Replace the fd_sets in struct fdtable with an array of unsigned longs and then
use the standard non-atomic bit operations rather than the FD_* macros.
This:
(1) Removes the abuses of struct fd_set:
(a) Since we don't want to allocate a full fd_set the vast majority of the
time, we actually, in effect, just allocate a just-big-enough array of
unsigned longs and cast it to an fd_set type - so why bother with the
fd_set at all?
(b) Some places outside of the core fdtable handling code (such as
SELinux) want to look inside the array of unsigned longs hidden inside
the fd_set struct for more efficient iteration over the entire set.
(2) Eliminates the use of FD_*() macros in the kernel completely.
(3) Permits the __FD_*() macros to be deleted entirely where not exposed to
userspace.
Signed-off-by: David Howells <dhowells@xxxxxxxxxx>
---
fs/exec.c | 4 ++--
fs/file.c | 46 ++++++++++++++++++++++------------------------
fs/select.c | 2 +-
include/linux/fdtable.h | 28 ++++++++++------------------
kernel/exit.c | 2 +-
security/selinux/hooks.c | 2 +-
6 files changed, 37 insertions(+), 47 deletions(-)
diff --git a/fs/exec.c b/fs/exec.c
index 22cc38d..cfd5e30 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -1026,10 +1026,10 @@ static void flush_old_files(struct files_struct * files)
fdt = files_fdtable(files);
if (i >= fdt->max_fds)
break;
- set = fdt->close_on_exec->fds_bits[j];
+ set = fdt->close_on_exec[j];
if (!set)
continue;
- fdt->close_on_exec->fds_bits[j] = 0;
+ fdt->close_on_exec[j] = 0;
spin_unlock(&files->file_lock);
for ( ; set ; i++,set >>= 1) {
if (set & 1) {
diff --git a/fs/file.c b/fs/file.c
index 114fea0..2d479dd 100644
--- a/fs/file.c
+++ b/fs/file.c
@@ -40,7 +40,7 @@ int sysctl_nr_open_max = 1024 * 1024; /* raised later */
*/
static DEFINE_PER_CPU(struct fdtable_defer, fdtable_defer_list);
-static void *alloc_fdmem(unsigned int size)
+static void *alloc_fdmem(size_t size)
{
/*
* Very large allocations can stress page reclaim, so fall back to
@@ -142,7 +142,7 @@ static void copy_fdtable(struct fdtable *nfdt, struct fdtable *ofdt)
static struct fdtable * alloc_fdtable(unsigned int nr)
{
struct fdtable *fdt;
- char *data;
+ void *data;
/*
* Figure out how many fds we actually want to support in this fdtable.
@@ -172,14 +172,15 @@ static struct fdtable * alloc_fdtable(unsigned int nr)
data = alloc_fdmem(nr * sizeof(struct file *));
if (!data)
goto out_fdt;
- fdt->fd = (struct file **)data;
- data = alloc_fdmem(max_t(unsigned int,
+ fdt->fd = data;
+
+ data = alloc_fdmem(max_t(size_t,
2 * nr / BITS_PER_BYTE, L1_CACHE_BYTES));
if (!data)
goto out_arr;
- fdt->open_fds = (fd_set *)data;
- data += nr / BITS_PER_BYTE;
- fdt->close_on_exec = (fd_set *)data;
+ fdt->open_fds = data;
+ data += nr / BITS_PER_LONG;
+ fdt->close_on_exec = data;
fdt->next = NULL;
return fdt;
@@ -275,11 +276,11 @@ static int count_open_files(struct fdtable *fdt)
int i;
/* Find the last open fd */
- for (i = size/(8*sizeof(long)); i > 0; ) {
- if (fdt->open_fds->fds_bits[--i])
+ for (i = size / BITS_PER_LONG; i > 0; ) {
+ if (fdt->open_fds[--i])
break;
}
- i = (i+1) * 8 * sizeof(long);
+ i = (i + 1) * BITS_PER_LONG;
return i;
}
@@ -306,8 +307,8 @@ struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
newf->next_fd = 0;
new_fdt = &newf->fdtab;
new_fdt->max_fds = NR_OPEN_DEFAULT;
- new_fdt->close_on_exec = (fd_set *)&newf->close_on_exec_init;
- new_fdt->open_fds = (fd_set *)&newf->open_fds_init;
+ new_fdt->close_on_exec = newf->close_on_exec_init;
+ new_fdt->open_fds = newf->open_fds_init;
new_fdt->fd = &newf->fd_array[0];
new_fdt->next = NULL;
@@ -350,10 +351,8 @@ struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
old_fds = old_fdt->fd;
new_fds = new_fdt->fd;
- memcpy(new_fdt->open_fds->fds_bits,
- old_fdt->open_fds->fds_bits, open_files/8);
- memcpy(new_fdt->close_on_exec->fds_bits,
- old_fdt->close_on_exec->fds_bits, open_files/8);
+ memcpy(new_fdt->open_fds, old_fdt->open_fds, open_files / 8);
+ memcpy(new_fdt->close_on_exec, old_fdt->close_on_exec, open_files / 8);
for (i = open_files; i != 0; i--) {
struct file *f = *old_fds++;
@@ -379,11 +378,11 @@ struct files_struct *dup_fd(struct files_struct *oldf, int *errorp)
memset(new_fds, 0, size);
if (new_fdt->max_fds > open_files) {
- int left = (new_fdt->max_fds-open_files)/8;
- int start = open_files / (8 * sizeof(unsigned long));
+ int left = (new_fdt->max_fds - open_files) / 8;
+ int start = open_files / BITS_PER_LONG;
- memset(&new_fdt->open_fds->fds_bits[start], 0, left);
- memset(&new_fdt->close_on_exec->fds_bits[start], 0, left);
+ memset(&new_fdt->open_fds[start], 0, left);
+ memset(&new_fdt->close_on_exec[start], 0, left);
}
rcu_assign_pointer(newf->fdt, new_fdt);
@@ -419,8 +418,8 @@ struct files_struct init_files = {
.fdtab = {
.max_fds = NR_OPEN_DEFAULT,
.fd = &init_files.fd_array[0],
- .close_on_exec = (fd_set *)&init_files.close_on_exec_init,
- .open_fds = (fd_set *)&init_files.open_fds_init,
+ .close_on_exec = init_files.close_on_exec_init,
+ .open_fds = init_files.open_fds_init,
},
.file_lock = __SPIN_LOCK_UNLOCKED(init_task.file_lock),
};
@@ -443,8 +442,7 @@ repeat:
fd = files->next_fd;
if (fd < fdt->max_fds)
- fd = find_next_zero_bit(fdt->open_fds->fds_bits,
- fdt->max_fds, fd);
+ fd = find_next_zero_bit(fdt->open_fds, fdt->max_fds, fd);
error = expand_files(files, fd);
if (error < 0)
diff --git a/fs/select.c b/fs/select.c
index d33418f..2e7fbe8 100644
--- a/fs/select.c
+++ b/fs/select.c
@@ -348,7 +348,7 @@ static int max_select_fd(unsigned long n, fd_set_bits *fds)
set = ~(~0UL << (n & (__NFDBITS-1)));
n /= __NFDBITS;
fdt = files_fdtable(current->files);
- open_fds = fdt->open_fds->fds_bits+n;
+ open_fds = fdt->open_fds + n;
max = 0;
if (set) {
set &= BITS(fds, n);
diff --git a/include/linux/fdtable.h b/include/linux/fdtable.h
index 7675da2..158a41e 100644
--- a/include/linux/fdtable.h
+++ b/include/linux/fdtable.h
@@ -21,51 +21,43 @@
*/
#define NR_OPEN_DEFAULT BITS_PER_LONG
-/*
- * The embedded_fd_set is a small fd_set,
- * suitable for most tasks (which open <= BITS_PER_LONG files)
- */
-struct embedded_fd_set {
- unsigned long fds_bits[1];
-};
-
struct fdtable {
unsigned int max_fds;
struct file __rcu **fd; /* current fd array */
- fd_set *close_on_exec;
- fd_set *open_fds;
+ unsigned long *close_on_exec;
+ unsigned long *open_fds;
struct rcu_head rcu;
struct fdtable *next;
};
static inline void __set_close_on_exec(int fd, struct fdtable *fdt)
{
- FD_SET(fd, fdt->close_on_exec);
+ __set_bit(fd, fdt->close_on_exec);
}
static inline void __clear_close_on_exec(int fd, struct fdtable *fdt)
{
- FD_CLR(fd, fdt->close_on_exec);
+ __clear_bit(fd, fdt->close_on_exec);
}
static inline bool close_on_exec(int fd, const struct fdtable *fdt)
{
- return FD_ISSET(fd, fdt->close_on_exec);
+ return test_bit(fd, fdt->close_on_exec);
}
static inline void __set_open_fd(int fd, struct fdtable *fdt)
{
- FD_SET(fd, fdt->open_fds);
+ __set_bit(fd, fdt->open_fds);
}
static inline void __clear_open_fd(int fd, struct fdtable *fdt)
{
- FD_CLR(fd, fdt->open_fds);
+ __clear_bit(fd, fdt->open_fds);
}
static inline bool fd_is_open(int fd, const struct fdtable *fdt)
{
- return FD_ISSET(fd, fdt->open_fds);
+ return test_bit(fd, fdt->open_fds);
}
/*
@@ -83,8 +75,8 @@ struct files_struct {
*/
spinlock_t file_lock ____cacheline_aligned_in_smp;
int next_fd;
- struct embedded_fd_set close_on_exec_init;
- struct embedded_fd_set open_fds_init;
+ unsigned long close_on_exec_init[1];
+ unsigned long open_fds_init[1];
struct file __rcu * fd_array[NR_OPEN_DEFAULT];
};
diff --git a/kernel/exit.c b/kernel/exit.c
index 4b4042f..4db0200 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -473,7 +473,7 @@ static void close_files(struct files_struct * files)
i = j * __NFDBITS;
if (i >= fdt->max_fds)
break;
- set = fdt->open_fds->fds_bits[j++];
+ set = fdt->open_fds[j++];
while (set) {
if (set & 1) {
struct file * file = xchg(&fdt->fd[i], NULL);
diff --git a/security/selinux/hooks.c b/security/selinux/hooks.c
index 6a3683e..421c990 100644
--- a/security/selinux/hooks.c
+++ b/security/selinux/hooks.c
@@ -2145,7 +2145,7 @@ static inline void flush_unauthorized_files(const struct cred *cred,
fdt = files_fdtable(files);
if (i >= fdt->max_fds)
break;
- set = fdt->open_fds->fds_bits[j];
+ set = fdt->open_fds[j];
if (!set)
continue;
spin_unlock(&files->file_lock);
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
