[PATCH] Define struct pspace
Sukadev Bhattiprolu [Mon, 2 Oct 2006 09:17:24 +0000 (02:17 -0700)]
Define a per-container pid space object.  And create one instance of this
object, init_pspace, to define the entire pid space.  Subsequent patches
will provide/use interfaces to create/destroy pid spaces.

Its a subset/rework of Eric Biederman's patch
http://lkml.org/lkml/2006/2/6/285 .

Signed-off-by: Eric Biederman <ebiederm@xmission.com>
Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Andrey Savochkin <saw@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/proc/proc_misc.c
include/linux/pspace.h
include/linux/sched.h
kernel/pid.c

index 66bc425..8d88e58 100644 (file)
@@ -45,6 +45,7 @@
 #include <linux/sysrq.h>
 #include <linux/vmalloc.h>
 #include <linux/crash_dump.h>
+#include <linux/pspace.h>
 #include <asm/uaccess.h>
 #include <asm/pgtable.h>
 #include <asm/io.h>
@@ -91,7 +92,7 @@ static int loadavg_read_proc(char *page, char **start, off_t off,
                LOAD_INT(a), LOAD_FRAC(a),
                LOAD_INT(b), LOAD_FRAC(b),
                LOAD_INT(c), LOAD_FRAC(c),
-               nr_running(), nr_threads, last_pid);
+               nr_running(), nr_threads, init_pspace.last_pid);
        return proc_calc_metrics(page, start, off, count, eof, len);
 }
 
index a8a064b..91d48b8 100644 (file)
@@ -13,4 +13,11 @@ struct pidmap {
 
 #define PIDMAP_ENTRIES         ((PID_MAX_LIMIT + 8*PAGE_SIZE - 1)/PAGE_SIZE/8)
 
+struct pspace {
+       struct pidmap pidmap[PIDMAP_ENTRIES];
+       int last_pid;
+};
+
+extern struct pspace init_pspace;
+
 #endif /* _LINUX_PSPACE_H */
index dd6c216..3b7c992 100644 (file)
@@ -118,7 +118,6 @@ extern unsigned long avenrun[];             /* Load averages */
 
 extern unsigned long total_forks;
 extern int nr_threads;
-extern int last_pid;
 DECLARE_PER_CPU(unsigned long, process_counts);
 extern int nr_processes(void);
 extern unsigned long nr_running(void);
index 8234bd0..89107b7 100644 (file)
@@ -34,7 +34,6 @@ static int pidhash_shift;
 static kmem_cache_t *pid_cachep;
 
 int pid_max = PID_MAX_DEFAULT;
-int last_pid;
 
 #define RESERVED_PIDS          300
 
@@ -43,7 +42,12 @@ int pid_max_max = PID_MAX_LIMIT;
 
 #define BITS_PER_PAGE          (PAGE_SIZE*8)
 #define BITS_PER_PAGE_MASK     (BITS_PER_PAGE-1)
-#define mk_pid(map, off)       (((map) - pidmap_array)*BITS_PER_PAGE + (off))
+
+static inline int mk_pid(struct pspace *pspace, struct pidmap *map, int off)
+{
+       return (map - pspace->pidmap)*BITS_PER_PAGE + off;
+}
+
 #define find_next_offset(map, off)                                     \
                find_next_zero_bit((map)->page, BITS_PER_PAGE, off)
 
@@ -53,8 +57,12 @@ int pid_max_max = PID_MAX_LIMIT;
  * value does not cause lots of bitmaps to be allocated, but
  * the scheme scales to up to 4 million PIDs, runtime.
  */
-static struct pidmap pidmap_array[PIDMAP_ENTRIES] =
-        { [ 0 ... PIDMAP_ENTRIES-1 ] = { ATOMIC_INIT(BITS_PER_PAGE), NULL } };
+struct pspace init_pspace = {
+       .pidmap = {
+               [ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
+       },
+       .last_pid = 0
+};
 
 /*
  * Note: disable interrupts while the pidmap_lock is held as an
@@ -69,40 +77,41 @@ static struct pidmap pidmap_array[PIDMAP_ENTRIES] =
  * irq handlers that take it we can leave the interrupts enabled.
  * For now it is easier to be safe than to prove it can't happen.
  */
+
 static  __cacheline_aligned_in_smp DEFINE_SPINLOCK(pidmap_lock);
 
-static fastcall void free_pidmap(int pid)
+static fastcall void free_pidmap(struct pspace *pspace, int pid)
 {
-       struct pidmap *map = pidmap_array + pid / BITS_PER_PAGE;
+       struct pidmap *map = pspace->pidmap + pid / BITS_PER_PAGE;
        int offset = pid & BITS_PER_PAGE_MASK;
 
        clear_bit(offset, map->page);
        atomic_inc(&map->nr_free);
 }
 
-static int alloc_pidmap(void)
+static int alloc_pidmap(struct pspace *pspace)
 {
-       int i, offset, max_scan, pid, last = last_pid;
+       int i, offset, max_scan, pid, last = pspace->last_pid;
        struct pidmap *map;
 
        pid = last + 1;
        if (pid >= pid_max)
                pid = RESERVED_PIDS;
        offset = pid & BITS_PER_PAGE_MASK;
-       map = &pidmap_array[pid/BITS_PER_PAGE];
+       map = &pspace->pidmap[pid/BITS_PER_PAGE];
        max_scan = (pid_max + BITS_PER_PAGE - 1)/BITS_PER_PAGE - !offset;
        for (i = 0; i <= max_scan; ++i) {
                if (unlikely(!map->page)) {
-                       unsigned long page = get_zeroed_page(GFP_KERNEL);
+                       void *page = kzalloc(PAGE_SIZE, GFP_KERNEL);
                        /*
                         * Free the page if someone raced with us
                         * installing it:
                         */
                        spin_lock_irq(&pidmap_lock);
                        if (map->page)
-                               free_page(page);
+                               kfree(page);
                        else
-                               map->page = (void *)page;
+                               map->page = page;
                        spin_unlock_irq(&pidmap_lock);
                        if (unlikely(!map->page))
                                break;
@@ -111,11 +120,11 @@ static int alloc_pidmap(void)
                        do {
                                if (!test_and_set_bit(offset, map->page)) {
                                        atomic_dec(&map->nr_free);
-                                       last_pid = pid;
+                                       pspace->last_pid = pid;
                                        return pid;
                                }
                                offset = find_next_offset(map, offset);
-                               pid = mk_pid(map, offset);
+                               pid = mk_pid(pspace, map, offset);
                        /*
                         * find_next_offset() found a bit, the pid from it
                         * is in-bounds, and if we fell back to the last
@@ -126,16 +135,16 @@ static int alloc_pidmap(void)
                                        (i != max_scan || pid < last ||
                                            !((last+1) & BITS_PER_PAGE_MASK)));
                }
-               if (map < &pidmap_array[(pid_max-1)/BITS_PER_PAGE]) {
+               if (map < &pspace->pidmap[(pid_max-1)/BITS_PER_PAGE]) {
                        ++map;
                        offset = 0;
                } else {
-                       map = &pidmap_array[0];
+                       map = &pspace->pidmap[0];
                        offset = RESERVED_PIDS;
                        if (unlikely(last == offset))
                                break;
                }
-               pid = mk_pid(map, offset);
+               pid = mk_pid(pspace, map, offset);
        }
        return -1;
 }
@@ -182,7 +191,7 @@ fastcall void free_pid(struct pid *pid)
        hlist_del_rcu(&pid->pid_chain);
        spin_unlock_irqrestore(&pidmap_lock, flags);
 
-       free_pidmap(pid->nr);
+       free_pidmap(&init_pspace, pid->nr);
        call_rcu(&pid->rcu, delayed_put_pid);
 }
 
@@ -196,7 +205,7 @@ struct pid *alloc_pid(void)
        if (!pid)
                goto out;
 
-       nr = alloc_pidmap();
+       nr = alloc_pidmap(&init_pspace);
        if (nr < 0)
                goto out_free;
 
@@ -363,10 +372,10 @@ void __init pidhash_init(void)
 
 void __init pidmap_init(void)
 {
-       pidmap_array->page = (void *)get_zeroed_page(GFP_KERNEL);
+       init_pspace.pidmap[0].page = kzalloc(PAGE_SIZE, GFP_KERNEL);
        /* Reserve PID 0. We never call free_pidmap(0) */
-       set_bit(0, pidmap_array->page);
-       atomic_dec(&pidmap_array->nr_free);
+       set_bit(0, init_pspace.pidmap[0].page);
+       atomic_dec(&init_pspace.pidmap[0].nr_free);
 
        pid_cachep = kmem_cache_create("pid", sizeof(struct pid),
                                        __alignof__(struct pid),