Skip to contentSkip to conversation

IA64 kexec/kdump 2.6.18-rc5 patch

PATCH Zou Nan hai 5 participants 12 messages Aug 29, 2006 – Sep 12, 2006

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

Hi,
Below is the IA64 kexec/kdump patch against 2.6.18-rc5.

Fixes and enhancements in this patch include:

1. Fix I/D cache coherence problem.
Kdump sometimes hit an I/D cache coherence issue on platform with separate I/D cache. Although there is fc.i instruction in relocate_kernel.S.
purgatory code and the second kernel code is copy into reserved region at kexec_load when running "kexec -p".
There need and an icache_flush when segments is copied into its destination.
2. Change elf_prstatus to an per_cpu value to save stack size at crash path according to Bob Montgomery's suggestion.
3. put AP to a loop of hint.pause instead of call pal_halt_light.

arch/ia64/hp/common/sba_iommu.c +22 -0
diff -Nraup linux-2.6.18-rc5/arch/ia64/hp/common/sba_iommu.c linux-2.6.18-rc5-kdump/arch/ia64/hp/common/sba_iommu.c
@@ -1623,6 +1623,28 @@ ioc_iova_init(struct ioc *ioc)
 	READ_REG(ioc->ioc_hpa + IOC_IBASE);
 }
 
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{
+	struct ioc *ioc;
+
+	ioc = ioc_list;
+
+	while (ioc != NULL) {
+		/* Disable IOVA translation */
+		WRITE_REG(ioc->ibase & 0xfffffffffffffffe, ioc->ioc_hpa + IOC_IBASE);
+		READ_REG(ioc->ioc_hpa + IOC_IBASE);
+
+		/* Clear I/O TLB of any possible entries */
+		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) + iovp_shift), ioc->ioc_hpa + IOC_PCOM);
+		READ_REG(ioc->ioc_hpa + IOC_PCOM);
+
+		ioc = ioc->next;
+	}
+}
+#endif
+
 static void __init
 ioc_resource_init(struct ioc *ioc)
 {
arch/ia64/Kconfig +23 -0
diff -Nraup linux-2.6.18-rc5/arch/ia64/Kconfig linux-2.6.18-rc5-kdump/arch/ia64/Kconfig
@@ -427,6 +427,29 @@ config SGI_SN
 
 source "drivers/sn/Kconfig"
 
+config KEXEC
+	bool "kexec system call (EXPERIMENTAL)"
+	depends on EXPERIMENTAL
+	help
+	  kexec is a system call that implements the ability to shutdown your
+	  current kernel, and to start another kernel.  It is like a reboot
+	  but it is indepedent of the system firmware.   And like a reboot
+	  you can start any kernel with it, not just Linux.
+
+	  The name comes from the similiarity to the exec system call.
+
+	  It is an ongoing process to be certain the hardware in a machine
+	  is properly shutdown, so do not be surprised if this code does not
+	  initially work for you.  It may help to enable device hotplugging
+	  support.  As of this writing the exact hardware interface is
+	  strongly in flux, so no good recommendation can be made.
+
+config CRASH_DUMP
+	  bool "kernel crash dumps (EXPERIMENTAL)"
+	  depends on EXPERIMENTAL
+	  help
+	    Generate crash dump after being started by kexec.
+
 source "drivers/firmware/Kconfig"
 
 source "fs/Kconfig.binfmt"
arch/ia64/kernel/crash.c +123 -0

123 lines not on this page · raw message

arch/ia64/kernel/efi.c +16 -3
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/efi.c linux-2.6.18-rc5-kdump/arch/ia64/kernel/efi.c
@@ -26,6 +26,7 @@
 #include <linux/types.h>
 #include <linux/time.h>
 #include <linux/efi.h>
+#include <linux/kexec.h>
 
 #include <asm/io.h>
 #include <asm/kregs.h>
@@ -41,7 +42,7 @@ extern efi_status_t efi_call_phys (void 
 struct efi efi;
 EXPORT_SYMBOL(efi);
 static efi_runtime_services_t *runtime;
-static unsigned long mem_limit = ~0UL, max_addr = ~0UL;
+static unsigned long mem_limit = ~0UL, max_addr = ~0UL, min_addr = 0UL;
 
 #define efi_call_virt(f, args...)	(*(f))(args)
 
@@ -421,6 +422,8 @@ efi_init (void)
 			mem_limit = memparse(cp + 4, &cp);
 		} else if (memcmp(cp, "max_addr=", 9) = 0) {
 			max_addr = GRANULEROUNDDOWN(memparse(cp + 9, &cp));
+		} else if (memcmp(cp, "min_addr=", 9) = 0) {
+			min_addr = GRANULEROUNDDOWN(memparse(cp + 9, &cp));
 		} else {
 			while (*cp != ' ' && *cp)
 				++cp;
@@ -428,6 +431,8 @@ efi_init (void)
 				++cp;
 		}
 	}
+	if (min_addr != 0UL)
+		printk(KERN_INFO "Ignoring memory below %luMB\n", min_addr >> 20);
 	if (max_addr != ~0UL)
 		printk(KERN_INFO "Ignoring memory above %luMB\n", max_addr >> 20);
 
@@ -894,7 +899,8 @@ find_memmap_space (void)
 		as = max(contig_low, md->phys_addr);
 		ae = min(contig_high, efi_md_end(md));
 
-		/* keep within max_addr= command line arg */
+		/* keep within max_addr= and min_addr= command line arg */
+		as = max(as, min_addr);
 		ae = min(ae, max_addr);
 		if (ae <= as)
 			continue;
@@ -1004,7 +1010,8 @@ efi_memmap_init(unsigned long *s, unsign
 		} else
 			ae = efi_md_end(md);
 
-		/* keep within max_addr= command line arg */
+		/* keep within max_addr= and min_addr= command line arg */
+		as = max(as, min_addr);
 		ae = min(ae, max_addr);
 		if (ae <= as)
 			continue;
@@ -1116,6 +1123,12 @@ efi_initialize_iomem_resources(struct re
 			 */
 			insert_resource(res, code_resource);
 			insert_resource(res, data_resource);
+#ifdef CONFIG_KEXEC
+                        insert_resource(res, &efi_memmap_res);
+                        insert_resource(res, &boot_param_res);
+			if (crashk_res.end > crashk_res.start)
+				insert_resource(res, &crashk_res);
+#endif
 		}
 	}
 }
arch/ia64/kernel/entry.S +1 -1
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/entry.S linux-2.6.18-rc5-kdump/arch/ia64/kernel/entry.S
@@ -1575,7 +1575,7 @@ sys_call_table:
 	data8 sys_mq_timedreceive		// 1265
 	data8 sys_mq_notify
 	data8 sys_mq_getsetattr
-	data8 sys_ni_syscall			// reserved for kexec_load
+	data8 sys_kexec_load
 	data8 sys_ni_syscall			// reserved for vserver
 	data8 sys_waitid			// 1270
 	data8 sys_add_key
arch/ia64/kernel/iosapic.c +19 -0
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/iosapic.c linux-2.6.18-rc5-kdump/arch/ia64/kernel/iosapic.c
@@ -288,6 +288,25 @@ nop (unsigned int irq)
 	/* do nothing... */
 }
 
+#ifdef CONFIG_CRASH_DUMP
+void
+kdump_disable_iosapic(void)
+{
+	u32 low32;
+	struct iosapic_intr_info *info;
+	struct iosapic_rte_info *rte;
+	for (info = iosapic_intr_info; info <
+			iosapic_intr_info + IA64_NUM_VECTORS; ++info) {
+		low32 = info->low32 |= IOSAPIC_MASK;
+		list_for_each_entry(rte, &info->rtes,
+				rte_list) {
+			iosapic_write(rte->addr,
+					IOSAPIC_RTE_LOW(rte->rte_index), low32);
+		}
+	}
+}
+#endif
+
 static void
 mask_irq (unsigned int irq)
 {
arch/ia64/kernel/machine_kexec.c +139 -0

139 lines not on this page · raw message

arch/ia64/kernel/Makefile +1 -0
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/Makefile linux-2.6.18-rc5-kdump/arch/ia64/kernel/Makefile
@@ -28,6 +28,7 @@ obj-$(CONFIG_IA64_CYCLONE)	+= cyclone.o
 obj-$(CONFIG_CPU_FREQ)		+= cpufreq/
 obj-$(CONFIG_IA64_MCA_RECOVERY)	+= mca_recovery.o
 obj-$(CONFIG_KPROBES)		+= kprobes.o jprobes.o
+obj-$(CONFIG_KEXEC)		+= machine_kexec.o relocate_kernel.o crash.o
 obj-$(CONFIG_IA64_UNCACHED_ALLOCATOR)	+= uncached.o
 obj-$(CONFIG_AUDIT)		+= audit.o
 mca_recovery-y			+= mca_drv.o mca_drv_asm.o
arch/ia64/kernel/relocate_kernel.S +490 -0

490 lines not on this page · raw message

arch/ia64/kernel/setup.c +42 -0

60 lines not on this page · raw message

arch/ia64/kernel/smp.c +55 -1

92 lines not on this page · raw message

include/asm-ia64/kexec.h +45 -0

45 lines not on this page · raw message

include/asm-ia64/machvec_hpzx1.h +2 -0
diff -Nraup linux-2.6.18-rc5/include/asm-ia64/machvec_hpzx1.h linux-2.6.18-rc5-kdump/include/asm-ia64/machvec_hpzx1.h
@@ -34,4 +34,6 @@ extern ia64_mv_dma_mapping_error	sba_dma
 #define platform_dma_supported			sba_dma_supported
 #define platform_dma_mapping_error		sba_dma_mapping_error
 
+extern void ioc_iova_disable(void);
+
 #endif /* _ASM_IA64_MACHVEC_HPZX1_h */
include/asm-ia64/meminit.h +2 -1
diff -Nraup linux-2.6.18-rc5/include/asm-ia64/meminit.h linux-2.6.18-rc5-kdump/include/asm-ia64/meminit.h
@@ -15,11 +15,12 @@
  * 	- initrd (optional)
  * 	- command line string
  * 	- kernel code & data
+ * 	- crash dumping code reserved region
  * 	- Kernel memory map built from EFI memory map
  *
  * More could be added if necessary
  */
-#define IA64_MAX_RSVD_REGIONS 6
+#define IA64_MAX_RSVD_REGIONS 7
 
 struct rsvd_region {
 	unsigned long start;	/* virtual address of beginning of element */
include/asm-ia64/smp.h +3 -0
diff -Nraup linux-2.6.18-rc5/include/asm-ia64/smp.h linux-2.6.18-rc5-kdump/include/asm-ia64/smp.h
@@ -128,6 +128,9 @@ extern void smp_send_reschedule (int cpu
 extern void lock_ipi_calllock(void);
 extern void unlock_ipi_calllock(void);
 extern void identify_siblings (struct cpuinfo_ia64 *);
+#ifdef CONFIG_KEXEC
+extern void kexec_stop_this_cpu(void *);
+#endif
 
 #else
 
include/linux/kexec.h +5 -0
diff -Nraup linux-2.6.18-rc5/include/linux/kexec.h linux-2.6.18-rc5-kdump/include/linux/kexec.h
@@ -108,6 +108,10 @@ int kexec_should_crash(struct task_struc
 extern struct kimage *kexec_image;
 extern struct kimage *kexec_crash_image;
 
+#ifndef kexec_flush_icache_page
+#define kexec_flush_icache_page(page)
+#endif
+
 #define KEXEC_ON_CRASH  0x00000001
 #define KEXEC_ARCH_MASK 0xffff0000
 
@@ -131,6 +135,7 @@ extern struct resource crashk_res;
 typedef u32 note_buf_t[MAX_NOTE_BYTES/4];
 extern note_buf_t *crash_notes;
 
+
 #else /* !CONFIG_KEXEC */
 struct pt_regs;
 struct task_struct;
kernel/irq/manage.c +0 -1
diff -Nraup linux-2.6.18-rc5/kernel/irq/manage.c linux-2.6.18-rc5-kdump/kernel/irq/manage.c
@@ -475,4 +475,3 @@ int request_irq(unsigned int irq,
 	return retval;
 }
 EXPORT_SYMBOL(request_irq);
-
kernel/kexec.c +1 -0
diff -Nraup linux-2.6.18-rc5/kernel/kexec.c linux-2.6.18-rc5-kdump/kernel/kexec.c
@@ -851,6 +851,7 @@ static int kimage_load_crash_segment(str
 			memset(ptr + uchunk, 0, mchunk - uchunk);
 		}
 		result = copy_from_user(ptr, buf, uchunk);
+		kexec_flush_icache_page(page);
 		kunmap(page);
 		if (result) {
 			result = (result < 0) ? result : -EIO;
#1 lore raw

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1629

Quoted 4 lines · Zou Nan hai
arch/ia64/hp/common/sba_iommu.c +4 -0
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{

Ugh. If you really need this functionality (which I have to say looks like a band-aid), it probably should be a platform vector. And should be split into a separate patch.

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1641

Quoted 12 lines
arch/ia64/hp/common/sba_iommu.c +12 -0
+	struct ioc *ioc;
+
+	ioc = ioc_list;
+
+	while (ioc != NULL) {
+		/* Disable IOVA translation */
+		WRITE_REG(ioc->ibase & 0xfffffffffffffffe, ioc->ioc_hpa + IOC_IBASE);
+		READ_REG(ioc->ioc_hpa + IOC_IBASE);
+
+		/* Clear I/O TLB of any possible entries */
+		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) + iovp_shift), ioc->ioc_hpa + IOC_PCOM);
+		READ_REG(ioc->ioc_hpa + IOC_PCOM);

This will just make any future device DMA attempts fail with an MCA, won't it? What problem does that solve? Don't you need the same for other IOMMUs like SGI's?

arch/ia64/Kconfig config SGI_SN · line 436

Quoted 7 lines
arch/ia64/Kconfig +7 -0
+config KEXEC
+	bool "kexec system call (EXPERIMENTAL)"
+	depends on EXPERIMENTAL
+	help
+	  kexec is a system call that implements the ability to shutdown your
+	  current kernel, and to start another kernel.  It is like a reboot
+	  but it is indepedent of the system firmware.   And like a reboot

independent

arch/ia64/Kconfig config SGI_SN · line 439

Quoted 3 lines
arch/ia64/Kconfig +3 -0
+	  you can start any kernel with it, not just Linux.
+
+	  The name comes from the similiarity to the exec system call.

similarity

arch/ia64/kernel/crash.c line 27

Quoted 2 lines
arch/ia64/kernel/crash.c +2 -0
+size_t copy_oldmem_page(unsigned long pfn, char *buf,
+                               size_t csize, unsigned long offset, int userbuf)

Doesn't seem to be used.

arch/ia64/kernel/crash.c line 48

Quoted 6 lines
arch/ia64/kernel/crash.c +6 -0
+static void device_shootdown(void)
+{
+	kdump_disable_iosapic();
+#ifdef CONFIG_IA64_HP_ZX1
+	ioc_iova_disable();
+#endif

Seems like sort of a heavy-handed way to shut down devices. But maybe you don't have any alternatives, I don't know. I guess you don't do the pci_disable_device() thing here just to avoid depending on more code?

arch/ia64/kernel/crash.c line 54

Quoted 6 lines
arch/ia64/kernel/crash.c +6 -0
+}
+
+static inline Elf64_Word
+*append_elf_note(Elf64_Word *buf, char *name, unsigned type, void *data,
+		size_t data_len)
+{

All this ELF stuff looks like something that could be split into a separate patch.

arch/ia64/kernel/crash.c line 97

Quoted 8 lines
arch/ia64/kernel/crash.c +8 -0
+	ia64_dump_cpu_regs(dst);
+        cfm = dst[43];
+        sol = (cfm >> 7) & 0x7f;
+        sof = cfm & 0x7f;
+        dst[46] = (unsigned long)ia64_rse_skip_regs((unsigned long *)dst[46],
+                        sof - sol);
+
+        buf = (u64 *) per_cpu_ptr(crash_notes, cpu);

Funny indentation above (spaces rather than tab, I guess).

arch/ia64/kernel/efi.c efi_init (void) · line 426

Quoted 11 lines
arch/ia64/kernel/efi.c +3 -1
-static unsigned long mem_limit = ~0UL, max_addr = ~0UL;
+static unsigned long mem_limit = ~0UL, max_addr = ~0UL, min_addr = 0UL;
 
 #define efi_call_virt(f, args...)	(*(f))(args)
@@ -421,6 +422,8 @@ efi_init (void)
 			mem_limit = memparse(cp + 4, &cp);
 		} else if (memcmp(cp, "max_addr=", 9) = 0) {
 			max_addr = GRANULEROUNDDOWN(memparse(cp + 9, &cp));
+		} else if (memcmp(cp, "min_addr=", 9) = 0) {
+			min_addr = GRANULEROUNDDOWN(memparse(cp + 9, &cp));

min_addr= looks like it could be a separate patch.

Quoted 8 lines
arch/ia64/kernel/iosapic.c +8 -0
+#ifdef CONFIG_CRASH_DUMP
+void
+kdump_disable_iosapic(void)
+{
+	u32 low32;
+	struct iosapic_intr_info *info;
+	struct iosapic_rte_info *rte;
+	for (info = iosapic_intr_info; info <

arch/ia64/kernel/iosapic.c nop (unsigned int irq) · line 308

Quoted 10 lines
arch/ia64/kernel/iosapic.c +10 -0
+			iosapic_intr_info + IA64_NUM_VECTORS; ++info) {
+		low32 = info->low32 |= IOSAPIC_MASK;
+		list_for_each_entry(rte, &info->rtes,
+				rte_list) {
+			iosapic_write(rte->addr,
+					IOSAPIC_RTE_LOW(rte->rte_index), low32);
+		}
+	}
+}
+#endif

Disabling the iosapic could be a separate patch.

Quoted 2 lines

+/*
+ * Do what every setup is needed on image and the

ever

arch/ia64/kernel/setup.c reserve_memory (void) · line 258

Quoted 4 lines
arch/ia64/kernel/setup.c +4 -0
+#ifdef CONFIG_KEXEC
+	/* crashkernel=size@addr specifies the location to reserve for
+	 * a crash kernel.  By reserving this memory we guarantee
+	 * that linux never set's it up as a DMA target.

sets, or better, s/set's it up/uses it/

arch/ia64/kernel/setup.c reserve_memory (void) · line 263

Quoted 5 lines
arch/ia64/kernel/setup.c +5 -0
+	 * Useful for holding code to do something appropriate
+	 * after a kernel panic.
+	 */
+	{
+		char *from = strstr(saved_command_line, "crashkernel=");

crashkernel= looks like it could be a separate patch.

arch/ia64/kernel/setup.c setup_arch (char **cmdline_p) · line 526

Quoted 4 lines
arch/ia64/kernel/setup.c +4 -0
+		char *from = strstr(saved_command_line, "elfcorehdr=");
+
+		if (from)
+			elfcorehdr_addr = memparse(from+11, &from);

elfcorehdr_addr isn't referenced anywhere else.

kernel/irq/manage.c int request_irq(unsigned int irq, · line 478

Quoted 8 lines

diff -Nraup linux-2.6.18-rc5/kernel/irq/manage.c linux-2.6.18-rc5-kdump/kernel/irq/manage.c

kernel/irq/manage.c +0 -1
@@ -475,4 +475,3 @@ int request_irq(unsigned int irq,
 	return retval;
 }
 EXPORT_SYMBOL(request_irq);
-

Extraneous whitespace change.

#2 lore raw
In reply to Zou Nan hai
Quoted 10 lines · Bjorn Helgaas

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1629

Quoted 4 lines · Zou Nan hai
arch/ia64/hp/common/sba_iommu.c +4 -0
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{

Ugh. If you really need this functionality (which I have to say looks like a band-aid), it probably should be a platform vector. And should be split into a separate patch.

Hi Bjorn,
The ioc_iova_disable code comes from Aziz in HP, I have almost no idea
of how IOMMU works on HP platform.
I am looking for an HP machine with IOMMU to test.

Quoted 8 lines
Quoted 8 lines
arch/ia64/hp/common/sba_iommu.c +8 -0
+	struct ioc *ioc;
+
+	ioc = ioc_list;
+
+	while (ioc != NULL) {
+		/* Disable IOVA translation */
+		WRITE_REG(ioc->ibase & 0xfffffffffffffffe, ioc->ioc_hpa + IOC_IBASE);
+		READ_REG(ioc->ioc_hpa + IOC_IBASE);
Quoted 9 lines

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1641

Quoted 4 lines
arch/ia64/hp/common/sba_iommu.c +4 -0
+
+		/* Clear I/O TLB of any possible entries */
+		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) + iovp_shift), ioc->ioc_hpa + IOC_PCOM);
+		READ_REG(ioc->ioc_hpa + IOC_PCOM);

This will just make any future device DMA attempts fail with an MCA, won't it? What problem does that solve? Don't you need the same for other IOMMUs like SGI's?

I guess we don't need IOMMU shutdown code. However it will be helpful
if people have machine with IOMMU to test the code and verify that.

Quoted 15 lines

arch/ia64/Kconfig config SGI_SN · line 436

Quoted 7 lines
arch/ia64/Kconfig +7 -0
+config KEXEC
+	bool "kexec system call (EXPERIMENTAL)"
+	depends on EXPERIMENTAL
+	help
+	  kexec is a system call that implements the ability to shutdown your
+	  current kernel, and to start another kernel.  It is like a reboot
+	  but it is indepedent of the system firmware.   And like a reboot

independent

arch/ia64/Kconfig config SGI_SN · line 439

Quoted 3 lines
arch/ia64/Kconfig +3 -0
+	  you can start any kernel with it, not just Linux.
+
+	  The name comes from the similiarity to the exec system call.

similarity

Quoted 4 lines

arch/ia64/kernel/crash.c line 27

Quoted 2 lines
arch/ia64/kernel/crash.c +2 -0
+size_t copy_oldmem_page(unsigned long pfn, char *buf,
+                               size_t csize, unsigned long offset, int userbuf)

Doesn't seem to be used.

This function is called when the crash dumping kernel dump memory from
first crashed kernel.

Quoted 13 lines

arch/ia64/kernel/crash.c line 48

Quoted 7 lines
arch/ia64/kernel/crash.c +6 -0
+static void device_shootdown(void)
+{
+	kdump_disable_iosapic();
+#ifdef CONFIG_IA64_HP_ZX1
+	ioc_iova_disable();
+#endif

Seems like sort of a heavy-handed way to shut down devices. But maybe you don't have any alternatives, I don't know. I guess you don't do the pci_disable_device() thing here just to avoid depending on more code?

pci_disable_device is too heavy to use at crash time.
I have plan to put kdump_disable_iosapic into purgatory code. However
it is a relatively light function. For the ioc_iova_disable code, I need HP people to verify it is safe to remove the code on HP platform with IOMMU.

Quoted 75 lines

arch/ia64/kernel/crash.c line 54

Quoted 6 lines
arch/ia64/kernel/crash.c +6 -0
+}
+
+static inline Elf64_Word
+*append_elf_note(Elf64_Word *buf, char *name, unsigned type, void *data,
+		size_t data_len)
+{

All this ELF stuff looks like something that could be split into a separate patch.

arch/ia64/kernel/crash.c line 97

Quoted 8 lines
arch/ia64/kernel/crash.c +8 -0
+	ia64_dump_cpu_regs(dst);
+        cfm = dst[43];
+        sol = (cfm >> 7) & 0x7f;
+        sof = cfm & 0x7f;
+        dst[46] = (unsigned long)ia64_rse_skip_regs((unsigned long *)dst[46],
+                        sof - sol);
+
+        buf = (u64 *) per_cpu_ptr(crash_notes, cpu);

Funny indentation above (spaces rather than tab, I guess).

arch/ia64/kernel/efi.c efi_init (void) · line 426

Quoted 11 lines
arch/ia64/kernel/efi.c +3 -1
-static unsigned long mem_limit = ~0UL, max_addr = ~0UL;
+static unsigned long mem_limit = ~0UL, max_addr = ~0UL, min_addr = 0UL;
 
 #define efi_call_virt(f, args...)	(*(f))(args)
@@ -421,6 +422,8 @@ efi_init (void)
 			mem_limit = memparse(cp + 4, &cp);
 		} else if (memcmp(cp, "max_addr=", 9) = 0) {
 			max_addr = GRANULEROUNDDOWN(memparse(cp + 9, &cp));
+		} else if (memcmp(cp, "min_addr=", 9) = 0) {
+			min_addr = GRANULEROUNDDOWN(memparse(cp + 9, &cp));

min_addr= looks like it could be a separate patch.

Quoted 8 lines
arch/ia64/kernel/iosapic.c +8 -0
+#ifdef CONFIG_CRASH_DUMP
+void
+kdump_disable_iosapic(void)
+{
+	u32 low32;
+	struct iosapic_intr_info *info;
+	struct iosapic_rte_info *rte;
+	for (info = iosapic_intr_info; info <

arch/ia64/kernel/iosapic.c nop (unsigned int irq) · line 308

Quoted 10 lines
arch/ia64/kernel/iosapic.c +10 -0
+			iosapic_intr_info + IA64_NUM_VECTORS; ++info) {
+		low32 = info->low32 |= IOSAPIC_MASK;
+		list_for_each_entry(rte, &info->rtes,
+				rte_list) {
+			iosapic_write(rte->addr,
+					IOSAPIC_RTE_LOW(rte->rte_index), low32);
+		}
+	}
+}
+#endif

Disabling the iosapic could be a separate patch.

Quoted 2 lines

+/*
+ * Do what every setup is needed on image and the

ever

arch/ia64/kernel/setup.c reserve_memory (void) · line 258

Quoted 4 lines
arch/ia64/kernel/setup.c +4 -0
+#ifdef CONFIG_KEXEC
+	/* crashkernel=size@addr specifies the location to reserve for
+	 * a crash kernel.  By reserving this memory we guarantee
+	 * that linux never set's it up as a DMA target.

sets, or better, s/set's it up/uses it/

arch/ia64/kernel/setup.c reserve_memory (void) · line 263

Quoted 5 lines
arch/ia64/kernel/setup.c +5 -0
+	 * Useful for holding code to do something appropriate
+	 * after a kernel panic.
+	 */
+	{
+		char *from = strstr(saved_command_line, "crashkernel=");

crashkernel= looks like it could be a separate patch.

Quoted 7 lines

arch/ia64/kernel/setup.c setup_arch (char **cmdline_p) · line 526

Quoted 4 lines
arch/ia64/kernel/setup.c +4 -0
+		char *from = strstr(saved_command_line, "elfcorehdr=");
+
+		if (from)
+			elfcorehdr_addr = memparse(from+11, &from);

elfcorehdr_addr isn't referenced anywhere else.

elfcorehdr_addr is referenced from vmcore proc filesystem to generate
elf headers for crashdump core file.

Quoted 10 lines

kernel/irq/manage.c int request_irq(unsigned int irq, · line 478

Quoted 8 lines

diff -Nraup linux-2.6.18-rc5/kernel/irq/manage.c linux-2.6.18-rc5-kdump/kernel/irq/manage.c

kernel/irq/manage.c +0 -1
@@ -475,4 +475,3 @@ int request_irq(unsigned int irq,
 	return retval;
 }
 EXPORT_SYMBOL(request_irq);
-

Extraneous whitespace change.

Thanks
Zou Nan hai

#3 lore raw
In reply to Zou Nan hai
Quoted 13 lines · Zou Nan hai
Quoted 12 lines · Bjorn Helgaas

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1629

Quoted 4 lines · Zou Nan hai
arch/ia64/hp/common/sba_iommu.c +4 -0
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{

Ugh. If you really need this functionality (which I have to say looks
like a band-aid), it probably should be a platform vector. And should
be split into a separate patch.

Quoted 5 lines

Hi Bjorn,
The ioc_iova_disable code comes from Aziz in HP, I have almost no idea
of how IOMMU works on HP platform.
I am looking for an HP machine with IOMMU to test.

Below are some minor cleanups against your latest patch. In particular the bogus kernel/irq/manage.c fragment is eliminated.

Signature · 194 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

IA64: kexec whitespace and 80 column fixes

Signed-OFf-By: Simon Horman <[email protected]>

 arch/ia64/hp/common/sba_iommu.c  |    6 ++++--
 arch/ia64/kernel/crash.c         |    6 +++---
 arch/ia64/kernel/efi.c           |    6 ++++--
 arch/ia64/kernel/iosapic.c       |    5 ++---
 arch/ia64/kernel/machine_kexec.c |   20 ++++++++++++--------
 arch/ia64/kernel/smp.c           |    2 +-
 include/linux/kexec.h            |    1 -
 kernel/irq/manage.c              |    1 +
 8 files changed, 27 insertions(+), 20 deletions(-)

Index: linux-2.6/arch/ia64/kernel/machine_kexec.c
=================================--- linux-2.6.orig/arch/ia64/kernel/machine_kexec.c	2006-08-30 16:48:14.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/machine_kexec.c	2006-08-30 16:58:02.000000000 +0900
@@ -57,9 +57,10 @@
 	/* Pre-load control code buffer to minimize work in kexec path */
 	control_code_buffer = page_address(image->control_code_page);
 	memcpy((void *)control_code_buffer, (const void *)func[0],
-			relocate_new_kernel_size);
+	       relocate_new_kernel_size);
 	flush_icache_range((unsigned long)control_code_buffer,
-			(unsigned long)control_code_buffer + relocate_new_kernel_size);
+   			   (unsigned long)control_code_buffer +
+   			   relocate_new_kernel_size);
 	ia64_kimage = image;
 
 	return 0;
@@ -81,14 +82,16 @@
 		}
 	}
 #elif defined(CONFIG_SMP)
-	smp_call_function(kexec_stop_this_cpu, (void *)ia64_kimage->start, 0, 0);
+	smp_call_function(kexec_stop_this_cpu, (void *)ia64_kimage->start,
+			  0, 0);
 #endif
 #ifdef CONFIG_PCI
 	{
 		struct pci_dev *dev = NULL;
 		irq_desc_t *idesc;
 		/* Disable all PCI devices */
-		while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+		while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev))
+		       != NULL) {
 			if (!(dev->is_enabled))
 				continue;
 			idesc = irq_desc + dev->irq;
@@ -112,12 +115,13 @@
  * We are past the point of no return, committed to rebooting now.
  */
 extern void *efi_get_pal_addr(void);
-static void ia64_machine_kexec(struct unw_frame_info *info, void *arg) 
+static void ia64_machine_kexec(struct unw_frame_info *info, void *arg)
 {
 	struct kimage *image = arg;
 	relocate_new_kernel_t rnk;
 	void *pal_addr = efi_get_pal_addr();
-	unsigned long code_addr = (unsigned long)page_address(image->control_code_page);
+	unsigned long code_addr = (unsigned long)
+				  page_address(image->control_code_page);
 	if (image->type = KEXEC_TYPE_CRASH) {
 		crash_save_this_cpu();
 		current->thread.ksp = (__u64)info->sw - 16;
@@ -128,9 +132,9 @@
 	local_irq_disable();
 	rnk = (relocate_new_kernel_t)&code_addr;
 	(*rnk)(image->head, image->start, ia64_boot_param,
-		     GRANULEROUNDDOWN((unsigned long) pal_addr));
+	       GRANULEROUNDDOWN((unsigned long) pal_addr));
 	BUG();
-} 
+}
 
 void machine_kexec(struct kimage *image)
 {
Index: linux-2.6/arch/ia64/kernel/smp.c
=================================--- linux-2.6.orig/arch/ia64/kernel/smp.c	2006-08-30 16:48:15.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/smp.c	2006-08-30 16:52:21.000000000 +0900
@@ -127,7 +127,7 @@
 }
 
 #ifdef CONFIG_CRASH_DUMP
-static void 
+static void
 kdump_cpu_freeze(struct unw_frame_info *info, void *arg)
 {
 	local_irq_disable();
Index: linux-2.6/include/linux/kexec.h
=================================--- linux-2.6.orig/include/linux/kexec.h	2006-08-30 16:48:33.000000000 +0900
+++ linux-2.6/include/linux/kexec.h	2006-08-30 16:49:48.000000000 +0900
@@ -135,7 +135,6 @@
 typedef u32 note_buf_t[MAX_NOTE_BYTES/4];
 extern note_buf_t *crash_notes;
 
-
 #else /* !CONFIG_KEXEC */
 struct pt_regs;
 struct task_struct;
Index: linux-2.6/arch/ia64/hp/common/sba_iommu.c
=================================--- linux-2.6.orig/arch/ia64/hp/common/sba_iommu.c	2006-08-30 16:53:08.000000000 +0900
+++ linux-2.6/arch/ia64/hp/common/sba_iommu.c	2006-08-30 16:54:21.000000000 +0900
@@ -1633,11 +1633,13 @@
 
 	while (ioc != NULL) {
 		/* Disable IOVA translation */
-		WRITE_REG(ioc->ibase & 0xfffffffffffffffe, ioc->ioc_hpa + IOC_IBASE);
+		WRITE_REG(ioc->ibase & 0xfffffffffffffffe,
+			  ioc->ioc_hpa + IOC_IBASE);
 		READ_REG(ioc->ioc_hpa + IOC_IBASE);
 
 		/* Clear I/O TLB of any possible entries */
-		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) + iovp_shift), ioc->ioc_hpa + IOC_PCOM);
+		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) +
+					iovp_shift), ioc->ioc_hpa + IOC_PCOM);
 		READ_REG(ioc->ioc_hpa + IOC_PCOM);
 
 		ioc = ioc->next;
Index: linux-2.6/arch/ia64/kernel/crash.c
=================================--- linux-2.6.orig/arch/ia64/kernel/crash.c	2006-08-30 16:53:08.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/crash.c	2006-08-30 16:55:28.000000000 +0900
@@ -24,7 +24,7 @@
 #include <asm/uaccess.h>
 
 size_t copy_oldmem_page(unsigned long pfn, char *buf,
-                               size_t csize, unsigned long offset, int userbuf)
+			size_t csize, unsigned long offset, int userbuf)
 {
         void  *vaddr;
 
@@ -50,7 +50,7 @@
 
 static inline Elf64_Word
 *append_elf_note(Elf64_Word *buf, char *name, unsigned type, void *data,
-		size_t data_len)
+ 		 size_t data_len)
 {
 	struct elf_note *note = (struct elf_note *)buf;
 	note->n_namesz = strlen(name) + 1;
@@ -92,7 +92,7 @@
         sol = (cfm >> 7) & 0x7f;
         sof = cfm & 0x7f;
         dst[46] = (unsigned long)ia64_rse_skip_regs((unsigned long *)dst[46],
-                        sof - sol);
+				    		    sof - sol);
 
         buf = (u64 *) per_cpu_ptr(crash_notes, cpu);
 	if (!buf)
Index: linux-2.6/arch/ia64/kernel/efi.c
=================================--- linux-2.6.orig/arch/ia64/kernel/efi.c	2006-08-30 16:53:08.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/efi.c	2006-08-30 16:55:55.000000000 +0900
@@ -432,9 +432,11 @@
 		}
 	}
 	if (min_addr != 0UL)
-		printk(KERN_INFO "Ignoring memory below %luMB\n", min_addr >> 20);
+		printk(KERN_INFO "Ignoring memory below %luMB\n",
+		       min_addr >> 20);
 	if (max_addr != ~0UL)
-		printk(KERN_INFO "Ignoring memory above %luMB\n", max_addr >> 20);
+		printk(KERN_INFO "Ignoring memory above %luMB\n",
+		       max_addr >> 20);
 
 	efi.systab = __va(ia64_boot_param->efi_systab);
 
Index: linux-2.6/arch/ia64/kernel/iosapic.c
=================================--- linux-2.6.orig/arch/ia64/kernel/iosapic.c	2006-08-30 16:53:08.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/iosapic.c	2006-08-30 16:56:45.000000000 +0900
@@ -298,10 +298,9 @@
 	for (info = iosapic_intr_info; info <
 			iosapic_intr_info + IA64_NUM_VECTORS; ++info) {
 		low32 = info->low32 |= IOSAPIC_MASK;
-		list_for_each_entry(rte, &info->rtes,
-				rte_list) {
+		list_for_each_entry(rte, &info->rtes, rte_list) {
 			iosapic_write(rte->addr,
-					IOSAPIC_RTE_LOW(rte->rte_index), low32);
+				      IOSAPIC_RTE_LOW(rte->rte_index), low32);
 		}
 	}
 }
Index: linux-2.6/kernel/irq/manage.c
=================================--- linux-2.6.orig/kernel/irq/manage.c	2006-08-30 16:53:08.000000000 +0900
+++ linux-2.6/kernel/irq/manage.c	2006-08-30 17:00:10.000000000 +0900
@@ -475,3 +475,4 @@
 	return retval;
 }
 EXPORT_SYMBOL(request_irq);
+
#4 lore raw
In reply to Zou Nan hai
Quoted 13 lines · Zou Nan hai
Quoted 12 lines · Bjorn Helgaas

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1629

Quoted 4 lines · Zou Nan hai
arch/ia64/hp/common/sba_iommu.c +4 -0
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{

Ugh. If you really need this functionality (which I have to say looks
like a band-aid), it probably should be a platform vector. And should
be split into a separate patch.

Quoted 5 lines

Hi Bjorn,
The ioc_iova_disable code comes from Aziz in HP, I have almost no idea
of how IOMMU works on HP platform.
I am looking for an HP machine with IOMMU to test.

Hi,

Below is an incremental version of your patch relative to the the V3 version that you posted about 10 days ago.

The entire series for Tony Luck's ia64-test and 2.6.18-rc5 can be found at:
http://www.vergenet.net/~horms/patches/ia64-kexec/kernel/

Signature · 5 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

IA64: kexec/kdump 2.6.18-rc5 patch
Quoted 19 lines

Archived-At: <http://permalink.gmane.org/gmane.linux.ports.ia64/14761>

Hi,
Below is the IA64 kexec/kdump patch against 2.6.18-rc5.

Fixes and enhancements in this patch include:

1. Fix I/D cache coherence problem.
Kdump sometimes hit an I/D cache coherence issue on platform with separate I/D cache. Although there is fc.i instruction in relocate_kernel.S. purgatory code and the second kernel code is copy into reserved region at kexec_load when running "kexec -p". There need and an icache_flush when segments is copied into its destination.
2. Change elf_prstatus to an per_cpu value to save stack size at crash path according
to Bob Montgomery's suggestion.
3. put AP to a loop of hint.pause instead of call pal_halt_light.

Quoted 1 line

Incremental version of the above patch

 arch/ia64/kernel/crash.c |   18 +++++++++++-------
 arch/ia64/kernel/smp.c   |    3 ++-
 include/asm-ia64/kexec.h |    5 +++++
 include/linux/kexec.h    |    5 +++++
 kernel/kexec.c           |    1 +
 5 files changed, 24 insertions(+), 8 deletions(-)

Index: linux-2.6/arch/ia64/kernel/crash.c
=================================--- linux-2.6.orig/arch/ia64/kernel/crash.c 2006-08-30 16:47:25.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/crash.c 2006-08-30 16:47:38.000000000 +0900
@@ -72,16 +72,20 @@

extern void ia64_dump_cpu_regs(void *);

+static DEFINE_PER_CPU(struct elf_prstatus, elf_prstatus); +
void
crash_save_this_cpu()
{
void *buf;
- struct elf_prstatus prstatus;
- int cpu = smp_processor_id();
unsigned long cfm, sof, sol;
- elf_greg_t *dst = (elf_greg_t *)&prstatus.pr_reg;
- memset(&prstatus, 0, sizeof(prstatus));
- prstatus.pr_pid = current->pid;
+
+ int cpu = smp_processor_id();
+ struct elf_prstatus *prstatus = &per_cpu(elf_prstatus, cpu);
+
+ elf_greg_t *dst = (elf_greg_t *)&(prstatus->pr_reg);
+ memset(prstatus, 0, sizeof(*prstatus));
+ prstatus->pr_pid = current->pid;

 	ia64_dump_cpu_regs(dst);
         cfm = dst[43];
@@ -93,8 +97,8 @@
         buf = (u64 *) per_cpu_ptr(crash_notes, cpu);

if (!buf)
return;
- buf = append_elf_note(buf, "CORE", NT_PRSTATUS, &prstatus,
- sizeof(prstatus));
+ buf = append_elf_note(buf, "CORE", NT_PRSTATUS, prstatus,
+ sizeof(*prstatus));
final_note(buf);
}

Index: linux-2.6/arch/ia64/kernel/smp.c
=================================--- linux-2.6.orig/arch/ia64/kernel/smp.c 2006-08-30 16:47:25.000000000 +0900
+++ linux-2.6/arch/ia64/kernel/smp.c 2006-08-30 16:47:38.000000000 +0900
@@ -133,7 +133,8 @@
local_irq_disable(); crash_save_this_cpu(); current->thread.ksp = (__u64)info->sw - 16;
- cpu_halt();
+ for (;;)
+ ia64_hint(ia64_hint_pause);
}
#endif

Index: linux-2.6/include/asm-ia64/kexec.h
=================================--- linux-2.6.orig/include/asm-ia64/kexec.h 2006-08-30 16:47:25.000000000 +0900
+++ linux-2.6/include/asm-ia64/kexec.h 2006-08-30 16:47:38.000000000 +0900
@@ -20,6 +20,11 @@
#define vmlpt_bits (impl_va_bits - PAGE_SHIFT + pte_bits)
#define POW2(n) (1ULL << (n))

+#define kexec_flush_icache_page(page) do { \
+ unsigned long page_addr = (unsigned long)page_address(page); \
+ flush_icache_range(page_addr, page_addr + PAGE_SIZE); \
+ } while(0)
+
DECLARE_PER_CPU(u64, ia64_mca_pal_base); const extern unsigned int relocate_new_kernel_size; volatile extern long kexec_rendez;
Index: linux-2.6/include/linux/kexec.h
=================================--- linux-2.6.orig/include/linux/kexec.h 2006-08-30 16:45:50.000000000 +0900
+++ linux-2.6/include/linux/kexec.h 2006-08-30 16:47:38.000000000 +0900
@@ -108,6 +108,10 @@
extern struct kimage *kexec_image;
extern struct kimage *kexec_crash_image;

+#ifndef kexec_flush_icache_page
+#define kexec_flush_icache_page(page)
+#endif
+
#define KEXEC_ON_CRASH 0x00000001
#define KEXEC_ARCH_MASK 0xffff0000

@@ -131,6 +135,7 @@
typedef u32 note_buf_t[MAX_NOTE_BYTES/4];
extern note_buf_t *crash_notes;

+
#else /* !CONFIG_KEXEC */
struct pt_regs;
struct task_struct;
Index: linux-2.6/kernel/kexec.c
=================================--- linux-2.6.orig/kernel/kexec.c 2006-08-30 16:45:50.000000000 +0900
+++ linux-2.6/kernel/kexec.c 2006-08-30 16:47:38.000000000 +0900
@@ -851,6 +851,7 @@
memset(ptr + uchunk, 0, mchunk - uchunk);
}
result = copy_from_user(ptr, buf, uchunk);
+ kexec_flush_icache_page(page);
kunmap(page);
if (result) {
result = (result < 0) ? result : -EIO;

#5 lore raw
In reply to Zou Nan hai
Quoted 13 lines · Zou Nan hai
Quoted 12 lines · Bjorn Helgaas

arch/ia64/hp/common/sba_iommu.c ioc_iova_init(struct ioc *ioc) · line 1629

Quoted 4 lines · Zou Nan hai
arch/ia64/hp/common/sba_iommu.c +4 -0
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{

Ugh. If you really need this functionality (which I have to say looks
like a band-aid), it probably should be a platform vector. And should
be split into a separate patch.

Quoted 5 lines

Hi Bjorn,
The ioc_iova_disable code comes from Aziz in HP, I have almost no idea
of how IOMMU works on HP platform.
I am looking for an HP machine with IOMMU to test.

That sounds like even more reason to break it out into a separate patch.

Actually, I really think that you sould either provide a set or smaller patches, or incremental patches. Its quite hard to follow what is changing with the current jumbo-patch format.

Its also quite unclear what if any portions you would like merged, and even which tree you want them merged into - you patches are against 2.6.18-rc5, but the ia64 tree seems the most likely path, and your code doesn't apply there because some of the changes are already present.

Signature · 3 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/
#6 lore raw
In reply to Zou Nan hai
Quoted 14 lines · Horms
Quoted 13 lines · Zou Nan hai
Quoted 12 lines · Bjorn Helgaas

Review

Quoted 4 lines · Zou Nan hai
(quoted diff) +4 -0
+#ifdef CONFIG_KEXEC
+void
+ioc_iova_disable(void)
+{

Ugh. If you really need this functionality (which I have to say looks
like a band-aid), it probably should be a platform vector. And should
be split into a separate patch.

Quoted 8 lines
Quoted 5 lines

Hi Bjorn,
The ioc_iova_disable code comes from Aziz in HP, I have almost no idea
of how IOMMU works on HP platform.
I am looking for an HP machine with IOMMU to test.

Below are some minor cleanups against your latest patch. In particular the bogus kernel/irq/manage.c fragment is eliminated.

Jes Sorensen pointed out to me that the kernel/irq/manage.c fragment of the previous version of this patch introduces a bogus blank line - and nothing else. This new version rectifies that problem.

Signature · 16 lines
Horms
  H: http://www.vergenet.net/~horms/
  W: http://www.valinux.co.jp/en/

IA64: kexec whitespace and 80 column fixes

Signed-OFf-By: Simon Horman <[email protected]>

 arch/ia64/hp/common/sba_iommu.c  |    6 ++++--
 arch/ia64/kernel/crash.c         |    6 +++---
 arch/ia64/kernel/efi.c           |    6 ++++--
 arch/ia64/kernel/iosapic.c       |    5 ++---
 arch/ia64/kernel/machine_kexec.c |   20 ++++++++++++--------
 arch/ia64/kernel/smp.c           |    2 +-
 include/linux/kexec.h            |    1 -
 7 files changed, 26 insertions(+), 20 deletions(-)
arch/ia64/hp/common/sba_iommu.c +4 -2
index aa4ef60..178e800 100644
@@ -1633,11 +1633,13 @@ ioc_iova_disable(void)
 
 	while (ioc != NULL) {
 		/* Disable IOVA translation */
-		WRITE_REG(ioc->ibase & 0xfffffffffffffffe, ioc->ioc_hpa + IOC_IBASE);
+		WRITE_REG(ioc->ibase & 0xfffffffffffffffe,
+			  ioc->ioc_hpa + IOC_IBASE);
 		READ_REG(ioc->ioc_hpa + IOC_IBASE);
 
 		/* Clear I/O TLB of any possible entries */
-		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) + iovp_shift), ioc->ioc_hpa + IOC_PCOM);
+		WRITE_REG(ioc->ibase | (get_iovp_order(ioc->iov_size) +
+					iovp_shift), ioc->ioc_hpa + IOC_PCOM);
 		READ_REG(ioc->ioc_hpa + IOC_PCOM);
 
 		ioc = ioc->next;
arch/ia64/kernel/crash.c +3 -3
index aa63c47..ae9d4ce 100644
@@ -24,7 +24,7 @@ #include <linux/device.h>
 #include <asm/uaccess.h>
 
 size_t copy_oldmem_page(unsigned long pfn, char *buf,
-                               size_t csize, unsigned long offset, int userbuf)
+			size_t csize, unsigned long offset, int userbuf)
 {
         void  *vaddr;
 
@@ -50,7 +50,7 @@ #endif
 
 static inline Elf64_Word
 *append_elf_note(Elf64_Word *buf, char *name, unsigned type, void *data,
-		size_t data_len)
+ 		 size_t data_len)
 {
 	struct elf_note *note = (struct elf_note *)buf;
 	note->n_namesz = strlen(name) + 1;
@@ -92,7 +92,7 @@ crash_save_this_cpu()
         sol = (cfm >> 7) & 0x7f;
         sof = cfm & 0x7f;
         dst[46] = (unsigned long)ia64_rse_skip_regs((unsigned long *)dst[46],
-                        sof - sol);
+				    		    sof - sol);
 
         buf = (u64 *) per_cpu_ptr(crash_notes, cpu);
 	if (!buf)
arch/ia64/kernel/efi.c +4 -2
index 6935452..ae4a115 100644
@@ -432,9 +432,11 @@ efi_init (void)
 		}
 	}
 	if (min_addr != 0UL)
-		printk(KERN_INFO "Ignoring memory below %luMB\n", min_addr >> 20);
+		printk(KERN_INFO "Ignoring memory below %luMB\n",
+		       min_addr >> 20);
 	if (max_addr != ~0UL)
-		printk(KERN_INFO "Ignoring memory above %luMB\n", max_addr >> 20);
+		printk(KERN_INFO "Ignoring memory above %luMB\n",
+		       max_addr >> 20);
 
 	efi.systab = __va(ia64_boot_param->efi_systab);
 
arch/ia64/kernel/iosapic.c +2 -3
index 9e9367a..9a3d0bf 100644
@@ -298,10 +298,9 @@ kdump_disable_iosapic(void)
 	for (info = iosapic_intr_info; info <
 			iosapic_intr_info + IA64_NUM_VECTORS; ++info) {
 		low32 = info->low32 |= IOSAPIC_MASK;
-		list_for_each_entry(rte, &info->rtes,
-				rte_list) {
+		list_for_each_entry(rte, &info->rtes, rte_list) {
 			iosapic_write(rte->addr,
-					IOSAPIC_RTE_LOW(rte->rte_index), low32);
+				      IOSAPIC_RTE_LOW(rte->rte_index), low32);
 		}
 	}
 }
arch/ia64/kernel/machine_kexec.c +12 -8
index 9deff9a..65b084f 100644
@@ -57,9 +57,10 @@ int machine_kexec_prepare(struct kimage 
 	/* Pre-load control code buffer to minimize work in kexec path */
 	control_code_buffer = page_address(image->control_code_page);
 	memcpy((void *)control_code_buffer, (const void *)func[0],
-			relocate_new_kernel_size);
+	       relocate_new_kernel_size);
 	flush_icache_range((unsigned long)control_code_buffer,
-			(unsigned long)control_code_buffer + relocate_new_kernel_size);
+   			   (unsigned long)control_code_buffer +
+   			   relocate_new_kernel_size);
 	ia64_kimage = image;
 
 	return 0;
@@ -81,14 +82,16 @@ #ifdef CONFIG_HOTPLUG_CPU
 		}
 	}
 #elif defined(CONFIG_SMP)
-	smp_call_function(kexec_stop_this_cpu, (void *)ia64_kimage->start, 0, 0);
+	smp_call_function(kexec_stop_this_cpu, (void *)ia64_kimage->start,
+			  0, 0);
 #endif
 #ifdef CONFIG_PCI
 	{
 		struct pci_dev *dev = NULL;
 		irq_desc_t *idesc;
 		/* Disable all PCI devices */
-		while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev)) != NULL) {
+		while ((dev = pci_get_device(PCI_ANY_ID, PCI_ANY_ID, dev))
+		       != NULL) {
 			if (!(dev->is_enabled))
 				continue;
 			idesc = irq_desc + dev->irq;
@@ -112,12 +115,13 @@ #endif
  * We are past the point of no return, committed to rebooting now.
  */
 extern void *efi_get_pal_addr(void);
-static void ia64_machine_kexec(struct unw_frame_info *info, void *arg) 
+static void ia64_machine_kexec(struct unw_frame_info *info, void *arg)
 {
 	struct kimage *image = arg;
 	relocate_new_kernel_t rnk;
 	void *pal_addr = efi_get_pal_addr();
-	unsigned long code_addr = (unsigned long)page_address(image->control_code_page);
+	unsigned long code_addr = (unsigned long)
+				  page_address(image->control_code_page);
 	if (image->type = KEXEC_TYPE_CRASH) {
 		crash_save_this_cpu();
 		current->thread.ksp = (__u64)info->sw - 16;
@@ -128,9 +132,9 @@ static void ia64_machine_kexec(struct un
 	local_irq_disable();
 	rnk = (relocate_new_kernel_t)&code_addr;
 	(*rnk)(image->head, image->start, ia64_boot_param,
-		     GRANULEROUNDDOWN((unsigned long) pal_addr));
+	       GRANULEROUNDDOWN((unsigned long) pal_addr));
 	BUG();
-} 
+}
 
 void machine_kexec(struct kimage *image)
 {
arch/ia64/kernel/smp.c +1 -1
index 154903d..0a4a763 100644
@@ -127,7 +127,7 @@ stop_this_cpu (void)
 }
 
 #ifdef CONFIG_CRASH_DUMP
-static void 
+static void
 kdump_cpu_freeze(struct unw_frame_info *info, void *arg)
 {
 	local_irq_disable();
include/linux/kexec.h +0 -1
index 75fbb7e..c790e08 100644
@@ -135,7 +135,6 @@ extern struct resource crashk_res;
 typedef u32 note_buf_t[MAX_NOTE_BYTES/4];
 extern note_buf_t *crash_notes;
 
-
 #else /* !CONFIG_KEXEC */
 struct pt_regs;
 struct task_struct;
kernel/irq/manage.c +0 -0
#7 lore raw
In reply to Zou Nan hai
Quoted 3 lines

Hi,
Below is the IA64 kexec/kdump patch against 2.6.18-rc5.

...

