replace cad_pid by a struct pid
RFC PATCH 1/1 replace cad_pid by a struct pid
There are a few places in the kernel where the init task is signaled. The ctrl+alt+del sequence is one them. It kills a task, usually init, using a cached pid (cad_pid).
This patch replaces the pid_t by a struct pid to avoid pid wrap around problem. The struct pid is initialized at boot time in init() and can be modified through systctl with
/proc/sys/kernel/cad_pid
[ I haven't found any distro using it ? ]
It also introduces a small helper routine kill_cad_pid() which is used where it seemed ok to use cad_pid instead of pid 1.
--- arch/mips/sgi-ip22/ip22-reset.c | 2 +- arch/mips/sgi-ip32/ip32-reset.c | 2 +- arch/powerpc/platforms/iseries/mf.c | 2 +- drivers/char/nwbutton.c | 2 +- drivers/char/snsc_event.c | 2 +- drivers/parisc/power.c | 2 +- drivers/s390/s390mach.c | 2 +- include/linux/sched.h | 7 +++++++ init/main.c | 2 ++ kernel/sys.c | 5 ++--- kernel/sysctl.c | 30 +++++++++++++++++++++++++++--- 11 files changed, 45 insertions(+), 13 deletions(-)
kernel/sysctl.c +27 -3
@@ -65,7 +65,6 @@ extern int sysrq_enabled;extern int core_uses_pid;extern int suid_dumpable;extern char core_pattern[]; -extern int cad_pid;extern int pid_max;extern int min_free_kbytes;extern int printk_ratelimit_jiffies;
@@ -148,6 +147,9 @@ static int parse_table(int __user *, intstatic int proc_do_uts_string (ctl_table *table,int write,struct file *filp,void __user *buffer,size_t *lenp,loff_t *ppos); +static int proc_do_cad_pid (ctl_table *table,int write,struct file *filp, +void __user *buffer,size_t *lenp,loff_t *ppos); +static ctl_table root_table[];static struct ctl_table_header root_table_header = { root_table,LIST_HEAD_INIT (root_table_header.ctl_entry) };
@@ -562,10 +564,10 @@ static ctl_table kern_table[] = { { .ctl_name = KERN_CADPID, .procname = "cad_pid", - .data = &cad_pid, + .data = NULL, .maxlen = sizeof (int), .mode = 0600, - .proc_handler = &proc_dointvec, + .proc_handler = &proc_do_cad_pid, }, { .ctl_name = KERN_MAX_THREADS,
@@ -2478,6 +2480,28 @@ proc_minmax: }#endif +static int proc_do_cad_pid (ctl_table *table,int write,struct file *filp, +void __user *buffer,size_t *lenp,loff_t *ppos) +{ +struct pid *new_pid; +pid_t tmp; +int r; + + tmp =pid_nr (cad_pid); + + r =__do_proc_dointvec (&tmp, table, write, filp, buffer, + lenp, ppos,NULL ,NULL ); +if (r || !write) +return r; + + new_pid =find_get_pid (tmp); +if (!new_pid) +return -ESRCH ; + +put_pid (xchg (&cad_pid, new_pid)); +return 0 ; +} +#else /* CONFIG_PROC_FS */ int proc_dostring(ctl_table *table, int write, struct file *filp,
include/linux/sched.h +7 -0
@@ -1096,6 +1096,8 @@ static inline int is_init(struct task_streturn tsk->pid ==1 ; } +extern struct pid * cad_pid; +extern void free_task (struct task_struct *tsk);#define get_task_struct (tsk) do { atomic_inc(&(tsk)->usage); } while(0)
@@ -1323,6 +1325,11 @@ extern int send_group_sigqueue(int, struextern int do_sigaction (int ,struct k_sigaction *,struct k_sigaction *);extern int do_sigaltstack (const stack_t __user *,stack_t __user *,unsigned long ); +static inline int kill_cad_pid (int sig,int priv) +{ +return kill_pid (cad_pid, sig, priv); +} +/* These can be the second arg to send_sig_info/send_group_sig_info. */ #define SEND_SIG_NOINFO ((struct siginfo *) 0)#define SEND_SIG_PRIV ((struct siginfo *) 1)
