superblock IO sometimes happens when memory might be tight,
and can lead to a deadlock if raid1d or raid5d initiate it.
With this patch, superblock IO happens without any memory
allocation.
----------- Diffstat output ------------
./drivers/md/md.c | 82 ++++++++++++++++++++++++--------------------
./include/linux/raid/md_k.h | 1
2 files changed, 47 insertions(+), 36 deletions(-)
diff ./drivers/md/md.c~current~ ./drivers/md/md.c
--- ./drivers/md/md.c~current~ 2002-12-16 16:29:19.000000000 +1100
+++ ./drivers/md/md.c 2002-12-16 16:29:27.000000000 +1100
@@ -445,21 +445,22 @@ static int alloc_disk_sb(mdk_rdev_t * rd
if (rdev->sb)
MD_BUG();
- rdev->sb = (mdp_super_t *) __get_free_page(GFP_KERNEL);
- if (!rdev->sb) {
+ rdev->sb_page = alloc_page(GFP_KERNEL);
+ if (!rdev->sb_page) {
printk(OUT_OF_MEM);
return -EINVAL;
}
- md_clear_page(rdev->sb);
+ rdev->sb = (mdp_super_t *) page_address(rdev->sb_page);
return 0;
}
static void free_disk_sb(mdk_rdev_t * rdev)
{
- if (rdev->sb) {
- free_page((unsigned long) rdev->sb);
+ if (rdev->sb_page) {
+ page_cache_release(rdev->sb_page);
rdev->sb = NULL;
+ rdev->sb_page = NULL;
rdev->sb_offset = 0;
rdev->size = 0;
} else {
@@ -468,12 +469,43 @@ static void free_disk_sb(mdk_rdev_t * rd
}
}
+
+static void bh_complete(struct buffer_head *bh, int uptodate)
+{
+
+ if (uptodate)
+ set_bit(BH_Uptodate, &bh->b_state);
+
+ complete((struct completion*)bh->b_private);
+}
+
+static int sync_page_io(kdev_t dev, unsigned long sector, int size,
+ struct page *page, int rw)
+{
+ struct buffer_head bh;
+ struct completion event;
+
+ init_completion(&event);
+ init_buffer(&bh, bh_complete, &event);
+ bh.b_rdev = dev;
+ bh.b_rsector = sector;
+ bh.b_state = (1 << BH_Req) | (1 << BH_Mapped);
+ bh.b_size = size;
+ bh.b_page = page;
+ bh.b_reqnext = NULL;
+ bh.b_data = page_address(page);
+ generic_make_request(rw, &bh);
+
+ run_task_queue(&tq_disk);
+ wait_for_completion(&event);
+
+ return test_bit(BH_Uptodate, &bh.b_state);
+}
+
static int read_disk_sb(mdk_rdev_t * rdev)
{
int ret = -EINVAL;
- struct buffer_head *bh = NULL;
kdev_t dev = rdev->dev;
- mdp_super_t *sb;
unsigned long sb_offset;
if (!rdev->sb) {
@@ -487,22 +519,14 @@ static int read_disk_sb(mdk_rdev_t * rde
*/
sb_offset = calc_dev_sboffset(rdev->dev, rdev->mddev, 1);
rdev->sb_offset = sb_offset;
- fsync_dev(dev);
- set_blocksize (dev, MD_SB_BYTES);
- bh = bread (dev, sb_offset / MD_SB_BLOCKS, MD_SB_BYTES);
-
- if (bh) {
- sb = (mdp_super_t *) bh->b_data;
- memcpy (rdev->sb, sb, MD_SB_BYTES);
- } else {
- printk(NO_SB,partition_name(rdev->dev));
- goto abort;
+
+ if (!sync_page_io(dev, sb_offset<<1, MD_SB_BYTES, rdev->sb_page, READ)) {
+ printk(NO_SB,partition_name(dev));
+ return -EINVAL;
}
printk(KERN_INFO " [events: %08lx]\n", (unsigned long)rdev->sb->events_lo);
ret = 0;
abort:
- if (bh)
- brelse (bh);
return ret;
}
@@ -890,10 +914,8 @@ static mdk_rdev_t * find_rdev_all(kdev_t
static int write_disk_sb(mdk_rdev_t * rdev)
{
- struct buffer_head *bh;
kdev_t dev;
unsigned long sb_offset, size;
- mdp_super_t *sb;
if (!rdev->sb) {
MD_BUG();
@@ -928,23 +950,11 @@ static int write_disk_sb(mdk_rdev_t * rd
}
printk(KERN_INFO "(write) %s's sb offset: %ld\n", partition_name(dev), sb_offset);
- fsync_dev(dev);
- set_blocksize(dev, MD_SB_BYTES);
- bh = getblk(dev, sb_offset / MD_SB_BLOCKS, MD_SB_BYTES);
- if (!bh) {
- printk(GETBLK_FAILED, partition_name(dev));
+
+ if (!sync_page_io(dev, sb_offset<<1, MD_SB_BYTES, rdev->sb_page, WRITE)) {
+ printk("md: write_disk_sb failed for device %s\n", partition_name(dev));
return 1;
}
- memset(bh->b_data,0,bh->b_size);
- sb = (mdp_super_t *) bh->b_data;
- memcpy(sb, rdev->sb, MD_SB_BYTES);
-
- mark_buffer_uptodate(bh, 1);
- mark_buffer_dirty(bh);
- ll_rw_block(WRITE, 1, &bh);
- wait_on_buffer(bh);
- brelse(bh);
- fsync_dev(dev);
skip:
return 0;
}
diff ./include/linux/raid/md_k.h~current~ ./include/linux/raid/md_k.h
--- ./include/linux/raid/md_k.h~current~ 2002-12-16 16:29:27.000000000 +1100
+++ ./include/linux/raid/md_k.h 2002-12-16 16:29:27.000000000 +1100
@@ -171,6 +171,7 @@ struct mdk_rdev_s
struct block_device *bdev; /* block device handle */
mdp_super_t *sb;
+ struct page *sb_page;
unsigned long sb_offset;
int alias_device; /* device alias to the same disk */
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
