Max Reitz [Thu, 5 Feb 2015 18:58:18 +0000 (13:58 -0500)]
qemu-img: Use BlockBackend as far as possible
Although qemu-img already creates BlockBackends, it does not do accesses
to the images through them. This patch converts all of the bdrv_* calls
for which this is currently possible to blk_* calls. Most of the
remaining calls will probably stay bdrv_* calls because they really do
operate on the BDS level instead of the BB level.
Max Reitz [Thu, 5 Feb 2015 18:58:15 +0000 (13:58 -0500)]
block/xen: Use blk_new_open() in blk_connect()
As part of the required changes, this fixes a bug where specifying an
invalid driver would result in the block layer probing the image format;
now it will result in an error, unless "<unset>" is specified as the
driver name. Fixing this would require further work on the xen_disk code
which does not seem worth it (at this point and for this patch).
Max Reitz [Thu, 5 Feb 2015 18:58:13 +0000 (13:58 -0500)]
iotests: Add test for driver=qcow2, format=qcow2
While specifying a different driver and format is obviously invalid,
specifying the same driver once through driver and once through format
is invalid as well. Add a test for it.
Max Reitz [Thu, 5 Feb 2015 18:58:12 +0000 (13:58 -0500)]
block: Add Error parameter to bdrv_find_protocol()
The argument given to bdrv_find_protocol() is just a file name, which
makes it difficult for the caller to reconstruct what protocol
bdrv_find_protocol() was hoping to find. This patch adds an Error
parameter to that function to solve this issue.
Max Reitz [Thu, 5 Feb 2015 18:58:11 +0000 (13:58 -0500)]
block: Add blk_new_open()
blk_new_with_bs() creates a BlockBackend with an empty BlockDriverState
attached to it. Empty BDSs are not nice, therefore add an alternative
function which combines blk_new_with_bs() with bdrv_open().
Note: In contrast to bdrv_open() which takes a BlockDriver parameter,
blk_new_open() does not take such a parameter. This is because
bdrv_open() opens a BlockDriverState, therefore it is natural to be able
to set the BlockDriver for that BDS. The fact that bdrv_open() can open
more than a single BDS is merely some form of a byproduct.
blk_new_open() on the other hand is intended to be used to create a
whole tree of BlockDriverStates. Therefore, setting a single BlockDriver
does not make much sense. Instead, the drivers to be used for each of
the nodes must be configured through the "options" QDict; including the
driver of the root BDS.
Max Reitz [Thu, 5 Feb 2015 18:58:10 +0000 (13:58 -0500)]
block: Lift some BDS functions to the BlockBackend
Create the blk_* counterparts for the following bdrv_* functions (which
make sense to call on the BlockBackend level):
- bdrv_co_write_zeroes()
- bdrv_write_compressed()
- bdrv_truncate()
- bdrv_nb_sectors()
- bdrv_discard()
- bdrv_load_vmstate()
- bdrv_save_vmstate()
Max Reitz [Wed, 11 Feb 2015 14:58:46 +0000 (09:58 -0500)]
qemu-img: Fix qemu-img convert -n
If -n is specified, it does not matter whether the output format and
protocol support image creation; building the creation options should
simply be skipped.
Fam Zheng [Fri, 30 Jan 2015 02:49:46 +0000 (10:49 +0800)]
qemu-iotests: Add 093 for IO throttling
This case utilizes qemu-io command "aio_{read,write} -q" to verify the
effectiveness of IO throttling options.
It's implemented by driving the vm timer from qtest protocol, so the
throttling timers are signaled with determinied time duration. Then we
verify the completed IO requests are within 10% error of bps and iops
limits.
"null" protocol is used as the disk backend so that no actual disk IO is
performed on host, this will make the blockstats much more
deterministic. Both "null-aio" and "null-co" are covered, which is also
a simple cross validation test for the driver code.
Fam Zheng [Mon, 19 Jan 2015 09:51:43 +0000 (17:51 +0800)]
qtest: Fix deadloop by running main loop AIO context's timers
qemu_clock_run_timers() only takes care of main_loop_tlg, we shouldn't
forget aio timer list groups.
Currently, the qemu_clock_deadline_ns_all (a few lines above) counts all
the timergroups of this clock type, including aio tlg, but we don't fire
them, so they are never cleared, which makes a dead loop.
For example, this function hangs when trying to drive throttled block
request queue with qtest clock_step.
Bin Wu [Tue, 10 Feb 2015 07:20:46 +0000 (15:20 +0800)]
nbd: fix the co_queue multi-adding bug
When we tested the VM migartion between different hosts with NBD
devices, we found if we sent a cancel command after the drive_mirror
was just started, a coroutine re-enter error would occur. The stack
was as follow:
(gdb) bt
00) 0x00007fdfc744d885 in raise () from /lib64/libc.so.6
01) 0x00007fdfc744ee61 in abort () from /lib64/libc.so.6
02) 0x00007fdfca467cc5 in qemu_coroutine_enter (co=0x7fdfcaedb400, opaque=0x0)
at qemu-coroutine.c:118
03) 0x00007fdfca467f6c in qemu_co_queue_run_restart (co=0x7fdfcaedb400) at
qemu-coroutine-lock.c:59
04) 0x00007fdfca467be5 in coroutine_swap (from=0x7fdfcaf3c4e8,
to=0x7fdfcaedb400) at qemu-coroutine.c:96
05) 0x00007fdfca467cea in qemu_coroutine_enter (co=0x7fdfcaedb400, opaque=0x0)
at qemu-coroutine.c:123
06) 0x00007fdfca467f6c in qemu_co_queue_run_restart (co=0x7fdfcaedbdc0) at
qemu-coroutine-lock.c:59
07) 0x00007fdfca467be5 in coroutine_swap (from=0x7fdfcaf3c4e8,
to=0x7fdfcaedbdc0) at qemu-coroutine.c:96
08) 0x00007fdfca467cea in qemu_coroutine_enter (co=0x7fdfcaedbdc0, opaque=0x0)
at qemu-coroutine.c:123
09) 0x00007fdfca4a1fa4 in nbd_recv_coroutines_enter_all (s=0x7fdfcaef7dd0) at
block/nbd-client.c:41
10) 0x00007fdfca4a1ff9 in nbd_teardown_connection (client=0x7fdfcaef7dd0) at
block/nbd-client.c:50
11) 0x00007fdfca4a20f0 in nbd_reply_ready (opaque=0x7fdfcaef7dd0) at
block/nbd-client.c:92
12) 0x00007fdfca45ed80 in aio_dispatch (ctx=0x7fdfcae15e90) at aio-posix.c:144
13) 0x00007fdfca45ef1b in aio_poll (ctx=0x7fdfcae15e90, blocking=false) at
aio-posix.c:222
14) 0x00007fdfca448c34 in aio_ctx_dispatch (source=0x7fdfcae15e90, callback=0x0,
user_data=0x0) at async.c:212
15) 0x00007fdfc8f2f69a in g_main_context_dispatch () from
/usr/lib64/libglib-2.0.so.0
16) 0x00007fdfca45c391 in glib_pollfds_poll () at main-loop.c:190
17) 0x00007fdfca45c489 in os_host_main_loop_wait (timeout=1483677098) at
main-loop.c:235
18) 0x00007fdfca45c57b in main_loop_wait (nonblocking=0) at main-loop.c:484
19) 0x00007fdfca25f403 in main_loop () at vl.c:2249
20) 0x00007fdfca266fc2 in main (argc=42, argv=0x7ffff517d638,
envp=0x7ffff517d790) at vl.c:4814
We find the nbd_recv_coroutines_enter_all function (triggered by a cancel
command or a network connection breaking down) will enter a coroutine which
is waiting for the sending lock. If the lock is still held by another coroutine,
the entering coroutine will be added into the co_queue again. Latter, when the
lock is released, a coroutine re-enter error will occur.
This bug can be fixed simply by delaying the setting of recv_coroutine as
suggested by paolo. After applying this patch, we have tested the cancel
operation in mirror phase looply for more than 5 hous and everything is fine.
Without this patch, a coroutine re-enter error will occur in 5 minutes.
John Snow [Thu, 5 Feb 2015 17:41:30 +0000 (12:41 -0500)]
qtest/ahci: Adding simple dma read-write test
Adds a test case for AHCI wherein we write a 4K
block of a changing pattern to sector 0, then
read back that 4K and compare the transmit and
receive buffers.
John Snow [Thu, 5 Feb 2015 17:41:29 +0000 (12:41 -0500)]
qtest/ahci: Assert sector size in identify test
A minor sanity check to assert that the sector size is 512.
The current block layer code deeply assumes that the IDE
sector size will be 512 bytes, so we carry forward that assumption
here.
This is useful for the DMA tests, which currently assume that
a sector will always be 512 bytes.
John Snow [Thu, 5 Feb 2015 17:41:28 +0000 (12:41 -0500)]
libqos/ahci: Add ahci_clean_mem
Clean up guest memory being used in ahci_clean_mem, to be
called during ahci_shutdown. With all guest memory leaks removed,
add an option to the allocator to throw an assertion if a leak
occurs.
This test adds some sanity to both the AHCI library and the
allocator.
John Snow [Thu, 5 Feb 2015 17:41:27 +0000 (12:41 -0500)]
libqos/ahci: add ahci_io
ahci_io is a wrapper around ahci_guest_io that takes a pointer to host
memory instead, and will create a guest memory buffer and copy the data
to/from as needed and as appropriate for a read/write command, such that
after a read, the guest data will be in a host buffer, and for a write,
the data will be transmitted to guest memory prior to the block operation.
Now that we have all the syntactic sugar functions in place for AHCI,
we can convert the identify test to be very, very short.
John Snow [Thu, 5 Feb 2015 17:41:26 +0000 (12:41 -0500)]
libqos/ahci: Add ahci_guest_io
ahci_guest_io is a shorthand function that will, in one shot,
execute a data command on the guest to the specified guest buffer
location, in the requested amount.
John Snow [Thu, 5 Feb 2015 17:41:23 +0000 (12:41 -0500)]
libqos/ahci: add ahci command functions
This patch adds the AHCICommand structure, and a set of functions to
operate on the structure.
ahci_command_create - Initialize and create a new AHCICommand in memory
ahci_command_free - Destroy this object.
ahci_command_set_buffer - Set where the guest memory DMA buffer is.
ahci_command_commit - Write this command to the AHCI HBA.
ahci_command_issue - Issue the committed command synchronously.
ahci_command_issue_async - Issue the committed command asynchronously.
ahci_command_wait - Wait for an asynchronous command to finish.
ahci_command_slot - Get the number of the command slot we committed to.
Helpers:
size_to_prdtl - Calculate the required minimum PRDTL size from
a buffer size.
ahci_command_find - Given an ATA command mnemonic, look it up in the
properties table to obtain info about the command.
command_header_init - Initialize the command header with sane values.
command_table_init - Initialize the command table with sane values.
[Peter Maydell <[email protected]> reported the following clang
warning:
tests/libqos/ahci.c:598:3: warning: redefinition
of typedef 'AHCICommand' is a C11 feature
[-Wtypedef-redefinition]
} AHCICommand;
I have replaced typedef struct ... AHCICommand; with struct ... ;
--Stefan]
John Snow [Thu, 5 Feb 2015 17:41:22 +0000 (12:41 -0500)]
libqos/ahci: Add ide cmd properties
Add a structure that defines some properties of various IDE commands.
These will be used to simplify the interface to the libqos AHCI calls,
lessening the redundancy of specifying and respecifying properties of
commands to various helper functions.
John Snow [Thu, 5 Feb 2015 17:41:21 +0000 (12:41 -0500)]
qtest/ahci: add ahci_write_fis
Similar to ahci_set_command_header, add a helper that takes an
in-memory representation of a command FIS and writes it to guest
memory, handling endianness as-needed.
John Snow [Thu, 5 Feb 2015 17:41:20 +0000 (12:41 -0500)]
qtest/ahci: Demagic ahci tests.
Add human-readable command names and other miscellaneous #defines
to help make the code more readable.
Some of these definitions are not yet used in this current series,
but for convenience and sanity they have been lumped together here,
as it's more trouble than it is worth in a test suite to hand-pick,
one-by-one, which preprocessor definitions are useful per-each test.
These definitions include:
ATA Command Mnemonics
Current expected AHCI sector size
FIS magic bytes
REG_H2D_FIS flags
Command Header flags
This patch adds a few helpers to help sanity-check the response of the
AHCI device after a command.
ahci_d2h_check_sanity inspects the D2H Register FIS,
ahci_pio_check_sanity inspects the PIO Setup FIS, and
ahci_cmd_check_sanity inspects the command header.
To support the PIO sanity check, a new structure is added for the
PIO Setup FIS type. Existing FIS types (H2D and D2H) have had their
members renamed slightly to condense reserved members into fewer
fields; and LBA fields are now represented by arrays of 8 byte chunks
instead of independent variables.
A helper that compares a given port's current interrupts and checks them
against a supplied list of expected interrupt bits, and throws an error
if they do not match.
The helper then resets the requested interrupts on this port, and asserts
that the interrupt register is now empty.
John Snow [Thu, 5 Feb 2015 17:41:15 +0000 (12:41 -0500)]
libqos/ahci: Add command header helpers
Adds command header helper functions:
-ahci_command_header_set
-ahci_command_header_get,
-ahci_command_destroy, and
-ahci_cmd_pick
These helpers help to quickly manage the command header information in
the AHCI device.
ahci_command_header_set and get will store or retrieve an AHCI command
header, respectively.
ahci_cmd_pick chooses the first available but least recently used
command slot to allow us to cycle through the available command slots.
ahci_command_destroy obliterates all information contained within a
given slot's command header, and frees its associated command table,
but not its DMA buffer!
Lastly, the command table pointer fields (dba and dbau) are merged into
a single 64bit value to make managing 64bit tests simpler.
John Snow [Thu, 5 Feb 2015 17:41:14 +0000 (12:41 -0500)]
qtest/ahci: rename 'Command' to 'CommandHeader'
The structure name is a bit of a misnomer; the structure currently named
command is actually the commandheader. A future patch in this series
will add an actual "Command" structure, so we'll rename it now before the
rest of the functions in this series try to use it.
In addition, rename the "b1" and "b2" fields
to be a unified uint16_t named "flags."
John Snow [Thu, 5 Feb 2015 17:41:13 +0000 (12:41 -0500)]
libqos/ahci: Add ahci_port_clear helper
Add a helper that assists in clearing out potentially old error and FIS
information from an AHCI port's data structures. This ensures we always
start with a blank slate for interrupt and FIS receipt information.
Cornelia Huck [Mon, 26 Jan 2015 16:26:42 +0000 (17:26 +0100)]
dataplane: endianness-aware accesses
The vring.c code currently assumes that guest and host endianness match,
which is not true for a number of cases:
- emulating targets with a different endianness than the host
- bi-endian targets, where the correct endianness depends on the virtio
device
- upcoming support for the virtio-1 standard mandates little-endian
accesses even for big-endian targets and hosts
Make sure to use accessors that depend on the virtio device.
Note that dataplane now needs to be built per-target.
