2019-05-29 07:12:40 -07:00
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-09-25 20:31:56 +00:00
|
|
|
|
2010-01-15 14:49:12 +01:00
|
|
|
|
|
|
|
|
|
2010-04-16 00:11:40 +02:00
|
|
|
|
|
|
|
|
|
2010-01-15 14:49:12 +01:00
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S: Allow XICS emulation to work in nested hosts using XIVE
Currently, the KVM code assumes that if the host kernel is using the
XIVE interrupt controller (the new interrupt controller that first
appeared in POWER9 systems), then the in-kernel XICS emulation will
use the XIVE hardware to deliver interrupts to the guest. However,
this only works when the host is running in hypervisor mode and has
full access to all of the XIVE functionality. It doesn't work in any
nested virtualization scenario, either with PR KVM or nested-HV KVM,
because the XICS-on-XIVE code calls directly into the native-XIVE
routines, which are not initialized and cannot function correctly
because they use OPAL calls, and OPAL is not available in a guest.
This means that using the in-kernel XICS emulation in a nested
hypervisor that is using XIVE as its interrupt controller will cause a
(nested) host kernel crash. To fix this, we change most of the places
where the current code calls xive_enabled() to select between the
XICS-on-XIVE emulation and the plain XICS emulation to call a new
function, xics_on_xive(), which returns false in a guest.
However, there is a further twist. The plain XICS emulation has some
functions which are used in real mode and access the underlying XICS
controller (the interrupt controller of the host) directly. In the
case of a nested hypervisor, this means doing XICS hypercalls
directly. When the nested host is using XIVE as its interrupt
controller, these hypercalls will fail. Therefore this also adds
checks in the places where the XICS emulation wants to access the
underlying interrupt controller directly, and if that is XIVE, makes
the code use the virtual mode fallback paths, which call generic
kernel infrastructure rather than doing direct XICS access.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2019-02-04 22:07:20 +11:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
2023-03-08 17:34:48 +11:00
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
2014-09-09 22:37:35 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-02-19 11:00:31 +01:00
|
|
|
|
2013-04-08 00:32:13 +00:00
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
|
|
|
|
|
2018-05-21 13:24:21 +08:00
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-20 13:52:36 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2020-06-23 21:14:16 +08:00
|
|
|
|
2009-10-30 05:47:07 +00:00
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
|
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
2014-01-09 11:51:16 +01:00
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2010-02-19 11:00:30 +01:00
|
|
|
|
2014-01-09 11:51:16 +01:00
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
KVM: PPC: Book3S: Add MMIO emulation for FP and VSX instructions
This patch provides the MMIO load/store emulation for instructions
of 'double & vector unsigned char & vector signed char & vector
unsigned short & vector signed short & vector unsigned int & vector
signed int & vector double '.
The instructions that this adds emulation for are:
- ldx, ldux, lwax,
- lfs, lfsx, lfsu, lfsux, lfd, lfdx, lfdu, lfdux,
- stfs, stfsx, stfsu, stfsux, stfd, stfdx, stfdu, stfdux, stfiwx,
- lxsdx, lxsspx, lxsiwax, lxsiwzx, lxvd2x, lxvw4x, lxvdsx,
- stxsdx, stxsspx, stxsiwx, stxvd2x, stxvw4x
[paulus@ozlabs.org - some cleanups, fixes and rework, make it
compile for Book E, fix build when PR KVM is built in]
Signed-off-by: Bin Lu <lblulb@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2017-02-21 21:12:36 +08:00
|
|
|
|
|
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2018-05-21 13:24:26 +08:00
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2018-05-21 13:24:26 +08:00
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2014-01-09 11:51:16 +01:00
|
|
|
|
|
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
KVM: PPC: Book3S: Add MMIO emulation for FP and VSX instructions
This patch provides the MMIO load/store emulation for instructions
of 'double & vector unsigned char & vector signed char & vector
unsigned short & vector signed short & vector unsigned int & vector
signed int & vector double '.
The instructions that this adds emulation for are:
- ldx, ldux, lwax,
- lfs, lfsx, lfsu, lfsux, lfd, lfdx, lfdu, lfdux,
- stfs, stfsx, stfsu, stfsux, stfd, stfdx, stfdu, stfdux, stfiwx,
- lxsdx, lxsspx, lxsiwax, lxsiwzx, lxvd2x, lxvw4x, lxvdsx,
- stxsdx, stxsspx, stxsiwx, stxvd2x, stxvw4x
[paulus@ozlabs.org - some cleanups, fixes and rework, make it
compile for Book E, fix build when PR KVM is built in]
Signed-off-by: Bin Lu <lblulb@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2017-02-21 21:12:36 +08:00
|
|
|
|
|
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
KVM: PPC: Fetch prefixed instructions from the guest
In order to handle emulation of prefixed instructions in the guest,
this first makes vcpu->arch.last_inst be an unsigned long, i.e. 64
bits on 64-bit platforms. For prefixed instructions, the upper 32
bits are used for the prefix and the lower 32 bits for the suffix, and
both halves are byte-swapped if the guest endianness differs from the
host.
Next, vcpu->arch.emul_inst is now 64 bits wide, to match the HEIR
register on POWER10. Like HEIR, for a prefixed instruction it is
defined to have the prefix is in the top 32 bits and the suffix in the
bottom 32 bits, with both halves in the correct byte order.
kvmppc_get_last_inst is extended on 64-bit machines to put the prefix
and suffix in the right places in the ppc_inst_t being returned.
kvmppc_load_last_inst now returns the instruction in an unsigned long
in the same format as vcpu->arch.last_inst. It makes the decision
about whether to fetch a suffix based on the SRR1_PREFIXED bit in the
MSR image stored in the vcpu struct, which generally comes from SRR1
or HSRR1 on an interrupt. This bit is defined in Power ISA v3.1B to
be set if the interrupt occurred due to a prefixed instruction and
cleared otherwise for all interrupts except for instruction storage
interrupt, which does not come to the hypervisor. It is set to zero
for asynchronous interrupts such as external interrupts. In previous
ISA versions it was always set to 0 for all interrupts except
instruction storage interrupt.
The code in book3s_hv_rmhandlers.S that loads the faulting instruction
on a HDSI is only used on POWER8 and therefore doesn't ever need to
load a suffix.
[npiggin@gmail.com - check that the is-prefixed bit in SRR1 matches the
type of instruction that was fetched.]
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Tested-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/ZAgsq9h1CCzouQuV@cleo
2023-03-08 17:35:23 +11:00
|
|
|
|
|
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
2014-06-20 13:58:16 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2014-06-18 14:53:49 +02:00
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2008-11-05 09:36:16 -06:00
|
|
|
|
2011-04-27 17:24:21 -05:00
|
|
|
|
2014-09-01 17:19:56 +03:00
|
|
|
|
2011-08-10 13:57:08 +02:00
|
|
|
|
2012-08-08 20:38:19 +00:00
|
|
|
|
|
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|
2009-01-03 16:22:59 -06:00
|
|
|
|
|
|
|
|
|
2008-12-02 15:51:53 -06:00
|
|
|
|
2008-12-02 15:51:55 -06:00
|
|
|
|
2008-07-25 13:54:53 -05:00
|
|
|
|
2009-01-03 16:23:03 -06:00
|
|
|
|
|
|
|
|
|
2009-01-03 16:23:02 -06:00
|
|
|
|
|
|
|
|
|
2009-01-03 16:23:11 -06:00
|
|
|
|
|
|
|
|
|
2014-06-20 13:52:36 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-11-05 09:36:14 -06:00
|
|
|
|
2019-12-18 13:55:00 -08:00
|
|
|
|
2008-11-05 09:36:18 -06:00
|
|
|
|
2008-11-05 09:36:17 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-11-05 09:36:14 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-02-16 14:07:37 +00:00
|
|
|
|
2008-11-05 09:36:14 -06:00
|
|
|
|
2023-03-30 20:32:23 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV P9: Stop handling hcalls in real-mode in the P9 path
In the interest of minimising the amount of code that is run in
"real-mode", don't handle hcalls in real mode in the P9 path. This
requires some new handlers for H_CEDE and xics-on-xive to be added
before xive is pulled or cede logic is checked.
This introduces a change in radix guest behaviour where radix guests
that execute 'sc 1' in userspace now get a privilege fault whereas
previously the 'sc 1' would be reflected as a syscall interrupt to the
guest kernel. That reflection is only required for hash guests that run
PR KVM.
Background:
In POWER8 and earlier processors, it is very expensive to exit from the
HV real mode context of a guest hypervisor interrupt, and switch to host
virtual mode. On those processors, guest->HV interrupts reach the
hypervisor with the MMU off because the MMU is loaded with guest context
(LPCR, SDR1, SLB), and the other threads in the sub-core need to be
pulled out of the guest too. Then the primary must save off guest state,
invalidate SLB and ERAT, and load up host state before the MMU can be
enabled to run in host virtual mode (~= regular Linux mode).
Hash guests also require a lot of hcalls to run due to the nature of the
MMU architecture and paravirtualisation design. The XICS interrupt
controller requires hcalls to run.
So KVM traditionally tries hard to avoid the full exit, by handling
hcalls and other interrupts in real mode as much as possible.
By contrast, POWER9 has independent MMU context per-thread, and in radix
mode the hypervisor is in host virtual memory mode when the HV interrupt
is taken. Radix guests do not require significant hcalls to manage their
translations, and xive guests don't need hcalls to handle interrupts. So
it's much less important for performance to handle hcalls in real mode on
POWER9.
One caveat is that the TCE hcalls are performance critical, real-mode
variants introduced for POWER8 in order to achieve 10GbE performance.
Real mode TCE hcalls were found to be less important on POWER9, which
was able to drive 40GBe networking without them (using the virt mode
hcalls) but performance is still important. These hcalls will benefit
from subsequent guest entry/exit optimisation including possibly a
faster "partial exit" that does not entirely switch to host context to
handle the hcall.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210528090752.3542186-14-npiggin@gmail.com
2021-05-28 19:07:33 +10:00
|
|
|
|
2023-03-30 20:32:23 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2008-11-05 09:36:14 -06:00
|
|
|
|
2009-12-21 20:21:24 +01:00
|
|
|
|
2008-11-05 09:36:14 -06:00
|
|
|
|
|
|
|
|
|
2013-02-14 14:00:25 +00:00
|
|
|
|
2023-03-30 20:32:23 +10:00
|
|
|
|
|
|
|
|
|
2014-06-18 21:56:55 +02:00
|
|
|
|
|
|
|
|
|
2023-03-30 20:32:23 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-06-18 21:56:55 +02:00
|
|
|
|
|
|
|
|
|
2023-03-30 20:32:23 +10:00
|
|
|
|
|
|
|
|
|
2012-07-31 00:19:50 +02:00
|
|
|
|
2012-08-13 12:50:35 +02:00
|
|
|
|
2008-11-05 09:36:16 -06:00
|
|
|
|
2008-11-05 09:36:18 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2010-07-29 14:47:48 +02:00
|
|
|
|
2011-06-14 18:34:41 -05:00
|
|
|
|
2008-11-24 11:37:38 -06:00
|
|
|
|
KVM: PPC: Book3S HV: Split HPT allocation from activation
Currently, kvmppc_alloc_hpt() both allocates a new hashed page table (HPT)
and sets it up as the active page table for a VM. For the upcoming HPT
resize implementation we're going to want to allocate HPTs separately from
activating them.
So, split the allocation itself out into kvmppc_allocate_hpt() and perform
the activation with a new kvmppc_set_hpt() function. Likewise we split
kvmppc_free_hpt(), which just frees the HPT, from kvmppc_release_hpt()
which unsets it as an active HPT, then frees it.
We also move the logic to fall back to smaller HPT sizes if the first try
fails into the single caller which used that behaviour,
kvmppc_hv_setup_htab_rma(). This introduces a slight semantic change, in
that previously if the initial attempt at CMA allocation failed, we would
fall back to attempting smaller sizes with the page allocator. Now, we
try first CMA, then the page allocator at each size. As far as I can tell
this change should be harmless.
To match, we make kvmppc_free_hpt() just free the actual HPT itself. The
call to kvmppc_free_lpid() that was there, we move to the single caller.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2016-12-20 16:49:02 +11:00
|
|
|
|
|
|
|
|
|
2023-02-08 15:01:00 +01:00
|
|
|
|
KVM: PPC: Book3S HV: Split HPT allocation from activation
Currently, kvmppc_alloc_hpt() both allocates a new hashed page table (HPT)
and sets it up as the active page table for a VM. For the upcoming HPT
resize implementation we're going to want to allocate HPTs separately from
activating them.
So, split the allocation itself out into kvmppc_allocate_hpt() and perform
the activation with a new kvmppc_set_hpt() function. Likewise we split
kvmppc_free_hpt(), which just frees the HPT, from kvmppc_release_hpt()
which unsets it as an active HPT, then frees it.
We also move the logic to fall back to smaller HPT sizes if the first try
fails into the single caller which used that behaviour,
kvmppc_hv_setup_htab_rma(). This introduces a slight semantic change, in
that previously if the initial attempt at CMA allocation failed, we would
fall back to attempting smaller sizes with the page allocator. Now, we
try first CMA, then the page allocator at each size. As far as I can tell
this change should be harmless.
To match, we make kvmppc_free_hpt() just free the actual HPT itself. The
call to kvmppc_free_lpid() that was there, we move to the single caller.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2016-12-20 16:49:02 +11:00
|
|
|
|
2017-09-13 16:00:10 +10:00
|
|
|
|
KVM: PPC: Only get pages when actually needed, not in prepare_memory_region()
This removes the code from kvmppc_core_prepare_memory_region() that
looked up the VMA for the region being added and called hva_to_page
to get the pfns for the memory. We have no guarantee that there will
be anything mapped there at the time of the KVM_SET_USER_MEMORY_REGION
ioctl call; userspace can do that ioctl and then map memory into the
region later.
Instead we defer looking up the pfn for each memory page until it is
needed, which generally means when the guest does an H_ENTER hcall on
the page. Since we can't call get_user_pages in real mode, if we don't
already have the pfn for the page, kvmppc_h_enter() will return
H_TOO_HARD and we then call kvmppc_virtmode_h_enter() once we get back
to kernel context. That calls kvmppc_get_guest_page() to get the pfn
for the page, and then calls back to kvmppc_h_enter() to redo the HPTE
insertion.
When the first vcpu starts executing, we need to have the RMO or VRMA
region mapped so that the guest's real mode accesses will work. Thus
we now have a check in kvmppc_vcpu_run() to see if the RMO/VRMA is set
up and if not, call kvmppc_hv_setup_rma(). It checks if the memslot
starting at guest physical 0 now has RMO memory mapped there; if so it
sets it up for the guest, otherwise on POWER7 it sets up the VRMA.
The function that does that, kvmppc_map_vrma, is now a bit simpler,
as it calls kvmppc_virtmode_h_enter instead of creating the HPTE itself.
Since we are now potentially updating entries in the slot_phys[]
arrays from multiple vcpu threads, we now have a spinlock protecting
those updates to ensure that we don't lose track of any references
to pages.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Avi Kivity <avi@redhat.com>
2011-12-12 12:31:00 +00:00
|
|
|
|
2011-12-12 12:31:41 +00:00
|
|
|
|
2011-06-29 00:22:05 +00:00
|
|
|
|
2017-03-22 15:21:56 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-09-13 16:00:10 +10:00
|
|
|
|
|
|
|
|
|
2017-11-22 14:38:53 +11:00
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
2023-02-08 15:01:00 +01:00
|
|
|
|
2016-03-01 17:54:40 +11:00
|
|
|
|
2017-03-22 15:21:55 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-03-15 21:58:34 +00:00
|
|
|
|
|
|
|
|
|
2016-02-15 12:55:09 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-02-21 16:31:10 +01:00
|
|
|
|
|
|
|
|
|
2016-12-20 16:48:59 +11:00
|
|
|
|
|
|
|
|
|
2011-06-29 00:19:22 +00:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:18:00 +05:30
|
|
|
|
2020-02-18 13:07:27 -08:00
|
|
|
|
2011-06-29 00:19:22 +00:00
|
|
|
|
2021-12-06 20:54:11 +01:00
|
|
|
|
|
|
|
|
|
2020-02-18 13:07:18 -08:00
|
|
|
|
2011-06-29 00:19:22 +00:00
|
|
|
|
2021-12-06 20:54:11 +01:00
|
|
|
|
2018-12-12 15:15:30 +11:00
|
|
|
|
|
|
|
|
|
2012-04-26 19:43:42 +00:00
|
|
|
|
|
|
|
|
|
2012-09-11 13:28:18 +00:00
|
|
|
|
|
|
|
|
|
2011-06-29 00:19:22 +00:00
|
|
|
|
2011-12-20 15:34:43 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-08-10 12:28:50 +02:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV: Provide a method for userspace to read and write the HPT
A new ioctl, KVM_PPC_GET_HTAB_FD, returns a file descriptor. Reads on
this fd return the contents of the HPT (hashed page table), writes
create and/or remove entries in the HPT. There is a new capability,
KVM_CAP_PPC_HTAB_FD, to indicate the presence of the ioctl. The ioctl
takes an argument structure with the index of the first HPT entry to
read out and a set of flags. The flags indicate whether the user is
intending to read or write the HPT, and whether to return all entries
or only the "bolted" entries (those with the bolted bit, 0x10, set in
the first doubleword).
This is intended for use in implementing qemu's savevm/loadvm and for
live migration. Therefore, on reads, the first pass returns information
about all HPTEs (or all bolted HPTEs). When the first pass reaches the
end of the HPT, it returns from the read. Subsequent reads only return
information about HPTEs that have changed since they were last read.
A read that finds no changed HPTEs in the HPT following where the last
read finished will return 0 bytes.
The format of the data provides a simple run-length compression of the
invalid entries. Each block of data starts with a header that indicates
the index (position in the HPT, which is just an array), the number of
valid entries starting at that index (may be zero), and the number of
invalid entries following those valid entries. The valid entries, 16
bytes each, follow the header. The invalid entries are not explicitly
represented.
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix documentation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-11-19 22:57:20 +00:00
|
|
|
|
2023-02-08 15:01:00 +01:00
|
|
|
|
2016-12-20 16:49:05 +11:00
|
|
|
|
2023-02-08 15:01:00 +01:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV: Provide a method for userspace to read and write the HPT
A new ioctl, KVM_PPC_GET_HTAB_FD, returns a file descriptor. Reads on
this fd return the contents of the HPT (hashed page table), writes
create and/or remove entries in the HPT. There is a new capability,
KVM_CAP_PPC_HTAB_FD, to indicate the presence of the ioctl. The ioctl
takes an argument structure with the index of the first HPT entry to
read out and a set of flags. The flags indicate whether the user is
intending to read or write the HPT, and whether to return all entries
or only the "bolted" entries (those with the bolted bit, 0x10, set in
the first doubleword).
This is intended for use in implementing qemu's savevm/loadvm and for
live migration. Therefore, on reads, the first pass returns information
about all HPTEs (or all bolted HPTEs). When the first pass reaches the
end of the HPT, it returns from the read. Subsequent reads only return
information about HPTEs that have changed since they were last read.
A read that finds no changed HPTEs in the HPT following where the last
read finished will return 0 bytes.
The format of the data provides a simple run-length compression of the
invalid entries. Each block of data starts with a header that indicates
the index (position in the HPT, which is just an array), the number of
valid entries starting at that index (may be zero), and the number of
invalid entries following those valid entries. The valid entries, 16
bytes each, follow the header. The invalid entries are not explicitly
represented.
Signed-off-by: Paul Mackerras <paulus@samba.org>
[agraf: fix documentation]
Signed-off-by: Alexander Graf <agraf@suse.de>
2012-11-19 22:57:20 +00:00
|
|
|
|
2013-04-12 14:08:46 +00:00
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:00 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:32:04 +00:00
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:00 +00:00
|
|
|
|
2014-08-13 14:39:44 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S: Add MMIO emulation for FP and VSX instructions
This patch provides the MMIO load/store emulation for instructions
of 'double & vector unsigned char & vector signed char & vector
unsigned short & vector signed short & vector unsigned int & vector
signed int & vector double '.
The instructions that this adds emulation for are:
- ldx, ldux, lwax,
- lfs, lfsx, lfsu, lfsux, lfd, lfdx, lfdu, lfdux,
- stfs, stfsx, stfsu, stfsux, stfd, stfdx, stfdu, stfdux, stfiwx,
- lxsdx, lxsspx, lxsiwax, lxsiwzx, lxvd2x, lxvw4x, lxvdsx,
- stxsdx, stxsspx, stxsiwx, stxvd2x, stxvw4x
[paulus@ozlabs.org - some cleanups, fixes and rework, make it
compile for Book E, fix build when PR KVM is built in]
Signed-off-by: Bin Lu <lblulb@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2017-02-21 21:12:36 +08:00
|
|
|
|
2018-05-21 13:24:26 +08:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-04-18 12:39:35 +02:00
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-10-07 22:18:01 +05:30
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-10-02 16:00:22 +10:00
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2019-12-18 13:55:00 -08:00
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-12-06 20:54:11 +01:00
|
|
|
|
|
|
|
|
|
2020-02-18 13:07:18 -08:00
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
2021-12-06 20:54:11 +01:00
|
|
|
|
2018-12-12 15:15:30 +11:00
|
|
|
|
|
|
|
|
|
2021-04-01 17:56:53 -07:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-02-18 13:07:27 -08:00
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-04-27 12:35:11 +08:00
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-02-08 15:01:00 +01:00
|
|
|
|
|
|
|
|
|
2014-06-02 11:03:00 +10:00
|
|
|
|
2016-08-19 15:35:47 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-01-30 21:21:41 +11:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV: Allow userspace to set the desired SMT mode
This allows userspace to set the desired virtual SMT (simultaneous
multithreading) mode for a VM, that is, the number of VCPUs that
get assigned to each virtual core. Previously, the virtual SMT mode
was fixed to the number of threads per subcore, and if userspace
wanted to have fewer vcpus per vcore, then it would achieve that by
using a sparse CPU numbering. This had the disadvantage that the
vcpu numbers can get quite large, particularly for SMT1 guests on
a POWER8 with 8 threads per core. With this patch, userspace can
set its desired virtual SMT mode and then use contiguous vcpu
numbering.
On POWER8, where the threading mode is "strict", the virtual SMT mode
must be less than or equal to the number of threads per subcore. On
POWER9, which implements a "loose" threading mode, the virtual SMT
mode can be any power of 2 between 1 and 8, even though there is
effectively one thread per subcore, since the threads are independent
and can all be in different partitions.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2017-02-06 13:24:41 +11:00
|
|
|
|
|
|
|
|
|
2018-05-21 13:24:22 +08:00
|
|
|
|
2018-09-21 20:02:01 +10:00
|
|
|
|
2018-12-14 16:29:06 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2020-03-19 15:29:55 +11:00
|
|
|
|
2019-11-25 08:36:30 +05:30
|
|
|
|
2020-12-16 16:12:19 +05:30
|
|
|
|
2021-02-05 13:41:54 -03:00
|
|
|
|
2022-01-11 11:54:04 +11:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
2013-10-07 22:18:01 +05:30
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
2023-03-08 17:34:48 +11:00
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Fetch prefixed instructions from the guest
In order to handle emulation of prefixed instructions in the guest,
this first makes vcpu->arch.last_inst be an unsigned long, i.e. 64
bits on 64-bit platforms. For prefixed instructions, the upper 32
bits are used for the prefix and the lower 32 bits for the suffix, and
both halves are byte-swapped if the guest endianness differs from the
host.
Next, vcpu->arch.emul_inst is now 64 bits wide, to match the HEIR
register on POWER10. Like HEIR, for a prefixed instruction it is
defined to have the prefix is in the top 32 bits and the suffix in the
bottom 32 bits, with both halves in the correct byte order.
kvmppc_get_last_inst is extended on 64-bit machines to put the prefix
and suffix in the right places in the ppc_inst_t being returned.
kvmppc_load_last_inst now returns the instruction in an unsigned long
in the same format as vcpu->arch.last_inst. It makes the decision
about whether to fetch a suffix based on the SRR1_PREFIXED bit in the
MSR image stored in the vcpu struct, which generally comes from SRR1
or HSRR1 on an interrupt. This bit is defined in Power ISA v3.1B to
be set if the interrupt occurred due to a prefixed instruction and
cleared otherwise for all interrupts except for instruction storage
interrupt, which does not come to the hypervisor. It is set to zero
for asynchronous interrupts such as external interrupts. In previous
ISA versions it was always set to 0 for all interrupts except
instruction storage interrupt.
The code in book3s_hv_rmhandlers.S that loads the faulting instruction
on a HDSI is only used on POWER8 and therefore doesn't ever need to
load a suffix.
[npiggin@gmail.com - check that the is-prefixed bit in SRR1 matches the
type of instruction that was fetched.]
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Tested-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/ZAgsq9h1CCzouQuV@cleo
2023-03-08 17:35:23 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
KVM: PPC: Fetch prefixed instructions from the guest
In order to handle emulation of prefixed instructions in the guest,
this first makes vcpu->arch.last_inst be an unsigned long, i.e. 64
bits on 64-bit platforms. For prefixed instructions, the upper 32
bits are used for the prefix and the lower 32 bits for the suffix, and
both halves are byte-swapped if the guest endianness differs from the
host.
