Re: super-block written got dislocation while 64K PAGE_SIZE enable.

[Zhengyuan Liu <liuzhengyuang521@xxxxxxxxx>]

To support arbitrary page size for superblock,include 4K/8K/64K currently, the off_set of SB_SECTOR should be a LCM(least common multiple) and that is 64. I redefine macro SB_SECTOR and BDEV_DATA_START_DEFAULT just as bellow patch showed:
        diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
        index ef567cd..1622f44 100644
        --- a/drivers/md/bcache/super.c
        +++ b/drivers/md/bcache/super.c
        @@ -74,7 +74,7 @@ static const char *read_super(struct cache_sb *sb, struct block_device *bdev,
         {
                const char *err;
                struct cache_sb *s;
        -       struct buffer_head *bh = __bread(bdev, 1, SB_SIZE);
        +       struct buffer_head *bh = __bread(bdev, 16, SB_SIZE);
                unsigned i;
 
                if (!bh)
        diff --git a/include/uapi/linux/bcache.h b/include/uapi/linux/bcache.h
        index 22b6ad3..eb948c8 100644
        --- a/include/uapi/linux/bcache.h
        +++ b/include/uapi/linux/bcache.h
        @@ -144,14 +144,14 @@ static inline struct bkey *bkey_idx(const struct bkey *k, unsigned nr_keys)
         #define BCACHE_SB_VERSION_BDEV_WITH_OFFSET 4
         #define BCACHE_SB_MAX_VERSION          4
 
        -#define SB_SECTOR                      8
        +#define SB_SECTOR                      128
         #define SB_SIZE                                4096
         #define SB_LABEL_SIZE                  32
         #define SB_JOURNAL_BUCKETS             256U
         /* SB_JOURNAL_BUCKETS must be divisible by BITS_PER_LONG */
         #define MAX_CACHES_PER_SET             8
 
        -#define BDEV_DATA_START_DEFAULT                16      /* sectors */
        +#define BDEV_DATA_START_DEFAULT                    256 /* sectors */
 
         struct cache_sb {
                __u64                   csum;

The second arg of __bread(bdev, 16, SB_SIZE)  has unit of block size and it need be relocated too.

To suitable this change, bcache-tools need be fixed up too. 

As for btree_node_size issues about page size, expecting somebody to commit a patch.

Kent, maybe you could use qemu to simulate ARM64 arch and there is always ubuntu image for ARM64 on the internet. 

2016-06-02 7:16 GMT+08:00 Kent Overstreet <kent.overstreet@xxxxxxxxx>:

On Wed, Jun 01, 2016 at 11:05:20AM -0700, Eric Wheeler wrote:
> On Wed, 1 Jun 2016, Zhengyuan Liu wrote:
>
> > Hi, I have created a mapped block device (bcach0) using make-bcache on
> > ARM64 server which has kernel enable 64K page size. However, the
> > bcach0 disappeared after the server reboot and there is no or dirty
> > metadata on super block of both cache device and back device . The
> > output of command  bcache-super-show was as bellow showed:
> >    [root@master Linux-4.4-LTS-storage]# bcache-super-show  /dev/sdb
> >    sb.magic bad magic
> >    Invalid superblock (bad magic)
> > /dev/sdb was the backing device and cache device got bad magic too.
> >
> > I tried to traced the written process of super block in bcache source
> > code and found that is the issue of PAGE_SIZE. It seems that the
> > bcache  was designed only considering for 4K PAGE_SIZE and it works
> > right only on 4K PAGE_SIZE exactly. To make bcache work correctly on
> > 64K PAGE_SIZE, I committed a patch as bellow showd:
> >     diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
> >     index 330cd6e..ef567cd 100644
> >     --- a/drivers/md/bcache/super.c
> >     +++ b/drivers/md/bcache/super.c
> >     @@ -224,6 +224,12 @@ static void __write_super(struct cache_sb
> > *sb, struct bio     *bio
> >         bio->bi_iter.bi_size    = SB_SIZE;
> >         bch_bio_map(bio, NULL);
> >
> >     +#ifdef CONFIG_ARM64_64K_PAGES
> >     +       out = (struct cache_sb *)((char *)out + (SB_SECTOR<<9));
> >     +       pr_debug("sb_page_adress %x, sb_address %x,page_size
> >           %d\n",page_address(bio
> >     +       bio->bi_io_vec[0].bv_offset = (SB_SECTOR<<9);
> >     +#endif
> >
> >         out->offset             = cpu_to_le64(sb->offset);
> >         out->version            = cpu_to_le64(sb->version);
> >
> > Does it not recommend to use bcache on 64K PAGE_SIZE? or it only
> > considers for 4K PAGE_SIZE for bcache currently?
> > Maybe it is more suitable  for me to redefine some macro such as
> > SB_SECTOR, BDEV_DATA_START_DEFAULT to make bcache work correctly on
> > both 64K PAGE_SIZE and 4K PAGE_SIZE.
>
> I think a patch to support arbitrary page size would be great.  Can
> you write the macros in terms of PAGE_SIZE or PAGE_SHIFT?

It shouldn't be referencing PAGE_SIZE at all - creating a new macro
(BCH_SB_SIZE, perhaps) is the correct approach.

The code that allocates the buffer for the superblock will have to be fixed
too - right now it's probably using __get_free_page(), it should probably just
be switched to kmalloc().

> (Out of curiosity, what ARM64 hardware are you using?)
>
> Kent, this may affect bcachefs too.  Can you think of any other places
> that might have PAGE_SIZE!=4k issues?

Oh, there's probably a couple. There's probably some stuff that'll break if
btree_node_size is smaller than PAGE_SIZE, too...

If it turns out to be too much for Zhengyuan, I can probably fix upstream too
(but I don't have any hardware to test with).

--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel