Linux-2.4.6-pre3

Linux-2.4.6-pre3

[Linus Torvalds]

User-noticeable things: if you are tired of not being able to NFS-export
your reiserfs tree, this should make you happy.

VM tuning has also happened, with Rik van Riel, Mike Galbraith, Marcelo
Tosatti and Andrew Morton all doing various tweaks. Give it a whirl.

		Linus

-----

-pre3:
 - remember to increment the version number
 - Chris Mason: reiserfs mark_journal_new and bh leak fix
 - Richard Gooch: devfs update
 - Alexander Viro: further FS cleanup (superblock list)
 - David Woodhouse: MTD update
 - Kai Germaschewski: ISDN update (stanford checker fixes etc)
 - Rich Baum: gcc-3.0 warning fixes
 - Jeff Garzik: network driver updates
 - Geert Uytterhoeven: m68k fbdev logo merge glitch fix
 - Andrea Arcangeli: fix signal return path
 - David Miller: Sparc updates
 - Johannes Erdfelt: USB update
 - Carsten Otte, Andries Brouwer: don't clear blk_size unconditionally
   on partition check
 - Martin Frey: alpha Sable irq fix
 - Paul Mackerras: PPC softirq update
 - Patrick Mochel: PCI power management infrastructure
 - Robert Siemer: miroSOUND driver update
 - Neil Brown: knfsd updates, including ability to export ReiserFS filesystems
 - Trond Myklebust: NFS readdir fixup, don't update atime on client
 - Andrew Morton: truncate_inode_pages speedup
 - Paul Menage: make inode quota count all inodes..

-pre2:
 - Takanori Kawano: brlock indexing bugfix
 - Ingo Molnar, Jeff Garzik: softirq updates and fixes
 - Al Viro: rampage of superblock cleanups.
 - Jean Tourrilhes: Orinoco driver update v6, IrNET update
 - Trond Myklebust: NFS brown-paper-bag thing
 - Tim Waugh: parport update
 - David Miller: networking and sparc updates
 - Jes Sorensen: m68k update.
 - Ben Fennema: UDF update
 - Geert Uytterhoeven: fbdev logo updates
 - Willem Riede: osst driver updates
 - Paul Mackerras: PPC update
 - Marcelo Tosatti: unlazy swap cache
 - Mikulas Patocka: hpfs update

-pre1:
 - Andreas Dilger: make ext2fs react more gracefully to inode disk
   errors
 - Andrea Arkangeli: fix up alpha compile issues
 - Ingo Molnar: io-apic MP table parsing update and softirq latency
   update
 - Johannes Erdfelt: USB updates
 - Richard Henderson: alpha rawhide irq handling fixes
 - Marcelo, Andrea, Rik: more VM issues
 - Al Viro: fix various ext2 directory handling checks by biting the
   bullet and using the page cache.

Re: Linux-2.4.6-pre3

[Keith Owens]

On Tue, 12 Jun 2001 18:42:45 -0700 (PDT), 
Linus Torvalds <torvalds@transmeta.com> wrote:
>-pre3:
> - Jeff Garzik: network driver updates

tulip_core.c:1756: warning: initialization from incompatible pointer type
tulip_core.c:1757: warning: initialization from incompatible pointer type

Re: Linux-2.4.6-pre3

[Jeff Garzik]

Keith Owens wrote:
> 
> On Tue, 12 Jun 2001 18:42:45 -0700 (PDT),
> Linus Torvalds <torvalds@transmeta.com> wrote:
> >-pre3:
> > - Jeff Garzik: network driver updates
> 
> tulip_core.c:1756: warning: initialization from incompatible pointer type
> tulip_core.c:1757: warning: initialization from incompatible pointer type

There are no network driver updates, including no tulip updates

The PCI API changed, there is breakage and cleanup is needed

-- 
Jeff Garzik      | Andre the Giant has a posse.
Building 1024    |
MandrakeSoft     |

Re: Linux-2.4.6-pre3

[Patrick Mochel]

On Wed, 13 Jun 2001, Keith Owens wrote:

> On Tue, 12 Jun 2001 18:42:45 -0700 (PDT), 
> Linus Torvalds <torvalds@transmeta.com> wrote:
> >-pre3:
> > - Jeff Garzik: network driver updates
> 
> tulip_core.c:1756: warning: initialization from incompatible pointer type
> tulip_core.c:1757: warning: initialization from incompatible pointer type

This is likely due to the updates to struct pci_driver.

The suspend callback was changed to take another parameter (the state it
is to enter) and to return an int.

The resume callback was changed to return an int.

Since these callbacks are rarely, if ever used, and since they don't
cause an actual compile error, the changes were considered benign.

	-pat

Re: Linux-2.4.6-pre3

[Tim Waugh]

[-- Attachment #1: Type: text/plain, Size: 170 bytes --]

On Tue, Jun 12, 2001 at 10:48:30PM -0400, Jeff Garzik wrote:

> The PCI API changed, there is breakage and cleanup is needed

..in a stable kernel series.. :-((

Tim.
*/

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Linux-2.4.6-pre3

[José Luis Domingo López]

On Tuesday, 12 June 2001, at 18:42:45 -0700,
Linus Torvalds wrote:

> 
> User-noticeable things: if you are tired of not being able to NFS-export
> your reiserfs tree, this should make you happy.
> 
> VM tuning has also happened, with Rik van Riel, Mike Galbraith, Marcelo
> Tosatti and Andrew Morton all doing various tweaks. Give it a whirl.
> 
Here is my rather simple VM stressing test. Just opening some applications
on the machine I use every day. Tested with 2.4.4 and 2.4.6-pre3.

Machine OFF. Power ON. Some processes running in the background, but iddle
(process list available on request, should it matters). Each section shows
output from "free" with additional applicatons started. Once KDE, Mozilla,
Konqueror and StarOffice are loaded, they are closed in the reverse order
as the were started.

------------------------------------------------------------------
Linux joseluis 2.4.4 #1 lun abr 30 12:02:46 CEST 2001 i686 unknown
------------------------------------------------------------------

Just booted
-----------
             total       used       free     shared    buffers     cached
Mem:        127124      30788      96336          0       3392      12360
-/+ buffers/cache:      15036     112088
Swap:       128516          0     128516

+ KDE 2.1.1 + rxvt
------------------
             total       used       free     shared    buffers     cached
Mem:        127124      85276      41848          0       5924      39656
-/+ buffers/cache:      39696      87428
Swap:       128516          0     128516

+ Konqueror (about:) + Mozilla 0.9.1 (www.engardelinux.com)
-----------------------------------------------------------
             total       used       free     shared    buffers     cached
Mem:        127124     125620       1504          0        524      45896
-/+ buffers/cache:      79200      47924
Swap:       128516      49072      79444

+ StarOffice 5.2 (ThreeDimensions.sdd)
--------------------------------------
             total       used       free     shared    buffers     cached
Mem:        127124     125656       1468          0        592      71964
-/+ buffers/cache:      53100      74024
Swap:       128516     104084      24432

+ Konqueror (about:) + Mozilla 0.9.1 (www.engardelinux.com)
-----------------------------------------------------------
             total       used       free     shared    buffers     cached
Mem:        127124     102812      24312          0        780      73900
-/+ buffers/cache:      28132      98992
Swap:       128516     102272      26244

KDE 2.1.1 + rxvt
----------------
             total       used       free     shared    buffers     cached
Mem:        127124     114660      12464          0       1176      98232
-/+ buffers/cache:      15252     111872
Swap:       128516      51068      77448

Back to the start
-----------------
             total       used       free     shared    buffers     cached
Mem:        127124     124100       3024          0       2692     110980
-/+ buffers/cache:      10428     116696
Swap:       128516      47972      80544

swapoff -a
----------
7 seconds, machine responsive

             total       used       free     shared    buffers     cached
Mem:        127124      91152      35972          0       2888      68640
-/+ buffers/cache:      19624     107500
Swap:            0          0          0


