|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-changelog] Allocate a request-id space per vbd in blkfront.
# HG changeset patch
# User kaf24@xxxxxxxxxxxxxxxxxxxx
# Node ID db61a0f346a831811ca5a820cec77a4c1748e1ac
# Parent 98de1d5fe5fbeedec536f3ed1618c6234e5186d1
Allocate a request-id space per vbd in blkfront.
This should fix the assertions people have been seeing
in domU blkfront where we run out of IDs (because they
were being shared by multiple vbds).
Signed-off-by: Keir Fraser <keir@xxxxxxxxxxxxx>
diff -r 98de1d5fe5fb -r db61a0f346a8
linux-2.6-xen-sparse/drivers/xen/blkfront/blkfront.c
--- a/linux-2.6-xen-sparse/drivers/xen/blkfront/blkfront.c Mon Aug 29
05:50:04 2005
+++ b/linux-2.6-xen-sparse/drivers/xen/blkfront/blkfront.c Mon Aug 29
13:56:53 2005
@@ -55,39 +55,32 @@
static unsigned int blkif_state = BLKIF_STATE_DISCONNECTED;
-#define BLK_RING_SIZE __RING_SIZE((blkif_sring_t *)0, PAGE_SIZE)
-
#define MAXIMUM_OUTSTANDING_BLOCK_REQS \
(BLKIF_MAX_SEGMENTS_PER_REQUEST * BLKIF_RING_SIZE)
#define GRANTREF_INVALID (1<<15)
-static struct blk_shadow {
- blkif_request_t req;
- unsigned long request;
- unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
-} blk_shadow[BLK_RING_SIZE];
-unsigned long blk_shadow_free;
-
static int recovery = 0; /* Recovery in progress: protected by blkif_io_lock */
static void kick_pending_request_queues(struct blkfront_info *info);
static void blkif_completion(struct blk_shadow *s);
-static inline int GET_ID_FROM_FREELIST(void)
-{
- unsigned long free = blk_shadow_free;
+static inline int GET_ID_FROM_FREELIST(
+ struct blkfront_info *info)
+{
+ unsigned long free = info->shadow_free;
BUG_ON(free > BLK_RING_SIZE);
- blk_shadow_free = blk_shadow[free].req.id;
- blk_shadow[free].req.id = 0x0fffffee; /* debug */
+ info->shadow_free = info->shadow[free].req.id;
+ info->shadow[free].req.id = 0x0fffffee; /* debug */
return free;
}
-static inline void ADD_ID_TO_FREELIST(unsigned long id)
-{
- blk_shadow[id].req.id = blk_shadow_free;
- blk_shadow[id].request = 0;
- blk_shadow_free = id;
+static inline void ADD_ID_TO_FREELIST(
+ struct blkfront_info *info, unsigned long id)
+{
+ info->shadow[id].req.id = info->shadow_free;
+ info->shadow[id].request = 0;
+ info->shadow_free = id;
}
static inline void pickle_request(struct blk_shadow *s, blkif_request_t *r)
@@ -213,8 +206,8 @@
/* Fill out a communications ring structure. */
ring_req = RING_GET_REQUEST(&info->ring, info->ring.req_prod_pvt);
- id = GET_ID_FROM_FREELIST();
- blk_shadow[id].request = (unsigned long)req;
+ id = GET_ID_FROM_FREELIST(info);
+ info->shadow[id].request = (unsigned long)req;
ring_req->id = id;
ring_req->operation = rq_data_dir(req) ?
