Linux 2.4.26-pre2

Linux 2.4.26-pre2

[Marcelo Tosatti]

Hi, 

Here goes -pre2 -- it contains networking updates, network drivers 
updates, an XFS update, amongst others.

Detailed changelog follows

Summary of changes from v2.4.26-pre1 to v2.4.26-pre2
============================================

<dolbeau:irisa.fr>:
  o Small fix to pm3fb & MAINTAINERS

<i.palsenberg:jdirmedia.nl>:
  o [QLOGIC]: Mark mbox_param[] as static to avoid namespace pollution

<jon:focalhost.com>:
  o [CRYPTO]: Add ARC4 module

<jpk:sgi.com>:
  o [XFS] Merge missing mount stripe-unit/width-alignment check over from IRIX

<mlord:pobox.com>:
  o Fix vmalloc() error handling

Chas Williams:
  o [ATM]: [lec] timer cleanup
  o [ATM]: [lec] send queued packets immediately after path switch

Christoph Hellwig:
  o [XFS] Simplify pagebuf_rele / pagebuf_free
  o [XFS] Stop using sleep_on
  o [XFS] Plug a pagebuf race that got bigger with the recent cleanup
  o [XFS] Fix gcc 3.5 compilation for real
  o [XFS] Fix buffer teardown on _pagebuf_lookup_pages failure
  o [XFS] Remove the lockable/not lockable buffer distinction.  All metada buffers are lockable these days.
  o [XFS] Remove PBF_MAPPABLE
  o [XFS] plug a pagebuf leak
  o [XFS] "backport" d_alloc_anon (this time for real)
  o [XFS] Avoid NULL returns from pagebuf_get
  o [XFS] use generic XFS stats and sysctl infrastructure in pagebuf
  o [XFS] Fix up daemon names
  o [XFS] only lock pagecache pages
  o [XFS] plug race in pagebuf freeing
  o [XFS] kill some dead constants from pagebuf

David S. Miller:
  o [SUNGEM]: At end of RX completion chain, double check OWN bit with completion register
  o [IPV4]: Do not return -EAGAIN on blocking UDP socket, noticed by Olaf Kirch
  o [NET]: Set default socket rmem/wmem values more sanely and consistently
  o [IPV6]: UDPv6 needs recvmsg csum error path fix too, thanks Olaf
  o [SCTP]: Ranem MSECS_TO_JIFFIES to avoid conflict with IRDA
  o [SCTP]: Comment out buggy ipv6 debugging printk
  o [SPARC64]: Update defconfig
  o [SPARC]: Pass a real page into do_mount() a final arg
  o [SPARC]: Update defconfig

David Stevens:
  o [IGMP/MLD]: Verify MSFILTER option length
  o [IGMP/MLD]: Check for numsrc overflow, plus temp buffer tweaks

David Woodhouse:
  o [IPV6]: Init ipv6 before ipv4 if built statically into kernel

Dean Roehrich:
  o [XFS] Change DM_SEM_FLAG to DM_SEM_FLAG_RD

Don Fry:
  o 2.4.25 pcnet32.c SLAB_DEBUG length error fix
  o 2.4.25 pcnet32.c transmit hang fix
  o 2.4.25 pcnet32.c wrong vendor ID fix
  o 2.4.25 pcnet32.c oops in rmmod
  o 2.4.25 pcnet32.c bus master arbitration failure fix
  o 2.4.25 pcnet32.c convert to use netif_msg_*
  o 2.4.25 pcnet32.c change to use ethtool_ops
  o pcnet32.c handle failures in open
  o pcnet32.c another diff error fix
  o pcnet32.c non-mii errors with ethtool
  o pcnet32.c add .remove to pci_driver
  o pcnet32.c adds loopback test
  o pcnet32.c avoid transmit hang for 79C791
  o pcnet32 non-mii link state fix

Eric Sandeen:
  o [XFS] Add switches to make xfs compile when the nptl patch is present
  o [XFS] Remove some dead debug code
  o [XFS] Make more xfs errors trappable with panic_mask
  o [XFS] zero log buffer during initialization at mount time

François Romieu:
  o [netdrvr r8169] fix TX descriptor overflow

Geert Uytterhoeven:
  o [netdrvr tulip] fix up 21041 media selection

Harald Welte:
  o [NETFILTER]: Kill bogus const in list helpers
  o [NETFILTER]: Fix ECN target cloned skb problem
  o [NETFILTER]: Remove unused structure member in NAT, from Patrick McHardy

James Morris:
  o [CRYPTO]: Backport Christophe Saout's 2.6.x scatterlist code extraction

Jean Delvare:
  o Identify Radeon Ya and Yd in radeonfb

Len Brown:
  o ACPI URL update
  o [ACPI] interrupt over-ride for nforce from Maciej W. Rozycki
  o [ACPI] delete unnecessary dmesg lines, fix spelling
  o [ACPI] include CONFIG_ACPI_RELAXED_AML code always add acpi=strict option to disable platform workarounds
  o [ACPI] ACPICA 20040220 from Bob Moore
  o [ACPI] fix ia64 build error (Bjorn Helgaas)

Marcelo Tosatti:
  o devfs: Fix truncation of mount data as 2.6
  o Changed EXTRAVERSION to -pre2

Matthias Andree:
  o [NET]: Export sysctl_optmem_max to modules

Nathan Scott:
  o [XFS] Fix a trivial compiler warning, remove some no-longer-used macros
  o [XFS] Use list_move for moving pagebufs between lists, not list_add/list_del
  o [XFS] Fix compile warning, ensure _pagebuf_lookup_pages return value is inited
  o [XFS] Fix data loss when writing into unwritten extents while memory is being reclaimed
  o [XFS] Remove bogus assert I added during testing of previous unwritten fix
  o [XFS] Add I/O path tracing code, twas useful in diagnosing that last unwritten extent problem
  o [XFS] Use a naming convention here thats more consistent with everything else
  o [XFS] Fix BUG in debug trace code, it was plain wrong for the unmapped page case
  o [XFS] Fix the by-handle attr list interface (used by xfsdump) for security attrs
  o [XFS] Fix length of mount argument path strings, off by one
  o [XFS] release i_sem before going into dmapi queues
  o [XFS] Remove PBF_SYNC buffer flag, unused for some time now
  o [XFS] Sort out some minor differences between trees
  o [XFS] Fix a compiler warning from a redefined symbol

Shmulik Hen:
  o bonding: Add support for HW accel. slaves
  o bonding: Add VLAN support in TLB mode
  o bonding: Add VLAN support in ALB mode

Simon Barber:
  o [NET]: Capture skb->protocol after invoking bridge

Simon Horman:
  o [JHASH]: Make key arg const in jhash()

Sridhar Samudrala:
  o [SCTP] Fix packed attribute usage
  o [SCTP] Fix NIP6 macro to take a ptr to struct in6_addr
  o [SCTP] Fix incorrect INIT process termination with sinit_max_init_timeo

Timothy Shimmin:
  o [XFS] Add XFS_FS_GOINGDOWN interface to xfs
  o [XFS] Fix log recovery case when have v2 log with size >32K and we have a Log Record wrapping around the physical log end. Need to reset the pb size back to what we were using and NOT just 32K.
  o [XFS] Version 2 log fixes - remove l_stripemask and add v2 log stripe padding to ic_roundoff to cater for pad in reservation cursor updates.
  o [XFS] fix up some log debug code for when XFS_LOUD_RECOVERY is turned on

Xose Vazquez Perez:
  o more RTL-8139 clone boards
  o more ne2k-pci clone boards

Yasuyuki Kozakai:
  o [IPV6]: Fix frag hdr parsing in ipv6_skip_exthdr()
  o [IPV6]: Fix ip6_tables TCP/UDP matching when ipv6 ext hdr exists

Re: Linux 2.4.26-pre2

[Eyal Lebedinsky]

[-- Attachment #1: Type: text/plain, Size: 528 bytes --]

Marcelo Tosatti wrote:
> 
> Hi, 
> 
> Here goes -pre2 -- it contains networking updates, network drivers 
> updates, an XFS update, amongst others.
 >
> <jon:focalhost.com>:
>   o [CRYPTO]: Add ARC4 module

In standard C we declare all variables at the top of a function. While
some compilers allow extension, it is not a good idea to get used to
them if we want portable code.

If fails for me on Debian/stable.

Attached is a trivial patch for crypto/arc4.c.

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

[-- Attachment #2: 2.4.26-pre2-arc4.patch --]
[-- Type: text/plain, Size: 475 bytes --]

--- linux/crypto/arc4.c.orig	Sun Mar  7 16:33:11 2004
+++ linux/crypto/arc4.c	Sun Mar  7 16:35:12 2004
@@ -55,14 +55,14 @@
 static void arc4_crypt(void *ctx_arg, u8 *out, const u8 *in)
 {
 	struct arc4_ctx *ctx = ctx_arg;
+	u8 *const S, x, y, a, b;
 
-	u8 *const S = ctx->S;
-	u8 x = ctx->x;
-	u8 y = ctx->y;
-
-	u8 a = S[x];
+	S = ctx->S;
+	x = ctx->x;
+	y = ctx->y;
+	a = S[x];
 	y = (y + a) & 0xff;
-	u8 b = S[y];
+	b = S[y];
 	S[x] = b;
 	S[y] = a;
 	x = (x + 1) & 0xff;

Re: Linux 2.4.26-pre2

[Horst von Brand]

Eyal Lebedinsky <eyal@eyal.emu.id.au> said:
> Marcelo Tosatti wrote:
> > 
> > Hi, 
> > 
> > Here goes -pre2 -- it contains networking updates, network drivers 
> > updates, an XFS update, amongst others.
>  >
> > <jon:focalhost.com>:
> >   o [CRYPTO]: Add ARC4 module
> 
> In standard C we declare all variables at the top of a function. While
> some compilers allow extension, it is not a good idea to get used to
> them if we want portable code.

Oh, come on. This is _kernel_ code, it won't ever be compiled with anything
not GCC-compatible.

> If fails for me on Debian/stable.

What compiler is in use there?
-- 
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: Linux 2.4.26-pre2

[Roland Dreier]

    Eyal> In standard C we declare all variables at the top of a
    Eyal> function. While some compilers allow extension, it is not a
    Eyal> good idea to get used to them if we want portable code.

    Horst> Oh, come on. This is _kernel_ code, it won't ever be
    Horst> compiled with anything not GCC-compatible.

gcc 2.95 rejects declarations after code.  The kernel, especially
kernel 2.4, shouldn't use this particular extension, even if gcc 3
accepts it.

 - R.

Re: Linux 2.4.26-pre2

[David Weinehall]

On Sun, Mar 07, 2004 at 01:19:46PM -0300, Horst von Brand wrote:
> Eyal Lebedinsky <eyal@eyal.emu.id.au> said:
> > Marcelo Tosatti wrote:
> > > 
> > > Hi, 
> > > 
> > > Here goes -pre2 -- it contains networking updates, network drivers 
> > > updates, an XFS update, amongst others.
> >  >
> > > <jon:focalhost.com>:
> > >   o [CRYPTO]: Add ARC4 module
> > 
> > In standard C we declare all variables at the top of a function. While
> > some compilers allow extension, it is not a good idea to get used to
> > them if we want portable code.
> 
> Oh, come on. This is _kernel_ code, it won't ever be compiled with anything
> not GCC-compatible.

Ugly warts don't become any less ugly just because gcc accepts them...


Regards: David Weinehall
-- 
 /) David Weinehall <tao@acc.umu.se> /) Northern lights wander      (\
//  Maintainer of the v2.0 kernel   //  Dance across the winter sky //
\)  http://www.acc.umu.se/~tao/    (/   Full colour fire           (/

Re: Linux 2.4.26-pre2

[Michael Frank]

On 07 Mar 2004 10:49:05 -0800, Roland Dreier <roland@topspin.com> wrote:

>     Eyal> In standard C we declare all variables at the top of a
>     Eyal> function. While some compilers allow extension, it is not a
>     Eyal> good idea to get used to them if we want portable code.
>
>     Horst> Oh, come on. This is _kernel_ code, it won't ever be
>     Horst> compiled with anything not GCC-compatible.
>
> gcc 2.95 rejects declarations after code.  The kernel, especially
> kernel 2.4, shouldn't use this particular extension, even if gcc 3
> accepts it.
>

I also use gcc 2.95 and have no intention to change compilers
for 2.4 at all as this compiler is most dependable and never
screwed up unlike 3.1.x and 3.2.2.

Documentation/Codingstyle also refers to K&R who historically
declare variables at the top of a function.

Excerpt:
	"all right-thinking people know that
	(a) K&R are _right_ and (b) K&R are right"

Regards
Michael

Re: Linux 2.4.26-pre2

[Rene Herman]

David Weinehall wrote:

>>>In standard C we declare all variables at the top of a function. While
>>>some compilers allow extension, it is not a good idea to get used to
>>>them if we want portable code.
>>
>>Oh, come on. This is _kernel_ code, it won't ever be compiled with anything
>>not GCC-compatible.
> 
> Ugly warts don't become any less ugly just because gcc accepts them...

