This diagram make user better understand COLO.
Suggested by Markus Armbruster.
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
COLO thread may sleep at qemu_sem_wait(&s->colo_checkpoint_sem),
while failover works begin, It's better to wakeup it to quick
the process.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Notify all net filters about the checkpoint and failover event.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
After one round of checkpoint, the states between PVM and SVM
become consistent, so it is unnecessary to adjust the sequence
of net packets for old connections, besides, while failover
happens, filter-rewriter will into failover mode that needn't
handle the new TCP connection.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Filter needs to process the event of checkpoint/failover or
other event passed by COLO frame.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Don't need to flush all VM's ram from cache, only
flush the dirty pages since last checkpoint
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
There are several stages during loadvm/savevm process. In different stage,
migration incoming processes different types of sections.
We want to control these stages more accuracy, it will benefit COLO
performance, we don't have to save type of QEMU_VM_SECTION_START
sections everytime while do checkpoint, besides, we want to separate
the process of saving/loading memory and devices state.
So we add three new helper functions: qemu_load_device_state() and
qemu_savevm_live_state() to achieve different process during migration.
Besides, we make qemu_loadvm_state_main() and qemu_save_device_state()
public, and simplify the codes of qemu_save_device_state() by calling the
wrapper qemu_savevm_state_header().
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Libvirt or other high level software can use this command query colo status.
You can test this command like that:
{'execute':'query-colo-status'}
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Suggested by Markus Armbruster rename COLO unknown mode to none mode.
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
If some errors happen during VM's COLO FT stage, it's important to
notify the users of this event. Together with 'x-colo-lost-heartbeat',
Users can intervene in COLO's failover work immediately.
If users don't want to get involved in COLO's failover verdict,
it is still necessary to notify users that we exited COLO mode.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
During the time of VM's running, PVM may dirty some pages, we will transfer
PVM's dirty pages to SVM and store them into SVM's RAM cache at next checkpoint
time. So, the content of SVM's RAM cache will always be same with PVM's memory
after checkpoint.
Instead of flushing all content of PVM's RAM cache into SVM's MEMORY,
we do this in a more efficient way:
Only flush any page that dirtied by PVM since last checkpoint.
In this way, we can ensure SVM's memory same with PVM's.
Besides, we must ensure flush RAM cache before load device state.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
We record the address of the dirty pages that received,
it will help flushing pages that cached into SVM.
Here, it is a trick, we record dirty pages by re-using migration
dirty bitmap. In the later patch, we will start the dirty log
for SVM, just like migration, in this way, we can record both
the dirty pages caused by PVM and SVM, we only flush those dirty
pages from RAM cache while do checkpoint.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
We should not load PVM's state directly into SVM, because there maybe some
errors happen when SVM is receving data, which will break SVM.
We need to ensure receving all data before load the state into SVM. We use
an extra memory to cache these data (PVM's ram). The ram cache in secondary side
is initially the same as SVM/PVM's memory. And in the process of checkpoint,
we cache the dirty pages of PVM into this ram cache firstly, so this ram cache
always the same as PVM's memory at every checkpoint, then we flush this cached ram
to SVM after we receive all PVM's state.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
We need to know if migration is going into COLO state for
incoming side before start normal migration.
Instead by using the VMStateDescription to send colo_state
from source side to destination side, we use MIG_CMD_ENABLE_COLO
to indicate whether COLO is enabled or not.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Make sure master start block replication after slave's block
replication started.
Besides, we need to activate VM's blocks before goes into
COLO state.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Li Zhijian <lizhijian@cn.fujitsu.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
For COLO FT, both the PVM and SVM run at the same time,
only sync the state while it needs.
So here, let SVM runs while not doing checkpoint, change
DEFAULT_MIGRATE_X_CHECKPOINT_DELAY to 200*100.
Besides, we forgot to release colo_checkpoint_semd and
colo_delay_timer, fix them here.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
It's a good idea to use notifier to notify COLO frame of
inconsistent packets comparing.
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
While do checkpoint, we need to flush all the unhandled packets,
By using the filter notifier mechanism, we can easily to notify
every compare object to do this process, which runs inside
of compare threads as a coroutine.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
We add almost full TCP state machine in filter-rewriter, except
TCPS_LISTEN and some simplify in VM active close FIN states.
The reason for this simplify job is because guest kernel will track
the TCP status and wait 2MSL time too, if client resend the FIN packet,
guest will resend the last ACK, so we needn't wait 2MSL time in filter-rewriter.
After a net connection is closed, we didn't clear its related resources
in connection_track_table, which will lead to memory leak.
Let's track the state of net connection, if it is closed, its related
resources will be cleared up.
Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Zhang Chen <zhangckid@gmail.com>
Signed-off-by: Zhang Chen <chen.zhang@intel.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
When op raises an exception, it may not have initialized the output
temps that would be written back by wout or cout.
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Emilio G. Cota <cota@braap.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Emilio G. Cota <cota@braap.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
GCC7+ will no longer advertise support for 16-byte __atomic operations
if only cmpxchg is supported, as for x86_64. Fortunately, x86_64 still
has support for __sync_compare_and_swap_16 and we can make use of that.
AArch64 does not have, nor ever has had such support, so open-code it.
Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Isolate the computation of an index from an address into a
helper before we change that function.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
[ cota: convert tlb_vaddr_to_host; use atomic_read on addr_write ]
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181009175129.17888-2-cota@braap.org>
Currently we rely on atomic operations for cross-CPU invalidations.
There are two cases that these atomics miss: cross-CPU invalidations
can race with either (1) vCPU threads flushing their TLB, which
happens via memset, or (2) vCPUs calling tlb_reset_dirty on their TLB,
which updates .addr_write with a regular store. This results in
undefined behaviour, since we're mixing regular and atomic ops
on concurrent accesses.
Fix it by using tlb_lock, a per-vCPU lock. All updaters of tlb_table
and the corresponding victim cache now hold the lock.
The readers that do not hold tlb_lock must use atomic reads when
reading .addr_write, since this field can be updated by other threads;
the conversion to atomic reads is done in the next patch.
Note that an alternative fix would be to expand the use of atomic ops.
However, in the case of TLB flushes this would have a huge performance
impact, since (1) TLB flushes can happen very frequently and (2) we
currently use a full memory barrier to flush each TLB entry, and a TLB
has many entries. Instead, acquiring the lock is barely slower than a
full memory barrier since it is uncontended, and with a single lock
acquisition we can flush the entire TLB.
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181009174557.16125-6-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181009174557.16125-5-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Paves the way for the addition of a per-TLB lock.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181009174557.16125-4-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
As far as I can tell tlb_flush does not need to be called
this early. tlb_flush is eventually called after the CPU
has been realized.
This change paves the way to the introduction of tlb_init,
which will be called from cpu_exec_realizefn.
Cc: Guan Xuetao <gxt@mprc.pku.edu.cn>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181009174557.16125-3-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
As far as I can tell tlb_flush does not need to be called
this early. tlb_flush is eventually called after the CPU
has been realized.
This change paves the way to the introduction of tlb_init,
which will be called from cpu_exec_realizefn.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181009174557.16125-2-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
When we implemented per-vCPU TCG contexts, we forgot to also
distribute the tcg_time counter, which has remained as a global
accessed without any serialization, leading to potentially missed
counts.
Fix it by distributing the field over the TCG contexts, embedding
it into TCGProfile with a field called "cpu_exec_time", which is more
descriptive than "tcg_time". Add a function to query this value
directly, and for completeness, fill in the field in
tcg_profile_snapshot, even though its callers do not use it.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181010144853.13005-5-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This plugs two 4-byte holes in 64-bit.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181010144853.13005-4-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We forgot to initialize n in commit 15fa08f845 ("tcg: Dynamically
allocate TCGOps", 2017-12-29).
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181010144853.13005-3-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Consistently access u16.high with atomics to avoid
undefined behaviour in MTTCG.
Note that icount_decr.u16.low is only used in icount mode,
so regular accesses to it are OK.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-Id: <20181010144853.13005-2-cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Rather than test NOCHAIN before linking, do not emit the
goto_tb opcode at all. We already do this for goto_ptr.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Fix misplaced 'break' in handling of NM_SHRA_R_PH. Found by
Coverity (CID 1395627).
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Stefan Markovic <smarkovic@wavecomp.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Fix emulation of microMIPS R6 <SELEQZ|SELNEZ>.<D|S> instructions.
Their handling was permuted.
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Matthew Fortune <matthew.fortune@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Implement hardware page table walker. This implementation is
limiter only to MIPS32.
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Add reset state for PWSize and PWField registers. The reset state
is different for pre-R6 and R6 (and post-R6) ISAa.
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Add PWCtl register (CP0 Register 5, Select 6).
The PWCtl register configures hardware page table walking for TLB
refills.
This register is required for the hardware page walker feature. It
exists only if Config3 PW bit is set to 1. It contains following
fields:
PWEn (31) - Hardware Page Table walker enable
PWDirExt (30) - If 1, 4-th level implemented (MIPS64 only)
XK (28) - If 1, walker handles xkseg (MIPS64 only)
XS (27) - If 1, walker handles xsseg (MIPS64 only)
XU (26) - If 1, walker handles xuseg (MIPS64 only)
DPH (7) - Dual Page format of Huge Page support
HugePg (6) - Huge Page PTE supported in Directory levels
PSn (5..0) - Bit position of PTEvld in Huge Page PTE
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Add PWSize register (CP0 Register 5, Select 7).
The PWSize register configures hardware page table walking for TLB
refills.
This register is required for the hardware page walker feature. It
exists only if Config3 PW bit is set to 1. It contains following
fields:
BDW (37..32) Base Directory index width (MIPS64 only)
GDW (29..24) Global Directory index width
UDW (23..18) Upper Directory index width
MDW (17..12) Middle Directory index width
PTW (11..6 ) Page Table index width
PTEW ( 5..0 ) Left shift applied to the Page Table index
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Add PWField register (CP0 Register 5, Select 6).
The PWField register configures hardware page table walking for TLB
refills.
This register is required for the hardware page walker feature. It
exists only if Config3 PW bit is set to 1. It contains following
fields:
MIPS64:
BDI (37..32) - Base Directory index
GDI (29..24) - Global Directory index
UDI (23..18) - Upper Directory index
MDI (17..12) - Middle Directory index
PTI (11..6 ) - Page Table index
PTEI ( 5..0 ) - Page Table Entry shift
MIPS32:
GDW (29..24) - Global Directory index
UDW (23..18) - Upper Directory index
MDW (17..12) - Middle Directory index
PTW (11..6 ) - Page Table index
PTEW ( 5..0 ) - Page Table Entry shift
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Add PWBase register (CP0 Register 5, Select 5).
The PWBase register contains the Page Table Base virtual address.
This register is required for the hardware page walker feature. It
exists only if Config3 PW bit is set to 1.
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Signed-off-by: Yongbok Kim <yongbok.kim@mips.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Add field corresponding to CP0 Config2 to DisasContext. This is
needed for availability control via Config2 bits.
Reviewed-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Stefan Markovic <smarkovic@wavecomp.com>
Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
