Peter Maydell [Wed, 7 May 2014 12:47:25 +0000 (13:47 +0100)]
Merge remote-tracking branch 'remotes/afaerber/tags/qom-devices-for-peter' into staging
QOM/QTest infrastructure fixes and device conversions
* -device / device_add assertion fix
* QEMUMachine conversion to MachineClass
* Device error handling improvements
* QTest cleanups and test cases for some more PCI devices
* PortIO memory leak fixes
# gpg: Signature made Mon 05 May 2014 19:59:16 BST using RSA key ID 3E7E013F
# gpg: Good signature from "Andreas Färber <[email protected]>"
# gpg: aka "Andreas Färber <[email protected]>"
* remotes/afaerber/tags/qom-devices-for-peter:
PortioList: Store PortioList in device state
tests: Add EHCI qtest
tests: Add ioh3420 qtest
tests: Add intel-hda qtests
tests: Add es1370 qtest
tests: Add ac97 qtest
qtest: Be paranoid about accept() addrlen argument
qtest: Add error reporting to socket_accept()
qtest: Assure that init_socket()'s listen() does not fail
MAINTAINERS: Document QOM
arm: Clean up fragile use of error_is_set() in realize() methods
qom: Clean up fragile use of error_is_set() in set() methods
hw: Consistently name Error ** objects errp, and not err
hw: Consistently name Error * objects err, and not errp
machine: Remove QEMUMachine indirection from MachineClass
machine: Replace QEMUMachine by MachineClass in accelerator configuration
vl.c: Replace QEMUMachine with MachineClass in QEMUMachineInitArgs
machine: Copy QEMUMachine's fields to MachineClass
machine: Remove obsoleted field from QEMUMachine
qdev: Fix crash by validating the object type
Peter Maydell [Tue, 6 May 2014 11:23:05 +0000 (12:23 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-smbios-2' into staging
smbios: make qemu generate smbios tables.
# gpg: Signature made Mon 05 May 2014 12:20:27 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg: aka "Gerd Hoffmann <[email protected]>"
# gpg: aka "Gerd Hoffmann (private) <[email protected]>"
* remotes/kraxel/tags/pull-smbios-2:
SMBIOS: Build aggregate smbios tables and entry point
SMBIOS: Use bitmaps to prevent incompatible comand line options
SMBIOS: Use macro to set smbios defaults
SMBIOS: Update header file definitions
SMBIOS: Rename symbols to better reflect future use
E820: Add interface for accessing e820 table
pc: add 2.1 machine type
Peter Maydell [Tue, 6 May 2014 09:56:38 +0000 (10:56 +0100)]
Merge remote-tracking branch 'remotes/riku/linux-user-for-upstream' into staging
* remotes/riku/linux-user-for-upstream:
linux-user: fix getrusage and wait4 failures with invalid rusage struct
linux-user/elfload.c: Support ARM HWCAP2 flags
linux-user/elfload.c: Fix A64 code which was incorrectly acting like A32
linux-user/elfload.c: Update ARM HWCAP bits
linux-user/elfload.c: Fix incorrect ARM HWCAP bits
linux-user: remove configure option for setting uname release
linux-user: move uname functions to uname.c
linux-user: rename cpu-uname -> uname
linux-user/signal.c: Set fault address in AArch64 signal info
linux-user: avoid using glibc internals in _syscall5 and in definition of target_sigevent struct
linux-user: Handle arches with llseek instead of _llseek
linux-user: Add support for SCM_CREDENTIALS.
linux-user: Move if-elses to a switch statement.
linux-user: Assert stack used for auxvec, envp, argv
linux-user: Add /proc/self/exe open forwarding
PortioList is an abstraction used for construction of MemoryRegionPortioList
from MemoryRegionPortio. It can be used later to unmap created memory regions.
It also requires proper cleanup because some of the memory inside is allocated
dynamically.
By moving PortioList ot device state we make it possible to cleanup later and
avoid leaking memory.
