any chance of 2.6.0-test*?

any chance of 2.6.0-test*?

[William Lee Irwin III]

Say, I've been having _smashing_ success with 2.5.x on the desktop and
on big fat highmem umpteen-way SMP (NUMA even!) boxen, and I was
wondering if you were considering 2.6.0-test* anytime soon.

I'd love to get this stuff out for users to hammer on ASAP, and things
are looking really good AFAICT.

Any specific concerns/issues/wishlist items you want taken care of
before doing it or is it a "generalized comfort level" kind of thing?
Let me know, I'd be much obliged for specific directions to move in.


Thanks,
Bill

Re: any chance of 2.6.0-test*?

[Dave Jones]

On Fri, Jan 10, 2003 at 08:10:12AM -0800, William Lee Irwin III wrote:
 > Say, I've been having _smashing_ success with 2.5.x on the desktop and
 > on big fat highmem umpteen-way SMP (NUMA even!) boxen, and I was
 > wondering if you were considering 2.6.0-test* anytime soon.

There's still a boatload of drivers that don't compile,
a metric shitload of bits that never came over from 2.4 after
I stopped doing it circa 2.4.18, a lot of little 'trivial'
patches that got left by the wayside, and a load of 'strange' bits
that still need nailing down (personally, I have two boxes
that won't boot a 2.5 kernel currently (One was pnpbios related,
other needs more investigation), and another that falls on its
face after 10 minutes idle uptime. My p4-ht desktop box is the only one
that runs 2.5 without any problems.

I think we're a way off from a '2.6-test' phase personally,
but instigating a harder 'code freeze' would probably be a
good thing to do[1]

		Dave

[1] Exemption granted for the bits not yet brought forward
    of course.

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: any chance of 2.6.0-test*?

[Zwane Mwaikambo]

On Fri, 10 Jan 2003, Dave Jones wrote:

> There's still a boatload of drivers that don't compile,
> a metric shitload of bits that never came over from 2.4 after
> I stopped doing it circa 2.4.18, a lot of little 'trivial'
> patches that got left by the wayside, and a load of 'strange' bits
> that still need nailing down (personally, I have two boxes
> that won't boot a 2.5 kernel currently (One was pnpbios related,

I had a problem with PCI init, pnpbios ordering at some point, but i
haven't tried a kernel with pnpbios in a while.

> other needs more investigation), and another that falls on its
> face after 10 minutes idle uptime. My p4-ht desktop box is the only one
> that runs 2.5 without any problems.

Thats interesting, i have a laptop experiencing the same symptoms, i'll be
looking at it over the weekend.

	Zwane
-- 
function.linuxpower.ca

Re: any chance of 2.6.0-test*?

[Dave Jones]

On Fri, Jan 10, 2003 at 11:41:48AM -0500, Zwane Mwaikambo wrote:

 > > other needs more investigation), and another that falls on its
 > > face after 10 minutes idle uptime. My p4-ht desktop box is the only one
 > > that runs 2.5 without any problems.
 > 
 > Thats interesting, i have a laptop experiencing the same symptoms, i'll be
 > looking at it over the weekend.

This is exactly the sort of thing I meant. There are still problems out
there which a lot of people haven't reported, yet when they see someone
else with that problem, suddenly the 'me too's come in.

I'm hoping the bugzilla will help out for corelating these as it starts to get
used more.

		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Fri, 2003-01-10 at 16:10, William Lee Irwin III wrote:
> Any specific concerns/issues/wishlist items you want taken care of
> before doing it or is it a "generalized comfort level" kind of thing?
> Let me know, I'd be much obliged for specific directions to move in.

IDE is all broken still and will take at least another three months to
fix - before we get to 'improve'. The entire tty layer locking is terminally
broken and nobody has even started fixing it. Just try a mass of parallel
tty/pty activity . It was problematic before, pre-empt has taken it  to dead, 
defunct and buried. 

Most of the drivers still don't build either.

I think its important that we get to the stage that we can actually say

- It compiles (as close to all the mainstream bits of it as possible)
- The stuff that is destined for the bitbucket is marked in Config and people
  agree it should go
- It works (certainly the common stuff)
- Its statistically unlikely to eat your computer
- It passes Cerberus uniprocessor and smp with/without pre-empt

Otherwise everyone wil rapidly decide that ".0-pre" means ".0 as in Windows"
at which point you've just destroyed your testing base.

Given all the new stuff should be in, I'd like to see a Linus the meanie
round of updating for a while which is simply about getting all the 2.4 fixes
and the 2.5 driver compile bugs nailed, and if it doesn't fix a compile bug
or a logic bug it doesn't go in.

No more "ISAPnP TNG" and module rewrites please

Alan

Re: any chance of 2.6.0-test*?

[Jeff Garzik]

On Fri, Jan 10, 2003 at 05:19:08PM +0000, Alan Cox wrote:
> No more "ISAPnP TNG" and module rewrites please

Highlighted and seconded...

Re: any chance of 2.6.0-test*?

[Dave Jones]

On Fri, Jan 10, 2003 at 05:19:08PM +0000, Alan Cox wrote:

 > Most of the drivers still don't build either.

or still lack 2.4 fixes.

 > - The stuff that is destined for the bitbucket is marked in Config and people
 >   agree it should go

What's happening with the OSS drivers ?
I'm still carrying a few hundred KB of changes from 2.4 for those.
I'm not going to spent a day splitting them up, commenting them and pushing
to Linus if we're going to be dropping various drivers.

 > - It passes Cerberus uniprocessor and smp with/without pre-empt

I think this should wait until at least some more of the 2.4 changes
come forward. Most of those are security issues and the likes, but there
are a few driver corner cases too.

 > Otherwise everyone wil rapidly decide that ".0-pre" means ".0 as in Windows"
 > at which point you've just destroyed your testing base.

agreed.

 > Given all the new stuff should be in, I'd like to see a Linus the meanie
 > round of updating for a while which is simply about getting all the 2.4 fixes
 > and the 2.5 driver compile bugs nailed, and if it doesn't fix a compile bug
 > or a logic bug it doesn't go in.

Seconded.

 > No more "ISAPnP TNG" and module rewrites please

Absolutly. Lets try and get 2.6 out the door _this_ year.

		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: any chance of 2.6.0-test*?

[Jeff Garzik]

On Fri, Jan 10, 2003 at 05:06:25PM +0000, Dave Jones wrote:
> On Fri, Jan 10, 2003 at 05:19:08PM +0000, Alan Cox wrote:
> What's happening with the OSS drivers ?
> I'm still carrying a few hundred KB of changes from 2.4 for those.
> I'm not going to spent a day splitting them up, commenting them and pushing
> to Linus if we're going to be dropping various drivers.

I've been updating the via audio every now and again.

If sound/oss is staying for 2.6.0, we might as well merge the 2.4.x
changes.

	Jeff

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Fri, 10 Jan 2003, Dave Jones wrote:
> 
> What's happening with the OSS drivers ?
> I'm still carrying a few hundred KB of changes from 2.4 for those.
> I'm not going to spent a day splitting them up, commenting them and pushing
> to Linus if we're going to be dropping various drivers.

I consider them to be old drivers, the same way "hd.c" was. Not
necessarily useful for most people, but neither was hd.c. And it was
around for a _long_ time (heh. I needed to check. The config option is 
still there ;)

So I don't see a huge reason to remove them from the sources, but we might
well make them harder to select by mistake, for example. Right now the
config help files aren't exactly helpful, and the OSS choice is before the
ALSA one, which looks wrong. 

They should probably be marked deprecated, and if they don't get a lot of 
maintenance, that's fine.

		Linus

Re: any chance of 2.6.0-test*?

[J.A. Magallon]

On 2003.01.10 Linus Torvalds wrote:
> 
> On Fri, 10 Jan 2003, Dave Jones wrote:
> > 
> > What's happening with the OSS drivers ?
> > I'm still carrying a few hundred KB of changes from 2.4 for those.
> > I'm not going to spent a day splitting them up, commenting them and pushing
> > to Linus if we're going to be dropping various drivers.
> 
> I consider them to be old drivers, the same way "hd.c" was. Not
> necessarily useful for most people, but neither was hd.c. And it was
> around for a _long_ time (heh. I needed to check. The config option is 
> still there ;)
> 
> So I don't see a huge reason to remove them from the sources, but we might
> well make them harder to select by mistake, for example. Right now the
> config help files aren't exactly helpful, and the OSS choice is before the
> ALSA one, which looks wrong. 
> 
> They should probably be marked deprecated, and if they don't get a lot of 
> maintenance, that's fine.
> 
> 		Linus

As there is a CONFIG_EXPERIMENTAL, how about a CONFIG_DEPRECATED for the
opposite edge ?

-- 
J.A. Magallon <jamagallon@able.es>      \                 Software is like sex:
werewolf.able.es                         \           It's better when it's free
Mandrake Linux release 9.1 (Cooker) for i586
Linux 2.4.21-pre2-jam2 (gcc 3.2.1 (Mandrake Linux 9.1 3.2.1-2mdk))

Re: any chance of 2.6.0-test*?

[Matthew D. Pitts]

> > They should probably be marked deprecated, and if they don't get a lot
of
> > maintenance, that's fine.
> >
> > Linus
>
> As there is a CONFIG_EXPERIMENTAL, how about a CONFIG_DEPRECATED for the
> opposite edge ?
>
I heartly second this. And would be willing to do it when time permits.

Matthew

Re: any chance of 2.6.0-test*?

[Jeff Garzik]

On Fri, Jan 10, 2003 at 07:47:39PM +0100, J.A. Magallon wrote:
> As there is a CONFIG_EXPERIMENTAL, how about a CONFIG_DEPRECATED for the
> opposite edge ?

CONFIG_OBSOLETE already exists for this.

Re: any chance of 2.6.0-test*?

[James H. Cloos Jr.]

[re OSS vs ALSA audio drivers]

Linus> So I don't see a huge reason to remove them from the sources,
Linus> but we might well make them harder to select by mistake, for
Linus> example. Right now the config help files aren't exactly helpful,
Linus> and the OSS choice is before the ALSA one, which looks wrong.

Linus> They should probably be marked deprecated, and if they don't
Linus> get a lot of maintenance, that's fine.

Like this?

Also available from:

        bk://cloos.bkbits.net/alsa-oss

ChangeSet@1.966, 2003-01-13 20:05:11-05:00, cloos@lugabout.jhcloos.org
  Move ALSA before OSS

 sound/Kconfig          |   26 ++++++++++++--------------
 arch/i386/defconfig    |   10 +++++-----
 arch/ia64/defconfig    |   10 +++++-----
 arch/parisc/defconfig  |    8 ++++----
 arch/sparc64/defconfig |   10 +++++-----
 5 files changed, 31 insertions(+), 33 deletions(-)


diff -Nru a/sound/Kconfig b/sound/Kconfig
--- a/sound/Kconfig	Mon Jan 13 20:32:29 2003
+++ b/sound/Kconfig	Mon Jan 13 20:32:29 2003
@@ -1,20 +1,6 @@
 # sound/Config.in
 #
 
-menu "Open Sound System"
-	depends on SOUND!=n
-
-config SOUND_PRIME
-	tristate "Open Sound System"
-	depends on SOUND
-	help
-	  Say 'Y' or 'M' to enable Open Sound System drivers.
-
-source "sound/oss/Kconfig"
-
-endmenu
-
-
 menu "Advanced Linux Sound Architecture"
 	depends on SOUND!=n
 
@@ -42,3 +28,15 @@
 
 endmenu
 
+menu "Open Sound System"
+	depends on SOUND!=n
+
+config SOUND_PRIME
+	tristate "Open Sound System (DEPRECATED)"
+	depends on SOUND
+	help
+	  Say 'Y' or 'M' to enable Open Sound System drivers.
+
+source "sound/oss/Kconfig"
+
+endmenu
diff -Nru a/arch/i386/defconfig b/arch/i386/defconfig
--- a/arch/i386/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/i386/defconfig	Mon Jan 13 20:32:29 2003
@@ -839,11 +839,6 @@
 CONFIG_SOUND=y
 
 #
-# Open Sound System
-#
-# CONFIG_SOUND_PRIME is not set
-
-#
 # Advanced Linux Sound Architecture
 #
 CONFIG_SND=y
@@ -922,6 +917,11 @@
 # ALSA USB devices
 #
 # CONFIG_SND_USB_AUDIO is not set
+
+#
+# Open Sound System
+#
+# CONFIG_SOUND_PRIME is not set
 
 #
 # USB support
diff -Nru a/arch/ia64/defconfig b/arch/ia64/defconfig
--- a/arch/ia64/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/ia64/defconfig	Mon Jan 13 20:32:29 2003
@@ -657,6 +657,11 @@
 CONFIG_SOUND=y
 
 #
+# Advanced Linux Sound Architecture
+#
+# CONFIG_SND is not set
+
+#
 # Open Sound System
 #
 CONFIG_SOUND_PRIME=y
@@ -679,11 +684,6 @@
 # CONFIG_SOUND_VIA82CXXX is not set
 # CONFIG_SOUND_OSS is not set
 # CONFIG_SOUND_TVMIXER is not set
-
-#
-# Advanced Linux Sound Architecture
-#
-# CONFIG_SND is not set
 
 #
 # USB support
diff -Nru a/arch/parisc/defconfig b/arch/parisc/defconfig
--- a/arch/parisc/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/parisc/defconfig	Mon Jan 13 20:32:29 2003
@@ -643,14 +643,14 @@
 CONFIG_SOUND=y
 
 #
-# Open Sound System
+# Advanced Linux Sound Architecture
 #
-# CONFIG_SOUND_PRIME is not set
+# CONFIG_SND is not set
 
 #
-# Advanced Linux Sound Architecture
+# Open Sound System
 #
-# CONFIG_SND is not set
+# CONFIG_SOUND_PRIME is not set
 
 #
 # USB support
diff -Nru a/arch/sparc64/defconfig b/arch/sparc64/defconfig
--- a/arch/sparc64/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/sparc64/defconfig	Mon Jan 13 20:32:29 2003
@@ -710,11 +710,6 @@
 CONFIG_SOUND=m
 
 #
-# Open Sound System
-#
-# CONFIG_SOUND_PRIME is not set
-
-#
 # Advanced Linux Sound Architecture
 #
 CONFIG_SND=m
@@ -777,6 +772,11 @@
 #
 CONFIG_SND_SUN_AMD7930=m
 CONFIG_SND_SUN_CS4231=m
+
+#
+# Open Sound System
+#
+# CONFIG_SOUND_PRIME is not set
 
 #
 # USB support

[PATCH][RESEND w/ reasonable Subject] alsa before oss in Kconfig

[James H. Cloos Jr.]

I should have changed the subject when I first sent this....

[re OSS vs ALSA audio drivers]
Linus> So I don't see a huge reason to remove them from the sources,
Linus> but we might well make them harder to select by mistake, for
Linus> example. Right now the config help files aren't exactly helpful,
Linus> and the OSS choice is before the ALSA one, which looks wrong.

Linus> They should probably be marked deprecated, and if they don't
Linus> get a lot of maintenance, that's fine.

Like this?

Also available from:

        bk://cloos.bkbits.net/alsa-oss

ChangeSet@1.966, 2003-01-13 20:05:11-05:00, cloos@lugabout.jhcloos.org
  Move ALSA before OSS

 sound/Kconfig          |   26 ++++++++++++--------------
 arch/i386/defconfig    |   10 +++++-----
 arch/ia64/defconfig    |   10 +++++-----
 arch/parisc/defconfig  |    8 ++++----
 arch/sparc64/defconfig |   10 +++++-----
 5 files changed, 31 insertions(+), 33 deletions(-)


diff -Nru a/sound/Kconfig b/sound/Kconfig
--- a/sound/Kconfig	Mon Jan 13 20:32:29 2003
+++ b/sound/Kconfig	Mon Jan 13 20:32:29 2003
@@ -1,20 +1,6 @@
 # sound/Config.in
 #
 
-menu "Open Sound System"
-	depends on SOUND!=n
-
-config SOUND_PRIME
-	tristate "Open Sound System"
-	depends on SOUND
-	help
-	  Say 'Y' or 'M' to enable Open Sound System drivers.
-
-source "sound/oss/Kconfig"
-
-endmenu
-
-
 menu "Advanced Linux Sound Architecture"
 	depends on SOUND!=n
 
@@ -42,3 +28,15 @@
 
 endmenu
 
+menu "Open Sound System"
+	depends on SOUND!=n
+
+config SOUND_PRIME
+	tristate "Open Sound System (DEPRECATED)"
+	depends on SOUND
+	help
+	  Say 'Y' or 'M' to enable Open Sound System drivers.
+
+source "sound/oss/Kconfig"
+
+endmenu
diff -Nru a/arch/i386/defconfig b/arch/i386/defconfig
--- a/arch/i386/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/i386/defconfig	Mon Jan 13 20:32:29 2003
@@ -839,11 +839,6 @@
 CONFIG_SOUND=y
 
 #
-# Open Sound System
-#
-# CONFIG_SOUND_PRIME is not set
-
-#
 # Advanced Linux Sound Architecture
 #
 CONFIG_SND=y
@@ -922,6 +917,11 @@
 # ALSA USB devices
 #
 # CONFIG_SND_USB_AUDIO is not set
+
+#
+# Open Sound System
+#
+# CONFIG_SOUND_PRIME is not set
 
 #
 # USB support
diff -Nru a/arch/ia64/defconfig b/arch/ia64/defconfig
--- a/arch/ia64/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/ia64/defconfig	Mon Jan 13 20:32:29 2003
@@ -657,6 +657,11 @@
 CONFIG_SOUND=y
 
 #
+# Advanced Linux Sound Architecture
+#
+# CONFIG_SND is not set
+
+#
 # Open Sound System
 #
 CONFIG_SOUND_PRIME=y
@@ -679,11 +684,6 @@
 # CONFIG_SOUND_VIA82CXXX is not set
 # CONFIG_SOUND_OSS is not set
 # CONFIG_SOUND_TVMIXER is not set
-
-#
-# Advanced Linux Sound Architecture
-#
-# CONFIG_SND is not set
 
 #
 # USB support
diff -Nru a/arch/parisc/defconfig b/arch/parisc/defconfig
--- a/arch/parisc/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/parisc/defconfig	Mon Jan 13 20:32:29 2003
@@ -643,14 +643,14 @@
 CONFIG_SOUND=y
 
 #
-# Open Sound System
+# Advanced Linux Sound Architecture
 #
-# CONFIG_SOUND_PRIME is not set
+# CONFIG_SND is not set
 
 #
-# Advanced Linux Sound Architecture
+# Open Sound System
 #
-# CONFIG_SND is not set
+# CONFIG_SOUND_PRIME is not set
 
 #
 # USB support
diff -Nru a/arch/sparc64/defconfig b/arch/sparc64/defconfig
--- a/arch/sparc64/defconfig	Mon Jan 13 20:32:29 2003
+++ b/arch/sparc64/defconfig	Mon Jan 13 20:32:29 2003
@@ -710,11 +710,6 @@
 CONFIG_SOUND=m
 
 #
-# Open Sound System
-#
-# CONFIG_SOUND_PRIME is not set
-
-#
 # Advanced Linux Sound Architecture
 #
 CONFIG_SND=m
@@ -777,6 +772,11 @@
 #
 CONFIG_SND_SUN_AMD7930=m
 CONFIG_SND_SUN_CS4231=m
+
+#
+# Open Sound System
+#
+# CONFIG_SOUND_PRIME is not set
 
 #
 # USB support

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Fri, 2003-01-10 at 17:06, Dave Jones wrote:
> What's happening with the OSS drivers ?
> I'm still carrying a few hundred KB of changes from 2.4 for those.
> I'm not going to spent a day splitting them up, commenting them and pushing
> to Linus if we're going to be dropping various drivers.

I'd hope they would go away but it seems that will be post 2.6. The
drivers do seem to mostly work in 2.5.x. I'm meaning the stuff that
doesn't compile and nobody gives a damn about.

Re: any chance of 2.6.0-test*?

[Jochen Friedrich]

On 10 Jan 2003, Alan Cox wrote:

> On Fri, 2003-01-10 at 16:10, William Lee Irwin III wrote:
> > Any specific concerns/issues/wishlist items you want taken care of
> > before doing it or is it a "generalized comfort level" kind of thing?
> > Let me know, I'd be much obliged for specific directions to move in.
>
> IDE is all broken still and will take at least another three months to
> fix - before we get to 'improve'.

As is the whole frame buffer mess. USB slowly seems to return to a working
state. ISDN seems to be a total mess, as well.

> No more "ISAPnP TNG" and module rewrites please

Full ACK. There are still archs without working module code, right now
(parisc and mips come to my mind).

--jochen

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Fri, 10 Jan 2003, Jochen Friedrich wrote:
> 
> Full ACK. There are still archs without working module code, right now
> (parisc and mips come to my mind).

Note that other architectures have never been an issue for releasing new
kernels, and that is _particularly_ true of architectures like parisc and
mips that haven't even _tried_ to track development kernels. In fact, mips
"anti-maintenance" has often actively discouraged people from even
bothering to update mips code when adding new features.

		Linus

Re: any chance of 2.6.0-test*?

[Greg KH]

On Fri, Jan 10, 2003 at 05:19:08PM +0000, Alan Cox wrote:
> The entire tty layer locking is terminally broken and nobody has even
> started fixing it. Just try a mass of parallel tty/pty activity . It
> was problematic before, pre-empt has taken it  to dead, defunct and
> buried. 

I've looked into this, and wow, it's not a simple fix :(

But this is really the first it's been mentioned, I can't see holding up
2.6 for this.  It's a 2.7 job at the earliest, unless someone wants to
step up and do it right now...

thanks,

greg k-h

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Sun, 2003-01-12 at 09:27, Greg KH wrote:
> I've looked into this, and wow, it's not a simple fix :(
> 
> But this is really the first it's been mentioned, I can't see holding up
> 2.6 for this.  It's a 2.7 job at the earliest, unless someone wants to
> step up and do it right now...

2.5.x crashes erratically and randomly under high tty/pty load. At the 
moment I'm assuming this is the tty code. That means we can't decide not
to fix it since its already fatally broken.

Alan

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Sun, 12 Jan 2003, Greg KH wrote:
> 
> But this is really the first it's been mentioned, I can't see holding up
> 2.6 for this.  It's a 2.7 job at the earliest, unless someone wants to
> step up and do it right now...

Hmm.. The tty layer still depends on the kernel lock, and we're clearly 
not changing that at this point any more.

I don't see what the fundamental problem is, it sounds like there's just
some part that got the lock yanked, probably as part of the normal VFS
kernel-lock-avoidance patches. Looking at tty_io.c (which pretty much
drives everything else), it does seem to take the lock _most_ of the time,
including read and write time.

The only place that looked like it _really_ didn't get the kernel lock was
apparently tty_open(), which is admittedly a fairly important part of it ;)

Alan, do you have a test-load? Might be worth just renaming "tty_open()" 
to "do_tty_open()", and then do a

	tty_open(..)
	{
		lock_kernel();
		retval = do_tty_open(..);
		unlock_kernel();
		return retval;
	}

thing.

Quite frankly, very little has changed in the 2.5.x tty layer (serial
stuff, yes, tty no), so its locking should work basically as well as 2.4.x
does. Except for infrastructure changes like the VFS lock removal (where
2.4.x will hold the kernel lock itself over open, 2.5.x won't).

Yeah, preemption has this tendency to show locking problems more easily,
and it could clearly be the case that it's broken in 2.4.x too but
basically impossible to trigger. But the _simple_ explanation is that
tty_open() (and perhaps some other cases like ioctl) just missed the
kernel lock.

		Linus

Re: any chance of 2.6.0-test*?

[Christoph Hellwig]

On Sun, Jan 12, 2003 at 09:05:16AM -0800, Linus Torvalds wrote:
> The only place that looked like it _really_ didn't get the kernel lock was
> apparently tty_open(), which is admittedly a fairly important part of it ;)

2.5 does hold the BKL on ->open of charater- (and block-) devices.

The real problem is that the big irqlock is gone and mingo just replaced
it with local_irq_save & friends, which is not enough.

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Sun, 12 Jan 2003, Christoph Hellwig wrote:
> 
> 2.5 does hold the BKL on ->open of charater- (and block-) devices.

Ahh, my bad.

> The real problem is that the big irqlock is gone and mingo just replaced
> it with local_irq_save & friends, which is not enough.

Ok, most of them should be fixable with a simple spinlock approach.

If the recursion is too nasty to handle, we could make some tty-specific 
recursive spinlock as a stop-gap measure, and mark it as being destined 
for the garbage-heap in 2.7.x:

	/*
	 * This isn't even _trying_ to be fast!
	 */
	struct recursive_spinlock {
		spinlock_t lock;
		int lock_count;
		struct task_struct *lock_owner;
	};

	static struct recursive_spinlock tty_lock = {
		.lock = SPIN_LOCK_UNLOCKED,
		.lock_count = 0,
		.lock_owner = NULL
	};

	unsigned long tty_spin_lock(void)
	{
		unsigned long flags;
		struct task_struct *tsk;

		local_irq_save(flags);
		preempt_disable();
		tsk = current;
		if (spin_trylock(&tty_lock.lock))
			goto got_lock;
		if (tsk == tty_lock.lock_owner) {
			WARN_ON(!tty_lock.lock_count);
			tty_lock.lock_count++;
			return flags;
		}
		spin_lock(&tty_lock.lock);
	got_lock:
		WARN_ON(tty_lock.lock_owner);
		WARN_ON(tty_lock.lock_count);
		tty_lock.lock_owner = tsk;
		tty_lock.lock_count = 1;
		return flags;
	}

	void tty_spin_unlock(unsigned long flags)
	{
		WARN_ON(tty_lock.lock_owner != current);
		WARN_ON(!tty_lock.lock_count);
		if (!--tty_lock.lock_count) {
			tty_lock.lock_owner = NULL;
			spin_unlock(&tty_lock.lock);
		}
		preempt_enable();
		local_irq_restore(flags);
	}

and be done with it.

Anybody willing to test it and see if the above works? 

			Linus

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

Linus,

I'm REALLY opposed to the use of the word "goto" in any code where it's
not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
to comment

Let me comment below the relevant code snippet below as to how I would
change it:

On Sun, 2003-01-12 at 14:15, Linus Torvalds wrote:
> 		if (spin_trylock(&tty_lock.lock))
> 			goto got_lock;
> 		if (tsk == tty_lock.lock_owner) {
> 			WARN_ON(!tty_lock.lock_count);
> 			tty_lock.lock_count++;
> 			return flags;
> 		}
> 		spin_lock(&tty_lock.lock);
> 	got_lock:
> 		WARN_ON(tty_lock.lock_owner);
	    	   <etc...>

I would change it to something like the following (without testing the
code through a compiler or anything to see if it's valid):

			if (!(spin_trylock(&tty_lock.lock))){
				if (tsk ==tty_lock.lock_owner){
					WRAN_ON(!tty_lock.lcok_count);
					tty_lock.lock_count++;
					return flags;
				}
			}
			WARN_ON(tty_lock.lock_owner);	
			<etc...>

Am I wrong that the above would do the same thing without generating the
sphagetti code that a goto would give you.  Gotos are BAD, very very
bad.  Please note also that the two if statements above could probably
even be combined further into one statement by using a short circuit &&
in the if.

If I'm misinterpreting the original code, then forgive me..  I just saw
a goto and gasped.  There's always a better option than goto.

-Rob

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

Minor change to my original message below.. I left a line out of a code
change suggestion.

On Sun, 2003-01-12 at 14:34, Rob Wilkens wrote:
> Linus,
> 
> I'm REALLY opposed to the use of the word "goto" in any code where it's
> not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> to comment
> 
> Let me comment below the relevant code snippet below as to how I would
> change it:
> 
> On Sun, 2003-01-12 at 14:15, Linus Torvalds wrote:
> > 		if (spin_trylock(&tty_lock.lock))
> > 			goto got_lock;
> > 		if (tsk == tty_lock.lock_owner) {
> > 			WARN_ON(!tty_lock.lock_count);
> > 			tty_lock.lock_count++;
> > 			return flags;
> > 		}
> > 		spin_lock(&tty_lock.lock);
> > 	got_lock:
> > 		WARN_ON(tty_lock.lock_owner);
> 	    	   <etc...>
> 
> I would change it to something like the following (without testing the
> code through a compiler or anything to see if it's valid):
> 
> 			if (!(spin_trylock(&tty_lock.lock))){
> 				if (tsk ==tty_lock.lock_owner){
> 					WRAN_ON(!tty_lock.lcok_count);
> 					tty_lock.lock_count++;
> 					return flags;
> 				}
> 			}
oops - Yes,  I forgot to add one line here (my point remains the same: 
			  spin_lock(&tty_lock.lock);
> 			WARN_ON(tty_lock.lock_owner);	
> 			<etc...>
> 
> Am I wrong that the above would do the same thing without generating the
> sphagetti code that a goto would give you.  Gotos are BAD, very very
> bad.  Please note also that the two if statements above could probably
> even be combined further into one statement by using a short circuit &&
> in the if.
> 
> If I'm misinterpreting the original code, then forgive me..  I just saw
> a goto and gasped.  There's always a better option than goto.
> 
> -Rob

Re: any chance of 2.6.0-test*?

[Tomas Szepe]

> [robw@optonline.net]
>
> Am I wrong that the above would do the same thing without generating the
> sphagetti code that a goto would give you.  Gotos are BAD, very very
> bad.

Whom do I pay to have this annoying clueless asshole shot?
OH MY GOD, I really can't take any more.

-- 
Tomas Szepe <szepe@pinerecords.com>

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 14:53, Tomas Szepe wrote:
> > [robw@optonline.net]
> >
> > Am I wrong that the above would do the same thing without generating the
> > sphagetti code that a goto would give you.  Gotos are BAD, very very
> > bad.
> 
> Whom do I pay to have this annoying clueless asshole shot?
> OH MY GOD, I really can't take any more.

Mail filters are a wonderful thing, use them on me and you want have to
have me shot.  We'll both be happier in the end.

What exactly is it that you can't take?  I made a minor suggestion in
the above message, very minor, and only a suggestion -- one that was
free to be ignorred.  There are over 300 messages a day on this list,
why do the few ones from _me_ bother you.

-Rob

Re: any chance of 2.6.0-test*?

[Terje Eggestad]

1) newbies has made the very same "minor" suggestion before; and had
their heads summarily  chopped of. 
2) It's not a minor suggestion.
3) and more to the point, you proved your selv why it's a dumb idea to
remove them, when you had to correct you own code 2 minutes after you
posted it.

-------------------- begin quote ---------------------------------
On Sun, 2003-01-12 at 14:34, Rob Wilkens wrote:
> I would change it to something like the following (without testing the
> code through a compiler or anything to see if it's valid):
> 
>                       if (!(spin_trylock(&tty_lock.lock))){
>                               if (tsk ==tty_lock.lock_owner){
>                                       WRAN_ON(!tty_lock.lcok_count);
>                                       tty_lock.lock_count++;
>                                       return flags;
>                               }
>                       }
oops - Yes,  I forgot to add one line here (my point remains the same: 
                          spin_lock(&tty_lock.lock);
>                       WARN_ON(tty_lock.lock_owner);   
>                       <etc...>
> 
-------------------- end quote ---------------------------------

OF all the reasons to use goto, spinlocks are alone reason enough. By
omitting the very little line you did
you created a race condition. Now I'm assuming that you know what a race
condition is.
IF YOU DON'T: DO NOT ASK!   L-O-O-K    I-T   U-P !!!!

Now since races are the worst kind of problems, in that they causes
seemingly random data corruption, or of you forget a spinunlock you get
a hung machine, minimize them is of paramount importance. 

Considering that doing kernel development is hard enough, new
development is almost always done on uni processors kernels that do only
one thing at the time. Then when you base logic is OK, you move to a
SMP, which means (adding and) debugging you spin locks.

Considering that fucking up spin locks are prone to corrupting your
machine, one very simple trick to makeing fewer mistakes to to have one,
and only one, unlock for every lock. 

In order to achive that you write your function to have one entry point
(which all C functions have) and one exit point, thus one and only one
return. 

In order to do that you can either concoct a set of if-then-else-orelse
logic that is incomprehensible, or you use goto!


Now you may look at Linus's code you buggified, and point out that the
lock and unlock are not in the same function, then read Linus's original
post CAREFULLY! And UNDERSTAND how these functions are used. 



 
On søn, 2003-01-12 at 21:03, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 14:53, Tomas Szepe wrote:
> > > [robw@optonline.net]
> > >
> > > Am I wrong that the above would do the same thing without generating the
> > > sphagetti code that a goto would give you.  Gotos are BAD, very very
> > > bad.
> > 
> > Whom do I pay to have this annoying clueless asshole shot?
> > OH MY GOD, I really can't take any more.
> 
> Mail filters are a wonderful thing, use them on me and you want have to
> have me shot.  We'll both be happier in the end.
> 
> What exactly is it that you can't take?  I made a minor suggestion in
> the above message, very minor, and only a suggestion -- one that was
> free to be ignorred.  There are over 300 messages a day on this list,
> why do the few ones from _me_ bother you.
> 
> -Rob
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
-- 
_________________________________________________________________________

Terje Eggestad                  mailto:terje.eggestad@scali.no
Scali Scalable Linux Systems    http://www.scali.com

Olaf Helsets Vei 6              tel:    +47 22 62 89 61 (OFFICE)
P.O.Box 150, Oppsal                     +47 975 31 574  (MOBILE)
N-0619 Oslo                     fax:    +47 22 62 89 51
NORWAY            
_________________________________________________________________________

Re: any chance of 2.6.0-test*?

[Jens Axboe]

On Mon, Jan 13 2003, Terje Eggestad wrote:
> Considering that doing kernel development is hard enough, new
> development is almost always done on uni processors kernels that do only
> one thing at the time. Then when you base logic is OK, you move to a
> SMP, which means (adding and) debugging you spin locks.

Goto's aside, I find the above extremely bad advise. You should _always_
develop with smp and for smp from the very start, or you will most
likely not get it right later on. With preempt, this becomes even more
important.

> Considering that fucking up spin locks are prone to corrupting your
> machine, one very simple trick to makeing fewer mistakes to to have one,
> and only one, unlock for every lock. 

Taking a spin lock twice will hard lock the machine, however on smp you
will typically have the luxury of an nmi watchdog which will help you
solve this quickly. Double unlock will oops immediately if you run with
spin lock debugging (you probably should, if you are developing kernel
code).

-- 
Jens Axboe

Re: any chance of 2.6.0-test*?

[Terje Eggestad]

On man, 2003-01-13 at 16:49, Jens Axboe wrote:
> On Mon, Jan 13 2003, Terje Eggestad wrote:
> > Considering that doing kernel development is hard enough, new
> > development is almost always done on uni processors kernels that do only
> > one thing at the time. Then when you base logic is OK, you move to a
> > SMP, which means (adding and) debugging you spin locks.
> 
> Goto's aside, I find the above extremely bad advise. You should _always_
> develop with smp and for smp from the very start, or you will most
> likely not get it right later on. With preempt, this becomes even more
> important.

You should, and I do, *design* with smp in mind, and I throw in
smplock/unlonk as I go, but I tend to make first runs on a UP. 

I see your point on preemt, though.
You do first runs on SMP? 

> > Considering that fucking up spin locks are prone to corrupting your
> > machine, one very simple trick to makeing fewer mistakes to to have one,
> > and only one, unlock for every lock. 
> 
> Taking a spin lock twice will hard lock the machine, however on smp you
> will typically have the luxury of an nmi watchdog which will help you
> solve this quickly. Double unlock will oops immediately if you run with
> spin lock debugging (you probably should, if you are developing kernel
> code).

I have the console on a serial port, and a terminal server. With kdb,
you can enter the kernel i kdb even when deadlocked.

TJ

-- 
_________________________________________________________________________

Terje Eggestad                  mailto:terje.eggestad@scali.no
Scali Scalable Linux Systems    http://www.scali.com

Olaf Helsets Vei 6              tel:    +47 22 62 89 61 (OFFICE)
P.O.Box 150, Oppsal                     +47 975 31 574  (MOBILE)
N-0619 Oslo                     fax:    +47 22 62 89 51
NORWAY            
_________________________________________________________________________

Re: any chance of 2.6.0-test*?

[Jens Axboe]

On Mon, Jan 13 2003, Terje Eggestad wrote:
> On man, 2003-01-13 at 16:49, Jens Axboe wrote:
> > On Mon, Jan 13 2003, Terje Eggestad wrote:
> > > Considering that doing kernel development is hard enough, new
> > > development is almost always done on uni processors kernels that do only
> > > one thing at the time. Then when you base logic is OK, you move to a
> > > SMP, which means (adding and) debugging you spin locks.
> > 
> > Goto's aside, I find the above extremely bad advise. You should _always_
> > develop with smp and for smp from the very start, or you will most
> > likely not get it right later on. With preempt, this becomes even more
> > important.
> 
> You should, and I do, *design* with smp in mind, and I throw in
> smplock/unlonk as I go, but I tend to make first runs on a UP. 
> 
> I see your point on preemt, though.
> You do first runs on SMP? 

Always, if for nothing else than the benefit of a better debugging
environment.

> > > Considering that fucking up spin locks are prone to corrupting your
> > > machine, one very simple trick to makeing fewer mistakes to to have one,
> > > and only one, unlock for every lock. 
> > 
> > Taking a spin lock twice will hard lock the machine, however on smp you
> > will typically have the luxury of an nmi watchdog which will help you
> > solve this quickly. Double unlock will oops immediately if you run with
> > spin lock debugging (you probably should, if you are developing kernel
> > code).
> 
> I have the console on a serial port, and a terminal server. With kdb,
> you can enter the kernel i kdb even when deadlocked.

Even if spinning with interrupt disabled?

-- 
Jens Axboe

Re: any chance of 2.6.0-test*?

[Terje Eggestad]

On man, 2003-01-13 at 17:26, Jens Axboe wrote:
> On Mon, Jan 13 2003, Terje Eggestad wrote:
> > On man, 2003-01-13 at 16:49, Jens Axboe wrote:
> > > On Mon, Jan 13 2003, Terje Eggestad wrote:
> > > > Considering that doing kernel development is hard enough, new
> > > > development is almost always done on uni processors kernels that do only
> > > > one thing at the time. Then when you base logic is OK, you move to a
> > > > SMP, which means (adding and) debugging you spin locks.
> > > 
> > > Goto's aside, I find the above extremely bad advise. You should _always_
> > > develop with smp and for smp from the very start, or you will most
> > > likely not get it right later on. With preempt, this becomes even more
> > > important.
> > 
> > You should, and I do, *design* with smp in mind, and I throw in
> > smplock/unlonk as I go, but I tend to make first runs on a UP. 
> > 
> > I see your point on preemt, though.
> > You do first runs on SMP? 
> 
> Always, if for nothing else than the benefit of a better debugging
> environment.
> 
> > > > Considering that fucking up spin locks are prone to corrupting your
> > > > machine, one very simple trick to makeing fewer mistakes to to have one,
> > > > and only one, unlock for every lock. 
> > > 
> > > Taking a spin lock twice will hard lock the machine, however on smp you
> > > will typically have the luxury of an nmi watchdog which will help you
> > > solve this quickly. Double unlock will oops immediately if you run with
> > > spin lock debugging (you probably should, if you are developing kernel
> > > code).
> > 
> > I have the console on a serial port, and a terminal server. With kdb,
> > you can enter the kernel i kdb even when deadlocked.
> 
> Even if spinning with interrupt disabled?

Haven't painted myself into that corner yet. Doubt it, very much.
 
-- 
_________________________________________________________________________

Terje Eggestad                  mailto:terje.eggestad@scali.no
Scali Scalable Linux Systems    http://www.scali.com

Olaf Helsets Vei 6              tel:    +47 22 62 89 61 (OFFICE)
P.O.Box 150, Oppsal                     +47 975 31 574  (MOBILE)
N-0619 Oslo                     fax:    +47 22 62 89 51
NORWAY            
_________________________________________________________________________

Re: any chance of 2.6.0-test*?

[Jens Axboe]

On Mon, Jan 13 2003, Terje Eggestad wrote:
> > > I have the console on a serial port, and a terminal server. With kdb,
> > > you can enter the kernel i kdb even when deadlocked.
> > 
> > Even if spinning with interrupt disabled?
> 
> Haven't painted myself into that corner yet. Doubt it, very much.

These are the nasty hangs, total lockup and no info at all if it wasn't
for the nmi watchdog triggering. That alone is reason enough for me :-)

-- 
Jens Axboe

Re: any chance of 2.6.0-test*?

[Zwane Mwaikambo]

On Mon, 13 Jan 2003, Jens Axboe wrote:

> On Mon, Jan 13 2003, Terje Eggestad wrote:
> > > > I have the console on a serial port, and a terminal server. With kdb,
> > > > you can enter the kernel i kdb even when deadlocked.
> > > 
> > > Even if spinning with interrupt disabled?
> > 
> > Haven't painted myself into that corner yet. Doubt it, very much.
> 
> These are the nasty hangs, total lockup and no info at all if it wasn't
> for the nmi watchdog triggering. That alone is reason enough for me :-)

It uses NMI's to break into the debugger, so it would also work with 
interrupts disabled and spinning on a lock, the same is also true for 
kgdb.

	Zwane
-- 
function.linuxpower.ca

Re: any chance of 2.6.0-test*?

[Jens Axboe]

On Mon, Jan 13 2003, Zwane Mwaikambo wrote:
> On Mon, 13 Jan 2003, Jens Axboe wrote:
> 
> > On Mon, Jan 13 2003, Terje Eggestad wrote:
> > > > > I have the console on a serial port, and a terminal server. With kdb,
> > > > > you can enter the kernel i kdb even when deadlocked.
> > > > 
> > > > Even if spinning with interrupt disabled?
> > > 
> > > Haven't painted myself into that corner yet. Doubt it, very much.
> > 
> > These are the nasty hangs, total lockup and no info at all if it wasn't
> > for the nmi watchdog triggering. That alone is reason enough for me :-)
> 
> It uses NMI's to break into the debugger, so it would also work with 
> interrupts disabled and spinning on a lock, the same is also true for 
> kgdb.

But still requiring up-apic, or smp with apic, right?

-- 
Jens Axboe

Re: any chance of 2.6.0-test*?

[Keith Owens]

On Mon, 13 Jan 2003 19:48:31 +0100, 
Jens Axboe <axboe@suse.de> wrote:
>On Mon, Jan 13 2003, Zwane Mwaikambo wrote:
>> It uses NMI's to break into the debugger, so it would also work with 
>> interrupts disabled and spinning on a lock, the same is also true for 
>> kgdb.
>
>But still requiring up-apic, or smp with apic, right?

nmi_watchdog=2 - uses performance counters, not apic.

Re: any chance of 2.6.0-test*?

[John Levon]

On Tue, Jan 14, 2003 at 09:14:52AM +1100, Keith Owens wrote:

> >But still requiring up-apic, or smp with apic, right?
> 
> nmi_watchdog=2 - uses performance counters, not apic.

which requires a local APIC to set up NMI delivery mode like Jens said

regards
john

-- 
Khendon's Law: If the same point is made twice by the same person,
the thread is over.

Re: any chance of 2.6.0-test*?

[Zwane Mwaikambo]

On Mon, 13 Jan 2003, Jens Axboe wrote:

> > It uses NMI's to break into the debugger, so it would also work with 
> > interrupts disabled and spinning on a lock, the same is also true for 
> > kgdb.
> 
> But still requiring up-apic, or smp with apic, right?

Well UP with Local APIC will suffice. So that works on a lot of i686 
machines.

	Zwane
-- 
function.linuxpower.ca

Re: any chance of 2.6.0-test*?

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 525 bytes --]

On Mon, 13 Jan 2003 17:49:12 EST, Zwane Mwaikambo said:
> On Mon, 13 Jan 2003, Jens Axboe wrote:
> 
> > > It uses NMI's to break into the debugger, so it would also work with 
> > > interrupts disabled and spinning on a lock, the same is also true for 
> > > kgdb.
> > 
> > But still requiring up-apic, or smp with apic, right?
> 
> Well UP with Local APIC will suffice. So that works on a lot of i686 
> machines.

This mean those of us that have a 'local_apic_kills_bios' flag in dmi_scan.c
are still out of luck, correct?

[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: any chance of 2.6.0-test*?

[Zwane Mwaikambo]

On Mon, 13 Jan 2003 Valdis.Kletnieks@vt.edu wrote:

> This mean those of us that have a 'local_apic_kills_bios' flag in dmi_scan.c
> are still out of luck, correct?

Yep, blacklisted systems still won't work.

	Zwane
-- 
function.linuxpower.ca

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Sun, 2003-01-12 at 19:53, Tomas Szepe wrote:
> > [robw@optonline.net]
> >
> > Am I wrong that the above would do the same thing without generating the
> > sphagetti code that a goto would give you.  Gotos are BAD, very very
> > bad.
> 
> Whom do I pay to have this annoying clueless asshole shot?
> OH MY GOD, I really can't take any more.

In evolution its

Tools->Create filter from message->Filter on sender

Add action
	Move to folder  QUAINT

and then just read it when drunk

Re: any chance of 2.6.0-test*?

[William Lee Irwin III]

[robw@optonline.net]
>> Am I wrong that the above would do the same thing without generating the
>> sphagetti code that a goto would give you.  Gotos are BAD, very very
>> bad.

On Sun, Jan 12, 2003 at 08:53:47PM +0100, Tomas Szepe wrote:
> Whom do I pay to have this annoying clueless asshole shot?
> OH MY GOD, I really can't take any more.

>From the looks of it you forgot to procmail the idiot to /dev/null

There are a lot more where he came from; you can't get rid of them
all, it'd be genocide. procmail and get it over with. It's been a sad
fact of life since they let those AOL and compuserve FPOS's on the net.


Bill

Re: any chance of 2.6.0-test*?

[Bernd Schmidt]

On Sun, 12 Jan 2003, William Lee Irwin III wrote:

> >From the looks of it you forgot to procmail the idiot to /dev/null
> 
> There are a lot more where he came from; you can't get rid of them
> all, it'd be genocide. procmail and get it over with. It's been a sad
> fact of life since they let those AOL and compuserve FPOS's on the net.

I find linux-kernel works best if you read it with a positive list in
.procmailrc rather than using it as a killfile.  This would work even
better if y'all didn't feed the trolls.


Bernd

Re: any chance of 2.6.0-test*?

[David Woodhouse]

alan@lxorguk.ukuu.org.uk said:
>  In evolution its
> Tools->Create filter from message->Filter on sender
> Add action
> 	Move to folder  QUAINT
> and then just read it when drunk 

That's fine for this particular troll, but some of the more well-known
trolls here are so good at what they do that you sometimes have to read
their messages while sober -- the errors can be subtle and well-hidden. So
much so, sometimes, that I suspect it's done on purpose.

Besides, filtering in the MUA is blatantly the wrong place to do it. The 
MTA (or MDA) should be doing the filtering.¹

--
dwmw2
¹ Yes, I'm bored of the other off-topic flames :)

RE: any chance of 2.6.0-test*?

[Folkert van Heusden]

> Am I wrong that the above would do the same thing without generating the
> sphagetti code that a goto would give you.  Gotos are BAD, very very
> bad.
TS> Whom do I pay to have this annoying clueless asshole shot?
TS> OH MY GOD, I really can't take any more.

Well well well, is that how we act these days if we see something we don't
like?

Anyway, I totally agree with Rob: if the code can be written without a goto
and produce the same efficient assembly (or better, see Robs change) it
should be written without it.
goto's are for lazy people

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Sun, 12 Jan 2003, Rob Wilkens wrote:
> 
> I'm REALLY opposed to the use of the word "goto" in any code where it's
> not needed.

I think goto's are fine, and they are often more readable than large
amounts of indentation. That's _especially_ true if the code flow isn't
actually naturally indented (in this case it is, so I don't think using
goto is in any way _clearer_ than not, but in general goto's can be quite
good for readability).

Of course, in stupid languages like Pascal, where labels cannot be 
descriptive, goto's can be bad. But that's not the fault of the goto, 
that's the braindamage of the language designer.

		Linus

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 14:38, Linus Torvalds wrote:
> I think goto's are fine

You're a relatively succesful guy, so I guess I shouldn't argue with
your style.

However, I have always been taught, and have always believed that
"goto"s are inherently evil.  They are the creators of spaghetti code
(you start reading through the code to understand it (months or years
after its written), and suddenly you jump to somewhere totally
unrelated, and then jump somewhere else backwards, and it all gets ugly
quickly).  This makes later debugging of code total hell.  

Would it be so terrible for you to change the code you had there to
_not_ use a goto and instead use something similar to what I suggested? 
Never mind the philosophical arguments, I'm just talking good coding
style for a relatively small piece of code.

If you want, but comments in your code to meaningfully describe what's
happening instead of goto labels.

In general, if you can structure your code properly, you should never
need a goto, and if you don't need a goto you shouldn't use it.  It's
just "common sense" as I've always been taught.  Unless you're
intentionally trying to write code that's harder for others to read.

-Rob

Re: any chance of 2.6.0-test*?

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 928 bytes --]

On Sun, 12 Jan 2003 14:59:57 EST, Rob Wilkens said:

> In general, if you can structure your code properly, you should never
> need a goto, and if you don't need a goto you shouldn't use it.  It's
> just "common sense" as I've always been taught.  Unless you're
> intentionally trying to write code that's harder for others to read.

Now, it's provable you never *NEED* a goto.  On the other hand, *judicious*
use of goto can prevent code that is so cluttered with stuff of the form:

        if(...) {
		...
		die_flag = 1;
		if (!die _flag) {...
		
Pretty soon, you have die_1_flag, die_2_flag, die_3_flag and so on,
rather than 3 or 4 "goto bail_now;".

The real problem is that C doesn't have a good multi-level "break" construct.
On the other hand, I don't know of any language that has a good one - some
allow "break 3;" to break 3 levels- but that's still bad because you get
screwed if somebody adds an 'if' clause....


[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: any chance of 2.6.0-test*?

[yodaiken]

On Sun, Jan 12, 2003 at 03:18:38PM -0500, Valdis.Kletnieks@vt.edu wrote:
> On Sun, 12 Jan 2003 14:59:57 EST, Rob Wilkens said:
> 
> > In general, if you can structure your code properly, you should never
> > need a goto, and if you don't need a goto you shouldn't use it.  It's
> > just "common sense" as I've always been taught.  Unless you're
> > intentionally trying to write code that's harder for others to read.
> 
> Now, it's provable you never *NEED* a goto.  On the other hand, *judicious*
> use of goto can prevent code that is so cluttered with stuff of the form:
> 
>         if(...) {
> 		...
> 		die_flag = 1;
> 		if (!die _flag) {...
> 		
> Pretty soon, you have die_1_flag, die_2_flag, die_3_flag and so on,
> rather than 3 or 4 "goto bail_now;".
> 
> The real problem is that C doesn't have a good multi-level "break" construct.

longjump. Used with good effect in the plan9 code.

Probably takes more coordination than is possible in Linux and has marginal
benefit, but it looks nice.






-- 
---------------------------------------------------------
Victor Yodaiken 
Finite State Machine Labs: The RTLinux Company.
www.fsmlabs.com  www.rtlinux.com
1+ 505 838 9109

Re: any chance of 2.6.0-test*?

[Tomas Szepe]

> [Valdis.Kletnieks@vt.edu]
> 
> The real problem is that C doesn't have a good multi-level "break" construct.
> On the other hand, I don't know of any language that has a good one - some
> allow "break 3;" to break 3 levels- but that's still bad because you get
> screwed if somebody adds an 'if' clause....

Not necessarily.  A good multilevel break/continue implementation will
only work on loop blocks, not conditionals.

-- 
Tomas Szepe <szepe@pinerecords.com>

Re: any chance of 2.6.0-test*?

[Sean Neakums]

commence  Valdis.Kletnieks@vt.edu quotation:

> The real problem is that C doesn't have a good multi-level "break"
> construct.  On the other hand, I don't know of any language that has
> a good one - some allow "break 3;" to break 3 levels- but that's
> still bad because you get screwed if somebody adds an 'if'
> clause....

Perl's facility for labelling blocks and jumping to the beginning or
end with 'next' and 'last' may be close to what you want, but I don't
know if it's ever been implemented in a language one could sensibly
use to write a kernel.

-- 
 /                          |
[|] Sean Neakums            |  Questions are a burden to others;
[|] <sneakums@zork.net>     |      answers a prison for oneself.
 \                          |

Re: any chance of 2.6.0-test*?

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 777 bytes --]

On Sun, 12 Jan 2003 20:32:01 GMT, Sean Neakums <sneakums@zork.net>  said:
> commence  Valdis.Kletnieks@vt.edu quotation:
> 
> > The real problem is that C doesn't have a good multi-level "break"
> > construct.  On the other hand, I don't know of any language that has
> > a good one - some allow "break 3;" to break 3 levels- but that's
> > still bad because you get screwed if somebody adds an 'if'
> > clause....
> 
> Perl's facility for labelling blocks and jumping to the beginning or
> end with 'next' and 'last' may be close to what you want, but I don't
> know if it's ever been implemented in a language one could sensibly
> use to write a kernel.

Hmm.. I'm not a Perl wizard.. <reads the docs> Yeah, that's what I meant,
and no, that's not a sane kernel language. ;)

[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: any chance of 2.6.0-test*?

[Randy Dunlap]

> commence  Valdis.Kletnieks@vt.edu quotation:
>
>> The real problem is that C doesn't have a good multi-level "break"
>> construct.  On the other hand, I don't know of any language that has a
>> good one - some allow "break 3;" to break 3 levels- but that's still bad
>> because you get screwed if somebody adds an 'if'
>> clause....
>
> Perl's facility for labelling blocks and jumping to the beginning or end
> with 'next' and 'last' may be close to what you want, but I don't know if
> it's ever been implemented in a language one could sensibly use to write a
> kernel.


Yes, Burroughs (who?) implemented it in System Development Language
(SDL) more than 20 years ago.  Worked very well.

~Randy  (I hope I don't find more of this thread to reply to. :(

Re: any chance of 2.6.0-test*?

[Kevin Puetz]

Valdis.Kletnieks@vt.edu wrote:

> On Sun, 12 Jan 2003 14:59:57 EST, Rob Wilkens said:
> 
>> In general, if you can structure your code properly, you should never
>> need a goto, and if you don't need a goto you shouldn't use it.  It's
>> just "common sense" as I've always been taught.  Unless you're
>> intentionally trying to write code that's harder for others to read.
> 
> Now, it's provable you never *NEED* a goto.  On the other hand,
> *judicious* use of goto can prevent code that is so cluttered with stuff
> of the form:
> 
>         if(...) {
> ...
> die_flag = 1;
> if (!die _flag) {...
> 
> Pretty soon, you have die_1_flag, die_2_flag, die_3_flag and so on,
> rather than 3 or 4 "goto bail_now;".
> 
> The real problem is that C doesn't have a good multi-level "break"
> construct. On the other hand, I don't know of any language that has a good
> one - some allow "break 3;" to break 3 levels- but that's still bad
> because you get screwed if somebody adds an 'if' clause....

perl has a pretty nice one... you can label each construct to which break is
applicable (ie, loop or whatever) then say "break foo" to escape that
struture (and any others you happen to be inside.

Re: any chance of 2.6.0-test*?

[Randy Dunlap]

> On Sun, 12 Jan 2003 14:59:57 EST, Rob Wilkens said:
>
>> In general, if you can structure your code properly, you should never need
>> a goto, and if you don't need a goto you shouldn't use it.  It's just
>> "common sense" as I've always been taught.  Unless you're
>> intentionally trying to write code that's harder for others to read.
>
> Now, it's provable you never *NEED* a goto.  On the other hand, *judicious*
> use of goto can prevent code that is so cluttered with stuff of the form:
>
>         if(...) {
> 		...
> 		die_flag = 1;
> 		if (!die _flag) {...
>
> Pretty soon, you have die_1_flag, die_2_flag, die_3_flag and so on, rather
> than 3 or 4 "goto bail_now;".

Right.

> The real problem is that C doesn't have a good multi-level "break"
> construct. On the other hand, I don't know of any language that has a good
> one - some allow "break 3;" to break 3 levels- but that's still bad because
> you get screwed if somebody adds an 'if' clause....


The one that I used in a previous life was like so.  No "while"
or "for" constructs, only "do thisloop forever" with conditionals
all being explicitly coded inside the loop(s).  All based on:
  do [loopname] [forever];
    {block};
  end [loopname];

with {block} possibly containing "undo [loopname]".
An unnamed undo just terminates the innermost loop.
Named undo's can be used to terminate any loop level.

~Randy

RE: any chance of 2.6.0-test*?

[Folkert van Heusden]

> Now, it's provable you never *NEED* a goto.  On the other hand,
*judicious*
> use of goto can prevent code that is so cluttered with stuff of the form:
>         if(...) {
> 		...
> 		die_flag = 1;
> 		if (!die _flag) {...
>
> Pretty soon, you have die_1_flag, die_2_flag, die_3_flag and so on,
> rather than 3 or 4 "goto bail_now;".

There's always the construction:

	for(;;)
	{
		/* do something */

		if (something_failed)
			break;

		/* do something */

		if (something_failed)
			break;

		...

		break;	/* the final break */
	}

	etc.

Re: any chance of 2.6.0-test*?

[David Ford]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Gotos aren't the root of code evil.  Poor design style is and poor 
preparation is the root of spaghetti code.  Use the right tool for the 
job which includes 'goto'.  Every teacher that I've met who has an 
objection to the use of gotos has been quite naive, often recommends 
avoiding pointers, frequently suggests you restrict your strings to 255 
characters or less, etc, etc.  The oddball extras vary from teacher to 
teacher, but the overall impression of the teacher makes me want to 
avoid them even with that 10 foot pole.

It's not the fault of the student for being taught wrong, but it is the 
fault of the student for ignoring possibile variances of others.  
Particularly the "worldly wise"; people who actually write code for the 
real world v.s. lectures.

David

Rob Wilkens wrote:

>On Sun, 2003-01-12 at 14:38, Linus Torvalds wrote:
>  
>
>>I think goto's are fine
>>    
>>
>
>You're a relatively succesful guy, so I guess I shouldn't argue with
>your style.
>
>However, I have always been taught, and have always believed that
>"goto"s are inherently evil.  They are the creators of spaghetti code
>(you start reading through the code to understand it (months or years
>after its written), and suddenly you jump to somewhere totally
>unrelated, and then jump somewhere else backwards, and it all gets ugly
>quickly).  This makes later debugging of code total hell.  
>
>Would it be so terrible for you to change the code you had there to
>_not_ use a goto and instead use something similar to what I suggested? 
>Never mind the philosophical arguments, I'm just talking good coding
>style for a relatively small piece of code.
>
>If you want, but comments in your code to meaningfully describe what's
>happening instead of goto labels.
>
>In general, if you can structure your code properly, you should never
>need a goto, and if you don't need a goto you shouldn't use it.  It's
>just "common sense" as I've always been taught.  Unless you're
>intentionally trying to write code that's harder for others to read.
>
>-Rob
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE+Ic4P74cGT/9uvgsRAt1QAKCb7ryMMG5iBwTefYgDB7HLuDkRngCeNSq/
M/euGkdIdXpv6IZ1Rw9ikEo=
=Yy9i
-----END PGP SIGNATURE-----

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Sun, 12 Jan 2003, Rob Wilkens wrote:
> 
> However, I have always been taught, and have always believed that
> "goto"s are inherently evil.  They are the creators of spaghetti code

No, you've been brainwashed by CS people who thought that Niklaus Wirth
actually knew what he was talking about. He didn't. He doesn't have a
frigging clue.

> (you start reading through the code to understand it (months or years
> after its written), and suddenly you jump to somewhere totally
> unrelated, and then jump somewhere else backwards, and it all gets ugly
> quickly).  This makes later debugging of code total hell.  

Any if-statement is a goto. As are all structured loops.

Ans sometimes structure is good. When it's good, you should use it.

And sometimes structure is _bad_, and gets into the way, and using a 
"goto" is just much clearer.

For example, it is quite common to have conditionals THAT DO NOT NEST.

In which case you have two possibilities

 - use goto, and be happy, since it doesn't enforce nesting

	This makes the code _more_ readable, since the code just does what 
	the algorithm says it should do.

 - duplicate the code, and rewrite it in a nesting form so that you can 
   use the structured jumps.

	This often makes the code much LESS readable, harder to maintain, 
	and bigger.

The Pascal language is a prime example of the latter problem. Because it 
doesn't have a "break" statement, loops in (traditional) Pascal end up 
often looking like total shit, because you have to add totally arbitrary 
logic to say "I'm done now".

		Linus

Re: any chance of 2.6.0-test*?

[Robert Love]

On Sun, 2003-01-12 at 15:22, Linus Torvalds wrote:

> No, you've been brainwashed by CS people who thought that Niklaus
> Wirth actually knew what he was talking about. He didn't. He
> doesn't have a frigging clue.

I thought Edsger Dijkstra coined the "gotos are evil" bit in his
structured programming push?

Nonetheless, they would both be wrong... 

	Robert Love

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On 12 Jan 2003, Robert Love wrote:
> On Sun, 2003-01-12 at 15:22, Linus Torvalds wrote:
> 
> > No, you've been brainwashed by CS people who thought that Niklaus
> > Wirth actually knew what he was talking about. He didn't. He
> > doesn't have a frigging clue.
> 
> I thought Edsger Dijkstra coined the "gotos are evil" bit in his
> structured programming push?

