hpfs: optimize quad buffer loading

HPFS needs to load 4 consecutive 512-byte sectors when accessing the
directory nodes or bitmaps.  We can't switch to 2048-byte block size
because files are allocated in the units of 512-byte sectors.

Previously, the driver would allocate a 2048-byte area using kmalloc,
copy the data from four buffers to this area and eventually copy them
back if they were modified.

In the current implementation of the buffer cache, buffers are allocated
in the pagecache.  That means that 4 consecutive 512-byte buffers are
stored in consecutive areas in the kernel address space.  So, we don't
need to allocate extra memory and copy the content of the buffers there.

This patch optimizes the code to avoid copying the buffers.  It checks
if the four buffers are stored in contiguous memory - if they are not,
it falls back to allocating a 2048-byte area and copying data there.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/fs/hpfs/buffer.c b/fs/hpfs/buffer.c
index 4d0a1afa058c8a830f00e36443a2f9fed03dd0a3..139ef1684d074bba059ff0c401f65273420e7c6a 100644 (file)
--- a/fs/hpfs/buffer.c
+++ b/fs/hpfs/buffer.c
@@ -86,7 +86,6 @@ void *hpfs_get_sector(struct super_block *s, unsigned secno, struct buffer_head
 void *hpfs_map_4sectors(struct super_block *s, unsigned secno, struct quad_buffer_head *qbh,
                   int ahead)
 {
-       struct buffer_head *bh;
        char *data;
 
        hpfs_lock_assert(s);
@@ -100,34 +99,32 @@ void *hpfs_map_4sectors(struct super_block *s, unsigned secno, struct quad_buffe
 
        hpfs_prefetch_sectors(s, secno, 4 + ahead);
 
+       if (!(qbh->bh[0] = sb_bread(s, secno + 0))) goto bail0;
+       if (!(qbh->bh[1] = sb_bread(s, secno + 1))) goto bail1;
+       if (!(qbh->bh[2] = sb_bread(s, secno + 2))) goto bail2;
+       if (!(qbh->bh[3] = sb_bread(s, secno + 3))) goto bail3;
+
+       if (likely(qbh->bh[1]->b_data == qbh->bh[0]->b_data + 1 * 512) &&
+           likely(qbh->bh[2]->b_data == qbh->bh[0]->b_data + 2 * 512) &&
+           likely(qbh->bh[3]->b_data == qbh->bh[0]->b_data + 3 * 512)) {
+               return qbh->data = qbh->bh[0]->b_data;
+       }
+
        qbh->data = data = kmalloc(2048, GFP_NOFS);
        if (!data) {
                printk("HPFS: hpfs_map_4sectors: out of memory\n");
-               goto bail;
+               goto bail4;
        }
 
-       qbh->bh[0] = bh = sb_bread(s, secno);
-       if (!bh)
-               goto bail0;
-       memcpy(data, bh->b_data, 512);
-
-       qbh->bh[1] = bh = sb_bread(s, secno + 1);
-       if (!bh)
-               goto bail1;
-       memcpy(data + 512, bh->b_data, 512);
-
-       qbh->bh[2] = bh = sb_bread(s, secno + 2);
-       if (!bh)
-               goto bail2;
-       memcpy(data + 2 * 512, bh->b_data, 512);
-
-       qbh->bh[3] = bh = sb_bread(s, secno + 3);
-       if (!bh)
-               goto bail3;
-       memcpy(data + 3 * 512, bh->b_data, 512);
+       memcpy(data + 0 * 512, qbh->bh[0]->b_data, 512);
+       memcpy(data + 1 * 512, qbh->bh[1]->b_data, 512);
+       memcpy(data + 2 * 512, qbh->bh[2]->b_data, 512);
+       memcpy(data + 3 * 512, qbh->bh[3]->b_data, 512);
 
        return data;
 
+ bail4:
+       brelse(qbh->bh[3]);
  bail3:
        brelse(qbh->bh[2]);
  bail2:
@@ -135,9 +132,6 @@ void *hpfs_map_4sectors(struct super_block *s, unsigned secno, struct quad_buffe
  bail1:
        brelse(qbh->bh[0]);
  bail0:
-       kfree(data);
-       printk("HPFS: hpfs_map_4sectors: read error\n");
- bail:
        return NULL;
 }
 
@@ -155,44 +149,54 @@ void *hpfs_get_4sectors(struct super_block *s, unsigned secno,
                return NULL;
        }
 
-       /*return hpfs_map_4sectors(s, secno, qbh, 0);*/
+       if (!hpfs_get_sector(s, secno + 0, &qbh->bh[0])) goto bail0;
+       if (!hpfs_get_sector(s, secno + 1, &qbh->bh[1])) goto bail1;
+       if (!hpfs_get_sector(s, secno + 2, &qbh->bh[2])) goto bail2;
+       if (!hpfs_get_sector(s, secno + 3, &qbh->bh[3])) goto bail3;
+
+       if (likely(qbh->bh[1]->b_data == qbh->bh[0]->b_data + 1 * 512) &&
+           likely(qbh->bh[2]->b_data == qbh->bh[0]->b_data + 2 * 512) &&
+           likely(qbh->bh[3]->b_data == qbh->bh[0]->b_data + 3 * 512)) {
+               return qbh->data = qbh->bh[0]->b_data;
+       }
+
        if (!(qbh->data = kmalloc(2048, GFP_NOFS))) {
                printk("HPFS: hpfs_get_4sectors: out of memory\n");
-               return NULL;
+               goto bail4;
        }
-       if (!(hpfs_get_sector(s, secno, &qbh->bh[0]))) goto bail0;
-       if (!(hpfs_get_sector(s, secno + 1, &qbh->bh[1]))) goto bail1;
-       if (!(hpfs_get_sector(s, secno + 2, &qbh->bh[2]))) goto bail2;
-       if (!(hpfs_get_sector(s, secno + 3, &qbh->bh[3]))) goto bail3;
-       memcpy(qbh->data, qbh->bh[0]->b_data, 512);
-       memcpy(qbh->data + 512, qbh->bh[1]->b_data, 512);
-       memcpy(qbh->data + 2*512, qbh->bh[2]->b_data, 512);
-       memcpy(qbh->data + 3*512, qbh->bh[3]->b_data, 512);
        return qbh->data;
 
-       bail3:  brelse(qbh->bh[2]);
-       bail2:  brelse(qbh->bh[1]);
-       bail1:  brelse(qbh->bh[0]);
-       bail0:
+bail4:
+       brelse(qbh->bh[3]);
+bail3:
+       brelse(qbh->bh[2]);
+bail2:
+       brelse(qbh->bh[1]);
+bail1:
+       brelse(qbh->bh[0]);
+bail0:
        return NULL;
 }
        
 
 void hpfs_brelse4(struct quad_buffer_head *qbh)
 {
-       brelse(qbh->bh[3]);
-       brelse(qbh->bh[2]);
-       brelse(qbh->bh[1]);
+       if (unlikely(qbh->data != qbh->bh[0]->b_data))
+               kfree(qbh->data);
        brelse(qbh->bh[0]);
-       kfree(qbh->data);
+       brelse(qbh->bh[1]);
+       brelse(qbh->bh[2]);
+       brelse(qbh->bh[3]);
 }      
 
 void hpfs_mark_4buffers_dirty(struct quad_buffer_head *qbh)
 {
-       memcpy(qbh->bh[0]->b_data, qbh->data, 512);
-       memcpy(qbh->bh[1]->b_data, qbh->data + 512, 512);
-       memcpy(qbh->bh[2]->b_data, qbh->data + 2 * 512, 512);
-       memcpy(qbh->bh[3]->b_data, qbh->data + 3 * 512, 512);
+       if (unlikely(qbh->data != qbh->bh[0]->b_data)) {
+               memcpy(qbh->bh[0]->b_data, qbh->data + 0 * 512, 512);
+               memcpy(qbh->bh[1]->b_data, qbh->data + 1 * 512, 512);
+               memcpy(qbh->bh[2]->b_data, qbh->data + 2 * 512, 512);
+               memcpy(qbh->bh[3]->b_data, qbh->data + 3 * 512, 512);
+       }
        mark_buffer_dirty(qbh->bh[0]);
        mark_buffer_dirty(qbh->bh[1]);
        mark_buffer_dirty(qbh->bh[2]);