Re: [PATCH] [revised -- version 2] Info dump on Oops or panic()

[Joshua Wise <jwise@google.com>]

On Thu, 28 Jun 2007, Kyle McMartin wrote:
> woah woah woah. This could push critical bits of the register dump or
> stacktrace off the screen... Barring any other problems with the patch,
> this should probably be dumped in the oops header, not trailing it where
> it could hide critical debugging info.

Okay, fair enough. Fixed version follows. I also fixed some checkpatch
issues that I missed before.

However, please note that in general, the info dumped by this will consist
of only a few lines. In our implementations, I believe that we only dump one
line per notifier.

joshua

--
From: Joshua Wise <jwise@google.com>

Version: 2

Background:
 When managing a large number of servers, as Google does, it's sometimes
 useful to get an "at-a-glance" view of a machine when it crashes. When no
 other post-mortem is possible, it's often useful to know how long the
 machine has been powered on, which kernel it was running, and possibly
 other information that a driver may have logged. Up until now, there was no
 real way to do this other than to keep statistics outside to the machine.

Description:
 This patch adds a call chain to be invoked when the system oopses or
 panics. Traps have been added to i386 and x86_64 kernels, but it should be
 trivial to add support for more architectures. Two sample notifiers have
 been added -- an uptime printer, and a utsname printer for showing various
 statistics about the running system.

Remaining issues:
 * Is do_posix_clock_monotonic_gettime really the right API?
 * Should this be on by default, or a disablable Kconfig entry?

Testing:
 I accidentally built my testing kernel without the IDE drivers. It panic()ed
 immediately, and informed me that my uptime was 0.38 seconds.

Credits:
 This code was mainly written by Mike Waychison <mikew@google.com> and
 Masoud Sharbiani <masouds@google.com>.

Changelog:
 v2 -- fixed some checkpatch issues. Moved dumping to before Oopses, as per
 the suggestion of Kyle McMartin <kyle@mcmartin.ca>.

Patch:
 This patch is against git 48d8d7ee5dd17c64833e0343ab4ae8ef01cc2648.

Signed-off-by: Joshua Wise <jwise@google.com>
Acked-by: Mike Waychison <mikew@google.com>

--

diff --git a/arch/i386/kernel/process.c b/arch/i386/kernel/process.c
index 06dfa65..f969c04 100644
--- a/arch/i386/kernel/process.c
+++ b/arch/i386/kernel/process.c
@@ -301,6 +301,8 @@ void show_regs(struct pt_regs * regs)
 {
 	unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L;

+	atomic_notifier_call_chain(&info_dumper_list, 0, NULL);
+
 	printk("\n");
 	printk("Pid: %d, comm: %20s\n", current->pid, current->comm);
 	printk("EIP: %04x:[<%08lx>] CPU: %d\n",0xffff & regs->xcs,regs->eip, smp_processor_id());
diff --git a/arch/i386/kernel/traps.c b/arch/i386/kernel/traps.c
index 90da057..2ad7f8c 100644
--- a/arch/i386/kernel/traps.c
+++ b/arch/i386/kernel/traps.c
@@ -289,6 +289,9 @@ void show_registers(struct pt_regs *regs
 		ss = regs->xss & 0xffff;
 	}
 	print_modules();
+
+	atomic_notifier_call_chain(&info_dumper_list, 0, NULL);
+
 	printk(KERN_EMERG "CPU:    %d\n"
 		KERN_EMERG "EIP:    %04x:[<%08lx>]    %s VLI\n"
 		KERN_EMERG "EFLAGS: %08lx   (%s %.*s)\n",
diff --git a/arch/x86_64/kernel/process.c b/arch/x86_64/kernel/process.c
index 5909039..e7f0298 100644
--- a/arch/x86_64/kernel/process.c
+++ b/arch/x86_64/kernel/process.c
@@ -353,6 +353,7 @@ void __show_regs(struct pt_regs * regs)

 void show_regs(struct pt_regs *regs)
 {
+	atomic_notifier_call_chain(&info_dumper_list, 0, NULL);
 	printk("CPU %d:", smp_processor_id());
 	__show_regs(regs);
 	show_trace(NULL, regs, (void *)(regs + 1));
diff --git a/arch/x86_64/kernel/traps.c b/arch/x86_64/kernel/traps.c
index aac1c0b..3d228d0 100644
--- a/arch/x86_64/kernel/traps.c
+++ b/arch/x86_64/kernel/traps.c
@@ -410,6 +410,8 @@ void show_registers(struct pt_regs *regs
 	const int cpu = smp_processor_id();
 	struct task_struct *cur = cpu_pda(cpu)->pcurrent;

