Commit 9eefe520 authored by Nadia Derbey's avatar Nadia Derbey Committed by Linus Torvalds

ipc: do not use a negative value to re-enable msgmni automatic recomputing

This patch proposes an alternative to the "magical
positive-versus-negative number trick" Andrew complained about last week
in http://lkml.org/lkml/2008/6/24/418.

This had been introduced with the patches that scale msgmni to the amount
of lowmem.  With these patches, msgmni has a registered notification
routine that recomputes msgmni value upon memory add/remove or ipc
namespace creation/ removal.

When msgmni is changed from user space (i.e.  value written to the proc
file), that notification routine is unregistered, and the way to make it
registered back is to write a negative value into the proc file.  This is
the "magical positive-versus-negative number trick".

To fix this, a new proc file is introduced: /proc/sys/kernel/auto_msgmni.
This file acts as ON/OFF for msgmni automatic recomputing.

With this patch, the process is the following:
1) kernel boots in "automatic recomputing mode"
   /proc/sys/kernel/msgmni contains the value that has been computed (depends
                           on lowmem)
   /proc/sys/kernel/automatic_msgmni contains "1"

2) echo <val> > /proc/sys/kernel/msgmni
   . sets msg_ctlmni to <val>
   . de-activates automatic recomputing (i.e. if, say, some memory is added
     msgmni won't be recomputed anymore)
   . /proc/sys/kernel/automatic_msgmni now contains "0"

3) echo "0" > /proc/sys/kernel/automatic_msgmni
   . de-activates msgmni automatic recomputing
     this has the same effect as 2) except that msg_ctlmni's value stays
     blocked at its current value)

3) echo "1" > /proc/sys/kernel/automatic_msgmni
   . recomputes msgmni's value based on the current available memory size
     and number of ipc namespaces
   . re-activates automatic recomputing for msgmni.
Signed-off-by: default avatarNadia Derbey <Nadia.Derbey@bull.net>
Cc: Solofo Ramangalahy <Solofo.Ramangalahy@bull.net>
Signed-off-by: default avatarAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: default avatarLinus Torvalds <torvalds@linux-foundation.org>
parent f1a43f93
...@@ -36,6 +36,7 @@ struct ipc_namespace { ...@@ -36,6 +36,7 @@ struct ipc_namespace {
int msg_ctlmni; int msg_ctlmni;
atomic_t msg_bytes; atomic_t msg_bytes;
atomic_t msg_hdrs; atomic_t msg_hdrs;
int auto_msgmni;
size_t shm_ctlmax; size_t shm_ctlmax;
size_t shm_ctlall; size_t shm_ctlall;
...@@ -53,7 +54,7 @@ extern atomic_t nr_ipc_ns; ...@@ -53,7 +54,7 @@ extern atomic_t nr_ipc_ns;
extern int register_ipcns_notifier(struct ipc_namespace *); extern int register_ipcns_notifier(struct ipc_namespace *);
extern int cond_register_ipcns_notifier(struct ipc_namespace *); extern int cond_register_ipcns_notifier(struct ipc_namespace *);
extern int unregister_ipcns_notifier(struct ipc_namespace *); extern void unregister_ipcns_notifier(struct ipc_namespace *);
extern int ipcns_notify(unsigned long); extern int ipcns_notify(unsigned long);
#else /* CONFIG_SYSVIPC */ #else /* CONFIG_SYSVIPC */
......
...@@ -27,15 +27,17 @@ static void *get_ipc(ctl_table *table) ...@@ -27,15 +27,17 @@ static void *get_ipc(ctl_table *table)
} }
/* /*
* Routine that is called when a tunable has successfully been changed by * Routine that is called when the file "auto_msgmni" has successfully been
* hand and it has a callback routine registered on the ipc namespace notifier * written.
* chain: we don't want such tunables to be recomputed anymore upon memory * Two values are allowed:
* add/remove or ipc namespace creation/removal. * 0: unregister msgmni's callback routine from the ipc namespace notifier
* They can come back to a recomputable state by being set to a <0 value. * chain. This means that msgmni won't be recomputed anymore upon memory
* add/remove or ipc namespace creation/removal.
* 1: register back the callback routine.
*/ */
static void tunable_set_callback(int val) static void ipc_auto_callback(int val)
{ {
if (val >= 0) if (!val)
unregister_ipcns_notifier(current->nsproxy->ipc_ns); unregister_ipcns_notifier(current->nsproxy->ipc_ns);
else { else {
/* /*
...@@ -71,7 +73,12 @@ static int proc_ipc_callback_dointvec(ctl_table *table, int write, ...@@ -71,7 +73,12 @@ static int proc_ipc_callback_dointvec(ctl_table *table, int write,
rc = proc_dointvec(&ipc_table, write, filp, buffer, lenp, ppos); rc = proc_dointvec(&ipc_table, write, filp, buffer, lenp, ppos);
if (write && !rc && lenp_bef == *lenp) if (write && !rc && lenp_bef == *lenp)
tunable_set_callback(*((int *)(ipc_table.data))); /*
* Tunable has successfully been changed by hand. Disable its
* automatic adjustment. This simply requires unregistering
* the notifiers that trigger recalculation.
*/
unregister_ipcns_notifier(current->nsproxy->ipc_ns);
return rc; return rc;
} }
...@@ -87,10 +94,39 @@ static int proc_ipc_doulongvec_minmax(ctl_table *table, int write, ...@@ -87,10 +94,39 @@ static int proc_ipc_doulongvec_minmax(ctl_table *table, int write,
lenp, ppos); lenp, ppos);
} }
static int proc_ipcauto_dointvec_minmax(ctl_table *table, int write,
struct file *filp, void __user *buffer, size_t *lenp, loff_t *ppos)
{
struct ctl_table ipc_table;
size_t lenp_bef = *lenp;
int oldval;
int rc;
memcpy(&ipc_table, table, sizeof(ipc_table));
ipc_table.data = get_ipc(table);
oldval = *((int *)(ipc_table.data));
rc = proc_dointvec_minmax(&ipc_table, write, filp, buffer, lenp, ppos);
if (write && !rc && lenp_bef == *lenp) {
int newval = *((int *)(ipc_table.data));
/*
* The file "auto_msgmni" has correctly been set.
* React by (un)registering the corresponding tunable, if the
* value has changed.
*/
if (newval != oldval)
ipc_auto_callback(newval);
}
return rc;
}
#else #else
#define proc_ipc_doulongvec_minmax NULL #define proc_ipc_doulongvec_minmax NULL
#define proc_ipc_dointvec NULL #define proc_ipc_dointvec NULL
#define proc_ipc_callback_dointvec NULL #define proc_ipc_callback_dointvec NULL
#define proc_ipcauto_dointvec_minmax NULL
#endif #endif
#ifdef CONFIG_SYSCTL_SYSCALL #ifdef CONFIG_SYSCTL_SYSCALL
...@@ -142,14 +178,11 @@ static int sysctl_ipc_registered_data(ctl_table *table, int __user *name, ...@@ -142,14 +178,11 @@ static int sysctl_ipc_registered_data(ctl_table *table, int __user *name,
rc = sysctl_ipc_data(table, name, nlen, oldval, oldlenp, newval, rc = sysctl_ipc_data(table, name, nlen, oldval, oldlenp, newval,
newlen); newlen);
if (newval && newlen && rc > 0) { if (newval && newlen && rc > 0)
/* /*
* Tunable has successfully been changed from userland * Tunable has successfully been changed from userland
*/ */
int *data = get_ipc(table); unregister_ipcns_notifier(current->nsproxy->ipc_ns);
tunable_set_callback(*data);
}
return rc; return rc;
} }
...@@ -158,6 +191,9 @@ static int sysctl_ipc_registered_data(ctl_table *table, int __user *name, ...@@ -158,6 +191,9 @@ static int sysctl_ipc_registered_data(ctl_table *table, int __user *name,
#define sysctl_ipc_registered_data NULL #define sysctl_ipc_registered_data NULL
#endif #endif
static int zero;
static int one = 1;
static struct ctl_table ipc_kern_table[] = { static struct ctl_table ipc_kern_table[] = {
{ {
.ctl_name = KERN_SHMMAX, .ctl_name = KERN_SHMMAX,
...@@ -222,6 +258,16 @@ static struct ctl_table ipc_kern_table[] = { ...@@ -222,6 +258,16 @@ static struct ctl_table ipc_kern_table[] = {
.proc_handler = proc_ipc_dointvec, .proc_handler = proc_ipc_dointvec,
.strategy = sysctl_ipc_data, .strategy = sysctl_ipc_data,
}, },
{
.ctl_name = CTL_UNNUMBERED,
.procname = "auto_msgmni",
.data = &init_ipc_ns.auto_msgmni,
.maxlen = sizeof(int),
.mode = 0644,
.proc_handler = proc_ipcauto_dointvec_minmax,
.extra1 = &zero,
.extra2 = &one,
},
{} {}
}; };
......
...@@ -55,25 +55,35 @@ static int ipcns_callback(struct notifier_block *self, ...@@ -55,25 +55,35 @@ static int ipcns_callback(struct notifier_block *self,
int register_ipcns_notifier(struct ipc_namespace *ns) int register_ipcns_notifier(struct ipc_namespace *ns)
{ {
int rc;
memset(&ns->ipcns_nb, 0, sizeof(ns->ipcns_nb)); memset(&ns->ipcns_nb, 0, sizeof(ns->ipcns_nb));
ns->ipcns_nb.notifier_call = ipcns_callback; ns->ipcns_nb.notifier_call = ipcns_callback;
ns->ipcns_nb.priority = IPCNS_CALLBACK_PRI; ns->ipcns_nb.priority = IPCNS_CALLBACK_PRI;
return blocking_notifier_chain_register(&ipcns_chain, &ns->ipcns_nb); rc = blocking_notifier_chain_register(&ipcns_chain, &ns->ipcns_nb);
if (!rc)
ns->auto_msgmni = 1;
return rc;
} }
int cond_register_ipcns_notifier(struct ipc_namespace *ns) int cond_register_ipcns_notifier(struct ipc_namespace *ns)
{ {
int rc;
memset(&ns->ipcns_nb, 0, sizeof(ns->ipcns_nb)); memset(&ns->ipcns_nb, 0, sizeof(ns->ipcns_nb));
ns->ipcns_nb.notifier_call = ipcns_callback; ns->ipcns_nb.notifier_call = ipcns_callback;
ns->ipcns_nb.priority = IPCNS_CALLBACK_PRI; ns->ipcns_nb.priority = IPCNS_CALLBACK_PRI;
return blocking_notifier_chain_cond_register(&ipcns_chain, rc = blocking_notifier_chain_cond_register(&ipcns_chain,
&ns->ipcns_nb); &ns->ipcns_nb);
if (!rc)
ns->auto_msgmni = 1;
return rc;
} }
int unregister_ipcns_notifier(struct ipc_namespace *ns) void unregister_ipcns_notifier(struct ipc_namespace *ns)
{ {
return blocking_notifier_chain_unregister(&ipcns_chain, blocking_notifier_chain_unregister(&ipcns_chain, &ns->ipcns_nb);
&ns->ipcns_nb); ns->auto_msgmni = 0;
} }
int ipcns_notify(unsigned long val) int ipcns_notify(unsigned long val)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment