It's either "GNU *Library* General Public version 2" or "GNU Lesser
General Public version *2.1*", but there was no "version 2.0" of the
"Lesser" library. So assume that version 2.1 is meant here.
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <1548252536-6242-5-git-send-email-thuth@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Include the cluster number in the hash we use to look
up TBs. This is important because a TB that is valid
for one cluster at a given physical address and set
of CPU flags is not necessarily valid for another:
the two clusters may have different views of physical
memory, or may have different CPU features (eg FPU
present or absent).
We put the cluster number in the high 8 bits of the
TB cflags. This gives us up to 256 clusters, which should
be enough for anybody. If we ever need more, or need
more bits in cflags for other purposes, we could make
tb_hash_func() take more data (and expand qemu_xxhash7()
to qemu_xxhash8()).
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Edgar E. Iglesias <edgar.iglesias@xilinx.com>
Message-id: 20190121152218.9592-4-peter.maydell@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>
When we support execution from non-RAM MMIO regions, get_page_addr_code()
will return -1 to indicate that there is no RAM at the requested address.
Handle this in the cpu-exec TB hashtable lookup code, treating it as
"no match found".
Note that the call to get_page_addr_code() in tb_lookup_cmp() needs
no changes -- a return of -1 will already correctly result in the
function returning false.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Emilio G. Cota <cota@braap.org>
Tested-by: Cédric Le Goater <clg@kaod.org>
Message-id: 20180710160013.26559-3-peter.maydell@linaro.org
Use mmap_lock in user-mode to protect TCG state and the page descriptors.
In !user-mode, each vCPU has its own TCG state, so no locks needed.
Per-page locks are used to protect the page descriptors.
Per-TB locks are used in both modes to protect TB jumps.
Some notes:
- tb_lock is removed from notdirty_mem_write by passing a
locked page_collection to tb_invalidate_phys_page_fast.
- tcg_tb_lookup/remove/insert/etc have their own internal lock(s),
so there is no need to further serialize access to them.
- do_tb_flush is run in a safe async context, meaning no other
vCPU threads are running. Therefore acquiring mmap_lock there
is just to please tools such as thread sanitizer.
- Not visible in the diff, but tb_invalidate_phys_page already
has an assert_memory_lock.
- cpu_io_recompile is !user-only, so no mmap_lock there.
- Added mmap_unlock()'s before all siglongjmp's that could
be called in user-mode while mmap_lock is held.
+ Added an assert for !have_mmap_lock() after returning from
the longjmp in cpu_exec, just like we do in cpu_exec_step_atomic.
Performance numbers before/after:
Host: AMD Opteron(tm) Processor 6376
ubuntu 17.04 ppc64 bootup+shutdown time
700 +-+--+----+------+------------+-----------+------------*--+-+
| + + + + + *B |
| before ***B*** ** * |
|tb lock removal ###D### *** |
600 +-+ *** +-+
| ** # |
| *B* #D |
| *** * ## |
500 +-+ *** ### +-+
| * *** ### |
| *B* # ## |
| ** * #D# |
400 +-+ ** ## +-+
| ** ### |
| ** ## |
| ** # ## |
300 +-+ * B* #D# +-+
| B *** ### |
| * ** #### |
| * *** ### |
200 +-+ B *B #D# +-+
| #B* * ## # |
| #* ## |
| + D##D# + + + + |
100 +-+--+----+------+------------+-----------+------------+--+-+
1 8 16 Guest CPUs 48 64
png: https://imgur.com/HwmBHXe
debian jessie aarch64 bootup+shutdown time
90 +-+--+-----+-----+------------+------------+------------+--+-+
| + + + + + + |
| before ***B*** B |
80 +tb lock removal ###D### **D +-+
| **### |
| **## |
70 +-+ ** # +-+
| ** ## |
| ** # |
60 +-+ *B ## +-+
| ** ## |
| *** #D |
50 +-+ *** ## +-+
| * ** ### |
| **B* ### |
40 +-+ **** # ## +-+
| **** #D# |
| ***B** ### |
30 +-+ B***B** #### +-+
| B * * # ### |
| B ###D# |
20 +-+ D ##D## +-+
| D# |
| + + + + + + |
10 +-+--+-----+-----+------------+------------+------------+--+-+
1 8 16 Guest CPUs 48 64
png: https://imgur.com/iGpGFtv
The gains are high for 4-8 CPUs. Beyond that point, however, unrelated
lock contention significantly hurts scalability.
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>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This applies to both user-mode and !user-mode emulation.
Instead of relying on a global lock, protect the list of incoming
jumps with tb->jmp_lock. This lock also protects tb->cflags,
so update all tb->cflags readers outside tb->jmp_lock to use
atomic reads via tb_cflags().
In order to find the destination TB (and therefore its jmp_lock)
from the origin TB, we introduce tb->jmp_dest[].
I considered not using a linked list of jumps, which simplifies
code and makes the struct smaller. However, it unnecessarily increases
memory usage, which results in a performance decrease. See for
instance these numbers booting+shutting down debian-arm:
Time (s) Rel. err (%) Abs. err (s) Rel. slowdown (%)
------------------------------------------------------------------------------
before 20.88 0.74 0.154512 0.
after 20.81 0.38 0.079078 -0.33524904
GTree 21.02 0.28 0.058856 0.67049808
GHashTable + xxhash 21.63 1.08 0.233604 3.5919540
Using a hash table or a binary tree to keep track of the jumps
doesn't really pay off, not only due to the increased memory usage,
but also because most TBs have only 0 or 1 jumps to them. The maximum
number of jumps when booting debian-arm that I measured is 35, but
as we can see in the histogram below a TB with that many incoming jumps
is extremely rare; the average TB has 0.80 incoming jumps.
n_jumps: 379208; avg jumps/tb: 0.801099
dist: [0.0,1.0)|▄█▁▁▁▁▁▁▁▁▁▁▁ ▁▁▁▁▁▁ ▁▁▁ ▁▁▁ ▁|[34.0,35.0]
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Use the recently-gained QHT feature of returning the matching TB if it
already exists. This allows us to get rid of the lookup we perform
right after acquiring tb_lock.
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The appended adds assertions to make sure we do not longjmp with page
locks held. Note that user-mode has nothing to check, since page_locks
are !user-mode only.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This paves the way for enabling scalable parallel generation of TCG code.
Instead of tracking TBs with a single binary search tree (BST), use a
BST for each TCG region, protecting it with a lock. This is as scalable
as it gets, since each TCG thread operates on a separate region.
The core of this change is the introduction of struct tcg_region_tree,
which contains a pointer to a GTree and an associated lock to serialize
accesses to it. We then allocate an array of tcg_region_tree's, adding
the appropriate padding to avoid false sharing based on
qemu_dcache_linesize.
Given a tc_ptr, we first find the corresponding region_tree. This
is done by special-casing the first and last regions first, since they
might be of size != region.size; otherwise we just divide the offset
by region.stride. I was worried about this division (several dozen
cycles of latency), but profiling shows that this is not a fast path.
Note that region.stride is not required to be a power of two; it
is only required to be a multiple of the host's page size.
Note that with this design we can also provide consistent snapshots
about all region trees at once; for instance, tcg_tb_foreach
acquires/releases all region_tree locks before/after iterating over them.
For this reason we now drop tb_lock in dump_exec_info().
As an alternative I considered implementing a concurrent BST, but this
can be tricky to get right, offers no consistent snapshots of the BST,
and performance and scalability-wise I don't think it could ever beat
having separate GTrees, given that our workload is insert-mostly (all
concurrent BST designs I've seen focus, understandably, on making
lookups fast, which comes at the expense of convoluted, non-wait-free
insertions/removals).
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>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
qht_lookup now uses the default cmp function. qht_lookup_custom is defined
to retain the old behaviour, that is a cmp function is explicitly provided.
qht_insert will gain use of the default cmp in the next patch.
Note that we move qht_lookup_custom's @func to be the last argument,
which makes the new qht_lookup as simple as possible.