init/main.c +2 -0
@@ -726,6 +726,8 @@ static int init(void * unused)*/ child_reaper = current; + cad_pid =task_pid (current); +smp_prepare_cpus (max_cpus);do_pre_smp_initcalls ();
kernel/sys.c +2 -3
@@ -92,7 +92,7 @@ EXPORT_SYMBOL(fs_overflowgid);*/ int C_A_D =1 ; -int cad_pid =1 ; +struct pid * cad_pid =NULL ;/* * Notifier list for kernel code which wants to be called
@@ -896,10 +896,9 @@ void ctrl_alt_del(void)if (C_A_D )schedule_work (&cad_work);else -kill_proc (cad_pid,SIGINT ,1 ); +kill_cad_pid (SIGINT ,1 ); } -/* * Unprivileged users may change the real gid to the effective gid * or vice versa. (BSD-style)
arch/mips/sgi-ip32/ip32-reset.c +1 -1
@@ -120,7 +120,7 @@ static inline void ip32_power_button(voiif (has_panicked)return ; -if (shuting_down ||kill_proc (1 ,SIGINT ,1 )) { +if (shuting_down ||kill_cad_pid (SIGINT ,1 )) {/* No init process or button pressed twice. */ ip32_machine_power_off (); }
arch/powerpc/platforms/iseries/mf.c +1 -1
@@ -357,7 +357,7 @@ static int dma_and_signal_ce_msg(char *c*/ static int shutdown (void ) { -int rc =kill_proc (1 ,SIGINT ,1 ); +int rc =kill_cad_pid (SIGINT ,1 );if (rc) { printk(KERN_ALERT "mf.c: SIGINT to init failed (%d), "
drivers/char/nwbutton.c +1 -1
@@ -128,7 +128,7 @@ static void button_sequence_finished (un {#ifdef CONFIG_NWBUTTON_REBOOT /* Reboot using button is enabled */ if (button_press_count == reboot_count) { -kill_proc (1 ,SIGINT ,1 );/* Ask init to reboot us */ +kill_cad_pid (SIGINT ,1 );/* Ask init to reboot us */ }#endif /* CONFIG_NWBUTTON_REBOOT */ button_consume_callbacks (button_press_count);
drivers/char/snsc_event.c +1 -1
@@ -220,7 +220,7 @@ scdrv_dispatch_event(char *event, int le " Sending SIGPWR to init...\n");/* give a SIGPWR signal to init proc */ -kill_proc (1 ,SIGPWR ,0 ); +kill_cad_pid (SIGPWR ,0 ); }else {/* print to system log */ printk ("%s|$(0x%x)%s\n" , severity, esp_code, desc);
drivers/parisc/power.c +1 -1
@@ -85,7 +85,7 @@static void deferred_poweroff (void *dummy) {extern int cad_pid;/* from kernel/sys.c */ -if (kill_proc (cad_pid,SIGINT ,1 )) { +if (kill_cad_pid (SIGINT ,1 )) {/* just in case killing init process failed */ machine_power_off (); }
drivers/s390/s390mach.c +1 -1
@@ -208,7 +208,7 @@ s390_handle_mcck(void)*/ __ctl_clear_bit (14 ,24 );/* Disable WARNING MCH */ if (xchg (&mchchk_wng_posted,1 ) ==0 ) - kill_proc(1, SIGPWR, 1); + kill_cad_pid(SIGPWR, 1); }#endif
arch/mips/sgi-ip22/ip22-reset.c +1 -1
@@ -123,7 +123,7 @@ static inline void power_button(void)if (machine_state &MACHINE_PANICED )return ; -if ((machine_state &MACHINE_SHUTTING_DOWN ) ||kill_proc (1 ,SIGINT ,1 )) { +if ((machine_state &MACHINE_SHUTTING_DOWN ) ||kill_cad_pid (SIGINT ,1 )) {/* No init process or button pressed twice. */ sgi_machine_power_off (); }
Quoted 9 lines · On Wed, 20 Sep 2006 11:38:35 +0200 Cedric Le Goater <[email protected]> wrote:
There are a few places in the kernel where the init task is signaled. The ctrl+alt+del sequence is one them. It kills a task, usually init, using a cached pid (
cad_pid).This patch replaces the
pid_tby a struct pid to avoid pid wrap around problem. The struct pid is initialized at boot time ininit()and can be modified through systctl with
/proc/sys/kernel/cad_pid
hm. Is there any sane scenario in which C-A-D would be directed to any other process?
What happens if/when the process which is identifier by /proc/sys/kernel/cad_pid exits? User error, I guess...
Quoted 1 line
+extern struct pid*
cad_pid;
Whitespace violation detected!
arch/mips/sgi-ip32/ip32-reset.c static inline void ip32_power_button(voi · line 123
Quoted 2 lines
arch/mips/sgi-ip32/ip32-reset.c +1
-1-if (shuting_down ||kill_proc (1 ,SIGINT ,1 )) { +if (shuting_down ||kill_cad_pid (SIGINT ,1 )) {
So your patch actually makes functional changes: lots of random init-signallers gain extra functionality: the process which they signal can now be configured via /proc/sys/kernel/cad_pid. But by default, things remain unchanged. Fair enough.
drivers/char/nwbutton.c
Quoted 2 lines
drivers/char/nwbutton.c +0
-0
This driver can be compiled as a module. I shall add the missing export...
(And I'll fix the parisc build too)
Hello,
Quoted 7 lines · Andrew Morton
Quoted 4 lines · On Wed, 20 Sep 2006 11:38:35 +0200 Cedric Le Goater <[email protected]> wrote:
There are a few places in the kernel where the init task is signaled. The ctrl+alt+del sequence is one them. It kills a task, usually init, using a cached pid (
cad_pid).
Quoted 8 lines
Quoted 5 lines
This patch replaces the
pid_tby a struct pid to avoid pid wrap around problem. The struct pid is initialized at boot time ininit()and can be modified through systctl with
/proc/sys/kernel/cad_pidhm. Is there any sane scenario in which C-A-D would be directed to any other process?
I don't know. I really looked hard to see where this was used and i didn't find any distro using it.
Quoted 2 lines
What happens if/when the process which is identifier by
/proc/sys/kernel/cad_pidexits? User error, I guess...
kill_pid() fails.
Quoted 11 lines
Quoted 1 line
+extern struct pid*
cad_pid;Whitespace violation detected!
arch/mips/sgi-ip32/ip32-reset.c static inline void ip32_power_button(voi · line 123
Quoted 2 lines
arch/mips/sgi-ip32/ip32-reset.c +1
-1-if (shuting_down ||kill_proc (1 ,SIGINT ,1 )) { +if (shuting_down ||kill_cad_pid (SIGINT ,1 )) {So your patch actually makes functional changes: lots of random init-signallers gain extra functionality: the process which they signal can now be configured via
/proc/sys/kernel/cad_pid. But by default, things remain unchanged. Fair enough.
yes.
Quoted 4 lines
drivers/char/nwbutton.c
Quoted 2 lines
drivers/char/nwbutton.c +0
-0This driver can be compiled as a module. I shall add the missing export...
oops thanks,
Quoted 1 line
(And I'll fix the parisc build too)
the only parisc i have still runs its native hpux 11.
do we still need /proc/sys/kernel/cad_pid ? If it is obsolete, i'd send a different patch to disable this feature with some warning in dmesg and the kill_proc(1) would stay as they are.
but may be some small distros for embedded linux use it.
thanks,
C.
Quoted 1 line · On Thu, 21 Sep 2006 15:21:23 +0200 Cedric Le Goater <[email protected]> wrote:
do we still need
/proc/sys/kernel/cad_pid?
Some searching indicates that it is used by anaconda, and by something called "rootskel".