More precisely, the "x-file-slots" count is bumped for all machine types
that:
(a) use fw_cfg, and
(b) are not versioned (hence migration is not expected to work for them
across QEMU releases anyway), or have version 2.9.
This affects machine types implemented in the following source files:
- "hw/arm/virt.c". The "virt-*" machine type is versioned, and the <= 2.8
versions already depend on HW_COMPAT_2_8 (see commit e353aac51b).
Therefore adding the "x-file-slots" compat values to HW_COMPAT_2_8
suffices.
- "hw/i386/pc.c". The "pc-i440fx-*" (including "pc-*") and "pc-q35-*"
machine types are versioned. Modifying HW_COMPAT_2_8 is sufficient here
too (see commit "pc: Add 2.9 machine-types"). The "isapc" machtype is
not versioned. The "xenfv" machine type, which uses fw_cfg for direct
kernel booting, is also not versioned.
- "hw/ppc/mac_newworld.c". The "mac99" machine type is not versioned.
- "hw/ppc/mac_oldworld.c". The "g3beige" machine type is not versioned.
- "hw/sparc/sun4m.c". None of the 9 machine types defined in this file
appear versioned.
- "hw/sparc64/sun4u.c". None of the 3 machine types defined in this file
appear versioned.
Cc: "Gabriel L. Somlo" <somlo@cmu.edu>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Alexander Graf <agraf@suse.de>
Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Artyom Tarasenko <atar4qemu@gmail.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
We'd like to raise the value of FW_CFG_FILE_SLOTS. Doing it naively could
lead to problems with backward migration: a more recent QEMU (running an
older machine type) would allow the guest, in fw_cfg_select(), to select a
high key value that is unavailable in the same machine type implemented by
the older (target) QEMU. On the target host, fw_cfg_data_read() for
example could dereference nonexistent entries.
As first step, size the FWCfgState.entries[*] and FWCfgState.entry_order
arrays dynamically. All three array sizes will be influenced by the new
field FWCfgState.file_slots (and matching device property).
Make the following changes:
- Replace the FW_CFG_FILE_SLOTS macro with FW_CFG_FILE_SLOTS_MIN (minimum
count of fw_cfg file slots) in the header file. The value remains 0x10.
- Replace all uses of FW_CFG_FILE_SLOTS with a helper function called
fw_cfg_file_slots(), returning the new property.
- Eliminate the macro FW_CFG_MAX_ENTRY, and replace all its uses with a
helper function called fw_cfg_max_entry().
- In the MMIO- and IO-mapped realize functions both, allocate all three
arrays dynamically, based on the new property.
- The new property defaults to FW_CFG_FILE_SLOTS_MIN. This is going to be
customized in the following patches.
Cc: "Gabriel L. Somlo" <somlo@cmu.edu>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Gabriel Somlo <somlo@cmu.edu>
Tested-by: Gabriel Somlo <somlo@cmu.edu>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
According to ISO C99 / N1256 (referenced in HACKING):
> 6.5.8 Relational operators
>
> 4 For the purposes of these operators, a pointer to an object that is
> not an element of an array behaves the same as a pointer to the first
> element of an array of length one with the type of the object as its
> element type.
>
> 5 When two pointers are compared, the result depends on the relative
> locations in the address space of the objects pointed to. If two
> pointers to object or incomplete types both point to the same object,
> or both point one past the last element of the same array object, they
> compare equal. If the objects pointed to are members of the same
> aggregate object, pointers to structure members declared later compare
> greater than pointers to members declared earlier in the structure,
> and pointers to array elements with larger subscript values compare
> greater than pointers to elements of the same array with lower
> subscript values. All pointers to members of the same union object
> compare equal. If the expression /P/ points to an element of an array
> object and the expression /Q/ points to the last element of the same
> array object, the pointer expression /Q+1/ compares greater than /P/.
> In all other cases, the behavior is undefined.
Our AddressSpace objects are allocated generally individually, and kept in
the "address_spaces" linked list, so we mustn't compare their addresses
with relops.
Convert the pointers subjected to the relop in rom_order_compare() to
"uintptr_t":
> 7.18.1.4 Integer types capable of holding object pointers
>
> 1 [...]
>
> The following type designates an unsigned integer type with the
> property that any valid pointer to void can be converted to this type,
> then converted back to pointer to void, and the result will compare
> equal to the original pointer:
>
> /uintptr_t/
>
> These types are optional.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Alistair Francis <alistair.francis@xilinx.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: qemu-devel@nongnu.org
Fixes: 3e76099aac
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
* Commit 3e76099aac ("loader: Allow a custom AddressSpace when loading
ROMs") introduced the "Rom.as" field:
(1) It modified the utility callers of rom_insert() to take "as" as a
new parameter from *their* callers, and set "rom->as" from that
parameter. The functions covered were rom_add_file() and
rom_add_elf_program().
(2) It also modified rom_insert() itself, to auto-assign
"&address_space_memory", in case the external caller passed -- and
the utility caller forwarded -- as=NULL.
Except, commit 3e76099aac forgot to update the third utility caller of
rom_insert(), under point (1), namely rom_add_blob().
* Later, commit 5e774eb3bd ("loader: Add AddressSpace loading support
to uImages") added the load_uimage_as() function, and the
rom_add_blob_fixed_as() function-like macro, with the necessary changes
elsewhere to propagate the new "as" parameter to rom_add_blob():
load_uimage_as()
load_uboot_image()
rom_add_blob_fixed_as()
rom_add_blob()
At this point, the signature (and workings) of rom_add_blob() had been
broken already, and the rom_add_blob_fixed_as() macro passed its "_as"
parameter to rom_add_blob() as "callback_opaque". Given that the
"fw_callback" parameter itself was set to NULL (correctly), this did no
additional damage (the opaque arg would never be used), but ultimately
it broke the new functionality of load_uimage_as().
* The load_uimage_as() function would be put to use in one of the later
patches, commit e481a1f63c ("generic-loader: Add a generic loader").
* We can fix this only in a unified patch now. Append "AddressSpace *as"
to the signature of rom_add_blob(), and handle the new parameter. Pass
NULL from all current callers, except from rom_add_blob_fixed_as(),
where "_as" has to be bumped to the proper position.
* Note that rom_add_file() rejects the case when both "mr" and "as" are
passed in as non-NULL. The action that this is apparently supposed to
prevent is the
rom->mr = mr;
assignment (that's the only place where the "mr" parameter is used in
rom_add_file()). In rom_add_blob() though, we have no "mr" parameter,
and the actions done on the fw_cfg branch:
if (fw_file_name && fw_cfg) {
if (mc->rom_file_has_mr) {
data = rom_set_mr(rom, OBJECT(fw_cfg), devpath);
mr = rom->mr;
} else {
data = rom->data;
}
reflect those that are performed by rom_add_file() too (with mr==NULL):
if (rom->fw_file && fw_cfg) {
if ((!option_rom || mc->option_rom_has_mr) &&
mc->rom_file_has_mr) {
data = rom_set_mr(rom, OBJECT(fw_cfg), devpath);
} else {
data = rom->data;
}
Hence we need no additional restrictions in rom_add_blob().
* Stable is not affected as both problematic commits appeared first in
v2.8.0-rc0.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Alistair Francis <alistair.francis@xilinx.com>
Cc: Igor Mammedov <imammedo@redhat.com>
Cc: Michael Walle <michael@walle.cc>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Shannon Zhao <zhaoshenglong@huawei.com>
Cc: qemu-arm@nongnu.org
Cc: qemu-devel@nongnu.org
Fixes: 3e76099aac
Fixes: 5e774eb3bd
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@xilinx.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
At the moment the following QEMU command line triggers an assertion
failure (minimal reproducer by Cole):
qemu-system-aarch64 \
-machine virt-2.6,accel=tcg \
-nodefaults \
-no-user-config \
-nographic -monitor stdio \
-device virtio-scsi-device,id=scsi0 \
-device virtio-scsi-device,id=scsi1 \
-drive file=foo.img,format=raw,if=none,id=d0 \
-device scsi-hd,bus=scsi0.0,drive=d0 \
-drive file=foo.img,format=raw,if=none,id=d1 \
-device scsi-hd,bus=scsi1.0,drive=d1
qemu-system-aarch64: migration/savevm.c:615:
vmstate_register_with_alias_id:
Assertion `!se->compat || se->instance_id == 0' failed.
The reason is that the vmstate sections for the two scsi-hd devices are
not uniquely identifiable by name.
The direct parent buses of the scsi-hd devices -- scsi0.0 and scsi1.0 --
support the BusClass.get_dev_path member function. scsibus_get_dev_path()
formats a device path prefix with the help of its topologically parent
bus, and then appends the chan:id:lun triplet to it. For both scsi-hd
devices, this triplet is 0:0:0.
(Here we use "device path" in the QEMU migration sense, for vmstate
section identification, not in the OFW or UEFI device path senses.)
The virtio-scsi HBA is plugged into the virtio-mmio bus (implemented by
the internal VirtIOMMIOProxy device). This bus class
(TYPE_VIRTIO_MMIO_BUS) inherits, as its get_dev_path() member function,
the virtio_bus_get_dev_path() method from its parent class
(TYPE_VIRTIO_BUS).
virtio_bus_get_dev_path() does not format any kind of device address on
its own; "virtio addresses" are transport-specific. Therefore
virtio_bus_get_dev_path() asks the topologically parent bus of the proxy
object (implementing the specific virtio transport) to format the address
of the proxy object.
(For virtio-pci devices (where the proxy is an instance of VirtIOPCIProxy,
plugged into a PCI bus), this ends up in pcibus_get_dev_path().)
However, VirtIOMMIOProxy is usually (in practice: always) plugged into
"main-system-bus", the singleton TYPE_SYSTEM_BUS object. This BusClass
does not support formatting QEMU vmstate device paths at all (as
SysBusDevice objects can have zero or more IO ports and zero or more MMIO
regions). Hence the formatting request delegated from
virtio_bus_get_dev_path() gets answered with NULL.
The end result is that the two scsi-hd devices end up with the same device
path "0:0:0", which triggers the assert.
