* [PATCH 0/3] x86/mce: Fix more noinstr fun @ 2022-02-04 8:30 Borislav Petkov 2022-02-04 8:30 ` [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper Borislav Petkov ` (3 more replies) 0 siblings, 4 replies; 11+ messages in thread From: Borislav Petkov @ 2022-02-04 8:30 UTC (permalink / raw) To: Tony Luck; +Cc: Jakub Kicinski, Linus Torvalds, Marco Elver, X86 ML, LKML From: Borislav Petkov <bp@suse.de> Hi all, here's a second small set of fixes for objtool noinstr validation issues in the MCE code. It goes ontop of tip:locking/core where there are some more commits changing generic helpers to be always inlined. As always, comments and suggestions are appreciated. Thx. Borislav Petkov (3): cpumask: Add a x86-specific cpumask_clear_cpu() helper x86/ptrace: Always inline v8086_mode() for instrumentation x86/mce: Use arch atomic and bit helpers arch/x86/include/asm/cpumask.h | 10 ++++++ arch/x86/include/asm/ptrace.h | 2 +- arch/x86/kernel/cpu/mce/core.c | 58 ++++++++++-------------------- arch/x86/kernel/cpu/mce/internal.h | 23 ++++++++++-- arch/x86/kernel/cpu/mce/severity.c | 2 +- include/linux/cpumask.h | 4 +-- 6 files changed, 54 insertions(+), 45 deletions(-) -- 2.29.2 ^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper 2022-02-04 8:30 [PATCH 0/3] x86/mce: Fix more noinstr fun Borislav Petkov @ 2022-02-04 8:30 ` Borislav Petkov 2022-02-04 16:46 ` Luck, Tony 2022-02-13 10:59 ` [tip: locking/core] " tip-bot2 for Borislav Petkov 2022-02-04 8:30 ` [PATCH 2/3] x86/ptrace: Always inline v8086_mode() for instrumentation Borislav Petkov ` (2 subsequent siblings) 3 siblings, 2 replies; 11+ messages in thread From: Borislav Petkov @ 2022-02-04 8:30 UTC (permalink / raw) To: Tony Luck; +Cc: Jakub Kicinski, Linus Torvalds, Marco Elver, X86 ML, LKML From: Borislav Petkov <bp@suse.de> Add a x86-specific cpumask_clear_cpu() helper which will be used in places where the explicit KASAN-instrumentation in the *_bit() helpers is unwanted. Also, always inline two more cpumask generic helpers. allyesconfig: text data bss dec hex filename 190553143 159425889 32076404 382055436 16c5b40c vmlinux.before 190551812 159424945 32076404 382053161 16c5ab29 vmlinux.after Signed-off-by: Borislav Petkov <bp@suse.de> --- arch/x86/include/asm/cpumask.h | 10 ++++++++++ include/linux/cpumask.h | 4 ++-- 2 files changed, 12 insertions(+), 2 deletions(-) diff --git a/arch/x86/include/asm/cpumask.h b/arch/x86/include/asm/cpumask.h index 3afa990d756b..c5aed9e9226c 100644 --- a/arch/x86/include/asm/cpumask.h +++ b/arch/x86/include/asm/cpumask.h @@ -20,11 +20,21 @@ static __always_inline bool arch_cpu_online(int cpu) { return arch_test_bit(cpu, cpumask_bits(cpu_online_mask)); } + +static __always_inline void arch_cpumask_clear_cpu(int cpu, struct cpumask *dstp) +{ + arch_clear_bit(cpumask_check(cpu), cpumask_bits(dstp)); +} #else static __always_inline bool arch_cpu_online(int cpu) { return cpu == 0; } + +static __always_inline void arch_cpumask_clear_cpu(int cpu, struct cpumask *dstp) +{ + return; +} #endif #define arch_cpu_is_offline(cpu) unlikely(!arch_cpu_online(cpu)) diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h index 6b06c698cd2a..fe29ac7cc469 100644 --- a/include/linux/cpumask.h +++ b/include/linux/cpumask.h @@ -102,7 +102,7 @@ extern atomic_t __num_online_cpus; extern cpumask_t cpus_booted_once_mask; -static inline void cpu_max_bits_warn(unsigned int cpu, unsigned int bits) +static __always_inline void cpu_max_bits_warn(unsigned int cpu, unsigned int bits) { #ifdef CONFIG_DEBUG_PER_CPU_MAPS WARN_ON_ONCE(cpu >= bits); @@ -110,7 +110,7 @@ static inline void cpu_max_bits_warn(unsigned int cpu, unsigned int bits) } /* verify cpu argument to cpumask_* operators */ -static inline unsigned int cpumask_check(unsigned int cpu) +static __always_inline unsigned int cpumask_check(unsigned int cpu) { cpu_max_bits_warn(cpu, nr_cpumask_bits); return cpu; -- 2.29.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* RE: [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper 2022-02-04 8:30 ` [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper Borislav Petkov @ 2022-02-04 16:46 ` Luck, Tony 2022-02-04 18:04 ` Borislav Petkov 2022-02-13 10:59 ` [tip: locking/core] " tip-bot2 for Borislav Petkov 1 sibling, 1 reply; 11+ messages in thread From: Luck, Tony @ 2022-02-04 16:46 UTC (permalink / raw) To: Borislav Petkov; +Cc: Jakub Kicinski, Linus Torvalds, Marco Elver, X86 ML, LKML > Also, always inline two more cpumask generic helpers. > > allyesconfig: > > text data bss dec hex filename > 190553143 159425889 32076404 382055436 16c5b40c vmlinux.before > 190551812 159424945 32076404 382053161 16c5ab29 vmlinux.after That looks odd. You added "always" to some inline functions, which I'd have thought might cause *more* inlining, and thus an increase in text size. But it actually went *down* by 1331 bytes. -Tony ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper 2022-02-04 16:46 ` Luck, Tony @ 2022-02-04 18:04 ` Borislav Petkov 0 siblings, 0 replies; 11+ messages in thread From: Borislav Petkov @ 2022-02-04 18:04 UTC (permalink / raw) To: Luck, Tony; +Cc: Jakub Kicinski, Linus Torvalds, Marco Elver, X86 ML, LKML On Fri, Feb 04, 2022 at 04:46:24PM +0000, Luck, Tony wrote: > > Also, always inline two more cpumask generic helpers. > > > > allyesconfig: > > > > text data bss dec hex filename > > 190553143 159425889 32076404 382055436 16c5b40c vmlinux.before > > 190551812 159424945 32076404 382053161 16c5ab29 vmlinux.after > > That looks odd. You added "always" to some inline functions, which I'd > have thought might cause *more* inlining, and thus an increase in text > size. But it actually went *down* by 1331 bytes. Or maybe the compiler did some optimizations after the inlining. Here are the numbers on an allmodconfig - 1073 bytes less with this patch. text data bss dec hex filename 27409158 31062465 26710144 85181767 513c547 vmlinux.before 27408085 31062821 26710144 85181050 513c27a vmlinux.after -- Regards/Gruss, Boris. https://people.kernel.org/tglx/notes-about-netiquette ^ permalink raw reply [flat|nested] 11+ messages in thread
* [tip: locking/core] cpumask: Add a x86-specific cpumask_clear_cpu() helper 2022-02-04 8:30 ` [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper Borislav Petkov 2022-02-04 16:46 ` Luck, Tony @ 2022-02-13 10:59 ` tip-bot2 for Borislav Petkov 1 sibling, 0 replies; 11+ messages in thread From: tip-bot2 for Borislav Petkov @ 2022-02-13 10:59 UTC (permalink / raw) To: linux-tip-commits; +Cc: Borislav Petkov, Marco Elver, x86, linux-kernel The following commit has been merged into the locking/core branch of tip: Commit-ID: f5c54f77b07b278cfde4a654e111c39996ac8b5b Gitweb: https://git.kernel.org/tip/f5c54f77b07b278cfde4a654e111c39996ac8b5b Author: Borislav Petkov <bp@suse.de> AuthorDate: Fri, 04 Feb 2022 09:30:13 +01:00 Committer: Borislav Petkov <bp@suse.de> CommitterDate: Sat, 12 Feb 2022 18:20:05 +01:00 cpumask: Add a x86-specific cpumask_clear_cpu() helper Add a x86-specific cpumask_clear_cpu() helper which will be used in places where the explicit KASAN-instrumentation in the *_bit() helpers is unwanted. Also, always inline two more cpumask generic helpers. allyesconfig: text data bss dec hex filename 190553143 159425889 32076404 382055436 16c5b40c vmlinux.before 190551812 159424945 32076404 382053161 16c5ab29 vmlinux.after Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Marco Elver <elver@google.com> Link: https://lore.kernel.org/r/20220204083015.17317-2-bp@alien8.de --- arch/x86/include/asm/cpumask.h | 10 ++++++++++ include/linux/cpumask.h | 4 ++-- 2 files changed, 12 insertions(+), 2 deletions(-) diff --git a/arch/x86/include/asm/cpumask.h b/arch/x86/include/asm/cpumask.h index 3afa990..c5aed9e 100644 --- a/arch/x86/include/asm/cpumask.h +++ b/arch/x86/include/asm/cpumask.h @@ -20,11 +20,21 @@ static __always_inline bool arch_cpu_online(int cpu) { return arch_test_bit(cpu, cpumask_bits(cpu_online_mask)); } + +static __always_inline void arch_cpumask_clear_cpu(int cpu, struct cpumask *dstp) +{ + arch_clear_bit(cpumask_check(cpu), cpumask_bits(dstp)); +} #else static __always_inline bool arch_cpu_online(int cpu) { return cpu == 0; } + +static __always_inline void arch_cpumask_clear_cpu(int cpu, struct cpumask *dstp) +{ + return; +} #endif #define arch_cpu_is_offline(cpu) unlikely(!arch_cpu_online(cpu)) diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h index 6b06c69..fe29ac7 100644 --- a/include/linux/cpumask.h +++ b/include/linux/cpumask.h @@ -102,7 +102,7 @@ extern atomic_t __num_online_cpus; extern cpumask_t cpus_booted_once_mask; -static inline void cpu_max_bits_warn(unsigned int cpu, unsigned int bits) +static __always_inline void cpu_max_bits_warn(unsigned int cpu, unsigned int bits) { #ifdef CONFIG_DEBUG_PER_CPU_MAPS WARN_ON_ONCE(cpu >= bits); @@ -110,7 +110,7 @@ static inline void cpu_max_bits_warn(unsigned int cpu, unsigned int bits) } /* verify cpu argument to cpumask_* operators */ -static inline unsigned int cpumask_check(unsigned int cpu) +static __always_inline unsigned int cpumask_check(unsigned int cpu) { cpu_max_bits_warn(cpu, nr_cpumask_bits); return cpu; ^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 2/3] x86/ptrace: Always inline v8086_mode() for instrumentation 2022-02-04 8:30 [PATCH 0/3] x86/mce: Fix more noinstr fun Borislav Petkov 2022-02-04 8:30 ` [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper Borislav Petkov @ 2022-02-04 8:30 ` Borislav Petkov 2022-02-13 10:59 ` [tip: locking/core] " tip-bot2 for Borislav Petkov 2022-02-04 8:30 ` [PATCH 3/3] x86/mce: Use arch atomic and bit helpers Borislav Petkov 2022-02-04 11:24 ` [PATCH 0/3] x86/mce: Fix more noinstr fun Marco Elver 3 siblings, 1 reply; 11+ messages in thread From: Borislav Petkov @ 2022-02-04 8:30 UTC (permalink / raw) To: Tony Luck; +Cc: Jakub Kicinski, Linus Torvalds, Marco Elver, X86 ML, LKML From: Borislav Petkov <bp@suse.de> Instrumentation glue like KASAN causes the following warning: vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.constprop.0() leaves .noinstr.text section due to gcc creating a function call for that oneliner. Force-inline it and even save some vmlinux bytes (.config is close to an allmodconfig): text data bss dec hex filename 209431677 208257651 34411048 452100376 1af28118 vmlinux.before 209431519 208257615 34411048 452100182 1af28056 vmlinux.after Signed-off-by: Borislav Petkov <bp@suse.de> --- arch/x86/include/asm/ptrace.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/include/asm/ptrace.h b/arch/x86/include/asm/ptrace.h index 703663175a5a..4357e0f2cd5f 100644 --- a/arch/x86/include/asm/ptrace.h +++ b/arch/x86/include/asm/ptrace.h @@ -137,7 +137,7 @@ static __always_inline int user_mode(struct pt_regs *regs) #endif } -static inline int v8086_mode(struct pt_regs *regs) +static __always_inline int v8086_mode(struct pt_regs *regs) { #ifdef CONFIG_X86_32 return (regs->flags & X86_VM_MASK); -- 2.29.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* [tip: locking/core] x86/ptrace: Always inline v8086_mode() for instrumentation 2022-02-04 8:30 ` [PATCH 2/3] x86/ptrace: Always inline v8086_mode() for instrumentation Borislav Petkov @ 2022-02-13 10:59 ` tip-bot2 for Borislav Petkov 0 siblings, 0 replies; 11+ messages in thread From: tip-bot2 for Borislav Petkov @ 2022-02-13 10:59 UTC (permalink / raw) To: linux-tip-commits; +Cc: Borislav Petkov, Marco Elver, x86, linux-kernel The following commit has been merged into the locking/core branch of tip: Commit-ID: b008893b08dcc8c30d756db05c229a1491bcb992 Gitweb: https://git.kernel.org/tip/b008893b08dcc8c30d756db05c229a1491bcb992 Author: Borislav Petkov <bp@suse.de> AuthorDate: Fri, 04 Feb 2022 09:30:14 +01:00 Committer: Borislav Petkov <bp@suse.de> CommitterDate: Sat, 12 Feb 2022 22:07:13 +01:00 x86/ptrace: Always inline v8086_mode() for instrumentation Instrumentation glue like KASAN causes the following warning: vmlinux.o: warning: objtool: mce_gather_info()+0x5f: call to v8086_mode.constprop.0() leaves .noinstr.text section due to gcc creating a function call for that oneliner. Force-inline it and even save some vmlinux bytes (.config is close to an allmodconfig): text data bss dec hex filename 209431677 208257651 34411048 452100376 1af28118 vmlinux.before 209431519 208257615 34411048 452100182 1af28056 vmlinux.after Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Marco Elver <elver@google.com> Link: https://lore.kernel.org/r/20220204083015.17317-3-bp@alien8.de --- arch/x86/include/asm/ptrace.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/arch/x86/include/asm/ptrace.h b/arch/x86/include/asm/ptrace.h index 7036631..4357e0f 100644 --- a/arch/x86/include/asm/ptrace.h +++ b/arch/x86/include/asm/ptrace.h @@ -137,7 +137,7 @@ static __always_inline int user_mode(struct pt_regs *regs) #endif } -static inline int v8086_mode(struct pt_regs *regs) +static __always_inline int v8086_mode(struct pt_regs *regs) { #ifdef CONFIG_X86_32 return (regs->flags & X86_VM_MASK); ^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH 3/3] x86/mce: Use arch atomic and bit helpers 2022-02-04 8:30 [PATCH 0/3] x86/mce: Fix more noinstr fun Borislav Petkov 2022-02-04 8:30 ` [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper Borislav Petkov 2022-02-04 8:30 ` [PATCH 2/3] x86/ptrace: Always inline v8086_mode() for instrumentation Borislav Petkov @ 2022-02-04 8:30 ` Borislav Petkov 2022-02-14 10:03 ` [tip: ras/core] " tip-bot2 for Borislav Petkov 2022-02-04 11:24 ` [PATCH 0/3] x86/mce: Fix more noinstr fun Marco Elver 3 siblings, 1 reply; 11+ messages in thread From: Borislav Petkov @ 2022-02-04 8:30 UTC (permalink / raw) To: Tony Luck; +Cc: Jakub Kicinski, Linus Torvalds, Marco Elver, X86 ML, LKML From: Borislav Petkov <bp@suse.de> The arch helpers do not have explicit KASAN instrumentation. Use them in noinstr code. Inline a couple more functions with single call sites, while at it: mce_severity_amd_smca() has a single call-site which is noinstr so force the inlining and fix: vmlinux.o: warning: objtool: mce_severity_amd.constprop.0()+0xca: call to \ mce_severity_amd_smca() leaves .noinstr.text section Always inline mca_msr_reg(): text data bss dec hex filename 16065240 128031326 36405368 180501934 ac23dae vmlinux.before 16065240 128031294 36405368 180501902 ac23d8e vmlinux.after and mce_no_way_out() as the latter one is used only once, to fix: vmlinux.o: warning: objtool: mce_read_aux()+0x53: call to mca_msr_reg() leaves .noinstr.text section vmlinux.o: warning: objtool: do_machine_check()+0xc9: call to mce_no_way_out() leaves .noinstr.text section Signed-off-by: Borislav Petkov <bp@suse.de> --- arch/x86/kernel/cpu/mce/core.c | 58 ++++++++++-------------------- arch/x86/kernel/cpu/mce/internal.h | 23 ++++++++++-- arch/x86/kernel/cpu/mce/severity.c | 2 +- 3 files changed, 41 insertions(+), 42 deletions(-) diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c index 5818b837fd4d..0e7147430ec0 100644 --- a/arch/x86/kernel/cpu/mce/core.c +++ b/arch/x86/kernel/cpu/mce/core.c @@ -173,27 +173,6 @@ void mce_unregister_decode_chain(struct notifier_block *nb) } EXPORT_SYMBOL_GPL(mce_unregister_decode_chain); -u32 mca_msr_reg(int bank, enum mca_msr reg) -{ - if (mce_flags.smca) { - switch (reg) { - case MCA_CTL: return MSR_AMD64_SMCA_MCx_CTL(bank); - case MCA_ADDR: return MSR_AMD64_SMCA_MCx_ADDR(bank); - case MCA_MISC: return MSR_AMD64_SMCA_MCx_MISC(bank); - case MCA_STATUS: return MSR_AMD64_SMCA_MCx_STATUS(bank); - } - } - - switch (reg) { - case MCA_CTL: return MSR_IA32_MCx_CTL(bank); - case MCA_ADDR: return MSR_IA32_MCx_ADDR(bank); - case MCA_MISC: return MSR_IA32_MCx_MISC(bank); - case MCA_STATUS: return MSR_IA32_MCx_STATUS(bank); - } - - return 0; -} - static void __print_mce(struct mce *m) { pr_emerg(HW_ERR "CPU %d: Machine Check%s: %Lx Bank %d: %016Lx\n", @@ -814,7 +793,8 @@ EXPORT_SYMBOL_GPL(machine_check_poll); * the severity assessment code. Pretend that EIPV was set, and take the * ip/cs values from the pt_regs that mce_gather_info() ignored earlier. */ -static void quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) +static __always_inline void +quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) { if (bank != 0) return; @@ -838,8 +818,8 @@ static void quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) * Do a quick check if any of the events requires a panic. * This decides if we keep the events around or clear them. */ -static int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, - struct pt_regs *regs) +static __always_inline int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, + struct pt_regs *regs) { char *tmp = *msg; int i; @@ -849,7 +829,7 @@ static int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, if (!(m->status & MCI_STATUS_VAL)) continue; - __set_bit(i, validp); + arch___set_bit(i, validp); if (mce_flags.snb_ifu_quirk) quirk_sandybridge_ifu(i, m, regs); @@ -1015,13 +995,13 @@ static noinstr int mce_start(int *no_way_out) if (!timeout) return ret; - atomic_add(*no_way_out, &global_nwo); + arch_atomic_add(*no_way_out, &global_nwo); /* * Rely on the implied barrier below, such that global_nwo * is updated before mce_callin. */ - order = atomic_inc_return(&mce_callin); - cpumask_clear_cpu(smp_processor_id(), &mce_missing_cpus); + order = arch_atomic_inc_return(&mce_callin); + arch_cpumask_clear_cpu(smp_processor_id(), &mce_missing_cpus); /* Enable instrumentation around calls to external facilities */ instrumentation_begin(); @@ -1029,10 +1009,10 @@ static noinstr int mce_start(int *no_way_out) /* * Wait for everyone. */ - while (atomic_read(&mce_callin) != num_online_cpus()) { + while (arch_atomic_read(&mce_callin) != num_online_cpus()) { if (mce_timed_out(&timeout, "Timeout: Not all CPUs entered broadcast exception handler")) { - atomic_set(&global_nwo, 0); + arch_atomic_set(&global_nwo, 0); goto out; } ndelay(SPINUNIT); @@ -1047,7 +1027,7 @@ static noinstr int mce_start(int *no_way_out) /* * Monarch: Starts executing now, the others wait. */ - atomic_set(&mce_executing, 1); + arch_atomic_set(&mce_executing, 1); } else { /* * Subject: Now start the scanning loop one by one in @@ -1055,10 +1035,10 @@ static noinstr int mce_start(int *no_way_out) * This way when there are any shared banks it will be * only seen by one CPU before cleared, avoiding duplicates. */ - while (atomic_read(&mce_executing) < order) { + while (arch_atomic_read(&mce_executing) < order) { if (mce_timed_out(&timeout, "Timeout: Subject CPUs unable to finish machine check processing")) { - atomic_set(&global_nwo, 0); + arch_atomic_set(&global_nwo, 0); goto out; } ndelay(SPINUNIT); @@ -1068,7 +1048,7 @@ static noinstr int mce_start(int *no_way_out) /* * Cache the global no_way_out state. */ - *no_way_out = atomic_read(&global_nwo); + *no_way_out = arch_atomic_read(&global_nwo); ret = order; @@ -1153,12 +1133,12 @@ static noinstr int mce_end(int order) return ret; } -static void mce_clear_state(unsigned long *toclear) +static __always_inline void mce_clear_state(unsigned long *toclear) { int i; for (i = 0; i < this_cpu_read(mce_num_banks); i++) { - if (test_bit(i, toclear)) + if (arch_test_bit(i, toclear)) mce_wrmsrl(mca_msr_reg(i, MCA_STATUS), 0); } } @@ -1208,8 +1188,8 @@ __mc_scan_banks(struct mce *m, struct pt_regs *regs, struct mce *final, int severity, i, taint = 0; for (i = 0; i < this_cpu_read(mce_num_banks); i++) { - __clear_bit(i, toclear); - if (!test_bit(i, valid_banks)) + arch___clear_bit(i, toclear); + if (!arch_test_bit(i, valid_banks)) continue; if (!mce_banks[i].ctl) @@ -1244,7 +1224,7 @@ __mc_scan_banks(struct mce *m, struct pt_regs *regs, struct mce *final, severity == MCE_UCNA_SEVERITY) && !no_way_out) continue; - __set_bit(i, toclear); + arch___set_bit(i, toclear); /* Machine check event was not enabled. Clear, but ignore. */ if (severity == MCE_NO_SEVERITY) diff --git a/arch/x86/kernel/cpu/mce/internal.h b/arch/x86/kernel/cpu/mce/internal.h index 52c633950b38..a04b61e27827 100644 --- a/arch/x86/kernel/cpu/mce/internal.h +++ b/arch/x86/kernel/cpu/mce/internal.h @@ -182,8 +182,6 @@ enum mca_msr { MCA_MISC, }; -u32 mca_msr_reg(int bank, enum mca_msr reg); - /* Decide whether to add MCE record to MCE event pool or filter it out. */ extern bool filter_mce(struct mce *m); @@ -209,4 +207,25 @@ static inline void winchip_machine_check(struct pt_regs *regs) {} noinstr u64 mce_rdmsrl(u32 msr); +static __always_inline u32 mca_msr_reg(int bank, enum mca_msr reg) +{ + if (mce_flags.smca) { + switch (reg) { + case MCA_CTL: return MSR_AMD64_SMCA_MCx_CTL(bank); + case MCA_ADDR: return MSR_AMD64_SMCA_MCx_ADDR(bank); + case MCA_MISC: return MSR_AMD64_SMCA_MCx_MISC(bank); + case MCA_STATUS: return MSR_AMD64_SMCA_MCx_STATUS(bank); + } + } + + switch (reg) { + case MCA_CTL: return MSR_IA32_MCx_CTL(bank); + case MCA_ADDR: return MSR_IA32_MCx_ADDR(bank); + case MCA_MISC: return MSR_IA32_MCx_MISC(bank); + case MCA_STATUS: return MSR_IA32_MCx_STATUS(bank); + } + + return 0; +} + #endif /* __X86_MCE_INTERNAL_H__ */ diff --git a/arch/x86/kernel/cpu/mce/severity.c b/arch/x86/kernel/cpu/mce/severity.c index 7aa2bda93cbb..ca0d775346fb 100644 --- a/arch/x86/kernel/cpu/mce/severity.c +++ b/arch/x86/kernel/cpu/mce/severity.c @@ -301,7 +301,7 @@ static noinstr int error_context(struct mce *m, struct pt_regs *regs) } } -static int mce_severity_amd_smca(struct mce *m, enum context err_ctx) +static __always_inline int mce_severity_amd_smca(struct mce *m, enum context err_ctx) { u64 mcx_cfg; -- 2.29.2 ^ permalink raw reply related [flat|nested] 11+ messages in thread
