Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'x86_urgent_for_v5.18_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 fixes from Borislav Petkov:
- Fix the MSI message data struct definition
- Use local labels in the exception table macros to avoid symbol
conflicts with clang LTO builds
- A couple of fixes to objtool checking of the relatively newly added
SLS and IBT code
- Rename a local var in the WARN* macro machinery to prevent shadowing
* tag 'x86_urgent_for_v5.18_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/msi: Fix msi message data shadow struct
x86/extable: Prefer local labels in .set directives
x86,bpf: Avoid IBT objtool warning
objtool: Fix SLS validation for kcov tail-call replacement
objtool: Fix IBT tail-call detection
x86/bug: Prevent shadowing in __WARN_FLAGS
x86/mm/tlb: Revert retpoline avoidance approach
+54
-57
+10
-10
arch/x86/include/asm/asm.h
+10
-10
arch/x86/include/asm/asm.h
···
154
154
155
155
# define DEFINE_EXTABLE_TYPE_REG \
156
156
".macro extable_type_reg type:req reg:req\n" \
157
-
".set found, 0\n" \
158
-
".set regnr, 0\n" \
157
+
".set .Lfound, 0\n" \
158
+
".set .Lregnr, 0\n" \
159
159
".irp rs,rax,rcx,rdx,rbx,rsp,rbp,rsi,rdi,r8,r9,r10,r11,r12,r13,r14,r15\n" \
160
160
".ifc \\reg, %%\\rs\n" \
161
-
".set found, found+1\n" \
162
-
".long \\type + (regnr << 8)\n" \
161
+
".set .Lfound, .Lfound+1\n" \
162
+
".long \\type + (.Lregnr << 8)\n" \
163
163
".endif\n" \
164
-
".set regnr, regnr+1\n" \
164
+
".set .Lregnr, .Lregnr+1\n" \
165
165
".endr\n" \
166
-
".set regnr, 0\n" \
166
+
".set .Lregnr, 0\n" \
167
167
".irp rs,eax,ecx,edx,ebx,esp,ebp,esi,edi,r8d,r9d,r10d,r11d,r12d,r13d,r14d,r15d\n" \
168
168
".ifc \\reg, %%\\rs\n" \
169
-
".set found, found+1\n" \
170
-
".long \\type + (regnr << 8)\n" \
169
+
".set .Lfound, .Lfound+1\n" \
170
+
".long \\type + (.Lregnr << 8)\n" \
171
171
".endif\n" \
172
-
".set regnr, regnr+1\n" \
172
+
".set .Lregnr, .Lregnr+1\n" \
173
173
".endr\n" \
174
-
".if (found != 1)\n" \
174
+
".if (.Lfound != 1)\n" \
175
175
".error \"extable_type_reg: bad register argument\"\n" \
176
176
".endif\n" \
177
177
".endm\n"
+2
-2
arch/x86/include/asm/bug.h
+2
-2
arch/x86/include/asm/bug.h
···
78
78
*/
79
79
#define __WARN_FLAGS(flags) \
80
80
do { \
81
-
__auto_type f = BUGFLAG_WARNING|(flags); \
81
+
__auto_type __flags = BUGFLAG_WARNING|(flags); \
82
82
instrumentation_begin(); \
83
-
_BUG_FLAGS(ASM_UD2, f, ASM_REACHABLE); \
83
+
_BUG_FLAGS(ASM_UD2, __flags, ASM_REACHABLE); \
84
84
instrumentation_end(); \
85
85
} while (0)
86
86
+11
-8
arch/x86/include/asm/msi.h
+11
-8
arch/x86/include/asm/msi.h
···
12
12
/* Structs and defines for the X86 specific MSI message format */
13
13
14
14
typedef struct x86_msi_data {
15
-
u32 vector : 8,
16
-
delivery_mode : 3,
17
-
dest_mode_logical : 1,
18
-
reserved : 2,
19
-
active_low : 1,
20
-
is_level : 1;
21
-
22
-
u32 dmar_subhandle;
15
+
union {
16
+
struct {
17
+
u32 vector : 8,
18
+
delivery_mode : 3,
19
+
dest_mode_logical : 1,
20
+
reserved : 2,
21
+
active_low : 1,
22
+
is_level : 1;
23
+
};
24
+
u32 dmar_subhandle;
25
+
};
23
26
} __attribute__ ((packed)) arch_msi_msg_data_t;
24
27
#define arch_msi_msg_data x86_msi_data
25
28
+5
-32
arch/x86/mm/tlb.c
+5
-32
arch/x86/mm/tlb.c
···
855
855
nr_invalidate);
856
856
}
857
857
858
-
static bool tlb_is_not_lazy(int cpu)
858
+
static bool tlb_is_not_lazy(int cpu, void *data)
859
859
{
860
860
return !per_cpu(cpu_tlbstate_shared.is_lazy, cpu);
861
861
}
862
-
863
-
static DEFINE_PER_CPU(cpumask_t, flush_tlb_mask);
864
862
865
863
DEFINE_PER_CPU_SHARED_ALIGNED(struct tlb_state_shared, cpu_tlbstate_shared);
866
864
EXPORT_PER_CPU_SYMBOL(cpu_tlbstate_shared);
···
888
890
* up on the new contents of what used to be page tables, while
889
891
* doing a speculative memory access.
890
892
*/
891
-
if (info->freed_tables) {
893
+
if (info->freed_tables)
892
894
on_each_cpu_mask(cpumask, flush_tlb_func, (void *)info, true);
893
-
} else {
894
-
/*
895
-
* Although we could have used on_each_cpu_cond_mask(),
896
-
* open-coding it has performance advantages, as it eliminates
897
-
* the need for indirect calls or retpolines. In addition, it
898
-
* allows to use a designated cpumask for evaluating the
899
-
* condition, instead of allocating one.
900
-
*
901
-
* This code works under the assumption that there are no nested
902
-
* TLB flushes, an assumption that is already made in
903
-
* flush_tlb_mm_range().
904
-
*
905
-
* cond_cpumask is logically a stack-local variable, but it is
906
-
* more efficient to have it off the stack and not to allocate
907
-
* it on demand. Preemption is disabled and this code is
908
-
* non-reentrant.
909
-
*/
910
-
struct cpumask *cond_cpumask = this_cpu_ptr(&flush_tlb_mask);
911
-
int cpu;
912
-
913
-
cpumask_clear(cond_cpumask);
914
-
915
-
for_each_cpu(cpu, cpumask) {
916
-
if (tlb_is_not_lazy(cpu))
917
-
__cpumask_set_cpu(cpu, cond_cpumask);
918
-
}
919
-
on_each_cpu_mask(cond_cpumask, flush_tlb_func, (void *)info, true);
920
-
}
895
+
else
896
+
on_each_cpu_cond_mask(tlb_is_not_lazy, flush_tlb_func,
897
+
(void *)info, 1, cpumask);
921
898
}
922
899
923
900
void flush_tlb_multi(const struct cpumask *cpumask,
+1
arch/x86/net/bpf_jit_comp.c
+1
arch/x86/net/bpf_jit_comp.c
+25
-5
tools/objtool/check.c
+25
-5
tools/objtool/check.c
···
1155
1155
: arch_nop_insn(insn->len));
1156
1156
1157
1157
insn->type = sibling ? INSN_RETURN : INSN_NOP;
1158
+
1159
+
if (sibling) {
1160
+
/*
1161
+
* We've replaced the tail-call JMP insn by two new
1162
+
* insn: RET; INT3, except we only have a single struct
1163
+
* insn here. Mark it retpoline_safe to avoid the SLS
1164
+
* warning, instead of adding another insn.
1165
+
*/
1166
+
insn->retpoline_safe = true;
1167
+
}
1168
+
1158
1169
return;
1159
1170
}
1160
1171
···
1250
1239
return insn1->func->pfunc == insn2->func->pfunc;
1251
1240
}
1252
1241
1253
-
static bool is_first_func_insn(struct instruction *insn)
1242
+
static bool is_first_func_insn(struct objtool_file *file, struct instruction *insn)
1254
1243
{
1255
-
return insn->offset == insn->func->offset ||
1256
-
(insn->type == INSN_ENDBR &&
1257
-
insn->offset == insn->func->offset + insn->len);
1244
+
if (insn->offset == insn->func->offset)
1245
+
return true;
1246
+
1247
+
if (ibt) {
1248
+
struct instruction *prev = prev_insn_same_sym(file, insn);
1249
+
1250
+
if (prev && prev->type == INSN_ENDBR &&
1251
+
insn->offset == insn->func->offset + prev->len)
1252
+
return true;
1253
+
}
1254
+
1255
+
return false;
1258
1256
}
1259
1257
1260
1258
/*
···
1347
1327
insn->jump_dest->func->pfunc = insn->func;
1348
1328
1349
1329
} else if (!same_function(insn, insn->jump_dest) &&
1350
-
is_first_func_insn(insn->jump_dest)) {
1330
+
is_first_func_insn(file, insn->jump_dest)) {
1351
1331
/* internal sibling call (without reloc) */
1352
1332
add_call_dest(file, insn, insn->jump_dest->func, true);
1353
1333
}