x86/microcode: Cleanup and late loading enhancements
Subsystems (6)
v3 cover x86/microcode: Cleanup and late loading enhancements
This is a follow up on:
https://lore.kernel.org/lkml/[email protected]
Late microcode loading is desired by enterprise users. Late loading is problematic as it requires detailed knowledge about the change and an analysis whether this change modifies something which is already in use by the kernel. Large enterprise customers have engineering teams and access to deep technical vendor support. The regular admin does not have such resources, so the kernel has always tainted the kernel after late loading.
Intel recently added a new previously reserved field to the microcode header which contains the minimal microcode revision which must be running on the CPU to make the load safe. This field is 0 in all older microcode revisions, which the kernel assumes to be unsafe. Minimal revision checking can be enforced via Kconfig or kernel command line. It then refuses to load an unsafe revision. The default loads unsafe revisions like before and taints the kernel. If a safe revision is loaded the kernel is not tainted.
But that does not solve all other known problems with late loading:
- Late loading on current Intel CPUs is unsafe vs. NMI when hyperthreading is enabled. If a NMI hits the secondary sibling while the primary loads the microcode, the machine can crash.
- Soft offline SMT siblings which are playing dead with MWAIT can cause damage too when the microcode update modifies MWAIT. That's a realistic scenario in the context of 'nosmt' mitigations. :(
Neither the core code nor the Intel specific code handles any of this at all.
While trying to implement this, I stumbled over disfunctional, horribly complex and redundant code, which I decided to clean up first so the new functionality can be added on a clean slate.
So the series has several sections:
1) Move the 32bit early loading after paging enable
2) Cleanup of the Intel specific code
3) Implementation of proper core control logic to handle the NMI safe requirements
4) Support for minimal revision check in the core and the Intel specific parts.
Changes vs. V2:
- Rebased on v6.5-rc1
- Removed the 32bit oddity of invoking the microcode loader before paging is enabled.
- Some minor improvements.
The series is also available from git:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.git ucode-v3
Thanks,
tglx
--- Documentation/admin-guide/kernel-parameters.txt | 5 arch/x86/Kconfig | 25 arch/x86/include/asm/apic.h | 5 arch/x86/include/asm/cpu.h | 20 arch/x86/include/asm/microcode.h | 16 arch/x86/kernel/Makefile | 1 arch/x86/kernel/apic/apic_flat_64.c | 2 arch/x86/kernel/apic/ipi.c | 8 arch/x86/kernel/apic/x2apic_cluster.c | 1 arch/x86/kernel/apic/x2apic_phys.c | 1 arch/x86/kernel/cpu/common.c | 12 arch/x86/kernel/cpu/microcode/amd.c | 57 -- arch/x86/kernel/cpu/microcode/core.c | 643 +++++++++++++++-------- arch/x86/kernel/cpu/microcode/intel.c | 659 ++++++------------------ arch/x86/kernel/cpu/microcode/internal.h | 32 - arch/x86/kernel/head32.c | 6 arch/x86/kernel/head_32.S | 10 arch/x86/kernel/nmi.c | 9 arch/x86/kernel/smpboot.c | 12 drivers/platform/x86/intel/ifs/load.c | 8 include/linux/cpuhotplug.h | 1 21 files changed, 740 insertions(+), 793 deletions(-)
PATCH v3 1/30 x86/microcode/32: Move early loading after paging enable
32-bit loads microcode before paging is enabled. The commit which introduced that has zero justification in the changelog. The cover letter has slightly more content, but it does not give any technical justification either:
"The problem in current microcode loading method is that we load a
microcode way, way too late; ideally we should load it before turning paging on. This may only be practical on 32 bits since we can't get to 64-bit mode without paging on, but we should still do it as early as at all possible."
Handwaving word salad with zero technical content.
Someone claimed in an offlist conversation that this is required for curing the ATOM erratum AAE44/AAF40/AAG38/AAH41. That erratum requires an microcode update in order to make the usage of PSE safe. But during early boot PSE is completely irrelevant and it is evaluated way later.
Neither is it relevant for the AP on single core HT enabled CPUs as the microcode loading on the AP is not doing anything.
On dual core CPUs there is a theoretical problem if a split of an executable large page between enabling paging including PSE and loading the microcode happens. But that's only theoretical, it's practically irrelevant because the affected dual core CPUs are 64bit enabled and therefore have paging and PSE enabled before loading the microcode on the second core. So why would it work on 64-bit but not on 32-bit?
The erratum:
"AAG38 Code Fetch May Occur to Incorrect Address After a Large Page is
Split Into 4-Kbyte Pages
Problem: If software clears the PS (page size) bit in a present PDE (page directory entry), that will cause linear addresses mapped through this PDE to use 4-KByte pages instead of using a large page after old TLB entries are invalidated. Due to this erratum, if a code fetch uses this PDE before the TLB entry for the large page is invalidated then it may fetch from a different physical address than specified by either the old large page translation or the new 4-KByte page translation. This erratum may also cause speculative code fetches from incorrect addresses."
The practical relevance for this is exactly zero because there is no splitting of large text pages during early boot-time, i.e. between paging enable and microcode loading, and neither during CPU hotplug.
IOW, this load microcode before paging enable is yet another voodoo programming solution in search of a problem. What's worse is that it causes at least two serious problems:
1) When stackprotector is enabled then the microcode loader code has the stackprotector mechanics enabled. The read from the per CPU variable __stack_chk_guard is always accessing the virtual address either directly on UP or via FS on SMP. In physical address mode this results in an access to memory above 3GB. So this works by chance as the hardware returns the same value when there is no RAM at this physical address. When there is RAM populated above 3G then the read is by chance the same as nothing changes that memory during the very early boot stage. That's not necessarily true during runtime CPU hotplug.
2) When function tracing is enabled, then the relevant microcode loader functions and the functions invoked from there will call into the tracing code and evaluate global and per CPU variables in physical address mode. What could potentially go wrong?
Cure this and move the microcode loading after the early paging enable and remove the gunk in the microcode loader which is required to handle physical address mode.
--- arch/x86/include/asm/microcode.h | 5 - arch/x86/kernel/cpu/common.c | 12 --- arch/x86/kernel/cpu/microcode/amd.c | 31 +------- arch/x86/kernel/cpu/microcode/core.c | 73 ++++---------------- arch/x86/kernel/cpu/microcode/intel.c | 108 +++---------------------------- arch/x86/kernel/cpu/microcode/internal.h | 2 arch/x86/kernel/head32.c | 3 arch/x86/kernel/head_32.S | 10 -- arch/x86/kernel/smpboot.c | 12 +-- 9 files changed, 41 insertions(+), 215 deletions(-)
arch/x86/include/asm/microcode.h +0 -5
@@ -68,11 +68,6 @@ static inline u32 intel_get_microcode_rereturn rev; } - -void show_ucode_info_early (void ); - -#else /* CONFIG_CPU_SUP_INTEL */ -static inline void show_ucode_info_early (void ) { }#endif /* !CONFIG_CPU_SUP_INTEL */ #endif /* _ASM_X86_MICROCODE_H */
arch/x86/kernel/cpu/common.c +0 -12
@@ -2123,8 +2123,6 @@ static inline void setup_getcpu(int cpu) }#ifdef CONFIG_X86_64 -static inline void ucode_cpu_init (int cpu) { } -static inline void tss_setup_ist (struct tss_struct *tss) {/* Set up the per-CPU TSS IST stacks */
@@ -2135,16 +2133,8 @@ static inline void tss_setup_ist(struct/* Only mapped when SEV-ES is active */ tss->x86_tss.ist[IST_INDEX_VC] = __this_cpu_ist_top_va(VC); } -#else /* CONFIG_X86_64 */ - -static inline void ucode_cpu_init (int cpu) -{ -show_ucode_info_early (); -} -static inline void tss_setup_ist (struct tss_struct *tss) { } -#endif /* !CONFIG_X86_64 */ static inline void tss_setup_io_bitmap (struct tss_struct *tss)
@@ -2200,8 +2190,6 @@ void cpu_init(void)struct task_struct *cur = current;int cpu =raw_smp_processor_id (); -ucode_cpu_init (cpu); -#ifdef CONFIG_NUMA if (this_cpu_read (numa_node) ==0 &&early_cpu_to_node (cpu) !=NUMA_NO_NODE )
arch/x86/kernel/cpu/microcode/amd.c +6 -25
@@ -456,14 +456,8 @@ static bool early_apply_microcode(u32 cp {struct cont_desc desc = {0 };struct microcode_amd *mc; -u32 rev, dummy, *new_rev;bool ret = false; - -#ifdef CONFIG_X86_32 - new_rev = (u32 *)__pa_nodebug (&ucode_new_rev); -#else - new_rev = &ucode_new_rev; -#endif +u32 rev, dummy; desc.cpuid_1_eax = cpuid_1_eax;
@@ -484,8 +478,8 @@ static bool early_apply_microcode(u32 cpreturn ret;if (!__apply_microcode_amd (mc)) { - *new_rev = mc->hdr.patch_id; - ret = true; + ucode_new_rev = mc->hdr.patch_id; + ret = true; }return ret;
@@ -514,26 +508,13 @@ static bool get_builtin_microcode(structstatic void find_blobs_in_containers (unsigned int cpuid_1_eax,struct cpio_data *ret) { -struct ucode_cpu_info *uci;struct cpio_data cp; -const char *path; -bool use_pa; - -if (IS_ENABLED (CONFIG_X86_32 )) { - uci = (struct ucode_cpu_info *)__pa_nodebug (ucode_cpu_info); - path = (const char *)__pa_nodebug (ucode_path); - use_pa = true; - }else { - uci = ucode_cpu_info; - path = ucode_path; - use_pa = false; - }if (!get_builtin_microcode (&cp,x86_family (cpuid_1_eax))) - cp =find_microcode_in_initrd (path, use_pa); + cp =find_microcode_in_initrd (ucode_path);/* Needed in load_microcode_amd() */ - uci->cpu_sig.sig = cpuid_1_eax; + ucode_cpu_info->cpu_sig.sig = cpuid_1_eax; *ret = cp; }
@@ -562,7 +543,7 @@ int __init save_microcode_in_initrd_amd(enum ucode_state ret;struct cpio_data cp; - cp =find_microcode_in_initrd (ucode_path, false); + cp =find_microcode_in_initrd (ucode_path);if (!(cp.data && cp.size))return -EINVAL ;
arch/x86/kernel/cpu/microcode/core.c +15 -58
142 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/intel.c +12 -96
156 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/internal.h +1 -1
@@ -44,7 +44,7 @@ struct microcode_ops { };extern struct ucode_cpu_info ucode_cpu_info[]; -struct cpio_data find_microcode_in_initrd (const char *path,bool use_pa); +struct cpio_data find_microcode_in_initrd (const char *path);#define MAX_UCODE_COUNT 128
arch/x86/kernel/head32.c +3 -0
@@ -19,6 +19,7 @@#include <asm/apic.h> #include <asm/io_apic.h> #include <asm/bios_ebda.h> +#include <asm/microcode.h> #include <asm/tlbflush.h> #include <asm/bootparam_utils.h>
@@ -34,6 +35,8 @@ asmlinkage __visible void __init __noret/* Make sure IDT is set up before any exception happens */ idt_setup_early_handler (); +load_ucode_bsp (); +cr4_init_shadow ();sanitize_boot_params (&boot_params);
arch/x86/kernel/head_32.S +0 -10
@@ -118,11 +118,6 @@ SYM_CODE_START(startup_32) movl %eax, pa(olpc_ofw_pgd) #endif -#ifdef CONFIG_MICROCODE - /* Early load ucode on BSP. */ - call load_ucode_bsp -#endif - /* Create early pagetables. */ call mk_early_pgtbl_32
@@ -157,11 +152,6 @@ SYM_FUNC_START(startup_32_smp) movl %eax,%ss leal -__PAGE_OFFSET(%ecx),%esp -#ifdef CONFIG_MICROCODE - /* Early load ucode on AP. */ - call load_ucode_ap -#endif - .Ldefault_entry: movl $(CR0_STATE & ~X86_CR0_PG),%eax movl %eax,%cr0
arch/x86/kernel/smpboot.c +4 -8
@@ -258,12 +258,9 @@ static void notrace start_secondary(voidcpu_init_exception_handling ();/* -* 32-bit systems load the microcode from the ASM startup code for -* historical reasons. -* -* On 64-bit systems load it before reaching the AP alive -* synchronization point below so it is not part of the full per -* CPU serialized bringup part when "parallel" bringup is enabled. +* Load the microcode before reaching the AP alive synchronization +* point below so it is not part of the full per CPU serialized +* bringup part when "parallel" bringup is enabled. * * That's even safe when hyperthreading is enabled in the CPU as * the core code starts the primary threads first and leaves the
@@ -276,8 +273,7 @@ static void notrace start_secondary(void* CPUID, MSRs etc. must be strictly serialized to maintain * software state correctness. */ -if (IS_ENABLED (CONFIG_X86_64 )) -load_ucode_ap (); +load_ucode_ap ();/* * Synchronization point with the hotplug core. Sets this CPUs
Borislav Petkov
maintainerRe: [patch V3 01/30] x86/microcode/32: Move early loading after paging enable
Quoted 17 lines · Thomas Gleixner
32-bit loads microcode before paging is enabled. The commit which introduced that has zero justification in the changelog. The cover letter has slightly more content, but it does not give any technical justification either:
"The problem in current microcode loading method is that we load a
microcode way, way too late; ideally we should load it before turning paging on. This may only be practical on 32 bits since we can't get to 64-bit mode without paging on, but we should still do it as early as at all possible."Handwaving word salad with zero technical content.
Someone claimed in an offlist conversation that this is required for curing the ATOM erratum AAE44/AAF40/AAG38/AAH41. That erratum requires an microcode update in order to make the usage of PSE safe. But during early boot PSE is completely irrelevant and it is evaluated way later.
...
Quoted 3 lines
Cure this and move the microcode loading after the early paging enable and remove the gunk in the microcode loader which is required to handle physical address mode.
That same someone from the offlist conversation loves this change - 32-bit PA loading gunk has been a major PITA ever since it got added - just look at the code you're removing and cringe.
Quoted 3 lines
Signed-off-by Thomas Gleixner <[email protected]>Cc Fenghua Yu <[email protected]>Cc Peter Anvin <[email protected]>
I guess those gents could chime in now - lemme put them in To:, perhaps they'll look at this then.
If not, I'd say we do this. The erratum in question is for pretty much dying hw anyway - Atom n270 of which I actually have *two* machines even. I can "accidentally" drop them on the concrete floor while carrying and problem solved => dying hardware.
:-P
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 01/30] x86/microcode/32: Move early loading after paging enable
In reply to Thomas GleixnerQuoted 3 lines · Thomas Gleixner
static int apply_microcode_early(struct ucode_cpu_info *uci, bool early)
{...
arch/x86/kernel/cpu/microcode/intel.c static int apply_microcode_early(struct · line 520
Quoted 5 lines
arch/x86/kernel/cpu/microcode/intel.c +0
-5-if (early) -print_ucode (old_rev, uci->cpu_sig.rev, mc->hdr.date); -else -print_ucode_info (old_rev, uci->cpu_sig.rev, mc->hdr.date); -
Nitpick:
with this change, there is no need for the second argument -- 'bool early'.
Thanks,
Chang
Re: [patch V3 01/30] x86/microcode/32: Move early loading after paging enable
Quoted 13 lines · Chang S. Bae
Quoted 3 lines · Thomas Gleixner
static int apply_microcode_early(struct ucode_cpu_info *uci, bool early) {...
arch/x86/kernel/cpu/microcode/intel.c static int apply_microcode_early(struct · line 520
Quoted 5 lines
arch/x86/kernel/cpu/microcode/intel.c +0
-5-if (early) -print_ucode (old_rev, uci->cpu_sig.rev, mc->hdr.date); -else -print_ucode_info (old_rev, uci->cpu_sig.rev, mc->hdr.date); -Nitpick:
Quoted 1 line
with this change, there is no need for the second argument -- 'bool early'.
Duh. Yes ...
PATCH v3 1/30 x86/microcode/32: Move early loading after paging enable
32-bit loads microcode before paging is enabled. The commit which introduced that has zero justification in the changelog. The cover letter has slightly more content, but it does not give any technical justification either:
"The problem in current microcode loading method is that we load a
microcode way, way too late; ideally we should load it before turning paging on. This may only be practical on 32 bits since we can't get to 64-bit mode without paging on, but we should still do it as early as at all possible."
Handwaving word salad with zero technical content.
Someone claimed in an offlist conversation that this is required for curing the ATOM erratum AAE44/AAF40/AAG38/AAH41. That erratum requires an microcode update in order to make the usage of PSE safe. But during early boot PSE is completely irrelevant and it is evaluated way later.
Neither is it relevant for the AP on single core HT enabled CPUs as the microcode loading on the AP is not doing anything.
On dual core CPUs there is a theoretical problem if a split of an executable large page between enabling paging including PSE and loading the microcode happens. But that's only theoretical, it's practically irrelevant because the affected dual core CPUs are 64bit enabled and therefore have paging and PSE enabled before loading the microcode on the second core. So why would it work on 64-bit but not on 32-bit?
The erratum:
"AAG38 Code Fetch May Occur to Incorrect Address After a Large Page is
Split Into 4-Kbyte Pages
Problem: If software clears the PS (page size) bit in a present PDE (page directory entry), that will cause linear addresses mapped through this PDE to use 4-KByte pages instead of using a large page after old TLB entries are invalidated. Due to this erratum, if a code fetch uses this PDE before the TLB entry for the large page is invalidated then it may fetch from a different physical address than specified by either the old large page translation or the new 4-KByte page translation. This erratum may also cause speculative code fetches from incorrect addresses."
The practical relevance for this is exactly zero because there is no splitting of large text pages during early boot-time, i.e. between paging enable and microcode loading, and neither during CPU hotplug.
IOW, this load microcode before paging enable is yet another voodoo programming solution in search of a problem. What's worse is that it causes at least two serious problems:
1) When stackprotector is enabled then the microcode loader code has the stackprotector mechanics enabled. The read from the per CPU variable __stack_chk_guard is always accessing the virtual address either directly on UP or via FS on SMP. In physical address mode this results in an access to memory above 3GB. So this works by chance as the hardware returns the same value when there is no RAM at this physical address. When there is RAM populated above 3G then the read is by chance the same as nothing changes that memory during the very early boot stage. That's not necessarily true during runtime CPU hotplug.
2) When function tracing is enabled, then the relevant microcode loader functions and the functions invoked from there will call into the tracing code and evaluate global and per CPU variables in physical address mode. What could potentially go wrong?
Cure this and move the microcode loading after the early paging enable and remove the gunk in the microcode loader which is required to handle physical address mode.
---
V3a: Remove the early arguments - Chang
Fixup the core code - 0day
--- arch/x86/include/asm/microcode.h | 5 - arch/x86/kernel/cpu/common.c | 12 --- arch/x86/kernel/cpu/microcode/amd.c | 103 ++++++++------------------- arch/x86/kernel/cpu/microcode/core.c | 73 ++++--------------- arch/x86/kernel/cpu/microcode/intel.c | 116 ++++--------------------------- arch/x86/kernel/cpu/microcode/internal.h | 2 arch/x86/kernel/head32.c | 3 arch/x86/kernel/head_32.S | 10 -- arch/x86/kernel/smpboot.c | 12 +-- 9 files changed, 71 insertions(+), 265 deletions(-)
arch/x86/include/asm/microcode.h +0 -5
@@ -68,11 +68,6 @@ static inline u32 intel_get_microcode_rereturn rev; } - -void show_ucode_info_early (void ); - -#else /* CONFIG_CPU_SUP_INTEL */ -static inline void show_ucode_info_early (void ) { }#endif /* !CONFIG_CPU_SUP_INTEL */ #endif /* _ASM_X86_MICROCODE_H */
arch/x86/kernel/cpu/common.c +0 -12
@@ -2166,8 +2166,6 @@ static inline void setup_getcpu(int cpu) }#ifdef CONFIG_X86_64 -static inline void ucode_cpu_init (int cpu) { } -static inline void tss_setup_ist (struct tss_struct *tss) {/* Set up the per-CPU TSS IST stacks */
@@ -2178,16 +2176,8 @@ static inline void tss_setup_ist(struct/* Only mapped when SEV-ES is active */ tss->x86_tss.ist[IST_INDEX_VC] = __this_cpu_ist_top_va(VC); } -#else /* CONFIG_X86_64 */ - -static inline void ucode_cpu_init (int cpu) -{ -show_ucode_info_early (); -} -static inline void tss_setup_ist (struct tss_struct *tss) { } -#endif /* !CONFIG_X86_64 */ static inline void tss_setup_io_bitmap (struct tss_struct *tss)
@@ -2243,8 +2233,6 @@ void cpu_init(void)struct task_struct *cur = current;int cpu =raw_smp_processor_id (); -ucode_cpu_init (cpu); -#ifdef CONFIG_NUMA if (this_cpu_read (numa_node) ==0 &&early_cpu_to_node (cpu) !=NUMA_NO_NODE )
arch/x86/kernel/cpu/microcode/amd.c +32 -71
255 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/core.c +15 -58
142 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/intel.c +16 -100
173 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/internal.h +1 -1
@@ -44,7 +44,7 @@ struct microcode_ops { };extern struct ucode_cpu_info ucode_cpu_info[]; -struct cpio_data find_microcode_in_initrd (const char *path,bool use_pa); +struct cpio_data find_microcode_in_initrd (const char *path);#define MAX_UCODE_COUNT 128
arch/x86/kernel/head32.c +3 -0
@@ -19,6 +19,7 @@#include <asm/apic.h> #include <asm/io_apic.h> #include <asm/bios_ebda.h> +#include <asm/microcode.h> #include <asm/tlbflush.h> #include <asm/bootparam_utils.h>
@@ -34,6 +35,8 @@ asmlinkage __visible void __init __noret/* Make sure IDT is set up before any exception happens */ idt_setup_early_handler (); +load_ucode_bsp (); +cr4_init_shadow ();sanitize_boot_params (&boot_params);
arch/x86/kernel/head_32.S +0 -10
@@ -118,11 +118,6 @@ SYM_CODE_START(startup_32) movl %eax, pa(olpc_ofw_pgd) #endif -#ifdef CONFIG_MICROCODE - /* Early load ucode on BSP. */ - call load_ucode_bsp -#endif - /* Create early pagetables. */ call mk_early_pgtbl_32
@@ -157,11 +152,6 @@ SYM_FUNC_START(startup_32_smp) movl %eax,%ss leal -__PAGE_OFFSET(%ecx),%esp -#ifdef CONFIG_MICROCODE - /* Early load ucode on AP. */ - call load_ucode_ap -#endif - .Ldefault_entry: movl $(CR0_STATE & ~X86_CR0_PG),%eax movl %eax,%cr0
arch/x86/kernel/smpboot.c +4 -8
@@ -258,12 +258,9 @@ static void notrace start_secondary(voidcpu_init_exception_handling ();/* -* 32-bit systems load the microcode from the ASM startup code for -* historical reasons. -* -* On 64-bit systems load it before reaching the AP alive -* synchronization point below so it is not part of the full per -* CPU serialized bringup part when "parallel" bringup is enabled. +* Load the microcode before reaching the AP alive synchronization +* point below so it is not part of the full per CPU serialized +* bringup part when "parallel" bringup is enabled. * * That's even safe when hyperthreading is enabled in the CPU as * the core code starts the primary threads first and leaves the
@@ -276,8 +273,7 @@ static void notrace start_secondary(void* CPUID, MSRs etc. must be strictly serialized to maintain * software state correctness. */ -if (IS_ENABLED (CONFIG_X86_64 )) -load_ucode_ap (); +load_ucode_ap ();/* * Synchronization point with the hotplug core. Sets this CPUs
PATCH v3 2/30 x86/boot/32: Disable stackprotector and tracing for mk_early_pgtbl_32()
In reply to Thomas GleixnerStackprotector cannot work before paging is enabled. The read from the per CPU variable __stack_chk_guard is always accessing the virtual address either directly on UP or via FS on SMP. In physical address mode this results in an access to memory above 3GB.
So this works by chance as the hardware returns the same value when there is no RAM at this physical address. When there is RAM populated above 3G then the read is by chance the same as nothing changes that memory during the very early boot stage.
Stop relying on pure luck and disable the stack protector for the only C function which is called during early boot before paging is enabled.
Remove function tracing from the whole source file as there is no way to trace this at all, but in case of CONFIG_DYNAMIC_FTRACE=n mk_early_pgtbl_32() would access global function tracer variables in physcial address mode which again might work by chance.
--- arch/x86/kernel/Makefile | 1 + arch/x86/kernel/head32.c | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-)
arch/x86/kernel/Makefile +1 -0
@@ -16,6 +16,7 @@ CFLAGS_REMOVE_kvmclock.o = -pg CFLAGS_REMOVE_ftrace.o = -pg CFLAGS_REMOVE_early_printk.o = -pg CFLAGS_REMOVE_head64.o = -pg +CFLAGS_REMOVE_head32.o = -pg CFLAGS_REMOVE_sev.o = -pg CFLAGS_REMOVE_rethook.o = -pg endif
arch/x86/kernel/head32.c +2 -1
@@ -73,7 +73,8 @@ asmlinkage __visible void __init __noret* always zero at this stage. */ void __init mk_early_pgtbl_32 (void ); -void __init mk_early_pgtbl_32 (void ) + +void __init __no_stack_protector mk_early_pgtbl_32(void) {#ifdef __pa#undef __pa
Borislav Petkov
maintainerRe: [patch V3 02/30] x86/boot/32: Disable stackprotector and tracing for mk_early_pgtbl_32()
Quoted 4 lines · Thomas Gleixner
Remove function tracing from the whole source file as there is no way to trace this at all, but in case of
CONFIG_DYNAMIC_FTRACE=nmk_early_pgtbl_32()would access global function tracer variables in physcial address mode which again might work by chance.
Unknown word [physcial] in commit message.
Suggestions: ['physical'
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 3/30 x86/microcode/intel: Rip out mixed stepping support for Intel CPUs
In reply to Thomas GleixnerFrom: Ashok Raj <[email protected]>
Mixed steppings aren't supported on Intel CPUs. Only one patch is required for the entire system. The caching of micro code blobs which match the family and model is therefore pointless and in fact it is disfunctional as CPU hotplug updates use only a single microcode blob, i.e. the one where *intel_ucode_patch points to.
Remove the microcode cache and make it an AMD local feature.
[ tglx: Save only at the end. Otherwise random microcode ends up in the
pointer for early loading ]
---
V2: Fix the bogus condition - Borislav
--- arch/x86/kernel/cpu/microcode/amd.c | 10 ++ arch/x86/kernel/cpu/microcode/core.c | 2 arch/x86/kernel/cpu/microcode/intel.c | 133 +++++-------------------------- arch/x86/kernel/cpu/microcode/internal.h | 10 -- 4 files changed, 35 insertions(+), 120 deletions(-)
---
arch/x86/kernel/cpu/microcode/amd.c +10 -0
@@ -37,6 +37,16 @@#include "internal.h" +struct ucode_patch { +struct list_head plist; +void *data; +unsigned int size; +u32 patch_id; +u16 equiv_cpu; +}; + +static LIST_HEAD (microcode_cache); +#define UCODE_MAGIC 0x00414d44#define UCODE_EQUIV_CPU_TABLE_TYPE 0x00000000#define UCODE_UCODE_TYPE 0x00000001
arch/x86/kernel/cpu/microcode/core.c +0 -2
@@ -46,8 +46,6 @@ static bool dis_ucode_ldr = true;bool initrd_gone; -LIST_HEAD (microcode_cache); -/* * Synchronization. *
arch/x86/kernel/cpu/microcode/intel.c +25 -108
207 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/internal.h +0 -10
@@ -8,16 +8,6 @@#include <asm/cpu.h> #include <asm/microcode.h> -struct ucode_patch { -struct list_head plist; -void *data;/* Intel uses only this one */ -unsigned int size; -u32 patch_id; -u16 equiv_cpu; -}; - -extern struct list_head microcode_cache; -struct device ;enum ucode_state {
Re: [patch V3 03/30] x86/microcode/intel: Rip out mixed stepping support for Intel CPUs
Quoted 4 lines
...
Mixed steppings aren't supported on Intel CPUs. Only one patch is required for the entire system. The caching of micro code blobs which match the
s/micro code/microcode/
Quoted 1 line
family and model is therefore pointless and in fact it is disfunctional as
s/disfunctional/dysfunctional/
Quoted 6 lines
CPU hotplug updates use only a single microcode blob, i.e. the one where
*intel_ucode_patchpoints to.Remove the microcode cache and make it an AMD local feature.
...
PATCH v3 4/30 x86/microcode/intel: Simplify scan_microcode()
In reply to Thomas GleixnerMake it readable and comprehensible.
--- arch/x86/kernel/cpu/microcode/intel.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)
arch/x86/kernel/cpu/microcode/intel.c +8 -22
@@ -265,25 +265,19 @@ static void save_microcode_patch(void *dreturn ;/* Save for early loading */ - intel_ucode_patch = (struct microcode_intel *)p; + intel_ucode_patch = (struct microcode_intel *)p; } -/* -* Get microcode matching with BSP's model. Only CPUs with the same model as -* BSP can stay in the platform. -*/ -static struct microcode_intel * -scan_microcode (void *data,size_t size,struct ucode_cpu_info *uci,bool save) +/* Scan CPIO for microcode matching the boot CPUs family, model, stepping */ +static struct microcode_intel *scan_microcode (void *data,size_t size, +struct ucode_cpu_info *uci,bool save) {struct microcode_header_intel *mc_header;struct microcode_intel *patch =NULL ;u32 cur_rev = uci->cpu_sig.rev;unsigned int mc_size; -while (size) { -if (size <sizeof (struct microcode_header_intel )) -break ; - +for (; size >=sizeof (struct microcode_header_intel ); size -= mc_size, data += mc_size) { mc_header = (struct microcode_header_intel *)data; mc_size =get_totalsize (mc_header);
@@ -291,27 +285,19 @@ scan_microcode(void *data, size_t size, intel_microcode_sanity_check(data, false, MC_HEADER_TYPE_MICROCODE) < 0) break; - size -= mc_size; - -if (!intel_find_matching_signature (data, uci->cpu_sig.sig, - uci->cpu_sig.pf)) { - data += mc_size; +if (!intel_find_matching_signature (data, uci->cpu_sig.sig, uci->cpu_sig.pf))continue ; - }/* BSP scan: Check whether there is newer microcode */ if (!save && cur_rev >= mc_header->rev) - goto next; +continue ;/* Save scan: Check whether there is newer or matching microcode */ if (save && cur_rev != mc_header->rev) - goto next; +continue ; patch = data; cur_rev = mc_header->rev; - -next: - data += mc_size; }if (size)
Borislav Petkov
maintainerRe: [patch V3 04/30] x86/microcode/intel: Simplify scan_microcode()
Quoted 10 lines · Thomas Gleixner
Make it readable and comprehensible.
Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/kernel/cpu/microcode/intel.c | 30 ++++++++---------------------- 1 file changed, 8 insertions(+), 22 deletions(-)
arch/x86/kernel/cpu/microcode/intel.c static void save_microcode_patch(void *d · line 268
Quoted 5 lines
arch/x86/kernel/cpu/microcode/intel.c +1
-1@@ -265,25 +265,19 @@ static void save_microcode_patch(void *dreturn ;/* Save for early loading */ - intel_ucode_patch = (struct microcode_intel *)p; + intel_ucode_patch = (struct microcode_intel *)p;
This got committed by mistake.
arch/x86/kernel/cpu/microcode/intel.c static void save_microcode_patch(void *d · line 271
Quoted 7 lines
arch/x86/kernel/cpu/microcode/intel.c +1
-6-/* -* Get microcode matching with BSP's model. Only CPUs with the same model as -* BSP can stay in the platform. -*/ -static struct microcode_intel * -scan_microcode (void *data,size_t size,struct ucode_cpu_info *uci,bool save) +/* Scan CPIO for microcode matching the boot CPUs family, model, stepping */
"... the boot CPU's... " or simply the "BSP's".
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 5/30 x86/microcode/intel: Simplify and rename generic_load_microcode()
In reply to Thomas Gleixnerso it becomes less obfuscated and rename it because there is nothing generic about it.
---
V3: Rename to parse_microcode_blobs() - Borislav
--- arch/x86/kernel/cpu/microcode/intel.c | 47 ++++++++++++---------------------- 1 file changed, 17 insertions(+), 30 deletions(-)
arch/x86/kernel/cpu/microcode/intel.c +17 -30
@@ -240,19 +240,6 @@ int intel_microcode_sanity_check(void *m }EXPORT_SYMBOL_GPL (intel_microcode_sanity_check); -/* -* Returns 1 if update has been found, 0 otherwise. -*/ -static int has_newer_microcode (void *mc,unsigned int csig,int cpf,int new_rev) -{ -struct microcode_header_intel *mc_hdr = mc; - -if (mc_hdr->rev <= new_rev) -return 0 ; - -return intel_find_matching_signature (mc, csig, cpf); -} -static void save_microcode_patch (void *data,unsigned int size) {struct microcode_header_intel *p;
@@ -561,14 +548,12 @@ static enum ucode_state apply_microcode_return ret; } -static enum ucode_state generic_load_microcode (int cpu,struct iov_iter *iter) +static enum ucode_state parse_microcode_blobs (int cpu,struct iov_iter *iter) {struct ucode_cpu_info *uci = ucode_cpu_info + cpu;unsigned int curr_mc_size =0 , new_mc_size =0 ; -enum ucode_state ret =UCODE_OK ; -int new_rev = uci->cpu_sig.rev; +int cur_rev = uci->cpu_sig.rev;u8 *new_mc =NULL , *mc =NULL ; -unsigned int csig, cpf;while (iov_iter_count (iter)) {struct microcode_header_intel mc_header;
@@ -585,6 +570,7 @@ static enum ucode_state generic_load_micpr_err ("error! Bad data in microcode data file (totalsize too small)\n" );break ; } + data_size = mc_size -sizeof (mc_header);if (data_size >iov_iter_count (iter)) {pr_err ("error! Bad data in microcode data file (truncated file?)\n" );
@@ -607,16 +593,17 @@ static enum ucode_state generic_load_micbreak ; } - csig = uci->cpu_sig.sig; - cpf = uci->cpu_sig.pf; -if (has_newer_microcode (mc, csig, cpf, new_rev)) { -vfree (new_mc); - new_rev = mc_header.rev; - new_mc = mc; - new_mc_size = mc_size; - mc =NULL ;/* trigger new vmalloc */ - ret =UCODE_NEW ; - } +if (cur_rev >= mc_header.rev) +continue ; + +if (!intel_find_matching_signature (mc, uci->cpu_sig.sig, uci->cpu_sig.pf)) +continue ; + +vfree (new_mc); + cur_rev = mc_header.rev; + new_mc = mc; + new_mc_size = mc_size; + mc =NULL ; }vfree (mc);
@@ -636,9 +623,9 @@ static enum ucode_state generic_load_micsave_microcode_patch (new_mc, new_mc_size);pr_debug ("CPU%d found a matching microcode update with version 0x%x (current=0x%x)\n" , - cpu, new_rev, uci->cpu_sig.rev); + cpu, cur_rev, uci->cpu_sig.rev); -return ret; +return UCODE_NEW ; }static bool is_blacklisted (unsigned int cpu)
@@ -687,7 +674,7 @@ static enum ucode_state request_microcod kvec.iov_base = (void *)firmware->data; kvec.iov_len = firmware->size;iov_iter_kvec (&iter,ITER_SOURCE , &kvec,1 , firmware->size); - ret =generic_load_microcode (cpu, &iter); + ret =parse_microcode_blobs (cpu, &iter);release_firmware (firmware);
PATCH v3 6/30 x86/microcode/intel: Cleanup code further
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Sanitize the microcode scan loop, fixup printks and move the initrd loading function next to the place where it is used and mark it __init.
---
V2: Fix changelog - Nikolay
--- arch/x86/kernel/cpu/microcode/intel.c | 76 ++++++++++++++-------------------- 1 file changed, 32 insertions(+), 44 deletions(-)
---
arch/x86/kernel/cpu/microcode/intel.c +32 -44
@@ -36,7 +36,7 @@ static const char ucode_path[] = "kernelstatic struct microcode_intel *intel_ucode_patch__read_mostly ;/* last level cache size per core */ -static int llc_size_per_core__ro_after_init ; +static unsigned int llc_size_per_core__ro_after_init ;/* microcode format is extended from prescott processors */ struct extended_signature {
@@ -296,29 +296,6 @@ static struct microcode_intel *scan_micrreturn patch; } -static bool load_builtin_intel_microcode (struct cpio_data *cp) -{ -unsigned int eax =1 , ebx, ecx =0 , edx; -struct firmware fw; -char name[30 ]; - -if (IS_ENABLED (CONFIG_X86_32 )) -return false; - -native_cpuid (&eax, &ebx, &ecx, &edx); - -sprintf (name,"intel-ucode/%02x-%02x-%02x" , -x86_family (eax),x86_model (eax),x86_stepping (eax)); - -if (firmware_request_builtin (&fw, name)) { - cp->size = fw.size; - cp->data = (void *)fw.data; -return true; - } - -return false; -} -static int apply_microcode_early (struct ucode_cpu_info *uci,bool early) {struct microcode_intel *mc;
@@ -362,6 +339,28 @@ static int apply_microcode_early(structreturn 0 ; } +static bool load_builtin_intel_microcode (struct cpio_data *cp) +{ +unsigned int eax =1 , ebx, ecx =0 , edx; +struct firmware fw; +char name[30 ]; + +if (IS_ENABLED (CONFIG_X86_32 )) +return false; + +native_cpuid (&eax, &ebx, &ecx, &edx); + +sprintf (name,"intel-ucode/%02x-%02x-%02x" , +x86_family (eax),x86_model (eax),x86_stepping (eax)); + +if (firmware_request_builtin (&fw, name)) { + cp->size = fw.size; + cp->data = (void *)fw.data; +return true; + } +return false; +} +int __init save_microcode_in_initrd_intel (void ) {struct ucode_cpu_info uci;
@@ -434,25 +433,16 @@ void load_ucode_intel_ap(void)apply_microcode_early (&uci, true); } -/* Accessor for microcode pointer */ -static struct microcode_intel *ucode_get_patch (void ) -{ -return intel_ucode_patch; -} -void reload_ucode_intel (void ) { -struct microcode_intel *p;struct ucode_cpu_info uci;intel_cpu_collect_info (&uci); - p =ucode_get_patch (); -if (!p) + uci.mc = intel_ucode_patch; +if (!uci.mc)return ; - uci.mc = p; -apply_microcode_early (&uci, false); }
@@ -490,8 +480,7 @@ static enum ucode_state apply_microcode_if (WARN_ON (raw_smp_processor_id () != cpu))return UCODE_ERROR ; -/* Look for a newer patch in our cache: */ - mc =ucode_get_patch (); + mc = intel_ucode_patch;if (!mc) { mc = uci->mc;if (!mc)
@@ -682,18 +671,17 @@ static enum ucode_state request_microcod }static struct microcode_ops microcode_intel_ops = { - .request_microcode_fw = request_microcode_fw, - .collect_cpu_info = collect_cpu_info, - .apply_microcode = apply_microcode_intel, + .request_microcode_fw = request_microcode_fw, + .collect_cpu_info = collect_cpu_info, + .apply_microcode = apply_microcode_intel, }; -static int __init calc_llc_size_per_core (struct cpuinfo_x86 *c) +static __init void calc_llc_size_per_core (struct cpuinfo_x86 *c) {u64 llc_size = c->x86_cache_size *1024ULL ;do_div (llc_size, c->x86_max_cores); - -return (int )llc_size; + llc_size_per_core = (unsigned int )llc_size; }struct microcode_ops *__init init_intel_microcode (void )
@@ -706,7 +694,7 @@ struct microcode_ops * __init init_intelreturn NULL ; } - llc_size_per_core =calc_llc_size_per_core (c); +calc_llc_size_per_core (c);return µcode_intel_ops; }
Borislav Petkov
maintainerRe: [patch V3 06/30] x86/microcode/intel: Cleanup code further
Quoted 3 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Sanitize the microcode scan loop, fixup printks and move the initrd loading
The "builtin". Yeah, we support builtin microcode loading too. ;-\
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 7/30 x86/microcode/intel: Simplify early loading
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
The early loading code is overly complicated:
- It scans the builtin/initrd for microcode not only on the BSP, but also on all APs during early boot and then later in the boot process it scans again to duplicate and save the microcode before initrd goes away.
That's a pointless exercise because this can be simply done before bringing up the APs when the memory allocator is up and running.
- Saving the microcode from within the scan loop is completely non-obvious and a left over of the microcode cache.
This can be done at the call site now which makes it obvious.
Rework the code so that only the BSP scans the builtin/initrd microcode once during early boot and save it away in an early initcall for later use.
--- arch/x86/kernel/cpu/microcode/core.c | 4 arch/x86/kernel/cpu/microcode/intel.c | 148 +++++++++++++------------------ arch/x86/kernel/cpu/microcode/internal.h | 2 3 files changed, 64 insertions(+), 90 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +0 -4
@@ -186,10 +186,6 @@ static int __init save_microcode_in_initint ret = -EINVAL ;switch (c->x86_vendor) { -case X86_VENDOR_INTEL : -if (c->x86 >=6 ) - ret =save_microcode_in_initrd_intel (); -break ;case X86_VENDOR_AMD :if (c->x86 >=0x10 ) ret =save_microcode_in_initrd_amd (cpuid_eax (1 ));
arch/x86/kernel/cpu/microcode/intel.c +64 -84
269 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/internal.h +0 -2
@@ -107,13 +107,11 @@ static inline void exit_amd_microcode(vo#ifdef CONFIG_CPU_SUP_INTEL void load_ucode_intel_bsp (void );void load_ucode_intel_ap (void ); -int save_microcode_in_initrd_intel (void );void reload_ucode_intel (void );struct microcode_ops *init_intel_microcode (void );#else /* CONFIG_CPU_SUP_INTEL */ static inline void load_ucode_intel_bsp (void ) { }static inline void load_ucode_intel_ap (void ) { } -static inline int save_microcode_in_initrd_intel (void ) {return -EINVAL ; }static inline void reload_ucode_intel (void ) { }static inline struct microcode_ops *init_intel_microcode (void ) {return NULL ; }#endif /* !CONFIG_CPU_SUP_INTEL */
Borislav Petkov
maintainerRe: [patch V3 07/30] x86/microcode/intel: Simplify early loading
arch/x86/kernel/cpu/microcode/intel.c static bool load_builtin_intel_microcode · line 359
Quoted 6 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/intel.c +1
-1@@ -361,89 +356,75 @@ static bool load_builtin_intel_microcodereturn false; } -int __init save_microcode_in_initrd_intel (void ) +static __init struct microcode_intel *get_ucode_from_cpio (struct ucode_cpu_info *uci)
This is not only cpio but the builtin crap too. Just call it load_microcode_blobs() or so.
...
Quoted 15 lines
arch/x86/kernel/cpu/microcode/intel.c +3
-10-static struct microcode_intel *__load_ucode_intel (struct ucode_cpu_info *uci) +static int __init save_microcode_from_cpio (void ) { -struct cpio_data cp; - -/* try built-in microcode first */ -if (!load_builtin_intel_microcode (&cp)) - cp =find_microcode_in_initrd (ucode_path); - -if (!(cp.data && cp.size)) -return NULL ; +struct microcode_intel *mc; -intel_cpu_collect_info (uci); +if (!ucode_early_pa)
arch/x86/kernel/cpu/microcode/intel.c static bool load_builtin_intel_microcode · line 396
Quoted 10 lines
arch/x86/kernel/cpu/microcode/intel.c +6
-1+return 0 ; -return scan_microcode (cp.data, cp.size, uci, false); + mc =__va ((void *)ucode_early_pa); +save_microcode_patch (mc); +return 0 ; } +early_initcall (save_microcode_from_cpio); +/* Load microcode on BSP from CPIO */
Yeah, no need to say "from CPIO" everywhere. We load it from somewhere, it can be cpio but it can be builtin too.
But I can fix up too when applying.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 07/30] x86/microcode/intel: Simplify early loading
Quoted 5 lines · Borislav Petkov
Quoted 1 line · Thomas Gleixner
+/* Load microcode on BSP from CPIO */
Yeah, no need to say "from CPIO" everywhere. We load it from somewhere, it can be cpio but it can be builtin too.
Makes sense.
Quoted 1 line
But I can fix up too when applying.
Thanks a lot!
Re: [patch V3 07/30] x86/microcode/intel: Simplify early loading
Quoted 4 lines · Thomas Gleixner
Quoted 3 lines · Borislav Petkov
Quoted 1 line · Thomas Gleixner
+/* Load microcode on BSP from CPIO */
Quoted 8 lines
Quoted 2 lines
Yeah, no need to say "from CPIO" everywhere. We load it from somewhere, it can be cpio but it can be builtin too.
Makes sense.
Quoted 1 line
But I can fix up too when applying.
Thanks a lot!
Let me fix that up and repost as the removal of the early argument in patch 1/N causes a metric ton of conflicts all over the series.
Thanks,
tglx
PATCH v3 8/30 x86/microcode/intel: Save the microcode only after a successful late-load
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
There are situations where the late microcode is loaded into memory, but is not applied:
1) The rendevouz fails
2) The microcode is rejected by the CPUs
If any of this happens then the pointer which was updated at firmware load time is stale and subsequent CPU hotplug operations either fail to update or create inconsistent microcode state.
Save the loaded microcode in a separate pointer from with the late load is attempted and when successful, update the hotplug pointer accordingly via a new micrcode_ops callback.
Remove the pointless fallback in the loader to a microcode pointer which is never populated.
--- arch/x86/kernel/cpu/microcode/core.c | 4 ++++ arch/x86/kernel/cpu/microcode/intel.c | 30 +++++++++++++++--------------- arch/x86/kernel/cpu/microcode/internal.h | 1 + 3 files changed, 20 insertions(+), 15 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +4 -0
@@ -400,6 +400,10 @@ static int microcode_reload_late(void)store_cpu_caps (&prev_info); ret =stop_machine_cpuslocked (__reload_late,NULL , cpu_online_mask); + +if (microcode_ops->finalize_late_load) + microcode_ops->finalize_late_load (ret); +if (!ret) {pr_info ("Reload succeeded, microcode revision: 0x%x -> 0x%x\n" , old, boot_cpu_data.microcode);
arch/x86/kernel/cpu/microcode/intel.c +15 -15
@@ -34,6 +34,7 @@ static const char ucode_path[] = "kernel/* Current microcode patch used in early patching on the APs. */ static struct microcode_intel *ucode_patch_va__read_mostly ; +static struct microcode_intel *ucode_patch_late__read_mostly ;/* last level cache size per core */ static unsigned int llc_size_per_core__ro_after_init ;
@@ -461,12 +462,9 @@ static enum ucode_state apply_microcode_if (WARN_ON (raw_smp_processor_id () != cpu))return UCODE_ERROR ; - mc = ucode_patch_va; -if (!mc) { - mc = uci->mc; -if (!mc) -return UCODE_NFOUND ; - } + mc = ucode_patch_late; +if (!mc) +return UCODE_NFOUND ;/* * Save us the MSR write below - which is a particular expensive
@@ -585,15 +583,7 @@ static enum ucode_state parse_microcode_if (!new_mc)return UCODE_NFOUND ; -/* Save for CPU hotplug */ -save_microcode_patch ((struct microcode_intel *)new_mc); - uci->mc = ucode_patch_va; - -vfree (new_mc); - -pr_debug ("CPU%d found a matching microcode update with version 0x%x (current=0x%x)\n" , - cpu, cur_rev, uci->cpu_sig.rev); - + ucode_patch_late = (struct microcode_intel *)new_mc;return UCODE_NEW ; }
@@ -650,10 +640,20 @@ static enum ucode_state request_microcodreturn ret; } +static void finalize_late_load (int result) +{ +if (!result) +save_microcode_patch (ucode_patch_late); + +vfree (ucode_patch_late); + ucode_patch_late =NULL ; +} +static struct microcode_ops microcode_intel_ops = { .request_microcode_fw = request_microcode_fw, .collect_cpu_info = collect_cpu_info, .apply_microcode = apply_microcode_intel, + .finalize_late_load = finalize_late_load, };static __init void calc_llc_size_per_core (struct cpuinfo_x86 *c)
arch/x86/kernel/cpu/microcode/internal.h +1 -0
@@ -31,6 +31,7 @@ struct microcode_ops {*/ enum ucode_state (*apply_microcode)(int cpu);int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); +void (*finalize_late_load)(int result); };extern struct ucode_cpu_info ucode_cpu_info[];
Borislav Petkov
maintainerRe: [patch V3 08/30] x86/microcode/intel: Save the microcode only after a successful late-load
Quoted 5 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
There are situations where the late microcode is loaded into memory, but is not applied:
Quoted 8 lines
1) The rendevouz fails
2) The microcode is rejected by the CPUsIf any of this happens then the pointer which was updated at firmware load time is stale and subsequent CPU hotplug operations either fail to update or create inconsistent microcode state.
Save the loaded microcode in a separate pointer from with the late load is
s/from with/before/
Quoted 2 lines
attempted and when successful, update the hotplug pointer accordingly via a new
micrcode_opscallback.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: Subject: [patch V3 08/30] x86/microcode/intel: Save the microcode only after a successful late-load
In reply to Thomas GleixnerQuoted 7 lines
...
From: Thomas Gleixner <[email protected]>There are situations where the late microcode is loaded into memory, but is not applied:
1) The rendevouz fails
s/rendevouz/rendezvous/
Quoted 9 lines
2) The microcode is rejected by the CPUs
If any of this happens then the pointer which was updated at firmware load time is stale and subsequent CPU hotplug operations either fail to update or create inconsistent microcode state.
Save the loaded microcode in a separate pointer from with the late load is attempted and when successful, update the hotplug pointer accordingly via a new
micrcode_opscallback.
s/micrcode_ops/microcode_ops
Quoted 4 lines
Remove the pointless fallback in the loader to a microcode pointer which is never populated.
...
PATCH v3 9/30 x86/microcode/intel: Switch to kvmalloc()
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Microcode blobs are getting larger and might soon reach the kmalloc() limit. Switch over kvmalloc().
--- arch/x86/kernel/cpu/microcode/intel.c | 50 +++++++++++++++++----------------- 1 file changed, 26 insertions(+), 24 deletions(-)
---
arch/x86/kernel/cpu/microcode/intel.c +26 -24
@@ -14,7 +14,6 @@#include <linux/earlycpio.h> #include <linux/firmware.h> #include <linux/uaccess.h> -#include <linux/vmalloc.h> #include <linux/initrd.h> #include <linux/kernel.h> #include <linux/slab.h>
@@ -243,7 +242,7 @@ EXPORT_SYMBOL_GPL(intel_microcode_sanitystatic void update_ucode_pointer (struct microcode_intel *mc) { -kfree (ucode_patch_va); +kvfree (ucode_patch_va);/* * Save the virtual address for early loading and for eventual free
@@ -254,11 +253,14 @@ static void update_ucode_pointer(structstatic void save_microcode_patch (struct microcode_intel *patch) { -struct microcode_intel *mc; +unsigned int size =get_totalsize (&patch->hdr); +struct microcode_intel *mc =NULL ; - mc =kmemdup (patch,get_totalsize (&patch->hdr),GFP_KERNEL ); + mc =kvmemdup (patch, size,GFP_KERNEL );if (mc)update_ucode_pointer (mc); +else +pr_err ("Unable to allocate microcode memory size: %u\n" , size); }/* Scan CPIO for microcode matching the boot CPUs family, model, stepping */
@@ -530,36 +532,34 @@ static enum ucode_state read_ucode_intelif (!copy_from_iter_full (&mc_header,sizeof (mc_header), iter)) {pr_err ("error! Truncated or inaccessible header in microcode data file\n" ); -break ; + goto fail; } mc_size =get_totalsize (&mc_header);if (mc_size <sizeof (mc_header)) {pr_err ("error! Bad data in microcode data file (totalsize too small)\n" ); -break ; + goto fail; } - data_size = mc_size -sizeof (mc_header);if (data_size >iov_iter_count (iter)) {pr_err ("error! Bad data in microcode data file (truncated file?)\n" ); -break ; + goto fail; }/* For performance reasons, reuse mc area when possible */ if (!mc || mc_size > curr_mc_size) { -vfree (mc); - mc =vmalloc (mc_size); +kvfree (mc); + mc =kvmalloc (mc_size,GFP_KERNEL );if (!mc) -break ; + goto fail; curr_mc_size = mc_size; }memcpy (mc, &mc_header,sizeof (mc_header)); data = mc +sizeof (mc_header);if (!copy_from_iter_full (data, data_size, iter) || -intel_microcode_sanity_check (mc, true,MC_HEADER_TYPE_MICROCODE ) <0 ) { -break ; - } +intel_microcode_sanity_check (mc, true,MC_HEADER_TYPE_MICROCODE ) <0 ) + goto fail;if (cur_rev >= mc_header.rev)continue ;
@@ -567,24 +567,26 @@ static enum ucode_state read_ucode_intelif (!intel_find_matching_signature (mc, uci->cpu_sig.sig, uci->cpu_sig.pf))continue ; -vfree (new_mc); +kvfree (new_mc); cur_rev = mc_header.rev; new_mc = mc; mc =NULL ; } -vfree (mc); - -if (iov_iter_count (iter)) { -vfree (new_mc); -return UCODE_ERROR ; - } +if (iov_iter_count (iter)) + goto fail; +kvfree (mc);if (!new_mc)return UCODE_NFOUND ; ucode_patch_late = (struct microcode_intel *)new_mc;return UCODE_NEW ; + +fail: +kvfree (mc); +kvfree (new_mc); +return UCODE_ERROR ; }static bool is_blacklisted (unsigned int cpu)
@@ -643,9 +645,9 @@ static enum ucode_state request_microcodstatic void finalize_late_load (int result) {if (!result) -save_microcode_patch (ucode_patch_late); - -vfree (ucode_patch_late); +update_ucode_pointer (ucode_patch_late); +else +kvfree (ucode_patch_late); ucode_patch_late =NULL ; }
Re: Subject: [patch V3 09/30] x86/microcode/intel: Switch to kvmalloc()
Quoted 14 lines
...
From: Thomas Gleixner <[email protected]>Microcode blobs are getting larger and might soon reach the
kmalloc()limit. Switch overkvmalloc().Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/kernel/cpu/microcode/intel.c | 50 +++++++++++++++++----------------- 1 file changed, 26 insertions(+), 24 deletions(-)---
arch/x86/kernel/cpu/microcode/intel.c static void update_ucode_pointer(struct · line 257
Quoted 6 lines
arch/x86/kernel/cpu/microcode/intel.c +2
-1static void save_microcode_patch (struct microcode_intel *patch) { -struct microcode_intel *mc; +unsigned int size =get_totalsize (&patch->hdr); +struct microcode_intel *mc =NULL ;
No need to initialize the 'mc' to NULL as it's unconditionally set by the following kmemdup().
Quoted 10 lines
arch/x86/kernel/cpu/microcode/intel.c +3
-1- mc =kmemdup (patch,get_totalsize (&patch->hdr),GFP_KERNEL ); + mc =kvmemdup (patch, size,GFP_KERNEL );if (mc)update_ucode_pointer (mc); +else +pr_err ("Unable to allocate microcode memory size: %u\n" , size); }
PATCH v3 10/30 x86/microcode/intel: Unify microcode apply() functions
In reply to Thomas GleixnerDeduplicate the early and late apply() functions.
--- arch/x86/kernel/cpu/microcode/intel.c | 105 +++++++++++----------------------- 1 file changed, 36 insertions(+), 69 deletions(-)
arch/x86/kernel/cpu/microcode/intel.c +36 -69
@@ -294,12 +294,11 @@ static __init struct microcode_intel *screturn size ?NULL : patch; } -static enum ucode_state apply_microcode_early (struct ucode_cpu_info *uci,bool early) +static enum ucode_state apply_microcode (struct ucode_cpu_info *uci,struct microcode_intel *mc, +u32 *cur_rev) { -struct microcode_intel *mc; -u32 rev, old_rev, date; +u32 rev; - mc = uci->mc;if (!mc)return UCODE_NFOUND ;
@@ -308,14 +307,12 @@ static enum ucode_state apply_microcode_* operation - when the other hyperthread has updated the microcode * already. */ - rev =intel_get_microcode_revision (); -if (rev >= mc->hdr.rev) { - uci->cpu_sig.rev = rev; + *cur_rev =intel_get_microcode_revision (); +if (*cur_rev >= mc->hdr.rev) { + uci->cpu_sig.rev = *cur_rev;return UCODE_OK ; } - old_rev = rev; -/* * Writeback and invalidate caches before updating microcode to avoid * internal issues depending on what the microcode is updating.
@@ -330,13 +327,24 @@ static enum ucode_state apply_microcode_return UCODE_ERROR ; uci->cpu_sig.rev = rev; - - date = mc->hdr.date; -pr_info_once ("updated early: 0x%x -> 0x%x, date = %04x-%02x-%02x\n" , - old_rev, rev, date &0xffff , date >>24 , (date >>16 ) &0xff );return UCODE_UPDATED ; } +static enum ucode_state apply_microcode_early (struct ucode_cpu_info *uci,bool early) +{ +struct microcode_intel *mc = uci->mc; +enum ucode_state ret; +u32 cur_rev, date; + + ret =apply_microcode (uci, mc, &cur_rev); +if (ret ==UCODE_UPDATED ) { + date = mc->hdr.date; +pr_info_once ("updated early: 0x%x -> 0x%x, date = %04x-%02x-%02x\n" , + cur_rev, mc->hdr.rev, date &0xffff , date >>24 , (date >>16 ) &0xff ); + } +return ret; +} +static __init bool load_builtin_intel_microcode (struct cpio_data *cp) {unsigned int eax =1 , ebx, ecx =0 , edx;
@@ -450,70 +458,29 @@ static int collect_cpu_info(int cpu_num,return 0 ; } -static enum ucode_state apply_microcode_intel (int cpu) +static enum ucode_state apply_microcode_late (int cpu) {struct ucode_cpu_info *uci = ucode_cpu_info + cpu; -struct cpuinfo_x86 *c = &cpu_data (cpu); -bool bsp = c->cpu_index == boot_cpu_data.cpu_index; -struct microcode_intel *mc; +struct microcode_intel *mc = ucode_patch_late;enum ucode_state ret; -static int prev_rev; -u32 rev; - -/* We should bind the task to the CPU */ -if (WARN_ON (raw_smp_processor_id () != cpu)) -return UCODE_ERROR ; - - mc = ucode_patch_late; -if (!mc) -return UCODE_NFOUND ; +u32 cur_rev; -/* -* Save us the MSR write below - which is a particular expensive -* operation - when the other hyperthread has updated the microcode -* already. -*/ - rev =intel_get_microcode_revision (); -if (rev >= mc->hdr.rev) { - ret =UCODE_OK ; - goto out; - } - -/* -* Writeback and invalidate caches before updating microcode to avoid -* internal issues depending on what the microcode is updating. -*/ -native_wbinvd (); - -/* write microcode via MSR 0x79 */ -wrmsrl (MSR_IA32_UCODE_WRITE , (unsigned long )mc->bits); - - rev =intel_get_microcode_revision (); - -if (rev != mc->hdr.rev) { -pr_err ("CPU%d update to revision 0x%x failed\n" , - cpu, mc->hdr.rev); +if (WARN_ON_ONCE (smp_processor_id () != cpu))return UCODE_ERROR ; - } -if (bsp && rev != prev_rev) { -pr_info ("updated to revision 0x%x, date = %04x-%02x-%02x\n" , - rev, - mc->hdr.date &0xffff , - mc->hdr.date >>24 , + ret =apply_microcode (uci, mc, &cur_rev); +if (ret !=UCODE_UPDATED && ret !=UCODE_OK ) +return ret; + +if (!cpu && uci->cpu_sig.rev != cur_rev) { +pr_info ("Updated to revision 0x%x, date = %04x-%02x-%02x\n" , + uci->cpu_sig.rev, mc->hdr.date &0xffff , mc->hdr.date >>24 , (mc->hdr.date >>16 ) &0xff ); - prev_rev = rev; } - ret =UCODE_UPDATED ; - -out: - uci->cpu_sig.rev = rev; - c->microcode = rev; - -/* Update boot_cpu_data's revision too, if we're on the BSP: */ -if (bsp) - boot_cpu_data.microcode = rev; +cpu_data (cpu).microcode = uci->cpu_sig.rev; +if (!cpu) + boot_cpu_data.microcode = uci->cpu_sig.rev;return ret; }
@@ -654,7 +621,7 @@ static void finalize_late_load(int resulstatic struct microcode_ops microcode_intel_ops = { .request_microcode_fw = request_microcode_fw, .collect_cpu_info = collect_cpu_info, - .apply_microcode = apply_microcode_intel, + .apply_microcode = apply_microcode_late, .finalize_late_load = finalize_late_load, };
Borislav Petkov
maintainerRe: [patch V3 10/30] x86/microcode/intel: Unify microcode apply() functions
Quoted 10 lines · Thomas Gleixner
Deduplicate the early and late
apply()functions.Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/kernel/cpu/microcode/intel.c | 105 +++++++++++----------------------- 1 file changed, 36 insertions(+), 69 deletions(-)
arch/x86/kernel/cpu/microcode/intel.c static __init struct microcode_intel *sc · line 297
Quoted 5 lines
arch/x86/kernel/cpu/microcode/intel.c +1
-1@@ -294,12 +294,11 @@ static __init struct microcode_intel *screturn size ?NULL : patch; } -static enum ucode_state apply_microcode_early (struct ucode_cpu_info *uci,bool early) +static enum ucode_state apply_microcode(struct ucode_cpu_info *uci, struct microcode_intel *mc,
Err, we have
struct microcode_ops.apply_microcode()
already. Can we disambiguate those pls?
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 11/30 x86/microcode/intel: Rework intel_cpu_collect_info()
In reply to Thomas GleixnerNothing needs struct ucode_cpu_info. Make it take struct cpu_signature, let it return a boolean and simplify the implementation. Rename it now that the silly name clash with collect_cpu_info() is gone.
---
V2: New patch
--- arch/x86/include/asm/cpu.h | 4 ++-- arch/x86/kernel/cpu/microcode/intel.c | 33 +++++++++------------------------ drivers/platform/x86/intel/ifs/load.c | 8 +++----- 3 files changed, 14 insertions(+), 31 deletions(-)
arch/x86/include/asm/cpu.h +2 -2
@@ -73,9 +73,9 @@ static inline void init_ia32_feat_ctl(st extern __noendbr void cet_disable(void); -struct ucode_cpu_info ; +struct cpu_signature ; -int intel_cpu_collect_info (struct ucode_cpu_info *uci); +void intel_collect_cpu_info (struct cpu_signature *sig);static inline bool intel_cpu_signatures_match (unsigned int s1,unsigned int p1,unsigned int s2,unsigned int p2)
arch/x86/kernel/cpu/microcode/intel.c +9 -24
@@ -66,36 +66,21 @@ static inline unsigned int exttable_sizereturn et->count *EXT_SIGNATURE_SIZE +EXT_HEADER_SIZE ; } -int intel_cpu_collect_info (struct ucode_cpu_info *uci) +void intel_collect_cpu_info (struct cpu_signature *sig) { -unsigned int val[2 ]; -unsigned int family, model; -struct cpu_signature csig = {0 }; -unsigned int eax, ebx, ecx, edx; + sig->sig =cpuid_eax (1 ); + sig->pf =0 ; + sig->rev =intel_get_microcode_revision (); -memset (uci,0 ,sizeof (*uci)); +if (x86_model (sig->sig) >=5 ||x86_family (sig->sig) >6 ) { +unsigned int val[2 ]; - eax =0x00000001 ; - ecx =0 ; -native_cpuid (&eax, &ebx, &ecx, &edx); - csig.sig = eax; - - family =x86_family (eax); - model =x86_model (eax); - -if (model >=5 || family >6 ) {/* get processor flags from MSR 0x17 */ native_rdmsr (MSR_IA32_PLATFORM_ID , val[0 ], val[1 ]); - csig.pf =1 << ((val[1 ] >>18 ) &7 ); + sig->pf =1 << ((val[1 ] >>18 ) &7 ); } - - csig.rev =intel_get_microcode_revision (); - - uci->cpu_sig = csig; - -return 0 ; } -EXPORT_SYMBOL_GPL (intel_cpu_collect_info); +EXPORT_SYMBOL_GPL (intel_collect_cpu_info);/* * Returns 1 if update has been found, 0 otherwise.
@@ -381,7 +366,7 @@ static __init struct microcode_intel *geif (!(cp.data && cp.size))return NULL ; -intel_cpu_collect_info (uci); +intel_collect_cpu_info (&uci->cpu_sig);return scan_microcode (cp.data, cp.size, uci); }
drivers/platform/x86/intel/ifs/load.c +3 -5
@@ -227,7 +227,7 @@ static int scan_chunks_sanity_check(strustatic int image_sanity_check (struct device *dev,const struct microcode_header_intel *data) { -struct ucode_cpu_info uci; +struct cpu_signature sig;/* Provide a specific error message when loading an older/unsupported image */ if (data->hdrver !=MC_HEADER_TYPE_IFS ) {
@@ -240,11 +240,9 @@ static int image_sanity_check(struct devreturn -EINVAL ; } -intel_cpu_collect_info (&uci); +intel_collect_cpu_info (&sig); -if (!intel_find_matching_signature ((void *)data, - uci.cpu_sig.sig, - uci.cpu_sig.pf)) { +if (!intel_find_matching_signature ((void *)data, sig.sig, sig.pf)) {dev_err (dev,"cpu signature, processor flags not matching\n" );return -EINVAL ; }
PATCH v3 12/30 x86/microcode/intel: Reuse intel_cpu_collect_info()
In reply to Thomas GleixnerNo point for an almost duplicate function.
---
V2: New patch
--- arch/x86/kernel/cpu/microcode/intel.c | 16 +--------------- 1 file changed, 1 insertion(+), 15 deletions(-)
arch/x86/kernel/cpu/microcode/intel.c +1 -15
@@ -425,21 +425,7 @@ void reload_ucode_intel(void)static int collect_cpu_info (int cpu_num,struct cpu_signature *csig) { -struct cpuinfo_x86 *c = &cpu_data (cpu_num); -unsigned int val[2 ]; - -memset (csig,0 ,sizeof (*csig)); - - csig->sig =cpuid_eax (0x00000001 ); - -if ((c->x86_model >=5 ) || (c->x86 >6 )) { -/* get processor flags from MSR 0x17 */ -rdmsr (MSR_IA32_PLATFORM_ID , val[0 ], val[1 ]); - csig->pf =1 << ((val[1 ] >>18 ) &7 ); - } - - csig->rev = c->microcode; - +intel_collect_cpu_info (csig);return 0 ; }
Borislav Petkov
maintainerRe: [patch V3 12/30] x86/microcode/intel: Reuse intel_cpu_collect_info()
Quoted 1 line · Thomas Gleixner
static int
collect_cpu_info(int cpu_num, struct cpu_signature *csig)
You can get rid of that silly wrapper too and use intel_collect_cpu_info() in the function pointer assignment directly.
Diff ontop:
---
arch/x86/include/asm/cpu.h +1 -1
@@ -75,7 +75,7 @@ extern __noendbr void cet_disable(void);struct cpu_signature ; -void intel_collect_cpu_info (struct cpu_signature *sig); +void intel_collect_cpu_info (int unused,struct cpu_signature *sig);static inline bool intel_cpu_signatures_match (unsigned int s1,unsigned int p1,unsigned int s2,unsigned int p2)
arch/x86/kernel/cpu/microcode/amd.c +1 -3
@@ -652,7 +652,7 @@ void reload_ucode_amd(unsigned int cpu) } } -static int collect_cpu_info_amd (int cpu,struct cpu_signature *csig) +static void collect_cpu_info_amd (int cpu,struct cpu_signature *csig) {struct cpuinfo_x86 *c = &cpu_data (cpu);struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
@@ -670,8 +670,6 @@ static int collect_cpu_info_amd(int cpu, struct cpu_signature *csig) uci->mc = p->data;pr_info ("CPU%d: patch_level=0x%08x\n" , cpu, csig->rev); - -return 0 ; }static enum ucode_state apply_microcode_amd (int cpu)
arch/x86/kernel/cpu/microcode/intel.c +3 -9
@@ -66,7 +66,7 @@ static inline unsigned int exttable_size(struct extended_sigtable *et)return et->count *EXT_SIGNATURE_SIZE +EXT_HEADER_SIZE ; } -void intel_collect_cpu_info (struct cpu_signature *sig) +void intel_collect_cpu_info (int unused,struct cpu_signature *sig) { sig->sig =cpuid_eax (1 ); sig->pf =0 ;
@@ -362,7 +362,7 @@ static __init struct microcode_intel *get_ucode_from_cpio(struct ucode_cpu_infoif (!(cp.data && cp.size))return NULL ; -intel_collect_cpu_info (&uci->cpu_sig); +intel_collect_cpu_info (0 , &uci->cpu_sig);return scan_microcode (cp.data, cp.size, uci); }
@@ -423,12 +423,6 @@ void reload_ucode_intel(void)apply_microcode_early (&uci, false); } -static int collect_cpu_info (int cpu_num,struct cpu_signature *csig) -{ -intel_collect_cpu_info (csig); -return 0 ; -} -static enum ucode_state apply_microcode_late (int cpu) {struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
@@ -591,7 +585,7 @@ static void finalize_late_load(int result)static struct microcode_ops microcode_intel_ops = { .request_microcode_fw = request_microcode_fw, - .collect_cpu_info = collect_cpu_info, + .collect_cpu_info = intel_collect_cpu_info, .apply_microcode = apply_microcode_late, .finalize_late_load = finalize_late_load, };
arch/x86/kernel/cpu/microcode/internal.h +1 -1
@@ -30,7 +30,7 @@ struct microcode_ops {* See also the "Synchronization" section in microcode_core.c. */ enum ucode_state (*apply_microcode)(int cpu); -int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); +void (*collect_cpu_info)(int cpu,struct cpu_signature *csig);void (*finalize_late_load)(int result); };
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 12/30] x86/microcode/intel: Reuse intel_cpu_collect_info()
Quoted 15 lines · Borislav Petkov
Quoted 1 line · Thomas Gleixner
static int
collect_cpu_info(int cpu_num, struct cpu_signature *csig)You can get rid of that silly wrapper too and use
intel_collect_cpu_info()in the function pointer assignment directly.Diff ontop:
---
arch/x86/include/asm/cpu.h +0
-0
arch/x86/include/asm/cpu.h extern __noendbr void cet_disable(void); · line 78
Quoted 5 lines
arch/x86/include/asm/cpu.h +1
-1@@ -75,7 +75,7 @@ extern __noendbr void cet_disable(void);struct cpu_signature ; -void intel_collect_cpu_info (struct cpu_signature *sig); +void intel_collect_cpu_info (int unused,struct cpu_signature *sig);
Eew. That's a function exposed to code outside of microcode and just grows that unused argument for no value and you obviously forgot to fixup the extern callsite :)
arch/x86/kernel/cpu/microcode/internal.h struct microcode_ops { · line 34
Quoted 11 lines
arch/x86/kernel/cpu/microcode/internal.h +1
-1@@ -30,7 +30,7 @@ struct microcode_ops {* See also the "Synchronization" section in microcode_core.c. */ enum ucode_state (*apply_microcode)(int cpu); -int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); +void (*collect_cpu_info)(int cpu,struct cpu_signature *csig);void (*finalize_late_load)(int result);
Making this void makes sense, but that's a separate change.
Thanks,
tglx
Borislav Petkov
maintainerRe: [patch V3 12/30] x86/microcode/intel: Reuse intel_cpu_collect_info()
Quoted 3 lines · Thomas Gleixner
Eew. That's a function exposed to code outside of microcode and just grows that unused argument for no value and you obviously forgot to fixup the extern callsite :)
It's used on AMD. Adding the below to the pile.
---
From: "Borislav Petkov (AMD)" <[email protected]>
Date: Tue, 3 Oct 2023 16:12:01 +0200
Subject: [PATCH] x86/microcode: Make microcode_ops.collect_cpu_info() return
void
Simplify code flow a bit more in the process.
No functional changes.
--- arch/x86/include/asm/cpu.h | 2 +- arch/x86/kernel/cpu/microcode/amd.c | 4 +--- arch/x86/kernel/cpu/microcode/intel.c | 12 +++--------- arch/x86/kernel/cpu/microcode/internal.h | 2 +- 4 files changed, 6 insertions(+), 14 deletions(-)
arch/x86/include/asm/cpu.h +1 -1
@@ -75,7 +75,7 @@ extern __noendbr void cet_disable(void);struct cpu_signature ; -void intel_collect_cpu_info (struct cpu_signature *sig); +void intel_collect_cpu_info (int unused,struct cpu_signature *sig);static inline bool intel_cpu_signatures_match (unsigned int s1,unsigned int p1,unsigned int s2,unsigned int p2)
arch/x86/kernel/cpu/microcode/amd.c +1 -3
@@ -632,7 +632,7 @@ void reload_ucode_amd(unsigned int cpu) } } -static int collect_cpu_info_amd (int cpu,struct cpu_signature *csig) +static void collect_cpu_info_amd (int cpu,struct cpu_signature *csig) {struct cpuinfo_x86 *c = &cpu_data (cpu);struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
@@ -650,8 +650,6 @@ static int collect_cpu_info_amd(int cpu, struct cpu_signature *csig) uci->mc = p->data;pr_info ("CPU%d: patch_level=0x%08x\n" , cpu, csig->rev); - -return 0 ; }static enum ucode_state apply_microcode_amd (int cpu)
arch/x86/kernel/cpu/microcode/intel.c +3 -9
@@ -66,7 +66,7 @@ static inline unsigned int exttable_size(struct extended_sigtable *et)return et->count *EXT_SIGNATURE_SIZE +EXT_HEADER_SIZE ; } -void intel_collect_cpu_info (struct cpu_signature *sig) +void intel_collect_cpu_info (int unused,struct cpu_signature *sig) { sig->sig =cpuid_eax (1 ); sig->pf =0 ;
@@ -363,7 +363,7 @@ static __init struct microcode_intel *get_microcode_blob(struct ucode_cpu_info *if (!(cp.data && cp.size))return NULL ; -intel_collect_cpu_info (&uci->cpu_sig); +intel_collect_cpu_info (0 , &uci->cpu_sig);return scan_microcode (cp.data, cp.size, uci); }
@@ -424,12 +424,6 @@ void reload_ucode_intel(void)apply_microcode_early (&uci); } -static int collect_cpu_info (int cpu_num,struct cpu_signature *csig) -{ -intel_collect_cpu_info (csig); -return 0 ; -} -static enum ucode_state apply_microcode_late (int cpu) {struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
@@ -592,7 +586,7 @@ static void finalize_late_load(int result)static struct microcode_ops microcode_intel_ops = { .request_microcode_fw = request_microcode_fw, - .collect_cpu_info = collect_cpu_info, + .collect_cpu_info = intel_collect_cpu_info, .apply_microcode = apply_microcode_late, .finalize_late_load = finalize_late_load, };
arch/x86/kernel/cpu/microcode/internal.h +1 -1
@@ -30,7 +30,7 @@ struct microcode_ops {* See also the "Synchronization" section in microcode_core.c. */ enum ucode_state (*apply_microcode)(int cpu); -int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); +void (*collect_cpu_info)(int cpu,struct cpu_signature *csig);void (*finalize_late_load)(int result); };
Signature · 8 lines
2.42.0.rc0.25.ga82fb66fed25
--
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteBorislav Petkov
maintainerRe: [patch V3 12/30] x86/microcode/intel: Reuse intel_cpu_collect_info()
Quoted 6 lines · Borislav Petkov
Quoted 3 lines · Thomas Gleixner
Eew. That's a function exposed to code outside of microcode and just grows that unused argument for no value and you obviously forgot to fixup the extern callsite :)
It's used on AMD. Adding the below to the pile.
And now that I look at it again, exposing that "unused" arg is uglier than having the local wrapper in the loader code. Yeah, lemme zap that.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 13/30 x86/microcode/intel: Rework intel_find_matching_signature()
In reply to Thomas GleixnerTake a cpu_signature argument and work from there. Move the match() helper next to the callsite as there is no point for having it in a header.
---
V2: New patch
--- arch/x86/include/asm/cpu.h | 16 +--------------- arch/x86/kernel/cpu/microcode/intel.c | 31 +++++++++++++++++++------------ drivers/platform/x86/intel/ifs/load.c | 2 +- 3 files changed, 21 insertions(+), 28 deletions(-)
arch/x86/include/asm/cpu.h +1 -15
@@ -77,22 +77,8 @@ struct cpu_signature;void intel_collect_cpu_info (struct cpu_signature *sig); -static inline bool intel_cpu_signatures_match (unsigned int s1,unsigned int p1, -unsigned int s2,unsigned int p2) -{ -if (s1 != s2) -return false; - -/* Processor flags are either both 0 ... */ -if (!p1 && !p2) -return true; - -/* ... or they intersect. */ -return p1 & p2; -} -extern u64 x86_read_arch_cap_msr (void ); -int intel_find_matching_signature (void *mc,unsigned int csig,int cpf); +bool intel_find_matching_signature (void *mc,struct cpu_signature *sig);int intel_microcode_sanity_check (void *mc,bool print_err,int hdr_type);extern struct cpumask cpus_stop_mask;
arch/x86/kernel/cpu/microcode/intel.c +19 -12
@@ -82,29 +82,36 @@ void intel_collect_cpu_info(struct cpu_s }EXPORT_SYMBOL_GPL (intel_collect_cpu_info); -/* -* Returns 1 if update has been found, 0 otherwise. -*/ -int intel_find_matching_signature (void *mc,unsigned int csig,int cpf) +static inline bool cpu_signatures_match (struct cpu_signature *s1,unsigned int sig2, +unsigned int pf2) +{ +if (s1->sig != sig2) +return false; + +/* Processor flags are either both 0 or they intersect. */ +return ((!s1->pf && !pf2) || (s1->pf & pf2)); +} + +bool intel_find_matching_signature (void *mc,struct cpu_signature *sig) {struct microcode_header_intel *mc_hdr = mc; -struct extended_sigtable *ext_hdr;struct extended_signature *ext_sig; +struct extended_sigtable *ext_hdr;int i; -if (intel_cpu_signatures_match (csig, cpf, mc_hdr->sig, mc_hdr->pf)) -return 1 ; +if (cpu_signatures_match (sig, mc_hdr->sig, mc_hdr->pf)) +return true;/* Look for ext. headers: */ if (get_totalsize (mc_hdr) <=intel_microcode_get_datasize (mc_hdr) +MC_HEADER_SIZE ) -return 0 ; +return false; ext_hdr = mc +intel_microcode_get_datasize (mc_hdr) +MC_HEADER_SIZE ; ext_sig = (void *)ext_hdr +EXT_HEADER_SIZE ;for (i =0 ; i < ext_hdr->count; i++) { -if (intel_cpu_signatures_match (csig, cpf, ext_sig->sig, ext_sig->pf)) -return 1 ; +if (cpu_signatures_match (sig, ext_sig->sig, ext_sig->pf)) +return true; ext_sig++; }return 0 ;
@@ -265,7 +272,7 @@ static __init struct microcode_intel *sc intel_microcode_sanity_check(data, false, MC_HEADER_TYPE_MICROCODE) < 0)break ; -if (!intel_find_matching_signature (data, uci->cpu_sig.sig, uci->cpu_sig.pf)) +if (!intel_find_matching_signature (data, &uci->cpu_sig))continue ;/* Check whether there is newer microcode */
@@ -502,7 +509,7 @@ static enum ucode_state read_ucode_intelif (cur_rev >= mc_header.rev)continue ; -if (!intel_find_matching_signature (mc, uci->cpu_sig.sig, uci->cpu_sig.pf)) +if (!intel_find_matching_signature (mc, &uci->cpu_sig))continue ;kvfree (new_mc);
drivers/platform/x86/intel/ifs/load.c +1 -1
@@ -242,7 +242,7 @@ static int image_sanity_check(struct devintel_collect_cpu_info (&sig); -if (!intel_find_matching_signature ((void *)data, sig.sig, sig.pf)) { +if (!intel_find_matching_signature ((void *)data, &sig)) {dev_err (dev,"cpu signature, processor flags not matching\n" );return -EINVAL ; }
PATCH v3 14/30 x86/microcode/amd: Read revision from hardware in collect_cpu_info_amd()
In reply to Thomas GleixnerPrepare to decrapify the core initialization logic which invokes microcode_ops::apply_microcode() several times just to set cpu_data::microcode.
---
V2: New patch
--- arch/x86/kernel/cpu/microcode/amd.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/kernel/cpu/microcode/amd.c +2 -2
@@ -673,12 +673,12 @@ void reload_ucode_amd(unsigned int cpu)static int collect_cpu_info_amd (int cpu,struct cpu_signature *csig) { -struct cpuinfo_x86 *c = &cpu_data (cpu);struct ucode_cpu_info *uci = ucode_cpu_info + cpu; +u32 dummy__always_unused ;struct ucode_patch *p; csig->sig =cpuid_eax (0x00000001 ); - csig->rev = c->microcode; +rdmsr (MSR_AMD64_PATCH_LEVEL , csig->rev, dummy);/* * a patch could have been loaded early, set uci->mc so that
PATCH v3 15/30 x86/microcode: Remove pointless apply() invocation
In reply to Thomas GleixnerMicrocode is applied on the APs during early bringup. There is no point in trying to apply the microcode again during the hotplug operations and neither at the point where the microcode device is initialized.
Collect CPU info and microcode revision in setup_online_cpu() for now. This will move to the CPU hotplug callback in the next step.
---
V2: New patch
--- arch/x86/kernel/cpu/microcode/core.c | 34 ++++++---------------------------- include/linux/cpuhotplug.h | 1 - 2 files changed, 6 insertions(+), 29 deletions(-)
arch/x86/kernel/cpu/microcode/core.c +6 -28
@@ -511,17 +511,6 @@ static void microcode_fini_cpu(int cpu) microcode_ops->microcode_fini_cpu (cpu); } -static enum ucode_state microcode_init_cpu (int cpu) -{ -struct ucode_cpu_info *uci = ucode_cpu_info + cpu; - -memset (uci,0 ,sizeof (*uci)); - - microcode_ops->collect_cpu_info (cpu, &uci->cpu_sig); - -return microcode_ops->apply_microcode (cpu); -} -/** * microcode_bsp_resume - Update boot CPU microcode during resume. */
@@ -540,15 +529,6 @@ static struct syscore_ops mc_syscore_ops .resume = microcode_bsp_resume, }; -static int mc_cpu_starting (unsigned int cpu) -{ -enum ucode_state err = microcode_ops->apply_microcode (cpu); - -pr_debug ("%s: CPU%d, err: %d\n" , __func__, cpu, err); - -return err ==UCODE_ERROR ; -} -static int mc_cpu_online (unsigned int cpu) {struct device *dev =get_cpu_device (cpu);
@@ -576,14 +556,14 @@ static int mc_cpu_down_prep(unsigned intstatic void setup_online_cpu (struct work_struct *work) {int cpu =smp_processor_id (); -enum ucode_state err; +struct ucode_cpu_info *uci = ucode_cpu_info + cpu; - err =microcode_init_cpu (cpu); -if (err ==UCODE_ERROR ) { -pr_err ("Error applying microcode on CPU%d\n" , cpu); -return ; - } +memset (uci,0 ,sizeof (*uci)); + microcode_ops->collect_cpu_info (cpu, &uci->cpu_sig); +cpu_data (cpu).microcode = uci->cpu_sig.rev; +if (!cpu) + boot_cpu_data.microcode = uci->cpu_sig.rev;mc_cpu_online (cpu); }
@@ -636,8 +616,6 @@ static int __init microcode_init(void)schedule_on_each_cpu (setup_online_cpu);register_syscore_ops (&mc_syscore_ops); -cpuhp_setup_state_nocalls (CPUHP_AP_MICROCODE_LOADER ,"x86/microcode:starting" , - mc_cpu_starting,NULL );cpuhp_setup_state_nocalls (CPUHP_AP_ONLINE_DYN ,"x86/microcode:online" , mc_cpu_online, mc_cpu_down_prep);
include/linux/cpuhotplug.h +0 -1
@@ -156,7 +156,6 @@ enum cpuhp_state {CPUHP_AP_IRQ_LOONGARCH_STARTING ,CPUHP_AP_IRQ_SIFIVE_PLIC_STARTING ,CPUHP_AP_ARM_MVEBU_COHERENCY , -CPUHP_AP_MICROCODE_LOADER ,CPUHP_AP_PERF_X86_AMD_UNCORE_STARTING ,CPUHP_AP_PERF_X86_STARTING , CPUHP_AP_PERF_X86_AMD_IBS_STARTING,
PATCH v3 16/30 x86/microcode: Get rid of the schedule work indirection
In reply to Thomas GleixnerScheduling work on all CPUs to collect the microcode information is just another extra step for no value. Let the CPU hotplug callback registration do it.
---
V2: New patch
--- arch/x86/kernel/cpu/microcode/core.c | 29 ++++++++++------------------- 1 file changed, 10 insertions(+), 19 deletions(-)
arch/x86/kernel/cpu/microcode/core.c +10 -19
@@ -531,8 +531,16 @@ static struct syscore_ops mc_syscore_opsstatic int mc_cpu_online (unsigned int cpu) { +struct ucode_cpu_info *uci = ucode_cpu_info + cpu;struct device *dev =get_cpu_device (cpu); +memset (uci,0 ,sizeof (*uci)); + + microcode_ops->collect_cpu_info (cpu, &uci->cpu_sig); +cpu_data (cpu).microcode = uci->cpu_sig.rev; +if (!cpu) + boot_cpu_data.microcode = uci->cpu_sig.rev; +if (sysfs_create_group (&dev->kobj, &mc_attr_group))pr_err ("Failed to create group for CPU%d\n" , cpu);return 0 ;
@@ -553,20 +561,6 @@ static int mc_cpu_down_prep(unsigned intreturn 0 ; } -static void setup_online_cpu (struct work_struct *work) -{ -int cpu =smp_processor_id (); -struct ucode_cpu_info *uci = ucode_cpu_info + cpu; - -memset (uci,0 ,sizeof (*uci)); - - microcode_ops->collect_cpu_info (cpu, &uci->cpu_sig); -cpu_data (cpu).microcode = uci->cpu_sig.rev; -if (!cpu) - boot_cpu_data.microcode = uci->cpu_sig.rev; -mc_cpu_online (cpu); -} -static struct attribute *cpu_root_microcode_attrs[] = { #ifdef CONFIG_MICROCODE_LATE_LOADING &dev_attr_reload.attr,
@@ -612,12 +606,9 @@ static int __init microcode_init(void) } } -/* Do per-CPU setup */ -schedule_on_each_cpu (setup_online_cpu); -register_syscore_ops (&mc_syscore_ops); -cpuhp_setup_state_nocalls (CPUHP_AP_ONLINE_DYN ,"x86/microcode:online" , - mc_cpu_online, mc_cpu_down_prep); +cpuhp_setup_state (CPUHP_AP_ONLINE_DYN ,"x86/microcode:online" , + mc_cpu_online, mc_cpu_down_prep);pr_info ("Microcode Update Driver: v%s." ,DRIVER_VERSION );
PATCH v3 17/30 x86/microcode: Clean up mc_cpu_down_prep()
In reply to Thomas GleixnerThis function has nothing to do with suspend. It's a hotplug callback. Remove the bogus comment.
Drop the pointless debug printk. The hotplug core provides tracepoints which track the invocation of those callbacks.
---
V2: New patch
--- arch/x86/kernel/cpu/microcode/core.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-)
arch/x86/kernel/cpu/microcode/core.c +1 -7
@@ -569,16 +569,10 @@ static int mc_cpu_online(unsigned int cpstatic int mc_cpu_down_prep (unsigned int cpu) { -struct device *dev; - - dev =get_cpu_device (cpu); +struct device *dev =get_cpu_device (cpu);microcode_fini_cpu (cpu); - -/* Suspend is in progress, only remove the interface */ sysfs_remove_group (&dev->kobj, &mc_attr_group); -pr_debug ("%s: CPU%d\n" , __func__, cpu); -return 0 ; }
PATCH v3 18/30 x86/microcode: Handle "nosmt" correctly
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
On CPUs where microcode loading is not NMI safe the SMT sibling which is parked in one of the play_dead() variants, these parked CPUs still react on NMIs. So if a NMI hits while the primary thread updates the microcode the resulting behaviour is undefined. The default play_dead() implementation on modern CPUs is using MWAIT, which is not guaranteed to be safe against an microcode update which affects MWAIT.
Take the cpus_booted_once_mask into account to detect this case and refuse to load late if the vendor specific driver does not advertise that late loading is NMI safe.
AMD stated that this is safe, so mark the AMD driver accordingly.
This requirement will be partially lifted in later changes.
--- arch/x86/Kconfig | 2 - arch/x86/kernel/cpu/microcode/amd.c | 9 +++-- arch/x86/kernel/cpu/microcode/core.c | 51 +++++++++++++++++++------------ arch/x86/kernel/cpu/microcode/internal.h | 13 +++---- 4 files changed, 44 insertions(+), 31 deletions(-)
---
arch/x86/Kconfig +1 -1
@@ -1314,7 +1314,7 @@ config MICROCODE config MICROCODE_LATE_LOADING bool "Late microcode loading (DANGEROUS)" default n - depends on MICROCODE + depends on MICROCODE && SMP help Loading microcode late, when the system is up and executing instructions is a tricky business and should be avoided if possible. Just the sequence
arch/x86/kernel/cpu/microcode/amd.c +5 -4
@@ -948,10 +948,11 @@ static void microcode_fini_cpu_amd(int c }static struct microcode_ops microcode_amd_ops = { - .request_microcode_fw = request_microcode_amd, - .collect_cpu_info = collect_cpu_info_amd, - .apply_microcode = apply_microcode_amd, - .microcode_fini_cpu = microcode_fini_cpu_amd, + .request_microcode_fw = request_microcode_amd, + .collect_cpu_info = collect_cpu_info_amd, + .apply_microcode = apply_microcode_amd, + .microcode_fini_cpu = microcode_fini_cpu_amd, + .nmi_safe = true, };struct microcode_ops *__init init_amd_microcode (void )
arch/x86/kernel/cpu/microcode/core.c +32 -19
70 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/internal.h +6 -7
@@ -20,18 +20,17 @@ enum ucode_state {struct microcode_ops {enum ucode_state (*request_microcode_fw)(int cpu,struct device *dev); -void (*microcode_fini_cpu)(int cpu);/* -* The generic 'microcode_core' part guarantees that -* the callbacks below run on a target cpu when they -* are being called. +* The generic 'microcode_core' part guarantees that the callbacks +* below run on a target cpu when they are being called. * See also the "Synchronization" section in microcode_core.c. */ -enum ucode_state (*apply_microcode)(int cpu); -int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); -void (*finalize_late_load)(int result); +enum ucode_state (*apply_microcode)(int cpu); +int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); +void (*finalize_late_load)(int result); +unsigned int nmi_safe :1 ; };extern struct ucode_cpu_info ucode_cpu_info[];
Borislav Petkov
maintainerRe: [patch V3 18/30] x86/microcode: Handle "nosmt" correctly
Just textual nitpicks. Otherwise looks nice.
Quoted 5 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
On CPUs where microcode loading is not NMI safe the SMT sibling which is parked in one of the
play_dead()variants, these parked CPUs still react on NMIs.
s/, these parked CPUs still react/still reacts/.
Simpler.
Quoted 4 lines
So if a NMI hits while the primary thread updates the microcode the resulting behaviour is undefined. The default
play_dead()implementation on modern CPUs is using MWAIT, which is not guaranteed to be safe against an microcode update which affects MWAIT.
s/an //
Quoted 8 lines
+/*
+ * Ensure that all required CPUs which are present and have been booted
+ * once are online.
+ *
+ * To pass this check, all primary threads must be online.
+ *
+ * If the microcode load is not safe against NMI then all SMT threads
+ * must be online as well because they still react on NMI when they are
s/react on NMI/react to NMIs/
Quoted 4 lines
+ * soft-offlined and parked in one of the play_dead() variants. So if a + * NMI hits while the primary thread updates the microcode the resulting + * behaviour is undefined. The default play_dead() implementation on + * modern CPUs is using MWAIT, which is also not guaranteed to be safe
s/is using/uses/
Quoted 5 lines
arch/x86/kernel/cpu/microcode/internal.h +0
-0@@ -20,18 +20,17 @@ enum ucode_state {struct microcode_ops {
arch/x86/kernel/cpu/microcode/internal.h enum ucode_state { · line 27
Quoted 10 lines
arch/x86/kernel/cpu/microcode/internal.h +2
-4enum ucode_state (*request_microcode_fw)(int cpu,struct device *dev); -void (*microcode_fini_cpu)(int cpu);/* -* The generic 'microcode_core' part guarantees that -* the callbacks below run on a target cpu when they -* are being called. +* The generic 'microcode_core' part guarantees that the callbacks +* below run on a target cpu when they are being called.
s/cpu/CPU/
while at it.
Quoted 10 lines
arch/x86/kernel/cpu/microcode/internal.h +4
-3* See also the "Synchronization" section in microcode_core.c. */ -enum ucode_state (*apply_microcode)(int cpu); -int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); -void (*finalize_late_load)(int result); +enum ucode_state (*apply_microcode)(int cpu); +int (*collect_cpu_info)(int cpu,struct cpu_signature *csig); +void (*finalize_late_load)(int result); + unsigned int nmi_safe : 1; };
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 19/30 x86/microcode: Clarify the late load logic
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
reload_store() is way too complicated. Split the inner workings out and make the following enhancements:
- Taint the kernel only when the microcode was actually updated. If. e.g. the rendevouz fails, then nothing happened and there is no reason for tainting.
- Return useful error codes
--- arch/x86/kernel/cpu/microcode/core.c | 39 +++++++++++++++-------------------- 1 file changed, 17 insertions(+), 22 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +17 -22
@@ -434,11 +434,11 @@ static int microcode_reload_late(void)pr_info ("Reload succeeded, microcode revision: 0x%x -> 0x%x\n" , old, boot_cpu_data.microcode);microcode_check (&prev_info); +add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); }else {pr_info ("Reload failed, current microcode revision: 0x%x\n" , boot_cpu_data.microcode); } -return ret; }
@@ -471,40 +471,35 @@ static bool ensure_cpus_are_online(void)return true; } +static int ucode_load_late_locked (void ) +{ +int ret; + +if (!ensure_cpus_are_online ()) +return -EBUSY ; + + ret = microcode_ops->request_microcode_fw (0 , µcode_pdev->dev); +if (ret !=UCODE_NEW ) +return ret ==UCODE_NFOUND ? -ENOENT : -EBADFD ; +return microcode_reload_late (); +} +static ssize_t reload_store (struct device *dev,struct device_attribute *attr,const char *buf,size_t size) { -enum ucode_state tmp_ret =UCODE_OK ; -int bsp = boot_cpu_data.cpu_index;unsigned long val; -ssize_t ret =0 ; +ssize_t ret; ret =kstrtoul (buf,0 , &val);if (ret || val !=1 )return -EINVAL ;cpus_read_lock (); - -if (!ensure_cpus_are_online ()) { - ret = -EBUSY ; - goto put; - } - - tmp_ret = microcode_ops->request_microcode_fw (bsp, µcode_pdev->dev); -if (tmp_ret !=UCODE_NEW ) - goto put; - - ret =microcode_reload_late (); -put: + ret =ucode_load_late_locked ();cpus_read_unlock (); -if (ret ==0 ) - ret = size; - -add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); - -return ret; +return ret ? : size; }static DEVICE_ATTR_WO (reload);
Borislav Petkov
maintainerRe: [patch V3 19/30] x86/microcode: Clarify the late load logic
arch/x86/kernel/cpu/microcode/core.c static bool ensure_cpus_are_online(void) · line 483
Quoted 3 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +3
-0+ ret = microcode_ops->request_microcode_fw (0 , µcode_pdev->dev); +if (ret !=UCODE_NEW ) +return ret ==UCODE_NFOUND ? -ENOENT : -EBADFD ;
I know this is short but let's make it more boring and readable pls:
if (ret != UCODE_NEW) {
if (ret == UCODE_NFOUND)
return -ENOENT;
else
return -EBADFD;
}Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 20/30 x86/microcode: Sanitize __wait_for_cpus()
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
The code is too complicated for no reason:
- The return value is pointless as this is a strict boolean.
- It's way simpler to count down from num_online_cpus() and check for zero.
- The timeout argument is pointless as this is always one second.
- Touching the NMI watchdog every 100ns does not make any sense, neither does checking every 100ns. This is really not a hotpath operation.
Preload the atomic counter with the number of online CPUs and simplify the whole timeout logic. Delay for one microsecond and touch the NMI watchdog once per millisecond.
--- arch/x86/kernel/cpu/microcode/core.c | 41 ++++++++++++++--------------------- 1 file changed, 17 insertions(+), 24 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +17 -24
@@ -324,31 +324,24 @@ static struct platform_device *microcode* requirement can be relaxed in the future. Right now, this is conservative * and good. */ -#define SPINUNIT 100/* 100 nsec */ +static atomic_t late_cpus_in, late_cpus_out; - -static atomic_t late_cpus_in; -static atomic_t late_cpus_out; - -static int __wait_for_cpus (atomic_t *t,long long timeout) +static bool wait_for_cpus (atomic_t *cnt) { -int all_cpus =num_online_cpus (); - -atomic_inc (t); - -while (atomic_read (t) < all_cpus) { -if (timeout <SPINUNIT ) { -pr_err ("Timeout while waiting for CPUs rendezvous, remaining: %d\n" , - all_cpus -atomic_read (t)); -return 1 ; - } +unsigned int timeout; -ndelay (SPINUNIT ); - timeout -=SPINUNIT ; +WARN_ON_ONCE (atomic_dec_return (cnt) <0 ); -touch_nmi_watchdog (); +for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { +if (!atomic_read (cnt)) +return true; +udelay (1 ); +if (!(timeout %1000 )) +touch_nmi_watchdog (); } -return 0 ; +/* Prevent the late comers to make progress and let them time out */ +atomic_inc (cnt); +return false; }/*
@@ -366,7 +359,7 @@ static int __reload_late(void *info)* Wait for all CPUs to arrive. A load will not be attempted unless all * CPUs show up. * */ -if (__wait_for_cpus (&late_cpus_in,NSEC_PER_SEC )) +if (!wait_for_cpus (&late_cpus_in))return -1 ;/*
@@ -389,7 +382,7 @@ static int __reload_late(void *info) } wait_for_siblings: -if (__wait_for_cpus (&late_cpus_out,NSEC_PER_SEC )) +if (!wait_for_cpus (&late_cpus_out))panic ("Timeout during microcode update!\n" );/*
@@ -416,8 +409,8 @@ static int microcode_reload_late(void)pr_err ("Attempting late microcode loading - it is dangerous and taints the kernel.\n" );pr_err ("You should switch to early loading, if possible.\n" ); -atomic_set (&late_cpus_in,0 ); -atomic_set (&late_cpus_out,0 ); +atomic_set (&late_cpus_in,num_online_cpus ()); +atomic_set (&late_cpus_out,num_online_cpus ());/* * Take a snapshot before the microcode update in order to compare and
Borislav Petkov
maintainerRe: [patch V3 20/30] x86/microcode: Sanitize __wait_for_cpus()
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 337
Quoted 3 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +3
-0+for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { +if (!atomic_read (cnt)) +return true;
<---- newline here.
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 338
Quoted 1 line
arch/x86/kernel/cpu/microcode/core.c +1
-0+udelay (1 );
And here.
Otherwise it looks too crammed.
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 339
Quoted 1 line
arch/x86/kernel/cpu/microcode/core.c +1
-0+if (!(timeout %1000 ))
MSEC_PER_SEC - no naked numbers pls.
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 342
Quoted 4 lines
arch/x86/kernel/cpu/microcode/core.c +2
-1+touch_nmi_watchdog (); } -return 0 ; +/* Prevent the late comers to make progress and let them time out */
s/to make progress/from making progress/
Nice, otherwise.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 20/30] x86/microcode: Sanitize __wait_for_cpus()
Quoted 7 lines · Borislav Petkov
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 337
Quoted 3 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +3
-0+for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { +if (!atomic_read (cnt)) +return true;<---- newline here.
Quoted 5 lines
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 338
Quoted 1 line
arch/x86/kernel/cpu/microcode/core.c +1
-0+udelay (1 );And here.
Otherwise it looks too crammed.
Oh well.
Quoted 3 lines
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 339
Quoted 1 line
arch/x86/kernel/cpu/microcode/core.c +1
-0+if (!(timeout %1000 ))
MSEC_PER_SEC- no naked numbers pls.
MSEC_PER_SEC? Thats really wrong because timeout counts in microseconds, no? So USEC_PER_MSEC.
Quoted 8 lines
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 342
Quoted 4 lines
arch/x86/kernel/cpu/microcode/core.c +2
-1+touch_nmi_watchdog (); } -return 0 ; +/* Prevent the late comers to make progress and let them time out */ s/to make progress/from making progress/
Nice, otherwise.
Borislav Petkov
maintainerRe: [patch V3 20/30] x86/microcode: Sanitize __wait_for_cpus()
Quoted 2 lines · Thomas Gleixner
MSEC_PER_SEC? Thats really wrong because timeout counts in microseconds, no? SoUSEC_PER_MSEC.
Doh, yes.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 21/30 x86/microcode: Add per CPU result state
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
The microcode rendevouz is purely acting on global state, which does not allow to analyze fails in a coherent way.
Introduce per CPU state where the results are written into, which allows to analyze the return codes of the individual CPUs.
Initialize the state when walking the cpu_present_mask in the online check to avoid another for_each_cpu() loop.
Enhance the result print out with that.
The structure is intentionally named ucode_ctrl as it will gain control fields in subsequent changes.
--- arch/x86/kernel/cpu/microcode/core.c | 108 ++++++++++++++++++------------- arch/x86/kernel/cpu/microcode/internal.h | 1 2 files changed, 65 insertions(+), 44 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +64 -44
@@ -324,6 +324,11 @@ static struct platform_device *microcode* requirement can be relaxed in the future. Right now, this is conservative * and good. */ +struct ucode_ctrl { +enum ucode_state result; +}; + +static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl);static atomic_t late_cpus_in, late_cpus_out;static bool wait_for_cpus (atomic_t *cnt)
@@ -344,23 +349,19 @@ static bool wait_for_cpus(atomic_t *cnt)return false; } -/* -* Returns: -* < 0 - on error -* 0 - success (no update done or microcode was updated) -*/ -static int __reload_late (void *info) +static int ucode_load_cpus_stopped (void *unused) {int cpu =smp_processor_id (); -enum ucode_state err; -int ret =0 ; +enum ucode_state ret;/* * Wait for all CPUs to arrive. A load will not be attempted unless all * CPUs show up. * */ -if (!wait_for_cpus (&late_cpus_in)) -return -1 ; +if (!wait_for_cpus (&late_cpus_in)) { +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +return 0 ; + }/* * On an SMT system, it suffices to load the microcode on one sibling of
@@ -369,17 +370,11 @@ static int __reload_late(void *info)* loading attempts happen on multiple threads of an SMT core. See * below. */ -if (cpumask_first (topology_sibling_cpumask (cpu)) == cpu) - err = microcode_ops->apply_microcode (cpu); -else +if (cpumask_first (topology_sibling_cpumask (cpu)) != cpu) goto wait_for_siblings; -if (err >=UCODE_NFOUND ) { -if (err ==UCODE_ERROR ) { -pr_warn ("Error reloading microcode on CPU %d\n" , cpu); - ret =-1 ; - } - } + ret = microcode_ops->apply_microcode (cpu); +this_cpu_write (ucode_ctrl.result, ret); wait_for_siblings:if (!wait_for_cpus (&late_cpus_out))
@@ -391,19 +386,18 @@ static int __reload_late(void *info)* per-cpu cpuinfo can be updated with right microcode * revision. */ -if (cpumask_first (topology_sibling_cpumask (cpu)) != cpu) - err = microcode_ops->apply_microcode (cpu); +if (cpumask_first (topology_sibling_cpumask (cpu)) == cpu) +return 0 ; -return ret; + ret = microcode_ops->apply_microcode (cpu); +this_cpu_write (ucode_ctrl.result, ret); +return 0 ; } -/* -* Reload microcode late on all CPUs. Wait for a sec until they -* all gather together. -*/ -static int microcode_reload_late (void ) +static int ucode_load_late_stop_cpus (void ) { -int old = boot_cpu_data.microcode, ret; +unsigned int cpu, updated =0 , failed =0 , timedout =0 , siblings =0 ; +int old_rev = boot_cpu_data.microcode;struct cpuinfo_x86 prev_info;pr_err ("Attempting late microcode loading - it is dangerous and taints the kernel.\n" );
@@ -418,26 +412,47 @@ static int microcode_reload_late(void)*/ store_cpu_caps (&prev_info); - ret =stop_machine_cpuslocked (__reload_late,NULL , cpu_online_mask); +stop_machine_cpuslocked (ucode_load_cpus_stopped,NULL , cpu_online_mask); + +/* Analyze the results */ +for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) { +switch (per_cpu (ucode_ctrl.result, cpu)) { +case UCODE_UPDATED : updated++;break ; +case UCODE_TIMEOUT : timedout++;break ; +case UCODE_OK : siblings++;break ; +default : failed++;break ; + } + }if (microcode_ops->finalize_late_load) - microcode_ops->finalize_late_load (ret); + microcode_ops->finalize_late_load (!updated); -if (!ret) { -pr_info ("Reload succeeded, microcode revision: 0x%x -> 0x%x\n" , - old, boot_cpu_data.microcode); -microcode_check (&prev_info); -add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); - }else { -pr_info ("Reload failed, current microcode revision: 0x%x\n" , - boot_cpu_data.microcode); +if (!updated) { +/* Nothing changed. */ +if (!failed && !timedout) +return 0 ; +pr_err ("Microcode update failed: %u CPUs failed %u CPUs timed out\n" , + failed, timedout); +return -EIO ; } -return ret; + +add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); +pr_info ("Microcode load: updated on %u primary CPUs with %u siblings\n" , updated, siblings); +if (failed || timedout) { +pr_err ("Microcode load incomplete. %u CPUs timed out or failed\n" , +num_online_cpus () - (updated + siblings)); + } +pr_info ("Microcode revision: 0x%x -> 0x%x\n" , old_rev, boot_cpu_data.microcode); +microcode_check (&prev_info); + +return updated + siblings ==num_online_cpus () ?0 : -EIO ; }/* -* Ensure that all required CPUs which are present and have been booted -* once are online. +* This function does two things: +* +* 1) Ensure that all required CPUs which are present and have been booted +* once are online. * * To pass this check, all primary threads must be online. *
@@ -448,9 +463,12 @@ static int microcode_reload_late(void)* behaviour is undefined. The default play_dead() implementation on * modern CPUs is using MWAIT, which is also not guaranteed to be safe * against a microcode update which affects MWAIT. +* +* 2) Initialize the per CPU control structure */ -static bool ensure_cpus_are_online (void ) +static bool ucode_setup_cpus (void ) { +struct ucode_ctrl ctrl = { .result =-1 , };unsigned int cpu;for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) {
@@ -460,6 +478,8 @@ static bool ensure_cpus_are_online(void)return false; } } +/* Initialize the per CPU state */ +per_cpu (ucode_ctrl, cpu) = ctrl; }return true; }
@@ -468,13 +488,13 @@ static int ucode_load_late_locked(void) {int ret; -if (!ensure_cpus_are_online ()) +if (!ucode_setup_cpus ())return -EBUSY ; ret = microcode_ops->request_microcode_fw (0 , µcode_pdev->dev);if (ret !=UCODE_NEW )return ret ==UCODE_NFOUND ? -ENOENT : -EBADFD ; -return microcode_reload_late (); +return ucode_load_late_stop_cpus (); } static ssize_t reload_store(struct device *dev,
arch/x86/kernel/cpu/microcode/internal.h +1 -0
@@ -16,6 +16,7 @@ enum ucode_state {UCODE_UPDATED ,UCODE_NFOUND ,UCODE_ERROR , + UCODE_TIMEOUT, }; struct microcode_ops {
Borislav Petkov
maintainerRe: [patch V3 21/30] x86/microcode: Add per CPU result state
Quoted 3 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
The microcode rendevouz is purely acting on global state, which does not
rendezvous
Quoted 23 lines
allow to analyze fails in a coherent way.
Introduce per CPU state where the results are written into, which allows to analyze the return codes of the individual CPUs.
Initialize the state when walking the
cpu_present_maskin the online check to avoid anotherfor_each_cpu()loop.Enhance the result print out with that.
The structure is intentionally named
ucode_ctrlas it will gain control fields in subsequent changes.Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/kernel/cpu/microcode/core.c | 108 ++++++++++++++++++------------- arch/x86/kernel/cpu/microcode/internal.h | 1 2 files changed, 65 insertions(+), 44 deletions(-)---
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 327
Quoted 4 lines
arch/x86/kernel/cpu/microcode/core.c +1
-0@@ -324,6 +324,11 @@ static struct platform_device *microcode* requirement can be relaxed in the future. Right now, this is conservative * and good. */ +struct ucode_ctrl {
microcode_ctrl
I know "ucode" is shorter but we already call everything new-er "microcode_" and this'll cause confusion.
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 331
Quoted 4 lines
arch/x86/kernel/cpu/microcode/core.c +4
-0+enum ucode_state result; +}; + +static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl);
You could do
static DEFINE_PER_CPU(struct microcode_ctrl, ucode_ctrl);
so that the naming is different too.
Quoted 4 lines
arch/x86/kernel/cpu/microcode/core.c +0
-0static atomic_t late_cpus_in, late_cpus_out;static bool wait_for_cpus (atomic_t *cnt)
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 352
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +1
-6@@ -344,23 +349,19 @@ static bool wait_for_cpus(atomic_t *cnt)return false; } -/* -* Returns: -* < 0 - on error -* 0 - success (no update done or microcode was updated) -*/ -static int __reload_late (void *info) +static int ucode_load_cpus_stopped (void *unused)
No need for "ucode_" prefixes to static functions.
Quoted 5 lines
arch/x86/kernel/cpu/microcode/core.c +1
-2{int cpu =smp_processor_id (); -enum ucode_state err; -int ret =0 ; +enum ucode_state ret;
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 363
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +3
-2/* * Wait for all CPUs to arrive. A load will not be attempted unless all * CPUs show up. * */ -if (!wait_for_cpus (&late_cpus_in)) -return -1 ; +if (!wait_for_cpus (&late_cpus_in)) { +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +return 0 ;
So the only value this function returns is 0 now. stop_machine_cpuslocked() still does pay attention at ret so I guess it should return non-null/negative on error or so?
Quoted 34 lines
arch/x86/kernel/cpu/microcode/core.c +6
-12/* * On an SMT system, it suffices to load the microcode on one sibling of @@ -369,17 +370,11 @@ static int __reload_late(void *info)* loading attempts happen on multiple threads of an SMT core. See * below. */ -if (cpumask_first (topology_sibling_cpumask (cpu)) == cpu) - err = microcode_ops->apply_microcode (cpu); -else +if (cpumask_first (topology_sibling_cpumask (cpu)) != cpu) goto wait_for_siblings; -if (err >=UCODE_NFOUND ) { -if (err ==UCODE_ERROR ) { -pr_warn ("Error reloading microcode on CPU %d\n" , cpu); - ret =-1 ; - } - } + ret = microcode_ops->apply_microcode (cpu); +this_cpu_write (ucode_ctrl.result, ret); wait_for_siblings:if (!wait_for_cpus (&late_cpus_out))@@ -391,19 +386,18 @@ static int __reload_late(void *info)* per-cpu cpuinfo can be updated with right microcode * revision. */ -if (cpumask_first (topology_sibling_cpumask (cpu)) != cpu) - err = microcode_ops->apply_microcode (cpu); +if (cpumask_first (topology_sibling_cpumask (cpu)) == cpu) +return 0 ; -return ret; + ret = microcode_ops->apply_microcode (cpu);
arch/x86/kernel/cpu/microcode/core.c static int __reload_late(void *info) · line 397
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +3
-5+this_cpu_write (ucode_ctrl.result, ret); +return 0 ; } -/* -* Reload microcode late on all CPUs. Wait for a sec until they -* all gather together. -*/ -static int microcode_reload_late (void ) +static int ucode_load_late_stop_cpus (void )
s/ucode_//
Quoted 12 lines
arch/x86/kernel/cpu/microcode/core.c +2
-2{ -int old = boot_cpu_data.microcode, ret; +unsigned int cpu, updated =0 , failed =0 , timedout =0 , siblings =0 ; +int old_rev = boot_cpu_data.microcode;struct cpuinfo_x86 prev_info;pr_err ("Attempting late microcode loading - it is dangerous and taints the kernel.\n" );@@ -418,26 +412,47 @@ static int microcode_reload_late(void)*/ store_cpu_caps (&prev_info); - ret =stop_machine_cpuslocked (__reload_late,NULL , cpu_online_mask);
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 424
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +10
-0+stop_machine_cpuslocked (ucode_load_cpus_stopped,NULL , cpu_online_mask); + +/* Analyze the results */ +for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) { +switch (per_cpu (ucode_ctrl.result, cpu)) { +case UCODE_UPDATED : updated++;break ; +case UCODE_TIMEOUT : timedout++;break ; +case UCODE_OK : siblings++;break ; +default : failed++;break ; + }
Align vertically.
Quoted 21 lines
arch/x86/kernel/cpu/microcode/core.c +9
-9+ }if (microcode_ops->finalize_late_load) - microcode_ops->finalize_late_load (ret); + microcode_ops->finalize_late_load (!updated); -if (!ret) { -pr_info ("Reload succeeded, microcode revision: 0x%x -> 0x%x\n" , - old, boot_cpu_data.microcode); -microcode_check (&prev_info); -add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); - }else { -pr_info ("Reload failed, current microcode revision: 0x%x\n" , - boot_cpu_data.microcode); +if (!updated) { +/* Nothing changed. */ +if (!failed && !timedout) +return 0 ; +pr_err ("Microcode update failed: %u CPUs failed %u CPUs timed out\n" , + failed, timedout); +return -EIO ;
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 445
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +8
-1} -return ret; + +add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); +pr_info ("Microcode load: updated on %u primary CPUs with %u siblings\n" , updated, siblings); +if (failed || timedout) { +pr_err ("Microcode load incomplete. %u CPUs timed out or failed\n" , +num_online_cpus () - (updated + siblings)); + } +pr_info ("Microcode revision: 0x%x -> 0x%x\n" , old_rev, boot_cpu_data.microcode);
You don't need "Microcode" in those strings - the pr_info has already "microcode:" as prefix.
Quoted 16 lines
arch/x86/kernel/cpu/microcode/core.c +7
-2+microcode_check (&prev_info); + +return updated + siblings ==num_online_cpus () ?0 : -EIO ; }/* -* Ensure that all required CPUs which are present and have been booted -* once are online. +* This function does two things: +* +* 1) Ensure that all required CPUs which are present and have been booted +* once are online. * * To pass this check, all primary threads must be online. *
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 469
Quoted 8 lines
arch/x86/kernel/cpu/microcode/core.c +3
-1@@ -448,9 +463,12 @@ static int microcode_reload_late(void)* behaviour is undefined. The default play_dead() implementation on * modern CPUs is using MWAIT, which is also not guaranteed to be safe * against a microcode update which affects MWAIT. +* +* 2) Initialize the per CPU control structure */ -static bool ensure_cpus_are_online (void ) +static bool ucode_setup_cpus (void )
s/ucode_//
Quoted 5 lines
arch/x86/kernel/cpu/microcode/core.c +1
-0{ +struct ucode_ctrl ctrl = { .result =-1 , };unsigned int cpu;for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) {
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 21/30] x86/microcode: Add per CPU result state
Quoted 7 lines · Borislav Petkov
Quoted 1 line · Thomas Gleixner
+struct
ucode_ctrl{
microcode_ctrlI know "ucode" is shorter but we already call everything new-er "
microcode_" and this'll cause confusion.
That starts to get silly. The struct is used only in the microcode realm and nothing which is globally visible. ucode is a pretty obvious and established shortcut. But so what....
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c static struct platform_device *microcode · line 331
Quoted 4 lines
arch/x86/kernel/cpu/microcode/core.c +4
-0+enum ucode_state result; +}; + +static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl);You could do
static
DEFINE_PER_CPU(struct microcode_ctrl, ucode_ctrl);so that the naming is different too.
And that solves what?
Quoted 7 lines
Quoted 7 lines
static
atomic_tlate_cpus_in,late_cpus_out;static bool
wait_for_cpus(atomic_t *cnt)arch/x86/kernel/cpu/microcode/core.c +0
-0@@ -344,23 +349,19 @@ static bool wait_for_cpus(atomic_t *cnt)return false; }
Quoted 9 lines
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 352
Quoted 7 lines
arch/x86/kernel/cpu/microcode/core.c +1
-6-/* -* Returns: -* < 0 - on error -* 0 - success (no update done or microcode was updated) -*/ -static int __reload_late (void *info) +static int ucode_load_cpus_stopped (void *unused)No need for "
ucode_" prefixes to static functions.
What's the problem with that prefix? The function name clearly says what this is doing.
Quoted 6 lines
Quoted 6 lines
arch/x86/kernel/cpu/microcode/core.c +1
-2{int cpu =smp_processor_id (); -enum ucode_state err; -int ret =0 ; +enum ucode_state ret;
Quoted 13 lines
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 363
Quoted 9 lines
arch/x86/kernel/cpu/microcode/core.c +3
-2/* * Wait for all CPUs to arrive. A load will not be attempted unless all * CPUs show up. * */ -if (!wait_for_cpus (&late_cpus_in)) -return -1 ; +if (!wait_for_cpus (&late_cpus_in)) { +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +return 0 ;So the only value this function returns is 0 now.
stop_machine_cpuslocked()still does pay attention at ret so I guess it should return non-null/negative on error or so?
Nope, because stop_machine_cpuslocked() does not usefully accumulate results from all involved CPUs. But it can return errors related to the invocation itself, which is a completely different story.
That's why ucode_ctrl.result is per CPU and has to be evaluated separately.
Quoted 5 lines
Quoted 5 lines
arch/x86/kernel/cpu/microcode/core.c +4
-1- ret =stop_machine_cpuslocked (__reload_late,NULL , cpu_online_mask); +stop_machine_cpuslocked (ucode_load_cpus_stopped,NULL , cpu_online_mask); + +/* Analyze the results */ +for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) {
Quoted 8 lines
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 424
Quoted 6 lines
arch/x86/kernel/cpu/microcode/core.c +6
-0+switch (per_cpu (ucode_ctrl.result, cpu)) { +case UCODE_UPDATED : updated++;break ; +case UCODE_TIMEOUT : timedout++;break ; +case UCODE_OK : siblings++;break ; +default : failed++;break ; + }Align vertically.
Align what?
Quoted 26 lines
Quoted 26 lines
arch/x86/kernel/cpu/microcode/core.c +12
-10+ }if (microcode_ops->finalize_late_load) - microcode_ops->finalize_late_load (ret); + microcode_ops->finalize_late_load (!updated); -if (!ret) { -pr_info ("Reload succeeded, microcode revision: 0x%x -> 0x%x\n" , - old, boot_cpu_data.microcode); -microcode_check (&prev_info); -add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); - }else { -pr_info ("Reload failed, current microcode revision: 0x%x\n" , - boot_cpu_data.microcode); +if (!updated) { +/* Nothing changed. */ +if (!failed && !timedout) +return 0 ; +pr_err ("Microcode update failed: %u CPUs failed %u CPUs timed out\n" , + failed, timedout); +return -EIO ; } -return ret; + +add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); +pr_info ("Microcode load: updated on %u primary CPUs with %u siblings\n" , updated, siblings);
Quoted 8 lines
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 445
Quoted 5 lines
arch/x86/kernel/cpu/microcode/core.c +5
-0+if (failed || timedout) { +pr_err ("Microcode load incomplete. %u CPUs timed out or failed\n" , +num_online_cpus () - (updated + siblings)); + } +pr_info ("Microcode revision: 0x%x -> 0x%x\n" , old_rev, boot_cpu_data.microcode);You don't need "Microcode" in those strings - the
pr_infohas already "microcode:" as prefix.