Next, vcpu->arch.emul_inst is now 64 bits wide, to match the HEIR
register on POWER10. Like HEIR, for a prefixed instruction it is
defined to have the prefix is in the top 32 bits and the suffix in the
bottom 32 bits, with both halves in the correct byte order.
kvmppc_get_last_inst is extended on 64-bit machines to put the prefix
and suffix in the right places in the ppc_inst_t being returned.
kvmppc_load_last_inst now returns the instruction in an unsigned long
in the same format as vcpu->arch.last_inst. It makes the decision
about whether to fetch a suffix based on the SRR1_PREFIXED bit in the
MSR image stored in the vcpu struct, which generally comes from SRR1
or HSRR1 on an interrupt. This bit is defined in Power ISA v3.1B to
be set if the interrupt occurred due to a prefixed instruction and
cleared otherwise for all interrupts except for instruction storage
interrupt, which does not come to the hypervisor. It is set to zero
for asynchronous interrupts such as external interrupts. In previous
ISA versions it was always set to 0 for all interrupts except
instruction storage interrupt.
The code in book3s_hv_rmhandlers.S that loads the faulting instruction
on a HDSI is only used on POWER8 and therefore doesn't ever need to
load a suffix.
[npiggin@gmail.com - check that the is-prefixed bit in SRR1 matches the
type of instruction that was fetched.]
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Tested-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/ZAgsq9h1CCzouQuV@cleo
2023-03-08 17:35:23 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-03-08 17:34:48 +11:00
|
|
|
|
KVM: PPC: Fetch prefixed instructions from the guest
In order to handle emulation of prefixed instructions in the guest,
this first makes vcpu->arch.last_inst be an unsigned long, i.e. 64
bits on 64-bit platforms. For prefixed instructions, the upper 32
bits are used for the prefix and the lower 32 bits for the suffix, and
both halves are byte-swapped if the guest endianness differs from the
host.
Next, vcpu->arch.emul_inst is now 64 bits wide, to match the HEIR
register on POWER10. Like HEIR, for a prefixed instruction it is
defined to have the prefix is in the top 32 bits and the suffix in the
bottom 32 bits, with both halves in the correct byte order.
kvmppc_get_last_inst is extended on 64-bit machines to put the prefix
and suffix in the right places in the ppc_inst_t being returned.
kvmppc_load_last_inst now returns the instruction in an unsigned long
in the same format as vcpu->arch.last_inst. It makes the decision
about whether to fetch a suffix based on the SRR1_PREFIXED bit in the
MSR image stored in the vcpu struct, which generally comes from SRR1
or HSRR1 on an interrupt. This bit is defined in Power ISA v3.1B to
be set if the interrupt occurred due to a prefixed instruction and
cleared otherwise for all interrupts except for instruction storage
interrupt, which does not come to the hypervisor. It is set to zero
for asynchronous interrupts such as external interrupts. In previous
ISA versions it was always set to 0 for all interrupts except
instruction storage interrupt.
The code in book3s_hv_rmhandlers.S that loads the faulting instruction
on a HDSI is only used on POWER8 and therefore doesn't ever need to
load a suffix.
[npiggin@gmail.com - check that the is-prefixed bit in SRR1 matches the
type of instruction that was fetched.]
Reviewed-by: Nicholas Piggin <npiggin@gmail.com>
Tested-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/ZAgsq9h1CCzouQuV@cleo
2023-03-08 17:35:23 +11:00
|
|
|
|
2014-07-23 19:06:21 +03:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:18:02 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-03-20 20:39:41 +11:00
|
|
|
|
|
|
|
|
|
2010-02-19 11:00:42 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-09-25 20:31:56 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
2011-04-27 17:24:21 -05:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
2011-04-27 17:24:21 -05:00
|
|
|
|
|
|
|
|
|
2011-12-12 12:26:50 +00:00
|
|
|
|
|
|
|
|
|
2012-09-25 20:31:56 +00:00
|
|
|
|
|
|
|
|
|
2011-12-12 12:26:50 +00:00
|
|
|
|
2011-04-27 17:24:21 -05:00
|
|
|
|
|
|
|
|
|
2013-04-12 14:08:46 +00:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:52 +05:30
|
|
|
|
2013-07-02 11:15:16 +05:30
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
|
|
|
|
|
2018-02-14 01:08:12 +10:00
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
KVM: PPC: Allocate RMAs (Real Mode Areas) at boot for use by guests
This adds infrastructure which will be needed to allow book3s_hv KVM to
run on older POWER processors, including PPC970, which don't support
the Virtual Real Mode Area (VRMA) facility, but only the Real Mode
Offset (RMO) facility. These processors require a physically
contiguous, aligned area of memory for each guest. When the guest does
an access in real mode (MMU off), the address is compared against a
limit value, and if it is lower, the address is ORed with an offset
value (from the Real Mode Offset Register (RMOR)) and the result becomes
the real address for the access. The size of the RMA has to be one of
a set of supported values, which usually includes 64MB, 128MB, 256MB
and some larger powers of 2.
Since we are unlikely to be able to allocate 64MB or more of physically
contiguous memory after the kernel has been running for a while, we
allocate a pool of RMAs at boot time using the bootmem allocator. The
size and number of the RMAs can be set using the kvm_rma_size=xx and
kvm_rma_count=xx kernel command line options.
KVM exports a new capability, KVM_CAP_PPC_RMA, to signal the availability
of the pool of preallocated RMAs. The capability value is 1 if the
processor can use an RMA but doesn't require one (because it supports
the VRMA facility), or 2 if the processor requires an RMA for each guest.
This adds a new ioctl, KVM_ALLOCATE_RMA, which allocates an RMA from the
pool and returns a file descriptor which can be used to map the RMA. It
also returns the size of the RMA in the argument structure.
Having an RMA means we will get multiple KMV_SET_USER_MEMORY_REGION
ioctl calls from userspace. To cope with this, we now preallocate the
kvm->arch.ram_pginfo array when the VM is created with a size sufficient
for up to 64GB of guest memory. Subsequently we will get rid of this
array and use memory associated with each memslot instead.
This moves most of the code that translates the user addresses into
host pfns (page frame numbers) out of kvmppc_prepare_vrma up one level
to kvmppc_core_prepare_memory_region. Also, instead of having to look
up the VMA for each page in order to check the page size, we now check
that the pages we get are compound pages of 16MB. However, if we are
adding memory that is mapped to an RMA, we don't bother with calling
get_user_pages_fast and instead just offset from the base pfn for the
RMA.
Typically the RMA gets added after vcpus are created, which makes it
inconvenient to have the LPCR (logical partition control register) value
in the vcpu->arch struct, since the LPCR controls whether the processor
uses RMA or VRMA for the guest. This moves the LPCR value into the
kvm->arch struct and arranges for the MER (mediated external request)
bit, which is the only bit that varies between vcpus, to be set in
assembly code when going into the guest if there is a pending external
interrupt request.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:25:44 +00:00
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-02-14 01:08:12 +10:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
KVM: PPC: Allocate RMAs (Real Mode Areas) at boot for use by guests
This adds infrastructure which will be needed to allow book3s_hv KVM to
run on older POWER processors, including PPC970, which don't support
the Virtual Real Mode Area (VRMA) facility, but only the Real Mode
Offset (RMO) facility. These processors require a physically
contiguous, aligned area of memory for each guest. When the guest does
an access in real mode (MMU off), the address is compared against a
limit value, and if it is lower, the address is ORed with an offset
value (from the Real Mode Offset Register (RMOR)) and the result becomes
the real address for the access. The size of the RMA has to be one of
a set of supported values, which usually includes 64MB, 128MB, 256MB
and some larger powers of 2.
Since we are unlikely to be able to allocate 64MB or more of physically
contiguous memory after the kernel has been running for a while, we
allocate a pool of RMAs at boot time using the bootmem allocator. The
size and number of the RMAs can be set using the kvm_rma_size=xx and
kvm_rma_count=xx kernel command line options.
KVM exports a new capability, KVM_CAP_PPC_RMA, to signal the availability
of the pool of preallocated RMAs. The capability value is 1 if the
processor can use an RMA but doesn't require one (because it supports
the VRMA facility), or 2 if the processor requires an RMA for each guest.
This adds a new ioctl, KVM_ALLOCATE_RMA, which allocates an RMA from the
pool and returns a file descriptor which can be used to map the RMA. It
also returns the size of the RMA in the argument structure.
Having an RMA means we will get multiple KMV_SET_USER_MEMORY_REGION
ioctl calls from userspace. To cope with this, we now preallocate the
kvm->arch.ram_pginfo array when the VM is created with a size sufficient
for up to 64GB of guest memory. Subsequently we will get rid of this
array and use memory associated with each memslot instead.
This moves most of the code that translates the user addresses into
host pfns (page frame numbers) out of kvmppc_prepare_vrma up one level
to kvmppc_core_prepare_memory_region. Also, instead of having to look
up the VMA for each page in order to check the page size, we now check
that the pages we get are compound pages of 16MB. However, if we are
adding memory that is mapped to an RMA, we don't bother with calling
get_user_pages_fast and instead just offset from the base pfn for the
RMA.
Typically the RMA gets added after vcpus are created, which makes it
inconvenient to have the LPCR (logical partition control register) value
in the vcpu->arch struct, since the LPCR controls whether the processor
uses RMA or VRMA for the guest. This moves the LPCR value into the
kvm->arch struct and arranges for the MER (mediated external request)
bit, which is the only bit that varies between vcpus, to be set in
assembly code when going into the guest if there is a pending external
interrupt request.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:25:44 +00:00
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:56 +05:30
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
2013-10-07 22:17:56 +05:30
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV: use smp_mb() when setting/clearing host_ipi flag
On a 2-socket Power9 system with 32 cores/128 threads (SMT4) and 1TB
of memory running the following guest configs:
guest A:
- 224GB of memory
- 56 VCPUs (sockets=1,cores=28,threads=2), where:
VCPUs 0-1 are pinned to CPUs 0-3,
VCPUs 2-3 are pinned to CPUs 4-7,
...
VCPUs 54-55 are pinned to CPUs 108-111
guest B:
- 4GB of memory
- 4 VCPUs (sockets=1,cores=4,threads=1)
with the following workloads (with KSM and THP enabled in all):
guest A:
stress --cpu 40 --io 20 --vm 20 --vm-bytes 512M
guest B:
stress --cpu 4 --io 4 --vm 4 --vm-bytes 512M
host:
stress --cpu 4 --io 4 --vm 2 --vm-bytes 256M
the below soft-lockup traces were observed after an hour or so and
persisted until the host was reset (this was found to be reliably
reproducible for this configuration, for kernels 4.15, 4.18, 5.0,
and 5.3-rc5):
[ 1253.183290] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1253.183319] rcu: 124-....: (5250 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=1941
[ 1256.287426] watchdog: BUG: soft lockup - CPU#105 stuck for 23s! [CPU 52/KVM:19709]
[ 1264.075773] watchdog: BUG: soft lockup - CPU#24 stuck for 23s! [worker:19913]
[ 1264.079769] watchdog: BUG: soft lockup - CPU#31 stuck for 23s! [worker:20331]
[ 1264.095770] watchdog: BUG: soft lockup - CPU#45 stuck for 23s! [worker:20338]
[ 1264.131773] watchdog: BUG: soft lockup - CPU#64 stuck for 23s! [avocado:19525]
[ 1280.408480] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1316.198012] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1316.198032] rcu: 124-....: (21003 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=8243
[ 1340.411024] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1379.212609] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1379.212629] rcu: 124-....: (36756 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=14714
[ 1404.413615] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1442.227095] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1442.227115] rcu: 124-....: (52509 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=21403
[ 1455.111787] INFO: task worker:19907 blocked for more than 120 seconds.
[ 1455.111822] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111833] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111884] INFO: task worker:19908 blocked for more than 120 seconds.
[ 1455.111905] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111925] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111966] INFO: task worker:20328 blocked for more than 120 seconds.
[ 1455.111986] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111998] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112048] INFO: task worker:20330 blocked for more than 120 seconds.
[ 1455.112068] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112097] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112138] INFO: task worker:20332 blocked for more than 120 seconds.
[ 1455.112159] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112179] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112210] INFO: task worker:20333 blocked for more than 120 seconds.
[ 1455.112231] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112282] INFO: task worker:20335 blocked for more than 120 seconds.
[ 1455.112303] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112332] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112372] INFO: task worker:20336 blocked for more than 120 seconds.
[ 1455.112392] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
CPUs 45, 24, and 124 are stuck on spin locks, likely held by
CPUs 105 and 31.
