libglusterfs: Implement API that provides page-aligned iobufs

One of the consumers of a page aligned buffer would be posix's
readv fop on O_DIRECT fds. Today the way it works is by getting
a page-aligned buffer through calloc, pread()ing into this buffer
and then copying its contents into a newly created iobuf's ptr.
This results in an extra memcpy() which can be avoided if we could
implement an api that would return an iobuf whose ptr is
page-aligned. That way the iobuf->ptr can be directly passed to
sys_pread() as a parameter by posix translator.

Change-Id: I385139bc6ee309fc501034b3af8f7814fab8cd65
BUG: 1343838
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: http://review.gluster.org/14672
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
Reviewed-by: Zhou Zhengping <johnzzpcrystal@gmail.com>
Reviewed-by: Jeff Darcy <jdarcy@redhat.com>

3 files changed, 50 insertions, 33 deletions

diff --git a/libglusterfs/src/iobuf.c b/libglusterfs/src/iobuf.c
index d1eb0acaf5e..17cd68fc206 100644
--- a/libglusterfs/src/iobuf.c
+++ b/libglusterfs/src/iobuf.c
@@ -679,6 +679,36 @@ unlock:
 }
 
 struct iobuf *
+iobuf_get_page_aligned (struct iobuf_pool *iobuf_pool, size_t page_size,
+                        size_t align_size)
+{
+        size_t              req_size     = 0;
+        struct iobuf       *iobuf        = NULL;
+
+        req_size = page_size;
+
+        if (req_size == 0) {
+                req_size = iobuf_pool->default_page_size;
+        }
+
+        iobuf = iobuf_get2 (iobuf_pool, req_size + align_size);
+        /* If std allocation was used, then free_ptr will be non-NULL. In this
+         * case, we do not want to modify the original free_ptr.
+         * On the other hand, if the buf was gotten through the available
+         * arenas, then we use iobuf->free_ptr to store the original
+         * pointer to the offset into the mmap'd block of memory and in turn
+         * reuse iobuf->ptr to hold the page-aligned address. And finally, in
+         * iobuf_put(), we copy iobuf->free_ptr into iobuf->ptr - back to where
+         * it was originally when __iobuf_get() returned this iobuf.
+         */
+        if (!iobuf->free_ptr)
+                iobuf->free_ptr = iobuf->ptr;
+        iobuf->ptr = GF_ALIGN_BUF (iobuf->ptr, align_size);
+
+        return iobuf;
+}
+
+struct iobuf *
 iobuf_get (struct iobuf_pool *iobuf_pool)
 {
         struct iobuf       *iobuf        = NULL;
@@ -745,6 +775,11 @@ __iobuf_put (struct iobuf *iobuf, struct iobuf_arena *iobuf_arena)
         list_del_init (&iobuf->list);
         iobuf_arena->active_cnt--;
 
+        if (iobuf->free_ptr) {
+                iobuf->ptr = iobuf->free_ptr;
+                iobuf->free_ptr = NULL;
+        }
+
         list_add (&iobuf->list, &iobuf_arena->passive.list);
         iobuf_arena->passive_cnt++;
 
diff --git a/libglusterfs/src/iobuf.h b/libglusterfs/src/iobuf.h
index 7e5cfe37a28..1578ceb84dd 100644
--- a/libglusterfs/src/iobuf.h
+++ b/libglusterfs/src/iobuf.h
@@ -169,4 +169,8 @@ void   iobuf_stats_dump (struct iobuf_pool *iobuf_pool);
 
 struct iobuf *
 iobuf_get2 (struct iobuf_pool *iobuf_pool, size_t page_size);
+
+struct iobuf *
+iobuf_get_page_aligned (struct iobuf_pool *iobuf_pool, size_t page_size,
+                        size_t align_size);
 #endif /* !_IOBUF_H_ */
diff --git a/xlators/storage/posix/src/posix.c b/xlators/storage/posix/src/posix.c
index deecb09a954..a362e633297 100644
--- a/xlators/storage/posix/src/posix.c
+++ b/xlators/storage/posix/src/posix.c
@@ -3043,8 +3043,6 @@ posix_readv (call_frame_t *frame, xlator_t *this,
         int32_t                op_ret     = -1;
         int32_t                op_errno   = 0;
         int                    _fd        = -1;
-        char                 *buf         = NULL;
-        char                 *alloc_buf   = NULL;
         struct posix_private * priv       = NULL;
         struct iobuf         * iobuf      = NULL;
         struct iobref        * iobref     = NULL;
@@ -3075,42 +3073,23 @@ posix_readv (call_frame_t *frame, xlator_t *this,
                 goto out;
         }
 
-        iobuf = iobuf_get2 (this->ctx->iobuf_pool, size);
+        iobuf = iobuf_get_page_aligned (this->ctx->iobuf_pool, size,
+                                        ALIGN_SIZE);
         if (!iobuf) {
                 op_errno = ENOMEM;
                 goto out;
         }
 
         _fd = pfd->fd;
-        if (pfd->flags & O_DIRECT) {
-                alloc_buf = _page_aligned_alloc (size, &buf);
-                if (!alloc_buf) {
-                        op_ret = -1;
-                        op_errno = errno;
-                        goto out;
-                }
-
-                op_ret = sys_pread (_fd, buf, size, offset);
-                if (op_ret == -1) {
-                        op_errno = errno;
-                        gf_msg (this->name, GF_LOG_ERROR, errno,
-                                P_MSG_READ_FAILED, "read failed on gfid=%s, "
-                                "fd=%p, offset=%"PRIu64" size=%"GF_PRI_SIZET", "
-                                "buf=%p", uuid_utoa (fd->inode->gfid), fd,
-                                offset, size, buf);
-                        goto out;
-                }
-                memcpy(iobuf->ptr, buf, size);
-        } else {
-                op_ret = sys_pread (_fd, iobuf->ptr, size, offset);
-                if (op_ret == -1) {
-                        op_errno = errno;
-                        gf_msg (this->name, GF_LOG_ERROR, errno, P_MSG_READ_FAILED,
-                                "read failed on gfid=%s, fd=%p, offset=%"PRIu64" "
-                                "size=%"GF_PRI_SIZET"", uuid_utoa (fd->inode->gfid), fd,
-                                offset, size);
-                        goto out;
-                }
+        op_ret = sys_pread (_fd, iobuf->ptr, size, offset);
+        if (op_ret == -1) {
+                op_errno = errno;
+                gf_msg (this->name, GF_LOG_ERROR, errno,
+                        P_MSG_READ_FAILED, "read failed on gfid=%s, "
+                        "fd=%p, offset=%"PRIu64" size=%"GF_PRI_SIZET", "
+                        "buf=%p", uuid_utoa (fd->inode->gfid), fd,
+                        offset, size, iobuf->ptr);
+                goto out;
         }
 
         LOCK (&priv->lock);
@@ -3153,7 +3132,6 @@ out:
                 iobref_unref (iobref);
         if (iobuf)
                 iobuf_unref (iobuf);
-        GF_FREE (alloc_buf);
 
         return 0;
 }