@@ -240,7 +233,7 @@
buffer_ma >> PAGE_SHIFT,
rq_data_dir(req) );
- blk_shadow[id].frame[ring_req->nr_segments] =
+ info->shadow[id].frame[ring_req->nr_segments] =
buffer_ma >> PAGE_SHIFT;
ring_req->frame_and_sects[ring_req->nr_segments] =
@@ -253,7 +246,7 @@
info->ring.req_prod_pvt++;
/* Keep a private copy so we can reissue requests when recovering. */
- pickle_request(&blk_shadow[id], ring_req);
+ pickle_request(&info->shadow[id], ring_req);
gnttab_free_grant_references(gref_head);
@@ -331,11 +324,11 @@
bret = RING_GET_RESPONSE(&info->ring, i);
id = bret->id;
- req = (struct request *)blk_shadow[id].request;
-
- blkif_completion(&blk_shadow[id]);
-
- ADD_ID_TO_FREELIST(id);
+ req = (struct request *)info->shadow[id].request;
+
+ blkif_completion(&info->shadow[id]);
+
+ ADD_ID_TO_FREELIST(info, id);
switch (bret->operation) {
case BLKIF_OP_READ:
@@ -387,16 +380,16 @@
int j;
/* Stage 1: Make a safe copy of the shadow state. */
- copy = (struct blk_shadow *)kmalloc(sizeof(blk_shadow), GFP_KERNEL);
+ copy = (struct blk_shadow *)kmalloc(sizeof(info->shadow), GFP_KERNEL);
BUG_ON(copy == NULL);
- memcpy(copy, blk_shadow, sizeof(blk_shadow));
+ memcpy(copy, info->shadow, sizeof(info->shadow));
/* Stage 2: Set up free list. */
- memset(&blk_shadow, 0, sizeof(blk_shadow));
+ memset(&info->shadow, 0, sizeof(info->shadow));
for (i = 0; i < BLK_RING_SIZE; i++)
- blk_shadow[i].req.id = i+1;
- blk_shadow_free = info->ring.req_prod_pvt;
- blk_shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
+ info->shadow[i].req.id = i+1;
+ info->shadow_free = info->ring.req_prod_pvt;
+ info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
/* Stage 3: Find pending requests and requeue them. */
for (i = 0; i < BLK_RING_SIZE; i++) {
@@ -410,8 +403,8 @@
unpickle_request(req, ©[i]);
/* We get a new request id, and must reset the shadow state. */
- req->id = GET_ID_FROM_FREELIST();
- memcpy(&blk_shadow[req->id], ©[i], sizeof(copy[i]));
+ req->id = GET_ID_FROM_FREELIST(info);
+ memcpy(&info->shadow[req->id], ©[i], sizeof(copy[i]));
/* Rewrite any grant references invalidated by susp/resume. */
for (j = 0; j < req->nr_segments; j++) {
@@ -420,13 +413,13 @@
blkif_gref_from_fas(
req->frame_and_sects[j]),
info->backend_id,
- blk_shadow[req->id].frame[j],
+ info->shadow[req->id].frame[j],
rq_data_dir(
(struct request *)
- blk_shadow[req->id].request));
+ info->shadow[req->id].request));
req->frame_and_sects[j] &= ~GRANTREF_INVALID;
}
- blk_shadow[req->id].req = *req;
+ info->shadow[req->id].req = *req;
info->ring.req_prod_pvt++;
}
@@ -628,9 +621,8 @@
static int blkfront_probe(struct xenbus_device *dev,
const struct xenbus_device_id *id)
{
- int err;
+ int err, vdevice, i;
struct blkfront_info *info;
- int vdevice;
/* FIXME: Use dynamic device id if this is not set. */
err = xenbus_scanf(dev->nodename, "virtual-device", "%i", &vdevice);
@@ -651,6 +643,12 @@
info->connected = BLKIF_STATE_DISCONNECTED;
info->mi = NULL;
INIT_WORK(&info->work, blkif_restart_queue, (void *)info);
+
+ info->shadow_free = 0;
+ memset(info->shadow, 0, sizeof(info->shadow));
+ for (i = 0; i < BLK_RING_SIZE; i++)
+ info->shadow[i].req.id = i+1;
+ info->shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
/* Front end dir is a number, which is used as the id. */
info->handle = simple_strtoul(strrchr(dev->nodename,'/')+1, NULL, 0);
@@ -752,19 +750,11 @@
static int __init xlblk_init(void)
{
- int i;
-
if ((xen_start_info.flags & SIF_INITDOMAIN)
|| (xen_start_info.flags & SIF_BLK_BE_DOMAIN) )
return 0;
IPRINTK("Initialising virtual block device driver\n");
-
- blk_shadow_free = 0;
- memset(blk_shadow, 0, sizeof(blk_shadow));
- for (i = 0; i < BLK_RING_SIZE; i++)
- blk_shadow[i].req.id = i+1;
- blk_shadow[BLK_RING_SIZE-1].req.id = 0x0fffffff;
init_blk_xenbus();
diff -r 98de1d5fe5fb -r db61a0f346a8
linux-2.6-xen-sparse/drivers/xen/blkfront/block.h
--- a/linux-2.6-xen-sparse/drivers/xen/blkfront/block.h Mon Aug 29 05:50:04 2005
+++ b/linux-2.6-xen-sparse/drivers/xen/blkfront/block.h Mon Aug 29 13:56:53 2005
@@ -96,6 +96,14 @@
struct xlbd_type_info *type;
};
+struct blk_shadow {
+ blkif_request_t req;
+ unsigned long request;
+ unsigned long frame[BLKIF_MAX_SEGMENTS_PER_REQUEST];
+};
+
+#define BLK_RING_SIZE __RING_SIZE((blkif_sring_t *)0, PAGE_SIZE)
+
/*
* We have one of these per vbd, whether ide, scsi or 'other'. They
* hang in private_data off the gendisk structure. We may end up
@@ -119,6 +127,8 @@
request_queue_t *rq;
struct work_struct work;
struct gnttab_free_callback callback;
+ struct blk_shadow shadow[BLK_RING_SIZE];
+ unsigned long shadow_free;
};
extern spinlock_t blkif_io_lock;
_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |