Re: 2.3.99-pre6 sound hfs

Your point being?

I was replying to Scott Knight [off-list ref]

        Alright, i have a couple questions.

...

Secondly is hfs support. Its nutty. I cant
mount any hfs cdroms. I can mount hfs hard drives but if something is
tried to copy to it it crashes the computer hard and emsse up the ahrd
drive a bit. Anyone know what the deal is with this? And yes I have hfs
support compiled in. ANy advice or help would be greatly apprecaited.

pointing out why hfs is not working, as he obviously was under the
assumption it was supposed to work in the latest devel kernels.

sorry for not quoting this originally if it added confusion, I tend to
edit out as much as possible in replies

	See You
	    Steve

--
sjh@wibble.net http://wibble.net/~sjh/
Look Up In The Sky
   Is it a bird?  No
      Is it a plane?  No
         Is it a small blue banana?
YES

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

quoted

Your point being?

I was replying to Scott Knight [off-list ref]

Sorry, missed that.

pointing out why hfs is not working, as he obviously was under the
assumption it was supposed to work in the latest devel kernels.

So was I until I saw Alan's laundry list. I haven't followed 2.3
development lately but I recall mention of further changes to the VFS.
Maybe a diff between 2.3.<last working> and 2.3.<recent> vfat or ext2 code
shows up something.

	Michael


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Michael Schmitz wrote:
[]

quoted

pointing out why hfs is not working, as he obviously was under the
assumption it was supposed to work in the latest devel kernels.

So was I until I saw Alan's laundry list. I haven't followed 2.3
development lately but I recall mention of further changes to the VFS.
Maybe a diff between 2.3.<last working> and 2.3.<recent> vfat or ext2 code
shows up something.

From my personal mail archive, I see that HFS was already broken in

2.3.18 last September.

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

quoted

So was I until I saw Alan's laundry list. I haven't followed 2.3
development lately but I recall mention of further changes to the VFS.
Maybe a diff between 2.3.<last working> and 2.3.<recent> vfat or ext2 code
shows up something.

quoted

From my personal mail archive, I see that HFS was already broken in

2.3.18 last September.

And it got fixed enough to be working again at least in 2.3.48 - that fix
(or rather workaround) may have been broken again lately.

I'll have a look at the diff as soon as I get around to rsyncing
2.3.<recent>. I don't really have much of a clue on VFS though :-(

	Michael


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Michael Schmitz wrote:
[]

quoted

quoted

From my personal mail archive, I see that HFS was already broken in

2.3.18 last September.

And it got fixed enough to be working again at least in 2.3.48 - that fix
(or rather workaround) may have been broken again lately.

This is a little mysterious. It looks almost as if HFS is working for
some people and not for others.

Around 2.3.45, there was a discussion with Ani Joshi who said that it
was working for him. For me it definitely wasn't. It hosed a ZIP disk so
thoroughly that MacOS died every time I just inserted it into the drive.
I had to reformat it first under Linux before MacOS would even touch it.

More recently (already in the 2.3.99's), Tom Rini claimed that HFS was
working for him. And it seems that even now it works for some people.
Weird.

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

quoted

quoted

quoted

From my personal mail archive, I see that HFS was already broken in

2.3.18 last September.

And it got fixed enough to be working again at least in 2.3.48 - that fix
(or rather workaround) may have been broken again lately.

This is a little mysterious. It looks almost as if HFS is working for
some people and not for others.

So it seems. I've no idea why I wasn't bitten by this bug - I've used the
approx. 300 MB HFS exchange partition a lot to copy kernels to MacOS but I
probably never ever did things like create a folder there (all the kernels
pile up in the top level directory :-)

And another thing: I recall initially having trouble with the kernel
panicking when I unmounted the HFS filesystem. Not each time, but on
occasion. Went away as soon as I started to sync before unmount (but I
haven't really pushed it).

Around 2.3.45, there was a discussion with Ani Joshi who said that it
was working for him. For me it definitely wasn't. It hosed a ZIP disk so
thoroughly that MacOS died every time I just inserted it into the drive.
I had to reformat it first under Linux before MacOS would even touch it.

I'll do my best to hose a ZIP disk, thanks for the tip. The 2.3.45 or so
fix must be what I remember (I had had a look at the HFS code and decided
that a few functions needed a complete rewrite, but someone found some
easier way to deal with it, though I never quite understood how that
worked).

More recently (already in the 2.3.99's), Tom Rini claimed that HFS was
working for him. And it seems that even now it works for some people.
Weird.

I'll give 2.3.99 a spin.

	Michael


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

As some of you have said before, the HFS issue in 2.3 is due to all the
VFS changes, page cache stuff being a big one.  anyhow, i poked around a
bit last nite and have stabalized it a bit more (though not perfect yet),
mounting and reads work fine, writes work sometimes and not, very strange.

below is a patch which makes hfs use generic file read/writes, obviously
more needs to be done then this, hopefully i'll post some more patches
soon...


ani

--- linux.orig/fs/hfs/file.c	Wed May  3 22:21:05 2000
+++ linux/fs/hfs/hfs/file.c	Thu May  4 23:24:04 2000

@@ -23,17 +23,14 @@

 /*================ Forward declarations ================*/

-static hfs_rwret_t hfs_file_read(struct file *, char *, hfs_rwarg_t,
-				 loff_t *);
-static hfs_rwret_t hfs_file_write(struct file *, const char *, hfs_rwarg_t,
-				  loff_t *);
+
 static void hfs_file_truncate(struct inode *);

 /*================ Global variables ================*/

 struct file_operations hfs_file_operations = {
-	read:		hfs_file_read,
-	write:		hfs_file_write,
+	read:		generic_file_read,
+	write:		generic_file_write,
 	mmap:		generic_file_mmap,
 	fsync:		file_fsync,
 };

@@ -174,46 +171,6 @@
 	return read;
 }

-/*
- * hfs_file_write()
- *
- * This is the write() entry in the file_operations structure for
- * "regular" files.  The purpose is to transfer up to 'count' bytes
- * to the file corresponding to 'inode' beginning at offset
- * 'file->f_pos' from user-space at the address 'buf'.  The return
- * value is the number of bytes actually transferred.
- */
-static hfs_rwret_t hfs_file_write(struct file * filp, const char * buf,
-				  hfs_rwarg_t count, loff_t *ppos)
-{
-        struct inode    *inode = filp->f_dentry->d_inode;
-	struct hfs_fork *fork = HFS_I(inode)->fork;
-	hfs_s32 written, pos;
-
-	if (!S_ISREG(inode->i_mode)) {
-		hfs_warn("hfs_file_write: mode = %07o\n", inode->i_mode);
-		return -EINVAL;
-	}
-
-	pos = (filp->f_flags & O_APPEND) ? inode->i_size : *ppos;
-
-	if (pos >= HFS_FORK_MAX) {
-		return 0;
-	}
-	if (count > HFS_FORK_MAX) {
-		count = HFS_FORK_MAX;
-	}
-	if ((written = hfs_do_write(inode, fork, pos, buf, count)) > 0)
-	        pos += written;
-
-	*ppos = pos;
-	if (*ppos > inode->i_size) {
-	        inode->i_size = *ppos;
-		mark_inode_dirty(inode);
-	}
-
-	return written;
-}

 /*
  * hfs_file_truncate()

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Ani Joshi wrote:

As some of you have said before, the HFS issue in 2.3 is due to all the
VFS changes, page cache stuff being a big one.  anyhow, i poked around a
bit last nite and have stabalized it a bit more (though not perfect yet),
mounting and reads work fine, writes work sometimes and not, very strange.

below is a patch which makes hfs use generic file read/writes, obviously
more needs to be done then this, hopefully i'll post some more patches
soon...

With this patch, I still cannot mount hfs CDs (the LinuxPPC 2000 CD),
and after a while of copying and gcc'ing on a hfs ZIP, it still freezes
the system. It seems, however, to survive somewhat longer than before,
and it freezes more gracefully, i.e. it puts me in a state (like vi
insert mode without Esc key) where I can type, even change consoles, but
it doesn't react. MagicSysRQ is still working, which wasn't the case
before.

--
Martin

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

quoted

quoted

quoted

From my personal mail archive, I see that HFS was already broken in

2.3.18 last September.

And it got fixed enough to be working again at least in 2.3.48 - that fix
(or rather workaround) may have been broken again lately.

This is a little mysterious. It looks almost as if HFS is working for
some people and not for others.

I've done a few tests and it looks like HFS in 2.3.48 is still pretty
broken. Creating a subdirectory, or copying a file to any directory except
for the root directory on a small zip partition is sure to panic. Even on
my 300 MB IDE partition, only copying to the root directory works more or
less reliably. Creating a subdirectory worked, copying the first file to
it worked, copying the second panicked. Sorry for causing a bit of
confusion...

I'll try 2.3.99 with Ani's patch next :-)

	Michael


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/