CPUs 105 and 31 are stuck in smp_call_function_many(), waiting on
target CPU 42. For instance:
# CPU 105 registers (via xmon)
R00 = c00000000020b20c R16 = 00007d1bcd800000
R01 = c00000363eaa7970 R17 = 0000000000000001
R02 = c0000000019b3a00 R18 = 000000000000006b
R03 = 000000000000002a R19 = 00007d537d7aecf0
R04 = 000000000000002a R20 = 60000000000000e0
R05 = 000000000000002a R21 = 0801000000000080
R06 = c0002073fb0caa08 R22 = 0000000000000d60
R07 = c0000000019ddd78 R23 = 0000000000000001
R08 = 000000000000002a R24 = c00000000147a700
R09 = 0000000000000001 R25 = c0002073fb0ca908
R10 = c000008ffeb4e660 R26 = 0000000000000000
R11 = c0002073fb0ca900 R27 = c0000000019e2464
R12 = c000000000050790 R28 = c0000000000812b0
R13 = c000207fff623e00 R29 = c0002073fb0ca808
R14 = 00007d1bbee00000 R30 = c0002073fb0ca800
R15 = 00007d1bcd600000 R31 = 0000000000000800
pc = c00000000020b260 smp_call_function_many+0x3d0/0x460
cfar= c00000000020b270 smp_call_function_many+0x3e0/0x460
lr = c00000000020b20c smp_call_function_many+0x37c/0x460
msr = 900000010288b033 cr = 44024824
ctr = c000000000050790 xer = 0000000000000000 trap = 100
CPU 42 is running normally, doing VCPU work:
# CPU 42 stack trace (via xmon)
[link register ] c00800001be17188 kvmppc_book3s_radix_page_fault+0x90/0x2b0 [kvm_hv]
[c000008ed3343820] c000008ed3343850 (unreliable)
[c000008ed33438d0] c00800001be11b6c kvmppc_book3s_hv_page_fault+0x264/0xe30 [kvm_hv]
[c000008ed33439d0] c00800001be0d7b4 kvmppc_vcpu_run_hv+0x8dc/0xb50 [kvm_hv]
[c000008ed3343ae0] c00800001c10891c kvmppc_vcpu_run+0x34/0x48 [kvm]
[c000008ed3343b00] c00800001c10475c kvm_arch_vcpu_ioctl_run+0x244/0x420 [kvm]
[c000008ed3343b90] c00800001c0f5a78 kvm_vcpu_ioctl+0x470/0x7c8 [kvm]
[c000008ed3343d00] c000000000475450 do_vfs_ioctl+0xe0/0xc70
[c000008ed3343db0] c0000000004760e4 ksys_ioctl+0x104/0x120
[c000008ed3343e00] c000000000476128 sys_ioctl+0x28/0x80
[c000008ed3343e20] c00000000000b388 system_call+0x5c/0x70
--- Exception: c00 (System Call) at 00007d545cfd7694
SP (7d53ff7edf50) is in userspace
It was subsequently found that ipi_message[PPC_MSG_CALL_FUNCTION]
was set for CPU 42 by at least 1 of the CPUs waiting in
smp_call_function_many(), but somehow the corresponding
call_single_queue entries were never processed by CPU 42, causing the
callers to spin in csd_lock_wait() indefinitely.
Nick Piggin suggested something similar to the following sequence as
a possible explanation (interleaving of CALL_FUNCTION/RESCHEDULE
IPI messages seems to be most common, but any mix of CALL_FUNCTION and
!CALL_FUNCTION messages could trigger it):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
105: smp_muxed_ipi_set_message():
105: smb_mb()
// STORE DEFERRED DUE TO RE-ORDERING
--105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 42: // returns to executing guest
| // RE-ORDERED STORE COMPLETES
->105: message[CALL_FUNCTION] = 1
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
This can be prevented with an smp_mb() at the beginning of
kvmppc_set_host_ipi(), such that stores to message[<type>] (or other
state indicated by the host_ipi flag) are ordered vs. the store to
to host_ipi.
However, doing so might still allow for the following scenario (not
yet observed):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
// STORE DEFERRED DUE TO RE-ORDERING
-- 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 105: smp_muxed_ipi_set_message():
| 105: smb_mb()
| 105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| // RE-ORDERED STORE COMPLETES
-> 42: kvmppc_set_host_ipi(42, 0)
42: // returns to executing guest
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
Fixing this scenario would require an smp_mb() *after* clearing
host_ipi flag in kvmppc_set_host_ipi() to order the store vs.
subsequent processing of IPI messages.
To handle both cases, this patch splits kvmppc_set_host_ipi() into
separate set/clear functions, where we execute smp_mb() prior to
setting host_ipi flag, and after clearing host_ipi flag. These
functions pair with each other to synchronize the sender and receiver
sides.
With that change in place the above workload ran for 20 hours without
triggering any lock-ups.
Fixes: 755563bc79c7 ("powerpc/powernv: Fixes for hypervisor doorbell handling") # v4.0
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911223155.16045-1-mdroth@linux.vnet.ibm.com
2019-09-11 17:31:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
KVM: PPC: Book3S HV: use smp_mb() when setting/clearing host_ipi flag
On a 2-socket Power9 system with 32 cores/128 threads (SMT4) and 1TB
of memory running the following guest configs:
guest A:
- 224GB of memory
- 56 VCPUs (sockets=1,cores=28,threads=2), where:
VCPUs 0-1 are pinned to CPUs 0-3,
VCPUs 2-3 are pinned to CPUs 4-7,
...
VCPUs 54-55 are pinned to CPUs 108-111
guest B:
- 4GB of memory
- 4 VCPUs (sockets=1,cores=4,threads=1)
with the following workloads (with KSM and THP enabled in all):
guest A:
stress --cpu 40 --io 20 --vm 20 --vm-bytes 512M
guest B:
stress --cpu 4 --io 4 --vm 4 --vm-bytes 512M
host:
stress --cpu 4 --io 4 --vm 2 --vm-bytes 256M
the below soft-lockup traces were observed after an hour or so and
persisted until the host was reset (this was found to be reliably
reproducible for this configuration, for kernels 4.15, 4.18, 5.0,
and 5.3-rc5):
[ 1253.183290] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1253.183319] rcu: 124-....: (5250 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=1941
[ 1256.287426] watchdog: BUG: soft lockup - CPU#105 stuck for 23s! [CPU 52/KVM:19709]
[ 1264.075773] watchdog: BUG: soft lockup - CPU#24 stuck for 23s! [worker:19913]
[ 1264.079769] watchdog: BUG: soft lockup - CPU#31 stuck for 23s! [worker:20331]
[ 1264.095770] watchdog: BUG: soft lockup - CPU#45 stuck for 23s! [worker:20338]
[ 1264.131773] watchdog: BUG: soft lockup - CPU#64 stuck for 23s! [avocado:19525]
[ 1280.408480] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1316.198012] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1316.198032] rcu: 124-....: (21003 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=8243
[ 1340.411024] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1379.212609] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1379.212629] rcu: 124-....: (36756 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=14714
[ 1404.413615] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1442.227095] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1442.227115] rcu: 124-....: (52509 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=21403
[ 1455.111787] INFO: task worker:19907 blocked for more than 120 seconds.
[ 1455.111822] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111833] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111884] INFO: task worker:19908 blocked for more than 120 seconds.
[ 1455.111905] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111925] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111966] INFO: task worker:20328 blocked for more than 120 seconds.
[ 1455.111986] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111998] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112048] INFO: task worker:20330 blocked for more than 120 seconds.
[ 1455.112068] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112097] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112138] INFO: task worker:20332 blocked for more than 120 seconds.
[ 1455.112159] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112179] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112210] INFO: task worker:20333 blocked for more than 120 seconds.
[ 1455.112231] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112282] INFO: task worker:20335 blocked for more than 120 seconds.
[ 1455.112303] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112332] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112372] INFO: task worker:20336 blocked for more than 120 seconds.
[ 1455.112392] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
CPUs 45, 24, and 124 are stuck on spin locks, likely held by
CPUs 105 and 31.
CPUs 105 and 31 are stuck in smp_call_function_many(), waiting on
target CPU 42. For instance:
# CPU 105 registers (via xmon)
R00 = c00000000020b20c R16 = 00007d1bcd800000
R01 = c00000363eaa7970 R17 = 0000000000000001
R02 = c0000000019b3a00 R18 = 000000000000006b
R03 = 000000000000002a R19 = 00007d537d7aecf0
R04 = 000000000000002a R20 = 60000000000000e0
R05 = 000000000000002a R21 = 0801000000000080
R06 = c0002073fb0caa08 R22 = 0000000000000d60
R07 = c0000000019ddd78 R23 = 0000000000000001
R08 = 000000000000002a R24 = c00000000147a700
R09 = 0000000000000001 R25 = c0002073fb0ca908
R10 = c000008ffeb4e660 R26 = 0000000000000000
R11 = c0002073fb0ca900 R27 = c0000000019e2464
R12 = c000000000050790 R28 = c0000000000812b0
R13 = c000207fff623e00 R29 = c0002073fb0ca808
R14 = 00007d1bbee00000 R30 = c0002073fb0ca800
R15 = 00007d1bcd600000 R31 = 0000000000000800
pc = c00000000020b260 smp_call_function_many+0x3d0/0x460
cfar= c00000000020b270 smp_call_function_many+0x3e0/0x460
lr = c00000000020b20c smp_call_function_many+0x37c/0x460
msr = 900000010288b033 cr = 44024824
ctr = c000000000050790 xer = 0000000000000000 trap = 100
CPU 42 is running normally, doing VCPU work:
# CPU 42 stack trace (via xmon)
[link register ] c00800001be17188 kvmppc_book3s_radix_page_fault+0x90/0x2b0 [kvm_hv]
[c000008ed3343820] c000008ed3343850 (unreliable)
[c000008ed33438d0] c00800001be11b6c kvmppc_book3s_hv_page_fault+0x264/0xe30 [kvm_hv]
[c000008ed33439d0] c00800001be0d7b4 kvmppc_vcpu_run_hv+0x8dc/0xb50 [kvm_hv]
[c000008ed3343ae0] c00800001c10891c kvmppc_vcpu_run+0x34/0x48 [kvm]
[c000008ed3343b00] c00800001c10475c kvm_arch_vcpu_ioctl_run+0x244/0x420 [kvm]
[c000008ed3343b90] c00800001c0f5a78 kvm_vcpu_ioctl+0x470/0x7c8 [kvm]
[c000008ed3343d00] c000000000475450 do_vfs_ioctl+0xe0/0xc70
[c000008ed3343db0] c0000000004760e4 ksys_ioctl+0x104/0x120
[c000008ed3343e00] c000000000476128 sys_ioctl+0x28/0x80
[c000008ed3343e20] c00000000000b388 system_call+0x5c/0x70
--- Exception: c00 (System Call) at 00007d545cfd7694
SP (7d53ff7edf50) is in userspace
It was subsequently found that ipi_message[PPC_MSG_CALL_FUNCTION]
was set for CPU 42 by at least 1 of the CPUs waiting in
smp_call_function_many(), but somehow the corresponding
call_single_queue entries were never processed by CPU 42, causing the
callers to spin in csd_lock_wait() indefinitely.
Nick Piggin suggested something similar to the following sequence as
a possible explanation (interleaving of CALL_FUNCTION/RESCHEDULE
IPI messages seems to be most common, but any mix of CALL_FUNCTION and
!CALL_FUNCTION messages could trigger it):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
105: smp_muxed_ipi_set_message():
105: smb_mb()
// STORE DEFERRED DUE TO RE-ORDERING
--105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 42: // returns to executing guest
| // RE-ORDERED STORE COMPLETES
->105: message[CALL_FUNCTION] = 1
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
This can be prevented with an smp_mb() at the beginning of
kvmppc_set_host_ipi(), such that stores to message[<type>] (or other
state indicated by the host_ipi flag) are ordered vs. the store to
to host_ipi.
However, doing so might still allow for the following scenario (not
yet observed):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
// STORE DEFERRED DUE TO RE-ORDERING
-- 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 105: smp_muxed_ipi_set_message():
| 105: smb_mb()
| 105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| // RE-ORDERED STORE COMPLETES
-> 42: kvmppc_set_host_ipi(42, 0)
42: // returns to executing guest
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
Fixing this scenario would require an smp_mb() *after* clearing
host_ipi flag in kvmppc_set_host_ipi() to order the store vs.
subsequent processing of IPI messages.
To handle both cases, this patch splits kvmppc_set_host_ipi() into
separate set/clear functions, where we execute smp_mb() prior to
setting host_ipi flag, and after clearing host_ipi flag. These
functions pair with each other to synchronize the sender and receiver
sides.
With that change in place the above workload ran for 20 hours without
triggering any lock-ups.
Fixes: 755563bc79c7 ("powerpc/powernv: Fixes for hypervisor doorbell handling") # v4.0
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911223155.16045-1-mdroth@linux.vnet.ibm.com
2019-09-11 17:31:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-05-10 13:31:14 +10:00
|
|
|
|
KVM: PPC: Book3S HV: use smp_mb() when setting/clearing host_ipi flag
On a 2-socket Power9 system with 32 cores/128 threads (SMT4) and 1TB
of memory running the following guest configs:
guest A:
- 224GB of memory
- 56 VCPUs (sockets=1,cores=28,threads=2), where:
VCPUs 0-1 are pinned to CPUs 0-3,
VCPUs 2-3 are pinned to CPUs 4-7,
...
VCPUs 54-55 are pinned to CPUs 108-111
guest B:
- 4GB of memory
- 4 VCPUs (sockets=1,cores=4,threads=1)
with the following workloads (with KSM and THP enabled in all):
guest A:
stress --cpu 40 --io 20 --vm 20 --vm-bytes 512M
guest B:
stress --cpu 4 --io 4 --vm 4 --vm-bytes 512M
host:
stress --cpu 4 --io 4 --vm 2 --vm-bytes 256M
the below soft-lockup traces were observed after an hour or so and
persisted until the host was reset (this was found to be reliably
reproducible for this configuration, for kernels 4.15, 4.18, 5.0,
and 5.3-rc5):
[ 1253.183290] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1253.183319] rcu: 124-....: (5250 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=1941
[ 1256.287426] watchdog: BUG: soft lockup - CPU#105 stuck for 23s! [CPU 52/KVM:19709]
[ 1264.075773] watchdog: BUG: soft lockup - CPU#24 stuck for 23s! [worker:19913]
[ 1264.079769] watchdog: BUG: soft lockup - CPU#31 stuck for 23s! [worker:20331]
[ 1264.095770] watchdog: BUG: soft lockup - CPU#45 stuck for 23s! [worker:20338]
[ 1264.131773] watchdog: BUG: soft lockup - CPU#64 stuck for 23s! [avocado:19525]
[ 1280.408480] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1316.198012] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1316.198032] rcu: 124-....: (21003 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=8243
[ 1340.411024] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1379.212609] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1379.212629] rcu: 124-....: (36756 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=14714
[ 1404.413615] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1442.227095] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1442.227115] rcu: 124-....: (52509 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=21403
[ 1455.111787] INFO: task worker:19907 blocked for more than 120 seconds.
[ 1455.111822] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111833] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111884] INFO: task worker:19908 blocked for more than 120 seconds.
[ 1455.111905] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111925] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111966] INFO: task worker:20328 blocked for more than 120 seconds.
[ 1455.111986] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111998] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112048] INFO: task worker:20330 blocked for more than 120 seconds.
[ 1455.112068] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112097] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112138] INFO: task worker:20332 blocked for more than 120 seconds.
[ 1455.112159] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112179] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112210] INFO: task worker:20333 blocked for more than 120 seconds.
[ 1455.112231] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112282] INFO: task worker:20335 blocked for more than 120 seconds.
[ 1455.112303] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112332] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112372] INFO: task worker:20336 blocked for more than 120 seconds.
[ 1455.112392] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
CPUs 45, 24, and 124 are stuck on spin locks, likely held by
CPUs 105 and 31.
CPUs 105 and 31 are stuck in smp_call_function_many(), waiting on
target CPU 42. For instance:
# CPU 105 registers (via xmon)
R00 = c00000000020b20c R16 = 00007d1bcd800000
R01 = c00000363eaa7970 R17 = 0000000000000001
R02 = c0000000019b3a00 R18 = 000000000000006b
R03 = 000000000000002a R19 = 00007d537d7aecf0
R04 = 000000000000002a R20 = 60000000000000e0
R05 = 000000000000002a R21 = 0801000000000080
R06 = c0002073fb0caa08 R22 = 0000000000000d60
R07 = c0000000019ddd78 R23 = 0000000000000001
R08 = 000000000000002a R24 = c00000000147a700
R09 = 0000000000000001 R25 = c0002073fb0ca908
R10 = c000008ffeb4e660 R26 = 0000000000000000
R11 = c0002073fb0ca900 R27 = c0000000019e2464
R12 = c000000000050790 R28 = c0000000000812b0
R13 = c000207fff623e00 R29 = c0002073fb0ca808
R14 = 00007d1bbee00000 R30 = c0002073fb0ca800
R15 = 00007d1bcd600000 R31 = 0000000000000800
pc = c00000000020b260 smp_call_function_many+0x3d0/0x460
cfar= c00000000020b270 smp_call_function_many+0x3e0/0x460
lr = c00000000020b20c smp_call_function_many+0x37c/0x460
msr = 900000010288b033 cr = 44024824
ctr = c000000000050790 xer = 0000000000000000 trap = 100
CPU 42 is running normally, doing VCPU work:
# CPU 42 stack trace (via xmon)
[link register ] c00800001be17188 kvmppc_book3s_radix_page_fault+0x90/0x2b0 [kvm_hv]
[c000008ed3343820] c000008ed3343850 (unreliable)
[c000008ed33438d0] c00800001be11b6c kvmppc_book3s_hv_page_fault+0x264/0xe30 [kvm_hv]
[c000008ed33439d0] c00800001be0d7b4 kvmppc_vcpu_run_hv+0x8dc/0xb50 [kvm_hv]
[c000008ed3343ae0] c00800001c10891c kvmppc_vcpu_run+0x34/0x48 [kvm]
[c000008ed3343b00] c00800001c10475c kvm_arch_vcpu_ioctl_run+0x244/0x420 [kvm]
[c000008ed3343b90] c00800001c0f5a78 kvm_vcpu_ioctl+0x470/0x7c8 [kvm]
[c000008ed3343d00] c000000000475450 do_vfs_ioctl+0xe0/0xc70
[c000008ed3343db0] c0000000004760e4 ksys_ioctl+0x104/0x120
[c000008ed3343e00] c000000000476128 sys_ioctl+0x28/0x80
[c000008ed3343e20] c00000000000b388 system_call+0x5c/0x70
--- Exception: c00 (System Call) at 00007d545cfd7694
SP (7d53ff7edf50) is in userspace
It was subsequently found that ipi_message[PPC_MSG_CALL_FUNCTION]
was set for CPU 42 by at least 1 of the CPUs waiting in
smp_call_function_many(), but somehow the corresponding
call_single_queue entries were never processed by CPU 42, causing the
callers to spin in csd_lock_wait() indefinitely.
Nick Piggin suggested something similar to the following sequence as
a possible explanation (interleaving of CALL_FUNCTION/RESCHEDULE
IPI messages seems to be most common, but any mix of CALL_FUNCTION and
!CALL_FUNCTION messages could trigger it):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
105: smp_muxed_ipi_set_message():
105: smb_mb()
// STORE DEFERRED DUE TO RE-ORDERING
--105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 42: // returns to executing guest
| // RE-ORDERED STORE COMPLETES
->105: message[CALL_FUNCTION] = 1
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
This can be prevented with an smp_mb() at the beginning of
kvmppc_set_host_ipi(), such that stores to message[<type>] (or other
state indicated by the host_ipi flag) are ordered vs. the store to
to host_ipi.
However, doing so might still allow for the following scenario (not
yet observed):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
// STORE DEFERRED DUE TO RE-ORDERING
-- 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 105: smp_muxed_ipi_set_message():
| 105: smb_mb()
| 105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| // RE-ORDERED STORE COMPLETES
-> 42: kvmppc_set_host_ipi(42, 0)
42: // returns to executing guest
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
Fixing this scenario would require an smp_mb() *after* clearing
host_ipi flag in kvmppc_set_host_ipi() to order the store vs.
subsequent processing of IPI messages.
To handle both cases, this patch splits kvmppc_set_host_ipi() into
separate set/clear functions, where we execute smp_mb() prior to
setting host_ipi flag, and after clearing host_ipi flag. These
functions pair with each other to synchronize the sender and receiver
sides.
With that change in place the above workload ran for 20 hours without
triggering any lock-ups.
Fixes: 755563bc79c7 ("powerpc/powernv: Fixes for hypervisor doorbell handling") # v4.0
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911223155.16045-1-mdroth@linux.vnet.ibm.com
2019-09-11 17:31:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-05-10 13:31:14 +10:00
|
|
|
|
KVM: PPC: Book3S HV: use smp_mb() when setting/clearing host_ipi flag
On a 2-socket Power9 system with 32 cores/128 threads (SMT4) and 1TB
of memory running the following guest configs:
guest A:
- 224GB of memory
- 56 VCPUs (sockets=1,cores=28,threads=2), where:
VCPUs 0-1 are pinned to CPUs 0-3,
VCPUs 2-3 are pinned to CPUs 4-7,
...
VCPUs 54-55 are pinned to CPUs 108-111
guest B:
- 4GB of memory
- 4 VCPUs (sockets=1,cores=4,threads=1)
with the following workloads (with KSM and THP enabled in all):
guest A:
stress --cpu 40 --io 20 --vm 20 --vm-bytes 512M
guest B:
stress --cpu 4 --io 4 --vm 4 --vm-bytes 512M
host:
stress --cpu 4 --io 4 --vm 2 --vm-bytes 256M
the below soft-lockup traces were observed after an hour or so and
persisted until the host was reset (this was found to be reliably
reproducible for this configuration, for kernels 4.15, 4.18, 5.0,
and 5.3-rc5):
[ 1253.183290] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1253.183319] rcu: 124-....: (5250 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=1941
[ 1256.287426] watchdog: BUG: soft lockup - CPU#105 stuck for 23s! [CPU 52/KVM:19709]
[ 1264.075773] watchdog: BUG: soft lockup - CPU#24 stuck for 23s! [worker:19913]
[ 1264.079769] watchdog: BUG: soft lockup - CPU#31 stuck for 23s! [worker:20331]
[ 1264.095770] watchdog: BUG: soft lockup - CPU#45 stuck for 23s! [worker:20338]
[ 1264.131773] watchdog: BUG: soft lockup - CPU#64 stuck for 23s! [avocado:19525]
[ 1280.408480] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1316.198012] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1316.198032] rcu: 124-....: (21003 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=8243
[ 1340.411024] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1379.212609] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1379.212629] rcu: 124-....: (36756 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=14714
[ 1404.413615] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1442.227095] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1442.227115] rcu: 124-....: (52509 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=21403
[ 1455.111787] INFO: task worker:19907 blocked for more than 120 seconds.
[ 1455.111822] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111833] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111884] INFO: task worker:19908 blocked for more than 120 seconds.
[ 1455.111905] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111925] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111966] INFO: task worker:20328 blocked for more than 120 seconds.
[ 1455.111986] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111998] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112048] INFO: task worker:20330 blocked for more than 120 seconds.
[ 1455.112068] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112097] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112138] INFO: task worker:20332 blocked for more than 120 seconds.
[ 1455.112159] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112179] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112210] INFO: task worker:20333 blocked for more than 120 seconds.
[ 1455.112231] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112282] INFO: task worker:20335 blocked for more than 120 seconds.
[ 1455.112303] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112332] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112372] INFO: task worker:20336 blocked for more than 120 seconds.
[ 1455.112392] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
CPUs 45, 24, and 124 are stuck on spin locks, likely held by
CPUs 105 and 31.
