Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 fixes from Ingo Molnar:
- misc fixes all around the map
- block non-root vm86(old) if mmap_min_addr != 0
- two small debuggability improvements
- removal of obsolete paravirt op
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/platform: Fix Geode LX timekeeping in the generic x86 build
x86/apic: Serialize LVTT and TSC_DEADLINE writes
x86/ioapic: Force affinity setting in setup_ioapic_dest()
x86/paravirt: Remove the unused pv_time_ops::get_tsc_khz method
x86/ldt: Fix small LDT allocation for Xen
x86/vm86: Fix the misleading CONFIG_VM86 Kconfig help text
x86/cpu: Print family/model/stepping in hex
x86/vm86: Block non-root vm86(old) if mmap_min_addr != 0
x86/alternatives: Make optimize_nops() interrupt safe and synced
x86/mm/srat: Print non-volatile flag in SRAT
x86/cpufeatures: Enable cpuid for Intel SHA extensions
+75
-29
+12
-11
arch/x86/Kconfig
+12
-11
arch/x86/Kconfig
···
1006
1006
depends on X86_MCE_INTEL
1007
1007
1008
1008
config X86_LEGACY_VM86
1009
-
bool "Legacy VM86 support (obsolete)"
1009
+
bool "Legacy VM86 support"
1010
1010
default n
1011
1011
depends on X86_32
1012
1012
---help---
···
1018
1018
available to accelerate real mode DOS programs. However, any
1019
1019
recent version of DOSEMU, X, or vbetool should be fully
1020
1020
functional even without kernel VM86 support, as they will all
1021
-
fall back to (pretty well performing) software emulation.
1021
+
fall back to software emulation. Nevertheless, if you are using
1022
+
a 16-bit DOS program where 16-bit performance matters, vm86
1023
+
mode might be faster than emulation and you might want to
1024
+
enable this option.
1022
1025
1023
-
Anything that works on a 64-bit kernel is unlikely to need
1024
-
this option, as 64-bit kernels don't, and can't, support V8086
1025
-
mode. This option is also unrelated to 16-bit protected mode
1026
-
and is not needed to run most 16-bit programs under Wine.
1026
+
Note that any app that works on a 64-bit kernel is unlikely to
1027
+
need this option, as 64-bit kernels don't, and can't, support
1028
+
V8086 mode. This option is also unrelated to 16-bit protected
1029
+
mode and is not needed to run most 16-bit programs under Wine.
1027
1030
1028
-
Enabling this option adds considerable attack surface to the
1029
-
kernel and slows down system calls and exception handling.
1031
+
Enabling this option increases the complexity of the kernel
1032
+
and slows down exception handling a tiny bit.
1030
1033
1031
-
Unless you use very old userspace or need the last drop of
1032
-
performance in your real mode DOS games and can't use KVM,
1033
-
say N here.
1034
+
If unsure, say N here.
1034
1035
1035
1036
config VM86
1036
1037
bool
+1
arch/x86/include/asm/cpufeature.h
+1
arch/x86/include/asm/cpufeature.h
···
241
241
#define X86_FEATURE_AVX512PF ( 9*32+26) /* AVX-512 Prefetch */
242
242
#define X86_FEATURE_AVX512ER ( 9*32+27) /* AVX-512 Exponential and Reciprocal */
243
243
#define X86_FEATURE_AVX512CD ( 9*32+28) /* AVX-512 Conflict Detection */
244
+
#define X86_FEATURE_SHA_NI ( 9*32+29) /* SHA1/SHA256 Instruction Extensions */
244
245
245
246
/* Extended state features, CPUID level 0x0000000d:1 (eax), word 10 */
246
247
#define X86_FEATURE_XSAVEOPT (10*32+ 0) /* XSAVEOPT */
-1
arch/x86/include/asm/paravirt_types.h
-1
arch/x86/include/asm/paravirt_types.h
+5
arch/x86/kernel/alternative.c
+5
arch/x86/kernel/alternative.c
···
338
338
339
339
static void __init_or_module optimize_nops(struct alt_instr *a, u8 *instr)
340
340
{
341
+
unsigned long flags;
342
+
341
343
if (instr[0] != 0x90)
342
344
return;
343
345
346
+
local_irq_save(flags);
344
347
add_nops(instr + (a->instrlen - a->padlen), a->padlen);
348
+
sync_core();
349
+
local_irq_restore(flags);
345
350
346
351
DUMP_BYTES(instr, a->instrlen, "%p: [%d:%d) optimized NOPs: ",
347
352
instr, a->instrlen - a->padlen, a->padlen);
+7
arch/x86/kernel/apic/apic.c
+7
arch/x86/kernel/apic/apic.c
···
336
336
apic_write(APIC_LVTT, lvtt_value);
337
337
338
338
if (lvtt_value & APIC_LVT_TIMER_TSCDEADLINE) {
339
+
/*
340
+
* See Intel SDM: TSC-Deadline Mode chapter. In xAPIC mode,
341
+
* writing to the APIC LVTT and TSC_DEADLINE MSR isn't serialized.
342
+
* According to Intel, MFENCE can do the serialization here.
343
+
*/
344
+
asm volatile("mfence" : : : "memory");
345
+
339
346
printk_once(KERN_DEBUG "TSC deadline timer enabled\n");
340
347
return;
341
348
}
+3
-2
arch/x86/kernel/apic/io_apic.c
+3
-2
arch/x86/kernel/apic/io_apic.c
···
2522
2522
int pin, ioapic, irq, irq_entry;
2523
2523
const struct cpumask *mask;
2524
2524
struct irq_data *idata;
2525
+
struct irq_chip *chip;
2525
2526
2526
2527
if (skip_ioapic_setup == 1)
2527
2528
return;
···
2546
2545
else
2547
2546
mask = apic->target_cpus();
2548
2547
2549
-
irq_set_affinity(irq, mask);
2548
+
chip = irq_data_get_irq_chip(idata);
2549
+
chip->irq_set_affinity(idata, mask, false);
2550
2550
}
2551
-
2552
2551
}
2553
2552
#endif
2554
2553
+2
-2
arch/x86/kernel/cpu/common.c
+2
-2
arch/x86/kernel/cpu/common.c
···
1110
1110
else
1111
1111
printk(KERN_CONT "%d86", c->x86);
1112
1112
1113
-
printk(KERN_CONT " (fam: %02x, model: %02x", c->x86, c->x86_model);
1113
+
printk(KERN_CONT " (family: 0x%x, model: 0x%x", c->x86, c->x86_model);
1114
1114
1115
1115
if (c->x86_mask || c->cpuid_level >= 0)
1116
-
printk(KERN_CONT ", stepping: %02x)\n", c->x86_mask);
1116
+
printk(KERN_CONT ", stepping: 0x%x)\n", c->x86_mask);
1117
1117
else
1118
1118
printk(KERN_CONT ")\n");
1119
1119
+2
-2
arch/x86/kernel/ldt.c
+2
-2
arch/x86/kernel/ldt.c
···
58
58
if (alloc_size > PAGE_SIZE)
59
59
new_ldt->entries = vzalloc(alloc_size);
60
60
else
61
-
new_ldt->entries = kzalloc(PAGE_SIZE, GFP_KERNEL);
61
+
new_ldt->entries = (void *)get_zeroed_page(GFP_KERNEL);
62
62
63
63
if (!new_ldt->entries) {
64
64
kfree(new_ldt);
···
95
95
if (ldt->size * LDT_ENTRY_SIZE > PAGE_SIZE)
96
96
vfree(ldt->entries);
97
97
else
98
-
kfree(ldt->entries);
98
+
free_page((unsigned long)ldt->entries);
99
99
kfree(ldt);
100
100
}
101
101
+10
-7
arch/x86/kernel/tsc.c
+10
-7
arch/x86/kernel/tsc.c
···
21
21
#include <asm/hypervisor.h>
22
22
#include <asm/nmi.h>
23
23
#include <asm/x86_init.h>
24
+
#include <asm/geode.h>
24
25
25
26
unsigned int __read_mostly cpu_khz; /* TSC clocks / usec, not used here */
26
27
EXPORT_SYMBOL(cpu_khz);
···
1014
1013
1015
1014
static void __init check_system_tsc_reliable(void)
1016
1015
{
1017
-
#ifdef CONFIG_MGEODE_LX
1018
-
/* RTSC counts during suspend */
1016
+
#if defined(CONFIG_MGEODEGX1) || defined(CONFIG_MGEODE_LX) || defined(CONFIG_X86_GENERIC)
1017
+
if (is_geode_lx()) {
1018
+
/* RTSC counts during suspend */
1019
1019
#define RTSC_SUSP 0x100
1020
-
unsigned long res_low, res_high;
1020
+
unsigned long res_low, res_high;
1021
1021
1022
-
rdmsr_safe(MSR_GEODE_BUSCONT_CONF0, &res_low, &res_high);
1023
-
/* Geode_LX - the OLPC CPU has a very reliable TSC */
1024
-
if (res_low & RTSC_SUSP)
1025
-
tsc_clocksource_reliable = 1;
1022
+
rdmsr_safe(MSR_GEODE_BUSCONT_CONF0, &res_low, &res_high);
1023
+
/* Geode_LX - the OLPC CPU has a very reliable TSC */
1024
+
if (res_low & RTSC_SUSP)
1025
+
tsc_clocksource_reliable = 1;
1026
+
}
1026
1027
#endif
1027
1028
if (boot_cpu_has(X86_FEATURE_TSC_RELIABLE))
1028
1029
tsc_clocksource_reliable = 1;
+27
arch/x86/kernel/vm86_32.c
+27
arch/x86/kernel/vm86_32.c
···
45
45
#include <linux/audit.h>
46
46
#include <linux/stddef.h>
47
47
#include <linux/slab.h>
48
+
#include <linux/security.h>
48
49
49
50
#include <asm/uaccess.h>
50
51
#include <asm/io.h>
···
232
231
struct kernel_vm86_regs vm86regs;
233
232
struct pt_regs *regs = current_pt_regs();
234
233
unsigned long err = 0;
234
+
235
+
err = security_mmap_addr(0);
236
+
if (err) {
237
+
/*
238
+
* vm86 cannot virtualize the address space, so vm86 users
239
+
* need to manage the low 1MB themselves using mmap. Given
240
+
* that BIOS places important data in the first page, vm86
241
+
* is essentially useless if mmap_min_addr != 0. DOSEMU,
242
+
* for example, won't even bother trying to use vm86 if it
243
+
* can't map a page at virtual address 0.
244
+
*
245
+
* To reduce the available kernel attack surface, simply
246
+
* disallow vm86(old) for users who cannot mmap at va 0.
247
+
*
248
+
* The implementation of security_mmap_addr will allow
249
+
* suitably privileged users to map va 0 even if
250
+
* vm.mmap_min_addr is set above 0, and we want this
251
+
* behavior for vm86 as well, as it ensures that legacy
252
+
* tools like vbetool will not fail just because of
253
+
* vm.mmap_min_addr.
254
+
*/
255
+
pr_info_once("Denied a call to vm86(old) from %s[%d] (uid: %d). Set the vm.mmap_min_addr sysctl to 0 and/or adjust LSM mmap_min_addr policy to enable vm86 if you are using a vm86-based DOS emulator.\n",
256
+
current->comm, task_pid_nr(current),
257
+
from_kuid_munged(&init_user_ns, current_uid()));
258
+
return -EPERM;
259
+
}
235
260
236
261
if (!vm86) {
237
262
if (!(vm86 = kzalloc(sizeof(*vm86), GFP_KERNEL)))
+3
-2
arch/x86/mm/srat.c
+3
-2
arch/x86/mm/srat.c
···
192
192
193
193
node_set(node, numa_nodes_parsed);
194
194
195
-
pr_info("SRAT: Node %u PXM %u [mem %#010Lx-%#010Lx]%s\n",
195
+
pr_info("SRAT: Node %u PXM %u [mem %#010Lx-%#010Lx]%s%s\n",
196
196
node, pxm,
197
197
(unsigned long long) start, (unsigned long long) end - 1,
198
-
hotpluggable ? " hotplug" : "");
198
+
hotpluggable ? " hotplug" : "",
199
+
ma->flags & ACPI_SRAT_MEM_NON_VOLATILE ? " non-volatile" : "");
199
200
200
201
/* Mark hotplug range in memblock. */
201
202
if (hotpluggable && memblock_mark_hotplug(start, ma->length))
+3
-2
tools/testing/selftests/x86/entry_from_vm86.c
+3
-2
tools/testing/selftests/x86/entry_from_vm86.c
···
116
116
v86->regs.eip = eip;
117
117
ret = vm86(VM86_ENTER, v86);
118
118
119
-
if (ret == -1 && errno == ENOSYS) {
120
-
printf("[SKIP]\tvm86 not supported\n");
119
+
if (ret == -1 && (errno == ENOSYS || errno == EPERM)) {
120
+
printf("[SKIP]\tvm86 %s\n",
121
+
errno == ENOSYS ? "not supported" : "not allowed");
121
122
return false;
122
123
}
123
124