[commit] r1162 - trunk/openbios-devel/fs/hfsplus

[repository service]

Author: agraf
Date: Tue Jun 25 19:23:09 2013
New Revision: 1162
URL: http://tracker.coreboot.org/trac/openbios/changeset/1162

Log:
hfsplus: Fix disk support for disks > 4GB

The HFS+ driver was truncating the end of the disk to 32bit implicitly
by multiplying a 32bit value with another 32bit value. Cast the first one
to 64bit, making the result 64bit wide as well.

This fixes booting from big disks for me.

Signed-off-by: Alexander Graf

Modified:
   trunk/openbios-devel/fs/hfsplus/hfsp_volume.c

Modified: trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
==============================================================================
--- trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 18:11:57 2013	(r1161)
+++ trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 19:23:09 2013	(r1162)
@@ -207,7 +207,7 @@
                 ret = volume_readbuf(vh, p);
    	if( !ret ) {
    	    /* When reading the initial partition we must use 512 byte blocks */
-		    vol_size = vh->blocksize * vh->total_blocks;
+		    vol_size = (uint64_t)vh->blocksize * vh->total_blocks;
    	    vol->maxblocks = vol_size / HFSP_BLOCKSZ;
    	}

[Programmingkid]

On Jun 25, 2013, at 1:23 PM, repository service wrote:
> Author: agraf
> Date: Tue Jun 25 19:23:09 2013
> New Revision: 1162
> URL: http://tracker.coreboot.org/trac/openbios/changeset/1162
>
> Log:
> hfsplus: Fix disk support for disks > 4GB
>
> The HFS+ driver was truncating the end of the disk to 32bit implicitly
> by multiplying a 32bit value with another 32bit value. Cast the first one
> to 64bit, making the result 64bit wide as well.
>
> This fixes booting from big disks for me.
>
> Signed-off-by: Alexander Graf
>
> Modified:
>   trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> Modified: trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
>
> --- trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 18:11:57 2013	(r1161)
> +++ trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 19:23:09 2013	(r1162)
> @@ -207,7 +207,7 @@
>                 ret = volume_readbuf(vh, p);
>    	if( !ret ) {
>    	    /* When reading the initial partition we must use 512 byte blocks */
>     vol_size = vh->blocksize * vh->total_blocks;
>     vol_size = (uint64_t)vh->blocksize * vh->total_blocks;
> vol->maxblocks = vol_size / HFSP_BLOCKSZ;
> }

I tried booting from a Mac OS 10.4 install dvd, but it didn't work. The DVD image is only 2.83GB. Is anyone able to boot from DVD images yet?

[Alexander Graf]

On 25.06.2013, at 23:16, Programmingkid wrote:
> On Jun 25, 2013, at 1:23 PM, repository service wrote:
> > Author: agraf
> > Date: Tue Jun 25 19:23:09 2013
> > New Revision: 1162
> > URL: http://tracker.coreboot.org/trac/openbios/changeset/1162
> >
> > Log:
> > hfsplus: Fix disk support for disks > 4GB
> >
> > The HFS+ driver was truncating the end of the disk to 32bit implicitly
> > by multiplying a 32bit value with another 32bit value. Cast the first one
> > to 64bit, making the result 64bit wide as well.
> >
> > This fixes booting from big disks for me.
> >
> > Signed-off-by: Alexander Graf
> >
> > Modified:
> >  trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > Modified: trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> >
> > --- trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 18:11:57 2013	(r1161)
> > +++ trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 19:23:09 2013	(r1162)
> > @@ -207,7 +207,7 @@
> >                ret = volume_readbuf(vh, p);
> >    	if( !ret ) {
> >    	    /* When reading the initial partition we must use 512 byte blocks */
> >     vol_size = vh->blocksize * vh->total_blocks;
> >     vol_size = (uint64_t)vh->blocksize * vh->total_blocks;
> > vol->maxblocks = vol_size / HFSP_BLOCKSZ;
> > }
>
> I tried booting from a Mac OS 10.4 install dvd, but it didn't work. The DVD image is only 2.83GB. Is anyone able to boot from DVD images yet?

