>Are you sure? If I remember the mcdx.c block strategy routine here
>I think one could get surprised here ...
The interface doesn't require the blockdevice driver to support requests
with blocks shorter than hardblocksize.
>So why comes it that it's possible to drive fs with less blocks through
>RAID?
raid remaps the device and the sector of the bh and that has nothing to do
with the size of the bh or with the blocksize. raid act like as a virtual
layer and the size doesn't change during the virt to phys translation.
>This doesn't answer one question: Why do all the other filesystems
>working fine
>without it? In esp. for example the minixfs.
Because it's buggy, so here the fix against 2.3.30pre3:
--- 2.3.30pre3-alpha/fs/minix/inode.c.~1~ Tue Sep 14 14:35:13 1999
+++ 2.3.30pre3-alpha/fs/minix/inode.c Mon Nov 29 17:45:02 1999
@@ -177,6 +177,7 @@
kdev_t dev = s->s_dev;
const char * errmsg;
struct inode *root_inode;
+ unsigned int hblock;
/* N.B. These should be compile-time tests.
Unfortunately that is impossible. */
@@ -186,6 +187,11 @@
panic("bad V2 i-node size");
MOD_INC_USE_COUNT;
+
+ hblock = get_hardblocksize(dev);
+ if (hblock && hblock > BLOCK_SIZE)
+ goto out_bad_hblock;
+
lock_super(s);
set_blocksize(dev, BLOCK_SIZE);
if (!(bh = bread(dev,1,BLOCK_SIZE)))
@@ -322,11 +328,16 @@
brelse(bh);
goto out_unlock;
+out_bad_hblock:
+ printk("MINIX-fs: blocksize too small for device.\n");
+ goto out;
+
out_bad_sb:
printk("MINIX-fs: unable to read superblock\n");
out_unlock:
- s->s_dev = 0;
unlock_super(s);
+ out:
+ s->s_dev = 0;
MOD_DEC_USE_COUNT;
return NULL;
}
>The NR_REQUESTS isn't used anywhere else and it's kernel internal
>thing that shouldn't be exposed to user space. (it wasn't).
When I gone to make the ll_rw_block layer bigmem capable I needed it from
the outside to make sure to avoid deadlocks due OOM. That's why I think
it's something that somebody (like the VM) may be interested about.
>I doubt seriously if the result of this procedure give good results on
>the average system out there. At least I would prefere much to have a
I dubited too but Dave says that testcase is been large enough to
represent to any kind of setup out there.
>deductive
>statement supporting it.
Me too.
Andrea
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/