Linux 2.4.22-pre5

Linux 2.4.22-pre5

[Marcelo Tosatti]

Hi,

Here goes -pre5.

It fixes a deadlock introduced by the IO fairness changes, fixes ACPI on
IBM's x440, has an uptodated cciss driver, a new ethernet driver for IBM
PPC's 4xx, amongst other fixes.

Please help testing.

Summary of changes from v2.4.22-pre4 to v2.4.22-pre5
============================================

<jcchen:icplus.com.tw>:
  o [netdrvr sundance] increase eeprom read timeout

<mike.miller:hp.com>:
  o cciss: change names and correct subsystem device ID for U320
  o cciss: PCI BAR fix
  o cciss: Fix potential overrun
  o cciss: update version
  o cciss: First part of PCI changes/driver cleanup
  o cciss: Second part of PCI changes/driver cleanup

Andi Kleen:
  o Fix compiling on x86-64

Benjamin Herrenschmidt:
  o radeonfb 0.1.8 + my stuffs

Chris Mason:
  o Fix deadlocks in IO scheduler changes

David Woodhouse:
  o Backport vsprintf/scanf fixes from 2.5.74

Geert Uytterhoeven:
  o Fix adbhid m68k screwup

J. A. Magallon:
  o hfsplus: group Apple FS's and help text

John Stultz:
  o Fix boot crash of x440's in full acpi mode
  o Cleanup x440 acpi fix

Marcelo Tosatti:
  o Changed EXTRAVERSION to -pre5

Petr Vandrovec:
  o Fix matroxfb on PPC64

Tom Rini:
  o An ethernet driver for the IBM PPC 4xx series of machines

Re: Linux 2.4.22-pre5

[Marcelo Tosatti]

On Fri, 11 Jul 2003, Trond Myklebust wrote:

>
> Is there still any chance for the NFS O_DIRECT support to make it?

Yes.

I guess the best way of doing so would be adding ->direct_io2 and
KERNEL24_HAS_ODIRECT_2 define.

Correct?

Re: Linux 2.4.22-pre5

[Trond Myklebust]

>>>>> " " == Marcelo Tosatti <marcelo@conectiva.com.br> writes:

    >> Is there still any chance for the NFS O_DIRECT support to make
    >> it?

     > I guess the best way of doing so would be adding ->direct_io2
     > and KERNEL24_HAS_ODIRECT_2 define.

That is what the last patch I sent you does (also sent to l-k). Should
I resend?

Cheers,
  Trond

Re: Linux 2.4.22-pre5

[Jeff Garzik]

Trond Myklebust wrote:
>>>>>>" " == Marcelo Tosatti <marcelo@conectiva.com.br> writes:
> 
> 
>     >> Is there still any chance for the NFS O_DIRECT support to make
>     >> it?
> 
>      > I guess the best way of doing so would be adding ->direct_io2
>      > and KERNEL24_HAS_ODIRECT_2 define.

IMO, yes.


> That is what the last patch I sent you does (also sent to l-k). Should
> I resend?

Sounds like it :)

Christoph, opinion?

	Jeff

Re: Linux 2.4.22-pre5

[Marcelo Tosatti]

On Fri, 11 Jul 2003, Trond Myklebust wrote:

> >>>>> " " == Marcelo Tosatti <marcelo@conectiva.com.br> writes:
>
>     >> Is there still any chance for the NFS O_DIRECT support to make
>     >> it?
>
>      > I guess the best way of doing so would be adding ->direct_io2
>      > and KERNEL24_HAS_ODIRECT_2 define.
>
> That is what the last patch I sent you does (also sent to l-k). Should

No, no need to resend. I have it.

I released -pre5 so quickly because of the IO hang fix, which is pretty
important. (Christoph: your vmap patch will go on -pre6 too once I read
it).

Well, Jeff, Christoph, do you have any comments on Trond's new
O_DIRECT patch?

I haven't looked at it closely, yet.

Re: Linux 2.4.22-pre5

[Jeff Garzik]

Marcelo Tosatti wrote:
> Well, Jeff, Christoph, do you have any comments on Trond's new
> O_DIRECT patch?


I looked at it, and it looked ok to me.  Basically I was wondering was 
Christoph thought at this point...

	Jeff

Re: Linux 2.4.22-pre5

[Christoph Hellwig]

On Fri, Jul 11, 2003 at 06:56:16PM -0300, Marcelo Tosatti wrote:
> Well, Jeff, Christoph, do you have any comments on Trond's new
> O_DIRECT patch?

Patch looks fine.  I don't really like the directfileio name, but
hey, this cludge is so ugly that it doesn't really matter anymore..

Re: Linux 2.4.22-pre5

[Trond Myklebust]

>>>>> " " == Christoph Hellwig <hch@infradead.org> writes:

     > Patch looks fine.  I don't really like the directfileio name,

Feel free to suggest a better one. I chose it 'cos I hate the idea of
version numbers on function names (i.e. direct_iO2).

     > but hey, this cludge is so ugly that it doesn't really matter
     > anymore..

Agreed...

cheers,
  Trond

Re: Linux 2.4.22-pre5

[Richard A Nelson]

On Fri, 11 Jul 2003, Marcelo Tosatti wrote:

> Here goes -pre5.
[...]
>
> Benjamin Herrenschmidt:
>   o radeonfb 0.1.8 + my stuffs

radeonfb.c:168:28: linux/radeonfb.h: No such file or directory

-- 
Rick Nelson
I still maintain the point that designing a monolithic kernel in 1991 is a
fundamental error.  Be thankful you are not my student.  You would not get a
high grade for such a design :-)
(Andrew Tanenbaum to Linus Torvalds)

Re: Linux 2.4.22-pre5

[Marcelo Tosatti]

Ben?


On Fri, 11 Jul 2003, Richard A Nelson wrote:

> On Fri, 11 Jul 2003, Marcelo Tosatti wrote:
>
> > Here goes -pre5.
> [...]
> >
> > Benjamin Herrenschmidt:
> >   o radeonfb 0.1.8 + my stuffs
>
> radeonfb.c:168:28: linux/radeonfb.h: No such file or directory
>
> --
> Rick Nelson
> I still maintain the point that designing a monolithic kernel in 1991 is a
> fundamental error.  Be thankful you are not my student.  You would not get a
> high grade for such a design :-)
> (Andrew Tanenbaum to Linus Torvalds)
>

Re: Linux 2.4.22-pre5

[Benjamin Herrenschmidt]

On Sat, 2003-07-12 at 00:38, Marcelo Tosatti wrote:
> Ben?

bk fuckage on my side, here it is

#ifndef __LINUX_RADEONFB_H__
#define __LINUX_RADEONFB_H__

#include <asm/ioctl.h>
#include <asm/types.h>

#define ATY_RADEON_LCD_ON	0x00000001
#define ATY_RADEON_CRT_ON	0x00000002


#define FBIO_RADEON_GET_MIRROR	_IOR('@', 3, __u32)
#define FBIO_RADEON_SET_MIRROR	_IOW('@', 4, __u32)

#endif

Re: Linux 2.4.22-pre5 - unresolved in hfsplus

[Eyal Lebedinsky]

Marcelo Tosatti wrote:
> Here goes -pre5.
> J. A. Magallon:
>   o hfsplus: group Apple FS's and help text

depmod: *** Unresolved symbols in
/lib/modules/2.4.22-pre5/kernel/fs/hfsplus/hfsplus.o
depmod:         mark_page_accessed


--
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>

Re: Linux 2.4.22-pre5 - unresolved in hfsplus

[J.A. Magallon]

On 07.12, Eyal Lebedinsky wrote:
> Marcelo Tosatti wrote:
> > Here goes -pre5.
> > J. A. Magallon:
> >   o hfsplus: group Apple FS's and help text
> 
> depmod: *** Unresolved symbols in
> /lib/modules/2.4.22-pre5/kernel/fs/hfsplus/hfsplus.o
> depmod:         mark_page_accessed
> 

I did not touch the code, I suppose nobody has tried to build it as a module.
Try with this...

--- mm/filemap.c.orig	2003-07-12 10:24:55.000000000 +0200
+++ mm/filemap.c	2003-07-12 10:25:17.000000000 +0200
@@ -1338,6 +1338,8 @@
 		SetPageReferenced(page);
 }
 
+EXPORT_SYMBOL(mark_page_accessed);
+
 /*
  * This is a generic file read routine, and uses the
  * inode->i_op->readpage() function for the actual low-level


-- 
J.A. Magallon <jamagallon@able.es>      \                 Software is like sex:
werewolf.able.es                         \           It's better when it's free
Mandrake Linux release 9.2 (Cooker) for i586
Linux 2.4.22-pre2-jam1 (gcc 3.3 (Mandrake Linux 9.2 3.3-2mdk))

Re: Linux 2.4.22-pre5 - unresolved in hfsplus

[Eyal Lebedinsky]

"J.A. Magallon" wrote:
> 
> On 07.12, Eyal Lebedinsky wrote:
> > Marcelo Tosatti wrote:
> > > Here goes -pre5.
> > > J. A. Magallon:
> > >   o hfsplus: group Apple FS's and help text
> >
> > depmod: *** Unresolved symbols in
> > /lib/modules/2.4.22-pre5/kernel/fs/hfsplus/hfsplus.o
> > depmod:         mark_page_accessed
> >
> 
> I did not touch the code, I suppose nobody has tried to build it as a module.

Something was touched since I did not have this problem with the
2.4.21-ac
series which included hfsplus. This problem arrived with 2.4.22-pre3
(when hfsplus was introduced) so there was some merge issue maybe?

Actually, diffing .22-pre3-ac1 and .22-pre3 shows a few more fixes with
operator precedence missing in -pre3, like

-       if (!file->f_flags & O_DIRECT)
+       if (!(file->f_flags & O_DIRECT))

Which I am sure must be fixed.

--
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>