Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'powerpc-7.0-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux
Pull powerpc fixes from Madhavan Srinivasan:
- Fix KUAP warning in VMX usercopy path
- Fix lockdep warning during PCI enumeration
- Fix to move CMA reservations to arch_mm_preinit
- Fix to check current->mm is alive before getting user callchain
Thanks to Aboorva Devarajan, Christophe Leroy (CS GROUP), Dan Horák,
Nicolin Chen, Nilay Shroff, Qiao Zhao, Ritesh Harjani (IBM), Saket Kumar
Bhaskar, Sayali Patil, Shrikanth Hegde, Venkat Rao Bagalkote, and Viktor
Malik.
* tag 'powerpc-7.0-3' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
powerpc/iommu: fix lockdep warning during PCI enumeration
powerpc/selftests/copyloops: extend selftest to exercise __copy_tofrom_user_power7_vmx
powerpc: fix KUAP warning in VMX usercopy path
powerpc, perf: Check that current->mm is alive before getting user callchain
powerpc/mem: Move CMA reservations to arch_mm_preinit
+109
-79
+47
-22
arch/powerpc/include/asm/uaccess.h
+47
-22
arch/powerpc/include/asm/uaccess.h
···
15
15
#define TASK_SIZE_MAX TASK_SIZE_USER64
16
16
#endif
17
17
18
+
/* Threshold above which VMX copy path is used */
19
+
#define VMX_COPY_THRESHOLD 3328
20
+
18
21
#include <asm-generic/access_ok.h>
19
22
20
23
/*
···
329
326
extern unsigned long __copy_tofrom_user(void __user *to,
330
327
const void __user *from, unsigned long size);
331
328
332
-
#ifdef __powerpc64__
329
+
unsigned long __copy_tofrom_user_base(void __user *to,
330
+
const void __user *from, unsigned long size);
331
+
332
+
unsigned long __copy_tofrom_user_power7_vmx(void __user *to,
333
+
const void __user *from, unsigned long size);
334
+
335
+
static __always_inline bool will_use_vmx(unsigned long n)
336
+
{
337
+
return IS_ENABLED(CONFIG_ALTIVEC) && cpu_has_feature(CPU_FTR_VMX_COPY) &&
338
+
n > VMX_COPY_THRESHOLD;
339
+
}
340
+
341
+
static __always_inline unsigned long
342
+
raw_copy_tofrom_user(void __user *to, const void __user *from,
343
+
unsigned long n, unsigned long dir)
344
+
{
345
+
unsigned long ret;
346
+
347
+
if (will_use_vmx(n) && enter_vmx_usercopy()) {
348
+
allow_user_access(to, dir);
349
+
ret = __copy_tofrom_user_power7_vmx(to, from, n);
350
+
prevent_user_access(dir);
351
+
exit_vmx_usercopy();
352
+
353
+
if (unlikely(ret)) {
354
+
allow_user_access(to, dir);
355
+
ret = __copy_tofrom_user_base(to, from, n);
356
+
prevent_user_access(dir);
357
+
}
358
+
return ret;
359
+
}
360
+
361
+
allow_user_access(to, dir);
362
+
ret = __copy_tofrom_user(to, from, n);
363
+
prevent_user_access(dir);
364
+
return ret;
365
+
}
366
+
367
+
#ifdef CONFIG_PPC64
333
368
static inline unsigned long
334
369
raw_copy_in_user(void __user *to, const void __user *from, unsigned long n)
335
370
{
336
-
unsigned long ret;
337
-
338
371
barrier_nospec();
339
-
allow_user_access(to, KUAP_READ_WRITE);
340
-
ret = __copy_tofrom_user(to, from, n);
341
-
prevent_user_access(KUAP_READ_WRITE);
342
-
return ret;
372
+
return raw_copy_tofrom_user(to, from, n, KUAP_READ_WRITE);
343
373
}
344
-
#endif /* __powerpc64__ */
374
+
#endif /* CONFIG_PPC64 */
345
375
346
-
static inline unsigned long raw_copy_from_user(void *to,
347
-
const void __user *from, unsigned long n)
376
+
static inline unsigned long raw_copy_from_user(void *to, const void __user *from, unsigned long n)
348
377
{
349
-
unsigned long ret;
350
-
351
-
allow_user_access(NULL, KUAP_READ);
352
-
ret = __copy_tofrom_user((__force void __user *)to, from, n);
353
-
prevent_user_access(KUAP_READ);
354
-
return ret;
378
+
return raw_copy_tofrom_user((__force void __user *)to, from, n, KUAP_READ);
355
379
}
356
380
357
381
static inline unsigned long
358
382
raw_copy_to_user(void __user *to, const void *from, unsigned long n)
359
383
{
360
-
unsigned long ret;
361
-
362
-
allow_user_access(to, KUAP_WRITE);
363
-
ret = __copy_tofrom_user(to, (__force const void __user *)from, n);
364
-
prevent_user_access(KUAP_WRITE);
365
-
return ret;
384
+
return raw_copy_tofrom_user(to, (__force const void __user *)from, n, KUAP_WRITE);
366
385
}
367
386
368
387
unsigned long __arch_clear_user(void __user *addr, unsigned long size);
+1
-1
arch/powerpc/kernel/iommu.c
+1
-1
arch/powerpc/kernel/iommu.c
···
1159
1159
struct device *dev,
1160
1160
struct iommu_domain *old)
1161
1161
{
1162
-
struct iommu_domain *domain = iommu_get_domain_for_dev(dev);
1162
+
struct iommu_domain *domain = iommu_driver_get_domain_for_dev(dev);
1163
1163
struct iommu_table_group *table_group;
1164
1164
struct iommu_group *grp;
1165
1165
-10
arch/powerpc/kernel/setup-common.c
-10
arch/powerpc/kernel/setup-common.c
···
35
35
#include <linux/of_irq.h>
36
36
#include <linux/hugetlb.h>
37
37
#include <linux/pgtable.h>
38
-
#include <asm/kexec.h>
39
38
#include <asm/io.h>
40
39
#include <asm/paca.h>
41
40
#include <asm/processor.h>
···
993
994
smp_release_cpus();
994
995
995
996
initmem_init();
996
-
997
-
/*
998
-
* Reserve large chunks of memory for use by CMA for kdump, fadump, KVM and
999
-
* hugetlb. These must be called after initmem_init(), so that
1000
-
* pageblock_order is initialised.
1001
-
*/
1002
-
fadump_cma_init();
1003
-
kdump_cma_reserve();
1004
-
kvm_cma_reserve();
1005
997
1006
998
early_memtest(min_low_pfn << PAGE_SHIFT, max_low_pfn << PAGE_SHIFT);
1007
999
+1
arch/powerpc/lib/copyuser_64.S
+1
arch/powerpc/lib/copyuser_64.S
+15
-30
arch/powerpc/lib/copyuser_power7.S
+15
-30
arch/powerpc/lib/copyuser_power7.S
···
5
5
*
6
6
* Author: Anton Blanchard <anton@au.ibm.com>
7
7
*/
8
+
#include <linux/export.h>
8
9
#include <asm/ppc_asm.h>
9
-
10
-
#ifndef SELFTEST_CASE
11
-
/* 0 == don't use VMX, 1 == use VMX */
12
-
#define SELFTEST_CASE 0
13
-
#endif
14
10
15
11
#ifdef __BIG_ENDIAN__
16
12
#define LVS(VRT,RA,RB) lvsl VRT,RA,RB
···
43
47
ld r15,STK_REG(R15)(r1)
44
48
ld r14,STK_REG(R14)(r1)
45
49
.Ldo_err3:
46
-
bl CFUNC(exit_vmx_usercopy)
50
+
ld r6,STK_REG(R31)(r1) /* original destination pointer */
51
+
ld r5,STK_REG(R29)(r1) /* original number of bytes */
52
+
