glusterfs: discard (hole punch) support

Add support for the DISCARD file operation. Discard punches a hole
in a file in the provided range. Block de-allocation is implemented
via fallocate() (as requested via fuse and passed on to the brick
fs) but a separate fop is created within gluster to emphasize the
fact that discard changes file data (the discarded region is
replaced with zeroes) and must invalidate caches where appropriate.

BUG: 963678
Change-Id: I34633a0bfff2187afeab4292a15f3cc9adf261af
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-on: http://review.gluster.org/5090
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

5 files changed, 190 insertions, 0 deletions

diff --git a/xlators/performance/io-cache/src/io-cache.c b/xlators/performance/io-cache/src/io-cache.c
index 30dc14a9cdb..d0667e46944 100644
--- a/xlators/performance/io-cache/src/io-cache.c
+++ b/xlators/performance/io-cache/src/io-cache.c
@@ -1422,6 +1422,33 @@ ioc_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
         return 0;
 }
 
+static int32_t
+ioc_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		int32_t op_ret, int32_t op_errno, struct iatt *pre,
+		struct iatt *post, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, pre, post, xdata);
+	return 0;
+}
+
+static int32_t
+ioc_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	    size_t len, dict_t *xdata)
+{
+	uint64_t ioc_inode = 0;
+
+	inode_ctx_get (fd->inode, this, &ioc_inode);
+
+	if (ioc_inode)
+		ioc_inode_flush ((ioc_inode_t *)(long)ioc_inode);
+
+	STACK_WIND(frame, ioc_discard_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+       return 0;
+}
+
+
+
 int32_t
 ioc_get_priority_list (const char *opt_str, struct list_head *first)
 {
@@ -2044,6 +2071,7 @@ struct xlator_fops fops = {
         .mknod       = ioc_mknod,
 
         .readdirp    = ioc_readdirp,
+	.discard     = ioc_discard,
 };
 
 
diff --git a/xlators/performance/io-threads/src/io-threads.c b/xlators/performance/io-threads/src/io-threads.c
index 33642bffb25..c6bdc375439 100644
--- a/xlators/performance/io-threads/src/io-threads.c
+++ b/xlators/performance/io-threads/src/io-threads.c
@@ -308,6 +308,7 @@ iot_schedule (call_frame_t *frame, xlator_t *this, call_stub_t *stub)
         case GF_FOP_FXATTROP:
         case GF_FOP_RCHECKSUM:
 	case GF_FOP_FALLOCATE:
+	case GF_FOP_DISCARD:
                 pri = IOT_PRI_LO;
                 break;
 
@@ -2459,6 +2460,56 @@ out:
 }
 
 int
+iot_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno,
+                  struct iatt *preop, struct iatt *postop, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, preop, postop,
+                             xdata);
+        return 0;
+}
+
+
+int
+iot_discard_wrapper(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+		    size_t len, dict_t *xdata)
+{
+        STACK_WIND (frame, iot_discard_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->discard, fd, offset, len, xdata);
+        return 0;
+}
+
+
+int
+iot_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	    size_t len, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+        int              ret = -1;
+
+        stub = fop_discard_stub(frame, iot_discard_wrapper, fd, offset, len,
+				xdata);
+        if (!stub) {
+                gf_log (this->name, GF_LOG_ERROR, "cannot create discard stub"
+                        "(out of memory)");
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        ret = iot_schedule (frame, this, stub);
+
+out:
+        if (ret < 0) {
+                STACK_UNWIND_STRICT (discard, frame, -1, -ret, NULL, NULL,
+                                     NULL);
+                if (stub != NULL) {
+                        call_stub_destroy (stub);
+                }
+        }
+        return 0;
+}
+
+int
 __iot_workers_scale (iot_conf_t *conf)
 {
         int       scale = 0;
@@ -2788,6 +2839,7 @@ struct xlator_fops fops = {
 	.fxattrop    = iot_fxattrop,
         .rchecksum   = iot_rchecksum,
 	.fallocate   = iot_fallocate,
+	.discard     = iot_discard,
 };
 
 struct xlator_cbks cbks;
diff --git a/xlators/performance/md-cache/src/md-cache.c b/xlators/performance/md-cache/src/md-cache.c
index 0b2c94319fb..2f52cbe5807 100644
--- a/xlators/performance/md-cache/src/md-cache.c
+++ b/xlators/performance/md-cache/src/md-cache.c
@@ -1889,6 +1889,45 @@ int mdc_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
 }
 