We can solve this by recognizing that virtio-mmio transports are
distinguished from each other by their base addresses in MMIO address
space. Implement virtio_mmio_bus_get_dev_path() as follows:
(1) The virtio device whose devpath is to be formatted resides on a
virtio-mmio bus that is implemented by a VirtIOMMIOProxy object. Ask
the parent bus of VirtIOMMIOProxy to format the device path of
VirtIOMMIOProxy, as a path prefix. (This is identical to what
virtio_bus_get_dev_path() does.)
(2) Append the base address of VirtIOMMIOProxy to the device path, such
as:
- virtio-mmio@000000000a003e00,
- virtio-mmio@000000000a003c00.
Given that these device paths are placed in the migration stream, step (2)
above, if done unconditionally, would break migration. So make that step
conditional on a new VirtIOMMIOProxy property, which is enabled for 2.7
machine types and later.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Cole Robinson <crobinso@redhat.com>
Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: Kevin Zhao <kevin.zhao@linaro.org>
Cc: Peter Maydell <peter.maydell@linaro.org>
Cc: Tom Hanson <thomas.hanson@linaro.org>
Reported-by: Kevin Zhao <kevin.zhao@linaro.org>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-id: 1467739394-28357-1-git-send-email-lersek@redhat.com
Fixes: https://bugs.launchpad.net/qemu/+bug/1594239
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The start_xmit() and e1000_receive_iov() functions implement DMA transfers
iterating over a set of descriptors that the guest's e1000 driver
prepares:
- the TDLEN and RDLEN registers store the total size of the descriptor
area,
- while the TDH and RDH registers store the offset (in whole tx / rx
descriptors) into the area where the transfer is supposed to start.
Each time a descriptor is processed, the TDH and RDH register is bumped
(as appropriate for the transfer direction).
QEMU already contains logic to deal with bogus transfers submitted by the
guest:
- Normally, the transmit case wants to increase TDH from its initial value
to TDT. (TDT is allowed to be numerically smaller than the initial TDH
value; wrapping at or above TDLEN bytes to zero is normal.) The failsafe
that QEMU currently has here is a check against reaching the original
TDH value again -- a complete wraparound, which should never happen.
- In the receive case RDH is increased from its initial value until
"total_size" bytes have been received; preferably in a single step, or
in "s->rxbuf_size" byte steps, if the latter is smaller. However, null
RX descriptors are skipped without receiving data, while RDH is
incremented just the same. QEMU tries to prevent an infinite loop
(processing only null RX descriptors) by detecting whether RDH assumes
its original value during the loop. (Again, wrapping from RDLEN to 0 is
normal.)
What both directions miss is that the guest could program TDLEN and RDLEN
so low, and the initial TDH and RDH so high, that these registers will
immediately be truncated to zero, and then never reassume their initial
values in the loop -- a full wraparound will never occur.
The condition that expresses this is:
xdh_start >= s->mac_reg[XDLEN] / sizeof(desc)
i.e., TDH or RDH start out after the last whole rx or tx descriptor that
fits into the TDLEN or RDLEN sized area.
This condition could be checked before we enter the loops, but
pci_dma_read() / pci_dma_write() knows how to fill in buffers safely for
bogus DMA addresses, so we just extend the existing failsafes with the
above condition.
This is CVE-2016-1981.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Petr Matousek <pmatouse@redhat.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Prasad Pandit <ppandit@redhat.com>
Cc: Michael Roth <mdroth@linux.vnet.ibm.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: qemu-stable@nongnu.org
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1296044
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Background on QEMU boot indices
-------------------------------
Normally, the "bootindex" property is configured for bootable devices
with:
DEVICE_instance_init()
device_add_bootindex_property(..., "bootindex", ...)
object_property_add(..., device_get_bootindex,
device_set_bootindex, ...)
and when the bootindex is set on the QEMU command line, with
-device DEVICE,...,bootindex=N
the setter that was configured above is invoked:
device_set_bootindex()
/* parse boot index */
visit_type_int32()
/* verify unicity */
check_boot_index()
/* store parsed boot index */
...
/* insert device path to boot order */
add_boot_device_path()
In the last step, add_boot_device_path() ensures that an OpenFirmware
device path will show up in the "bootorder" fw_cfg file, at a position
corresponding to the device's boot index. Thus guest firmware (SeaBIOS and
OVMF) can try to boot off the device with the right priority.
NVMe boot index
---------------
In QEMU commit 33739c7129,
nvma: ide: add bootindex to qom property
the following generic setters / getters:
- device_set_bootindex()
- device_get_bootindex()
were open-coded for NVMe, under the names
- nvme_set_bootindex()
- nvme_get_bootindex()
Plus nvme_instance_init() was added to configure the "bootindex" property
manually, designating the open-coded getter & setter, rather than calling
device_add_bootindex_property().
Crucially, nvme_set_bootindex() avoided the final add_boot_device_path()
call. This fact is spelled out in the message of commit 33739c7129, and
it was presumably the entire reason for all of the code duplication.
Now, Vladislav filed an RFE for OVMF
<https://github.com/tianocore/edk2/issues/48>; OVMF should boot off NVMe
devices. It is simple to build edk2's existent NvmExpressDxe driver into
OVMF, but the boot order matching logic in OVMF can only handle NVMe if
the "bootorder" fw_cfg file includes such devices.
Therefore this patch converts the NVMe device model to
device_set_bootindex() all the way.
Device paths
------------
device_set_bootindex() accepts an optional parameter called "suffix". When
present, it is expected to take the form of an OpenFirmware device path
node, and it gets appended as last node to the otherwise auto-generated
OFW path.
For NVMe, the auto-generated part is
/pci@i0cf8/pci8086,5845@6[,1]
^ ^ ^ ^
| | PCI slot and (present when nonzero)
| | function of the NVMe controller, both hex
| "driver name" component, built from PCI vendor & device IDs
PCI root at system bus port, PIO
to which here we append the suffix
/namespace@1,0
^ ^
| big endian (MSB at lowest address) numeric interpretation
| of the 64-bit IEEE Extended Unique Identifier, aka EUI-64,
| hex
32-bit NVMe namespace identifier, aka NSID, hex
resulting in the OFW device path
/pci@i0cf8/pci8086,5845@6[,1]/namespace@1,0
The reason for including the NSID and the EUI-64 is that an NVMe device
can in theory produce several different namespaces (distinguished by
NSID). Additionally, each of those may (optionally) have an EUI-64 value.
For now, QEMU only provides namespace 1.
Furthermore, QEMU doesn't even represent the EUI-64 as a standalone field;
it is embedded (and left unused) inside the "NvmeIdNs.res30" array, at the
last eight bytes. (Which is fine, since EUI-64 can be left zero-filled if
unsupported by the device.)
Based on the above, we set the "unit address" part of the last
("namespace") node to fixed "1,0".
OVMF will then map the above OFW device path to the following UEFI device
path fragment, for boot order processing:
PciRoot(0x0)/Pci(0x6,0x1)/NVMe(0x1,00-00-00-00-00-00-00-00)
^ ^ ^ ^ ^ ^
| | | | | octets of the EUI-64 in address order
| | | | NSID
| | | NVMe namespace messaging device path node
| PCI slot and function
PCI root bridge
Cc: Keith Busch <keith.busch@intel.com> (supporter:nvme)
Cc: Kevin Wolf <kwolf@redhat.com> (supporter:Block layer core)
Cc: qemu-block@nongnu.org (open list:nvme)
Cc: Gonglei <arei.gonglei@huawei.com>
Cc: Vladislav Vovchenko <vladislav.vovchenko@sk.com>
Cc: Feng Tian <feng.tian@intel.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Kevin O'Connor <kevin@koconnor.net>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Acked-by: Gonglei <arei.gonglei@huawei.com>
Acked-by: Keith Busch <keith.busch@intel.com>
Tested-by: Vladislav Vovchenko <vladislav.vovchenko@sk.com>
Message-id: 1453850483-27511-1-git-send-email-lersek@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The ACPI specification (minimally versions 1.0b through 6.0) define the
FADT.CENTURY field as:
The RTC CMOS RAM index to the century of data value (hundred and
thousand year decimals). If this field contains a zero, then the RTC
centenary feature is not supported. If this field has a non-zero value,
then this field contains an index into RTC RAM space that OSPM can use
to program the centenary field.
The x86 targets generate ACPI payload, emulate an RTC
(CONFIG_MC146818RTC), and that RTC supports the "centenary feature" (see
occurrences of RTC_CENTURY in cmos_ioport_write() and cmos_ioport_read()
in "hw/timer/mc146818rtc.c".)
However, FADT.CENTURY is left at zero currently:
[06Ch 0108 1] RTC Century Index : 00
which -- according to analysis done by Ruiyu Ni at Intel -- should cause
Linux and Windows 8+ to think the RTC centenary feature is unavailable,
and cause Windows 7 to (incorrectly) assume that the offset to use is
constant 0x32. (0x32 happens to be the right value on QEMU, but Windows 7
is wrong to assume anything at all).
Exposing the right nonzero offset in FADT.CENTURY informs Linux and
Windows 8+ about the right capabilities of the hardware, plus it retrofits
our FADT to Windows 7's behavior.
Regression tested with the following guests (all UEFI installs):
- i386 Q35: Fedora 21 ("Fedlet" edition)
- x86_64:
- i440fx:
- Fedora 21
- RHEL 6 and 7
- Windows 7 and 10
- Windows Server 2008 R2 and 2012 R2
- Q35:
- Fedora 22
- Windows 8.1
Cc: "Michael S. Tsirkin" <mst@redhat.com> (supporter:ACPI/SMBIOS)
Cc: Igor Mammedov <imammedo@redhat.com> (supporter:ACPI/SMBIOS)
Cc: Paolo Bonzini <pbonzini@redhat.com> (maintainer:X86)
Cc: Richard Henderson <rth@twiddle.net> (maintainer:X86)
Cc: Eduardo Habkost <ehabkost@redhat.com> (maintainer:X86)
Cc: Ruiyu Ni <ruiyu.ni@intel.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 4d00636e97 ("ich9: Add the lpc chip", Nov 14 2012) added the
ich9_apm_ctrl_changed() ioport write callback function such that it would
inject the SMI, in response to a write to the APM_CNT register, on the
first CPU, invariably.
Since this register is used by guest code to trigger an SMI synchronously,
the interrupt should be injected on the VCPU that is performing the write.
apm_ioport_writeb() is the .write callback of the "apm_ops"
MemoryRegionOps [hw/isa/apm.c]; it is parametrized to call
ich9_apm_ctrl_changed() by ich9_lpc_init() [hw/isa/lpc_ich9.c], via
apm_init(). Therefore this change affects no other board.
ich9_generate_smi() is an unrelated function that is called by the TCO
watchdog; a watchdog is likely in its right to (asynchronously) inject
interrupts on the first CPU only.
This patch allows the combined edk2/OVMF SMM driver stack to work with
multiple VCPUs on TCG, using both qemu-system-i386 and qemu-system-x86_64.
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Michael Kinney <michael.d.kinney@intel.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
The current trace prototypes and (matching) trace calls lead to
"unorthodox" PCI BDF notation in at least the stderr trace backend. For
example, the four BARs of a QXL video card at 00:01.0 (bus 0, slot 1,
function 0) are traced like this (PID and timestamps removed):
pci_update_mappings_add d=0x7f14a73bf890 00:00.1 0,0x84000000+0x4000000
pci_update_mappings_add d=0x7f14a73bf890 00:00.1 1,0x80000000+0x4000000
pci_update_mappings_add d=0x7f14a73bf890 00:00.1 2,0x88200000+0x2000
pci_update_mappings_add d=0x7f14a73bf890 00:00.1 3,0xd060+0x20
The slot and function values are in reverse order.
Stick with the conventional BDF notation.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Don Koch <dkoch@verizon.com>
Cc: qemu-trivial@nongnu.org
Fixes: 7828d75045
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Thanks to the last patch, pc_cmos_init() doesn't need the (optional)
board-default FDC any longer as an input parameter. Update
pc_basic_device_init() not to hand it back to pc_init1() / pc_q35_init(),
and update the latter not to carry the FDC to pc_cmos_init(). This
simplifies the code.
pc_init1() | pc_q35_init()
pc_basic_device_init()
pc_cmos_init()
Cc: Jan Tomko <jtomko@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
With the pc-q35-2.4 machine type, if the user creates an ISA FDC manually:
-device isa-fdc,driveA=drive-fdc0-0-0 \
-drive file=...,if=none,id=drive-fdc0-0-0,format=raw
then the board-default FDC will be skipped, and only the explicitly
requested FDC will exist. qtree-wise, this is correct; however such an FDC
is currently not registered in the CMOS, because that code is only reached
for the board-default FDC.
The pc_cmos_init_late() one-shot reset handler -- one-shot because the
CMOS is not reprogrammed during warm reset -- should search for any ISA
FDC devices, created implicitly (by board code) or explicitly, and set the
CMOS accordingly to the ISA FDC(s) with iobase=0x3f0:
- if there is no such FDC, report both drives absent,
- if there is exactly one such FDC, report its drives in the CMOS,
- if there are more than one such FDCs, then pick one (it is not specified
which one), and print a warning about the ambiguity.
Cc: Jan Tomko <jtomko@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Reported-by: Jan Tomko <jtomko@redhat.com>
Suggested-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Extract the pc_cmos_init_floppy() function from pc_cmos_init(). The
function sets two RTC registers: floppy drive types (0x10), overwriting
the earlier value in there), and REG_EQUIPMENT_BYTE (0x14), setting bits
in the prior value.
Cc: Jan Tomko <jtomko@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
We have agreed that OpenFirmware device paths in the "bootorder" fw_cfg
file should follow the pattern
/pci@i0cf8,%x/...
for devices that live behind an extra root bus. The extra root bus in
question is the %x'th among the extra root buses. (In other words, %x
gives the position of the affected extra root bus relative to the other
extra root buses, in bus_nr order.) %x starts at 1, and is formatted in
hex.
The portion of the unit address that comes before the comma is dynamically
taken from the main host bridge, similarly to sysbus_get_fw_dev_path().
Cc: Kevin O'Connor <kevin@koconnor.net>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The sysbus_get_fw_dev_path() function formats OpenFirmware device path
nodes ("driver-name@unit-address") for sysbus devices. The first choice
for "unit-address" is the base address of the device's first MMIO region.
The second choice is its first IO port.
However, if two sysbus devices with the same "driver-name" lack both MMIO
and PIO resources, then there is no good way to distinguish them based on
their OFW nodes, because in this case unit-address is omitted completely
for both devices. An example is TYPE_PXB_HOST ("pxb-host").
For the sake of such devices, introduce the explicit_ofw_unit_address()
"virtual member function". With this function, each sysbus device in the
same SysBusDeviceClass can state its own address.
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Tested-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
OVMF downloads the ACPI linker/loader script from QEMU when the edk2 PCI
Bus driver globally signals the firmware that PCI enumeration and resource
allocation have completed. At this point QEMU regenerates the ACPI payload
in an fw_cfg read callback, and this is when the PXB's _CRS gets
populated.
Unfortunately, when this happens, the PCI_COMMAND_MEMORY bit is clear in
the root bus's command register, *unlike* under SeaBIOS. The consequences
unfold as follows:
- When build_crs() fetches dev->io_regions[i].addr, it is all-bits-one,
because pci_update_mappings() --> pci_bar_address() calculated it as
PCI_BAR_UNMAPPED, due to the PCI_COMMAND_MEMORY bit being clear.
- Consequently, the SHPC MMIO BAR (bar 0) of the bridge is not added to
the _CRS, *despite* having been programmed in PCI config space.
- Similarly, the SHPC MMIO BAR of the PXB is not removed from the main
root bus's DWordMemory descriptor.
- Guest OSes (Linux and Windows alike) notice the pre-programmed SHPC BAR
within the PXB's config space, and notice that it conflicts with the
main root bus's memory resource descriptors. Linux reports
pci 0000:04:00.0: BAR 0: can't assign mem (size 0x100)
pci 0000:04:00.0: BAR 0: trying firmware assignment [mem
0x88200000-0x882000ff 64bit]
pci 0000:04:00.0: BAR 0: [mem 0x88200000-0x882000ff 64bit] conflicts
with PCI Bus 0000:00 [mem
0x88200000-0xfebfffff]
While Windows Server 2012 R2 reports
https://technet.microsoft.com/en-us/library/cc732199%28v=ws.10%29.aspx
This device cannot find enough free resources that it can use. If you
want to use this device, you will need to disable one of the other
devices on this system. (Code 12)
This issue was apparently encountered earlier, see the "hack" in:
https://lists.nongnu.org/archive/html/qemu-devel/2015-01/msg02983.html
and the current hole-punching logic in build_crs() and build_ssdt() is
probably supposed to remedy exactly that problem -- however, for OVMF they
don't work, because at the end of the PCI enumeration and resource
allocation, which cues the ACPI linker/loader client, the command register
is clear.
The "shpc" property of "pci-bridge", introduced in the previous patches,
allows us to disable the standard hotplug controller cleanly, eliminating
the SHPC bar and the conflict.
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
In the PCI expander bridge, we will want to disable those features of
pci-bridge that relate to SHPC (standard hotplug controller):
- SHPC bar and underlying MemoryRegion
- interrupt (INTx or MSI)
- effective hotplug callbacks
- other SHPC hooks (initialization, cleanup, migration etc)
Introduce a new feature request bit in the PCIBridgeDev.flags field, and
turn off the above if the bit is explicitly cleared.
Suggested-by: Michael S. Tsirkin <mst@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Change the signature of the function-like macro SHPC_VMSTATE(), so that we
can produce and expect this field conditionally in the migration stream,
starting with an upcoming patch.
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
There is no _TEST() variant of VMSTATE_BUFFER_UNSAFE_INFO() yet, but we'll
soon need it. Introduce it and rebase the original
VMSTATE_BUFFER_UNSAFE_INFO() on top.
The parameter order of the new function-like macro follows that of
VMSTATE_SINGLE_TEST(): "_test" is introduced between "_state" and
"_version".
Cc: Juan Quintela <quintela@redhat.com>
Cc: Amit Shah <amit.shah@redhat.com>
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This is done mainly for improving readability, and in preparation for the
next patch, but Markus pointed out another bonus for the string being
returned:
"No arbitrary length limit. Before the patch, it's 39 characters, and the
code breaks catastrophically when qdev_fw_name() is longer: the second
snprintf() is called with its first argument pointing beyond path[], and
its second argument underflowing to a huge size."
Cc: qemu-stable@nongnu.org
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Tested-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The patch
apci: fix PXB behaviour if used with unsupported BIOS
uses the following condition to see if a "PXB mem/IO chunk" has *not* been
configured by the BIOS:
(!range_base || range_base > range_limit)
When this condition evaluates to true, said patch *omits* the
corresponding entry from the _CRS.
Later on the patch checks for the opposite condition (with the intent of
*adding* entries to the _CRS if the "PXB mem/IO chunks" *have* been
configured). Unfortunately, the condition was negated incorrectly: only
the first ! operator was removed, which led to the nonsensical expression
(range_base || range_base > range_limit)
leading to bogus entries in the _CRS, and causing BSOD in Windows Server
2012 R2 when it runs on OVMF.
The correct negative of the condition seen at the top is
(range_base && range_base <= range_limit)
Fix the expressions.
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The ACPI specification permits the _HID and _UID objects to evaluate to
strings. (See "6.1.5 _HID (Hardware ID)" and "6.1.12 _UID (Unique ID)" in
the ACPI v6.0 spec.)
With regard to related standards, the UEFI specification can also express
a device address composed from string _HID and _UID identifiers, inside
the Expanded ACPI Device Path Node. (See "9.3.3 ACPI Device Path", Table
49, in the UEFI v2.5 spec.)
However, numeric (integer) contents for both _HID and _UID are more
traditional. They are recommended by the UEFI spec for size reasons:
[...] the ACPI Device Path node is smaller and should be used if
possible to reduce the size of device paths that may potentially be
stored in nonvolatile storage [...]
External tools support them better (for example the --acpi_hid and
--acpi_uid options of "efibootmgr" only take numeric identifiers).
Finally, numeric _HID and _UID contents are existing practice in the QEMU
source.
This patch was tested with a Fedora 20 LiveCD and a preexistent Windows
Server 2012 R2 guest. Using "acpidump" and "iasl" in the Fedora guest, we
get, in the SSDT:
> Scope (\_SB)
> {
> Device (PC04)
> {
> Name (_UID, 0x04) // _UID: Unique ID
> Name (_HID, EisaId ("PNP0A03") /* PCI Bus */) // _HID: Hardware ID
Cc: Marcel Apfelbaum <marcel@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This PIIX4 init function has no more reason to receive a pointer to the
FwCfg object. Remove the parameter from the prototype, and update callers.
As a result, the pc_init1() function no longer needs to save the return
value of pc_memory_init() and xen_load_linux(), which makes it more
similar to pc_q35_init().
The return type & value of pc_memory_init() and xen_load_linux() are not
changed themselves; maybe we'll need their return values sometime later.
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1204696
Cc: Amit Shah <amit.shah@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
The acpi_pm1_cnt_init() core function is responsible for setting up the
register block that will ultimately react to S3 and S4 requests (see
acpi_pm1_cnt_write()). It makes sense to advertise this configuration to
the guest firmware via an easy to parse fw_cfg file (ACPI is too complex
for firmware to parse), and indeed PIIX4 does that. However, since
acpi_pm1_cnt_init() is not specific to PIIX4, neither should be the fw_cfg
file.
This patch makes "etc/system-states" appear on all chipsets modified in
the previous patch, not just PIIX4 (assuming they have fw_cfg at all).
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1204696
Cc: Amit Shah <amit.shah@redhat.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: Leon Alrae <leon.alrae@imgtec.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
The "no_floppy = 1" machine class setting causes "default_floppy" in
main() to become zero. Consequently, default_drive() will not call
drive_add() and drive_new() for IF_FLOPPY, index=0, meaning that no
default floppy drive will be created for the virtual machine. In that
case, board code should also not insist on the creation of the
board-default FDC.
The board-default FDC will still be created if the user requests a floppy
drive with "-drive if=floppy".
Additionally, separate FDCs can be specified manually with "-device
isa-fdc". They allow the
-device isa-fdc,driveA=...
syntax that is more flexible than the one required by the board-default
FDC:
-global isa-fdc.driveA=...
This patch doesn't change the behavior observably, as all Q35 machine
types have "no_floppy = 0".
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Cc: "Gabriel L. Somlo" <gsomlo@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: qemu-block@nongnu.org
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Even if board code decides not to request the creation of the FDC (keyed
off board-level factors, to be determined later), we should create the FDC
nevertheless if the user passes '-drive if=floppy' on the command line.
Otherwise '-drive if=floppy' would break without explicit '-device
isa-fdc' on such boards.
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Cc: "Gabriel L. Somlo" <gsomlo@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: qemu-block@nongnu.org
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
This patch introduces no observable change, but it allows the callers of
pc_basic_device_init(), ie. pc_init1() and pc_q35_init(), to request (or
not request) the creation of the FDC explicitly.
At the moment both callers pass constant create_fdctrl=true (hence no
observable change).
Assuming a board passes create_fdctrl=false, "floppy" will be NULL on
output, and (beyond the FDC not being created) that NULL will be passed on
to pc_cmos_init(). Luckily, pc_cmos_init() already handles that case.
Cc: Markus Armbruster <armbru@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: John Snow <jsnow@redhat.com>
Cc: "Gabriel L. Somlo" <gsomlo@gmail.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: qemu-block@nongnu.org
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
At the moment, when the XHCI driver in edk2
(MdeModulePkg/Bus/Pci/XhciDxe/XhciDxe.inf) runs on QEMU, with the options
-device nec-usb-xhci -device usb-kbd
it crashes with:
ASSERT MdeModulePkg/Bus/Pci/XhciDxe/XhciSched.c(1759):
TrsRing != ((void*) 0)
The crash hits in the following edk2 call sequence (all files under
MdeModulePkg/Bus/):
UsbEnumerateNewDev() [Usb/UsbBusDxe/UsbEnumer.c]
UsbBuildDescTable() [Usb/UsbBusDxe/UsbDesc.c]
UsbGetDevDesc() [Usb/UsbBusDxe/UsbDesc.c]
UsbCtrlGetDesc(USB_REQ_GET_DESCRIPTOR) [Usb/UsbBusDxe/UsbDesc.c]
UsbCtrlRequest() [Usb/UsbBusDxe/UsbDesc.c]
UsbHcControlTransfer() [Usb/UsbBusDxe/UsbUtility.c]
XhcControlTransfer() [Pci/XhciDxe/Xhci.c]
XhcCreateUrb() [Pci/XhciDxe/XhciSched.c]
XhcCreateTransferTrb() [Pci/XhciDxe/XhciSched.c]
XhcExecTransfer() [Pci/XhciDxe/XhciSched.c]
XhcCheckUrbResult() [Pci/XhciDxe/XhciSched.c]
//
// look for TRB_TYPE_DATA_STAGE event [1]
//
//
// Store a copy of the device descriptor, as the hub device
// needs this info to configure endpoint. [2]
//
UsbSetConfig() [Usb/UsbBusDxe/UsbDesc.c]
UsbCtrlRequest(USB_REQ_SET_CONFIG) [Usb/UsbBusDxe/UsbDesc.c]
UsbHcControlTransfer() [Usb/UsbBusDxe/UsbUtility.c]
XhcControlTransfer() [Pci/XhciDxe/Xhci.c]
XhcSetConfigCmd() [Pci/XhciDxe/XhciSched.c]
XhcInitializeEndpointContext() [Pci/XhciDxe/XhciSched.c]
//
// allocate transfer ring for the endpoint [3]
//
USBKeyboardDriverBindingStart() [Usb/UsbKbDxe/EfiKey.c]
UsbIoAsyncInterruptTransfer() [Usb/UsbBusDxe/UsbBus.c]
UsbHcAsyncInterruptTransfer() [Usb/UsbBusDxe/UsbUtility.c]
XhcAsyncInterruptTransfer() [Pci/XhciDxe/Xhci.c]
XhcCreateUrb() [Pci/XhciDxe/Xhci.c]
XhcCreateTransferTrb() [Pci/XhciDxe/XhciSched.c]
XhcSyncTrsRing() [Pci/XhciDxe/XhciSched.c]
ASSERT (TrsRing != NULL) [4]
UsbEnumerateNewDev() in the USB bus driver issues a GET_DESCRIPTOR
request, in order to determine the number of configurations that the
endpoint supports. The requests consists of three stages (three TRBs),
setup, data, and status. The length of the response is determined in [1],
namely from the transfer event that the host controller generates in
response to the request's middle stage (ie. the data stage).
If the length of the answer is correct (a full GET_DESCRIPTOR request
takes 18 bytes), then the XHCI driver that underlies the USB bus driver
"snoops" (caches) the descriptor data for later [2].
Later, the USB bus driver sends a SET_CONFIG request. The underlying XHCI
driver allocates a transfer ring for the endpoint, relying on the data
snooped and cached in step [2].
Finally, the USB keyboard driver submits an asynchronous interrupt
transfer to manage the keyboard. As part of this it asserts [4] that the
ring has been allocated in step [3].
And this ASSERT() fires. The root cause can be found in the way QEMU
handles the initial GET_DESCRIPTOR request.
Again, that request consists of three stages (TRBs, Transfer Request
Blocks), "setup", "data", and "status". The XhcCreateTransferTrb()
function sets the IOC ("Interrupt on Completion") flag in each of these
TRBs.
According to the XHCI specification, the host controller shall generate a
Transfer Event in response to *each* individual TRB of the request that
had the IOC flag set. This means that QEMU should queue three events:
setup, data, and status, for edk2's XHCI driver.
However, QEMU only generates two events:
- one for the setup (ie. 1st) stage,
- another for the status (ie. 3rd) stage.
No event is generated for the middle (ie. data) stage. The loop in QEMU's
xhci_xfer_report() function runs three times, but due to the "reported"
variable, only the first and the last TRBs elicit events, the middle (data
stage) results in no event queued.
As a consequence:
- When handling the GET_DESCRIPTOR request, XhcCheckUrbResult() in [1]
does not update the response length from zero.
- XhcControlTransfer() thinks that the response is invalid (it has zero
length payload instead of 18 bytes), hence [2] is not reached; the
device descriptor is not stashed for later, and the number of possible
configurations is left at zero.
- When handling the SET_CONFIG request, (NumConfigurations == 0) from
above prevents the allocation of the endpoint's transfer ring.
- When the keyboard driver tries to use the endpoint, the ASSERT() blows
up.
The solution is to correct the emulation in QEMU, and to generate a
transfer event whenever IOC is set in a TRB.
The patch replaces
!reported && (IOC || foo) == !reported && IOC ||
!reported && foo
with
IOC || (!reported && foo) == IOC ||
!reported && foo
which only changes how
reported && IOC
is handled. (Namely, it now generates an event.)
Tested with edk2 built for "qemu-system-aarch64 -M virt" (ie.
"ArmVirtualizationQemu.dsc", aka "AAVMF"), and guest Linux.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(1) Let's contemplate what device endianness means, for a memory mapped
device register (independently of QEMU -- that is, on physical hardware).
It determines the byte order that the device will put on the data bus when
the device is producing a *numerical value* for the CPU. This byte order
may differ from the CPU's own byte order, therefore when software wants to
consume the *numerical value*, it may have to swap the byte order first.
For example, suppose we have a device that exposes in a 2-byte register
the number of sheep we have to count before falling asleep. If the value
is decimal 37 (0x0025), then a big endian register will produce [0x00,
0x25], while a little endian register will produce [0x25, 0x00].
If the device register is big endian, but the CPU is little endian, the
numerical value will read as 0x2500 (decimal 9472), which software has to
byte swap before use.
However... if we ask the device about who stole our herd of sheep, and it
answers "XY", then the byte representation coming out of the register must
be [0x58, 0x59], regardless of the device register's endianness for
numeric values. And, software needs to copy these bytes into a string
field regardless of the CPU's own endianness.
(2) QEMU's device register accessor functions work with *numerical values*
exclusively, not strings:
The emulated register's read accessor function returns the numerical value
(eg. 37 decimal, 0x0025) as a *host-encoded* uint64_t. QEMU translates
this value for the guest to the endianness of the emulated device register
(which is recorded in MemoryRegionOps.endianness). Then guest code must
translate the numerical value from device register to guest CPU
endianness, before including it in any computation (see (1)).
(3) However, the data register of the fw_cfg device shall transfer strings
*only* -- that is, opaque blobs. Interpretation of any given blob is
subject to further agreement -- it can be an integer in an independently
determined byte order, or a genuine string, or an array of structs of
integers (in some byte order) and fixed size strings, and so on.
Because register emulation in QEMU is integer-preserving, not
string-preserving (see (2)), we have to jump through a few hoops.
(3a) We defined the memory mapped fw_cfg data register as
DEVICE_BIG_ENDIAN.
The particular choice is not really relevant -- we picked BE only for
consistency with the control register, which *does* transfer integers --
but our choice affects how we must host-encode values from fw_cfg strings.
(3b) Since we want the fw_cfg string "XY" to appear as the [0x58, 0x59]
array on the data register, *and* we picked DEVICE_BIG_ENDIAN, we must
compose the host (== C language) value 0x5859 in the read accessor
function.
(3c) When the guest performs the read access, the immediate uint16_t value
will be 0x5958 (in LE guests) and 0x5859 (in BE guests). However, the
uint16_t value does not matter. The only thing that matters is the byte
pattern [0x58, 0x59], which the guest code must copy into the target
string *without* any byte-swapping.
(4) Now I get to explain where I screwed up. :(
When we decided for big endian *integer* representation in the MMIO data
register -- see (3a) --, I mindlessly added an indiscriminate
byte-swizzling step to the (little endian) guest firmware.
This was a grave error -- it violates (3c) --, but I didn't realize it. I
only saw that the code I otherwise intended for fw_cfg_data_mem_read():
value = 0;
for (i = 0; i < size; ++i) {
value = (value << 8) | fw_cfg_read(s);
}
didn't produce the expected result in the guest.
In true facepalm style, instead of blaming my guest code (which violated
(3c)), I blamed my host code (which was correct). Ultimately, I coded
ldX_he_p() into fw_cfg_data_mem_read(), because that happened to work.
Obviously (...in retrospect) that was wrong. Only because my host happened
to be LE, ldX_he_p() composed the (otherwise incorrect) host value 0x5958
from the fw_cfg string "XY". And that happened to compensate for the bogus
indiscriminate byte-swizzling in my guest code.
Clearly the current code leaks the host endianness through to the guest,
which is wrong. Any device should work the same regardless of host
endianness.
The solution is to compose the host-endian representation (2) of the big
endian interpretation (3a, 3b) of the fw_cfg string, and to drop the wrong
byte-swizzling in the guest (3c).
Brown paper bag time for me.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Message-id: 1420024880-15416-1-git-send-email-lersek@redhat.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Introduce the new boolean field "arm_boot_info.firmware_loaded". When this
field is set, it means that the portion of guest DRAM that the VCPU
normally starts to execute, or the pflash chip that the VCPU normally
starts to execute, has been populated by board-specific code with
full-fledged guest firmware code, before the board calls
arm_load_kernel().
Simultaneously, "arm_boot_info.firmware_loaded" guarantees that the board
code has set up the global firmware config instance, for arm_load_kernel()
to find with fw_cfg_find().
Guest kernel (-kernel) and guest firmware (-bios, -pflash) has always been
possible to specify independently on the command line. The following cases
should be considered:
nr -bios -pflash -kernel description
unit#0
-- ------- ------- ------- -------------------------------------------
1 present present absent Board code rejects this case, -bios and
present present present -pflash unit#0 are exclusive. Left intact
by this patch.
2 absent absent present Traditional kernel loading, with qemu's
minimal board firmware. Left intact by this
patch.
3 absent present absent Preexistent case for booting guest firmware
present absent absent loaded with -bios or -pflash. Left intact
by this patch.
4 absent absent absent Preexistent case for not loading any
firmware or kernel up-front. Left intact by
this patch.
5 present absent present New case introduced by this patch: kernel
absent present present image is passed to externally loaded
firmware in unmodified form, using fw_cfg.
An easy way to see that this patch doesn't interfere with existing cases
is to realize that "info->firmware_loaded" is constant zero at this point.
Which makes the "outer" condition unchanged, and the "inner" condition
(with the fw_cfg-related code) dead.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-11-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
In the next patch we'd like to reuse the image decompression facility
without installing the output as a ROM at a specific guest-phys address.
In addition, expose LOAD_IMAGE_MAX_GUNZIP_BYTES, because that's a
straightforward "max_sz" argument for the new load_image_gzipped_buffer().
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-10-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
fw_cfg already supports exposure over MMIO (used in ppc/mac_newworld.c,
ppc/mac_oldworld.c, sparc/sun4m.c); we can easily add it to the "virt"
board.
Because MMIO access is slow on ARM KVM, we enable the guest, with
fw_cfg_init_mem_wide(), to transfer up to 8 bytes with a single access.
This has been measured to speed up transfers up to 7.5-fold, relative to
single byte data access, on both ARM KVM and x86_64 TCG.
The MMIO register block of fw_cfg is advertized in the device tree. As
base address we pick 0x09020000, which conforms to the comment preceding
"a15memmap": it falls in the miscellaneous device I/O range 128MB..256MB,
and it is aligned at 64KB. The DTB properties follow the documentation in
the Linux source file "Documentation/devicetree/bindings/arm/fw-cfg.txt".
fw_cfg automatically exports a number of files to the guest; for example,
"bootorder" (see fw_cfg_machine_reset()).
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-9-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
We rebase fw_cfg_init_mem() to the new function for compatibility with
current callers.
The behavior of the (big endian) multi-byte data reads is best shown
with a qtest session. Here, we are reading the first six bytes of
the UUID
$ arm-softmmu/qemu-system-arm -M virt -machine accel=qtest \
-qtest stdio -uuid 4600cb32-38ec-4b2f-8acb-81c6ea54f2d8
>>> writew 0x9020008 0x0200
<<< OK
>>> readl 0x9020000
<<< OK 0x000000004600cb32
Remember this is big endian. On big endian machines, it is stored
directly as 0x46 0x00 0xcb 0x32.
On a little endian machine, we have to first swap it, so that it becomes
0x32cb0046. When written to memory, it becomes 0x46 0x00 0xcb 0x32
again.
Reading byte-by-byte works too, of course:
>>> readb 0x9020000
<<< OK 0x0000000000000038
>>> readb 0x9020000
<<< OK 0x00000000000000ec
Here only a single byte is read at a time, so they are read in order
similar to the 1-byte data port that is already in PPC and SPARC
machines.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-8-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The "data_width" property is capable of changing the maximum valid access
size to the MMIO data register, and resizes the memory region similarly,
at device realization time.
The default value of "data_memwidth" is set so that we don't yet diverge
from "fw_cfg_data_mem_ops".
Most of the fw_cfg_mem users will stick with the default, and for them we
should continue using the statically allocated "fw_cfg_data_mem_ops". This
is beneficial for debugging because gdb can resolve pointers referencing
static objects to the names of those objects.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-7-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The standalone selector port (fw_cfg_ctl_mem_ops) is only used by big
endian guests to date (*), hence this change doesn't regress them. Paolo
and Alex have suggested / requested an explicit DEVICE_BIG_ENDIAN setting
here, for clarity.
(*) git grep -l fw_cfg_init_mem
hw/nvram/fw_cfg.c
hw/ppc/mac_newworld.c
hw/ppc/mac_oldworld.c
hw/sparc/sun4m.c
include/hw/nvram/fw_cfg.h
The standalone data port (fw_cfg_data_mem_ops) has max_access_size 1 (for
now), hence changing its endianness doesn't change behavior for existing
guest code.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-5-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
We are going to introduce a wide data register for fw_cfg, but only for
the MMIO mapped device. The wide data register will also require the
tightening of endiannesses.
However we don't want to touch the I/O port mapped fw_cfg device at all.
Currently QEMU provides a single fw_cfg device type that can handle both
I/O port and MMIO mapping. This flexibility is not actually exploited by
any board in the tree, but it renders restricting the above changes to
MMIO very hard.
Therefore, let's derive two classes from TYPE_FW_CFG: TYPE_FW_CFG_IO and
TYPE_FW_CFG_MEM.
TYPE_FW_CFG_IO incorporates the base I/O port and the related combined
MemoryRegion. (NB: all boards in the tree that use the I/O port mapped
flavor opt for the combined mapping; that is, when the data port overlays
the high address byte of the selector port. Therefore we can drop the
capability to map those I/O ports separately.)
TYPE_FW_CFG_MEM incorporates the base addresses for the MMIO selector and
data registers, and their respective MemoryRegions.
The "realize" and "props" class members are specific to each new derived
class, and become unused for the base class. The base class retains the
"reset" member and the "vmsd" member, because the reset functionality and
the set of migrated data are not specific to the mapping.
The new functions fw_cfg_init_io() and fw_cfg_init_mem() expose the
possible mappings in separation. For now fw_cfg_init() is retained as a
compatibility shim that enforces the above assumptions.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-id: 1419250305-31062-2-git-send-email-pbonzini@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
With this patch applied, the output of -M \? is
> Supported machines are:
> pc Standard PC (i440FX + PIIX, 1996) (alias of pc-i440fx-2.2)
> pc-i440fx-2.2 Standard PC (i440FX + PIIX, 1996) (default)
> pc-i440fx-2.1 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-2.0 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.7 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.6 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.5 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.4 Standard PC (i440FX + PIIX, 1996)
> pc-1.3 Standard PC (i440FX + PIIX, 1996)
> pc-1.2 Standard PC (i440FX + PIIX, 1996)
> pc-1.1 Standard PC (i440FX + PIIX, 1996)
> pc-1.0 Standard PC (i440FX + PIIX, 1996)
> pc-0.15 Standard PC (i440FX + PIIX, 1996)
> pc-0.14 Standard PC (i440FX + PIIX, 1996)
> pc-0.13 Standard PC (i440FX + PIIX, 1996)
> pc-0.12 Standard PC (i440FX + PIIX, 1996)
> pc-0.11 Standard PC (i440FX + PIIX, 1996)
> pc-0.10 Standard PC (i440FX + PIIX, 1996)
> q35 Standard PC (Q35 + ICH9, 2009) (alias of pc-q35-2.2)
> pc-q35-2.2 Standard PC (Q35 + ICH9, 2009)
> pc-q35-2.1 Standard PC (Q35 + ICH9, 2009)
> pc-q35-2.0 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.7 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.6 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.5 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.4 Standard PC (Q35 + ICH9, 2009)
> isapc ISA-only PC
> none empty machine
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1145042
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel.a@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Commit 261747f1 ("vl: Use MachineClass instead of global QEMUMachine
list") broke the ordering of the machine types in the user-visible output
of
qemu-system-XXXX -M \?
This occurred because registration was rebased from a manually maintained
linked list to GLib hash tables:
qemu_register_machine()
type_register()
type_register_internal()
type_table_add()
g_hash_table_insert()
and because the listing was rebased accordingly, from the traversal of the
list to the traversal of the hash table (rendered as an ad-hoc list):
machine_parse()
object_class_get_list(TYPE_MACHINE)
object_class_foreach()
g_hash_table_foreach()
The current order is a "random" one, for practical purposes, which is
annoying for users.
Introduce new members QEMUMachine.family and MachineClass.family, allowing
machine types to be "clustered". Introduce a comparator function that
establishes a total ordering between machine types, ordering machine types
in the same family next to each other. In machine_parse(), list the
supported machine types sorted with the comparator function.
The comparator function:
- sorts whole families before standalone machine types,
- sorts whole families between each other in alphabetically increasing
order,
- sorts machine types inside the same family in alphabetically decreasing
order,
- sorts standalone machine types between each other in alphabetically
increasing order.
After this patch, all machine types are considered standalone, and
accordingly, the output is alphabetically ascending. This will be refined
in the following patches.
Effects on the x86_64 output:
Before:
> Supported machines are:
> pc-0.13 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-2.0 Standard PC (i440FX + PIIX, 1996)
> pc-1.0 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-2.1 Standard PC (i440FX + PIIX, 1996)
> pc-q35-1.7 Standard PC (Q35 + ICH9, 2009)
> pc-1.1 Standard PC (i440FX + PIIX, 1996)
> pc-0.14 Standard PC (i440FX + PIIX, 1996)
> pc-q35-2.0 Standard PC (Q35 + ICH9, 2009)
> pc-i440fx-1.4 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.5 Standard PC (i440FX + PIIX, 1996)
> pc-0.15 Standard PC (i440FX + PIIX, 1996)
> pc-q35-1.4 Standard PC (Q35 + ICH9, 2009)
> isapc ISA-only PC
> pc Standard PC (i440FX + PIIX, 1996) (alias of pc-i440fx-2.2)
> pc-i440fx-2.2 Standard PC (i440FX + PIIX, 1996) (default)
> pc-1.2 Standard PC (i440FX + PIIX, 1996)
> pc-0.10 Standard PC (i440FX + PIIX, 1996)
> pc-0.11 Standard PC (i440FX + PIIX, 1996)
> pc-q35-2.1 Standard PC (Q35 + ICH9, 2009)
> q35 Standard PC (Q35 + ICH9, 2009) (alias of pc-q35-2.2)
> pc-q35-2.2 Standard PC (Q35 + ICH9, 2009)
> pc-i440fx-1.6 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.7 Standard PC (i440FX + PIIX, 1996)
> none empty machine
> pc-q35-1.5 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.6 Standard PC (Q35 + ICH9, 2009)
> pc-0.12 Standard PC (i440FX + PIIX, 1996)
> pc-1.3 Standard PC (i440FX + PIIX, 1996)
After:
> Supported machines are:
> isapc ISA-only PC
> none empty machine
> pc-0.10 Standard PC (i440FX + PIIX, 1996)
> pc-0.11 Standard PC (i440FX + PIIX, 1996)
> pc-0.12 Standard PC (i440FX + PIIX, 1996)
> pc-0.13 Standard PC (i440FX + PIIX, 1996)
> pc-0.14 Standard PC (i440FX + PIIX, 1996)
> pc-0.15 Standard PC (i440FX + PIIX, 1996)
> pc-1.0 Standard PC (i440FX + PIIX, 1996)
> pc-1.1 Standard PC (i440FX + PIIX, 1996)
> pc-1.2 Standard PC (i440FX + PIIX, 1996)
> pc-1.3 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.4 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.5 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.6 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-1.7 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-2.0 Standard PC (i440FX + PIIX, 1996)
> pc-i440fx-2.1 Standard PC (i440FX + PIIX, 1996)
> pc Standard PC (i440FX + PIIX, 1996) (alias of pc-i440fx-2.2)
> pc-i440fx-2.2 Standard PC (i440FX + PIIX, 1996) (default)
> pc-q35-1.4 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.5 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.6 Standard PC (Q35 + ICH9, 2009)
> pc-q35-1.7 Standard PC (Q35 + ICH9, 2009)
> pc-q35-2.0 Standard PC (Q35 + ICH9, 2009)
> pc-q35-2.1 Standard PC (Q35 + ICH9, 2009)
> q35 Standard PC (Q35 + ICH9, 2009) (alias of pc-q35-2.2)
> pc-q35-2.2 Standard PC (Q35 + ICH9, 2009)
Effects on the aarch64 output:
Before:
> Supported machines are:
> lm3s811evb Stellaris LM3S811EVB
> canon-a1100 Canon PowerShot A1100 IS
> vexpress-a15 ARM Versatile Express for Cortex-A15
> vexpress-a9 ARM Versatile Express for Cortex-A9
> xilinx-zynq-a9 Xilinx Zynq Platform Baseboard for Cortex-A9
> connex Gumstix Connex (PXA255)
> n800 Nokia N800 tablet aka. RX-34 (OMAP2420)
> lm3s6965evb Stellaris LM3S6965EVB
> versatileab ARM Versatile/AB (ARM926EJ-S)
> borzoi Borzoi PDA (PXA270)
> tosa Tosa PDA (PXA255)
> cheetah Palm Tungsten|E aka. Cheetah PDA (OMAP310)
> midway Calxeda Midway (ECX-2000)
> mainstone Mainstone II (PXA27x)
> n810 Nokia N810 tablet aka. RX-44 (OMAP2420)
> terrier Terrier PDA (PXA270)
> highbank Calxeda Highbank (ECX-1000)
> cubieboard cubietech cubieboard
> sx1-v1 Siemens SX1 (OMAP310) V1
> sx1 Siemens SX1 (OMAP310) V2
> realview-eb-mpcore ARM RealView Emulation Baseboard (ARM11MPCore)
> kzm ARM KZM Emulation Baseboard (ARM1136)
> akita Akita PDA (PXA270)
> z2 Zipit Z2 (PXA27x)
> musicpal Marvell 88w8618 / MusicPal (ARM926EJ-S)
> realview-pb-a8 ARM RealView Platform Baseboard for Cortex-A8
> versatilepb ARM Versatile/PB (ARM926EJ-S)
> realview-eb ARM RealView Emulation Baseboard (ARM926EJ-S)
> realview-pbx-a9 ARM RealView Platform Baseboard Explore for Cortex-A9
> spitz Spitz PDA (PXA270)
> none empty machine
> virt ARM Virtual Machine
> collie Collie PDA (SA-1110)
> smdkc210 Samsung SMDKC210 board (Exynos4210)
> verdex Gumstix Verdex (PXA270)
> nuri Samsung NURI board (Exynos4210)
> integratorcp ARM Integrator/CP (ARM926EJ-S)
After:
> Supported machines are:
> akita Akita PDA (PXA270)
> borzoi Borzoi PDA (PXA270)
> canon-a1100 Canon PowerShot A1100 IS
> cheetah Palm Tungsten|E aka. Cheetah PDA (OMAP310)
> collie Collie PDA (SA-1110)
> connex Gumstix Connex (PXA255)
> cubieboard cubietech cubieboard
> highbank Calxeda Highbank (ECX-1000)
> integratorcp ARM Integrator/CP (ARM926EJ-S)
> kzm ARM KZM Emulation Baseboard (ARM1136)
> lm3s6965evb Stellaris LM3S6965EVB
> lm3s811evb Stellaris LM3S811EVB
> mainstone Mainstone II (PXA27x)
> midway Calxeda Midway (ECX-2000)
> musicpal Marvell 88w8618 / MusicPal (ARM926EJ-S)
> n800 Nokia N800 tablet aka. RX-34 (OMAP2420)
> n810 Nokia N810 tablet aka. RX-44 (OMAP2420)
> none empty machine
> nuri Samsung NURI board (Exynos4210)
> realview-eb ARM RealView Emulation Baseboard (ARM926EJ-S)
> realview-eb-mpcore ARM RealView Emulation Baseboard (ARM11MPCore)
> realview-pb-a8 ARM RealView Platform Baseboard for Cortex-A8
> realview-pbx-a9 ARM RealView Platform Baseboard Explore for Cortex-A9
> smdkc210 Samsung SMDKC210 board (Exynos4210)
> spitz Spitz PDA (PXA270)
> sx1 Siemens SX1 (OMAP310) V2
> sx1-v1 Siemens SX1 (OMAP310) V1
> terrier Terrier PDA (PXA270)
> tosa Tosa PDA (PXA255)
> verdex Gumstix Verdex (PXA270)
> versatileab ARM Versatile/AB (ARM926EJ-S)
> versatilepb ARM Versatile/PB (ARM926EJ-S)
> vexpress-a15 ARM Versatile Express for Cortex-A15
> vexpress-a9 ARM Versatile Express for Cortex-A9
> virt ARM Virtual Machine
> xilinx-zynq-a9 Xilinx Zynq Platform Baseboard for Cortex-A9
> z2 Zipit Z2 (PXA27x)
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1145042
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel.a@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
A drive that backs a pflash device is special:
- it is very small,
- its entire contents are kept in a RAMBlock at all times, covering the
guest-phys address range that provides the guest's view of the emulated
flash chip.
The pflash device model keeps the drive (the host-side file) and the
guest-visible flash contents in sync. When migrating the guest, the
guest-visible flash contents (the RAMBlock) is migrated by default, but on
the target host, the drive (the host-side file) remains in full sync with
the RAMBlock only if:
- the source and target hosts share the storage underlying the pflash
drive,
- or the migration requests full or incremental block migration too, which
then covers all drives.