-----------------------------------------------------------------------
Linux joseluis 2.4.6-pre3 #1 mié jun 13 11:22:51 CEST 2001 i686 unknown
-----------------------------------------------------------------------

Just booted
-----------
             total       used       free     shared    buffers     cached
Mem:        127120      32108      95012          0       1468      14908
-/+ buffers/cache:      15732     111388
Swap:       128516          0     128516

KDE 2.1.1 + rxvt
----------------
             total       used       free     shared    buffers     cached
Mem:        127120      85316      41804          0       3036      42464
-/+ buffers/cache:      39816      87304
Swap:       128516          0     128516

+ Konqueror (about:) + Mozilla 0.9.1 (www.engardelinux.com)
-----------------------------------------------------------
             total       used       free     shared    buffers     cached
Mem:        127120     124804       2316          0        356      69264
-/+ buffers/cache:      55184      71936
Swap:       128516      59236      69280

+ StarOffice 5.2 (ThreeDimensions.sdd)
--------------------------------------
             total       used       free     shared    buffers     cached
Mem:        127120     124308       2812          0       1408      88592
-/+ buffers/cache:      34308      92812
Swap:       128516     109104      19412

+ Konqueror (about:) + Mozilla 0.9.1 (www.engardelinux.com)
-----------------------------------------------------------
             total       used       free     shared    buffers     cached
Mem:        127120     104168      22952          0       1488      87884
-/+ buffers/cache:      14796     112324
Swap:       128516     104516      24000

KDE 2.1.1 + rxvt
----------------
             total       used       free     shared    buffers     cached
Mem:        127120     106760      20360          0       1664      91188
-/+ buffers/cache:      13908     113212
Swap:       128516      35896      92620

Back to the start
-----------------
             total       used       free     shared    buffers     cached
Mem:        127120     103584      23536          0       1896      92168
-/+ buffers/cache:       9520     117600
Swap:       128516      22652     105864

swapoff -a
----------
4 seconds, machine responsive

             total       used       free     shared    buffers     cached
Mem:        127120      96656      30464          0       2020      75532
-/+ buffers/cache:      19104     108016
Swap:            0          0          0


Hope it helps.

--
José Luis Domingo López
Linux Registered User #189436     Debian GNU/Linux Potato (P166 64 MB RAM)
 
jdomingo EN internautas PUNTO org  => ¿ Spam ? Atente a las consecuencias
jdomingo AT internautas DOT   org  => Spam at your own risk

Re: Linux-2.4.6-pre3

[Jeff Golds]

Patrick Mochel wrote:
> 
> On Wed, 13 Jun 2001, Keith Owens wrote:
> 
> > tulip_core.c:1756: warning: initialization from incompatible pointer type
> > tulip_core.c:1757: warning: initialization from incompatible pointer type
> 
> This is likely due to the updates to struct pci_driver.
> 
> The suspend callback was changed to take another parameter (the state it
> is to enter) and to return an int.
> 
> The resume callback was changed to return an int.
> 
> Since these callbacks are rarely, if ever used, and since they don't
> cause an actual compile error, the changes were considered benign.
> 

No offense, but that kind of attitude is harmful to the progression of making Linux a good operating system.

Our products NEED suspend/resume to work properly.  PERIOD.  I just spent a few days fixing some bugs in the Tulip driver with suspend/resume, and now your telling me that it's ok to change the API because no one is using it?  So if people ARE using the suspend/resume calls they are just SOL?  Sorry, I don't think things work well that way.

These sort of changes should either wait until 2.5 OR wait until EVERYONE has time to change ALL the drivers so that things don't get broken when you change the API.

2.4.x is supposed to be a "stable" series of kernels.  Thus far, I have NOT been impressed with its stability.