* [tip: ras/core] x86/mce: Use arch atomic and bit helpers 2022-02-04 8:30 ` [PATCH 3/3] x86/mce: Use arch atomic and bit helpers Borislav Petkov @ 2022-02-14 10:03 ` tip-bot2 for Borislav Petkov 0 siblings, 0 replies; 11+ messages in thread From: tip-bot2 for Borislav Petkov @ 2022-02-14 10:03 UTC (permalink / raw) To: linux-tip-commits; +Cc: Borislav Petkov, Marco Elver, x86, linux-kernel The following commit has been merged into the ras/core branch of tip: Commit-ID: f11445ba7a1160b87615e3f863a9e66c85189399 Gitweb: https://git.kernel.org/tip/f11445ba7a1160b87615e3f863a9e66c85189399 Author: Borislav Petkov <bp@suse.de> AuthorDate: Fri, 04 Feb 2022 09:30:15 +01:00 Committer: Borislav Petkov <bp@suse.de> CommitterDate: Sun, 13 Feb 2022 22:08:27 +01:00 x86/mce: Use arch atomic and bit helpers The arch helpers do not have explicit KASAN instrumentation. Use them in noinstr code. Inline a couple more functions with single call sites, while at it: mce_severity_amd_smca() has a single call-site which is noinstr so force the inlining and fix: vmlinux.o: warning: objtool: mce_severity_amd.constprop.0()+0xca: call to \ mce_severity_amd_smca() leaves .noinstr.text section Always inline mca_msr_reg(): text data bss dec hex filename 16065240 128031326 36405368 180501934 ac23dae vmlinux.before 16065240 128031294 36405368 180501902 ac23d8e vmlinux.after and mce_no_way_out() as the latter one is used only once, to fix: vmlinux.o: warning: objtool: mce_read_aux()+0x53: call to mca_msr_reg() leaves .noinstr.text section vmlinux.o: warning: objtool: do_machine_check()+0xc9: call to mce_no_way_out() leaves .noinstr.text section Signed-off-by: Borislav Petkov <bp@suse.de> Acked-by: Marco Elver <elver@google.com> Link: https://lore.kernel.org/r/20220204083015.17317-4-bp@alien8.de --- arch/x86/kernel/cpu/mce/core.c | 58 +++++++++-------------------- arch/x86/kernel/cpu/mce/internal.h | 23 ++++++++++-- arch/x86/kernel/cpu/mce/severity.c | 2 +- 3 files changed, 41 insertions(+), 42 deletions(-) diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c index 5818b83..0e71474 100644 --- a/arch/x86/kernel/cpu/mce/core.c +++ b/arch/x86/kernel/cpu/mce/core.c @@ -173,27 +173,6 @@ void mce_unregister_decode_chain(struct notifier_block *nb) } EXPORT_SYMBOL_GPL(mce_unregister_decode_chain); -u32 mca_msr_reg(int bank, enum mca_msr reg) -{ - if (mce_flags.smca) { - switch (reg) { - case MCA_CTL: return MSR_AMD64_SMCA_MCx_CTL(bank); - case MCA_ADDR: return MSR_AMD64_SMCA_MCx_ADDR(bank); - case MCA_MISC: return MSR_AMD64_SMCA_MCx_MISC(bank); - case MCA_STATUS: return MSR_AMD64_SMCA_MCx_STATUS(bank); - } - } - - switch (reg) { - case MCA_CTL: return MSR_IA32_MCx_CTL(bank); - case MCA_ADDR: return MSR_IA32_MCx_ADDR(bank); - case MCA_MISC: return MSR_IA32_MCx_MISC(bank); - case MCA_STATUS: return MSR_IA32_MCx_STATUS(bank); - } - - return 0; -} - static void __print_mce(struct mce *m) { pr_emerg(HW_ERR "CPU %d: Machine Check%s: %Lx Bank %d: %016Lx\n", @@ -814,7 +793,8 @@ EXPORT_SYMBOL_GPL(machine_check_poll); * the severity assessment code. Pretend that EIPV was set, and take the * ip/cs values from the pt_regs that mce_gather_info() ignored earlier. */ -static void quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) +static __always_inline void +quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) { if (bank != 0) return; @@ -838,8 +818,8 @@ static void quirk_sandybridge_ifu(int bank, struct mce *m, struct pt_regs *regs) * Do a quick check if any of the events requires a panic. * This decides if we keep the events around or clear them. */ -static int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, - struct pt_regs *regs) +static __always_inline int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, + struct pt_regs *regs) { char *tmp = *msg; int i; @@ -849,7 +829,7 @@ static int mce_no_way_out(struct mce *m, char **msg, unsigned long *validp, if (!(m->status & MCI_STATUS_VAL)) continue; - __set_bit(i, validp); + arch___set_bit(i, validp); if (mce_flags.snb_ifu_quirk) quirk_sandybridge_ifu(i, m, regs); @@ -1015,13 +995,13 @@ static noinstr int mce_start(int *no_way_out) if (!timeout) return ret; - atomic_add(*no_way_out, &global_nwo); + arch_atomic_add(*no_way_out, &global_nwo); /* * Rely on the implied barrier below, such that global_nwo * is updated before mce_callin. */ - order = atomic_inc_return(&mce_callin); - cpumask_clear_cpu(smp_processor_id(), &mce_missing_cpus); + order = arch_atomic_inc_return(&mce_callin); + arch_cpumask_clear_cpu(smp_processor_id(), &mce_missing_cpus); /* Enable instrumentation around calls to external facilities */ instrumentation_begin(); @@ -1029,10 +1009,10 @@ static noinstr int mce_start(int *no_way_out) /* * Wait for everyone. */ - while (atomic_read(&mce_callin) != num_online_cpus()) { + while (arch_atomic_read(&mce_callin) != num_online_cpus()) { if (mce_timed_out(&timeout, "Timeout: Not all CPUs entered broadcast exception handler")) { - atomic_set(&global_nwo, 0); + arch_atomic_set(&global_nwo, 0); goto out; } ndelay(SPINUNIT); @@ -1047,7 +1027,7 @@ static noinstr int mce_start(int *no_way_out) /* * Monarch: Starts executing now, the others wait. */ - atomic_set(&mce_executing, 1); + arch_atomic_set(&mce_executing, 1); } else { /* * Subject: Now start the scanning loop one by one in @@ -1055,10 +1035,10 @@ static noinstr int mce_start(int *no_way_out) * This way when there are any shared banks it will be * only seen by one CPU before cleared, avoiding duplicates. */ - while (atomic_read(&mce_executing) < order) { + while (arch_atomic_read(&mce_executing) < order) { if (mce_timed_out(&timeout, "Timeout: Subject CPUs unable to finish machine check processing")) { - atomic_set(&global_nwo, 0); + arch_atomic_set(&global_nwo, 0); goto out; } ndelay(SPINUNIT); @@ -1068,7 +1048,7 @@ static noinstr int mce_start(int *no_way_out) /* * Cache the global no_way_out state. */ - *no_way_out = atomic_read(&global_nwo); + *no_way_out = arch_atomic_read(&global_nwo); ret = order; @@ -1153,12 +1133,12 @@ out: return ret; } -static void mce_clear_state(unsigned long *toclear) +static __always_inline void mce_clear_state(unsigned long *toclear) { int i; for (i = 0; i < this_cpu_read(mce_num_banks); i++) { - if (test_bit(i, toclear)) + if (arch_test_bit(i, toclear)) mce_wrmsrl(mca_msr_reg(i, MCA_STATUS), 0); } } @@ -1208,8 +1188,8 @@ __mc_scan_banks(struct mce *m, struct pt_regs *regs, struct mce *final, int severity, i, taint = 0; for (i = 0; i < this_cpu_read(mce_num_banks); i++) { - __clear_bit(i, toclear); - if (!test_bit(i, valid_banks)) + arch___clear_bit(i, toclear); + if (!arch_test_bit(i, valid_banks)) continue; if (!mce_banks[i].ctl) @@ -1244,7 +1224,7 @@ __mc_scan_banks(struct mce *m, struct pt_regs *regs, struct mce *final, severity == MCE_UCNA_SEVERITY) && !no_way_out) continue; - __set_bit(i, toclear); + arch___set_bit(i, toclear); /* Machine check event was not enabled. Clear, but ignore. */ if (severity == MCE_NO_SEVERITY) diff --git a/arch/x86/kernel/cpu/mce/internal.h b/arch/x86/kernel/cpu/mce/internal.h index 52c6339..a04b61e 100644 --- a/arch/x86/kernel/cpu/mce/internal.h +++ b/arch/x86/kernel/cpu/mce/internal.h @@ -182,8 +182,6 @@ enum mca_msr { MCA_MISC, }; -u32 mca_msr_reg(int bank, enum mca_msr reg); - /* Decide whether to add MCE record to MCE event pool or filter it out. */ extern bool filter_mce(struct mce *m); @@ -209,4 +207,25 @@ static inline void winchip_machine_check(struct pt_regs *regs) {} noinstr u64 mce_rdmsrl(u32 msr); +static __always_inline u32 mca_msr_reg(int bank, enum mca_msr reg) +{ + if (mce_flags.smca) { + switch (reg) { + case MCA_CTL: return MSR_AMD64_SMCA_MCx_CTL(bank); + case MCA_ADDR: return MSR_AMD64_SMCA_MCx_ADDR(bank); + case MCA_MISC: return MSR_AMD64_SMCA_MCx_MISC(bank); + case MCA_STATUS: return MSR_AMD64_SMCA_MCx_STATUS(bank); + } + } + + switch (reg) { + case MCA_CTL: return MSR_IA32_MCx_CTL(bank); + case MCA_ADDR: return MSR_IA32_MCx_ADDR(bank); + case MCA_MISC: return MSR_IA32_MCx_MISC(bank); + case MCA_STATUS: return MSR_IA32_MCx_STATUS(bank); + } + + return 0; +} + #endif /* __X86_MCE_INTERNAL_H__ */ diff --git a/arch/x86/kernel/cpu/mce/severity.c b/arch/x86/kernel/cpu/mce/severity.c index 7aa2bda..ca0d775 100644 --- a/arch/x86/kernel/cpu/mce/severity.c +++ b/arch/x86/kernel/cpu/mce/severity.c @@ -301,7 +301,7 @@ static noinstr int error_context(struct mce *m, struct pt_regs *regs) } } -static int mce_severity_amd_smca(struct mce *m, enum context err_ctx) +static __always_inline int mce_severity_amd_smca(struct mce *m, enum context err_ctx) { u64 mcx_cfg; ^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH 0/3] x86/mce: Fix more noinstr fun 2022-02-04 8:30 [PATCH 0/3] x86/mce: Fix more noinstr fun Borislav Petkov ` (2 preceding siblings ...) 2022-02-04 8:30 ` [PATCH 3/3] x86/mce: Use arch atomic and bit helpers Borislav Petkov @ 2022-02-04 11:24 ` Marco Elver 2022-02-04 14:48 ` Borislav Petkov 3 siblings, 1 reply; 11+ messages in thread From: Marco Elver @ 2022-02-04 11:24 UTC (permalink / raw) To: Borislav Petkov; +Cc: Tony Luck, Jakub Kicinski, Linus Torvalds, X86 ML, LKML On Fri, 4 Feb 2022 at 09:30, Borislav Petkov <bp@alien8.de> wrote: > > From: Borislav Petkov <bp@suse.de> > > Hi all, > > here's a second small set of fixes for objtool noinstr validation issues > in the MCE code. > > It goes ontop of tip:locking/core where there are some more commits > changing generic helpers to be always inlined. > > As always, comments and suggestions are appreciated. > > Thx. > > Borislav Petkov (3): > cpumask: Add a x86-specific cpumask_clear_cpu() helper > x86/ptrace: Always inline v8086_mode() for instrumentation > x86/mce: Use arch atomic and bit helpers I guess to solve noinstr "calling" something else the only solution is to make the other function noinstr as well, or __always_inline. Have you considered making some of these other functions 'noinstr' as well? I guess __always_inline works, esp. if there's just 1 caller. And by the looks of it you're getting a net .text reduction, so Acked-by: Marco Elver <elver@google.com> ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 0/3] x86/mce: Fix more noinstr fun 2022-02-04 11:24 ` [PATCH 0/3] x86/mce: Fix more noinstr fun Marco Elver @ 2022-02-04 14:48 ` Borislav Petkov 0 siblings, 0 replies; 11+ messages in thread From: Borislav Petkov @ 2022-02-04 14:48 UTC (permalink / raw) To: Marco Elver; +Cc: Tony Luck, Jakub Kicinski, Linus Torvalds, X86 ML, LKML On Fri, Feb 04, 2022 at 12:24:00PM +0100, Marco Elver wrote: > I guess to solve noinstr "calling" something else the only solution is > to make the other function noinstr as well, or __always_inline. > > Have you considered making some of these other functions 'noinstr' as > well? I guess __always_inline works, esp. if there's just 1 caller. > And by the looks of it you're getting a net .text reduction, so Yeah, I started doing that and the savings were the persuasive argument. Even more so if the function has one caller only and gets inlined, normally. I guess it doesn't get inlined when there's KASAN instrumentation or so but I haven't verified it fully why. Because even for oneliners like v8086_mode() which should get inlined trivially, the compiler ends up doing this constprop thing - I guess some constants folding optimization thing... it probably doesn't even matter so much whether oneliners get inlined in KASAN-enabled builds so I guess we might just as well force-inline them for the other configs. > Acked-by: Marco Elver <elver@google.com> Thx. -- Regards/Gruss, Boris. https://people.kernel.org/tglx/notes-about-netiquette ^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2022-02-14 10:39 UTC | newest] Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2022-02-04 8:30 [PATCH 0/3] x86/mce: Fix more noinstr fun Borislav Petkov 2022-02-04 8:30 ` [PATCH 1/3] cpumask: Add a x86-specific cpumask_clear_cpu() helper Borislav Petkov 2022-02-04 16:46 ` Luck, Tony 2022-02-04 18:04 ` Borislav Petkov 2022-02-13 10:59 ` [tip: locking/core] " tip-bot2 for Borislav Petkov 2022-02-04 8:30 ` [PATCH 2/3] x86/ptrace: Always inline v8086_mode() for instrumentation Borislav Petkov 2022-02-13 10:59 ` [tip: locking/core] " tip-bot2 for Borislav Petkov 2022-02-04 8:30 ` [PATCH 3/3] x86/mce: Use arch atomic and bit helpers Borislav Petkov 2022-02-14 10:03 ` [tip: ras/core] " tip-bot2 for Borislav Petkov 2022-02-04 11:24 ` [PATCH 0/3] x86/mce: Fix more noinstr fun Marco Elver 2022-02-04 14:48 ` Borislav Petkov
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.