John Snow [Mon, 19 Jan 2015 20:16:03 +0000 (15:16 -0500)]
libqos/ahci: create libqos/ahci.c
With global state removed, code responsible for booting up,
verifying, and initializing the AHCI HBA is extracted and
inserted into libqos/ahci.c, which would allow for other
qtests in the future to quickly grab a meaningfully initialized
reference to an AHCI HBA.
Even without other users, functionalizing and isolating the code
assists future AHCI tests that exercise Q35 migration.
For now, libqos/ahci.o will be PC-only, but can be expanded into
something arch-agnostic in the future, if needed.
John Snow [Mon, 19 Jan 2015 20:16:02 +0000 (15:16 -0500)]
qtest/ahci: Bookmark FB and CLB pointers
Instead of re-querying the AHCI device for the FB and CLB buffers, save
the pointer we gave to the device during initialization and reference
these values instead.
[Peter Maydell <[email protected]> reported the following clang
compiler warnings:
tests/libqos/ahci.c:256:40: warning: format specifies type 'unsigned
long' but the argument has type 'uint64_t'
(aka 'unsigned long long') [-Wformat]
g_test_message("CLB: 0x%08lx", ahci->port[i].clb);
tests/libqos/ahci.c:264:39: warning: format specifies type 'unsigned
long' but the argument has type 'uint64_t'
(aka 'unsigned long long') [-Wformat]
g_test_message("FB: 0x%08lx", ahci->port[i].fb);
The commit moved from uint32_t to uint64_t, so PRIx64 should be used for
the format specifier.
--Stefan]
John Snow [Mon, 19 Jan 2015 20:16:01 +0000 (15:16 -0500)]
qtest/ahci: remove getter/setter macros
These macros were a bad idea: They relied upon certain arguments being
present locally with a specific name.
With the endgoal being to factor out AHCI helper functions outside of
the test file itself, these have to be replaced by more explicit helper
setter/getter functions.
John Snow [Mon, 19 Jan 2015 20:16:00 +0000 (15:16 -0500)]
libqos/ahci: Functional register helpers
Introduce a set of "static inline" register helpers that are intended to
replace the current set of macros with more functional versions that are
better suited to inclusion in libqos than porcelain macros.
As a stopgap measure before eliminating the porcelain macros, define them
to use the new functions defined in the ahci.h header.
John Snow [Mon, 19 Jan 2015 20:15:56 +0000 (15:15 -0500)]
qtest/ahci: Store hba_base in AHCIQState
Store the HBA memory base address in the new state object, to simplify
function prototypes and encourage a more functional testing style.
This causes a lot of churn, but this patch is as "simplified" as I could
get it to be. This patch is therefore fairly mechanical and straightforward:
Any case where we pass "hba_base" has been consolidated into the AHCIQState
object and we pass the one unified parameter.
Any case where we reference "ahci" and "hba_state" have been modified to use
"ahci->dev" for the PCIDevice and "ahci->hba_state" to get at the base memory
address, accordingly.
Notes:
- A needless return is removed from start_ahci_device.
- For ease of reviewing, this patch can be reproduced (mostly) by:
# Replace (ahci, hba_base) prototypes with unified parameter
's/(QPCIDevice \*ahci, void \*\?\*hba_base/(AHCIQState *ahci/'
# Replace (ahci->dev, hba_base) calls with unified parameter
's/(ahci->dev, &\?hba_base)/(ahci)/'
# Replace calls to PCI config space using "ahci" with "ahci->dev"
's/qpci_config_\(read\|write\)\(.\)(ahci,/qpci_config_\1\2(ahci->dev,/'
After these, the remaining differences are easy to review by hand.
John Snow [Mon, 19 Jan 2015 20:15:55 +0000 (15:15 -0500)]
libqos: add pc specific interface
Create an operations structure so that the libqos interface can be
architecture agnostic, and create a pc-specific interface to functions
like qtest_boot.
Move the libqos object in the Makefile from being ahci-test only to
being linked with all tests that utilize the libqos features.
John Snow [Mon, 19 Jan 2015 20:15:54 +0000 (15:15 -0500)]
libqos: Update QGuestAllocator to be opaque
To avoid the architecture-specific implementations of the generic qtest
allocator having to know about fields within the allocator, add a
page_size setter method for users or arch specializations to use.
The allocator will assume a default page_size for general use, but it
can always be overridden.
Since this was the last instance of code directly using properties of the
QGuestAllocator object directly, modify the type to be opaque and move
the structure inside of malloc.c.
mlist_new, which was previously exported, is made static local to malloc.c,
as it has no external users.
[Peter Maydell <[email protected]> reported the following clang
warning:
tests/libqos/malloc.c:35:3: warning:
redefinition of typedef 'QGuestAllocator' is a C11 feature
[-Wtypedef-redefinition]
} QGuestAllocator;
I converted typedef struct ... QGuestAllocator; to struct ...;
--Stefan]
John Snow [Mon, 19 Jan 2015 20:15:51 +0000 (15:15 -0500)]
libqos: create libqos.c
The intent of this file is to serve as a misc. utilities file to be
shared amongst tests that are utilizing libqos facilities.
In a later patch, migration test helpers will be added to libqos.c that
will allow simplified testing of migration cases where libqos is
"Just Enough OS" for migrations testing.
The addition of the AHCIQState structure will also allow us to eliminate
global variables inside of qtests to manage allocators and test instances
in a better, more functional way.
libqos.c:
- Add qtest_boot
- Add qtest_shutdown
libqos.h:
- Create QOSState structure for allocator and QTestState.
ahci-test.c:
- Move qtest_boot and qtest_shutdown to libqos.c/h
- Create AHCIQState to interface with new qtest_boot/shutdown prototypes
- Modify tests slightly to use new types.
For now, the new object file is only linked to ahci-test, because it still
relies on pc architecture specific code in libqos. The next two patches will
reorganize the code to be more general.
Max Reitz [Fri, 6 Feb 2015 21:06:18 +0000 (16:06 -0500)]
iotests: Add test for drive-mirror with NBD target
When the drive-mirror block job is completed, it will call bdrv_swap()
on the source and the target BDS; this should obviously not result in a
segmentation fault.
Max Reitz [Fri, 6 Feb 2015 21:06:16 +0000 (16:06 -0500)]
nbd: Drop BDS backpointer
Before this patch, the "opaque" pointer in an NBD BDS points to a
BDRVNBDState, which contains an NbdClientSession object, which in turn
contains a pointer to the BDS. This pointer may become invalid due to
bdrv_swap(), so drop it, and instead pass the BDS directly to the
nbd-client.c functions which then retrieve the NbdClientSession object
from there.
Peter Maydell [Fri, 13 Feb 2015 11:44:49 +0000 (11:44 +0000)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20150212' into staging
Convert to linked list.
# gpg: Signature made Fri 13 Feb 2015 05:40:41 GMT using RSA key ID 4DD0279B
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: aka "Richard Henderson <[email protected]>"
# gpg: aka "Richard Henderson <[email protected]>"
* remotes/rth/tags/pull-tcg-20150212:
tcg: Remove unused opcodes
tcg: Implement insert_op_before
tcg: Remove opcodes instead of noping them out
tcg: Put opcodes in a linked list
tcg: Introduce tcg_op_buf_count and tcg_op_buf_full
tcg: Move emit of INDEX_op_end into gen_tb_end
tcg: Reduce ifdefs in tcg-op.c
tcg: Move some opcode generation functions out of line
Peter Maydell [Fri, 13 Feb 2015 05:46:09 +0000 (05:46 +0000)]
target-arm: A64: Avoid left shifting negative integers in disas_pc_rel_addr
Shifting a negative integer left is undefined behaviour in C.
Avoid it by assembling and shifting the offset fields as
unsigned values and then sign extending as the final action.
Peter Maydell [Fri, 13 Feb 2015 05:46:09 +0000 (05:46 +0000)]
target-arm: A64: Fix handling of rotate in logic_imm_decode_wmask
The code in logic_imm_decode_wmask attempts to rotate a mask
value within the bottom 'e' bits of the value with
mask = (mask >> r) | (mask << (e - r));
This has two issues:
* if the element size is 64 then a rotate by zero results
in a shift left by 64, which is undefined behaviour
* if the element size is smaller than 64 then this will
leave junk in the value at bit 'e' and above, which is
not valid input to bitfield_replicate(). As it happens,
the bits at bit 'e' to '2e - r' are exactly the ones
which bitfield_replicate is going to copy in there,
so this isn't a "wrong code generated" bug, but it's
confusing and if we ever put an assert in
bitfield_replicate it would fire on valid guest code.
Fix the former by not doing anything if r is zero, and
the latter by masking with bitmask64(e).
Greg Bellows [Fri, 13 Feb 2015 05:46:08 +0000 (05:46 +0000)]
target-arm: Add 32/64-bit register sync
Add AArch32 to AArch64 register sychronization functions.
Replace manual register synchronization with new functions in
aarch64_cpu_do_interrupt() and HELPER(exception_return)().
Greg Bellows [Fri, 13 Feb 2015 05:46:08 +0000 (05:46 +0000)]
target-arm: Add CPU property to disable AArch64
Adds registration and get/set functions for enabling/disabling the AArch64
execution state on AArch64 CPUs. By default AArch64 execution state is enabled
on AArch64 CPUs, setting the property to off, will disable the execution state.
The below QEMU invocation would have AArch64 execution state disabled.
Alexander Graf [Fri, 13 Feb 2015 05:46:07 +0000 (05:46 +0000)]
pci: Add generic PCIe host bridge
With simple exposure of MMFG, ioport window, mmio window and an IRQ line we
can successfully create a workable PCIe host bridge that can be mapped anywhere
and only needs to get described to the OS using whatever means it likes.
This patch implements such a "generic" host bridge. It handles 4 legacy IRQ
lines. MSIs need to be handled external to the host bridge.
This device is particularly useful for the "pci-host-ecam-generic" driver in
Linux.
Alexander Graf [Fri, 13 Feb 2015 05:46:07 +0000 (05:46 +0000)]
pci: Allocate PCIe host bridge PCI ID
We are going to introduce a PCIe host controller that doesn't exist that
way in real hardware, but still needs to expose some PCIe root device which
has PCI IDs.
Allocate a PCI ID in the Red Hat space that we use for other devices of this
kind.
tcg: Move some opcode generation functions out of line
Some of these functions are really quite large. We have a number of
things that ought to be circularly dependent, but we duplicated code
to break that chain for the inlines.
This saved 25% of the code size of one of the translators I examined.
Alex Williamson [Tue, 10 Feb 2015 17:25:44 +0000 (10:25 -0700)]
vfio: Use vfio type1 v2 IOMMU interface
The difference between v1 and v2 is fairly subtle, simply more
deterministic behavior for unmaps. The v1 interface allows the user
to attempt to unmap sub-regions of previous mappings, returning
success with zero size if unable to comply. This was a reflection of
the underlying IOMMU API. The v2 interface requires that the user
may only unmap fully contained mappings, ie. an unmap cannot intersect
or bisect a previous mapping, but may cover multiple mappings. QEMU
never made use of the sub-region v1 support anyway, so we can support
either v1 or v2. We'll favor v2 since it's newer.
Paolo Bonzini [Tue, 10 Feb 2015 17:25:44 +0000 (10:25 -0700)]
vfio: unmap and free BAR data in instance_finalize
In the case of VFIO, the unrealize callback is too early to munmap the
BARs. The munmap must be delayed until memory accesses are complete.
