[PATCH] kdump: Use real pt_regs from exception

Makes kexec_crashdump() take a pt_regs * as an argument.  This allows to
get exact register state at the point of the crash.  If we come from direct
panic assertion NULL will be passed and the current registers saved before
crashdump.

This hooks into two places:
die(): check the conditions under which we will panic when calling
do_exit and go there directly with the pt_regs that caused the fatal
fault.

die_nmi(): If we receive an NMI lockup while in the kernel use the
pt_regs and go directly to crash_kexec(). We're probably nested up badly
at this point so this might be the only chance to escape with proper
information.

Signed-off-by: Alexander Nyberg <alexn@telia.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

6 files changed, 45 insertions, 16 deletions

diff --git a/arch/i386/kernel/crash.c b/arch/i386/kernel/crash.c
index a021681d21f8..8bdb4b6af0ff 100644
--- a/arch/i386/kernel/crash.c
+++ b/arch/i386/kernel/crash.c
@@ -100,12 +100,31 @@ static void crash_get_current_regs(struct pt_regs *regs)
 	regs->eip = (unsigned long)current_text_addr();
 }
 
-static void crash_save_self(void)
+/* CPU does not save ss and esp on stack if execution is already
+ * running in kernel mode at the time of NMI occurrence. This code
+ * fixes it.
+ */
+static void crash_setup_regs(struct pt_regs *newregs, struct pt_regs *oldregs)
+{
+	memcpy(newregs, oldregs, sizeof(*newregs));
+	newregs->esp = (unsigned long)&(oldregs->esp);
+	__asm__ __volatile__("xorl %eax, %eax;");
+	__asm__ __volatile__ ("movw %%ss, %%ax;" :"=a"(newregs->xss));
+}
+
+/* We may have saved_regs from where the error came from
+ * or it is NULL if via a direct panic().
+ */
+static void crash_save_self(struct pt_regs *saved_regs)
 {
 	struct pt_regs regs;
 	int cpu;
 	cpu = smp_processor_id();
-	crash_get_current_regs(&regs);
+
+	if (saved_regs)
+		crash_setup_regs(&regs, saved_regs);
+	else
+		crash_get_current_regs(&regs);
 	crash_save_this_cpu(&regs, cpu);
 }
 
@@ -124,15 +143,8 @@ static int crash_nmi_callback(struct pt_regs *regs, int cpu)
 		return 1;
 	local_irq_disable();
 
-	/* CPU does not save ss and esp on stack if execution is already
-	 * running in kernel mode at the time of NMI occurrence. This code
-	 * fixes it.
-	 */
 	if (!user_mode(regs)) {
-		memcpy(&fixed_regs, regs, sizeof(*regs));
-		fixed_regs.esp = (unsigned long)&(regs->esp);
-		__asm__ __volatile__("xorl %eax, %eax;");
-		__asm__ __volatile__ ("movw %%ss, %%ax;" :"=a"(fixed_regs.xss));
+		crash_setup_regs(&fixed_regs, regs);
 		regs = &fixed_regs;
 	}
 	crash_save_this_cpu(regs, cpu);
@@ -184,7 +196,7 @@ static void nmi_shootdown_cpus(void)
 }
 #endif
 
-void machine_crash_shutdown(void)
+void machine_crash_shutdown(struct pt_regs *regs)
 {
 	/* This function is only called after the system
 	 * has paniced or is otherwise in a critical state.
@@ -204,5 +216,5 @@ void machine_crash_shutdown(void)
 #if defined(CONFIG_X86_IO_APIC)
 	disable_IO_APIC();
 #endif
-	crash_save_self();
+	crash_save_self(regs);
 }
diff --git a/arch/i386/kernel/traps.c b/arch/i386/kernel/traps.c
index 207ea8ba7169..e458463ebc05 100644
--- a/arch/i386/kernel/traps.c
+++ b/arch/i386/kernel/traps.c
@@ -27,6 +27,7 @@
 #include <linux/ptrace.h>
 #include <linux/utsname.h>
 #include <linux/kprobes.h>
+#include <linux/kexec.h>
 
 #ifdef CONFIG_EISA
 #include <linux/ioport.h>
@@ -294,6 +295,9 @@ bug:
 	printk("Kernel BUG\n");
 }
 
+/* This is gone through when something in the kernel
+ * has done something bad and is about to be terminated.
+*/
 void die(const char * str, struct pt_regs * regs, long err)
 {
 	static struct {
@@ -341,6 +345,10 @@ void die(const char * str, struct pt_regs * regs, long err)
 	bust_spinlocks(0);
 	die.lock_owner = -1;
 	spin_unlock_irq(&die.lock);
+
+	if (kexec_should_crash(current))
+		crash_kexec(regs);
+
 	if (in_interrupt())
 		panic("Fatal exception in interrupt");
 
@@ -570,6 +578,15 @@ void die_nmi (struct pt_regs *regs, const char *msg)
 	console_silent();
 	spin_unlock(&nmi_print_lock);
 	bust_spinlocks(0);
+
+	/* If we are in kernel we are probably nested up pretty bad
+	 * and might aswell get out now while we still can.
+	*/
+	if (!user_mode(regs)) {
+		current->thread.trap_no = 2;
+		crash_kexec(regs);
+	}
+
 	do_exit(SIGSEGV);
 }
 
diff --git a/arch/ppc/kernel/machine_kexec.c b/arch/ppc/kernel/machine_kexec.c
index 435ad9ea0a83..b82535357d6d 100644
--- a/arch/ppc/kernel/machine_kexec.c
+++ b/arch/ppc/kernel/machine_kexec.c
@@ -34,7 +34,7 @@ void machine_shutdown(void)
 	}
 }
 
-void machine_crash_shutdown(void)
+void machine_crash_shutdown(struct pt_regs *regs)
 {
 	if (ppc_md.machine_crash_shutdown) {
 		ppc_md.machine_crash_shutdown();
diff --git a/arch/ppc64/kernel/machine_kexec.c b/arch/ppc64/kernel/machine_kexec.c
index 217965d60a45..06b25b59c8a8 100644
--- a/arch/ppc64/kernel/machine_kexec.c
+++ b/arch/ppc64/kernel/machine_kexec.c
@@ -34,7 +34,7 @@ note_buf_t crash_notes[NR_CPUS];
  * and if what it will achieve. Letting it be now to compile the code
  * in generic kexec environment
  */
-void machine_crash_shutdown(void)
+void machine_crash_shutdown(struct pt_regs *regs)
 {
 	/* do nothing right now */
 	/* smp_relase_cpus() if we want smp on panic kernel */
diff --git a/arch/s390/kernel/crash.c b/arch/s390/kernel/crash.c
index db38283c1f27..7bd169c58b0c 100644
--- a/arch/s390/kernel/crash.c
+++ b/arch/s390/kernel/crash.c
@@ -12,6 +12,6 @@
 
 note_buf_t crash_notes[NR_CPUS];
 
-void machine_crash_shutdown(void)
+void machine_crash_shutdown(struct pt_regs *regs)
 {
 }
diff --git a/arch/x86_64/kernel/crash.c b/arch/x86_64/kernel/crash.c
index 6183bcb85257..d7fa4248501c 100644
--- a/arch/x86_64/kernel/crash.c
+++ b/arch/x86_64/kernel/crash.c
@@ -22,7 +22,7 @@
 
 note_buf_t crash_notes[NR_CPUS];
 
-void machine_crash_shutdown(void)
+void machine_crash_shutdown(struct pt_regs *regs)
 {
 	/* This function is only called after the system
 	 * has paniced or is otherwise in a critical state.