Re: [PATCH v3 14/26] xfs: parent pointer attribute creation

[Allison Henderson <allison.henderson@oracle.com>]

On Fri, 2022-09-23 at 14:11 -0700, Darrick J. Wong wrote:
> On Wed, Sep 21, 2022 at 10:44:46PM -0700,
> allison.henderson@oracle.com wrote:
> > From: Allison Henderson <allison.henderson@oracle.com>
> > 
> > Add parent pointer attribute during xfs_create, and subroutines to
> > initialize attributes
> > 
> > Signed-off-by: Dave Chinner <dchinner@redhat.com>
> > Signed-off-by: Allison Henderson <allison.henderson@oracle.com>
> > ---
> >  fs/xfs/Makefile            |   1 +
> >  fs/xfs/libxfs/xfs_attr.c   |   4 +-
> >  fs/xfs/libxfs/xfs_attr.h   |   4 +-
> >  fs/xfs/libxfs/xfs_parent.c | 135
> > +++++++++++++++++++++++++++++++++++++
> >  fs/xfs/libxfs/xfs_parent.h |  32 +++++++++
> >  fs/xfs/xfs_inode.c         |  37 ++++++++--
> >  fs/xfs/xfs_xattr.c         |   2 +-
> >  fs/xfs/xfs_xattr.h         |   1 +
> >  8 files changed, 207 insertions(+), 9 deletions(-)
> > 
> > diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
> > index 03135a1c31b6..e2b2cf50ffcf 100644
> > --- a/fs/xfs/Makefile
> > +++ b/fs/xfs/Makefile
> > @@ -40,6 +40,7 @@ xfs-y                         += $(addprefix
> > libxfs/, \
> >                                    xfs_inode_fork.o \
> >                                    xfs_inode_buf.o \
> >                                    xfs_log_rlimit.o \
> > +                                  xfs_parent.o \
> >                                    xfs_ag_resv.o \
> >                                    xfs_rmap.o \
> >                                    xfs_rmap_btree.o \
> > diff --git a/fs/xfs/libxfs/xfs_attr.c b/fs/xfs/libxfs/xfs_attr.c
> > index 0c9589261990..805aaa5639d2 100644
> > --- a/fs/xfs/libxfs/xfs_attr.c
> > +++ b/fs/xfs/libxfs/xfs_attr.c
> > @@ -886,7 +886,7 @@ xfs_attr_lookup(
> >         return error;
> >  }
> >  
> > -static int
> > +int
> >  xfs_attr_intent_init(
> >         struct xfs_da_args      *args,
> >         unsigned int            op_flags,       /* op flag (set or
> > remove) */
> > @@ -904,7 +904,7 @@ xfs_attr_intent_init(
> >  }
> >  
> >  /* Sets an attribute for an inode as a deferred operation */
> > -static int
> > +int
> >  xfs_attr_defer_add(
> >         struct xfs_da_args      *args)
> >  {
> > diff --git a/fs/xfs/libxfs/xfs_attr.h b/fs/xfs/libxfs/xfs_attr.h
> > index b79dae788cfb..0cf23f5117ad 100644
> > --- a/fs/xfs/libxfs/xfs_attr.h
> > +++ b/fs/xfs/libxfs/xfs_attr.h
> > @@ -544,6 +544,7 @@ int xfs_inode_hasattr(struct xfs_inode *ip);
> >  bool xfs_attr_is_leaf(struct xfs_inode *ip);
> >  int xfs_attr_get_ilocked(struct xfs_da_args *args);
> >  int xfs_attr_get(struct xfs_da_args *args);
> > +int xfs_attr_defer_add(struct xfs_da_args *args);
> >  int xfs_attr_set(struct xfs_da_args *args);
> >  int xfs_attr_set_iter(struct xfs_attr_intent *attr);
> >  int xfs_attr_remove_iter(struct xfs_attr_intent *attr);
> > @@ -552,7 +553,8 @@ bool xfs_attr_namecheck(struct xfs_mount *mp,
> > const void *name, size_t length,
> >  int xfs_attr_calc_size(struct xfs_da_args *args, int *local);
> >  void xfs_init_attr_trans(struct xfs_da_args *args, struct
> > xfs_trans_res *tres,
> >                          unsigned int *total);
> > -
> > +int xfs_attr_intent_init(struct xfs_da_args *args, unsigned int
> > op_flags,
> > +                        struct xfs_attr_intent  **attr);
> >  /*
> >   * Check to see if the attr should be upgraded from non-existent
> > or shortform to
> >   * single-leaf-block attribute list.
> > diff --git a/fs/xfs/libxfs/xfs_parent.c
> > b/fs/xfs/libxfs/xfs_parent.c
> > new file mode 100644
> > index 000000000000..dddbf096a4b5
> > --- /dev/null
> > +++ b/fs/xfs/libxfs/xfs_parent.c
> > @@ -0,0 +1,135 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (c) 2022 Oracle, Inc.
> > + * All rights reserved.
> > + */
> > +#include "xfs.h"
> > +#include "xfs_fs.h"
> > +#include "xfs_format.h"
> > +#include "xfs_da_format.h"
> > +#include "xfs_log_format.h"
> > +#include "xfs_shared.h"
> > +#include "xfs_trans_resv.h"
> > +#include "xfs_mount.h"
> > +#include "xfs_bmap_btree.h"
> > +#include "xfs_inode.h"
> > +#include "xfs_error.h"
> > +#include "xfs_trace.h"
> > +#include "xfs_trans.h"
> > +#include "xfs_da_btree.h"
> > +#include "xfs_attr.h"
> > +#include "xfs_da_btree.h"
> > +#include "xfs_attr_sf.h"
> > +#include "xfs_bmap.h"
> > +#include "xfs_defer.h"
> > +#include "xfs_log.h"
> > +#include "xfs_xattr.h"
> > +#include "xfs_parent.h"
> > +
> > +/*
> > + * Parent pointer attribute handling.
> > + *
> > + * Because the attribute value is a filename component, it will
> > never be longer
> > + * than 255 bytes. This means the attribute will always be a local
> > format
> > + * attribute as it is xfs_attr_leaf_entsize_local_max() for v5
> > filesystems will
> > + * always be larger than this (max is 75% of block size).
> > + *
> > + * Creating a new parent attribute will always create a new
> > attribute - there
> > + * should never, ever be an existing attribute in the tree for a
> > new inode.
> > + * ENOSPC behavior is problematic - creating the inode without the
> > parent
> > + * pointer is effectively a corruption, so we allow parent
> > attribute creation
> > + * to dip into the reserve block pool to avoid unexpected ENOSPC
> > errors from
> > + * occurring.
> 
> Hoisting a discussion (that I never replied to; sorry... :( ) from
> last
> time around:
> 
> > > Shouldn't we increase XFS_LINK_SPACE_RES to avoid this?  The
> > > reserve
> > > pool isn't terribly large (8192 blocks) and was really only
> > > supposed
> > > to save us from an ENOSPC shutdown if an unwritten extent
> > > conversion
> > > in the writeback endio handler needs a few more blocks.
> > > 
> > Did you maybe mean XFS_IALLOC_SPACE_RES?  That looks like the macro
> > that's getting used below in xfs_create
> 
> I meant modifying XFS_MKDIR_SPACE_RES and XFS_CREATE_SPACE_RES:
> 
> unsigned int
> xfs_pptr_calc_space_res(
>         struct xfs_mount        *mp,
>         unsigned int            namelen)
> {
>         /*
>          * Code lifted from xfs_attr_calc_size, check this for
>          * correctness since I assumed that a pptr never requires rmt
>          * blocks...
>          */
>         return XFS_DAENTER_SPACE_RES(mp, XFS_ATTR_FORK) +
>                XFS_NEXTENTADD_SPACE_RES(mp, namelen, XFS_ATTR_FORK);
> }
> 
> unsigned int
> xfs_create_space_res(
>         struct xfs_mount        *mp,
>         unsigned int            namelen)
> {
>         unsigned int            ret;
> 
>         ret = XFS_IALLOC_SPACE_RES(mp) + XFS_DIRENTER_SPACE_RES(mp,
> namelen);
>         if (xfs_has_parent(mp))
>                 ret += xfs_pptr_calc_space_res(mp, namelen);
> 
>         return ret;
> }
> 
> unsigned int
> xfs_mkdir_space_res(
>         struct xfs_mount        *mp,
>         unsigned int            namelen)
> {
>         return xfs_create_space_res(mp, namelen);
> }
> 
> (and then change the function case names as necessary)
> 
Oooh ok...  So you're saying you want wrappers to XFS_IALLOC_SPACE_RES,
right?  Do you want these helpers corralled in a common area, or do you
just want them to be static in the same file where they are used?