This change spans several target platforms. The following testcases cover all
changed lines:
qemu-system-ppc -M prep
qemu-system-i386 -vga qxl
qemu-system-i386 -M isapc -soundhw adlib -device ib700,id=watchdog0,bus=isa.0
Andreas Färber [Thu, 17 Apr 2014 17:21:12 +0000 (19:21 +0200)]
qtest: Be paranoid about accept() addrlen argument
POSIX specifies that address_len shall on output specify the length of
the stored address; it does not however specify whether it may get
updated on failure as well to, e.g., zero.
In case EINTR occurs, re-initialize the variable to the desired value.
arm: Clean up fragile use of error_is_set() in realize() methods
Using error_is_set(ERRP) to find out whether a function failed is
either wrong, fragile, or unnecessarily opaque. It's wrong when ERRP
may be null, because errors go undetected when it is. It's fragile
when proving ERRP non-null involves a non-local argument. Else, it's
unnecessarily opaque (see commit 84d18f0).
I guess the error_is_set(errp) in the DeviceClass realize() methods
are merely fragile right now, because I can't find a call chain that
passes a null errp argument.
Make the code more robust and more obviously correct: receive the
error in a local variable, then propagate it through the parameter.
qom: Clean up fragile use of error_is_set() in set() methods
Using error_is_set(ERRP) to find out whether a function failed is
either wrong, fragile, or unnecessarily opaque. It's wrong when ERRP
may be null, because errors go undetected when it is. It's fragile
when proving ERRP non-null involves a non-local argument. Else, it's
unnecessarily opaque (see commit 84d18f0).
I guess the error_is_set(errp) in the ObjectProperty set() methods are
merely fragile right now, because I can't find a call chain that
passes a null errp argument.
Make the code more robust and more obviously correct: receive the
error in a local variable, then propagate it through the parameter.
vl.c: Replace QEMUMachine with MachineClass in QEMUMachineInitArgs
QEMUMachine's fields are already in MachineClass. We can safely
make the switch because we copy them in machine_class_init() and
spapr_machine_class_init().
machine: Copy QEMUMachine's fields to MachineClass
In order to eliminate the QEMUMachine indirection,
add its fields directly to MachineClass.
Do not yet remove qemu_machine field because it is
still in use by sPAPR.
Signed-off-by: Marcel Apfelbaum <[email protected]>
[AF: Copied fields for sPAPR, too] Signed-off-by: Andreas Färber <[email protected]>
The machine registration flow is refactored to use the QOM functionality.
Instead of linking the machines into a list, each machine has a type
and the types can be traversed in the QOM way.
Amos Kong [Wed, 16 Apr 2014 01:57:14 +0000 (09:57 +0800)]
qdev: Fix crash by validating the object type
QEMU crashed when I try to list device parameters and the driver name is
actually an available bus name.
# qemu -device virtio-pci-bus,?
# qemu -device virtio-bus,?
# qemu -device virtio-serial-bus,?
qdev-monitor.c:212:qdev_device_help: Object 0x7fd932f50620 is not an
instance of type device
Aborted (core dumped)
We can also reproduce this bug by adding device from monitor, so it's
worth to fix the crash.
(qemu) device_add virtio-serial-bus
qdev-monitor.c:491:qdev_device_add: Object 0x7f5e89530920 is not an
instance of type device
Aborted (core dumped)
Petar Jovanovic [Tue, 8 Apr 2014 17:24:30 +0000 (19:24 +0200)]
linux-user: fix getrusage and wait4 failures with invalid rusage struct
Implementations of system calls getrusage and wait4 have not previously
handled correctly cases when incorrect address of struct rusage is
passed.
This change makes sure return values are correctly set for these cases.
Gabriel L. Somlo [Wed, 23 Apr 2014 13:42:42 +0000 (09:42 -0400)]
SMBIOS: Build aggregate smbios tables and entry point
Build an aggregate set of smbios tables and an entry point structure.
Insert tables and entry point into fw_cfg respectively under
"etc/smbios/smbios-tables" and "etc/smbios/smbios-anchor".
Machine types <= 2.0 will for now continue using field-by-field
overrides to SeaBIOS defaults, but for machine types 2.1 and up we
expect the BIOS to look for and use the aggregate tables generated
by this patch.
This defines a descriptor for OHCIState.
This changes some OHCIState field types to be migration compatible.
This adds a descriptor for OHCIPort.
This migrates the EOF timer if the USB was started at the time of
migration.
Gabriel L. Somlo [Wed, 23 Apr 2014 13:42:41 +0000 (09:42 -0400)]
SMBIOS: Use bitmaps to prevent incompatible comand line options
Replace existing smbios_check_collision() functionality with
a pair of bitmaps: have_binfile_bitmap and have_fields_bitmap.
Bits corresponding to each smbios type are set by smbios_entry_add(),
which also uses the bitmaps to ensure that binary blobs and field
values are never accepted for the same type.
These bitmaps will also be used in the future to decide whether
or not to build a full table for a given smbios type.
Gabriel L. Somlo [Wed, 23 Apr 2014 13:42:38 +0000 (09:42 -0400)]
SMBIOS: Rename symbols to better reflect future use
Rename the following symbols:
- smbios_set_type1_defaults() to the more general smbios_set_defaults();
- bool smbios_type1_defaults to the more general smbios_defaults;
- smbios_get_table() to smbios_get_table_legacy();
At the moment, 2.1 and 2.0 machines are identical.
As several people are working on incompatible changes
to the PC machine, collaboration will be made easier
by merging this place-holder.
Peter Maydell [Fri, 2 May 2014 13:45:15 +0000 (14:45 +0100)]
linux-user/elfload.c: Support ARM HWCAP2 flags
The ARM kernel has chosen to spill into the HWCAP2 ELF feature bit flags
early, even though it hasn't yet exhausted all 32 bits of the HWCAP word.
Add support for setting this in the same way we do for HWCAP.
Peter Maydell [Fri, 2 May 2014 13:45:14 +0000 (14:45 +0100)]
linux-user/elfload.c: Fix A64 code which was incorrectly acting like A32
The ARM target-specific code in elfload.c was incorrectly allowing
the 64-bit ARM target to use most of the existing 32-bit definitions:
most noticably this meant that our HWCAP bits passed to the guest
were wrong, and register handling when dumping core was totally
broken. Fix this by properly separating the 64 and 32 bit code,
since they have more differences than similarities.
Peter Maydell [Fri, 2 May 2014 13:45:13 +0000 (14:45 +0100)]
linux-user/elfload.c: Update ARM HWCAP bits
The kernel has added support for a number of new ARM HWCAP bits;
add them to QEMU, including support for setting them where we have
a corresponding CPU feature bit.
We were also incorrectly setting the VFPv3D16 HWCAP -- this means
"only 16 D registers", not "supports 16-bit floating point format";
since QEMU always has 32 D registers for VFPv3, we can just remove
the line that incorrectly set this bit.
The kernel does not set the HWCAP_FPA even if it is providing FPA
emulation via nwfpe, so don't set this bit in QEMU either.
Peter Maydell [Fri, 2 May 2014 13:45:12 +0000 (14:45 +0100)]
linux-user/elfload.c: Fix incorrect ARM HWCAP bits
The ELF HWCAP bits for ARM features THUMBEE, NEON, VFPv3 and VFPv3D16 are
all off by one compared to the kernel definitions. Fix this discrepancy
and add in the missing CRUNCH bit which was the cause of the off-by-one
error. (We don't emulate any of the CPUs which have that weird hardware,
so it's otherwise uninteresting to us.)
Riku Voipio [Tue, 4 Mar 2014 02:28:43 +0000 (04:28 +0200)]
linux-user: remove configure option for setting uname release
--enable-uname-release was a rather heavyweight hammer, as it allows
providing values less that UNAME_MINIMUM_RELEASE. Also, it affects
all built linux-user targets, which in most cases is not what user
wants.
Now that we have UNAME_MINIMUM_RELEASE for all linux-user platforms,
we can drop --enable-uname-release and the related CONFIG_UNAME_RELEASE
define.
Users can still override the variable with QEMU_UNAME=2.6.32 or -r
command line option. If distributors need to update a minimum version
for a specific target, it can be done by updating UNAME_MINIMUM_RELEASE.
James Hogan [Tue, 25 Mar 2014 21:51:08 +0000 (21:51 +0000)]
linux-user: Handle arches with llseek instead of _llseek
Recently merged kernel ports (such as OpenRISC and Meta) have an llseek
system call instead of _llseek. This is handled for the host
architecture by defining __NR__llseek as __NR_llseek, but not for the
target architecture.
Handle it in the same way for these architectures, defining
TARGET_NR__llseek as TARGET_NR_llseek.
James Hogan [Tue, 25 Mar 2014 23:21:50 +0000 (23:21 +0000)]
linux-user: Assert stack used for auxvec, envp, argv
Assert that the amount of stack space used for auxvec, envp & argv
exactly matches the amount allocated. This catches if DLINFO_ITEMS isn't
updated when another NEW_AUX_ENT is added.
Peter Maydell [Fri, 2 May 2014 10:32:00 +0000 (11:32 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20140501' into staging
target-arm queue:
* implement XScale cache lockdown cp15 ops
* fix v7M CPUID base register
* implement WFE and YIELD as yields for A64
* fix A64 "BLR LR"
* support Cortex-A57 in virt machine model
* a few other minor AArch64 bugfixes
# gpg: Signature made Thu 01 May 2014 15:42:17 BST using RSA key ID 14360CDE
# gpg: Good signature from "Peter Maydell <[email protected]>"
* remotes/pmaydell/tags/pull-target-arm-20140501:
hw/arm/virt: Add support for Cortex-A57
hw/arm/virt: Put GIC register banks on 64K boundaries
hw/arm/virt: Create the GIC ourselves rather than (ab)using a15mpcore_priv
target-arm: Correct a comment refering to EL0
target-arm: A64: Fix a typo when declaring TLBI ops
target-arm: A64: Handle blr lr
target-arm: Make vbar_write 64bit friendly on 32bit hosts
target-arm: implement WFE/YIELD as a yield for AArch64
armv7m_nvic: fix CPUID Base Register
target-arm: Implement XScale cache lockdown operations as NOPs
Peter Maydell [Fri, 2 May 2014 09:50:58 +0000 (10:50 +0100)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block patches
# gpg: Signature made Wed 30 Apr 2014 19:19:32 BST using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
* remotes/kevin/tags/for-upstream: (31 commits)
curl: Fix hang reading from slow connections
curl: Ensure all informationals are checked for completion
curl: Eliminate unnecessary use of curl_multi_socket_all
curl: Remove unnecessary explicit calls to internal event handler
curl: Remove erroneous sleep waiting for curl completion
curl: Fix return from curl_read_cb with invalid state
curl: Remove unnecessary use of goto
curl: Fix long line
block/vdi: Error out immediately in vdi_create()
block/bochs: Fix error handling for seek_to_sector()
qcow2: Check min_size in qcow2_grow_l1_table()
qcow2: Catch bdrv_getlength() error
block: Use correct width in format strings
qcow2: Avoid overflow in alloc_clusters_noref()
block: Use error_abort in bdrv_image_info_specific_dump()
block: Fix open_flags in bdrv_reopen()
Revert "block: another bdrv_append fix"
block: Unlink temporary files in raw-posix/win32
block: Remove BDRV_O_COPY_ON_READ for bs->file
block: Create bdrv_backing_flags()
...
Peter Maydell [Thu, 24 Apr 2014 17:54:42 +0000 (18:54 +0100)]
hw/arm/virt: Put GIC register banks on 64K boundaries
For an AArch64 CPU which supports 64K pages, having the GIC
register banks at 4K offsets is potentially awkward. Move
them out to being at 64K offsets. (This is harmless for
AArch32 CPUs and for AArch64 CPUs with 4K pages, so it is simpler
to use the same offsets everywhere than to try to use 64K offsets
only for AArch64 host CPUs.)
Peter Maydell [Thu, 1 May 2014 14:24:46 +0000 (15:24 +0100)]
hw/arm/virt: Create the GIC ourselves rather than (ab)using a15mpcore_priv
Rather than having the virt machine model create an a15mpcore_priv
device regardless of the actual CPU type in order to instantiate the GIC,
move to having the machine model create the GIC directly. This
corresponds to a system which uses a standalone GIC (eg the GIC-400)
rather than the one built in to the CPU core.
The primary motivation for this is to support the Cortex-A57,
which for a KVM configuration will use a GICv2, which is not
built into the CPU.
Peter Maydell [Thu, 1 May 2014 14:24:44 +0000 (15:24 +0100)]
target-arm: Implement XScale cache lockdown operations as NOPs
XScale defines some implementation-specific coprocessor registers
for doing cache lockdown operations. Since QEMU doesn't model a
cache no proper implementation is possible, but NOP out the
registers so that guest code like u-boot that tries to use them
doesn't crash.
Peter Maydell [Thu, 1 May 2014 13:17:33 +0000 (14:17 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-6' into staging
gtk: collection of fixes and cleanups by Cole Robinson
# gpg: Signature made Tue 29 Apr 2014 10:44:44 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg: aka "Gerd Hoffmann <[email protected]>"
# gpg: aka "Gerd Hoffmann (private) <[email protected]>"
* remotes/kraxel/tags/pull-gtk-6:
gtk: Fix accelerators being triggered twice with gtk3
gtk: Fix -serial vc
gtk: Use ctrl+alt+q for quit accelerator
gtk: Remove use of deprecated stock items
gtk: Don't use deprecated vte_terminal_get_adjustment
gtk: Don't use deprecated gtk_image_menu_item_new_with_mnemonic
configure: Re-run make if gtkabi/sdlabi is changed
configure: Document --with-gtkabi
Matthew Booth [Tue, 29 Apr 2014 15:03:32 +0000 (16:03 +0100)]
curl: Fix hang reading from slow connections
When receiving a new aio read request, we first look for an existing
transaction whose range will cover the read request by the time it
completes. However, we weren't checking that the existing transaction
was still active. If it had timed out, we were adding the request to a
transaction which would never complete and had already been cancelled,
resulting in a hang.
Matthew Booth [Tue, 29 Apr 2014 15:03:31 +0000 (16:03 +0100)]
curl: Ensure all informationals are checked for completion
According to the documentation, the correct way to ensure all
informationals have been returned by curl_multi_info_read is to loop
until it returns NULL.
Matthew Booth [Tue, 29 Apr 2014 15:03:30 +0000 (16:03 +0100)]
curl: Eliminate unnecessary use of curl_multi_socket_all
curl_multi_socket_all is a deprecated catch-all which checks for
activities on all open curl sockets. We have enough information from
the event loop to check only the sockets with activity. This change
removes use of curl_multi_socket_all in favour of
curl_multi_socket_action called with the relevant handle.
At the same time, it also ensures that the driver only checks for
completion of read operations after reading from a socket, rather than
both reading and writing.
Matthew Booth [Tue, 29 Apr 2014 15:03:28 +0000 (16:03 +0100)]
curl: Remove erroneous sleep waiting for curl completion
The driver will not start more than a fixed number of curl sessions.
If it needs more, it must wait for the completion of an existing one.
The driver was sleeping, which will prevent the main loop from
running, and therefore the event it's waiting on. It was also directly
calling its internal handler rather than waiting on existing
registered handlers to be called from the main loop.
This change causes it simply to wait for a period of time whilst
allowing the main loop to execute.
Matthew Booth [Tue, 29 Apr 2014 15:03:27 +0000 (16:03 +0100)]
curl: Fix return from curl_read_cb with invalid state
A curl write callback is supposed to return the number of bytes it
handled. curl_read_cb would have erroneously reported it had handled
all bytes in the event that the internal curl state was invalid.
Max Reitz [Tue, 29 Apr 2014 17:03:16 +0000 (19:03 +0200)]
block/vdi: Error out immediately in vdi_create()
Currently, if an error occurs during the part of vdi_create() which
actually writes the image, the function stores -errno, but continues
anyway.
Instead of trying to write data which (if it can be written at all) does
not make any sense without the operations before succeeding (e.g.,
writing the image header), just error out immediately.
Max Reitz [Tue, 29 Apr 2014 17:03:15 +0000 (19:03 +0200)]
block/bochs: Fix error handling for seek_to_sector()
Currently, seek_to_sector() returns -1 both for errors and unallocated
sectors, resulting in silent errors. As 0 is an invalid offset of data
clusters (bitmap_offset is greater than 0 because s->data_offset is
greater than 0), just return 0 for unallocated sectors and -errno in
case of error. This should then be propagated by bochs_read(), the sole
user of seek_to_sector().
That function also has a case of "return -1 in case of error", which is
fixed by this patch as well.
bochs_read() is called by bochs_co_read() which passes the return value
through, therefore it is indeed correct for bochs_read() to return
-errno.
Max Reitz [Tue, 29 Apr 2014 17:03:14 +0000 (19:03 +0200)]
qcow2: Check min_size in qcow2_grow_l1_table()
First, new_l1_size is an int64_t, whereas min_size is a uint64_t.
Therefore, during the loop which adjusts new_l1_size until it equals or
exceeds min_size, new_l1_size might overflow and become negative. The
comparison in the loop condition however will take it as an unsigned
value (because min_size is unsigned) and therefore recognize it as
exceeding min_size. Therefore, the loop is left with a negative
new_l1_size, which is not correct. This could be fixed by making
new_l1_size uint64_t.
On the other hand, however, by doing this, the while loop may take
forever. If min_size is e.g. UINT64_MAX, it will take new_l1_size
probably multiple overflows to reach the exact same value (if it reaches
it at all). Then, right after the loop, new_l1_size will be recognized
as being too big anyway.
Both problems require a ridiculously high min_size value, which is very
unlikely to occur; but both problems are also simply avoided by checking
whether min_size is sane before calculating new_l1_size (which should
still be checked separately, though).
Max Reitz [Tue, 29 Apr 2014 17:03:12 +0000 (19:03 +0200)]
block: Use correct width in format strings
Instead of blindly relying on a normal integer having a width of 32 bits
(which is a pretty good assumption, but we should not rely on it if
there is no need), use the correct format string macros.
Max Reitz [Tue, 29 Apr 2014 17:03:11 +0000 (19:03 +0200)]
qcow2: Avoid overflow in alloc_clusters_noref()
alloc_clusters_noref() stores the cluster index in a uint64_t. However,
offsets are often represented as int64_t (as for example the return
value of alloc_clusters_noref() itself demonstrates). Therefore, we
should make sure all offsets in the allocated range of clusters are
representable using int64_t without overflows.
Max Reitz [Tue, 29 Apr 2014 16:32:25 +0000 (18:32 +0200)]
block: Use error_abort in bdrv_image_info_specific_dump()
Currently, bdrv_image_info_specific_dump() uses an error variable for
visit_type_ImageInfoSpecific, but ignores the result. As this function
is used here with an output visitor to transform the ImageInfoSpecific
object to a generic QDict, an error should actually be impossible. It is
however better to assert that this is indeed the case. This is done by
this patch using error_abort instead of an unused local Error variable.
Kevin Wolf [Fri, 25 Apr 2014 17:04:55 +0000 (19:04 +0200)]
block: Fix open_flags in bdrv_reopen()
Use the same function as bdrv_open() for determining what the right
flags for bs->file are. Without doing this, a reopen means that
bs->file loses BDRV_O_CACHE_WB or BDRV_O_UNMAP if bs doesn't have it as
well.
When creating live snapshots, the new image file is opened with
BDRV_O_NO_BACKING because the whole backing chain is already opened.
It is then appended to the chain using bdrv_append(). The result of
this was that the image had a backing file, but BDRV_O_NO_BACKING
was still set. This is obviously inconsistent.
There used to be some places in qemu that closed and image and then
opened it again, with its old flags (a bdrv_open()/close() sequence
involves reopening the whole backing file chain, too). In this case
the BDRV_O_NO_BACKING flag meant that the backing chain wasn't
reopened and only the top layer was left.
(Most, but not all of these places are replaced by bdrv_reopen()
today, which doesn't touch the backing files at all.)
Other places that looked at bs->open_flags weren't interested in
BDRV_O_NO_BACKING, so no breakage there.
What it actually did:
The commit moved the BDRV_O_NO_BACKING away to the backing file.
Because the bdrv_open()/close() sequences only looked at the flags
of the top level BlockDriverState and used it for the whole chain,
the flag didn't hurt there any more. Obviously, it is still
inconsistent because the backing file may have another backing file,
but without practical impact.
At the same time, it swapped all other flags. This is practically
irrelevant as long as live snapshots only allow opening the new
layer with the same flags as the old top layer. It still doesn't
make any sense, and it is a time bomb that explodes as soon as the
flags can differ.
bdrv_append_temp_snapshot() is such a case: It adds the new flag
BDRV_O_TEMPORARY for the temporary snapshot. The swapping of commit 3a389e79 results in the following nonsensical configuration:
bs->open_flags: BDRV_O_TEMPORARY cleared
bs->file->open_flags: BDRV_O_TEMPORARY set
bs->backing_hd->open_flags: BDRV_O_TEMPORARY set
bs->backing_hd->file->open_flags: BDRV_O_TEMPORARY cleared
We're still lucky because the format layer ignores the flag and the
protocol layer happens to get the right value, but sooner or later
this is bound to go wrong...
What the right fix would have been:
Simply clear the BDRV_O_NO_BACKING flag when the BlockDriverState is
appended to an existing backing file chain, because now it does have
a backing file.
Commit 4ddc07ca already implemented this silently in bdrv_append(),
so we don't have to come up with a new fix.
Kevin Wolf [Fri, 11 Apr 2014 17:16:36 +0000 (19:16 +0200)]
block: Unlink temporary files in raw-posix/win32
Instead of having unlink() calls in the generic block layer, where we
aren't even guarateed to have a file name, move them to those block
drivers that are actually used and that always have a filename. Gets us
rid of some #ifdefs as well.
The patch also converts bs->is_temporary to a new BDRV_O_TEMPORARY open
flag so that it is inherited in the protocol layer and the raw-posix and
raw-win32 drivers can unlink the file.
Kevin Wolf [Fri, 11 Apr 2014 19:36:45 +0000 (21:36 +0200)]
block: Remove BDRV_O_COPY_ON_READ for bs->file
Copy on Read makes sense on the format level where backing files are
implemented, but it's not required on the protocol level. While it
shouldn't actively break anything to have COR enabled on both layers,
needless serialisation and allocation checks may impact performance.
Kevin Wolf [Fri, 25 Apr 2014 11:27:34 +0000 (13:27 +0200)]
block: Create bdrv_backing_flags()
Instead of manipulation flags inline, move the derivation of the flags
of a backing file into a new function next to the existing functions
that derive flags for bs->file and for the block driver open function.
Kevin Wolf [Fri, 11 Apr 2014 19:29:52 +0000 (21:29 +0200)]
block: Create bdrv_inherited_flags()
Instead of having bdrv_open_flags() as a function that creates flags for
several unrelated places and then adding open-coded flags on top, create
a new function that derives the flags for bs->file from the flags for bs.
projects / qemu.git / log