linux-xfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Brian Foster <bfoster@redhat.com>
To: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: linux-xfs@vger.kernel.org
Subject: Re: [PATCH 1/4] xfs: introduce fake roots for ag-rooted btrees
Date: Wed, 16 Oct 2019 11:26:29 -0400	[thread overview]
Message-ID: <20191016152629.GB41077@bfoster> (raw)
In-Reply-To: <157063968551.2912204.15634530264967900662.stgit@magnolia>

On Wed, Oct 09, 2019 at 09:48:05AM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
> 
> Create an in-core fake root for AG-rooted btree types so that callers
> can generate a whole new btree using the upcoming btree bulk load
> function without making the new tree accessible from the rest of the
> filesystem.  It is up to the individual btree type to provide a function
> to create a staged cursor (presumably with the appropriate callouts to
> update the fakeroot) and then commit the staged root back into the
> filesystem.
> 
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> ---

>  fs/xfs/libxfs/xfs_btree.c |  115 +++++++++++++++++++++++++++++++++++++++++++++
>  fs/xfs/libxfs/xfs_btree.h |   43 +++++++++++++++--
>  fs/xfs/xfs_trace.h        |   28 +++++++++++
>  3 files changed, 181 insertions(+), 5 deletions(-)
> 
> 
> diff --git a/fs/xfs/libxfs/xfs_btree.c b/fs/xfs/libxfs/xfs_btree.c
> index 71de937f9e64..7b18f0fa5e99 100644
> --- a/fs/xfs/libxfs/xfs_btree.c
> +++ b/fs/xfs/libxfs/xfs_btree.c
...
> @@ -4930,3 +4932,116 @@ xfs_btree_has_more_records(
...
> +/* Initialize a pointer to the root block from the fakeroot. */
> +STATIC void
> +xfs_btree_fakeroot_init_ptr_from_cur(
> +	struct xfs_btree_cur	*cur,
> +	union xfs_btree_ptr	*ptr)
> +{
> +	struct xbtree_afakeroot	*afake;
> +
> +	ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
> +
> +	if (cur->bc_flags & XFS_BTREE_LONG_PTRS) {
> +		ptr->l = cpu_to_be64(0);

Why zero here? Is this just not supported?

Otherwise this seems straightforward code-wise. I think I get the
general idea, but it's hard to reason further about the pieces until I
see the broader context..

Brian

> +	} else {
> +		afake = cur->bc_private.a.afake;
> +		ptr->s = cpu_to_be32(afake->af_root);
> +	}
> +}
> +
> +/* Set the root block when our tree has a fakeroot. */
> +STATIC void
> +xfs_btree_afakeroot_set_root(
> +	struct xfs_btree_cur	*cur,
> +	union xfs_btree_ptr	*ptr,
> +	int			inc)
> +{
> +	struct xbtree_afakeroot	*afake = cur->bc_private.a.afake;
> +
> +	ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
> +	afake->af_root = be32_to_cpu(ptr->s);
> +	afake->af_levels += inc;
> +}
> +
> +/*
> + * Initialize a AG-rooted btree cursor with the given AG @fakeroot, and prepare
> + * @ops for overriding by duplicating them into @new_ops.  The caller can
> + * replace pointers in @new_ops as necessary once this call completes.  Note
> + * that staging cursors can only be used for bulk loading.
> + */
> +void
> +xfs_btree_stage_afakeroot(
> +	struct xfs_btree_cur		*cur,
> +	struct xbtree_afakeroot		*afake,
> +	const struct xfs_btree_ops	*ops,
> +	struct xfs_btree_ops		**new_ops)
> +{
> +	ASSERT(!(cur->bc_flags & XFS_BTREE_STAGING));
> +	ASSERT(!(cur->bc_flags & XFS_BTREE_ROOT_IN_INODE));
> +
> +	*new_ops = kmem_alloc(sizeof(struct xfs_btree_ops), KM_NOFS);
> +	memcpy(*new_ops, ops, sizeof(struct xfs_btree_ops));
> +	(*new_ops)->alloc_block = xfs_btree_fakeroot_alloc_block;
> +	(*new_ops)->free_block = xfs_btree_fakeroot_free_block;
> +	(*new_ops)->init_ptr_from_cur = xfs_btree_fakeroot_init_ptr_from_cur;
> +	(*new_ops)->set_root = xfs_btree_afakeroot_set_root;
> +	(*new_ops)->dup_cursor = xfs_btree_fakeroot_dup_cursor;
> +
> +	cur->bc_private.a.afake = afake;
> +	cur->bc_nlevels = afake->af_levels;
> +	cur->bc_ops = *new_ops;
> +	cur->bc_flags |= XFS_BTREE_STAGING;
> +}
> +
> +/*
> + * Transform an AG-rooted staging btree cursor back into a regular btree
> + * cursor.  Caller is responsible for logging changes before this.
> + */
> +void
> +xfs_btree_commit_afakeroot(
> +	struct xfs_btree_cur		*cur,
> +	struct xfs_buf			*agbp,
> +	const struct xfs_btree_ops	*ops)
> +{
> +	ASSERT(cur->bc_flags & XFS_BTREE_STAGING);
> +
> +	trace_xfs_btree_commit_afakeroot(cur);
> +
> +	kmem_free((void *)cur->bc_ops);
> +	cur->bc_private.a.agbp = agbp;
> +	cur->bc_ops = ops;
> +	cur->bc_flags &= ~XFS_BTREE_STAGING;
> +}
> diff --git a/fs/xfs/libxfs/xfs_btree.h b/fs/xfs/libxfs/xfs_btree.h
> index ced1e65d1483..453de8a49e96 100644
> --- a/fs/xfs/libxfs/xfs_btree.h
> +++ b/fs/xfs/libxfs/xfs_btree.h
> @@ -185,6 +185,16 @@ union xfs_btree_cur_private {
>  	} refc;
>  };
>  
> +/* Private information for a AG-rooted btree. */
> +struct xfs_btree_priv_ag {			/* needed for BNO, CNT, INO */
> +	union {
> +		struct xfs_buf		*agbp;	/* agf/agi buffer pointer */
> +		struct xbtree_afakeroot	*afake;	/* fake ag header root */
> +	};
> +	xfs_agnumber_t			agno;	/* ag number */
> +	union xfs_btree_cur_private	priv;
> +};
> +
>  /*
>   * Btree cursor structure.
>   * This collects all information needed by the btree code in one place.
> @@ -206,11 +216,7 @@ typedef struct xfs_btree_cur
>  	xfs_btnum_t	bc_btnum;	/* identifies which btree type */
>  	int		bc_statoff;	/* offset of btre stats array */
>  	union {
> -		struct {			/* needed for BNO, CNT, INO */
> -			struct xfs_buf	*agbp;	/* agf/agi buffer pointer */
> -			xfs_agnumber_t	agno;	/* ag number */
> -			union xfs_btree_cur_private	priv;
> -		} a;
> +		struct xfs_btree_priv_ag a;
>  		struct {			/* needed for BMAP */
>  			struct xfs_inode *ip;	/* pointer to our inode */
>  			int		allocated;	/* count of alloced */
> @@ -229,6 +235,12 @@ typedef struct xfs_btree_cur
>  #define XFS_BTREE_LASTREC_UPDATE	(1<<2)	/* track last rec externally */
>  #define XFS_BTREE_CRC_BLOCKS		(1<<3)	/* uses extended btree blocks */
>  #define XFS_BTREE_OVERLAPPING		(1<<4)	/* overlapping intervals */
> +/*
> + * The root of this btree is a fakeroot structure so that we can stage a btree
> + * rebuild without leaving it accessible via primary metadata.  The ops struct
> + * is dynamically allocated and must be freed when the cursor is deleted.
> + */
> +#define XFS_BTREE_STAGING		(1<<5)
>  
>  
>  #define	XFS_BTREE_NOERROR	0
> @@ -514,4 +526,25 @@ int xfs_btree_has_record(struct xfs_btree_cur *cur, union xfs_btree_irec *low,
>  		union xfs_btree_irec *high, bool *exists);
>  bool xfs_btree_has_more_records(struct xfs_btree_cur *cur);
>  
> +/* Fake root for an AG-rooted btree. */
> +struct xbtree_afakeroot {
> +	/* AG block number of the new btree root. */
> +	xfs_agblock_t		af_root;
> +
> +	/* Height of the new btree. */
> +	unsigned int		af_levels;
> +
> +	/* Number of blocks used by the btree. */
> +	unsigned int		af_blocks;
> +};
> +
> +/* Cursor interactions with with fake roots for AG-rooted btrees. */
> +void xfs_btree_stage_afakeroot(struct xfs_btree_cur *cur,
> +		struct xbtree_afakeroot *afake,
> +		const struct xfs_btree_ops *ops,
> +		struct xfs_btree_ops **new_ops);
> +void xfs_btree_commit_afakeroot(struct xfs_btree_cur *cur,
> +		struct xfs_buf *agbp,
> +		const struct xfs_btree_ops *ops);
> +
>  #endif	/* __XFS_BTREE_H__ */
> diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
> index eaae275ed430..e04ed5324696 100644
> --- a/fs/xfs/xfs_trace.h
> +++ b/fs/xfs/xfs_trace.h
> @@ -3609,6 +3609,34 @@ DEFINE_KMEM_EVENT(kmem_alloc_large);
>  DEFINE_KMEM_EVENT(kmem_realloc);
>  DEFINE_KMEM_EVENT(kmem_zone_alloc);
>  
> +TRACE_EVENT(xfs_btree_commit_afakeroot,
> +	TP_PROTO(struct xfs_btree_cur *cur),
> +	TP_ARGS(cur),
> +	TP_STRUCT__entry(
> +		__field(dev_t, dev)
> +		__field(xfs_btnum_t, btnum)
> +		__field(xfs_agnumber_t, agno)
> +		__field(xfs_agblock_t, agbno)
> +		__field(unsigned int, levels)
> +		__field(unsigned int, blocks)
> +	),
> +	TP_fast_assign(
> +		__entry->dev = cur->bc_mp->m_super->s_dev;
> +		__entry->btnum = cur->bc_btnum;
> +		__entry->agno = cur->bc_private.a.agno;
> +		__entry->agbno = cur->bc_private.a.afake->af_root;
> +		__entry->levels = cur->bc_private.a.afake->af_levels;
> +		__entry->blocks = cur->bc_private.a.afake->af_blocks;
> +	),
> +	TP_printk("dev %d:%d btree %s ag %u levels %u blocks %u root %u",
> +		  MAJOR(__entry->dev), MINOR(__entry->dev),
> +		  __print_symbolic(__entry->btnum, XFS_BTNUM_STRINGS),
> +		  __entry->agno,
> +		  __entry->levels,
> +		  __entry->blocks,
> +		  __entry->agbno)
> +)
> +
>  #endif /* _TRACE_XFS_H */
>  
>  #undef TRACE_INCLUDE_PATH
> 

  reply	other threads:[~2019-10-16 15:26 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-10-09 16:47 [PATCH 0/4] xfs: btree bulk loading Darrick J. Wong
2019-10-09 16:48 ` [PATCH 1/4] xfs: introduce fake roots for ag-rooted btrees Darrick J. Wong
2019-10-16 15:26   ` Brian Foster [this message]
2019-10-16 16:53     ` Darrick J. Wong
2019-10-16 17:36       ` Brian Foster
2019-10-25 19:04         ` Darrick J. Wong
2019-10-09 16:48 ` [PATCH 2/4] xfs: introduce fake roots for inode-rooted btrees Darrick J. Wong
2019-10-09 16:48 ` [PATCH 3/4] xfs: support bulk loading of staged btrees Darrick J. Wong
2019-10-16 15:26   ` Brian Foster
2019-10-16 18:15     ` Darrick J. Wong
2019-10-16 21:07       ` Brian Foster
2019-10-17  0:40         ` Darrick J. Wong
2019-10-17  9:32           ` Brian Foster
2019-10-17 19:06             ` Darrick J. Wong
2019-10-18 14:38               ` Brian Foster
2019-10-09 16:48 ` [PATCH 4/4] xfs: support staging cursors for per-AG btree types Darrick J. Wong
2019-10-16 15:28   ` Brian Foster
2019-10-16 16:57     ` Darrick J. Wong
  -- strict thread matches above, loose matches on Subject: below --
2019-10-29 23:30 [PATCH v2 0/4] xfs: btree bulk loading Darrick J. Wong
2019-10-29 23:30 ` [PATCH 1/4] xfs: introduce fake roots for ag-rooted btrees Darrick J. Wong
2020-01-01  1:00 [PATCH v2 0/4] xfs: btree bulk loading Darrick J. Wong
2020-01-01  1:00 ` [PATCH 1/4] xfs: introduce fake roots for ag-rooted btrees Darrick J. Wong
2020-03-04  3:28 [PATCH v3 0/4] xfs: btree bulk loading Darrick J. Wong
2020-03-04  3:28 ` [PATCH 1/4] xfs: introduce fake roots for ag-rooted btrees Darrick J. Wong
2020-03-04 18:21   ` Brian Foster
2020-03-04 23:34     ` Darrick J. Wong
2020-03-04 23:53       ` Dave Chinner
2020-03-05  1:23         ` Darrick J. Wong
2020-03-05 14:30       ` Brian Foster
2020-03-05 17:39         ` Darrick J. Wong
2020-03-05  1:20   ` Dave Chinner
2020-03-05  1:23     ` Darrick J. Wong

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20191016152629.GB41077@bfoster \
    --to=bfoster@redhat.com \
    --cc=darrick.wong@oracle.com \
    --cc=linux-xfs@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).