[PATCH] fdtable: Make fdarray and fdsets equal in size

Currently, each fdtable supports three dynamically-sized arrays of data: the
fdarray and two fdsets.  The code allows the number of fds supported by the
fdarray (fdtable->max_fds) to differ from the number of fds supported by each
of the fdsets (fdtable->max_fdset).

In practice, it is wasteful for these two sizes to differ: whenever we hit a
limit on the smaller-capacity structure, we will reallocate the entire fdtable
and all the dynamic arrays within it, so any delta in the memory used by the
larger-capacity structure will never be touched at all.

Rather than hogging this excess, we shouldn't even allocate it in the first
place, and keep the capacities of the fdarray and the fdsets equal.  This
patch removes fdtable->max_fdset.  As an added bonus, most of the supporting
code becomes simpler.

Signed-off-by: Vadim Lobanov <vlobanov@speakeasy.net>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dipankar Sarma <dipankar@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

6 files changed, 36 insertions, 50 deletions

diff --git a/fs/compat.c b/fs/compat.c
index b766964a625c..0ec70e3cee0a 100644
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -1679,19 +1679,19 @@ int compat_core_sys_select(int n, compat_ulong_t __user *inp,
 {
 	fd_set_bits fds;
 	char *bits;
-	int size, max_fdset, ret = -EINVAL;
+	int size, max_fds, ret = -EINVAL;
 	struct fdtable *fdt;
 
 	if (n < 0)
 		goto out_nofds;
 
-	/* max_fdset can increase, so grab it once to avoid race */
+	/* max_fds can increase, so grab it once to avoid race */
 	rcu_read_lock();
 	fdt = files_fdtable(current->files);
-	max_fdset = fdt->max_fdset;
+	max_fds = fdt->max_fds;
 	rcu_read_unlock();
-	if (n > max_fdset)
-		n = max_fdset;
+	if (n > max_fds)
+		n = max_fds;
 
 	/*
 	 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),
diff --git a/fs/exec.c b/fs/exec.c
index 12d8cd461b41..11fe93f7363c 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -783,7 +783,7 @@ static void flush_old_files(struct files_struct * files)
 		j++;
 		i = j * __NFDBITS;
 		fdt = files_fdtable(files);
-		if (i >= fdt->max_fds || i >= fdt->max_fdset)
+		if (i >= fdt->max_fds)
 			break;
 		set = fdt->close_on_exec->fds_bits[j];
 		if (!set)
diff --git a/fs/fcntl.c b/fs/fcntl.c
index 2bdaef35da54..8e382a5d51bd 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -77,10 +77,9 @@ repeat:
 		start = files->next_fd;
 
 	newfd = start;
-	if (start < fdt->max_fdset) {
+	if (start < fdt->max_fds)
 		newfd = find_next_zero_bit(fdt->open_fds->fds_bits,
-			fdt->max_fdset, start);
-	}
+					   fdt->max_fds, start);
 	
 	error = -EMFILE;
 	if (newfd >= current->signal->rlim[RLIMIT_NOFILE].rlim_cur)
diff --git a/fs/file.c b/fs/file.c
index 51aef675470f..fb3d2038dc21 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -68,8 +68,8 @@ void free_fd_array(struct file **array, int num)
 
 static void __free_fdtable(struct fdtable *fdt)
 {
-	free_fdset(fdt->open_fds, fdt->max_fdset);
-	free_fdset(fdt->close_on_exec, fdt->max_fdset);
+	free_fdset(fdt->open_fds, fdt->max_fds);
+	free_fdset(fdt->close_on_exec, fdt->max_fds);
 	free_fd_array(fdt->fd, fdt->max_fds);
 	kfree(fdt);
 }
@@ -98,7 +98,7 @@ static void free_fdtable_rcu(struct rcu_head *rcu)
 	struct fdtable_defer *fddef;
 
