Pull locking/atomics update from Thomas Gleixner:
"The locking, atomics and memory model brains delivered:
- A larger update to the atomics code which reworks the ordering
barriers, consolidates the atomic primitives, provides the new
atomic64_fetch_add_unless() primitive and cleans up the include
hell.
- Simplify cmpxchg() instrumentation and add instrumentation for
xchg() and cmpxchg_double().
- Updates to the memory model and documentation"
* 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (48 commits)
locking/atomics: Rework ordering barriers
locking/atomics: Instrument cmpxchg_double*()
locking/atomics: Instrument xchg()
locking/atomics: Simplify cmpxchg() instrumentation
locking/atomics/x86: Reduce arch_cmpxchg64*() instrumentation
tools/memory-model: Rename litmus tests to comply to norm7
tools/memory-model/Documentation: Fix typo, smb->smp
sched/Documentation: Update wake_up() & co. memory-barrier guarantees
locking/spinlock, sched/core: Clarify requirements for smp_mb__after_spinlock()
sched/core: Use smp_mb() in wake_woken_function()
tools/memory-model: Add informal LKMM documentation to MAINTAINERS
locking/atomics/Documentation: Describe atomic_set() as a write operation
tools/memory-model: Make scripts executable
tools/memory-model: Remove ACCESS_ONCE() from model
tools/memory-model: Remove ACCESS_ONCE() from recipes
locking/memory-barriers.txt/kokr: Update Korean translation to fix broken DMA vs. MMIO ordering example
MAINTAINERS: Add Daniel Lustig as an LKMM reviewer
tools/memory-model: Fix ISA2+pooncelock+pooncelock+pombonce name
tools/memory-model: Add litmus test for full multicopy atomicity
locking/refcount: Always allow checked forms
...
In m68k the physical memory is described by [memory_start, memory_end] for
!MMU variant and by m68k_memory array of memory ranges for the MMU version.
This information is directly use to register the physical memory with
memblock.
The reserve_bootmem() calls are replaced with memblock_reserve() and the
bootmap bitmap allocation is simply dropped.
Since the MMU variant creates early mappings only for the small part of the
memory we force bottom-up allocations in memblock.
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Add explicit casting to unsigned long to the __va() parameter
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
The generic bitops declare __ffs as
static inline unsigned long __ffs(unsigned long word);
Convert the m68k version to match the generic declaration.
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
The dummy functions defined in <asm/io_mm.h> can be provided by
<asm-generic/io.h>.
As nommu already uses <asm-generic/io.h>, move its inclusion to
<asm/io.h>, and add/adjust include guards where appropriate.
This gets rid of lots of "statement with no effect" and "unused
variable" warnings when compile-testing.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
The mem*io define guards are applicable to all users of <asm/kmap.h>.
Hence move them, and drop the #ifdef.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
- Add missing define guard for ioremap_wt(),
- Move ARCH_HAS_IOREMAP_WT from <asm/io_mm.h> to <asm/kmap.h>, as it
is applicable to Coldfire with MMU, too,
- Fix typo s/ioremap_fillcache/ioremap_fullcache/,
- Add define guard for iounmap() for consistency with other
architectures.
Fixes: 9746882f54 ("m68k: group io mapping definitions and functions")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
Commit 397ac99c6c ("m68k: remove dead timer code") removed set_rtc_mmss()
because it was unused in 2012. However, this was itself the only user of the
mach_set_clock_mmss() callback and the many implementations of that callback,
which are equally unused.
This removes all of those as well.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
The real-time clock on m68k (and powerpc) mac systems uses an unsigned
32-bit value starting in 1904, which overflows in 2040, about two years
later than everyone else, but this gets wrapped around in the Linux
code in 2038 already because of the deprecated usage of time_t and/or
long in the conversion.
Getting rid of the deprecated interfaces makes it work until 2040 as
documented, and it could be easily extended by reinterpreting
the resulting time64_t as a positive number. For the moment, I'm
adding a WARN_ON() that triggers if we encounter a time before 1970
or after 2040 (the two are indistinguishable).
This brings it in line with the corresponding code that we have on
powerpc macintosh.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
[fthain: Adopt __u32 for the union in via_read_time(), consistent with
changes to via_write_time()]
[fthain: Use lower_32_bits() in via_write_time(), consistent with changes
to pmu_write_time() and cuda_write_time()]
[fthain: Have via_read_time() return a time64_t, consistent with changes
to pmu_read_time() and cuda_read_time()]
[fthain: Drop the pointless wraparound conditional in via_read_time()]
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
[geert: Drop WARN_ON(), as it is reported to trigger on powermac]
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
I can confirm that mac_scsi PDMA now works on these machines.
This increases sequential read throughput by a factor of 4.5.
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Booting a ColdFire m68k core with MMU enabled causes a "bad page state"
oops since commit 1d40a5ea01 ("mm: mark pages in use for page tables"):
BUG: Bad page state in process sh pfn:01ce2
page:004fefc8 count:0 mapcount:-1024 mapping:00000000 index:0x0
flags: 0x0()
raw: 00000000 00000000 00000000 fffffbff 00000000 00000100 00000200 00000000
raw: 039c4000
page dumped because: nonzero mapcount
Modules linked in:
CPU: 0 PID: 22 Comm: sh Not tainted 4.17.0-07461-g1d40a5ea01d5 #13
Fix by calling pgtable_page_dtor() in our __pte_free_tlb() code path,
so that the PG_table flag is cleared before we free the pte page.
Note that I had to change the type of pte_free() to be static from
extern. Otherwise you get a lot of warnings like this:
./arch/m68k/include/asm/mcf_pgalloc.h:80:2: warning: ‘pgtable_page_dtor’ is static but used in inline function ‘pte_free’ which is not static
pgtable_page_dtor(page);
^
And making it static is consistent with our use of this in the other
m68k pgalloc definitions of pte_free().
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
CC: Matthew Wilcox <willy@infradead.org>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Many of the inc/dec ops are mandatory, but for most architectures inc/dec are
simply trivial wrappers around their corresponding add/sub ops.
Let's make all the inc/dec ops optional, so that we can get rid of these
boilerplate wrappers.
The instrumented atomics are updated accordingly.
There should be no functional change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Palmer Dabbelt <palmer@sifive.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/lkml/20180621121321.4761-17-mark.rutland@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Some of the atomics return the result of a test applied after the atomic
operation, and almost all architectures implement these as trivial
wrappers around the underlying atomic. Specifically:
* <atomic>_inc_and_test(v) is (<atomic>_inc_return(v) == 0)
* <atomic>_dec_and_test(v) is (<atomic>_dec_return(v) == 0)
* <atomic>_sub_and_test(i, v) is (<atomic>_sub_return(i, v) == 0)
* <atomic>_add_negative(i, v) is (<atomic>_add_return(i, v) < 0)
Rather than have these definitions duplicated in all architectures, with
minor inconsistencies in formatting and documentation, let's make these
operations optional, with default fallbacks as above. Implementations
must now provide a preprocessor symbol.
The instrumented atomics are updated accordingly.
Both x86 and m68k have custom implementations, which are left as-is,
given preprocessor symbols to avoid being overridden.
There should be no functional change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Palmer Dabbelt <palmer@sifive.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/lkml/20180621121321.4761-16-mark.rutland@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Several architectures these have a near-identical implementation based
on atomic_read() and atomic_cmpxchg() which we can instead define in
<linux/atomic.h>, so let's do so, using something close to the existing
x86 implementation with try_cmpxchg().
Where an architecture provides its own atomic_fetch_add_unless(), it
must define a preprocessor symbol for it. The instrumented atomics are
updated accordingly.
Note that arch/arc's existing atomic_fetch_add_unless() had redundant
barriers, as these are already present in its atomic_cmpxchg()
implementation.
There should be no functional change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Palmer Dabbelt <palmer@sifive.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Link: https://lore.kernel.org/lkml/20180621121321.4761-7-mark.rutland@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
While __atomic_add_unless() was originally intended as a building-block
for atomic_add_unless(), it's now used in a number of places around the
kernel. It's the only common atomic operation named __atomic*(), rather
than atomic_*(), and for consistency it would be better named
atomic_fetch_add_unless().
This lack of consistency is slightly confusing, and gets in the way of
scripting atomics. Given that, let's clean things up and promote it to
an official part of the atomics API, in the form of
atomic_fetch_add_unless().
This patch converts definitions and invocations over to the new name,
including the instrumented version, using the following script:
----
git grep -w __atomic_add_unless | while read line; do
sed -i '{s/\<__atomic_add_unless\>/atomic_fetch_add_unless/}' "${line%%:*}";
done
git grep -w __arch_atomic_add_unless | while read line; do
sed -i '{s/\<__arch_atomic_add_unless\>/arch_atomic_fetch_add_unless/}' "${line%%:*}";
done
----
Note that we do not have atomic{64,_long}_fetch_add_unless(), which will
be introduced by later patches.
There should be no functional change as a result of this patch.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Palmer Dabbelt <palmer@sifive.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/lkml/20180621121321.4761-2-mark.rutland@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
<asm-generic/bitops/lock.h> is shortly going to be built on top of the
atomic_long_*() API, which introduces a nasty circular dependency for
m68k where <linux/atomic.h> pulls in <linux/bitops.h> via:
linux/atomic.h
asm/atomic.h
linux/irqflags.h
asm/irqflags.h
linux/preempt.h
asm/preempt.h
asm-generic/preempt.h
linux/thread_info.h
asm/thread_info.h
asm/page.h
asm-generic/getorder.h
linux/log2.h
linux/bitops.h
Since m68k isn't SMP and doesn't support ACQUIRE/RELEASE barriers, we
can just define the lock bitops in terms of the atomic bitops in the
<asm/bitops.h> header.
Signed-off-by: Will Deacon <will.deacon@arm.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-arm-kernel@lists.infradead.org
Cc: yamada.masahiro@socionext.com
Link: https://lore.kernel.org/lkml/1529412794-17720-3-git-send-email-will.deacon@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull m68knommu updates from Greg Ungerer:
"These changes all relate to converting the IO access functions for the
ColdFire (and all other non-MMU m68k) platforms to use asm-generic IO
instead.
This makes the IO support the same on all ColdFire (regardless of MMU
enabled or not) and means we can now support PCI in non-MMU mode.
As a bonus these changes remove more code than they add"
* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu:
m68k: fix ColdFire PCI config reads and writes
m68k: introduce iomem() macro for __iomem conversions
m68k: allow ColdFire PCI bus on MMU and non-MMU configuration
m68k: fix ioremapping for internal ColdFire peripherals
m68k: fix read/write multi-byte IO for PCI on ColdFire
m68k: don't redefine access functions if we have PCI
m68k: remove old ColdFire IO access support code
m68k: use io_no.h for MMU and non-MMU enabled ColdFire
m68k: setup PCI support code in io_no.h
m68k: group io mapping definitions and functions
m68k: rework raw access macros for the non-MMU case
m68k: use asm-generic/io.h for non-MMU io access functions
m68k: put definition guards around virt_to_phys and phys_to_virt
m68k: move *_relaxed macros into io_no.h and io_mm.h
Pull timers and timekeeping updates from Thomas Gleixner:
- Core infrastucture work for Y2038 to address the COMPAT interfaces:
+ Add a new Y2038 safe __kernel_timespec and use it in the core
code
+ Introduce config switches which allow to control the various
compat mechanisms
+ Use the new config switch in the posix timer code to control the
32bit compat syscall implementation.
- Prevent bogus selection of CPU local clocksources which causes an
endless reselection loop
- Remove the extra kthread in the clocksource code which has no value
and just adds another level of indirection
- The usual bunch of trivial updates, cleanups and fixlets all over the
place
- More SPDX conversions
* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (24 commits)
clocksource/drivers/mxs_timer: Switch to SPDX identifier
clocksource/drivers/timer-imx-tpm: Switch to SPDX identifier
clocksource/drivers/timer-imx-gpt: Switch to SPDX identifier
clocksource/drivers/timer-imx-gpt: Remove outdated file path
clocksource/drivers/arc_timer: Add comments about locking while read GFRC
clocksource/drivers/mips-gic-timer: Add pr_fmt and reword pr_* messages
clocksource/drivers/sprd: Fix Kconfig dependency
clocksource: Move inline keyword to the beginning of function declarations
timer_list: Remove unused function pointer typedef
timers: Adjust a kernel-doc comment
tick: Prefer a lower rating device only if it's CPU local device
clocksource: Remove kthread
time: Change nanosleep to safe __kernel_* types
time: Change types to new y2038 safe __kernel_* types
time: Fix get_timespec64() for y2038 safe compat interfaces
time: Add new y2038 safe __kernel_timespec
posix-timers: Make compat syscalls depend on CONFIG_COMPAT_32BIT_TIME
time: Introduce CONFIG_COMPAT_32BIT_TIME
time: Introduce CONFIG_64BIT_TIME in architectures
compat: Enable compat_get/put_timespec64 always
...
- A few time-related fixes:
- off-by-one calendar month on some classes of machines,
- Y2038 preparation,
- Build fix for ndelay() being called with a 64-bit type,
- Revive 64-bit get_user(), which is used by some Android code,
- Defconfig updates,
- Fix for a long-standing fatal bug in iounmap() on '020/030, which
was actually fixed in 2.4.23, but never in 2.5.x and later,
- Default DMA mask to avoid warning splats,
- Minor fixes and cleanups.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iF4EABYIAAYFAlsU83oACgkQisJQ/WRJ8XC+tAD/TC++gQlSUbz5ZSjZ95aGZl1O
nHjOHQpnae/R9pNFjogA/imtboTg7ukyx9Qnv7q47G8w7wGyxS+QHVTa6Hr2tEEF
=8PWk
-----END PGP SIGNATURE-----
Merge tag 'm68k-for-v4.18-tag1' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k
Pull m68k updates from Geert Uytterhoeven:
- a few time-related fixes:
- off-by-one calendar month on some classes of machines
- Y2038 preparation
- build fix for ndelay() being called with a 64-bit type
- revive 64-bit get_user(), which is used by some Android code
- defconfig updates
- fix for a long-standing fatal bug in iounmap() on '020/030, which was
actually fixed in 2.4.23, but never in 2.5.x and later
- default DMA mask to avoid warning splats
- minor fixes and cleanups
* tag 'm68k-for-v4.18-tag1' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k:
m68k: Set default dma mask for platform devices
m68k/mm: Adjust VM area to be unmapped by gap size for __iounmap()
m68k/defconfig: Update defconfigs for v4.17-rc3
m68k/uaccess: Revive 64-bit get_user()
m68k: Implement ndelay() as an inline function to force type checking/casting
zorro: Add a blank line after declarations
m68k: Use read_persistent_clock64() consistently
m68k: Fix off-by-one calendar month
m68k: Fix style, spelling, and grammar in siginfo_build_tests()
m68k/mac: Fix SWIM memory resource end address
Pull siginfo updates from Eric Biederman:
"This set of changes close the known issues with setting si_code to an
invalid value, and with not fully initializing struct siginfo. There
remains work to do on nds32, arc, unicore32, powerpc, arm, arm64, ia64
and x86 to get the code that generates siginfo into a simpler and more
maintainable state. Most of that work involves refactoring the signal
handling code and thus careful code review.
Also not included is the work to shrink the in kernel version of
struct siginfo. That depends on getting the number of places that
directly manipulate struct siginfo under control, as it requires the
introduction of struct kernel_siginfo for the in kernel things.
Overall this set of changes looks like it is making good progress, and
with a little luck I will be wrapping up the siginfo work next
development cycle"
* 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: (46 commits)
signal/sh: Stop gcc warning about an impossible case in do_divide_error
signal/mips: Report FPE_FLTUNK for undiagnosed floating point exceptions
signal/um: More carefully relay signals in relay_signal.
signal: Extend siginfo_layout with SIL_FAULT_{MCEERR|BNDERR|PKUERR}
signal: Remove unncessary #ifdef SEGV_PKUERR in 32bit compat code
signal/signalfd: Add support for SIGSYS
signal/signalfd: Remove __put_user from signalfd_copyinfo
signal/xtensa: Use force_sig_fault where appropriate
signal/xtensa: Consistenly use SIGBUS in do_unaligned_user
signal/um: Use force_sig_fault where appropriate
signal/sparc: Use force_sig_fault where appropriate
signal/sparc: Use send_sig_fault where appropriate
signal/sh: Use force_sig_fault where appropriate
signal/s390: Use force_sig_fault where appropriate
signal/riscv: Replace do_trap_siginfo with force_sig_fault
signal/riscv: Use force_sig_fault where appropriate
signal/parisc: Use force_sig_fault where appropriate
signal/parisc: Use force_sig_mceerr where appropriate
signal/openrisc: Use force_sig_fault where appropriate
signal/nios2: Use force_sig_fault where appropriate
...
- replaceme the force_dma flag with a dma_configure bus method.
(Nipun Gupta, although one patch is іncorrectly attributed to me
due to a git rebase bug)
- use GFP_DMA32 more agressively in dma-direct. (Takashi Iwai)
- remove PCI_DMA_BUS_IS_PHYS and rely on the dma-mapping API to do the
right thing for bounce buffering.
- move dma-debug initialization to common code, and apply a few cleanups
to the dma-debug code.
- cleanup the Kconfig mess around swiotlb selection
- swiotlb comment fixup (Yisheng Xie)
- a trivial swiotlb fix. (Dan Carpenter)
- support swiotlb on RISC-V. (based on a patch from Palmer Dabbelt)
- add a new generic dma-noncoherent dma_map_ops implementation and use
it for arc, c6x and nds32.
- improve scatterlist validity checking in dma-debug. (Robin Murphy)
- add a struct device quirk to limit the dma-mask to 32-bit due to
bridge/system issues, and switch x86 to use it instead of a local
hack for VIA bridges.
- handle devices without a dma_mask more gracefully in the dma-direct
code.
-----BEGIN PGP SIGNATURE-----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=6Acj
-----END PGP SIGNATURE-----
Merge tag 'dma-mapping-4.18' of git://git.infradead.org/users/hch/dma-mapping
Pull dma-mapping updates from Christoph Hellwig:
- replace the force_dma flag with a dma_configure bus method. (Nipun
Gupta, although one patch is іncorrectly attributed to me due to a
git rebase bug)
- use GFP_DMA32 more agressively in dma-direct. (Takashi Iwai)
- remove PCI_DMA_BUS_IS_PHYS and rely on the dma-mapping API to do the
right thing for bounce buffering.
- move dma-debug initialization to common code, and apply a few
cleanups to the dma-debug code.
- cleanup the Kconfig mess around swiotlb selection
- swiotlb comment fixup (Yisheng Xie)
- a trivial swiotlb fix. (Dan Carpenter)
- support swiotlb on RISC-V. (based on a patch from Palmer Dabbelt)
- add a new generic dma-noncoherent dma_map_ops implementation and use
it for arc, c6x and nds32.
- improve scatterlist validity checking in dma-debug. (Robin Murphy)
- add a struct device quirk to limit the dma-mask to 32-bit due to
bridge/system issues, and switch x86 to use it instead of a local
hack for VIA bridges.
- handle devices without a dma_mask more gracefully in the dma-direct
code.
* tag 'dma-mapping-4.18' of git://git.infradead.org/users/hch/dma-mapping: (48 commits)
dma-direct: don't crash on device without dma_mask
nds32: use generic dma_noncoherent_ops
nds32: implement the unmap_sg DMA operation
nds32: consolidate DMA cache maintainance routines
x86/pci-dma: switch the VIA 32-bit DMA quirk to use the struct device flag
x86/pci-dma: remove the explicit nodac and allowdac option
x86/pci-dma: remove the experimental forcesac boot option
Documentation/x86: remove a stray reference to pci-nommu.c
core, dma-direct: add a flag 32-bit dma limits
dma-mapping: remove unused gfp_t parameter to arch_dma_alloc_attrs
dma-debug: check scatterlist segments
c6x: use generic dma_noncoherent_ops
arc: use generic dma_noncoherent_ops
arc: fix arc_dma_{map,unmap}_page
arc: fix arc_dma_sync_sg_for_{cpu,device}
arc: simplify arc_dma_sync_single_for_{cpu,device}
dma-mapping: provide a generic dma-noncoherent implementation
dma-mapping: simplify Kconfig dependencies
riscv: add swiotlb support
riscv: only enable ZONE_DMA32 for 64-bit
...
This avoids a WARNING splat when loading the macsonic or macmace driver.
Please see commit 205e1b7f51 ("dma-mapping: warn when there is no
coherent_dma_mask").
This implementation of arch_setup_pdev_archdata() differs from the
powerpc one, in that this one avoids clobbering a device dma mask
which has already been initialized.
Cc: Christoph Hellwig <hch@lst.de>
Cc: Greg Ungerer <gerg@linux-m68k.org>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Acked-by: Greg Ungerer <gerg@linux-m68k.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
The ColdFire PCI configuration space access functions swap addressing
regions to do their work. Just letting the read/write cycles exit
the CPU core (via the ColdFire "nop" instruction) is not enough to
guarantee that the address region remapping has actually completed.
Insert a read back of the mapping register to be absolutely sure
that the remapping has completed.
This fixes an occasional boot hang during the ColdFire PCI initialization
phase.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
A lot of the ColdFire internal peripherals (clocks, timers, interrupt
controllers, etc) are addressed using constants. The only problem with
that is they are not type clean when used with __raw_read/__raw_write
and read/write - they should be of type "void __iomem". This isn't
a problem currently because the IO access functions are local macros.
To switch to using the asm-generic implementations of these we need to
clean up the types. Otherwise you get warnings like this:
In file included from ./arch/m68k/include/asm/mcfsim.h:24:0,
from arch/m68k/coldfire/intc-simr.c:20:
arch/m68k/coldfire/intc-simr.c: In function ‘init_IRQ’:
./arch/m68k/include/asm/m520xsim.h:40:29: warning: passing argument 2 of ‘__raw_writeb’ makes pointer from integer without a cast [-Wint-conversion]
#define MCFINTC0_SIMR (MCFICM_INTC0 + MCFINTC_SIMR)
^
arch/m68k/coldfire/intc-simr.c:182:21: note: in expansion of macro ‘MCFINTC0_SIMR’
__raw_writeb(0xff, MCFINTC0_SIMR);
^
In file included from ./arch/m68k/include/asm/io_no.h:120:0,
from ./arch/m68k/include/asm/io.h:3,
from ./include/linux/io.h:25,
from ./include/linux/irq.h:25,
from ./include/asm-generic/hardirq.h:13,
from ./arch/m68k/include/asm/hardirq.h:25,
from ./include/linux/hardirq.h:9,
from ./include/linux/interrupt.h:13,
from arch/m68k/coldfire/intc-simr.c:16:
./include/asm-generic/io.h:71:22: note: expected ‘volatile void *’ but argument is of type ‘unsigned int’
#define __raw_writeb __raw_writeb
^
./include/asm-generic/io.h:72:20: note: in expansion of macro ‘__raw_writeb’
static inline void __raw_writeb(u8 value, volatile void __iomem *addr)
^
To start this clean up process introduce a macro, iomem(), that converts
a constant address to the correct "void __iomem *" type.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Tested-by: Angelo Dureghello <angelo@sysam.it>
Up to now we have only had support for the PCI bus when running the
ColdFire CPU family with the MMU enabled. The only reason for this was
the incomplete state of the IO remapping and access functions when
running with the MMU disabled.
Recent fixes and improvements to the ColdFire IO access code means we
can now support the PCI bus when running non-MMU enabled as well.
So modify the configuration support to allow it to be selected no matter
what choice of MMU mode is used.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
The ColdFire SoC internal peripherals are mapped into virtual address
space using the ACR registers of the cache control unit. This means we
are using a 1:1 physical:virtual mapping for them that does not rely on
page table mappings. We can quickly determine if we are accessing an
internal peripheral device given the physical or vitrual address using
the same range check.
The implications of this mapping is that an ioremap should return the
physical address as the virtual mapping __iomem cookie as well. So fix
ioremap() to deal with this on ColdFire. Of course you need to take
care of this in the iounmap() path as well.
Reported-by: Angelo Dureghello <angelo@sysam.it>
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
We need to treat built-in peripherals and bus based address ranges
differently. Local built-in peripherals (and the ColdFire SoC parts
have quite a lot of them) are always native endian - which is big
endian on m68k/ColdFire. Bus based address ranges, like the PCI bus,
are accessed little endian - so we need to byte swap those.
So implement readw/writew and readl/writel functions to deal with
memory mapped accesses correctly based on the address range being
accessed.
This fixes readw/writew and readl/writel so that they can be used in
drivers for native SoC hardware modules (many of which are shared with
other architectures (ARM in Freescale SoC parts for example). And also
drivers for PCI devices.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Tested-by: Angelo Dureghello <angelo@sysam.it>
Some ColdFire platforms do have real PCI buses, so we should not be
re-defining or otherwise mangling the IO access functions for them.
So when CONFIG_PCI is true use the real io.h support.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
All the ColdFire IO access support code has been moved to io_no.h.
