| Message ID | 1464080368-29584-1-git-send-email-pl@kamp.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Tue, 05/24 10:59, Peter Lieven wrote: > at least in the path via virtio-blk the maximum size is not > restricted. > > Cc: qemu-stable@nongnu.org > Signed-off-by: Peter Lieven <pl@kamp.de> > --- > block/iscsi.c | 7 +++++++ > 1 file changed, 7 insertions(+) > > diff --git a/block/iscsi.c b/block/iscsi.c > index 2ca8e72..e7d5f7b 100644 > --- a/block/iscsi.c > +++ b/block/iscsi.c > @@ -833,6 +833,13 @@ static BlockAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, > return &acb->common; > } > > + if (acb->ioh->cmd_len > SCSI_CDB_MAX_SIZE) { > + error_report("iSCSI: ioctl error CDB exceeds max size (%d > %d)", > + acb->ioh->cmd_len, SCSI_CDB_MAX_SIZE); > + qemu_aio_unref(acb); > + return NULL; > + } > + > acb->task = malloc(sizeof(struct scsi_task)); > if (acb->task == NULL) { > error_report("iSCSI: Failed to allocate task for scsi command. %s", Is it better to invoke the cb and report -EINVAL to the caller? Fam
Am 31.05.2016 um 08:44 hat Fam Zheng geschrieben: > On Tue, 05/24 10:59, Peter Lieven wrote: > > at least in the path via virtio-blk the maximum size is not > > restricted. > > > > Cc: qemu-stable@nongnu.org > > Signed-off-by: Peter Lieven <pl@kamp.de> > > --- > > block/iscsi.c | 7 +++++++ > > 1 file changed, 7 insertions(+) > > > > diff --git a/block/iscsi.c b/block/iscsi.c > > index 2ca8e72..e7d5f7b 100644 > > --- a/block/iscsi.c > > +++ b/block/iscsi.c > > @@ -833,6 +833,13 @@ static BlockAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, > > return &acb->common; > > } > > > > + if (acb->ioh->cmd_len > SCSI_CDB_MAX_SIZE) { > > + error_report("iSCSI: ioctl error CDB exceeds max size (%d > %d)", > > + acb->ioh->cmd_len, SCSI_CDB_MAX_SIZE); > > + qemu_aio_unref(acb); > > + return NULL; > > + } > > + > > acb->task = malloc(sizeof(struct scsi_task)); > > if (acb->task == NULL) { > > error_report("iSCSI: Failed to allocate task for scsi command. %s", > > Is it better to invoke the cb and report -EINVAL to the caller? You need to implement the BH manually then. The difference is -EINVAL vs. -ENOTSUP, which don't result in a different guest behaviour. So I think returning NULL is simpler and therefore better. Kevin
Am 31.05.2016 um 08:44 hat Fam Zheng geschrieben: > On Tue, 05/24 10:59, Peter Lieven wrote: > > at least in the path via virtio-blk the maximum size is not > > restricted. > > > > Cc: qemu-stable@nongnu.org > > Signed-off-by: Peter Lieven <pl@kamp.de> > > --- > > block/iscsi.c | 7 +++++++ > > 1 file changed, 7 insertions(+) > > > > diff --git a/block/iscsi.c b/block/iscsi.c > > index 2ca8e72..e7d5f7b 100644 > > --- a/block/iscsi.c > > +++ b/block/iscsi.c > > @@ -833,6 +833,13 @@ static BlockAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, > > return &acb->common; > > } > > > > + if (acb->ioh->cmd_len > SCSI_CDB_MAX_SIZE) { > > + error_report("iSCSI: ioctl error CDB exceeds max size (%d > %d)", > > + acb->ioh->cmd_len, SCSI_CDB_MAX_SIZE); > > + qemu_aio_unref(acb); > > + return NULL; > > + } > > + > > acb->task = malloc(sizeof(struct scsi_task)); > > if (acb->task == NULL) { > > error_report("iSCSI: Failed to allocate task for scsi command. %s", > > Is it better to invoke the cb and report -EINVAL to the caller? By the way, when returning NULL, it looks like bdrv_co_do_ioctl() leaks its BdrvIoctlCompletionData. This code was introduced by your commit 5c5ae76a ("block: Emulate bdrv_ioctl with bdrv_aio_ioctl and track both"). Kevin
On Tue, 05/31 09:35, Kevin Wolf wrote: > Am 31.05.2016 um 08:44 hat Fam Zheng geschrieben: > > On Tue, 05/24 10:59, Peter Lieven wrote: > > > at least in the path via virtio-blk the maximum size is not > > > restricted. > > > > > > Cc: qemu-stable@nongnu.org > > > Signed-off-by: Peter Lieven <pl@kamp.de> > > > --- > > > block/iscsi.c | 7 +++++++ > > > 1 file changed, 7 insertions(+) > > > > > > diff --git a/block/iscsi.c b/block/iscsi.c > > > index 2ca8e72..e7d5f7b 100644 > > > --- a/block/iscsi.c > > > +++ b/block/iscsi.c > > > @@ -833,6 +833,13 @@ static BlockAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, > > > return &acb->common; > > > } > > > > > > + if (acb->ioh->cmd_len > SCSI_CDB_MAX_SIZE) { > > > + error_report("iSCSI: ioctl error CDB exceeds max size (%d > %d)", > > > + acb->ioh->cmd_len, SCSI_CDB_MAX_SIZE); > > > + qemu_aio_unref(acb); > > > + return NULL; > > > + } > > > + > > > acb->task = malloc(sizeof(struct scsi_task)); > > > if (acb->task == NULL) { > > > error_report("iSCSI: Failed to allocate task for scsi command. %s", > > > > Is it better to invoke the cb and report -EINVAL to the caller? > > You need to implement the BH manually then. The difference is -EINVAL > vs. -ENOTSUP, which don't result in a different guest behaviour. So I > think returning NULL is simpler and therefore better. Makes sense. Thanks for explaining! Reviewed-by: Fam Zheng <famz@redhat.com>
On Tue, 05/31 09:39, Kevin Wolf wrote: > Am 31.05.2016 um 08:44 hat Fam Zheng geschrieben: > > On Tue, 05/24 10:59, Peter Lieven wrote: > > > at least in the path via virtio-blk the maximum size is not > > > restricted. > > > > > > Cc: qemu-stable@nongnu.org > > > Signed-off-by: Peter Lieven <pl@kamp.de> > > > --- > > > block/iscsi.c | 7 +++++++ > > > 1 file changed, 7 insertions(+) > > > > > > diff --git a/block/iscsi.c b/block/iscsi.c > > > index 2ca8e72..e7d5f7b 100644 > > > --- a/block/iscsi.c > > > +++ b/block/iscsi.c > > > @@ -833,6 +833,13 @@ static BlockAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, > > > return &acb->common; > > > } > > > > > > + if (acb->ioh->cmd_len > SCSI_CDB_MAX_SIZE) { > > > + error_report("iSCSI: ioctl error CDB exceeds max size (%d > %d)", > > > + acb->ioh->cmd_len, SCSI_CDB_MAX_SIZE); > > > + qemu_aio_unref(acb); > > > + return NULL; > > > + } > > > + > > > acb->task = malloc(sizeof(struct scsi_task)); > > > if (acb->task == NULL) { > > > error_report("iSCSI: Failed to allocate task for scsi command. %s", > > > > Is it better to invoke the cb and report -EINVAL to the caller? > > By the way, when returning NULL, it looks like bdrv_co_do_ioctl() > leaks its BdrvIoctlCompletionData. This code was introduced by your > commit 5c5ae76a ("block: Emulate bdrv_ioctl with bdrv_aio_ioctl and > track both"). I'll send a patch to fix it now. Fam
diff --git a/block/iscsi.c b/block/iscsi.c index 2ca8e72..e7d5f7b 100644 --- a/block/iscsi.c +++ b/block/iscsi.c @@ -833,6 +833,13 @@ static BlockAIOCB *iscsi_aio_ioctl(BlockDriverState *bs, return &acb->common; } + if (acb->ioh->cmd_len > SCSI_CDB_MAX_SIZE) { + error_report("iSCSI: ioctl error CDB exceeds max size (%d > %d)", + acb->ioh->cmd_len, SCSI_CDB_MAX_SIZE); + qemu_aio_unref(acb); + return NULL; + } + acb->task = malloc(sizeof(struct scsi_task)); if (acb->task == NULL) { error_report("iSCSI: Failed to allocate task for scsi command. %s",
at least in the path via virtio-blk the maximum size is not restricted. Cc: qemu-stable@nongnu.org Signed-off-by: Peter Lieven <pl@kamp.de> --- block/iscsi.c | 7 +++++++ 1 file changed, 7 insertions(+)