 int
+mdc_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno,
+                struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        mdc_local_t  *local = NULL;
+
+        local = frame->local;
+
+        if (op_ret != 0)
+                goto out;
+
+        if (!local)
+                goto out;
+
+        mdc_inode_iatt_set_validate(this, local->fd->inode, prebuf, postbuf);
+
+out:
+        MDC_STACK_UNWIND(discard, frame, op_ret, op_errno, prebuf, postbuf,
+                         xdata);
+
+        return 0;
+}
+
+int mdc_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+		size_t len, dict_t *xdata)
+{
+	mdc_local_t *local;
+
+	local = mdc_local_get(frame);
+	local->fd = fd_ref(fd);
+
+	STACK_WIND(frame, mdc_discard_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->discard, fd, offset, len,
+		   xdata);
+
+	return 0;
+}
+
+int
 mdc_forget (xlator_t *this, inode_t *inode)
 {
         mdc_inode_wipe (this, inode);
@@ -2017,6 +2056,7 @@ struct xlator_fops fops = {
 	.readdirp    = mdc_readdirp,
 	.readdir     = mdc_readdir,
 	.fallocate   = mdc_fallocate,
+	.discard     = mdc_discard,
 };
 
 
diff --git a/xlators/performance/open-behind/src/open-behind.c b/xlators/performance/open-behind/src/open-behind.c
index 5ac813132eb..26b684e888c 100644
--- a/xlators/performance/open-behind/src/open-behind.c
+++ b/xlators/performance/open-behind/src/open-behind.c
@@ -701,6 +701,25 @@ err:
 }
 
 int
+ob_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	   size_t len, dict_t *xdata)
+{
+	call_stub_t *stub;
+
+	stub = fop_discard_stub(frame, default_discard_resume, fd, offset, len,
+				xdata);
+	if (!stub)
+		goto err;
+
+	open_and_resume(this, fd, stub);
+
+	return 0;
+err:
+	STACK_UNWIND_STRICT(discard, frame, -1, ENOMEM, NULL, NULL, NULL);
+	return 0;
+}
+
+int
 ob_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflags,
 	   dict_t *xdata)
 {
@@ -922,6 +941,7 @@ struct xlator_fops fops = {
 	.fxattrop    = ob_fxattrop,
 	.fsetattr    = ob_fsetattr,
 	.fallocate   = ob_fallocate,
+	.discard     = ob_discard,
 	.unlink      = ob_unlink,
 	.rename      = ob_rename,
 	.lk          = ob_lk,
diff --git a/xlators/performance/read-ahead/src/read-ahead.c b/xlators/performance/read-ahead/src/read-ahead.c
index 522fa52b8e9..241fa477fda 100644
--- a/xlators/performance/read-ahead/src/read-ahead.c
+++ b/xlators/performance/read-ahead/src/read-ahead.c
@@ -942,6 +942,55 @@ unwind:
         return 0;
 }
 
+int
+ra_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+               struct iatt *postbuf, dict_t *xdata)
+{
+        GF_ASSERT (frame);
+
+        STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+        return 0;
+}
+
+static int
+ra_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	     size_t len, dict_t *xdata)
+{
+        ra_file_t *file    = NULL;
+        fd_t      *iter_fd = NULL;
+        inode_t   *inode   = NULL;
+        uint64_t  tmp_file = 0;
+        int32_t   op_errno = EINVAL;
+
+        GF_ASSERT (frame);
+        GF_VALIDATE_OR_GOTO (frame->this->name, this, unwind);
+        GF_VALIDATE_OR_GOTO (frame->this->name, fd, unwind);
+
+        inode = fd->inode;
+
+        LOCK (&inode->lock);
+        {
+                list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+                        fd_ctx_get (iter_fd, this, &tmp_file);
+                        file = (ra_file_t *)(long)tmp_file;
+                        if (!file)
+                                continue;
+
+                        flush_region(frame, file, offset, len, 1);
+                }
+        }
+        UNLOCK (&inode->lock);
+
+        STACK_WIND (frame, ra_discard_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->discard, fd, offset, len, xdata);
+        return 0;
+
+unwind:
+        STACK_UNWIND_STRICT (discard, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
 
 int
 ra_priv_dump (xlator_t *this)
@@ -1123,6 +1172,7 @@ struct xlator_fops fops = {
         .truncate    = ra_truncate,
         .ftruncate   = ra_ftruncate,
         .fstat       = ra_fstat,
+	.discard     = ra_discard,
 };
 
 struct xlator_cbks cbks = {