> > > IOWs, we really ought to ENOSPC at transaction reservation time
> > > instead of draining the reserve pool.
> > It looks like we do that in most cases.  I dont actually see rsvd
> > getting set, other than in xfs_attr_set.  Which isnt used in parent
> > pointer updating, and should probably be removed.  I suspect it's a
> > relic of the pre-larp version of the set. So perhaps the comment is
> > stale and should be removed as well.
> 
> The block reservations for create/mkdir/link/unlink all need to be
> adjusted upwards, which will eliminate the need for the comment.
> 
> > + */
> > +
> > +
> > +/* Initializes a xfs_parent_name_rec to be stored as an attribute
> > name */
> > +void
> > +xfs_init_parent_name_rec(
> > +       struct xfs_parent_name_rec      *rec,
> > +       struct xfs_inode                *ip,
> > +       uint32_t                        p_diroffset)
> > +{
> > +       xfs_ino_t                       p_ino = ip->i_ino;
> > +       uint32_t                        p_gen = VFS_I(ip)-
> > >i_generation;
> > +
> > +       rec->p_ino = cpu_to_be64(p_ino);
> > +       rec->p_gen = cpu_to_be32(p_gen);
> > +       rec->p_diroffset = cpu_to_be32(p_diroffset);
> > +}
> > +
> > +/* Initializes a xfs_parent_name_irec from an xfs_parent_name_rec
> > */
> > +void
> > +xfs_init_parent_name_irec(
> > +       struct xfs_parent_name_irec     *irec,
> > +       struct xfs_parent_name_rec      *rec)
> > +{
> > +       irec->p_ino = be64_to_cpu(rec->p_ino);
> > +       irec->p_gen = be32_to_cpu(rec->p_gen);
> > +       irec->p_diroffset = be32_to_cpu(rec->p_diroffset);
> > +}
> > +
> > +int
> > +xfs_parent_init(
> > +       xfs_mount_t                     *mp,
> 
> Nit: typedef usage.
Will fix
> 
> Everything else below here looks good though. :)
> 
Thanks for the reviews!

Allison
> --D
> 
> > +       struct xfs_parent_defer         **parentp)
> > +{
> > +       struct xfs_parent_defer         *parent;
> > +       int                             error;
> > +
> > +       if (!xfs_has_parent(mp))
> > +               return 0;
> > +
> > +       error = xfs_attr_grab_log_assist(mp);
> > +       if (error)
> > +               return error;
> > +
> > +       parent = kzalloc(sizeof(*parent), GFP_KERNEL);
> > +       if (!parent)
> > +               return -ENOMEM;
> > +
> > +       /* init parent da_args */
> > +       parent->args.geo = mp->m_attr_geo;
> > +       parent->args.whichfork = XFS_ATTR_FORK;
> > +       parent->args.attr_filter = XFS_ATTR_PARENT;
> > +       parent->args.op_flags = XFS_DA_OP_OKNOENT |
> > XFS_DA_OP_LOGGED;
> > +       parent->args.name = (const uint8_t *)&parent->rec;
> > +       parent->args.namelen = sizeof(struct xfs_parent_name_rec);
> > +
> > +       *parentp = parent;
> > +       return 0;
> > +}
> > +
> > +int
> > +xfs_parent_defer_add(
> > +       struct xfs_trans        *tp,
> > +       struct xfs_parent_defer *parent,
> > +       struct xfs_inode        *dp,
> > +       struct xfs_name         *parent_name,
> > +       xfs_dir2_dataptr_t      diroffset,
> > +       struct xfs_inode        *child)
> > +{
> > +       struct xfs_da_args      *args = &parent->args;
> > +
> > +       xfs_init_parent_name_rec(&parent->rec, dp, diroffset);
> > +       args->hashval = xfs_da_hashname(args->name, args->namelen);
> > +
> > +       args->trans = tp;
> > +       args->dp = child;
> > +       if (parent_name) {
> > +               parent->args.value = (void *)parent_name->name;
> > +               parent->args.valuelen = parent_name->len;
> > +       }
> > +
> > +       return xfs_attr_defer_add(args);
> > +}
> > +
> > +void
> > +xfs_parent_cancel(
> > +       xfs_mount_t             *mp,
> > +       struct xfs_parent_defer *parent)
> > +{
> > +       xlog_drop_incompat_feat(mp->m_log);
> > +       kfree(parent);
> > +}
> > +
> > diff --git a/fs/xfs/libxfs/xfs_parent.h
> > b/fs/xfs/libxfs/xfs_parent.h
> > new file mode 100644
> > index 000000000000..971044458f8a
> > --- /dev/null
> > +++ b/fs/xfs/libxfs/xfs_parent.h
> > @@ -0,0 +1,32 @@
> > +// SPDX-License-Identifier: GPL-2.0
> > +/*
> > + * Copyright (c) 2022 Oracle, Inc.
> > + * All Rights Reserved.
> > + */
> > +#ifndef        __XFS_PARENT_H__
> > +#define        __XFS_PARENT_H__
> > +
> > +/*
> > + * Dynamically allocd structure used to wrap the needed data to
> > pass around
> > + * the defer ops machinery
> > + */
> > +struct xfs_parent_defer {
> > +       struct xfs_parent_name_rec      rec;
> > +       struct xfs_da_args              args;
> > +};
> > +
> > +/*
> > + * Parent pointer attribute prototypes
> > + */
> > +void xfs_init_parent_name_rec(struct xfs_parent_name_rec *rec,
> > +                             struct xfs_inode *ip,
> > +                             uint32_t p_diroffset);
> > +void xfs_init_parent_name_irec(struct xfs_parent_name_irec *irec,
> > +                              struct xfs_parent_name_rec *rec);
> > +int xfs_parent_init(xfs_mount_t *mp, struct xfs_parent_defer
> > **parentp);
> > +int xfs_parent_defer_add(struct xfs_trans *tp, struct
> > xfs_parent_defer *parent,
> > +                        struct xfs_inode *dp, struct xfs_name
> > *parent_name,
> > +                        xfs_dir2_dataptr_t diroffset, struct
> > xfs_inode *child);
> > +void xfs_parent_cancel(xfs_mount_t *mp, struct xfs_parent_defer
> > *parent);
> > +
> > +#endif /* __XFS_PARENT_H__ */
> > diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> > index 6eb264598517..181d6417412e 100644
> > --- a/fs/xfs/xfs_inode.c
> > +++ b/fs/xfs/xfs_inode.c
> > @@ -37,6 +37,8 @@
> >  #include "xfs_reflink.h"
> >  #include "xfs_ag.h"
> >  #include "xfs_log_priv.h"
> > +#include "xfs_parent.h"
> > +#include "xfs_xattr.h"
> >  
> >  struct kmem_cache *xfs_inode_cache;
> >  
> > @@ -949,7 +951,7 @@ xfs_bumplink(
> >  int
> >  xfs_create(
> >         struct user_namespace   *mnt_userns,
> > -       xfs_inode_t             *dp,
> > +       struct xfs_inode        *dp,
> >         struct xfs_name         *name,
> >         umode_t                 mode,
> >         dev_t                   rdev,
> > @@ -961,7 +963,7 @@ xfs_create(
> >         struct xfs_inode        *ip = NULL;
> >         struct xfs_trans        *tp = NULL;
> >         int                     error;
> > -       bool                    unlock_dp_on_error = false;
> > +       bool                    unlock_dp_on_error = false;
> >         prid_t                  prid;
> >         struct xfs_dquot        *udqp = NULL;
> >         struct xfs_dquot        *gdqp = NULL;
> > @@ -969,6 +971,8 @@ xfs_create(
> >         struct xfs_trans_res    *tres;
> >         uint                    resblks;
> >         xfs_ino_t               ino;
> > +       xfs_dir2_dataptr_t      diroffset;
> > +       struct xfs_parent_defer *parent = NULL;
> >  
> >         trace_xfs_create(dp, name);
> >  
> > @@ -995,6 +999,12 @@ xfs_create(
> >                 tres = &M_RES(mp)->tr_create;
> >         }
> >  
> > +       if (xfs_has_parent(mp)) {
> > +               error = xfs_parent_init(mp, &parent);
> > +               if (error)
> > +                       goto out_release_dquots;
> > +       }
> > +
> >         /*
> >          * Initially assume that the file does not exist and
> >          * reserve the resources for that case.  If that is not
> > @@ -1010,7 +1020,7 @@ xfs_create(
> >                                 resblks, &tp);
> >         }
> >         if (error)
> > -               goto out_release_dquots;
> > +               goto drop_incompat;
> >  
> >         xfs_ilock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
> >         unlock_dp_on_error = true;
> > @@ -1020,6 +1030,7 @@ xfs_create(
> >          * entry pointing to them, but a directory also the "."
> > entry
> >          * pointing to itself.
> >          */
> > +       init_xattrs = init_xattrs || xfs_has_parent(mp);
> >         error = xfs_dialloc(&tp, dp->i_ino, mode, &ino);
> >         if (!error)
> >                 error = xfs_init_new_inode(mnt_userns, tp, dp, ino,
> > mode,
> > @@ -1034,11 +1045,12 @@ xfs_create(
> >          * the transaction cancel unlocking dp so don't do it
> > explicitly in the
> >          * error path.
> >          */
> > -       xfs_trans_ijoin(tp, dp, XFS_ILOCK_EXCL);
> > +       xfs_trans_ijoin(tp, dp, 0);
> >         unlock_dp_on_error = false;
> >  
> >         error = xfs_dir_createname(tp, dp, name, ip->i_ino,
> > -                                  resblks -
> > XFS_IALLOC_SPACE_RES(mp), NULL);
> > +                                  resblks -
> > XFS_IALLOC_SPACE_RES(mp),
> > +                                  &diroffset);
> >         if (error) {
> >                 ASSERT(error != -ENOSPC);
> >                 goto out_trans_cancel;
> > @@ -1054,6 +1066,17 @@ xfs_create(
> >                 xfs_bumplink(tp, dp);
> >         }
> >  
> > +       /*
> > +        * If we have parent pointers, we need to add the attribute
> > containing
> > +        * the parent information now.
> > +        */
> > +       if (parent) {
> > +               error = xfs_parent_defer_add(tp, parent, dp, name,
> > diroffset,
> > +                                            ip);
> > +               if (error)
> > +                       goto out_trans_cancel;
> > +       }
> > +
> >         /*
> >          * If this is a synchronous mount, make sure that the
> >          * create transaction goes to disk before returning to
> > @@ -1079,6 +1102,7 @@ xfs_create(
> >  
> >         *ipp = ip;
> >         xfs_iunlock(ip, XFS_ILOCK_EXCL);
> > +       xfs_iunlock(dp, XFS_ILOCK_EXCL | XFS_ILOCK_PARENT);
> >         return 0;
> >  
> >   out_trans_cancel:
> > @@ -1093,6 +1117,9 @@ xfs_create(
> >                 xfs_finish_inode_setup(ip);
> >                 xfs_irele(ip);
> >         }
> > + drop_incompat:
> > +       if (parent)
> > +               xfs_parent_cancel(mp, parent);
> >   out_release_dquots:
> >         xfs_qm_dqrele(udqp);
> >         xfs_qm_dqrele(gdqp);
> > diff --git a/fs/xfs/xfs_xattr.c b/fs/xfs/xfs_xattr.c
> > index c325a28b89a8..d9067c5f6bd6 100644
> > --- a/fs/xfs/xfs_xattr.c
> > +++ b/fs/xfs/xfs_xattr.c
> > @@ -27,7 +27,7 @@
> >   * they must release the permission by calling
> > xlog_drop_incompat_feat
> >   * when they're done.
> >   */
> > -static inline int
> > +int
> >  xfs_attr_grab_log_assist(
> >         struct xfs_mount        *mp)
> >  {
> > diff --git a/fs/xfs/xfs_xattr.h b/fs/xfs/xfs_xattr.h
> > index 2b09133b1b9b..3fd6520a4d69 100644
> > --- a/fs/xfs/xfs_xattr.h
> > +++ b/fs/xfs/xfs_xattr.h
> > @@ -7,6 +7,7 @@
> >  #define __XFS_XATTR_H__
> >  
> >  int xfs_attr_change(struct xfs_da_args *args);
> > +int xfs_attr_grab_log_assist(struct xfs_mount *mp);
> >  
> >  extern const struct xattr_handler *xfs_xattr_handlers[];
> >  
> > -- 
> > 2.25.1
> >