Now we use printf style alignment there is no need to manually space
these fields.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Right now we allocate a cacheline sized NR_CPUS array for xics IPI
communication. Use DECLARE_PER_CPU_SHARED_ALIGNED to put it in percpu
data in its own cacheline since it is written to by other cpus.
On a kernel with NR_CPUS=1024, this saves quite a lot of memory:
text data bss dec hex filename
8767779 2944260 1505724 13217763 c9afe3 vmlinux.irq_cpustat
8767555 2813444 1505724 13086723 c7b003 vmlinux.xics
A saving of around 128kB.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Frans Pop <elendil@planet.nl>
Cc: linuxppc-dev@ozlabs.org
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The code to track the CPPR values added by commit
49bd364713 ("powerpc/pseries: Track previous
CPPR values to correctly EOI interrupts") broke kexec on pseries because
the kexec code in xics.c calls xics_set_cpu_priority() before the IPI has
been EOI'ed. This wasn't a problem previously but it now triggers a BUG_ON
in xics_set_cpu_priority() because os_cppr->index isn't 0.
Fix this problem by setting the index on the CPPR stack to 0 before calling
xics_set_cpu_priority() in xics_teardown_cpu().
Also make it clear that we only want to set the priority when there's just
one CPPR value in the stack, and enforce it by updating the value of
os_cppr->stack[0] rather than os_cppr->stack[os_cppr->index].
While we're at it change the BUG_ON to a WARN_ON.
Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Mark Nelson <markn@au1.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
desc->affinity doesn't exit in that case. Let's use a macro for
the UP variant of get_irq_server(), it's the easiest way, avoids
evaluating arguments.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Commit 57b150cce8 (irq: only update affinity if
->set_affinity() is sucessfull) broke xics irq affinity.
We need to use the cpumask passed in, instead of accessing ->affinity directly.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Convert locks which cannot be sleeping locks in preempt-rt to
raw_spinlocks.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Ingo Molnar <mingo@elte.hu>
At the moment when we EOI an interrupt we set the CPPR back to 0xFF
regardless of its previous value. This could lead to problems if we
take an interrupt with a priority of 5, but before EOIing it we get
an IPI which has a priority of 4. The problem is that at the moment
when we EOI the IPI we will set the CPPR to 0xFF, but it should
really be set back to 5 (the previous priority).
To keep track of the previous CPPR values we create the xics_cppr
structure that has an array for CPPR values and an index pointing
to the current priority. This can easily grow if new priorities get
added in the future.
This will also be useful because the partition adjunct option of
upcoming machines will update the H_XIRR hcall to accept the CPPR
as a parameter.
Signed-off-by: Mark Nelson <markn@au1.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Writing a driver using SCLPC on the MPC5200B I detected, that the
intspec arrays to map irqs to Linux virq cannot be const, because the
mapping and xlate functions only take non const pointers. All those
functions do not modify the intspec, so a const pointer could be used.
Signed-off-by: Roman Fietze <roman.fietze@telemotive.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The typename member of struct irq_chip was kept for migration purposes
and is obsolete since more than 2 years. Fix up the leftovers.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: linuxppc-dev@ozlabs.org
Acked-by: Geoff Levand <geoffrey.levand@am.sony.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Patch f598282f51 exposed a problem in
powerpc MSI-X functionality, making network interfaces such as ixgbe
and cxgb3 stop to work when MSI-X is enabled. RX interrupts were not
being generated.
The problem was caused because MSI irq was not being effectively
unmasked after device initialization.
Signed-off-by: Andre Detsch <adetsch@br.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
get_irq_desc() is a powerpc-specific version of irq_to_desc(). That
is reason enough to remove it, but it also doesn't know about sparse
irq_desc support which irq_to_desc() does (when we enable it).
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
pr_debug() can now result in code being generated even when DEBUG
is not defined. That's not really desirable in some places.
With CONFIG_DYNAMIC_DEBUG=y:
size before:
text data bss dec hex filename
7720 5488 296 13504 34c0 platforms/pseries/xics.o
size after:
text data bss dec hex filename
7535 5456 296 13287 33e7 platforms/pseries/xics.o
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
according to Ingo, change set_affinity() in irq_chip should return int,
because that way we can handle failure cases in a much cleaner way, in
the genirq layer.
v2: fix two typos
[ Impact: extend API ]
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: linux-arch@vger.kernel.org
LKML-Reference: <49F654E9.4070809@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Impact: cleanup, update to new cpumask API
Irq_desc.affinity and irq_desc.pending_mask are now cpumask_var_t's
so access to them should be using the new cpumask API.