subf r7,r6,r3 /* #bytes copied */
53
+
subf r3,r7,r5 /* #bytes not copied in r3 */
47
54
ld r0,STACKFRAMESIZE+16(r1)
48
55
mtlr r0
49
-
b .Lexit
56
+
addi r1,r1,STACKFRAMESIZE
57
+
blr
50
58
#endif /* CONFIG_ALTIVEC */
51
59
52
60
.Ldo_err2:
···
74
74
75
75
_GLOBAL(__copy_tofrom_user_power7)
76
76
cmpldi r5,16
77
-
cmpldi cr1,r5,3328
78
77
79
78
std r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
80
79
std r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
···
81
82
82
83
blt .Lshort_copy
83
84
84
-
#ifdef CONFIG_ALTIVEC
85
-
test_feature = SELFTEST_CASE
86
-
BEGIN_FTR_SECTION
87
-
bgt cr1,.Lvmx_copy
88
-
END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
89
-
#endif
90
85
91
86
.Lnonvmx_copy:
92
87
/* Get the source 8B aligned */
···
256
263
15: li r3,0
257
264
blr
258
265
259
-
.Lunwind_stack_nonvmx_copy:
260
-
addi r1,r1,STACKFRAMESIZE
261
-
b .Lnonvmx_copy
262
-
263
-
.Lvmx_copy:
264
266
#ifdef CONFIG_ALTIVEC
267
+
_GLOBAL(__copy_tofrom_user_power7_vmx)
265
268
mflr r0
266
269
std r0,16(r1)
267
270
stdu r1,-STACKFRAMESIZE(r1)
268
-
bl CFUNC(enter_vmx_usercopy)
269
-
cmpwi cr1,r3,0
270
-
ld r0,STACKFRAMESIZE+16(r1)
271
-
ld r3,STK_REG(R31)(r1)
272
-
ld r4,STK_REG(R30)(r1)
273
-
ld r5,STK_REG(R29)(r1)
274
-
mtlr r0
275
271
272
+
std r3,STK_REG(R31)(r1)
273
+
std r5,STK_REG(R29)(r1)
276
274
/*
277
275
* We prefetch both the source and destination using enhanced touch
278
276
* instructions. We use a stream ID of 0 for the load side and
···
283
299
ori r10,r7,1 /* stream=1 */
284
300
285
301
DCBT_SETUP_STREAMS(r6, r7, r9, r10, r8)
286
-
287
-
beq cr1,.Lunwind_stack_nonvmx_copy
288
302
289
303
/*
290
304
* If source and destination are not relatively aligned we use a
···
460
478
err3; stb r0,0(r3)
461
479
462
480
15: addi r1,r1,STACKFRAMESIZE
463
-
b CFUNC(exit_vmx_usercopy) /* tail call optimise */
481
+
li r3,0
482
+
blr
464
483
465
484
.Lvmx_unaligned_copy:
466
485
/* Get the destination 16B aligned */
···
664
681
err3; stb r0,0(r3)
665
682
666
683
15: addi r1,r1,STACKFRAMESIZE
667
-
b CFUNC(exit_vmx_usercopy) /* tail call optimise */
684
+
li r3,0
685
+
blr
686
+
EXPORT_SYMBOL(__copy_tofrom_user_power7_vmx)
668
687
#endif /* CONFIG_ALTIVEC */
+2
arch/powerpc/lib/vmx-helper.c
+2
arch/powerpc/lib/vmx-helper.c
+14
arch/powerpc/mm/mem.c
+14
arch/powerpc/mm/mem.c
···
30
30
#include <asm/setup.h>
31
31
#include <asm/fixmap.h>
32
32
33
+
#include <asm/fadump.h>
34
+
#include <asm/kexec.h>
35
+
#include <asm/kvm_ppc.h>
36
+
33
37
#include <mm/mmu_decl.h>
34
38
35
39
unsigned long long memory_limit __initdata;
···
272
268
273
269
void __init arch_mm_preinit(void)
274
270
{
271
+
272
+
/*
273
+
* Reserve large chunks of memory for use by CMA for kdump, fadump, KVM
274
+
* and hugetlb. These must be called after pageblock_order is
275
+
* initialised.
276
+
*/
277
+
fadump_cma_init();
278
+
kdump_cma_reserve();
279
+
kvm_cma_reserve();
280
+
275
281
/*
276
282
* book3s is limited to 16 page sizes due to encoding this in
277
283
* a 4-bit field for slices.
+5
arch/powerpc/perf/callchain.c
+5
arch/powerpc/perf/callchain.c
···
103
103
void
104
104
perf_callchain_user(struct perf_callchain_entry_ctx *entry, struct pt_regs *regs)
105
105
{
106
+
perf_callchain_store(entry, perf_arch_instruction_pointer(regs));
107
+
108
+
if (!current->mm)
109
+
return;
110
+
106
111
if (!is_32bit_task())
107
112
perf_callchain_user_64(entry, regs);
108
113
else
-1
arch/powerpc/perf/callchain_32.c
-1
arch/powerpc/perf/callchain_32.c
-1
arch/powerpc/perf/callchain_64.c
-1
arch/powerpc/perf/callchain_64.c
+2
-2
tools/testing/selftests/powerpc/copyloops/.gitignore
+2
-2
tools/testing/selftests/powerpc/copyloops/.gitignore
+8
-3
tools/testing/selftests/powerpc/copyloops/Makefile
+8
-3
tools/testing/selftests/powerpc/copyloops/Makefile
···
1
1
# SPDX-License-Identifier: GPL-2.0
2
2
TEST_GEN_PROGS := copyuser_64_t0 copyuser_64_t1 copyuser_64_t2 \
3
-
copyuser_p7_t0 copyuser_p7_t1 \
3
+
copyuser_p7 copyuser_p7_vmx \
4
4
memcpy_64_t0 memcpy_64_t1 memcpy_64_t2 \
5
5
memcpy_p7_t0 memcpy_p7_t1 copy_mc_64 \
6
6
copyuser_64_exc_t0 copyuser_64_exc_t1 copyuser_64_exc_t2 \
···
28
28
-D SELFTEST_CASE=$(subst copyuser_64_t,,$(notdir $@)) \
29
29
-o $@ $^
30
30
31
-
$(OUTPUT)/copyuser_p7_t%: copyuser_power7.S $(EXTRA_SOURCES)
31
+
$(OUTPUT)/copyuser_p7: copyuser_power7.S $(EXTRA_SOURCES)
32
32
$(CC) $(CPPFLAGS) $(CFLAGS) \
33
33
-D COPY_LOOP=test___copy_tofrom_user_power7 \
34
-
-D SELFTEST_CASE=$(subst copyuser_p7_t,,$(notdir $@)) \
34
+
-o $@ $^
35
+
36
+
$(OUTPUT)/copyuser_p7_vmx: copyuser_power7.S $(EXTRA_SOURCES) ../utils.c
37
+
$(CC) $(CPPFLAGS) $(CFLAGS) \
38
+
-D COPY_LOOP=test___copy_tofrom_user_power7_vmx \
39
+
-D VMX_TEST \
35
40
-o $@ $^
36
41
37
42
# Strictly speaking, we only need the memcpy_64 test cases for big-endian
-8
tools/testing/selftests/powerpc/copyloops/stubs.S
-8
tools/testing/selftests/powerpc/copyloops/stubs.S
+14
-1
tools/testing/selftests/powerpc/copyloops/validate.c
+14
-1
tools/testing/selftests/powerpc/copyloops/validate.c
···
12
12
#define BUFLEN (MAX_LEN+MAX_OFFSET+2*MIN_REDZONE)
13
13
#define POISON 0xa5
14
14
15
+
#ifdef VMX_TEST
16
+
#define VMX_COPY_THRESHOLD 3328
17
+
#endif
18
+
15
19
unsigned long COPY_LOOP(void *to, const void *from, unsigned long size);
16
20
17
21
static void do_one(char *src, char *dst, unsigned long src_off,
···
85
81
/* Fill with sequential bytes */
86
82
for (i = 0; i < BUFLEN; i++)
87
83
fill[i] = i & 0xff;
88
-
84
+
#ifdef VMX_TEST
85
+
/* Force sizes above kernel VMX threshold (3328) */
86
+
for (len = VMX_COPY_THRESHOLD + 1; len < MAX_LEN; len++) {
87
+
#else
89
88
for (len = 1; len < MAX_LEN; len++) {
89
+
#endif
90
90
for (src_off = 0; src_off < MAX_OFFSET; src_off++) {
91
91
for (dst_off = 0; dst_off < MAX_OFFSET; dst_off++) {
92
92
do_one(src, dst, src_off, dst_off, len,
···
104
96
105
97
int main(void)
106
98
{
99
+
#ifdef VMX_TEST
100
+
/* Skip if Altivec not present */
101
+
SKIP_IF_MSG(!have_hwcap(PPC_FEATURE_HAS_ALTIVEC), "ALTIVEC not supported");
102
+
#endif
103
+
107
104
return test_harness(test_copy_loop, str(COPY_LOOP));
108
105
}