From: George Beshers <gbeshers@redhat.com>
Subject: [RHEL5 - BZ#219091] Kexec, Kdump on SGI IA64 NUMA machines
Date: Mon, 18 Dec 2006 13:32:43 -0500
Bugzilla: 219091
Message-Id: <4586DECB.4040702@redhat.com>
Changelog: ia64: Kexec, Kdump on SGI IA64 NUMA machines fixes



BZ#219091 works in conjunction with RH#218105;
RH#219722 is covered by this commit:

Problem:
Supporting Kdump, Kexec on SGIs IA64 NUMA architectures which
is critical for field maintenance.

This is upstream as part of patch ae390618a9a4 by Zou Nan hai.

   [IA64] IA64 Kexec/kdump

   Changes and updates.

   1. Remove fake rendz path and related code according to discuss with
Khalid Aziz.
   2. fc.i offset fix in relocate_kernel.S.
   3. iospic shutdown code eoi and mask race fix from Fujitsu.
   4. Warm boot hook in machine_kexec to SN SAL code from Jack Steiner.
   5. Send slave to SAL slave loop patch from Jay Lan.
   6. Kdump on non-recoverable MCA event patch from Jay Lan
   7. Use CTL_UNNUMBERED in kdump_on_init sysctl.

   Signed-off-by: Zou Nan hai <nanhai.zou@intel.com 
   <mailto:nanhai.zou@intel.com>>
   Signed-off-by: Tony Luck <tony.luck@intel.com 
   <mailto:tony.luck@intel.com>>


The proposed patch has been built both and tested against 
kernel-2.6.18-1.2876.el5.

George




--- linux-2.6.18.ia64/arch/ia64/kernel/crash.c.orig	2006-12-11 11:13:29.000000000 -0500
+++ linux-2.6.18.ia64/arch/ia64/kernel/crash.c	2006-12-11 11:13:56.000000000 -0500
@@ -24,6 +24,7 @@
 int kdump_status[NR_CPUS];
 atomic_t kdump_cpu_freezed;
 int kdump_on_init = 1;
+extern int kdump_in_progress;
 
 ssize_t
 copy_oldmem_page(unsigned long pfn, char *buf,
@@ -146,14 +147,21 @@
 void
 kdump_cpu_freeze(struct unw_frame_info *info, void *arg)
 {
+	int cpuid = smp_processor_id();
+
 	local_irq_disable();
 	crash_save_this_cpu();
 	current->thread.ksp = (__u64)info->sw - 16;
 	atomic_inc(&kdump_cpu_freezed);
-	kdump_status[smp_processor_id()] = 1;
+	kdump_status[cpuid] = 1;
 	mb();
-	for (;;)
-		cpu_relax();
+        /* return cpus (except cpu0) to SAL slave loop */
+        if (cpuid == 0) {
+                for (;;)
+                        cpu_relax();
+        } else {
+                ia64_jump_to_sal(&sal_boot_rendez_state[cpuid]);
+        }
 }
 
 static int
@@ -165,12 +173,16 @@
 	if (!kdump_on_init)
 		return NOTIFY_DONE;
 
-	if (val != DIE_INIT_MONARCH_ENTER && val != DIE_INIT_SLAVE_ENTER)
+	if (val != DIE_INIT_MONARCH_ENTER &&
+	    val != DIE_INIT_SLAVE_ENTER &&
+	    val != DIE_MCA_RENDZVOUS_LEAVE &&
+	    val != DIE_MCA_MONARCH_LEAVE)
 		return NOTIFY_DONE;
 
 	nd = (struct ia64_mca_notify_die *)args->err;
 	/* Reason code 1 means machine check rendezous*/
-	if (nd->sos->rv_rc == 1)
+	if ((val==DIE_INIT_MONARCH_ENTER || val==DIE_INIT_SLAVE_ENTER) &&
+	    nd->sos->rv_rc == 1)
 		return NOTIFY_DONE;
 