Yes, I'm booting from the 10.4 install image for a while now. It just doesn't find its root device yet. I found that particular issue with a ready-installed 10.4.11 image that I had created earlier with MOL. I was planning to use that to debug the guest with self-compiled kexts.

As a nice side-effect, I can also boot my leopard DVD as well now. Though that one really does complain about the g3beige ;).

Alex

[Alexander Graf]

On 26.06.2013, at 01:25, Alexander Graf wrote:
> On 25.06.2013, at 23:16, Programmingkid wrote:
> > On Jun 25, 2013, at 1:23 PM, repository service wrote:
> > > Author: agraf
> > > Date: Tue Jun 25 19:23:09 2013
> > > New Revision: 1162
> > > URL: http://tracker.coreboot.org/trac/openbios/changeset/1162
> > >
> > > Log:
> > > hfsplus: Fix disk support for disks > 4GB
> > >
> > > The HFS+ driver was truncating the end of the disk to 32bit implicitly
> > > by multiplying a 32bit value with another 32bit value. Cast the first one
> > > to 64bit, making the result 64bit wide as well.
> > >
> > > This fixes booting from big disks for me.
> > >
> > > Signed-off-by: Alexander Graf
> > >
> > > Modified:
> > > trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > > Modified: trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > >
> > > --- trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 18:11:57 2013	(r1161)
> > > +++ trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 19:23:09 2013	(r1162)
> > > @@ -207,7 +207,7 @@
> > >               ret = volume_readbuf(vh, p);
> > >    	if( !ret ) {
> > >    	    /* When reading the initial partition we must use 512 byte blocks */
> > >     vol_size = vh->blocksize * vh->total_blocks;
> > >     vol_size = (uint64_t)vh->blocksize * vh->total_blocks;
> > > vol->maxblocks = vol_size / HFSP_BLOCKSZ;
> > > }
> >
> > I tried booting from a Mac OS 10.4 install dvd, but it didn't work. The DVD image is only 2.83GB. Is anyone able to boot from DVD images yet?
>
> Yes, I'm booting from the 10.4 install image for a while now. It just doesn't find its root device yet. I found that particular issue with a ready-installed 10.4.11 image that I had created earlier with MOL. I was planning to use that to debug the guest with self-compiled kexts.
>
> As a nice side-effect, I can also boot my leopard DVD as well now. Though that one really does complain about the g3beige ;).

It looks like whether booting works is very timing sensitive. It almost feels as if DMA transfers are racy, but that's really just a wild guess.

Alex

[Programmingkid]

On Jun 25, 2013, at 10:10 PM, Alexander Graf wrote:
> On 26.06.2013, at 01:25, Alexander Graf wrote:
> > On 25.06.2013, at 23:16, Programmingkid wrote:
> > > On Jun 25, 2013, at 1:23 PM, repository service wrote:
> > > > Author: agraf
> > > > Date: Tue Jun 25 19:23:09 2013
> > > > New Revision: 1162
> > > > URL: http://tracker.coreboot.org/trac/openbios/changeset/1162
> > > >
> > > > Log:
> > > > hfsplus: Fix disk support for disks > 4GB
> > > >
> > > > The HFS+ driver was truncating the end of the disk to 32bit implicitly
> > > > by multiplying a 32bit value with another 32bit value. Cast the first one
> > > > to 64bit, making the result 64bit wide as well.
> > > >
> > > > This fixes booting from big disks for me.
> > > >
> > > > Signed-off-by: Alexander Graf
> > > >
> > > > Modified:
> > > > trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > > > Modified: trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > > >
> > > > --- trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 18:11:57 2013	(r1161)
> > > > +++ trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 19:23:09 2013	(r1162)
> > > > @@ -207,7 +207,7 @@
> > > >              ret = volume_readbuf(vh, p);
> > > >    	if( !ret ) {
> > > >    	    /* When reading the initial partition we must use 512 byte blocks */
> > > >     vol_size = vh->blocksize * vh->total_blocks;
> > > >     vol_size = (uint64_t)vh->blocksize * vh->total_blocks;
> > > > vol->maxblocks = vol_size / HFSP_BLOCKSZ;
> > > > }
> > >
> > > I tried booting from a Mac OS 10.4 install dvd, but it didn't work. The DVD image is only 2.83GB. Is anyone able to boot from DVD images yet?
> >
> > Yes, I'm booting from the 10.4 install image for a while now. It just doesn't find its root device yet. I found that particular issue with a ready-installed 10.4.11 image that I had created earlier with MOL. I was planning to use that to debug the guest with self-compiled kexts.
> >
> > As a nice side-effect, I can also boot my leopard DVD as well now. Though that one really does complain about the g3beige ;).
>
> It looks like whether booting works is very timing sensitive. It almost feels as if DMA transfers are racy, but that's really just a wild guess.

There might be a timing issue involved.

How did you make your DVD image files? I used Disk Utility and made a DVD Master file. Then just changed the extension to iso. This way worked with CDs. Just not with DVDs. Are you able to view the contents of your DVDs in OpenBIOS using the dir cd:,\  command?

[Alexander Graf]

On 26.06.2013, at 04:19, Programmingkid wrote:
> On Jun 25, 2013, at 10:10 PM, Alexander Graf wrote:
> > On 26.06.2013, at 01:25, Alexander Graf wrote:
> > > On 25.06.2013, at 23:16, Programmingkid wrote:
> > > > On Jun 25, 2013, at 1:23 PM, repository service wrote:
> > > > > Author: agraf
> > > > > Date: Tue Jun 25 19:23:09 2013
> > > > > New Revision: 1162
> > > > > URL: http://tracker.coreboot.org/trac/openbios/changeset/1162
> > > > >
> > > > > Log:
> > > > > hfsplus: Fix disk support for disks > 4GB
> > > > >
> > > > > The HFS+ driver was truncating the end of the disk to 32bit implicitly
> > > > > by multiplying a 32bit value with another 32bit value. Cast the first one
> > > > > to 64bit, making the result 64bit wide as well.
> > > > >
> > > > > This fixes booting from big disks for me.
> > > > >
> > > > > Signed-off-by: Alexander Graf
> > > > >
> > > > > Modified:
> > > > > trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > > > > Modified: trunk/openbios-devel/fs/hfsplus/hfsp_volume.c
> > > > >
> > > > > --- trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 18:11:57 2013	(r1161)
> > > > > +++ trunk/openbios-devel/fs/hfsplus/hfsp_volume.c	Tue Jun 25 19:23:09 2013	(r1162)
> > > > > @@ -207,7 +207,7 @@
> > > > >             ret = volume_readbuf(vh, p);
> > > > >    	if( !ret ) {
> > > > >    	    /* When reading the initial partition we must use 512 byte blocks */
> > > > >     vol_size = vh->blocksize * vh->total_blocks;
> > > > >     vol_size = (uint64_t)vh->blocksize * vh->total_blocks;
> > > > > vol->maxblocks = vol_size / HFSP_BLOCKSZ;
> > > > > }
> > > >
> > > > I tried booting from a Mac OS 10.4 install dvd, but it didn't work. The DVD image is only 2.83GB. Is anyone able to boot from DVD images yet?
> > >
> > > Yes, I'm booting from the 10.4 install image for a while now. It just doesn't find its root device yet. I found that particular issue with a ready-installed 10.4.11 image that I had created earlier with MOL. I was planning to use that to debug the guest with self-compiled kexts.
> > >
> > > As a nice side-effect, I can also boot my leopard DVD as well now. Though that one really does complain about the g3beige ;).
> >
> > It looks like whether booting works is very timing sensitive. It almost feels as if DMA transfers are racy, but that's really just a wild guess.
>
> There might be a timing issue involved.
>
> How did you make your DVD image files?

I simply dd'ed the DVD from Linux :).
> I used Disk Utility and made a DVD Master file.

That works too - just make sure that you pass -drive ...,format=raw into QEMU.

Alex