Skip to content

VMState: Fix design faults, i2c port, x86 cpu port, ...

Juan Quintela qemu-devel 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 Latest49 patches 17 repliesIn review

Changes in v1 from the cover letter

I know you are not going to believe it, but VMState design have flaws:

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/49vmstate: remove const for put operationshw/hw.h hw/pci.c hw/ptimer.c +1 more+28 -280 replies
2/49vmstate: add version_id argument to post_loadexec.c hw/acpi.c hw/cirrus_vga.c +10 more+13 -130 replies
3/49vmstate: remove const from pre_save() functionsexec.c hw/fdc.c hw/hpet.c +2 more+10 -100 replies
4/49vmstate: remove ps2_kbd_load_old()hw/ps2.c+4 -120 replies
5/49vmstate: Add support for sending partial arrayshw/hw.h+12 -00 replies
6/49qdev: Add support for uint8_thw/qdev-properties.c hw/qdev.h+38 -00 replies
7/49i2c: addresses are load/save as uint8_t values, change types to reflect thishw/i2c.c hw/i2c.h hw/mips_malta.c +3 more+35 -340 replies
8/49vmstate: port i2c_bus devicehw/i2c.c+19 -120 replies
9/49vmstate: port i2c_slave devicehw/i2c.c+26 -70 replies
10/49vmstate: add uint8 arrayhw/hw.h+6 -00 replies
11/49vmstate: create VMSTATE_I2C_SLAVEhw/hw.h hw/i2c.c+12 -10 replies
12/49vmstate: port wm8750 devicehw/wm8750.c+45 -750 replies
13/49vmstate: port max7310 devicehw/max7310.c+18 -330 replies
14/49vmstate: create VMSTATE_STRUCT_POINTERhw/hw.h+9 -00 replies
15/49vmstate: port pxa2xx_i2c devicehw/pxa2xx.c+26 -280 replies
16/49vmstate: port ssd0303 devicehw/ssd0303.c+22 -430 replies
17/49vmstate: create VMSTATE_INT16_ARRAYhw/hw.h+6 -00 replies
18/49tmp105: change len and alorm to uint8_thw/tmp105.c+2 -20 replies
19/49vmstate: port tmp105 devicehw/tmp105.c+25 -310 replies
20/49twl92230: change pwrbtn_state to uint8_thw/twl92230.c+1 -10 replies
21/49vmstate: port twl92230 devicehw/twl92230.c+88 -1100 replies
22/49vmstate: add support for arrays of pointershw/hw.h savevm.c+18 -00 replies
23/49lm832x: make fields to have the same types that they are saved/loadedhw/lm832x.c+22 -190 replies
24/49vmstate: port lm832x devicehw/lm832x.c+37 -700 replies
25/49vmstate: remove i2c_slave_load/savehw/i2c.c hw/i2c.h+0 -130 replies
26/49vmstate: port audio/audio.caudio/audio.c+9 -170 replies
27/49ac97: add active to the statehw/ac97.c+20 -2217 replies
28/49vmstate: port ac97 devicehw/ac97.c+43 -550 replies
29/49x86: hflags is not modified at all, just save it directlytarget-i386/machine.c+2 -80 replies
30/49x86: make a20_mask int32_ttarget-i386/cpu.h target-i386/helper.c target-i386/machine.c+6 -110 replies
31/49x86: fpuc is uint16_t not unsigned inttarget-i386/cpu.h target-i386/machine.c+5 -70 replies
32/49x86: fpus is uint16_t not unsigned inttarget-i386/cpu.h target-i386/machine.c+10 -90 replies
33/49x86: add fptag_vmstate to the statetarget-i386/cpu.h target-i386/machine.c+10 -90 replies
34/49x86: add pending_irq_vmstate to the statetarget-i386/cpu.h target-i386/machine.c+8 -90 replies
35/49x86: add fpregs_format_vmstatetarget-i386/cpu.h target-i386/machine.c+6 -70 replies
36/49x86: mce_banks always have the same sizetarget-i386/cpu.h target-i386/helper.c target-i386/machine.c+4 -50 replies
37/49x86: send mce_banks as an arraytarget-i386/machine.c+4 -100 replies
38/49x86: mcg_cap is never 0target-i386/machine.c+8 -120 replies
39/49x86: split FPReg uniontarget-i386/cpu.h+10 -80 replies
40/49x86: split MTRRVar uniontarget-i386/cpu.h+6 -40 replies
41/49Add *TL functions to vmstatehw/hw.h+19 -00 replies
42/49x86: port segments to vmstatetarget-i386/machine.c+16 -80 replies
43/49x86: factor out cpu_pre_save()target-i386/machine.c+32 -210 replies
44/49x86: factor out cpu_pre/post_load()target-i386/machine.c+41 -200 replies
45/49x86: factor out cpu_get/put_xmm_reg()target-i386/machine.c+24 -40 replies
46/49x86: factor out cpu_get/put_mttr_var()target-i386/machine.c+24 -40 replies
47/49x86: factor out cpu_get/put_fpreg()target-i386/machine.c+137 -720 replies
48/49vmstate: Add suppot for field_exist() testhw/hw.h savevm.c+49 -170 replies
49/49x86: port cpu to vmstatetarget-i386/machine.c+238 -2530 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [Qemu-devel] [PATCH 00/49] VMState: Fix design faults, i2c port, x86 cpu port, ...

Hi

This is the whole VMState patches from this last two weeks. It has integrated the VMState design flaws and the i2c devices port.

New is:
- remove const from *_pre_save(), and port all users
- x86 cpu is ported to vmstate.
- fpregs: it allows load/save mix of 64/80 bits floats, ugly as hell (it was already ugly as hell)
- field_exist() method, now we can have optional fields
- ac97 got integrated in the buch.

Comments?

Later, Juan.

Quoted 1 line

From older series.

Subject: [PATCH v2 0/4] Fix VMState design flaws

v2:
- fix old_version field for ps2 keyboard (thanks paolo)

v1:
I know you are not going to believe it, but VMState design have flaws:

(everybody) Ooohhhhhhh!!!!

This patches fixes (some) of them:

* When we added .pre_save() and .post_save(), we have it basically to change values in the variable pointed by the "void *opaque". Without this functions, it made sense that this variables were const, with them, it means that we have to do
FooState *s = (void *)opaque
in every pre/post_save() function. Just remove the const.
* Add version_id field to post_load(). Now we can assign default values in post_load for old versions of the state (ps2_kbd as example). This one is also needed for ioapic in qemu-kvm.
* Add support for sending partial struct arrays (i.e. only some fields starting from the beggining) fdc cleanups for getting pc98 in need it.

Quoted 1 line

From Subject: [PATCH 00/20] VMState: port all i2c devices

This ports all i2c_devices to vmstate. Big changes:

- i2c->address now are uint8_t, my review of all uses is that they are always used as uint8_t (and that is the type that is passed on the value). If you know i2c, please check. Change 0002 looks big, but it is because as I was auditing the paths, I changed the types of the functios to reflect that they use uint8_t. Real part of the patch is changing the struct definitions.

- qdev: there was not support for uint8_t (the i2c addresses) in qdev, now it is I missed this change in my 1st pass (on mips_malta)

- qdev_prop_set_uint32(eeprom, "address", 0x50 + i);
+ qdev_prop_set_uint8(eeprom, "address", 0x50 + i);

And qdev complained during compilation. qdev++

- I had to fix lots of places where values of one type were sent with a different type. qemu_put_byte/qemu_put_be16/qemu_put_be32 should be just baned. Yesterday.

- tmp105: We have a winner, the 1st user of post_save(). Can anyone explain me what we need to do _anything_ after saving

     s->faults = tmp105_faultq[(s->config >> 3) & 3];		/* F */

I can't see why saving have to change faults, and no, I don't understand what that line does.

- twl922230: here we go.
It pass all the fields of a struct tm (they are ints) as uint16_t. All solutions are bad (tm):
* marshalling the struct tm in a struct tm_16bits fields, and use normal vmstate
* up version and declare all previous versions baned. No forward migration for you.
* Do the hack that I did, new type:
vmstate_hack_int32_as_uint16
local to that function, and be done with it. It is a big hack, but the function were already abusing the format.