Mixing code and declarations is also c99. For (a sane) gcc specifically, 
you have to tell it -std=c89 -pedantic to have it even complain.

Rene.

Re: Linux 2.4.26-pre2

[Jeff Garzik]

Rene Herman wrote:
> David Weinehall wrote:
> 
>>>> In standard C we declare all variables at the top of a function. While
>>>> some compilers allow extension, it is not a good idea to get used to
>>>> them if we want portable code.
>>>
>>>
>>> Oh, come on. This is _kernel_ code, it won't ever be compiled with 
>>> anything
>>> not GCC-compatible.
>>
>>
>> Ugly warts don't become any less ugly just because gcc accepts them...
> 
> 
> Mixing code and declarations is also c99. For (a sane) gcc specifically, 
> you have to tell it -std=c89 -pedantic to have it even complain.

Agreed, with the proviso s/sane/new/

We want to support older gccs that do not support the C99/C++ syntax, 
for now.

	Jeff

Re: Linux 2.4.26-pre2

[David Weinehall]

On Sun, Mar 07, 2004 at 08:52:45PM +0100, Rene Herman wrote:
> David Weinehall wrote:
> 
> >>>In standard C we declare all variables at the top of a function. While
> >>>some compilers allow extension, it is not a good idea to get used to
> >>>them if we want portable code.
> >>
> >>Oh, come on. This is _kernel_ code, it won't ever be compiled with 
> >>anything
> >>not GCC-compatible.
> >
> >Ugly warts don't become any less ugly just because gcc accepts them...
> 
> Mixing code and declarations is also c99. For (a sane) gcc specifically, 
> you have to tell it -std=c89 -pedantic to have it even complain.

Ok, didn't know that.  Still doesn't make it any less ugly, though.
There are quite a lot of things that a valid in C.  That doesn't mean
they should be used...


Regards: David Weinehall
-- 
 /) David Weinehall <tao@acc.umu.se> /) Northern lights wander      (\
//  Maintainer of the v2.0 kernel   //  Dance across the winter sky //
\)  http://www.acc.umu.se/~tao/    (/   Full colour fire           (/

Re: Linux 2.4.26-pre2

[Michal Schmidt]

Hello,

> <mlord:pobox.com>:
>   o Fix vmalloc() error handling
> 

This change has a problem. In __vmalloc_area_pages when pmd_alloc fails, 
the spinlock will not be released. Maybe the following patch is needed? 
(only compile tested, I don't have SMP)

Michal Schmidt


--- linux-2.4.26-pre2/mm/vmalloc.c      Sun Mar  7 20:44:27 2004
+++ linux-2.4.26-pre2.mich/mm/vmalloc.c Sun Mar  7 20:54:02 2004
@@ -183,6 +183,7 @@ static inline int __vmalloc_area_pages (
  err:
         if (address > start)
                 vmfree_area_pages((address - start), address - start);
+       spin_unlock(&init_mm.page_table_lock);
         return -ENOMEM;
  }

Re: Linux 2.4.26-pre2

[Michael Frank]

On Sun, 7 Mar 2004 21:05:09 +0100, David Weinehall <tao@acc.umu.se> wrote:

> On Sun, Mar 07, 2004 at 08:52:45PM +0100, Rene Herman wrote:
>> David Weinehall wrote:
>>
>> >>>In standard C we declare all variables at the top of a function. While
>> >>>some compilers allow extension, it is not a good idea to get used to
>> >>>them if we want portable code.
>> >>
>> >>Oh, come on. This is _kernel_ code, it won't ever be compiled with
>> >>anything
>> >>not GCC-compatible.
>> >
>> >Ugly warts don't become any less ugly just because gcc accepts them...
>>
>> Mixing code and declarations is also c99. For (a sane) gcc specifically,
>> you have to tell it -std=c89 -pedantic to have it even complain.
>
> Ok, didn't know that.  Still doesn't make it any less ugly, though.
> There are quite a lot of things that a valid in C.  That doesn't mean
> they should be used...

C99 is C++ish. I have my experience with these methods after following
the PopularCppWays for some time. HellIsThisCrap which I wrote a pain
to maintain...

When a function is short and concise, there is no need to worry about
the few variables at the top.

Regards
Michael