Re: RFC: [patch] log fatal signals like SIGSEGV

[Marcin Slusarz]

On Thu, Sep 18, 2008 at 12:10:39PM +0200, Thomas Jarosch wrote:
> > It looks much better now. But I don't think it will go in as is.
> > Maybe you can disable it by default and create a sysctl switch?
> 
> Thinking about it some more, I've added a "signal-log-level" sysctl var
> including documentation. The patch applies nicely to 2.6.26 and 2.6 HEAD.
> 
> The idea is to default to log level 1 and log fatal signals only.
> Log output should be close to zero during normal system operation.
> 
> There is a bit of a naming clash with "print-fatal-signals", though that
> should be called "debug-fatal-signals" because of all the register dumps etc.
> I don't want to rename it as it would unnecessarily cause issues
> and it's debug-only (Documentation/kernel-parameters.txt) anyway.
> 
> Enjoy.
> 
> ------------------------------------------------------
> From: Thomas Jarosch <thomas.jarosch@xxxxxxxxxxxxx>
> 
> Log signals like SIGSEGV, SIGILL, SIGBUS or SIGFPE to aid tracing
> of obscure problems. Also logs the sender of the signal.
> 
> The log message looks like this:
> "kernel: signal 9 sent to freezed[2634] uid:100,
>  parent init[1] uid:0 by bash[3168] uid:0, parent sshd[3164] uid:0"
> 
> You can control the degree of logging via sysctl: "signal-log-level"
>     0 - Signal logging disabled
>     1 - Log SIGSEGV, SIGILL, SIGBUS and SIGPFE (default)
                                          ^^^^^^

>     2 - Log SIGKILL and SIGABRT and all signals from log level 1
>     3 or higher: Log all signals
> 
> The printing code is based on grsecurity's signal logger.
> 
> Signed-off-by: Thomas Jarosch <thomas.jarosch@xxxxxxxxxxxxx>
> Signed-off-by: Gerd v. Egidy <gve@xxxxxxxxxxxxx>
> 
> diff -u -r -p linux-2.6.26.vanilla/kernel/signal.c linux-2.6.26/kernel/signal.c
> --- linux-2.6.26.vanilla/kernel/signal.c	Tue Sep 16 13:45:34 2008
> +++ linux-2.6.26/kernel/signal.c	Thu Sep 18 10:43:27 2008
> @@ -796,6 +796,35 @@ static void complete_signal(int sig, str
>  	return;
>  }
>  
> +int signal_log_level __read_mostly = 1;
> +
> +static void log_signal(const int sig, const struct task_struct *t)
> +{
> +	bool log_signal = false;
> +
> +	if(signal_log_level >= 1 && (sig == SIGSEGV || sig == SIGILL
> +			|| sig == SIGBUS || sig == SIGFPE))
> +		log_signal = true;
> +	else if (signal_log_level >= 2 && (sig == SIGKILL || sig == SIGABRT))
> +		log_signal = true;
> +	else if (signal_log_level >= 3)
> +		log_signal = true;
> +
> +	if (!log_signal)
> +		return;
> +
> +	if (printk_ratelimit()) {
> +		/* Don't lock "tasklist_lock" here as it's already locked by "siglock" */
> +		printk(KERN_WARNING "signal %d sent to %.30s[%d] uid:%u, "
> +				"parent %.30s[%d] uid:%u by %.30s[%d] uid:%u, "
> +				"parent %.30s[%d] uid:%u\n", sig, t->comm,
> +				t->pid, t->uid, t->parent->comm, t->parent->pid,
> +				t->parent->uid, current->comm, current->pid,
> +				current->uid, current->parent->comm,
> +				current->parent->pid, current->parent->uid);
> +	}
> +}
> +
>  static inline int legacy_queue(struct sigpending *signals, int sig)
>  {
>  	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
> @@ -810,6 +839,8 @@ static int send_signal(int sig, struct s
>  	assert_spin_locked(&t->sighand->siglock);
>  	if (!prepare_signal(sig, t))
>  		return 0;
> +
> +	log_signal(sig, t);
>  
>  	pending = group ? &t->signal->shared_pending : &t->pending;
>  	/*
> diff -u -r -p linux-2.6.26.vanilla/kernel/sysctl.c linux-2.6.26/kernel/sysctl.c
> --- linux-2.6.26.vanilla/kernel/sysctl.c	Sun Jul 13 23:51:29 2008
> +++ linux-2.6.26/kernel/sysctl.c	Thu Sep 18 10:08:47 2008
> @@ -63,6 +63,7 @@ static int deprecated_sysctl_warning(str
>  /* External variables not in a header file. */
>  extern int C_A_D;
>  extern int print_fatal_signals;
> +extern int signal_log_level;
>  extern int sysctl_overcommit_memory;
>  extern int sysctl_overcommit_ratio;
>  extern int sysctl_panic_on_oom;
> @@ -398,6 +428,14 @@ static struct ctl_table kern_table[] = {
>  		.ctl_name	= CTL_UNNUMBERED,
>  		.procname	= "print-fatal-signals",
>  		.data		= &print_fatal_signals,
> +		.maxlen		= sizeof(int),
> +		.mode		= 0644,
> +		.proc_handler	= &proc_dointvec,
> +	},
> +	{
> +		.ctl_name	= CTL_UNNUMBERED,
> +		.procname	= "signal-log-level",
> +		.data		= &signal_log_level,
>  		.maxlen		= sizeof(int),
>  		.mode		= 0644,
>  		.proc_handler	= &proc_dointvec,
> diff -u -r linux-2.6.26.vanilla/Documentation/sysctl/kernel.txt linux-2.6.26/Documentation/sysctl/kernel.txt
> --- linux-2.6.26.vanilla/Documentation/sysctl/kernel.txt	Sun Jul 13 23:51:29 2008
> +++ linux-2.6.26/Documentation/sysctl/kernel.txt	Thu Sep 18 10:50:13 2008
> @@ -47,6 +47,7 @@
>  - rtsig-max
>  - rtsig-nr
>  - sem
> +- signal-log-level
>  - sg-big-buff                 [ generic SCSI device (sg) ]
>  - shmall
>  - shmmax                      [ sysv ipc ]
> @@ -349,6 +350,21 @@
>  
>  ==============================================================
>  
> +signal-log-level: 
> +
> +Brief logging of signal and sender to aid
> +tracing of obsucure problems later on.
              ^^^^^^^^

> +
> +  0 - Signal logging disabled
> +
> +  1 - Log SIGSEGV, SIGILL, SIGBUS and SIGPFE (default)
                                         ^^^^^^

> +
> +  2 - Log SIGKILL and SIGABRT and all signals from log level 1
> +  
> +  3 or higher: Log all signals
> +
> +==============================================================
> +
>  softlockup_thresh:
>  
>  This value can be used to lower the softlockup tolerance
> 
> 

When fix typos you can add:
Reviewed-by: Marcin Slusarz <marcin.slusarz@xxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/