Yeah, he did, but he's dead, and we shouldn't talk ill of the dead. So 
these days I can only rant about Niklaus Wirth, who took the "structured 
programming" thing and enforced it in his languages (Pascal and Modula-2), 
and thus forced his evil on untold generations of poor CS students who had 
to learn langauges that weren't actually useful for real work.

(Yeah, yeah, most _practical_ versions of Pascal ended up having all the 
stuff necessary to break structure, but as you may be able to tell, I was 
one of the unwashed masses who had to write in "standard Pascal" in my 
youth. I'm scarred for life).

			Linus

Re: any chance of 2.6.0-test*?

[Horst von Brand]

Linus Torvalds <torvalds@transmeta.com> said:
> On 12 Jan 2003, Robert Love wrote:
> > On Sun, 2003-01-12 at 15:22, Linus Torvalds wrote:
> > 
> > > No, you've been brainwashed by CS people who thought that Niklaus
> > > Wirth actually knew what he was talking about. He didn't. He
> > > doesn't have a frigging clue.
> > 
> > I thought Edsger Dijkstra coined the "gotos are evil" bit in his
> > structured programming push?

> Yeah, he did, but he's dead, and we shouldn't talk ill of the dead. So 
> these days I can only rant about Niklaus Wirth, who took the "structured 
> programming" thing and enforced it in his languages (Pascal and Modula-2), 
> and thus forced his evil on untold generations of poor CS students who had 
> to learn langauges that weren't actually useful for real work.

You did not live through the horror that was FORTRAN IV, I learned
programming in it (no structure except DO, a primitive for). Sure, Pascal
strictures are overkill, but in a sense they were needed to implant a bit
of structure into the mess that usually came before.

> (Yeah, yeah, most _practical_ versions of Pascal ended up having all the 
> stuff necessary to break structure, but as you may be able to tell, I was 
> one of the unwashed masses who had to write in "standard Pascal" in my 
> youth. I'm scarred for life).

And you do break structure only when necessary, so the dosis of Pascal did
you some good after all ;-)
--
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: any chance of 2.6.0-test*?

[Eric W. Biederman]

Linus Torvalds <torvalds@transmeta.com> writes:

> On 12 Jan 2003, Robert Love wrote:
> > On Sun, 2003-01-12 at 15:22, Linus Torvalds wrote:
> > 
> > > No, you've been brainwashed by CS people who thought that Niklaus
> > > Wirth actually knew what he was talking about. He didn't. He
> > > doesn't have a frigging clue.
> > 
> > I thought Edsger Dijkstra coined the "gotos are evil" bit in his
> > structured programming push?
> 

Actually Edsger Dijkstra wrote a paper entitled "Goto considered Harmful"
He never called them evil.  And the title was more for shock value to
grab peoples attention.  For the beginners Dijkstra did not
distinguish between goto, break, return, and continue.

You can find many of his papers at:
http://www.cs.utexas.edu/users/EWD/

What he was after was simple and maintainable code, and from
everything I have read, I think he would have no major problems with
the Linux kernel code. 

In the category of maintainability, and simplicity Dijkstra made
certain the first hardware implementation of an interrupt handler he
worked with had the ability to completely save the interrupt state so
he could later return to the interrupted code.  To make writing to a
printer a non-realtime task he invented the semaphore. 

> Yeah, he did, but he's dead, and we shouldn't talk ill of the dead.

Linus I think if he was alive you would have enjoyed talking to him,
your view points seem to mesh quite well.

>  So 
> these days I can only rant about Niklaus Wirth, who took the "structured 
> programming" thing and enforced it in his languages (Pascal and Modula-2), 
> and thus forced his evil on untold generations of poor CS students who had 
> to learn langauges that weren't actually useful for real work.

Now that is something worth criticizing! 

Though standard Pascal was not unusable because of the lack of a goto,  
that was the one feature it actually had.  But I suspect a lot of
teachers failed to mention it.

> 
> (Yeah, yeah, most _practical_ versions of Pascal ended up having all the 
> stuff necessary to break structure, but as you may be able to tell, I was 
> one of the unwashed masses who had to write in "standard Pascal" in my 
> youth. I'm scarred for life).


Well you seem to be coping well.  Just a little Pascal phobia there...

Eric

Re: any chance of 2.6.0-test*?

[Terje Eggestad]

On man, 2003-01-13 at 12:09, Eric W. Biederman wrote:
> Linus Torvalds <torvalds@transmeta.com> writes:
> 
> > On 12 Jan 2003, Robert Love wrote:
> > > On Sun, 2003-01-12 at 15:22, Linus Torvalds wrote:
> > > 
> > > > No, you've been brainwashed by CS people who thought that Niklaus
> > > > Wirth actually knew what he was talking about. He didn't. He
> > > > doesn't have a frigging clue.
> > > 
> > > I thought Edsger Dijkstra coined the "gotos are evil" bit in his
> > > structured programming push?
> > 
> 
> Actually Edsger Dijkstra wrote a paper entitled "Goto considered Harmful"
> He never called them evil.  And the title was more for shock value to
> grab peoples attention.  For the beginners Dijkstra did not
> distinguish between goto, break, return, and continue.
> 
> You can find many of his papers at:
> http://www.cs.utexas.edu/users/EWD/
> 

http://www.cs.utexas.edu/users/EWD/ewd02xx/EWD215.PDF
to be exact. Reading it you can tell exactly how much of an
mathematician Dijkstra really was.  At these times It's best to keep in
mind a quote:

"I used to understand the Theory of Relativity, but then the
mathematicians got hold of it."
  -- Albert Einstein

> What he was after was simple and maintainable code, and from
> everything I have read, I think he would have no major problems with
> the Linux kernel code. 
> 

Well, in the formentioned paper he made the case that while - do, and
do-while are superfluous; we should all use recursion instead. 
Out of respect of the dead, I'm not going to say what I think of that.

Since C don't allow goto beyond it's function, most of what was
problematic of goto's aren't legal anyway. 
But I submit that those that think goto's are evil never had to deal
with smp locks.
 
Most goto's I've seen deal with error handling, and I guess much could
be "solved/hidden" with a "on_return { };" clause, but...
An on_return would be a simplified try/trow/catch that is limited to the
function. 
IMHO: a trow  that is catch'ed somewhere up the call stack, is just as
much an evil as goto. 

> >  So 
> > these days I can only rant about Niklaus Wirth, who took the "structured 
> > programming" thing and enforced it in his languages (Pascal and Modula-2), 
> > and thus forced his evil on untold generations of poor CS students who had 
> > to learn langauges that weren't actually useful for real work.
> 
> Now that is something worth criticizing! 
> 

Considering that students are bottled on Java these day, they're still
learning a language unusable for real work.

> > 
> > (Yeah, yeah, most _practical_ versions of Pascal ended up having all the 
> > stuff necessary to break structure, but as you may be able to tell, I was 
> > one of the unwashed masses who had to write in "standard Pascal" in my 
> > youth. I'm scarred for life).
> 
> 
> Well you seem to be coping well.  Just a little Pascal phobia there...
> 
> Eric
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
_________________________________________________________________________

Terje Eggestad                  mailto:terje.eggestad@scali.no
Scali Scalable Linux Systems    http://www.scali.com

Olaf Helsets Vei 6              tel:    +47 22 62 89 61 (OFFICE)
P.O.Box 150, Oppsal                     +47 975 31 574  (MOBILE)
N-0619 Oslo                     fax:    +47 22 62 89 51
NORWAY            
_________________________________________________________________________

Re: any chance of 2.6.0-test*?

[Bob Taylor]

In message <m14r8dgvbe.fsf@frodo.biederman.org>, Eric W. 
Biederman writes:

[snip]

> Though standard Pascal was not unusable because of the lack of a goto,  
> that was the one feature it actually had.  But I suspect a lot of
> teachers failed to mention it.

If my memory serves me, Niklaus Wirth invented Pascal as a 
teaching language not meant for production work. Modula, I 
think, was meant for production work. I don't care for either 
also.

[snip]

Bob

-- 
+---------------------------------------------------------------+
| Bob Taylor             Email: brtaylor@sanfelipe.com.mx       |
|---------------------------------------------------------------|
| Like the ad says, at 300 dpi you can tell she's wearing a     |
| swimsuit. At 600 dpi you can tell it's wet. At 1200 dpi you   |
| can tell it's painted on. I suppose at 2400 dpi you can tell  |
| if the paint is giving her a rash. (So says Joshua R. Poulson)|
+---------------------------------------------------------------+

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 15:22, Linus Torvalds wrote:
> No, you've been brainwashed by CS people who thought that Niklaus Wirth
> actually knew what he was talking about. He didn't. He doesn't have a
> frigging clue.

Ok, so you disagree with his style is what you're saying.  You are free
to disagree, that's your right.  We're just from different schools.

> Any if-statement is a goto. As are all structured loops.

"structured" being a key word.  I like structure in code.  Let's face
it, everyone likes code to be well structured, that's why there's the
"CodingStyle" document -- people like the code to look a certain way so
they can read it later.  Without structure, you don't have that

> Ans sometimes structure is good. When it's good, you should use it.

Ok, we agree there.

> And sometimes structure is _bad_, and gets into the way, and using a 
> "goto" is just much clearer.

I can't foresee many scenarios where this would be the case.  Someone
spoke, for example, about being indented too far and needing to jump out
because there isn't a good "break" mechanism... Well, as it says in the
coding style document -- if you're indented more then 3 levels in,
"you're screwed anyway, and should fix your program".  

In the code snippet you put up, I don't see structure as having been bad
and a goto as having been clearer.

> For example, it is quite common to have conditionals THAT DO NOT NEST.

Could someone (Linus, I know your busy, so I won't ask you -- in fact,
I'm amazed you'd bother to write to _me_ at all) please point me to
sample code that illustrates this.  I'm trying to envision "conditionals
that do not nest that need goto".  As near as I can tell, there must be
some condition that the goto nests all the conditions in otherwise the
goto wouldn't be needed.  In some cases, Maybe what's between the goto
and whats gone to should be broken up into a new function (modularized)
and called independently of function that the goto was in.

As someone else pointed out, it's provable that goto isn't needed, and
given that C is a minimalist language, I'm not sure why it was included.

> The Pascal language is a prime example of the latter problem. Because it 
> doesn't have a "break" statement, loops in (traditional) Pascal end up 
> often looking like total shit, because you have to add totally arbitrary 
> logic to say "I'm done now".

But at least the code is "readable" when you do that.  Sure it's a
little more work on the part of the programmer.  But anyone reading the
code can say "oh, so the loop exits when condition x is met", rather
than digging through the code looking for any place their might be a
goto.

Of course, I'm not about to argue that the kernel should be rewritten 
in PASCAL, nor am I interested in reinventing the wheel.  We've got a
good system here, let's just keep it good and make it better.  (Notice I
say "we" as if I've contributed something to it, whereas really I mean
we including myself as an end-user.)

-Rob

Re: any chance of 2.6.0-test*?

[Dimitrie O. Paun]

On January 12, 2003 03:48 pm, Rob Wilkens wrote:
> As someone else pointed out, it's provable that goto isn't needed, and
> given that C is a minimalist language, I'm not sure why it was included.

Rob,
You should really try to post one controversial thing at a time... :)
Anyway, all programming languages (minimalistic or not) are Turing
complete, maybe you want to program in Turing-machine language?

On a side note, I do hope you realize that Linus & Co. are considered
by smart people in the business as being some of the best. And for good
reason, too. That is to say, they are held in high esteem by a lot of
very, _very_ smart people, based on their continuous review of the
exceptional work done on the kernel. Now, doesn't it strike you a little
bit odd that you come in here and find all this obvious "problems"?
How come you are the only one seeing all thing things that any 1st year
CS student should be able to spot, yet hundreds of the brightest minds
in the business failed to notice?

-- 
Dimi.

Re: any chance of 2.6.0-test*? -> goto example

[Willy Tarreau]

Hi !

On Sun, Jan 12, 2003 at 03:48:24PM -0500, Rob Wilkens wrote:
[...] 
> > For example, it is quite common to have conditionals THAT DO NOT NEST.
> 
> Could someone (Linus, I know your busy, so I won't ask you -- in fact,
> I'm amazed you'd bother to write to _me_ at all) please point me to
> sample code that illustrates this.  I'm trying to envision "conditionals
> that do not nest that need goto".  As near as I can tell, there must be
> some condition that the goto nests all the conditions in otherwise the
> goto wouldn't be needed.  In some cases, Maybe what's between the goto
> and whats gone to should be broken up into a new function (modularized)
> and called independently of function that the goto was in.

Look at any tree walking function or macro. Try to write an efficient one
yourself, with lots of time the same tests and code to walk left, right, up or
down. Always the same thing : just move pointers around and go on to the loop.
Of course you can write it in your "clear structured" code, but I bet many
people won't understand it because of the multiple nested tests. But a simple
goto in each if statement to go to a common place is far cleaner. It's about
the same as the final instruction in the for loop.
 
> As someone else pointed out, it's provable that goto isn't needed, and
> given that C is a minimalist language, I'm not sure why it was included.

Of course it's provable. Anyway, any algorithm can also be written with a
very complex S-K program ! But if you'd know how a processor works, you'd
understand the evidence : about 1/5 of instructions is a branch (goto). The
basics of computing are test, set and branch. Your while, for, if ... all
do a mix of these. Why do you want to put the branch as a special case ? it's
as much needed as others. And the only instruction which does only branch
without test nor set is goto.

What you do with your methods, is replacing a branch with a set, test then
branch. More efficient ? I don't think so ! More readable ? I don't think so.
Linus is right, you've been brainwashed :-)

In fact, at school, thet tell you not to use goto to force you to learn how to
use useful things such as the while loop. If you had used BASIC in the past,
you'd remember how awful these programs were, not because of all the gotos, but
because since we were never told the beauty of other methods, we simply used
crappy gotos everywhere within left-justified code.
 
> > The Pascal language is a prime example of the latter problem. Because it 
> > doesn't have a "break" statement, loops in (traditional) Pascal end up 
> > often looking like total shit, because you have to add totally arbitrary 
> > logic to say "I'm done now".
> 
> But at least the code is "readable" when you do that.  Sure it's a
> little more work on the part of the programmer.

and the CPU ! Fortunately, Borland added the break and continue statements so
that their compiler could be used to produce good programs.

> But anyone reading the
> code can say "oh, so the loop exits when condition x is met", rather
> than digging through the code looking for any place their might be a
> goto.

I prefer a "goto end_of_loop" than a "end_of_loop=1" followed by nested
"if (!end_of_loop)" for each following blocks ! This way, you focus on
what the code really does and not how the programmer could trick the
compiler to tell the CPU what he wanted to really do.

> -Rob

Now, you asked for an example. Here is one. Please recode it without the
gotos, try to keep it at least as readable, and then bench it !
I'll send you public apologies if you can make it either faster and as
much readable, or as fast and much more readable. (yes, all the code
sits within a for condition, just to allow 'break' within the loop !).
As you see, I don't really mind writing awful primitives if they help
upper levels being much cleaner and efficient.

Cheers,
Willy

#define tree_foreach_destructive(__type, __root, __data, __stack, __slen)               \
    for (__slen = 0, __stack[0] = __root, __data = NULL; ({             \
        __label__ __left, __right, __again, __end;                      \
        typeof(__root) __ptr = __stack[__slen];                         \
__again:                                                                \
        __data = __ptr->data;                                           \
        if (__data != NULL) {                                           \
            __ptr->data = NULL;                                         \
            goto __end;                                                 \
        }                                                               \
        else if (__ptr->left != NULL) {                                 \
            __stack[++__slen] = __ptr = __ptr->left;                    \
            goto __again;                                               \
        }                                                               \
        else                                                            \
__left:                                                                 \
        if (__ptr->right != NULL) {                                     \
            __stack[++__slen] = __ptr = __ptr->right;                   \
            goto __again;                                               \
        }                                                               \
        else                                                            \
__right:                                                                \
        if (!__slen--)                                                  \
            goto __end; /* nothing left, don't delete the root node */  \
        else {                                                          \
            typeof (__root) __old;                                      \
            pool_free(__type, __ptr);                                   \
            __old = __ptr;                                              \
            __ptr = __stack[__slen];                                    \
            if (__ptr->left == __old) {                                 \
                /* unlink this node from its parent */                  \
                __ptr->left = NULL;                                     \
                goto __left;                                            \
            }                                                           \
            else {                                                      \
                /* no need to unlink, the parent will also die */       \
                goto __right;                                           \
            }                                                           \
        }                                                               \
__end:                                                                  \
        (__slen >= 0); /* nothing after loop */}); )

Re: any chance of 2.6.0-test*? -> goto example

[Rob Wilkens]

On Sun, 2003-01-12 at 17:48, Willy Tarreau wrote:
> Now, you asked for an example. Here is one. Please recode it without the

The example you gave me wasn't valid C.. Well, from what I can tell it
wasn't valid C, because the "for" statement had an opening "(", but no
visible closing ")" .. And the "{" looked like it was in the wrong
spot.  If you want me to have something to benchmark against, I have to
first be able to test and use your code, and if I can't use your code
it's pointless to try to make it quicker.  

If you could provide (preferrably via web download or ftp) sample source
that uses that function, it would speed up my ability to benchmark,
thereafter I could submit improved code.

-Rob

Re: any chance of 2.6.0-test*? -> goto example

[Willy Tarreau]

On Sun, Jan 12, 2003 at 07:53:57PM -0500, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 17:48, Willy Tarreau wrote:
> > Now, you asked for an example. Here is one. Please recode it without the
> 

[...]
> In fact, no where on the printed page did I
> see a closing parenthesis for your for.  Your giving me bad code to work
> from.
[...]
> The example you gave me wasn't valid C..
[...]

Well, first, I must say I've never seen such an arrogant beginner ! You even
felt the need to send me two posts, one private, then nearly the same one to
LKML. Poor guy, I gave you a piece of code to show you that your concepts were
wrong when applied to reality, and all you try to do is to gain street credit.
You will never learn because you whine before you think or search. Don't start
a code review of the kernel or you'll get lots of ennemies !

> Well, from what I can tell it wasn't valid C, because the "for" statement
> had an opening "(", but no visible closing ")" .. And the "{" looked like it
> was in the wrong spot.

>From what I can tell, you didn't even try to "info gcc" to check those wonderful
gcc extensions which allow a complex instruction block inside an expression.
The special construct ({ blabla }) is an expression which gets the value of the
last evaluated expression. Now I'm fedup with explaining to you how to access
basic information. Please unsubscribe, forget us and reboot under windows this
way (I'm sure you're already reading your mail as root) :

# lilo -u
# reboot

> If you want me to have something to benchmark against, I have to
> first be able to test and use your code, and if I can't use your code
> it's pointless to try to make it quicker.  

you'd need more code to be able to compile something. At least, I thought you
were smart enough to imagine how to rework it as you did on Linus' code, but
apparently I was wrong.

> If you could provide (preferrably via web download or ftp) sample source
> that uses that function, it would speed up my ability to benchmark,
> thereafter I could submit improved code.

I don't have all the parts of the code right at hand, and I don't know when/if
I'll send them. For the moment, just try to understand what I sent to you and
you can use this structure definition to understand better :

struct ultree {
    unsigned long low;          /* 32 bits low value of this node */
    unsigned long high;         /* 32 bits high value of this node, not used in 32 bits */
    int level;                  /* bit level of this node */
    void *data;                 /* carried data */
    struct ultree *left, *right;        /* children : left and right. NULL = leaf */
    struct ultree *up;          /* parent node. NULL = root */
};

Oh and if you have complaints about comments after the 80th column, remove them.
And if the order of the declaration doesn't match your habits, just know that
this is the order which gives me best performance, since swapping any 2 of its
members induces a loss of about 20%. You know, L1, cache line size... all the
things that your teachers didn't tell you because the story of the evil goto was
better to keep children quiet !

Bye,
Willy

Re: any chance of 2.6.0-test*? -> goto example

[Alexander Kellett]

On Mon, Jan 13, 2003 at 02:31:33AM +0100, Willy Tarreau wrote:
> Oh and if you have complaints about comments after the 80th column, remove them.
> And if the order of the declaration doesn't match your habits, just know that
> this is the order which gives me best performance, since swapping any 2 of its
> members induces a loss of about 20%. You know, L1, cache line size... all the
> things that your teachers didn't tell you because the story of the evil goto was
> better to keep children quiet !

:)

As much as I absolutely love the utility of this 
piece of code you really do have to admit that it 
_is_ rather difficult to understand :)

Is it so flawed of me to expect that some day this
code could be rewritten in an extremely clean
fashion and compilers made to do the work that
was put in to make this fast?

mvg,
Alex

-- 
"[...] Konqueror open source project. Weighing in at less than
            one tenth the size of another open source renderer"
Apple,  Jan 2003 (http://www.apple.com/safari/)

Re: any chance of 2.6.0-test*? -> goto example

[Helge Hafting]

Alexander Kellett wrote:
[...]
> As much as I absolutely love the utility of this
> piece of code you really do have to admit that it
> _is_ rather difficult to understand :)
> 
It is hard for everybody - until we get used to it.  Then it
is just another normal thing. :-)

> Is it so flawed of me to expect that some day this
> code could be rewritten in an extremely clean
> fashion and compilers made to do the work that
> was put in to make this fast?

This is not flawed - but we actually need that compiler
to exist, and become _common_ before replacing
hand-optimizations with clean code.  Feel free to join
the gcc team and make the world better...

Helge Hafting

Coding style - (Was Re: any chance of 2.6.0-test*?)

[John Bradford]

[Most of a discussion about coding style removed]

> As someone else pointed out, it's provable that goto isn't needed, and
> given that C is a minimalist language, I'm not sure why it was included.

It's worth keeping in mind, when coding in *any* language, that
concepts such as loops, functions, and any kind of 'structured'
programming, are all completely artificial concepts.

Microprocessors are essentially based around:

Loading discrete values to registers.
Loading discrete values into memory locations.
Copying a value from a register or a memory location, to another
register or a memory location.
Pushing and poping values on to and off of the stack.
Jumping to another location based on whether a register contains zero
or a non zero value.

The only thing that can really be considered anything like structured
programming is the stack, (and possibly interupts).

> But at least the code is "readable" when you do that.

Assember for a simple microprocessor such as a Z80 is arguable easier
to understand than C source code for something like the Linux kernel.

It's not the language, or how it is indented, that determines how
difficult code is to read.  It's the logic in the code.

John.

Re: any chance of 2.6.0-test*?

[Helge Hafting]

Rob Wilkens wrote:

> > Any if-statement is a goto. As are all structured loops.
> 
> "structured" being a key word.  I like structure in code.  Let's face
> it, everyone likes code to be well structured, that's why there's the
> "CodingStyle" document -- people like the code to look a certain way so
> they can read it later.  Without structure, you don't have that
> 
If you like "structure" so much, consider the fact that gotos are
used in a very structured way in the kernel.  Books won't list
goto as a "structured keyword" but that don't prevent
it from being used in a structured and diciplined manner.

"Avoid gotos" is only a rule of thumb, for those that don't
know the finer points.  Gotos have a enormous potential for abuse,
but it isn't abused in the kernel because it is used by experts who
limit themselves to only using it right.  (Well, most of the time, but
that isn't what you complain about. :-)

One cannot expect a newbie to get this right while learning the
language,
so beginners get a "don't use gotos" rule.  And therefore it takes time
before they learn to read code containing gotos too - even _good_ code.

Please note that rules of thumb aren't absolute.  There is another
well known - don't play with fire.  Generally good, but some people
have to do a lot of it anyway in order to design better fire
extinguishers
and train firefighters. 

> > Ans sometimes structure is good. When it's good, you should use it.
> 
> Ok, we agree there.
> 
> > And sometimes structure is _bad_, and gets into the way, and using a
> > "goto" is just much clearer.
> 
> I can't foresee many scenarios where this would be the case.  Someone
> spoke, for example, about being indented too far and needing to jump out
> because there isn't a good "break" mechanism... Well, as it says in the
> coding style document -- if you're indented more then 3 levels in,
> "you're screwed anyway, and should fix your program".

Sometimes using gotos makes things clearer, _for those that have some
experience with this particular way of using gotos_. 

_Of course_ it isn't the clearest to _you_, because you, as a newbie,
have noe training or experience in this.  Get experience, learn to
recognise the very limited use of goto in the kernel, and you'll
probably agree.

Consider someone who takes a C course where they don't teach "switch"
because "if" can do the same thing.  He would consider "switch"
unnecessary.  And hard to read, due to lack of experience.  And just
like
you, he would urge people to not use switch because _he_ finds it harder
to read.  That doesn't make avoiding switch the right thing to do
though.

The same applies to goto.  It is a useful thing - if used right.
[...]
> As someone else pointed out, it's provable that goto isn't needed, and
> given that C is a minimalist language, I'm not sure why it was included.

Again, switch is as unnecessary, yet nobody wants to remove it.  And
yes,
switch can be abused too (replace all ifs with switch for example.)
The rule against goto is a beginners simplification, because the real
rules for using goto are harder to express.  They are wasted on people
who struggle with "for" and pointers - and then the beginners course
ends.

> > The Pascal language is a prime example of the latter problem. Because it
> > doesn't have a "break" statement, loops in (traditional) Pascal end up
> > often looking like total shit, because you have to add totally arbitrary
> > logic to say "I'm done now".
> 
> But at least the code is "readable" when you do that.  

Readable for who, is the question.  Some limited use of goto don't take
that long to learn.  I used to simply think "urgh, a goto" but didn't
complain - it wasn't _my_ kernel.  Now I understand that kind of code
and approve.  And I shudder thinking about some code I've written for
pay - doing ten allocations in a row, testing each, and freeing
everything in a nested fashion upon possible failure.

> Sure it's a
> little more work on the part of the programmer.  
Now, if we can avoid that extra work _without_ harm, then it is 
certainly worth it, no?

> But anyone reading the
> code can say "oh, so the loop exits when condition x is met", rather
> than digging through the code looking for any place their might be a
> goto.

What you describe is the problem with undiciplined use of goto.  Kernel
code don't use gotos in any arbitrary way.  Check, and you'll see
that 99% of it it is very diciplined, using gotos in one
well-defined way.  Much of it is explicitly coded function-internal
"exceptions", a structured concept newer than C.

(Yyou can surely find some bad examples too.  Either hacks with a
special reason, or things that really ought to be cleaned up.)

 > Of course, I'm not about to argue that the kernel should be rewritten
> in PASCAL, nor am I interested in reinventing the wheel.  We've got a
> good system here, let's just keep it good and make it better. 

Making it better is interesting.  Note that performance sometimes 
overrules style, and yes - there are cases where goto wins
on raw performance.  And that might pay off if it is a very
often executed piece of code.

Also, avoiding goto at all cost is extra work for programmers
_every time_.  Learning to use gotos "right" (I.e. advanced use of C)
is one-time learning and then it is free - everytime.

The use of gotos may be a learning curve for every kernel newbie, but
so is understanding the VFS or the VM system or any other big
complicated piece.  Wether or not that piece uses gotos. :-)

Helge Hafting

Re: any chance of 2.6.0-test*?

[yodaiken]

See Brian Kernighan's explanation in 1981. Still good.

	http://bit.csc.lsu.edu/tutorial/ten-commandments/bwk-on-pascal.html


On Sun, Jan 12, 2003 at 12:22:26PM -0800, Linus Torvalds wrote:
> 
> On Sun, 12 Jan 2003, Rob Wilkens wrote:
> > 
> > However, I have always been taught, and have always believed that
> > "goto"s are inherently evil.  They are the creators of spaghetti code
> 
> No, you've been brainwashed by CS people who thought that Niklaus Wirth
> actually knew what he was talking about. He didn't. He doesn't have a
> frigging clue.
> 
> > (you start reading through the code to understand it (months or years
> > after its written), and suddenly you jump to somewhere totally
> > unrelated, and then jump somewhere else backwards, and it all gets ugly
> > quickly).  This makes later debugging of code total hell.  
> 
> Any if-statement is a goto. As are all structured loops.
> 
> Ans sometimes structure is good. When it's good, you should use it.
> 
> And sometimes structure is _bad_, and gets into the way, and using a 
> "goto" is just much clearer.
> 
> For example, it is quite common to have conditionals THAT DO NOT NEST.
> 
> In which case you have two possibilities
> 
>  - use goto, and be happy, since it doesn't enforce nesting
> 
> 	This makes the code _more_ readable, since the code just does what 
> 	the algorithm says it should do.
> 
>  - duplicate the code, and rewrite it in a nesting form so that you can 
>    use the structured jumps.
> 
> 	This often makes the code much LESS readable, harder to maintain, 
> 	and bigger.
> 
> The Pascal language is a prime example of the latter problem. Because it 
> doesn't have a "break" statement, loops in (traditional) Pascal end up 
> often looking like total shit, because you have to add totally arbitrary 
> logic to say "I'm done now".
> 
> 		Linus
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
---------------------------------------------------------
Victor Yodaiken 
Finite State Machine Labs: The RTLinux Company.
www.fsmlabs.com  www.rtlinux.com
1+ 505 838 9109

Re: any chance of 2.6.0-test*?

[Rik van Riel]

On Sun, 12 Jan 2003, Rob Wilkens wrote:

> However, I have always been taught, and have always believed that
> "goto"s are inherently evil.  They are the creators of spaghetti code

If the main flow of the code is through a bunch of hard to trace
gotos and you choose to blame the tool instead of the programmer,
I guess you could blame goto.

However, the goto can also be a great tool to make the code more
readable.  The goto statement is, IMHO, one of the more elegant
ways to code exceptions into a C function; that is, dealing with
error situations that don't happen very often, in such a way that
the error handling code doesn't clutter up the main code path.

As an example, you could look at fs/super.c::do_kern_mount()

        mnt = alloc_vfsmnt(name);
        if (!mnt)
                goto out;
        sb = type->get_sb(type, flags, name, data);
        if (IS_ERR(sb))
                goto out_mnt;

Do you see how the absence of the error handling cleanup code
makes the normal code path easier to read ?

regards,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".
http://www.surriel.com/		http://guru.conectiva.com/
Current spamtrap:  <a href=mailto:"october@surriel.com">october@surriel.com</a>

RE: any chance of 2.6.0-test*?

[Scott Robert Ladd]

Rob Wilken wrote:
> Would it be so terrible for you to change the code you had there to
> _not_ use a goto and instead use something similar to what I suggested?
> Never mind the philosophical arguments, I'm just talking good coding
> style for a relatively small piece of code.
>
> If you want, but comments in your code to meaningfully describe what's
> happening instead of goto labels.
>
> In general, if you can structure your code properly, you should never
> need a goto, and if you don't need a goto you shouldn't use it.  It's
> just "common sense" as I've always been taught.  Unless you're
> intentionally trying to write code that's harder for others to read.

