kernel update (relative to 2.6.0-test4)

The latest ia64 patch is now available at ftp://www.kernel.org/pub/linux/kernel/ports/ia64/v2.5/ in file:

linux-2.6.0-test4-ia64-030826.diff.gz

Mainly a collection of bug-fixes plus sync'ing with 2.6.0-test4 (e.g., fadvise64_64() is now hooked up). Also, I included a patch by Arun which makes x86 GDT/LDT accesses work again. The patch was supposed to be in Linus' 2.6.0-test4 but got overlooked somehow.

The kernel seems to work fine on various zx1-based machines and the HP Simulator. I didn't try Lion/Big Sur but the ACPI bug that was keeping earlier kernels from working has been folded into the main tree, so in theory, the kernel should work on those platforms again.

Oh, -test3 had some nasty race conditions in the timer code and that caused it to crash on MP machines under heavy load. -test4 seems to be much better in this regard.

Enjoy,

	--david

David,

The second chunk of the "UP cmc/cpe polling fix" seems to have
gotten lost. Please apply the attached for the cpe side of the fix. Thanks,

Alex

Alex Williamson                             HP Linux & Open Source Lab

@@ -1225,7 +1225,7 @@
 
 	ia64_mca_cpe_int_handler(cpe_irq, arg, ptregs);
 
-	for (++cpuid ; !cpu_online(cpuid) && cpuid < NR_CPUS ; cpuid++);
+	for (++cpuid ; cpuid < NR_CPUS && !cpu_online(cpuid) ; cpuid++);
 
 	if (cpuid < NR_CPUS) {
 		platform_send_ipi(cpuid, IA64_CPEP_VECTOR, IA64_IPI_DM_INT, 0);

Last few lines of the build output (building for Tiger):

  KSYM    .tmp_kallsyms2.S
  AS      .tmp_kallsyms2.o
  LD      vmlinux

OBJCOPY arch/ia64/hp/sim/boot/vmlinux.bin
BFD: arch/ia64/hp/sim/boot/vmlinux.bin: warning: allocated section `.data.nosave' not in segment
GZIP arch/ia64/hp/sim/boot/vmlinux.gz
LN vmlinux.gz
Kernel: vmlinux.gz is ready

Is that .../hp/sim/boot/... path right? I don't have CONFIG_IA64_HP_SIM set in my .config.

Any clues on what the warning from OBJCOPY really means?

-Tony

Tony> Last few lines of the build output (building for Tiger):

  Tony> KSYM    .tmp_kallsyms2.S
  Tony> AS      .tmp_kallsyms2.o
  Tony> LD      vmlinux

Tony> OBJCOPY arch/ia64/hp/sim/boot/vmlinux.bin
Tony> BFD: arch/ia64/hp/sim/boot/vmlinux.bin: warning: allocated section `.data.nosave' not in segment
Tony> GZIP arch/ia64/hp/sim/boot/vmlinux.gz
Tony> LN vmlinux.gz
Tony> Kernel: vmlinux.gz is ready

Tony> Is that .../hp/sim/boot/... path right?

Not really, but it doesn't hurt, AFAIK (I'll try to remember to clean it up, though).

Tony> Any clues on what the warning from OBJCOPY really means?

Nope. I don't get that (building for zx1):

  KSYM    .tmp_kallsyms2.S
  AS      .tmp_kallsyms2.o
  LD      vmlinux

Building modules, stage 2.
OBJCOPY arch/ia64/hp/sim/boot/vmlinux.bin
GZIP arch/ia64/hp/sim/boot/vmlinux.gz
MODPOST
LN vmlinux.gz
Kernel: vmlinux.gz is ready

	--david

Ok, I found it. I somehow ended up with CONFIG_PM=y, so "kernel/power" got built, and inside kernel/power/main.c there is:

  static int in_suspend __nosavedata = 0;

where __nosavedata is defined in include/linux/init.h as:

#define __nosavedata __attribute__ ((__section__ (".data.nosave")))

We don't have an entry in vmlinux.lds.S to handle this section.

-Tony

CONFIG_PM is set without a menu entry to turn it on/off.

@@ -264,7 +264,7 @@
 	  unsure, answer Y.
 
 config PM
-	bool
+	bool "Power Management support"
 	depends on IA64_GENERIC || IA64_DIG || IA64_HP_ZX1
 	default y
 	---help---

Yes, that's right. David and I had a little discussion about that a couple of days ago. It's a harmless change.

"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk