Paolo Bonzini [Wed, 29 Oct 2014 10:17:19 +0000 (11:17 +0100)]
virtio: link the rng backend through an alias property
The virtio-rng backend is currently linked twice, once in the proxy
device (e.g. virtio-rng-pci) and once in virtio-rng-device. This causes
a double unref of the backend when the parent device is unplugged.
To fix this, make the proxy device use an alias, similar to what is
already being done for the iothread link.
Jan Kiszka [Mon, 27 Oct 2014 14:13:02 +0000 (14:13 +0000)]
Revert "main-loop.c: Handle SIGINT, SIGHUP and SIGTERM synchronously"
This reverts commit 15124e142034d21341ec9f1a304a1dc5a6c25681. It breaks
debuggability of qemu and is no longer needed as the problem has
now been addressed in a different way.
Instead we provide a comment about why these signals must be
handled asynchronously.
Jan Kiszka [Mon, 27 Oct 2014 14:05:10 +0000 (14:05 +0000)]
Make qemu_shutdown_requested signal-safe
qemu_shutdown_requested may be interrupted by qemu_system_killed. If the
latter sets shutdown_requested after qemu_shutdown_requested has read it
but before it was cleared, the shutdown event is lost. Fix this by using
atomic_xchg.
This provides a different fix for the problem which commit 15124e142
attempts to deal with. That commit breaks use of ^C to drop into gdb,
and so this approach is better (and 15124e142 can be reverted).
Peter Maydell [Fri, 24 Oct 2014 11:40:28 +0000 (12:40 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20141024' into staging
target-arm queue:
* remove pointless 'info pcmcia' and a lot of now-dead code
* register ARM cpu reset handlers even if not using -kernel
* update to libvixl 1.6
* various minor code cleanups
* support PSCI under TCG ('virt' machine can now be shut down,
SMP configurations work)
* correct the sense of the AArch64 DCZID DZP bit
* report a valid L1Ip field in CTR_EL0 for CPU type "any"
* correctly UNDEF writes to FPINST/FPINST2 from EL0
* more preparatory code refactoring for EL2/EL3 support
# gpg: Signature made Fri 24 Oct 2014 12:35:52 BST using RSA key ID 14360CDE
# gpg: Good signature from "Peter Maydell <[email protected]>"
* remotes/pmaydell/tags/pull-target-arm-20141024: (23 commits)
target-arm: A32: Emulate the SMC instruction
target-arm: make arm_current_el() return EL3
target-arm: rename arm_current_pl to arm_current_el
target-arm: reject switching to monitor mode
target-arm: add arm_is_secure() function
target-arm: increase arrays of registers R13 & R14
target-arm: correctly UNDEF writes to FPINST/FPINST2 from EL0
target-arm: Report a valid L1Ip field in CTR_EL0 for CPU type "any"
target-arm: Correct sense of the DCZID DZP bit
arm/virt: enable PSCI emulation support for system emulation
target-arm: add emulation of PSCI calls for system emulation
target-arm: Add support for A32 and T32 HVC and SMC insns
target-arm: Handle SMC/HVC undef-if-no-ELx in pre_* helpers
target-arm: add missing PSCI constants needed for PSCI emulation
target-arm: do not set do_interrupt handlers for ARM and AArch64 user modes
target-arm: add powered off cpu state
omap_gpmc.c: Remove duplicate assignment
disas/libvixl/a64/instructions-a64.h: Remove unused constants
arm_gic: remove unused parameter.
disas/libvixl: Update to libvixl 1.6
...
Fabian Aggeler [Fri, 24 Oct 2014 11:19:14 +0000 (12:19 +0100)]
target-arm: make arm_current_el() return EL3
Make arm_current_el() return EL3 for secure PL1 and monitor mode.
Increase MMU modes since mmu_index is directly inferred from arm_
current_el(). Change assertion in arm_el_is_aa64() to allow EL3.
Fabian Aggeler [Fri, 24 Oct 2014 11:19:14 +0000 (12:19 +0100)]
target-arm: add arm_is_secure() function
arm_is_secure() function allows to determine CPU security state
if the CPU implements Security Extensions/EL3.
arm_is_secure_below_el3() returns true if CPU is in secure state
below EL3.
Peter Maydell [Fri, 24 Oct 2014 11:19:14 +0000 (12:19 +0100)]
target-arm: correctly UNDEF writes to FPINST/FPINST2 from EL0
The ARM ARM requires that the FPINST and FPINST2 VFP control
registers are not accessible to code at EL0. We were already
correctly implementing this for reads of these registers; add
the missing check for the write code path.
Peter Maydell [Fri, 24 Oct 2014 11:19:13 +0000 (12:19 +0100)]
target-arm: Report a valid L1Ip field in CTR_EL0 for CPU type "any"
For the CPU type "any" (only used with linux-user) we were reporting
the L1Ip field as 0b00, which is reserved. Change this field to 0b10
instead, indicating a VIPT icache as the comment describes.
Peter Maydell [Fri, 24 Oct 2014 11:19:13 +0000 (12:19 +0100)]
target-arm: Correct sense of the DCZID DZP bit
The DZP bit in the DCZID system register should be set if
the control bits which prohibit use of the DC ZVA instruction
have been set (it stands for Data Zero Prohibited). However
we had the sense of the test inverted; fix this so that the
bit reads correctly.
To avoid this regressing the behaviour of the user-mode
emulator, we must set the DZE bit in the SCTLR for that
config so that userspace continues to see DZP as zero (it
was getting the correct result by accident previously).
Rob Herring [Fri, 24 Oct 2014 11:19:13 +0000 (12:19 +0100)]
arm/virt: enable PSCI emulation support for system emulation
Now that we have PSCI emulation, enable it for the virt platform.
This simplifies the virt machine a bit now that PSCI no longer
needs to be a KVM only feature.
Rob Herring [Fri, 24 Oct 2014 11:19:13 +0000 (12:19 +0100)]
target-arm: add emulation of PSCI calls for system emulation
Add support for handling PSCI calls in system emulation. Both version
0.1 and 0.2 of the PSCI spec are supported. Platforms can enable support
by setting the "psci-conduit" QOM property on the cpus to SMC or HVC
emulation and having a PSCI binding in their dtb.
Signed-off-by: Rob Herring <[email protected]> Signed-off-by: Ard Biesheuvel <[email protected]> Signed-off-by: Peter Maydell <[email protected]>
Message-id: 1412865028[email protected]
[PMM: made system reset/off PSCI functions power down the CPU so
we obey the PSCI API requirement never to return from them;
rearranged how the code is plumbed into the exception system,
so that we split "is this a valid call?" from "do the call"] Signed-off-by: Peter Maydell <[email protected]>
Peter Maydell [Fri, 24 Oct 2014 11:19:13 +0000 (12:19 +0100)]
target-arm: Add support for A32 and T32 HVC and SMC insns
Add support for HVC and SMC instructions to the A32 and
T32 decoder. Using these for real exceptions to EL2 or EL3
is currently not supported (the do_interrupt routine does
not handle them) but we require the instruction support to
implement PSCI.
Peter Maydell [Fri, 24 Oct 2014 11:19:12 +0000 (12:19 +0100)]
target-arm: Handle SMC/HVC undef-if-no-ELx in pre_* helpers
SMC must UNDEF if EL3 is not implemented; similarly HVC UNDEFs
if EL2 is not implemented. Move the handling of this from
translate-a64.c into the pre_smc and pre_hvc helper functions.
This is necessary because use of these instructions for PSCI
takes precedence over this UNDEF case, and we can't tell if
this is a PSCI call until runtime.
Rob Herring [Fri, 24 Oct 2014 11:19:12 +0000 (12:19 +0100)]
target-arm: do not set do_interrupt handlers for ARM and AArch64 user modes
User mode emulation should never get interrupts and thus should not
use the system emulation exception handler function. Remove the reference,
and '#ifndef USER_MODE_ONLY' the function itself as well, so that we can add
system mode only functionality to it.
Rob Herring [Fri, 24 Oct 2014 11:19:12 +0000 (12:19 +0100)]
target-arm: add powered off cpu state
Add tracking of cpu power state in order to support powering off of
cores in system emuluation. The initial state is determined by the
start-powered-off QOM property.
