Since kernel 4.11 the thread and irq stacks on parisc randomly overflow
the default size of 16k. The reason why stack usage suddenly grew is yet
unknown.
Signed-off-by: Helge Deller <deller@gmx.de>
Cc: stable@vger.kernel.org # 4.11+
Signed-off-by: Helge Deller <deller@gmx.de>
Before the irq handler detects a low stack and then panics the kernel, disable
further stack checks to avoid recursive panics.
Reported-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
When detecting a serial port on newer PA-RISC machines (with iosapic) we have a
long way to go to find the right IRQ line, registering it, then registering the
serial port and the irq handler for the serial port. During this phase spurious
interrupts for the serial port may happen which then crashes the kernel because
the action handler might not have been set up yet.
So, basically it's a race condition between the serial port hardware and the
CPU which sets up the necessary fields in the irq sructs. The main reason for
this race is, that we unmask the serial port irqs too early without having set
up everything properly before (which isn't easily possible because we need the
IRQ number to register the serial ports).
This patch is a work-around for this problem. It adds checks to the CPU irq
handler to verify if the IRQ action field has been initialized already. If not,
we just skip this interrupt (which isn't critical for a serial port at bootup).
The real fix would probably involve rewriting all PA-RISC specific IRQ code
(for CPU, IOSAPIC, GSC and EISA) to use IRQ domains with proper parenting of
the irq chips and proper irq enabling along this line.
This bug has been in the PA-RISC port since the beginning, but the crashes
happened very rarely with currently used hardware. But on the latest machine
which I bought (a C8000 workstation), which uses the fastest CPUs (4 x PA8900,
1GHz) and which has the largest possible L1 cache size (64MB each), the kernel
crashed at every boot because of this race. So, without this patch the machine
would currently be unuseable.
For the record, here is the flow logic:
1. serial_init_chip() in 8250_gsc.c calls iosapic_serial_irq().
2. iosapic_serial_irq() calls txn_alloc_irq() to find the irq.
3. iosapic_serial_irq() calls cpu_claim_irq() to register the CPU irq
4. cpu_claim_irq() unmasks the CPU irq (which it shouldn't!)
5. serial_init_chip() then registers the 8250 port.
Problems:
- In step 4 the CPU irq shouldn't have been registered yet, but after step 5
- If serial irq happens between 4 and 5 have finished, the kernel will crash
Signed-off-by: Helge Deller <deller@gmx.de>
The [user space] interface does not filter out offline cpus. It merily
guarantees that the mask contains at least one online cpu.
So the selector in the irq chip implementation needs to make sure to
pick only an online cpu because otherwise:
Offline Core 1
Set affinity to 0xe (is valid due to online mask 0xd)
cpumask_first will pick core 1, which is offline
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: James E.J. Bottomley <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-parisc@vger.kernel.org
Link: http://lkml.kernel.org/r/20140304203100.859489993@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Pull IRQ changes from Ingo Molnar:
"The biggest change this cycle are the softirq/hardirq stack
interaction and nesting fixes, cleanups and reorganizations from
Frederic. This is the longer followup story to the softirq nesting
fix that is already upstream (commit ded7975475: "irq: Force hardirq
exit's softirq processing on its own stack")"
* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
irqchip: bcm2835: Convert to use IRQCHIP_DECLARE macro
powerpc: Tell about irq stack coverage
x86: Tell about irq stack coverage
irq: Optimize softirq stack selection in irq exit
irq: Justify the various softirq stack choices
irq: Improve a bit softirq debugging
irq: Optimize call to softirq on hardirq exit
irq: Consolidate do_softirq() arch overriden implementations
x86/irq: Correct comment about i8259 initialization
The number of IPI calls is already visible as per-cpu IPI irq counters
in/proc/cpuinfo, so let's drop this additional counting.
This partly reverts:
cd85d55 parisc: more irq statistics in /proc/interrupts
Signed-off-by: Helge Deller <deller@gmx.de>
All arch overriden implementations of do_softirq() share the following
common code: disable irqs (to avoid races with the pending check),
check if there are softirqs pending, then execute __do_softirq() on
a specific stack.
Consolidate the common parts such that archs only worry about the
stack switch.
