Commit af3cbfbe8018ccc16fb3a0048e928f66f0d05e87 hoisted some "common"
loads of the temporary type, forgetting that the types could differ
during truncating moves. This affects the correctness of the memory
offset on big-endian hosts.
exec-all.h: Use stl_le_p instead of stl_p patching x86
We got the wrong version of stl_p, the one that bswaps as appropriate
for the target. Since x86 is always little-endian, the "_le_" routine
will resolve to what we want.
Peter Maydell [Fri, 9 May 2014 15:06:41 +0000 (16:06 +0100)]
bsd-user: Remove reference to CONFIG_UNAME_RELEASE
Commit e586822a5 broke the bsd-user build when it removed the
CONFIG_UNAME_RELEASE define but forgot to remove the use of it
in bsd-user. Fix this in the simplest possible way (bsd-user
doesn't make any use at all of the qemu_uname_release variable
except to allow it to be pointlessly set by the user, so this
is all we need to do.)
Peter Maydell [Tue, 13 May 2014 10:30:07 +0000 (11:30 +0100)]
Merge remote-tracking branch 'remotes/bonzini/configure' into staging
* remotes/bonzini/configure:
libcacard: remove libcacard-specific CFLAGS and LIBS from global vars
build: simplify and fix fix-obj-vars
build: convert some obj-specific CFLAGS to use new foo.o-cflags syntax
build: add support for per-object -cflags and -libs to all rules
Makefile: use $(INSTALL_LIB) for modules not $(INSTALL_PROG)
Makefile.target: use $(INSTALL_PROG) for installing, not $(INSTALL)
Makefile: strip tools and modules too
build: simplify Makefile.target around unnest-vars invocations
build: simplify Makefile.target a bit, use just one rule for softmmu
build: Fix per-object variables for Makefile.target
Peter Maydell [Tue, 13 May 2014 09:35:46 +0000 (10:35 +0100)]
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
Block pull request
# gpg: Signature made Fri 09 May 2014 19:57:53 BST using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <[email protected]>"
# gpg: aka "Stefan Hajnoczi <[email protected]>"
* remotes/stefanha/tags/block-pull-request:
glib: fix g_poll early timeout on windows
block: qemu-iotests - test for live migration
block: qemu-iotests - update 085 to use common.qemu
block: qemu-iotests - add common.qemu, for bash-controlled qemu tests
block/raw-posix: Try both FIEMAP and SEEK_HOLE
gluster: Correctly propagate errors when volume isn't accessible
vl.c: remove init_clocks call from main
block: Fix open flags with BDRV_O_SNAPSHOT
qemu-iotests: Test converting to streamOptimized from small cluster size
vmdk: Implement .bdrv_get_info()
vmdk: Implement .bdrv_write_compressed
qemu-img: Convert by cluster size if target is compressed
block/iscsi: bump year in copyright notice
block/nfs: Check for NULL server part
qemu-img: sort block formats in help message
iotests: Use configured python
qcow2: Fix alloc_clusters_noref() overflow detection
There was some modulo logic to ensure that Microblaze always booted into
physical RAM regardless of the elf entry. Removed it, as QEMU should fail
gracefully when given a bad elf, rather than attempt to run it.
The TCSR register has only 11 valid bits. This is now used by the
linux kernel to auto-detect endianness, and causes Linux 3.15-rc1
and later to hang when run under qemu-microblaze. Mask valid bits
before writing the register to solve the problem.
The MER register only has two valid bits. This is now used by
the linux kernel to auto-detect endianness, and causes Linux 3.15-rc1
and later to hang when run under qemu-microblaze. Mask valid bits before
writing the register to solve the problem.
tcg: Remove unreachable code in tcg_out_op and op_defs
The INDEX_op_call case has just been obsoleted; the mov and movi
cases have not been reachable for years. Attempt to document this
both in each tcg_out_op switch, and via TCG_OPF_NOT_PRESENT.
Because of the TCG_OPF_NOT_PRESENT change, this must be done for
all targets in a single commit.
tcg: Use tcg_target_available_regs in tcg_reg_alloc_mov
The move opcodes are special in that their constraints must cover
all available registers. So instead of checking the constraints,
just use the available registers.
Using a 16-byte aligned structure achieves best results, both for code
cleanliness and compiled code size. However, this means that we can't
use the trick of encoding the slot number into the low 2 bits.
Thankfully, we only ever use slot2, so make that explicit in the names
of the relocation functions, and drop the code for other slots.
To be defined by the tcg backend based on the elemental unit of the ISA.
During the transition, allow TCG_TARGET_INSN_UNIT_SIZE to be undefined,
which allows us to default tcg_insn_unit to the current uint8_t.
Peter Maydell [Fri, 28 Mar 2014 15:29:48 +0000 (15:29 +0000)]
tcg: Avoid undefined behaviour patching code at unaligned addresses
To avoid C undefined behaviour when patching generated code,
provide wrappers tcg_patch8/16/32/64 which use the usual memcpy
trick, and use them in the i386 backend.
Peter Maydell [Fri, 28 Mar 2014 15:29:47 +0000 (15:29 +0000)]
tcg: Avoid stores to unaligned addresses
Avoid stores to unaligned addresses in TCG code generation, by using the
usual memcpy() approach. (Using bswap.h would drag a lot of QEMU baggage
into TCG, so it's simpler just to do direct memcpy() here.)
Peter Maydell [Fri, 28 Mar 2014 15:29:46 +0000 (15:29 +0000)]
exec-all.h: Use stl_p to avoid undefined behaviour patching x86 jumps
The code which patches x86 jump instructions assumes it can do an
unaligned write of a uint32_t. This is actually safe on x86, but it's
still undefined behaviour. We have infrastructure for doing efficient
unaligned accesses which doesn't engage in undefined behaviour, so
use it.
This is technically fractionally less efficient, at least with gcc 4.6;
instead of one instruction:
7b2: 89 3e mov %edi,(%rsi)
we get an extra spurious store to the stack slot:
7b2: 89 7c 24 64 mov %edi,0x64(%rsp)
7b6: 89 3e mov %edi,(%rsi)
Michael Tokarev [Thu, 8 May 2014 12:48:27 +0000 (16:48 +0400)]
libcacard: remove libcacard-specific CFLAGS and LIBS from global vars
Currently all what's needed for single file libcacard/vcard_emul_nss.c
(libnss cflags) and hw/usb/ccid-card-emulated.c (libcacard includes)
together with the libs is added to global QEMU_CFLAGS and libs_softmmu.
Use the cflags only where really used (for two mentioned files), and
libs only where needed.
While at it, rename variables to better reflect reality: libcacard_*
is really nss_*.
This needs a bit more tweaking: $(NSS_LIBS) should not contain $glib_libs
(ditto for _cflags). But in order to fix it, some more preparations
should be made first. So add a FIXME comment.
Paolo Bonzini [Thu, 8 May 2014 13:02:48 +0000 (15:02 +0200)]
build: simplify and fix fix-obj-vars
fix-obj-vars has the undesired side effect of breaking -cflags
-objs and -libs variables in the toplevel Makefile.objs. The
variables in the toplevel Makefile.objs do not need any fix,
so fix-obj-vars need not do anything.
Since we are touching it, remove the now unnecessary $(if)
in the callers.
Sangho Park [Thu, 8 May 2014 08:47:10 +0000 (12:47 +0400)]
glib: fix g_poll early timeout on windows
g_poll has a problem on Windows when using
timeouts < 10ms, in glib/gpoll.c:
/* If not, and we have a significant timeout, poll again with
* timeout then. Note that this will return indication for only
* one event, or only for messages. We ignore timeouts less than
* ten milliseconds as they are mostly pointless on Windows, the
* MsgWaitForMultipleObjectsEx() call will timeout right away
* anyway.
*/
if (retval == 0 && (timeout == INFINITE || timeout >= 10))
retval = poll_rest (poll_msgs, handles, nhandles, fds, nfds, timeout);
so whenever g_poll is called with timeout < 10ms it does
a quick poll instead of wait, this causes significant performance
degradation of QEMU, thus we should use WaitForMultipleObjectsEx
directly
Jeff Cody [Wed, 30 Apr 2014 14:55:09 +0000 (10:55 -0400)]
block: qemu-iotests - update 085 to use common.qemu
The new functionality of common.qemu implements the QEMU control
and communication functionality that was originally in test 085.
This removes that now-duplicate functionality, and uses the
common.qemu functions.
The QEMU commandline changes slightly due to this; in addition to
monitor and qmp i/o options, the new QEMU commandline from inside
common.qemu now introduces -machine accel=qtest.
Jeff Cody [Wed, 30 Apr 2014 14:55:08 +0000 (10:55 -0400)]
block: qemu-iotests - add common.qemu, for bash-controlled qemu tests
This creates some common functions for bash language qemu-iotests
to control, and communicate with, a running QEMU process.
4 functions are introduced:
1. _launch_qemu()
This launches the QEMU process(es), and sets up the file
descriptors and fifos for communication. You can choose to
launch each QEMU process listening for either QMP or HMP
monitor. You can call this function multiple times, and
save the handle returned from each. The returned handle is
in $QEMU_HANDLE. You must copy this value.
Commands 2 and 3 use the handle received from _launch_qemu(), to talk
to the appropriate process.
2. _send_qemu_cmd()
Sends a command string, specified by $2, to QEMU. If $3 is
non-NULL, _send_qemu_cmd() will wait to receive $3 as a
required result string from QEMU. Failure to receive $3 will
cause the test to fail. The command can optionally be retried
$qemu_cmd_repeat number of times. Set $qemu_error_no_exit
to not force the test the fail on exit; in this case,
$QEMU_STATUS[$1] will be set to -1 on failure.
3. _timed_wait_for()
Waits for a response, for up to a default of 10 seconds. If
$2 is not seen in that time (anywhere in the response), then
the test fails. Primarily used by _send_qemu_cmd, but could
be useful standalone, as well. To prevent automatic exit
(and therefore test failure), set $qemu_error_no_exit to a
non-NULL value. If $silent is a non-NULL value, then output
to stdout will be suppressed.
4. _cleanup_qemu()
Kills the running QEMU processes, and removes the fifos.
Max Reitz [Thu, 8 May 2014 18:57:55 +0000 (20:57 +0200)]
block/raw-posix: Try both FIEMAP and SEEK_HOLE
The current version of raw-posix always uses ioctl(FS_IOC_FIEMAP) if
FIEMAP is available; lseek with SEEK_HOLE/SEEK_DATA are not even
compiled in in this case. However, there may be implementations which
support the latter but not the former (e.g., NFSv4.2) as well as vice
versa.
To cover both cases, try FIEMAP first (as this will return -ENOTSUP if
not supported instead of returning a failsafe value (everything
allocated as a single extent)) and if that does not work, fall back to
SEEK_HOLE/SEEK_DATA.
Peter Krempa [Fri, 9 May 2014 10:08:10 +0000 (12:08 +0200)]
gluster: Correctly propagate errors when volume isn't accessible
The docs for glfs_init suggest that the function sets errno on every
failure. In fact it doesn't. As other functions such as
qemu_gluster_open() in the gluster block code report their errors based
on this fact we need to make sure that errno is set on each failure.
This fixes a crash of qemu-img/qemu when a gluster brick isn't
accessible from given host while the server serving the volume
description is.
Thread 1 (Thread 0x7ffff7fba740 (LWP 203880)):
#0 0x00007ffff77673f8 in glfs_lseek () from /usr/lib64/libgfapi.so.0
#1 0x0000555555574a68 in qemu_gluster_getlength ()
#2 0x0000555555565742 in refresh_total_sectors ()
#3 0x000055555556914f in bdrv_open_common ()
#4 0x000055555556e8e8 in bdrv_open ()
#5 0x000055555556f02f in bdrv_open_image ()
#6 0x000055555556e5f6 in bdrv_open ()
#7 0x00005555555c5775 in bdrv_new_open ()
#8 0x00005555555c5b91 in img_info ()
#9 0x00007ffff62c9c05 in __libc_start_main () from /lib64/libc.so.6
#10 0x00005555555648ad in _start ()
Kirill Batuzov [Tue, 6 May 2014 12:59:53 +0000 (16:59 +0400)]
vl.c: remove init_clocks call from main
Clocks are initialized in qemu_init_main_loop. They are not needed before it.
Initializing them twice is not only unnecessary but is harmful: it results in
memory leak and potentially can lead to a situation where different parts of
QEMU use different sets of timers.
To avoid it remove init_clocks call from main and add an assertion to
qemu_clock_init that corresponding clock has not been initialized yet.
Kevin Wolf [Tue, 6 May 2014 10:11:42 +0000 (12:11 +0200)]
block: Fix open flags with BDRV_O_SNAPSHOT
The immediately visible effect of this patch is that it fixes committing
a temporary snapshot to its backing file. Previously, it would fail with
a "permission denied" error because bdrv_inherited_flags() forced the
backing file to be read-only, ignoring the r/w reopen of bdrv_commit().
The bigger problem this revealed is that the original open flags must
actually only be applied to the temporary snapshot, and the original
image file must be treated as a backing file of the temporary snapshot
and get the right flags for that.
Peter Maydell [Fri, 9 May 2014 14:46:34 +0000 (15:46 +0100)]
Merge remote-tracking branch 'remotes/qmp-unstable/queue/qmp' into staging
* remotes/qmp-unstable/queue/qmp: (38 commits)
Revert "qapi: Clean up superfluous null check in qapi_dealloc_type_str()"
qapi: Document optional arguments' backwards compatibility
qmp: use valid JSON in transaction example
qmp: Don't use error_is_set() to suppress additional errors
dump: Drop pointless error_is_set(), DumpState member errp
qemu-option: Clean up fragile use of error_is_set()
qga: Drop superfluous error_is_set()
qga: Clean up fragile use of error_is_set()
qapi: Clean up fragile use of error_is_set()
tests/qapi-schema: Drop superfluous error_is_set()
qapi: Drop redundant, unclean error_is_set()
hmp: Guard against misuse of hmp_handle_error()
qga: Use return values instead of error_is_set(errp)
error: Consistently name Error ** objects errp, and not err
qmp: Consistently name Error ** objects errp, and not err
qga: Consistently name Error ** objects errp, and not err
qmp hmp: Consistently name Error * objects err, and not errp
pci-assign: assigned_initfn(): set monitor error in common error handler
pci-assign: propagate errors from assign_intx()
pci-assign: propagate errors from assign_device()
...
qmp: Don't use error_is_set() to suppress additional errors
Using error_is_set(errp) that way can sweep programming errors under
the carpet when we get called incorrectly with an error set.
encrypted_bdrv_it() does it, because there's no way to make
bdrv_iterate() break its loop. Actually safe, because qmp_cont()
clears the error before the loop. Clean it up anyway: replace
bdrv_iterate() by bdrv_next(), break the loop on error.
dump: Drop pointless error_is_set(), DumpState member errp
In qmp_dump_guest_memory(), the error must be clear on entry, and we
always bail out after setting it, directly or via dump_init().
Therefore, both error_is_set() are always false. Drop them.
DumpState member errp is now write-only. Drop it, too.
qemu-option: Clean up fragile use of error_is_set()
Using error_is_set(ERRP) to find out whether to bail out due to
previous error 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).
The error_is_set(state->errp) in qemu_opts_from_qdict_1() is merely
fragile, because the callers never pass state argument with null
state->errp.
Make the code more robust and more obviously correct: test
*state->errp directly.
acquire_privilege(), execute_async() and check_suspend_mode() do
nothing when called with an error set. Callers shouldn't do that, and
no caller does. Drop the superfluous tests.
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).
The error_is_set(errp) in the guest agent command handler functions
are merely fragile, because all chall chains (do_qmp_dispatch() via
the generated marshalling functions) pass a non-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.
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).
The error_is_set(errp) in do_qmp_dispatch() is merely fragile, because
the caller never 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.
do_qmp_dispatch()'s test for qmp_dispatch_check_obj() failure examines
both the return value and the error object. The latter part is
unclean; it works only when do_qmp_dispatch()'s caller passes a
non-null errp argument. That's the case, but it's not locally
obvious. Unclean.
Cleanup would be easy enough, but since the unclean code is also
redundant, let's just drop it.
qga: Use return values instead of error_is_set(errp)
Using error_is_set(errp) to check whether a function call failed is
fragile: it breaks when errp is null. ga_get_fd_handle() and
guest_file_handle_add() don't return a useful value when they fail,
but that's just stupid. Fix that, and check them instead. As far
as I can tell, errp can't be null there, but this is more robust and
more obviously correct.
Fam Zheng [Tue, 6 May 2014 13:08:45 +0000 (21:08 +0800)]
vmdk: Implement .bdrv_get_info()
This will return cluster_size and needs_compressed_writes to caller, if all the
extents have the same value (or there's only one extent). Otherwise return
-ENOTSUP.
Fam Zheng [Tue, 6 May 2014 13:08:44 +0000 (21:08 +0800)]
vmdk: Implement .bdrv_write_compressed
Add a wrapper function to support "compressed" path in qemu-img convert.
Only support streamOptimized subformat case for now (num_extents == 1
and extent compression is true).
Mike Day [Mon, 5 May 2014 16:53:34 +0000 (12:53 -0400)]
qemu-img: sort block formats in help message
The help message for qemu-img lists the supported block formats, of
which there are 27 as of version 2.0.50. The formats are printed in
the order of their driver's position in a linked list, which appears
random. This patch prints the formats in sorted order, making it
easier to read and to find a specific format in the list.
[Added suggestions from Fam Zheng <[email protected]> to declare variables
at the top of the scope in help() and to omit explicit cast for void*
opaque.
--Stefan]
Max Reitz [Sat, 3 May 2014 14:47:08 +0000 (16:47 +0200)]
iotests: Use configured python
Currently, QEMU's iotests rely on /usr/bin/env to start the correct
Python (that is, at least Python 2.4, but not 3). On systems where
Python 3 is the default, the user has no clean way of making the iotests
use the correct binary.
This commit makes the iotests use the Python selected by configure.
If the very first allocation has a length of 0, the free_cluster_index
is still 0 after the for loop, which means that subtracting one from it
will underflow and signal an invalid range of clusters by returning
-EFBIG. However, there is no such range, as its length is 0.
Fix this by preventing underflows on free_cluster_index during the
check.
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:
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.
pci-assign: propagate errors from assigned_dev_register_msix_mmio()
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).
pci-assign: accept Error from pci_add_capability2()
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).
pci-assign: propagate Error from check_irqchip_in_kernel()
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.
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.
pci-assign: accept Error from monitor_handle_fd_param2()
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).
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)).
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
Amos Kong [Mon, 28 Apr 2014 05:53:49 +0000 (13:53 +0800)]
qapi: treat all negative return of strtosz_suffix() as error
strtosz_suffix() might return negative error, this patch fixes
the error handling.
This patch also changes to handle error in the if statement
rather than handle success specially, this will make this use
of strtosz_suffix consistent with all other uses.
projects / qemu.git / log