Linux 2.4.28-rc3

Linux 2.4.28-rc3

[Marcelo Tosatti]

Hi,

Here goes the third release candidate.

It contains a v2.6 backport of the binfmt_elf potential vulnerabilities 
disclosed this week, an enhanced smbfs client overflow fix, an ACPI update 
fixing a couple of nasty bugs, a NFS client bugfix and a network update
from Davem.



Summary of changes from v2.4.28-rc2 to v2.4.28-rc3
============================================

Aaron Grothe:
  o [CRYPTO]: Add Anubis support

Chris Wright:
  o binfmt_elf: handle partial reads gracefully

David S. Miller:
  o [TG3]: Use ioremap_nocache()
  o [TG3]: Bump driver version and reldate
  o Cset exclude: davem@nuts.davemloft.net|ChangeSet|20040831000448|00808
  o Cset exclude: pablo@eurodev.net|ChangeSet|20040831000223|00117
  o [ATM]: Put back mistakedly removed LEC procfs code

Herbert Xu:
  o [NET]: Fix tbl->entries race

Len Brown:
  o [ACPI] fix ASUS boot crash http://bugzilla.kernel.org/show_bug.cgi?id=2755
  o [ACPI] fix poweroff regression backport from 2.6 and ACPICA 20040427 http://bugzilla.kernel.org/show_bug.cgi?id=2109

Marcelo Tosatti:
  o Changed EXTRAVERSION to -rc3

Mike Miller:
  o cleans up warnings in 32/64-bit conversions

Mike Waychison:
  o [TG3]: Fix fiber hw autoneg bounces

Patrick McHardy:
  o [PKT_SCHED]: Don't try to destroy builtin qdiscs

Stefan Esser:
  o Improved smbfs client overflow fix

Thomas Graf:
  o [NET]: Fix neighbour/arp build
  o [PKT_SCHED]: break is not enough to stop walking

Trond Myklebust:
  o NFS: Always wake up tasks that are waiting on the sillyrenamed file to complete

Wensong Zhang:
  o [IPVS]: Update version to 1.2.1

Re: Linux 2.4.28-rc3

[Andrey Melnikoff]

In article <20041112180052.GE23215@logos.cnet> you wrote:
> Hi,

> Here goes the third release candidate.

> It contains a v2.6 backport of the binfmt_elf potential vulnerabilities
> disclosed this week, an enhanced smbfs client overflow fix, an ACPI update
> fixing a couple of nasty bugs, a NFS client bugfix and a network update
> from Davem.

Any chance to apply this patch before release?

Prevent NMI oopser kill kernel thread when megearid2 driver wating abort or
reset command completion. 

Signed-off-by: Andrey Melnikov <temnota+kernel@kmv.ru>

--- linux-2.4.28-rc3/drivers/scsi/megaraid2.c~	Thu Nov 11 19:37:13 2004
+++ linux-2.4.28-rc3/drivers/scsi/megaraid2.c	Sat Nov 13 19:20:23 2004
@@ -39,6 +39,7 @@
 #include <linux/reboot.h>
 #include <linux/module.h>
 #include <linux/list.h>
+#include <linux/nmi.h>
 
 #include "sd.h"
 #include "scsi.h"
@@ -2820,6 +2821,7 @@
 
 		if( iter++ < MBOX_ABORT_SLEEP*1000 ) {
 			mdelay(1);
+			touch_nmi_watchdog();
 		}
 		else {
 			printk(KERN_WARNING
@@ -2900,6 +2902,7 @@
 
 		if( iter++ < MBOX_RESET_SLEEP*1000 ) {
 			mdelay(1);
+			touch_nmi_watchdog();
 		}
 		else {
 			printk(KERN_WARNING


-- 
 Best regards, TEMHOTA-RIPN aka MJA13-RIPE
 System Administrator. mailto:temnota@kmv.ru

Re: Linux 2.4.28-rc3

[Marcelo Tosatti]

On Sat, Nov 13, 2004 at 07:27:09PM +0300, Andrey Melnikoff wrote:
> In article <20041112180052.GE23215@logos.cnet> you wrote:
> > Hi,
> 
> > Here goes the third release candidate.
> 
> > It contains a v2.6 backport of the binfmt_elf potential vulnerabilities
> > disclosed this week, an enhanced smbfs client overflow fix, an ACPI update
> > fixing a couple of nasty bugs, a NFS client bugfix and a network update
> > from Davem.
> 
> Any chance to apply this patch before release?
> 
> Prevent NMI oopser kill kernel thread when megearid2 driver wating abort or
> reset command completion. 

Hi Andrey,

I talked to Atul and Arjan about this one - the correct thing to do is to 
replace mdelay() with CPU yielding msleep(). 

We should backport msleep() in 2.4.29-pre1. 


> Signed-off-by: Andrey Melnikov <temnota+kernel@kmv.ru>
> 
> --- linux-2.4.28-rc3/drivers/scsi/megaraid2.c~	Thu Nov 11 19:37:13 2004
> +++ linux-2.4.28-rc3/drivers/scsi/megaraid2.c	Sat Nov 13 19:20:23 2004
> @@ -39,6 +39,7 @@
>  #include <linux/reboot.h>
>  #include <linux/module.h>
>  #include <linux/list.h>
> +#include <linux/nmi.h>
>  
>  #include "sd.h"
>  #include "scsi.h"
> @@ -2820,6 +2821,7 @@
>  
>  		if( iter++ < MBOX_ABORT_SLEEP*1000 ) {
>  			mdelay(1);
> +			touch_nmi_watchdog();
>  		}
>  		else {
>  			printk(KERN_WARNING
> @@ -2900,6 +2902,7 @@
>  
>  		if( iter++ < MBOX_RESET_SLEEP*1000 ) {
>  			mdelay(1);
> +			touch_nmi_watchdog();
>  		}
>  		else {
>  			printk(KERN_WARNING
> 
> 
> -- 
>  Best regards, TEMHOTA-RIPN aka MJA13-RIPE
>  System Administrator. mailto:temnota@kmv.ru

Re: megaraid2.c PATCH (was: Linux 2.4.28-rc3)

[Andrey J. Melnikoff (TEMHOTA)]

Hi Marcelo Tosatti!
 On Sat, Nov 13, 2004 at 01:24:50PM -0200, Marcelo Tosatti wrote next:

> On Sat, Nov 13, 2004 at 07:27:09PM +0300, Andrey Melnikoff wrote:
> > In article <20041112180052.GE23215@logos.cnet> you wrote:
> > > Hi,
> > 
> > > Here goes the third release candidate.
> > 
> > > It contains a v2.6 backport of the binfmt_elf potential vulnerabilities
> > > disclosed this week, an enhanced smbfs client overflow fix, an ACPI update
> > > fixing a couple of nasty bugs, a NFS client bugfix and a network update
> > > from Davem.
> > 
> > Any chance to apply this patch before release?
> > 
> > Prevent NMI oopser kill kernel thread when megearid2 driver wating abort or
> > reset command completion. 
> 
> I talked to Atul and Arjan about this one - the correct thing to do is to 
> replace mdelay() with CPU yielding msleep(). 
> 
> We should backport msleep() in 2.4.29-pre1. 

Ok, msleep() backported, but driver isn't fixed. This patch acceptable?

Prevent NMI oopser kill kernel thread when megearid2 driver wating 
abort or reset command completion. 

Signed-off-by: Andrey Melnikov <temnota+kernel@kmv.ru>

--- linux-2.4.29/drivers/scsi/megaraid2.c~	Wed Jan 19 20:39:04 2005
+++ linux-2.4.29/drivers/scsi/megaraid2.c	Wed Jan 19 20:44:21 2005
@@ -2819,7 +2819,7 @@
 		}
 
 		if( iter++ < MBOX_ABORT_SLEEP*1000 ) {
-			mdelay(1);
+			msleep(1);
 		}
 		else {
 			printk(KERN_WARNING
@@ -2899,7 +2899,7 @@
 		}
 
 		if( iter++ < MBOX_RESET_SLEEP*1000 ) {
-			mdelay(1);
+			msleep(1);
 		}
 		else {
 			printk(KERN_WARNING
@@ -4040,10 +4040,10 @@
 	printk(KERN_INFO "megaraid: cache flush delay:   ");
 	for( i = 9; i >= 0; i-- ) {
 		printk("\b\b\b[%d]", i);
-		mdelay(1000);
+		msleep(1000);
 	}
 	printk("\b\b\b[done]\n");
-	mdelay(1000);
+	msleep(1000);
 
 	return NOTIFY_DONE;
 }

-- 
 Best regards, TEMHOTA-RIPN aka MJA13-RIPE
 System Administrator. mailto:temnota@kmv.ru

Re: megaraid2.c PATCH (was: Linux 2.4.28-rc3)

[Marcelo Tosatti]

On Wed, Jan 19, 2005 at 08:50:16PM +0300, Andrey J. Melnikoff (TEMHOTA) wrote:
> Hi Marcelo Tosatti!
>  On Sat, Nov 13, 2004 at 01:24:50PM -0200, Marcelo Tosatti wrote next:
> 
> > On Sat, Nov 13, 2004 at 07:27:09PM +0300, Andrey Melnikoff wrote:
> > > In article <20041112180052.GE23215@logos.cnet> you wrote:
> > > > Hi,
> > > 
> > > > Here goes the third release candidate.
> > > 
> > > > It contains a v2.6 backport of the binfmt_elf potential vulnerabilities
> > > > disclosed this week, an enhanced smbfs client overflow fix, an ACPI update
> > > > fixing a couple of nasty bugs, a NFS client bugfix and a network update
> > > > from Davem.
> > > 
> > > Any chance to apply this patch before release?
> > > 
> > > Prevent NMI oopser kill kernel thread when megearid2 driver wating abort or
> > > reset command completion. 
> > 
> > I talked to Atul and Arjan about this one - the correct thing to do is to 
> > replace mdelay() with CPU yielding msleep(). 
> > 
> > We should backport msleep() in 2.4.29-pre1. 
> 
> Ok, msleep() backported, but driver isn't fixed. This patch acceptable?
> 
> Prevent NMI oopser kill kernel thread when megearid2 driver wating 
> abort or reset command completion. 

The megaraid maintainers seem to have an scheduled driver update but they
have taken too long.

Patch applied, thanks.

> Signed-off-by: Andrey Melnikov <temnota+kernel@kmv.ru>
> 
> --- linux-2.4.29/drivers/scsi/megaraid2.c~	Wed Jan 19 20:39:04 2005
> +++ linux-2.4.29/drivers/scsi/megaraid2.c	Wed Jan 19 20:44:21 2005
> @@ -2819,7 +2819,7 @@
>  		}
>  
>  		if( iter++ < MBOX_ABORT_SLEEP*1000 ) {
> -			mdelay(1);
> +			msleep(1);
>  		}
>  		else {
>  			printk(KERN_WARNING
> @@ -2899,7 +2899,7 @@
>  		}
>  
>  		if( iter++ < MBOX_RESET_SLEEP*1000 ) {
> -			mdelay(1);
> +			msleep(1);
>  		}
>  		else {
>  			printk(KERN_WARNING
> @@ -4040,10 +4040,10 @@
>  	printk(KERN_INFO "megaraid: cache flush delay:   ");
>  	for( i = 9; i >= 0; i-- ) {
>  		printk("\b\b\b[%d]", i);
> -		mdelay(1000);
> +		msleep(1000);
>  	}
>  	printk("\b\b\b[done]\n");
> -	mdelay(1000);
> +	msleep(1000);
>  
>  	return NOTIFY_DONE;
>  }

[patch] 2.4.28-rc3: neigh_for_each must be EXPORT_SYMBOL'ed

[Adrian Bunk]

I'm getting the following error:

<--  snip  -->

depmod: *** Unresolved symbols in /lib/modules/2.4.28-rc3/kernel/net/decnet/decnet.o
depmod:         neigh_for_each

<--  snip  -->


This was caused by Harald's backport of the neighbour scalability fixes 
from 2.6 .


neigh_for_each must be EXPORT_SYMBOL'ed (as it is in 2.6):


Signed-off-by: Adrian Bunk <bunk@stusta.de>

--- linux-2.4.28-rc3-modular/net/core/Makefile.old	2004-11-13 19:40:36.000000000 +0100
+++ linux-2.4.28-rc3-modular/net/core/Makefile	2004-11-13 19:40:50.000000000 +0100
@@ -9,7 +9,7 @@
 
 O_TARGET := core.o
 
-export-objs := netfilter.o profile.o ethtool.o
+export-objs := netfilter.o profile.o ethtool.o neighbour.o
 
 obj-y := sock.o skbuff.o iovec.o datagram.o scm.o
 
--- linux-2.4.28-rc3-modular/net/core/neighbour.c.old	2004-11-13 19:35:26.000000000 +0100
+++ linux-2.4.28-rc3-modular/net/core/neighbour.c	2004-11-13 19:35:59.000000000 +0100
@@ -1569,6 +1569,7 @@
 	}
 	read_unlock_bh(&tbl->lock);
 }
+EXPORT_SYMBOL(neigh_for_each);
 
 /* The tbl->lock must be held as a writer and BH disabled. */
 void __neigh_for_each_release(struct neigh_table *tbl,

Re: [patch] 2.4.28-rc3: neigh_for_each must be EXPORT_SYMBOL'ed

[David S. Miller]

On Sat, 13 Nov 2004 21:07:35 +0100
Adrian Bunk <bunk@stusta.de> wrote:

> neigh_for_each must be EXPORT_SYMBOL'ed (as it is in 2.6):

Good catch Adrian, I'll apply this and push it to
Marcelo.

