Now that cpu_exit() actually kicks all accelerators, use it whenever
the message to another thread is processed in qemu_wait_io_event().
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(H)DEC and PURR get reset before icount does, which causes them to be
skewed and not match the init state. This can cause replay to not
match the recorded trace exactly. For DEC and HDEC this is usually not
noticable since they tend to get programmed before affecting the
target machine. PURR has been observed to cause replay bugs when
running Linux.
Fix this by resetting using a time of 0.
Message-ID: <20241219034035.1826173-2-npiggin@gmail.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Headers in include/sysemu/ are not only related to system
*emulation*, they are also used by virtualization. Rename
as system/ which is clearer.
Files renamed manually then mechanical change using sed tool.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Message-Id: <20241203172445.28576-1-philmd@linaro.org>
The BookE decrementer stops at 0, meaning that it won't decremented
towards "negative" values. However, the current logic is inverted: decr
is updated solely when the resulting value would be negative.
Signed-off-by: Clément Chigot <chigot@adacore.com>
Fixes: 8e0a5ac878 ("hw/ppc: Avoid decrementer rounding errors")
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
The H_GUEST_RUN_VCPU hcall is used to start execution of a Guest VCPU.
The Hypervisor will update the state of the Guest VCPU based on the
input buffer, restore the saved Guest VCPU state, and start its
execution.
The Guest VCPU can stop running for numerous reasons including HCALLs,
hypervisor exceptions, or an outstanding Host Partition Interrupt.
The reason that the Guest VCPU stopped running is communicated through
R4 and the output buffer will be filled in with any relevant state.
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
The Big QEMU Lock (BQL) has many names and they are confusing. The
actual QemuMutex variable is called qemu_global_mutex but it's commonly
referred to as the BQL in discussions and some code comments. The
locking APIs, however, are called qemu_mutex_lock_iothread() and
qemu_mutex_unlock_iothread().
The "iothread" name is historic and comes from when the main thread was
split into into KVM vcpu threads and the "iothread" (now called the main
loop thread). I have contributed to the confusion myself by introducing
a separate --object iothread, a separate concept unrelated to the BQL.
The "iothread" name is no longer appropriate for the BQL. Rename the
locking APIs to:
- void bql_lock(void)
- void bql_unlock(void)
- bool bql_locked(void)
There are more APIs with "iothread" in their names. Subsequent patches
will rename them. There are also comments and documentation that will be
updated in later patches.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Acked-by: Fabiano Rosas <farosas@suse.de>
Acked-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Acked-by: Peter Xu <peterx@redhat.com>
Acked-by: Eric Farman <farman@linux.ibm.com>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Acked-by: Hyman Huang <yong.huang@smartx.com>
Reviewed-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-id: 20240102153529.486531-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
It's unnecessary for non-KVM accelerators(TCG, for example),
to call this function, so change the order of kvm_enable() judgment.
The static inline function that returns -1 directly does not work
in TCG's situation.
Signed-off-by: jianchunfu <chunfu.jian@shingroup.cn>
Tested-by: Gautam Menghani <gautam@linux.ibm.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Timebase save uses a random number for a legacy vmstate field, which
makes rr snapshot loading unbalanced. The easiest way to deal with this
is just to skip the rng if record-replay is active.
Reviewed-by: Pavel Dovgalyuk <Pavel.Dovgalyuk@ispras.ru>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reading the time more than once to perform an operation always increases
complexity and fragility due to introduced deltas. Simplify the
decrementer write by reading the clock once for the operation.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Lower interrupts, delete timers, and set time facility registers
back to initial state on machine reset.
This is not so important for record-replay since timebase and
decrementer are migrated, but it gives a cleaner reset state.
Cc: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Cc: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[ clg: checkpatch.pl fixes ]
Signed-off-by: Cédric Le Goater <clg@kaod.org>
When writing a value to the decrementer that raises an exception, the
irq is raised, but the value is not stored so the store doesn't appear
to have changed the register when it is read again.
Always store the write value to the register.
Fixes: e81a982aa5 ("PPC: Clean up DECR implementation")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
When storing a large decrementer value with the most significant
implemented bit set, it is to be treated as a negative and sign
extended.
This isn't hit for book3s DEC because of another bug, fixing it
in the next patch exposes this one and can cause additional
problems, so fix this first. It can be hit with HDECR and other
edge triggered types.
Fixes: a8dafa5251 ("target/ppc: Implement large decrementer support for TCG")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
[ clg: removed extra cpu and pcc variables shadowing local variables ]
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The decrementer register contains a relative time in timebase units.
When writing to DECR this is converted and stored as an absolute value
in nanosecond units, reading DECR converts back to relative timebase.
The tb<->ns conversion of the relative part can cause rounding such that
a value writen to the decrementer can read back a different, with time
held constant. This is a particular problem for a deterministic icount
and record-replay trace.
Fix this by storing the absolute value in timebase units rather than
nanoseconds. The math before:
store: decr_next = now_ns + decr * ns_per_sec / tb_per_sec
load: decr = (decr_next - now_ns) * tb_per_sec / ns_per_sec
load(store): decr = decr * ns_per_sec / tb_per_sec * tb_per_sec /
ns_per_sec
After:
store: decr_next = now_ns * tb_per_sec / ns_per_sec + decr
load: decr = decr_next - now_ns * tb_per_sec / ns_per_sec
load(store): decr = decr
Fixes: 9fddaa0c0c ("PowerPC merge: real time TB and decrementer - faster and simpler exception handling (Jocelyn Mayer)")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The rule of timers is typically that they should never expire before the
timeout, but some time afterward. Rounding timer intervals up when doing
conversion is the right thing to do.
Under most circumstances it is impossible observe the decrementer
interrupt before the dec register has triggered. However with icount
timing, problems can arise. For example setting DEC to 0 can schedule
the timer for now, causing it to fire before any more instructions
have been executed and DEC is still 0.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
These calculations are repeated several times, and they will become
a little more complicated with subsequent changes.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The clock update logic reads the clock twice to compute the new clock
value, with a value derived from the later time subtracted from a value
derived from the earlier time. The delta causes time to be lost.
This can ultimately result in time becoming unsynchronized between CPUs
and that can cause OS lockups, timeouts, watchdogs, etc. This can be
seen running a KVM guest (that causes lots of TB updates) on a powernv
SMP machine.
Fix this by reading the clock once.
Cc: qemu-stable@nongnu.org
Fixes: dbdd25065e ("Implement time-base start/stop helpers.")
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Frederic Barrat <fbarrat@linux.ibm.com>
Message-ID: <20230629020713.327745-1-npiggin@gmail.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
HDEC interrupts are edge-triggered on HDECR underflow (notably different
from DEC which is level-triggered).
HDEC interrupts already clear the irq on delivery so that does not need
to be changed.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-ID: <20230625122045.15544-1-npiggin@gmail.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Doorbells in SMT need to coordinate msgsnd/msgclr and DPDES access from
multiple threads that affect the same state.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The decrementer store function has logic that short-cuts the timer if a
very small value is stored (0, 1, or 2) and raises an interrupt
directly. There are two problem with this on BookE.
First is that BookE says a decrementer interrupt should not be raised
on a store of 0, only of a decrement from 1. Second is that raising
the irq directly will bypass the auto-reload logic in the booke decr
timer function, breaking autoreload when 1 or 2 is stored.
Fix this by removing that small-value special case. It makes this
tricky logic even more difficult to reason about, and it hardly matters
for performance.
Cc: sdicaro@DDCI.com
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20230530131214.373524-2-npiggin@gmail.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