Signed-off-by: Mike Travis <travis@sgi.com>
* 'cpus4096-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (66 commits)
x86: export vector_used_by_percpu_irq
x86: use logical apicid in x2apic_cluster's x2apic_cpu_mask_to_apicid_and()
sched: nominate preferred wakeup cpu, fix
x86: fix lguest used_vectors breakage, -v2
x86: fix warning in arch/x86/kernel/io_apic.c
sched: fix warning in kernel/sched.c
sched: move test_sd_parent() to an SMP section of sched.h
sched: add SD_BALANCE_NEWIDLE at MC and CPU level for sched_mc>0
sched: activate active load balancing in new idle cpus
sched: bias task wakeups to preferred semi-idle packages
sched: nominate preferred wakeup cpu
sched: favour lower logical cpu number for sched_mc balance
sched: framework for sched_mc/smt_power_savings=N
sched: convert BALANCE_FOR_xx_POWER to inline functions
x86: use possible_cpus=NUM to extend the possible cpus allowed
x86: fix cpu_mask_to_apicid_and to include cpu_online_mask
x86: update io_apic.c to the new cpumask code
x86: Introduce topology_core_cpumask()/topology_thread_cpumask()
x86: xen: use smp_call_function_many()
x86: use work_on_cpu in x86/kernel/cpu/mcheck/mce_amd_64.c
...
Fixed up trivial conflict in kernel/time/tick-sched.c manually
Since "Factor out cpu joining/unjoining the GIQ"
(b4963255ad) the WARN_ON in
xics_set_cpu_giq() is being triggered during boot on JS20 because the
GIQ indicator is not available on that platform. While the warning is
harmless and the system runs normally, it's nicer to check for the
existence of the indicator before trying to manipulate it.
Implement rtas_indicator_present(), which searches the
/rtas/rtas-indicators property for the given indicator token, and use
this function in xics_set_cpu_giq().
Also use a WARN statement in xics_set_cpu_giq to get better
information on failure.
Signed-off-by: Nathan Lynch <ntl@pobox.com>
Acked-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Impact: change existing irq_chip API
Not much point with gentle transition here: the struct irq_chip's
setaffinity method signature needs to change.
Fortunately, not widely used code, but hits a few architectures.
Note: In irq_select_affinity() I save a temporary in by mangling
irq_desc[irq].affinity directly. Ingo, does this break anything?
(Folded in fix from KOSAKI Motohiro)
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Reviewed-by: Grant Grundler <grundler@parisc-linux.org>
Acked-by: Ingo Molnar <mingo@redhat.com>
Cc: ralf@linux-mips.org
Cc: grundler@parisc-linux.org
Cc: jeremy@xensource.com
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Impact: change calling convention of existing cpumask APIs
Most cpumask functions started with cpus_: these have been replaced by
cpumask_ ones which take struct cpumask pointers as expected.
These four functions don't have good replacement names; fortunately
they're rarely used, so we just change them over.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Cc: paulus@samba.org
Cc: mingo@redhat.com
Cc: tony.luck@intel.com
Cc: ralf@linux-mips.org
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: cl@linux-foundation.org
Cc: srostedt@redhat.com
The 'ibm,interrupt-server#-size' properties are not in the cpu nodes,
which is where we currently look for them, but rather live under the
interrupt source controller nodes (which have "ibm,ppc-xics" in their
compatible property).
This moves the code that looks for the ibm,interrupt-server#-size
properties from xics_update_irq_servers() into xics_init_IRQ().
Also this adds a check for mismatched sizes across the interrupt
source controller nodes. Not sure this is necessary as in this case
the firmware might be seriously busted.
This property only appears on POWER6 boxes and is only used in the
set-indicator(gqirm) call, and apparently firmware currently ignores
the value we pass. Nevertheless we need to fix it in case future
firmware versions use it.
Signed-off-by: Sebastien Dugue <sebastien.dugue@bull.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
A single full sync (mb()) is requrired to order the mmio to the qirr reg
with the set or clear of the message word. However, test_and_clear_bit
has the effect of smp_mb() and we are not doing any other io from here,
so we don't need a mb per bit processed.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Several printks were broken at word boundaries for line length. Some
even referred to old function names. Using __func__ and changing the
text slightly for the format allows these printk formats to fit on one
line.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
It is physically per-cpu, and we want the irq layer to treat it that way.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
EOI normally has the side effect of returning the cpu to the base
priority to recieve the next interrupt. This is actually controlled
by the top byte of the xirr register. When we are exiting the
kernel in kexec we must eoi the ipi for the next kernel because we
never return from the handler, but we want to leave interrupt
delivery blocked until the next kernel takes action.