I've spent some time looking through the kernel source code, getting a feel
for the style and process before attempting to contribute something of my
own. In most ways, the quality of Linux code equals or exceeds that of
commercial products I've worked on. It may not be perfect, but I'd prefer
that the maintainers focus on features and bug fixes, not religious issues.

Your attitude against "goto" is perhaps based upon an excellent but dated
article, "Goto Considered Harmful", written by Edsger W. Dijkstra, and
published by the ACM in 1968. (A recent reprint can be found at
http://www.acm.org/classics/oct95/.) As you can tell from the date, this
article predates modern programming languages and idioms; it comes from a
time when Fortran ruled, and before Fortran 77 provided significant tools
for avoiding spaghetti code.

A "goto" is not, in and of itself, dangerous -- it is a language feature,
one that directly translates to the jump instructions implemented in machine
code. Like pointers, operator overloading, and a host of other "perceived"
evils in programming, "goto" is widely hated by those who've been bitten by
poor programming. Bad code is the product of bad programmers; in my
experience, a poor programmer will write a poor program, regardless of the
availability of "goto."

If you think people can't write spaghetti code in a "goto-less" language, I
can send you some *lovely* examples to disabuse you of that notion. ;)

Used over short distances with well-documented labels, a "goto" can be more
effective, faster, and cleaner than a series of complex flags or other
constructs. The "goto" may also be safer and more intuitive than the
alternative. A "break" is a goto; a "continue" is a "goto" -- these are
statements that move the point of execution explicitly.

That said, I have used exactly two "goto" statements in all the lines of C,
C++, Fortran 95, and (yes) COBOL I've written since leaving BASIC and
Fortran IV behind. In one case, a single "goto" doubled the speed of a
time-critical application; in the other case, "goto" shortens a segment of
code by half and makes the algorithm much clearer. I would not use a goto
willy-nilly for the fun of it -- unless I was entering an obfuscated code
contest ;)

We keep lowering the bar for technical prowess, it seems; if something has
the potential to be used "wrong", high-minded designers remove the offending
syntax rather than find or train competent programmers. This is why Java
removes pointers (among other things) -- it's not that pointers aren't
useful or efficient, it's that they require discipline from programmers.

Just because something is dogma doesn't mean it is absolute truth. If
anything, dogma should be sniffed quite carefully, since it tends to be
rather rank if you get close enough. Removing goto is a religious choice,
not a technical one.

I could draw parallels with idiotic laws in general society, but this
message is already marginal for this list.

..Scott

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Professional programming for science and engineering;
Interesting and unusual bits of very free code.

RE: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 19:03, Scott Robert Ladd wrote:
> I've spent some time looking through the kernel source code, getting a feel
> for the style and process before attempting to contribute something of my
> own. 

In part that's why i haven't contributed too much of any actual code
yet.  I'm trying to get a feel for the code and the users.  I figure the
best way is to join the discussion list and follow the patches going in.
> Your attitude against "goto" is perhaps based upon an excellent but dated
> article, "Goto Considered Harmful", written by Edsger W. Dijkstra, and
> published by the ACM in 1968. (A recent reprint can be found at
> http://www.acm.org/classics/oct95/.) As you can tell from the date, this
> article predates modern programming languages and idioms; it comes from a
> time when Fortran ruled, and before Fortran 77 provided significant tools
> for avoiding spaghetti code.

It only goes to show that the core of computer science hasn't changed
over the years.  While some technology changes, the science itself stays
the same.  Much like a physicist should know the outdated theories of
relativity by Albert Einstein (made way back in the 20th century as
well) a good computer scientist should appreciate the contributions of
its founders, and have a good grasp of their core contributions.  I'm
not claiming to be an expert in either field, however.

Of course, this isn't a computer science mailing list, this is a linux
kernel mailing list, so I apologize for getting off topic.  Let's agree
to drop it here.

-Rob

[OT] Re: any chance of 2.6.0-test*?

[J Sloan]

Rob Wilkens wrote:

>Of course, this isn't a computer science mailing list, this is a linux
>kernel mailing list, so I apologize for getting off topic. 
>
But it IS a computer science mailing list.

Applied, practical, real world computer science -
This is where the rubber meets the road - and
interesting theories that don't cut the mustard
in real life are discarded.

:-)

Best Regards,

Joe

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

On 01:38, lunedì 13 gennaio 2003, Rob Wilkens wrote:
> It only goes to show that the core of computer science hasn't changed
> over the years.  While some technology changes, the science itself stays
> the same.  Much like a physicist should know the outdated theories of
> relativity by Albert Einstein (made way back in the 20th century as
> well) a good computer scientist should appreciate the contributions of
> its founders, and have a good grasp of their core contributions.  I'm
> not claiming to be an expert in either field, however.

You think Relativity is outdated ?!?  I can't beleave !


-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Randy Dunlap]

> On Sun, 2003-01-12 at 14:38, Linus Torvalds wrote:
>> I think goto's are fine
>
> You're a relatively succesful guy, so I guess I shouldn't argue with your
> style.

Good.  (although I don't know why I'm replying as this thread
is way overdone....:)

> However, I have always been taught, and have always believed that
> "goto"s are inherently evil.  They are the creators of spaghetti code (you
> start reading through the code to understand it (months or years after its
> written), and suddenly you jump to somewhere totally
> unrelated, and then jump somewhere else backwards, and it all gets ugly
> quickly).  This makes later debugging of code total hell.
>
> Would it be so terrible for you to change the code you had there to _not_
> use a goto and instead use something similar to what I suggested?  Never
> mind the philosophical arguments, I'm just talking good coding style for a
> relatively small piece of code.
>
> If you want, but comments in your code to meaningfully describe what's
               or put
> happening instead of goto labels.
>
> In general, if you can structure your code properly, you should never need a
> goto, and if you don't need a goto you shouldn't use it.  It's just "common
> sense" as I've always been taught.  Unless you're
> intentionally trying to write code that's harder for others to read.

There are goto-less languages, even Algol-like ones.
And OSes can be written in them.
Well, they just have different names for JUMP.

~Randy

Re: any chance of 2.6.0-test*?

[Christoph Hellwig]

On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> Linus,
> 
> I'm REALLY opposed to the use of the word "goto" in any code where it's
> not needed.

Who cares?

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 14:41, Christoph Hellwig wrote:
> On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> > Linus,
> > 
> > I'm REALLY opposed to the use of the word "goto" in any code where it's
> > not needed.
> 
> Who cares?
> 

I do.

-Rob

Re: any chance of 2.6.0-test*?

[David Ford]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The horrific response to the use of "goto" is deprecated in favor of 
proper use.  A function overloaded with gotos probably should be 
reworked.  But there there is no need to outright avoid the use of 
'goto'.  if()  is simply a conditional test with a goto A or goto B logic.

There is a reason for the implementation of goto.  When it all boils 
down to it, in assembler it's all a matter of JMP with or without a 
condition.

David

Rob Wilkens wrote:

>On Sun, 2003-01-12 at 14:41, Christoph Hellwig wrote:
>  
>
>>On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
>>    
>>
>>>Linus,
>>>
>>>I'm REALLY opposed to the use of the word "goto" in any code where it's
>>>not needed.
>>>      
>>>
>>Who cares?
>>
>>    
>>
>
>I do.
>
>-Rob
>  
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE+IcjV74cGT/9uvgsRAtPFAKDATzN7dkRyfRk7WXEDGyYe0oGiqACffj+X
iXkDBQntdGnHeJFKjfmp0Lg=
=l9hb
-----END PGP SIGNATURE-----

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 14:58, David Ford wrote:
> The horrific response to the use of "goto" is deprecated in favor of 
> proper use.  A function overloaded with gotos probably should be 
> reworked.  But there there is no need to outright avoid the use of 
> 'goto'.  if()  is simply a conditional test with a goto A or goto B logic.

But if the if() is already there, there is no need to have an additional
goto in the statement,  My version may even be faster than linus'
version because it may save an extra assembler line of code (the beq or
whatever branch statement generated by the if would do the branch during
the if rather than after the if was evaluated and got to the first
statement which was the goto).  Of course, compiler optimizatoins
probably make that particular point pointless to discuss.

> There is a reason for the implementation of goto.  When it all boils 
> down to it, in assembler it's all a matter of JMP with or without a 
> condition.

But in this case, the condition was already there..  My point is that if
you don't need a goto, you shouldn't use it.  Whenever there's an
obvious way not to use it, don't.

-Rob

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

> But in this case, the condition was already there..  My point is that if
> you don't need a goto, you shouldn't use it.  Whenever there's an
> obvious way not to use it, don't.

You should not code to avoid a construction that's valid C.
Just use the control structure that comes natural.
Of course "if (c % 2) goto a;" is a crime, but things
like "if (!buffer) goto err_out_nomem;" are better because
the tell you why the code jumps here, it includes an implicit
comment.
if (buffer) {
	/* do stuff */
} else {
	return -ENOMEM;
}
just obscures the logic of the code.
Or look at it another way, if you dislike goto you should
crusade against conditional return as well.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[David Ford]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Naturally. Evaluate the assembler when it's done.  Normally an 
unconditional jump takes less cycles than a conditional jump.  It's hard 
to argue with that.  Use the right tool for the job.  If the code 
expression necessitates a goto within a block of code to keep things 
neat and orderly, then do it.

Regarding the if () { ... }, there are no braces around the assembler 
code.  It's all a matter of linear instructions with jumps.  An if() { 
... } is a prettiness for our human eyes to organize code.

David

Rob Wilkens wrote:

>On Sun, 2003-01-12 at 14:58, David Ford wrote:
>  
>
>>The horrific response to the use of "goto" is deprecated in favor of 
>>proper use.  A function overloaded with gotos probably should be 
>>reworked.  But there there is no need to outright avoid the use of 
>>'goto'.  if()  is simply a conditional test with a goto A or goto B logic.
>>    
>>
>
>But if the if() is already there, there is no need to have an additional
>goto in the statement,  My version may even be faster than linus'
>version because it may save an extra assembler line of code (the beq or
>whatever branch statement generated by the if would do the branch during
>the if rather than after the if was evaluated and got to the first
>statement which was the goto).  Of course, compiler optimizatoins
>probably make that particular point pointless to discuss.
>
>  
>
>>There is a reason for the implementation of goto.  When it all boils 
>>down to it, in assembler it's all a matter of JMP with or without a 
>>condition.
>>    
>>
>
>But in this case, the condition was already there..  My point is that if
>you don't need a goto, you shouldn't use it.  Whenever there's an
>obvious way not to use it, don't.
>
>-Rob
>  
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE+IdFM74cGT/9uvgsRAuMkAJ4nKED1gEVvNKJiYqXg6J73aQgTRQCfWrrJ
zYoHzuMjZ+ikKHCQ1O9E2C4=
=PBlz
-----END PGP SIGNATURE-----

Re: any chance of 2.6.0-test*?

[Robert Love]

On Sun, 2003-01-12 at 14:34, Rob Wilkens wrote:

> I'm REALLY opposed to the use of the word "goto" in any code where
> it's not needed.  OF course, I'm a linux kernel newbie, so I'm in
> no position to comment

We use goto's extensively in the kernel.  They generate fast, tight
code.

There is nothing wrong with being a newbie, but there is something wrong
with voicing ignorance.  "Learn first, form opinions later".

When you start posting more than most of the developers here, combined,
and none of your posts are contributive kernel code, there is a problem.

	Robert Love

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

Mr. Love,

I posted a minor change to the floppy driver which no one commented on
earlier (probably in wrong form)-- three times now (once in unified diff
format).. And you might even say that the message you quoted was kernel
code that I was posting since I posted alternate code to linus' code.

As per whether I post more than the developers here, it's only because I
don't have a job, so I've got more free time than they do.  I can't help
it.  If they're busier than me, I wouldn't expect them to waste time
writing conversational messages about the kernel.  

-Rob

On Sun, 2003-01-12 at 15:31, Robert Love wrote:
> On Sun, 2003-01-12 at 14:34, Rob Wilkens wrote:
> 
> > I'm REALLY opposed to the use of the word "goto" in any code where
> > it's not needed.  OF course, I'm a linux kernel newbie, so I'm in
> > no position to comment
> 
> We use goto's extensively in the kernel.  They generate fast, tight
> code.
> 
> There is nothing wrong with being a newbie, but there is something wrong
> with voicing ignorance.  "Learn first, form opinions later".
> 
> When you start posting more than most of the developers here, combined,
> and none of your posts are contributive kernel code, there is a problem.
> 
> 	Robert Love
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: any chance of 2.6.0-test*?

[Matti Aarnio]

On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> Linus,
> 
> I'm REALLY opposed to the use of the word "goto" in any code where it's
> not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> to comment

Bob,

At first, read   Documentation/CodingStyle   of the kernel.
Then have a look into:

    fs/open.c  file    do_sys_truncate()  function.

Explain how you can do that cleanly, understandably, and without
code duplication, or ugly kludges  without using those goto ?
(And sticking to coding-style.)

Also take into account, that there the most common execution path
is direct, and exceptions are redirected elsewere.


Original "gotos are evil" comes from days of FORTRAN (and BASIC)
coding, where labels were all numeric, and definitely non-descriptive.
Where nested conditionals didn't exist, etc. Source-code nesting didn't
exist either...


Doing things in moderation will usually produce better code,
than strict sticking into some paradigm or other.


Of course in schools there is definite need for severe whaking of
sense into heads of the young would-be coders.  Clean reusable
and maintainable code is _hard_ to produce without any sort of
role-models and rules, especially when kids teach coding for
themselves by looking into other kids code of "look how clever
spaghetti I made" stuff.

When you learn the ropes (rules), and code enough (some years),
you will finally learn where you can relax the rules a bit.
Especially where the rules _must_ be altered to allow some
formerly boo-boo styles.


...
> If I'm misinterpreting the original code, then forgive me..  I just saw
> a goto and gasped.  There's always a better option than goto.

Yes there are, but we are coding in C, nor C++ or Java...

> -Rob

/Matti Aarnio

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 16:15, Matti Aarnio wrote:
> On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> > Linus,
> > 
> > I'm REALLY opposed to the use of the word "goto" in any code where it's
> > not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> > to comment
> 
> Bob,
> 
> At first, read   Documentation/CodingStyle   of the kernel.
> Then have a look into:
> 
>     fs/open.c  file    do_sys_truncate()  function.
> 
> Explain how you can do that cleanly, understandably, and without
> code duplication, or ugly kludges  without using those goto ?
> (And sticking to coding-style.)

I've only compiled (and haven't tested this code), but it should be much
faster than the original code.  Why?  Because we're eliminating an extra
"jump" in several places in the code every time open would be called. 
Yes, it's more code, so the kernel is a little bigger, but it should be
faster at the same time, and memory should be less of an issue nowadays.

Here's the patch if you want to apply it (i have only compile tested it,
I haven't booted with it).. This patch applied to the 2.5.56 kernel.

--- open.c.orig	2003-01-12 16:17:01.000000000 -0500
+++ open.c	2003-01-12 16:22:32.000000000 -0500
@@ -100,44 +100,58 @@
 
 	error = -EINVAL;
 	if (length < 0)	/* sorry, but loff_t says... */
-		goto out;
+		return error;
 
 	error = user_path_walk(path, &nd);
 	if (error)
-		goto out;
+		return error;
 	inode = nd.dentry->d_inode;
 
 	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
 	error = -EISDIR;
-	if (S_ISDIR(inode->i_mode))
-		goto dput_and_out;
+	if (S_ISDIR(inode->i_mode)){
+		path_release(&nd);
+		return error;
+	}
 
 	error = -EINVAL;
-	if (!S_ISREG(inode->i_mode))
-		goto dput_and_out;
+	if (!S_ISREG(inode->i_mode)){
+		path_release(&nd);
+		return error;
+	}
 
 	error = permission(inode,MAY_WRITE);
-	if (error)
-		goto dput_and_out;
+	if (error){
+		path_release(&nd);
+		return error;
+	}
 
 	error = -EROFS;
-	if (IS_RDONLY(inode))
-		goto dput_and_out;
+	if (IS_RDONLY(inode)){
+		path_release(&nd);
+		return error;
+	}
 
 	error = -EPERM;
-	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
-		goto dput_and_out;
+	if (IS_IMMUTABLE(inode) || IS_APPEND(inode)){
+		path_release(&nd);
+		return error;
+	}
 
 	/*
 	 * Make sure that there are no leases.
 	 */
 	error = break_lease(inode, FMODE_WRITE);
-	if (error)
-		goto dput_and_out;
+	if (error){
+		path_release(&nd);
+		return error;
+	}
 
 	error = get_write_access(inode);
-	if (error)
-		goto dput_and_out;
+	if (error){
+		path_release(&nd);
+		return error;
+	}
 
 	error = locks_verify_truncate(inode, NULL, length);
 	if (!error) {
@@ -146,9 +160,7 @@
 	}
 	put_write_access(inode);
 
-dput_and_out:
 	path_release(&nd);
-out:
 	return error;
 }

Re: any chance of 2.6.0-test*?

[Rik van Riel]

On Sun, 12 Jan 2003, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 16:15, Matti Aarnio wrote:

> > At first, read   Documentation/CodingStyle   of the kernel.
> > Then have a look into:
> >
> >     fs/open.c  file    do_sys_truncate()  function.

> -	if (S_ISDIR(inode->i_mode))
> -		goto dput_and_out;
> +	if (S_ISDIR(inode->i_mode)){
> +		path_release(&nd);
> +		return error;
> +	}

	[snip same change in a few more changes]

OK, now imagine the dcache locking changing a little bit.
You need to update this piece of (duplicated) code in half
a dozen places in just this function and no doubt in dozens
of other places all over fs/*.c.

>From a maintenance point of view, a goto to a single block
of error handling code is easier to maintain.

regards,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".
http://www.surriel.com/		http://guru.conectiva.com/
Current spamtrap:  <a href=mailto:"october@surriel.com">october@surriel.com</a>

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 16:40, Rik van Riel wrote:
> OK, now imagine the dcache locking changing a little bit.
> You need to update this piece of (duplicated) code in half
> a dozen places in just this function and no doubt in dozens
> of other places all over fs/*.c.
> 
> >From a maintenance point of view, a goto to a single block
> of error handling code is easier to maintain.
> 

There's no reason, though, that the error handling/cleanup code can't be
in an entirely separate function, and if speed is needed, there's no
reason it can't be an "inline" function.  Or am I oversimplifying things
again?

-Rob

Re: any chance of 2.6.0-test*?

[Aaron Lehmann]

On Sun, Jan 12, 2003 at 04:44:05PM -0500, Rob Wilkens wrote:
> There's no reason, though, that the error handling/cleanup code can't be
> in an entirely separate function, and if speed is needed, there's no
> reason it can't be an "inline" function.  Or am I oversimplifying things
> again?

Remind me why this is better than a goto?

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

In the case of an inline function, you're saving a jump, because the
code that you would "goto" is right there in sequence with the code you
are executing as far as the processor is concerned.  In essence, you're
duplicating code, but you're not retyping code, and your keeping code
consistent accross all uses of it (keeping it modular).

It's trivial, but where you're trying to cut down on the total number of
instructions executed in kernel mode, you would think even where you
could save one instructon (and branches are expensive, no?) you would
want to.

-Rob

On Sun, 2003-01-12 at 16:49, Aaron Lehmann wrote:
> On Sun, Jan 12, 2003 at 04:44:05PM -0500, Rob Wilkens wrote:
> > There's no reason, though, that the error handling/cleanup code can't be
> > in an entirely separate function, and if speed is needed, there's no
> > reason it can't be an "inline" function.  Or am I oversimplifying things
> > again?
> 
> Remind me why this is better than a goto?

Re: any chance of 2.6.0-test*?

[Aaron Lehmann]

On Sun, Jan 12, 2003 at 05:07:19PM -0500, Rob Wilkens wrote:
> In the case of an inline function, you're saving a jump, because the
> code that you would "goto" is right there in sequence with the code you
> are executing as far as the processor is concerned.  In essence, you're
> duplicating code, but you're not retyping code, and your keeping code
> consistent accross all uses of it (keeping it modular).

These are usually error conditions. If you inline them, you will have
to jump *over* them as part of the normal code path. You don't save
any instructions, and you end up with a kernel which has much more
duplicated code and thus thrashes the cache more. It also makes the
code harder to read. goto makes it easy "stack" error handlers and not
worry about the order in which you do clean up.

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 17:18, Aaron Lehmann wrote:
> These are usually error conditions. If you inline them, you will have
> to jump *over* them as part of the normal code path. You don't save

You're wrong.  You wouldn't have to jump over them any more than you
have to jump over the "goto" statement.  They would be where the goto
statement is.  Instead of the goto you would have the function.

> any instructions, and you end up with a kernel which has much more
> duplicated code and thus thrashes the cache more. It also makes the

If that argument was taken to it's logical conclusion (and I did, in my
mind just now), no one should add any code the grows the kernel at all.

-Rob

Re: any chance of 2.6.0-test*?

[Aaron Lehmann]

On Sun, Jan 12, 2003 at 05:34:58PM -0500, Rob Wilkens wrote:
> You're wrong.  You wouldn't have to jump over them any more than you
> have to jump over the "goto" statement.

The goto is a conditional jump. You propose replacing it with a
conditional jump past the error handling code predicated on the
opposite condition. Where's the improvement?

> > any instructions, and you end up with a kernel which has much more
> > duplicated code and thus thrashes the cache more. It also makes the
> 
> If that argument was taken to it's logical conclusion (and I did, in my
> mind just now), no one should add any code the grows the kernel at all.

Not without good reason.

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 17:52, Aaron Lehmann wrote:
> On Sun, Jan 12, 2003 at 05:34:58PM -0500, Rob Wilkens wrote:
> > You're wrong.  You wouldn't have to jump over them any more than you
> > have to jump over the "goto" statement.
> 
> The goto is a conditional jump. You propose replacing it with a
> conditional jump past the error handling code predicated on the
> opposite condition. Where's the improvement?

The goto is absolutely not a conditional jump.  The if that precedes it
is conditional.  The goto is not.  The if is already there.

-Rob

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Montag, 13. Januar 2003 00:11 schrieb Rob Wilkens:
> On Sun, 2003-01-12 at 17:52, Aaron Lehmann wrote:
> > On Sun, Jan 12, 2003 at 05:34:58PM -0500, Rob Wilkens wrote:
> > > You're wrong.  You wouldn't have to jump over them any more than you
> > > have to jump over the "goto" statement.
> >
> > The goto is a conditional jump. You propose replacing it with a
> > conditional jump past the error handling code predicated on the
> > opposite condition. Where's the improvement?
>
> The goto is absolutely not a conditional jump.  The if that precedes it
> is conditional.  The goto is not.  The if is already there.

Oh, well.
Apologies first, my assembler is rusty.

if (a == NULL)
	goto err_out;

bad compiler ->
tst $D0 ; evaluate a == NULL
bne L1 ; this is the if
bra err_out ; this is the goto
L1:

good compiler ->
tst $D0
beq err_out ; obvious optimisation

	Oliver

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Rob Wilkens">
> On Sun, 2003-01-12 at 17:52, Aaron Lehmann wrote:
>> On Sun, Jan 12, 2003 at 05:34:58PM -0500, Rob Wilkens wrote:
>> > You're wrong.  You wouldn't have to jump over them any more than you have to jump
>> over the "goto" statement.
>>
>> The goto is a conditional jump. You propose replacing it with a
>> conditional jump past the error handling code predicated on the
>> opposite condition. Where's the improvement?
>
> The goto is absolutely not a conditional jump.  The if that precedes it is
> conditional.  The goto is not.  The if is already there.
>
> -Rob


Oh my God ... really you are the only one on this list not to understand that hi has
mispelled ?!?

Now I understand the reason of your trollness !!

-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 18:39, Emiliano Gabrielli wrote:
> Now I understand the reason of your trollness !!

Enough with the personal attacks.. They get you know where, I'm not
botherred by them, and they clog the list.

-Rob

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Rob Wilkens">
> On Sun, 2003-01-12 at 18:39, Emiliano Gabrielli wrote:
>> Now I understand the reason of your trollness !!
>
> Enough with the personal attacks.. They get you know where, I'm not botherred by them,
> and they clog the list.
>
> -Rob


Ok, my apologises... what I would say is "why don't you sit for a moment and think about
what a lot of skilled coder said you ???"

What I can't understand in you is the reason of your insistence, ...

cheers :-)

-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 18:57, Emiliano Gabrielli wrote:
> Ok, my apologises... what I would say is "why don't you sit for a moment and think about
> what a lot of skilled coder said you ???"

Believe me when I say, I'm learning a lot from these discussions.  Hey,
that's what I'm here for.  And that's what the list is about in some
respect (discussion).  Call me a troll if you want, call me a cockroach
if it makes you feel better.  I don't care.  But in the end, I'm saying
what I think is right, and when I'm wrong, I'm learning the whys.  If I
still think I'm right, I'm saying so, and people explain vividly exactly
why it is that I'm wrong, and believe it or not, I do learn.  Sometimes
I learn at a different rate than others, but that's only because I have
a different background and a different set of learning disabilities.

-Rob

Re: any chance of 2.6.0-test*?

[Terje Eggestad]

On man, 2003-01-13 at 01:08, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 18:57, Emiliano Gabrielli wrote:
> > Ok, my apologises... what I would say is "why don't you sit for a moment and think about
> > what a lot of skilled coder said you ???"
> 
> Believe me when I say, I'm learning a lot from these discussions.  Hey,
I'm sure
> that's what I'm here for.  And that's what the list is about in some
> respect (discussion). 
THat's NOT why we're here. People are on this list to do work, not to be
teaching assistants. That's why You're considered disrespectful. 

>  Call me a troll if you want, call me a cockroach
> if it makes you feel better.  I don't care.  But in the end, I'm saying
> what I think is right, and when I'm wrong, I'm learning the whys.  If I
> still think I'm right, I'm saying so, and people explain vividly exactly
> why it is that I'm wrong, and believe it or not, I do learn.  Sometimes
> I learn at a different rate than others, but that's only because I have
> a different background and a different set of learning disabilities.
> 
> -Rob
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
-- 
_________________________________________________________________________

Terje Eggestad                  mailto:terje.eggestad@scali.no
Scali Scalable Linux Systems    http://www.scali.com

Olaf Helsets Vei 6              tel:    +47 22 62 89 61 (OFFICE)
P.O.Box 150, Oppsal                     +47 975 31 574  (MOBILE)
N-0619 Oslo                     fax:    +47 22 62 89 51
NORWAY            
_________________________________________________________________________

Re: any chance of 2.6.0-test*?

[Sean Neakums]

commence  Rob Wilkens quotation:

> On Sun, 2003-01-12 at 17:18, Aaron Lehmann wrote:
>> These are usually error conditions. If you inline them, you will have
>> to jump *over* them as part of the normal code path. You don't save
>
> You're wrong.  You wouldn't have to jump over them any more than you
> have to jump over the "goto" statement.  They would be where the goto
> statement is.  Instead of the goto you would have the function.

He said the *normal* path.  Jumping to error code is rarely a normal
path.  By replacing the gotos with inlined functions, you turn the
fast path into a bunch of jumps over duplicated error code, and the
error path into a straight line.

>> any instructions, and you end up with a kernel which has much more
>> duplicated code and thus thrashes the cache more. It also makes the
>
> If that argument was taken to it's logical conclusion (and I did, in my
> mind just now), no one should add any code the grows the kernel at all.

The interesting thing about taking arguments to their logical
conclusion is that it rarely makes any sense to do so.

-- 
 /                          |
[|] Sean Neakums            |  Questions are a burden to others;
[|] <sneakums@zork.net>     |      answers a prison for oneself.
 \                          |

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Sonntag, 12. Januar 2003 23:34 schrieb Rob Wilkens:
> On Sun, 2003-01-12 at 17:18, Aaron Lehmann wrote:
> > These are usually error conditions. If you inline them, you will have
> > to jump *over* them as part of the normal code path. You don't save
>
> You're wrong.  You wouldn't have to jump over them any more than you
> have to jump over the "goto" statement.  They would be where the goto
> statement is.  Instead of the goto you would have the function.

That exactly is the problem. If they are where the goto would be
they needlessly fill the CPU's pipeline, take up space in L1 and use
up bandwidth on the busses.
A goto is much shorter than a cleanup.

And you would have to jump. Any control structure will result in one
or more conditional jumps in the assembler code (or conditional instructions)

Turning "if (a) goto b;"  into a single branch instruction is trivial.
The best the compiler can do with
if (a) {
	cleanup();
	return err;
}
is putting the conditional code on the end of the function.

So in the best case the compiler can generate almost equivalent code
at a cost of maintainability.

> > any instructions, and you end up with a kernel which has much more
> > duplicated code and thus thrashes the cache more. It also makes the
>
> If that argument was taken to it's logical conclusion (and I did, in my
> mind just now), no one should add any code the grows the kernel at all.

Correct. If you mean that literally, you've grasped an important concept.
You grow the kernel only with good cause.

And you look where you add the code. Additional device drivers don't hurt.
A computed jump is still only one computed jump. Additional code in common
code paths of the core hurts a lot.
For the inner loops of core code there are two considerations, size and
reducing jumps.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[Rik van Riel]

On Sun, 12 Jan 2003, Rob Wilkens wrote:

> ... jump over the "goto" statement.

Umm, an if + goto IS a conditional jump.  A modern compiler
will simply turn the goto into a conditional jump, meaning
that the normal code path doesn't do any jumps.

cheers,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".
http://www.surriel.com/		http://guru.conectiva.com/
Current spamtrap:  <a href=mailto:"october@surriel.com">october@surriel.com</a>

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Rob Wilkens">
> On Sun, 2003-01-12 at 16:40, Rik van Riel wrote:
>> OK, now imagine the dcache locking changing a little bit.
>> You need to update this piece of (duplicated) code in half
>> a dozen places in just this function and no doubt in dozens
>> of other places all over fs/*.c.
>>
>> >From a maintenance point of view, a goto to a single block
>> of error handling code is easier to maintain.
>>
>
> There's no reason, though, that the error handling/cleanup code can't be in an
> entirely separate function, and if speed is needed, there's no reason it can't be an
> "inline" function.  Or am I oversimplifying things again?
>
> -Rob
>

you do, if you inline the code and every drive writer use this tecnique the kernel will
be much bigger don't you think ?!?

Makeing a simple function instead is quite slower I think... don't forget that goto are
used only in error recovery routines ...

You can simply build a "stack" of labels .. IMHO this is a great way to be sure of the
right order we are performing cleanup/recovery ...
-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 16:58, Emiliano Gabrielli wrote:
> you do, if you inline the code and every drive writer use this tecnique the kernel will
> be much bigger don't you think ?!?

