| Age | Commit message (Collapse) | Author |
|---|
|
commit 85c1fafd7262e68ad821ee1808686b1392b1167d upstream.
On ppc64 we support 4K hash pte with 64K page size. That requires
us to track the hash pte slot information on a per 4k basis. We do that
by storing the slot details in the second half of pte page. The pte bit
_PAGE_COMBO is used to indicate whether the second half need to be
looked while building real_pte. We need to use read memory barrier while
doing that so that load of hidx is not reordered w.r.t _PAGE_COMBO
check. On the store side we already do a lwsync in __hash_page_4K
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: include <asm/system.h> to ensure smp_rmb()
is defined; cell_defconfig fails to build without this]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 83d5e64b7efa7f39b10ff5e92792e807a720289c upstream.
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit b00fc6ec1f24f9d7af9b8988b6a198186eb3408c upstream.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=81631
Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: Andrey Utkin <andrey.krieger.utkin@gmail.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 4badad352a6bb202ec68afa7a574c0bb961e5ebc upstream.
The optimistic spin code assumes regular stores and cmpxchg() play nice;
this is found to not be true for at least: parisc, sparc32, tile32,
metag-lock1, arc-!llsc and hexagon.
There is further wreckage, but this in particular seemed easy to
trigger, so blacklist this.
Opt in for known good archs.
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: David Miller <davem@davemloft.net>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: James Bottomley <James.Bottomley@hansenpartnership.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Waiman Long <waiman.long@hp.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: John David Anglin <dave.anglin@bell.net>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Davidlohr Bueso <davidlohr@hp.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Cc: linuxppc-dev@lists.ozlabs.org
Cc: sparclinux@vger.kernel.org
Link: http://lkml.kernel.org/r/20140606175316.GV13930@laptop.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.2:
- Adjust context
- Drop arm64 change]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 59a53afe70fd530040bdc69581f03d880157f15a upstream.
OPAL will mark a CPU that is guarded as "bad" in the status property of the CPU
node.
Unfortunatley Linux doesn't check this property and will put the bad CPU in the
present map. This has caused hangs on booting when we try to unsplit the core.
This patch checks the CPU is avaliable via this status property before putting
it in the present map.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit c4cad90f9e9dcb85afc5e75a02ae3522ed077296 upstream.
We had a mix & match of flags used when creating legacy ports
depending on where we found them in the device-tree. Among others
we were missing UPF_SKIP_TEST for some kind of ISA ports which is
a problem as quite a few UARTs out there don't support the loopback
test (such as a lot of BMCs).
Let's pick the set of flags used by the SoC code and generalize it
which means autoconf, no loopback test, irq maybe shared and fixed
port.
Sending to stable as the lack of UPF_SKIP_TEST is breaking
serial on some machines so I want this back into distros
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 7998eb3dc700aaf499f93f50b3d77da834ef9e1d upstream.
With binutils 2.24, various 64 bit builds fail with relocation errors
such as
arch/powerpc/kernel/built-in.o: In function `exc_debug_crit_book3e':
(.text+0x165ee): relocation truncated to fit: R_PPC64_ADDR16_HI
against symbol `interrupt_base_book3e' defined in .text section
in arch/powerpc/kernel/built-in.o
arch/powerpc/kernel/built-in.o: In function `exc_debug_crit_book3e':
(.text+0x16602): relocation truncated to fit: R_PPC64_ADDR16_HI
against symbol `interrupt_end_book3e' defined in .text section
in arch/powerpc/kernel/built-in.o
The assembler maintainer says:
I changed the ABI, something that had to be done but unfortunately
happens to break the booke kernel code. When building up a 64-bit
value with lis, ori, shl, oris, ori or similar sequences, you now
should use @high and @higha in place of @h and @ha. @h and @ha
(and their associated relocs R_PPC64_ADDR16_HI and R_PPC64_ADDR16_HA)
now report overflow if the value is out of 32-bit signed range.
ie. @h and @ha assume you're building a 32-bit value. This is needed
to report out-of-range -mcmodel=medium toc pointer offsets in @toc@h
and @toc@ha expressions, and for consistency I did the same for all
other @h and @ha relocs.
Replacing @h with @high in one strategic location fixes the relocation
errors. This has to be done conditionally since the assembler either
supports @h or @high but not both.
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 8fe9c93e7453e67b8bd09f263ec1bb0783c733fc upstream.
GCC 4.8 now generates out-of-line vr save/restore functions when
optimizing for size. They are needed for the raid6 altivec support.
Signed-off-by: Andreas Schwab <schwab@linux-m68k.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit a5b2cf5b1af424ee3dd9e3ce6d5cea18cb927e67 upstream.
The 64bit relocation code places a few symbols in the text segment.
These symbols are only 4 byte aligned where they need to be 8 byte
aligned. Add an explicit alignment.
Signed-off-by: Anton Blanchard <anton@samba.org>
Tested-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit f5295bd8ea8a65dc5eac608b151386314cb978f1 upstream.
In copy_oldmem_page, the current check using max_pfn and min_low_pfn to
decide if the page is backed or not, is not valid when the memory layout is
not continuous.
This happens when running as a QEMU/KVM guest, where RTAS is mapped higher
in the memory. In that case max_pfn points to the end of RTAS, and a hole
between the end of the kdump kernel and RTAS is not backed by PTEs. As a
consequence, the kdump kernel is crashing in copy_oldmem_page when accessing
in a direct way the pages in that hole.
This fix relies on the memblock's service memblock_is_region_memory to
check if the read page is part or not of the directly accessible memory.
Signed-off-by: Laurent Dufour <ldufour@linux.vnet.ibm.com>
Tested-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 91b973f90c1220d71923e7efe1e61f5329806380 upstream.
The code in remove_cache_dir() is supposed to remove the "cache"
subdirectory from the sysfs directory for a CPU when that CPU is
being offlined. It tries to do this by calling kobject_put() on
the kobject for the subdirectory. However, the subdirectory only
gets removed once the last reference goes away, and the reference
being put here may well not be the last reference. That means
that the "cache" subdirectory may still exist when the offlining
operation has finished. If the same CPU subsequently gets onlined,
the code tries to add a new "cache" subdirectory. If the old
subdirectory has not yet been removed, we get a WARN_ON in the
sysfs code, with stack trace, and an error message printed on the
console. Further, we ultimately end up with an online cpu with no
"cache" subdirectory.
This fixes it by doing an explicit kobject_del() at the point where
we want the subdirectory to go away. kobject_del() removes the sysfs
directory even though the object still exists in memory. The object
will get freed at some point in the future. A subsequent onlining
operation can create a new sysfs directory, even if the old object
still exists in memory, without causing any problems.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 70713fe315ed14cd1bb07d1a7f33e973d136ae3d upstream.
Use gva_t instead of unsigned int for eaddr in deliver_tlb_miss().
Signed-off-by: Mihai Caraman <mihai.caraman@freescale.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 90ff5d688e61f49f23545ffab6228bd7e87e6dc7 upstream.
In EXCEPTION_PROLOG_COMMON() we check to see if the stack pointer (r1)
is valid when coming from the kernel. If it's not valid, we die but
with a nice oops message.
Currently we allocate a stack frame (subtract INT_FRAME_SIZE) before we
check to see if the stack pointer is negative. Unfortunately, this
won't detect a bad stack where r1 is less than INT_FRAME_SIZE.
This patch fixes the check to compare the modified r1 with
-INT_FRAME_SIZE. With this, bad kernel stack pointers (including NULL
pointers) are correctly detected again.
Kudos to Paulus for finding this.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit ec67ad82814bee92251fd963bf01c7a173856555 upstream.
In a recent patch:
commit c13f20ac48328b05cd3b8c19e31ed6c132b44b42
Author: Michael Neuling <mikey@neuling.org>
powerpc/signals: Mark VSX not saved with small contexts
We fixed an issue but an improved solution was later discussed after the patch
was merged.
Firstly, this patch doesn't handle the 64bit signals case, which could also hit
this issue (but has never been reported).
Secondly, the original patch isn't clear what MSR VSX should be set to. The
new approach below always clears the MSR VSX bit (to indicate no VSX is in the
context) and sets it only in the specific case where VSX is available (ie. when
VSX has been used and the signal context passed has space to provide the
state).
This reverts the original patch and replaces it with the improved solution. It
also adds a 64 bit version.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit c13f20ac48328b05cd3b8c19e31ed6c132b44b42 upstream.
The VSX MSR bit in the user context indicates if the context contains VSX
state. Currently we set this when the process has touched VSX at any stage.
Unfortunately, if the user has not provided enough space to save the VSX state,
we can't save it but we currently still set the MSR VSX bit.
This patch changes this to clear the MSR VSX bit when the user doesn't provide
enough space. This indicates that there is no valid VSX state in the user
context.
This is needed to support get/set/make/swapcontext for applications that use
VSX but only provide a small context. For example, getcontext in glibc
provides a smaller context since the VSX registers don't need to be saved over
the glibc function call. But since the program calling getcontext may have
used VSX, the kernel currently says the VSX state is valid when it's not. If
the returned context is then used in setcontext (ie. a small context without
VSX but with MSR VSX set), the kernel will refuse the context. This situation
has been reported by the glibc community.
Based on patch from Carlos O'Donell.
Tested-by: Haren Myneni <haren@linux.vnet.ibm.com>
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 84b073868b9d9e754ae48b828337633d1b386482 upstream.
When reading from the dispatch trace log (dtl) userspace interface, I
sometimes see duplicate entries. One example:
# hexdump -C dtl.out
00000000 07 04 00 0c 00 00 48 44 00 00 00 00 00 00 00 00
00000010 00 0c a0 b4 16 83 6d 68 00 00 00 00 00 00 00 00
00000020 00 00 00 00 10 00 13 50 80 00 00 00 00 00 d0 32
00000030 07 04 00 0c 00 00 48 44 00 00 00 00 00 00 00 00
00000040 00 0c a0 b4 16 83 6d 68 00 00 00 00 00 00 00 00
00000050 00 00 00 00 10 00 13 50 80 00 00 00 00 00 d0 32
The problem is in scan_dispatch_log() where we call dtl_consumer()
but bail out before incrementing the index.
To fix this I moved dtl_consumer() after the timebase comparison.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 411cabf79e684171669ad29a0628c400b4431e95 upstream.
Commit e82b89a6f19bae73fb064d1b3dd91fcefbb478f4 used strcat instead of
strcpy which can result in an overflow of newlines on the buffer.
Signed-off-by: Prarit Bhargava
Cc: benh@kernel.crashing.org
Cc: ben@decadent.org.uk
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit e82b89a6f19bae73fb064d1b3dd91fcefbb478f4 upstream.
modalias_show() should return an empty string on error, not -ENODEV.
This causes the following false and annoying error:
> find /sys/devices -name modalias -print0 | xargs -0 cat >/dev/null
cat: /sys/devices/vio/4000/modalias: No such device
cat: /sys/devices/vio/4001/modalias: No such device
cat: /sys/devices/vio/4002/modalias: No such device
cat: /sys/devices/vio/4004/modalias: No such device
cat: /sys/devices/vio/modalias: No such device
Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 3f0116c3238a96bc18ad4b4acefe4e7be32fa861 upstream.
Fengguang Wu, Oleg Nesterov and Peter Zijlstra tracked down
a kernel crash to a GCC bug: GCC miscompiles certain 'asm goto'
constructs, as outlined here:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58670
Implement a workaround suggested by Jakub Jelinek.
Reported-and-tested-by: Fengguang Wu <fengguang.wu@intel.com>
Reported-by: Oleg Nesterov <oleg@redhat.com>
Reported-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Suggested-by: Jakub Jelinek <jakub@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.2:
- Drop inapplicable changes
- Adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit cfc860253abd73e1681696c08ea268d33285a2c4 upstream.
This fixes a typo in the code that saves the guest DSCR (Data Stream
Control Register) into the kvm_vcpu_arch struct on guest exit. The
effect of the typo was that the DSCR value was saved in the wrong place,
so changes to the DSCR by the guest didn't persist across guest exit
and entry, and some host kernel memory got corrupted.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Alexander Graf <agraf@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 5676005acf26ab7e924a8438ea4746e47d405762 upstream.
need set '\0' for 'local_buffer'.
SPLPAR_MAXLENGTH is 1026, RTAS_DATA_BUF_SIZE is 4096. so the contents of
rtas_data_buf may truncated in memcpy.
if contents are really truncated.
the splpar_strlen is more than 1026. the next while loop checking will
not find the end of buffer. that will cause memory access violation.
Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 1cf389df090194a0976dc867b7fffe99d9d490cb upstream.
Under heavy (DLPAR?) stress, we tripped this panic() in
arch/powerpc/kernel/iommu.c::iommu_init_table():
page = alloc_pages_node(nid, GFP_ATOMIC, get_order(sz));
if (!page)
panic("iommu_init_table: Can't allocate %ld bytes\n", sz);
Before the panic() we got a page allocation failure for an order-2
allocation. There appears to be memory free, but perhaps not in the
ATOMIC context. I looked through all the call-sites of
iommu_init_table() and didn't see any obvious reason to need an ATOMIC
allocation. Most call-sites in fact have an explicit GFP_KERNEL
allocation shortly before the call to iommu_init_table(), indicating we
are not in an atomic context. There is some indirection for some paths,
but I didn't see any locks indicating that GFP_KERNEL is inappropriate.
With this change under the same conditions, we have not been able to
reproduce the panic.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit d1211af3049f4c9c1d8d4eb8f8098cc4f4f0d0c7 upstream.
arch/powerpc/kernel/sysfs.c exports PURR with write permission.
This may be valid for kernel in phyp mode. But writing to
the file in guest mode causes crash due to a priviledge violation
Signed-off-by: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2:
- Adjust context
- CPUs are sysdev and we must use the sysdev API]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 8f21bd0090052e740944f9397e2be5ac7957ded7 upstream.
The csum_partial_copy_generic() function saves the PowerPC non-volatile
r14, r15, and r16 registers for the main checksum-and-copy loop.
Unfortunately, it fails to restore them upon error exit from this loop,
which results in silent corruption of these registers in the presumably
rare event of an access exception within that loop.
This commit therefore restores these register on error exit from the loop.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: register name macros use lower-case 'r']
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit d9813c3681a36774b254c0cdc9cce53c9e22c756 upstream.
The csum_partial_copy_generic() uses register r7 to adjust the remaining
bytes to process. Unfortunately, r7 also holds a parameter, namely the
address of the flag to set in case of access exceptions while reading
the source buffer. Lacking a quantum implementation of PowerPC, this
commit instead uses register r9 to do the adjusting, leaving r7's
pointer uncorrupted.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 230aef7a6a23b6166bd4003bfff5af23c9bd381f upstream.
Normally when we haven't implemented an alignment handler for
a load or store instruction the process will be terminated.
The alignment handler uses the DSISR (or a pseudo one) to locate
the right handler. Unfortunately ldbrx and stdbrx overlap lfs and
stfs so we incorrectly think ldbrx is an lfs and stdbrx is an
stfs.
This bug is particularly nasty - instead of terminating the
process we apply an incorrect fixup and continue on.
With more and more overlapping instructions we should stop
creating a pseudo DSISR and index using the instruction directly,
but for now add a special case to catch ldbrx/stdbrx.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit bdbc29c19b2633b1d9c52638fb732bcde7a2031a upstream.
On 64-bit, __pa(&static_var) gets miscompiled by recent versions of
gcc as something like:
addis 3,2,.LANCHOR1+4611686018427387904@toc@ha
addi 3,3,.LANCHOR1+4611686018427387904@toc@l
This ends up effectively ignoring the offset, since its bottom 32 bits
are zero, and means that the result of __pa() still has 0xC in the top
nibble. This happens with gcc 4.8.1, at least.
To work around this, for 64-bit we make __pa() use an AND operator,
and for symmetry, we make __va() use an OR operator. Using an AND
operator rather than a subtraction ends up with slightly shorter code
since it can be done with a single clrldi instruction, whereas it
takes three instructions to form the constant (-PAGE_OFFSET) and add
it on. (Note that MEMORY_START is always 0 on 64-bit.)
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit f5f6cbb61610b7bf9d9d96db9c3979d62a424bab upstream.
/proc/powerpc/lparcfg is an ancient facility (though still actively used)
which allows access to some informations relative to the partition when
running underneath a PAPR compliant hypervisor.
It makes no sense on non-pseries machines. However, currently, not only
can it be created on these if the kernel has pseries support, but accessing
it on such a machine will crash due to trying to do hypervisor calls.
In fact, it should also not do HV calls on older pseries that didn't have
an hypervisor either.
Finally, it has the plumbing to be a module but is a "bool" Kconfig option.
This fixes the whole lot by turning it into a machine_device_initcall
that is only created on pseries, and adding the necessary hypervisor
check before calling the H_GET_EM_PARMS hypercall
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: lparcfg_cleanup() was a bit different]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 0e0ed6406e61434d3f38fb58aa8464ec4722b77e upstream.
Module CRCs are implemented as absolute symbols that get resolved by
a linker script. We build an intermediate .o that contains an
unresolved symbol for each CRC. genksysms parses this .o, calculates
the CRCs and writes a linker script that "resolves" the symbols to
the calculated CRC.
Unfortunately the ppc64 relocatable kernel sees these CRCs as symbols
that need relocating and relocates them at boot. Commit d4703aef
(module: handle ppc64 relocating kcrctabs when CONFIG_RELOCATABLE=y)
added a hook to reverse the bogus relocations. Part of this patch
created a symbol at 0x0:
# head -2 /proc/kallsyms
0000000000000000 T reloc_start
c000000000000000 T .__start
This reloc_start symbol is causing lots of confusion to perf. It
thinks reloc_start is a massive function that stretches from 0x0 to
0xc000000000000000 and we get various cryptic errors out of perf,
including:
problem incrementing symbol count, skipping event
This patch removes the reloc_start linker script label and instead
defines it as PHYSICAL_START. We also need to wrap it with
CONFIG_PPC64 because the ppc32 kernel can set a non zero
PHYSICAL_START at compile time and we wouldn't want to subtract
it from the CRCs in that case.
Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
spinning_secondaries
commit 8246aca7058f3f2c2ae503081777965cd8df7b90 upstream.
the smp_release_cpus is a normal funciton and called in normal environments,
but it calls the __initdata spinning_secondaries.
need modify spinning_secondaries to match smp_release_cpus.
the related warning:
(the linker report boot_paca.33377, but it should be spinning_secondaries)
-----------------------------------------------------------------------------
WARNING: arch/powerpc/kernel/built-in.o(.text+0x23176): Section mismatch in reference from the function .smp_release_cpus() to the variable .init.data:boot_paca.33377
The function .smp_release_cpus() references
the variable __initdata boot_paca.33377.
This is often because .smp_release_cpus lacks a __initdata
annotation or the annotation of boot_paca.33377 is wrong.
WARNING: arch/powerpc/kernel/built-in.o(.text+0x231fe): Section mismatch in reference from the function .smp_release_cpus() to the variable .init.data:boot_paca.33377
The function .smp_release_cpus() references
the variable __initdata boot_paca.33377.
This is often because .smp_release_cpus lacks a __initdata
annotation or the annotation of boot_paca.33377 is wrong.
-----------------------------------------------------------------------------
Signed-off-by: Chen Gang <gang.chen@asianux.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit bf593907f7236e95698a76b7c7a2bbf8b1165327 upstream.
Normally, the kernel emulates a few instructions that are unimplemented
on some processors (e.g. the old dcba instruction), or privileged (e.g.
mfpvr). The emulation of unimplemented instructions is currently not
working on the PowerNV platform. The reason is that on these machines,
unimplemented and illegal instructions cause a hypervisor emulation
assist interrupt, rather than a program interrupt as on older CPUs.
Our vector for the emulation assist interrupt just calls
program_check_exception() directly, without setting the bit in SRR1
that indicates an illegal instruction interrupt. This fixes it by
making the emulation assist interrupt set that bit before calling
program_check_interrupt(). With this, old programs that use no-longer
implemented instructions such as dcba now work again.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 120496ac2d2d60aee68d3123a68169502a85f4b5 upstream.
This patch brings online all threads which are present but not online
prior to migration/hibernation. After migration/hibernation those
threads are taken back offline.
During migration/hibernation all online CPUs must call H_JOIN, this is
required by the hypervisor. Without this patch, threads that are offline
(H_CEDE'd) will not be woken to make the H_JOIN call and the OS will be
deadlocked (all threads either JOIN'd or CEDE'd).
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 444080d13d05dc38d07dd3bf751d38bce7ab7c72 upstream.
This fixes a hang that was observed during live partition migration.
Since stop_topology_update must not be called from an interrupt
context, call it earlier in the migration process. The hang observed
can be seen below:
WARNING: at kernel/timer.c:1011
Modules linked in: ip6t_LOG xt_tcpudp xt_pkttype ipt_LOG xt_limit ip6t_REJECT nf_conntrack_ipv6 nf_defrag_ipv6 ip6table_raw xt_NOTRACK ipt_REJECT xt_state iptable_raw iptable_filter ip6table_mangle nf_conntrack_netbios_ns nf_conntrack_broadcast nf_conntrack_ipv4 nf_conntrack nf_defrag_ipv4 ip_tables ip6table_filter ip6_tables x_tables ipv6 fuse loop ibmveth sg ext3 jbd mbcache raid456 async_raid6_recov async_pq raid6_pq async_xor xor async_memcpy async_tx raid10 raid1 raid0 scsi_dh_alua scsi_dh_rdac scsi_dh_hp_sw scsi_dh_emc dm_round_robin dm_multipath scsi_dh sd_mod crc_t10dif ibmvfc scsi_transport_fc scsi_tgt scsi_mod dm_snapshot dm_mod
NIP: c0000000000c52d8 LR: c00000000004be28 CTR: 0000000000000000
REGS: c00000005ffd77d0 TRAP: 0700 Not tainted (3.2.0-git-00001-g07d106d)
MSR: 8000000000021032 <ME,CE,IR,DR> CR: 48000084 XER: 00000001
CFAR: c00000000004be20
TASK = c00000005ec78860[0] 'swapper/3' THREAD: c00000005ec98000 CPU: 3
GPR00: 0000000000000001 c00000005ffd7a50 c000000000fbbc98 c000000000ec8340
GPR04: 00000000282a0020 0000000000000000 0000000000004000 0000000000000101
GPR08: 0000000000000012 c00000005ffd4000 0000000000000020 c000000000f3ba88
GPR12: 0000000000000000 c000000007f40900 0000000000000001 0000000000000004
GPR16: 0000000000000001 0000000000000000 0000000000000000 c000000001022310
GPR20: 0000000000000001 0000000000000000 0000000000200200 c000000001029e14
GPR24: 0000000000000000 0000000000000001 0000000000000040 c00000003f74bc80
GPR28: c00000003f74bc84 c000000000f38038 c000000000f16b58 c000000000ec8340
NIP [c0000000000c52d8] .del_timer_sync+0x28/0x60
LR [c00000000004be28] .stop_topology_update+0x20/0x38
Call Trace:
[c00000005ffd7a50] [c00000005ec78860] 0xc00000005ec78860 (unreliable)
[c00000005ffd7ad0] [c00000000004be28] .stop_topology_update+0x20/0x38
[c00000005ffd7b40] [c000000000028378] .__rtas_suspend_last_cpu+0x58/0x260
[c00000005ffd7bf0] [c0000000000fa230] .generic_smp_call_function_interrupt+0x160/0x358
[c00000005ffd7cf0] [c000000000036ec8] .smp_ipi_demux+0x88/0x100
[c00000005ffd7d80] [c00000000005c154] .icp_hv_ipi_action+0x5c/0x80
[c00000005ffd7e00] [c00000000012a088] .handle_irq_event_percpu+0x100/0x318
[c00000005ffd7f00] [c00000000012e774] .handle_percpu_irq+0x84/0xd0
[c00000005ffd7f90] [c000000000022ba8] .call_handle_irq+0x1c/0x2c
[c00000005ec9ba20] [c00000000001157c] .do_IRQ+0x22c/0x2a8
[c00000005ec9bae0] [c0000000000054bc] hardware_interrupt_entry+0x18/0x1c
Exception: 501 at .cpu_idle+0x194/0x2f8
LR = .cpu_idle+0x194/0x2f8
[c00000005ec9bdd0] [c000000000017e58] .cpu_idle+0x188/0x2f8 (unreliable)
[c00000005ec9be90] [c00000000067ec18] .start_secondary+0x3e4/0x524
[c00000005ec9bf90] [c0000000000093e8] .start_secondary_prolog+0x10/0x14
Instruction dump:
ebe1fff8 4e800020 fbe1fff8 7c0802a6 f8010010 7c7f1b78 f821ff81 78290464
80090014 5400019e 7c0000d0 78000fe0 <0b000000> 4800000c 7c210b78 7c421378
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 7122beeee7bc1757682049780179d7c216dd1c83 upstream.
