| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
The late_alloc() PTE allocation function used by create_mapping_late()
does not call pgtable_page_ctor() on PTE pages it allocates, leaving
the per-page spinlock uninitialized.
Since generic page table manipulation code may assume that translation
table pages that are not owned by init_mm are covered by fully
constructed struct pages, the following crash may occur with the new
UEFI memory attributes table code.
efi: memattr: Processing EFI Memory Attributes table:
efi: memattr: 0x0000ffa16000-0x0000ffa82fff [Runtime Code |RUN| | |XP| | | | | | | | ]
Unable to handle kernel NULL pointer dereference at virtual address 00000010
pgd = c0204000
[00000010] *pgd=00000000
Internal error: Oops: 5 [#1] SMP ARM
Modules linked in:
CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.7.0-rc4-00063-g3882aa7b340b #361
Hardware name: Generic DT based system
task: ed858000 ti: ed842000 task.ti: ed842000
PC is at __lock_acquire+0xa0/0x19a8
...
[<c038c830>] (__lock_acquire) from [<c038e4f8>] (lock_acquire+0x6c/0x88)
[<c038e4f8>] (lock_acquire) from [<c0c06134>] (_raw_spin_lock+0x2c/0x3c)
[<c0c06134>] (_raw_spin_lock) from [<c0410384>] (apply_to_page_range+0xe8/0x238)
[<c0410384>] (apply_to_page_range) from [<c1205f34>] (efi_set_mapping_permissions+0x54/0x5c)
[<c1205f34>] (efi_set_mapping_permissions) from [<c1247474>] (efi_memattr_apply_permissions+0x2b8/0x378)
[<c1247474>] (efi_memattr_apply_permissions) from [<c1248258>] (arm_enable_runtime_services+0x1f0/0x22c)
[<c1248258>] (arm_enable_runtime_services) from [<c0301f0c>] (do_one_initcall+0x44/0x174)
[<c0301f0c>] (do_one_initcall) from [<c1200d10>] (kernel_init_freeable+0x90/0x1e8)
[<c1200d10>] (kernel_init_freeable) from [<c0bff690>] (kernel_init+0x8/0x114)
[<c0bff690>] (kernel_init) from [<c0307ed0>] (ret_from_fork+0x14/0x24)
The crash is due to the fact that the UEFI page tables are not owned by
init_mm, but are not covered by fully constructed struct pages.
Given that the UEFI subsystem is currently the only user of
create_mapping_late(), add an unconditional call to pgtable_page_ctor() to
late_alloc().
Fixes: 9fc68b717c24 ("ARM/efi: Apply strict permissions for UEFI Runtime Services regions")
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
To limit the amount of mapped low memory, we determine a physical address
boundary based on the start of the vmalloc area using __pa().
Strictly speaking, the vmalloc area location is arbitrary and does not
necessarily corresponds to a valid physical address. For example, if
PAGE_OFFSET = 0x80000000
PHYS_OFFSET = 0x90000000
vmalloc_min = 0xf0000000
then __pa(vmalloc_min) overflows and returns a wrapped 0 when phys_addr_t
is a 32-bit type. Then the code that follows determines that the entire
physical memory is above that boundary and no low memory gets mapped at
all:
|[...]
|Machine model: Freescale i.MX51 NA04 Board
|Ignoring RAM at 0x90000000-0xb0000000 (!CONFIG_HIGHMEM)
|Consider using a HIGHMEM enabled kernel.
To avoid this problem let's make vmalloc_limit a 64-bit value all the
time and determine that boundary explicitly without using __pa().
Reported-by: Emil Renner Berthing <kernel@esmil.dk>
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Tested-by: Emil Renner Berthing <kernel@esmil.dk>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
The dma-mapping core and the implementations do not change the DMA
attributes passed by pointer. Thus the pointer can point to const data.
However the attributes do not have to be a bitfield. Instead unsigned
long will do fine:
1. This is just simpler. Both in terms of reading the code and setting
attributes. Instead of initializing local attributes on the stack
and passing pointer to it to dma_set_attr(), just set the bits.
2. It brings safeness and checking for const correctness because the
attributes are passed by value.
Semantic patches for this change (at least most of them):
virtual patch
virtual context
@r@
identifier f, attrs;
@@
f(...,
- struct dma_attrs *attrs
+ unsigned long attrs
, ...)
{
...
}
@@
identifier r.f;
@@
f(...,
- NULL
+ 0
)
and
// Options: --all-includes
virtual patch
virtual context
@r@
identifier f, attrs;
type t;
@@
t f(..., struct dma_attrs *attrs);
@@
identifier r.f;
@@
f(...,
- NULL
+ 0
)
Link: http://lkml.kernel.org/r/1468399300-5399-2-git-send-email-k.kozlowski@samsung.com
Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Acked-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Hans-Christian Noren Egtvedt <egtvedt@samfundet.no>
Acked-by: Mark Salter <msalter@redhat.com> [c6x]
Acked-by: Jesper Nilsson <jesper.nilsson@axis.com> [cris]
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch> [drm]
Reviewed-by: Bart Van Assche <bart.vanassche@sandisk.com>
Acked-by: Joerg Roedel <jroedel@suse.de> [iommu]
Acked-by: Fabien Dessenne <fabien.dessenne@st.com> [bdisp]
Reviewed-by: Marek Szyprowski <m.szyprowski@samsung.com> [vb2-core]
Acked-by: David Vrabel <david.vrabel@citrix.com> [xen]
Acked-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> [xen swiotlb]
Acked-by: Joerg Roedel <jroedel@suse.de> [iommu]
Acked-by: Richard Kuo <rkuo@codeaurora.org> [hexagon]
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org> [m68k]
Acked-by: Gerald Schaefer <gerald.schaefer@de.ibm.com> [s390]
Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Acked-by: Hans-Christian Noren Egtvedt <egtvedt@samfundet.no> [avr32]
Acked-by: Vineet Gupta <vgupta@synopsys.com> [arc]
Acked-by: Robin Murphy <robin.murphy@arm.com> [arm64 and dma-iommu]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull smp hotplug updates from Thomas Gleixner:
"This is the next part of the hotplug rework.
- Convert all notifiers with a priority assigned
- Convert all CPU_STARTING/DYING notifiers
The final removal of the STARTING/DYING infrastructure will happen
when the merge window closes.
Another 700 hundred line of unpenetrable maze gone :)"
* 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (70 commits)
timers/core: Correct callback order during CPU hot plug
leds/trigger/cpu: Move from CPU_STARTING to ONLINE level
powerpc/numa: Convert to hotplug state machine
arm/perf: Fix hotplug state machine conversion
irqchip/armada: Avoid unused function warnings
ARC/time: Convert to hotplug state machine
clocksource/atlas7: Convert to hotplug state machine
clocksource/armada-370-xp: Convert to hotplug state machine
clocksource/exynos_mct: Convert to hotplug state machine
clocksource/arm_global_timer: Convert to hotplug state machine
rcu: Convert rcutree to hotplug state machine
KVM/arm/arm64/vgic-new: Convert to hotplug state machine
smp/cfd: Convert core to hotplug state machine
x86/x2apic: Convert to CPU hotplug state machine
profile: Convert to hotplug state machine
timers/core: Convert to hotplug state machine
hrtimer: Convert to hotplug state machine
x86/tboot: Convert to hotplug state machine
arm64/armv8 deprecated: Convert to hotplug state machine
hwtracing/coresight-etm4x: Convert to hotplug state machine
...
|
|
Pull ARM updates from Russell King:
"Included in this update are:
- Patches from Gregory Clement to fix the coherent DMA cases in our
dma-mapping code.
- A number of CPU errata updates and fixes.
- ARM cpuidle improvements from Jisheng Zhang.
- Fix from Kees for the location of _etext.
- Cleanups from Masahiro Yamada to avoid duplicated messages during
the kernel build, and remove CONFIG_ARCH_HAS_BARRIERS.
- Remove a udelay loop limitation, allowing for faster CPUs to
calibrate the delay correctly.
- Cleanup some left-overs from the SW PAN implementation.
- Ensure that a modified address limit is not visible to exception
handlers"
* 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm: (21 commits)
ARM: 8586/1: cpuidle: make arm_cpuidle_suspend() a bit more efficient
ARM: 8585/1: cpuidle: fix !cpuidle_ops[cpu].init case during init
ARM: 8561/4: dma-mapping: Fix the coherent case when iommu is used
ARM: 8561/3: dma-mapping: Don't use outer_flush_range when the L2C is coherent
ARM: 8560/1: errata: Workaround errata A12 825619 / A17 852421
ARM: 8559/1: errata: Workaround erratum A12 821420
ARM: 8558/1: errata: Workaround errata A12 818325/852422 A17 852423
ARM: save and reset the address limit when entering an exception
ARM: 8577/1: Fix Cortex-A15 798181 errata initialization
ARM: 8584/1: floppy: avoid gcc-6 warning
ARM: 8583/1: mm: fix location of _etext
ARM: 8582/1: remove unused CONFIG_ARCH_HAS_BARRIERS
ARM: 8306/1: loop_udelay: remove bogomips value limitation
ARM: 8581/1: add missing <asm/prom.h> to arch/arm/kernel/devtree.c
ARM: 8576/1: avoid duplicating "Kernel: arch/arm/boot/*Image is ready"
ARM: 8556/1: on a generic DT system: do not touch l2x0
ARM: uaccess: remove put_user() code duplication
ARM: 8580/1: Remove orphaned __addr_ok() definition
ARM: get rid of horrible *(unsigned int *)(regs + 1)
ARM: introduce svc_pt_regs structure
...
|
|
We always have vma->vm_mm around.
Link: http://lkml.kernel.org/r/1466021202-61880-8-git-send-email-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
__GFP_REPEAT has a rather weak semantic but since it has been introduced
around 2.6.12 it has been ignored for low order allocations.
PGALLOC_GFP uses __GFP_REPEAT but none of the allocation which uses this
flag is for more than order-2. This means that this flag has never been
actually useful here because it has always been used only for
PAGE_ALLOC_COSTLY requests.
Link: http://lkml.kernel.org/r/1464599699-30131-5-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Install the callbacks via the state machine and let the core invoke
the callbacks on the already online CPUs.
Signed-off-by: Richard Cochran <rcochran@linutronix.de>
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Brad Mouring <brad.mouring@ni.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Linus Walleij <linus.walleij@linaro.org>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rob Herring <robh@kernel.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-arm-kernel@lists.infradead.org
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160713153336.801270887@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
When doing dma allocation with IOMMU the __iommu_alloc_atomic() was
used even when the system was coherent. However, this function
allocates from a non-cacheable pool, which is fine when the device is
not cache coherent but won't work as expected in the device is cache
coherent. Indeed, the CPU and device must access the memory using the
same cacheability attributes.
Moreover when the devices are coherent, the mmap call must not change
the pg_prot flags in the vma struct. The arm_coherent_iommu_mmap_attrs
has been updated in the same way that it was done for the arm_dma_mmap
in commit 55af8a91640d ("ARM: 8387/1: arm/mm/dma-mapping.c: Add
arm_coherent_dma_mmap").
Suggested-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
When a L2 cache controller is used in a system that provides hardware
coherency, the entire outer cache operations are useless, and can be
skipped. Moreover, on some systems, it is harmful as it causes
deadlocks between the Marvell coherency mechanism, the Marvell PCIe
controller and the Cortex-A9.
In the current kernel implementation, the outer cache flush range
operation is triggered by the dma_alloc function.
This operation can be take place during runtime and in some
circumstances may lead to the PCIe/PL310 deadlock on Armada 375/38x
SoCs.
This patch extends the __dma_clear_buffer() function to receive a
boolean argument related to the coherency of the system. The same
things is done for the calling functions.
Reported-by: Nadav Haklai <nadavh@marvell.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Cc: <stable@vger.kernel.org> # v3.16+
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
The workaround for both errata is to set bit 24 in the diagnostic
register. There are no known end-user bugs solved by fixing this
errata, but the fix is trivial and it seems sane to apply it.
The arguments for why this needs to be in the kernel are similar to the
arugments made in the patch "Workaround errata A12 818325/852422 A17
852423".
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
This erratum has a very simple workaround (set a bit in a register), so
let's apply it. Apparently the workaround's downside is a very slight
power impact.
Note that applying this errata fixes deadlocks that are easy to
reproduce with real world applications.
The arguments for why this needs to be in the kernel are similar to the
arugments made in the patch "Workaround errata A12 818325/852422 A17
852423".
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
There are several similar errata on Cortex A12 and A17 that all have the same workaround: setting bit[12] of the Feature Register.
Technically the list of errata are:
- A12 818325: Execution of an UNPREDICTABLE STR or STM instruction
might deadlock. Fixed in r0p1.
- A12 852422: Execution of a sequence of instructions might lead to
either a data corruption or a CPU deadlock. Not fixed in any A12s
yet.
- A17 852423: Execution of a sequence of instructions might lead to
either a data corruption or a CPU deadlock. Not fixed in any A17s
yet.
Since A12 got renamed to A17 it seems likely that there won't be any
future Cortex-A12 cores, so we'll enable for all Cortex-A12.
For Cortex-A17 I believe that all known revisions are affected and that all knows revisions means <= r1p2. Presumably if a new A17 was
released it would have this problem fixed.
Note that in <https://patchwork.kernel.org/patch/4735341/> folks
previously expressed opposition to this change because:
A) It was thought to only apply to r0p0 and there were no known r0p0
boards supported in mainline.
B) It was argued that such a workaround beloned in firmware.
Now that this same fix solves other errata on real boards (like
rk3288) point A) is addressed.
Point B) is impossible to address on boards like rk3288. On rk3288
the firmware doesn't stay resident in RAM and isn't involved at all in
the suspend/resume process nor in the SMP bringup process. That means
that the most the firmware could do would be to set the bit on "core
0" and this bit would be lost at suspend/resume time. It is true that
we could write a "generic" solution that saved the boot-time "core 0"
value of this register and applied it at SMP bringup / resume time.
However, since this register (described as the "Feature Register" in
errata) appears to be undocumented (as far as I can tell) and is only
modified for these errata, that "generic" solution seems questionably
cleaner. The generic solution also won't fix existing users that
haven't happened to do a FW update.
Note that in ARM64 presumably PSCI will be universal and fixes like
this will end up in ATF. Hopefully we are nearing the end of this
style of errata workaround.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Huang Tao <huangtao@rock-chips.com>
Signed-off-by: Kever Yang <kever.yang@rock-chips.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Since commit 2b749cb3a515 ("ARM: realview: remove private barrier
implementation"), this config is not used by any platform.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
The binary GCD algorithm is based on the following facts:
1. If a and b are all evens, then gcd(a,b) = 2 * gcd(a/2, b/2)
2. If a is even and b is odd, then gcd(a,b) = gcd(a/2, b)
3. If a and b are all odds, then gcd(a,b) = gcd((a-b)/2, b) = gcd((a+b)/2, b)
Even on x86 machines with reasonable division hardware, the binary
algorithm runs about 25% faster (80% the execution time) than the
division-based Euclidian algorithm.
On platforms like Alpha and ARMv6 where division is a function call to
emulation code, it's even more significant.
There are two variants of the code here, depending on whether a fast
__ffs (find least significant set bit) instruction is available. This
allows the unpredictable branches in the bit-at-a-time shifting loop to
be eliminated.
If fast __ffs is not available, the "even/odd" GCD variant is used.
I use the following code to benchmark:
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <time.h>
#include <unistd.h>
#define swap(a, b) \
do { \
a ^= b; \
b ^= a; \
a ^= b; \
} while (0)
unsigned long gcd0(unsigned long a, unsigned long b)
{
unsigned long r;
if (a < b) {
swap(a, b);
}
if (b == 0)
return a;
while ((r = a % b) != 0) {
a = b;
b = r;
}
return b;
}
unsigned long gcd1(unsigned long a, unsigned long b)
{
unsigned long r = a | b;
if (!a || !b)
return r;
b >>= __builtin_ctzl(b);
for (;;) {
a >>= __builtin_ctzl(a);
if (a == b)
return a << __builtin_ctzl(r);
if (a < b)
swap(a, b);
a -= b;
}
}
unsigned long gcd2(unsigned long a, unsigned long b)
{
unsigned long r = a | b;
if (!a || !b)
return r;
r &= -r;
while (!(b & r))
b >>= 1;
for (;;) {
while (!(a & r))
a >>= 1;
if (a == b)
return a;
if (a < b)
swap(a, b);
a -= b;
a >>= 1;
if (a & r)
a += b;
a >>= 1;
}
}
unsigned long gcd3(unsigned long a, unsigned long b)
{
unsigned long r = a | b;
if (!a || !b)
return r;
b >>= __builtin_ctzl(b);
if (b == 1)
return r & -r;
for (;;) {
a >>= __builtin_ctzl(a);
if (a == 1)
return r & -r;
if (a == b)
return a << __builtin_ctzl(r);
if (a < b)
swap(a, b);
a -= b;
}
}
unsigned long gcd4(unsigned long a, unsigned long b)
{
unsigned long r = a | b;
if (!a || !b)
return r;
r &= -r;
while (!(b & r))
b >>= 1;
if (b == r)
return r;
for (;;) {
while (!(a & r))
a >>= 1;
if (a == r)
return r;
if (a == b)
return a;
if (a < b)
swap(a, b);
a -= b;
a >>= 1;
if (a & r)
a += b;
a >>= 1;
}
}
static unsigned long (*gcd_func[])(unsigned long a, unsigned long b) = {
gcd0, gcd1, gcd2, gcd3, gcd4,
};
#define TEST_ENTRIES (sizeof(gcd_func) / sizeof(gcd_func[0]))
#if defined(__x86_64__)
#define rdtscll(val) do { \
unsigned long __a,__d; \
__asm__ __volatile__("rdtsc" : "=a" (__a), "=d" (__d)); \
(val) = ((unsigned long long)__a) | (((unsigned long long)__d)<<32); \
} while(0)
static unsigned long long benchmark_gcd_func(unsigned long (*gcd)(unsigned long, unsigned long),
unsigned long a, unsigned long b, unsigned long *res)
{
unsigned long long start, end;
unsigned long long ret;
unsigned long gcd_res;
rdtscll(start);
gcd_res = gcd(a, b);
rdtscll(end);
if (end >= start)
ret = end - start;
else
ret = ~0ULL - start + 1 + end;
*res = gcd_res;
return ret;
}
#else
static inline struct timespec read_time(void)
{
struct timespec time;
clock_gettime(CLOCK_PROCESS_CPUTIME_ID, &time);
return time;
}
static inline unsigned long long diff_time(struct timespec start, struct timespec end)
{
struct timespec temp;
if ((end.tv_nsec - start.tv_nsec) < 0) {
temp.tv_sec = end.tv_sec - start.tv_sec - 1;
temp.tv_nsec = 1000000000ULL + end.tv_nsec - start.tv_nsec;
} else {
temp.tv_sec = end.tv_sec - start.tv_sec;
temp.tv_nsec = end.tv_nsec - start.tv_nsec;
}
return temp.tv_sec * 1000000000ULL + temp.tv_nsec;
}
static unsigned long long benchmark_gcd_func(unsigned long (*gcd)(unsigned long, unsigned long),
unsigned long a, unsigned long b, unsigned long *res)
{
struct timespec start, end;
unsigned long gcd_res;
start = read_time();
gcd_res = gcd(a, b);
end = read_time();
*res = gcd_res;
return diff_time(start, end);
}
#endif
static inline unsigned long get_rand()
{
if (sizeof(long) == 8)
return (unsigned long)rand() << 32 | rand();
else
return rand();
}
int main(int argc, char **argv)
{
unsigned int seed = time(0);
int loops = 100;
int repeats = 1000;
unsigned long (*res)[TEST_ENTRIES];
unsigned long long elapsed[TEST_ENTRIES];
int i, j, k;
for (;;) {
int opt = getopt(argc, argv, "n:r:s:");
/* End condition always first */
if (opt == -1)
break;
switch (opt) {
case 'n':
loops = atoi(optarg);
break;
case 'r':
repeats = atoi(optarg);
break;
case 's':
seed = strtoul(optarg, NULL, 10);
break;
default:
/* You won't actually get here. */
break;
}
}
res = malloc(sizeof(unsigned long) * TEST_ENTRIES * loops);
memset(elapsed, 0, sizeof(elapsed));
srand(seed);
for (j = 0; j < loops; j++) {
unsigned long a = get_rand();
/* Do we have args? */
unsigned long b = argc > optind ? strtoul(argv[optind], NULL, 10) : get_rand();
unsigned long long min_elapsed[TEST_ENTRIES];
for (k = 0; k < repeats; k++) {
for (i = 0; i < TEST_ENTRIES; i++) {
unsigned long long tmp = benchmark_gcd_func(gcd_func[i], a, b, &res[j][i]);
if (k == 0 || min_elapsed[i] > tmp)
min_elapsed[i] = tmp;
}
}
for (i = 0; i < TEST_ENTRIES; i++)
elapsed[i] += min_elapsed[i];
}
for (i = 0; i < TEST_ENTRIES; i++)
printf("gcd%d: elapsed %llu\n", i, elapsed[i]);
k = 0;
srand(seed);
for (j = 0; j < loops; j++) {
unsigned long a = get_rand();
unsigned long b = argc > optind ? strtoul(argv[optind], NULL, 10) : get_rand();
for (i = 1; i < TEST_ENTRIES; i++) {
if (res[j][i] != res[j][0])
break;
}
if (i < TEST_ENTRIES) {
if (k == 0) {
k = 1;
fprintf(stderr, "Error:\n");
}
fprintf(stderr, "gcd(%lu, %lu): ", a, b);
for (i = 0; i < TEST_ENTRIES; i++)
fprintf(stderr, "%ld%s", res[j][i], i < TEST_ENTRIES - 1 ? ", " : "\n");
}
}
if (k == 0)
fprintf(stderr, "PASS\n");
free(res);
return 0;
}
Compiled with "-O2", on "VirtualBox 4.4.0-22-generic #38-Ubuntu x86_64" got:
zhaoxiuzeng@zhaoxiuzeng-VirtualBox:~/develop$ ./gcd -r 500000 -n 10
gcd0: elapsed 10174
gcd1: elapsed 2120
gcd2: elapsed 2902
gcd3: elapsed 2039
gcd4: elapsed 2812
PASS
zhaoxiuzeng@zhaoxiuzeng-VirtualBox:~/develop$ ./gcd -r 500000 -n 10
gcd0: elapsed 9309
gcd1: elapsed 2280
gcd2: elapsed 2822
gcd3: elapsed 2217
gcd4: elapsed 2710
PASS
zhaoxiuzeng@zhaoxiuzeng-VirtualBox:~/develop$ ./gcd -r 500000 -n 10
gcd0: elapsed 9589
gcd1: elapsed 2098
gcd2: elapsed 2815
gcd3: elapsed 2030
gcd4: elapsed 2718
PASS
zhaoxiuzeng@zhaoxiuzeng-VirtualBox:~/develop$ ./gcd -r 500000 -n 10
gcd0: elapsed 9914
gcd1: elapsed 2309
gcd2: elapsed 2779
gcd3: elapsed 2228
gcd4: elapsed 2709
PASS
[akpm@linux-foundation.org: avoid #defining a CONFIG_ variable]
Signed-off-by: Zhaoxiu Zeng <zhaoxiu.zeng@gmail.com>
Signed-off-by: George Spelvin <linux@horizon.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Pull ARM updates from Russell King:
"Changes included in this pull request:
- revert pxa2xx-flash back to using ioremap_cached() and switch
memremap() to use arch_memremap_wb()
- remove pci=firmware command line argument handling
- remove unnecessary arm_dma_set_mask() implementation, the generic
implementation will do for ARM
- removal of the ARM kallsyms "hack" to work around mode switching
veneers and vectors located below PAGE_OFFSET
- tidy up build system output a little
- add L2 cache power management DT bindings
- remove duplicated local_irq_disable() in reboot paths
- handle AMBA primecell devices better at registration time with PM
domains (needed for Samsung SoCs)
- ARM specific preparation to support Keystone II kexec"
* 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
ARM: 8567/1: cache-uniphier: activate ways for secondary CPUs
ARM: 8570/2: Documentation: devicetree: Add PL310 PM bindings
ARM: 8569/1: pl2x0: Add OF control of cache power management
ARM: 8568/1: reboot: remove duplicated local_irq_disable()
ARM: 8566/1: drivers: amba: properly handle devices with power domains
ARM: provide arm_has_idmap_alias() helper
ARM: kexec: remove 512MB restriction on kexec crashdump
ARM: provide improved virt_to_idmap() functionality
ARM: kexec: fix crashkernel= handling
ARM: 8557/1: specify install, zinstall, and uinstall as PHONY targets
ARM: 8562/1: suppress "include/generated/mach-types.h is up to date."
ARM: 8553/1: kallsyms: remove --page-offset command line option
ARM: 8552/1: kallsyms: remove special lower address limit for CONFIG_ARM
ARM: 8555/1: kallsyms: ignore ARM mode switching veneers
ARM: 8548/1: dma-mapping: remove arm_dma_set_mask()
ARM: 8554/1: kernel: pci: remove pci=firmware command line parameter handling
ARM: memremap: implement arch_memremap_wb()
memremap: add arch specific hook for MEMREMAP_WB mappings
mtd: pxa2xx-flash: switch back from memremap to ioremap_cached
ARM: reintroduce ioremap_cached() for creating cached I/O mappings
|
|
for-linus
|
|
As a set of driver-provided callbacks and static data, there is no
compelling reason for struct iommu_ops to be mutable in core code, so
enforce const-ness throughout.
Acked-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
|
|
Pull ARM fixes from Russell King:
"These are a number of updates to fix a few problems found in the ARM
nommu code over the last couple of years, caused mostly by changes on
the mmu side"
* 'fixes' of git://git.armlinux.org.uk/~rmk/linux-arm:
ARM: 8573/1: domain: move {set,get}_domain under config guard
ARM: 8572/1: nommu: change memory reserve for the vectors
ARM: 8571/1: nommu: fix PMSAv7 setup
|
|
This outer cache allows to control active ways independently for
each CPU, but currently nothing is done for secondary CPUs. In
other words, all the ways are locked for secondary CPUs by default.
This commit fixes it to fully bring out the performance of this
outer cache.
There would be two possible ways to achieve this:
[1] Each CPU initializes active ways for itself. This can be done
via the SSCLPDAWCR register. This is a banked register, so each
CPU sees a different instance of the register for its own.
[2] The master CPU initializes active ways for all the CPUs. This
is available via SSCDAWCARMR(N) registers, where all instances
of SSCLPDAWCR are mirrored. They are mapped at the address
SSCDAWCARMR + 4 * N, where N is the CPU number.
The outer cache frame work does not support a per-CPU init callback.
So this commit adopts [2]; the master CPU iterates over possible CPUs
setting up SSCDAWCARMR(N) registers.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Commit 19accfd3 (ARM: move vector stubs) moved the vector stubs in an
additional page above the base vector one. This change wasn't taken into
account by the nommu memreserve.
This patch ensures that the kernel won't overwrite any vector stub on
nommu.
[changed the MPU side too]
Signed-off-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Commit 1c2f87c (ARM: 8025/1: Get rid of meminfo) broke the support for
MPU on ARMv7-R. This patch adapts the code inside CONFIG_ARM_MPU to use
memblocks appropriately.
MPU initialisation only uses the first memory region, and removes all
subsequent ones. Because looping over all regions that need removal is
inefficient, and memblock_remove already handles memory ranges, we can
flatten the 'for_each_memblock' part.
Signed-off-by: Jean-Philippe Brucker <jean-philippe.brucker@arm.com>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Add ability to override power management bits of 310 controllers
(dynamic clock gating and standby mode) through OF entries. As the
saved register is only applied when working on a supported controller,
it is safe to save the settings.
In order to maintain existing behavior, if the settings are not found
in the DT, the corresponding feature will be enabled.
Signed-off-by: Brad Mouring <brad.mouring@ni.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
For kexec, we need more functionality from the IDMAP system. We need to
be able to convert physical addresses to their identity mappped versions
as well as virtual addresses.
Convert the existing arch_virt_to_idmap() to deal with physical
addresses instead.
Acked-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Pull ARM fixes from Russell King:
"Three further fixes for ARM.
Alexandre Courbot was having problems with DMA allocations with the
GFP flags affecting where the tracking data was being allocated from.
Vladimir Murzin noticed that the CPU feature code was not entirely
correct, which can cause some features to be misreported"
* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
ARM: 8564/1: fix cpu feature extracting helper
ARM: 8563/1: fix demoting HWCAP_SWP
ARM: 8551/2: DMA: Fix kzalloc flags in __dma_alloc
|
|
git://git.linaro.org/people/ard.biesheuvel/linux-arm into devel-stable
This series wires up the generic memremap() function for ARM in a way
that allows it to be used as intended, i.e., without regard for whether
the region being mapped is covered by a struct page and/or the linear
mapping (lowmem)
|
|
Commit 19e6e5e5392b ("ARM: 8547/1: dma-mapping: store buffer
information") allocates a structure meant for internal buffer management
with the GFP flags of the buffer itself. This can trigger the following
safeguard in the slab/slub allocator:
if (unlikely(flags & GFP_SLAB_BUG_MASK)) {
pr_emerg("gfp: %un", flags & GFP_SLAB_BUG_MASK);
BUG();
}
Fix this by filtering the flags that make the slab allocator unhappy.
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Acked-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Pull ARM fixes from Russell King:
"A couple of small fixes, and wiring up the new syscalls which appeared
during the merge window"
* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
ARM: 8550/1: protect idiv patching against undefined gcc behavior
ARM: wire up preadv2 and pwritev2 syscalls
ARM: SMP enable of cache maintanence broadcast
|
|
arm_dma_set_mask() implements exactly the same behavior as the fallback
that dma_set_mask() takes if the set_dma_mask op is not set. Remove it
and use that fallback instead like what is already done for
dma_get_mask().
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time
ago with promise that one day it will be possible to implement page
cache with bigger chunks than PAGE_SIZE.
This promise never materialized. And unlikely will.
We have many places where PAGE_CACHE_SIZE assumed to be equal to
PAGE_SIZE. And it's constant source of confusion on whether
PAGE_CACHE_* or PAGE_* constant should be used in a particular case,
especially on the border between fs and mm.
Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much
breakage to be doable.
Let's stop pretending that pages in page cache are special. They are
not.
The changes are pretty straight-forward:
- <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>;
- PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN};
- page_cache_get() -> get_page();
- page_cache_release() -> put_page();
This patch contains automated changes generated with coccinelle using
script below. For some reason, coccinelle doesn't patch header files.
I've called spatch for them manually.
The only adjustment after coccinelle is revert of changes to
PAGE_CAHCE_ALIGN definition: we are going to drop it later.
There are few places in the code where coccinelle didn't reach. I'll
fix them manually in a separate patch. Comments and documentation also
will be addressed with the separate patch.
virtual patch
@@
expression E;
@@
- E << (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
expression E;
@@
- E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT)
+ E
@@
@@
- PAGE_CACHE_SHIFT
+ PAGE_SHIFT
@@
@@
- PAGE_CACHE_SIZE
+ PAGE_SIZE
@@
@@
- PAGE_CACHE_MASK
+ PAGE_MASK
@@
expression E;
@@
- PAGE_CACHE_ALIGN(E)
+ PAGE_ALIGN(E)
@@
expression E;
@@
- page_cache_get(E)
+ get_page(E)
@@
expression E;
@@
- page_cache_release(E)
+ put_page(E)
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
The generic memremap() falls back to using ioremap_cache() to create
MEMREMAP_WB mappings if the requested region is not already covered
by the linear mapping, unless the architecture provides an implementation
of arch_memremap_wb().
Since ioremap_cache() is not appropriate on ARM to map memory with the
same attributes used for the linear mapping, implement arch_memremap_wb()
which does exactly that. Also, relax the WARN() check to allow MT_MEMORY_RW
mappings of pfn_valid() pages.
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
|
|
The original ARM-only ioremap flavor 'ioremap_cached' has been renamed
to 'ioremap_cache' to align with other architectures, and subsequently
abused in generic code to map things like firmware tables in memory.
For that reason, there is currently an effort underway to deprecate
ioremap_cache, whose semantics are poorly defined, and which is typed
with an __iomem annotation that is inappropriate for mappings of ordinary
memory.
However, original users of ioremap_cached() used it in a context where
the I/O connotation is appropriate, and replacing those instances with
memremap() does not make sense. So let's revive ioremap_cached(), so
that we can change back those original users before we drop ioremap_cache
entirely in favor of memremap.
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
|
|
Masahiro Yamada reports that we can fail to set the FW bit in the
auxiliary control register, which enables broadcasting the cache
maintanence operations. This occurs because we only check that the
SMP/nAMP bit is set, rather than checking whether all the bits we
want to be set are set.
Rearrange the code to ensure that all desired bits are set, and only
update the register if we discover some required bits are not set.
Tested-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Pull ARM updates from Russell King:
"Another mixture of changes this time around:
- Split XIP linker file from main linker file to make it more
maintainable, and various XIP fixes, and clean up a resulting
macro.
- Decompressor cleanups from Masahiro Yamada
- Avoid printing an error for a missing L2 cache
- Remove some duplicated symbols in System.map, and move
vectors/stubs back into kernel VMA
- Various low priority fixes from Arnd
- Updates to allow bus match functions to return negative errno
values, touching some drivers and the driver core. Greg has acked
these changes.
- Virtualisation platform udpates form Jean-Philippe Brucker.
- Security enhancements from Kees Cook
- Rework some Kconfig dependencies and move PSCI idle management code
out of arch/arm into drivers/firmware/psci.c
- ARM DMA mapping updates, touching media, acked by Mauro.
- Fix places in ARM code which should be using virt_to_idmap() so
that Keystone2 can work.
- Fix Marvell Tauros2 to work again with non-DT boots.
- Provide a delay timer for ARM Orion platforms"
* 'for-linus' of git://ftp.arm.linux.org.uk/~rmk/linux-arm: (45 commits)
ARM: 8546/1: dma-mapping: refactor to fix coherent+cma+gfp=0
ARM: 8547/1: dma-mapping: store buffer information
ARM: 8543/1: decompressor: rename suffix_y to compress-y
ARM: 8542/1: decompressor: merge piggy.*.S and simplify Makefile
ARM: 8541/1: decompressor: drop redundant FORCE in Makefile
ARM: 8540/1: decompressor: use clean-files instead of extra-y to clean files
ARM: 8539/1: decompressor: drop more unneeded assignments to "targets"
ARM: 8538/1: decompressor: drop unneeded assignments to "targets"
ARM: 8532/1: uncompress: mark putc as inline
ARM: 8531/1: turn init_new_context into an inline function
ARM: 8530/1: remove VIRT_TO_BUS
ARM: 8537/1: drop unused DEBUG_RODATA from XIP_KERNEL
ARM: 8536/1: mm: hide __start_rodata_section_aligned for non-debug builds
ARM: 8535/1: mm: DEBUG_RODATA makes no sense with XIP_KERNEL
ARM: 8534/1: virt: fix hyp-stub build for pre-ARMv7 CPUs
ARM: make the physical-relative calculation more obvious
ARM: 8512/1: proc-v7.S: Adjust stack address when XIP_KERNEL
ARM: 8411/1: Add default SPARSEMEM settings
ARM: 8503/1: clk_register_clkdev: remove format string interface
ARM: 8529/1: remove 'i' and 'zi' targets
...
|
|
The define has a comment from Nick Piggin from 2007:
/* For backwards compat. Remove me quickly. */
I guess 9 years should not be too hurried sense of 'quickly' even for
kernel measures.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
There are few things about *pte_alloc*() helpers worth cleaning up:
- 'vma' argument is unused, let's drop it;
- most __pte_alloc() callers do speculative check for pmd_none(),
before taking ptl: let's introduce pte_alloc() macro which does
the check.
The only direct user of __pte_alloc left is userfaultfd, which has
different expectation about atomicity wrt pmd.
- pte_alloc_map() and pte_alloc_map_lock() are redefined using
pte_alloc().
[sudeep.holla@arm.com: fix build for arm64 hugetlbpage]
[sfr@canb.auug.org.au: fix arch/arm/mm/mmu.c some more]
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Pull ARM fixes from Russell King:
"Just two ARM fixes this time: one to fix the hyp-stub for older ARM
CPUs, and another to fix the set_memory_xx() permission functions to
deal with zero sizes correctly"
* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
ARM: 8544/1: set_memory_xx fixes
ARM: 8534/1: virt: fix hyp-stub build for pre-ARMv7 CPUs
|
|
|
|
Given a device which uses arm_coherent_dma_ops and on which
dev_get_cma_area(dev) returns non-NULL, the following usage of the DMA
API with gfp=0 results in memory corruption and a memory leak.
p = dma_alloc_coherent(dev, sz, &dma, 0);
if (p)
dma_free_coherent(dev, sz, p, dma);
The memory leak is because the alloc allocates using
__alloc_simple_buffer() but the free attempts
dma_release_from_contiguous() which does not do free anything since the
page is not in the CMA area.
The memory corruption is because the free calls __dma_remap() on a page
which is backed by only first level page tables. The
apply_to_page_range() + __dma_update_pte() loop ends up interpreting the
section mapping as an addresses to a second level page table and writing
the new PTE to memory which is not used by page tables.
We don't have access to the GFP flags used for allocation in the free
function. Fix this by adding allocator backends and using this
information in the free function so that we always use the correct
release routine.
Fixes: 21caf3a7 ("ARM: 8398/1: arm DMA: Fix allocation from CMA for coherent DMA")
Signed-off-by: Rabin Vincent <rabin.vincent@axis.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Keep a list of allocated DMA buffers so that we can store metadata in
alloc() which we later need in free().
Signed-off-by: Rabin Vincent <rabin.vincent@axis.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Allow zero size updates. This makes set_memory_xx() consistent with x86, s390 and arm64 and makes apply_to_page_range() not to BUG() when loading modules.
Signed-off-by: Mika Penttilä mika.penttila@nextfour.com
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
Replace calls to get_random_int() followed by a cast to (unsigned long)
with calls to get_random_long(). Also address shifting bug which, in
case of x86 removed entropy mask for mmap_rnd_bits values > 31 bits.
Signed-off-by: Daniel Cashman <dcashman@android.com>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: David S. Miller <davem@davemloft.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Nick Kralevich <nnk@google.com>
Cc: Jeff Vander Stoep <jeffv@google.com>
Cc: Mark Salyzyn <salyzyn@android.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
When CONFIG_DEBUG_ALIGN_RODATA is set, we get a link error:
arch/arm/mm/built-in.o:(.data+0x4bc): undefined reference to `__start_rodata_section_aligned'
However, this combination is useless, as XIP_KERNEL implies that all the
RODATA is already marked readonly, so both CONFIG_DEBUG_RODATA and
CONFIG_DEBUG_ALIGN_RODATA (which depends on the other) are not
needed with XIP_KERNEL, and this patches enforces that using a Kconfig
dependency.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 25362dc496ed ("ARM: 8501/1: mm: flip priority of CONFIG_DEBUG_RODATA")
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
The physical-relative calculation between the XIP text and data sections
introduced by the previous patch was far from obvious. Let's simplify it
by turning it into a macro which takes the two (virtual) addresses.
This allows us to arrange the calculation in a more obvious manner - we
can make it two sub-expressions which calculate the physical address for
each symbol, and then takes the difference of those physical addresses.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
When XIP_KERNEL is enabled, the virt to phys address translation for RAM
is not the same as the virt to phys address translation for .text.
The only way to know where physical RAM is located is to use
PLAT_PHYS_OFFSET.
The MACRO will be useful for other places where there is a similar problem.
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Chris Brandt <chris.brandt@renesas.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
When rodata is large enough that it crosses a section boundary after the
kernel text, mark the rest NX. This is as close to full NX of rodata as
we can get without splitting page tables or doing section alignment via
CONFIG_DEBUG_ALIGN_RODATA.
When the config is:
CONFIG_DEBUG_RODATA=y
# CONFIG_DEBUG_ALIGN_RODATA is not set
Before:
---[ Kernel Mapping ]---
0x80000000-0x80100000 1M RW NX SHD
0x80100000-0x80a00000 9M ro x SHD
0x80a00000-0xa0000000 502M RW NX SHD
After:
---[ Kernel Mapping ]---
0x80000000-0x80100000 1M RW NX SHD
0x80100000-0x80700000 6M ro x SHD
0x80700000-0x80a00000 3M ro NX SHD
0x80a00000-0xa0000000 502M RW NX SHD
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
For an XIP build, _etext does not represent the end of the
binary image that needs to stay mapped into the MODULES_VADDR area.
Years ago, data came before text in the memory map. However,
now that the order is text/init/data, an XIP_KERNEL needs to map
up to the data location in order to keep from cutting off
parts of the kernel that are needed.
We only map up to the beginning of data because data has already been
copied, so there's no reason to keep it around anymore.
A new symbol is created to make it clear what it is we are referring
to.
This fixes the bug where you might lose the end of your kernel area
after page table setup is complete.
Signed-off-by: Chris Brandt <chris.brandt@renesas.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
If we know that TLB efficiency will not be an issue when memory is
accessed then it's not terribly important to allocate big chunks of
memory. The whole point of allocating the big chunks was that it would
make TLB usage efficient.
As Marek Szyprowski indicated:
Please note that mapping memory with larger pages significantly
improves performance, especially when IOMMU has a little TLB
cache. This can be easily observed when multimedia devices do
processing of RGB data with 90/270 degree rotation
Image rotation is distinctly an operation that needs to bounce around
through memory, so it makes sense that TLB efficiency is important
there.
Video decoding, on the other hand, is a fairly sequential operation.
During video decoding it's not expected that we'll be jumping all over
memory. Decoding video is also pretty heavy and the TLB misses aren't a
huge deal. Presumably most HW video acceleration users of dma-mapping
will not care about huge pages and will set DMA_ATTR_ALLOC_SINGLE_PAGES.
Allocating big chunks of memory is quite expensive, especially if we're
doing it repeadly and memory is full. In one (out of tree) usage model
it is common that arm_iommu_alloc_attrs() is called 16 times in a row,
each one trying to allocate 4 MB of memory. This is called whenever the
system encounters a new video, which could easily happen while the
memory system is stressed out. In fact, on certain social media
websites that auto-play video and have infinite scrolling, it's quite
common to see not just one of these 16x4MB allocations but 2 or 3 right
after another. Asking the system even to do a small amount of extra
work to give us big chunks in this case is just not a good use of time.
Allocating big chunks of memory is also expensive indirectly. Even if
we ask the system not to do ANY extra work to allocate _our_ memory,
we're still potentially eating up all big chunks in the system.
Presumably there are other users in the system that aren't quite as
flexible and that actually need these big chunks. By eating all the big
chunks we're causing extra work for the rest of the system. We also may
start making other memory allocations fail. While the system may be
robust to such failures (as is the case with dwc2 USB trying to allocate
buffers for Ethernet data and with WiFi trying to allocate buffers for
WiFi data), it is yet another big performance hit.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Tested-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
The __iommu_alloc_buffer() is expected to be called to allocate pretty
sizeable buffers. Upon simple tests of video I saw it trying to
allocate 4,194,304 bytes. The function tries to allocate large chunks
in order to optimize IOMMU TLB usage.
The current function is very, very slow.
One problem is the way it keeps trying and trying to allocate big
chunks. Imagine a very fragmented memory that has 4M free but no
contiguous pages at all. Further imagine allocating 4M (1024 pages).
We'll do the following memory allocations:
- For page 1:
- Try to allocate order 10 (no retry)
- Try to allocate order 9 (no retry)
- ...
- Try to allocate order 0 (with retry, but not needed)
- For page 2:
- Try to allocate order 9 (no retry)
- Try to allocate order 8 (no retry)
- ...
- Try to allocate order 0 (with retry, but not needed)
- ...
- ...
Total number of calls to alloc() calls for this case is:
sum(int(math.log(i, 2)) + 1 for i in range(1, 1025))
=> 9228
The above is obviously worse case, but given how slow alloc can be we
really want to try to avoid even somewhat bad cases. I timed the old
code with a device under memory pressure and it wasn't hard to see it
take more than 120 seconds to allocate 4 megs of memory! (NOTE: testing
was done on kernel 3.14, so possibly mainline would behave
differently).
A second problem is that allocating big chunks under memory pressure
when we don't need them is just not a great idea anyway unless we really
need them. We can make due pretty well with smaller chunks so it's
probably wise to leave bigger chunks for other users once memory
pressure is on.
Let's adjust the allocation like this:
1. If a big chunk fails, stop trying to hard and bump down to lower
order allocations.
2. Don't try useless orders. The whole point of big chunks is to
optimize the TLB and it can really only make use of 2M, 1M, 64K and
4K sizes.
We'll still tend to eat up a bunch of big chunks, but that might be the
right answer for some users. A future patch could possibly add a new
DMA_ATTR that would let the caller decide that TLB optimization isn't
important and that we should use smaller chunks. Presumably this would
be a sane strategy for some callers.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Reviewed-by: Tomasz Figa <tfiga@chromium.org>
Tested-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
|
The use of CONFIG_DEBUG_RODATA is generally seen as an essential part of
kernel self-protection:
http://www.openwall.com/lists/kernel-hardening/2015/11/30/13
Additionally, its name has grown to mean things beyond just rodata. To
get ARM closer to this, we ought to rearrange the names of the configs
that control how the kernel protects its memory. What was called
CONFIG_ARM_KERNMEM_PERMS is realy doing the work that other architectures
call CONFIG_DEBUG_RODATA.
This redefines CONFIG_DEBUG_RODATA to actually do the bulk of the
ROing (and NXing). In the place of the old CONFIG_DEBUG_RODATA, use
CONFIG_DEBUG_ALIGN_RODATA, since that's what the option does: adds
section alignment for making rodata explicitly NX, as arm does not split
the page tables like arm64 does without _ALIGN_RODATA.
Also adds human readable names to the sections so I could more easily
debug my typos, and makes CONFIG_DEBUG_RODATA default "y" for CPU_V7.
Results in /sys/kernel/debug/kernel_page_tables for each config state:
# CONFIG_DEBUG_RODATA is not set
# CONFIG_DEBUG_ALIGN_RODATA is not set
---[ Kernel Mapping ]---
0x80000000-0x80900000 9M RW x SHD
0x80900000-0xa0000000 503M RW NX SHD
CONFIG_DEBUG_RODATA=y
CONFIG_DEBUG_ALIGN_RODATA=y
---[ Kernel Mapping ]---
0x80000000-0x80100000 1M RW NX SHD
0x80100000-0x80700000 6M ro x SHD
0x80700000-0x80a00000 3M ro NX SHD
0x80a00000-0xa0000000 502M RW NX SHD
CONFIG_DEBUG_RODATA=y
# CONFIG_DEBUG_ALIGN_RODATA is not set
---[ Kernel Mapping ]---
0x80000000-0x80100000 1M RW NX SHD
0x80100000-0x80a00000 9M ro x SHD
0x80a00000-0xa0000000 502M RW NX SHD
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Laura Abbott <labbott@fedoraproject.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
|
