block/file-posix.c | 11 +++++++++++ 1 file changed, 11 insertions(+)
People get surprised when, after "qemu-imc create -f raw /dev/sdX", they
still see qcow2 with "qemu-img info", if previously the bdev had a qcow2
header. While this is natural because raw doesn't need to write any
magic bytes during creation, hdev_create is free to clear out the first
sector to make sure the stale qcow2 header doesn't cause such a
confusion.
Signed-off-by: Fam Zheng <famz@redhat.com>
---
block/file-posix.c | 11 +++++++++++
1 file changed, 11 insertions(+)
diff --git a/block/file-posix.c b/block/file-posix.c
index f4de022ae0..1d8ef6f873 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -2703,6 +2703,17 @@ static int hdev_create(const char *filename, QemuOpts *opts,
ret = -ENOSPC;
}
+ if (total_size) {
+ int64_t zero_size = MIN(BDRV_SECTOR_SIZE, total_size);
+ uint8_t *buf;
+ if (lseek(fd, 0, SEEK_SET) == -1) {
+ ret = -errno;
+ } else {
+ buf = g_malloc0(zero_size);
+ ret = qemu_write_full(fd, buf, zero_size);
+ g_free(buf);
+ }
+ }
qemu_close(fd);
return ret;
}
--
2.13.3
On 08/10/2017 03:01 AM, Fam Zheng wrote: > People get surprised when, after "qemu-imc create -f raw /dev/sdX", they > still see qcow2 with "qemu-img info", if previously the bdev had a qcow2 > header. While this is natural because raw doesn't need to write any > magic bytes during creation, hdev_create is free to clear out the first > sector to make sure the stale qcow2 header doesn't cause such a > confusion. > > Signed-off-by: Fam Zheng <famz@redhat.com> > --- > block/file-posix.c | 11 +++++++++++ > 1 file changed, 11 insertions(+) > > diff --git a/block/file-posix.c b/block/file-posix.c > index f4de022ae0..1d8ef6f873 100644 > --- a/block/file-posix.c > +++ b/block/file-posix.c > @@ -2703,6 +2703,17 @@ static int hdev_create(const char *filename, QemuOpts *opts, > ret = -ENOSPC; > } > > + if (total_size) { > + int64_t zero_size = MIN(BDRV_SECTOR_SIZE, total_size); > + uint8_t *buf; Since BDRV_SECTOR_SIZE is small enough to stack-allocate, you could skip the malloc by doing: uint8_t buf[BDRV_SECTOR_SIZE] = ""; > + if (lseek(fd, 0, SEEK_SET) == -1) { > + ret = -errno; > + } else { > + buf = g_malloc0(zero_size); > + ret = qemu_write_full(fd, buf, zero_size); Instead of doing lseek + qemu_write_full, can we just use qemu_pwritev(fd, &iov, 1, 0) with an iov set up to point to the appropriate amount of buf? At any rate, my ideas are micro-optimizations, so I can also live with how you wrote it. Reviewed-by: Eric Blake <eblake@redhat.com> Are you arguing that this is a bug-fix worthy of inclusion in 2.10, because it helps avoid user confusion? Or are you delaying it to 2.11, because we've had the existing behavior for longer than one release, so one release more won't hurt? -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
On Thu, 08/10 08:58, Eric Blake wrote: > On 08/10/2017 03:01 AM, Fam Zheng wrote: > > People get surprised when, after "qemu-imc create -f raw /dev/sdX", they > > still see qcow2 with "qemu-img info", if previously the bdev had a qcow2 > > header. While this is natural because raw doesn't need to write any > > magic bytes during creation, hdev_create is free to clear out the first > > sector to make sure the stale qcow2 header doesn't cause such a > > confusion. > > > > Signed-off-by: Fam Zheng <famz@redhat.com> > > --- > > block/file-posix.c | 11 +++++++++++ > > 1 file changed, 11 insertions(+) > > > > diff --git a/block/file-posix.c b/block/file-posix.c > > index f4de022ae0..1d8ef6f873 100644 > > --- a/block/file-posix.c > > +++ b/block/file-posix.c > > @@ -2703,6 +2703,17 @@ static int hdev_create(const char *filename, QemuOpts *opts, > > ret = -ENOSPC; > > } > > > > + if (total_size) { > > + int64_t zero_size = MIN(BDRV_SECTOR_SIZE, total_size); > > + uint8_t *buf; > > Since BDRV_SECTOR_SIZE is small enough to stack-allocate, you could skip > the malloc by doing: > > uint8_t buf[BDRV_SECTOR_SIZE] = ""; > > > + if (lseek(fd, 0, SEEK_SET) == -1) { > > + ret = -errno; > > + } else { > > + buf = g_malloc0(zero_size); > > + ret = qemu_write_full(fd, buf, zero_size); > > Instead of doing lseek + qemu_write_full, can we just use > qemu_pwritev(fd, &iov, 1, 0) with an iov set up to point to the > appropriate amount of buf? Neat, will update. > > At any rate, my ideas are micro-optimizations, so I can also live with > how you wrote it. > > Reviewed-by: Eric Blake <eblake@redhat.com> > > Are you arguing that this is a bug-fix worthy of inclusion in 2.10, > because it helps avoid user confusion? Or are you delaying it to 2.11, > because we've had the existing behavior for longer than one release, so > one release more won't hurt? IMO no need to rush for 2.10, it can wait for 2.11. Thanks! Fam
On Fri, 08/11 15:28, Fam Zheng wrote: > On Thu, 08/10 08:58, Eric Blake wrote: > > On 08/10/2017 03:01 AM, Fam Zheng wrote: > > > + if (total_size) { > > > + int64_t zero_size = MIN(BDRV_SECTOR_SIZE, total_size); > > > + uint8_t *buf; > > > > Since BDRV_SECTOR_SIZE is small enough to stack-allocate, you could skip > > the malloc by doing: > > > > uint8_t buf[BDRV_SECTOR_SIZE] = ""; > > > > > + if (lseek(fd, 0, SEEK_SET) == -1) { > > > + ret = -errno; > > > + } else { > > > + buf = g_malloc0(zero_size); > > > + ret = qemu_write_full(fd, buf, zero_size); > > > > Instead of doing lseek + qemu_write_full, can we just use > > qemu_pwritev(fd, &iov, 1, 0) with an iov set up to point to the > > appropriate amount of buf? > > Neat, will update. For stack-allocation yes, but qemu_write_full turns out to be better because of short write handling. So I'll keep it in v2. Fam
在 8/10/2017 4:01 PM, Fam Zheng 写道: > People get surprised when, after "qemu-imc create -f raw /dev/sdX", they s/qemu-imc/qemu-img/ > still see qcow2 with "qemu-img info", if previously the bdev had a qcow2 -- Thanks -Xie
On Fri, 08/11 15:44, Xie Changlong wrote: > 在 8/10/2017 4:01 PM, Fam Zheng 写道: > > People get surprised when, after "qemu-imc create -f raw /dev/sdX", they > > s/qemu-imc/qemu-img/ Thanks, will fix in v2. Fam
© 2016 - 2024 Red Hat, Inc.