diff options
| author | Markus Armbruster <armbru@redhat.com> | 2013-04-12 17:25:03 +0200 |
|---|---|---|
| committer | Anthony Liguori <aliguori@us.ibm.com> | 2013-04-24 13:23:51 -0500 |
| commit | 9953f8822cc316eec9962f0a2858c3439a80adec (patch) | |
| tree | bf8e999fa32255d6bac73f24f4c41ef35838fecd /include | |
| parent | 6fd028f64f662c801fd5a54d0e3a1d2baeee93ea (diff) | |
| download | qemu-9953f8822cc316eec9962f0a2858c3439a80adec.tar.gz qemu-9953f8822cc316eec9962f0a2858c3439a80adec.tar.bz2 qemu-9953f8822cc316eec9962f0a2858c3439a80adec.zip | |
pc: Kill the "use flash device for BIOS unless KVM" misfeature
Use of a flash memory device for the BIOS was added in series "[PATCH
v10 0/8] PC system flash support", commit 4732dca..1b89faf, v1.1.
Flash vs. ROM is a guest-visible difference. Thus, flash use had to
be suppressed for machine types pc-1.0 and older. This was
accomplished by adding a dummy device "pc-sysfw" with property
"rom_only":
* Non-zero rom_only means "use ROM". Default for pc-1.0 and older.
* Zero rom_only means "maybe use flash". Default for newer machines.
Not only is the dummy device ugly, it was also retroactively added to
the older machine types! Fortunately, it's not guest-visible (thus no
immediate guest ABI breakage), and has no vmstate (thus no immediate
migration breakage). Breakage occurs only if the user unwisely
enables flash by setting rom_only to zero. Patch review FAIL #1.
Why "maybe use flash"? Flash didn't (and still doesn't) work with
KVM. Therefore, rom_only=0 really means "use flash, except when KVM
is enabled, use ROM". This is a Bad Idea, because it makes enabling/
disabling KVM guest-visible. Patch review FAIL #2.
Aside: it also precludes migrating between KVM on and off, but that's
not possible for other reasons anyway.
Fix as follows:
1. Change the meaning of rom_only=0 to mean "use flash, no ifs, buts,
or maybes" for pc-i440fx-1.5 and pc-q35-1.5. Don't change anything
for older machines (to remain bug-compatible).
2. Change the default value from 0 to 1 for these machines.
Necessary, because 0 doesn't work with KVM. Once it does, we can flip
the default back to 0.
3. Don't revert the retroactive addition of device "pc-sysfw" to older
machine types. Seems not worth the trouble.
4. Add a TODO comment asking for device "pc-sysfw" to be dropped once
flash works with KVM.
Net effect is that you get a BIOS ROM again even when KVM is disabled,
just like for machines predating the introduction of flash.
To get flash instead, use "--global pc-sysfw.rom_only=0".
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1365780303-26398-4-git-send-email-armbru@redhat.com
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Diffstat (limited to 'include')
| -rw-r--r-- | include/hw/i386/pc.h | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/include/hw/i386/pc.h b/include/hw/i386/pc.h index 9bcc8193b7..14b504c74d 100644 --- a/include/hw/i386/pc.h +++ b/include/hw/i386/pc.h @@ -168,6 +168,7 @@ static inline bool isa_ne2000_init(ISABus *bus, int base, int irq, NICInfo *nd) } /* pc_sysfw.c */ +extern bool pc_sysfw_flash_vs_rom_bug_compatible; void pc_system_firmware_init(MemoryRegion *rom_memory); /* e820 types */ @@ -233,6 +234,10 @@ int e820_add_entry(uint64_t, uint64_t, uint32_t); .driver = "virtio-net-pci",\ .property = "romfile",\ .value = "pxe-virtio.rom",\ + },{\ + .driver = "pc-sysfw",\ + .property = "rom_only",\ + .value = stringify(0),\ } #endif |