2002-03-09 01:47:00

by Alan

[permalink] [raw]
Subject: Linux 2.2.21pre4

Please save any 2.2 updates for 2.2.22pre from now on unless they are
clear bug fixes. I hope to do a 2.2.21rc1 next

2.2.21pre4
o Fix FAT breakage in pre3 (Dmitry Levin)
o Add S/390 LCS driver (IBM opensourced it now) (DJ Barrow,
Frank Pavlic)
o Update COPYING file to match FSF update (Dan Quinlann)
| basically swap 19xx example for this century..
o Fix a file name comment (William Stearns)
o Add realtek phy support to 2.2 sis900 driver (Allan Jacobsen)
o Fix MCE address reporting order, fix oops with (Dave Jones)
newer gcc due to bad asm constraints
o Starfire update (Ion Badulescu)
o Always victimise the dcache a little when (John Lash, me)
short of memory

2.2.21pre3
o Fix a case where a non blocking tty write could (Peter Benie)
get stuck
o Fix non blocking midi close on es1370, es1371 (me)
sonicvibes right this time
o Fix menu/xconfig warnings (Ren? Scharfe)
o Fix non blocking midi close on cmpci, cs4281, (me)
esssolo, trident.
o Add eepro100VE ident (Hanno Boeck)
o Fix DRM oops case (Herbert Xu)
o Fix an oops causing datagram AF_UNIX race (Paul Menage)
o Support newer geodes using new CPUID properly (Hiroshi Miura)
o Fix up RTC build for non pmac ppc boxes (Tom Rini)
o Fix MCE address reporting (Pete Wyckoff)
o Vibra16 docs update (Neale Banks)
o Eicon include file fix (Herbert Xu)
o ISDN loop and header fixes (Kai Germaschewski)
o Fix eepro100 out of memory during init path (Neale Banks)
o Fix BSD partition table handling breakage (Andries Brouwer)
o Add WD XD signature to xd driver (Paul)
o 3Ware driver update (Adam Radford)
o S/390 debugging updates (Carsten Otte)
o S/390 DASD updates (Carsten Otte)
o S/390 CIO updates (Carsten Otte)
o Update USB serial, belkin, digi_acceleport, (Greg Kroah-Hartmann)
empeg, ftdsio, edgeport, keyspan, mctu232,
omninet, prolific, visor
o Cyberjack USB driver (Matthias Bruestle)
o USB ir dongle driver (Greg Kroah-Hartmann)
o Support very large FAT file systems (Vijay Kumar)
o Backport 2.4 modversions build fix (Mikael Pettersson)
o Backport 2.4 es1371 init for new revs (Julian Anastasov)
o 3c507 driver fixes (Mark Mackenzie)
o ext2 obscure group descriptor corruption fix (Daniel Phillips,
Al Viro)
o Correct a problem where rpciod didnt give up (Andreas Haumer)
its current dir

2.2.21pre2
o Fix non blocking midi close on es1370, es1371 (me)
sonicvibes
o Update osst driver (Willem Riede)
o Update machine check support in 2.2 to match 2.4(Dave Jones)
o Additional P4, Rise, Winchip handling for setup (Dave Jones)
o Fix extended MMX initialisation on Cyrix MII (me)
o Backport a lot of x86 setup (cache size etc) (Dave Jones)
o ISDN cleanups (Kai Germaschewski)
o Backport eicon driver fixes (Kai Germaschewski)
o ISDN ppp fixes (Andre Beck)
o Fix timeout handling in eicon driver (Kai Germaschewski)
o Fix null pointer bug in isdnloop (Kai Germaschewski)
o Menuconfig refresh fixup (Willy Tarreau)
o Modular ati frame buffer build fix (Krzysztof Taraszka)
o Backport VIA chipset fixes to 2.2 (me)
o Make DCD high->low work on SX16 with CLOCAL set (Ado Arnolds)

2.2.21pre1
o Fix potential corruption with vmalloc on (Ralf Baechle)
virtually cached boxes
o Small PPC build fixups (Tom Rini)
o zImage booting fix (Kalev Soikonen)
o EIO on NFS read fixup (Trond Myklebust)
o Update 3ware raid driver (Adam Radford)
o page_alloc race fix (Andrea Arcangeli)
o Update USB maintainers (Greg Kroah-Hartmann)
o bttv clipcount=0 fix (Solar Designer)
o Fix multiple eepro driver bugs (Aris)
o Sym53c8xx queue handling fix (Gerard Roudier)
o Update SubmittingDrivers document (Michal Svec)
o 8139too performance tune (Jens David)
o procfs follow link return fix (Solar Designer)
o Backport SEM_UNDO overflow fix from 2.4 (Leonid Igolnik)
o VM86 fixes (Manfred Spraul)
o Fix alpha build (Kim Heino)

--
Bistromathematics also has exception and out of range errors. Fortunately
NaN normally only occurs in an Indian resturant


2002-03-09 13:31:43

by Barry K. Nathan

[permalink] [raw]
Subject: [BUG][PATCH] boot failure Re: Linux 2.2.21pre4

> Please save any 2.2 updates for 2.2.22pre from now on unless they are
> clear bug fixes. I hope to do a 2.2.21rc1 next
>
> 2.2.21pre4
[snip]
> o Fix MCE address reporting order, fix oops with (Dave Jones)
> newer gcc due to bad asm constraints

Hmmm... I have a computer here that freezes at this point in the boot
process with 2.2.21-pre4:

...
CPU: L2 cache: 128K
Intel machine check architecture supported.

Reversing the following change to bluesmoke.c lets this computer boot
again like it does with 2.2.21-pre3. The computer in question has a 533MHz
Intel Celeron (the Pentium II-based kind).

-Barry K. Nathan <[email protected]>

diff -ruN linux-2.2.21-pre3/arch/i386/kernel/bluesmoke.c linux-2.2.21-pre4/arch/i386/kernel/bluesmoke.c
--- linux-2.2.21-pre3/arch/i386/kernel/bluesmoke.c Sun Mar 3 23:20:11 2002
+++ linux-2.2.21-pre4/arch/i386/kernel/bluesmoke.c Sat Mar 9 03:58:57 2002
@@ -165,7 +164,7 @@
if(l&(1<<8))
wrmsr(0x17b, 0xffffffff, 0xffffffff);
banks = l&0xff;
- for(i=1;i<banks;i++)
+ for(i=0;i<banks;i++)
{
wrmsr(0x400+4*i, 0xffffffff, 0xffffffff);
}

2002-03-09 14:59:30

by Mikael Pettersson

[permalink] [raw]
Subject: Re: [BUG][PATCH] boot failure Re: Linux 2.2.21pre4

Barry K. Nathan wrote:
>Reversing the following change to bluesmoke.c lets this computer boot
>again like it does with 2.2.21-pre3. The computer in question has a 533MHz
>Intel Celeron (the Pentium II-based kind).
>
>-Barry K. Nathan <[email protected]>
>
>diff -ruN linux-2.2.21-pre3/arch/i386/kernel/bluesmoke.c linux-2.2.21-pre4/arch/i386/kernel/bluesmoke.c
>--- linux-2.2.21-pre3/arch/i386/kernel/bluesmoke.c Sun Mar 3 23:20:11 2002
>+++ linux-2.2.21-pre4/arch/i386/kernel/bluesmoke.c Sat Mar 9 03:58:57 2002
>@@ -165,7 +164,7 @@
> if(l&(1<<8))
> wrmsr(0x17b, 0xffffffff, 0xffffffff);
> banks = l&0xff;
>- for(i=1;i<banks;i++)
>+ for(i=0;i<banks;i++)
> {
> wrmsr(0x400+4*i, 0xffffffff, 0xffffffff);
> }

Same here. s/i=0/i=1/ i the for() and my PII boots again.

/Mikael

2002-03-10 10:12:02

by Barry K. Nathan

[permalink] [raw]
Subject: [PATCH] 2.2.21pre[234] misreports Pentium II model names

In 2.2.21-pre2 through -pre4, many Intel CPUs are misreported in
/proc/cpuinfo and dmesg as Mobile Pentium II's, even when they are
Celerons in desktops or whatever.

In init_intel, the correct CPU name is being assigned to p, but p isn't
being copied into c->x86_model_id. As a result, the name from the tables,
"Mobile Pentium II", is always being used.

This patch (backported from 2.4) fixes the problem. Alan, please apply.

-Barry K. Nathan <[email protected]>

--- linux-2.2.21-pre4/arch/i386/kernel/setup.c Sat Mar 9 03:58:57 2002
+++ linux-2.2.21-pre4-bknA-2/arch/i386/kernel/setup.c Sat Mar 9 22:31:34 2002
@@ -1232,6 +1232,9 @@
break;
}
}
+
+ if ( p )
+ strcpy(c->x86_model_id, p);
}

struct cpu_model_info {