[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH] hw/scsi/vmw_pvscsi.c: Fix wrong usage of gboolean types in P
From: |
Dr. David Alan Gilbert |
Subject: |
Re: [PATCH] hw/scsi/vmw_pvscsi.c: Fix wrong usage of gboolean types in PVSCSIState |
Date: |
Tue, 15 Sep 2020 20:00:09 +0100 |
User-agent: |
Mutt/1.14.6 (2020-07-11) |
* Philippe Mathieu-Daudé (philmd@redhat.com) wrote:
> Hi Amey,
>
> On 9/14/20 1:44 AM, Amey Narkhede wrote:
> > rings_info_valid, msg_ring_info_valid and use_msg fields of struct
> > PVSCSIState are using gboolean TRUE/FALSE values with the type uint8_t.
> > Change their type to bool along with the usage of initialization macro
> > VMSTATE_BOOL during initialization of vmstate_pvscsi and
> > pvscsi_properties.
> >
> > Signed-off-by: Amey Narkhede <ameynarkhede03@gmail.com>
> > ---
> > hw/scsi/vmw_pvscsi.c | 22 +++++++++++-----------
> > 1 file changed, 11 insertions(+), 11 deletions(-)
> >
> > diff --git a/hw/scsi/vmw_pvscsi.c b/hw/scsi/vmw_pvscsi.c
> > index c071e0c7aa..86f00e3d7e 100644
> > --- a/hw/scsi/vmw_pvscsi.c
> > +++ b/hw/scsi/vmw_pvscsi.c
> > @@ -123,9 +123,9 @@ typedef struct {
> > /* Collector for current command data */
> > uint32_t curr_cmd_data[PVSCSI_MAX_CMD_DATA_WORDS];
> >
> > - uint8_t rings_info_valid; /* Whether data rings initialized
> > */
> > - uint8_t msg_ring_info_valid; /* Whether message ring
> > initialized */
> > - uint8_t use_msg; /* Whether to use message ring
> > */
> > + bool rings_info_valid; /* Whether data rings initialized
> > */
> > + bool msg_ring_info_valid; /* Whether message ring initialized
> > */
> > + bool use_msg; /* Whether to use message ring
> > */
> >
> > uint8_t msi_used; /* For migration compatibility
> > */
> > PVSCSIRingInfo rings; /* Data transfer rings manager
> > */
> > @@ -349,8 +349,8 @@ pvscsi_reset_state(PVSCSIState *s)
> > s->reg_command_status = PVSCSI_COMMAND_PROCESSING_SUCCEEDED;
> > s->reg_interrupt_status = 0;
> > pvscsi_ring_cleanup(&s->rings);
> > - s->rings_info_valid = FALSE;
> > - s->msg_ring_info_valid = FALSE;
> > + s->rings_info_valid = false;
> > + s->msg_ring_info_valid = false;
> > QTAILQ_INIT(&s->pending_queue);
> > QTAILQ_INIT(&s->completion_queue);
> > }
> > @@ -792,7 +792,7 @@ pvscsi_on_cmd_setup_rings(PVSCSIState *s)
> > pvscsi_dbg_dump_tx_rings_config(rc);
> > pvscsi_ring_init_data(&s->rings, rc);
> >
> > - s->rings_info_valid = TRUE;
> > + s->rings_info_valid = true;
> > return PVSCSI_COMMAND_PROCESSING_SUCCEEDED;
> > }
> >
> > @@ -874,7 +874,7 @@ pvscsi_on_cmd_setup_msg_ring(PVSCSIState *s)
> > if (pvscsi_ring_init_msg(&s->rings, rc) < 0) {
> > return PVSCSI_COMMAND_PROCESSING_FAILED;
> > }
> > - s->msg_ring_info_valid = TRUE;
> > + s->msg_ring_info_valid = true;
> > }
> > return sizeof(PVSCSICmdDescSetupMsgRing) / sizeof(uint32_t);
> > }
> > @@ -1232,9 +1232,9 @@ static const VMStateDescription vmstate_pvscsi = {
> > VMSTATE_UINT32(curr_cmd_data_cntr, PVSCSIState),
> > VMSTATE_UINT32_ARRAY(curr_cmd_data, PVSCSIState,
> > ARRAY_SIZE(((PVSCSIState
> > *)NULL)->curr_cmd_data)),
> > - VMSTATE_UINT8(rings_info_valid, PVSCSIState),
> > - VMSTATE_UINT8(msg_ring_info_valid, PVSCSIState),
> > - VMSTATE_UINT8(use_msg, PVSCSIState),
> > + VMSTATE_BOOL(rings_info_valid, PVSCSIState),
> > + VMSTATE_BOOL(msg_ring_info_valid, PVSCSIState),
> > + VMSTATE_BOOL(use_msg, PVSCSIState),
>
> I believe this change the migration data structure. This
> area is described in "Changing migration data structures"
> in docs/devel/migration.rst.
>
> If this structure were not affected, your change would be
> a good cleanup. However changing migration can become a
> nightmare, so ... cleaning this is hard.
>
> Cc'ing Dave (a migration maintainer) as I'm not sure there
> already is a document describing easily this problem.
Well; the migration structures are...very unstructured.
It turns out that VMSTATE_BOOL already reads/writes a raw
byte to the file - so it's the same size as the VMSTATE_UINT8 at least.
Disappointingly get_bool and put_bool in vmstate-types.c don't do any
sanitation either; so you get whatever the raw value is.
glib's FALSE/TRUE is:
#define FALSE (0)
#define TRUE (!FALSE)
so that's probably fine; as long as no one does anything silly like
compare the value with TRUE or true.
Dave
> >
> > VMSTATE_UINT64(rings.rs_pa, PVSCSIState),
> > VMSTATE_UINT32(rings.txr_len_mask, PVSCSIState),
> > @@ -1255,7 +1255,7 @@ static const VMStateDescription vmstate_pvscsi = {
> > };
> >
> > static Property pvscsi_properties[] = {
> > - DEFINE_PROP_UINT8("use_msg", PVSCSIState, use_msg, 1),
> > + DEFINE_PROP_BOOL("use_msg", PVSCSIState, use_msg, true),
> > DEFINE_PROP_BIT("x-old-pci-configuration", PVSCSIState, compat_flags,
> > PVSCSI_COMPAT_OLD_PCI_CONFIGURATION_BIT, false),
> > DEFINE_PROP_BIT("x-disable-pcie", PVSCSIState, compat_flags,
> > --
> > 2.28.0
> >
> > This is my first qemu patch. Let know if there are any mistakes
> >
>
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK