The __cpu_logical_map array is statically initialized to 0, which is a valid
MPIDR value. To prevent issues with the current implementation, this patch
defines an MPIDR_INVALID value, and statically initializes the
__cpu_logical_map[] array to it. Entries in the arm_dt_init_cpu_maps()
tmp_map array used to stash DT reg properties while parsing DT are initialized
with the MPIDR_INVALID value as well for consistency.
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Current implementation of cpu_{suspend}/cpu_{resume} relies on the MPIDR
to index the array of pointers where the context is saved and restored.
The current approach works as long as the MPIDR can be considered a
linear index, so that the pointers array can simply be dereferenced by
using the MPIDR[7:0] value.
On ARM multi-cluster systems, where the MPIDR may not be a linear index,
to properly dereference the stack pointer array, a mapping function should
be applied to it so that it can be used for arrays look-ups.
This patch adds code in the cpu_{suspend}/cpu_{resume} implementation
that relies on shifting and ORing hashing method to map a MPIDR value to a
set of buckets precomputed at boot to have a collision free mapping from
MPIDR to context pointers.
The hashing algorithm must be simple, fast, and implementable with few
instructions since in the cpu_resume path the mapping is carried out with
the MMU off and the I-cache off, hence code and data are fetched from DRAM
with no-caching available. Simplicity is counterbalanced with a little
increase of memory (allocated dynamically) for stack pointers buckets, that
should be anyway fairly limited on most systems.
Memory for context pointers is allocated in a early_initcall with
size precomputed and stashed previously in kernel data structures.
Memory for context pointers is allocated through kmalloc; this
guarantees contiguous physical addresses for the allocated memory which
is fundamental to the correct functioning of the resume mechanism that
relies on the context pointer array to be a chunk of contiguous physical
memory. Virtual to physical address conversion for the context pointer
array base is carried out at boot to avoid fiddling with virt_to_phys
conversions in the cpu_resume path which is quite fragile and should be
optimized to execute as few instructions as possible.
Virtual and physical context pointer base array addresses are stashed in a
struct that is accessible from assembly using values generated through the
asm-offsets.c mechanism.
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Colin Cross <ccross@android.com>
Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Amit Kucheria <amit.kucheria@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Dave Martin <Dave.Martin@arm.com>
Reviewed-by: Nicolas Pitre <nico@linaro.org>
Tested-by: Shawn Guo <shawn.guo@linaro.org>
Tested-by: Kevin Hilman <khilman@linaro.org>
Tested-by: Stephen Warren <swarren@wwwdotorg.org>
On ARM SMP systems, cores are identified by their MPIDR register.
The MPIDR guidelines in the ARM ARM do not provide strict enforcement of
MPIDR layout, only recommendations that, if followed, split the MPIDR
on ARM 32 bit platforms in three affinity levels. In multi-cluster
systems like big.LITTLE, if the affinity guidelines are followed, the
MPIDR can not be considered an index anymore. This means that the
association between logical CPU in the kernel and the HW CPU identifier
becomes somewhat more complicated requiring methods like hashing to
associate a given MPIDR to a CPU logical index, in order for the look-up
to be carried out in an efficient and scalable way.
This patch provides a function in the kernel that starting from the
cpu_logical_map, implement collision-free hashing of MPIDR values by checking
all significative bits of MPIDR affinity level bitfields. The hashing
can then be carried out through bits shifting and ORing; the resulting
hash algorithm is a collision-free though not minimal hash that can be
executed with few assembly instructions. The mpidr is filtered through a
mpidr mask that is built by checking all bits that toggle in the set of
MPIDRs corresponding to possible CPUs. Bits that do not toggle do not carry
information so they do not contribute to the resulting hash.