The following commit breaks numa distance setup for old powerpc
systems that use form0 encoding in device tree.
commit 41eab6f88f24124df89e38067b3766b7bef06ddb
powerpc/numa: Use form 1 affinity to setup node distance
Device tree node /rtas/ibm,associativity-reference-points would
index into /cpus/PowerPCxxxx/ibm,associativity based on form0 or
form1 encoding detected by ibm,architecture-vec-5 property.
All modern systems use form1 and current kernel code is correct.
However, on older systems with form0 encoding, the numa distance
will get hard coded as LOCAL_DISTANCE for all nodes. This causes
task scheduling anomaly since scheduler will skip building numa
level domain (topmost domain with all cpus) if all numa distances
are same. (value of 'level' in sched_init_numa() will remain 0)
Prior to the above commit:
((from) == (to) ? LOCAL_DISTANCE : REMOTE_DISTANCE)
Restoring compatible behavior with this patch for old powerpc systems
with device tree where numa distance are encoded as form0.
Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 29ce3c5073057991217916abc25628e906911757 upstream.
In __after_prom_start we copy the kernel down to zero in two calls to
copy_and_flush. After the first call (copy from 0 to copy_to_here:)
we jump to the newly copied code soon after.
Unfortunately there's no isync between the copy of this code and the
jump to it. Hence it's possible that stale instructions could still be
in the icache or pipeline before we branch to it.
We've seen this on real machines and it's results in no console output
after:
calling quiesce...
returning from prom_init
The below adds an isync to ensure that the copy and flushing has
completed before any branching to the new instructions occurs.
Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 6747e83235caecd30b186d1282e4eba7679f81b7 upstream.
In commit 85fe402 (fs: do not assign default i_ino in new_inode), the
initialisation of i_ino was removed from new_inode() and pushed down
into the callers. However spufs_new_inode() was not updated.
This exhibits as no files appearing in /spu, because all our dirents
have a zero inode, which readdir() seems to dislike.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
performed before the ANDCOND test
commit 9fb2640159f9d4f5a2a9d60e490482d4cbecafdb upstream.
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.
Signed-off-by: Michael Wolf <mjw@linux.vnet.ibm.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
flush_signal_handlers() needs to know whether sigaction::sa_restorer
is defined, not whether SA_RESTORER is defined. Define the
__ARCH_HAS_SA_RESTORER macro to indicate this.
Vaguely based on upstream commit 574c4866e33d 'consolidate kernel-side
struct sigaction declarations'.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
|
|
commit d63ac5f6cf31c8a83170a9509b350c1489a7262b upstream.
Commit 44ae3ab3358e962039c36ad4ae461ae9fb29596c forgot to update
the entry for the 970MP rev 1.0 processor when moving some CPU
features bits to the MMU feature bit mask. This breaks booting
on some rare G5 models using that chip revision.
Reported-by: Phileas Fogg <phileas-fogg@mail.ru>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
This reverts commit 066f289835f09a3f744d6bac96f25e25d20b3ded which was
6a040ce72598159a74969a2d01ab0ba5ee6536b3 upstream.
This was not needed and is not suitable for 3.2.y.
Reported-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 8520e443aa56cc157b015205ea53e7b9fc831291 upstream.
Disable hard IRQ before kexec a new kernel image.
Not doing it can result in corrupted data in the memory segments
reserved for the new kernel.
Signed-off-by: Phileas Fogg <phileas-fogg@mail.ru>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 6a040ce72598159a74969a2d01ab0ba5ee6536b3 upstream.
The DDW code uses a eeh_dev struct from the pci_dev. However, this is
not set until eeh_add_device_late is called.
Since pci_bus_add_devices is called before eeh_add_device_late, the PCI
devices are added to the bus, making drivers' probe hooks to be called.
These will call set_dma_mask, which will call the DDW code, which will
require the eeh_dev struct from pci_dev. This would result in a crash,
due to a NULL dereference.
Calling eeh_add_device_late after pci_bus_add_devices would make the
system BUG, because device files shouldn't be added to devices there
were not added to the system. So, a new function is needed to add such
files only after pci_bus_add_devices have been called.
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Acked-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit d3286144c92ec876da9e30320afa875699b7e0f1 upstream.
Guests can trigger MMIO exits using dcbf. Since we don't emulate cache
incoherent MMIO, just do nothing and move on.
Reported-by: Ben Collins <ben.c@servergy.com>
Signed-off-by: Alexander Graf <agraf@suse.de>
Tested-by: Ben Collins <ben.c@servergy.com>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit e6449c9b2d90c1bd9a5985bf05ddebfd1631cd6b upstream.
The missing NULL terminator can cause a panic on
PPC405 boards during boot:
Linux/PowerPC load: console=ttyS0,115200 root=/dev/mtdblock1 rootfstype=squashfs,jffs2 noinitrd init=/etc/preinit
Finalizing device tree... flat tree at 0x6a5160
bootconsole [udbg0] enabled
Page fault in user mode with in_atomic() = 1 mm = (null)
NIP = c0275f50 MSR = fffffffe
Oops: Weird page fault, sig: 11 [#1]
PowerPC 40x Platform
Modules linked in:
NIP: c0275f50 LR: c0275f60 CTR: c0280000
REGS: c0275eb0 TRAP: 636f7265 Not tainted (3.7.1)
MSR: fffffffe <VEC,VSX,EE,PR,FP,ME,SE,BE,IR,DR,PMM,RI> CR: c06a6190 XER: 00000001
TASK = c02662a8[0] 'swapper' THREAD: c0274000
GPR00: c0275ec0 c000c658 c027c4bf 00000000 c0275ee0 c000a0ec c020a1a8 c020a1f0
GPR08: c020f631 c020f404 c025f078 c025f080 c0275f10
Call Trace:
---[ end trace 31fd0ba7d8756001 ]---
Kernel panic - not syncing: Attempted to kill the idle task!
The panic happens since commit 9597abe00c1bab2aedce6b49866bf6d1e81c9eed
(sections: fix section conflicts in arch/powerpc), however the root
cause of this is that the NULL terminator were not added in commit
a4f740cf33f7f6c164bbde3c0cdbcc77b0c4997c (of/flattree: Add of_flat_dt_match()
helper function).
Cc: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit ce73ec6db47af84d1466402781ae0872a9e7873c upstream.
The locking in update_vsyscall_tz() is not only unnecessary because the vdso
code copies the data unproteced in __kernel_gettimeofday() but also
introduces a hard to reproduce race condition between update_vsyscall()
and update_vsyscall_tz(), which causes user space process to loop
forever in vdso code.
The following patch removes the locking from update_vsyscall_tz().
Locking is not only unnecessary because the vdso code copies the data
unprotected in __kernel_gettimeofday() but also erroneous because updating
the tb_update_count is not atomic and introduces a hard to reproduce race
condition between update_vsyscall() and update_vsyscall_tz(), which further
causes user space process to loop forever in vdso code.
The below scenario describes the race condition,
x==0 Boot CPU other CPU
proc_P: x==0
timer interrupt
update_vsyscall
x==1 x++;sync settimeofday
update_vsyscall_tz
x==2 x++;sync
x==3 sync;x++
sync;x++
proc_P: x==3 (loops until x becomes even)
Because the ++ operator would be implemented as three instructions and not
atomic on powerpc.
A similar change was made for x86 in commit 6c260d58634
("x86: vdso: Remove bogus locking in update_vsyscall_tz")
Signed-off-by: Shan Hai <shan.hai@windriver.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 11ee7e99f35ecb15f59b21da6a82d96d2cd3fcc8 upstream.
If we build a kernel with CONFIG_RELOCATABLE=y CONFIG_CRASH_DUMP=n,
the kernel fails when we run at a non zero offset. It turns out
we were incorrectly wrapping some of the relocatable kernel code
with CONFIG_CRASH_DUMP.
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit e43a028752fed049e4bd94ef895542f96d79fa74 upstream.
When remembering the direction of a DCR transaction, we should write
to the same variable that we interpret on later when doing vcpu_run
again.
Signed-off-by: Alexander Graf <agraf@suse.de>
[bwh: Backported to 3.2: adjust context, indentation]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
Fix wii_memory_fixups() the following compile error on 3.0.y tree with
wii_defconfig on 3.0.y tree.
