floppy issue also in 2.4.18 ? / 2.5 solution

floppy issue also in 2.4.18 ? / 2.5 solution

[Nico Schottelius]

[-- Attachment #1: Type: text/plain, Size: 966 bytes --]

Hello again!

The floppy driver and me just married...
Two questions [hopefully first last]:

1) Why does 2.4.18 kernel hangs when I try to boot it from floppy ?
   Symptoms: After 'Loading ... ' progress finished, the disk drive's
   led is still lighting and nothing happens anymore.
   What can be wrong ? CPU is correct (I even tried 386,586,p-classic,p-mmx,..),
   it's a P133. And howto fix it?

2) Somebody explain me what changed in vfs, where to find doc and give me some
   time to fix that damn 2.5 floppy driver. I'm gonna fight it.

3) oops, one more: Can somebody else please fix the apm issue ?

Sincerly,

Nico /* tired, but alive */

-- 
Changing mail address: please forget all known @pcsystems.de addresses.

Please send your messages pgp-signed and/or pgp-encrypted (don't encrypt mails
to mailing list!). If you don't know what pgp is visit www.gnupg.org.
(public pgp key: ftp.schottelius.org/pub/familiy/nico/pgp-key)

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: floppy issue also in 2.4.18 ? / 2.5 solution

[Thunder from the hill]

Hi,

On Sun, 4 Aug 2002, Nico Schottelius wrote:
> 1) Why does 2.4.18 kernel hangs when I try to boot it from floppy ?
>    Symptoms: After 'Loading ... ' progress finished, the disk drive's
>    led is still lighting and nothing happens anymore.
>    What can be wrong ? CPU is correct (I even tried 386,586,p-classic,p-mmx,..),
>    it's a P133. And howto fix it?

Yo. Some days ago there was a patch that fixed the floppy I/O driver. I'll 
append it, against 2.5.29.

> 2) Somebody explain me what changed in vfs, where to find doc and give me some
>    time to fix that damn 2.5 floppy driver. I'm gonna fight it.

It was in 2.5.12. Just look at the changes, I don't have a pointer right 
now. (I'm busy writing something about the downturn of the first ancient 
egypt empire...)

			Thunder
-- 
.-../../-./..-/-..- .-./..-/.-.././.../.-.-.-

Re: floppy issue also in 2.4.18 ? / 2.5 solution

[Thunder from the hill]

diff -ruN linux-2.5.29/drivers/block/floppy.c linux-2.5.29.fix-floppy/drivers/block/floppy.c
--- linux-2.5.29/drivers/block/floppy.c	Thu Jul 25 01:27:30 2002
+++ linux-2.5.29.fix-floppy/drivers/block/floppy.c	Sat Jul 27 14:14:29 2002
@@ -3699,6 +3699,10 @@
 		UDRS->fd_ref = 0;
 	}
 	floppy_release_irq_and_dma();
+
+	/* undo ++bd_openers in floppy_open() */
+	--inode->i_bdev->bd_openers;
+
 	return 0;
 }
 
@@ -3791,6 +3795,43 @@
 		invalidate_buffers(mk_kdev(FLOPPY_MAJOR,old_dev));
 	}
 
+	/* Problems:
+	 * 1. floppy_open() triggers a call to submit_bio(), but
+	 *    bdev->bd_queue may be NULL since it is not set up until
+	 *    after floppy_open() has returned. Prior to 2.5.18, NULL
+	 *    queues were initialised when needed.
+	 * 2. fs/block_dev.c:do_open() changes bdev's block size
+	 *    after floppy_open() has returned. For some reason, this
+	 *    causes data corruption durings writes.
+	 * 3. fs/block_dev.c:do_open() sets bdev->bd_inode->i_size to 0
+	 *    after floppy_open() has returned. This cases ENOSPC errors
+	 *    on writes.
+	 * Workarounds:
+	 * 1. bdev->bd_queue is initialised here.
+	 * 2. Set bdev block size equal to the hardsect/queue size.
+	 *    This seems to cure the data corruption problem for raw
+	 *    accesses. VFS accesses to mounted floppies still cause
+	 *    data corruption for unknown reasons.
+	 * 3. Set i_size to 1.44M and bd_offset to 0.
+	 * 2&3. ++bdev->bd_openers to bypass do_open()'s changes.
+	 *    floppy_release() does a --bdev->bd_openers.
+	 */
+	{
+		struct block_device *bdev = inode->i_bdev;
+		kdev_t dev = inode->i_rdev;
+		struct blk_dev_struct *p = blk_dev + major(dev);
+
+		if (p->queue)
+			bdev->bd_queue = p->queue(dev);
+		else
+			bdev->bd_queue = &p->request_queue;
+		bdev->bd_block_size = bdev_hardsect_size(bdev);
+		bdev->bd_inode->i_blkbits = blksize_bits(block_size(bdev));
+		bdev->bd_offset = 0;
+		bdev->bd_inode->i_size = 1440 * 1024;
+		++bdev->bd_openers;
+	}
+
 	/* Allow ioctls if we have write-permissions even if read-only open.
 	 * Needed so that programs such as fdrawcmd still can work on write
 	 * protected disks */
@@ -4228,8 +4269,6 @@
 {
 	int i,unit,drive;
 
-	register_sys_device(&device_floppy);
-
 	raw_cmd = NULL;
 
 	devfs_handle = devfs_mk_dir (NULL, "floppy", NULL);
@@ -4356,6 +4395,9 @@
 			register_disk(NULL, mk_kdev(MAJOR_NR,TOMINOR(drive)+i*4),
 					1, &floppy_fops, 0);
 	}
+
+	register_sys_device(&device_floppy);
+
 	return have_no_fdc;
 }
 
@@ -4538,6 +4580,7 @@
 {
 	int dummy;
 		
+	unregister_sys_device(&device_floppy);
 	devfs_unregister (devfs_handle);
 	devfs_unregister_blkdev(MAJOR_NR, "fd");