Instead of this (i.e. keeping @func 2nd):
0000000000010750 <qht_lookup>:
10750: 89 d1 mov %edx,%ecx
10752: 48 89 f2 mov %rsi,%rdx
10755: 48 8b 77 08 mov 0x8(%rdi),%rsi
10759: e9 22 ff ff ff jmpq 10680 <qht_lookup_custom>
1075e: 66 90 xchg %ax,%ax
We get:
0000000000010740 <qht_lookup>:
10740: 48 8b 4f 08 mov 0x8(%rdi),%rcx
10744: e9 37 ff ff ff jmpq 10680 <qht_lookup_custom>
10749: 0f 1f 80 00 00 00 00 nopl 0x0(%rax)
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>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Usually the logging of the CPU state produced by -d cpu is sufficient
to diagnose problems, but sometimes you want to see the state of
the floating point registers as well. We don't want to enable that
by default as it adds a lot of extra data to the log; instead,
allow it to be optionally enabled via -d fpu.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180510130024.31678-1-peter.maydell@linaro.org
In icount mode, instructions that access io memory spaces in the middle
of the translation block invoke TB recompilation. After recompilation,
such instructions become last in the TB and are allowed to access io
memory spaces.
When the code includes instruction like i386 'xchg eax, 0xffffd080'
which accesses APIC, QEMU goes into an infinite loop of the recompilation.
This instruction includes two memory accesses - one read and one write.
After the first access, APIC calls cpu_report_tpr_access, which restores
the CPU state to get the current eip. But cpu_restore_state_from_tb
resets the cpu->can_do_io flag which makes the second memory access invalid.
Therefore the second memory access causes a recompilation of the block.
Then these operations repeat again and again.
This patch moves resetting cpu->can_do_io flag from
cpu_restore_state_from_tb to cpu_loop_exit* functions.
It also adds a parameter for cpu_restore_state which controls restoring
icount. There is no need to restore icount when we only query CPU state
without breaking the TB. Restoring it in such cases leads to the
incorrect flow of the virtual time.
In most cases new parameter is true (icount should be recalculated).
But there are two cases in i386 and openrisc when the CPU state is only
queried without the need to break the TB. This patch fixes both of
these cases.
Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgaluk@ispras.ru>
Message-Id: <20180409091320.12504.35329.stgit@pasha-VirtualBox>
[rth: Make can_do_io setting unconditional; move from cpu_exec;
make cpu_loop_exit_{noexc,restore} call cpu_loop_exit.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Function cpu_handle_interrupt calls cc->cpu_exec_interrupt to process
pending hardware interrupts. Under the hood cpu_exec_interrupt uses
cpu->exception_index to pass information to the internal function which
is usually common for exception and interrupt processing.
But this value is not reset after return and may be processed again
by cpu_handle_exception. This does not happen due to overwriting
the exception_index at the end of cpu_handle_interrupt.
But this branch may also overwrite the valid exception_index in some cases.
Therefore this patch:
1. resets exception_index just after the call to cpu_exec_interrupt
2. prevents overwriting the meaningful value of exception_index
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20180227095140.1060.61357.stgit@pasha-VirtualBox>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Pavel Dovgalyuk <Pavel.Dovgaluk@ispras.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20171217055023.29225-1-pbonzini@redhat.com>
[rth: Also change the Chain logging in helper_lookup_tb_ptr.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The conditional memory barrier not only looks strange but actually is
wrong.
On s390x, I can reproduce interrupts via cpu_interrupt() not leading to
a proper kick out of emulation every now and then. cpu_interrupt() is
especially used for inter CPU communication via SIGP (esp. external
calls and emergency interrupts).
With this patch, I was not able to reproduce. (esp. no stalls or hangs
in the guest).
My setup is s390x MTTCG with 16 VCPUs on 8 CPU host, running make -j16.
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20171129191319.11483-1-david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch ensures that icount_decr.u32.high is clear before calling
cpu_exec_nocache when exception is pending. Because the exception is
caused by the first instruction in the block and it cannot be executed
without resetting the flag.
There are two parts in the fix. First, clear icount_decr.u32.high in
cpu_handle_interrupt (just before processing the "dependent" request,
stored in cpu->interrupt_request or cpu->exit_request) rather than
cpu_loop_exec_tb; this ensures that cpu_handle_exception is always
reached with zero icount_decr.u32.high unless another interrupt has
happened in the meanwhile.
Second, try to cause the exception at the beginning of
cpu_handle_exception, and exit immediately if the TB cannot
execute. With this change, interrupts are processed and
cpu_exec_nocache can make process.
Signed-off-by: Maria Klimushenkova <maria.klimushenkova@ispras.ru>
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Message-Id: <20171114081818.27640.33165.stgit@pasha-VirtualBox>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This patch adds a condition before overwriting exception_index fiels.
