Linux 2.2.20-pre4

Linux 2.2.20-pre4

[Alan Cox]

Linux 2.2 is now firmly into maintainance state. Patches for neat new ideas
belong in 2.4. Generally new drivers belong in 2.4 (possibly in 2.2 as well
after 2.4 shows them stable). Expect me to be very picky on changes to the
core code now. 

2.2.20pre4
o	Fix small corruption bug in 82596		(Andries Brouwer)
o	Fix usb printer probing				(Pete Zaitcev)
o	Fix swapon/procfs race				(Paul Menage)
o	Handle ide dma bug in the CS5530		(Mark Lord)
o	Backport 2.4 ipv6 neighbour discovery changes	(Dave Miller)
o	FIx sock_wmalloc error handling			(Dave Miller)
o	Enter quickack mode for out of window TCP data	(Andi Kleen)
o	Fix Established v SYN-ACK TCP state error	(Alexey Kuznetsov)
o	Sparc updates, ptrace changes etc		(Dave Miller)
o	Fix wrong printk in vdolive masq		(Keitaro Yosimura)
o	Fix core dump handling bugs in 2.2		(Al Viro)
o	Update hdlc and synclink drivers		(Paul Fulghum)
o	Update netlink help texts			(Magnus Damm)
o	Fix rtl8139 keeping files open			(Andrew Morton)
o	Further sk98 driver updates. fix wrong license	(Mirko Lindner)
	text in files
o	Jonathan Woithe has moved			(Jonathan Woithe)
o	Update cpqarray driver				(Charles White)
o	Update cciss driver				(Charles White)
o	Don't delete directories on an fs that reports	(Ingo Oeser)
	then 0 size when doing distclean
o	Add support for the 2.4 boot extensions to 2.2	(H Peter Anvin)
o	Fix nfs cache locking corruption on SMP		(Craig Hagan)
o	Add missing check to cdrom readaudio ioctl	(Jani Jaakkola)
o	Fix refclock build with newer gcc		(Jari Ruusu)
o	koi8-r fixes					(Andy Rysin)
o	Spelling fixes for documentation		(Andries Brouwer)

2.2.20pre3
o	FPU/ptrace corruption fixes			(Victor Zandy)
o	Resync belkin usb serial with 2.4		(Greg Kroah-Hartmann)
o	Resync digiport usb serial with 2.4		(Greg Kroah-Hartmann)
o	Rsync empeg usb serial with 2.4			(Greg Kroah-Hartmann)
o	Resync ftdi_sio against 2.4			(Greg Kroah-Hartmann)
o	Bring keyscan usb back into line with 2.4	(Greg Kroah-Hartmann)
o	Resync keyspan_pda usb with 2.4			(Greg Kroah-Hartmann)
o	Resync omninet usb with 2.4.5			(Greg Kroah-Hartmann)
o	Resync usb-serial driver with 2.4.5		(Greg Kroah-Hartmann)
o	Resync visor usb driver with 2.4.5		(Greg Kroah-Hartmann)
o	Rsync whiteheat driver with 2.4.5		(Greg Kroah-Hartmann)
o	Add edgeport USB serial				(Greg Kroah-Hartmann)
o	Add mct_u232 USB serial				(Greg Kroah-Hartmann)
o	Update usb storage device list		(Stas Bekman, Kaz Sasayma)
o	Bring usb acm driver into line with 2.4.5	(Greg Kroah-Hartmann)
o	Bring bluetooth driver into line with 2.4.5	(Greg Kroah-Hartmann)
o	Bring dabusb driver into line with 2.4.5	(Greg Kroah-Hartmann)
o	Bring usb dc2xx driver into line with 2.4.5	(Greg Kroah-Hartmann)
o	Bring mdc800 usb driver into line with 2.4.5	(Greg Kroah-Hartmann)
o	Bring rio driver into line with 2.4.5		(Greg Kroah-Hartmann)
o	Bring USB scanner drivers into line with 2.4.5	(Greg Kroah-Hartmann)
o	Update ov511 driver to match 2.4.5		(Greg Kroah-Hartmann)
o	Update PCIIOC ioctls (esp for sparc)		(Dave Miller)
o	General sparc bugfixes				(Dave Miller)
o	Fix possible oops in fbmem ioctls		(Dave Miller)
o	Fix reboot/halt bug on "Alcor" Alpha boxes	(Tom Vier)
o	Update osst driver 				(Willem Riede)
o	Fix syncppp negotiation bug			(Bob Dunlop)
o	SMBfs bug fixes from 2.4 series			(Urban Widmark)
o	3ware IDE raid driver updates			(Adam Radford)
o	Fix incorrect use of bitops on non long types	(Dave Miller)
o	Fix reboot/halt bug on 'Miata' Alpha boxes	(Tom Vier)
o	Update Tim Waugh's contact info			(Tim Waugh)
o	Add TIOCGSERIAL to sun serial on PCI sparc32	(Lars Kellogg-Stedman)
o	ov511 check user data more carefully		(Marc McClelland)
o	Fix netif_wake_queue compatibility macro	(Andi Kleen)

2.2.20pre2
o	Fix ip_decrease_ttl as per 2.4			(Dave Miller)
o	Fix tcp retransmit state bug			(Alexey Kuznetsov)
o	Fix a few obscure sparc tree bugs		(Dave Miller)
o	Fix fb /proc bug and OF fb name size bug	(Segher Boessenkool)
o	Fix complie with CONFIG_INTEL_RNG=y		(Andrzej Krzysztofowicz)
o	Fix rio driver when HZ!=100			(Andrzej Krzysztofowicz)
o	Stop 3c509 grabbing other EISA boards		(Andrzej Krzysztofowicz)
o	Remove surplus defines for root= names		(Andrzej Krzysztofowicz)
o	Revert pre1 APIC change

2.2.20pre1
o	Fix SMP deadlock in NFS				(Trond Myklebust)
o	Fix missing printk in bluesmoke handler		(me)
o	Fix sparc64 nfs					(Dave Miller)
o	Update io_apic code to avoid breaking dual	(Johannes Erdfelt)
	Athlon 760MP
o	Fix includes bugs in toshiba driver		(Justin Keene,
							 Greg Kroah-Hartmann)
o	Fix wanpipe cross compile			(Phil Blundell)
o	AGPGART copy_from_user fix			(Dawson Engler)
o	Fix alpha resource setup error			(Allan Frank)
o	Eicon driver updates				(Armind Schindler)
o	PC300 driver update				(Daniela Squassoni)
o	Show lock owner on flocks			(Jim Mintha)
o	Update cciss driver to 1.0.3			(Charles White)
o	Backport cciss/cpqarray security fixes		(me)
o	Update i810 random number generator		(Jeff Garzik)
o	Update sk98 driver				(Mirko Lindner)
o	Update sis900 ethernet driver			(Hui-Fen Hsu)
o	Fix checklist glitch in make menuconfig		(Moritz Schulte)
o	Update synclink driver				(Paul Fulghum)
o	Update advansys scsi driver			(Bob Frey)
o	Ver_linux fixes for 2.2				(Steven Cole)
o	Bring 2.2 back into line with the master ISDN	(Kai Germaschewski)
o	Whiteheat usb driver update			(Greg Kroah-Hartmann)
o	Fix via_rhine byte counters			(Adam Lackorzynski)
o	Fix modem control on rio serial			(Rogier Wolff)
o	Add more Iomega Zip to the usb storage list	(Wim Coekaerts)
o	Add ZF Micro watchdog 				(Fernando Fuganti)

Re: Linux 2.2.20-pre4

[Jochen Striepe]

        Hi,

On 19 Jun 2001, Alan Cox <laughing@shared-source.org> wrote:
> 
> 2.2.20pre4

Just to keep you informed... (I think there was a saying that there was
interest in experiences with compiling the kernel with non-recommended
gcc's ...)

I tried the newly released gcc-3.0 compiling 2.2.20pre4 (yes, I _know_ it
is not recommended): 

/usr/src/linux/include/linux/signal.h: In function `siginitset':
/usr/src/linux/include/linux/signal.h:193: warning: deprecated use of label at end of compound statement
/usr/src/linux/include/linux/signal.h: In function `siginitsetinv':
/usr/src/linux/include/linux/signal.h:205: warning: deprecated use of label at end of compound statement
sched.c: At top level:
sched.c:52: conflicting types for `xtime'
/usr/src/linux/include/linux/sched.h:509: previous declaration of `xtime'
sched.c: In function `schedule':
sched.c:739: warning: deprecated use of label at end of compound statement
make[2]: *** [sched.o] Error 1
make[2]: Leaving directory `/usr/src/linux-2.2.20pre4/kernel'
make[1]: *** [first_rule] Error 2
make[1]: Leaving directory `/usr/src/linux-2.2.20pre4/kernel'
make: *** [_dir_kernel] Error 2


$ sh /usr/src/linux/scripts/ver_linux 
If some fields are empty or look unusual you may have an old version.
Compare to the current minimal requirements in Documentation/Changes.
 
Linux tolot 2.4.6-pre3 #1 Wed Jun 13 09:55:57 CEST 2001 i586 unknown
 
Gnu C                  3.0
Gnu make               3.79.1
binutils               2.11.1
util-linux             2.11f
mount                  2.11f
modutils               2.4.6
e2fsprogs              1.21
PPP                    2.4.1
Linux C Library        2.2.3
Dynamic linker (ldd)   2.2.3
Procps                 2.0.7
Net-tools              1.60
Kbd                    1.06
Sh-utils               2.0.11
Modules Loaded         nls_utf8 nls_iso8859-15 nls_iso8859-2
nls_iso8859-1 nls_cp852 nls_cp850 nls_cp437 floppy sr_mod sg isofs
ne2k-pci 8390 ide-cd cdrom adlib_card opl3 sb sb_lib uart401 sound
soundcore ppp_generic slhc lp parport serial


So long,

Jochen.

-- 
Cahn's Axiom:
        When all else fails, read the instructions.

Re: Linux 2.2.20-pre4

[Alan Cox]

> Just to keep you informed... (I think there was a saying that there was
> interest in experiences with compiling the kernel with non-recommended
> gcc's ...)


> sched.c:52: conflicting types for `xtime'
> /usr/src/linux/include/linux/sched.h:509: previous declaration of `xtime'

Stick a volatile in the declaration. Thats a real bug it found

Re: Linux 2.2.20-pre4

[Jochen Striepe]

        Hi,

On 19 Jun 2001, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
> 
> > sched.c:52: conflicting types for `xtime'
> > /usr/src/linux/include/linux/sched.h:509: previous declaration of `xtime'
> 
> Stick a volatile in the declaration. Thats a real bug it found

Um...

I made it

extern volatile struct timeval xtime;

Now it stops with

/usr/src/linux/include/linux/sched.h: At top level:
/usr/src/linux/include/linux/sched.h:509: warning: useless keyword or
type name in empty declaration
In file included from /usr/src/linux/include/linux/blkdev.h:6,
                 from ksyms.c:15:
/usr/src/linux/include/linux/genhd.h: In function `ptype':
/usr/src/linux/include/linux/genhd.h:83: warning: deprecated use of
label at end of compound statement
ksyms.c: At top level:
ksyms.c:352: `xtime' undeclared here (not in a function)
ksyms.c:352: initializer element is not constant
ksyms.c:352: (near initialization for `__ksymtab_xtime.value')
make[2]: *** [ksyms.o] Error 1
make[2]: Leaving directory `/usr/src/linux-2.2.20pre4/kernel'
make[1]: *** [first_rule] Error 2
make[1]: Leaving directory `/usr/src/linux-2.2.20pre4/kernel'
make: *** [_dir_kernel] Error 2


So long,

Jochen.

-- 
The number of UNIX installations has grown to 10, with more expected.
                     - Dennis Ritchie and Ken Thompson, June 1972

Re: Linux 2.2.20-pre4

[Jochen Striepe]

        Hi again,

On 19 Jun 2001, Jochen Striepe <jochen@tolot.escape.de> wrote:
> 
> Now it stops with

OK, this resolved to nothing (my mistake). Now it works fine. Until it
reaches

ld -m elf_i386 -T /usr/src/linux/arch/i386/vmlinux.lds -e stext arch/i386/kernel/head.o arch/i386/kernel/init_task.o init/main.o init/version.o \
        --start-group \
        arch/i386/kernel/kernel.o arch/i386/mm/mm.o kernel/kernel.o mm/mm.o fs/fs.o ipc/ipc.o \
        fs/filesystems.a \
        net/network.a \
        drivers/block/block.a drivers/char/char.o drivers/misc/misc.a drivers/net/net.a drivers/scsi/scsi.a drivers/cdrom/cdrom.a drivers/pci/pci.a drivers/pnp/pnp.a drivers/video/video.a \
        /usr/src/linux/arch/i386/lib/lib.a /usr/src/linux/lib/lib.a /usr/src/linux/arch/i386/lib/lib.a \
        --end-group \
        -o vmlinux
drivers/scsi/scsi.a(aic7xxx.o): In function `aic7xxx_load_seeprom':
aic7xxx.o(.text+0x12a76): undefined reference to `memcpy'
make: *** [vmlinux] Error 1


HAND,

Jochen.

-- 
"Gosh that takes me back ... or forward.  That's the trouble with time
travel, you never can tell."
                -- Dr. Who

Re: Linux 2.2.20-pre4

[Tomasz Kłoczko]

On Tue, 19 Jun 2001, Alan Cox wrote:
[..]
> o	Fix refclock build with newer gcc		(Jari Ruusu)

Is it mean now kernel 2.2 with prepatch is (or will be) gcc 3.0 ready ?
If not what must be fixed/chenged to be ready ?

kloczek
-- 
-----------------------------------------------------------
*Ludzie nie mają problemów, tylko sobie sami je stwarzają*
-----------------------------------------------------------
Tomasz Kłoczko, sys adm @zie.pg.gda.pl|*e-mail: kloczek@rudy.mif.pg.gda.pl*

Re: Linux 2.2.20-pre4

[Alan Cox]

> On Tue, 19 Jun 2001, Alan Cox wrote:
> [..]
> > o	Fix refclock build with newer gcc		(Jari Ruusu)
> 
> Is it mean now kernel 2.2 with prepatch is (or will be) gcc 3.0 ready ?
> If not what must be fixed/chenged to be ready ?

It wont build with gcc 3.0 yet. To start with gcc 3.0 will assume it can
insert calls to 'memcpy' 

Re: Linux 2.2.20-pre4

[Jeff Garzik]

Alan Cox wrote:
> 
> > On Tue, 19 Jun 2001, Alan Cox wrote:
> > [..]
> > > o   Fix refclock build with newer gcc               (Jari Ruusu)
> >
> > Is it mean now kernel 2.2 with prepatch is (or will be) gcc 3.0 ready ?
> > If not what must be fixed/chenged to be ready ?
> 
> It wont build with gcc 3.0 yet. To start with gcc 3.0 will assume it can
> insert calls to 'memcpy'

IMHO omitting -fno-builtin when compiling the kernel was always a risky
proposition...  Since we provide our own copies of many of the builtins
[which are used in the kernel] anyway... why not always -fno-builtin,
and then call __builtin_foo when we really want the compiler's version..

gcc 3.0 without -fno-builtin is perfectly allowed to assume it can
insert calls to memcpy..

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

Re: Linux 2.2.20-pre4

[Alan Cox]

> IMHO omitting -fno-builtin when compiling the kernel was always a risky
> proposition...  Since we provide our own copies of many of the builtins
> [which are used in the kernel] anyway... why not always -fno-builtin,
> and then call __builtin_foo when we really want the compiler's version..

That may well be the right thing to do. Of course we rely on the compiler
providing some of them too

> gcc 3.0 without -fno-builtin is perfectly allowed to assume it can
> insert calls to memcpy..

I have no argument about its correctness there, but -fno-builtin will still
give a kernel that dosnt link due to abs() and other problems.. 8)

Re: Linux 2.2.20-pre4

[Philip Blundell]

[-- Attachment #1: Type: text/plain, Size: 849 bytes --]

>> It wont build with gcc 3.0 yet. To start with gcc 3.0 will assume it can
>> insert calls to 'memcpy'
>
>IMHO omitting -fno-builtin when compiling the kernel was always a risky
>proposition...  Since we provide our own copies of many of the builtins
>[which are used in the kernel] anyway... why not always -fno-builtin,
>and then call __builtin_foo when we really want the compiler's version..
>
>gcc 3.0 without -fno-builtin is perfectly allowed to assume it can
>insert calls to memcpy..

I don't think -fno-builtin has any bearing on whether gcc will emit calls to 
memcpy; instead it prevents gcc from open-coding them when it thinks it 
understands what's going on.

Try this with gcc -O2 -S, and again with -fno-builtin:

struct s { int a[200]; };

f(struct s *a, struct s *b)
{
  *b = *a;
}

g(int *a, int *b)
{
  memcpy(b, a, 4);
}

p.



[-- Attachment #2: Type: application/pgp-signature, Size: 237 bytes --]

Re: Linux 2.2.20-pre4

[Jeff Garzik]

Alan Cox wrote:
> > why not always -fno-builtin,
> > and then call __builtin_foo when we really want the compiler's version..

> That may well be the right thing to do. Of course we rely on the compiler
> providing some of them too

true, it wouldn't be a completely transparent switchover, but it seems
like the best way to produce expected results across a bunch of
different compilers.


> but -fno-builtin will still
> give a kernel that dosnt link due to abs() and other problems.. 8)

Any others come to mind?  abs is definitely special in that the compiler
[potentially] can do additional magic with the type information it has. 
Maybe -fno-builtin plus
	#undef abs
	#define abs __builtin_abs

Thanks,

	Jeff


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

Re: Linux 2.2.20-pre4

[Jeff Garzik]

Philip Blundell wrote:
> I don't think -fno-builtin has any bearing on whether gcc will emit calls to
> memcpy;

Good point.  The subject was about the compiler adding function calls to
code, and I started talking about the compiler removing them...

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

Re: Linux 2.2.20-pre4

[Eric Lammerts]

On Tue, 19 Jun 2001, Alan Cox wrote:
> > Is it mean now kernel 2.2 with prepatch is (or will be) gcc 3.0 ready ?
> > If not what must be fixed/chenged to be ready ?
>
> It wont build with gcc 3.0 yet. To start with gcc 3.0 will assume it can
> insert calls to 'memcpy'

I tried it, but didn't run into problems (apart from the volatile
xtime thing)

Linux version 2.2.18 (eric@andredvb) (gcc version 3.0 (Debian))
#1 Wed Jun 20 23:15:46 CEST 2001

(Tons of warnings, though)

Eric

Re: Linux 2.2.20-pre4

[Luigi Genoni]

Tried this too, but i have the feeling the kernel compiled with this gcc
3.0 is somehow slower. context switch is slower....
no benchs (no time to make them) to sustain my feeling, just a feeling...

On Wed, 20 Jun 2001, Eric Lammerts wrote:

>
> On Tue, 19 Jun 2001, Alan Cox wrote:
> > > Is it mean now kernel 2.2 with prepatch is (or will be) gcc 3.0 ready ?
> > > If not what must be fixed/chenged to be ready ?
> >
> > It wont build with gcc 3.0 yet. To start with gcc 3.0 will assume it can
> > insert calls to 'memcpy'
>
> I tried it, but didn't run into problems (apart from the volatile
> xtime thing)
>
> Linux version 2.2.18 (eric@andredvb) (gcc version 3.0 (Debian))
> #1 Wed Jun 20 23:15:46 CEST 2001
>
> (Tons of warnings, though)
>
> 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/
>