Am 02.10.2016 um 21:13 hat Tomáš Golembiovský geschrieben:
> Added two new options 'offset' and 'size'. This makes it possible to use
> only part of the file as a device. This can be used e.g. to limit the
> access only to single partition in a disk image or use a disk inside a
> tar archive (like OVA).
>
> For now this is only possible for files in read-only mode. It should be
> possible to extend it later to allow read-write mode, but would probably
> require that the size of the device is kept constant (i.e. no resizing).
>
> Signed-off-by: Tomáš Golembiovský <tgole...@redhat.com>
> ---
> block/raw-posix.c | 97
> +++++++++++++++++++++++++++++++++++++++++++++++++++----
> 1 file changed, 91 insertions(+), 6 deletions(-)
Sounds like a very useful feature. I agree with comments made by others
that we should include write support from the start, and that the code
should be moved to raw_bsd.c.
You also need to update the options for the blockdev-add QMP command in
qapi/block-core.json. The relevant type is BlockdevOptionsFile.
> diff --git a/block/raw-posix.c b/block/raw-posix.c
> index 6ed7547..36f2712 100644
> --- a/block/raw-posix.c
> +++ b/block/raw-posix.c
> @@ -136,6 +136,8 @@ typedef struct BDRVRawState {
> int type;
> int open_flags;
> size_t buf_align;
> + uint64_t offset;
> + uint64_t assumed_size;
>
> #ifdef CONFIG_XFS
> bool is_xfs:1;
> @@ -154,6 +156,7 @@ typedef struct BDRVRawReopenState {
>
> static int fd_open(BlockDriverState *bs);
> static int64_t raw_getlength(BlockDriverState *bs);
> +static int64_t raw_getlength_real(BlockDriverState *bs);
>
> typedef struct RawPosixAIOData {
> BlockDriverState *bs;
> @@ -399,6 +402,16 @@ static QemuOptsList raw_runtime_opts = {
> .type = QEMU_OPT_STRING,
> .help = "File name of the image",
> },
> + {
> + .name = "offset",
> + .type = QEMU_OPT_SIZE,
> + .help = "Offset into the file"
> + },
> + {
> + .name = BLOCK_OPT_SIZE,
> + .type = QEMU_OPT_SIZE,
> + .help = "Virtual disk size"
> + },
BLOCK_OPT_SIZE is an option for bdrv_create(). I'd either create a
BDRV_OPT_SIZE to be consistent with other bdrv_open() option name
macros, or just use a string literal here.
> { /* end of list */ }
> },
> };
> @@ -412,6 +425,7 @@ static int raw_open_common(BlockDriverState *bs, QDict
> *options,
> const char *filename = NULL;
> int fd, ret;
> struct stat st;
> + int64_t real_size;
>
> opts = qemu_opts_create(&raw_runtime_opts, NULL, 0, &error_abort);
> qemu_opts_absorb_qdict(opts, options, &local_err);
> @@ -421,6 +435,17 @@ static int raw_open_common(BlockDriverState *bs, QDict
> *options,
> goto fail;
> }
>
> + s->offset = qemu_opt_get_size(opts, "offset", 0);
> + s->assumed_size = qemu_opt_get_size(opts, BLOCK_OPT_SIZE, 0);
> +
> + if (((bs->drv != &bdrv_file) || !bs->read_only) &&
> + ((s->offset > 0) || (s->assumed_size > 0))) {
> + error_setg(errp, "offset and size options are allowed only for "
> + "files in read-only mode");
> + ret = -EINVAL;
> + goto fail;
> + }
> +
> filename = qemu_opt_get(opts, "filename");
>
> ret = raw_normalize_devicepath(&filename);
> @@ -443,6 +468,23 @@ static int raw_open_common(BlockDriverState *bs, QDict
> *options,
> }
> s->fd = fd;
>
> + /* Check size and offset */
> + real_size = raw_getlength_real(bs);
raw_getlength() can fail, so real_size < 0 should be handled separately
(otherwise we end up with an error message that doesn't make a lot of
sense).
> + if (real_size < (s->offset + s->assumed_size)) {
> + if (s->assumed_size == 0) {
> + error_setg(errp, "The offset has to be smaller than actual size "
> + "of the containing file (%ld) ",
> + real_size);
> + } else {
> + error_setg(errp, "The sum of offset (%lu) and size (%lu) has to "
> + "be smaller than actual size of the containing "
> + "file (%ld) ",
> + s->offset, s->assumed_size, real_size);
I think the need for PRId64/PRIu64 has been mentioned.
Also s/has to be smaller/must not be larger/. Equality is allowed.
> + }
> + ret = -EINVAL;
> + goto fail;
> + }
> +
> #ifdef CONFIG_LINUX_AIO
> if (!raw_use_aio(bdrv_flags) && (bdrv_flags & BDRV_O_NATIVE_AIO)) {
> error_setg(errp, "aio=native was specified, but it requires "
> @@ -1271,6 +1313,19 @@ static int coroutine_fn raw_co_preadv(BlockDriverState
> *bs, uint64_t offset,
> uint64_t bytes, QEMUIOVector *qiov,
> int flags)
> {
> + BDRVRawState *s = bs->opaque;
> + if (s->assumed_size > 0) {
> + if (offset > s->assumed_size) {
> + /* Attempt to read beyond EOF */
> + return 0;
> + } else if ((offset + bytes) > s->assumed_size) {
> + /* Trying to read more than is available */
> + bytes = s->assumed_size - offset;
> + }
> + }
As long as we make sure to return the right values in .bdrv_getlength(),
we can assert(!s->assumed_size || offset + bytes <= s->assumed_size)
instead of doing all of this.
The block layer already handles reads across and beyond EOF (by
returning a zeroed area, not ignoring half of the request as this code
would do).
> + offset += s->offset;
> +
> return raw_co_prw(bs, offset, bytes, qiov, QEMU_AIO_READ);
> }
>
> @@ -1348,7 +1403,7 @@ static int raw_truncate(BlockDriverState *bs, int64_t
> offset)
> }
>
> #ifdef __OpenBSD__
> -static int64_t raw_getlength(BlockDriverState *bs)
> +static int64_t raw_getlength_real(BlockDriverState *bs)
> {
> BDRVRawState *s = bs->opaque;
> int fd = s->fd;
> @@ -1367,7 +1422,7 @@ static int64_t raw_getlength(BlockDriverState *bs)
> return st.st_size;
> }
> #elif defined(__NetBSD__)
> -static int64_t raw_getlength(BlockDriverState *bs)
> +static int64_t raw_getlength_real(BlockDriverState *bs)
> {
> BDRVRawState *s = bs->opaque;
> int fd = s->fd;
> @@ -1392,7 +1447,7 @@ static int64_t raw_getlength(BlockDriverState *bs)
> return st.st_size;
> }
> #elif defined(__sun__)
> -static int64_t raw_getlength(BlockDriverState *bs)
> +static int64_t raw_getlength_real(BlockDriverState *bs)
> {
> BDRVRawState *s = bs->opaque;
> struct dk_minfo minfo;
> @@ -1423,7 +1478,7 @@ static int64_t raw_getlength(BlockDriverState *bs)
> return size;
> }
> #elif defined(CONFIG_BSD)
> -static int64_t raw_getlength(BlockDriverState *bs)
> +static int64_t raw_getlength_real(BlockDriverState *bs)
> {
> BDRVRawState *s = bs->opaque;
> int fd = s->fd;
> @@ -1497,7 +1552,7 @@ again:
> return size;
> }
> #else
> -static int64_t raw_getlength(BlockDriverState *bs)
> +static int64_t raw_getlength_real(BlockDriverState *bs)
> {
> BDRVRawState *s = bs->opaque;
> int ret;
> @@ -1516,6 +1571,28 @@ static int64_t raw_getlength(BlockDriverState *bs)
> }
> #endif
>
> +static int64_t raw_getlength(BlockDriverState *bs)
> +{
> + BDRVRawState *s = bs->opaque;
> +
> + if (s->assumed_size > 0) {
> + return (int64_t)s->assumed_size;
> + }
> +
> + int64_t size = raw_getlength_real(bs);
> + if (s->offset > 0) {
> + if (s->offset > size) {
> + /* The size has changed! We didn't expect that. */
> + return -EIO;
You're overwriting any error return code with -EIO here. I think the
correct way to phrase it would be:
if (size >= 0 && s->offset > size) {
return -EIO;
}
return size - s->offset;
You can subtract s->offset unconditionally as subtracting 0 doesn't
hurt.
>
> + }
> + size -= s->offset;
> + }
> +
> + return size;
> +}
> +
> +
> +
Why three newlines?
> static int64_t raw_get_allocated_file_size(BlockDriverState *bs)
> {
> struct stat st;
> @@ -1524,7 +1601,15 @@ static int64_t
> raw_get_allocated_file_size(BlockDriverState *bs)
> if (fstat(s->fd, &st) < 0) {
> return -errno;
> }
> - return (int64_t)st.st_blocks * 512;
> + uint64_t size = st.st_blocks * 512;
> + /* If the file is sparse we have no idea which part of the file is
> + * allocated and which is not. So we just make sure the returned value is
> + * not greater than what we're working with.
> + */
> + if (s->assumed_size > 0 && s->assumed_size < size) {
> + size = s->assumed_size;
> + }
> + return (int64_t)size;
> }
This function isn't on a hot path, so maybe even using a
bdrv_get_block_status() loop to determine the actual allocation status
could be defendable here. I won't insist on it, though.
Kevin
