#include <linux/highmem.h>
#include <linux/workqueue.h>
#include <linux/security.h>
+#include <linux/rcuref.h>
#include <asm/kmap_types.h>
#include <asm/uaccess.h>
#endif
/*------ sysctl variables----*/
-atomic_t aio_nr = ATOMIC_INIT(0); /* current system wide number of aio requests */
-unsigned aio_max_nr = 0x10000; /* system wide maximum number of aio requests */
+static DEFINE_SPINLOCK(aio_nr_lock);
+unsigned long aio_nr; /* current system wide number of aio requests */
+unsigned long aio_max_nr = 0x10000; /* system wide maximum number of aio requests */
/*----end sysctl variables---*/
static kmem_cache_t *kiocb_cachep;
return ERR_PTR(-EINVAL);
}
- if (nr_events > aio_max_nr)
+ if ((unsigned long)nr_events > aio_max_nr)
return ERR_PTR(-EAGAIN);
ctx = kmem_cache_alloc(kioctx_cachep, GFP_KERNEL);
goto out_freectx;
/* limit the number of system wide aios */
- atomic_add(ctx->max_reqs, &aio_nr); /* undone by __put_ioctx */
- if (unlikely(atomic_read(&aio_nr) > aio_max_nr))
+ spin_lock(&aio_nr_lock);
+ if (aio_nr + ctx->max_reqs > aio_max_nr ||
+ aio_nr + ctx->max_reqs < aio_nr)
+ ctx->max_reqs = 0;
+ else
+ aio_nr += ctx->max_reqs;
+ spin_unlock(&aio_nr_lock);
+ if (ctx->max_reqs == 0)
goto out_cleanup;
/* now link into global list. kludge. FIXME */
return ctx;
out_cleanup:
- atomic_sub(ctx->max_reqs, &aio_nr);
- ctx->max_reqs = 0; /* prevent __put_ioctx from sub'ing aio_nr */
__put_ioctx(ctx);
return ERR_PTR(-EAGAIN);
pr_debug("__put_ioctx: freeing %p\n", ctx);
kmem_cache_free(kioctx_cachep, ctx);
- atomic_sub(nr_events, &aio_nr);
+ if (nr_events) {
+ spin_lock(&aio_nr_lock);
+ BUG_ON(aio_nr - nr_events > aio_nr);
+ aio_nr -= nr_events;
+ spin_unlock(&aio_nr_lock);
+ }
}
/* aio_get_req
if (unlikely(!req))
return NULL;
- req->ki_flags = 1 << KIF_LOCKED;
+ req->ki_flags = 0;
req->ki_users = 2;
req->ki_key = 0;
req->ki_ctx = ctx;
/* Must be done under the lock to serialise against cancellation.
* Call this aio_fput as it duplicates fput via the fput_work.
*/
- if (unlikely(atomic_dec_and_test(&req->ki_filp->f_count))) {
+ if (unlikely(rcuref_dec_and_test(&req->ki_filp->f_count))) {
get_ioctx(ctx);
spin_lock(&fput_lock);
list_add(&req->ki_list, &fput_head);
ret = retry(iocb);
current->io_wait = NULL;
- if (-EIOCBRETRY != ret) {
- if (-EIOCBQUEUED != ret) {
- BUG_ON(!list_empty(&iocb->ki_wait.task_list));
- aio_complete(iocb, ret, 0);
- /* must not access the iocb after this */
- }
- } else {
- /*
- * Issue an additional retry to avoid waiting forever if
- * no waits were queued (e.g. in case of a short read).
- */
- if (list_empty(&iocb->ki_wait.task_list))
- kiocbSetKicked(iocb);
+ if (ret != -EIOCBRETRY && ret != -EIOCBQUEUED) {
+ BUG_ON(!list_empty(&iocb->ki_wait.task_list));
+ aio_complete(iocb, ret, 0);
}
out:
spin_lock_irq(&ctx->ctx_lock);
* and if required activate the aio work queue to process
* it
*/
-static void queue_kicked_iocb(struct kiocb *iocb)
+static void try_queue_kicked_iocb(struct kiocb *iocb)
{
struct kioctx *ctx = iocb->ki_ctx;
unsigned long flags;
int run = 0;
- WARN_ON((!list_empty(&iocb->ki_wait.task_list)));
+ /* We're supposed to be the only path putting the iocb back on the run
+ * list. If we find that the iocb is *back* on a wait queue already
+ * than retry has happened before we could queue the iocb. This also
+ * means that the retry could have completed and freed our iocb, no
+ * good. */
+ BUG_ON((!list_empty(&iocb->ki_wait.task_list)));
spin_lock_irqsave(&ctx->ctx_lock, flags);
- run = __queue_kicked_iocb(iocb);
+ /* set this inside the lock so that we can't race with aio_run_iocb()
+ * testing it and putting the iocb on the run list under the lock */
+ if (!kiocbTryKick(iocb))
+ run = __queue_kicked_iocb(iocb);
spin_unlock_irqrestore(&ctx->ctx_lock, flags);
if (run)
aio_queue_work(ctx);
return;
}
- /* If its already kicked we shouldn't queue it again */
- if (!kiocbTryKick(iocb)) {
- queue_kicked_iocb(iocb);
- }
+ try_queue_kicked_iocb(iocb);
}
EXPORT_SYMBOL(kick_iocb);
goto out;
ret = -EINVAL;
- if (unlikely(ctx || (int)nr_events <= 0)) {
- pr_debug("EINVAL: io_setup: ctx or nr_events > max\n");
+ if (unlikely(ctx || nr_events == 0)) {
+ pr_debug("EINVAL: io_setup: ctx %lu nr_events %u\n",
+ ctx, nr_events);
goto out;
}
}
/*
- * Default retry method for aio_read (also used for first time submit)
- * Responsible for updating iocb state as retries progress
+ * aio_p{read,write} are the default ki_retry methods for
+ * IO_CMD_P{READ,WRITE}. They maintains kiocb retry state around potentially
+ * multiple calls to f_op->aio_read(). They loop around partial progress
+ * instead of returning -EIOCBRETRY because they don't have the means to call
+ * kick_iocb().
*/
static ssize_t aio_pread(struct kiocb *iocb)
{
struct inode *inode = mapping->host;
ssize_t ret = 0;
- ret = file->f_op->aio_read(iocb, iocb->ki_buf,
- iocb->ki_left, iocb->ki_pos);
+ do {
+ ret = file->f_op->aio_read(iocb, iocb->ki_buf,
+ iocb->ki_left, iocb->ki_pos);
+ /*
+ * Can't just depend on iocb->ki_left to determine
+ * whether we are done. This may have been a short read.
+ */
+ if (ret > 0) {
+ iocb->ki_buf += ret;
+ iocb->ki_left -= ret;
+ }
- /*
- * Can't just depend on iocb->ki_left to determine
- * whether we are done. This may have been a short read.
- */
- if (ret > 0) {
- iocb->ki_buf += ret;
- iocb->ki_left -= ret;
/*
- * For pipes and sockets we return once we have
- * some data; for regular files we retry till we
- * complete the entire read or find that we can't
- * read any more data (e.g short reads).
+ * For pipes and sockets we return once we have some data; for
+ * regular files we retry till we complete the entire read or
+ * find that we can't read any more data (e.g short reads).
*/
- if (!S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode))
- ret = -EIOCBRETRY;
- }
+ } while (ret > 0 && iocb->ki_left > 0 &&
+ !S_ISFIFO(inode->i_mode) && !S_ISSOCK(inode->i_mode));
/* This means we must have transferred all that we could */
/* No need to retry anymore */
return ret;
}
-/*
- * Default retry method for aio_write (also used for first time submit)
- * Responsible for updating iocb state as retries progress
- */
+/* see aio_pread() */
static ssize_t aio_pwrite(struct kiocb *iocb)
{
struct file *file = iocb->ki_filp;
ssize_t ret = 0;
- ret = file->f_op->aio_write(iocb, iocb->ki_buf,
- iocb->ki_left, iocb->ki_pos);
-
- if (ret > 0) {
- iocb->ki_buf += ret;
- iocb->ki_left -= ret;
-
- ret = -EIOCBRETRY;
- }
+ do {
+ ret = file->f_op->aio_write(iocb, iocb->ki_buf,
+ iocb->ki_left, iocb->ki_pos);
+ if (ret > 0) {
+ iocb->ki_buf += ret;
+ iocb->ki_left -= ret;
+ }
+ } while (ret > 0 && iocb->ki_left > 0);
- /* This means we must have transferred all that we could */
- /* No need to retry anymore */
if ((ret == 0) || (iocb->ki_left == 0))
ret = iocb->ki_nbytes - iocb->ki_left;
if (unlikely(!access_ok(VERIFY_WRITE, kiocb->ki_buf,
kiocb->ki_left)))
break;
+ ret = security_file_permission(file, MAY_READ);
+ if (unlikely(ret))
+ break;
ret = -EINVAL;
if (file->f_op->aio_read)
kiocb->ki_retry = aio_pread;
if (unlikely(!access_ok(VERIFY_READ, kiocb->ki_buf,
kiocb->ki_left)))
break;
+ ret = security_file_permission(file, MAY_WRITE);
+ if (unlikely(ret))
+ break;
ret = -EINVAL;
if (file->f_op->aio_write)
kiocb->ki_retry = aio_pwrite;
goto out_put_req;
spin_lock_irq(&ctx->ctx_lock);
- if (likely(list_empty(&ctx->run_list))) {
- aio_run_iocb(req);
- } else {
- list_add_tail(&req->ki_run_list, &ctx->run_list);
+ aio_run_iocb(req);
+ if (!list_empty(&ctx->run_list)) {
/* drain the run list */
while (__aio_run_iocbs(ctx))
;