On 2023/4/21 3:04, Al Viro wrote:
On Tue, Apr 18, 2023 at 10:05:12PM +0800, Jason Yan wrote:
On 2023/4/18 20:28, Kanchan Joshi wrote:
On Tue, Apr 18, 2023 at 11:00:12AM +0100, David Howells wrote:
Hi Christoph,
It seems that my use of i_size_read(file_inode(in)) in
filemap_splice_read()
to get the size of the file to be spliced from doesn't work in the
case of
blockdevs and it always returns 0.
What would be the best way to get the blockdev size? Look at
file->f_mapping->i_size maybe?
bdev_nr_bytes(I_BDEV(file->f_mapping->host))
should work I suppose.
This needs the caller to check if the file is a blockdev. Can we fill the
upper inode size which belongs to devtmpfs so that the generic code do not
need to aware the low level inode type?
We do:
static inline loff_t bdev_nr_bytes(struct block_device *bdev)
{
return (loff_t)bdev_nr_sectors(bdev) << SECTOR_SHIFT;
}
static inline sector_t bdev_nr_sectors(struct block_device *bdev)
{
return bdev->bd_nr_sectors;
}
$ git grep -n -w bd_nr_sectors
block/genhd.c:64: bdev->bd_nr_sectors = sectors;
block/partitions/core.c:92: bdev->bd_nr_sectors = sectors;
include/linux/blk_types.h:42: sector_t bd_nr_sectors;
include/linux/blkdev.h:785: return bdev->bd_nr_sectors;
and if you look into the functions with those assignments you'll see
void set_capacity(struct gendisk *disk, sector_t sectors)
{
struct block_device *bdev = disk->part0;
spin_lock(&bdev->bd_size_lock);
i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
bdev->bd_nr_sectors = sectors;
spin_unlock(&bdev->bd_size_lock);
}
and
static void bdev_set_nr_sectors(struct block_device *bdev, sector_t sectors)
{
spin_lock(&bdev->bd_size_lock);
i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT);
bdev->bd_nr_sectors = sectors;
spin_unlock(&bdev->bd_size_lock);
}
As you can see, both do i_size_write() on bdev->bd_inode with the value
equal to what bdev_nr_bytes() will return after the store to ->bd_nr_sectors.
Now, bdev->bd_inode always points to inode coallocated with bdev
(see bdev_alloc() for details) and that's what we have
->f_mapping->host pointing to for opened file after
blkdev_open() - see
filp->f_mapping = bdev->bd_inode->i_mapping;
in there combined with inode->i_mapping.host set to inode by
inode_init_always()<-alloc_inode()<-new_inode_pseudo()<-new_inode()<-bdev_alloc().
IOW, i_size_read(file->f_mapping->host) is correct answer for any kind of file.
Ah, yes. Thanks for your patient explanation.
Jason