The instructions-a64.h header defines a number of floating point
constants whose initializers are function calls. gcc 5 will warn
if these constants are not used by the C or C++ file which includes
the header, because they imply a runtime cost. Since for the files
QEMU uses from libvixl we don't use these constants at all, just
remove them.
Upstream intend to fix these by shifting to an 'extern const' in
the header plus definition in a suitable source file, so we can
drop this patch when we sync with the upcoming libvixl 1.7.
The related compiling error:
CXX disas/arm-a64.o
In file included from /upstream/qemu/disas/libvixl/a64/disasm-a64.h:32:0,
from disas/arm-a64.cc:20:
disas/libvixl/a64/instructions-a64.h:98:13: error: 'vixl::kFP32PositiveInfinity' defined but not used [-Werror=unused-variable]
const float kFP32PositiveInfinity = rawbits_to_float(0x7f800000);
^
disas/libvixl/a64/instructions-a64.h:99:13: error: 'vixl::kFP32NegativeInfinity' defined but not used [-Werror=unused-variable]
const float kFP32NegativeInfinity = rawbits_to_float(0xff800000);
^
disas/libvixl/a64/instructions-a64.h:100:14: error: 'vixl::kFP64PositiveInfinity' defined but not used [-Werror=unused-variable]
const double kFP64PositiveInfinity =
^
disas/libvixl/a64/instructions-a64.h:102:14: error: 'vixl::kFP64NegativeInfinity' defined but not used [-Werror=unused-variable]
const double kFP64NegativeInfinity =
^
disas/libvixl/a64/instructions-a64.h:107:21: error: 'vixl::kFP64SignallingNaN' defined but not used [-Werror=unused-variable]
static const double kFP64SignallingNaN =
^
disas/libvixl/a64/instructions-a64.h:109:20: error: 'vixl::kFP32SignallingNaN' defined but not used [-Werror=unused-variable]
static const float kFP32SignallingNaN = rawbits_to_float(0x7f800001);
^
disas/libvixl/a64/instructions-a64.h:112:21: error: 'vixl::kFP64QuietNaN' defined but not used [-Werror=unused-variable]
static const double kFP64QuietNaN =
^
disas/libvixl/a64/instructions-a64.h:114:20: error: 'vixl::kFP32QuietNaN' defined but not used [-Werror=unused-variable]
static const float kFP32QuietNaN = rawbits_to_float(0x7fc00001);
^
disas/libvixl/a64/instructions-a64.h:117:21: error: 'vixl::kFP64DefaultNaN' defined but not used [-Werror=unused-variable]
static const double kFP64DefaultNaN =
^
disas/libvixl/a64/instructions-a64.h:119:20: error: 'vixl::kFP32DefaultNaN' defined but not used [-Werror=unused-variable]
static const float kFP32DefaultNaN = rawbits_to_float(0x7fc00000);
^
cc1plus: all warnings being treated as errors
make: *** [disas/arm-a64.o] Error 1
Signed-off-by: Chen Gang <[email protected]>
[PMM: Rewrote the commit message a little] Signed-off-by: Peter Maydell <[email protected]>
Peter Maydell [Fri, 24 Oct 2014 11:19:11 +0000 (12:19 +0100)]
disas/libvixl: Update to libvixl 1.6
Update our copy of libvixl to upstream 1.6. There are no
changes of any particular interest to QEMU, so this is simply
keeping up with current upstream.
Ard Biesheuvel [Fri, 24 Oct 2014 11:19:11 +0000 (12:19 +0100)]
hw/arm/boot: register cpu reset handlers if using -bios
Move the registering of CPU reset handlers to before the point where
we leave the function in the -bios (not -kernel) case, so CPU reset
works correctly with -bios as well.
This command lists PCMCIA sockets and cards. Only a few ARM boards
have sockets (akita, borzoi, connex, mainstone, spitz, terrier, tosa,
verdex, z2), the only card is the DSCM-1xxxx Hitachi Microdrive (qdev
"microdrive"), and it is only inserted during machine init, if ever.
So this command doesn't really tell anybody anything new so far.
Moreover, pcmcia_socket_unregister() has a use-after-free bug, flagged
by Coverity. Has never been used, because there has never been code
to eject a PCMCIA card.
Peter Maydell [Fri, 24 Oct 2014 10:33:46 +0000 (11:33 +0100)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block patches
# gpg: Signature made Thu 23 Oct 2014 18:56:05 BST using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
* remotes/kevin/tags/for-upstream: (32 commits)
qemu-img: Print error if check failed
block: char devices on FreeBSD are not behind a pager
iotests: Add test for qcow2 L1 table update
qcow2: Do not overflow when writing an L1 sector
iotests: Add test for map commands
qemu-io: Respect early image end for map
block: Respect underlying file's EOF
docs/qcow2: Limit refcount_order to [0, 6]
docs/qcow2: Correct refcount_block_entries
qcow2: Drop REFCOUNT_SHIFT
iotests: Add test for potentially damaging repairs
iotests: Fix test outputs
qcow2: Clean up after refcount rebuild
qcow2: Rebuild refcount structure during check
qcow2: Do not perform potentially damaging repairs
qcow2: Fix refcount blocks beyond image end
qcow2: Reuse refcount table in calculate_refcounts()
qcow2: Let inc_refcounts() resize the reftable
qcow2: Let inc_refcounts() return -errno
qcow2: Split fail code in L1 and L2 checks
...
Max Reitz [Thu, 23 Oct 2014 13:29:12 +0000 (15:29 +0200)]
qemu-img: Print error if check failed
Currently, if bdrv_check() fails either by returning -errno or having
check_errors set, qemu-img check just exits with 1 after having told the
user that there were no errors on the image. This is bad.
Instead of printing the check result if there were internal errors which
were so bad that bdrv_check() could not even complete with 0 as a return
value, qemu-img check should inform the user about the error.
Peter Maydell [Thu, 23 Oct 2014 16:05:15 +0000 (17:05 +0100)]
Merge remote-tracking branch 'remotes/qmp-unstable/tags/for-upstream' into staging
QMP patches
# gpg: Signature made Thu 23 Oct 2014 16:05:52 BST using RSA key ID E24ED5A7
# gpg: Good signature from "Luiz Capitulino <[email protected]>"
* remotes/qmp-unstable/tags/for-upstream:
monitor: delete device_del_bus_completion
monitor: add del completion for peripheral device
qdev: add qdev_build_hotpluggable_device_list helper
MAINTAINERS: add entry for qobject files
dump: Turn some functions to void to make code cleaner
dump: Propagate errors into qmp_dump_guest_memory()
virtio-balloon: Tweak recent fix for integer overflow
Roger Pau Monne [Tue, 21 Oct 2014 14:03:03 +0000 (16:03 +0200)]
block: char devices on FreeBSD are not behind a pager
Introduce a new flag to mark devices that require requests to be aligned and
replace the usage of BDRV_O_NOCACHE and O_DIRECT with this flag when
appropriate.
If a character device is used as a backend on a FreeBSD host set this flag
unconditionally.
Max Reitz [Thu, 16 Oct 2014 13:25:56 +0000 (15:25 +0200)]
qcow2: Do not overflow when writing an L1 sector
While writing an L1 table sector, qcow2_write_l1_entry() copies the
respective range from s->l1_table to the local "buf" array. The size of
s->l1_table does not have to be a multiple of L1_ENTRIES_PER_SECTOR;
thus, limit the index which is used for copying all entries to the L1
size.
Max Reitz [Wed, 22 Oct 2014 15:00:16 +0000 (17:00 +0200)]
qemu-io: Respect early image end for map
bdrv_is_allocated() may report zero clusters which most probably means
the image (file) is shorter than expected. Respect this case in order to
avoid an infinite loop.
Max Reitz [Wed, 22 Oct 2014 15:00:15 +0000 (17:00 +0200)]
block: Respect underlying file's EOF
When falling through to the underlying file in
bdrv_co_get_block_status(), if it returns that the query offset is
beyond the file end (by setting *pnum to 0), return the range to be
zero and do not let the number of sectors for which information could be
obtained be overwritten.
Max Reitz [Tue, 2 Sep 2014 22:25:09 +0000 (00:25 +0200)]
docs/qcow2: Limit refcount_order to [0, 6]
Specify the upper limit of refcount_order to be 6 (that is,
refcount_bits = 64). Any larger value does not make much sense when all
offsets, sizes, cluster counts etc. "only" have a width of 64 bit as
well, and very large values would be very difficult to support.
Therefore, just cap it at the largest reasonable value.
Max Reitz [Tue, 2 Sep 2014 22:25:08 +0000 (00:25 +0200)]
docs/qcow2: Correct refcount_block_entries
A refblock entry may have a different size than 16 bits, it may even be
smaller than a byte. Correct the refcount_block_entries calculation
accordingly.
Max Reitz [Wed, 22 Oct 2014 12:09:43 +0000 (14:09 +0200)]
iotests: Add test for potentially damaging repairs
There are certain cases where repairing a qcow2 image might actually
damage it further (or rather, where repairing it has in fact damaged it
further with the old qcow2 check implementation). This should not
happen, so add a test for these cases.
Furthermore, the repair function now repairs refblocks beyond the image
end by resizing the image accordingly. Add several tests for this as
well.
Max Reitz [Wed, 22 Oct 2014 12:09:42 +0000 (14:09 +0200)]
iotests: Fix test outputs
039, 060 and 061 all create images with referenced clusters having a
refcount of 0. Because previous commits changed handling of such errors,
these tests now have a different output. Fix it.
Furthermore, 060 created a refblock with a refcount greater than one
which now results in having to rebuild the refcount structure as well.
Max Reitz [Wed, 22 Oct 2014 12:09:41 +0000 (14:09 +0200)]
qcow2: Clean up after refcount rebuild
Because the old refcount structure will be leaked after having rebuilt
it, we need to recalculate the refcounts and run a leak-fixing operation
afterwards (if leaks should be fixed at all).
Max Reitz [Wed, 22 Oct 2014 12:09:40 +0000 (14:09 +0200)]
qcow2: Rebuild refcount structure during check
The previous commit introduced the "rebuild" variable to qcow2's
implementation of the image consistency check. Now make use of this by
adding a function which creates a completely new refcount structure
based solely on the in-memory information gathered before.
The old refcount structure will be leaked, however. This leak will be
dealt with in a follow-up commit.
Max Reitz [Wed, 22 Oct 2014 12:09:39 +0000 (14:09 +0200)]
qcow2: Do not perform potentially damaging repairs
If a referenced cluster has a refcount of 0, increasing its refcount may
result in clusters being allocated for the refcount structures. This may
overwrite the referenced cluster, therefore we cannot simply increase
the refcount then.
In such cases, we can either try to replicate all the refcount
operations solely for the check operation, basing the allocations on the
in-memory refcount table; or we can simply rebuild the whole refcount
structure based on the in-memory refcount table. Since the latter will
be much easier, do that.
To prepare for this, introduce a "rebuild" boolean which should be set
to true whenever a fix is rather dangerous or too complicated using the
current refcount structures. Another example for this is refcount blocks
being referenced more than once.
Max Reitz [Wed, 22 Oct 2014 12:09:38 +0000 (14:09 +0200)]
qcow2: Fix refcount blocks beyond image end
If the qcow2 check function detects a refcount block located beyond the
image end, grow the image appropriately. This cannot break anything and
is the logical fix for such a case.
Max Reitz [Wed, 22 Oct 2014 12:09:36 +0000 (14:09 +0200)]
qcow2: Let inc_refcounts() resize the reftable
Now that the refcount table can be passed around by reference, do that
for inc_refcounts() (and subsequently check_refcounts_l1() and
check_refcounts_l2()) and use it for resizing it when a cluster after
the image end is encountered.
Max Reitz [Wed, 22 Oct 2014 12:09:35 +0000 (14:09 +0200)]
qcow2: Let inc_refcounts() return -errno
As of a future patch, inc_refcounts() will have to throw errors which
are generally signaled by returning -errno. Therefore, let it return an
integer which is either 0 for success or -errno and handle the -errno
case in all callers.
Max Reitz [Wed, 22 Oct 2014 12:09:34 +0000 (14:09 +0200)]
qcow2: Split fail code in L1 and L2 checks
Instead of printing out an error message, incrementing check_errors and
returning a fixed -errno, just do cleanups and return -ret, with ret set
by the code which threw the exception (jumped to the fail label).
Also, increment check_errors on error in check_refcounts_l2().
Max Reitz [Wed, 22 Oct 2014 12:09:31 +0000 (14:09 +0200)]
qcow2: Use sizeof(**refcount_table)
When implementing variable refcounts, we want to be able to easily find
all the places in qemu which are tied to a certain refcount order.
Replace sizeof(uint16_t) in the check code by sizeof(**refcount_table)
so we can later find it more easily.
Max Reitz [Wed, 22 Oct 2014 12:09:28 +0000 (14:09 +0200)]
qcow2: Calculate refcount block entry count
The size of a refblock entry is (in theory) variable; calculate
therefore the number of entries per refblock and the according bit shift
(1 << x == entry count) when opening an image.
Peter Lieven [Mon, 20 Oct 2014 11:47:11 +0000 (13:47 +0200)]
block: qemu-iotests change _supported_proto to file once more.
In preparation to possible automatic regression and performance
testing for the block layer I found that the iotests don't work
for all protocols anymore.
In commit 1f7bf7d0 I started to change supported protocols from
generic to file for various tests. Unfortunately, some tests
added in the meantime again carry generic protocol altough they
can only work with file because they require local file access.
The other way around for some tests that only support file I added
NFS protocol after confirming they work.
zhanghailiang [Thu, 9 Oct 2014 06:13:11 +0000 (14:13 +0800)]
dump: Turn some functions to void to make code cleaner
Functions shouldn't return an error code and an Error object at the same time.
Turn all these functions that returning Error object to void.
We also judge if a function success or fail by reference to the local_err.
zhanghailiang [Thu, 9 Oct 2014 06:13:10 +0000 (14:13 +0800)]
dump: Propagate errors into qmp_dump_guest_memory()
The code calls dump_error() on error, and even passes it a suitable
message. However, the message is thrown away, and its callers pass
up only success/failure. All qmp_dump_guest_memory() can do is set
a generic error.
Propagate the errors properly, so qmp_dump_guest_memory() can return
a more useful error.
virtio-balloon: Tweak recent fix for integer overflow
Commit 1f9296b avoids "other kinds of overflow" by limiting the
polling interval to UINT_MAX. The computations to protect are done in
64 bits. This is indeed safe when unsigned is 32 bits, as it commonly
is. It isn't when unsigned is 64 bits. Purely theoretical; I'm not
aware of such a system. Limit it to UINT32_MAX instead.
Peter Maydell [Wed, 22 Oct 2014 20:42:33 +0000 (21:42 +0100)]
Merge remote-tracking branch 'remotes/mdroth/tags/qga-pull-2014-10-22-tag' into staging
qga: remove readdir_r usage and fix use-after-free
Signed-off-by: Michael Roth <[email protected]>
# gpg: Signature made Wed 22 Oct 2014 13:56:19 BST using RSA key ID F108B584
# gpg: Can't check signature: public key not found
* remotes/mdroth/tags/qga-pull-2014-10-22-tag:
qga: Rewrite code where using readdir_r
Peter Maydell [Wed, 22 Oct 2014 17:43:35 +0000 (18:43 +0100)]
Merge remote-tracking branch 'remotes/bkoppelmann/tags/pull-tricore-20141021' into staging
TriCore ABS, ABSB, B, BIT, BO instructions added
# gpg: Signature made Tue 21 Oct 2014 17:47:32 BST using RSA key ID 6B69CA14
# gpg: Good signature from "Bastian Koppelmann <[email protected]>"
* remotes/bkoppelmann/tags/pull-tricore-20141021:
target-tricore: Add instructions of BO opcode format
target-tricore: Add instructions of BIT opcode format
target-tricore: Add instructions of B opcode format
target-tricore: Add instructions of ABS, ABSB opcode format
target-tricore: Cleanup and Bugfixes
Peter Maydell [Wed, 22 Oct 2014 15:39:49 +0000 (16:39 +0100)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block patches
# gpg: Signature made Mon 20 Oct 2014 13:04:09 BST using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
* remotes/kevin/tags/for-upstream: (28 commits)
block: Make device model's references to BlockBackend strong
block: Lift device model API into BlockBackend
blockdev: Convert qmp_eject(), qmp_change_blockdev() to BlockBackend
block/qapi: Convert qmp_query_block() to BlockBackend
blockdev: Fix blockdev-add not to create DriveInfo
blockdev: Drop superfluous DriveInfo member id
pc87312: Drop unused members of PC87312State
ide: Complete conversion from BlockDriverState to BlockBackend
hw: Convert from BlockDriverState to BlockBackend, mostly
virtio-blk: Rename VirtIOBlkConf variables to conf
virtio-blk: Drop redundant VirtIOBlock member conf
block: Rename BlockDriverCompletionFunc to BlockCompletionFunc
block: Rename BlockDriverAIOCB* to BlockAIOCB*
block: Eliminate DriveInfo member bdrv, use blk_by_legacy_dinfo()
block: Merge BlockBackend and BlockDriverState name spaces
block: Eliminate BlockDriverState member device_name[]
block: Eliminate bdrv_iterate(), use bdrv_next()
blockdev: Eliminate drive_del()
block: Make BlockBackend own its BlockDriverState
block: Code motion to get rid of stubs/blockdev.c
...
Move device model attachment / detachment and the BlockDevOps device
model callbacks and their wrappers from BlockDriverState to
BlockBackend.
Wrapper calls in block.c change from
bdrv_dev_FOO_cb(bs, ...)
to
if (bs->blk) {
bdrv_dev_FOO_cb(bs->blk, ...);
}
No change, because both bdrv_dev_change_media_cb() and
bdrv_dev_resize_cb() do nothing when no device model is attached, and
a device model can be attached only when bs->blk.
blockdev: Convert qmp_eject(), qmp_change_blockdev() to BlockBackend
Much more command code needs conversion. I'm converting these now
because they're using bdrv_dev_* functions, which I'm about to lift
into BlockBackend.
blockdev: Fix blockdev-add not to create DriveInfo
blockdev_init() always creates a DriveInfo, but only drive_new() fills
it in. qmp_blockdev_add() leaves it blank. This results in a drive
with type = IF_IDE, bus = 0, unit = 0. Screwed up in commit ee13ed1c.
Board initialization code looking for IDE drive (0,0) can pick up one
of these bogus drives. The QMP command has to execute really early to
be visible. Not sure how likely that is in practice.
Fix by creating DriveInfo in drive_new(). Block backends created by
blockdev-add don't get one.
Breaks the test for "has been created by qmp_blockdev_add()" in
blockdev_mark_auto_del() and do_drive_del(), because it changes the
value of dinfo && !dinfo->enable_auto_del from true to false. Simply
test !dinfo instead.
Leaves DriveInfo member enable_auto_del unused. Drop it.
A few places assume a block backend always has a DriveInfo. Fix them
up.
hw: Convert from BlockDriverState to BlockBackend, mostly
Device models should access their block backends only through the
block-backend.h API. Convert them, and drop direct includes of
inappropriate headers.
Just four uses of BlockDriverState are left:
* The Xen paravirtual block device backend (xen_disk.c) opens images
itself when set up via xenbus, bypassing blockdev.c. I figure it
should go through qmp_blockdev_add() instead.
* Device model "usb-storage" prompts for keys. No other device model
does, and this one probably shouldn't do it, either.
* ide_issue_trim_cb() uses bdrv_aio_discard() instead of
blk_aio_discard() because it fishes its backend out of a BlockAIOCB,
which has only the BlockDriverState.
* PC87312State has an unused BlockDriverState[] member.
I'll use BlockDriverAIOCB with block backends shortly, and the name is
going to fit badly there. It's a block layer thing anyway, not just a
block driver thing.
projects / qemu.git / log