| Message ID | 20110325062741.23726.36621.stgit@dhcp-91-7.nay.redhat.com.englab.nay.redhat.com |
|---|---|
| State | New |
| Headers | show |
Jason Wang <jasowang@redhat.com> wrote: > We need to keep DIR register unchanged across migration, but currently it > depends on the media_changed flags from block layer and we do not save/restore > it which could let the guest driver think the floppy have changed after > migration. To fix this, a new filed media_changed in FDrive strcutre was > introduced in order to save and restore the it from block layer through > pre_save/post_load callbacks. > > Signed-off-by: Jason Wang <jasowang@redhat.com> > --- > hw/fdc.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ > 1 files changed, 46 insertions(+), 0 deletions(-) > > diff --git a/hw/fdc.c b/hw/fdc.c > index 9fdbc75..8257d51 100644 > --- a/hw/fdc.c > +++ b/hw/fdc.c > @@ -36,6 +36,7 @@ > #include "qdev-addr.h" > #include "blockdev.h" > #include "sysemu.h" > +#include "block_int.h" > > /********************************************************/ > /* debug Floppy devices */ > @@ -82,6 +83,7 @@ typedef struct FDrive { > uint8_t max_track; /* Nb of tracks */ > uint16_t bps; /* Bytes per sector */ > uint8_t ro; /* Is read-only */ > + uint8_t media_changed; /* Is media changed */ > } FDrive; > > static void fd_init(FDrive *drv) > @@ -533,6 +535,42 @@ static CPUWriteMemoryFunc * const fdctrl_mem_write_strict[3] = { > NULL, > }; > > +static void fdrive_media_changed_pre_save(void *opaque) > +{ > + FDrive *drive = opaque; > + > + drive->media_changed = drive->bs->media_changed; > +} > + > +static int fdrive_media_changed_post_load(void *opaque, int version_id) > +{ > + FDrive *drive = opaque; > + > + drive->bs->media_changed = drive->media_changed; shouldn't this be something like: if (!drive->bs) return 1; drive->bs->media_changed = drive->media_changed. ??? Otherwise, we can can have troubles if floppy is not opened in the other side. Rest of code looks nice, thanks. Later, Juan.
On 03/25/2011 07:27 AM, Jason Wang wrote: > We need to keep DIR register unchanged across migration, but currently it > depends on the media_changed flags from block layer and we do not save/restore > it which could let the guest driver think the floppy have changed after > migration. To fix this, a new filed media_changed in FDrive strcutre was > introduced in order to save and restore the it from block layer through > pre_save/post_load callbacks. I guess you can avoid saving if the media changed flag is zero, too (which should be the common case after the guest has booted, right?). Paolo
Paolo Bonzini writes: > On 03/25/2011 07:27 AM, Jason Wang wrote: > > We need to keep DIR register unchanged across migration, but currently it > > depends on the media_changed flags from block layer and we do not save/restore > > it which could let the guest driver think the floppy have changed after > > migration. To fix this, a new filed media_changed in FDrive strcutre was > > introduced in order to save and restore the it from block layer through > > pre_save/post_load callbacks. > > I guess you can avoid saving if the media changed flag is zero, too > (which should be the common case after the guest has booted, right?). > > Paolo > Yes, zero is the common case, but the bdrv_open() called by listening qemu in dest mode would always set the media_changed to one, so we must save and restore it during migration.