Pseudo code:
/* check all bits that toggle, so they are required */
for (i = 1, mpidr_mask = 0; i < num_possible_cpus(); i++)
mpidr_mask |= (cpu_logical_map(i) ^ cpu_logical_map(0));
/*
* Build shifts to be applied to aff0, aff1, aff2 values to hash the mpidr
* fls() returns the last bit set in a word, 0 if none
* ffs() returns the first bit set in a word, 0 if none
*/
fs0 = mpidr_mask[7:0] ? ffs(mpidr_mask[7:0]) - 1 : 0;
fs1 = mpidr_mask[15:8] ? ffs(mpidr_mask[15:8]) - 1 : 0;
fs2 = mpidr_mask[23:16] ? ffs(mpidr_mask[23:16]) - 1 : 0;
ls0 = fls(mpidr_mask[7:0]);
ls1 = fls(mpidr_mask[15:8]);
ls2 = fls(mpidr_mask[23:16]);
bits0 = ls0 - fs0;
bits1 = ls1 - fs1;
bits2 = ls2 - fs2;
aff0_shift = fs0;
aff1_shift = 8 + fs1 - bits0;
aff2_shift = 16 + fs2 - (bits0 + bits1);
u32 hash(u32 mpidr) {
u32 l0, l1, l2;
u32 mpidr_masked = mpidr & mpidr_mask;
l0 = mpidr_masked & 0xff;
l1 = mpidr_masked & 0xff00;
l2 = mpidr_masked & 0xff0000;
return (l0 >> aff0_shift | l1 >> aff1_shift | l2 >> aff2_shift);
}
The hashing algorithm relies on the inherent properties set in the ARM ARM
recommendations for the MPIDR. Exotic configurations, where for instance the
MPIDR values at a given affinity level have large holes, can end up requiring
big hash tables since the compression of values that can be achieved through
shifting is somewhat crippled when holes are present. Kernel warns if
the number of buckets of the resulting hash table exceeds the number of
possible CPUs by a factor of 4, which is a symptom of a very sparse HW
MPIDR configuration.
The hash algorithm is quite simple and can easily be implemented in assembly
code, to be used in code paths where the kernel virtual address space is
not set-up (ie cpu_resume) and instruction and data fetches are strongly
ordered so code must be compact and must carry out few data accesses.
Cc: Will Deacon <will.deacon@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Colin Cross <ccross@android.com>
Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Amit Kucheria <amit.kucheria@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Dave Martin <Dave.Martin@arm.com>
Reviewed-by: Nicolas Pitre <nico@linaro.org>
Tested-by: Shawn Guo <shawn.guo@linaro.org>
Tested-by: Kevin Hilman <khilman@linaro.org>
Tested-by: Stephen Warren <swarren@wwwdotorg.org>
nommu platforms do not perform address translation and therefore clearly
don't have TLBs. However, some SMP code assumes the presence of the TLB
flushing routines and will therefore fail to compile for a nommu system.
This patch defines dummy local_* TLB operations and #defines
tlb_ops_need_broadcast() as 0, therefore causing the usual ARM SMP TLB
operations to call the local variants instead.
Signed-off-by: Will Deacon <will.deacon@arm.com>
CC: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
CC: Nicolas Pitre <nico@linaro.org>
In ARM SMP systems the MPIDR register ([23:0] bits) is used to uniquely
identify CPUs.
In order to retrieve the logical CPU index corresponding to a given
MPIDR value and guarantee a consistent translation throughout the kernel,
this patch adds a look-up based on the MPIDR[23:0] so that kernel subsystems
can use it whenever the logical cpu index corresponding to a given MPIDR
value is needed.
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Acked-by: Nicolas Pitre <nico@linaro.org>
It turns out that the logical CPU mapping is useful even when !CONFIG_SMP
for manipulation of devices like interrupt and power controllers when
running a UP kernel on a CPU other than 0. This can happen when kexecing
a UP image from an SMP kernel.
In the future, multi-cluster systems running AMP configurations will
require something similar for mapping cluster IDs, so it makes sense to
decouple this logic in preparation for this support.
Acked-by: Yang Bai <hamo.by@gmail.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Reported-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Broadcast should not be needed when running SMP kernel on UP systems.
Also, this fixes an undefined instruction for SMP_ON_UP on earlier ARM
cores without the extended CPUID_EXT_MMFR3 register.
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
UP systems do not implement all the instructions that SMP systems have,
so in order to boot a SMP kernel on a UP system, we need to rewrite
parts of the kernel.
Do this using an 'alternatives' scheme, where the kernel code and data
is modified prior to initialization to replace the SMP instructions,
thereby rendering the problematical code ineffectual. We use the linker
to generate a list of 32-bit word locations and their replacement values,
and run through these replacements when we detect a UP system.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
ARMv7 processors like Cortex-A9 broadcast the cache maintenance
operations in hardware. This patch allows the
flush_dcache_page/update_mmu_cache pair to work in lazy flushing mode
similar to the UP case.