This means that all ColdFire support is at least now consistent no
matter whether the MMU is enabled or not for them.
Now that io_mm.h has reverted to only support the traditional m68k MMU
enabled processors we can remove the ColdFire specific definitions.
We can also remove the old ColdFire PCI bus IO access functions.
The new io_no.h uses asm-generic/io.h to provide all the basic support.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
Use the io_no.h IO access support for all ColdFire systems, no matter
whether configured with MMU enabled or disabled. Previously there was
subtle differences in IO access functions used in both cases, and these
resulted in broken behavior for some drivers.
As observed and reported by Angelo when using MMU enabled systems the
read/write family of functions was using little endian access, while the
non-MMU enabled systems were using native endian. This results in drivers
that are shared across Freescale processors (for some of the common
internal SoC peripherals) not working - since they are wired up for native
endian access.
This problem brings to light issues with PCI bus access and local
peripheral access - but these are not addressed with this fix.
Reported-by: Angelo Dureghello <angelo@sysam.it>
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
Ultimately we want the ColdFire IO access support to be consisent no matter
whether it is configured with MMU enabled or disabled. To acheive that we
need to get all the ColdFire IO access support code together in one place,
in this case io_no.h. The last big piece not in io_no.h is the PCI bus
support functions.
Define the IO mapping addresses required to use the asm-generic IO
access functions. They can provide everything we need - no need for us
to duplicate or have local in/out or read/write access functions.
Note that this support is not active yet, since we haven't done the
full switch over to using the asm-generic functions yet. And also note
that we do not yet remove the old PCI functions from io_mm.h yet.
Consolodating all this IO access support in a single place will make
it easier in the future to enable PCI bus support for non-MMU enabled
ColdFire (which we currently cannot do).
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
Create a new header file, kmap.h, that groups all the definitions and
functions associated with the io mapping and remapping.
Currently the functions are spread across raw_io.h and io_mm.h. And in
the future we will want to use these in io_no.h as well. So it makes
sense to move them all together into a single header file.
It is named after the arch/m68k/mm/kmap.c file that actually implements
many of the exported functions.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Tested-by: Angelo Dureghello <angelo@sysam.it>
The primary and fundamental access macros are really the __raw versions.
So make them the actual implementation for access, and not the read/write
access macros. The read/write macros and functions are built on top of
the raw access (with byte swapping or other actions as required).
This in itself causes no functional change right now. But it will make it
easier to fix and resolve problems with PCI bus access in the future.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
There is nothing really special about the non-MMU m68k IO access functions.
So we can easily switch to using the asm-generic/io.h functions.
The only thing we do need to handle is that historically the m68k IO access
functions for readw/readl/writew/writel use native CPU endian ordering. So
for us on m68k/ColdFire that means they are big-endian. Leave the existing
set of _raw_read/__raw_write and read/write macros in place to deal with
them. (They are ripe for later cleanup, but that is for another patch).
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
The non-MMU and ColdFire IO access functions will be moving to using the
asm-generic/io.h in the near future. To make that possible we need define
guards around the m68k specific virt_to_phys() and phys_to_virt()
functions.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
Move a copy of the definitions of the *_relaxed() macros into io_no.h
and io_mm.h. This precedes a change to the io_no.h file to use
asm-generic/io.h. They will be removed from io_no.h at that point.
Signed-off-by: Greg Ungerer <gerg@linux-m68k.org>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Angelo Dureghello <angelo@sysam.it>
Tested-by: Angelo Dureghello <angelo@sysam.it>
If 020/030 support is enabled, get_io_area() leaves an IO_SIZE gap
between mappings which is added to the vm_struct representing the
mapping. __ioremap() uses the actual requested size (after alignment),
while __iounmap() is passed the size from the vm_struct.
On 020/030, early termination descriptors are used to set up mappings of
extent 'size', which are validated on unmapping. The unmapped gap of
size IO_SIZE defeats the sanity check of the pmd tables, causing
__iounmap() to loop forever on 030.
On 040/060, unmapping of page table entries does not check for a valid
mapping, so the umapping loop always completes there.
Adjust size to be unmapped by the gap that had been added in the
vm_struct prior.
This fixes the hang in atari_platform_init() reported a long time ago,
and a similar one reported by Finn recently (addressed by removing
ioremap() use from the SWIM driver.
Tested on my Falcon in 030 mode - untested but should work the same on
040/060 (the extra page tables cleared there would never have been set
up anyway).
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
[geert: Minor commit description improvements]
[geert: This was fixed in 2.4.23, but not in 2.5.x]
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: stable@vger.kernel.org
Revive support for 64-bit get_user(), which was disabled in commit
d94af931af ("[PATCH] m68k: clean up uaccess.h") due to a "broken"
typeof in (then brand new) gcc-4.1.
- Keep on using u64 for the temporary, as __typeof__() doesn't drop
the const qualifier,
- Move it into a union (like mips32 does) to get rid of the cast, as
using get_user() to fetch a __user pointer would cause a "cast to
pointer from integer of different size" warning otherwise.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Martijn Coenen <maco@android.com>
ndelay() is supposed to take an unsigned long, but if you define
ndelay() as a macro and the caller pass an unsigned long long instead
of an unsigned long, the unsigned long long to unsigned long cast is
not done and we end up with an "undefined reference to `__udivdi3'"
error at link time.
Fix that by making ndelay() an inline function and then defining dummy
ndelay() macro that redirects to the ndelay() function (it's how most
archs do to implement ndelay()).
Fixes: c8ee038bd1 ("m68k: Implement ndelay() based on the existing udelay() logic")
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
[geert: Remove comment now it is no longer a macro]
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
We have two ways of getting the current time from a platform at boot
or during suspend: either using read_persistent_clock() or the rtc
class operation. We never need both, so I'm hiding the
read_persistent_clock variant when the generic RTC is enabled.
Since read_persistent_clock() and mktime() are deprecated because of
the y2038 overflow of time_t, we should use the time64_t based
replacements here.
Finally, the dependency on CONFIG_ARCH_USES_GETTIMEOFFSET looks
completely bogus in this case, so let's remove that. It was
added in commit b13b3f51ff ("m68k: fix inclusion of
arch_gettimeoffset for non-MMU 68k classic CPU types") to deal
with arch_gettimeoffset(), which has since been removed from
this file and is unrelated to the RTC functions.
The rtc accessors are only used by classic machines, while
coldfire uses proper RTC drivers, so we can put the old
ifdef back around both functions.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
This fixes a bug in read_persistent_clock() which causes the system
clock to lag the Real Time Clock by one month. The problem was noticed
on a Mac, but theoretically it must also affect Atari, BVME6000 and Q40.
The tm_mon value in the struct rtc_time passed to mach_hwclk() is
zero-based, and atari_mste_hwclk(), atari_tt_hwclk(), bvme6000_hwclk(),
mac_hwclk() and q40_hwclk() all make this adjustment. Unfortunately,
dn_dummy_hwclk(), mvme147_hwclk(), mvme16x_hwclk(), sun3_hwclk() and
sun3x_hwclk() fail to decrement tm_mon. Also m68328_hwclk() assumes
a one-based tm_mon.
Bring these platforms into line and fix read_persistent_clock() so it
works correctly on all m68k platforms.
The datasheets for the RTC devices found on the affected platforms
all confirm that the year is stored as a value in the range 0-99 and
the month is stored as a value in the range 1-12. Please refer to the
datasheets for MC146818 (Apollo), DS1643 (MVME), ICM7170 (Sun 3)
and M48T02 (Sun 3x).
Reported-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Variants of proc_create{,_data} that directly take a seq_file show
callback and drastically reduces the boilerplate code in the callers.
All trivial callers converted over.
Signed-off-by: Christoph Hellwig <hch@lst.de>
This was used by the ide, scsi and networking code in the past to
determine if they should bounce payloads. Now that the dma mapping
always have to support dma to all physical memory (thanks to swiotlb
for non-iommu systems) there is no need to this crude hack any more.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Palmer Dabbelt <palmer@sifive.com> (for riscv)
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Filling in struct siginfo before calling force_sig_info a tedious and
error prone process, where once in a great while the wrong fields
are filled out, and siginfo has been inconsistently cleared.
Simplify this process by using the helper force_sig_fault. Which
takes as a parameters all of the information it needs, ensures
all of the fiddly bits of filling in struct siginfo are done properly
and then calls force_sig_info.
In short about a 5 line reduction in code for every time force_sig_info
is called, which makes the calling function clearer.
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: linux-m68k@lists.linux-m68k.org
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