- lm832x: I got an unused command (0xff) to be send as an error. it was using an int, -1 means an error, and a small number meaned a command. As the numbers of commands is very limited, I think this is the best solution. Notice that we were doing this already when we saved/loaded the value. Name the constant instead of -1 to make things easier.

- vmstate arrays shortened the save/load code for this series _a lot_.

Juan Quintela (49):
vmstate: remove const for put operations
vmstate: add version_id argument to post_load
vmstate: remove const from pre_save() functions
vmstate: remove ps2_kbd_load_old()
vmstate: Add support for sending partial arrays
qdev: Add support for uint8_t
i2c: addresses are load/save as uint8_t values, change types to
reflect this
vmstate: port i2c_bus device
vmstate: port i2c_slave device
vmstate: add uint8 array
vmstate: create VMSTATE_I2C_SLAVE
vmstate: port wm8750 device
vmstate: port max7310 device
vmstate: create VMSTATE_STRUCT_POINTER
vmstate: port pxa2xx_i2c device
vmstate: port ssd0303 device
vmstate: create VMSTATE_INT16_ARRAY
tmp105: change len and alorm to uint8_t
vmstate: port tmp105 device
twl92230: change pwrbtn_state to uint8_t
vmstate: port twl92230 device
vmstate: add support for arrays of pointers
lm832x: make fields to have the same types that they are saved/loaded
vmstate: port lm832x device
vmstate: remove i2c_slave_load/save
vmstate: port audio/audio.c
ac97: add active to the state
vmstate: port ac97 device
x86: hflags is not modified at all, just save it directly
x86: make a20_mask int32_t
x86: fpuc is uint16_t not unsigned int
x86: fpus is uint16_t not unsigned int
x86: add fptag_vmstate to the state
x86: add pending_irq_vmstate to the state
x86: add fpregs_format_vmstate
x86: mce_banks always have the same size
x86: send mce_banks as an array
x86: mcg_cap is never 0
x86: split FPReg union
x86: split MTRRVar union
Add *TL functions to vmstate
x86: port segments to vmstate
x86: factor out cpu_pre_save()
x86: factor out cpu_pre/post_load()
x86: factor out cpu_get/put_xmm_reg()
x86: factor out cpu_get/put_mttr_var()
x86: factor out cpu_get/put_fpreg()
vmstate: Add suppot for field_exist() test
x86: port cpu to vmstate

 audio/audio.c         |   26 +--
 exec.c                |    6 +-
 hw/ac97.c             |  136 ++++-----
 hw/acpi.c             |    2 +-
 hw/cirrus_vga.c       |    2 +-
 hw/dma.c              |    2 +-
 hw/fdc.c              |    6 +-
 hw/hpet.c             |    6 +-
 hw/hw.h               |  114 ++++++++-
 hw/i2c.c              |   64 +++--
 hw/i2c.h              |   10 +-
 hw/lm832x.c           |  148 ++++------
 hw/max7310.c          |   51 ++---
 hw/mips_malta.c       |    2 +-
 hw/pc.c               |    2 +-
 hw/pci.c              |    2 +-
 hw/piix_pci.c         |    2 +-
 hw/ps2.c              |   16 +-
 hw/ptimer.c           |    4 +-
 hw/pxa2xx.c           |   54 ++--
 hw/qdev-properties.c  |   33 +++
 hw/qdev.h             |    5 +
 hw/serial.c           |    6 +-
 hw/slavio_intctl.c    |    2 +-
 hw/smbus.c            |   18 +-
 hw/smbus.h            |   18 +-
 hw/ssd0303.c          |   65 ++---
 hw/tcx.c              |    2 +-
 hw/tmp105.c           |   60 ++---
 hw/twl92230.c         |  200 ++++++--------
 hw/vmmouse.c          |    2 +-
 hw/wm8750.c           |  120 +++-----
 savevm.c              |   87 ++++---
 target-i386/cpu.h     |   40 ++-
 target-i386/helper.c  |    9 +-
 target-i386/machine.c |  725 ++++++++++++++++++++++++++++---------------------
 36 files changed, 1083 insertions(+), 964 deletions(-)