Quoted 4 lines

3. put AP to a loop of hint.pause instead of call pal_halt_light.
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/smp.c linux-2.6.18-rc5-kdump/arch/ia64/kernel/smp.c

arch/ia64/kernel/smp.c +0 -0

...

Quoted 12 lines
arch/ia64/kernel/smp.c +12 -0
+void
+kexec_stop_this_cpu (void *func)
+{
+	unsigned long pta, impl_va_bits, pal_base;
+
+	/*
+	 * Remove this CPU by putting it into fake SAL rendezvous
+	 */
+	cpu_clear(smp_processor_id(), cpu_online_map);
+	max_xtp();
+	ia64_eoi();
+

arch/ia64/kernel/smp.c unlock_ipi_calllock(void) · line 115

Quoted 10 lines
arch/ia64/kernel/smp.c +10 -0
+	/* Disable VHPT */
+	impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
+	pta = POW2(61) - POW2(vmlpt_bits);
+	ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | 0);
+
+	local_irq_disable();
+	pal_base = __get_cpu_var(ia64_mca_pal_base);
+	kexec_fake_sal_rendez(func, ap_wakeup_vector, pal_base);
+}
+#endif

What was the reason for introducing the kexec_fake_sal_rendez() function instead of actually returning to the real SAL slave loop. The HOTPLUG_CPU code in play_dead() in arch/ia64/kernel/process.c is very similar to what is needed.

I'm sure the problem is platform specific, but on the SN platform, the other cpus must be sent back to the real SAL slave loops. Otherwise, targeting of IO interrupts will not work correctly in the new kexec'd kernel.

IO interrupts are distributed across cpus that are not in the SAL slave loop. If cpus are idled in the OS instead of SAL, interrrupts are incorrected targeted to cpus that cannot respond.

-- jack

#8 lore raw
In reply to Zou Nan hai
Quoted 27 lines · Jack Steiner
Quoted 3 lines

Hi,
Below is the IA64 kexec/kdump patch against 2.6.18-rc5.

...

Quoted 4 lines

3. put AP to a loop of hint.pause instead of call pal_halt_light.
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/smp.c linux-2.6.18-rc5-kdump/arch/ia64/kernel/smp.c

arch/ia64/kernel/smp.c +0 -0

...

Quoted 18 lines
arch/ia64/kernel/smp.c +18 -0
+void
+kexec_stop_this_cpu (void *func)
+{
+	unsigned long pta, impl_va_bits, pal_base;
+
+	/*
+	 * Remove this CPU by putting it into fake SAL rendezvous
+	 */
+	cpu_clear(smp_processor_id(), cpu_online_map);
+	max_xtp();
+	ia64_eoi();
+
+	/* Disable VHPT */
+	impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
+	pta = POW2(61) - POW2(vmlpt_bits);
+	ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | 0);
+
+	local_irq_disable();
Quoted 8 lines

arch/ia64/kernel/smp.c unlock_ipi_calllock(void) · line 115

Quoted 4 lines
arch/ia64/kernel/smp.c +4 -0
+	pal_base = __get_cpu_var(ia64_mca_pal_base);
+	kexec_fake_sal_rendez(func, ap_wakeup_vector, pal_base);
+}
+#endif

What was the reason for introducing the kexec_fake_sal_rendez() function instead of actually returning to the real SAL slave loop. The HOTPLUG_CPU code in play_dead() in arch/ia64/kernel/process.c is very similar to what is needed.

Hmmm. I may have answered at least part of my question. It appears that the boot cpu cannot exit back to the SAL slave loop since it was never in the slave loop to start with.

This will take some thought..... More later.

Quoted 13 lines

I'm sure the problem is platform specific, but on the SN platform, the other cpus must be sent back to the real SAL slave loops. Otherwise, targeting of IO interrupts will not work correctly in the new kexec'd kernel.

IO interrupts are distributed across cpus that are not in the SAL slave loop. If cpus are idled in the OS instead of SAL, interrrupts are incorrected targeted to cpus that cannot respond.

-- jack

#9 lore raw
In reply to Zou Nan hai
Quoted 4 lines

Hmmm. I may have answered at least part of my question. It appears that the boot cpu cannot exit back to the SAL slave loop since it was never in the slave loop to start with.

This will take some thought..... More later.

Yes. cpu0 is a special case as there is no way to return it to SAL. Linux hotplug code has a hack where we borrow the return details from some other cpu in the case that someone wants to take cpu0 offline. Will this work for Altix? Would we have to be careful to get the return details from some other cpu on the same node?

-Tony

#10 lore raw
In reply to Zou Nan hai
Quoted 5 lines · Tony
Quoted 4 lines

Hmmm. I may have answered at least part of my question. It appears that the boot cpu cannot exit back to the SAL slave loop since it was never in the slave loop to start with.

This will take some thought..... More later.

Quoted 7 lines

Yes. cpu0 is a special case as there is no way to return it to SAL. Linux hotplug code has a hack where we borrow the return details from some other cpu in the case that someone wants to take cpu0 offline. Will this work for Altix? Would we have to be careful to get the return details from some other cpu on the same node?

-Tony

Interesting idea. We might be able to make this work. It looks like we need to make some changes to our BIOS to make this work but it looks possible.

I'll investigate this some more.....

-- jack

#11 lore raw
In reply to Zou Nan hai
Quoted 27 lines · Jack Steiner
Quoted 3 lines

Hi,
Below is the IA64 kexec/kdump patch against 2.6.18-rc5.

...

Quoted 4 lines

3. put AP to a loop of hint.pause instead of call pal_halt_light.
diff -Nraup linux-2.6.18-rc5/arch/ia64/kernel/smp.c linux-2.6.18-rc5-kdump/arch/ia64/kernel/smp.c

arch/ia64/kernel/smp.c +0 -0

...

Quoted 18 lines
arch/ia64/kernel/smp.c +18 -0
+void
+kexec_stop_this_cpu (void *func)
+{
+	unsigned long pta, impl_va_bits, pal_base;
+
+	/*
+	 * Remove this CPU by putting it into fake SAL rendezvous
+	 */
+	cpu_clear(smp_processor_id(), cpu_online_map);
+	max_xtp();
+	ia64_eoi();
+
+	/* Disable VHPT */
+	impl_va_bits = ffz(~(local_cpu_data->unimpl_va_mask | (7UL << 61)));
+	pta = POW2(61) - POW2(vmlpt_bits);
+	ia64_set_pta(pta | (0 << 8) | (vmlpt_bits << 2) | 0);
+
+	local_irq_disable();
Quoted 9 lines

arch/ia64/kernel/smp.c unlock_ipi_calllock(void) · line 115

Quoted 4 lines
arch/ia64/kernel/smp.c +4 -0
+	pal_base = __get_cpu_var(ia64_mca_pal_base);
+	kexec_fake_sal_rendez(func, ap_wakeup_vector, pal_base);
+}
+#endif

What was the reason for introducing the kexec_fake_sal_rendez() function instead of actually returning to the real SAL slave loop. The HOTPLUG_CPU code in play_dead() in arch/ia64/kernel/process.c is very similar to what is needed.

the fake_sal_rendez code is from Aziz to be used at time of kexec -l
if CONFIG_HOTPLUG_CPU is not defined. They are not executed at the time of crash dump.

Quoted 10 lines

I'm sure the problem is platform specific, but on the SN platform, the other cpus must be sent back to the real SAL slave loops. Otherwise, targeting of IO interrupts will not work correctly in the new kexec'd kernel.

IO interrupts are distributed across cpus that are not in the SAL slave loop. If cpus are idled in the OS instead of SAL, interrrupts are incorrected targeted to cpus that cannot respond.

At the time of crash, neither fake nor real SAL rendez state are
entered. I just put all the other cpus into cpu_relax loop to simplify the code. Does it work on SN2 if you call ia64_jump_to_sal at kdump_cpu_freeze?

Thanks
Zou Nan hai

Quoted 2 lines

-- jack