get rid of NR_OPEN and introduce a sysctl_nr_open

NR_OPEN (historically set to 1024*1024) actually forbids processes to open
more than 1024*1024 handles.

Unfortunatly some production servers hit the not so 'ridiculously high
value' of 1024*1024 file descriptors per process.

Changing NR_OPEN is not considered safe because of vmalloc space potential
exhaust.

This patch introduces a new sysctl (/proc/sys/fs/nr_open) wich defaults to
1024*1024, so that admins can decide to change this limit if their workload
needs it.

[akpm@linux-foundation.org: export it for sparc64]
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation/filesystems/proc.txt

@@ -1029,6 +1029,14 @@ nr_inodes
 Denotes the  number  of  inodes the system has allocated. This number will
 grow and shrink dynamically.
 
+nr_open
+-------
+
+Denotes the maximum number of file-handles a process can
+allocate. Default value is 1024*1024 (1048576) which should be
+enough for most machines. Actual limit depends on RLIMIT_NOFILE
+resource limit.
+
 nr_free_inodes
 --------------
 

Documentation/sysctl/fs.txt

@@ -23,6 +23,7 @@ Currently, these files are in /proc/sys/fs:
 - inode-max
 - inode-nr
 - inode-state
+- nr_open
 - overflowuid
 - overflowgid
 - suid_dumpable
@@ -91,6 +92,15 @@ usage of file handles and you don't need to increase the maximum.
 
 ==============================================================
 
+nr_open:
+
+This denotes the maximum number of file-handles a process can
+allocate. Default value is 1024*1024 (1048576) which should be
+enough for most machines. Actual limit depends on RLIMIT_NOFILE
+resource limit.
+
+==============================================================
+
 inode-max, inode-nr & inode-state:
 
 As with file handles, the kernel allocates the inode structures

arch/alpha/kernel/osf_sys.c

@@ -430,7 +430,7 @@ sys_getpagesize(void)
 asmlinkage unsigned long
 sys_getdtablesize(void)
 {
-	return NR_OPEN;
+	return sysctl_nr_open;
 }
 
 /*

arch/mips/kernel/sysirix.c

@@ -356,7 +356,7 @@ asmlinkage int irix_syssgi(struct pt_regs *regs)
 			retval = NGROUPS_MAX;
 			goto out;
 		case 5:
-			retval = NR_OPEN;
+			retval = sysctl_nr_open;
 			goto out;
 		case 6:
 			retval = 1;

arch/sparc64/kernel/sparc64_ksyms.c

@@ -277,6 +277,7 @@ EXPORT_SYMBOL(sys_getpid);
 EXPORT_SYMBOL(sys_geteuid);
 EXPORT_SYMBOL(sys_getuid);
 EXPORT_SYMBOL(sys_getegid);
+EXPORT_SYMBOL(sysctl_nr_open);
 EXPORT_SYMBOL(sys_getgid);
 EXPORT_SYMBOL(svr4_getcontext);
 EXPORT_SYMBOL(svr4_setcontext);

arch/sparc64/solaris/fs.c

@@ -624,7 +624,7 @@ asmlinkage int solaris_ulimit(int cmd, int val)
 	case 3: /* UL_GMEMLIM */
 		return current->signal->rlim[RLIMIT_DATA].rlim_cur;
 	case 4: /* UL_GDESLIM */
-		return NR_OPEN;
+		return sysctl_nr_open;
 	}
 	return -EINVAL;
 }

arch/sparc64/solaris/timod.c

@@ -859,7 +859,8 @@ asmlinkage int solaris_getmsg(unsigned int fd, u32 arg1, u32 arg2, u32 arg3)
 
 	SOLD("entry");
 	lock_kernel();
-	if(fd >= NR_OPEN) goto out;
+	if (fd >= sysctl_nr_open)
+		goto out;
 
 	fdt = files_fdtable(current->files);
 	filp = fdt->fd[fd];
@@ -927,7 +928,8 @@ asmlinkage int solaris_putmsg(unsigned int fd, u32 arg1, u32 arg2, u32 arg3)
 
 	SOLD("entry");
 	lock_kernel();
-	if(fd >= NR_OPEN) goto out;
+	if (fd >= sysctl_nr_open)
+		goto out;
 
 	fdt = files_fdtable(current->files);
 	filp = fdt->fd[fd];

fs/file.c

@@ -24,6 +24,8 @@ struct fdtable_defer {
 	struct fdtable *next;
 };
 
+int sysctl_nr_open __read_mostly = 1024*1024;
+
 /*
  * We use this list to defer free fdtables that have vmalloced
  * sets/arrays. By keeping a per-cpu list, we avoid having to embed
@@ -147,8 +149,8 @@ static struct fdtable * alloc_fdtable(unsigned int nr)
 	nr /= (1024 / sizeof(struct file *));
 	nr = roundup_pow_of_two(nr + 1);
 	nr *= (1024 / sizeof(struct file *));
-	if (nr > NR_OPEN)
-		nr = NR_OPEN;
+	if (nr > sysctl_nr_open)
+		nr = sysctl_nr_open;
 
 	fdt = kmalloc(sizeof(struct fdtable), GFP_KERNEL);
 	if (!fdt)
@@ -233,7 +235,7 @@ int expand_files(struct files_struct *files, int nr)
 	if (nr < fdt->max_fds)
 		return 0;
 	/* Can we expand? */
-	if (nr >= NR_OPEN)
+	if (nr >= sysctl_nr_open)
 		return -EMFILE;
 
 	/* All good, so we try */

include/linux/fs.h

@@ -21,7 +21,7 @@
 
 /* Fixed constants first: */
 #undef NR_OPEN
-#define NR_OPEN (1024*1024)	/* Absolute upper limit on fd num */
+extern int sysctl_nr_open;
 #define INR_OPEN 1024		/* Initial setting for nfile rlimits */
 
 #define BLOCK_SIZE_BITS 10

kernel/sys.c

@@ -1472,7 +1472,7 @@ asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
 	if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
 	    !capable(CAP_SYS_RESOURCE))
 		return -EPERM;
-	if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
+	if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > sysctl_nr_open)
 		return -EPERM;
 
 	retval = security_task_setrlimit(resource, &new_rlim);

kernel/sysctl.c

@@ -1202,6 +1202,14 @@ static struct ctl_table fs_table[] = {
 		.mode		= 0644,
 		.proc_handler	= &proc_dointvec,
 	},
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "nr_open",
+		.data		= &sysctl_nr_open,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+	},
 	{
 		.ctl_name	= FS_DENTRY,
 		.procname	= "dentry-state",