On 3/1/23 8:59 PM, Jia Zhu wrote:
> In order to support mounting multi-blobs container image as a single
> block device, add flattened block device feature for EROFS.
>
> In this mode, all meta/data contents will be mapped into one block
> address. User could compose a block device(by nbd/ublk/virtio-blk/
> vhost-user-blk) from multiple sources and mount the block device by
> EROFS directly. It can reduce the number of block devices used, and
> it's also benefits in both VM file passthrough and distributed storage
> scenarios.
>
> You can test this using the method mentioned by:
> https://github.com/dragonflyoss/image-service/pull/1111
> 1. Compose a (nbd)block device from multi-blobs.
> 2. Mount EROFS on mntdir/.
> 3. Compare the md5sum between source dir and mntdir/.
>
> Later, we could also use it to refer original tar blobs.
>
> Signed-off-by: Jia Zhu <[email protected]>
> Signed-off-by: Xin Yin <[email protected]>
> ---
> v2:
> 1. Supplement commit message.
> 2. Add a bool field in erofs_dev_context to indicate flattened block
> device mode.
> ---
> fs/erofs/data.c | 8 ++++++--
> fs/erofs/internal.h | 1 +
> fs/erofs/super.c | 6 +++++-
> 3 files changed, 12 insertions(+), 3 deletions(-)
>
> diff --git a/fs/erofs/data.c b/fs/erofs/data.c
> index e16545849ea7..818f78ce648c 100644
> --- a/fs/erofs/data.c
> +++ b/fs/erofs/data.c
> @@ -197,7 +197,6 @@ int erofs_map_dev(struct super_block *sb, struct
> erofs_map_dev *map)
> struct erofs_device_info *dif;
> int id;
>
> - /* primary device by default */
> map->m_bdev = sb->s_bdev;
> map->m_daxdev = EROFS_SB(sb)->dax_dev;
> map->m_dax_part_off = EROFS_SB(sb)->dax_part_off;
> @@ -210,12 +209,17 @@ int erofs_map_dev(struct super_block *sb, struct
> erofs_map_dev *map)
> up_read(&devs->rwsem);
> return -ENODEV;
> }
> + if (devs->flatdev) {
> + map->m_pa += blknr_to_addr(dif->mapped_blkaddr);
> + up_read(&devs->rwsem);
> + return 0;
> + }
> map->m_bdev = dif->bdev;
> map->m_daxdev = dif->dax_dev;
> map->m_dax_part_off = dif->dax_part_off;
> map->m_fscache = dif->fscache;
> up_read(&devs->rwsem);
> - } else if (devs->extra_devices) {
> + } else if (devs->extra_devices && !devs->flatdev) {
> down_read(&devs->rwsem);
> idr_for_each_entry(&devs->tree, dif, id) {
> erofs_off_t startoff, length;
> diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
> index 3f3561d37d1b..4fee380a98d9 100644
> --- a/fs/erofs/internal.h
> +++ b/fs/erofs/internal.h
> @@ -81,6 +81,7 @@ struct erofs_dev_context {
> struct rw_semaphore rwsem;
>
> unsigned int extra_devices;
> + bool flatdev;
> };
>
> struct erofs_fs_context {
> diff --git a/fs/erofs/super.c b/fs/erofs/super.c
> index 19b1ae79cec4..307b3d2392cf 100644
> --- a/fs/erofs/super.c
> +++ b/fs/erofs/super.c
> @@ -248,7 +248,7 @@ static int erofs_init_device(struct erofs_buf *buf,
> struct super_block *sb,
> if (IS_ERR(fscache))
> return PTR_ERR(fscache);
> dif->fscache = fscache;
> - } else {
> + } else if (!sbi->devs->flatdev) {
> bdev = blkdev_get_by_path(dif->path, FMODE_READ | FMODE_EXCL,
> sb->s_type);
> if (IS_ERR(bdev))
> @@ -281,6 +281,10 @@ static int erofs_scan_devices(struct super_block *sb,
> else
> ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
>
> + if (!sbi->devs->extra_devices && ondisk_extradevs &&
> + !erofs_is_fscache_mode(sb))
> + sbi->devs->flatdev = true;
> +
I would move this check down after all sanity checks, e.g.
if (!ondisk_extradevs)
return 0;
+ if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
+ sbi->devs->flatdev = true;
Otherwise LGTM.
--
Thanks,
Jingbo
