Linux 2.5.23-dj2

Lots of bits got thrown out this time, as Christoph Hellwig went through the patch and picked up on quite a few obviously wrong bits. In addition, this patch introduces the mad axemen, who come to carve up all that is monolithic. Patrick's MTRR split-up has been around for a while, and could use a bit more testing before it goes to Linus. The AGPGART changes I did this afternoon, and haven't seen much testing at all yet.

Finally, another round of compile fixes and the likes from Linux Kernel.

As usual,..

Patch against 2.5.23 vanilla is available from:
ftp://ftp.kernel.org/pub/linux/kernel/people/davej/patches/2.5/

Merged patch archive: http://www.codemonkey.org.uk/patches/merged/

Check http://www.codemonkey.org.uk/Linux-2.5.html before reporting known bugs that are also in mainline.

-- Davej.

2.5.23-dj2

o   Drop lots of obsolete/reversed changes.		(Christoph Hellwig)
o   Split up agpgart backends into per vendor files.	(Me)
o   Split up IA32 MTRR driver into per-CPU files.	(Patrick Mochel, Me)
o   Nuke 2 strtok() calls that crept back in.		(Me)
o   Add more missing tqueue includes.		(Andy Pfiffer, Matthew Harrell,
						 Stelian Pop, Adrian Bunk)
o   Mark some x86 SMP variables as initdata.		(Robert Love)
o   Offer P4 thermal monitoring when CONFIG_SMP=y	(Zwane Mwaikambo)
o   Add missing kmalloc check to iphase driver.		([email protected])
o   Poll/Select fast path optimisation take 2.		(Andi Kleen)
o   Oops fix in tcp_v6_get_port().			(Carl Ritson)
o   Various janitor work in megaraid driver.		(William Stinson)
o   Move software suspend to power management menu.	(Brad Hards)
o   Shrink stack usage of check_nmi_watchdog()		(Mikael Pettersson)
o   Nuke unneeded headers from mm/page_alloc.c		(William Lee Irwin)
o   Various janitor work on ixj telephony driver.	(Sam Ravnborg)
o   Workaround for lockd deadlock.			(Daniel Forrest)
o   Update reference to MIPS documentation.		(Rolf Eike Beer)
o   Convert SAA7110 driver to new i2c.			(Frank Davis)

2.5.23-dj1

o   Small UP optimisation in the scheduler.		(James Bottomley)
o   Update x86 cpufreq scaling code.			(Dominik Brodowski)
o   Export ioremap_nocache() for modules.		(Andi Kleen)
o   Export default_wake_function() for modules.		(Benjamin LaHaise)
o   Compaq hotplug compile fixes.			(Felipe Contreras)
o   Fix migration thread for non linear numbered CPUs.	(Ingo Molnar)
o   Framebuffer updates.				(James Simmons)
o   Introduce CONFIG_ISA option for i386.		(Andi Kleen)
o   Fix bad locking in driver/ core.			(Arnd Bergmann)

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

I was busy testing it with 2.5.23-dj1...
got a panic, but could not save the output (and did not liked the idea to write it all down 8), also I thought it had notinhg to do with the agpgart split and wanted to try to run 2.5.23-dj1 first before reporting... ah well will try it with -dj2

Rudmer

Chipset type and the output of "grep AGP .config" may be something to begin with. Did it crash on load at boot time? or during agp usage?

Dave.

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

got this error with make xconfig (which worked in -dj1):

make[1]: Entering directory `/usr/src/kernel/linux-2.5.23-dj2/scripts'
Generating kconfig.tk
-: 172: incorrect argument
chmod 755 kconfig.tk
make[1]: Leaving directory `/usr/src/kernel/linux-2.5.23-dj2/scripts' wish -f scripts/kconfig.tk
Error in startup script: invalid command name "clear_choices"
while executing
"clear_choices"
(procedure "read_config" line 3)
invoked from within
"read_config .config"
invoked from within

"if { [file readable .config] == 1} then {
        if { $argc > 0 } then {
                if { [lindex $argv 0] != "-D" } then {
                        read_config .config
                }
                else
                {
                        r..."
    (file "scripts/kconfig.tk" line 646)

make: *** [xconfig] Error 1

Rudmer

Another obviously wrong bit seems to be the patch below that is still in -dj2:

- it adds a function that isn't used in neither plain 2.5.23 nor by anything in the -dj2 patch
- it doesn't compile

cu
Adrian

@@ -59,6 +59,21 @@
 
 /* ------------------------------------------------------------- */
 
+void b1_set_revision(struct capi_driver *driver, char *rev)
+{
+	char *p;
+
+	if ((p = strchr(rev, ':')) != 0 && p[1]) {
+		strncpy(driver->revision, p + 2, sizeof(driver->revision));
+		driver->revision[sizeof(driver->revision)-1] = 0;
+		if ((p = strchr(driver->revision, '$')) != 0 && p > driver->revision)
+			*(p-1) = 0;
+	}
+	printk(KERN_INFO "%s: revision %s\n", driver->name, driver->revision);
+}
+
+/* ------------------------------------------------------------- */
+
 avmcard *b1_alloc_card(int nr_controllers)
 {
 	avmcard *card;

sorry, was a bit short 8-)

It crashed during boot, at the time X was starting (as reported earlier by ??) booting without starting X worked, and the following oops occured while staring X (using the bootscript)

Kernel BUG at exit.c: 562

invalid operand: 0000
<snip> (did not handcopied this part)
<0>Kernel Panic: attempted to kill the idle task!
in idle task - not syncing

gandalf:~ # lspci
00:00.0 Host bridge: Silicon Integrated Systems [SiS]: Unknown device 0735 (rev
01)
00:01.0 PCI bridge: Silicon Integrated Systems [SiS] 5591/5592 AGP 00:02.0 ISA bridge: Silicon Integrated Systems [SiS] 85C503/5513 00:02.5 IDE interface: Silicon Integrated Systems [SiS] 5513 [IDE] (rev d0) 00:0b.0 Ethernet controller: Winbond Electronics Corp W89C940 00:0f.0 Multimedia audio controller: Ensoniq 5880 AudioPCI (rev 02) 00:11.0 Ethernet controller: Winbond Electronics Corp W89C940 01:00.0 VGA compatible controller: Matrox Graphics, Inc. MGA G200 AGP (rev 01)

gandalf:/usr/src/kernel/linux-2.5.23-dj2 # grep AGP .config CONFIG_AGP=y
# CONFIG_AGP_INTEL is not set
# CONFIG_AGP_I810 is not set
# CONFIG_AGP_VIA is not set
# CONFIG_AGP_AMD is not set
CONFIG_AGP_SIS=y
# CONFIG_AGP_ALI is not set
# CONFIG_AGP_SWORKS is not set

Rudmer

PS. will try -dj2 in a moment (just finished compiling)

Bad voodoo in arch/i386/config.in
Change the == on line 172 to a single =

Dave.

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

Ok I can run -dj2, but I cannot use X 8-( although this time no BUG or panic.

I got these errors during boot:

Jun 19 23:22:10 gandalf kdm[269]: IO Error in XOpenDisplay
Jun 19 23:22:10 gandalf kdm[259]: Server for display :0 terminated 
unexpectedly Jun 19 23:22:10 gandalf kdm[259]: Display :0 cannot be opened
Jun 19 23:22:13 gandalf kdm[284]: IO Error in XOpenDisplay
Jun 19 23:22:13 gandalf kdm[259]: Server for display :0 terminated 
unexpectedly Jun 19 23:22:13 gandalf kdm[259]: Display :0 cannot be opened
Jun 19 23:22:17 gandalf kdm[291]: IO Error in XOpenDisplay
Jun 19 23:22:17 gandalf kdm[259]: Server for display :0 terminated 
unexpectedly Jun 19 23:22:17 gandalf kdm[259]: Display :0 cannot be opened
Jun 19 23:22:20 gandalf kdm[298]: IO Error in XOpenDisplay
Jun 19 23:22:20 gandalf kdm[259]: Server for display :0 terminated 
unexpectedly Jun 19 23:22:20 gandalf kdm[259]: Display :0 cannot be opened
Jun 19 23:22:20 gandalf kdm[259]: Display :0 is being disabled (restarting 

too fast)

and whem starting X with startx:
<X startup messages>
XIO: Fatal IO error 104 (connection reset by peer) on X server ":0.0"
after 0 requests (0 known processed) with 0 events remaining.

when X is starting I see the normal 'flash' as the screen resolution is adjusted but the screen remains black and then the console returns.

same box (SIS and MGA)

Rudmer

1, any agpgart related messages in the logs/dmesg ?
2. Can you disable agpgart, and try again. I'm fairly certain this is the cause, but just in case..

Dave

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

output from dmesg (didn't cut too much I hope):
Linux version 2.5.23-dj2 (rudmer@gandalf) (gcc version 2.95.3 20010315 (release)) #1 Wed Jun 19 23:16:06 CEST 2002
Video mode to be used for restore is ffff
<snip>
255MB LOWMEM available.
<snip>
Kernel command line: BOOT_IMAGE=lin2.5 ro root=302 hdc=ide-scsi 2 <snip>
PCI: PCI BIOS revision 2.10 entry at 0xfdb01, last bus=1
PCI: Using configuration type 1
PCI: Probing PCI hardware
PCI: Probing PCI hardware (bus 00)
Unknown bridge resource 0: assuming transparent
PCI: Using IRQ router SIS [1039/0008] at 00:02.0
__iounmap: bad address d0802030
apm: BIOS version 1.2 Flags 0x03 (Driver version 1.16) <snip>
Linux agpgart interface v0.99 (c) Jeff Hartmann
agpgart: Maximum main memory to use for agp memory: 203M
agpgart: Detected SiS 735 chipset
agpgart: AGP aperture is 32M @ 0xd0000000
[drm] AGP 0.99 on SiS @ 0xd0000000 32MB
[drm] Initialized mga 3.0.2 20010321 on minor 0
block: 256 slots per queue, batch=32
RAMDISK driver initialized: 16 RAM disks of 512K size 1024 blocksize

I just saw the iounmap error, maybe related??

Did not try this yet (will do it tomorrow evening), but the X failure was related to my mouse setup: /dev/mouse pointed to /dev/psaux which is not available in 2.5.xx-djX. when I corrected this to /dev/mouse0 I got a working X again 8-)

however, when I started X from the bootscript, that is the bootscript starts kdm which in turn starts the X server, I got the same oops as before... the process that causes the oops appears to be chmod, if you want the whole oops, please tell and I will write it down (cannot use a serial console...).

so the agpgart split seems to work fine here, but there is clearly something wrong when kde2 tries to start.

Rudmer

No, that happens earlier. No idea what causes it, but it's obviously a problem somewhere..

Please do. And feed it through ksymoops please.

Finger of suspicion points to..
http://www.codemonkey.org.uk/patches/merged/2.5.23/dj2/poll-select-fast-path.diff

Apply this (with -R), and see if it goes away.

Dave

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

just checked 2 but no improvement, also checked without drm again no solution...

otherwise the system seems stable (also when running X) but I also saw the 'spurious 8259A interrupt: IRQ7.' message after a couple of minutes. I know that this has got something to do with local apic, but nothing more...

this is from dmesg:
Local APIC disabled by BIOS -- reenabling. Found and enabled local APIC!
Initializing CPU#0
<snip>
Using local APIC timer interrupts. calibrating APIC timer ...
..... CPU clock speed is 1128.5606 MHz. ..... host bus clock speed is 265.5435 MHz.
cpu: 0, clocks: 2655435, slice: 1327717 CPU0<T0:2655424,T1:1327696,D:11,S:1327717,C:2655435>

(interresting: this differs from 2.4.19-pre10-ac2... but there I get the same spurious interrupt)

UP system, athlon 1133, mobo: ecs k7s5am

gandalf:~ # lspci
00:00.0 Host bridge: Silicon Integrated Systems [SiS]: Unknown device 0735 (rev
01)
00:01.0 PCI bridge: Silicon Integrated Systems [SiS] 5591/5592 AGP 00:02.0 ISA bridge: Silicon Integrated Systems [SiS] 85C503/5513 00:02.5 IDE interface: Silicon Integrated Systems [SiS] 5513 [IDE] (rev d0) 00:0b.0 Ethernet controller: Winbond Electronics Corp W89C940 00:0f.0 Multimedia audio controller: Ensoniq 5880 AudioPCI (rev 02) 00:11.0 Ethernet controller: Winbond Electronics Corp W89C940 01:00.0 VGA compatible controller: Matrox Graphics, Inc. MGA G200 AGP (rev 01)

Rudmer

Well, that's sort of good in a way.. it means the agpgart changes aren't to blame. 8-)

As for your crash in exit.c, I'm puzzled by that one. Are you using preempt ? if so, does disabling that fix it ?

Dave

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