CC arch/powerpc/platforms/embedded6xx/wii.o
arch/powerpc/platforms/embedded6xx/wii.c: In function ‘wii_memory_fixups’:
arch/powerpc/platforms/embedded6xx/wii.c:88:2: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 2 has type ‘phys_addr_t’ [-Werror=format]
arch/powerpc/platforms/embedded6xx/wii.c:88:2: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 3 has type ‘phys_addr_t’ [-Werror=format]
arch/powerpc/platforms/embedded6xx/wii.c:90:2: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 2 has type ‘phys_addr_t’ [-Werror=format]
arch/powerpc/platforms/embedded6xx/wii.c:90:2: error: format ‘%llx’ expects argument of type ‘long long unsigned int’, but argument 3 has type ‘phys_addr_t’ [-Werror=format]
cc1: all warnings being treated as errors
make[2]: *** [arch/powerpc/platforms/embedded6xx/wii.o] Error 1
make[1]: *** [arch/powerpc/platforms/embedded6xx] Error 2
make: *** [arch/powerpc/platforms] Error 2
Signed-off-by: Shuah Khan <shuah.khan@hp.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit feadf7c0a1a7c08c74bebb4a13b755f8c40e3bbc upstream.
The EEH core is talking with the PCI device driver to determine the
action (purely reset, or PCI device removal). During the period, the
driver might be unloaded and in turn causes kernel crash as follows:
EEH: Detected PCI bus error on PHB#4-PE#10000
EEH: This PCI device has failed 3 times in the last hour
lpfc 0004:01:00.0: 0:2710 PCI channel disable preparing for reset
Unable to handle kernel paging request for data at address 0x00000490
Faulting instruction address: 0xd00000000e682c90
cpu 0x1: Vector: 300 (Data Access) at [c000000fc75ffa20]
pc: d00000000e682c90: .lpfc_io_error_detected+0x30/0x240 [lpfc]
lr: d00000000e682c8c: .lpfc_io_error_detected+0x2c/0x240 [lpfc]
sp: c000000fc75ffca0
msr: 8000000000009032
dar: 490
dsisr: 40000000
current = 0xc000000fc79b88b0
paca = 0xc00000000edb0380 softe: 0 irq_happened: 0x00
pid = 3386, comm = eehd
enter ? for help
[c000000fc75ffca0] c000000fc75ffd30 (unreliable)
[c000000fc75ffd30] c00000000004fd3c .eeh_report_error+0x7c/0xf0
[c000000fc75ffdc0] c00000000004ee00 .eeh_pe_dev_traverse+0xa0/0x180
[c000000fc75ffe70] c00000000004ffd8 .eeh_handle_event+0x68/0x300
[c000000fc75fff00] c0000000000503a0 .eeh_event_handler+0x130/0x1a0
[c000000fc75fff90] c000000000020138 .kernel_thread+0x54/0x70
1:mon>
The patch increases the reference of the corresponding driver modules
while EEH core does the negotiation with PCI device driver so that the
corresponding driver modules can't be unloaded during the period and
we're safe to refer the callbacks.
Reported-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[bwh: Backported to 3.2:
- Adjust context
- Reporting functions return int (success = 0), not void * (success = NULL)
- Assume that the 'dev' arguments are non-null]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
|
commit 9fb1b36ca1234e64a5d1cc573175303395e3354d upstream.
We have been observing hangs, both of KVM guest vcpu tasks and more
generally, where a process that is woken doesn't properly wake up and
continue to run, but instead sticks in TASK_WAKING state. This
happens because the update of rq->wake_list in ttwu_queue_remote()
is not ordered with the update of ipi_message in
smp_muxed_ipi_message_pass(), and the reading of rq->wake_list in
scheduler_ipi() is not ordered with the reading of ipi_message in
smp_ipi_demux(). Thus it is possible for the IPI receiver not to see
the updated rq->wake_list and therefore conclude that there is nothing
for it to do.
In order to make sure that anything done before smp_send_reschedule()
is ordered before anything done in the resulting call to scheduler_ipi(),
this adds barriers in smp_muxed_message_pass() and smp_ipi_demux().
The barrier in smp_muxed_message_pass() is a full barrier to ensure that
there is a full ordering between the smp_send_reschedule() caller and
scheduler_ipi(). In smp_ipi_demux(), we use xchg() rather than
xchg_local() because xchg() includes release and acquire barriers.
Using xchg() rather than xchg_local() makes sense given that
ipi_message is not just accessed locally.
This moves the barrier between setting the message and calling the
cause_ipi() function into the individual cause_ipi implementations.
Most of them -- those that used outb, out_8 or similar -- already had
a full barrier because out_8 etc. include a sync before the MMIO
store. This adds an explicit barrier in the two remaining cases.
These changes made no measurable difference to the speed of IPIs as
measured using a simple ping-pong latency test across two CPUs on
different cores of a POWER7 machine.
The analysis of the reason why processes were not waking up properly
is due to Milton Miller.
Reported-by: Milton Miller <miltonm@bga.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
|