Note that cache flushing on SMP systems now takes place via the
set_pte_at() call (__sync_icache_dcache) and there is no race with other
CPUs executing code from the new PTE before the cache flushing took
place.
Tested-by: Rabin Vincent <rabin.vincent@stericsson.com>
Cc: Nicolas Pitre <nicolas.pitre@linaro.org>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
We suffer an unfortunate combination of "features" which makes highmem
support on platforms without hardware TLB maintainence broadcast difficult:
- we need kmap_high_get() support for DMA cache coherence
- this requires kmap_high() to take a spinlock with IRQs disabled
- kmap_high() occasionally calls flush_all_zero_pkmaps() to clear
out old mappings
- flush_all_zero_pkmaps() calls flush_tlb_kernel_range(), which
on s/w IPI'd systems eventually calls smp_call_function_many()
- smp_call_function_many() must not be called with IRQs disabled:
WARNING: at kernel/smp.c:380 smp_call_function_many+0xc4/0x240()
Modules linked in:
Backtrace:
[<c00306f0>] (dump_backtrace+0x0/0x108) from [<c0286e6c>] (dump_stack+0x18/0x1c)
r6:c007cd18 r5:c02ff228 r4:0000017c
[<c0286e54>] (dump_stack+0x0/0x1c) from [<c0053e08>] (warn_slowpath_common+0x50/0x80)
[<c0053db8>] (warn_slowpath_common+0x0/0x80) from [<c0053e50>] (warn_slowpath_null+0x18/0x1c)
r7:00000003 r6:00000001 r5:c1ff4000 r4:c035fa34
[<c0053e38>] (warn_slowpath_null+0x0/0x1c) from [<c007cd18>] (smp_call_function_many+0xc4/0x240)
[<c007cc54>] (smp_call_function_many+0x0/0x240) from [<c007cec0>] (smp_call_function+0x2c/0x38)
[<c007ce94>] (smp_call_function+0x0/0x38) from [<c005980c>] (on_each_cpu+0x1c/0x38)
[<c00597f0>] (on_each_cpu+0x0/0x38) from [<c0031788>] (flush_tlb_kernel_range+0x50/0x58)
r6:00000001 r5:00000800 r4:c05f3590
[<c0031738>] (flush_tlb_kernel_range+0x0/0x58) from [<c009c600>] (flush_all_zero_pkmaps+0xc0/0xe8)
[<c009c540>] (flush_all_zero_pkmaps+0x0/0xe8) from [<c009c6b4>] (kmap_high+0x8c/0x1e0)
[<c009c628>] (kmap_high+0x0/0x1e0) from [<c00364a8>] (kmap+0x44/0x5c)
[<c0036464>] (kmap+0x0/0x5c) from [<c0109dfc>] (cramfs_readpage+0x3c/0x194)
[<c0109dc0>] (cramfs_readpage+0x0/0x194) from [<c0090c14>] (__do_page_cache_readahead+0x1f0/0x290)
[<c0090a24>] (__do_page_cache_readahead+0x0/0x290) from [<c0090ce4>] (ra_submit+0x30/0x38)
[<c0090cb4>] (ra_submit+0x0/0x38) from [<c0089384>] (filemap_fault+0x3dc/0x438)
r4:c1819988
[<c0088fa8>] (filemap_fault+0x0/0x438) from [<c009d21c>] (__do_fault+0x58/0x43c)
[<c009d1c4>] (__do_fault+0x0/0x43c) from [<c009e8cc>] (handle_mm_fault+0x104/0x318)
[<c009e7c8>] (handle_mm_fault+0x0/0x318) from [<c0033c98>] (do_page_fault+0x188/0x1e4)
[<c0033b10>] (do_page_fault+0x0/0x1e4) from [<c0033ddc>] (do_translation_fault+0x7c/0x84)
[<c0033d60>] (do_translation_fault+0x0/0x84) from [<c002b474>] (do_DataAbort+0x40/0xa4)
r8:c1ff5e20 r7:c0340120 r6:00000805 r5:c1ff5e54 r4:c03400d0
[<c002b434>] (do_DataAbort+0x0/0xa4) from [<c002bcac>] (__dabt_svc+0x4c/0x60)
...
So we disable highmem support on these systems.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Russell King
Lorenzo Pieralisi
Will Deacon
Tony Lindgren
Catalin Marinas