CPUs 105 and 31 are stuck in smp_call_function_many(), waiting on
target CPU 42. For instance:
# CPU 105 registers (via xmon)
R00 = c00000000020b20c R16 = 00007d1bcd800000
R01 = c00000363eaa7970 R17 = 0000000000000001
R02 = c0000000019b3a00 R18 = 000000000000006b
R03 = 000000000000002a R19 = 00007d537d7aecf0
R04 = 000000000000002a R20 = 60000000000000e0
R05 = 000000000000002a R21 = 0801000000000080
R06 = c0002073fb0caa08 R22 = 0000000000000d60
R07 = c0000000019ddd78 R23 = 0000000000000001
R08 = 000000000000002a R24 = c00000000147a700
R09 = 0000000000000001 R25 = c0002073fb0ca908
R10 = c000008ffeb4e660 R26 = 0000000000000000
R11 = c0002073fb0ca900 R27 = c0000000019e2464
R12 = c000000000050790 R28 = c0000000000812b0
R13 = c000207fff623e00 R29 = c0002073fb0ca808
R14 = 00007d1bbee00000 R30 = c0002073fb0ca800
R15 = 00007d1bcd600000 R31 = 0000000000000800
pc = c00000000020b260 smp_call_function_many+0x3d0/0x460
cfar= c00000000020b270 smp_call_function_many+0x3e0/0x460
lr = c00000000020b20c smp_call_function_many+0x37c/0x460
msr = 900000010288b033 cr = 44024824
ctr = c000000000050790 xer = 0000000000000000 trap = 100
CPU 42 is running normally, doing VCPU work:
# CPU 42 stack trace (via xmon)
[link register ] c00800001be17188 kvmppc_book3s_radix_page_fault+0x90/0x2b0 [kvm_hv]
[c000008ed3343820] c000008ed3343850 (unreliable)
[c000008ed33438d0] c00800001be11b6c kvmppc_book3s_hv_page_fault+0x264/0xe30 [kvm_hv]
[c000008ed33439d0] c00800001be0d7b4 kvmppc_vcpu_run_hv+0x8dc/0xb50 [kvm_hv]
[c000008ed3343ae0] c00800001c10891c kvmppc_vcpu_run+0x34/0x48 [kvm]
[c000008ed3343b00] c00800001c10475c kvm_arch_vcpu_ioctl_run+0x244/0x420 [kvm]
[c000008ed3343b90] c00800001c0f5a78 kvm_vcpu_ioctl+0x470/0x7c8 [kvm]
[c000008ed3343d00] c000000000475450 do_vfs_ioctl+0xe0/0xc70
[c000008ed3343db0] c0000000004760e4 ksys_ioctl+0x104/0x120
[c000008ed3343e00] c000000000476128 sys_ioctl+0x28/0x80
[c000008ed3343e20] c00000000000b388 system_call+0x5c/0x70
--- Exception: c00 (System Call) at 00007d545cfd7694
SP (7d53ff7edf50) is in userspace
It was subsequently found that ipi_message[PPC_MSG_CALL_FUNCTION]
was set for CPU 42 by at least 1 of the CPUs waiting in
smp_call_function_many(), but somehow the corresponding
call_single_queue entries were never processed by CPU 42, causing the
callers to spin in csd_lock_wait() indefinitely.
Nick Piggin suggested something similar to the following sequence as
a possible explanation (interleaving of CALL_FUNCTION/RESCHEDULE
IPI messages seems to be most common, but any mix of CALL_FUNCTION and
!CALL_FUNCTION messages could trigger it):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
105: smp_muxed_ipi_set_message():
105: smb_mb()
// STORE DEFERRED DUE TO RE-ORDERING
--105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 42: // returns to executing guest
| // RE-ORDERED STORE COMPLETES
->105: message[CALL_FUNCTION] = 1
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
This can be prevented with an smp_mb() at the beginning of
kvmppc_set_host_ipi(), such that stores to message[<type>] (or other
state indicated by the host_ipi flag) are ordered vs. the store to
to host_ipi.
However, doing so might still allow for the following scenario (not
yet observed):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
// STORE DEFERRED DUE TO RE-ORDERING
-- 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 105: smp_muxed_ipi_set_message():
| 105: smb_mb()
| 105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| // RE-ORDERED STORE COMPLETES
-> 42: kvmppc_set_host_ipi(42, 0)
42: // returns to executing guest
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
Fixing this scenario would require an smp_mb() *after* clearing
host_ipi flag in kvmppc_set_host_ipi() to order the store vs.
subsequent processing of IPI messages.
To handle both cases, this patch splits kvmppc_set_host_ipi() into
separate set/clear functions, where we execute smp_mb() prior to
setting host_ipi flag, and after clearing host_ipi flag. These
functions pair with each other to synchronize the sender and receiver
sides.
With that change in place the above workload ran for 20 hours without
triggering any lock-ups.
Fixes: 755563bc79c7 ("powerpc/powernv: Fixes for hypervisor doorbell handling") # v4.0
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911223155.16045-1-mdroth@linux.vnet.ibm.com
2019-09-11 17:31:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
|
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
|
|
|
|
|
2013-10-07 22:18:01 +05:30
|
|
|
|
2013-10-07 22:17:53 +05:30
|
|
|
|
KVM: PPC: Allocate RMAs (Real Mode Areas) at boot for use by guests
This adds infrastructure which will be needed to allow book3s_hv KVM to
run on older POWER processors, including PPC970, which don't support
the Virtual Real Mode Area (VRMA) facility, but only the Real Mode
Offset (RMO) facility. These processors require a physically
contiguous, aligned area of memory for each guest. When the guest does
an access in real mode (MMU off), the address is compared against a
limit value, and if it is lower, the address is ORed with an offset
value (from the Real Mode Offset Register (RMOR)) and the result becomes
the real address for the access. The size of the RMA has to be one of
a set of supported values, which usually includes 64MB, 128MB, 256MB
and some larger powers of 2.
Since we are unlikely to be able to allocate 64MB or more of physically
contiguous memory after the kernel has been running for a while, we
allocate a pool of RMAs at boot time using the bootmem allocator. The
size and number of the RMAs can be set using the kvm_rma_size=xx and
kvm_rma_count=xx kernel command line options.
KVM exports a new capability, KVM_CAP_PPC_RMA, to signal the availability
of the pool of preallocated RMAs. The capability value is 1 if the
processor can use an RMA but doesn't require one (because it supports
the VRMA facility), or 2 if the processor requires an RMA for each guest.
This adds a new ioctl, KVM_ALLOCATE_RMA, which allocates an RMA from the
pool and returns a file descriptor which can be used to map the RMA. It
also returns the size of the RMA in the argument structure.
Having an RMA means we will get multiple KMV_SET_USER_MEMORY_REGION
ioctl calls from userspace. To cope with this, we now preallocate the
kvm->arch.ram_pginfo array when the VM is created with a size sufficient
for up to 64GB of guest memory. Subsequently we will get rid of this
array and use memory associated with each memslot instead.
This moves most of the code that translates the user addresses into
host pfns (page frame numbers) out of kvmppc_prepare_vrma up one level
to kvmppc_core_prepare_memory_region. Also, instead of having to look
up the VMA for each page in order to check the page size, we now check
that the pages we get are compound pages of 16MB. However, if we are
adding memory that is mapped to an RMA, we don't bother with calling
get_user_pages_fast and instead just offset from the base pfn for the
RMA.
Typically the RMA gets added after vcpus are created, which makes it
inconvenient to have the LPCR (logical partition control register) value
in the vcpu->arch struct, since the LPCR controls whether the processor
uses RMA or VRMA for the guest. This moves the LPCR value into the
kvm->arch struct and arranges for the MER (mediated external request)
bit, which is the only bit that varies between vcpus, to be set in
assembly code when going into the guest if there is a pending external
interrupt request.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:25:44 +00:00
|
|
|
|
2014-05-23 18:15:25 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-23 19:52:20 +10:00
|
|
|
|
2019-04-29 19:00:40 +10:00
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
2013-07-02 11:15:16 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Allocate RMAs (Real Mode Areas) at boot for use by guests
This adds infrastructure which will be needed to allow book3s_hv KVM to
run on older POWER processors, including PPC970, which don't support
the Virtual Real Mode Area (VRMA) facility, but only the Real Mode
Offset (RMO) facility. These processors require a physically
contiguous, aligned area of memory for each guest. When the guest does
an access in real mode (MMU off), the address is compared against a
limit value, and if it is lower, the address is ORed with an offset
value (from the Real Mode Offset Register (RMOR)) and the result becomes
the real address for the access. The size of the RMA has to be one of
a set of supported values, which usually includes 64MB, 128MB, 256MB
and some larger powers of 2.
Since we are unlikely to be able to allocate 64MB or more of physically
contiguous memory after the kernel has been running for a while, we
allocate a pool of RMAs at boot time using the bootmem allocator. The
size and number of the RMAs can be set using the kvm_rma_size=xx and
kvm_rma_count=xx kernel command line options.
KVM exports a new capability, KVM_CAP_PPC_RMA, to signal the availability
of the pool of preallocated RMAs. The capability value is 1 if the
processor can use an RMA but doesn't require one (because it supports
the VRMA facility), or 2 if the processor requires an RMA for each guest.
This adds a new ioctl, KVM_ALLOCATE_RMA, which allocates an RMA from the
pool and returns a file descriptor which can be used to map the RMA. It
also returns the size of the RMA in the argument structure.
Having an RMA means we will get multiple KMV_SET_USER_MEMORY_REGION
ioctl calls from userspace. To cope with this, we now preallocate the
kvm->arch.ram_pginfo array when the VM is created with a size sufficient
for up to 64GB of guest memory. Subsequently we will get rid of this
array and use memory associated with each memslot instead.
This moves most of the code that translates the user addresses into
host pfns (page frame numbers) out of kvmppc_prepare_vrma up one level
to kvmppc_core_prepare_memory_region. Also, instead of having to look
up the VMA for each page in order to check the page size, we now check
that the pages we get are compound pages of 16MB. However, if we are
adding memory that is mapped to an RMA, we don't bother with calling
get_user_pages_fast and instead just offset from the base pfn for the
RMA.
Typically the RMA gets added after vcpus are created, which makes it
inconvenient to have the LPCR (logical partition control register) value
in the vcpu->arch struct, since the LPCR controls whether the processor
uses RMA or VRMA for the guest. This moves the LPCR value into the
kvm->arch struct and arranges for the MER (mediated external request)
bit, which is the only bit that varies between vcpus, to be set in
assembly code when going into the guest if there is a pending external
interrupt request.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:25:44 +00:00
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV: use smp_mb() when setting/clearing host_ipi flag
On a 2-socket Power9 system with 32 cores/128 threads (SMT4) and 1TB
of memory running the following guest configs:
guest A:
- 224GB of memory
- 56 VCPUs (sockets=1,cores=28,threads=2), where:
VCPUs 0-1 are pinned to CPUs 0-3,
VCPUs 2-3 are pinned to CPUs 4-7,
...
VCPUs 54-55 are pinned to CPUs 108-111
guest B:
- 4GB of memory
- 4 VCPUs (sockets=1,cores=4,threads=1)
with the following workloads (with KSM and THP enabled in all):
guest A:
stress --cpu 40 --io 20 --vm 20 --vm-bytes 512M
guest B:
stress --cpu 4 --io 4 --vm 4 --vm-bytes 512M
host:
stress --cpu 4 --io 4 --vm 2 --vm-bytes 256M
the below soft-lockup traces were observed after an hour or so and
persisted until the host was reset (this was found to be reliably
reproducible for this configuration, for kernels 4.15, 4.18, 5.0,
and 5.3-rc5):
[ 1253.183290] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1253.183319] rcu: 124-....: (5250 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=1941
[ 1256.287426] watchdog: BUG: soft lockup - CPU#105 stuck for 23s! [CPU 52/KVM:19709]
[ 1264.075773] watchdog: BUG: soft lockup - CPU#24 stuck for 23s! [worker:19913]
[ 1264.079769] watchdog: BUG: soft lockup - CPU#31 stuck for 23s! [worker:20331]
[ 1264.095770] watchdog: BUG: soft lockup - CPU#45 stuck for 23s! [worker:20338]
[ 1264.131773] watchdog: BUG: soft lockup - CPU#64 stuck for 23s! [avocado:19525]
[ 1280.408480] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1316.198012] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1316.198032] rcu: 124-....: (21003 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=8243
[ 1340.411024] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1379.212609] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1379.212629] rcu: 124-....: (36756 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=14714
[ 1404.413615] watchdog: BUG: soft lockup - CPU#124 stuck for 22s! [ksmd:791]
[ 1442.227095] rcu: INFO: rcu_sched self-detected stall on CPU
[ 1442.227115] rcu: 124-....: (52509 ticks this GP) idle=10a/1/0x4000000000000002 softirq=5408/5408 fqs=21403
[ 1455.111787] INFO: task worker:19907 blocked for more than 120 seconds.
[ 1455.111822] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111833] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111884] INFO: task worker:19908 blocked for more than 120 seconds.
[ 1455.111905] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111925] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.111966] INFO: task worker:20328 blocked for more than 120 seconds.
[ 1455.111986] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.111998] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112048] INFO: task worker:20330 blocked for more than 120 seconds.
[ 1455.112068] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112097] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112138] INFO: task worker:20332 blocked for more than 120 seconds.
[ 1455.112159] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112179] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112210] INFO: task worker:20333 blocked for more than 120 seconds.
[ 1455.112231] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112242] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112282] INFO: task worker:20335 blocked for more than 120 seconds.
[ 1455.112303] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
[ 1455.112332] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 1455.112372] INFO: task worker:20336 blocked for more than 120 seconds.
[ 1455.112392] Tainted: G L 5.3.0-rc5-mdr-vanilla+ #1
CPUs 45, 24, and 124 are stuck on spin locks, likely held by
CPUs 105 and 31.
