Re: [PATCH 6/9] xfs: wire up the new v5 bulkstat_single ioctl

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Fri, Jul 05, 2019 at 07:05:33AM -0400, Brian Foster wrote:
> On Wed, Jul 03, 2019 at 01:01:43PM -0700, Darrick J. Wong wrote:
> > On Wed, Jul 03, 2019 at 12:11:36PM -0400, Brian Foster wrote:
> > > On Wed, Jul 03, 2019 at 07:52:56AM -0700, Darrick J. Wong wrote:
> > > > On Wed, Jul 03, 2019 at 09:24:41AM -0400, Brian Foster wrote:
> > > > > On Wed, Jun 26, 2019 at 01:46:13PM -0700, Darrick J. Wong wrote:
> > > > > > From: Darrick J. Wong <darrick.wong@oracle.com>
> > > > > > 
> > > > > > Wire up the V5 BULKSTAT_SINGLE ioctl and rename the old one V1.
> > > > > > 
> > > > > > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > > > > > ---
> > > > > >  fs/xfs/libxfs/xfs_fs.h |   16 ++++++++++
> > > > > >  fs/xfs/xfs_ioctl.c     |   79 ++++++++++++++++++++++++++++++++++++++++++++++++
> > > > > >  fs/xfs/xfs_ioctl32.c   |    1 +
> > > > > >  fs/xfs/xfs_ondisk.h    |    1 +
> > > > > >  4 files changed, 97 insertions(+)
> > > > > > 
> > > > > > 
> > > > > > diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
> > > > > > index 960f3542e207..95d0411dae9b 100644
> > > > > > --- a/fs/xfs/libxfs/xfs_fs.h
> > > > > > +++ b/fs/xfs/libxfs/xfs_fs.h
> > > > > > @@ -468,6 +468,16 @@ struct xfs_bulk_ireq {
> > > > > >  
> > > > > >  #define XFS_BULK_IREQ_FLAGS_ALL	(0)
> > > > > >  
> > > > > > +/* Header for a single inode request. */
> > > > > > +struct xfs_ireq {
> > > > > > +	uint64_t	ino;		/* I/O: start with this inode	*/
> > > > > > +	uint32_t	flags;		/* I/O: operation flags		*/
> > > > > > +	uint32_t	reserved32;	/* must be zero			*/
> > > > > > +	uint64_t	reserved[2];	/* must be zero			*/
> > > > > > +};
> > > > > > +
> > > > > > +#define XFS_IREQ_FLAGS_ALL	(0)
> > > > > > +
> > > > > >  /*
> > > > > >   * ioctl structures for v5 bulkstat and inumbers requests
> > > > > >   */
> > > > > > @@ -478,6 +488,11 @@ struct xfs_bulkstat_req {
> > > > > >  #define XFS_BULKSTAT_REQ_SIZE(nr)	(sizeof(struct xfs_bulkstat_req) + \
> > > > > >  					 (nr) * sizeof(struct xfs_bulkstat))
> > > > > >  
> > > > > > +struct xfs_bulkstat_single_req {
> > > > > > +	struct xfs_ireq		hdr;
> > > > > > +	struct xfs_bulkstat	bulkstat;
> > > > > > +};
> > > > > > +
> > > > > 
> > > > > What's the reasoning for separate data structures when the single
> > > > > command is basically a subset of standard bulkstat (similar to the older
> > > > > interface)?
> > > > 
> > > > I split them up to avoid having irrelevant bulk_req fields (specifically
> > > > icount and ocount) cluttering up the single_req header.  In patch 9 the
> > > > bulkstat single command grows the ability to request the root inode to
> > > > fix xfsdump's inability to correctly guess the root directory.
> > > > 
> > > 
> > > Ok, figured as much once I got to the magic inode flag patch. It's
> > > probably reasonable either way, the tradeoff to this approach is the
> > > extra boilerplate code associated with processing the separate data
> > > structure.
> > 
> > <nod>
> > 
> > > Thinking about it a bit more.. what was the purpose of the separate
> > > BULKSTAT_SINGLE command in the first place? AFAICT it just toggles the
> > > hacky ->lastip semantic. If we've addressed that with an entirely new
> > > data structure, do we need the separate ioctl at all anymore?
> > 
> > At the moment, somewhat simpler setup in userspace:
> > 
> > 	struct xfs_bulkstat_single_req b = { 0 };
> > 	int ret;
> > 
> > 	b.hdr.ino = some_ino;
> > 	ret = ioctl(fd, XFS_IOC_BULKSTAT_SINGLE, &b);
> > 	if (ret)
> > 		abort();
> > 
> > 	printf("%d\n", b.bulkstat.bs_ino);
> > 
> > vs. a single-inode bulkstat:
> > 
> > 	struct xfs_bulkstat_req *b;
> > 	int ret;
> > 
> > 	b = calloc(1, XFS_BULKSTAT_REQ_SIZE(1));
> > 	if (!b)
> > 		abort();
> > 
> > 	b->hdr.ino = some_ino;
> > 	ret = ioctl(fd, XFS_IOC_BULKSTAT, b);
> > 	if (ret)
> > 		abort();
> > 
> > 	if (b->hdr.ocount)
> > 		printf("%d\n", b->bulkstat[0].bs_ino);
> > 
> > 	free(b);
> > 
> > AFAICT, the main advantages are (1) replacing the heap allocation with a
> > stack allocation and (2) not cluttering the regular bulkstat with the
> > "just grab this one symbolic inode" interface.  We could very well just
> > drop this one, particularly since I'm in the process of wrapping most of
> > the ioctl invocation complexity in libfrog wrapper functions anyway...
> > 
> 
> That seems preferable to me. I'm not that invested in the semantics
> between bulkstat and bulkstat_single, but if the only reason we
> originally had the latter as a separate ioctl was to deal with the
> ->lastip interface wart then I think it would be unfortunate to
> unnecessarily carry that mess forward through another generation of the
> interface that doesn't have the original problem. Userspace should be
> able to easily abstract the single case.

Indeed, it was trivially easy to rewrite the xfrog_bulkstat_single
wrapper. :)

--D

> Brian
> 
> > --D
> > 
> > > Brian
> > > 
> > > > --D
> > > > 
> > > > > Brian
> > > > > 
> > > > > >  /*
> > > > > >   * Error injection.
> > > > > >   */
> > > > > > @@ -780,6 +795,7 @@ struct xfs_scrub_metadata {
> > > > > >  #define XFS_IOC_GOINGDOWN	     _IOR ('X', 125, uint32_t)
> > > > > >  #define XFS_IOC_FSGEOMETRY	     _IOR ('X', 126, struct xfs_fsop_geom)
> > > > > >  #define XFS_IOC_BULKSTAT	     _IOR ('X', 127, struct xfs_bulkstat_req)
> > > > > > +#define XFS_IOC_BULKSTAT_SINGLE	     _IOR ('X', 128, struct xfs_bulkstat_single_req)
> > > > > >  /*	XFS_IOC_GETFSUUID ---------- deprecated 140	 */
> > > > > >  
> > > > > >  
> > > > > > diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
> > > > > > index cf6a38c2a3ed..2c821fa601a4 100644
> > > > > > --- a/fs/xfs/xfs_ioctl.c
> > > > > > +++ b/fs/xfs/xfs_ioctl.c
> > > > > > @@ -922,6 +922,83 @@ xfs_ioc_bulkstat(
> > > > > >  	return 0;
> > > > > >  }
> > > > > >  
> > > > > > +/*
> > > > > > + * Check the incoming singleton request @hdr from userspace and initialize the
> > > > > > + * internal @breq bulk request appropriately.  Returns 0 if the bulk request
> > > > > > + * should proceed; or the usual negative error code.
> > > > > > + */
> > > > > > +static int
> > > > > > +xfs_ireq_setup(
> > > > > > +	struct xfs_mount	*mp,
> > > > > > +	struct xfs_ireq		*hdr,
> > > > > > +	struct xfs_ibulk	*breq,
> > > > > > +	void __user		*ubuffer)
> > > > > > +{
> > > > > > +	if ((hdr->flags & ~XFS_IREQ_FLAGS_ALL) ||
> > > > > > +	    hdr->reserved32 ||
> > > > > > +	    memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
> > > > > > +		return -EINVAL;
> > > > > > +
> > > > > > +	if (XFS_INO_TO_AGNO(mp, hdr->ino) >= mp->m_sb.sb_agcount)
> > > > > > +		return -EINVAL;
> > > > > > +
> > > > > > +	breq->ubuffer = ubuffer;
> > > > > > +	breq->icount = 1;
> > > > > > +	breq->startino = hdr->ino;
> > > > > > +	return 0;
> > > > > > +}
> > > > > > +
> > > > > > +/*
> > > > > > + * Update the userspace singleton request @hdr to reflect the end state of the
> > > > > > + * internal bulk request @breq.  If @error is negative then we return just
> > > > > > + * that; otherwise we copy the state so that userspace can discover what
> > > > > > + * happened.
> > > > > > + */
> > > > > > +static void
> > > > > > +xfs_ireq_teardown(
> > > > > > +	struct xfs_ireq		*hdr,
> > > > > > +	struct xfs_ibulk	*breq)
> > > > > > +{
> > > > > > +	hdr->ino = breq->startino;
> > > > > > +}
> > > > > > +
> > > > > > +/* Handle the v5 bulkstat_single ioctl. */
> > > > > > +STATIC int
> > > > > > +xfs_ioc_bulkstat_single(
> > > > > > +	struct xfs_mount	*mp,
> > > > > > +	unsigned int		cmd,
> > > > > > +	struct xfs_bulkstat_single_req __user *arg)
> > > > > > +{
> > > > > > +	struct xfs_ireq		hdr;
> > > > > > +	struct xfs_ibulk	breq = {
> > > > > > +		.mp		= mp,
> > > > > > +	};
> > > > > > +	int			error;
> > > > > > +
> > > > > > +	if (!capable(CAP_SYS_ADMIN))
> > > > > > +		return -EPERM;
> > > > > > +
> > > > > > +	if (XFS_FORCED_SHUTDOWN(mp))
> > > > > > +		return -EIO;
> > > > > > +
> > > > > > +	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
> > > > > > +		return -EFAULT;
> > > > > > +
> > > > > > +	error = xfs_ireq_setup(mp, &hdr, &breq, &arg->bulkstat);
> > > > > > +	if (error)
> > > > > > +		return error;
> > > > > > +
> > > > > > +	error = xfs_bulkstat_one(&breq, xfs_bulkstat_fmt);
> > > > > > +	if (error)
> > > > > > +		return error;
> > > > > > +
> > > > > > +	xfs_ireq_teardown(&hdr, &breq);
> > > > > > +	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
> > > > > > +		return -EFAULT;
> > > > > > +
> > > > > > +	return 0;
> > > > > > +}
> > > > > > +
> > > > > >  STATIC int
> > > > > >  xfs_ioc_fsgeometry(
> > > > > >  	struct xfs_mount	*mp,
> > > > > > @@ -2088,6 +2165,8 @@ xfs_file_ioctl(
> > > > > >  
> > > > > >  	case XFS_IOC_BULKSTAT:
> > > > > >  		return xfs_ioc_bulkstat(mp, cmd, arg);
> > > > > > +	case XFS_IOC_BULKSTAT_SINGLE:
> > > > > > +		return xfs_ioc_bulkstat_single(mp, cmd, arg);
> > > > > >  
> > > > > >  	case XFS_IOC_FSGEOMETRY_V1:
> > > > > >  		return xfs_ioc_fsgeometry(mp, arg, 3);
> > > > > > diff --git a/fs/xfs/xfs_ioctl32.c b/fs/xfs/xfs_ioctl32.c
> > > > > > index df107adbdbf3..6fa0f41dbae5 100644
> > > > > > --- a/fs/xfs/xfs_ioctl32.c
> > > > > > +++ b/fs/xfs/xfs_ioctl32.c
> > > > > > @@ -581,6 +581,7 @@ xfs_file_compat_ioctl(
> > > > > >  	case FS_IOC_GETFSMAP:
> > > > > >  	case XFS_IOC_SCRUB_METADATA:
> > > > > >  	case XFS_IOC_BULKSTAT:
> > > > > > +	case XFS_IOC_BULKSTAT_SINGLE:
> > > > > >  		return xfs_file_ioctl(filp, cmd, p);
> > > > > >  #if !defined(BROKEN_X86_ALIGNMENT) || defined(CONFIG_X86_X32)
> > > > > >  	/*
> > > > > > diff --git a/fs/xfs/xfs_ondisk.h b/fs/xfs/xfs_ondisk.h
> > > > > > index 954484c6eb96..fa1252657b08 100644
> > > > > > --- a/fs/xfs/xfs_ondisk.h
> > > > > > +++ b/fs/xfs/xfs_ondisk.h
> > > > > > @@ -150,6 +150,7 @@ xfs_check_ondisk_structs(void)
> > > > > >  	XFS_CHECK_STRUCT_SIZE(struct xfs_bulkstat,		192);
> > > > > >  	XFS_CHECK_STRUCT_SIZE(struct xfs_inumbers,		24);
> > > > > >  	XFS_CHECK_STRUCT_SIZE(struct xfs_bulkstat_req,		64);
> > > > > > +	XFS_CHECK_STRUCT_SIZE(struct xfs_bulkstat_single_req,	224);
> > > > > >  }
> > > > > >  
> > > > > >  #endif /* __XFS_ONDISK_H */
> > > > > >