Jack Ren and Eric Miao tracked down the following long standing
problem in the NOHZ code:
scheduler switch to idle task
enable interrupts
Window starts here
----> interrupt happens (does not set NEED_RESCHED)
irq_exit() stops the tick
----> interrupt happens (does set NEED_RESCHED)
return from schedule()
cpu_idle(): preempt_disable();
Window ends here
The interrupts can happen at any point inside the race window. The
first interrupt stops the tick, the second one causes the scheduler to
rerun and switch away from idle again and we end up with the tick
disabled.
The fact that it needs two interrupts where the first one does not set
NEED_RESCHED and the second one does made the bug obscure and extremly
hard to reproduce and analyse. Kudos to Jack and Eric.
Solution: Limit the NOHZ functionality to the idle loop to make sure
that we can not run into such a situation ever again.
cpu_idle()
{
preempt_disable();
while(1) {
tick_nohz_stop_sched_tick(1); <- tell NOHZ code that we
are in the idle loop
while (!need_resched())
halt();
tick_nohz_restart_sched_tick(); <- disables NOHZ mode
preempt_enable_no_resched();
schedule();
preempt_disable();
}
}
In hindsight we should have done this forever, but ...
/me grabs a large brown paperbag.
Debugged-by: Jack Ren <jack.ren@marvell.com>,
Debugged-by: eric miao <eric.y.miao@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Andrew Morton reported this against linux-next:
ERROR: ".save_stack_trace" [tests/backtracetest.ko] undefined!
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Expand the per-process PGDs so that they cover the kernel virtual
memory area as well. This simplifies the TLB miss handler fastpath
since it doesn't have to check for kernel addresses anymore.
If a TLB miss happens on a kernel address and a second-level page
table can't be found, we check swapper_pg_dir and copy the PGD entry
into the user PGD if it can be found there.
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Instead of storing physical addresses along with page flags in the
PGD, store virtual addresses and use NULL to indicate a not present
second-level page table. A non-page-aligned page table indicates a bad
PMD.
This simplifies the TLB miss handler since it no longer has to check
the Present bit and no longer has to convert the PGD entry from
physical to virtual address. Instead, it has to check for a NULL
entry, which is slightly cheaper than either.
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Kill the special exception handler sections .tlbx.ex.text,
.tlbr.ex.text, tlbw.ex.text and .scall.text. Use .org instead to place
the handlers at the required offsets from EVBA.
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
A signal handler should be able to change the signal stack used for the
next signal by altering the ucontext_t passed as a parameter to the
handler. This does not currently work on avr32 since it doesn't update
the in-kernel signal context from the ucontext_t upon signal handler
return.
Fix it by adding a call to do_sigaltstack() from sys_rt_sigreturn(),
bringing it in line with most other architectures.
Signed-off-by: Martin Koegler <mkoegler@auto.tuwien.ac.at>
[haavard.skinnemoen@atmel.com: changed patch description]
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
This replaces the duplicated arch-specific versions of "sys_pipe()" with
one unified implementation. This removes almost 250 lines of duplicated
code.
It's marked __weak, so that *if* an architecture wants to override the
default implementation it can do so by simply having its own replacement
version, since many architectures use alternate calling conventions for
the 'pipe()' system call for legacy reasons (ie traditional UNIX
implementations often return the two file descriptors in registers)
I still haven't changed the cris version even though Linus says the BKL
isn't needed. The arch maintainer can easily do it if there are really
no obstacles.
Signed-off-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If the user specified a fixed framebuffer address on the command line, it may
have been initialized already with a splash image or something, so we
shouldn't clear it.
Therefore, we should only initialize the framebuffer if we allocated it
ourselves. This patch also updates the AVR32 setup code to clear the
framebuffer if it allocated it itself, i.e. the user didn't provide a fixed
address or the reservation failed.
I've updated the at91 platform code as well so that it initializes the
framebuffer if it is located in SRAM, but I haven't tested that it actually
works.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Cc: Nicolas FERRE <nicolas.ferre@rfo.atmel.com>
Cc: Andrew Victor <andrew@sanpeople.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch adds in the indirect call to pm_power_off(), as is done in
other architectures (e.g. ARM).
Tested on NGW100, with custom board with GPIO control over main DC
power.
Signed-off-by: Peter Ma <pma@mediamatech.com>
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
This combines three patches from David Brownell:
* avr32: tclib support
* avr32: simplify clocksources
* avr32: Turn count/compare into a oneshot clockevent device
Register both TC blocks (instead of just the first one) so that
the AT32/AT91 tclib code will pick them up (instead of just the
avr32-only PIT-style clocksource).
Rename the first one and its resources appropriately.
More cleanups to the cycle counter clocksource code
- Disable all the weak symbol magic; remove the AVR32-only TCB-based
clocksource code (source and header).
- Mark the __init code properly.
- Don't forget to report IRQF_TIMER.
- Make the system work properly with this clocksource, by preventing
use of the CPU "idle" sleep state in the idle loop when it's used.
Package the avr32 count/compare timekeeping support as a oneshot
clockevent device, so it supports NO_HZ and high res timers.
This means it also supports plugging in other clockevent devices
and clocksources.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Create a new file, pm-at32ap700x.S, in mach-at32ap and move the CPU
idle sleep code there. Make it possible to disable the sleep code.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Start cleaning up the AVR32 clocksource mess, starting with the cycle
counter clocksource: remove unneeded pseudo-RTC (just inline that
call to mktime) and associated build warning, and unused sysdev.
Add comment about the problem using the cycle counter register,
and adjust the clocksource rating accordingly. Later patches can
make this usable again (by disabling use of the idle state and
providing a proper clocksource without the weak binding hacks)
and move towards TCB-based clockevent support (including high
resolution timers) that's shared between AT91 and AVR32.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Semaphores are no longer performance-critical, so a generic C
implementation is better for maintainability, debuggability and
extensibility. Thanks to Peter Zijlstra for fixing the lockdep
warning. Thanks to Harvey Harrison for pointing out that the
unlikely() was unnecessary.
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
add_reserved_region() tries to keep the resource list sorted, so when
looking for a place to insert the new resource, it may break out
before the last entry.
When this happens, the list is broken in two because the sibling field
of the new entry doesn't point to the next resource. Fix it by
updating the new resource's sibling field appropriately.
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Don't include the BUG trap handling code when CONFIG_BUG is not set.
This fixes allnoconfig.
Signed-off-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Iff the parent has TIF_DEBUG set, _and_ clone_flags includes
CLONE_PTRACE we should set the TIF_DEBUG flag for the child and
increment the ocd refcount. Otherwise, the TIF_DEBUG flag must be
unset.
Currently, the child inherits TIF_DEBUG from the parent before
copy_thread is called, so TIF_DEBUG may be already be set before we
determine whether the child is supposed to inherit debugging
capabilities from the parent or not. This means that ocd_enable()
won't increment the refcount, because TIF_DEBUG is already set, and
that TIF_DEBUG will be set for processes that aren't being debugged.
This leads to a refcounting asymmetry, which may show up as
------------[ cut here ]------------
Badness at arch/avr32/kernel/ocd.c:73
PC is at ocd_disable+0x34/0x60
LR is at put_lock_stats+0xa/0x20
as reported by David Brownell. Happens when strace'ing a process that
forks a new child process, e.g. "strace mount -tjffs2 mtd1 /mnt", and
subsequently killing the child process (e.g. "umount /mnt".)
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
This fixes a hang on boot with nohz enabled. nohz is not actually
supported in mainline yet, but patches that add support for it are
currently under review.
When nohz is compiled out, the functions are no-ops, so this patch
results in no functional change, but it arguably makes the code more
correct.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Sometimes simple attributes might need to return an error, e.g. for
acquiring a mutex interruptibly. In fact we have that situation in
spufs already which is the original user of the simple attributes. This
patch merged the temporarily forked attributes in spufs back into the
main ones and allows to return errors.
[akpm@linux-foundation.org: build fix]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: <stefano.brivio@polimi.it>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg KH <greg@kroah.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patchset adds a flags variable to reserve_bootmem() and uses the
BOOTMEM_EXCLUSIVE flag in crashkernel reservation code to detect collisions
between crashkernel area and already used memory.
This patch:
Change the reserve_bootmem() function to accept a new flag BOOTMEM_EXCLUSIVE.
If that flag is set, the function returns with -EBUSY if the memory already
has been reserved in the past. This is to avoid conflicts.
Because that code runs before SMP initialisation, there's no race condition
inside reserve_bootmem_core().
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix powerpc build]
Signed-off-by: Bernhard Walle <bwalle@suse.de>
Cc: <linux-arch@vger.kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch consolidate all definitions of .init.text, .init.data
and .exit.text, .exit.data section definitions in
the generic vmlinux.lds.h.
This is a preparational patch - alone it does not buy
us much good.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6:
[AVR32] extint: Set initial irq type to low level
[AVR32] extint: change set_irq_type() handling
[AVR32] NMI debugging
[AVR32] constify function pointer tables
[AVR32] ATNGW100: Update defconfig
[AVR32] ATSTK1002: Update defconfig
[AVR32] Kconfig: Choose daughterboard instead of CPU
[AVR32] Add support for ATSTK1003 and ATSTK1004
[AVR32] Clean up external DAC setup code
[AVR32] ATSTK1000: Move gpio-leds setup to setup.c
[AVR32] Add support for AT32AP7001 and AT32AP7002
[AVR32] Provide more CPU information in /proc/cpuinfo and dmesg
[AVR32] Oprofile support
[AVR32] Include instrumentation menu
Disable VGA text console for AVR32 architecture
[AVR32] Enable debugging only when needed
ptrace: Call arch_ptrace_attach() when request=PTRACE_TRACEME
[AVR32] Remove redundant try_to_freeze() call from do_signal()
[AVR32] Drop GFP_COMP for DMA memory allocations
Change the NMI handler to use the die notifier chain to signal anyone
who cares. Add a simple "nmi debugger" which hooks into this chain and
that may dump registers, task state, etc. when it happens.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Add the following fields to /proc/cpuinfo:
* chip type and revision (from the JTAG chip id)
* cpu MHz (from clk_get_rate())
* features (from the CONFIG0 register)
Also rename "cpu family" to "cpu arch" and "cpu type" to "cpu core" to
remove some ambiguity.
Show chip type and revision at bootup, and clarify that the other
kinds of IDs that we're already printing are for the cpu core and
architecture. Rename "AP7000" to "AP7" since that's the name of the
core.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Keep track of processes being debugged (including the kernel itself)
and turn the OCD system on and off as appropriate. Since enabling
debugging turns off some optimizations in the CPU core, this fixes the
issue that enabling KProbes support or simply running a program under
gdbserver will reduce system performance significantly until the next
reboot.
The CPU performance will still be reduced for all processes while a
process is being debugged, but this is a lot better than reducing the
performance forever.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
get_signal_to_deliver() will call try_to_freeze(), so there's no point
in do_signal() doing it as well.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
All kobjects require a dynamically allocated name now. We no longer
need to keep track if the name is statically assigned, we can just
unconditionally free() all kobject names on cleanup.
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It's not like it really matters at this point since the system is
dying anyway, but handle_critical pushes too few registers on the
stack so the register dump, which makes the register dump look a bit
strange. This patch fixes it.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
The current debug trap handling code does a number of things that are
illegal according to the AVR32 Architecture manual. Most importantly,
it may try to schedule from Debug Mode, thus clearing the D bit, which
can lead to "undefined behaviour".
It seems like this works in most cases, but several people have
observed somewhat unstable behaviour when debugging programs,
including soft lockups. So there's definitely something which is not
right with the existing code.
The new code will never schedule from Debug mode, it will always exit
Debug mode with a "retd" instruction, and if something not running in
Debug mode needs to do something debug-related (like doing a single
step), it will enter debug mode through a "breakpoint" instruction.
The monitor code will then return directly to user space, bypassing
its own saved registers if necessary (since we don't actually care
about the trapped context, only the one that came before.)
This adds three instructions to the common exception handling code,
including one branch. It does not touch super-hot paths like the TLB
miss handler.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Generate a new set of OCD register definitions in asm/ocd.h and rename
__mfdr() and __mtdr() to ocd_read() and ocd_write() respectively.
The bitfield definitions are a lot more complete now, and they are
entirely based on bit numbers, not masks. This is because OCD
registers are frequently accessed from assembly code, where bit
numbers are a lot more useful (can be fed directly to sbr, bfins,
etc.)
Bitfields that consist of more than one bit have two definitions:
_START, which indicates the number of the first bit, and _SIZE, which
indicates the number of bits. These directly correspond to the
parameters taken by the bfextu, bfexts and bfins instructions.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
The 'H' bit is bit 29, while the 'R' bit doesn't exist. Luckily, we
don't actually use any of the bits in question.
Also update show_regs() to show the Debug Mask and Debug state bits.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
is_init() is an ambiguous name for the pid==1 check. Split it into
is_global_init() and is_container_init().
A cgroup init has it's tsk->pid == 1.
A global init also has it's tsk->pid == 1 and it's active pid namespace
is the init_pid_ns. But rather than check the active pid namespace,
compare the task structure with 'init_pid_ns.child_reaper', which is
initialized during boot to the /sbin/init process and never changes.
Changelog:
2.6.22-rc4-mm2-pidns1:
- Use 'init_pid_ns.child_reaper' to determine if a given task is the
global init (/sbin/init) process. This would improve performance
and remove dependence on the task_pid().
2.6.21-mm2-pidns2:
- [Sukadev Bhattiprolu] Changed is_container_init() calls in {powerpc,
ppc,avr32}/traps.c for the _exception() call to is_global_init().
This way, we kill only the cgroup if the cgroup's init has a
bug rather than force a kernel panic.
[akpm@linux-foundation.org: fix comment]
[sukadev@us.ibm.com: Use is_global_init() in arch/m32r/mm/fault.c]
[bunk@stusta.de: kernel/pid.c: remove unused exports]
[sukadev@us.ibm.com: Fix capability.c to work with threaded init]
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Acked-by: Pavel Emelianov <xemul@openvz.org>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Herbert Poetzel <herbert@13thfloor.at>
Cc: Kirill Korotaev <dev@sw.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Introduce architecture dependent kretprobe blacklists to prohibit users
from inserting return probes on the function in which kprobes can be
inserted but kretprobes can not.
This patch also removes "__kprobes" mark from "__switch_to" on x86_64 and
registers "__switch_to" to the blacklist on x86-64, because that mark is to
prohibit user from inserting only kretprobe.
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Prasanna S Panchamukhi <prasanna@in.ibm.com>
Acked-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Identical handlers of PTRACE_DETACH go into ptrace_request().
Not touching compat code.
Not touching archs that don't call ptrace_request.
Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
Acked-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As explained on:
http://www.avrfreaks.net/index.php?nameÿphpBB2&fileÿewtopic&tS307
If the current process is preempted before it can copy RAR_SUP and
RSR_SUP both register are lost and the process will segfault as soon
as it return from the syscall since the return adress will be
corrupted.
This patch disable IRQ as soon as we enter the syscall path and
reenable them when the copy is done.
In the interrupt handlers, check if we are interrupting the srrf
instruction, if so disable interrupts and return. The interrupt
handler will be re-called immediatly when the interrupts are
reenabled.
After some stressing workload:
- find / > /dev/null in loop
- top (in ssh)
- ping -f avr32
The segfaults are not seen anymore.
Signed-off-by: Philippe Rétornaz <philippe.retornaz@epfl.ch>
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
__init_end, which comes immediately before .text, is already page
aligned, and that should be more than enough for the .text section.
The reason why we need to align the .text section is because the
interrupt handler offset is ORed with EVBA, so we need to provide
enough alignment of EVBA that this OR operation works as an ADD.
Currently, the last interrupt handler is not nearly a full page away
from EVBA, so it won't be a problem.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Use PAGE_SIZE, THREAD_SIZE and L1_CACHE_BYTES instead of harcoded
constants in places where that's what we really mean.
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