Since the hardware ipi vector is fixed, its easiest to just do the
eoi explicitly.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This factors out processors joining and unjoining the Global Interrupt
Queue into a separate function.
There is a bit of math to calculate the arguments to rtas to join
or leave the global interrupt queue, and a warning on failure
afterwards. Make a helper for the 3 callers.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
We only need to check the ibm,interrupt-server#-size property once, not
once per global server and thread.
We can use !CONFIG_SMP cpu masks and hard_smp_processor_id() to avoid an ifdef.
Put the node when breaking out of the loop on lpar systems.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Trim unneeded includes from xics.c. We don't use signals or gfp
flags, we use only OF functions and don't need prom, and the 8259
is now handled by our caller.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
The xirr is 32 bits in hardware, but the hypervisor requries the upper
bits of the register to be clear on the hcall. By changing the type
from signed to unsigned int we can drop masking it back to 32 bits.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Now that xics_update_irq_servers is called only from init and hotplug
code, it becomes possible to clean up the ordering of functions in the
file, grouping them but the interfaces they implement.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
xics supports only one ipi per cpu, and expects software to use some
queue to know why the interrupt was sent. In Linux, we use a an array
of bitmaps indexed by cpu to identify the message. Currently the bits
are set in smp.c and decoded in xics.c, with the data structure in a
header file. Consolidate the code in xics.c similar to mpic and other
interrupt controllers.
Also, while making the the array static, the message word doesn't need
to be volatile as set_bit and test_clear_bit take care of it for us, and
put it under ifdef smp.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Currently, every time we determine which irq server to use, we check if
default_server, which is the id of the bootcpu, is still online. But
default_server is a hardware cpu, not the logical cpu id needed to index
cpu_online_map.
Since the default server can only go offline during a cpu hotplug event,
explicitly check the default server and choose the new one when we move
irqs away from the cpu being offlined.
This has the added benefit of only needing the boot_cpuid to be updated
and not relying on the cpu being marked offline during migrate_irqs_away.
Also, since xics_update_irq_servers only reads device tree information, we
can call it before xics_init_host in xics_init_IRQ and then default_server
will always be valid when we can reach get_irq_server via the host ops.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
When reciving an irq vector that does not have a linux mapping, the kernel
prints a message and calls RTAS to disable the irq source. Previously
the kernel did not EOI the interrupt, causing the source to think it is
still being processed by software. While this does add an additional
layer of protection against interrupt storms had RTAS failed to disable
the source, it also prevents the interrupt from working when a driver
later enables it. (We could alternatively send an EOI on startup, but
that strategy would likely fail on an emulated xics.)
All interrupts should be disabled when the kernel starts, but this can
be observed if a driver does not shutdown an interrupt in its reboot
hook before starting a new kernel with kexec.
Michael reports this can be reproduced trivially by banging the keyboard
while kexec'ing on a P5 LPAR: even though the hvc_console driver request's
the console irq later in boot, the console is non-functional because
we're receiving no console interrupts.
Reported-By: Michael Ellerman
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
irq_radix_revmap() currently serves 2 purposes, irq mapping lookup
and insertion which happen in interrupt and process context respectively.
Separate the function into its 2 components, one for lookup only and one
for insertion only.
Fix the only user of the revmap tree (XICS) to use the new functions.
Also, move the insertion into the radix tree of those irqs that were
requested before it was initialized at said tree initialization.
Mutual exclusion between the tree initialization and readers/writers is
handled via a state variable (revmap_trees_allocated) set to 1 when the tree
has been initialized and set to 2 after the already requested irqs have been
inserted in the tree by the init path. This state is checked before any reader
or writer access just like we used to check for tree.gfp_mask != 0 before.
Finally, now that we're not any longer inserting nodes into the radix-tree
in interrupt context, turn the GFP_ATOMIC allocations into GFP_KERNEL ones.
Signed-off-by: Sebastien Dugue <sebastien.dugue@bull.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This converts ppc to use the new helpers for smp_call_function() and
friends, and adds support for smp_call_function_single().
ppc loses the timeout functionality of smp_call_function_mask() with
this change, as the generic code does not provide that.
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Add a DEBUG config setting which turns on all (most) of the debugging
under platforms/pseries.