 	if (kdump_sending_init)
@@ -183,6 +195,15 @@
 		case DIE_INIT_SLAVE_ENTER:
 			unw_init_running(kdump_cpu_freeze, NULL);
 			break;
+		case DIE_MCA_RENDZVOUS_LEAVE:
+			if (kdump_in_progress)
+				unw_init_running(kdump_cpu_freeze, NULL);
+			break;
+		case DIE_MCA_MONARCH_LEAVE:
+			/* die_register->signr indicate if MCA is recoverable */
+			if (!args->signr)
+				machine_kdump_on_init();
+			break;
 	}
 	return NOTIFY_DONE;
 }
--- linux-2.6.18.ia64/include/asm-ia64/machvec.h.orig	2006-09-19 23:42:06.000000000 -0400
+++ linux-2.6.18.ia64/include/asm-ia64/machvec.h	2006-12-11 11:13:41.000000000 -0500
@@ -35,6 +35,7 @@
 typedef int ia64_mv_pci_legacy_write_t (struct pci_bus *, u16 port, u32 val,
 					u8 size);
 typedef void ia64_mv_migrate_t(struct task_struct * task);
+typedef void ia64_mv_kernel_launch_event_t(void);
 
 /* DMA-mapping interface: */
 typedef void ia64_mv_dma_init (void);
@@ -205,6 +206,7 @@
 	ia64_mv_readq_relaxed_t *readq_relaxed;
 	ia64_mv_migrate_t *migrate;
 	ia64_mv_msi_init_t *msi_init;
+	ia64_mv_kernel_launch_event_t *kernel_launch_event;
 } __attribute__((__aligned__(16))); /* align attrib? see above comment */
 
 #define MACHVEC_INIT(name)			\
@@ -251,6 +253,7 @@
 	platform_readq_relaxed,			\
 	platform_migrate,			\
 	platform_msi_init,			\
+	platform_kernel_launch_event,		\
 }
 
 extern struct ia64_machine_vector ia64_mv;
@@ -303,6 +306,9 @@
 #ifndef platform_tlb_migrate_finish
 # define platform_tlb_migrate_finish	machvec_noop_mm
 #endif
+#ifndef platform_kernel_launch_event
+# define platform_kernel_launch_event	machvec_noop
+#endif
 #ifndef platform_dma_init
 # define platform_dma_init		swiotlb_init
 #endif
--- linux-2.6.18.ia64/include/asm-ia64/machvec_sn2.h.orig	2006-09-19 23:42:06.000000000 -0400
+++ linux-2.6.18.ia64/include/asm-ia64/machvec_sn2.h	2006-12-11 11:13:42.000000000 -0500
@@ -67,6 +67,7 @@
 extern ia64_mv_dma_mapping_error	sn_dma_mapping_error;
 extern ia64_mv_dma_supported		sn_dma_supported;
 extern ia64_mv_migrate_t		sn_migrate;
+extern ia64_mv_kernel_launch_event_t	sn_kernel_launch_event;
 extern ia64_mv_msi_init_t		sn_msi_init;
 
 
@@ -119,6 +120,7 @@
 #define platform_dma_mapping_error		sn_dma_mapping_error
 #define platform_dma_supported		sn_dma_supported
 #define platform_migrate		sn_migrate
+#define platform_kernel_launch_event	sn_kernel_launch_event
 #ifdef CONFIG_PCI_MSI
 #define platform_msi_init		sn_msi_init
 #else
--- linux-2.6.18.ia64/include/asm-ia64/sn/sn_sal.h.orig	2006-09-19 23:42:06.000000000 -0400
+++ linux-2.6.18.ia64/include/asm-ia64/sn/sn_sal.h	2006-12-11 11:13:42.000000000 -0500
@@ -87,6 +87,8 @@
 #define  SN_SAL_INJECT_ERROR			   0x02000067
 #define  SN_SAL_SET_CPU_NUMBER			   0x02000068
 
+#define SN_SAL_KERNEL_LAUNCH_EVENT		   0x02000069
+
 /*
  * Service-specific constants
  */
@@ -1154,4 +1156,13 @@
 	SAL_CALL_NOLOCK(rv, SN_SAL_SET_CPU_NUMBER, cpu, 0, 0, 0, 0, 0, 0);
 	return rv.status;
 }
+
+static inline int
+ia64_sn_kernel_launch_event(void)
+{
+	struct ia64_sal_retval rv;
+
+	SAL_CALL_NOLOCK(rv, SN_SAL_KERNEL_LAUNCH_EVENT, 0, 0, 0, 0, 0, 0, 0);
+	return rv.status;
+}
 #endif /* _ASM_IA64_SN_SN_SAL_H */
--- linux-2.6.18.ia64/arch/ia64/sn/kernel/setup.c.orig	2006-09-19 23:42:06.000000000 -0400
+++ linux-2.6.18.ia64/arch/ia64/sn/kernel/setup.c	2006-12-11 11:13:42.000000000 -0500
@@ -765,3 +765,11 @@
 }
 EXPORT_SYMBOL(sn_prom_feature_available);
 
+void
+sn_kernel_launch_event(void)
+{
+	/* ignore status until we understand possible failures, if any */
+	if (ia64_sn_kernel_launch_event())
+		printk(KERN_ERR "KEXEC is not supported in this PROM. Please update the PROM.\n");
+}
+
--- linux-2.6.18.ia64/arch/ia64/kernel/machine_kexec.c.orig	2006-12-11 11:09:49.000000000 -0500
+++ linux-2.6.18.ia64/arch/ia64/kernel/machine_kexec.c	2006-12-11 11:13:42.000000000 -0500
@@ -18,6 +18,7 @@
 #include <asm/setup.h>
 #include <asm/delay.h>
 #include <asm/meminit.h>
+#include <asm/machvec.h>
 
 typedef void (*relocate_new_kernel_t)(unsigned long, unsigned long,
 		struct ia64_boot_param *, unsigned long);
@@ -125,6 +126,7 @@
 		ia64_eoi();
 		vector = ia64_get_ivr();
 	}
+	platform_kernel_launch_event();
 	rnk = (relocate_new_kernel_t)&code_addr;
 	(*rnk)(image->head, image->start, ia64_boot_param,
 		     GRANULEROUNDDOWN((unsigned long) pal_addr));
--- linux-2.6.18.ia64/arch/ia64/kernel/mca.c.orig	2006-12-11 11:09:49.000000000 -0500
+++ linux-2.6.18.ia64/arch/ia64/kernel/mca.c	2006-12-11 11:13:56.000000000 -0500
@@ -92,6 +92,11 @@
 # define IA64_MCA_DEBUG(fmt...)
 #endif
 
+#ifdef CONFIG_KEXEC
+/* Used by arch/ia64/kernel/crash.c */
+int kdump_in_progress;
+#endif
+
 /* Used by mca_asm.S */
 u32				ia64_mca_serialize;
 DEFINE_PER_CPU(u64, ia64_mca_data); /* == __per_cpu_mca[smp_processor_id()] */
@@ -1066,6 +1071,15 @@
 		rh->severity = sal_log_severity_corrected;
 		ia64_sal_clear_state_info(SAL_INFO_TYPE_MCA);
 		sos->os_status = IA64_MCA_CORRECTED;
+#ifdef CONFIG_KEXEC
+	} else {
+		kdump_in_progress = 1;
+		/* In the case of (!recover), notify_die(DIE_MCA_MONARCH_LEAVE)
+		   will not return. A dump kernel will be booted. Need to set
+		   nonarch_cpu here to get slave cpus out of looping in OS.
+		 */
+		monarch_cpu = -1;
+#endif
 	}
 	if (notify_die(DIE_MCA_MONARCH_LEAVE, "MCA", regs, (long)&nd, 0, recover)
 			== NOTIFY_STOP)