Skip to contentSkip to conversation

UML - Send pointers instead of structures to I/O thread

PATCH In review Jeff Dike 2 participants 3 messages Apr 11, 2007 User-mode Linux (UML)

lore
Share & tools
lore

Jeff Dike

author
#0 lore raw

PATCH 2/4 UML - Send pointers instead of structures to I/O thread

Instead of writing entire structures between UML and the I/O thread, we send pointers. This cuts down on the amount of data being copied and possibly allows more requests to be pending between the two.

This requires that the requests be kmalloced and freed instead of living on the stack.

--

 arch/um/drivers/ubd_kern.c |   41 ++++++++++++++++++++++++++---------------
 1 file changed, 26 insertions(+), 15 deletions(-)
arch/um/drivers/ubd_kern.c +26 -15
Index: linux-2.6.21-mm/arch/um/drivers/ubd_kern.c
===================================================================
@@ -503,7 +503,7 @@ static struct list_head restart = LIST_H
 /* Called without dev->lock held, and only in interrupt context. */
 static void ubd_handler(void)
 {
-	struct io_thread_req req;
+	struct io_thread_req *req;
 	struct request *rq;
 	struct ubd *ubd;
 	struct list_head *list, *next_ele;
@@ -511,7 +511,8 @@ static void ubd_handler(void)
 	int n;
 
 	while(1){
-		n = os_read_file_k(thread_fd, &req, sizeof(req));
+		n = os_read_file_k(thread_fd, &req,
+				   sizeof(struct io_thread_req *));
 		if(n != sizeof(req)){
 			if(n == -EAGAIN)
 				break;
@@ -520,10 +521,11 @@ static void ubd_handler(void)
 			return;
 		}
 
-		rq = req.req;
-		rq->nr_sectors -= req.length >> 9;
+		rq = req->req;
+		rq->nr_sectors -= req->length >> 9;
 		if(rq->nr_sectors == 0)
 			ubd_finish(rq, rq->hard_nr_sectors << 9);
+		kfree(req);
 	}
 	reactivate_fd(thread_fd, UBD_IRQ);
 
@@ -1078,7 +1080,7 @@ static void prepare_request(struct reque
 /* Called with dev->lock held */
 static void do_ubd_request(request_queue_t *q)
 {
-	struct io_thread_req io_req;
+	struct io_thread_req *io_req;
 	struct request *req;
 	int n;
 
@@ -1099,13 +1101,20 @@ static void do_ubd_request(request_queue
 		while(dev->start_sg < dev->end_sg){
 			struct scatterlist *sg = &dev->sg[dev->start_sg];
 
-			prepare_request(req, &io_req,
+			io_req = kmalloc(sizeof(struct io_thread_req),
+					 GFP_KERNEL | GFP_ATOMIC);
+			if(io_req == NULL){
+				if(list_empty(&dev->restart))
+					list_add(&dev->restart, &restart);
+				return;
+			}
+			prepare_request(req, io_req,
 					(unsigned long long) req->sector << 9,
 					sg->offset, sg->length, sg->page);
 
-			n = os_write_file_k(thread_fd, (char *) &io_req,
-					    sizeof(io_req));
-			if(n != sizeof(io_req)){
+			n = os_write_file_k(thread_fd, &io_req,
+					    sizeof(struct io_thread_req *));
+			if(n != sizeof(struct io_thread_req *)){
 				if(n != -EAGAIN)
 					printk("write to io thread failed, "
 					       "errno = %d\n", -n);
@@ -1437,13 +1446,14 @@ static int io_count = 0;
 
 int io_thread(void *arg)
 {
-	struct io_thread_req req;
+	struct io_thread_req *req;
 	int n;
 
 	ignore_sigwinch_sig();
 	while(1){
-		n = os_read_file_k(kernel_fd, &req, sizeof(req));
-		if(n != sizeof(req)){
+		n = os_read_file_k(kernel_fd, &req,
+				 sizeof(struct io_thread_req *));
+		if(n != sizeof(struct io_thread_req *)){
 			if(n < 0)
 				printk("io_thread - read failed, fd = %d, "
 				       "err = %d\n", kernel_fd, -n);
@@ -1454,9 +1464,10 @@ int io_thread(void *arg)
 			continue;
 		}
 		io_count++;
-		do_io(&req);
-		n = os_write_file_k(kernel_fd, &req, sizeof(req));
-		if(n != sizeof(req))
+		do_io(req);
+		n = os_write_file_k(kernel_fd, &req,
+				  sizeof(struct io_thread_req *));
+		if(n != sizeof(struct io_thread_req *))
 			printk("io_thread - write failed, fd = %d, err = %d\n",
 			       kernel_fd, -n);
 	}

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
User-mode-linux-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

#1 lore raw
Quoted 3 lines · Jeff Dike

Instead of writing entire structures between UML and the I/O thread, we send pointers. This cuts down on the amount of data being copied and possibly allows more requests to be pending between the two.

<snip>

arch/um/drivers/ubd_kern.c static void do_ubd_request(request_queue · line 1117

Quoted 7 lines
arch/um/drivers/ubd_kern.c +3 -3
-                       n = os_write_file_k(thread_fd, (char *) &io_req,
-                                           sizeof(io_req));
-                       if(n != sizeof(io_req)){
+                       n = os_write_file_k(thread_fd, &io_req,
+                                           sizeof(struct io_thread_req *));
+                       if(n != sizeof(struct io_thread_req *)){

Is this wrong? you declare io_reg as a pointer but then you pass him to os_write_file_k by reference of a pointer? Should you just do this instead:

Review

Quoted 2 lines
(quoted diff) +2 -0
+                       n = os_write_file_k(thread_fd, io_req,
+                                           sizeof(struct io_thread_req *));

Was this intentional, or am I just an idiot. :)

-Jerry

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
User-mode-linux-devel mailing list [email protected]
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

author
#2 lore raw
Quoted 3 lines · Gerb Stralko

Is this wrong? you declare io_reg as a pointer but then you pass him to os_write_file_k by reference of a pointer? Should you just do this instead:

Intentional, I'm writing pointers, not structures, down the pipe.

Tx for the review, BTW.

Jeff

Signature · 11 lines
Work email - jdike at linux dot intel dot com

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
User-mode-linux-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel