
+       /*
+        * Trying to reassemble free fragments into block
+        */
+       blkno = ufs_fragstoblks (bbase);
+       if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno)) {
+               fs32_sub(sb, &ucg->cg_cs.cs_nffree, uspi->s_fpb);
+               uspi->cs_total.cs_nffree -= uspi->s_fpb;
+               fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, uspi->s_fpb);
+               if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
+                       ufs_clusteracct (sb, ucpi, blkno, 1);
+               fs32_add(sb, &ucg->cg_cs.cs_nbfree, 1);
+               uspi->cs_total.cs_nbfree++;
+               fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nbfree, 1);
+               if (uspi->fs_magic != UFS2_MAGIC) {
+                       unsigned cylno = ufs_cbtocylno (bbase);
+
+                       fs16_add(sb, &ubh_cg_blks(ucpi, cylno,
+                                                 ufs_cbtorpos(bbase)), 1);
+                       fs32_add(sb, &ubh_cg_blktot(ucpi, cylno), 1);
+               }
+       }
+       
+       ubh_mark_buffer_dirty (USPI_UBH(uspi));
+       ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
+       if (sb->s_flags & MS_SYNCHRONOUS)
+               ubh_sync_block(UCPI_UBH(ucpi));
+       ufs_mark_sb_dirty(sb);
+
+       mutex_unlock(&UFS_SB(sb)->s_lock);
+       UFSD("EXIT\n");
+       return;
+
+failed:
+       mutex_unlock(&UFS_SB(sb)->s_lock);
+       UFSD("EXIT (FAILED)\n");
+       return;
+}
+
+/*
+ * Free 'count' fragments from fragment number 'fragment' (free whole blocks)
+ */
+void ufs_free_blocks(struct inode *inode, u64 fragment, unsigned count)
+{
+       struct super_block * sb;
+       struct ufs_sb_private_info * uspi;
+       struct ufs_cg_private_info * ucpi;
+       struct ufs_cylinder_group * ucg;
+       unsigned overflow, cgno, bit, end_bit, i;
+       u64 blkno;
+       
+       sb = inode->i_sb;
+       uspi = UFS_SB(sb)->s_uspi;
+
+       UFSD("ENTER, fragment %llu, count %u\n",
+            (unsigned long long)fragment, count);
+       
+       if ((fragment & uspi->s_fpbmask) || (count & uspi->s_fpbmask)) {
+               ufs_error (sb, "ufs_free_blocks", "internal error, "
+                          "fragment %llu, count %u\n",
+                          (unsigned long long)fragment, count);
+               goto failed;
+       }
+
+       mutex_lock(&UFS_SB(sb)->s_lock);
+       
+do_more:
+       overflow = 0;
+       cgno = ufs_dtog(uspi, fragment);
+       bit = ufs_dtogd(uspi, fragment);
+       if (cgno >= uspi->s_ncg) {
+               ufs_panic (sb, "ufs_free_blocks", "freeing blocks are outside device");
+               goto failed_unlock;
+       }
+       end_bit = bit + count;
+       if (end_bit > uspi->s_fpg) {
+               overflow = bit + count - uspi->s_fpg;
+               count -= overflow;
+               end_bit -= overflow;
+       }
+
+       ucpi = ufs_load_cylinder (sb, cgno);
+       if (!ucpi) 
+               goto failed_unlock;
+       ucg = ubh_get_ucg (UCPI_UBH(ucpi));
+       if (!ufs_cg_chkmagic(sb, ucg)) {
+               ufs_panic (sb, "ufs_free_blocks", "internal error, bad magic number on cg %u", cgno);
+               goto failed_unlock;
+       }
+
+       for (i = bit; i < end_bit; i += uspi->s_fpb) {
+               blkno = ufs_fragstoblks(i);
+               if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno)) {
+                       ufs_error(sb, "ufs_free_blocks", "freeing free fragment");
+               }
+               ubh_setblock(UCPI_UBH(ucpi), ucpi->c_freeoff, blkno);
+               if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
+                       ufs_clusteracct (sb, ucpi, blkno, 1);
+
+               fs32_add(sb, &ucg->cg_cs.cs_nbfree, 1);
+               uspi->cs_total.cs_nbfree++;
+               fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nbfree, 1);
+
+               if (uspi->fs_magic != UFS2_MAGIC) {
+                       unsigned cylno = ufs_cbtocylno(i);
+
+                       fs16_add(sb, &ubh_cg_blks(ucpi, cylno,
+                                                 ufs_cbtorpos(i)), 1);
+                       fs32_add(sb, &ubh_cg_blktot(ucpi, cylno), 1);
+               }
+       }
+
+       ubh_mark_buffer_dirty (USPI_UBH(uspi));
+       ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
+       if (sb->s_flags & MS_SYNCHRONOUS)
+               ubh_sync_block(UCPI_UBH(ucpi));
+
+       if (overflow) {
+               fragment += count;
+               count = overflow;
+               goto do_more;
+       }
+
+       ufs_mark_sb_dirty(sb);
+       mutex_unlock(&UFS_SB(sb)->s_lock);
+       UFSD("EXIT\n");
+       return;
+
+failed_unlock:
+       mutex_unlock(&UFS_SB(sb)->s_lock);
+failed:
+       UFSD("EXIT (FAILED)\n");
+       return;
+}
+
+/*
+ * Modify inode page cache in such way:
+ * have - blocks with b_blocknr equal to oldb...oldb+count-1
+ * get - blocks with b_blocknr equal to newb...newb+count-1
+ * also we suppose that oldb...oldb+count-1 blocks
+ * situated at the end of file.
+ *
+ * We can come here from ufs_writepage or ufs_prepare_write,
+ * locked_page is argument of these functions, so we already lock it.
+ */
+static void ufs_change_blocknr(struct inode *inode, sector_t beg,
+                              unsigned int count, sector_t oldb,
+                              sector_t newb, struct page *locked_page)
+{
+       const unsigned blks_per_page =
+               1 << (PAGE_CACHE_SHIFT - inode->i_blkbits);
+       const unsigned mask = blks_per_page - 1;
+       struct address_space * const mapping = inode->i_mapping;
+       pgoff_t index, cur_index, last_index;
+       unsigned pos, j, lblock;
+       sector_t end, i;
+       struct page *page;
+       struct buffer_head *head, *bh;
+
+       UFSD("ENTER, ino %lu, count %u, oldb %llu, newb %llu\n",
+             inode->i_ino, count,
+            (unsigned long long)oldb, (unsigned long long)newb);
+
+       BUG_ON(!locked_page);
+       BUG_ON(!PageLocked(locked_page));
+
+       cur_index = locked_page->index;
+       end = count + beg;
+       last_index = end >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
+       for (i = beg; i < end; i = (i | mask) + 1) {
+               index = i >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
+
+               if (likely(cur_index != index)) {
+                       page = ufs_get_locked_page(mapping, index);
+                       if (!page)/* it was truncated */
+                               continue;
+                       if (IS_ERR(page)) {/* or EIO */
+                               ufs_error(inode->i_sb, __func__,
+                                         "read of page %llu failed\n",
+                                         (unsigned long long)index);
+                               continue;
+                       }
+               } else
+                       page = locked_page;
+
+               head = page_buffers(page);
+               bh = head;
+               pos = i & mask;
+               for (j = 0; j < pos; ++j)
+                       bh = bh->b_this_page;
+
+
+               if (unlikely(index == last_index))
+                       lblock = end & mask;
+               else
+                       lblock = blks_per_page;
+
+               do {
+                       if (j >= lblock)
+                               break;
+                       pos = (i - beg) + j;
+
+                       if (!buffer_mapped(bh))
+                                       map_bh(bh, inode->i_sb, oldb + pos);
+                       if (!buffer_uptodate(bh)) {
+                               ll_rw_block(READ, 1, &bh);
+                               wait_on_buffer(bh);
+                               if (!buffer_uptodate(bh)) {
+                                       ufs_error(inode->i_sb, __func__,
+                                                 "read of block failed\n");
+                                       break;
+                               }
+                       }
+
+                       UFSD(" change from %llu to %llu, pos %u\n",
+                            (unsigned long long)(pos + oldb),
+                            (unsigned long long)(pos + newb), pos);
+
+                       bh->b_blocknr = newb + pos;
+                       unmap_underlying_metadata(bh->b_bdev,
+                                                 bh->b_blocknr);
+                       mark_buffer_dirty(bh);
+                       ++j;
+                       bh = bh->b_this_page;
+               } while (bh != head);
+
+               if (likely(cur_index != index))
+                       ufs_put_locked_page(page);
+       }
+       UFSD("EXIT\n");
+}
+
+static void ufs_clear_frags(struct inode *inode, sector_t beg, unsigned int n,
+                           int sync)
+{
+       struct buffer_head *bh;
+       sector_t end = beg + n;
+
+       for (; beg < end; ++beg) {
+               bh = sb_getblk(inode->i_sb, beg);
+               lock_buffer(bh);
+               memset(bh->b_data, 0, inode->i_sb->s_blocksize);
+               set_buffer_uptodate(bh);
+               mark_buffer_dirty(bh);
+               unlock_buffer(bh);
+               if (IS_SYNC(inode) || sync)
+                       sync_dirty_buffer(bh);
+               brelse(bh);
+       }
+}
+
+u64 ufs_new_fragments(struct inode *inode, void *p, u64 fragment,
+                          u64 goal, unsigned count, int *err,
+                          struct page *locked_page)
