commit 30d6e0a4190d37740e9447e4e4815f06992dd8c3 upstream.
There is code duplicated over all architecture's headers for
futex_atomic_op_inuser. Namely op decoding, access_ok check for uaddr,
and comparison of the result.
Remove this duplication and leave up to the arches only the needed
assembly which is now in arch_futex_atomic_op_inuser.
This effectively distributes the Will Deacon's arm64 fix for undefined
behaviour reported by UBSAN to all architectures. The fix was done in
commit 5f16a046f8e1 (arm64: futex: Fix undefined behaviour with
FUTEX_OP_OPARG_SHIFT usage). Look there for an example dump.
And as suggested by Thomas, check for negative oparg too, because it was
also reported to cause undefined behaviour report.
Note that s390 removed access_ok check in d12a29703 ("s390/uaccess:
remove pointless access_ok() checks") as access_ok there returns true.
We introduce it back to the helper for the sake of simplicity (it gets
optimized away anyway).
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Russell King <rmk+kernel@armlinux.org.uk>
Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com> [s390]
Acked-by: Chris Metcalf <cmetcalf@mellanox.com> [for tile]
Reviewed-by: Darren Hart (VMware) <dvhart@infradead.org>
Reviewed-by: Will Deacon <will.deacon@arm.com> [core/arm64]
Cc: linux-mips@linux-mips.org
Cc: Rich Felker <dalias@libc.org>
Cc: linux-ia64@vger.kernel.org
Cc: linux-sh@vger.kernel.org
Cc: peterz@infradead.org
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: sparclinux@vger.kernel.org
Cc: Jonas Bonn <jonas@southpole.se>
Cc: linux-s390@vger.kernel.org
Cc: linux-arch@vger.kernel.org
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: linux-hexagon@vger.kernel.org
Cc: Helge Deller <deller@gmx.de>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Matt Turner <mattst88@gmail.com>
Cc: linux-snps-arc@lists.infradead.org
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: linux-xtensa@linux-xtensa.org
Cc: Stefan Kristiansson <stefan.kristiansson@saunalahti.fi>
Cc: openrisc@lists.librecores.org
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Stafford Horne <shorne@gmail.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: Richard Henderson <rth@twiddle.net>
Cc: Chris Zankel <chris@zankel.net>
Cc: Michal Simek <monstr@monstr.eu>
Cc: Tony Luck <tony.luck@intel.com>
Cc: linux-parisc@vger.kernel.org
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Kuo <rkuo@codeaurora.org>
Cc: linux-alpha@vger.kernel.org
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: linuxppc-dev@lists.ozlabs.org
Cc: "David S. Miller" <davem@davemloft.net>
Link: http://lkml.kernel.org/r/20170824073105.3901-1-jslaby@suse.cz
Cc: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 79435ac78d160e4c245544d457850a56f805ac0d upstream.
This used to setup the LP_COUNT register automatically, but now has been
removed.
There was an earlier fix 3c7c7a2fc8811 which fixed instance in delay.h but
somehow missed this one as gcc change had not made its way into
production toolchains and was not pedantic as it is now !
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1ee55a8f7f6b7ca4c0c59e0b4b4e3584a085c2d3 upstream.
I recently came upon a scenario where I would get a double fault
machine check exception tiriggered by a kernel module.
However the ensuing crash stacktrace (ksym lookup) was not working
correctly.
Turns out that machine check auto-disables MMU while modules are allocated
in kernel vaddr spapce.
This patch re-enables the MMU before start printing the stacktrace
making stacktracing of modules work upon a fatal exception.
Signed-off-by: Jose Abreu <joabreu@synopsys.com>
Reviewed-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
[vgupta: moved code into low level handler to avoid in 2 places]
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 7d79cee2c6540ea64dd917a14e2fd63d4ac3d3c0 upstream.
It is necessary to explicitly set both SLC_AUX_RGN_START1 and SLC_AUX_RGN_END1
which hold MSB bits of the physical address correspondingly of region start
and end otherwise SLC region operation is executed in unpredictable manner
Without this patch, SLC flushes on HSDK (IOC disabled) were taking
seconds.
Reported-by: Vladimir Kondratiev <vladimir.kondratiev@intel.com>
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
[vgupta: PAR40 regs only written if PAE40 exist]
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1be7107fbe18eed3e319a6c3e83c78254b693acb upstream.
Stack guard page is a useful feature to reduce a risk of stack smashing
into a different mapping. We have been using a single page gap which
is sufficient to prevent having stack adjacent to a different mapping.
But this seems to be insufficient in the light of the stack usage in
userspace. E.g. glibc uses as large as 64kB alloca() in many commonly
used functions. Others use constructs liks gid_t buffer[NGROUPS_MAX]
which is 256kB or stack strings with MAX_ARG_STRLEN.
This will become especially dangerous for suid binaries and the default
no limit for the stack size limit because those applications can be
tricked to consume a large portion of the stack and a single glibc call
could jump over the guard page. These attacks are not theoretical,
unfortunatelly.
Make those attacks less probable by increasing the stack guard gap
to 1MB (on systems with 4k pages; but make it depend on the page size
because systems with larger base pages might cap stack allocations in
the PAGE_SIZE units) which should cover larger alloca() and VLA stack
allocations. It is obviously not a full fix because the problem is
somehow inherent, but it should reduce attack space a lot.
One could argue that the gap size should be configurable from userspace,
but that can be done later when somebody finds that the new 1MB is wrong
for some special case applications. For now, add a kernel command line
option (stack_guard_gap) to specify the stack gap size (in page units).
Implementation wise, first delete all the old code for stack guard page:
because although we could get away with accounting one extra page in a
stack vma, accounting a larger gap can break userspace - case in point,
a program run with "ulimit -S -v 20000" failed when the 1MB gap was
counted for RLIMIT_AS; similar problems could come with RLIMIT_MLOCK
and strict non-overcommit mode.
Instead of keeping gap inside the stack vma, maintain the stack guard
gap as a gap between vmas: using vm_start_gap() in place of vm_start
(or vm_end_gap() in place of vm_end if VM_GROWSUP) in just those few
places which need to respect the gap - mainly arch_get_unmapped_area(),
and and the vma tree's subtree_gap support for that.
Original-patch-by: Oleg Nesterov <oleg@redhat.com>
Original-patch-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Tested-by: Helge Deller <deller@gmx.de> # parisc
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[wt: backport to 4.11: adjust context]
[wt: backport to 4.9: adjust context ; kernel doc was not in admin-guide]
[wt: backport to 4.4: adjust context ; drop ppc hugetlb_radix changes]
Signed-off-by: Willy Tarreau <w@1wt.eu>
[gkh: minor build fixes for 4.4]
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ecd43afdbe72017aefe48080631eb625e177ef4d upstream.
This is not exposed to userspace debugers yet, which can be done
independently as a seperate patch !
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9aed02feae57bf7a40cb04ea0e3017cb7a998db4 upstream.
After emulating an unaligned access in delay slot of a branch, we
pretend as the delay slot never happened - so return back to actual
branch target (or next PC if branch was not taken).
Curently we did this by handling STATUS32.DE, we also need to clear the
BTA.T bit, which is disregarded when returning from original misaligned
exception, but could cause weirdness if it took the interrupt return
path (in case interrupt was acive too)
One ARC700 customer ran into this when enabling unaligned access fixup
for kernel mode accesses as well
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 36425cd67052e3becf325fd4d3ba5691791ef7e4 upstream.
commit 3c7c7a2fc8811bc ("ARC: Don't use "+l" inline asm constraint")
modified the inline assembly to setup LP_COUNT register manually and NOT
rely on gcc to do it (with the +l inline assembler contraint hint, now
being retired in the compiler)
However the fix was flawed as we didn't add LP_COUNT to asm clobber list,
meaning gcc doesn't know that LP_COUNT or zero-delay-loops are in action
in the inline asm.
This resulted in some fun - as nested ZOL loops were being generared
| mov lp_count,250000 ;16 # tmp235,
| lp .L__GCC__LP14 # <======= OUTER LOOP (gcc generated)
| .L14:
| ld r2, [r5] # MEM[(volatile u32 *)prephitmp_43], w
| dmb 1
| breq r2, -1, @.L21 #, w,,
| bbit0 r2,1,@.L13 # w,,
| ld r4,[r7] ;25 # loops_per_jiffy, loops_per_jiffy
| mpymu r3,r4,r6 #, loops_per_jiffy, tmp234
|
| mov lp_count, r3 # <====== INNER LOOP (from inline asm)
| lp 1f
| nop
| 1:
| nop_s
| .L__GCC__LP14: ; loop end, start is @.L14 #,
This caused issues with drivers relying on sane behaviour of udelay
friends.
With LP_COUNT added to clobber list, gcc doesn't generate the outer
loop in say above case.
Addresses STAR 9001146134
Reported-by: Joao Pinto <jpinto@synopsys.com>
Fixes: 3c7c7a2fc8811bc ("ARC: Don't use "+l" inline asm constraint")
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 08fe007968b2b45e831daf74899f79a54d73f773 upstream.
An ARC700 customer reported linux boot crashes when upgrading to bigger
L1 dcache (64K from 32K). Turns out they had an aliasing VIPT config and
current code only assumed 2 colours, while theirs had 4. So default to 4
colours and complain if there are fewer. Ideally this needs to be a
Kconfig option, but heck that's too much of hassle for a single user.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3c7c7a2fc8811bc7097479f69acf2527693d7562 upstream.
Apparenty this is coming in the way of gcc fix which inhibits the usage
of LP_COUNT as a gpr.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 922cc171998ac3dbe74d57011ef7ed57e9b0d7df upstream.
The current code doesn't even compile as somehow the inline assembly
can't see the register names defined as ARC_RTC_*
I'm pretty sure It worked when I first got it merged, but the tools were
definitely different then.
So better to write this in "C" anyways.
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 7798bf2140ebcc36eafec6a4194fffd8d585d471 upstream.
On faulting sigreturn we do get SIGSEGV, all right, but anything
we'd put into pt_regs could end up in the coredump. And since
__copy_from_user() never zeroed on arc, we'd better bugger off
on its failure without copying random uninitialized bits of
kernel stack into pt_regs...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1c3c909303924d30145601f47b6c058fdd2cbc2e upstream.
| CC mm/memory.o
| In file included from ../mm/memory.c:53:0:
| ../include/linux/pfn_t.h: In function ‘pfn_t_pte’:
| ../include/linux/pfn_t.h:78:2: error: conversion to non-scalar type requested
| return pfn_pte(pfn_t_to_pfn(pfn), pgprot);
With STRICT_MM_TYPECHECKS pte_t is a struct and the offending code
forces a cast which ends up shifting a struct and hence the gcc warning.
Note that in recent past some of the arches (aarch64, s390) made
STRICT_MM_TYPECHECKS default, but we don't for ARC as this leads to slightly
worse generated code, given ARC ABI definition of returning structs
(which pte_t would become)
Quoting from ARC ABI...
"Results of type struct are returned in a caller-supplied temporary
variable whose address is passed in r0.
For such functions, the arguments are shifted so that they are
passed in r1 and up."
So
- struct to be returned would be allocated on stack requiring extra
code at call sites
- callee updates stack memory to facilitate the return (vs. simple
MOV into return reg r0)
Hence STRICT_MM_TYPECHECKS is not enabled by default for ARC
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 45c3b08a117e2232fc8d7b9e849ead36386f4f96 upstream.
For resources shared by all cores such as SLC and IOC, only the master
core needs to do any setups / enabling / disabling etc.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 18b43e89d295cc65151c505c643c98fb2c320e59 upstream.
trace_hardirqs_on_caller() in lockdep.c expects to be called before, not
after interrupts are actually enabled.
The following comment in kernel/locking/lockdep.c substantiates this
claim:
"
/*
* We're enabling irqs and according to our state above irqs weren't
* already enabled, yet we find the hardware thinks they are in fact
* enabled.. someone messed up their IRQ state tracing.
*/
"
An example can be found in include/linux/irqflags.h:
do { trace_hardirqs_on(); raw_local_irq_enable(); } while (0)
Without this change, we hit the following DEBUG_LOCKS_WARN_ON.
[ 7.760000] ------------[ cut here ]------------
[ 7.760000] WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:2711 resume_user_mode_begin+0x48/0xf0
[ 7.770000] DEBUG_LOCKS_WARN_ON(!irqs_disabled())
[ 7.780000] Modules linked in:
[ 7.780000] CPU: 0 PID: 1 Comm: init Not tainted 4.7.0-00003-gc668bb9-dirty #366
[ 7.790000]
[ 7.790000] Stack Trace:
[ 7.790000] arc_unwind_core.constprop.1+0xa4/0x118
[ 7.800000] warn_slowpath_fmt+0x72/0x158
[ 7.800000] resume_user_mode_begin+0x48/0xf0
[ 7.810000] ---[ end trace 6f6a7a8fae20d2f0 ]---
Signed-off-by: Daniel Mentz <danielmentz@google.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 86147e3cfa5e118b61e78f4f0bf29e920dcbd477 upstream.
User mode callee regs are explicitly collected before signal delivery or
breakpoint trap. r25 is special for kernel as it serves as task pointer,
so user mode value is clobbered very early. It is saved in pt_regs where
generally only scratch (aka caller saved) regs are saved.
The code to access the corresponding pt_regs location had a subtle bug as
it was using load/store with scaling of offset, whereas the offset was already
byte wise correct. So fix this by replacing LD.AS with a standard LD
Signed-off-by: Liav Rehana <liavr@mellanox.com>
Reviewed-by: Alexey Brodkin <abrodkin@synopsys.com>
[vgupta: rewrote title and commit log]
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 20d780374c81cf237834af2202c26df2100ddd69 upstream.
ARC architecture has 2 instruction sets: ARCompact/ARCv2.
While same gcc supports compiling for either (using appropriate toggles),
we can't use the same toolchain to build kernel because libgcc needs
to be unique and the toolchian (uClibc based) is not multilibed.
uClibc toolchain is convenient since it allows all userspace and
kernel to be built with a single install for an ISA.
This however means 2 gnu installs (with same triplet prefix) are needed
for building for 2 ISA and need to be in PATH.
As developers we keep switching the builds, but would occassionally fail
to update the PATH leading to usage of wrong tools. And this would only
show up at the end of kernel build when linking incompatible libgcc.
So the initial solution was to have gcc define a special preprocessor macro
DEFAULT_CPU_xxx which is unique for default toolchain configuration.
Claudiu proposed using grep for an existing preprocessor macro which is
again uniquely defined per ISA.
Cc: Michal Marek <mmarek@suse.cz>
Suggested-by: Claudiu Zissulescu <claziss@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3925a16ae980c79d1a8fd182d7f9487da1edd4dc upstream.
LTP madvise05 was generating mm splat
| [ARCLinux]# /sd/ltp/testcases/bin/madvise05
| BUG: Bad page map in process madvise05 pte:80e08211 pmd:9f7d4000
| page:9fdcfc90 count:1 mapcount:-1 mapping: (null) index:0x0 flags: 0x404(referenced|reserved)
| page dumped because: bad pte
| addr:200b8000 vm_flags:00000070 anon_vma: (null) mapping: (null) index:1005c
| file: (null) fault: (null) mmap: (null) readpage: (null)
| CPU: 2 PID: 6707 Comm: madvise05
And for newer kernels, the system was rendered unusable afterwards.
The problem was mprotect->pte_modify() clearing PTE_SPECIAL (which is
set to identify the special zero page wired to the pte).
When pte was finally unmapped, special casing for zero page was not
done, and instead it was treated as a "normal" page, tripping on the
map counts etc.
This fixes ARC STAR 9001053308
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit f52e126cc7476196f44f3c313b7d9f0699a881fc upstream.
With recent binutils update to support dwarf CFI pseudo-ops in gas, we
now get .eh_frame vs. .debug_frame. Although the call frame info is
exactly the same in both, the CIE differs, which the current kernel
unwinder can't cope with.
This broke both the kernel unwinder as well as loadable modules (latter
because of a new unhandled relo R_ARC_32_PCREL from .rela.eh_frame in
the module loader)
The ideal solution would be to switch unwinder to .eh_frame.
For now however we can make do by just ensureing .debug_frame is
generated by removing -fasynchronous-unwind-tables
.eh_frame generated with -gdwarf-2 -fasynchronous-unwind-tables
.debug_frame generated with -gdwarf-2
Fixes STAR 9001058196
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 9bd54517ee86cb164c734f72ea95aeba4804f10b upstream.
If CONFIG_ARC_DW2_UNWIND is disabled every time arc_unwind_core()
gets called following message gets printed in debug console:
----------------->8---------------
CONFIG_ARC_DW2_UNWIND needs to be enabled
----------------->8---------------
That message makes sense if user indeed wants to see a backtrace or
get nice function call-graphs in perf but what if user disabled
unwinder for the purpose? Why pollute his debug console?
So instead we'll warn user about possibly missing feature once and
let him decide if that was what he or she really wanted.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit e5bc0478ab6cf565619224536d75ecb2aedca43b upstream.
While reviewing a different change to asm-generic/io.h Arnd spotted that
ARC ioread32 and ioread32be both of which come from asm-generic versions
are not symmetrical in terms of calling the io barriers.
generic ioread32 -> ARC readl() [ has barriers]
generic ioread32be -> __be32_to_cpu(__raw_readl()) [ lacks barriers]
While generic ioread32be is being remediated to call readl(), that involves
a swab32(), causing double swaps on ioread32be() on Big Endian systems.
So provide our versions of big endian IO accessors to ensure io barrier
calls while also keeping them optimal
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit f778cc65717687a3d3f26dd21bef62cd059f1b8b upstream.
read{l,w}() write{l,w}() primitives should use le{16,32}_to_cpu() and
cpu_to_le{16,32}() respectively to ensure device registers are read
correctly in Big Endian CPU configuration.
Per Arnd Bergmann
| Most drivers using readl() or readl_relaxed() expect those to perform byte
| swaps on big-endian architectures, as the registers tend to be fixed endian
This was needed for getting UART to work correctly on a Big Endian ARC.
The ARC accessors originally were fine, and the bug got introduced
inadventently by commit b8a0330239 ("ARCv2: barriers")
Fixes: b8a0330239 ("ARCv2: barriers")
Link: http://lkml.kernel.org/r/201603100845.30602.arnd@arndb.de
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Lada Trimasova <ltrimas@synopsys.com>
[vgupta: beefed up changelog, added Fixes/stable tags]
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit bb143f814ea488769ca2e79e0b376139cb5f134b upstream.
ARConnect/MCIP Inter-Core-Interrupt module can't send interrupt to
local core. So use core intc capability to trigger software
interrupt to self, using an unsued IRQ #21.
This showed up as csd deadlock with LTP trace_sched on a dual core
system. This test acts as scheduler fuzzer, triggering all sorts of
schedulting activity. Trouble starts with IPI to self, which doesn't get
delivered (effectively lost due to H/w capability), but the msg intended
to be sent remain enqueued in per-cpu @ipi_data.
All subsequent IPIs to this core from other cores get elided due to the
IPI coalescing optimization in ipi_send_msg_one() where a pending msg
implies an IPI already sent and assumes other core is yet to ack it.
After the elided IPI, other core simply goes into csd_lock_wait()
but never comes out as this core never sees the interrupt.
Fixes STAR 9001008624
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit cbfe74a753e877b49dc54e9b04d5d42230ca0aed upstream.
Returning to delay slot, riding an interrupti, had one loose end.
AUX_USER_SP used for restoring user mode SP upon RTIE was not being
setup from orig task's saved value, causing task to use wrong SP,
leading to ProtV errors.
The reason being:
- INTERRUPT_EPILOGUE returns to a kernel trampoline, thus not expected to restore it
- EXCEPTION_EPILOGUE is not used at all
Fix that by restoring AUX_USER_SP explicitly in the trampoline.
This was broken in the original workaround, but the error scenarios got
reduced considerably since v3.14 due to following:
1. The Linuxthreads.old based userspace at the time caused many more
exceptions in delay slot than the current NPTL based one.
Infact with current userspace the error doesn't happen at all.
2. Return from interrupt (delay slot or otherwise) doesn't get exercised much
after commit 4de0e52867 ("Really Re-enable interrupts to avoid deadlocks")
since IRQ_ACTIVE.active being clear means most returns are as if from pure
kernel (even for active interrupts)
Infact the issue only happened in an experimental branch where I was tinkering with
reverted 4de0e52867
Fixes: 4255b07f2c ("ARCv2: STAR 9000793984: Handle return from intr to Delay Slot")
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Blingly ignoring CIE.version != 1 was a bad idea.
It still leaves "desirability" when running perf with callgraphing where libgcc
symbols might show in hotspot.
More importantly, basic CIE.version == 3 support already exists in code:
|
| retAddrReg = state.version <= 1 ? *ptr++ : get_uleb128(&ptr, end);
|
Next commit with simply add continue-not-bail for CIE.version != 1
This reverts commit 323f41f9e7.
At -Os, ARC gcc generates millicode thunk for function prologue/epilogue,
which are served by libgcc.
Modules historically are NOT linked with libgcc to avoid code bloat, reducing
runtime relocation fixups etc. I even once tried doing that but got lost
in makefile intricacies.
This means modules at -Os don't get the millicode thunks, causing build
failures below:
| MODPOST 5 modules
| ERROR: "__ld_r13_to_r18" [crypto/sha256_generic.ko] undefined!
| ERROR: "__ld_r13_to_r18_ret" [crypto/sha256_generic.ko] undefined!
| ERROR: "__st_r13_to_r18" [crypto/sha256_generic.ko] undefined!
| ERROR: "__ld_r13_to_r17_ret" [crypto/sha256_generic.ko] undefined!
| ERROR: "__st_r13_to_r17" [crypto/sha256_generic.ko] undefined!
| ERROR: "__ld_r13_to_r16_ret" [crypto/sha256_generic.ko] undefined!
| ERROR: "__st_r13_to_r16" [crypto/sha256_generic.ko] undefined!
|....
|....
Workaround that by inhibiting millicode thunks for loadable modules
Fixes STAR 9000641864:
("Linux built with optimizations for size emits errors for modules")
Reported-by: Anton Kolesov <akolesov@synosys.com>
Cc: Michal Marek <mmarek@suse.cz>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
ARC700 cores with MMU v2 don't have IC_PTAG AUX register and so we only
define ARC_REG_IC_PTAG for MMU versions >= 3.
But current implementation of cache_line_loop_vX() routines assumes
availability of all of them (v2, v3 and v4) simultaneously.
And given undefined ARC_REG_IC_PTAG if CONFIG_MMU_VER=2 we're seeing
compilation problem:
---------------------------------->8-------------------------------
CC arch/arc/mm/cache.o
arch/arc/mm/cache.c: In function '__cache_line_loop_v3':
arch/arc/mm/cache.c:270:13: error: 'ARC_REG_IC_PTAG' undeclared (first use in this function)
aux_tag = ARC_REG_IC_PTAG;
^
arch/arc/mm/cache.c:270:13: note: each undeclared identifier is reported only once for each function it appears in
scripts/Makefile.build:258: recipe for target 'arch/arc/mm/cache.o' failed
---------------------------------->8-------------------------------
The simples fix is to have ARC_REG_IC_PTAG defined regardless MMU
version being used.
We don't use it in cache_line_loop_v2() anyways so who cares.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
| WARNING: vmlinux.o(.text+0xd6c2): Section mismatch in reference from the function alloc_kmap_pgtable() to the function
| .init.text:__alloc_bootmem_low()
The function alloc_kmap_pgtable() references the function __init __alloc_bootmem_low().
This is often because alloc_kmap_pgtable lacks a __init annotation or the annotation of __alloc_bootmem_low is wrong.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
This will better reflect its description i.e. "any needed setup..."
and not just do an "IPI request".
Signed-off-by: Noam Camus <noamc@ezchip.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
ARC dwarf unwinder only supports CIE version == 1
The boot time dwarf sanitizer (part of binary lookup table constructor)
would simply bail if it saw CIE version == 3, rendering unwinder with a
NULL lookup table.
It seems libgcc linked with kernel does have such entries.
With fallback linear search removed, and a NULL binary lookup table,
unwinder fails to generate any stack trace.
So allow graceful ignoring of unsupported CIE entries.
This problem was initially seen in Alexey's setup (and not mine) as he
was using buildroot built toolchain (libgcc) which doesn't get built with
CFLAGS_FOR_TARGET="-gdwarf-2 which is my default
Fixes STAR 9000985048: "kernel unwinder broken with stock tools"
Fixes: 2e22502c08 ARC: dw2 unwind: Remove falllback linear search thru FDE entries
Reported-by Alexey Brodkin <abrodkin@synopsys.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
The fix which removed linear searching of dwarf (because binary lookup
data always exists) missed out on the fact that modules don't get the
binary lookup tables info. This caused unwinding out of modules to stop
working.
So add binary lookup header setup (equivalent of eh_frame_hdr setup) to
modules as well.
While at it, confine the header setup to within unwinder code,
reducing one API exposed out of unwinder code.
Fixes: 2e22502c08 ARC: dw2 unwind: Remove falllback linear search thru FDE entries
Cc: <stable@vger.kernel.org>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
HIGHMEM support bumped the default memory size for nsim platform to 1G.
Thus total memory ended at the very edge of start of peripherals address
space. With linux link base shifted, memory started bleeding into
peripheral space which caused early boot bad_page spew !
Fixes: 29e332261d ("ARC: mm: HIGHMEM: populate high memory from DT")
Reported-by: Anton Kolesov <akolesov@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
This was the second perf intr issue
perf sampling on multicore requires intr to be enabled on all cores.
ARC perf probe code used helper arc_request_percpu_irq() which calls
- request_percpu_irq() on core0
- enable_percpu_irq() on all all cores (including core0)
genirq requires that request be made ahead of enable call.
However if perf probe happened on non core0 (observed on a 3.18 kernel),
enable would get called ahead of request, failing obviously and
rendering perf intr disabled on all such cores
[ 11.120000] 1 ARC perf : 8 counters (48 bits), 113 conditions, [overflow IRQ support]
[ 11.130000] 1 -----> enable_percpu_irq() IRQ 20 failed
[ 11.140000] 3 -----> enable_percpu_irq() IRQ 20 failed
[ 11.140000] 2 -----> enable_percpu_irq() IRQ 20 failed
[ 11.140000] 0 =====> request_percpu_irq() IRQ 20
[ 11.140000] 0 -----> enable_percpu_irq() IRQ 20
Fix this fragility, by calling request_percpu_irq() on whatever core
calls probe (there is no requirement on which core calls this anyways)
and then calling enable on each cores.
Interestingly this started as invesigation of STAR 9000838902:
"sporadically IRQs enabled on perf prob"
which was about occassional boot spew as request_percpu_irq got called
non-locally (from an IPI), and re-enabled interrupts in following path
proc_mkdir -> spin_unlock_irq()
which the irq work code didn't like.
| ARC perf : 8 counters (48 bits), 113 conditions, [overflow IRQ support]
|
| BUG: failure at ../kernel/irq_work.c:135/irq_work_run_list()!
| CPU: 0 PID: 0 Comm: swapper/0 Not tainted 3.18.10-01127-g285efb8e66d1 #2
|
| Stack Trace:
| arc_unwind_core.constprop.1+0x94/0x104
| dump_stack+0x62/0x98
| irq_work_run_list+0xb0/0xb4
| irq_work_run+0x22/0x3c
| do_IPI+0x74/0x9c
| handle_irq_event_percpu+0x34/0x164
| handle_percpu_irq+0x58/0x78
| generic_handle_irq+0x1e/0x2c
| arch_do_IRQ+0x3c/0x60
| ret_from_exception+0x0/0x8
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-snps-arc@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: <stable@vger.kernel.org> #4.2+
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
arc_request_percpu_irq() is called by all cores to request/enable percpu
irq. It has some "prep" calls needed by genirq:
- setup percpu devid
- disable IRQ_NOAUTOEN
However given that enable_percpu_irq() is called enayways, latter can be
avoided.
We are now left with irq_set_percpu_devid() quirk and that too for
ARCompact builds only, since previous patch updated ARCv2 intc to do this
in the "right" place, i.e. irq map function.
By next release, this will ultimately be fixed for ARCompact as well.
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexey Brodkin <abrodkin@synopsys.com>
Cc: linux-snps-arc@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Current ARC SDP boards cannot reliably handle 1Gbit
Ethernet connections due to limitations in hardware.
To make sure networking is stable on the board we're
limiting phy to 100 Mbit.
Signed-off-by: Alexey Brodkin <abrodkin@synopsys.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Fixes STAR 9000953410: "perf callgraph profiling causing RCU stalls"
| perf record -g -c 15000 -e cycles /sbin/hackbench
|
| INFO: rcu_preempt self-detected stall on CPU
| 1: (1 GPs behind) idle=609/140000000000002/0 softirq=2914/2915 fqs=603
| Task dump for CPU 1:
in-kernel dwarf unwinder has a fast binary lookup and a fallback linear
search (which iterates thru each of ~11K entries) thus takes 2 orders of
magnitude longer (~3 million cycles vs. 2000). Routines written in hand
assembler lack dwarf info (as we don't support assembler CFI pseudo-ops
yet) fail the unwinder binary lookup, hit linear search, failing
nevertheless in the end.
However the linear search is pointless as binary lookup tables are created
from it in first place. It is impossible to have binary lookup fail while
succeed the linear search. It is pure waste of cycles thus removed by
this patch.
This manifested as RCU stalls / NMI watchdog splat when running
hackbench under perf with callgraph profiling. The triggering condition
was perf counter overflowing in routine lacking dwarf info (like memset)
leading to patheic 3 million cycle unwinder slow path and by the time it
returned new interrupts were already pending (Timer, IPI) and taken
rightaway. The original memset didn't make forward progress, system kept
accruing more interrupts and more unwinder delayes in a vicious feedback
loop, ultimately triggering the NMI diagnostic.
Cc: stable@vger.kernel.org
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
SYNC in __switch_to() is a historic relic and not needed at all.
- In UP context it is obviously useless, why would we want to stall
the core for all updates to stack memory of t0 to complete before
loading kernel mode callee registers from t1 stack's memory.
- In SMP, there could be potential race in which outgoing task could
be concurrently picked for running on a different core, thus writes
to stack here need to be visible before the reads from stack on
other core. Peter confirmed that generic schedular already has needed
barriers (by way of rq lock) so there is no need for additional arch
barrier.
This came up when Noam was trying to replace this SYNC with EZChip
specific hardware thread scheduling instruction for their platform
support.
Link: http://lkml.kernel.org/r/20151102092654.GM17308@twins.programming.kicks-ass.net
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: linux-kernel@vger.kernel.org
Cc: Noam Camus <noamc@ezchip.com>
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Although kernel doesn't support the multiple IRQ priority levels provided
by HS38x core intc yet, ensure that the default prio value is used
anyways by relevant code.
SLEEP insn needs to be provided the IRQ priority level which can
interrupt it. This needs to be the default level which maynot
necessarily be 0 as assumed by current code.
This change allows a kernel with ARCV2_IRQ_DEF_PRIO = 1 to boot fine.
Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