 	BUG_ON(!fdt);
-	fdset_size = fdt->max_fdset / 8;
+	fdset_size = fdt->max_fds / 8;
 	fdarray_size = fdt->max_fds * sizeof(struct file *);
 
 	if (fdt->free_files) {
@@ -110,13 +110,11 @@ static void free_fdtable_rcu(struct rcu_head *rcu)
 		kmem_cache_free(files_cachep, fdt->free_files);
 		return;
 	}
-	if (fdt->max_fdset <= EMBEDDED_FD_SET_SIZE &&
-		fdt->max_fds <= NR_OPEN_DEFAULT) {
+	if (fdt->max_fds <= NR_OPEN_DEFAULT)
 		/*
 		 * The fdtable was embedded
 		 */
 		return;
-	}
 	if (fdset_size <= PAGE_SIZE && fdarray_size <= PAGE_SIZE) {
 		kfree(fdt->open_fds);
 		kfree(fdt->close_on_exec);
@@ -136,9 +134,7 @@ static void free_fdtable_rcu(struct rcu_head *rcu)
 
 void free_fdtable(struct fdtable *fdt)
 {
-	if (fdt->free_files ||
-		fdt->max_fdset > EMBEDDED_FD_SET_SIZE ||
-		fdt->max_fds > NR_OPEN_DEFAULT)
+	if (fdt->free_files || fdt->max_fds > NR_OPEN_DEFAULT)
 		call_rcu(&fdt->rcu, free_fdtable_rcu);
 }
 
@@ -151,12 +147,11 @@ static void copy_fdtable(struct fdtable *nfdt, struct fdtable *fdt)
 	int i;
 	int count;
 
-	BUG_ON(nfdt->max_fdset < fdt->max_fdset);
 	BUG_ON(nfdt->max_fds < fdt->max_fds);
 	/* Copy the existing tables and install the new pointers */
 
-	i = fdt->max_fdset / (sizeof(unsigned long) * 8);
-	count = (nfdt->max_fdset - fdt->max_fdset) / 8;
+	i = fdt->max_fds / (sizeof(unsigned long) * 8);
+	count = (nfdt->max_fds - fdt->max_fds) / 8;
 
 	/*
 	 * Don't copy the entire array if the current fdset is
@@ -164,9 +159,9 @@ static void copy_fdtable(struct fdtable *nfdt, struct fdtable *fdt)
 	 */
 	if (i) {
 		memcpy (nfdt->open_fds, fdt->open_fds,
-						fdt->max_fdset/8);
+						fdt->max_fds/8);
 		memcpy (nfdt->close_on_exec, fdt->close_on_exec,
-						fdt->max_fdset/8);
+						fdt->max_fds/8);
 		memset (&nfdt->open_fds->fds_bits[i], 0, count);
 		memset (&nfdt->close_on_exec->fds_bits[i], 0, count);
 	}
@@ -201,7 +196,7 @@ fd_set * alloc_fdset(int num)
 
 void free_fdset(fd_set *array, int num)
 {
-	if (num <= EMBEDDED_FD_SET_SIZE) /* Don't free an embedded fdset */
+	if (num <= NR_OPEN_DEFAULT) /* Don't free an embedded fdset */
 		return;
 	else if (num <= 8 * PAGE_SIZE)
 		kfree(array);
@@ -220,18 +215,6 @@ static struct fdtable *alloc_fdtable(int nr)
 	if (!fdt)
   		goto out;
 
-	nfds = max_t(int, 8 * L1_CACHE_BYTES, roundup_pow_of_two(nr + 1));
-	if (nfds > NR_OPEN)
-		nfds = NR_OPEN;
-
-  	new_openset = alloc_fdset(nfds);
-  	new_execset = alloc_fdset(nfds);
-  	if (!new_openset || !new_execset)
-  		goto out;
-	fdt->open_fds = new_openset;
-	fdt->close_on_exec = new_execset;
-	fdt->max_fdset = nfds;
-
 	nfds = NR_OPEN_DEFAULT;
 	/*
 	 * Expand to the max in easy steps, and keep expanding it until
@@ -251,15 +234,21 @@ static struct fdtable *alloc_fdtable(int nr)
 				nfds = NR_OPEN;
   		}
 	} while (nfds <= nr);
+
+  	new_openset = alloc_fdset(nfds);
+  	new_execset = alloc_fdset(nfds);
+  	if (!new_openset || !new_execset)
+  		goto out;
+	fdt->open_fds = new_openset;
+	fdt->close_on_exec = new_execset;
+
 	new_fds = alloc_fd_array(nfds);
 	if (!new_fds)
-		goto out2;
+		goto out;
 	fdt->fd = new_fds;
 	fdt->max_fds = nfds;
 	fdt->free_files = NULL;
 	return fdt;
-out2:
-	nfds = fdt->max_fdset;
 out:
 	free_fdset(new_openset, nfds);
 	free_fdset(new_execset, nfds);
@@ -290,7 +279,7 @@ static int expand_fdtable(struct files_struct *files, int nr)
 	 * we dropped the lock
 	 */
 	cur_fdt = files_fdtable(files);
-	if (nr >= cur_fdt->max_fds || nr >= cur_fdt->max_fdset) {
+	if (nr >= cur_fdt->max_fds) {
 		/* Continue as planned */
 		copy_fdtable(new_fdt, cur_fdt);
 		rcu_assign_pointer(files->fdt, new_fdt);
@@ -316,11 +305,10 @@ int expand_files(struct files_struct *files, int nr)
 
 	fdt = files_fdtable(files);
 	/* Do we need to expand? */
-	if (nr < fdt->max_fdset && nr < fdt->max_fds)
+	if (nr < fdt->max_fds)
 		return 0;
 	/* Can we expand? */
-	if (fdt->max_fdset >= NR_OPEN || fdt->max_fds >= NR_OPEN ||
-	    nr >= NR_OPEN)
+	if (nr >= NR_OPEN)
 		return -EMFILE;
 
 	/* All good, so we try */
diff --git a/fs/open.c b/fs/open.c
index 0d94319e8681..c989fb4cf7b9 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -864,8 +864,7 @@ int get_unused_fd(void)
 
 repeat:
 	fdt = files_fdtable(files);
- 	fd = find_next_zero_bit(fdt->open_fds->fds_bits,
-				fdt->max_fdset,
+	fd = find_next_zero_bit(fdt->open_fds->fds_bits, fdt->max_fds,
 				files->next_fd);
 
 	/*
diff --git a/fs/select.c b/fs/select.c
index dcbc1112b7ec..fe0893afd931 100644
--- a/fs/select.c
+++ b/fs/select.c
@@ -311,7 +311,7 @@ static int core_sys_select(int n, fd_set __user *inp, fd_set __user *outp,
 {
 	fd_set_bits fds;
 	void *bits;
-	int ret, max_fdset;
+	int ret, max_fds;
 	unsigned int size;
 	struct fdtable *fdt;
 	/* Allocate small arguments on the stack to save memory and be faster */
@@ -321,13 +321,13 @@ static int core_sys_select(int n, fd_set __user *inp, fd_set __user *outp,
 	if (n < 0)
 		goto out_nofds;
 
-	/* max_fdset can increase, so grab it once to avoid race */
+	/* max_fds can increase, so grab it once to avoid race */
 	rcu_read_lock();
 	fdt = files_fdtable(current->files);
-	max_fdset = fdt->max_fdset;
+	max_fds = fdt->max_fds;
 	rcu_read_unlock();
-	if (n > max_fdset)
-		n = max_fdset;
+	if (n > max_fds)
+		n = max_fds;
 
 	/*
 	 * We need 6 bitmaps (in/out/ex for both incoming and outgoing),