+{
+       struct super_block * sb;
+       struct ufs_sb_private_info * uspi;
+       struct ufs_super_block_first * usb1;
+       unsigned cgno, oldcount, newcount;
+       u64 tmp, request, result;
+       
+       UFSD("ENTER, ino %lu, fragment %llu, goal %llu, count %u\n",
+            inode->i_ino, (unsigned long long)fragment,
+            (unsigned long long)goal, count);
+       
+       sb = inode->i_sb;
+       uspi = UFS_SB(sb)->s_uspi;
+       usb1 = ubh_get_usb_first(uspi);
+       *err = -ENOSPC;
+
+       mutex_lock(&UFS_SB(sb)->s_lock);
+       tmp = ufs_data_ptr_to_cpu(sb, p);
+
+       if (count + ufs_fragnum(fragment) > uspi->s_fpb) {
+               ufs_warning(sb, "ufs_new_fragments", "internal warning"
+                           " fragment %llu, count %u",
+                           (unsigned long long)fragment, count);
+               count = uspi->s_fpb - ufs_fragnum(fragment); 
+       }
+       oldcount = ufs_fragnum (fragment);
+       newcount = oldcount + count;
+
+       /*
+        * Somebody else has just allocated our fragments
+        */
+       if (oldcount) {
+               if (!tmp) {
+                       ufs_error(sb, "ufs_new_fragments", "internal error, "
+                                 "fragment %llu, tmp %llu\n",
+                                 (unsigned long long)fragment,
+                                 (unsigned long long)tmp);
+                       mutex_unlock(&UFS_SB(sb)->s_lock);
+                       return INVBLOCK;
+               }
+               if (fragment < UFS_I(inode)->i_lastfrag) {
+                       UFSD("EXIT (ALREADY ALLOCATED)\n");
+                       mutex_unlock(&UFS_SB(sb)->s_lock);
+                       return 0;
+               }
+       }
+       else {
+               if (tmp) {
+                       UFSD("EXIT (ALREADY ALLOCATED)\n");
+                       mutex_unlock(&UFS_SB(sb)->s_lock);
+                       return 0;
+               }
+       }
+
+       /*
+        * There is not enough space for user on the device
+        */
+       if (!capable(CAP_SYS_RESOURCE) && ufs_freespace(uspi, UFS_MINFREE) <= 0) {
+               mutex_unlock(&UFS_SB(sb)->s_lock);
+               UFSD("EXIT (FAILED)\n");
+               return 0;
+       }
+
+       if (goal >= uspi->s_size) 
+               goal = 0;
+       if (goal == 0) 
+               cgno = ufs_inotocg (inode->i_ino);
+       else
+               cgno = ufs_dtog(uspi, goal);
+        
+       /*
+        * allocate new fragment
+        */
+       if (oldcount == 0) {
+               result = ufs_alloc_fragments (inode, cgno, goal, count, err);
+               if (result) {
+                       ufs_cpu_to_data_ptr(sb, p, result);
+                       *err = 0;
+                       UFS_I(inode)->i_lastfrag =
+                               max(UFS_I(inode)->i_lastfrag, fragment + count);
+                       ufs_clear_frags(inode, result + oldcount,
+                                       newcount - oldcount, locked_page != NULL);
+               }
+               mutex_unlock(&UFS_SB(sb)->s_lock);
+               UFSD("EXIT, result %llu\n", (unsigned long long)result);
+               return result;
+       }
+
+       /*
+        * resize block
+        */
+       result = ufs_add_fragments(inode, tmp, oldcount, newcount);
+       if (result) {
+               *err = 0;
+               UFS_I(inode)->i_lastfrag = max(UFS_I(inode)->i_lastfrag,
+                                               fragment + count);
+               ufs_clear_frags(inode, result + oldcount, newcount - oldcount,
+                               locked_page != NULL);
+               mutex_unlock(&UFS_SB(sb)->s_lock);
+               UFSD("EXIT, result %llu\n", (unsigned long long)result);
+               return result;
+       }
+
+       /*
+        * allocate new block and move data
+        */
+       switch (fs32_to_cpu(sb, usb1->fs_optim)) {
+           case UFS_OPTSPACE:
+               request = newcount;
+               if (uspi->s_minfree < 5 || uspi->cs_total.cs_nffree
+                   > uspi->s_dsize * uspi->s_minfree / (2 * 100))
+                       break;
+               usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
+               break;
+           default:
+               usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
+       
+           case UFS_OPTTIME:
+               request = uspi->s_fpb;
+               if (uspi->cs_total.cs_nffree < uspi->s_dsize *
+                   (uspi->s_minfree - 2) / 100)
+                       break;
+               usb1->fs_optim = cpu_to_fs32(sb, UFS_OPTTIME);
+               break;
+       }
+       result = ufs_alloc_fragments (inode, cgno, goal, request, err);
+       if (result) {
+               ufs_clear_frags(inode, result + oldcount, newcount - oldcount,
+                               locked_page != NULL);
+               ufs_change_blocknr(inode, fragment - oldcount, oldcount,
+                                  uspi->s_sbbase + tmp,
+                                  uspi->s_sbbase + result, locked_page);
+               ufs_cpu_to_data_ptr(sb, p, result);
+               *err = 0;
+               UFS_I(inode)->i_lastfrag = max(UFS_I(inode)->i_lastfrag,
+                                               fragment + count);
+               mutex_unlock(&UFS_SB(sb)->s_lock);
+               if (newcount < request)
+                       ufs_free_fragments (inode, result + newcount, request - newcount);
+               ufs_free_fragments (inode, tmp, oldcount);
+               UFSD("EXIT, result %llu\n", (unsigned long long)result);
+               return result;
+       }
+
+       mutex_unlock(&UFS_SB(sb)->s_lock);
+       UFSD("EXIT (FAILED)\n");
+       return 0;
+}              
+
+static u64 ufs_add_fragments(struct inode *inode, u64 fragment,
+                            unsigned oldcount, unsigned newcount)
+{
+       struct super_block * sb;
+       struct ufs_sb_private_info * uspi;
+       struct ufs_cg_private_info * ucpi;
+       struct ufs_cylinder_group * ucg;
+       unsigned cgno, fragno, fragoff, count, fragsize, i;
+       
+       UFSD("ENTER, fragment %llu, oldcount %u, newcount %u\n",
+            (unsigned long long)fragment, oldcount, newcount);
+       
+       sb = inode->i_sb;
+       uspi = UFS_SB(sb)->s_uspi;
+       count = newcount - oldcount;
+       
+       cgno = ufs_dtog(uspi, fragment);
+       if (fs32_to_cpu(sb, UFS_SB(sb)->fs_cs(cgno).cs_nffree) < count)
+               return 0;
+       if ((ufs_fragnum (fragment) + newcount) > uspi->s_fpb)
+               return 0;
+       ucpi = ufs_load_cylinder (sb, cgno);
+       if (!ucpi)
+               return 0;
+       ucg = ubh_get_ucg (UCPI_UBH(ucpi));
+       if (!ufs_cg_chkmagic(sb, ucg)) {
+               ufs_panic (sb, "ufs_add_fragments",
+                       "internal error, bad magic number on cg %u", cgno);
+               return 0;
+       }
+
+       fragno = ufs_dtogd(uspi, fragment);
+       fragoff = ufs_fragnum (fragno);
+       for (i = oldcount; i < newcount; i++)
+               if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, fragno + i))
+                       return 0;
+       /*
+        * Block can be extended
+        */
+       ucg->cg_time = cpu_to_fs32(sb, get_seconds());
+       for (i = newcount; i < (uspi->s_fpb - fragoff); i++)
+               if (ubh_isclr (UCPI_UBH(ucpi), ucpi->c_freeoff, fragno + i))
+                       break;
+       fragsize = i - oldcount;
+       if (!fs32_to_cpu(sb, ucg->cg_frsum[fragsize]))
+               ufs_panic (sb, "ufs_add_fragments",
+                       "internal error or corrupted bitmap on cg %u", cgno);
+       fs32_sub(sb, &ucg->cg_frsum[fragsize], 1);
+       if (fragsize != count)
+               fs32_add(sb, &ucg->cg_frsum[fragsize - count], 1);
+       for (i = oldcount; i < newcount; i++)
+               ubh_clrbit (UCPI_UBH(ucpi), ucpi->c_freeoff, fragno + i);
+
+       fs32_sub(sb, &ucg->cg_cs.cs_nffree, count);
+       fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
+       uspi->cs_total.cs_nffree -= count;
+       
+       ubh_mark_buffer_dirty (USPI_UBH(uspi));
+       ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
+       if (sb->s_flags & MS_SYNCHRONOUS)
+               ubh_sync_block(UCPI_UBH(ucpi));
+       ufs_mark_sb_dirty(sb);
+
+       UFSD("EXIT, fragment %llu\n", (unsigned long long)fragment);
+       
+       return fragment;
+}
+
+#define UFS_TEST_FREE_SPACE_CG \
+       ucg = (struct ufs_cylinder_group *) UFS_SB(sb)->s_ucg[cgno]->b_data; \
+       if (fs32_to_cpu(sb, ucg->cg_cs.cs_nbfree)) \
+               goto cg_found; \
+       for (k = count; k < uspi->s_fpb; k++) \
+               if (fs32_to_cpu(sb, ucg->cg_frsum[k])) \
+                       goto cg_found; 
+
+static u64 ufs_alloc_fragments(struct inode *inode, unsigned cgno,
+                              u64 goal, unsigned count, int *err)
+{
+       struct super_block * sb;
+       struct ufs_sb_private_info * uspi;
+       struct ufs_cg_private_info * ucpi;
+       struct ufs_cylinder_group * ucg;
+       unsigned oldcg, i, j, k, allocsize;
+       u64 result;
+       
+       UFSD("ENTER, ino %lu, cgno %u, goal %llu, count %u\n",
+            inode->i_ino, cgno, (unsigned long long)goal, count);
+
+       sb = inode->i_sb;
+       uspi = UFS_SB(sb)->s_uspi;
+       oldcg = cgno;
+       
+       /*
+        * 1. searching on preferred cylinder group
+        */
+       UFS_TEST_FREE_SPACE_CG
+
+       /*
+        * 2. quadratic rehash
+        */
+       for (j = 1; j < uspi->s_ncg; j *= 2) {
+               cgno += j;
+               if (cgno >= uspi->s_ncg) 
+                       cgno -= uspi->s_ncg;
+               UFS_TEST_FREE_SPACE_CG
+       }
+
+       /*
+        * 3. brute force search
+        * We start at i = 2 ( 0 is checked at 1.step, 1 at 2.step )
+        */
+       cgno = (oldcg + 1) % uspi->s_ncg;
+       for (j = 2; j < uspi->s_ncg; j++) {
+               cgno++;
+               if (cgno >= uspi->s_ncg)
+                       cgno = 0;
+               UFS_TEST_FREE_SPACE_CG
+       }
+       
+       UFSD("EXIT (FAILED)\n");
+       return 0;
+
+cg_found:
+       ucpi = ufs_load_cylinder (sb, cgno);
+       if (!ucpi)
+               return 0;
+       ucg = ubh_get_ucg (UCPI_UBH(ucpi));
+       if (!ufs_cg_chkmagic(sb, ucg)) 
+               ufs_panic (sb, "ufs_alloc_fragments",
+                       "internal error, bad magic number on cg %u", cgno);
+       ucg->cg_time = cpu_to_fs32(sb, get_seconds());
+
+       if (count == uspi->s_fpb) {
+               result = ufs_alloccg_block (inode, ucpi, goal, err);
+               if (result == INVBLOCK)
+                       return 0;
+               goto succed;
+       }
+
+       for (allocsize = count; allocsize < uspi->s_fpb; allocsize++)
+               if (fs32_to_cpu(sb, ucg->cg_frsum[allocsize]) != 0)
+                       break;
+       
+       if (allocsize == uspi->s_fpb) {
+               result = ufs_alloccg_block (inode, ucpi, goal, err);
+               if (result == INVBLOCK)
+                       return 0;
+               goal = ufs_dtogd(uspi, result);
+               for (i = count; i < uspi->s_fpb; i++)
+                       ubh_setbit (UCPI_UBH(ucpi), ucpi->c_freeoff, goal + i);
+               i = uspi->s_fpb - count;
+
+               fs32_add(sb, &ucg->cg_cs.cs_nffree, i);
+               uspi->cs_total.cs_nffree += i;
+               fs32_add(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, i);
+               fs32_add(sb, &ucg->cg_frsum[i], 1);
+               goto succed;
+       }
+
+       result = ufs_bitmap_search (sb, ucpi, goal, allocsize);
+       if (result == INVBLOCK)
+               return 0;
+       for (i = 0; i < count; i++)
+               ubh_clrbit (UCPI_UBH(ucpi), ucpi->c_freeoff, result + i);
+       
+       fs32_sub(sb, &ucg->cg_cs.cs_nffree, count);
+       uspi->cs_total.cs_nffree -= count;
+       fs32_sub(sb, &UFS_SB(sb)->fs_cs(cgno).cs_nffree, count);
+       fs32_sub(sb, &ucg->cg_frsum[allocsize], 1);
+
+       if (count != allocsize)
+               fs32_add(sb, &ucg->cg_frsum[allocsize - count], 1);
+
+succed:
+       ubh_mark_buffer_dirty (USPI_UBH(uspi));
+       ubh_mark_buffer_dirty (UCPI_UBH(ucpi));
+       if (sb->s_flags & MS_SYNCHRONOUS)
+               ubh_sync_block(UCPI_UBH(ucpi));
+       ufs_mark_sb_dirty(sb);
+
+       result += cgno * uspi->s_fpg;
+       UFSD("EXIT3, result %llu\n", (unsigned long long)result);
+       return result;
+}
+
+static u64 ufs_alloccg_block(struct inode *inode,
+                            struct ufs_cg_private_info *ucpi,
+                            u64 goal, int *err)
+{
+       struct super_block * sb;
+       struct ufs_sb_private_info * uspi;
+       struct ufs_cylinder_group * ucg;
+       u64 result, blkno;
+
+       UFSD("ENTER, goal %llu\n", (unsigned long long)goal);
+
+       sb = inode->i_sb;
+       uspi = UFS_SB(sb)->s_uspi;
+       ucg = ubh_get_ucg(UCPI_UBH(ucpi));
+
+       if (goal == 0) {
+               goal = ucpi->c_rotor;
+               goto norot;
+       }
+       goal = ufs_blknum (goal);
+       goal = ufs_dtogd(uspi, goal);
+       
+       /*
+        * If the requested block is available, use it.
+        */
+       if (ubh_isblockset(UCPI_UBH(ucpi), ucpi->c_freeoff, ufs_fragstoblks(goal))) {
+               result = goal;
+               goto gotit;
+       }
+       
+norot: 
+       result = ufs_bitmap_search (sb, ucpi, goal, uspi->s_fpb);
+       if (result == INVBLOCK)
+               return INVBLOCK;
+       ucpi->c_rotor = result;
+gotit:
+       blkno = ufs_fragstoblks(result);
+       ubh_clrblock (UCPI_UBH(ucpi), ucpi->c_freeoff, blkno);
+       if ((UFS_SB(sb)->s_flags & UFS_CG_MASK) == UFS_CG_44BSD)
+               ufs_clusteracct (sb, ucpi, blkno, -1);
+
+       fs32_sub(sb, &ucg->cg_cs.cs_nbfree, 1);
+       uspi->cs_total.cs_nbfree--;
+       fs32_sub(sb, &UFS_SB(sb)->fs_cs(ucpi->c_cgx).cs_nbfree, 1);
+
+       if (uspi->fs_magic != UFS2_MAGIC) {
+               unsigned cylno = ufs_cbtocylno((unsigned)result);
+
+               fs16_sub(sb, &ubh_cg_blks(ucpi, cylno,
+                                         ufs_cbtorpos((unsigned)result)), 1);
+               fs32_sub(sb, &ubh_cg_blktot(ucpi, cylno), 1);
+       }
+       
+       UFSD("EXIT, result %llu\n", (unsigned long long)result);
+
+       return result;
+}
+
+static unsigned ubh_scanc(struct ufs_sb_private_info *uspi,
+                         struct ufs_buffer_head *ubh,
+                         unsigned begin, unsigned size,
+                         unsigned char *table, unsigned char mask)
+{
+       unsigned rest, offset;
+       unsigned char *cp;
+       
+
+       offset = begin & ~uspi->s_fmask;
+       begin >>= uspi->s_fshift;
+       for (;;) {
+               if ((offset + size) < uspi->s_fsize)
+                       rest = size;
+               else
+                       rest = uspi->s_fsize - offset;
+               size -= rest;
+               cp = ubh->bh[begin]->b_data + offset;
+               while ((table[*cp++] & mask) == 0 && --rest)
+                       ;
+               if (rest || !size)
+                       break;
+               begin++;
+               offset = 0;
+       }
+       return (size + rest);
+}
+
+/*
+ * Find a block of the specified size in the specified cylinder group.
+ * @sp: pointer to super block
+ * @ucpi: pointer to cylinder group info
+ * @goal: near which block we want find new one
+ * @count: specified size
+ */
+static u64 ufs_bitmap_search(struct super_block *sb,
+                            struct ufs_cg_private_info *ucpi,
+                            u64 goal, unsigned count)
+{
+       /*
+        * Bit patterns for identifying fragments in the block map
+        * used as ((map & mask_arr) == want_arr)
+        */
+       static const int mask_arr[9] = {
+               0x3, 0x7, 0xf, 0x1f, 0x3f, 0x7f, 0xff, 0x1ff, 0x3ff
+       };
+       static const int want_arr[9] = {
+               0x0, 0x2, 0x6, 0xe, 0x1e, 0x3e, 0x7e, 0xfe, 0x1fe
+       };
+       struct ufs_sb_private_info *uspi = UFS_SB(sb)->s_uspi;
+       unsigned start, length, loc;
+       unsigned pos, want, blockmap, mask, end;
+       u64 result;
+
+       UFSD("ENTER, cg %u, goal %llu, count %u\n", ucpi->c_cgx,
+            (unsigned long long)goal, count);
+
+       if (goal)
+               start = ufs_dtogd(uspi, goal) >> 3;
+       else
+               start = ucpi->c_frotor >> 3;
+               
+       length = ((uspi->s_fpg + 7) >> 3) - start;
+       loc = ubh_scanc(uspi, UCPI_UBH(ucpi), ucpi->c_freeoff + start, length,
+               (uspi->s_fpb == 8) ? ufs_fragtable_8fpb : ufs_fragtable_other,
+               1 << (count - 1 + (uspi->s_fpb & 7))); 
+       if (loc == 0) {
+               length = start + 1;
+               loc = ubh_scanc(uspi, UCPI_UBH(ucpi), ucpi->c_freeoff, length,
+                               (uspi->s_fpb == 8) ? ufs_fragtable_8fpb :
+                               ufs_fragtable_other,
+                               1 << (count - 1 + (uspi->s_fpb & 7)));
+               if (loc == 0) {
+                       ufs_error(sb, "ufs_bitmap_search",
+                                 "bitmap corrupted on cg %u, start %u,"
+                                 " length %u, count %u, freeoff %u\n",
+                                 ucpi->c_cgx, start, length, count,
+                                 ucpi->c_freeoff);
+                       return INVBLOCK;
+               }
+               start = 0;
+       }
+       result = (start + length - loc) << 3;
+       ucpi->c_frotor = result;
+
+       /*
+        * found the byte in the map
+        */
+
+       for (end = result + 8; result < end; result += uspi->s_fpb) {
+               blockmap = ubh_blkmap(UCPI_UBH(ucpi), ucpi->c_freeoff, result);
+               blockmap <<= 1;
+               mask = mask_arr[count];
+               want = want_arr[count];
+               for (pos = 0; pos <= uspi->s_fpb - count; pos++) {
+                       if ((blockmap & mask) == want) {
+                               UFSD("EXIT, result %llu\n",
+                                    (unsigned long long)result);
+                               return result + pos;
+                       }
+                       mask <<= 1;
+                       want <<= 1;
+               }
+       }
+
+       ufs_error(sb, "ufs_bitmap_search", "block not in map on cg %u\n",
+                 ucpi->c_cgx);
+       UFSD("EXIT (FAILED)\n");
+       return INVBLOCK;
+}
+
+static void ufs_clusteracct(struct super_block * sb,
+       struct ufs_cg_private_info * ucpi, unsigned blkno, int cnt)
+{
+       struct ufs_sb_private_info * uspi;
+       int i, start, end, forw, back;
+       
+       uspi = UFS_SB(sb)->s_uspi;
+       if (uspi->s_contigsumsize <= 0)
+               return;
+
+       if (cnt > 0)
+               ubh_setbit(UCPI_UBH(ucpi), ucpi->c_clusteroff, blkno);
+       else
+               ubh_clrbit(UCPI_UBH(ucpi), ucpi->c_clusteroff, blkno);
+
+       /*
+        * Find the size of the cluster going forward.
+        */
+       start = blkno + 1;
+       end = start + uspi->s_contigsumsize;
+       if ( end >= ucpi->c_nclusterblks)
+               end = ucpi->c_nclusterblks;
+       i = ubh_find_next_zero_bit (UCPI_UBH(ucpi), ucpi->c_clusteroff, end, start);
+       if (i > end)
+               i = end;
+       forw = i - start;
+       
+       /*
+        * Find the size of the cluster going backward.
+        */
+       start = blkno - 1;
+       end = start - uspi->s_contigsumsize;
+       if (end < 0 ) 
+               end = -1;
+       i = ubh_find_last_zero_bit (UCPI_UBH(ucpi), ucpi->c_clusteroff, start, end);
+       if ( i < end) 
+               i = end;
+       back = start - i;
+       
+       /*
+        * Account for old cluster and the possibly new forward and
+        * back clusters.
+        */
+       i = back + forw + 1;
+       if (i > uspi->s_contigsumsize)
+               i = uspi->s_contigsumsize;
+       fs32_add(sb, (__fs32*)ubh_get_addr(UCPI_UBH(ucpi), ucpi->c_clustersumoff + (i << 2)), cnt);
+       if (back > 0)
+               fs32_sub(sb, (__fs32*)ubh_get_addr(UCPI_UBH(ucpi), ucpi->c_clustersumoff + (back << 2)), cnt);
+       if (forw > 0)
+               fs32_sub(sb, (__fs32*)ubh_get_addr(UCPI_UBH(ucpi), ucpi->c_clustersumoff + (forw << 2)), cnt);
+}
+
+
+static unsigned char ufs_fragtable_8fpb[] = {
+       0x00, 0x01, 0x01, 0x02, 0x01, 0x01, 0x02, 0x04, 0x01, 0x01, 0x01, 0x03, 0x02, 0x03, 0x04, 0x08,
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x02, 0x03, 0x03, 0x02, 0x04, 0x05, 0x08, 0x10,
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09,
+       0x02, 0x03, 0x03, 0x02, 0x03, 0x03, 0x02, 0x06, 0x04, 0x05, 0x05, 0x06, 0x08, 0x09, 0x10, 0x20,
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09, 
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x03, 0x03, 0x03, 0x03, 0x05, 0x05, 0x09, 0x11,
+       0x02, 0x03, 0x03, 0x02, 0x03, 0x03, 0x02, 0x06, 0x03, 0x03, 0x03, 0x03, 0x02, 0x03, 0x06, 0x0A,
+       0x04, 0x05, 0x05, 0x06, 0x05, 0x05, 0x06, 0x04, 0x08, 0x09, 0x09, 0x0A, 0x10, 0x11, 0x20, 0x40,
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09,
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x03, 0x03, 0x03, 0x03, 0x05, 0x05, 0x09, 0x11,
+       0x01, 0x01, 0x01, 0x03, 0x01, 0x01, 0x03, 0x05, 0x01, 0x01, 0x01, 0x03, 0x03, 0x03, 0x05, 0x09,
+       0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x07, 0x05, 0x05, 0x05, 0x07, 0x09, 0x09, 0x11, 0x21,
+       0x02, 0x03, 0x03, 0x02, 0x03, 0x03, 0x02, 0x06, 0x03, 0x03, 0x03, 0x03, 0x02, 0x03, 0x06, 0x0A,
+       0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x03, 0x07, 0x02, 0x03, 0x03, 0x02, 0x06, 0x07, 0x0A, 0x12,
+       0x04, 0x05, 0x05, 0x06, 0x05, 0x05, 0x06, 0x04, 0x05, 0x05, 0x05, 0x07, 0x06, 0x07, 0x04, 0x0C,
+       0x08, 0x09, 0x09, 0x0A, 0x09, 0x09, 0x0A, 0x0C, 0x10, 0x11, 0x11, 0x12, 0x20, 0x21, 0x40, 0x80,
+};
+
+static unsigned char ufs_fragtable_other[] = {
+       0x00, 0x16, 0x16, 0x2A, 0x16, 0x16, 0x26, 0x4E, 0x16, 0x16, 0x16, 0x3E, 0x2A, 0x3E, 0x4E, 0x8A,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x2A, 0x3E, 0x3E, 0x2A, 0x3E, 0x3E, 0x2E, 0x6E, 0x3E, 0x3E, 0x3E, 0x3E, 0x2A, 0x3E, 0x6E, 0xAA,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x26, 0x36, 0x36, 0x2E, 0x36, 0x36, 0x26, 0x6E, 0x36, 0x36, 0x36, 0x3E, 0x2E, 0x3E, 0x6E, 0xAE,
+       0x4E, 0x5E, 0x5E, 0x6E, 0x5E, 0x5E, 0x6E, 0x4E, 0x5E, 0x5E, 0x5E, 0x7E, 0x6E, 0x7E, 0x4E, 0xCE,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x16, 0x16, 0x16, 0x3E, 0x16, 0x16, 0x36, 0x5E, 0x16, 0x16, 0x16, 0x3E, 0x3E, 0x3E, 0x5E, 0x9E,
+       0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0xBE,
+       0x2A, 0x3E, 0x3E, 0x2A, 0x3E, 0x3E, 0x2E, 0x6E, 0x3E, 0x3E, 0x3E, 0x3E, 0x2A, 0x3E, 0x6E, 0xAA,
+       0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x3E, 0x7E, 0xBE,
+       0x4E, 0x5E, 0x5E, 0x6E, 0x5E, 0x5E, 0x6E, 0x4E, 0x5E, 0x5E, 0x5E, 0x7E, 0x6E, 0x7E, 0x4E, 0xCE,
+       0x8A, 0x9E, 0x9E, 0xAA, 0x9E, 0x9E, 0xAE, 0xCE, 0x9E, 0x9E, 0x9E, 0xBE, 0xAA, 0xBE, 0xCE, 0x8A,
+};