On 03/28/2011 04:40 AM, Jason Wang wrote: > > On 03/25/2011 07:27 AM, Jason Wang wrote: > > > We need to keep DIR register unchanged across migration, but currently it > > > depends on the media_changed flags from block layer and we do not save/restore > > > it which could let the guest driver think the floppy have changed after > > > migration. To fix this, a new filed media_changed in FDrive strcutre was > > > introduced in order to save and restore the it from block layer through > > > pre_save/post_load callbacks. > > > > I guess you can avoid saving if the media changed flag is zero, too > > (which should be the common case after the guest has booted, right?). > > > > Paolo > > Yes, zero is the common case, but the bdrv_open() called by listening qemu in > dest mode would always set the media_changed to one, so we must save and restore > it during migration. If zero is the common case, you can do something like this to avoid saving the subsection in the common case and allow seamless migration from 0.15 to 0.14: static void fdrive_pre_save(void *opaque) { FDrive *drive = opaque; if (drive->bs != NULL) { drive->media_changed = drive->bs->media_changed; } else { drive->media_changed = 2; } } static void fdrive_pre_load(void *opaque) { FDrive *drive = opaque; /* Set the default value in case there is no subsection. */ if (drive->bs != NULL) { drive->media_changed = 0; } else { drive->media_changed = 2; } } static void fdrive_post_load(void *opaque) { /* Fail if the source machine had a disk and we don't. */ if (drive->bs == NULL && drive->media_changed != 2) { return 1; } /* Reset the field if both us and the source machine have a disk. */ if (drive->bs != NULL && drive->media_changed != 2) { /* bdrv_open() called in dest node may set the media_changed flag when trying to open floppy image. Copy it here, so that it is reset even if the subsection was not saved. */ drive->bs->media_changed = drive->media_changed; } return 0; } static bool fdrive_media_changed_needed(void *opaque) { FDrive *drive = opaque; return (drive->bs != NULL && drive->media_changed != 0); } Note that there is a small change here: if the media changed flag is 1 and you migrate 0.14 to 0.15, the destination machine will see the flag reset to 0 (while 0.14 to 0.14 will see the flag set to 1 just because that's how bdrv_open leaves it). To fix this, you can define a boolean qdev property default_migration_media_changed and use it instead of the two hard-coded "0" values. Then, register the default value of the property in hw/pc_piix.c to be 0 in pc-0.15, and 1 in pc-0.14 and older. Paolo
diff --git a/hw/fdc.c b/hw/fdc.c index 9fdbc75..8257d51 100644 --- a/hw/fdc.c +++ b/hw/fdc.c @@ -36,6 +36,7 @@ #include "qdev-addr.h" #include "blockdev.h" #include "sysemu.h" +#include "block_int.h" /********************************************************/ /* debug Floppy devices */ @@ -82,6 +83,7 @@ typedef struct FDrive { uint8_t max_track; /* Nb of tracks */ uint16_t bps; /* Bytes per sector */ uint8_t ro; /* Is read-only */ + uint8_t media_changed; /* Is media changed */ } FDrive; static void fd_init(FDrive *drv) @@ -533,6 +535,42 @@ static CPUWriteMemoryFunc * const fdctrl_mem_write_strict[3] = { NULL, }; +static void fdrive_media_changed_pre_save(void *opaque) +{ + FDrive *drive = opaque; + + drive->media_changed = drive->bs->media_changed; +} + +static int fdrive_media_changed_post_load(void *opaque, int version_id) +{ + FDrive *drive = opaque; + + drive->bs->media_changed = drive->media_changed; + + return 0; +} + +static bool fdrive_has_driver(void *opaque) +{ + FDrive *drive = opaque; + + return drive->bs != NULL; +} + +static const VMStateDescription vmstate_fdrive_media_changed = { + .name = "fdrive/media_changed", + .version_id = 1, + .minimum_version_id = 1, + .minimum_version_id_old = 1, + .pre_save = fdrive_media_changed_pre_save, + .post_load = fdrive_media_changed_post_load, + .fields = (VMStateField [] ) { + VMSTATE_UINT8(media_changed, FDrive), + VMSTATE_END_OF_LIST() + } +}; + static const VMStateDescription vmstate_fdrive = { .name = "fdrive", .version_id = 1, @@ -543,6 +581,14 @@ static const VMStateDescription vmstate_fdrive = { VMSTATE_UINT8(track, FDrive), VMSTATE_UINT8(sect, FDrive), VMSTATE_END_OF_LIST() + }, + .subsections = (VMStateSubsection []) { + { + .vmsd = &vmstate_fdrive_media_changed, + .needed = &fdrive_has_driver, + } , { + /* empty */ + } } };
We need to keep DIR register unchanged across migration, but currently it depends on the media_changed flags from block layer and we do not save/restore it which could let the guest driver think the floppy have changed after migration. To fix this, a new filed media_changed in FDrive strcutre was introduced in order to save and restore the it from block layer through pre_save/post_load callbacks. Signed-off-by: Jason Wang <jasowang@redhat.com> --- hw/fdc.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 46 insertions(+), 0 deletions(-)