CPUs 105 and 31 are stuck in smp_call_function_many(), waiting on
target CPU 42. For instance:
# CPU 105 registers (via xmon)
R00 = c00000000020b20c R16 = 00007d1bcd800000
R01 = c00000363eaa7970 R17 = 0000000000000001
R02 = c0000000019b3a00 R18 = 000000000000006b
R03 = 000000000000002a R19 = 00007d537d7aecf0
R04 = 000000000000002a R20 = 60000000000000e0
R05 = 000000000000002a R21 = 0801000000000080
R06 = c0002073fb0caa08 R22 = 0000000000000d60
R07 = c0000000019ddd78 R23 = 0000000000000001
R08 = 000000000000002a R24 = c00000000147a700
R09 = 0000000000000001 R25 = c0002073fb0ca908
R10 = c000008ffeb4e660 R26 = 0000000000000000
R11 = c0002073fb0ca900 R27 = c0000000019e2464
R12 = c000000000050790 R28 = c0000000000812b0
R13 = c000207fff623e00 R29 = c0002073fb0ca808
R14 = 00007d1bbee00000 R30 = c0002073fb0ca800
R15 = 00007d1bcd600000 R31 = 0000000000000800
pc = c00000000020b260 smp_call_function_many+0x3d0/0x460
cfar= c00000000020b270 smp_call_function_many+0x3e0/0x460
lr = c00000000020b20c smp_call_function_many+0x37c/0x460
msr = 900000010288b033 cr = 44024824
ctr = c000000000050790 xer = 0000000000000000 trap = 100
CPU 42 is running normally, doing VCPU work:
# CPU 42 stack trace (via xmon)
[link register ] c00800001be17188 kvmppc_book3s_radix_page_fault+0x90/0x2b0 [kvm_hv]
[c000008ed3343820] c000008ed3343850 (unreliable)
[c000008ed33438d0] c00800001be11b6c kvmppc_book3s_hv_page_fault+0x264/0xe30 [kvm_hv]
[c000008ed33439d0] c00800001be0d7b4 kvmppc_vcpu_run_hv+0x8dc/0xb50 [kvm_hv]
[c000008ed3343ae0] c00800001c10891c kvmppc_vcpu_run+0x34/0x48 [kvm]
[c000008ed3343b00] c00800001c10475c kvm_arch_vcpu_ioctl_run+0x244/0x420 [kvm]
[c000008ed3343b90] c00800001c0f5a78 kvm_vcpu_ioctl+0x470/0x7c8 [kvm]
[c000008ed3343d00] c000000000475450 do_vfs_ioctl+0xe0/0xc70
[c000008ed3343db0] c0000000004760e4 ksys_ioctl+0x104/0x120
[c000008ed3343e00] c000000000476128 sys_ioctl+0x28/0x80
[c000008ed3343e20] c00000000000b388 system_call+0x5c/0x70
--- Exception: c00 (System Call) at 00007d545cfd7694
SP (7d53ff7edf50) is in userspace
It was subsequently found that ipi_message[PPC_MSG_CALL_FUNCTION]
was set for CPU 42 by at least 1 of the CPUs waiting in
smp_call_function_many(), but somehow the corresponding
call_single_queue entries were never processed by CPU 42, causing the
callers to spin in csd_lock_wait() indefinitely.
Nick Piggin suggested something similar to the following sequence as
a possible explanation (interleaving of CALL_FUNCTION/RESCHEDULE
IPI messages seems to be most common, but any mix of CALL_FUNCTION and
!CALL_FUNCTION messages could trigger it):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
105: smp_muxed_ipi_set_message():
105: smb_mb()
// STORE DEFERRED DUE TO RE-ORDERING
--105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 42: // returns to executing guest
| // RE-ORDERED STORE COMPLETES
->105: message[CALL_FUNCTION] = 1
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
This can be prevented with an smp_mb() at the beginning of
kvmppc_set_host_ipi(), such that stores to message[<type>] (or other
state indicated by the host_ipi flag) are ordered vs. the store to
to host_ipi.
However, doing so might still allow for the following scenario (not
yet observed):
CPU
X: smp_muxed_ipi_set_message():
X: smp_mb()
X: message[RESCHEDULE] = 1
X: doorbell_global_ipi(42):
X: kvmppc_set_host_ipi(42, 1)
X: ppc_msgsnd_sync()/smp_mb()
X: ppc_msgsnd() -> 42
42: doorbell_exception(): // from CPU X
42: ppc_msgsync()
// STORE DEFERRED DUE TO RE-ORDERING
-- 42: kvmppc_set_host_ipi(42, 0)
| 42: smp_ipi_demux_relaxed()
| 105: smp_muxed_ipi_set_message():
| 105: smb_mb()
| 105: message[CALL_FUNCTION] = 1
| 105: doorbell_global_ipi(42):
| 105: kvmppc_set_host_ipi(42, 1)
| // RE-ORDERED STORE COMPLETES
-> 42: kvmppc_set_host_ipi(42, 0)
42: // returns to executing guest
105: ppc_msgsnd_sync()/smp_mb()
105: ppc_msgsnd() -> 42
42: local_paca->kvm_hstate.host_ipi == 0 // IPI ignored
105: // hangs waiting on 42 to process messages/call_single_queue
Fixing this scenario would require an smp_mb() *after* clearing
host_ipi flag in kvmppc_set_host_ipi() to order the store vs.
subsequent processing of IPI messages.
To handle both cases, this patch splits kvmppc_set_host_ipi() into
separate set/clear functions, where we execute smp_mb() prior to
setting host_ipi flag, and after clearing host_ipi flag. These
functions pair with each other to synchronize the sender and receiver
sides.
With that change in place the above workload ran for 20 hours without
triggering any lock-ups.
Fixes: 755563bc79c7 ("powerpc/powernv: Fixes for hypervisor doorbell handling") # v4.0
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20190911223155.16045-1-mdroth@linux.vnet.ibm.com
2019-09-11 17:31:55 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-05-23 18:15:25 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
|
|
|
|
|
2022-04-03 22:12:52 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2022-04-03 22:12:52 +10:00
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-08-19 15:35:48 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-08-19 15:35:54 +10:00
|
|
|
|
2016-08-19 15:35:48 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-17 14:59:06 -06:00
|
|
|
|
|
|
|
|
|
2016-08-19 15:35:48 +10:00
|
|
|
|
2016-08-19 15:35:52 +10:00
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV P9: Stop handling hcalls in real-mode in the P9 path
In the interest of minimising the amount of code that is run in
"real-mode", don't handle hcalls in real mode in the P9 path. This
requires some new handlers for H_CEDE and xics-on-xive to be added
before xive is pulled or cede logic is checked.
This introduces a change in radix guest behaviour where radix guests
that execute 'sc 1' in userspace now get a privilege fault whereas
previously the 'sc 1' would be reflected as a syscall interrupt to the
guest kernel. That reflection is only required for hash guests that run
PR KVM.
Background:
In POWER8 and earlier processors, it is very expensive to exit from the
HV real mode context of a guest hypervisor interrupt, and switch to host
virtual mode. On those processors, guest->HV interrupts reach the
hypervisor with the MMU off because the MMU is loaded with guest context
(LPCR, SDR1, SLB), and the other threads in the sub-core need to be
pulled out of the guest too. Then the primary must save off guest state,
invalidate SLB and ERAT, and load up host state before the MMU can be
enabled to run in host virtual mode (~= regular Linux mode).
Hash guests also require a lot of hcalls to run due to the nature of the
MMU architecture and paravirtualisation design. The XICS interrupt
controller requires hcalls to run.
So KVM traditionally tries hard to avoid the full exit, by handling
hcalls and other interrupts in real mode as much as possible.
By contrast, POWER9 has independent MMU context per-thread, and in radix
mode the hypervisor is in host virtual memory mode when the HV interrupt
is taken. Radix guests do not require significant hcalls to manage their
translations, and xive guests don't need hcalls to handle interrupts. So
it's much less important for performance to handle hcalls in real mode on
POWER9.
One caveat is that the TCE hcalls are performance critical, real-mode
variants introduced for POWER8 in order to achieve 10GbE performance.
Real mode TCE hcalls were found to be less important on POWER9, which
was able to drive 40GBe networking without them (using the virt mode
hcalls) but performance is still important. These hcalls will benefit
from subsequent guest entry/exit optimisation including possibly a
faster "partial exit" that does not entirely switch to host context to
handle the hcall.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210528090752.3542186-14-npiggin@gmail.com
2021-05-28 19:07:33 +10:00
|
|
|
|
2013-04-17 20:32:26 +00:00
|
|
|
|
|
|
|
|
|
2013-04-27 00:28:37 +00:00
|
|
|
|
|
|
|
|
|
2015-12-17 14:59:09 -06:00
|
|
|
|
KVM: PPC: Book3S HV: Set server for passed-through interrupts
When a guest has a PCI pass-through device with an interrupt, it
will direct the interrupt to a particular guest VCPU. In fact the
physical interrupt might arrive on any CPU, and then get
delivered to the target VCPU in the emulated XICS (guest interrupt
controller), and eventually delivered to the target VCPU.
Now that we have code to handle device interrupts in real mode
without exiting to the host kernel, there is an advantage to having
the device interrupt arrive on the same sub(core) as the target
VCPU is running on. In this situation, the interrupt can be
delivered to the target VCPU without any exit to the host kernel
(using a hypervisor doorbell interrupt between threads if
necessary).
This patch aims to get passed-through device interrupts arriving
on the correct core by setting the interrupt server in the real
hardware XICS for the interrupt to the first thread in the (sub)core
where its target VCPU is running. We do this in the real-mode H_EOI
code because the H_EOI handler already needs to look at the
emulated ICS state for the interrupt (whereas the H_XIRR handler
doesn't), and we know we are running in the target VCPU context
at that point.
We set the server CPU in hardware using an OPAL call, regardless of
what the IRQ affinity mask for the interrupt says, and without
updating the affinity mask. This amounts to saying that when an
interrupt is passed through to a guest, as a matter of policy we
allow the guest's affinity for the interrupt to override the host's.
This is inspired by an earlier patch from Suresh Warrier, although
none of this code came from that earlier patch.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2016-08-19 15:35:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV: Use OPAL XICS emulation on POWER9
POWER9 includes a new interrupt controller, called XIVE, which is
quite different from the XICS interrupt controller on POWER7 and
POWER8 machines. KVM-HV accesses the XICS directly in several places
in order to send and clear IPIs and handle interrupts from PCI
devices being passed through to the guest.
In order to make the transition to XIVE easier, OPAL firmware will
include an emulation of XICS on top of XIVE. Access to the emulated
XICS is via OPAL calls. The one complication is that the EOI
(end-of-interrupt) function can now return a value indicating that
another interrupt is pending; in this case, the XIVE will not signal
an interrupt in hardware to the CPU, and software is supposed to
acknowledge the new interrupt without waiting for another interrupt
to be delivered in hardware.
This adapts KVM-HV to use the OPAL calls on machines where there is
no XICS hardware. When there is no XICS, we look for a device-tree
node with "ibm,opal-intc" in its compatible property, which is how
OPAL indicates that it provides XICS emulation.
In order to handle the EOI return value, kvmppc_read_intr() has
become kvmppc_read_one_intr(), with a boolean variable passed by
reference which can be set by the EOI functions to indicate that
another interrupt is pending. The new kvmppc_read_intr() keeps
calling kvmppc_read_one_intr() until there are no more interrupts
to process. The return value from kvmppc_read_intr() is the
largest non-zero value of the returns from kvmppc_read_one_intr().
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2016-11-18 09:02:08 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2015-12-21 16:33:57 -06:00
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
2016-08-19 15:35:48 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-01-25 17:53:38 +08:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-08-19 15:35:52 +10:00
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:26 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
KVM: PPC: Book3S HV P9: Stop handling hcalls in real-mode in the P9 path
In the interest of minimising the amount of code that is run in
"real-mode", don't handle hcalls in real mode in the P9 path. This
requires some new handlers for H_CEDE and xics-on-xive to be added
before xive is pulled or cede logic is checked.
This introduces a change in radix guest behaviour where radix guests
that execute 'sc 1' in userspace now get a privilege fault whereas
previously the 'sc 1' would be reflected as a syscall interrupt to the
guest kernel. That reflection is only required for hash guests that run
PR KVM.
Background:
In POWER8 and earlier processors, it is very expensive to exit from the
HV real mode context of a guest hypervisor interrupt, and switch to host
virtual mode. On those processors, guest->HV interrupts reach the
hypervisor with the MMU off because the MMU is loaded with guest context
(LPCR, SDR1, SLB), and the other threads in the sub-core need to be
pulled out of the guest too. Then the primary must save off guest state,
invalidate SLB and ERAT, and load up host state before the MMU can be
enabled to run in host virtual mode (~= regular Linux mode).
Hash guests also require a lot of hcalls to run due to the nature of the
MMU architecture and paravirtualisation design. The XICS interrupt
controller requires hcalls to run.
So KVM traditionally tries hard to avoid the full exit, by handling
hcalls and other interrupts in real mode as much as possible.
By contrast, POWER9 has independent MMU context per-thread, and in radix
mode the hypervisor is in host virtual memory mode when the HV interrupt
is taken. Radix guests do not require significant hcalls to manage their
translations, and xive guests don't need hcalls to handle interrupts. So
it's much less important for performance to handle hcalls in real mode on
POWER9.
One caveat is that the TCE hcalls are performance critical, real-mode
variants introduced for POWER8 in order to achieve 10GbE performance.
Real mode TCE hcalls were found to be less important on POWER9, which
was able to drive 40GBe networking without them (using the virt mode
hcalls) but performance is still important. These hcalls will benefit
from subsequent guest entry/exit optimisation including possibly a
faster "partial exit" that does not entirely switch to host context to
handle the hcall.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210528090752.3542186-14-npiggin@gmail.com
2021-05-28 19:07:33 +10:00
|
|
|
|
|
|
|
|
|
KVM: PPC: Allow book3s_hv guests to use SMT processor modes
This lifts the restriction that book3s_hv guests can only run one
hardware thread per core, and allows them to use up to 4 threads
per core on POWER7. The host still has to run single-threaded.
This capability is advertised to qemu through a new KVM_CAP_PPC_SMT
capability. The return value of the ioctl querying this capability
is the number of vcpus per virtual CPU core (vcore), currently 4.
To use this, the host kernel should be booted with all threads
active, and then all the secondary threads should be offlined.
This will put the secondary threads into nap mode. KVM will then
wake them from nap mode and use them for running guest code (while
they are still offline). To wake the secondary threads, we send
them an IPI using a new xics_wake_cpu() function, implemented in
arch/powerpc/sysdev/xics/icp-native.c. In other words, at this stage
we assume that the platform has a XICS interrupt controller and
we are using icp-native.c to drive it. Since the woken thread will
need to acknowledge and clear the IPI, we also export the base
physical address of the XICS registers using kvmppc_set_xics_phys()
for use in the low-level KVM book3s code.
When a vcpu is created, it is assigned to a virtual CPU core.
The vcore number is obtained by dividing the vcpu number by the
number of threads per core in the host. This number is exported
to userspace via the KVM_CAP_PPC_SMT capability. If qemu wishes
to run the guest in single-threaded mode, it should make all vcpu
numbers be multiples of the number of threads per core.
We distinguish three states of a vcpu: runnable (i.e., ready to execute
the guest), blocked (that is, idle), and busy in host. We currently
implement a policy that the vcore can run only when all its threads
are runnable or blocked. This way, if a vcpu needs to execute elsewhere
in the kernel or in qemu, it can do so without being starved of CPU
by the other vcpus.
When a vcore starts to run, it executes in the context of one of the
vcpu threads. The other vcpu threads all go to sleep and stay asleep
until something happens requiring the vcpu thread to return to qemu,
or to wake up to run the vcore (this can happen when another vcpu
thread goes from busy in host state to blocked).
It can happen that a vcpu goes from blocked to runnable state (e.g.
because of an interrupt), and the vcore it belongs to is already
running. In that case it can start to run immediately as long as
the none of the vcpus in the vcore have started to exit the guest.
We send the next free thread in the vcore an IPI to get it to start
to execute the guest. It synchronizes with the other threads via
the vcore->entry_exit_count field to make sure that it doesn't go
into the guest if the other vcpus are exiting by the time that it
is ready to actually enter the guest.
Note that there is no fixed relationship between the hardware thread
number and the vcpu number. Hardware threads are assigned to vcpus
as they become runnable, so we will always use the lower-numbered
hardware threads in preference to higher-numbered threads if not all
the vcpus in the vcore are runnable, regardless of which vcpus are
runnable.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Alexander Graf <agraf@suse.de>
2011-06-29 00:23:08 +00:00
|
|
|
|
|
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-07-01 15:27:32 +02:00
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
2021-07-01 15:27:32 +02:00
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-10-08 16:30:55 +11:00
|
|
|
|
2021-05-28 19:07:28 +10:00
|
|
|
|
2022-03-03 15:33:12 +10:00
|
|
|
|
2019-04-18 12:39:27 +02:00
|
|
|
|
2019-04-18 12:39:28 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-04-18 12:39:35 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-08-26 16:21:21 +10:00
|
|
|
|
2019-04-18 12:39:27 +02:00
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2018-10-08 16:30:55 +11:00
|
|
|
|
2021-05-28 19:07:28 +10:00
|
|
|
|
2022-03-03 15:33:12 +10:00
|
|
|
|
2019-04-18 12:39:27 +02:00
|
|
|
|
2019-04-18 12:39:28 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-04-18 12:39:35 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-04-18 12:39:27 +02:00
|
|
|
|
2017-04-05 17:54:56 +10:00
|
|
|
|
|
|
|
|
|
2019-02-25 14:35:06 +11:00
|
|
|
|
KVM: PPC: Book3S: Allow XICS emulation to work in nested hosts using XIVE
Currently, the KVM code assumes that if the host kernel is using the
XIVE interrupt controller (the new interrupt controller that first
appeared in POWER9 systems), then the in-kernel XICS emulation will
use the XIVE hardware to deliver interrupts to the guest. However,
this only works when the host is running in hypervisor mode and has
full access to all of the XIVE functionality. It doesn't work in any
nested virtualization scenario, either with PR KVM or nested-HV KVM,
because the XICS-on-XIVE code calls directly into the native-XIVE
routines, which are not initialized and cannot function correctly
because they use OPAL calls, and OPAL is not available in a guest.
This means that using the in-kernel XICS emulation in a nested
hypervisor that is using XIVE as its interrupt controller will cause a
(nested) host kernel crash. To fix this, we change most of the places
where the current code calls xive_enabled() to select between the
XICS-on-XIVE emulation and the plain XICS emulation to call a new
function, xics_on_xive(), which returns false in a guest.
However, there is a further twist. The plain XICS emulation has some
functions which are used in real mode and access the underlying XICS
controller (the interrupt controller of the host) directly. In the
case of a nested hypervisor, this means doing XICS hypercalls
directly. When the nested host is using XIVE as its interrupt
controller, these hypercalls will fail. Therefore this also adds
checks in the places where the XICS emulation wants to access the
underlying interrupt controller directly, and if that is XIVE, makes
the code use the virtual mode fallback paths, which call generic
kernel infrastructure rather than doing direct XICS access.
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
2019-02-04 22:07:20 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-05-28 19:07:44 +10:00
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
KVM: PPC: Book3S HV: Simplify machine check handling
This makes the handling of machine check interrupts that occur inside
a guest simpler and more robust, with less done in assembler code and
in real mode.
Now, when a machine check occurs inside a guest, we always get the
machine check event struct and put a copy in the vcpu struct for the
vcpu where the machine check occurred. We no longer call
machine_check_queue_event() from kvmppc_realmode_mc_power7(), because
on POWER8, when a vcpu is running on an offline secondary thread and
we call machine_check_queue_event(), that calls irq_work_queue(),
which doesn't work because the CPU is offline, but instead triggers
the WARN_ON(lazy_irq_pending()) in pnv_smp_cpu_kill_self() (which
fires again and again because nothing clears the condition).
All that machine_check_queue_event() actually does is to cause the
event to be printed to the console. For a machine check occurring in
the guest, we now print the event in kvmppc_handle_exit_hv()
instead.
The assembly code at label machine_check_realmode now just calls C
code and then continues exiting the guest. We no longer either
synthesize a machine check for the guest in assembly code or return
to the guest without a machine check.
The code in kvmppc_handle_exit_hv() is extended to handle the case
where the guest is not FWNMI-capable. In that case we now always
synthesize a machine check interrupt for the guest. Previously, if
the host thinks it has recovered the machine check fully, it would
return to the guest without any notification that the machine check
had occurred. If the machine check was caused by some action of the
guest (such as creating duplicate SLB entries), it is much better to
tell the guest that it has caused a problem. Therefore we now always
generate a machine check interrupt for guests that are not
FWNMI-capable.
Reviewed-by: Aravinda Prasad <aravinda@linux.vnet.ibm.com>
Reviewed-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
2019-02-21 13:38:49 +11:00
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-11-23 19:52:31 +10:00
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2021-04-12 11:48:40 +10:00
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2019-03-22 17:05:45 +11:00
|
|
|
|
|
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
|
|
|
|
|
2018-10-08 16:30:50 +11:00
|
|
|
|
2016-12-01 14:03:46 +11:00
|
|
|
|
2015-12-17 14:59:06 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-17 17:01:40 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-01-04 18:12:48 +01:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-12 14:08:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-12 14:08:47 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-12 14:08:46 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-12 14:08:47 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-12 14:08:46 +00:00
|
|
|
|
|
|
|
|
|
2011-08-18 15:25:21 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2011-12-20 15:34:20 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
kvm: rename pfn_t to kvm_pfn_t
To date, we have implemented two I/O usage models for persistent memory,
PMEM (a persistent "ram disk") and DAX (mmap persistent memory into
userspace). This series adds a third, DAX-GUP, that allows DAX mappings
to be the target of direct-i/o. It allows userspace to coordinate
DMA/RDMA from/to persistent memory.
The implementation leverages the ZONE_DEVICE mm-zone that went into
4.3-rc1 (also discussed at kernel summit) to flag pages that are owned
and dynamically mapped by a device driver. The pmem driver, after
mapping a persistent memory range into the system memmap via
devm_memremap_pages(), arranges for DAX to distinguish pfn-only versus
page-backed pmem-pfns via flags in the new pfn_t type.
The DAX code, upon seeing a PFN_DEV+PFN_MAP flagged pfn, flags the
resulting pte(s) inserted into the process page tables with a new
_PAGE_DEVMAP flag. Later, when get_user_pages() is walking ptes it keys
off _PAGE_DEVMAP to pin the device hosting the page range active.
Finally, get_page() and put_page() are modified to take references
against the device driver established page mapping.
Finally, this need for "struct page" for persistent memory requires
memory capacity to store the memmap array. Given the memmap array for a
large pool of persistent may exhaust available DRAM introduce a
mechanism to allocate the memmap from persistent memory. The new
"struct vmem_altmap *" parameter to devm_memremap_pages() enables
arch_add_memory() to use reserved pmem capacity rather than the page
allocator.
This patch (of 18):
The core has developed a need for a "pfn_t" type [1]. Move the existing
pfn_t in KVM to kvm_pfn_t [2].
[1]: https://lists.01.org/pipermail/linux-nvdimm/2015-September/002199.html
[2]: https://lists.01.org/pipermail/linux-nvdimm/2015-September/002218.html
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-01-15 16:56:11 -08:00
|
|
|
|
2012-08-03 13:56:33 +02:00
|
|
|
|
2023-08-02 16:13:49 +01:00
|
|
|
|
2013-04-25 06:33:57 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-08-02 16:13:49 +01:00
|
|
|
|
2025-08-05 18:22:51 +01:00
|
|
|
|
2023-08-02 16:13:49 +01:00
|
|
|
|
2025-08-05 18:22:51 +01:00
|
|
|
|
2012-08-03 13:56:33 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:52 +10:00
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
2014-07-30 15:03:56 +05:30
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:52 +10:00
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
2014-07-30 15:03:56 +05:30
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2023-10-11 16:37:06 +11:00
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2023-10-11 16:37:06 +11:00
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-10-11 16:37:06 +11:00
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2023-10-11 16:37:06 +11:00
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2023-09-14 13:05:52 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
2023-09-14 13:05:52 +10:00
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
2014-07-17 17:01:35 +05:30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2023-09-14 13:05:59 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-07-10 17:47:39 -05:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-08-13 01:04:19 +02:00
|
|
|
|
2013-07-10 17:47:39 -05:00
|
|
|
|
|
|
|
|
|
2012-08-13 01:04:19 +02:00
|
|
|
|
2014-01-09 19:18:40 -06:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-08-13 01:04:19 +02:00
|
|
|
|
|
|
|
|
|
2017-12-20 09:25:50 +05:30
|
|
|
|
2012-08-13 01:04:19 +02:00
|
|
|
|
|
|
|
|
|
2012-08-03 13:56:33 +02:00
|
|
|
|
2022-11-27 22:49:26 +10:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-10-11 06:13:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-10-11 06:13:23 +00:00
|
|
|
|
2012-10-11 06:13:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2012-10-11 06:13:23 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2014-04-24 13:46:24 +02:00
|
|
|
|
2012-10-11 06:13:23 +00:00
|
|
|
|
|
|
|
|
|
2012-10-11 06:13:22 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2013-04-17 20:30:50 +00:00
|
|
|
|
|
|
|
|
|
2008-04-16 23:28:09 -05:00
|
|
|
|