| /qemu/docs/devel/migration/ |
| H A D | compatibility.rst | 7 When we do migration, we have two QEMU processes: the source and the
18 Let's start with a practical example, we start with:
36 I am going to list the number of combinations that we can have. Let's
50 This are the easiest ones, we will not talk more about them in this
53 Now we start with the more interesting cases. Consider the case where
54 we have the same QEMU version in both sides (qemu-5.2) but we are using
72 because we have the limitation than qemu-5.1 doesn't know pc-5.2. So
77 This migration is known as newer to older. We need to make sure
78 when we are developing 5.2 we need to take care about not to break
79 migration to qemu-5.1. Notice that we can't make updates to
[all …]
|
| H A D | main.rst | 10 Restoring a guest is just the opposite operation: we need to load the
16 be relaxed a bit, but for now we can consider that configuration has
19 Once that we are able to save/restore a guest, a new functionality is
122 (which we do to varying degrees in the existing code). Check that offsets
170 We are declaring the state with name "pckbd". The ``version_id`` is
171 3, and there are 4 uint8_t fields in the KBDState structure. We
182 For devices that are ``qdev`` based, we can register the device in the class
232 When we migrate a device, we save/load the state as a series
233 of fields. Sometimes, due to bugs or new functionality, we need to
254 On the receiving side, if we found a subsection for a device that we
[all …]
|
| /qemu/docs/system/ |
| H A D | introduction.rst | 82 For a non-x86 system where we emulate a broad range of machine types,
88 command line to launch VMs, we do want to highlight that there are a
152 In the following example we first define a ``virt`` machine which is a
153 general purpose platform for running Aarch64 guests. We enable
154 virtualisation so we can use KVM inside the emulated guest. As the
155 ``virt`` machine comes with some built in pflash devices we give them
156 names so we can override the defaults later.
164 We then define the 4 vCPUs using the ``max`` option which gives us all
165 the Arm features QEMU is capable of emulating. We enable a more
167 algorithm. We explicitly specify TCG acceleration even though QEMU
[all …]
|
| /qemu/target/hexagon/ |
| H A D | README | 2 processor(DSP). We also support Hexagon Vector eXtensions (HVX). HVX
12 We presented an overview of the project at the 2019 KVM Forum.
38 We start with scripts that generate a bunch of include files. This
109 cases this is necessary for correct execution. We can also override for
113 The gen_tcg.h file has any overrides. For example, we could write
118 C semantics are specified only with macros, we can override the default with
125 In gen_tcg.h, we use the shortcode
129 There are also cases where we brute force the TCG code generation.
134 won't fit in a TCGv or TCGv_i64, so we pass TCGv_ptr variables to pass the
158 Notice that we also generate a variable named <operand>_off for each operand of
[all …]
|
| H A D | fma_emu.c | 163 * On the add/sub, we need to be able to shift out lots of bits, but need a
199 /* Keep around shifted out bits... we might need those later */ in accum_sub()
252 /* Keep around shifted out bits... we might need those later */ in accum_add()
327 * We want DF_MANTBITS bits of mantissa plus the leading one. in accum_round_float64()
328 * That means that we want DF_MANTBITS+1 bits, or 0x000000000000FF_FFFF in accum_round_float64()
329 * So we need to normalize right while the high word is non-zero and in accum_round_float64()
338 * We want to normalize left until we have a leading one in bit 24 in accum_round_float64()
339 * Theoretically, we only need to shift a maximum of one to the left if we in accum_round_float64()
340 * shifted out lots of bits from B, or if we had no shift / 1 shift sticky in accum_round_float64()
347 * OK, now we might need to denormalize because of potential underflow. in accum_round_float64()
[all …]
|
| /qemu/tests/tcg/ |
| H A D | Makefile.target | 5 # These are complicated by the fact we want to build them for guest
6 # systems. This requires knowing what guests we are building and which
7 # ones we have cross-compilers for or docker images with
14 # We only include the host build system for SRC_PATH and we don't
15 # bother with the common rules.mk. We expect the following:
19 # BUILD_STATIC - are we building static binaries
27 # We also accept SPEED=slow to enable slower running tests
29 # We also expect to be in the tests build dir for the FOO-(linux-user|softmmu).
65 # to work around the pipe squashing the status we only pipe the result if
66 # we know it failed and then force failure at the end.
[all …]
|
| /qemu/docs/devel/ |
| H A D | s390-dasd-ipl.rst | 34 the real operating system is loaded into memory and we are ready to hand
49 should contain the needed flags for the operating system we have loaded. The
50 psw's instruction address will point to the location in memory where we want
68 In theory we should merely have to do the following to IPL/boot a guest
79 When we start a channel program we pass the channel subsystem parameters via an
95 it from the disk. So we need to be able to handle this case.
100 Since we are forced to live with prefetch we cannot use the very simple IPL
101 procedure we defined in the preceding section. So we compensate by doing the
112 to read the very next record which will be IPL2. But since we are not reading
113 both IPL1 and IPL2 as part of the same channel program we must manually set
[all …]
|
| H A D | tcg-icount.rst | 38 until the next timer will expire. We store this budget as part of a
44 In the case of icount, before the flag is checked we subtract the
46 would cause the instruction budget to go negative we exit the main
49 was due to expire will expire exactly when we exit the main run loop.
54 While we can adjust the instruction budget for known events like timer
55 expiry we cannot do the same for MMIO. Every load/store we execute
56 might potentially trigger an I/O event, at which point we will need an
59 To deal with this case, when an I/O access is made we:
70 MMIO isn't the only type of operation for which we might need a
|
| /qemu/migration/ |
| H A D | migration-stats.h | 32 * based on MigrationStats. We change to Stat64 any counter that
38 * Number of bytes that were dirty last time that we synced with
39 * the guest memory. We use that to calculate the downtime. As
40 * the remaining dirty amounts to what we know that is still dirty
42 * since we synchronized bitmaps.
50 * Number of times we have synchronized guest bitmaps.
76 * Number of postcopy page faults that we have handled during
93 * Maximum amount of data we can send in a cycle.
118 * This is called when we know we start a new transfer cycle.
134 * Returns how many bytes have we transferred since the beginning of
|
| H A D | migration.h | 63 * big enough and make sure we won't overflow easily.
75 * This points to the host page we're going to install for this temp page.
76 * It tells us after we've received the whole page, where we should put it.
100 * Used to sync thread creations. Note that we can't create threads in
116 /* Set this when we want the fault thread to quit */
127 QemuMutex rp_mutex; /* We send replies from multiple threads */
187 /* The coroutine we should enter (back) after failover */
215 /* A tree of pages that we requested to the source VM */
223 * The mutex helps to maintain the requested pages that we sent to the
285 * used for both the start or recover of a postcopy migration. We'll
[all …]
|
| /qemu/tests/tcg/multiarch/gdbstub/ |
| H A D | interrupt.py | 16 Check that, if thread is resumed, we go back to the same thread when the
20 # Switch to the thread we're going to be running the test in.
26 # While there are cleaner ways to do this, we want to minimize the number of
28 # Ideally, there should be no difference between what we're doing here and
31 # For this to be safe, we only need the prologue of loop() to not have
32 # instructions that may have problems with what we're doing here. We don't
40 # Check whether the thread we're in after the interruption is the same we
|
| /qemu/include/user/ |
| H A D | safe-syscall.h | 31 * (Errnos are host errnos; we rely on QEMU_ERESTARTSYS not clashing
59 * for which we need to either return EINTR or arrange for the guest
72 * signal could arrive just before we make the host syscall inside libc,
81 * which are only technically blocking (ie which we know in practice won't
92 * The basic setup is that we make the host syscall via a known
96 * instruction then we change the PC to point at a "return
99 * Then in the main.c loop if we see this magic return value we adjust
104 * (1) signal came in just before we took the host syscall (a race);
105 * in this case we'll take the guest signal and have another go
111 * in this case we want to restart the guest syscall also, and so
[all …]
|
| /qemu/target/arm/tcg/ |
| H A D | m_helper.c | 139 * user-only emulation we don't have the MPU. in HELPER()
140 * Luckily since we know we are NonSecure unprivileged (and that in in HELPER()
203 * What kind of stack write are we doing? This affects how exceptions
276 * By pending the exception at this point we are making in v7m_stack_write()
280 * later if we have two derived exceptions. in v7m_stack_write()
281 * The only case when we must not pend the exception but instead in v7m_stack_write()
282 * throw it away is if we are doing the push of the callee registers in v7m_stack_write()
283 * and we've already generated a derived exception (this is indicated in v7m_stack_write()
284 * by the caller passing STACK_IGNFAULTS). Even in this case we will in v7m_stack_write()
348 * By pending the exception at this point we are making in v7m_stack_read()
[all …]
|
| /qemu/target/hexagon/idef-parser/ |
| H A D | README.rst | 10 To better understand the scope of the idef-parser, we'll explore an applicative
39 passed explicitly as function parameters. Among the passed parameters we will
51 After the instruction identifier, we have a series of parameters that represents
55 We will leverage this information to infer several information:
93 Then, we are generating the sum tinycode operator between the selected
100 The result of the addition is now stored in the temporary, we move it into the
109 instruction semantics in ``semantics_generated.pyinc`` that we need to consider.
130 we obtain the pseudo code
146 To finish the above example, after preprocessing ``J2_jumpr`` we obtain
224 With this initial portion of the grammar we are defining the instruction, its'
[all …]
|
| /qemu/docs/ |
| H A D | rdma.txt | 162 on the receiver side is registered and pinned, we're
185 At this point, we define a control channel on top of SEND messages
226 After ram block exchange is completed, we have two protocol-level
234 1. We transmit a READY command to let the sender know that
235 we are *ready* to receive some data bytes on the control channel.
236 2. Before attempting to receive the expected command, we post another
237 RQ work request to replace the one we just used up.
240 5. Verify that the command-type and version received matches the one we expected.
247 2. Optionally: if we are expecting a response from the command
248 (that we have not yet transmitted), let's post an RQ
[all …]
|
| /qemu/tests/qemu-iotests/tests/ |
| H A D | block-status-cache | 41 # which we are going to query to provoke a block-status inquiry with
63 We can provoke a want_zero=false call with `qemu-img map` over NBD with
64 x-dirty-bitmap=qemu:allocation-depth, so we first run a normal `map`
90 # We need to run this map twice: On the first call, we probably still
95 # If we did a want_zero=true call at this point, we would thus get
97 # we would get fresh block-status information from the driver, which
101 # Therefore, we need a second want_zero=false map to reproduce:
109 # subsequent map operation will be served from the cache, and so we can
116 # whether we get correct information (i.e. the same as we got on our
133 # we can only use the raw format
|
| /qemu/include/qemu/ |
| H A D | stats64.h | 131 /* High 32 bits are equal. Read low after high, otherwise we in stat64_min()
133 * 0x1234,0x8000 and we read it as 0x1234,0x0000). Pairs with in stat64_min()
142 /* See if we were lucky and a writer raced against us. The in stat64_min()
143 * barrier is theoretically unnecessary, but if we remove it in stat64_min()
144 * we may miss being lucky. in stat64_min()
153 /* If the value changes in any way, we have to take the lock. */ in stat64_min()
171 /* High 32 bits are equal. Read low after high, otherwise we in stat64_max()
173 * 0x1235,0x0000 and we read it as 0x1235,0x8000). Pairs with in stat64_max()
182 /* See if we were lucky and a writer raced against us. The in stat64_max()
183 * barrier is theoretically unnecessary, but if we remove it in stat64_max()
[all …]
|
| /qemu/target/arm/ |
| H A D | arm-powerctl.c | 67 /* Initialize the cpu we are turning on */ in arm_set_cpu_on_async_work()
72 /* We check if the started CPU is now at the correct level */ in arm_set_cpu_on_async_work()
114 * if we are booting in AArch64 mode then "entry" needs to be 4 bytes in arm_set_cpu_on()
120 /* Retrieve the cpu we are powering up */ in arm_set_cpu_on()
150 * For now we don't support booting an AArch64 CPU in AArch32 mode in arm_set_cpu_on()
151 * TODO: We should add this support later in arm_set_cpu_on()
161 * If another CPU has powered the target on we are in the state in arm_set_cpu_on()
173 /* To avoid racing with a CPU we are just kicking off we do the in arm_set_cpu_on()
186 /* We are good to go */ in arm_set_cpu_on()
195 /* Initialize the cpu we are turning on */ in arm_set_cpu_on_and_reset_async_work()
[all …]
|
| /qemu/tests/qtest/migration/ |
| H A D | framework.h | 63 * Our goal is to ensure that we run a single full migration
67 * We can't directly synchronize with the start of a migration
68 * so we have to apply some tricks monitoring memory that is
71 * Initially we set the migration bandwidth to an insanely
76 * so we can't let the entire migration pass run at this speed.
77 * Our intent is to let it run just long enough that we can
81 * Before migration starts, we write a 64-bit magic marker
88 * Finally we go back to the source and read a byte just
89 * before the marker until we see it flip in value. This
93 * IOW, we're guaranteed at least a 2nd migration pass
[all …]
|
| /qemu/tests/tcg/aarch64/system/ |
| H A D | boot.S | 100 * Initialise the stack for whatever EL we are in before
101 * anything else, we need it to be able to _exit cleanly.
102 * It's smaller than the stack we pass to the C code but we
111 * EL needed for the test. However if no semihosting args are set we will
141 /* Are we already in a lower EL than we want? */
158 /* Ensure we trap if we get anything wrong */
173 * We need to check if EL2 is actually enabled via ID_AA64PFR0_EL1,
174 * otherwise we should just jump straight to EL1.
208 /* Ensure we trap if we get anything wrong */
236 * At the target EL for the test, usually EL1. Note we still
[all …]
|
| /qemu/tests/qemu-iotests/ |
| H A D | 245 | 75 # Once the VM is shut down we can parse the log and see if qemu-io
108 # If key has the form "foo.bar" then we need to do
142 # We can reopen the image passing the same options
145 # We can also reopen passing a child reference in 'file'
148 # We cannot change any of these
184 # We can't reopen the image passing the same options, 'backing' is mandatory
187 # Everything works if we pass 'backing' using the existing node name
193 # We can't use a non-existing or empty (non-NULL) node as the backing image
197 # We can reopen the image just fine if we specify the backing options
203 # We cannot change any of these options
[all …]
|
| /qemu/block/ |
| H A D | graph-lock.c | 120 * reader_count >= 1: we don't know if writer read has_writer == 0 or 1, in bdrv_graph_wrlock()
121 * but we need to wait. in bdrv_graph_wrlock()
126 * has_writer must be 0 while polling, otherwise we get a deadlock if in bdrv_graph_wrlock()
135 * We want to only check reader_count() after has_writer = 1 is visible in bdrv_graph_wrlock()
136 * to other threads. That way no more readers can sneak in after we've in bdrv_graph_wrlock()
181 /* make sure writer sees reader_count before we check has_writer */ in bdrv_graph_co_rdlock()
186 * has_writer == 1: we don't know if writer read reader_count == 0 in bdrv_graph_co_rdlock()
187 * or > 0, but we need to wait anyways because in bdrv_graph_co_rdlock()
206 * we will enter this critical section and call aio_wait_kick(). in bdrv_graph_co_rdlock()
210 * Additional check when we use the above lock to synchronize in bdrv_graph_co_rdlock()
[all …]
|
| /qemu/util/ |
| H A D | lockcnt.c | 76 * only getting out of the loop if we can have another shot at the in qemu_lockcnt_cmpxchg_or_wait()
77 * fast path. Once we can, get out to compute the new destination in qemu_lockcnt_cmpxchg_or_wait()
135 /* If we were woken by another thread, we should also wake one because in qemu_lockcnt_inc()
136 * we are effectively releasing the lock that was given to us. This is in qemu_lockcnt_inc()
178 /* At this point we do not know if there are more waiters. Assume in qemu_lockcnt_dec_and_lock()
186 /* If we were woken by another thread, but we're returning in unlocked in qemu_lockcnt_dec_and_lock()
187 * state, we should also wake a thread because we are effectively in qemu_lockcnt_dec_and_lock()
219 /* At this point we do not know if there are more waiters. Assume in qemu_lockcnt_dec_if_lock()
226 /* If we were woken by another thread, but we're returning in unlocked in qemu_lockcnt_dec_if_lock()
227 * state, we should also wake a thread because we are effectively in qemu_lockcnt_dec_if_lock()
[all …]
|
| /qemu/docs/sphinx/ |
| H A D | hxtool.py | 34 # We parse hx files with a state machine which may be in one of two
52 # empty we ignore the directive -- these are used only to add
55 # Return the heading text. We strip out any trailing ':' for
68 # Return the heading text. We strip out any trailing ':' for
100 # We build up lines of rST in this ViewList, which we will
142 # We had some rST fragments before the first
143 # DEFHEADING. We don't have a section to put
156 # Not a directive: put in output if we are in rST fragment
162 # We don't have multiple sections, so just parse the rst
163 # fragments into a dummy node so we can return the children.
[all …]
|
| /qemu/accel/tcg/ |
| H A D | cpu-exec.c | 137 /* Print every 2s max if the guest is late. We limit the number in init_delay_params()
176 * We know that the first page matched, and an otherwise valid TB in tb_lookup_cmp()
178 * therefore we know that generating a new TB from the current PC in tb_lookup_cmp()
232 /* we should never be trying to look up an INVALID tb */ in tb_lookup()
301 * we would fail to make forward progress in reverse-continue. in check_for_breakpoints_slow()
313 * If we have an exact pc match, trigger the breakpoint. in check_for_breakpoints_slow()
371 * the tcg epilogue so that we return into cpu_tb_exec.
379 * By definition we've just finished a TB, so I/O is OK. in HELPER()
383 * The next TB, if we chain to it, will clear the flag again. in HELPER()
443 * until we actually need to modify the TB. The read-only copy, in cpu_tb_exec()
[all …]
|