We support THP only with book3s_64 and 64K page size. Move
THP details to hash64-64k.h to clarify the same.
Acked-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
In LE kernel, we currently have a hack for kexec that resets the exception
endian before starting a new kernel as the kernel that is loaded could be a
big endian or a little endian kernel. In kdump case, resetting exception
endian fails when one or more cpus is disabled. But we can ignore the failure
and still go ahead, as in most cases crashkernel will be of same endianess
as primary kernel and reseting endianess is not even needed in those cases.
This patch adds a new inline function to say if this is kdump path. This
function is used at places where such a check is needed.
Signed-off-by: Hari Bathini <hbathini@linux.vnet.ibm.com>
[mpe: Rename to kdump_in_progress(), use bool, and edit comment]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
upatepp can get called for a nohpte fault when we find from the linux
page table that the translation was hashed before. In that case
we are sure that there is no existing translation, hence we could
avoid doing tlbie.
We could possibly race with a parallel fault filling the TLB. But
that should be ok because updatepp is only ever relaxing permissions.
We also look at linux pte permission bits when filling hash pte
permission bits. We also hold the linux pte busy bits while
inserting/updating a hashpte entry, hence a paralle update of
linux pte is not possible. On the other hand mprotect involves
ptep_modify_prot_start which cause a hpte invalidate and not updatepp.
Performance number:
We use randbox_access_bench written by Anton.
Kernel with THP disabled and smaller hash page table size.
86.60% random_access_b [kernel.kallsyms] [k] .native_hpte_updatepp
2.10% random_access_b random_access_bench [.] doit
1.99% random_access_b [kernel.kallsyms] [k] .do_raw_spin_lock
1.85% random_access_b [kernel.kallsyms] [k] .native_hpte_insert
1.26% random_access_b [kernel.kallsyms] [k] .native_flush_hash_range
1.18% random_access_b [kernel.kallsyms] [k] .__delay
0.69% random_access_b [kernel.kallsyms] [k] .native_hpte_remove
0.37% random_access_b [kernel.kallsyms] [k] .clear_user_page
0.34% random_access_b [kernel.kallsyms] [k] .__hash_page_64K
0.32% random_access_b [kernel.kallsyms] [k] fast_exception_return
0.30% random_access_b [kernel.kallsyms] [k] .hash_page_mm
With Fix:
27.54% random_access_b random_access_bench [.] doit
22.90% random_access_b [kernel.kallsyms] [k] .native_hpte_insert
5.76% random_access_b [kernel.kallsyms] [k] .native_hpte_remove
5.20% random_access_b [kernel.kallsyms] [k] fast_exception_return
5.12% random_access_b [kernel.kallsyms] [k] .__hash_page_64K
4.80% random_access_b [kernel.kallsyms] [k] .hash_page_mm
3.31% random_access_b [kernel.kallsyms] [k] data_access_common
1.84% random_access_b [kernel.kallsyms] [k] .trace_hardirqs_on_caller
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
If we know that user address space has never executed on other cpus
we could use tlbiel.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This still has not been merged and now powerpc is the only arch that does
not have this change. Sorry about missing linuxppc-dev before.
V2->V2
- Fix up to work against 3.18-rc1
__get_cpu_var() is used for multiple purposes in the kernel source. One of
them is address calculation via the form &__get_cpu_var(x). This calculates
the address for the instance of the percpu variable of the current processor
based on an offset.
Other use cases are for storing and retrieving data from the current
processors percpu area. __get_cpu_var() can be used as an lvalue when
writing data or on the right side of an assignment.
__get_cpu_var() is defined as :
__get_cpu_var() always only does an address determination. However, store
and retrieve operations could use a segment prefix (or global register on
other platforms) to avoid the address calculation.
this_cpu_write() and this_cpu_read() can directly take an offset into a
percpu area and use optimized assembly code to read and write per cpu
variables.
This patch converts __get_cpu_var into either an explicit address
calculation using this_cpu_ptr() or into a use of this_cpu operations that
use the offset. Thereby address calculations are avoided and less registers
are used when code is generated.
At the end of the patch set all uses of __get_cpu_var have been removed so
the macro is removed too.
The patch set includes passes over all arches as well. Once these operations
are used throughout then specialized macros can be defined in non -x86
arches as well in order to optimize per cpu access by f.e. using a global
register that may be set to the per cpu base.
Transformations done to __get_cpu_var()
1. Determine the address of the percpu instance of the current processor.
DEFINE_PER_CPU(int, y);
int *x = &__get_cpu_var(y);
Converts to
int *x = this_cpu_ptr(&y);
2. Same as #1 but this time an array structure is involved.
DEFINE_PER_CPU(int, y[20]);
int *x = __get_cpu_var(y);
Converts to
int *x = this_cpu_ptr(y);
3. Retrieve the content of the current processors instance of a per cpu
variable.
DEFINE_PER_CPU(int, y);
int x = __get_cpu_var(y)
Converts to
int x = __this_cpu_read(y);
4. Retrieve the content of a percpu struct
DEFINE_PER_CPU(struct mystruct, y);
struct mystruct x = __get_cpu_var(y);
Converts to
memcpy(&x, this_cpu_ptr(&y), sizeof(x));
5. Assignment to a per cpu variable
DEFINE_PER_CPU(int, y)
__get_cpu_var(y) = x;
Converts to
__this_cpu_write(y, x);
6. Increment/Decrement etc of a per cpu variable
DEFINE_PER_CPU(int, y);
__get_cpu_var(y)++
Converts to
__this_cpu_inc(y)
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: Paul Mackerras <paulus@samba.org>
Signed-off-by: Christoph Lameter <cl@linux.com>
[mpe: Fix build errors caused by set/or_softirq_pending(), and rework
assignment in __set_breakpoint() to use memcpy().]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
CONFIG_JUMP_LABEL doesn't ensure HAVE_JUMP_LABEL, if it
is not the case use maintainers's own mutex to guard
the modification of global values.
Signed-off-by: Zhouyi Zhou <yizhouzhou@ict.ac.cn>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
There were a number of prototypes for functions that no longer
exist. Remove them.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
The segment identifier and segment size will remain the same in
the loop, So we can compute it outside. We also change the
hugepage_invalidate interface so that we can use it the later patch
CC: <stable@vger.kernel.org>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
hcall tracepoints add quite a few instructions to our hcall path:
plpar_hcall:
mr r2,r2
mfcr r0
stw r0,8(r1)
b 164 <---- start
ld r12,0(r2)
std r12,32(r1)
cmpdi r12,0
beq 164 <---- end
...
We have an unconditional branch that gets noped out during boot and
a load/compare/branch. We also store the tracepoint value to the
stack for the hcall_exit path to use.
By using jump labels we can simplify this to just a single nop that
gets replaced with a branch when the tracepoint is enabled:
plpar_hcall:
mr r2,r2
mfcr r0
stw r0,8(r1)
nop <----
...
If jump labels are not enabled, we fall back to the old method.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Even though we have same value for linux PTE bits and hash PTE pits
use the hash pte bits wen updating hash pte
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The only external user of slb_shadow is the pseries lpar code, and it
can access through the paca array instead.
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On little endian builds call H_SET_MODE so exceptions have the
correct endianness. We need to reset the endian during kexec
so do that in the MMU hashtable clear callback.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The spec says it "may be problematic" if CPU x registers the VPA of
CPU y. Add a warning in case we ever do that.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
As a part of pseries_idle backend driver cleanup to make
the code common to both pseries and powernv platforms, it
is necessary to move the backend-driver code to drivers/cpuidle.
As a pre-requisite for that, it is essential to move plpar_wrapper.h
to include/asm.
Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The lppaca, slb_shadow and dtl_entry hypervisor structures are
big endian, so we have to byte swap them in little endian builds.
LE KVM hosts will also need to be fixed but for now add an #error
to remind us.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Address some of the trivial sparse warnings in arch/powerpc.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Commit 801eb73f45 introduced
a bug while checking PTE flags. We have to drop the _PAGE_COHERENT flag
when __PAGE_NO_CACHE is set and the cache update policy is not write-through
(i.e. _PAGE_WRITETHRU is not set)
Signed-off-by: Denis Kirjanov <kda@linux-powerpc.org>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
CC: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On LPAR systems we need to inform the hypervisor that we are using the
EBB registers. We do this by setting a bit in the Virtual Processor Area
(VPA) - formerly known as the lppaca.
For now we do this always, ie. we do not dynamically enable/disable.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Hugepage invalidate involves invalidating multiple hpte entries.
Optimize the operation using H_BULK_REMOVE on lpar platforms.
On native, reduce the number of tlb flush.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
If a hash bucket gets full, we "evict" a more/less random entry from it.
When we do that we don't invalidate the TLB (hpte_remove) because we assume
the old translation is still technically "valid". This implies that when
we are invalidating or updating pte, even if HPTE entry is not valid
we should do a tlb invalidate. With hugepages, we need to pass the correct
actual page size value for tlb invalidation.
This change update the patch 0608d69246
"powerpc/mm: Always invalidate tlb on hpte invalidate and update" to handle
transparent hugepages correctly.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We look at both the segment base page size and actual page size and store
the pte-lp-encodings in an array per base page size.
We also update all relevant functions to take actual page size argument
so that we can use the correct PTE LP encoding in HPTE. This should also
get the basic Multiple Page Size per Segment (MPSS) support. This is needed
to enable THP on ppc64.
[Fixed PR KVM build --BenH]
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
PAPR defines these errors as negative values. So print them accordingly
for easy debugging.
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Some versions of pHyp will perform the adjunct partition test before the
ANDCOND test. The result of this is that H_RESOURCE can be returned and
cause the BUG_ON condition to occur. The HPTE is not removed. So add a
check for H_RESOURCE, it is ok if this HPTE is not removed as
pSeries_lpar_hpte_remove is looking for an HPTE to remove and not a
specific HPTE to remove. So it is ok to just move on to the next slot
and try again.
Cc: stable@vger.kernel.org
Signed-off-by: Michael Wolf <mjw@linux.vnet.ibm.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
This patch convert different functions to take virtual page number
instead of virtual address. Virtual page number is virtual address
shifted right by VPN_SHIFT (12) bits. This enable us to have an
address range of upto 76 bits.
Reviewed-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
It's empty now, apart from other includes.
Fixup a few files that were getting things via this header.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This is no longer selectable, so just remove all the dependent code.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Tracepoints should not be called inside an rcu_idle_enter/rcu_idle_exit
region. Since pSeries calls H_CEDE in the idle loop, we were violating
this rule.
commit a7b152d534 (powerpc: Tell RCU about idle after hcall tracing)
tried to work around it by delaying the rcu_idle_enter until after we
called the hcall tracepoint, but there are a number of issues with it.
The hcall tracepoint trampoline code is called conditionally when the
tracepoint is enabled. If the tracepoint is not enabled we never call
rcu_idle_enter. The idle_uses_rcu check was also done at compile time
which breaks multiplatform builds.
The simple fix is to avoid tracing H_CEDE and rely on other tracepoints
and the hypervisor dispatch trace log to work out if we called H_CEDE.
This fixes a hang during boot on pSeries.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc: (185 commits)
powerpc: fix compile error with 85xx/p1010rdb.c
powerpc: fix compile error with 85xx/p1023_rds.c
powerpc/fsl: add MSI support for the Freescale hypervisor
arch/powerpc/sysdev/fsl_rmu.c: introduce missing kfree
powerpc/fsl: Add support for Integrated Flash Controller
powerpc/fsl: update compatiable on fsl 16550 uart nodes
powerpc/85xx: fix PCI and localbus properties in p1022ds.dts
powerpc/85xx: re-enable ePAPR byte channel driver in corenet32_smp_defconfig
powerpc/fsl: Update defconfigs to enable some standard FSL HW features
powerpc: Add TBI PHY node to first MDIO bus
sbc834x: put full compat string in board match check
powerpc/fsl-pci: Allow 64-bit PCIe devices to DMA to any memory address
powerpc: Fix unpaired probe_hcall_entry and probe_hcall_exit
offb: Fix setting of the pseudo-palette for >8bpp
offb: Add palette hack for qemu "standard vga" framebuffer
offb: Fix bug in calculating requested vram size
powerpc/boot: Change the WARN to INFO for boot wrapper overlap message
powerpc/44x: Fix build error on currituck platform
powerpc/boot: Change the load address for the wrapper to fit the kernel
powerpc/44x: Enable CRASH_DUMP for 440x
...
Fix up a trivial conflict in arch/powerpc/include/asm/cputime.h due to
the additional sparse-checking code for cputime_t.
Unpaired calling of probe_hcall_entry and probe_hcall_exit might happen
as following, which could cause incorrect preempt count.
__trace_hcall_entry => trace_hcall_entry -> probe_hcall_entry =>
get_cpu_var => preempt_disable
__trace_hcall_exit => trace_hcall_exit -> probe_hcall_exit =>
put_cpu_var => preempt_enable
where:
A => B and A -> B means A calls B, but
=> means A will call B through function name, and B will definitely be
called.
-> means A will call B through function pointer, so B might not be
called if the function pointer is not set.
So error happens when only one of probe_hcall_entry and probe_hcall_exit
get called during a hcall.
This patch tries to move the preempt count operations from
probe_hcall_entry and probe_hcall_exit to its callers.
Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Tested-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
CC: stable@kernel.org [v2.6.32+]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The PowerPC pSeries platform (CONFIG_PPC_PSERIES=y) enables
hypervisor-call tracing for CONFIG_TRACEPOINTS=y kernels. One of the
hypervisor calls that is traced is the H_CEDE call in the idle loop
that tells the hypervisor that this OS instance no longer needs the
current CPU. However, tracing uses RCU, so this combination of kernel
configuration variables needs to avoid telling RCU about the current CPU's
idleness until after the H_CEDE-entry tracing completes on the one hand,
and must tell RCU that the the current CPU is no longer idle before the
H_CEDE-exit tracing starts.
In all other cases, it suffices to inform RCU of CPU idleness upon
idle-loop entry and exit.
This commit makes the required adjustments.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
With module.h being implicitly everywhere via device.h, the absence
of explicitly including something for EXPORT_SYMBOL went unnoticed.
Since we are heading to fix things up and clean module.h from the
device.h file, we need to explicitly include these files now.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Make the VPA, SLB shadow and DTL registration and deregistration
functions print consistent messages on error. I needed the firmware
error code while chasing a kexec bug but we weren't printing it.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On a box with 8TB of RAM the MMU hashtable is 64GB in size. That
means we have 4G PTEs. pSeries_lpar_hptab_clear was using a signed
int to store the index which will overflow at 2G.
Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: <stable@kernel.org>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
On pseries machines, consoles are provided by the hypervisor using
a low level get_chars/put_chars type interface. However, this is
really just a transport to the service processor which implements
them either as "raw" console (networked consoles, HMC, ...) or as
"hvsi" serial ports.
The later is a simple packet protocol on top of the raw character
interface that is supposed to convey additional "serial port" style
semantics. In practice however, all it does is provide a way to
read the CD line and set/clear our DTR line, that's it.
We currently implement the "raw" protocol as an hvc console backend
(/dev/hvcN) and the "hvsi" protocol using a separate tty driver
(/dev/hvsi0).
However this is quite impractical. The arbitrary difference between
the two type of devices has been a major source of user (and distro)
confusion. Additionally, there's an additional mini -hvsi implementation
in the pseries platform code for our low level debug console and early
boot kernel messages, which means code duplication, though that low
level variant is impractical as it's incapable of doing the initial
protocol negociation to establish the link to the FSP.
This essentially replaces the dedicated hvsi driver and the platform
udbg code completely by extending the existing hvc_vio backend used
in "raw" mode so that:
- It now supports HVSI as well
- We add support for hvc backend providing tiocm{get,set}
- It also provides a udbg interface for early debug and boot console
This is overall less code, though this will only be obvious once we
remove the old "hvsi" driver, which is still available for now. When
the old driver is enabled, the new code still kicks in for the low
level udbg console, replacing the old mini implementation in the platform
code, it just doesn't provide the higher level "hvc" interface.
In addition to producing generally simler code, this has several benefits
over our current situation:
- The user/distro only has to deal with /dev/hvcN for the hypervisor
console, avoiding all sort of confusion that has plagued us in the past
- The tty, kernel and low level debug console all use the same code
base which supports the full protocol establishment process, thus the
console is now available much earlier than it used to be with the
old HVSI driver. The kernel console works much earlier and udbg is
available much earlier too. Hackers can enable a hard coded very-early
debug console as well that works with HVSI (previously that was only
supported for the "raw" mode).
I've tried to keep the same semantics as hvsi relative to how I react
to things like CD changes, with some subtle differences though:
- I clear DTR on close if HUPCL is set
- Current hvsi triggers a hangup if it detects a up->down transition
on CD (you can still open a console with CD down). My new implementation
triggers a hangup if the link to the FSP is severed, and severs it upon
detecting a up->down transition on CD.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
When CONFIG_PPC_EARLY_DEBUG is set, call register_early_udbg_console()
early from generic code.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Adds support for page coalescing, which is a feature on IBM Power servers
which allows for coalescing identical pages between logical partitions.
Hint text pages as coalesce candidates, since they are the most likely
pages to be able to be coalesced between partitions. This patch also
exports some page coalescing statistics available from firmware via
lparcfg.
[BenH: Moved a couple of things around to fix compile problems]
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Some of the 64bit PPC CPU features are MMU-related, so this patch moves
them to MMU_FTR_ bits. All cpu_has_feature()-style tests are moved to
mmu_has_feature(), and seven feature bits are freed as a result.
Signed-off-by: Matt Evans <matt@ozlabs.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Spinlocks on shared processor partitions use H_YIELD to notify the
hypervisor we are waiting on another virtual CPU. Unfortunately this means
the hcall tracepoints can recurse.
The patch below adds a percpu depth and checks it on both the entry and
exit hcall tracepoints.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
CC: stable@kernel.org
This introduces a pair of kernel parameters that can be used to disable
the MULTITCE and BULK_REMOVE h-calls.
By default, those hcalls are enabled, active, and good for throughput
and performance. The ability to disable them will be useful for some of
the PREEMPT_RT related investigation and work occurring on Power.
Signed-off-by: Will Schmidt <will_schmidt@vnet.ibm.com>
cc: Olof Johansson <olof@lixom.net>
cc: Anton Blanchard <anton@samba.org>
cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently, when CONFIG_VIRT_CPU_ACCOUNTING is enabled, we use the
PURR register for measuring the user and system time used by
processes, as well as other related times such as hardirq and
softirq times. This turns out to be quite confusing for users
because it means that a program will often be measured as taking
less time when run on a multi-threaded processor (SMT2 or SMT4 mode)
than it does when run on a single-threaded processor (ST mode), even
though the program takes longer to finish. The discrepancy is
accounted for as stolen time, which is also confusing, particularly
when there are no other partitions running.
This changes the accounting to use the timebase instead, meaning that
the reported user and system times are the actual number of real-time
seconds that the program was executing on the processor thread,
regardless of which SMT mode the processor is in. Thus a program will
generally show greater user and system times when run on a
multi-threaded processor than on a single-threaded processor.
On pSeries systems on POWER5 or later processors, we measure the
stolen time (time when this partition wasn't running) using the
hypervisor dispatch trace log. We check for new entries in the
log on every entry from user mode and on every transition from
kernel process context to soft or hard IRQ context (i.e. when
account_system_vtime() gets called). So that we can correctly
distinguish time stolen from user time and time stolen from system
time, without having to check the log on every exit to user mode,
we store separate timestamps for exit to user mode and entry from
user mode.
On systems that have a SPURR (POWER6 and POWER7), we read the SPURR
in account_system_vtime() (as before), and then apportion the SPURR
ticks since the last time we read it between scaled user time and
scaled system time according to the relative proportions of user
time and system time over the same interval. This avoids having to
read the SPURR on every kernel entry and exit. On systems that have
PURR but not SPURR (i.e., POWER5), we do the same using the PURR
rather than the SPURR.
This disables the DTL user interface in /sys/debug/kernel/powerpc/dtl
for now since it conflicts with the use of the dispatch trace log
by the time accounting code.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently we have the lppaca structs as a simple array of NR_CPUS
entries, taking up space in the data section of the kernel image.
In future we would like to allocate them dynamically, so this
abstracts out the accesses to the array, making it easier to
change how we locate the lppaca for a given cpu in future.
Specifically, lppaca[cpu] changes to lppaca_of(cpu).
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently for kexec the PTE tear down on 1TB segment systems normally
requires 3 hcalls for each PTE removal. On a machine with 32GB of
memory it can take around a minute to remove all the PTEs.
This optimises the path so that we only remove PTEs that are valid.
It also uses the read 4 PTEs at once HCALL. For the common case where
a PTEs is invalid in a 1TB segment, this turns the 3 HCALLs per PTE
down to 1 HCALL per 4 PTEs.
This gives an > 10x speedup in kexec times on PHYP, taking a 32GB
machine from around 1 minute down to a few seconds.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
While most users of the hcall tracepoints will only want the opcode
and return code, some will want all the arguments. To avoid the
complexity of using varargs we pass a pointer to the register save
area, which contains all the arguments.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Add hcall_entry and hcall_exit tracepoints. This replaces the inline
assembly HCALL_STATS code and converts it to use the new tracepoints.
To keep the disabled case as quick as possible, we embed a status word
in the TOC so we can get at it with a single load. By doing so we
keep the overhead at a minimum. Time taken for a null hcall:
No tracepoint code: 135.79 cycles
Disabled tracepoints: 137.95 cycles
For reference, before this patch enabling HCALL_STATS resulted in a null
hcall of 201.44 cycles!
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
pr_debug() can now result in code being generated even when DEBUG
is not defined. That's not really desirable in some places.
In particular, pSeries_lpar_hpte_insert() goes from 185 instructions
to 77 instructions as a result of this patch. Luckily that code
isn't called very often ...
With CONFIG_DYNAMIC_DEBUG=y:
size before:
text data bss dec hex filename
7284 1552 296 9132 23ac platforms/pseries/lpar.o
size after:
text data bss dec hex filename
5806 1096 296 7198 1c1e platforms/pseries/lpar.o
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Adds support for the "unused" page hint which can be used in shared
memory partitions to flag pages not in use, which will then be stolen
before active pages by the hypervisor when memory needs to be moved to
LPARs in need of additional memory. Failure to mark pages as 'unused'
makes the LPAR slower to give up unused memory to other partitions.
This adds the kernel parameter 'cmo_free_hint' to disable this
functionality.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
It is okay for both _PAGE_GUARDED and _PAGE_COHERENT (G and M) to be set
in the same pte. In fact, even if that were not the case, there doesn't
seem to be any place where G is set without also setting I (_PAGE_NO_CACHE),
so the test for I is sufficient as a condition to clear _PAGE_COHERENT
when filling the hash table.
Signed-off-by: Dave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