To do this, split vfio_unmap_bars in two. The removal step, now called
vfio_unregister_bars, remains in vfio_exitfn. The reclamation step
is vfio_unmap_bars and is moved to the instance_finalize callback.
Similarly, quirk MemoryRegions have to be removed during
vfio_unregister_bars, but freeing the data structure must be delayed
to vfio_unmap_bars.
Paolo Bonzini [Tue, 10 Feb 2015 17:25:44 +0000 (10:25 -0700)]
vfio: free dynamically-allocated data in instance_finalize
In order to enable out-of-BQL address space lookup, destruction of
devices needs to be split in two phases.
Unrealize is the first phase; once it complete no new accesses will
be started, but there may still be pending memory accesses can still
be completed.
The second part is freeing the device, which only happens once all memory
accesses are complete. At this point the reference count has dropped to
zero, an RCU grace period must have completed (because the RCU-protected
FlatViews hold a reference to the device via memory_region_ref). This is
when instance_finalize is called.
Freeing data belongs in an instance_finalize callback, because the
dynamically allocated memory can still be used after unrealize by the
pending memory accesses.
This starts the process by creating an instance_finalize callback and
freeing most of the dynamically-allocated data in instance_finalize.
Because instance_finalize is also called on error paths or also when
the device is actually not realized, the common code needs some changes
to be ready for this. The error path in vfio_initfn can be simplified too.
Now that vfio_put_base_device is called unconditionally at instance_finalize
time, it can be called twice if vfio_populate_device fails. This works
but it is slightly harder to follow.
Change vfio_get_device to not touch the vbasedev struct until it will
definitely succeed, moving the vfio_populate_device call back to vfio-pci.
This way, vfio_put_base_device will only be called once.
Paolo Bonzini [Tue, 10 Feb 2015 17:25:44 +0000 (10:25 -0700)]
memory: unregister AddressSpace MemoryListener within BQL
address_space_destroy_dispatch is called from an RCU callback and hence
outside the iothread mutex (BQL). However, after address_space_destroy
no new accesses can hit the destroyed AddressSpace so it is not necessary
to observe changes to the memory map. Move the memory_listener_unregister
call earlier, to make it thread-safe again.
Stefan Weil [Fri, 6 Feb 2015 21:43:13 +0000 (22:43 +0100)]
serial: Fix warnings caused by missing 'static' attribute
Warnings from the Sparse static analysis tool:
hw/char/serial.c:630:26: warning: symbol
'vmstate_serial_thr_ipending' was not declared. Should it be static?
hw/char/serial.c:646:26: warning: symbol
'vmstate_serial_tsr' was not declared. Should it be static?
hw/char/serial.c:665:26: warning: symbol
'vmstate_serial_recv_fifo' was not declared. Should it be static?
hw/char/serial.c:681:26: warning: symbol
'vmstate_serial_xmit_fifo' was not declared. Should it be static?
hw/char/serial.c:697:26: warning: symbol
'vmstate_serial_fifo_timeout_timer' was not declared. Should it be static?
hw/char/serial.c:713:26: warning: symbol
'vmstate_serial_timeout_ipending' was not declared. Should it be static?
hw/char/serial.c:729:26: warning: symbol
'vmstate_serial_poll' was not declared. Should it be static?
Stefan Weil [Fri, 6 Feb 2015 21:43:11 +0000 (22:43 +0100)]
migration: Fix warnings caused by missing 'static' attribute
Warnings from the Sparse static analysis tool:
migration-rdma.c:151:12: warning:
symbol 'wrid_desc' was not declared. Should it be static?
migration-rdma.c:190:12: warning:
symbol 'control_desc' was not declared. Should it be static?
migration-rdma.c:3301:19: warning:
symbol 'rdma_read_ops' was not declared. Should it be static?
migration-rdma.c:3308:19: warning:
symbol 'rdma_write_ops' was not declared. Should it be static?
projects / qemu.git / log