Re: [PATCH -next] block: remove field 'bd_inode' from block_device

[Yu Kuai <yukuai1@xxxxxxxxxxxxxxx>]

Hi,

在 2023/11/25 22:32, Greg KH 写道:
> On Sat, Nov 25, 2023 at 05:39:12PM +0800, Yu Kuai wrote:
> > From: Yu Kuai <yukuai3@xxxxxxxxxx>
> >
> > block_devcie is allocated from bdev_alloc() by bdev_alloc_inode(), and
> > currently block_device contains a pointer that point to the address of
> > inode, while such inode is allocated together:
> >
> > bdev_alloc
> >   inode = new_inode()
> >    // inode is &bdev_inode->vfs_inode
> >   bdev = I_BDEV(inode)
> >    // bdev is &bdev_inode->bdev
> >   bdev->inode = inode
> >
> > Add a new helper to get address of inode from bdev by add operation
> > instead of memory access, which is more efficiency. Also prepare to
> > add a new field 'bd_flags' in the first cacheline(64 bytes).
> >
> > Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>
> > ---
> >   block/bdev.c                       | 39 +++++++++++++++++-------------
> >   block/blk-zoned.c                  |  4 +--
> >   block/fops.c                       |  4 +--
> >   block/genhd.c                      |  8 +++---
> >   block/ioctl.c                      |  8 +++---
> >   block/partitions/core.c            |  9 ++++---
> >   drivers/block/xen-blkback/xenbus.c |  2 +-
> >   drivers/md/bcache/super.c          |  2 +-
> >   drivers/mtd/devices/block2mtd.c    | 12 ++++-----
> >   drivers/s390/block/dasd_ioctl.c    |  2 +-
> >   drivers/scsi/scsicam.c             |  2 +-
> >   fs/bcachefs/util.h                 |  2 +-
> >   fs/btrfs/disk-io.c                 |  6 ++---
> >   fs/btrfs/volumes.c                 |  4 +--
> >   fs/btrfs/zoned.c                   |  2 +-
> >   fs/buffer.c                        |  8 +++---
> >   fs/cramfs/inode.c                  |  2 +-
> >   fs/erofs/data.c                    |  2 +-
> >   fs/ext4/dir.c                      |  2 +-
> >   fs/ext4/ext4_jbd2.c                |  2 +-
> >   fs/ext4/super.c                    |  8 +++---
> >   fs/gfs2/glock.c                    |  2 +-
> >   fs/gfs2/ops_fstype.c               |  2 +-
> >   fs/jbd2/journal.c                  |  3 ++-
> >   fs/jbd2/recovery.c                 |  2 +-
> >   fs/nilfs2/segment.c                |  2 +-
> >   include/linux/blk_types.h          | 10 ++++++--
> >   include/linux/blkdev.h             |  4 +--
> >   include/linux/buffer_head.h        |  4 +--
> >   29 files changed, 86 insertions(+), 73 deletions(-)
> You should do this as a patch series, add the helper function that does
> nothing, convert all the different portions of the kernel as different
> patches, and _then_ change the implementation of the block layer to
> handle the change in the structure.
>
> Otherwise this is going to be hard to get accepted.
Okay, thanks for the adivce, I'll do that in v2.

By the way, I was thinking that this patch is quite simple, and doesn't
worth spliting into 10+ patches,
> Also, one note:
>
> > @@ -85,6 +84,13 @@ struct block_device {
> >   #define bdev_kobj(_bdev) \
> >         (&((_bdev)->bd_device.kobj))
> >   
> > +static inline struct inode *bdev_inode(struct block_device *bdev)
> > +{
> > +       void *inode = bdev + 1;
> That's crazy, if something changes, this will keep working yet the
> kernel will break and no one will know why.
>
> Please use container_of(), that's what it is there for, this exact type
> of thing.  Or if not, are you just assuming that the memory location
> right after bdev is the inode?  That's a tough assumption, how are you
> going to assure it really stays there?
Struct bdev_inode never changes since commit 8fbd544cbca5 ("[PATCH]
bdev: add I_BDEV()") from 2004, and I think it won't change unless
there is a different way to manage lifetime of block_device.

And the 'bdev + 1' is copied from blk_mq_rq_to_pdu(), however, I aggre
that use container_of() is better and I will use it in v2.

Thanks,
Kuai

> thanks,
>
> greg k-h
> .