It is needed when exception_index is already set to some meaningful value.
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Message-Id: <20171114081812.27640.26372.stgit@pasha-VirtualBox>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit ac03ee5331 narrowed the scope of the exclusive
region so it only covers when we're executing the TB, not when
we're generating it. However it missed that there is more than
one execution path out of cpu_tb_exec -- if the atomic insn
causes an exception then the code will longjmp out, skipping
the code to end the exclusive region. This causes QEMU to hang
the next time the CPU calls start_exclusive(), waiting for
itself to exit the region.
Move the "end the region" code out to the end of the
function so that it is run for both normal exit and also
for exit-via-longjmp. We have to use a volatile bool flag
to decide whether we need to end the region, because we
can longjump out of the codegen as well as the execution.
(For some reason this only reproduces for me with a clang
optimized build, not a gcc debug build.)
Reviewed-by: Emilio G. Cota <cota@braap.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Fixes: ac03ee5331
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <1509640536-32160-1-git-send-email-peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Groundwork for supporting multiple TCG contexts.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We don't really free anything in this function anymore; we just remove
the TB from the binary search tree.
Suggested-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Now that we have curr_cflags, we can include CF_USE_ICOUNT
early and then remove it as necessary.
Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Now that all code generation has been converted to check CF_PARALLEL, we can
generate !CF_PARALLEL code without having yet set !parallel_cpus --
and therefore without having to be in the exclusive region during
cpu_exec_step_atomic.
While at it, merge cpu_exec_step into cpu_exec_step_atomic.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
We were generating code during tb_invalidate_phys_page_range,
check_watchpoint, cpu_io_recompile, and (seemingly) discarding
the TB, assuming that it would magically be picked up during
the next iteration through the cpu_exec loop.
Instead, record the desired cflags in CPUState so that we request
the proper TB so that there is no more magic.
Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This will enable us to decouple code translation from the value
of parallel_cpus at any given time. It will also help us minimize
TB flushes when generating code via EXCP_ATOMIC.
Note that the declaration of parallel_cpus is brought to exec-all.h
to be able to define there the "curr_cflags" inline.
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
In preparation for adding tc.size to be able to keep track of
TB's using the binary search tree implementation from glib.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This gets rid of a hole in struct TranslationBlock.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reusing the have_tb_lock name, which is also defined in translate-all.c,
makes code reviewing unnecessarily harder.
Avoid potential confusion by renaming the local have_tb_lock variable
to something else.
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Replace the USE_DIRECT_JUMP ifdef with a TCG_TARGET_HAS_direct_jump
boolean test. Replace the tb_set_jmp_target1 ifdef with an unconditional
function tb_target_set_jmp_target.
While we're touching all backends, add a parameter for tb->tc_ptr;
we're going to need it shortly for some backends.
Move tb_set_jmp_target and tb_add_jump from exec-all.h to cpu-exec.c.
This opens the possibility for TCG_TARGET_HAS_direct_jump to be
a runtime decision -- based on host cpu capabilities, the size of
code_gen_buffer, or a future debugging switch.
Signed-off-by: Richard Henderson <rth@twiddle.net>
Every vCPU now uses a separate set of TBs for each set of dynamic
tracing event state values. Each set of TBs can be used by any number of
vCPUs to maximize TB reuse when vCPUs have the same tracing state.
This feature is later used by tracetool to optimize tracing of guest
code events.
The maximum number of TB sets is defined as 2^E, where E is the number
of events that have the 'vcpu' property (their state is stored in
CPUState->trace_dstate).
For this to work, a change on the dynamic tracing state of a vCPU will
force it to flush its virtual TB cache (which is only indexed by
address), and fall back to the physical TB cache (which now contains the
vCPU's dynamic tracing state as part of the hashing function).
Signed-off-by: Lluís Vilanova <vilanova@ac.upc.edu>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-id: 149915775266.6295.10060144081246467690.stgit@frigg.lan
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
move cputlb.c, cpu-exec-common.c and cpu-exec.c related tcg exec
file into accel/tcg/ subdirectory.
Signed-off-by: Yang Zhong <yang.zhong@intel.com>
Message-Id: <1496383606-18060-3-git-send-email-yang.zhong@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