Kernel size (footprint in memory) would grow a tad bit (not much), but
it's overall speed would also go up.  

> Makeing a simple function instead is quite slower I think... don't forget that goto are
> used only in error recovery routines ...

That wasn't the case in Torvalds' sample code which started this
thread.  That was spin-locking code, I believe.  Of course, in that
case, there was no need for an inline function or anything, but rather
restructuring of an if statement.

> You can simply build a "stack" of labels .. IMHO this is a great way to be sure of the
> right order we are performing cleanup/recovery ...

If the kernel developer "knows" always that they have to follow the
stack order when adding additional cases, it's fine.  But it is more
obvious in the patch I wrote what is going on than in the original
"goto" version of the code with the fs/open.c code someone asked me to
look at.  This makes it harder for people unfamiliar with the complete
linux kernel design in this area to get started.  Of course, I'm getting
a crash course it seems.

-Rob

Re: any chance of 2.6.0-test*?

[Olivier Galibert]

On Sun, Jan 12, 2003 at 05:12:42PM -0500, Rob Wilkens wrote:
> Kernel size (footprint in memory) would grow a tad bit (not much), but
> it's overall speed would also go up.  

No, it would go down.  The damage may not be too great thanks to GCSE,
but even that isn't sure especially if asms (via locking code) are
involved.  You really need to brush up on current computer
architectures.  Look in particular the L1 caches structures, cacheline
sizes, cache sizes and branch prediction.  And of course current
compiler technology, so that you actually know what code is generated.

  OG.

PS: I know, don't feed the troll, sorry.

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Sun, 2003-01-12 at 22:12, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 16:58, Emiliano Gabrielli wrote:
> > you do, if you inline the code and every drive writer use this tecnique the kernel will
> > be much bigger don't you think ?!?
> 
> Kernel size (footprint in memory) would grow a tad bit (not much), but
> it's overall speed would also go up.  

Wrong. Thats the trouble with programming an abstract level. People just don't
understand the real world. Cache is everything, small cache footprint is king.

Alan

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Sonntag, 12. Januar 2003 22:44 schrieb Rob Wilkens:
> On Sun, 2003-01-12 at 16:40, Rik van Riel wrote:
> > OK, now imagine the dcache locking changing a little bit.
> > You need to update this piece of (duplicated) code in half
> > a dozen places in just this function and no doubt in dozens
> > of other places all over fs/*.c.
> >
> > >From a maintenance point of view, a goto to a single block
> >
> > of error handling code is easier to maintain.
>
> There's no reason, though, that the error handling/cleanup code can't be
> in an entirely separate function, and if speed is needed, there's no
> reason it can't be an "inline" function.  Or am I oversimplifying things
> again?

Yes. Typical error cleanup looks like:
err_out:
	up(&sem);
	return err;

Releasing a lock in another function is a crime punished by slow death.
(Some might even resort to voodoo to make sure your shadow suffers
in the beyond.)
It makes code absolutely unreadable.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[Aaron Lehmann]

On Sun, Jan 12, 2003 at 11:12:41PM +0100, Oliver Neukum wrote:
> Yes. Typical error cleanup looks like:
> err_out:
> 	up(&sem);
> 	return err;
> 
> Releasing a lock in another function is a crime punished by slow death.

Not to mention that the 'return err;' statement is hard to move to an
inline function meaningfully.

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 17:18, Aaron Lehmann wrote:
> On Sun, Jan 12, 2003 at 11:12:41PM +0100, Oliver Neukum wrote:
> > Yes. Typical error cleanup looks like:
> > err_out:
> > 	up(&sem);
> > 	return err;
> > 
> > Releasing a lock in another function is a crime punished by slow death.
> 
> Not to mention that the 'return err;' statement is hard to move to an
> inline function meaningfully.

Not that hard:  You just "return functionname()" where functionname is
the name of your inline function that returns the value you want to
return.

-Rob

Re: any chance of 2.6.0-test*?

[Jens Axboe]

On Sun, Jan 12 2003, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 16:40, Rik van Riel wrote:
> > OK, now imagine the dcache locking changing a little bit.
> > You need to update this piece of (duplicated) code in half
> > a dozen places in just this function and no doubt in dozens
> > of other places all over fs/*.c.
> > 
> > >From a maintenance point of view, a goto to a single block
> > of error handling code is easier to maintain.
> > 
> 
> There's no reason, though, that the error handling/cleanup code can't be
> in an entirely separate function, and if speed is needed, there's no
> reason it can't be an "inline" function.  Or am I oversimplifying things
> again?

*plonk*

-- 
Jens Axboe

Re: any chance of 2.6.0-test*?

[David Ford]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

<sarcasm> An even worse misuse, exiting the function from random 
locations avoiding logic completion </sarcasm>

David

Rob Wilkens wrote:

> 	error = -EINVAL;
> 	if (length < 0)	/* sorry, but loff_t says... */
>-		goto out;
>+		return error;
>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQE+IeLc74cGT/9uvgsRAmVKAJ0eNNXPL9vjCdkoaSGo0V9ZCihdNQCgg5np
WKKq3YLs+Iz3mbDynww2Zk0=
=R9xZ
-----END PGP SIGNATURE-----

Re: [OffTopic] [apology] any chance of 2.6.0-test*?

[Rob Wilkens]

I'm still getting messages in my inbox about it, but I'm dropping this whole
goto thread nonsense.

Some people (ok, at least 4 on this thread) have accused me of being "an
arrogant beginner" (some by private e-mail, some with less friendly words
but with the same basic meaning)..  (about 2 or 3 others told me "not to
leave" but rather to "tone down")

I must apologize, but also comment that arrogance is not my intention.  It's
a de facto symptom of a biological illness I have.  I hope you wouldn't hold
it against me.  I am taking medications to control it (maybe not so
succssfully, but I'm up to 14 pills a day now), and apologize when it gets
in the way of the normal flow of discussions here.

It (The symptom in question) is called "Grandiosity" or "Inflated Sense of
Self Worth" depending on the reference but here's some places where you can
find the symptoms listed:

http://bipolar.about.com/library/weekly/aa980408.htm
    Refers to "Grandiose Thinking", "Inflated Self Esteem", and something
others may have seen on thist list, "Lack of insight"...

http://www.frii.com/~parrot/dsm.html
    Refers more officially to "inflated self-esteem or grandiosity", and is
listed as the first symptom under "mania"

I'd suggest reading to educate yourself on the fact that you should no more
make fun of me for being grandiose than you should make fun of a retarded
person for being slow.  It's a legitimate disabiliity, and I can actually
collect social secutiy for it (and might go that route, it pays around
$1500/month to stay home).

Anyway, my point is that if I seem all that high and mighty in my tone, I'm
not really that way in general (I'm really a nice guy).  I don't know how to
communicate that effectively, and I don't know that it matters.

I also thank those who have pointed out the arrogance in my tone if indeed
it was there.  I have a therapist appointment in three hours and will
definitely bring it up.

Please do _not_ respond to this message on the list -- or if you do, don't
quote it, it's off topic.  I just wanted to generally send an apology to the
few people who haven't already filterred me out as a result of my behavior.

-Rob

p.s. If I've gotten "black listed" (filtered,. procmailed, etc) so quickly
by so many on this list so quickly, can you imagine how quickly I've
probably (before even getting on this list) I've been blacklisted in the job
market by former employers based on my behaviors there prior to quitting
(they could not fire me for same behaviors due to the americans with
disabilities act).  A friend of mine claims he's seen my name on a corporate
blacklist.  I didn't ask for proof, but presumed he was right.  It's only
fair that social security exists to protect people like me from becoming
completely homeless, and that they offer section 8 housing for mentally ill
folk like myself whch one day (after my parents grow old and die) I will
probably have to take advantage of..

p.p.s. I'm writing from window now as anyone looking at headers might figure
out.. I know, "bad".. But I'm spending about 80% of my time in linux and 20%
in windows which I think is fair.  The thought of apologizing came to me
now, and I happenned to have been in windows, so this is where I'm writing
from.

Re: any chance of 2.6.0-test*?

[Adam Kropelin]

On Sun, Jan 12, 2003 at 04:27:30PM -0500, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 16:15, Matti Aarnio wrote:
> > On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> > > Linus,
> > > 
> > > I'm REALLY opposed to the use of the word "goto" in any code where it's
> > > not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> > > to comment
> > 
> >     fs/open.c  file    do_sys_truncate()  function.
> > 
> > Explain how you can do that cleanly, understandably, and without
> > code duplication, or ugly kludges  without using those goto ?
> > (And sticking to coding-style.)
> 
> I've only compiled (and haven't tested this code), but it should be much
> faster than the original code.  Why?  Because we're eliminating an extra
> "jump" in several places in the code every time open would be called. 

Congratulations. You've possibly increased the speed of an error path by 
an infintessimal amount at the expense of increasing the size of kernel
image and making the code harder to read and maintain. (I say "possibly"
since with caching effects you may have actually slowed the code down.)

Harder to read: The primary code path is polluted with repetative code
that has no bearing on its primary mission.

Harder to maintain: Add an extra resource allocation near the top and
now you have to hunt out every failure case and manually update them all
(with yet more duplicate code) instead of just amending the cleanup code
at the end of the function.

>  	error = -EINVAL;
>  	if (length < 0)	/* sorry, but loff_t says... */
> -		goto out;
> +		return error;

Possibly an improvement, though there are maintenance and consistency
arguments against it.

>  	/* For directories it's -EISDIR, for other non-regulars - -EINVAL */
>  	error = -EISDIR;
> -	if (S_ISDIR(inode->i_mode))
> -		goto dput_and_out;
> +	if (S_ISDIR(inode->i_mode)){
> +		path_release(&nd);
> +		return error;
> +	}

Nope. This code is precisely what the on-error-goto method is used
to avoid.

The use of goto in the kernel surprised me the first time I saw it, too.
However, rather than hurry to point out how much more I knew about C
style than the kernel hackers, I stayed quiet and made a learning
experience of it. I suggest you do the same.

--Adam

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 16:59, Adam Kropelin wrote:
> Congratulations. You've possibly increased the speed of an error path by 
> an infintessimal amount at the expense of increasing the size of kernel
> image and making the code harder to read and maintain. (I say "possibly"
> since with caching effects you may have actually slowed the code down.)

Hey, if the compiler does it's job right, I increased the speed of
something in the kernel.  And, as a kernel newbie, I'm proud of that.  I
also did it in under 12 minutes (from time stamp of message received to
time stamp of message sent after code compiled and diff'd).  

> Harder to read: The primary code path is polluted with repetative code
> that has no bearing on its primary mission.

I thought it was easier to read.  For me, I can read "ok, this condition
happens, I fail"... Or "if this other condition happens, I release my
path, then I fail"...  

Whereas the "goto out" was very unclear.  It made me page down to figure
out what was going on.  

That's the whole point.. To just browse the code.. I shouldn't have to
page down to understand what the code right in front of me is doing. 
"goto out" is unclear.  "retun error" is clear.  "path_release" seems
like a relatively plain english function name and I can guess what it
does without knowing exactly what it does.  I can also surmise that if I
go beyond a certain point in the function that I need to path_release()
the same way a non-kernel programmer might need to free memory allocated
inside of a function before returning to the calling function. 

It really is that simple.  

> Harder to maintain: Add an extra resource allocation near the top and
> now you have to hunt out every failure case and manually update them all
> (with yet more duplicate code) instead of just amending the cleanup code
> at the end of the function.

It took me 12 minutes from message received to message sent to update
the entire block of code to handle the new case.  How long do you think
it would take to make a minor modification that would only have to do a
portion of what I did?  Is it such a burden on the developer to make the
code more readable?

-Rob

Re: any chance of 2.6.0-test*?

[Tomas Szepe]

> [robw@optonline.net]
> 
> That's the whole point.. To just browse the code.. I shouldn't have to
> page down to understand what the code right in front of me is doing. 
> "goto out" is unclear.  "retun error" is clear.  "path_release" seems
> like a relatively plain english function name and I can guess what it
> does without knowing exactly what it does.  I can also surmise that if I
> go beyond a certain point in the function that I need to path_release()
> the same way a non-kernel programmer might need to free memory allocated
> inside of a function before returning to the calling function. 
> 
> It really is that simple.  

"Never argue with an idiot.  They drag you down to their level,
then beat you with experience."  -- Martin Josefsson

T.

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Rob Wilkens">
> On Sun, 2003-01-12 at 16:59, Adam Kropelin wrote:
>> Congratulations. You've possibly increased the speed of an error path by  an
>> infintessimal amount at the expense of increasing the size of kernel image and
>> making the code harder to read and maintain. (I say "possibly" since with caching
>> effects you may have actually slowed the code down.)
>
> Hey, if the compiler does it's job right, I increased the speed of something in the
> kernel.  And, as a kernel newbie, I'm proud of that.  I also did it in under 12
> minutes (from time stamp of message received to time stamp of message sent after code
> compiled and diff'd).
>
>> Harder to read: The primary code path is polluted with repetative code that has no
>> bearing on its primary mission.
>
> I thought it was easier to read.  For me, I can read "ok, this condition happens, I
> fail"... Or "if this other condition happens, I release my path, then I fail"...
>
> Whereas the "goto out" was very unclear.  It made me page down to figure out what was
> going on.
>
> That's the whole point.. To just browse the code.. I shouldn't have to page down to
> understand what the code right in front of me is doing.  "goto out" is unclear.
> "retun error" is clear.  "path_release" seems like a relatively plain english function
> name and I can guess what it does without knowing exactly what it does.

goto out_path_release is finer to you ?!?  :)

> I can also
> surmise that if I go beyond a certain point in the function that I need to
> path_release() the same way a non-kernel programmer might need to free memory
> allocated inside of a function before returning to the calling function.
>
> It really is that simple.
>
>> Harder to maintain: Add an extra resource allocation near the top and now you have
>> to hunt out every failure case and manually update them all (with yet more duplicate
>> code) instead of just amending the cleanup code at the end of the function.
>
> It took me 12 minutes from message received to message sent to update the entire block
> of code to handle the new case.  How long do you think it would take to make a minor
> modification that would only have to do a portion of what I did?  Is it such a burden
> on the developer to make the code more readable?
>

I think you have no idea of the mole of the linux kernel and the number of daily patches
the mantainers receive ...


I'm also a beginner, and me too at the very first time hated the goto (every one have
told me they are evil !!) but after aving taken a look at the kernel and reading
LinuxDevice Driver I think that the style the linux kernel is coded is cryptic for
beginners, but it is perfect !

It is a wonderful experience to browse the linux kernel sources .. and every time I
didn't understand why a thing was done in such a way .. well I discover later (with
experience) that it was done in the best way it could ever be done !!!




-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Sonntag, 12. Januar 2003 23:21 schrieb Rob Wilkens:
> On Sun, 2003-01-12 at 16:59, Adam Kropelin wrote:
> > Congratulations. You've possibly increased the speed of an error path by
> > an infintessimal amount at the expense of increasing the size of kernel
> > image and making the code harder to read and maintain. (I say "possibly"
> > since with caching effects you may have actually slowed the code down.)
>
> Hey, if the compiler does it's job right, I increased the speed of
> something in the kernel.  And, as a kernel newbie, I'm proud of that.  I
> also did it in under 12 minutes (from time stamp of message received to
> time stamp of message sent after code compiled and diff'd).

Nope you didn't.
Any compiler that can move outline code like
if (con) {
	cleanup()
	return err;
}
can easily simplify
if (con)
	goto err_out;
into a single branch. Still you have enlarged the kernel a bit.
With the same number of jumps, as a rule of thumb, the smaller
code is faster.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[romieu]

Rob Wilkens <robw@optonline.net> :
[...]
> It took me 12 minutes from message received to message sent to update
> the entire block of code to handle the new case.  How long do you think
> it would take to make a minor modification that would only have to do a
> portion of what I did?  Is it such a burden on the developer to make the
> code more readable?

Herbal colon cleanser doesn't count as an efficient drug for linux kernel
code programming.

--
Ueimor

Re: any chance of 2.6.0-test*?

[Randy Dunlap]

>> Harder to read: The primary code path is polluted with repetative code
>> that has no bearing on its primary mission.
>
> I thought it was easier to read.  For me, I can read "ok, this condition
> happens, I fail"... Or "if this other condition happens, I release my path,
> then I fail"...

over and over, with additions each time?

> Whereas the "goto out" was very unclear.  It made me page down to figure out
> what was going on.
>
> That's the whole point.. To just browse the code.. I shouldn't have to page
> down to understand what the code right in front of me is doing.  "goto out"
> is unclear.  "retun error" is clear.  "path_release" seems like a relatively
> plain english function name and I can guess what it does without knowing
> exactly what it does.  I can also surmise that if I go beyond a certain
> point in the function that I need to path_release() the same way a
> non-kernel programmer might need to free memory allocated inside of a
> function before returning to the calling function.
So you choose to agree with your lessons that goto is bad, but
ignore the other lesson that functions shouldn't have multiple
return paths?  Or didn't have that lesson?
~Randy

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 20:06, Randy Dunlap wrote:
> So you choose to agree with your lessons that goto is bad, but
> ignore the other lesson that functions shouldn't have multiple
> return paths?  Or didn't have that lesson?
> ~Randy

I didn't have a lesson about multiple return paths.. I actually before
posting checked out the coding style document..

:/usr/src/linux-2.5.56/Documentation# grep -i return CodingStyle

Returned nothing..  There was no mention of using return in the middle
or end of a function in the coding style document, though perhaps there
should be.

-Rob

Re: any chance of 2.6.0-test*?

[Nuno Monteiro]

On 13.01.03 01:21 Rob Wilkens wrote:
> 
> I didn't have a lesson about multiple return paths.. I actually before
> posting checked out the coding style document..
> 
> :/usr/src/linux-2.5.56/Documentation# grep -i return CodingStyle
> 
> Returned nothing..  There was no mention of using return in the middle
> or end of a function in the coding style document, though perhaps there
> should be.
> 

You dont realistically expect to learn how to program reading 
Documentation/CodingStyle, do you? CodingStyle merely serves as general 
guidelines of what is considered good practice among kernel development, 
no more no less. If you dont know how to code in the first place (which 
you have proved spectacularly, by now), you're SOL.

Go educate yourself and stop trolling. Somewhere, along the bucketloads 
of vomit you insisted in throwing here, you asked who were the other 
top-rank trolls -- sincerely, and as far as I can remember, you're ahead 
by a fair margin. Not even imel96@trustix.co.id with his delireous 
"single user linux" crap came anywhere near you, although he was, too, 
very resistant to being clued in.

See things from this end -- when 99% of the people tell you you're wrong, 
consider the possibility you are _actually_ wrong, and not the other way 
around.

Now, off to ~/.killfile you go.


Regards,


	Nuno

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 20:51, Nuno Monteiro wrote:
> You dont realistically expect to learn how to program reading 
> Documentation/CodingStyle, do you? CodingStyle merely serves as general 

No, but I would expect rules of kernel coding to be in there.  For
example, we were talking about using return mid-function.  I always
considerred that perfectly OK.  Others were saying that's a no-no.  If
it's a no-no, it should be documented.  That is, if it's a no-no.  That
may be a situational thing though.

> See things from this end -- when 99% of the people tell you you're wrong, 
> consider the possibility you are _actually_ wrong, and not the other way 
> around.

99% of the people are running a microsoft based OS (It has 95% of the
market in terms of sales, but factor in pirated copies and your figure
will get close), of the 0.5% running linux, 0.025% of them are probably
subscribed to this mailing list.  Of that, probably 5% read it
regularly.  Now, of those 5% (of 0.25% of 0.5%), I'm probably getting
5-10 people reacting negatively (the rest are neutral reactions or no
reactions), and claiming to speak (in a rather -- shall we say --
arrogant and self-important manor) for the list as a whole.  

> Now, off to ~/.killfile you go.

Yeah, yeah, should've done it before..

-Rob

p.s. Yes, all statistics are made up on the spot for comical purposes
only and are only half-way intended to represent the real world.  p.p.s.
given those percentages, 5-10 people may be a lot percentagewise, I
don't know, I haven't done the math.

Re: any chance of 2.6.0-test*?

[Marcelo Pacheco]

>
>
>99% of the people are running a microsoft based OS (It has 95% of the
>market in terms of sales, but factor in pirated copies and your figure
>will get close), of the 0.5% running linux, 0.025% of them are probably
>subscribed to this mailing list.  Of that, probably 5% read it
>regularly.  Now, of those 5% (of 0.25% of 0.5%), I'm probably getting
>5-10 people reacting negatively (the rest are neutral reactions or no
>reactions), and claiming to speak (in a rather -- shall we say --
>arrogant and self-important manor) for the list as a whole.
>
No Rob, most people don't bother responding because they read the FAQ 
for the list and they realize that this is waste of bandwidth of a lot 
of people that are important to Linux. You're wasting their times and 
this is why this should stop NOW.

You're soo arrogant you don't want to see how much YOU'RE wrong ?
See, I'm a computer geek that started at 11, that learned everything 
from the practical viewpoint (first assembly then Basic then structured 
language) and even going through structured programming classes 7 years 
latter they still managed to get brainwashed to the point that just now 
I realize that goto's are really a good thing, and that it should be 
used as standard practice (college has been 10+ years ago). So open your 
mind to the fact that you were brainwashed, and the fact that they 
brainwash so many people that aren't smart enough to question them, that 
everything is taken at face value.

And Linux is not a hobbist thing anymore. From the point a lot of people 
trust their businesses to Linux like I do everyday (I'm MIS with about 
12 Penguins, with about 30 more Microshits to be converted to Penguins), 
then you should respect people here. Me and I'm sure a lot others read 
this list daily, and they don't want to spend their time on this thread. 
I apreciate very much the work of everybody here, considering that they 
could probably make a hell of a lot more money through other ways, but 
they decided to go the open source way, and the whole world's society is 
benefiting from their work, instead of mostly benefiting big, greedy 
companies like microshit.

By the way, you're not being paid my Microshit or other companies that 
have a lot to benefit from stalling Linux development to troll this 
list, aren't you ?

Marcelo Pacheco

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 21:51, Marcelo Pacheco wrote:
> No Rob, most people don't bother responding because they read the FAQ 
> for the list and they realize that this is waste of bandwidth of a lot 
> of people that are important to Linux. You're wasting their times and 
> this is why this should stop NOW.

Actually, we were talking in private e-mail, so if we were discussing
the FAQ, you would've continued this there, but since you've decided to
bring this discussion out here, I'll continue it here for now..

I have unlimitted bandwidth, as do many others.  IF they don't they
shouldn't subscribe to this list, they should subscribe to the newsgroup
and pick-and-choose by sbject/thread.

> You're soo arrogant you don't want to see how much YOU'RE wrong ?

I'm perfectly willing to see how much I'm wrong.  No one has yet shown
me that I'm wrong, though.  I'd be perfectly willing to see it if it
were shown.

> See, I'm a computer geek that started at 11, that learned everything 

I have a head start on you then, I started around age 8.  TI-99/4A & the
BASIC programming language back then (I even had a voice box for the
computer so I was writing programs that could talk which I thought was
the coolest thing.)  Moved on to the Commodore 64, 128, then moved into
the PC world kind of late, in 1990 I bought an XT 10MHz machine that
happenned to work with the commodore 128 monitor, and since then have
incrementally upgrade part by part to my current machine (haven't
upgraded in three years).

> mind to the fact that you were brainwashed, and the fact that they 
> brainwash so many people that aren't smart enough to question them, that 
> everything is taken at face value.

If I'm questioning people on this list, what makes you think I wasn't
smart enough to question them then??  Think about that one for a second?

> And Linux is not a hobbist thing anymore. From the point a lot of people 
> trust their businesses to Linux like I do everyday (I'm MIS with about 
> 12 Penguins, with about 30 more Microshits to be converted to Penguins), 

Serverwise or backofficewise, penguins have their place.  On the
desktop, penguins have to learn to adapt to a different climate.  That's
off topic for the kernel list though.

> then you should respect people here. Me and I'm sure a lot others read 
> this list daily, and they don't want to spend their time on this thread. 

Then don't read it, skip over to the next message and move on with
life.  As someone else put it, life is too short.

> I apreciate very much the work of everybody here, considering that they 
> could probably make a hell of a lot more money through other ways, but 
> they decided to go the open source way, and the whole world's society is 
> benefiting from their work, instead of mostly benefiting big, greedy 
> companies like microshit.

Few would be doing it if there weren't SOME benefit for themselves. 
That benefit _may_ be a better system for themselves.. Some may be the
intellectual satisfaction of a job well done.. Some may work for
companies like red hat or suse.. Some may work for companies that want
to replace old UNIX systems with cheaper PC hardware that does the same
thing...  Everyone has a motivating factor.

> By the way, you're not being paid my Microshit or other companies that 
> have a lot to benefit from stalling Linux development to troll this 
> list, aren't you ?

The answer to this is that I'm not in any way shape or form employed by
any company what so ever at present.  I'm considering part time work,
but will not say more about that because the last time I mentioned
something about an employer on a list where people thought I was a troll
(alt.support.depression.manic around 2000 when I was unmedicated) they
were sending faxes and calling the office where I worked at with
complaints.  I'd rather not forewarn any employer that I'm a little
off-center, though I guess it wouldn't hurt my chances of eventually
getting on social security disability which would cover my bills.

-Rob

Re: any chance of 2.6.0-test*?

[Billy Rose]

At 12:51 AM 1/13/2003 -0200, Marcelo Pacheco wrote:
>And Linux is not a hobbist thing anymore. From the point a lot of people 
>trust their businesses to Linux like I do everyday (I'm MIS with about 12 
>Penguins, with about 30 more Microshits to be converted to Penguins), then 
>you should respect people here. Me and I'm sure

20,000+ boxes here brother. constructive argument helps to drive 
innovation. this list was founded upon it, was it not?

br 

Re: any chance of 2.6.0-test*?

[Jochen Striepe]

    Hi,

On 12 Jan 2003, Rob Wilkens wrote:
> 
> [...] Now, of those 5% (of 0.25% of 0.5%), I'm probably getting
> 5-10 people reacting negatively (the rest are neutral reactions or no
> reactions), [...]

You don't think people might be silent because they think everything
relevant was said by those 5-10 people? 

Please do assume long-time kernel hackers *know* what they do. Asking
of course is OK, but please do not consider anything they do plain
wrong just because you don't understand after some seconds of thinking.
If you imply they're too dumb to do their job, you *will* get negative
reactions, like it or not.

If you think there's not enough documentation, try the web (including
the mailing list archive) or some kernel hacking books. If some things
are *not* documented at all (that is, you don't find the proper
documentation), ask and write some documentation yourself or ask
yourself were to get a grip on techniques your not familiar with.


Thanks,

Jochen.

-- 
The sequence is important. People just learning to pee sometimes
put the steps in the wrong order, with frustrating results.
Maybe a checklist would be useful to those just starting out.
  -- Brien K. Meehan; Msg-ID <tkolgcfimtn387@news.supernews.com>

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 22:06, Jochen Striepe wrote:
> If you imply they're too dumb to do their job, you *will* get negative
> reactions, like it or not.

I never implied anyone was too dumb to do their jobs.  IF that was read
into in any of my messages, I apologize.

-Rob

Re: any chance of 2.6.0-test*?

[Matti Aarnio]

On Sun, Jan 12, 2003 at 09:19:28PM -0500, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 20:51, Nuno Monteiro wrote:
> > You dont realistically expect to learn how to program reading 
> > Documentation/CodingStyle, do you? CodingStyle merely serves as general 
> 
> No, but I would expect rules of kernel coding to be in there.  For
> example, we were talking about using return mid-function.  I always
> considerred that perfectly OK.  Others were saying that's a no-no.  If
> it's a no-no, it should be documented.  That is, if it's a no-no.  That
> may be a situational thing though.


That document is short for several reasons.

- It has been written for particular needs, and updated rarely

- Most long-time kernel coders really don't need detailed guidelines,
  although newcomers (like yourself) apparently want something..



Things you need to understand include at least:

- Linux kernel is one of the most complicated codes that gcc compiler
  is ever put to work on, and over the years we have unearthed tons of
  optimizer bugs.    Many of somewhat odd looking coding details do
  originate from circumventions of those compiler bugs.

- Advanced optimizer hinting features, like   unlikely()  attribute
  are very new in (this) compiler, and while they in theory move the 
  "unlikely" codes out of the fast-path, such things have been buggy
  in the past, and we are worried of bug effects...

- Why code in a manner, where the optimizer needs to work very hard to
  produce fast code, when simplified structures, like those with 'goto'
  in them, can result the same with much less effort ?

- Knowing effects of those optimizations requires profiling, and studying
  produced assembly code on different target systems.

- Linux is multi-CPU-architecture kernel, looking at it via "intel-only,
  lattest Pentium-XIV" viewpoint is not going to cover all cases.  Making
  changes that bloat code will make system that much harder to fit into
  smaller embedded machines.

- Sometimes you need to code differently to achieve higher performance
  at different processors.  See   drivers/md/xor.c   and header files
  it includes.  Especially all  include/asm-*/xor.h  files, including
  include/asm-generic/xor.h ...   (That shows to you how different
  approaches work at different systems.  What works fast in register-
  rich processors does not necessarily be fast in register-poor things,
  like i386 series.)

- Adding function classifiers like "static inline" do change rules 
  somewhat. (Especially things like "mid-function return".)

- Larger code, either with manually replicated exit cleanup (which
  are pain to maintain, when some large-scale change is needed), or
  via excessive use of "static inline", will pressure caches, and
  makes it more likely to drop soon to be again needed code (or data)
  from cache.    Cases where cache-miss costs your tens, or hundreds
  of processor clock cycles that is really important question.

- Explaining things like GCC asm() statement syntaxes isn't a thing
  for Linux kernel documentation.  (Those vary from target processor
  to another!)


I repeat:

To understand effects of various coding details, you need to study produced
assembly code, and keep in mind that in modern processors a branch out of
straight-line execution path (after optimizations) is expensive (in time),
and cache-misses are even more expensive.

Then you need to benchmark those codes with different loads (where 
applicable) to find out what caches affect them -- and when you must
bypass caches (they are limited resource, after all) in order to give
overall improved system performance, when your code does not cause
actively used code or data to be evicted from caches unnecessarily.


> -Rob

/Matti Aarnio

Re: any chance of 2.6.0-test*?

[Ryan Anderson]

On Sun, Jan 12, 2003 at 08:21:50PM -0500, Rob Wilkens wrote:
> :/usr/src/linux-2.5.56/Documentation# grep -i return CodingStyle
> 
> Returned nothing..  

As annoying as this thread is, there are occassional valid points.