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@au1.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@au1.ibm.com>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: James E.J. Bottomley <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
The logic to detect if the irq stack was already in use with
raw_spin_trylock() is wrong, because it will generate a "trylock failure
on UP" error message with CONFIG_SMP=n and CONFIG_DEBUG_SPINLOCK=y.
arch_spin_trylock() can't be used either since in the CONFIG_SMP=n case
no atomic protection is given and we are reentrant here. A mutex didn't
worked either and brings more overhead by turning off interrupts.
So, let's use the fastest path for parisc which is the ldcw instruction.
Counting how often the irq stack was used is pretty useless, so just
drop this piece of code.
Signed-off-by: Helge Deller <deller@gmx.de>
This patch fixes few build issues which were introduced with the last
irq stack patch, e.g. the combination of stack overflow check and irq
stack.
Furthermore we now do proper locking and change the irq bh handler
to use the irq stack as well.
In /proc/interrupts one now can monitor how huge the irq stack has grown
and how often it was preferred over the kernel stack.
IRQ stacks are now enabled by default just to make sure that we not
overflow the kernel stack by accident.
Signed-off-by: Helge Deller <deller@gmx.de>
Default kernel stack size on parisc is 16k. During tests we found that the
kernel stack can easily grow beyond 13k, which leaves 3k left for irq
processing.
This patch adds the possibility to activate an additional stack of 16k per CPU
which is being used during irq processing. This implementation does not yet
uses this irq stack for the irq bh handler.
The assembler code for call_on_stack was heavily cleaned up by John
David Anglin.
CC: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
Add the CONFIG_DEBUG_STACKOVERFLOW config option to enable checks to
detect kernel stack overflows.
Stack overflows can not be detected reliable since we do not want to
introduce too much overhead.
Instead, during irq processing in do_cpu_irq_mask() we check kernel
stack usage of the interrupted kernel process. Kernel threads can be
easily detected by checking the value of space register 7 (sr7) which
is zero when running inside the kernel.
Since THREAD_SIZE is 16k and PAGE_SIZE is 4k, reduce the alignment of
the init thread to the lower value (PAGE_SIZE) in the kernel
vmlinux.ld.S linker script.
Signed-off-by: Helge Deller <deller@gmx.de>
People are playing odd games with IRQF_DISABLED, remove it.
Its not reliable, since shared interrupt lines could disable it for you,
and its possible and allowed for archs to disable IRQs to limit IRQ nesting.
Therefore, simply mandate that _ALL_ IRQ handlers are run with IRQs disabled.
[ This _should_ not break anything, since we've mandated that IRQ handlers
_must_ be able to deal with this for a _long_ time ]
IRQ handlers should be fast, no if buts and any other exceptions. We also have
plenty instrumentation to find any offending IRQ latency sources.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Helge Deller <deller@gmx.de>
The CPU irqs (timer and IPI) are not shared and only need to be claimed once.
A mismatch error occurs if they are claimed more than once.
Signed-off-by: John David Anglin <dave.anglin@nrc-cnrc.gc.ca>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Helge Deller <deller@gmx.de>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Signed-off-by: Helge Deller <deller@gmx.de>
Convert to the new function names. Scripted with coccinelle.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: linux-parisc@vger.kernel.org
1) As promised in the comment, the core does not copy cpumask anymore
when the arch code returns -EINVAL
2) Get the per cpu information from irq_data
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: linux-parisc@vger.kernel.org
Convert all the parisc driver interrupt handlers (dino, eisa, gsc,
iosapic and superio) as well as the cpu interrupts. Prepare
show_interrupts for GENERIC_HARDIRQS_NO_DEPRECATED and finally selects
that Kconfig option
[jejb: compile and testing fixes]
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The CHECK_IRQ_PER_CPU is wrong, it should be checking
irq_to_desc(irq)->status not just irq.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@kernel.org
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
The generic conversion eliminates the spurious no_ack and no_end
routines, converts all the cascaded handlers to handle_simple_irq() and
makes iosapic use a modified handle_percpu_irq() to become the same as
the CPU irq's. This isn't an essential change, but it eliminates the
mask/unmask overhead of handle_level_irq().
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Tested-by: Helge Deller <deller@gmx.de>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
The essential problem we're currently having is that dino (and gsc) is a
cascaded CPU interrupt. Under the old __do_IRQ() handler, our CPU
interrupts basically did an ack followed by an end. In the new scheme,
we replaced them with level handlers which do a mask, an ack and then an
unmask (but no end). Instead, with the renaming of end to eoi, we
actually want to call the percpu flow handlers, because they actually
have all the characteristics we want.
This patch does the conversion and gets my C360 booting again.
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
Use proper accessors and handlers for generic irq cleanups. We just
call back into __do_IRQ through desc->handler now, and remove the
explicit calls.
Signed-off-by: Kyle McMartin <kyle@redhat.com>
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>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
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>
The defines and typedefs (hw_interrupt_type, no_irq_type, irq_desc_t) have
been kept around for migration reasons. After more than two years it's
time to remove them finally.
This patch cleans up one of the remaining users. When all such patches
hit mainline we can remove the defines and typedefs finally.
Impact: cleanup
Convert the last remaining users to struct irq_chip and remove the
define.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
Fix miscompilation in arch/parisc/kernel/irq.c:
123: warning: passing arg 1 of `cpumask_setall' from incompatible pointer type
141: warning: passing arg 1 of `cpumask_copy' from incompatible pointer type
300: warning: passing arg 1 of `cpumask_copy' from incompatible pointer type
357: warning: passing arg 2 of `cpumask_copy' from incompatible pointer type
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
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, futureproof
In fact, all cpumask ops will only be valid (in general) for bit
numbers < nr_cpu_ids. So use that instead of NR_CPUS in various
places.
This is always safe: no cpu number can be >= nr_cpu_ids, and
nr_cpu_ids is initialized to NR_CPUS at boot.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Acked-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>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
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>
Impact: build fix
Ingo Molnar wrote:
> tip/arch/blackfin/kernel/irqchip.c: In function 'show_interrupts':
> tip/arch/blackfin/kernel/irqchip.c:85: error: 'struct kernel_stat' has no member named 'irqs'
> make[2]: *** [arch/blackfin/kernel/irqchip.o] Error 1
> make[2]: *** Waiting for unfinished jobs....
>
So could move kstat_irqs array to irq_desc struct.
(s390, m68k, sparc) are not touched yet, because they don't support genirq
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
trivial fixes:
- use KERN_WARNING for printk()
- use BUG_ON() instead of "if (xx) BUG();"
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
--
To unsubscribe from this list: send the line "unsubscribe linux-parisc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
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>
The timer interrupt had accidentally been marked IRQ_DISABLED since
IRQ_PER_CPU had been OR-ed in, instead of set. This had been working
by accident for quite a while.
Commit c642b8391c changed the behaviour of
IRQ_PER_CPU interrupts, which previously weren't checked for IRQ_DISABLED.
Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
Kudos to Thibaut Varene for spotting the (mis)use of appropriately named
global_ack_eiem. This took a long time to figure out and both insight
from myself, Kyle McMartin, and James Bottomley were required to narrow
down which bit of code could have this race condition.
The symptom was interrupts stopped getting delivered while some workload
was generating IO interrupts on two different CPUs. One of the interrupt
sources would get masked off and stay unmasked. Problem was global_ack_eiem
was accessed with read/modified/write sequence and not protected by
a spinlock.
PA-RISC doesn't need a global ack flag though. External Interrupts
are _always_ delivered to a single CPU (except for "global broadcast
interrupt" which AFAIK currently is not used.) So we don't have to worry
about any given IRQ vector getting delivered to more than one CPU.
Tested on a500 and rp34xx boxen. rsync to/from gsyprf11 (a500)
would lock up the box since NIC (tg3) interrupt and SCSI (sym2)
were on "opposite" CPUs (2 CPU system). Put them on the same CPU
or apply this patch and 10GB of data would rsync completely.
Please apply the following critical patch.
thanks,
grant
Signed-off-by: Grant Grundler <grundler@parisc-linux.org>
Acked-by: Thibaut VARENE <T-Bone@parisc-linux.org>
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
Add IRQF_IRQPOLL to the timer interrupt on parisc.
Signed-off-by: Bernhard Walle <bwalle@suse.de>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Matthew Wilcox <willy@debian.org>
Cc: Grant Grundler <grundler@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