Due to the special nature of pflash drives, the following scenario makes
sense as well:
- no full nor incremental block migration, covering all drives, alongside
the base migration (justified eg. by shared storage for "normal" (big)
drives),
- non-shared storage for pflash drives.
In this case, currently only those portions of the flash drive are updated
on the target disk that the guest reprograms while running on the target
host.
In order to restore accord, dump the entire flash contents to the bdrv in
a post_load() callback.
- The read-only check follows the other call-sites of pflash_update();
- both "pfl->ro" and pflash_update() reflect / consider the case when
"pfl->bs" is NULL;
- the total size of the flash device is calculated as in
pflash_cfi01_realize().
When using shared storage, or requesting full or incremental block
migration along with the normal migration, the patch should incur a
harmless rewrite from the target side.
It is assumed that, on the target host, RAM is loaded ahead of the call to
pflash_post_load().
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
"vmstate_xhci_event" was introduced in commit 37352df3 ("xhci: add live
migration support"), and first released in v1.6.0. The field list in this
VMSD is not terminated with the VMSTATE_END_OF_LIST() macro.
During normal use (ie. migration), the issue is practically invisible,
because the "vmstate_xhci_event" object (with the unterminated field list)
is only ever referenced -- via "vmstate_xhci_intr" -- if xhci_er_full()
returns true, for the "ev_buffer" test. Since that field_exists() check
(apparently) almost always returns false, we almost never traverse
"vmstate_xhci_event" during migration, which hides the bug.
However, Amit's vmstate checker forces recursion into this VMSD as well,
and the lack of VMSTATE_END_OF_LIST() breaks the field list terminator
check (field->name != NULL) in dump_vmstate_vmsd(). The result is
undefined behavior, which in my case translates to infinite recursion
(because the loop happens to overflow into "vmstate_xhci_intr", which then
links back to "vmstate_xhci_event").
Add the missing terminator.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
In addition to the on-line reporting added in the previous patch, allow
libvirt to query frontend state independently of events.
Libvirt's path to identify the guest agent channel it cares about differs
between the event added in the previous patch and the QMP response field
added here. The event identifies the frontend device, by "id". The
'query-chardev' QMP command identifies the backend device (again by "id").
The association is under libvirt's control.
RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1080376
Reviewed-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
We can (and should) rely on the fact that s->flag_compress is exactly one
of DUMP_DH_COMPRESSED_ZLIB, DUMP_DH_COMPRESSED_LZO, and
DUMP_DH_COMPRESSED_SNAPPY.
This is ensured by the QMP schema and dump_init() in combination.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
qmp_dump_guest_memory()
dump_init()
lzo_init() <---------+
create_kdump_vmcore() |
write_dump_pages() |
get_len_buf_out() |
lzo_init() ------+
This patch doesn't change the fact that lzo_init() is called for every
LZO-compressed dump, but it makes get_len_buf_out() more focused (single
responsibility).
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
The specific ELF architecture (d_machine) carries Too Much Information
(TM) for deciding between create_header32() and create_header64(), use
"d_class" instead (ELFCLASS32 vs. ELFCLASS64).
This change adapts write_dump_header() to write_elf_loads(), dump_begin()
etc. that also rely on the ELF class of the target for bitness selection.
Considering the current targets that support dumping, cpu_get_dump_info()
works as follows:
- target-s390x/arch_dump.c: (EM_S390, ELFCLASS64) only
- target-ppc/arch_dump.c (EM_PPC64, ELFCLASS64) only
- target-i386/arch_dump.c: sets (EM_X86_64, ELFCLASS64) vs. (EM_386,
ELFCLASS32) keying off the same Long Mode Active flag.
Hence no observable change.
Approximately-suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Use TARGET_PAGE_SIZE and ~TARGET_PAGE_MASK instead.
"DumpState.page_size" has type "size_t", whereas TARGET_PAGE_SIZE has type
"int". TARGET_PAGE_MASK is of type "int" and has negative value. The patch
affects the implicit type conversions as follows:
- create_header32() and create_header64(): assigned to "block_size", which
has type "uint32_t". No change.
- get_next_page(): "block->target_start", "block->target_end" and "addr"
have type "hwaddr" (uint64_t).
Before the patch,
- if "size_t" was "uint64_t", then no additional conversion was done as
part of the usual arithmetic conversions,
- If "size_t" was "uint32_t", then it was widened to uint64_t as part of
the usual arithmetic conversions,
for the remainder and addition operators.
After the patch,
- "~TARGET_PAGE_MASK" expands to ~~((1 << TARGET_PAGE_BITS) - 1). It
has type "int" and positive value (only least significant bits set).
That's converted (widened) to "uint64_t" for the bit-ands. No visible
change.
- The same holds for the (addr + TARGET_PAGE_SIZE) addition.
- write_dump_pages():
- TARGET_PAGE_SIZE passed as argument to a bunch of functions that all
have prototypes. No change.
- When incrementing "offset_data" (of type "off_t"): given that we never
build for ILP32_OFF32 (see "-D_FILE_OFFSET_BITS=64" in configure),
"off_t" is always "int64_t", and we only need to consider:
- ILP32_OFFBIG: "size_t" is "uint32_t".
- before: int64_t += uint32_t. Page size converted to int64_t for
the addition.
- after: int64_t += int32_t. No change.
- LP64_OFF64: "size_t" is "uint64_t".
- before: int64_t += uint64_t. Offset converted to uint64_t for the
addition, then the uint64_t result is converted to int64_t for
storage.
- after: int64_t += int32_t. Same as the ILP32_OFFBIG/after case.
No visible change.
- (size_out < s->page_size) comparisons, and (size_out = s->page_size)
assignment:
- before: "size_out" is of type "size_t", no implicit conversion for
either operator.
- after: TARGET_PAGE_SIZE (of type "int" and positive value) is
converted to "size_t" (for the relop because the latter is
one of "uint32_t" and "uint64_t"). No visible change.
- dump_init():
- DIV_ROUND_UP(DIV_ROUND_UP(s->max_mapnr, CHAR_BIT), s->page_size): The
innermost "DumpState.max_mapnr" field has type uint64_t, which
propagates through all implicit conversions at hand:
#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
regardless of the page size macro argument's type. In the outer macro
replacement, the page size is converted from uint32_t and int32_t
alike to uint64_t.
- (tmp * s->page_size) multiplication: "tmp" has size "uint64_t"; the
RHS is converted to that type from uint32_t and int32_t just the same
if it's not uint64_t to begin with.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Just use TARGET_PAGE_BITS.
"DumpState.page_shift" used to have type "uint32_t", while the replacement
TARGET_PAGE_BITS has type "int". Since "DumpState.page_shift" was only
used as bit shift counts in the paddr_to_pfn() and pfn_to_paddr() macros,
this is safe.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Currently, the function
- defines and populates an auto variable of type MakedumpfileHeader
- allocates and zeroes a buffer of size MAX_SIZE_MDF_HEADER (4096)
- copies the former into the latter (covering an initial portion of the
latter)
Fill in the MakedumpfileHeader structure in its final place (the alignment
is OK because the structure lives at the address returned by g_malloc0()).
Approximately-suggested-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
The "mh.signature" array field has size 16, and is zeroed by the preceding
memset(). MAKEDUMPFILE_SIGNATURE expands to a string literal with string
length 12 (size 13). There's no need to measure the length of
MAKEDUMPFILE_SIGNATURE at runtime, nor for the extra zero-filling of
"mh.signature" with strncpy().
Use memcpy() with MIN(sizeof, sizeof) for robustness (which is an integer
constant expression, evaluable at compile time.)
Approximately-suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Among the callers, only assigned_initfn() should set the monitor's stored
error. Other callers may run in contexts where the monitor's stored error
makes no sense. For example:
assigned_dev_pci_write_config()
assigned_dev_update_msix()
assign_intx()
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Also, change the return type to "void"; the function is static (with a
sole caller) and the negative errno values are not distinguished from each
other.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
The return type is also changed from "int" to "void", because it was used
in a success vs. failure sense only (the caller didn't distinguish error
codes from each other, and even assigned_dev_register_msix_mmio() masked
mmap()'s errno values with a common -EFAULT).
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Propagate any errors while adding PCI capabilities to
assigned_device_pci_cap_init(). We'll continue the propagation upwards
when assigned_device_pci_cap_init() becomes a leaf itself (when none of
its callees will report errors internally any longer when detecting and
returning them).
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Rename check_irqchip_in_kernel() to verify_irqchip_in_kernel(), so that
the name reflects our expectation better. Rather than returning a bool,
make it do nothing or set an Error.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
get_real_id() has two thin wrappers (and no other callers),
get_real_vendor_id() and get_real_device_id(); it's easiest to convert
them in one fell swoop.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
This allows us to report the entire error with one error_report() call,
easing future error propagation.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Propagate any errors in monitor fd handling up to get_real_device(), and
report them there. We'll continue the propagation upwards when
get_real_device() becomes a leaf itself (when none of its callees will
report errors internally any longer when detecting and returning an
error).
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
eviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
and rebase monitor_handle_fd_param() to it. (Note that this will slightly
change the behavior when the qemu_parse_fd() branch is selected and it
fails: we now report (and in case of QMP, set) the error immediately,
rather than allowing the caller to set its own error message (if any)).
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
qemu_parse_fd() used to handle at least the following strings incorrectly:
o "-2": simply let through
o "2147483648": returned as LONG_MAX==INT_MAX on ILP32 (with ERANGE
ignored); implementation-defined behavior on LP64
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Building on the previous patch, raise the maximal count of processor
objects / NTFY branches / CPON elements from 255 to 256. This allows the
VCPU with APIC ID 0xFF to be hotplugged.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The build_ssdt() function builds a number of AML objects that are related
to CPU hotplug, and whose IDs form a contiguous sequence of APIC IDs.
(APIC IDs are in fact discontiguous, but this is the traditional
interface: build a contiguous sequence from zero up that covers all
possible APIC IDs.) These objects are:
- a Processor() object for each VCPU,
- a NTFY method, with one branch for each VCPU,
- a CPON package with one element (hotplug status byte) for each VCPU.
The build_ssdt() function currently limits the *count* of processor
objects, and NTFY branches, and CPON elements, in 0xFF (see the assignment
to "acpi_cpus"). This allows for an inclusive APIC ID range of [0..254].
This is incorrect, because the highest APIC ID that we otherwise allow a
VCPU to take is 255.
In order to extend the maximum count to 256, and the traversed APIC ID
range correspondingly to [0..255]:
- the Processor() objects need no change,
- the NTFY method also needs no change,
- the CPON package must be updated, because it is defined with a
DefPackage, and the number of elements in such a package can be at most
255. We pick a DefVarPackage instead.
We replace the Op byte, and the encoding of the number of elements.
Compare:
DefPackage := PackageOp PkgLength NumElements PackageElementList
DefVarPackage := VarPackageOp PkgLength VarNumElements PackageElementList
PackageOp := 0x12
VarPackageOp := 0x13
NumElements := ByteData
VarNumElements := TermArg => Integer
The build_append_int() function implements precisely the following TermArg
encodings (a subset of what the ACPI spec describes):
TermArg := DataObject
DataObject := ComputationalData
ComputationalData := ConstObj | ByteConst | WordConst | DWordConst
directly encoded in the function, with build_append_byte():
ConstObj := ZeroOp | OneOp
ZeroOp := 0x00
OneOp := 0x01
call to build_append_value(..., 1):
ByteConst := BytePrefix ByteData
BytePrefix := 0x0A
ByteData := 0x00 - 0xFF
call to build_append_value(..., 2):
WordConst := WordPrefix WordData
WordPrefix := 0x0B
WordData := ByteData[0:7] ByteData[8:15]
call to build_append_value(..., 4):
DWordConst := DWordPrefix DWordData
DWordPrefix := 0x0C
DWordData := WordData[0:15] WordData[16:31]
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The "keytab" specification in "qemu.sasl" only makes sense if "gssapi" is
selected in "mech_list". Even if the latter is not done (ie. "gssapi" is
not selected), the cyrus-sasl library tries to open the specified keytab
file, although nothing has a use for it outside the gssapi backend.
Since the default keytab file "/etc/qemu/krb5.tab" is usually absent, the
cyrus-sasl library emits a warning to syslog at startup, which tends to
annoy users (who didn't ask for gssapi in the first place).
Comment out the keytab specification per default.
"qemu-doc.texi" already correctly explains how to use "mech_list: gssapi"
together with "keytab:".
See also:
- upstream libvirt commit fe772f24,
- Red Hat Bugzilla <https://bugzilla.redhat.com/show_bug.cgi?id=1018434>.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
ACKed-By: Cole Robinson <crobinso@redhat.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Check whether the firmware is not hidden by other memory regions.
Qemu is started in paused mode: it shouldn't try to interpret generated
garbage.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The blob is 64K in size and contains 0x00..0xFF repeatedly.
The client code added to main() wouldn't make much sense in the long term.
It helps with debugging and it silences gcc about create_blob_file() being
unused, and we'll replace it in the next patch anyway.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The current two GTest cases, /i440fx/defaults and /i440fx/pam can share a
qemu process, but the next two cases will need dedicated instances. It is
messy (and order-dependent) to dynamically configure GTest cases one by
one to start, stop, or keep the current qtest (*); let's just have each
GTest work with its own qtest. The performance difference should be
negligible.
(*) As g_test_run() can be invoked at most once per process startup, and
it runs GTest cases in sequence, we'd need clumsy data structures to
control each GTest case to start/stop/keep the qemu instance. Or, we'd
have to code the same information into the test methods themselves, which
would make them even more order-dependent.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Similarly to commit 1d9358e6
("libqtest: New qtest_end() to go with qtest_start()").
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This patch allows the user to usefully specify
-drive file=img_1,if=pflash,format=raw,readonly \
-drive file=img_2,if=pflash,format=raw
on the command line. The flash images will be mapped under 4G in their
reverse unit order -- that is, with their base addresses progressing
downwards, in increasing unit order.
(The unit number increases with command line order if not explicitly
specified.)
This accommodates the following use case: suppose that OVMF is split in
two parts, a writeable host file for non-volatile variable storage, and a
read-only part for bootstrap and decompressible executable code.
The binary code part would be read-only, centrally managed on the host
system, and passed in as unit 0. The variable store would be writeable,
VM-specific, and passed in as unit 1.
00000000ffe00000-00000000ffe1ffff (prio 0, R-): system.flash1
00000000ffe20000-00000000ffffffff (prio 0, R-): system.flash0
(If the guest tries to write to the flash range that is backed by the
read-only drive, pflash_update() is never called; various flash
programming/erase errors are returned to the guest instead. See the
callers of pflash_update(), and the initialization of "pfl->ro", in
"hw/block/pflash_cfi01.c".)
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
qemu_opts_parse() can always return NULL, even if the QemuOptsList.desc in
question would be trivial to satisfy (eg. because it's empty). For
example:
qemu_opts_parse()
opts_parse()
qemu_opts_create()
id_wellformed()
In practice:
$ .../qemu-system-x86_64 -acpitable id=3
qemu-system-x86_64: -acpitable id=3: Parameter 'id' expects an identifier
**
ERROR:vl.c:3491:main: assertion failed: (opts != NULL)
Aborted (core dumped)
$ .../qemu-system-x86_64 -smbios id=3
qemu-system-x86_64: -smbios id=3: Parameter 'id' expects an identifier
Segmentation fault (core dumped)
I checked all qemu_opts_parse() invocations (and all drive_def()
invocations too, because it blindly forwards the former's retval). Only
the two above examples look problematic.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-id: 1385658779-7529-1-git-send-email-lersek@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
"Incoming" function prototypes and "outgoing" function calls must match
reality. Implemented using the "struct BlockDriver" definition in
"include/block/block_int.h", and gcc errors & warnings.
v1->v2:
On 08/20/13 09:51, Kevin Wolf wrote:
> Am 18.08.2013 um 16:29 hat Paolo Bonzini geschrieben:
>> Il 16/08/2013 16:15, Laszlo Ersek ha scritto:
>>> +static int raw_reopen_prepare(BDRVReopenState *reopen_state,
>>> + BlockReopenQueue *queue, Error **errp)
>>> {
>>> - return bdrv_reopen_prepare(bs->file);
>>> + BDRVReopenState tmp = *reopen_state;
>>> +
>>> + tmp.bs = tmp.bs->file;
>>> + return bdrv_reopen_prepare(&tmp, queue, errp);
>>> }
>>
>> This should just return zero, my fault.
>
> Which is because bdrv_reopen_queue() already queues bs->file for reopen.
> The simple return 0; implementation is shared by all other format drivers
> that support reopening images.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On 08/05/13 15:03, Paolo Bonzini wrote:
>
> [...]
>
> 5) Formats are registered with bdrv_register (takes a BlockDriver*). You
> also need to pass the caller of bdrv_register to block_init.
Fill in the BlockDriver structure with the raw_*() functions that have
been added to "block/raw_bsd.c", in the order the fields are defined in
"include/block/block_int.h".
I needed more explanation / naming examples for registering the driver
than what Paolo gave me, so I copied / adapted from "block/qcow2.c". The
parts I took as basis for modification are blamed on
commit 5efa9d5a8b
Author: Anthony Liguori <aliguori@us.ibm.com>
Date: Sat May 9 17:03:42 2009 -0500
Convert block infrastructure to use new module init functionality
commit 20d97356c9
Author: Blue Swirl <blauwirbel@gmail.com>
Date: Fri Apr 23 20:19:47 2010 +0000
Fix OpenBSD build
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On 08/05/13 15:03, Paolo Bonzini wrote:
>
> [...]
>
> 4) There is another member, .create_options, which is an array of
> QEMUOptionParameter structs, terminated by an all-zero item. The only
> option you need is for the virtual disk size. You will find something
> to copy from in other block drivers, for example block/qcow2.c.
Code taken and adapted from "block/qcow2.c", as suggested. The code being
copied/modified is blamed on
commit 20d97356c9
Author: Blue Swirl <blauwirbel@gmail.com>
Date: Fri Apr 23 20:19:47 2010 +0000
Fix OpenBSD build
and
commit 7c80ab3f21
Author: Jes Sorensen <Jes.Sorensen@redhat.com>
Date: Fri Dec 17 16:02:39 2010 +0100
block/qcow2.c: rename qcow_ functions to qcow2_
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On 08/05/13 15:03, Paolo Bonzini wrote:
>
> [...]
>
> 3) These members are special
>
> .format_name is the string "raw"
> .bdrv_open raw_open should set bs->sg to bs->file->sg and return 0
> .bdrv_close raw_close should do nothing
> .bdrv_probe raw_probe should just return 1.
v1->v2:
On 08/20/13 10:11, Kevin Wolf wrote:
> Am 16.08.2013 um 16:15 hat Laszlo Ersek geschrieben:
>> +static int raw_probe(void)
>> +{
>> + return 1;
>> +}
>
> Maybe add a comment here like "smallest possible positive score so that
> raw is used if and only if no other block driver works".
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On 08/05/13 15:03, Paolo Bonzini wrote:
>
> [...]
>
> 2) This is also a simple forwarder function:
>
> .bdrv_create
>
> but there is no BlockDriverState argument so the forwarded-to function
> does not have a bs->file argument either. The forwarded-to function is
> bdrv_create_file.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On 08/05/13 15:03, Paolo Bonzini wrote:
>
> [...]
>
> 1) BlockDriver is a struct in which these function members are
> interesting:
>
> .bdrv_reopen_prepare
> .bdrv_co_readv
> .bdrv_co_writev
> .bdrv_co_is_allocated
> .bdrv_co_write_zeroes
> .bdrv_co_discard
> .bdrv_getlength
> .bdrv_get_info
> .bdrv_truncate
> .bdrv_is_inserted
> .bdrv_media_changed
> .bdrv_eject
> .bdrv_lock_medium
> .bdrv_ioctl
> .bdrv_aio_ioctl
> .bdrv_has_zero_init
>
> They should be implemented as simple forwarders (see above). There are
> 16 functions listed here, you can easily see how this already accounts
> for 100+ SLOC roughly...
>
> The implementations of bdrv_co_readv and bdrv_co_writev should also call
> BLKDBG_EVENT on bs->file too, before forwarding to bs->file. The events
> to be generated are BLKDBG_READ_AIO and BLKDBG_WRITE_AIO.
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Laszlo Ersek