Thanks.

Re: Linux 2.4.28-rc3

[Willy Tarreau]

Hi Marcelo,

On Fri, Nov 12, 2004 at 04:00:52PM -0200, Marcelo Tosatti wrote:
> Here goes the third release candidate.

Just compiled it right now (with Adrian's patch applied).

results :
  - it builds and runs on athlon SMP+scsi (aic7xxx)
  - it builds and runs on sparc64 SMP+scsi (ncr), except for aic7xxx,
    which needs the attached patch to remove an unused function (aic7xxx
    compiles with -Werror).

Regards,
Willy

AIC7xxx build error on sparc64 :

sparc64-linux-gcc -D__KERNEL__ -I/usr/src/linux-2.4.28-rc3-fix/include -Wall -Wstrict-prototypes -Wno-trigraphs -O2 -fno-strict-aliasing -fno-common -fomit-frame-pointer -m64 -pipe -mno-fpu -mcpu=ultrasparc -mcmodel=medlow -ffixed-g4 -fcall-used-g5 -fcall-used-g7 -Wno-sign-compare -Wa,--undeclared-regs -finline-limit=100000 -DMODULE -I/usr/src/linux-2.4.28-rc3-fix/drivers/scsi -Werror -nostdinc -iwithprefix include -DKBUILD_BASENAME=aic79xx_osm_pci  -c -o aic79xx_osm_pci.o aic79xx_osm_pci.c
cc1: warnings being treated as errors
aic79xx_osm_pci.c:278: warning: `ahd_linux_pci_reserve_mem_region' defined but not used
make[1]: *** [aic79xx_osm_pci.o] Error 1
make[1]: Leaving directory `/usr/src/linux-2.4.28-rc3-fix/drivers/scsi/aic7xxx'
make: *** [_mod_drivers/scsi/aic7xxx] Error 2


Trivial patch (the function is used only if MMAPIO is defined) :

--- ./drivers/scsi/aic7xxx/aic79xx_osm_pci.c.orig	Sun Nov 14 10:17:41 2004
+++ ./drivers/scsi/aic7xxx/aic79xx_osm_pci.c	Sun Nov 14 10:19:35 2004
@@ -52,9 +52,12 @@
 					const struct pci_device_id *ent);
 static int	ahd_linux_pci_reserve_io_regions(struct ahd_softc *ahd,
 						 u_long *base, u_long *base2);
+#ifdef MMAPIO
 static int	ahd_linux_pci_reserve_mem_region(struct ahd_softc *ahd,
 						 u_long *bus_addr,
 						 uint8_t **maddr);
+#endif
+
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0)
 static void	ahd_linux_pci_dev_remove(struct pci_dev *pdev);
 
@@ -271,6 +274,7 @@
 	return (0);
 }
 
+#ifdef MMAPIO
 static int
 ahd_linux_pci_reserve_mem_region(struct ahd_softc *ahd,
 				 u_long *bus_addr,
@@ -318,6 +322,7 @@
 		error = ENOMEM;
 	return (error);
 }
+#endif
 
 int
 ahd_pci_map_registers(struct ahd_softc *ahd)

Re: [patch] 2.4.28-rc3: neigh_for_each must be EXPORT_SYMBOL'ed

[O.Sezer]

A similar export should also be needed for __neigh_for_each_release :

/sbin/depmod -ae -F System.map 2.4.28-rc3aac2
depmod: *** Unresolved symbols in 
/lib/modules/2.4.28-rc3aac2/kernel/net/atm/clip.o
depmod: 	__neigh_for_each_release

Ozkan Sezer

Re: [patch] 2.4.28-rc3: neigh_for_each must be EXPORT_SYMBOL'ed

[David S. Miller]

On Mon, 15 Nov 2004 17:27:56 +0200
"O.Sezer" <sezeroz@ttnet.net.tr> wrote:

> A similar export should also be needed for __neigh_for_each_release :
> 
> /sbin/depmod -ae -F System.map 2.4.28-rc3aac2
> depmod: *** Unresolved symbols in 
> /lib/modules/2.4.28-rc3aac2/kernel/net/atm/clip.o
> depmod: 	__neigh_for_each_release

Good catch, I've fixed this in my tree and will push
upstream.