On 2019-10-30, Omar Sandoval <osandov@xxxxxxxxxxx> wrote:
> On Tue, Oct 22, 2019 at 01:02:15PM +1100, Aleksa Sarai wrote:
> > On 2019-10-21, Darrick J. Wong <darrick.wong@xxxxxxxxxx> wrote:
> > > On Tue, Oct 22, 2019 at 05:38:31AM +1100, Aleksa Sarai wrote:
> > > > On 2019-10-21, Darrick J. Wong <darrick.wong@xxxxxxxxxx> wrote:
> > > > > On Tue, Oct 15, 2019 at 11:42:40AM -0700, Omar Sandoval wrote:
> > > > > > From: Omar Sandoval <osandov@xxxxxx>
> > > > > >
> > > > > > Btrfs supports transparent compression: data written by the user can be
> > > > > > compressed when written to disk and decompressed when read back.
> > > > > > However, we'd like to add an interface to write pre-compressed data
> > > > > > directly to the filesystem, and the matching interface to read
> > > > > > compressed data without decompressing it. This adds support for
> > > > > > so-called "encoded I/O" via preadv2() and pwritev2().
> > > > > >
> > > > > > A new RWF_ENCODED flags indicates that a read or write is "encoded". If
> > > > > > this flag is set, iov[0].iov_base points to a struct encoded_iov which
> > > > > > is used for metadata: namely, the compression algorithm, unencoded
> > > > > > (i.e., decompressed) length, and what subrange of the unencoded data
> > > > > > should be used (needed for truncated or hole-punched extents and when
> > > > > > reading in the middle of an extent). For reads, the filesystem returns
> > > > > > this information; for writes, the caller provides it to the filesystem.
> > > > > > iov[0].iov_len must be set to sizeof(struct encoded_iov), which can be
> > > > > > used to extend the interface in the future. The remaining iovecs contain
> > > > > > the encoded extent.
> > > > > >
> > > > > > Filesystems must indicate that they support encoded writes by setting
> > > > > > FMODE_ENCODED_IO in ->file_open().
> > > > > >
> > > > > > Signed-off-by: Omar Sandoval <osandov@xxxxxx>
> > > > > > ---
> > > > > > include/linux/fs.h | 14 +++++++
> > > > > > include/uapi/linux/fs.h | 26 ++++++++++++-
> > > > > > mm/filemap.c | 82 ++++++++++++++++++++++++++++++++++-------
> > > > > > 3 files changed, 108 insertions(+), 14 deletions(-)
> > > > > >
> > > > > > diff --git a/include/linux/fs.h b/include/linux/fs.h
> > > > > > index e0d909d35763..54681f21e05e 100644
> > > > > > --- a/include/linux/fs.h
> > > > > > +++ b/include/linux/fs.h
> > > > > > @@ -175,6 +175,9 @@ typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
> > > > > > /* File does not contribute to nr_files count */
> > > > > > #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000)
> > > > > >
> > > > > > +/* File supports encoded IO */
> > > > > > +#define FMODE_ENCODED_IO ((__force fmode_t)0x40000000)
> > > > > > +
> > > > > > /*
> > > > > > * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
> > > > > > * that indicates that they should check the contents of the iovec are
> > > > > > @@ -314,6 +317,7 @@ enum rw_hint {
> > > > > > #define IOCB_SYNC (1 << 5)
> > > > > > #define IOCB_WRITE (1 << 6)
> > > > > > #define IOCB_NOWAIT (1 << 7)
> > > > > > +#define IOCB_ENCODED (1 << 8)
> > > > > >
> > > > > > struct kiocb {
> > > > > > struct file *ki_filp;
> > > > > > @@ -3088,6 +3092,11 @@ extern int sb_min_blocksize(struct super_block *, int);
> > > > > > extern int generic_file_mmap(struct file *, struct vm_area_struct *);
> > > > > > extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
> > > > > > extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
> > > > > > +struct encoded_iov;
> > > > > > +extern int generic_encoded_write_checks(struct kiocb *, struct encoded_iov *);
> > > > > > +extern ssize_t check_encoded_read(struct kiocb *, struct iov_iter *);
> > > > > > +extern int import_encoded_write(struct kiocb *, struct encoded_iov *,
> > > > > > + struct iov_iter *);
> > > > > > extern int generic_remap_checks(struct file *file_in, loff_t pos_in,
> > > > > > struct file *file_out, loff_t pos_out,
> > > > > > loff_t *count, unsigned int remap_flags);
> > > > > > @@ -3403,6 +3412,11 @@ static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
> > > > > > return -EOPNOTSUPP;
> > > > > > ki->ki_flags |= IOCB_NOWAIT;
> > > > > > }
> > > > > > + if (flags & RWF_ENCODED) {
> > > > > > + if (!(ki->ki_filp->f_mode & FMODE_ENCODED_IO))
> > > > > > + return -EOPNOTSUPP;
> > > > > > + ki->ki_flags |= IOCB_ENCODED;
> > > > > > + }
> > > > > > if (flags & RWF_HIPRI)
> > > > > > ki->ki_flags |= IOCB_HIPRI;
> > > > > > if (flags & RWF_DSYNC)
> > > > > > diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> > > > > > index 379a612f8f1d..ed92a8a257cb 100644
> > > > > > --- a/include/uapi/linux/fs.h
> > > > > > +++ b/include/uapi/linux/fs.h
> > > > > > @@ -284,6 +284,27 @@ struct fsxattr {
> > > > > >
> > > > > > typedef int __bitwise __kernel_rwf_t;
> > > > > >
> > > > > > +enum {
> > > > > > + ENCODED_IOV_COMPRESSION_NONE,
> > > > > > + ENCODED_IOV_COMPRESSION_ZLIB,
> > > > > > + ENCODED_IOV_COMPRESSION_LZO,
> > > > > > + ENCODED_IOV_COMPRESSION_ZSTD,
> > > > > > + ENCODED_IOV_COMPRESSION_TYPES = ENCODED_IOV_COMPRESSION_ZSTD,
> > > > > > +};
> > > > > > +
> > > > > > +enum {
> > > > > > + ENCODED_IOV_ENCRYPTION_NONE,
> > > > > > + ENCODED_IOV_ENCRYPTION_TYPES = ENCODED_IOV_ENCRYPTION_NONE,
> > > > > > +};
> > > > > > +
> > > > > > +struct encoded_iov {
> > > > > > + __u64 len;
> > > > > > + __u64 unencoded_len;
> > > > > > + __u64 unencoded_offset;
> > > > > > + __u32 compression;
> > > > > > + __u32 encryption;
> > > > >
> > > > > Can we add some must-be-zero padding space at the end here for whomever
> > > > > comes along next wanting to add more encoding info?
> > > >
> > > > I would suggest to copy the extension design of copy_struct_from_user().
> > > > Adding must-be-zero padding is a less-ideal solution to the extension
> > > > problem than length-based extension.
> > >
> > > Come to think of it, you /do/ have to specify iov_len so... yeah, do
> > > that instead; we can always extend the structure later.
> >
> > Just to clarify -- if we want to make the interface forward-compatible
> > from the outset (programs built 4 years from now running on 5.5), we
> > will need to implement this in the original merge. Otherwise userspace
> > will need to handle backwards-compatibility themselves once new features
> > are added.
> >
> > @Omar: If it'd make your life easier, I can send some draft patches
> > which port copy_struct_from_user() to iovec-land.
>
> You're right, I didn't think about the case of newer programs on older
> kernels. I can do that for the next submission. RWF_ENCODED should
> probably translate the E2BIG from copy_struct_from_user() to EINVAL,
> though, to avoid ambiguity with the case that the buffer wasn't big
> enough to return the encoded data.
Yeah, that seems fair enough. I would've preferred to keep the error
semantics the same everywhere, but adding additional ambiguity to such
error cases isn't a good idea.
It's a bit of a shame we don't have more granular EINVALs to make it
easier to figure out *why* you got an EINVAL (then we wouldn't have had
to abuse E2BIG to indicate to userspace "you're using a new feature on
an old kernel") -- but that's a more generic problem that probably won't
be solved any time soon.
--
Aleksa Sarai
Senior Software Engineer (Containers)
SUSE Linux GmbH
<https://www.cyphar.com/>
Attachment:
signature.asc
Description: PGP signature