A first pass at an addition:

--- local/Documentation/CodingStyle.orig	2002-11-19 03:02:32.000000000 -0500
+++ local/Documentation/CodingStyle	2003-01-12 22:44:39.000000000 -0500
@@ -264,3 +264,27 @@
 
 Remember: if another thread can find your data structure, and you don't
 have a reference count on it, you almost certainly have a bug.
+
+		Chapter 9: Error handling
+
+Error handling in functions needs to follow a few simple rules.  If the
+function has allocated resources (irqs, memory), taken a reference, 
+grabbed locks, etc, all of those allocations must be reversed when 
+returning an error.
+
+In functions that depend on several allocations, the preferred way to 
+return the error is with with the use of a few "goto"s that point at an
+error block at the end of the function, after the normal, successful 
+return.
+
+Several consecutive lables can be used, reversing the order of the 
+allocations.  For example, if memory is allocated, a lock taken, and 
+an irq activated, the error labels might be labeled "err_noirq", 
+"err_nolock", "err_nomem", in order.  The final step would be to
+return the error.
+
+The reason for this style is very simple: Multiple return paths from
+the same block of code are extremely confusing, and make verification
+that locks are balanced, memory is properly freed, and that reference
+counts are not leaked very difficult.
+


-- 

Ryan Anderson
  sometimes Pug Majere

Re: any chance of 2.6.0-test*?

[Billy Rose]

At 04:59 PM 1/12/2003 -0500, Adam Kropelin wrote:
>The use of goto in the kernel surprised me the first time I saw it, too.
>However, rather than hurry to point out how much more I knew about C
>style than the kernel hackers, I stayed quiet and made a learning
>experience of it. I suggest you do the same.
>
>--Adam

every article i have read (an agreed with) on operating system internals,
has stated that the goto is considered taboo in C... _except_ in operating
system code where speed is critical, and goto can increase readability.
i taught myself how to program by single stepping through assembly
code on a DOS machine. i have found that too much structure leads
not to modular code that is easy to maintain, but a nightmare of
conditionals that cause your head to throb trying to trace the state
of an execution. being that i am partial to assembly, perhaps i am
biased in the matter.

br 

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

> I've only compiled (and haven't tested this code), but it should be much
> faster than the original code.  Why?  Because we're eliminating an extra
> "jump" in several places in the code every time open would be called.
> Yes, it's more code, so the kernel is a little bigger, but it should be
> faster at the same time, and memory should be less of an issue nowadays.
>
> Here's the patch if you want to apply it (i have only compile tested it,
> I haven't booted with it).. This patch applied to the 2.5.56 kernel.
>
> --- open.c.orig	2003-01-12 16:17:01.000000000 -0500
> +++ open.c	2003-01-12 16:22:32.000000000 -0500
> @@ -100,44 +100,58 @@
>
>  	error = -EINVAL;
>  	if (length < 0)	/* sorry, but loff_t says... */
> -		goto out;
> +		return error;

Please don't do such things. The next time locking is changed and a lock
is needed here, some poor guy has to go through that and change all
back to goto.
This may not be applicable here, but as a general rule, don't do it.
I speak from experience.

As for efficiency, that is the compiler's job.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 17:06, Oliver Neukum wrote:
> Please don't do such things. The next time locking is changed and a lock
> is needed here, some poor guy has to go through that and change all
> back to goto.
> This may not be applicable here, but as a general rule, don't do it.
> I speak from experience.
> 
> As for efficiency, that is the compiler's job.

I say "please don't use goto" and instead have a "cleanup_lock" function
and add that before all the return statements..  It should not be a
burden.  Yes, it's asking the developer to work a little harder, but the
end result is better code.

-Rob

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Sonntag, 12. Januar 2003 23:22 schrieb Rob Wilkens:
> On Sun, 2003-01-12 at 17:06, Oliver Neukum wrote:
> > Please don't do such things. The next time locking is changed and a lock
> > is needed here, some poor guy has to go through that and change all
> > back to goto.
> > This may not be applicable here, but as a general rule, don't do it.
> > I speak from experience.
> >
> > As for efficiency, that is the compiler's job.
>
> I say "please don't use goto" and instead have a "cleanup_lock" function
> and add that before all the return statements..  It should not be a
> burden.  Yes, it's asking the developer to work a little harder, but the
> end result is better code.

It's code that causes added hardship for anybody checking the locking.
It becomes impossible to see whether locks are balanced and turns into
a nightmare as soon as you need error exits from several depths of locking
or with and without memory to be freed.

Please listen to experience.

err_out_buffer:
	kfree(buffer);
err_out_nobuffer:
	up(&sem);
err_out_nolock:
	return err;

is pretty much the only readable construction.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
> It's code that causes added hardship for anybody checking the locking.

I can certainly see where it would seem "easier" to match "one lock" to
"one unlock" if your troubleshooting a locking issue.

"easier" is the motivation behind using goto.. Laziness.. that's all it
is.

-Rob

Re: any chance of 2.6.0-test*?

[Robert Love]

On Sun, 2003-01-12 at 17:51, Rob Wilkens wrote:

> "easier" is the motivation behind using goto.. Laziness.. that's
> all it is.

And faster, cleaner, simpler, and smaller.  Yup.

In the history of trolls on lkml, you are rising to the top.  Are you
proud?

	Robert Love

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 18:00, Robert Love wrote:
> In the history of trolls on lkml, you are rising to the top.  Are you
> proud?

Let's see, I subscribed Thursday, it's now Sunday.. That's 3 or 4 days
on the list.. I'm impressed if I've shot myself to the top of the entire
troll list that quickly.

Are there any other famous trolls to know about? :-)  Does labelling or
classifying someone as a troll make it any easier to dismiss what
they're saying?

-Rob

Re: any chance of 2.6.0-test*?

[Nicolas Pitre]

On Sun, 12 Jan 2003, Robert Love wrote:

> On Sun, 2003-01-12 at 17:51, Rob Wilkens wrote:
> 
> > [...]
> 
> And faster, cleaner, simpler, and smaller.  Yup.
> 
> In the history of trolls on lkml, you are rising to the top.  Are you
> proud?

What amazes me the most is the ability this guy has to have people generate
replies.  Even Linus got caught.  That's nearly better than RMS postings so 
far...


Nicolas

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Rob Wilkens">
> On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
>> It's code that causes added hardship for anybody checking the locking.
>
> I can certainly see where it would seem "easier" to match "one lock" to "one unlock"
> if your troubleshooting a locking issue.
>
> "easier" is the motivation behind using goto.. Laziness.. that's all it is.

In italy we say "Non c'è peggior sordo di chi non vuole sentire"

Till to this point the thread was useful to beginners becouse of a lot of smarty coders
explained a point that everyone has heard but that rarely knows at deep...

but now Let's ignore him ! He is not a beginner he is a lamer!


-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Montag, 13. Januar 2003 00:05 schrieb Emiliano Gabrielli:
> <quote who="Rob Wilkens">
>
> > On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
> >> It's code that causes added hardship for anybody checking the locking.
> >
> > I can certainly see where it would seem "easier" to match "one lock" to
> > "one unlock" if your troubleshooting a locking issue.
> >
> > "easier" is the motivation behind using goto.. Laziness.. that's all it
> > is.
>
> In italy we say "Non c'è peggior sordo di chi non vuole sentire"

There's no worse fate than that of those who don't want to listen?
(My Italian is bad.)

> Till to this point the thread was useful to beginners becouse of a lot of
> smarty coders explained a point that everyone has heard but that rarely
> knows at deep...
>
> but now Let's ignore him ! He is not a beginner he is a lamer!

Very well. What could be said has been said. But please don't insult him.
Time spent explaining is usually well spent. Perhaps somebody learned.

OK, maybe  I have a soft spot.

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Oliver Neukum">
> Am Montag, 13. Januar 2003 00:05 schrieb Emiliano Gabrielli:
>> <quote who="Rob Wilkens">
>>
>> > On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
>> >> It's code that causes added hardship for anybody checking the locking.
>> >
>> > I can certainly see where it would seem "easier" to match "one lock" to "one
>> unlock" if your troubleshooting a locking issue.
>> >
>> > "easier" is the motivation behind using goto.. Laziness.. that's all it is.
>>
>> In italy we say "Non c'è peggior sordo di chi non vuole sentire"
>
> There's no worse fate than that of those who don't want to listen? (My Italian is
> bad.)

It is better than my english :-))

>
>> Till to this point the thread was useful to beginners becouse of a lot of smarty
>> coders explained a point that everyone has heard but that rarely knows at deep...
>>
>> but now Let's ignore him ! He is not a beginner he is a lamer!
>
> Very well. What could be said has been said. But please don't insult him. Time spent
> explaining is usually well spent. Perhaps somebody learned.
>

I agree .. my apologises .. I use currently the word "lamer" in IRC, but I don't really
know the true translation.. if it is an insult .. plz adcept my apologises

> OK, maybe  I have a soft spot.
>

:-)

> 	Regards
> 		Oliver


-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Emiliano Gabrielli]

<quote who="Oliver Neukum">
> Am Montag, 13. Januar 2003 00:05 schrieb Emiliano Gabrielli:
>> <quote who="Rob Wilkens">
>>
>> > On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
>> >> It's code that causes added hardship for anybody checking the locking.
>> >
>> > I can certainly see where it would seem "easier" to match "one lock" to "one
>> unlock" if your troubleshooting a locking issue.
>> >
>> > "easier" is the motivation behind using goto.. Laziness.. that's all it is.
>>
>> In italy we say "Non c'è peggior sordo di chi non vuole sentire"
>
> There's no worse fate than that of those who don't want to listen? (My Italian is
> bad.)

It is better than my english :-))

>
>> Till to this point the thread was useful to beginners becouse of a lot of smarty
>> coders explained a point that everyone has heard but that rarely knows at deep...
>>
>> but now Let's ignore him ! He is not a beginner he is a lamer!
>
> Very well. What could be said has been said. But please don't insult him. Time spent
> explaining is usually well spent. Perhaps somebody learned.
>

I agree .. my apologises .. I use currently the word "lamer" in IRC, but I don't really
know the true translation.. if it is an insult .. plz adcept my apologises

> OK, maybe  I have a soft spot.
>

:-)

> 	Regards
> 		Oliver


-- 
Emiliano Gabrielli

dip. di Fisica
2° Università di Roma "Tor Vergata"

Re: any chance of 2.6.0-test*?

[Horst von Brand]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 655 bytes --]

Oliver Neukum <oliver@neukum.name> said:
> Am Montag, 13. Januar 2003 00:05 schrieb Emiliano Gabrielli:

[...]

> > In italy we say "Non c'è peggior sordo di chi non vuole sentire"

I guess in Chile it's "No hay peor sordo que el que no quiere oir".

> There's no worse fate than that of those who don't want to listen?

There is no worse deaf than the one who refuses to listen.
--
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Sonntag, 12. Januar 2003 23:51 schrieb Rob Wilkens:
> On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
> > It's code that causes added hardship for anybody checking the locking.
>
> I can certainly see where it would seem "easier" to match "one lock" to
> "one unlock" if your troubleshooting a locking issue.
>
> "easier" is the motivation behind using goto.. Laziness.. that's all it
> is.

Yes, exactly. Lazyness is good. Any second spend on hitting Ctrl+F
and typing in the name of a cleanup function is lost. Any time
you overlook something because the cleanup and the code are
not on the same screen is a bug that is overlooked.
Any time one of the cleanup functions is overlooked can cause
hours lost in debugging after a change.

	Oliver

Re: any chance of 2.6.0-test*?

[Ryan Anderson]

On Sun, Jan 12, 2003 at 05:51:57PM -0500, Rob Wilkens wrote:
> On Sun, 2003-01-12 at 17:43, Oliver Neukum wrote:
> > It's code that causes added hardship for anybody checking the locking.
> 
> I can certainly see where it would seem "easier" to match "one lock" to
> "one unlock" if your troubleshooting a locking issue.
> 
> "easier" is the motivation behind using goto.. Laziness.. that's all it
> is.

Have you ever read "Programming Perl"?

Laziness is a virtue.  (At least I think that's the right book -
possibly the 3rd edition.)

-- 

Ryan Anderson
  sometimes Pug Majere

Re: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 21:00, Ryan Anderson wrote:
> On Sun, Jan 12, 2003 at 05:51:57PM -0500, Rob Wilkens wrote:
> > "easier" is the motivation behind using goto.. Laziness.. that's all it
> > is.
> 
> Have you ever read "Programming Perl"?
> 
> Laziness is a virtue.  (At least I think that's the right book -
> possibly the 3rd edition.)

Actually, it's one of the few books I actually still own (the perl
one)..  Laziness is indeed a virtue..  One I live my life by.  Of
course, that's why I haven't gotten off my ass all day and have been
doing nothing but reading/writing mails from this list.  :-0  If the
list accepted binaries, I'd take a photo of my room and workspace, and
you'd get the impression that I was one of the laziest people you've
ever met.  Granted, one of my computer science professors (not one of
the greatest professors I ever had) said that the best programmers in
the world are the laziest sons of bitches you'll ever meet, and I think
I may qualify (as a lazy son of a bitch, not necessarily as a great
programmer).

Aww.. Heck.. Go ahead and use goto all you want.  I'm not in a position
to stop anyone anyway.

-Rob

Re: any chance of 2.6.0-test*?

[Rik van Riel]

On Sun, 12 Jan 2003, Rob Wilkens wrote:

> Granted, one of my computer science professors (not one of the greatest
> professors I ever had) said that the best programmers in the world are
> the laziest sons of bitches you'll ever meet,

I think this is generally marketed under the term "code reuse"

cheers,

Rik
-- 
Bravely reimplemented by the knights who say "NIH".
http://www.surriel.com/		http://guru.conectiva.com/
Current spamtrap:  <a href=mailto:"october@surriel.com">october@surriel.com</a>

Re: any chance of 2.6.0-test*?

[Robert Love]

On Sun, 2003-01-12 at 17:22, Rob Wilkens wrote:

> I say "please don't use goto" and instead have a "cleanup_lock" function
> and add that before all the return statements..  It should not be a
> burden.  Yes, it's asking the developer to work a little harder, but the
> end result is better code.

No, it is gross and it bloats the kernel.  It inlines a bunch of junk
for N error paths, as opposed to having the exit code once at the end. 
Cache footprint is key and you just killed it.

Nor is it easier to read.

As a final argument, it does not let us cleanly do the usual stack-esque
wind and unwind, i.e.

	do A
	if (error)
		goto out_a;
	do B
	if (error)
		goto out_b;
	do C
	if (error)
		goto out_c;
	goto out;
	out_c:
	undo C
	out_b:
	undo B:
	out_a:
	undo A
	out:
	return ret;

Now stop this.

	Robert Love

Re: any chance of 2.6.0-test*?

[Matti Aarnio]

On Sun, Jan 12, 2003 at 04:27:30PM -0500, Rob Wilkens wrote:
> > Explain how you can do that cleanly, understandably, and without
> > code duplication, or ugly kludges  without using those goto ?
> > (And sticking to coding-style.)
> 
> I've only compiled (and haven't tested this code), but it should be much
> faster than the original code.  Why?  Because we're eliminating an extra
> "jump" in several places in the code every time open would be called. 

Benchmark them - original and your revised.   (Their equivalents, that is,
in userspace with rdtscll() ) 

Put them into a test-harness calling them million times (or some such).
Look also what assembly the compiler produced, e.g. that it didn't optimize
away error paths.

> Yes, it's more code, so the kernel is a little bigger, but it should be
> faster at the same time, and memory should be less of an issue nowadays.

That is common fallacy.  Learn to code in environments where you
have to account for every byte of code and ram usage, and your code
will be fast even in bigger systems.

> Here's the patch if you want to apply it (i have only compile tested it,
> I haven't booted with it).. This patch applied to the 2.5.56 kernel.
....

/Matti Aarnio

Re: any chance of 2.6.0-test*?

[David Lang]

> I've only compiled (and haven't tested this code), but it should be much
> faster than the original code.  Why?  Because we're eliminating an extra
> "jump" in several places in the code every time open would be called.
> Yes, it's more code, so the kernel is a little bigger, but it should be
> faster at the same time, and memory should be less of an issue nowadays.

Rob, one thing you may not have noticed since you haven't been following
the list for a while is that with the current generation of computers size
frequently translates directly into speed and a lot of the time honored
performance tricks that trade size for fewer commands executed end up
being losses.

this can be seen by compiling code with -O2 and with -Os and frequently
the -Os will actually be faster.

This is becouse not all memory is equal, main memory is very slow compared
to the CPU cache, so code that is slightly larger can cause more cache
misses and therefor be slower, even if significantly fewer commands are
executed.

in addition frequently the effect isn't direct (i.e. no noticable
difference on the code you are changing, but instead the change makes
other code slower as it gets evicted from the cache)

unfortunantly while this effect is known the rules of when to optimize for
space and when to optimize for fewer cpu cycles for code execution are not
clear and vary from CPU to CPU frequently within variations of the same
family)

if you google for -Os you should find one of the several discussions on
the list in the last year on the subject.

David Lang

Re: any chance of 2.6.0-test*?

[William Lee Irwin III]

On Sun, Jan 12, 2003 at 05:26:48PM -0800, David Lang wrote:
> This is becouse not all memory is equal, main memory is very slow compared
> to the CPU cache, so code that is slightly larger can cause more cache
> misses and therefor be slower, even if significantly fewer commands are
> executed.

Not all memory is equal here, either i.e. I'm on NUMA boxen.


Bill

How to avoid the woord 'goto' (was Re: any chance of 2.6.0-test*?)

[Bernd Petrovitsch]

Rob Wilkens <robw@optonline.net> wrote:
[...]
>Here's the patch if you want to apply it (i have only compile tested it,
>I haven't booted with it).. This patch applied to the 2.5.56 kernel.
>
>--- open.c.orig	2003-01-12 16:17:01.000000000 -0500
>+++ open.c	2003-01-12 16:22:32.000000000 -0500
>@@ -100,44 +100,58 @@
> 
> 	error = -EINVAL;
> 	if (length < 0)	/* sorry, but loff_t says... */
>-		goto out;
>+		return error;
> 
> 	error = user_path_walk(path, &nd);
> 	if (error)
>-		goto out;
>+		return error;
> 	inode = nd.dentry->d_inode;
[ snipped the rest ]

You just copied the logic to "cleanup and leave" the function several 
times. The (current, next and subsequent) maintainers at the next 
change in that function simply _have_ to check all cases as if they 
are different.
Yes, _now_ you (and all others) know that they are identical. But in 6 
month after tons of patches?

Perhaps you want to avoid goto's with:
----  snip (yes, it is purposely not a `diff -urN`)  ----
switch(0==0) {
default:
    error = -EINVAL;
    if (length < 0) /* sorry, but loff_t says... */
        break;
    error = user_path_walk(path, &nd);
    if (error)
        break;
    inode = nd.dentry->d_inode;

    switch(0==0) {
    default:
        /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
        error = -EISDIR;
        if (S_ISDIR(inode->i_mode))
	    break;
        error = -EINVAL;
        if (!S_ISREG(inode->i_mode))
            break;

        error = permission(inode,MAY_WRITE);
        if (error)
            break;

        error = -EROFS;
        if(IS_RDONLY(inode))
            break;

        /* the rest omitted - the pattern should be clear */

        put_write_access(inode);
        break;
    }
    path_release(&nd);
    break;
}
return error;
----  snip  ----

FYI, backward goto's can be rewritten with:
----  snip  ----

    for(;;) {
        <do something>
        if(i_want_to_go_back)
            continue;
        <do something_else>
        break;
    }
----  snip  ----

Are these two more understandable because the avoid the 'goto'?
And try to see it not from the viewpoint of a first time reader, but
from the viewpoint of a/the maintainer/developer (who reads this code
probably quite often)?

	Bernd

-- 
Bernd Petrovitsch                              Email : bernd@gams.at
g.a.m.s gmbh                                  Fax : +43 1 205255-900
Prinz-Eugen-Straße 8                    A-1040 Vienna/Austria/Europe
                     LUGA : http://www.luga.at

Re: How to avoid the woord 'goto' (was Re: any chance of 2.6.0-test*?)

[Rob Wilkens]

Yeah, but..

My changes won't actually be incorporated (i hope) into the kernel. 
I've since learned that a goto will optimize to be just as quick as what
I wrote (by studying the optimized assembler output at other
suggestion).  Therefore, it was silly and pointless for me to spend the
10-12 minutes I did changing the code and testing it and e-mailing back
the changes.

So, there's nothing to worry about.  Goto's are here forever.

-Rob

On Mon, 2003-01-13 at 06:08, Bernd Petrovitsch wrote:
> Rob Wilkens <robw@optonline.net> wrote:
> [...]
> >Here's the patch if you want to apply it (i have only compile tested it,
> >I haven't booted with it).. This patch applied to the 2.5.56 kernel.
> >
> >--- open.c.orig	2003-01-12 16:17:01.000000000 -0500
> >+++ open.c	2003-01-12 16:22:32.000000000 -0500
> >@@ -100,44 +100,58 @@
> > 
> > 	error = -EINVAL;
> > 	if (length < 0)	/* sorry, but loff_t says... */
> >-		goto out;
> >+		return error;
> > 
> > 	error = user_path_walk(path, &nd);
> > 	if (error)
> >-		goto out;
> >+		return error;
> > 	inode = nd.dentry->d_inode;
> [ snipped the rest ]
> 
> You just copied the logic to "cleanup and leave" the function several 
> times. The (current, next and subsequent) maintainers at the next 
> change in that function simply _have_ to check all cases as if they 
> are different.
> Yes, _now_ you (and all others) know that they are identical. But in 6 
> month after tons of patches?
> 
> Perhaps you want to avoid goto's with:
> ----  snip (yes, it is purposely not a `diff -urN`)  ----
> switch(0==0) {
> default:
>     error = -EINVAL;
>     if (length < 0) /* sorry, but loff_t says... */
>         break;
>     error = user_path_walk(path, &nd);
>     if (error)
>         break;
>     inode = nd.dentry->d_inode;
> 
>     switch(0==0) {
>     default:
>         /* For directories it's -EISDIR, for other non-regulars - -EINVAL */
>         error = -EISDIR;
>         if (S_ISDIR(inode->i_mode))
> 	    break;
>         error = -EINVAL;
>         if (!S_ISREG(inode->i_mode))
>             break;
> 
>         error = permission(inode,MAY_WRITE);
>         if (error)
>             break;
> 
>         error = -EROFS;
>         if(IS_RDONLY(inode))
>             break;
> 
>         /* the rest omitted - the pattern should be clear */
> 
>         put_write_access(inode);
>         break;
>     }
>     path_release(&nd);
>     break;
> }
> return error;
> ----  snip  ----
> 
> FYI, backward goto's can be rewritten with:
> ----  snip  ----
> 
>     for(;;) {
>         <do something>
>         if(i_want_to_go_back)
>             continue;
>         <do something_else>
>         break;
>     }
> ----  snip  ----
> 
> Are these two more understandable because the avoid the 'goto'?
> And try to see it not from the viewpoint of a first time reader, but
> from the viewpoint of a/the maintainer/developer (who reads this code
> probably quite often)?
> 
> 	Bernd

Re: any chance of 2.6.0-test*?

[Val Henson]

On Sun, Jan 12, 2003 at 04:27:30PM -0500, Rob Wilkens wrote:
> 
> I've only compiled (and haven't tested this code), but it should be much
> faster than the original code.  Why?  Because we're eliminating an extra
> "jump" in several places in the code every time open would be called. 
> Yes, it's more code, so the kernel is a little bigger, but it should be
> faster at the same time, and memory should be less of an issue nowadays.

To the tune of "Teen Angst" from Cracker:

  Cause, what the world needs now,
  is another theoretical optimizer
  like I need a hole in my head.

You'd think I'd be resigned to the l-k tradition of the unbenchmarked
"optimization" patch, but apparently not yet...

-VAL

gotos in kernel [Was: Re: any chance of 2.6.0-test*?]

[Horst von Brand]

Rob Wilkens <robw@optonline.net> said:
> I'm REALLY opposed to the use of the word "goto" in any code where it's
> not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> to comment

My, my, my, an anti-goto zealot now.

> Let me comment below the relevant code snippet below as to how I would
> change it:
> 
> On Sun, 2003-01-12 at 14:15, Linus Torvalds wrote:
> > 		if (spin_trylock(&tty_lock.lock))
> > 			goto got_lock;
> > 		if (tsk == tty_lock.lock_owner) {
> > 			WARN_ON(!tty_lock.lock_count);
> > 			tty_lock.lock_count++;
> > 			return flags;
> > 		}
> > 		spin_lock(&tty_lock.lock);
> > 	got_lock:
> > 		WARN_ON(tty_lock.lock_owner);
> 	    	   <etc...>
> 
> I would change it to something like the following (without testing the
> code through a compiler or anything to see if it's valid):

It just to happens that sometimes the compiler generates stupid code for
goto-less solutions. While the famous "no goto needed" theorem is certainly
true, it duplicates lots of code to get rid of gotos, and that is a no -
never - only over my dead body proposition in a kernel that tries to be as
fast as humanly possible.

And if used with care and good taste, a goto can lead to clearer, more
understandable (and thus more probably correct) code. Just look up what
sort of "programming style" Dijkstra was complaining about when asking to
ban gotos, and what D. E. Knuth has to say on structured programming with
gotos.
--
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: gotos in kernel [Was: Re: any chance of 2.6.0-test*?]

[Rob Wilkens]

On Mon, 2003-01-13 at 05:32, Horst von Brand wrote:
> Rob Wilkens <robw@optonline.net> said:
> > I'm REALLY opposed to the use of the word "goto" in any code where it's
> > not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> > to comment
> 
> My, my, my, an anti-goto zealot now.

I pick out _one_ piece of code that _one_ random person posts publicly
(so, it happens to be one of the leaders of the free open source world)
and suddenly I'm a zealot.  I'm a nobody and proud of that fact.   I was
just explaining my dislike of a particular code structure based on my
earlier training and experience.

In the time since my writing of that post and your response, though, I
have since been "Enlightenned" by all of the wonderful people on this
discussion forum.

-Rob

Re: any chance of 2.6.0-test*?

[Dave Jones]

On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:

 > I'm REALLY opposed to the use of the word "goto" in any code where it's
 > not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
 > to comment

Someone want to add this one to the lkml faq (does anyone actually
read that these days?) or maybe http://www.kernelnewbies.org/faq/

Wow, one week later, and this would tie in with the fourth anniversary
of someone else[2] making an ass of himself on this issue[1]

		Dave

[1] http://www.uwsg.iu.edu/hypermail/linux/kernel/9901.2/0939.html
[2] Complete with quaint 'cool' l33t handle. How 90s! *bows head in shame*

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: any chance of 2.6.0-test*?

[Andrew Walrond]

Well well. Who'd have thought! Idealistic youth; those were the days; 
etc etc ;)

Dave Jones wrote:
> On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> 
>  > I'm REALLY opposed to the use of the word "goto" in any code where it's
>  > not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
>  > to comment
> 
> Someone want to add this one to the lkml faq (does anyone actually
> read that these days?) or maybe http://www.kernelnewbies.org/faq/
> 
> Wow, one week later, and this would tie in with the fourth anniversary
> of someone else[2] making an ass of himself on this issue[1]
> 
> 		Dave
> 
> [1] http://www.uwsg.iu.edu/hypermail/linux/kernel/9901.2/0939.html
> [2] Complete with quaint 'cool' l33t handle. How 90s! *bows head in shame*
> 

Re: any chance of 2.6.0-test*?

[Oliver Neukum]

Am Montag, 13. Januar 2003 14:08 schrieb Dave Jones:
> On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
>  > I'm REALLY opposed to the use of the word "goto" in any code where it's
>  > not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
>  > to comment
>
> Someone want to add this one to the lkml faq (does anyone actually
> read that these days?) or maybe http://www.kernelnewbies.org/faq/
>
> Wow, one week later, and this would tie in with the fourth anniversary
> of someone else[2] making an ass of himself on this issue[1]

So let us rejoice in a story of human growth.
In the exuberant spirit of this joyfull occasion, I'd like you to find
a place for this memorative piece of code:

if (a->sibling->present)
	goto your_place;
else
	goto my_place;

	Regards
		Oliver

Re: any chance of 2.6.0-test*?

[Bruce Harada]

On Mon, 13 Jan 2003 13:08:42 +0000
Dave Jones <davej@codemonkey.org.uk> wrote:

> On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
> 
> Wow, one week later, and this would tie in with the fourth anniversary
> of someone else[2] making an ass of himself on this issue[1]
> 
> 		Dave
> 
> [1] http://www.uwsg.iu.edu/hypermail/linux/kernel/9901.2/0939.html
> [2] Complete with quaint 'cool' l33t handle. How 90s! *bows head in shame*

So does that mean that we can look forward to Rob being a very productive
leading Linux developer in four years time?

Naaaah.... ;)

Re: any chance of 2.6.0-test*?

[Brian Tinsley]

What's happened to Rob??? He's been unusually quiet today! Too many 
 tranquilizers? He must have passed out face down on his keyboard and 
the drool shorted it out ;)



Bruce Harada wrote:

>On Mon, 13 Jan 2003 13:08:42 +0000
>Dave Jones <davej@codemonkey.org.uk> wrote:
>
>  
>
>>On Sun, Jan 12, 2003 at 02:34:54PM -0500, Rob Wilkens wrote:
>>
>>Wow, one week later, and this would tie in with the fourth anniversary
>>of someone else[2] making an ass of himself on this issue[1]
>>
>>		Dave
>>
>>[1] http://www.uwsg.iu.edu/hypermail/linux/kernel/9901.2/0939.html
>>[2] Complete with quaint 'cool' l33t handle. How 90s! *bows head in shame*
>>    
>>
>
>So does that mean that we can look forward to Rob being a very productive
>leading Linux developer in four years time?
>
>Naaaah.... ;)
>
>
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at  http://www.tux.org/lkml/
>  
>

-- 

Brian Tinsley
Chief Systems Engineer
Emageon

Re: any chance of 2.6.0-test*?

[Richard B. Johnson]

On Sun, 12 Jan 2003, Rob Wilkens wrote:

> Linus,
> 
> I'm REALLY opposed to the use of the word "goto" in any code where it's
> not needed.  OF course, I'm a linux kernel newbie, so I'm in no position
> to comment
>
I'm sure Linus can hold his own on this one, but this has become
a FAQ or FSB (Frequently stated bitch).


Programming 101:
All known microprocessors execute instructions called
"op-codes". The logic of program-flow is implemented
by testing condition "flags" and jumping somewhere else,
based upon the condition. These jumps are "GOTOs".

To implement logic, you have three, and only three choices.
You can jump to some code that's earlier in the instruction
stream, you can jump to some code that's later in the
instruction-stream, or you can just keep executing the
current instruction stream.

Assemblers generate op-codes that programmers program.
They are not allowed to change anything a programmer inputs.
For all practical purposes, they simply change programmer-
defined names to processor-specific op-codes.

Compilers have more latitude. Compilers, including 'C' compilers
attempt to convert logic that a programmer defines to op-codes.
Since a compiler needs to generate code that will give the
correct logic, regardless of the previous logic or the following
logic, it is constrained in its code generation capabilities in
a manner that an assembly-language programmer would not necessarily
be. However, an astute programmer can, using the tools provided
by the compiler, including 'goto', cause code-generation in which
the usual case allows the processor to continue executing the
current execution-stream without jumping.

When the processor is forced to jump on a condition, this usually
costs CPU cycles. An astute kernel programmer may spend hours
mucking with a single piece of code trying to make it execute as
fast as possible. When a new-be comes along and complains about
the methods used, he is going to get his feet wet real quickly!

Here is an example of how some junior instructors in college
teach...
        if(...){
        {
            if(...){
            {
                if(...){
                 {

They just don't get it. When you get to something that's not
true you can't readily "get out". Further, the normal program-
flow ends up with all those "ifs". The above code should be:

         if(!...) goto get_out;
         if(!...) goto get_out;
         if(!...) goto get_out;
         {do something()...}
         get_out:

Cheers,
Dick Johnson
Penguin : Linux version 2.4.18 on an i686 machine (797.90 BogoMips).
Why is the government concerned about the lunatic fringe? Think about it.

Re: any chance of 2.6.0-test*?

[Greg KH]

On Sun, Jan 12, 2003 at 11:15:52AM -0800, Linus Torvalds wrote:
> 
> Anybody willing to test it and see if the above works? 

Hm, I just did, and it seems to work for me.  I ran a bunch of different
usb-serial drivers through some tests, but these worked for me without
this patch.  I don't know what stress tests Alan was running to show up
a problem.

Anyway, here's a patch with your new lock, if you want to apply it.

thanks,

greg k-h


diff -Nru a/drivers/char/tty_io.c b/drivers/char/tty_io.c
--- a/drivers/char/tty_io.c	Sun Jan 12 22:27:42 2003
+++ b/drivers/char/tty_io.c	Sun Jan 12 22:27:42 2003
@@ -159,6 +159,58 @@
 extern void tx3912_rs_init(void);
 extern void hvc_console_init(void);
 
+
+/*
+* This isn't even _trying_ to be fast!
+*/
+struct recursive_spinlock {
+	spinlock_t lock;
+	int lock_count;
+	struct task_struct *lock_owner;
+};
+
+static struct recursive_spinlock tty_lock = {
+	.lock = SPIN_LOCK_UNLOCKED,
+	.lock_count = 0,
+	.lock_owner = NULL
+};
+
+unsigned long tty_spin_lock(void)
+{
+	unsigned long flags;
+	struct task_struct *tsk;
+
+	local_irq_save(flags);
+	preempt_disable();
+	tsk = current;
+	if (spin_trylock(&tty_lock.lock))
+		goto got_lock;
+	if (tsk == tty_lock.lock_owner) {
+		WARN_ON(!tty_lock.lock_count);
+		tty_lock.lock_count++;
+		return flags;
+	}
+	spin_lock(&tty_lock.lock);
+got_lock:
+	WARN_ON(tty_lock.lock_owner);
+	WARN_ON(tty_lock.lock_count);
+	tty_lock.lock_owner = tsk;
+	tty_lock.lock_count = 1;
+	return flags;
+}
+
+void tty_spin_unlock(unsigned long flags)
+{
+	WARN_ON(tty_lock.lock_owner != current);
+	WARN_ON(!tty_lock.lock_count);
+	if (!--tty_lock.lock_count) {
+		tty_lock.lock_owner = NULL;
+		spin_unlock(&tty_lock.lock);
+	}
+	preempt_enable();
+	local_irq_restore(flags);
+}
+
 static struct tty_struct *alloc_tty_struct(void)
 {
 	struct tty_struct *tty;
@@ -460,7 +512,7 @@
 	{
 		unsigned long flags;
 
-		local_irq_save(flags); // FIXME: is this safe?
+		flags = tty_spin_lock();
 		if (tty->ldisc.flush_buffer)
 			tty->ldisc.flush_buffer(tty);
 		if (tty->driver.flush_buffer)
@@ -468,7 +520,7 @@
 		if ((test_bit(TTY_DO_WRITE_WAKEUP, &tty->flags)) &&
 		    tty->ldisc.write_wakeup)
 			(tty->ldisc.write_wakeup)(tty);
-		local_irq_restore(flags); // FIXME: is this safe?
+		tty_spin_unlock(flags);
 	}
 
 	wake_up_interruptible(&tty->write_wait);
@@ -1926,7 +1978,7 @@
 		fp = tty->flip.flag_buf + TTY_FLIPBUF_SIZE;
 		tty->flip.buf_num = 0;
 
-		local_irq_save(flags); // FIXME: is this safe?
+		flags = tty_spin_lock();
 		tty->flip.char_buf_ptr = tty->flip.char_buf;
 		tty->flip.flag_buf_ptr = tty->flip.flag_buf;
 	} else {
@@ -1934,13 +1986,13 @@
 		fp = tty->flip.flag_buf;
 		tty->flip.buf_num = 1;
 
-		local_irq_save(flags); // FIXME: is this safe?
+		flags = tty_spin_lock();
 		tty->flip.char_buf_ptr = tty->flip.char_buf + TTY_FLIPBUF_SIZE;
 		tty->flip.flag_buf_ptr = tty->flip.flag_buf + TTY_FLIPBUF_SIZE;
 	}
 	count = tty->flip.count;
 	tty->flip.count = 0;
-	local_irq_restore(flags); // FIXME: is this safe?
+	tty_spin_unlock(flags);
 	
 	tty->ldisc.receive_buf(tty, cp, fp, count);
 }

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On Sun, 12 Jan 2003, Greg KH wrote:
> 
> Anyway, here's a patch with your new lock, if you want to apply it.

I'd like to have some verification (or some test-suite) to see whether it 
makes any difference at all before applying it.

Alan, what's your load?

		Linus

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Mon, 2003-01-13 at 16:35, Linus Torvalds wrote:
> On Sun, 12 Jan 2003, Greg KH wrote:
> > 
> > Anyway, here's a patch with your new lock, if you want to apply it.
> 
> I'd like to have some verification (or some test-suite) to see whether it 
> makes any difference at all before applying it.
> 
> Alan, what's your load?

Lots of serial activity (standard PC serial ports) with carrier drops 
present and random oopses appear. I've seen ppp oopses too but don't know
if they are related. I tried duplicating it with pty/tty traffic on a dual
PPro200 and suprisingly that did the same.

Ages ago I chased serial bugs down by doing data transfers between two
PC's while one of them was strobing the carrier up and down on the test
PC with varying frequencies

I've not had time to try that paticular abuse again alas. 

Alan

Re: any chance of 2.6.0-test*?

[Linus Torvalds]

On 13 Jan 2003, Alan Cox wrote:
> 
> Lots of serial activity (standard PC serial ports) with carrier drops 
> present and random oopses appear.

Well, that would kind of match the locking place - the hangup might well
race with IO activity on another CPU. That's exactly the sequence that was
protected by the global irq lock.

> I tried duplicating it with pty/tty traffic on a dual PPro200 and
> suprisingly that did the same.

Just regular IO on its own shouldn't trigger this, I _think_ (no hangup
event).

Although I can actually imagine the "flush_to_ldisc()" racing with itself
on a pty (with the master flushing the slave as the slave flushes itself),
so maybe it could actually happen.. But if your pty tests closed the
master and forced hangups that way, the race would be more likely.

Do you still have the pty stress-test program? 

> Ages ago I chased serial bugs down by doing data transfers between two
> PC's while one of them was strobing the carrier up and down on the test
> PC with varying frequencies

Yeah, that hangup path is one of the nastier tty events, and it's also oen 
that doesn't get much testing in many "normal" loads.

		Linus

Re: any chance of 2.6.0-test*?

[Andrew McGregor]

I believe I may have seen it on 2.5.53 just surfing on an Alcatel 
Speedtouch DSL modem (using PPPoATM) while compiling in a background 
gnome-terminal (and probably running a few other things, maybe including 
top, in terminals).  I'll try to confirm that.

Andrew

--On Monday, January 13, 2003 18:33:31 +0000 Alan Cox 
<alan@lxorguk.ukuu.org.uk> wrote:

> On Mon, 2003-01-13 at 16:35, Linus Torvalds wrote:
>> On Sun, 12 Jan 2003, Greg KH wrote:
>> >
>> > Anyway, here's a patch with your new lock, if you want to apply it.
>>
>> I'd like to have some verification (or some test-suite) to see whether
>> it  makes any difference at all before applying it.
>>
>> Alan, what's your load?
>
> Lots of serial activity (standard PC serial ports) with carrier drops
> present and random oopses appear. I've seen ppp oopses too but don't know
> if they are related. I tried duplicating it with pty/tty traffic on a dual
> PPro200 and suprisingly that did the same.
>
> Ages ago I chased serial bugs down by doing data transfers between two
> PC's while one of them was strobing the carrier up and down on the test
> PC with varying frequencies
>
> I've not had time to try that paticular abuse again alas.
>
> Alan
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
>

Re: any chance of 2.6.0-test*?

[Russell King]

On Sun, Jan 12, 2003 at 09:05:16AM -0800, Linus Torvalds wrote:
> The only place that looked like it _really_ didn't get the kernel lock was
> apparently tty_open(), which is admittedly a fairly important part of it ;)

chrdev_open() kindly takes the BKL for tty_open() so there isn't an issue
here.

There were, however, parts of the code which were using the global-cli(),
and these merely got converted to local_irq_save() without (iirc) too much
auditing.  They're all marked with FIXME in my current sources.

One thing that regularly bugs me about our current tty implementation,
however, is the handling of SIGWINCH.  This is sent to the process group
rather than the session.  Take the following example:

- you're running an xterm, with a shell inside.
- you're running 3 ssh connections, 2 suspended, one running.
- you resize the xterm

The WINCH signal will be sent to just the one foreground ssh connection.

- you suspend the current ssh.

The shell believes that the terminal is the _old_ size.

- you resume one of the old ssh sessions.

Likewise, ssh doesn't believe the terminal has changed size.

My current "workaround" is to re-tweak the xterm size each time I resume
or suspend an old process.

POSIX (at least signal and general terminal interface sections) seems to
be rather quiet on the behaviour of WINCH.

Are there any reasons why we shouldn't send WINCH to the whole session
rather than just the pgrp associated with the tty?

-- 
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: any chance of 2.6.0-test*?

[Michael Kingsbury]

> Say, I've been having _smashing_ success with 2.5.x on the desktop and
> on big fat highmem umpteen-way SMP (NUMA even!) boxen, and I was
> wondering if you were considering 2.6.0-test* anytime soon.
> 
> I'd love to get this stuff out for users to hammer on ASAP, and things
> are looking really good AFAICT.
> 
> Any specific concerns/issues/wishlist items you want taken care of
> before doing it or is it a "generalized comfort level" kind of thing?
> Let me know, I'd be much obliged for specific directions to move in.

As long as I can't unzip the latest, do a quick

	'make xconfig' w/o changing the defaults,

	 followed by

	'make bzImage',

and actually get the sucker to compile, I'd say go for it.  But there's 
still too much breakage.  And until that breakage is fixed, you won't get 
the next wave of people (users) who are curious to poke around with it.  
(some of us don't care to take half a day just to figure out what to do to make something compile because 
there's problems in the source.)

(Yes, I realize I have to customize the kernel some.  But at the very 
least the defaults shouldn't cause it to fall flat on its face.)

-mike

Re: any chance of 2.6.0-test*?

[William Lee Irwin III]

At some point in the past, I wrote:
>> Any specific concerns/issues/wishlist items you want taken care of
>> before doing it or is it a "generalized comfort level" kind of thing?
>> Let me know, I'd be much obliged for specific directions to move in.

On Sun, Jan 12, 2003 at 08:08:42PM -0500, Michael Kingsbury wrote:
> As long as I can't unzip the latest, do a quick
> 	'make xconfig' w/o changing the defaults,
> 	 followed by
> 	'make bzImage',
> and actually get the sucker to compile, I'd say go for it.  But there's 
> still too much breakage.  And until that breakage is fixed, you won't get 
> the next wave of people (users) who are curious to poke around with it.  
> (some of us don't care to take half a day just to figure out what to do to make something compile because 
> there's problems in the source.)
> (Yes, I realize I have to customize the kernel some.  But at the very 
> least the defaults shouldn't cause it to fall flat on its face.)

Hmm, I never do this, I've been recycling my own .configs for some time.
Sounds like a valid concern, I'll give it a shot.

Bill

Re: any chance of 2.6.0-test*?

[Shawn Starr]

There will be a new kernel tree that will fit this purpose soon called -xlk 
(eXtendable or Extended Linux Kernel). The hope to make it an 'official' like 
-ac, -mm tree for stuffing experimental stuff into a post 2.6 (or just before 
2.6 goes live) kernel. I will need help in getting this to become a reality 
in the coming months to 2.6.

The purpose of the tree is to get experimental code ready for the 2.7 (2.9?) 
tree. We want code that will add new drivers / devices and general 
improvements to the kernel. The goal is once these are stabilized they can be 
submitted to Linus and friends for blessings and inclusion into 2.7 dev 
*early* so we won't have a mad rush for features before the next feature 
freeze.

>Subject:  any chance of 2.6.0-test*?
From:     William Lee Irwin III <wli () holomorphy ! com>
>Date:     2003-01-10 16:10:12

>Say, I've been having _smashing_ success with 2.5.x on the desktop and
>on big fat highmem umpteen-way SMP (NUMA even!) boxen, and I was
>wondering if you were considering 2.6.0-test* anytime soon.

>I'd love to get this stuff out for users to hammer on ASAP, and things
>are looking really good AFAICT.

>Any specific concerns/issues/wishlist items you want taken care of
>before doing it or is it a "generalized comfort level" kind of thing?
>Let me know, I'd be much obliged for specific directions to move in.


>Thanks,
>Bill
 
-- 
Shawn Starr
UNIX Systems Administrator, Operations
Datawire Communication Networks Inc.
10 Carlson Court, Suite 300
Toronto, ON, M9W 6L2
T: 416-213-2001 ext 179  F: 416-213-2008
shawn.starr@datawire.net
"The power to Transact" - http://www.datawire.net

Re: any chance of 2.6.0-test*?

[Dave Jones]

On Fri, Jan 10, 2003 at 11:39:57AM -0500, Shawn Starr wrote:
 > There will be a new kernel tree that will fit this purpose soon called -xlk 
 > (eXtendable or Extended Linux Kernel). The hope to make it an 'official' like 
 > -ac, -mm tree for stuffing experimental stuff into a post 2.6 (or just before 
 > 2.6 goes live) kernel. I will need help in getting this to become a reality 
 > in the coming months to 2.6.

The effort is really much better spent trying to get to 2.6 first before
worrying about things like 2.7.  I hope 2.6 doesn't turn into the
"heres my 2.4+preempt+rmap patchset" monster that we saw six months ago.
 
 > We want code that will add new drivers / devices and general 
 > improvements to the kernel.

non-core changes (ie, new drivers) still get added during code freeze,
and during 2.6.x, there's no need for a specific tree just for this.
Adding a new driver doesn't (or at least shouldn't) impact any existing
users if done right.
 
 > The goal is once these are stabilized they can be 
 > submitted to Linus and friends for blessings and inclusion into 2.7 dev 
 > *early* so we won't have a mad rush for features before the next feature 
 > freeze.

Nice try. It'll still happen regardless. Bombing Linus with ten zillion
patches when he opens up 2.7.x with "has been tested in 2.6-xyz" isn't
the way to do it. Everything has to happen incrementally, or you end
up with a mess.

		Dave

-- 
| Dave Jones.        http://www.codemonkey.org.uk
| SuSE Labs

Re: any chance of 2.6.0-test*?

[Horst von Brand]

Shawn Starr <shawn.starr@datawire.net> said:
> There will be a new kernel tree that will fit this purpose soon called
> -xlk (eXtendable or Extended Linux Kernel). The hope to make it an
> 'official' like -ac, -mm tree for stuffing experimental stuff into a post
> 2.6 (or just before 2.6 goes live) kernel. I will need help in getting
> this to become a reality in the coming months to 2.6.

Thanks, but please don't. The idea of the feature/code freezes, and then
the stable version running alone for a while, is precisely to focus the
hacker comunity on fixing the bugs and cleaning up, and then stabilizing
the whole. Please don't distract them.
--
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: any chance of 2.6.0-test*?

[Matthew Wilcox]

Linus Torvalds wrote:
> Note that other architectures have never been an issue for releasing new
> kernels, and that is _particularly_ true of architectures like parisc and
> mips that haven't even _tried_ to track development kernels. In fact, mips
> "anti-maintenance" has often actively discouraged people from even
> bothering to update mips code when adding new features.

Um.. hey!  Just because we weren't trying to merge with you during 2.5
for a long time doesn't mean we weren't tracking development.  Looking
at our CVS history, we've merged your tree into ours in:

 2.5.26
 2.5.32
 2.5.41
 2.5.43
 2.5.44
 2.5.45
 2.5.46
 2.5.47
 2.5.50
 2.5.51
 2.5.53
 2.5.54

Our outstanding diff vs your tree is about 200k gzipped and it's almost
all drivers.  Off the top of my head, I can't think of any core changes
we still need.  I don't think we're using any deprecated interfaces
(eg flush_page_to_ram, unlike m68k, mips, mips64, sparc32 & v850).

Of course, I don't consider having working PA-RISC in your tree to be
a prerequisite for release.  I just object to being used as an example :-P

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk

Re: any chance of 2.6.0-test*?

[Andi Kleen]

Alan Cox <alan@lxorguk.ukuu.org.uk> writes:

> On Fri, 2003-01-10 at 16:10, William Lee Irwin III wrote:
> > Any specific concerns/issues/wishlist items you want taken care of
> > before doing it or is it a "generalized comfort level" kind of thing?
> > Let me know, I'd be much obliged for specific directions to move in.
> 
> IDE is all broken still and will take at least another three months to
> fix - before we get to 'improve'. The entire tty layer locking is terminally

Can you quickly summarize what is broken with IDE ?

Are just some low level drivers broken or are there some generic
nasty problems.

If it is just some broken low level drivers I guess they can 
be marked dangerous or CONFIG_EXPERIMENTAL.

How does it differ from the code that was just merged into 2.4.21pre3
(has the later all the problems fixed?)

> broken and nobody has even started fixing it. Just try a mass of parallel
> tty/pty activity . It was problematic before, pre-empt has taken it  to dead, 
> defunct and buried. 

Can someone shortly describe what is the main problem with TTY?

>From what I can see the high level tty code mostly takes lock_kernel
before doing anything.
On reads access to file->private_data is not serialized, but it at 
least shouldn't go away because VFS takes care of struct file
reference counting.

The tty_drivers list does seem to need a spinlock, but I guess
just taking lock_kernel in tty_open would fix that for now.

[i didn't look at low level ldiscs]

Any particular test cases that break ?

If yes I would recommend to post them as scripts and their oopses so 
that people can start working on them.


The appended untested patch adds some lock_kernel()s that appear to be missing
to tty_io.c. The rest seems to already run under BKL or not access
any global data
(except tty_paranoia_check, but is probably ok with the reference counting
in the VFS)

> 
> Most of the drivers still don't build either.

In UP most did last time I tried.
On SMP a lot of problems are caused by the cli removal

My personal (i386) problem list is relatively short.
I use used 2.5.54 on my desktop without any problems (without preempt)

- BIO still oopses when XFS tries replay a log on RAID-0

-Andi

--- linux-2.5.56-work/drivers/char/tty_io.c-o	2003-01-02 05:13:12.000000000 +0100
+++ linux-2.5.56-work/drivers/char/tty_io.c	2003-01-11 13:23:15.000000000 +0100
@@ -1329,6 +1329,8 @@
 		int major, minor;
 		struct tty_driver *driver;
 
+		lock_kernel(); 
+
 		/* find a device that is not in use. */
 		retval = -1;
 		for ( major = 0 ; major < UNIX98_NR_MAJORS ; major++ ) {
@@ -1340,6 +1342,8 @@
 				if (!init_dev(device, &tty)) goto ptmx_found; /* ok! */
 			}
 		}
+
+		unlock_kernel();
 		return -EIO; /* no free ptys */
 	ptmx_found:
 		set_bit(TTY_PTY_LOCK, &tty->flags); /* LOCK THE SLAVE */
@@ -1357,6 +1361,8 @@
 #endif  /* CONFIG_UNIX_98_PTYS */
 	}
 
+	lock_kernel();
+
 	retval = init_dev(device, &tty);
 	if (retval)
 		return retval;
@@ -1389,6 +1395,8 @@
 #endif
 
 		release_dev(filp);
+
+		unlock_kernel(); 
 		if (retval != -ERESTARTSYS)
 			return retval;
 		if (signal_pending(current))
@@ -1397,6 +1405,7 @@
 		/*
 		 * Need to reset f_op in case a hangup happened.
 		 */
+		lock_kernel();
 		filp->f_op = &tty_fops;
 		goto retry_open;
 	}
@@ -1424,6 +1433,7 @@
 			nr_warns++;
 		}
 	}
+	unlock_kernel();
 	return 0;
 }
 
@@ -1444,8 +1454,13 @@
 	if (tty_paranoia_check(tty, filp->f_dentry->d_inode->i_rdev, "tty_poll"))
 		return 0;
 
-	if (tty->ldisc.poll)
-		return (tty->ldisc.poll)(tty, filp, wait);
+	if (tty->ldisc.poll) { 
+		int ret;
+		lock_kernel();
+		ret = (tty->ldisc.poll)(tty, filp, wait);
+		unlock_kernel();
+		return ret;
+	}
 	return 0;
 }
 

Re: any chance of 2.6.0-test*?

[Russell King]

> --- linux-2.5.56-work/drivers/char/tty_io.c-o	2003-01-02 05:13:12.000000000 +0100
> +++ linux-2.5.56-work/drivers/char/tty_io.c	2003-01-11 13:23:15.000000000 +0100
> @@ -1329,6 +1329,8 @@
>  		int major, minor;
>  		struct tty_driver *driver;
>  
> +		lock_kernel(); 
> +

Deadlock.  chrdev_open() calls lock_kernel() and then the fops->open
method, which is tty_open().

>  		/* find a device that is not in use. */
>  		retval = -1;
>  		for ( major = 0 ; major < UNIX98_NR_MAJORS ; major++ ) {
> @@ -1340,6 +1342,8 @@
>  				if (!init_dev(device, &tty)) goto ptmx_found; /* ok! */
>  			}
>  		}
> +
> +		unlock_kernel();
>  		return -EIO; /* no free ptys */
>  	ptmx_found:
>  		set_bit(TTY_PTY_LOCK, &tty->flags); /* LOCK THE SLAVE */
> @@ -1357,6 +1361,8 @@
>  #endif  /* CONFIG_UNIX_98_PTYS */
>  	}
>  
> +	lock_kernel();
> +

Deadlock.  See chrdev_open() note above.

>  	retval = init_dev(device, &tty);
>  	if (retval)
>  		return retval;
> @@ -1389,6 +1395,8 @@
>  #endif
>  
>  		release_dev(filp);
> +
> +		unlock_kernel(); 
>  		if (retval != -ERESTARTSYS)
>  			return retval;
>  		if (signal_pending(current))
> @@ -1397,6 +1405,7 @@
>  		/*
>  		 * Need to reset f_op in case a hangup happened.
>  		 */
> +		lock_kernel();

Deadlock.  See chrdev_open() note above.

>  		filp->f_op = &tty_fops;
>  		goto retry_open;
>  	}
> @@ -1424,6 +1433,7 @@
>  			nr_warns++;
>  		}
>  	}
> +	unlock_kernel();
>  	return 0;
>  }
>  
> @@ -1444,8 +1454,13 @@
>  	if (tty_paranoia_check(tty, filp->f_dentry->d_inode->i_rdev, "tty_poll"))
>  		return 0;
>  
> -	if (tty->ldisc.poll)
> -		return (tty->ldisc.poll)(tty, filp, wait);
> +	if (tty->ldisc.poll) { 
> +		int ret;
> +		lock_kernel();
> +		ret = (tty->ldisc.poll)(tty, filp, wait);
> +		unlock_kernel();
> +		return ret;
> +	}

This one needs deeper review.

-- 
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: any chance of 2.6.0-test*?

[Andi Kleen]

On Sat, Jan 11, 2003 at 02:01:51PM +0100, Russell King wrote:
> > --- linux-2.5.56-work/drivers/char/tty_io.c-o	2003-01-02 05:13:12.000000000 +0100
> > +++ linux-2.5.56-work/drivers/char/tty_io.c	2003-01-11 13:23:15.000000000 +0100
> > @@ -1329,6 +1329,8 @@
> >  		int major, minor;
> >  		struct tty_driver *driver;
> >  
> > +		lock_kernel(); 
> > +
> 
> Deadlock.  chrdev_open() calls lock_kernel() and then the fops->open
> method, which is tty_open().

No problem, lock_kernel is recursive and dropped on schedule.

It is very very hard to get a BKL deadlock.

> This one needs deeper review.

I agree, but one has to start somewhere. Please submit any fixes,
perhaps we can take then close these issues for good.

Was looking at n_tty.c now, looks like it has some more race 
problems.


-Andi

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Sat, 2003-01-11 at 12:27, Andi Kleen wrote:
> Can you quickly summarize what is broken with IDE ?
> Are just some low level drivers broken or are there some generic
> nasty problems.
> 
> If it is just some broken low level drivers I guess they can 
> be marked dangerous or CONFIG_EXPERIMENTAL.

Low level drivers are basically sorted.

The main problems are
  - Incorrect locking all over the place
  - Incorrect timings on some phases
  - Some ioctls can cause crashes due to locking
  - ISAPnP IDE doesn't work right now
  - Flaws in error recovery paths in certain situations
  - Lots of random oopses on boot/remove that were apparently
    introduced by the kobject/sysfs people and need chasing
    down. (There are some non sysfs ones mostly fixed)
  - ide-scsi needs some cleanup to fix switchover ide-cd/scsi
    (We can't dump ide-scsi)
  - Unregister path has races which cause all the long
    standing problems with pcmcia and prevents pci unreg
  - PCI IDE driver registration needs busy checks
  - PCI layer needs some stuff from 2.4
  - PCI layer in 2.4/2.5 needs an IRQ bug fixing
  - ACPI doesn't seem to handle compatibility IRQ mode
  - We don't handle a few errata (MWDMA on 450NX for example)
  - IDE raid hasn't been ported to 2.5 at all yet

Thats off the top of my head right now.

> How does it differ from the code that was just merged into 2.4.21pre3
> (has the later all the problems fixed?)

No, although some don't show up in the same ways in 2.4 - eg the pcmcia
unload race is rare in 2.4 for other reasons. Endianism should all
be cured in 2.4, and I sent that to Linus. The PCI i/o assignment code
in 2.4 is done. I hope to have some of the locking and also the timing
path work Ross Biro has done in for 2.4.22 proper

> > broken and nobody has even started fixing it. Just try a mass of parallel
> > tty/pty activity . It was problematic before, pre-empt has taken it  to dead, 
> > defunct and buried. 
> 
> Can someone shortly describe what is the main problem with TTY?
> 
> >From what I can see the high level tty code mostly takes lock_kernel
> before doing anything.

Which works really well with all the IRQ paths on it

> On reads access to file->private_data is not serialized, but it at 
> least shouldn't go away because VFS takes care of struct file
> reference counting.

The refcounting bugs are/were in the ldisc stuff. I thought the
bluetooth folks (Max and co) fixed that one

If we can lock_kernel the tty layer for now (I'm bothered about
the ldisc end of tht which is IRQ context) then great, tty scaling
is suddenelly a 2.7 problem.

Re: any chance of 2.6.0-test*?

[Andi Kleen]

> The main problems are
>   - Incorrect locking all over the place

Hmm bad. Is it that hard to fix ?

>   - Incorrect timings on some phases

Can't you just take out the timings in that case? 
My (not very informed) understanding is: 
everything should work with the BIOS timings and generic IDE,
having own timings is just an optimization to squeeze out a bit
more speed.

>   - ISAPnP IDE doesn't work right now

Hardly a release show stopper I guess

(ISA support in general being rather broken, but I doubt many people care
still) 

>   - Flaws in error recovery paths in certain situations
>   - Lots of random oopses on boot/remove that were apparently
>     introduced by the kobject/sysfs people and need chasing
>     down. (There are some non sysfs ones mostly fixed)

I guess the kobject/sysfs stuff could be ripped out if it doesn't
work - it is probably not a "must have" feature.


>   - ide-scsi needs some cleanup to fix switchover ide-cd/scsi
>     (We can't dump ide-scsi)
>   - Unregister path has races which cause all the long
>     standing problems with pcmcia and prevents pci unreg

Can't you just disable module unloading for the release ?
(not nice, but has worked for other subsystems like IPv6 too in the past)
Also 2.4 didn't have much problems without modular IDE, 2.6 will
be probably able to tolerate it too.

>   - IDE raid hasn't been ported to 2.5 at all yet

Vendor problem ?
> 
> > > broken and nobody has even started fixing it. Just try a mass of parallel
> > > tty/pty activity . It was problematic before, pre-empt has taken it  to dead, 
> > > defunct and buried. 
> > 
> > Can someone shortly describe what is the main problem with TTY?
> > 
> > >From what I can see the high level tty code mostly takes lock_kernel
> > before doing anything.
> 
> Which works really well with all the IRQ paths on it

Then 2.0/2.2/2.4 would have been racy too :-) Apparently they worked though.

High level tty code doesn't touch any ttys as far as I can see.

It is done in n_tty.c, which apparently needs a bit of work
(found some races together with Rik there)

> > On reads access to file->private_data is not serialized, but it at 
> > least shouldn't go away because VFS takes care of struct file
> > reference counting.
> 
> The refcounting bugs are/were in the ldisc stuff. I thought the
> bluetooth folks (Max and co) fixed that one

I was thinking about races with tty_release (= own tty_struct suddenly
going away). VFS would prevent that.


-Andi

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Sat, 2003-01-11 at 14:06, Andi Kleen wrote:
> > The main problems are
> >   - Incorrect locking all over the place
> 
> Hmm bad. Is it that hard to fix ?

Very. Just read the settings/proc code for an example

> >   - Incorrect timings on some phases
> 
> Can't you just take out the timings in that case? 
> My (not very informed) understanding is: 
> everything should work with the BIOS timings and generic IDE,
> having own timings is just an optimization to squeeze out a bit
> more speed.

These are timing for phases not drive timings. Drive timings from
the BIOS are also frequently questionable. These have to be fixed
and as boxes get faster it becomes more important they are

> >   - Lots of random oopses on boot/remove that were apparently
> >     introduced by the kobject/sysfs people and need chasing
> >     down. (There are some non sysfs ones mostly fixed)
> 
> I guess the kobject/sysfs stuff could be ripped out if it doesn't
> work - it is probably not a "must have" feature.

Without a doubt. 

> >   - ide-scsi needs some cleanup to fix switchover ide-cd/scsi
> >     (We can't dump ide-scsi)
> >   - Unregister path has races which cause all the long
> >     standing problems with pcmcia and prevents pci unreg
> 
> Can't you just disable module unloading for the release ?

Only if I can also nail shut your PCMCIA slot, disallow SATA and remove
some ioctls people use for docking.

> >   - IDE raid hasn't been ported to 2.5 at all yet
> 
> Vendor problem ?

It needs a rewrite to the new bio layer or rewriting to use the device
mapper, which is a distinct option here.


> > Which works really well with all the IRQ paths on it
> 
> Then 2.0/2.2/2.4 would have been racy too :-) Apparently they worked though.

Long ago lock_kernel had meaning against IRQ paths. TTY never caught up.

Re: any chance of 2.6.0-test*?

[Andi Kleen]

On Sat, Jan 11, 2003 at 04:31:00PM +0100, Alan Cox wrote:
> > >   - ide-scsi needs some cleanup to fix switchover ide-cd/scsi
> > >     (We can't dump ide-scsi)
> > >   - Unregister path has races which cause all the long
> > >     standing problems with pcmcia and prevents pci unreg
> > 
> > Can't you just disable module unloading for the release ?
> 
> Only if I can also nail shut your PCMCIA slot, disallow SATA and remove
> some ioctls people use for docking.

Hmm? Didn't that all work in 2.4 with monolithic IDE ?

-Andi

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Sat, 2003-01-11 at 15:25, Andi Kleen wrote:
> On Sat, Jan 11, 2003 at 04:31:00PM +0100, Alan Cox wrote:
> > > >   - ide-scsi needs some cleanup to fix switchover ide-cd/scsi
> > > >     (We can't dump ide-scsi)
> > > >   - Unregister path has races which cause all the long
> > > >     standing problems with pcmcia and prevents pci unreg
> > > 
> > > Can't you just disable module unloading for the release ?
> > 
> > Only if I can also nail shut your PCMCIA slot, disallow SATA and remove
> > some ioctls people use for docking.
> 
> Hmm? Didn't that all work in 2.4 with monolithic IDE ?

The pcmcia stuff crashes erratically with the old IDE the bugs have not
changed but the problem now shows up far more often. With a 2Gb type II
PCMCIA IDE disk I get a crash about 1 in 3 ejects now.

Re: any chance of 2.6.0-test*?

[Paul Mackerras]

Alan Cox writes:

> The pcmcia stuff crashes erratically with the old IDE the bugs have not
> changed but the problem now shows up far more often. With a 2Gb type II
> PCMCIA IDE disk I get a crash about 1 in 3 ejects now.

I'm using the patch below, which makes sure that ide_release (in
ide-cs.c) runs in process context not interrupt context.  The specific
problem I saw was that ide_unregister calls various devfs routines
that don't like being called in interrupt context, but there may well
be other thing ide_unregister does that aren't safe in interrupt
context.

I am also hitting another problem on my powerbook that I would like
your advice on.  The thing is that my powerbook (a 500MHz G4 titanium
powerbook) has 3 IDE interfaces, but only the first two have anything
attached.  The pmac IDE driver goes through and initializes all three,
including setting hwif->selectproc for each one.  Then, when I plug in
a CF card in the pcmcia/cardbus slot, ide_register_hw goes and looks
for a free ide_hwifs[] entry.  Slot 2 appears to be available because
there are no devices attached, and so it uses that.  In the process of
setting up ide_hwifs[2], it sets hwif->io_ports but it doesn't reset
selectproc (or hwif->OUTB etc., AFAICS).  The result is that it calls
the pmac IDE selectproc, which crashes (since we munged hwif->io_ports
without telling it).

I have "fixed" the problem for now by adding a call to
init_hwif_data(index) in ide_register_hw just before the first
memcpy.  I'd be interested to know what the right fix is. :)

Regards,
Paul.

diff -urN linux-2.4/drivers/ide/legacy/ide-cs.c pmac/drivers/ide/legacy/ide-cs.c
--- linux-2.4/drivers/ide/legacy/ide-cs.c	2002-12-19 11:50:35.000000000 +1100
+++ pmac/drivers/ide/legacy/ide-cs.c	2003-01-13 14:02:31.000000000 +1100
@@ -92,10 +92,11 @@
     int		ndev;
     dev_node_t	node;
     int		hd;
+    struct tq_struct rel_task;
 } ide_info_t;
 
 static void ide_config(dev_link_t *link);
-static void ide_release(u_long arg);
+static void ide_release(void *arg);
 static int ide_event(event_t event, int priority,
 		     event_callback_args_t *args);
 
@@ -136,9 +137,8 @@
     if (!info) return NULL;
     memset(info, 0, sizeof(*info));
     link = &info->link; link->priv = info;
+    INIT_TQUEUE(&info->rel_task, ide_release, link);
 
-    link->release.function = &ide_release;
-    link->release.data = (u_long)link;
     link->io.Attributes1 = IO_DATA_PATH_WIDTH_AUTO;
     link->io.Attributes2 = IO_DATA_PATH_WIDTH_8;
     link->io.IOAddrLines = 3;
@@ -187,6 +187,7 @@
 static void ide_detach(dev_link_t *link)
 {
     dev_link_t **linkp;
+    ide_info_t *info = link->priv;
     int ret;
 
     DEBUG(0, "ide_detach(0x%p)\n", link);
@@ -197,9 +198,10 @@
     if (*linkp == NULL)
 	return;
 
-    del_timer(&link->release);
-    if (link->state & DEV_CONFIG)
-	ide_release((u_long)link);
+    if (link->state & DEV_CONFIG) {
+	schedule_task(&info->rel_task);
+	flush_scheduled_tasks();
+    }
     
     if (link->handle) {
 	ret = CardServices(DeregisterClient, link->handle);
@@ -209,7 +211,7 @@
     
     /* Unlink, free device structure */
     *linkp = link->next;
-    kfree(link->priv);
+    kfree(info);
     
 } /* ide_detach */
 
@@ -381,7 +383,7 @@
 cs_failed:
     cs_error(link->handle, last_fn, last_ret);
 failed:
-    ide_release((u_long)link);
+    ide_release(link);
 
 } /* ide_config */
 
@@ -393,12 +395,15 @@
     
 ======================================================================*/
 
-void ide_release(u_long arg)
+static void ide_release(void *arg)
 {
-    dev_link_t *link = (dev_link_t *)arg;
+    dev_link_t *link = arg;
     ide_info_t *info = link->priv;
     
-    DEBUG(0, "ide_release(0x%p)\n", link);
+    if (!(link->state & DEV_CONFIG))
+	return;
+
+    DEBUG(0, "ide_do_release(0x%p)\n", link);
 
     if (info->ndev) {
         /* FIXME: if this fails we need to queue the cleanup somehow
@@ -435,6 +440,7 @@
 	      event_callback_args_t *args)
 {
     dev_link_t *link = args->client_data;
+    ide_info_t *info = link->priv;
 
     DEBUG(1, "ide_event(0x%06x)\n", event);
     
@@ -442,7 +448,7 @@
     case CS_EVENT_CARD_REMOVAL:
 	link->state &= ~DEV_PRESENT;
 	if (link->state & DEV_CONFIG)
-	    mod_timer(&link->release, jiffies + HZ/20);
+	    schedule_task(&info->rel_task);
 	break;
     case CS_EVENT_CARD_INSERTION:
 	link->state |= DEV_PRESENT | DEV_CONFIG_PENDING;

Re: any chance of 2.6.0-test*?

[Alan Cox]

On Mon, 2003-01-13 at 03:33, Paul Mackerras wrote:
> I'm using the patch below, which makes sure that ide_release (in
> ide-cs.c) runs in process context not interrupt context.  The specific
> problem I saw was that ide_unregister calls various devfs routines
> that don't like being called in interrupt context, but there may well
> be other thing ide_unregister does that aren't safe in interrupt
> context.

The ide release code isnt safe in any context.

> I have "fixed" the problem for now by adding a call to
> init_hwif_data(index) in ide_register_hw just before the first
> memcpy.  I'd be interested to know what the right fix is. :)

The code is currently written on the basis that people don't mangle
free interfaces (the free up code restores stuff which I grant is
kind of ass backwards). It also needs serious review and 2.5 testing
before I'd want to move it to the right spot.


Also note that freeing the IDE can fail. If it fails then the code
should probably be a lot smarter. Right now it 'loses' the interface.
Really it should set a timer and try again. It might also want to
add a null iops (out does nothing in returns FFFFFFFF) to stop
further I/O cycles.

Re: any chance of 2.6.0-test*?

[Benjamin Herrenschmidt]

On Mon, 2003-01-13 at 15:59, Alan Cox wrote:
> On Mon, 2003-01-13 at 03:33, Paul Mackerras wrote:
> > I'm using the patch below, which makes sure that ide_release (in
> > ide-cs.c) runs in process context not interrupt context.  The specific
> > problem I saw was that ide_unregister calls various devfs routines
> > that don't like being called in interrupt context, but there may well
> > be other thing ide_unregister does that aren't safe in interrupt
> > context.
> 
> The ide release code isnt safe in any context.

Yup, though Paul's patch is a first step in the right way as I don't
think anybody sane plan to fix the IDE release code to make it interrupt
safe ;) At least I don't...
 
> > I have "fixed" the problem for now by adding a call to
> > init_hwif_data(index) in ide_register_hw just before the first
> > memcpy.  I'd be interested to know what the right fix is. :)
> 
> The code is currently written on the basis that people don't mangle
> free interfaces (the free up code restores stuff which I grant is
> kind of ass backwards). It also needs serious review and 2.5 testing
> before I'd want to move it to the right spot.

The code is indeed strangely backward, I had to deal with that in 2.4
for the PowerBook hotswap CD bay and will soon have to review that for
2.4.21-pre & 2.5.

> Also note that freeing the IDE can fail. If it fails then the code
> should probably be a lot smarter. Right now it 'loses' the interface.
> Really it should set a timer and try again. It might also want to
> add a null iops (out does nothing in returns FFFFFFFF) to stop
> further I/O cycles.

Yup, this have been a problem for me too, as ide_unregister for example
fails with a mounted fs. So the user had effectively removed the drive
from the bay, but I couldn't free the interface... nasty. Especially if
the user then plugs some different IDE device in the bay, the kernel
will get completely confused.

Re: any chance of 2.6.0-test*?

[Pavel Machek]

Hi!

> >   - Flaws in error recovery paths in certain situations
> >   - Lots of random oopses on boot/remove that were apparently
> >     introduced by the kobject/sysfs people and need chasing
> >     down. (There are some non sysfs ones mostly fixed)
> 
> I guess the kobject/sysfs stuff could be ripped out if it doesn't
> work - it is probably not a "must have" feature.

sysfs is  needed to properly flush caches on powerdown and for
S3/S4 suspend/resume to work.
								Pavel
-- 
Worst form of spam? Adding advertisment signatures ala sourceforge.net.
What goes next? Inserting advertisment *into* email?

Re: any chance of 2.6.0-test*?

[David Truog]

Alan Cox wrote:
>> I've looked into this, and wow, it's not a simple
fix :(
>> 
>> But this is really the first it's been mentioned, I
can't see holding up
>> 2.6 for this.  It's a 2.7 job at the earliest,
unless someone wants to
>> step up and do it right now...
>
>2.5.x crashes erratically and randomly under high
tty/pty load. At the 
>moment I'm assuming this is the tty code. That means
we can't decide not
>to fix it since its already fatally broken.

Is there a list of things broken and "homeless"?  I
have seen a number of
people say (and do) fix things when they are
specificaly pointed out as
being broken.  if a list doesn't exist, could linus or
alan post one 
somewhere?  like the LKML maintainers list backwards.

just a thought

David Truog

RE: any chance of 2.6.0-test*?

[Rob Wilkens]

On Sun, 2003-01-12 at 15:23, Mikael Starvik wrote:
> Why do you think Linus would not be capable of deciding
> the best code to do a certain sequence of operations
> in Linux?

Did I imply he was incapable?  Of course, making decisions requires
input, and I was simply offerring input.  Whether or not he reads my
input is entirely up to him.  My advice is free, and as such is given at
face value.

> In short: Linus and others does not care at all what you 

If he didn't care he wouldn't write back with such strong words as
[Nicklaus Wirth] "doesn't have a frigging clue."  

Obviously Mr. Torvalds cares about the topic enough to say something. 
He has no obligation to explain his position to me, though.  I'm a
nobody.

-Rob

Re: any chance of 2.6.0-test*?

[Manfred Spraul]

[-- Attachment #1: Type: text/plain, Size: 784 bytes --]

Jens wrote:

>On Mon, Jan 13 2003, Terje Eggestad wrote:
>> > > I have the console on a serial port, and a terminal server. With kdb,
>> > > you can enter the kernel i kdb even when deadlocked.
>> > 
>> > Even if spinning with interrupt disabled?
>> 
>> Haven't painted myself into that corner yet. Doubt it, very much.
>
>These are the nasty hangs, total lockup and no info at all if it wasn't
>for the nmi watchdog triggering. That alone is reason enough for me :-)
>  
>
One alternative are the UP debug spinlocks: They detect double-locks.
2.4 contains macros, but I'm not sure if they work.
The macros were removed in 2.5, but if anyone is interested, I've 
attached my debug spinlock patch: for double locks, it even prints the 
filename/lineno of the offender.

--
    Manfred

[-- Attachment #2: patch-debug-sp --]
[-- Type: text/plain, Size: 4189 bytes --]

--- 2.5/include/linux/spinlock.h	2002-12-14 10:06:57.000000000 +0100
+++ build-2.5/include/linux/spinlock.h	2002-12-21 14:36:42.000000000 +0100
@@ -37,30 +37,120 @@
 #ifdef CONFIG_SMP
 #include <asm/spinlock.h>
 
-/*
- * !CONFIG_SMP and spin_lock_init not previously defined
- * (e.g. by including include/asm/spinlock.h)
- */
-#elif !defined(spin_lock_init)
+#else
 
-#ifndef CONFIG_PREEMPT
+#if !defined(CONFIG_PREEMPT) && !defined(CONFIG_DEBUG_SPINLOCK)
 # define atomic_dec_and_lock(atomic,lock) atomic_dec_and_test(atomic)
 # define ATOMIC_DEC_AND_LOCK
 #endif
 
+#ifdef CONFIG_DEBUG_SPINLOCK
+ 
+#define SPINLOCK_MAGIC	0x1D244B3C
+typedef struct {
+	unsigned long magic;
+	volatile unsigned long lock;
+	volatile unsigned int babble;
+	const char *module;
+	char *owner;
+	int oline;
+} spinlock_t;
+#define SPIN_LOCK_UNLOCKED (spinlock_t) { SPINLOCK_MAGIC, 0, 10, __FILE__ , NULL, 0}
+
+#define spin_lock_init(x) \
+	do { \
+		(x)->magic = SPINLOCK_MAGIC; \
+		(x)->lock = 0; \
+		(x)->babble = 5; \
+		(x)->module = __FILE__; \
+		(x)->owner = NULL; \
+		(x)->oline = 0; \
+	} while (0)
+
+#define CHECK_LOCK(x) \
+	do { \
+	 	if ((x)->magic != SPINLOCK_MAGIC) { \
+			printk(KERN_ERR "%s:%d: spin_is_locked on uninitialized spinlock %p.\n", \
+					__FILE__, __LINE__, (x)); \
+		} \
+	} while(0)
+
+#define _raw_spin_lock(x)		\
+	do { \
+	 	CHECK_LOCK(x); \
+		if ((x)->lock&&(x)->babble) { \
+			printk("%s:%d: spin_lock(%s:%p) already locked by %s/%d\n", \
+					__FILE__,__LINE__, (x)->module, \
+					(x), (x)->owner, (x)->oline); \
+			(x)->babble--; \
+		} \
+		(x)->lock = 1; \
+		(x)->owner = __FILE__; \
+		(x)->oline = __LINE__; \
+	} while (0)
+
+/* without debugging, spin_is_locked on UP always says
+ * FALSE. --> printk if already locked. */
+#define spin_is_locked(x) \
+	({ \
+	 	CHECK_LOCK(x); \
+		if ((x)->lock&&(x)->babble) { \
+			printk("%s:%d: spin_is_locked(%s:%p) already locked by %s/%d\n", \
+					__FILE__,__LINE__, (x)->module, \
+					(x), (x)->owner, (x)->oline); \
+			(x)->babble--; \
+		} \
+		0; \
+	})
+
+/* without debugging, spin_trylock on UP always says
+ * TRUE. --> printk if already locked. */
+#define _raw_spin_trylock(x) \
+	({ \
+	 	CHECK_LOCK(x); \
+		if ((x)->lock&&(x)->babble) { \
+			printk("%s:%d: spin_trylock(%s:%p) already locked by %s/%d\n", \
+					__FILE__,__LINE__, (x)->module, \
+					(x), (x)->owner, (x)->oline); \
+			(x)->babble--; \
+		} \
+		(x)->lock = 1; \
+		(x)->owner = __FILE__; \
+		(x)->oline = __LINE__; \
+		1; \
+	})
+
+#define spin_unlock_wait(x)	\
+	do { \
+	 	CHECK_LOCK(x); \
+		if ((x)->lock&&(x)->babble) { \
+			printk("%s:%d: spin_unlock_wait(%s:%p) owned by %s/%d\n", \
+					__FILE__,__LINE__, (x)->module, (x), \
+					(x)->owner, (x)->oline); \
+			(x)->babble--; \
+		}\
+	} while (0)
+
+#define _raw_spin_unlock(x) \
+	do { \
+	 	CHECK_LOCK(x); \
+		if (!(x)->lock&&(x)->babble) { \
+			printk("%s:%d: spin_unlock(%s:%p) not locked\n", \
+					__FILE__,__LINE__, (x)->module, (x));\
+			(x)->babble--; \
+		} \
+		(x)->lock = 0; \
+	} while (0)
+#else
 /*
  * gcc versions before ~2.95 have a nasty bug with empty initializers.
  */
 #if (__GNUC__ > 2)
   typedef struct { } spinlock_t;
-  typedef struct { } rwlock_t;
   #define SPIN_LOCK_UNLOCKED (spinlock_t) { }
-  #define RW_LOCK_UNLOCKED (rwlock_t) { }
 #else
   typedef struct { int gcc_is_buggy; } spinlock_t;
-  typedef struct { int gcc_is_buggy; } rwlock_t;
   #define SPIN_LOCK_UNLOCKED (spinlock_t) { 0 }
-  #define RW_LOCK_UNLOCKED (rwlock_t) { 0 }
 #endif
 
 /*
@@ -72,6 +162,18 @@
 #define _raw_spin_trylock(lock)	((void)(lock), 1)
 #define spin_unlock_wait(lock)	do { (void)(lock); } while(0)
 #define _raw_spin_unlock(lock)	do { (void)(lock); } while(0)
+#endif /* CONFIG_DEBUG_SPINLOCK */
+
+/* RW spinlocks: No debug version */
+
+#if (__GNUC__ > 2)
+  typedef struct { } rwlock_t;
+  #define RW_LOCK_UNLOCKED (rwlock_t) { }
+#else
+  typedef struct { int gcc_is_buggy; } rwlock_t;
+  #define RW_LOCK_UNLOCKED (rwlock_t) { 0 }
+#endif
+
 #define rwlock_init(lock)	do { (void)(lock); } while(0)
 #define _raw_read_lock(lock)	do { (void)(lock); } while(0)
 #define _raw_read_unlock(lock)	do { (void)(lock); } while(0)

Re: any chance of 2.6.0-test*?

[Albert D. Cahalan]

Matti Aarnio writes:

> Advanced optimizer hinting features, like unlikely() attribute
> are very new in (this) compiler, and while they in theory move the
> "unlikely" codes out of the fast-path, such things have been buggy
> in the past, and we are worried of bug effects... 

I've been wondering about this as the goto-thread spewed by.

As I recall, gcc recently started moving basic blocks around.
This destroyed most of the careful goto-based optimizations.
Now we're supposed to use likely() and unlikely() instead.

Hmmm?

BTW, what I'd like is a way to change optimization settings
on a per-function or even per-block basis. Telling gcc to
unroll a specific loop or pack a function into a tiny space
would be really cool.   __attribute__((__opt__("-Os")))
I could go for an "assume default case can't happen" too.

Re: any chance of 2.6.0-test*?

[Kai Henningsen]

torvalds@transmeta.com (Linus Torvalds)  wrote on 12.01.03 in <20030112204005$7287@gated-at.bofh.it>:

> No, you've been brainwashed by CS people who thought that Niklaus Wirth
> actually knew what he was talking about. He didn't. He doesn't have a
> frigging clue.

Well ... he did have some nice ideas. Unfortunately, they usually don't  
get palatable until someone else has worked on them.

If you look into Wirth's books and see that the example code in there ...

* uses one(!) space indentation
* routinely puts several statements on one line
* typically uses one- or two-character variable names

... then you realize that *every* Wirth source needs rewriting.

---
Niklaus Wirth has lamented that, whereas Europeans pronounce his name correctly
(Ni-klows Virt), Americans invariably mangle it into (Nick-les Worth).  Which
is to say that Europeans call him by name, but Americans call him by value.
---
#if _FP_W_TYPE_SIZE < 32
#error "Here's a nickle kid.  Go buy yourself a real computer."
#endif
	-- linux/arch/sparc64/double.h
---

MfG Kai

RE: any chance of 2.6.0-test*?

[Scott Robert Ladd]

Kai Henningsen wrote:
> Well ... he did have some nice ideas. Unfortunately, they usually don't
> get palatable until someone else has worked on them.

I actually *liked* Modula-2... but I haven't used it since the very early
90s. I guess I didn't like it *that* much, huh? ;)

> If you look into Wirth's books and see that the example code in there ...
>
> * uses one(!) space indentation
> * routinely puts several statements on one line

I've got a copy of the classic "Algorithms + Data Structures = Programs"; in
it, Wirth uses four-space indentation but a proportional font. The style
isn't all that bad. Hell, I've got worse stuff in some of *my* books --
oops, shouldn't have typed that... ;)

One-space indents may very well be an artifact of idiot copy editors, and
not the author.

> * typically uses one- or two-character variable names

There is one instance when one/two-character variable names make sense:
mathematical code that directly implements numericla algorithms from the
text. In such a case, short variable names correspond directly to standard
notation; using longer names would actually obscure the correspondence
between text and code. I also don't see the point of using "array_index"
over a plain old traditional "i" in a loop.

Rarely is any coding "rule" absolute. The point is clarity; if a "goto" or
one-character identifier make sense, use'em.

For those who growl -- I think this kind of discussion *has* value in the
kernel mailing list. Kernel newbies and such can learn a great deal from
rational, calm debates among experts; if they learn, their contributions to
the kernel will be better.

Of course, note the "rational" and "calm" above, which does not apply to the
Stallman debate... ;)

I enjoy the implementation debates; they give me a better idea of where the
kernel is going, so I can figure out where to stick my oar in the waters.

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Professional programming for science and engineering;
Interesting and unusual bits of very free code.

Re: any chance of 2.6.0-test*?

[Marcelo Pacheco]

>
>
>For those who growl -- I think this kind of discussion *has* value in the
>kernel mailing list. Kernel newbies and such can learn a great deal from
>rational, calm debates among experts; if they learn, their contributions to
>the kernel will be better.
>
>Of course, note the "rational" and "calm" above, which does not apply to the
>Stallman debate... ;)
>
>I enjoy the implementation debates; they give me a better idea of where the
>kernel is going, so I can figure out where to stick my oar in the waters.
>
Yes, but how many times this discussion already happened ?
I have nothing against the discussion happening ... As long as it 
doesn't happen a couple of times a year. That's what a list archive is 
for. The latest generation of Internet users, including some latest 
generation developers are in general LAZY about looking up archives, 
they want to post the question now and get an answer, while they could 
have gotten the answer without bothering anyone. Sure, it takes one a 
little more effort to search the archives, but that's one person's 
effort, while posting one message wastes hundreds (perhaps thousands) of 
people's time.
And also, the FAQ for the list states it quite clearly that the people 
doing usefull work here is only really interested in proof that 
something is better, performance and/or space wise, so if someone has an 
opinion, first go prove it on some piece of the kernel, the post their 
opinions along with the results to the list. That's the way to go. 
Maintenability is important, after speed, memory usage and portability. 
That's sound engineering, instead of C.S. dogmas.

Marcelo Pacheco

Re: any chance of 2.6.0-test*?

[Kai Henningsen]

scott@coyotegulch.com (Scott Robert Ladd)  wrote on 15.01.03 in <FKEAJLBKJCGBDJJIPJLJKEJFEDAA.scott@coyotegulch.com>:

> Kai Henningsen wrote:
> > Well ... he did have some nice ideas. Unfortunately, they usually don't
> > get palatable until someone else has worked on them.
>
> I actually *liked* Modula-2... but I haven't used it since the very early
> 90s. I guess I didn't like it *that* much, huh? ;)

Stuck shift key syndrome.

> > If you look into Wirth's books and see that the example code in there ...
> >
> > * uses one(!) space indentation
> > * routinely puts several statements on one line
>
> I've got a copy of the classic "Algorithms + Data Structures = Programs"; in
> it, Wirth uses four-space indentation but a proportional font. The style
> isn't all that bad. Hell, I've got worse stuff in some of *my* books --
> oops, shouldn't have typed that... ;)
>
> One-space indents may very well be an artifact of idiot copy editors, and
> not the author.

I think I've actually seen that in more than one Wirth book, but it's been  
quite a while.

> > * typically uses one- or two-character variable names
>
> There is one instance when one/two-character variable names make sense:
> mathematical code that directly implements numericla algorithms from the
> text. In such a case, short variable names correspond directly to standard
> notation; using longer names would actually obscure the correspondence
> between text and code. I also don't see the point of using "array_index"
> over a plain old traditional "i" in a loop.

B-trees is the example I most remember. I once carefully typed his source  
in, then spent several days reformatting the source and putting in  
sensible variable names; then when that source was finally readable, I  
decided that several things in there needed doing differently - I don't  
remember the details, but IIRC some parameter/result passing stuff was  
rather poorly thought out.

If *I* were an IT prof and a student handed me code like that, I'd be  
sorely tempted to give that a "fail". Or at the very least a "just barely  
passed". Publishing it in a book means the editor fell down on his job.

> Rarely is any coding "rule" absolute. The point is clarity; if a "goto" or
> one-character identifier make sense, use'em.

And so I do.

> For those who growl -- I think this kind of discussion *has* value in the
> kernel mailing list. Kernel newbies and such can learn a great deal from
> rational, calm debates among experts; if they learn, their contributions to
> the kernel will be better.

That part is fine - what's bad is obvious newbies insisting they know  
better, two dozen times a day, and the rest of the world not treating them  
as the obvious kooks they are. Those are the painful parts of this thread.

> Of course, note the "rational" and "calm" above, which does not apply to the
> Stallman debate... ;)

Let's just say that my ideas about successful public relations (note I'm  
not talking about programming here) seem to be diametrally opposite to  
those of rms, and leave it at that. (Ok, I feel I need to add that while I  
don't personally like esr, his PR efforts seem to have been a lot more  
successful - and I completely reject the notion that Open Source anything  
more than a different name for Free Software. A rose by any other name  
...)

> I enjoy the implementation debates; they give me a better idea of where the
> kernel is going, so I can figure out where to stick my oar in the waters.

Similar here.


MfG Kai

Re: any chance of 2.6.0-test*?

[Werner Almesberger]

Kai Henningsen wrote:
> * typically uses one- or two-character variable names

At ETH, we jokingly called this WNF ("Wirth's Normal Form". Well,
actually "Wirthsche Normalform"): shortening any identifier name
would make it conflict with another one.

Isn't it funny how irrelevant those Pascal-family languages
become, once you grok C ? :-)

- Werner

-- 
  _________________________________________________________________________
 / Werner Almesberger, Buenos Aires, Argentina         wa@almesberger.net /
/_http://www.almesberger.net/____________________________________________/

Re: any chance of 2.6.0-test*?

[Kai Henningsen]

terje.eggestad@scali.com (Terje Eggestad)  wrote on 13.01.03 in <20030113144017$68de@gated-at.bofh.it>:

> On man, 2003-01-13 at 12:09, Eric W. Biederman wrote:

> http://www.cs.utexas.edu/users/EWD/ewd02xx/EWD215.PDF
> to be exact. Reading it you can tell exactly how much of an
> mathematician Dijkstra really was.  At these times It's best to keep in
> mind a quote:
>
> "I used to understand the Theory of Relativity, but then the
> mathematicians got hold of it."
>   -- Albert Einstein

Wasn't Dijkstra one of the guys behind Algol 68? Yes, the one where some  
Geek test asks if you've read the revised report? ... and did you  
understand it? (I did, even found a bug in the library chapter (a simple  
case of exporting a pointer to local storage), but boy was that the  
hardest-to-understand language definition I've ever seen! Did complete  
symbol definition-and-usage checks purely in the (context-dependant  
meta-)grammar.)

MfG Kai

Re: any chance of 2.6.0-test*?

[Kai Henningsen]

robw@optonline.net (Rob Wilkens)  wrote on 13.01.03 in <20030113035007$451e@gated-at.bofh.it>:

> I have unlimitted bandwidth, as do many others.  IF they don't they
> shouldn't subscribe to this list, they should subscribe to the newsgroup
> and pick-and-choose by sbject/thread.

Just your typical clueless troll. What are you trying to do, become most- 
plonked guy on l-k?

MfG Kai

Re: any chance of 2.6.0-test*?

[Kai Henningsen]

Emiliano.Gabrielli@roma2.infn.it (Emiliano Gabrielli)  wrote on 13.01.03 in <20030113170014$20a1@gated-at.bofh.it>:

> On 01:38, lunedì 13 gennaio 2003, Rob Wilkens wrote:
> > It only goes to show that the core of computer science hasn't changed
> > over the years.  While some technology changes, the science itself stays
> > the same.  Much like a physicist should know the outdated theories of
> > relativity by Albert Einstein (made way back in the 20th century as
> > well) a good computer scientist should appreciate the contributions of
> > its founders, and have a good grasp of their core contributions.  I'm
> > not claiming to be an expert in either field, however.
>
> You think Relativity is outdated ?!?  I can't beleave !

I seem to recall that there are some areas in the world (carefully not  
naming any examples here) where some religious extremists have managed to  
sabotage public education, especially the sciences. If he's a victim of  
those guys, that would explain a lot (and shift some part of the blame  
elsewhere) ...

MfG Kai