To have this take effect we need to remove all the #undef DEBUG's, in
various files. We leave the #undef DEBUG in platforms/pseries/lpar.c,
as this enables debugging printks from the low-level hash table routines,
and tends to make your system unusable. If you want those enabled you
still have to turn them on by hand.
Also some of the RAS code has a DEBUG block which causes a functional
change, so I've keyed this off a different (non-existant) debug #define.
This is only enabled if you have PPC_EARLY_DEBUG enabled also.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The xics code currently has a direct and lpar variant of
xics_host_map, the only difference being which irq_chip they use. If
we remember which irq_chip we're using we can combine these two
routines. That also allows us to have a single irq_host_ops instead
of two.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The code in xics.c to setup the i8259 cascaded irq handler is not really
xics specific, so move it into setup.c - we will clean this up further in
a subsequent patch.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The xics code does update the default server information when the boot
cpu is removed. This patch recognizes when the boot cpu is being
removed and updates the appropriate information based on the new 'boot
cpu'.
Failure to update this information can causes us to leave irqs pinned
to cpus that are being removed, especially when removing the boot cpu.
The cpu is removed from the kernel, but cpu dlpar remove operations
fail since we cannot return the cpu to the hypervisor.
Signed-off-by: Nathan Fonteno <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
It appears that xics.c has its own of_get_cpu_node(). Remove this and
use the common one from prom.c.
Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
This splits off the kexec path bits of the xics_teardown_cpu() routine
into its own xics_kexec_teardown_cpu() routine. With the previous
combined routine the CPPR for a cpu that is being removed may have its
CPPR reset in the plpar_eoi() call (which explicitly sets the CPPR to
a non-zero value). Splitting of the kexec bits of the code prevents
this from happening in the cpu remove path.
Once again, this does not cause the cpu remove from the kernel to
fail, but it does cause cpu dlpar operations to not be able to return
the cpu to the hypervisor.
Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The affinity mask in the virq descriptor needs to be set before we
reset the affinity for the virq. Without doing this the call to get
the new irq server fails and we end up leaving the virq pinned to the
cpu we are removing.
This does not fail the cpu remove from the kernel, but it does prevent
cpu dlpar remove operations from returning the cpu to the hypervisor.
Signed-off-by: Nathan Fontenot <nfont@austin.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The per-processor interrupt request register and current processor
priority register are only accessed on the current cpu. In fact the
hypervisor doesn't even let us choose which cpu's registers to access.
The only function to use cpu twice is xics_migrate_irqs_away, not a fast
path. But we can cache the result of get_hard_processor_id() instead of
calling get_hard_smp_processor_id(cpu) in a loop across the call to rtas.
Years ago the irq code passed smp_processor_id into get_irq, I thought
we might initialize the CPPR third party at boot as an extra measure of
saftey, and it made the code symmetric with the qirr (queued interrupt
for software generated interrupts), but now it is just extra and
sometimes unneeded work to pass it down.
Signed-off-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
On a POWER6 machine running 2.6.23-rc8 I sometimes see the following error:
xics_set_affinity: No online cpus in the mask 00000000,00000000,00000000,00000001 for irq 20
In a desperate attempt to get a changelog entry in 2.6.23, I took a look
into it.
It turns out we are passing a real and not a virtual irq into
get_irq_server. This works for the case where hwirq < NR_IRQS and we
set virq = hwirq. In my case however hwirq = 590082 and we try and
access irq_desc[590082], slightly past the end at 512 entries.
Lucky we ship lots of memory with our machines.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The majority of irq_host implementations (3 out of 4) are associated
with a device_node, and need to stash it somewhere. Rather than having
it somewhere different for each host, add an optional device_node pointer
to the irq_host structure.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Paul Mackerras <paulus@samba.org>
In some of the PPC970 based systems, interrupt would be distributed to
offline cpus also even when booted with "maxcpus=1". So check whether
cpu online map and cpu present map are equal or not. If they are equal
default_distrib_server is used as interrupt server otherwise boot cpu
(default_server) used as interrupt server.
In addition to this, if an interrupt is assigned to a specific cpu (ie
smp affinity) and if that cpu is not online, the earlier code used to
return the default_distrib_server as interrupt server. This
introduces an additional parameter to the get_irq function, called
strict_check. Based on this parameter, if the cpu is not online
either default_distrib_server or -1 is returned.
Signed-off-by: Mohan Kumar M <mohan@in.ibm.com>
Cc: Michael Ellerman <michael@ellerman.id.au>
Acked-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
In 616883df78 request_irq was marked as
__must_check so we must... er... check it.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Anton Blanchard