+	atomic_notifier_call_chain(&info_dumper_list, 0, NULL);
+
 	rsp = regs->rsp;
 	printk("CPU %d ", cpu);
 	__show_regs(regs);
diff --git a/include/asm-blackfin/macros.h b/include/asm-blackfin/macros.h
deleted file mode 100644
index e69de29..0000000
diff --git a/include/linux/kernel.h b/include/linux/kernel.h
index 7a48525..2058e58 100644
--- a/include/linux/kernel.h
+++ b/include/linux/kernel.h
@@ -104,6 +104,7 @@ #define abs(x) ({				\
 	})

 extern struct atomic_notifier_head panic_notifier_list;
+extern struct atomic_notifier_head info_dumper_list;
 extern long (*panic_blink)(long time);
 NORET_TYPE void panic(const char * fmt, ...)
 	__attribute__ ((NORET_AND format (printf, 1, 2)));
diff --git a/kernel/panic.c b/kernel/panic.c
index 623d182..8bae618 100644
--- a/kernel/panic.c
+++ b/kernel/panic.c
@@ -97,6 +97,7 @@ #ifdef CONFIG_SMP
 #endif

 	atomic_notifier_call_chain(&panic_notifier_list, 0, buf);
+	atomic_notifier_call_chain(&info_dumper_list, 0, NULL);

 	if (!panic_blink)
 		panic_blink = no_blink;
diff --git a/kernel/sys.c b/kernel/sys.c
index 872271c..21b9ca1 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -98,6 +98,39 @@ struct pid *cad_pid;
 EXPORT_SYMBOL(cad_pid);

 /*
+ *	Notifier list for kernel code which wants to printk hardware specific
+ *	information at Oops or panic time.
+ */
+
+ATOMIC_NOTIFIER_HEAD(info_dumper_list);
+EXPORT_SYMBOL(info_dumper_list);
+
+/*
+ * Dump out UTS info on oops / panic.
+ */
+
+static int dump_utsname(struct notifier_block *self, unsigned long v, void *p)
+{
+	printk (KERN_EMERG "%s %s %s %s %s %s\n",
+		utsname()->sysname,
+		utsname()->nodename,
+		utsname()->release,
+		utsname()->version,
+		utsname()->machine,
+		utsname()->domainname);
+	return 0;
+}
+
+static struct notifier_block utsname_notifier;
+
+static int __init register_utsname_dump(void) {
+	utsname_notifier.notifier_call = dump_utsname;
+	atomic_notifier_chain_register(&info_dumper_list, &utsname_notifier);
+	return 0;
+}
+__initcall(register_utsname_dump);
+
+/*
  *	Notifier list for kernel code which wants to be called
  *	at shutdown. This is used to stop any idling DMA operations
  *	and the like. 
diff --git a/kernel/time.c b/kernel/time.c
index f04791f..9c81504 100644
--- a/kernel/time.c
+++ b/kernel/time.c
@@ -31,6 +31,7 @@ #include <linux/module.h>
 #include <linux/timex.h>
 #include <linux/capability.h>
 #include <linux/errno.h>
+#include <linux/notifier.h>
 #include <linux/syscalls.h>
 #include <linux/security.h>
 #include <linux/fs.h>
@@ -744,3 +745,34 @@ EXPORT_SYMBOL(get_jiffies_64);
 #endif

 EXPORT_SYMBOL(jiffies);
+
+/*
+ * Dump out uptime info on oops / panic.
+ *
+ * FIXME: Should I be just using get_jiffies64()? what if the lock there
+ * is damaged beyond any recognition?
+ */
+
+static int dump_uptime(struct notifier_block *self, unsigned long v, void *p)
+{
+	struct timespec uptime;
+	/* The logic below is very much like how kernel
+	 * prepares /proc/uptime.
+	 */
+	do_posix_clock_monotonic_gettime(&uptime);
+	printk(KERN_EMERG "Uptime (seconds): %lu.%02lu\n",
+		(unsigned long) uptime.tv_sec,
+		(uptime.tv_nsec / (NSEC_PER_SEC / 100)));
+
+	return 0;
+}
+
+static struct notifier_block uptime_notifier;
+
+static int __init register_uptime_dump(void) {
+	uptime_notifier.notifier_call = dump_uptime;
+	atomic_notifier_chain_register(&info_dumper_list, &uptime_notifier);
+	return 0;
+}
+__initcall(register_uptime_dump);
+
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/