-Jeff

-- 
Jeff Golds
Sr. Software Engineer
jgolds@resilience.com

Re: Linux-2.4.6-pre3

[Patrick Mochel]

On Wed, 13 Jun 2001, Jeff Golds wrote:

> Patrick Mochel wrote:
> > 
> > On Wed, 13 Jun 2001, Keith Owens wrote:
> > 
> > > tulip_core.c:1756: warning: initialization from incompatible pointer type
> > > tulip_core.c:1757: warning: initialization from incompatible pointer type
> > 
> > This is likely due to the updates to struct pci_driver.
> > 
> > The suspend callback was changed to take another parameter (the state it
> > is to enter) and to return an int.
> > 
> > The resume callback was changed to return an int.
> > 
> > Since these callbacks are rarely, if ever used, and since they don't
> > cause an actual compile error, the changes were considered benign.
> > 
> 
> No offense, but that kind of attitude is harmful to the progression of making Linux a good operating system.
> 
> Our products NEED suspend/resume to work properly.  PERIOD.  I just spent a few days fixing some bugs in the Tulip driver with suspend/resume, and now your telling me that it's ok to change the API because no one is using it?  So if people ARE using the suspend/resume calls they are just SOL?  Sorry, I don't think things work well that way.
> 
> These sort of changes should either wait until 2.5 OR wait until EVERYONE has time to change ALL the drivers so that things don't get broken when you change the API.
> 
> 2.4.x is supposed to be a "stable" series of kernels.  Thus far, I have NOT been impressed with its stability.

First off, the patch went into a pre-release of the kernel. Never would I
trust a pre-release to be stable. Other issues with stability should be
addressed through appropriate channels, not piggy-backing on another rant.

Second of all, if you look at the big picture, you may see the benefit in
the change that was made. There is an effort to make Linux support power
management, both at the system-wide level and at the driver level, in a
much better way than it ever has. These changes are a manifestation of
that effort. 

You want to know what power state the device has been requested to go in.
And, you want to be able to return some sort of error code from that
operation. And, you want to be able to do the same from resume. The
infrastructure is not there to take note of it yet; but it will be.

Thirdly, these functions are rarely used. There are 1609 .c files in the
drivers directory. Of these, 68 of these have a "struct pci_driver" in
them. Of those, 15 actually define a suspend callback. It is a fairly
trivial task to add the parameter and the proper return values to each of
those affected. Would it make you feel better if I promised that each one
would be free of compile warnings before 2.4.6 (final)?

It wouldn't me, because I know there is a lot of work to be done still.
How many of those drivers properly suspend and resume? With comments
like:

	/* I'm absolutely uncertain if this part of code may work.
	...

it doesn't give me much confidence. Nor do the several completely empty
suspend/resume implementations.

(Btw, 3 of these were updated to support the API changes in -pre3)

Next, since you know the tulip code so well, then you shouldn't have much
of a problem making sure that it can suspend to all of the states that it
supports, and that it can incrementally do so. See a previous post about
the possible state transitions. Also, make sure that it can handle
completely reinitialization on a D3->D0 transition. 

Also, there will be another function added to the callback - save_state,
which will the routine that handles both saving the device's state and
actually failing to notify the system that it can't enter a global suspend
state if the device can't make the transition for some reason. (This will
of course all be documented in pci.txt)

Lastly, politics is a game that cannot be won. Someone could have said
"We will change the suspend/resume API; please update drivers
accordingly." Someone else would have complained, unless the operation
happened completely atomically. And, I'm sure someone would have
complained even then. 

This change was relatively minor, and easy to remedy the warnings before
the next _stable_ version of the kernel was released. Plus, it will 
benefit everyone, anyway, in the end (hopefully).

	-pat

Re: Linux-2.4.6-pre3

[Tim Waugh]

[-- Attachment #1: Type: text/plain, Size: 899 bytes --]

On Wed, Jun 13, 2001 at 11:39:49PM -0700, Patrick Mochel wrote:

> First off, the patch went into a pre-release of the kernel. Never would I
> trust a pre-release to be stable.

The issue is that of interface stability, as I'm sure you know.

> Second of all, if you look at the big picture, you may see the benefit in
> the change that was made. There is an effort to make Linux support power
> management, both at the system-wide level and at the driver level, in a
> much better way than it ever has. These changes are a manifestation of
> that effort. 

I've no doubt that new and better functions were necessary: but
couldn't it have been done in a source-compatible way?  New functions
rather than changed ones?

> Thirdly, these functions are rarely used.

Not every single driver is in Linus' tree.

I didn't see a feature-test macro either: did I miss it?

Tim.
*/

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: Linux-2.4.6-pre3

[Jeff Garzik]

Patrick Mochel wrote:
> First off, the patch went into a pre-release of the kernel. Never would I
> trust a pre-release to be stable. Other issues with stability should be
> addressed through appropriate channels, not piggy-backing on another rant.
[...]
> This change was relatively minor, and easy to remedy the warnings before
> the next _stable_ version of the kernel was released. Plus, it will
> benefit everyone, anyway, in the end (hopefully).

You are totally missing the point.

First of all, I agree with the change (not surprising eh?), but the
other posters are definitely right.  If you are in a stable series,
having a patch arrive in a pre-release is totally immaterial.  A stable
series is a stable series.  API changes should be weighed far more
carefully than that.  You don't make changes just because there is
little existing breakage.  How do you know what you broke outside the
kernel tree?  Further, the change will might OOPS not just cause a
warning, if a driver is left in unmodified state.

And yes we need a feature macro too, as another poster mentioned.

I have converted the net drivers and ymfpci already to fix the
suspend/resume callbacks, and sent to Linus, so nobody needs to bother
with those.  Patches sent to Linus are archived at
ftp://ftp.us.kernel.org/pub/linux/kernel/people/jgarzik/patches/2.4.6/

Anyway I beg you -- please consider API changes more carefully in the
future, even if Quick Draw Torvalds does not.  The changes that occured
here are immaterial:  the principle of the stable series is what is at
stake here.

Finally, when you modify drivers, please CC the maintainer.  ie. when
you patched eepro100, you should have CC'd Andrey Savochin.  Read
MAINTAINERS for an e-mail address, or the source code if not in
MAINTAINERS.

Regards,

	Jeff



-- 
Jeff Garzik      | Andre the Giant has a posse.
Building 1024    |
MandrakeSoft     |

Re: Linux-2.4.6-pre3

[Alan Cox]

> On Tue, 12 Jun 2001 18:42:45 -0700 (PDT), 
> Linus Torvalds <torvalds@transmeta.com> wrote:
> >-pre3:
> > - Jeff Garzik: network driver updates
> 
> tulip_core.c:1756: warning: initialization from incompatible pointer type
> tulip_core.c:1757: warning: initialization from incompatible pointer type

Use pre2. Linus applied a patch that changed the PCI power management stuff
and broke all the drivers. In fact you were lucky you noticed this - it'll
compile with warnings and most users will never realise its totally broken,
or that every third party non kernel 2.4 driver using PM just broke too.

Re: Linux-2.4.6-pre3

[Linus Torvalds]

In article <E15AbRo-00053O-00@the-village.bc.nu>,
Alan Cox  <alan@lxorguk.ukuu.org.uk> wrote:
>
>Use pre2. Linus applied a patch that changed the PCI power management stuff
>and broke all the drivers.

It shouldn't have broken anything.  The warning happens, but the
function call ends up doing the same thing as it used to - old drivers
will just ignore the new argument. 

It was a necessary step in working ACPI suspend.  Which Patrick has
working - with caveats.  And the fact that Pat happens to work at the
same company I do probably has more to do with the fact that Transmeta
is obviously interested in suspend issues more than most - and not so
much with the fact that he would exert undue influence on me. 

		Linus