True.
Quoted 11 lines
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 469
Quoted 9 lines
arch/x86/kernel/cpu/microcode/core.c +3
-1@@ -448,9 +463,12 @@ static int microcode_reload_late(void)* behaviour is undefined. The default play_dead() implementation on * modern CPUs is using MWAIT, which is also not guaranteed to be safe * against a microcode update which affects MWAIT. +* +* 2) Initialize the per CPU control structure */ -static bool ensure_cpus_are_online (void ) +static bool ucode_setup_cpus (void )s/ucode_//
and setup_cpus() then tells what?
Borislav Petkov
maintainerRe: [patch V3 21/30] x86/microcode: Add per CPU result state
Quoted 3 lines · Thomas Gleixner
That starts to get silly. The struct is used only in the microcode realm and nothing which is globally visible. ucode is a pretty obvious and established shortcut. But so what....
Ok, which prefix do you propose?
"microcode_", "ucode_"?
And I chose "microcode_" a while back and planned on converting stuff gradually when touching the code and not do solely a renaming patch.
All I'm saying is, we should be consistent.
Quoted 7 lines
Quoted 5 lines
You could do
static
DEFINE_PER_CPU(struct microcode_ctrl, ucode_ctrl);so that the naming is different too.
And that solves what?
I find it somewhat confusing when the variable name is called the same name as the struct and I try to have the struct names be more expressive than the variables of the same type.
But not a big deal.
Quoted 4 lines
Quoted 1 line
No need for "
ucode_" prefixes to static functions.What's the problem with that prefix? The function name clearly says what this is doing.
Giving proper prefixes only to the externally visible functions is, I think, a nice way of showing what is what. The static, internally used symbols, OTOH, don't need a prefix and when you look at the name, you know immediately whether it is a static symbol or an externally visible and potentially used by other things. We do that already for other code, like global variables, for example.
Quoted 3 lines
Nope, because
stop_machine_cpuslocked()does not usefully accumulate results from all involved CPUs. But it can return errors related to the invocation itself, which is a completely different story.
Ah, I see what you mean:
" * RETURNS:
* -ENOENT if @fn(@arg) was not executed at all because all cpus in
* @cpumask were offline; otherwise, 0 if all executions of @fn
* returned 0, any non zero return value if any returned non zero."
So we have to return 0 here. Oh well.
Quoted 2 lines
That's why ucode_ctrl.result is per CPU and has to be evaluated separately.
Right.
Quoted 11 lines
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c static int microcode_reload_late(void) · line 424
Quoted 8 lines
arch/x86/kernel/cpu/microcode/core.c +8
-0+/* Analyze the results */ +for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) { +switch (per_cpu (ucode_ctrl.result, cpu)) { +case UCODE_UPDATED : updated++;break ; +case UCODE_TIMEOUT : timedout++;break ; +case UCODE_OK : siblings++;break ; +default : failed++;break ; + }Align vertically.
Quoted 1 line
Align what?
switch (per_cpu(ucode_ctrl.result, cpu)) {
case UCODE_UPDATED: updated++; break;
case UCODE_TIMEOUT: timedout++; break;
case UCODE_OK: siblings++; break;
default: failed++; break;But meh, it's ok either way.
Quoted 1 line
and
setup_cpus()then tells what?
See above. I think there's a merit in distinguishing the symbol scope based on the naming only but I'm sure you have an opinion... :-)
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 22/30 x86/microcode: Add per CPU control field
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Add a per CPU control field to ucode_ctrl and define constants for it:
SCTRL_WAIT indicates that the CPU needs to spinwait with timeoutSCTRL_APPLY indicates that the CPU needs to invoke the microcode_apply()
callbackSCTRL_DONE indicates that the CPU can proceed without invoking the
microcode_apply() callback.
In theory this could be a global control field, but a global control does not cover the following case:
15 primary CPUs load microcode successfully
1 primary CPU fails and returns with an error code
With global control the sibling of the failed CPU would either try again or the whole operation would be aborted with the consequence that the 15 siblings do not invoke the apply path and end up with inconsistent software state. The result in dmesg would be inconsistent too.
There are two additional fields added and initialized:
ctrl_cpu and secondaries. ctrl_cpu is the CPU number of the primary thread for now, but with the upcoming uniform loading at package or system scope this will be one CPU per package or just one CPU. Secondaries hands the control CPU a CPU mask which will be required to release the secondary CPUs out of the wait loop.
Preparatory change for implementing a properly split control flow for primary and secondary CPUs.
--- arch/x86/kernel/cpu/microcode/core.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +18 -2
@@ -324,8 +324,16 @@ static struct platform_device *microcode* requirement can be relaxed in the future. Right now, this is conservative * and good. */ +enum sibling_ctrl { +SCTRL_WAIT , +SCTRL_APPLY , +SCTRL_DONE , +}; +struct ucode_ctrl { +enum sibling_ctrl ctrl;enum ucode_state result; +unsigned int ctrl_cpu; }; static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl);
@@ -468,7 +476,7 @@ static int ucode_load_late_stop_cpus(voi*/ static bool ucode_setup_cpus (void ) { -struct ucode_ctrl ctrl = { .result =-1 , }; +struct ucode_ctrl ctrl = { .ctrl =SCTRL_WAIT , .result =-1 , };unsigned int cpu;for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) {
@@ -478,7 +486,15 @@ static bool ucode_setup_cpus(void)return false; } } -/* Initialize the per CPU state */ + +/* +* Initialize the per CPU state. This is core scope for now, +* but prepared to take package or system scope into account. +*/ +if (topology_is_primary_thread (cpu)) + ctrl.ctrl_cpu = cpu; +else + ctrl.ctrl_cpu =cpumask_first (topology_sibling_cpumask (cpu));per_cpu (ucode_ctrl, cpu) = ctrl; }return true;
Borislav Petkov
maintainerRe: [patch V3 22/30] x86/microcode: Add per CPU control field
Quoted 9 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
Add a per CPU control field to
ucode_ctrland define constants for it:
SCTRL_WAITindicates that the CPU needs to spinwait with timeoutSCTRL_APPLYindicates that the CPU needs to invoke themicrocode_apply()
callbackSCTRL_DONEindicates that the CPU can proceed without invoking the
microcode_apply()callback.
Can we put those explanations over the enum definition pls?
Also, s/indicates that //g when you do.
Quoted 11 lines
In theory this could be a global control field, but a global control does not cover the following case:
15 primary CPUs load microcode successfully
1 primary CPU fails and returns with an error codeWith global control the sibling of the failed CPU would either try again or the whole operation would be aborted with the consequence that the 15 siblings do not invoke the apply path and end up with inconsistent software state. The result in dmesg would be inconsistent too.
Quoted 7 lines
There are two additional fields added and initialized:
ctrl_cpuand secondaries.ctrl_cpuis the CPU number of the primary thread for now, but with the upcoming uniform loading at package or system scope this will be one CPU per package or just one CPU. Secondaries hands the control CPU a CPU mask which will be required to release the secondary CPUs out of the wait loop.
Also as a comment above their definitions pls.
Quoted 42 lines
Preparatory change for implementing a properly split control flow for primary and secondary CPUs.
Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/kernel/cpu/microcode/core.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-)---
arch/x86/kernel/cpu/microcode/core.c +9
-1@@ -324,8 +324,16 @@ static struct platform_device *microcode* requirement can be relaxed in the future. Right now, this is conservative * and good. */ +enum sibling_ctrl { +SCTRL_WAIT , +SCTRL_APPLY , +SCTRL_DONE , +}; +struct ucode_ctrl { +enum sibling_ctrl ctrl;enum ucode_state result; +unsigned int ctrl_cpu; }; static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl);@@ -468,7 +476,7 @@ static int ucode_load_late_stop_cpus(voi*/ static bool ucode_setup_cpus (void ) { -struct ucode_ctrl ctrl = { .result =-1 , }; +struct ucode_ctrl ctrl = { .ctrl =SCTRL_WAIT , .result =-1 , };unsigned int cpu;for_each_cpu_and (cpu, cpu_present_mask, &cpus_booted_once_mask) {@@ -478,7 +486,15 @@ static bool ucode_setup_cpus(void)return false; } }
arch/x86/kernel/cpu/microcode/core.c static bool ucode_setup_cpus(void) · line 497
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +9
-1-/* Initialize the per CPU state */ + +/* +* Initialize the per CPU state. This is core scope for now, +* but prepared to take package or system scope into account. +*/ +if (topology_is_primary_thread (cpu)) + ctrl.ctrl_cpu = cpu; +else + ctrl.ctrl_cpu =cpumask_first (topology_sibling_cpumask (cpu));
<---- newline here.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 23/30 x86/microcode: Provide new control functions
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
The current all in one code is unreadable and really not suited for adding future features like uniform loading with package or system scope.
Provide a set of new control functions which split the handling of the primary and secondary CPUs. These will replace the current rendevouz all in one function in the next step. This is intentionally a separate change because diff makes an complete unreadable mess otherwise.
So the flow separates the primary and the secondary CPUs into their own functions, which use the control field in the per CPU ucode_ctrl struct.
primary() secondary()
wait_for_all() wait_for_all()
apply_ucode() wait_for_release()
release() apply_ucode()--- arch/x86/kernel/cpu/microcode/core.c | 86 +++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+)
---
arch/x86/kernel/cpu/microcode/core.c +86 -0
@@ -357,6 +357,92 @@ static bool wait_for_cpus(atomic_t *cnt)return false; } +static bool wait_for_ctrl (void ) +{ +unsigned int timeout; + +for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { +if (this_cpu_read (ucode_ctrl.ctrl) !=SCTRL_WAIT ) +return true; +udelay (1 ); +if (!(timeout %1000 )) +touch_nmi_watchdog (); + } +return false; +} + +static __maybe_unused void ucode_load_secondary (unsigned int cpu) +{ +unsigned int ctrl_cpu =this_cpu_read (ucode_ctrl.ctrl_cpu); +enum ucode_state ret; + +/* Initial rendevouz to ensure that all CPUs have arrived */ +if (!wait_for_cpus (&late_cpus_in)) { +pr_err_once ("Microcode load: %d CPUs timed out\n" , +atomic_read (&late_cpus_in) -1 ); +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +return ; + } + +/* +* Wait for primary threads to complete. If one of them hangs due +* to the update, there is no way out. This is non-recoverable +* because the CPU might hold locks or resources and confuse the +* scheduler, watchdogs etc. There is no way to safely evacuate the +* machine. +*/ +if (!wait_for_ctrl ()) +panic ("Microcode load: Primary CPU %d timed out\n" , ctrl_cpu); + +/* +* If the primary succeeded then invoke the apply() callback, +* otherwise copy the state from the primary thread. +*/ +if (this_cpu_read (ucode_ctrl.ctrl) ==SCTRL_APPLY ) + ret = microcode_ops->apply_microcode (cpu); +else + ret =per_cpu (ucode_ctrl.result, ctrl_cpu); + +this_cpu_write (ucode_ctrl.result, ret); +this_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE ); +} + +static __maybe_unused void ucode_load_primary (unsigned int cpu) +{ +struct cpumask *secondaries =topology_sibling_cpumask (cpu); +enum sibling_ctrl ctrl; +enum ucode_state ret; +unsigned int sibling; + +/* Initial rendevouz to ensure that all CPUs have arrived */ +if (!wait_for_cpus (&late_cpus_in)) { +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +pr_err_once ("Microcode load: %d CPUs timed out\n" , +atomic_read (&late_cpus_in) -1 ); +return ; + } + + ret = microcode_ops->apply_microcode (cpu); +this_cpu_write (ucode_ctrl.result, ret); +this_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE ); + +/* +* If the update was successful, let the siblings run the apply() +* callback. If not, tell them it's done. This also covers the +* case where the CPU has uniform loading at package or system +* scope implemented but does not advertise it. +*/ +if (ret ==UCODE_UPDATED || ret ==UCODE_OK ) + ctrl =SCTRL_APPLY ; +else + ctrl =SCTRL_DONE ; + +for_each_cpu (sibling, secondaries) { +if (sibling != cpu) +per_cpu (ucode_ctrl.ctrl, sibling) = ctrl; + } +} +static int ucode_load_cpus_stopped (void *unused) {int cpu =smp_processor_id ();
Borislav Petkov
maintainerRe: [patch V3 23/30] x86/microcode: Provide new control functions
Quoted 7 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
The current all in one code is unreadable and really not suited for adding future features like uniform loading with package or system scope.
Provide a set of new control functions which split the handling of the primary and secondary CPUs. These will replace the current rendevouz all in
rendezvous
In the comments below too.
Quoted 28 lines
one function in the next step. This is intentionally a separate change because diff makes an complete unreadable mess otherwise.
So the flow separates the primary and the secondary CPUs into their own functions, which use the control field in the per CPU
ucode_ctrlstruct.primary() secondary() wait_for_all() wait_for_all() apply_ucode() wait_for_release() release() apply_ucode()Signed-off-by Thomas Gleixner <[email protected]>--- arch/x86/kernel/cpu/microcode/core.c | 86 +++++++++++++++++++++++++++++++++++ 1 file changed, 86 insertions(+)---
arch/x86/kernel/cpu/microcode/core.c +5
-0@@ -357,6 +357,92 @@ static bool wait_for_cpus(atomic_t *cnt)return false; } +static bool wait_for_ctrl (void ) +{ +unsigned int timeout; + +for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) {
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 374
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +10
-0+if (this_cpu_read (ucode_ctrl.ctrl) !=SCTRL_WAIT ) +return true; +udelay (1 ); +if (!(timeout %1000 )) +touch_nmi_watchdog (); + } +return false; +} + +static __maybe_unused void ucode_load_secondary (unsigned int cpu)
s/ucode_//
ucode_load_primary() too.
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 381
Quoted 7 lines
arch/x86/kernel/cpu/microcode/core.c +7
-0+{ +unsigned int ctrl_cpu =this_cpu_read (ucode_ctrl.ctrl_cpu); +enum ucode_state ret; + +/* Initial rendevouz to ensure that all CPUs have arrived */ +if (!wait_for_cpus (&late_cpus_in)) { + pr_err_once("Microcode load: %d CPUs timed out\n",
Make that look like "microcode: Late loading: ..."
And I think we should use "Late loading" or similar prefix for all those operations here so that it is easily greppable in the logs.
Quoted 36 lines
arch/x86/kernel/cpu/microcode/core.c +36
-0+ atomic_read(&late_cpus_in) - 1); +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +return ; + } + +/* +* Wait for primary threads to complete. If one of them hangs due +* to the update, there is no way out. This is non-recoverable +* because the CPU might hold locks or resources and confuse the +* scheduler, watchdogs etc. There is no way to safely evacuate the +* machine. +*/ +if (!wait_for_ctrl ()) +panic ("Microcode load: Primary CPU %d timed out\n" , ctrl_cpu); + +/* +* If the primary succeeded then invoke the apply() callback, +* otherwise copy the state from the primary thread. +*/ +if (this_cpu_read (ucode_ctrl.ctrl) ==SCTRL_APPLY ) + ret = microcode_ops->apply_microcode (cpu); +else + ret =per_cpu (ucode_ctrl.result, ctrl_cpu); + +this_cpu_write (ucode_ctrl.result, ret); +this_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE ); +} + +static __maybe_unused void ucode_load_primary (unsigned int cpu) +{ +struct cpumask *secondaries =topology_sibling_cpumask (cpu); +enum sibling_ctrl ctrl; +enum ucode_state ret; +unsigned int sibling; + +/* Initial rendevouz to ensure that all CPUs have arrived */
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 427
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +10
-0+if (!wait_for_cpus (&late_cpus_in)) { +this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); +pr_err_once ("Microcode load: %d CPUs timed out\n" , +atomic_read (&late_cpus_in) -1 ); +return ; + } + + ret = microcode_ops->apply_microcode (cpu); +this_cpu_write (ucode_ctrl.result, ret); +this_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE );
Do that update...
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 438
Quoted 11 lines
arch/x86/kernel/cpu/microcode/core.c +11
-0+ +/* +* If the update was successful, let the siblings run the apply() +* callback. If not, tell them it's done. This also covers the +* case where the CPU has uniform loading at package or system +* scope implemented but does not advertise it. +*/ +if (ret ==UCODE_UPDATED || ret ==UCODE_OK ) + ctrl =SCTRL_APPLY ; +else + ctrl =SCTRL_DONE ;
... here, after having checked ret.
Quoted 11 lines
arch/x86/kernel/cpu/microcode/core.c +7
-0+ +for_each_cpu (sibling, secondaries) { +if (sibling != cpu) +per_cpu (ucode_ctrl.ctrl, sibling) = ctrl; + } +} +static int ucode_load_cpus_stopped (void *unused) {int cpu =smp_processor_id ();
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 23/30] x86/microcode: Provide new control functions
Quoted 8 lines · Borislav Petkov
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 407
Quoted 4 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +4
-0+ + ret = microcode_ops->apply_microcode (cpu); +this_cpu_write (ucode_ctrl.result, ret); +this_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE );Do that update...
Quoted 13 lines
arch/x86/kernel/cpu/microcode/core.c static bool wait_for_cpus(atomic_t *cnt) · line 438
Quoted 11 lines
arch/x86/kernel/cpu/microcode/core.c +11
-0+ +/* +* If the update was successful, let the siblings run the apply() +* callback. If not, tell them it's done. This also covers the +* case where the CPU has uniform loading at package or system +* scope implemented but does not advertise it. +*/ +if (ret ==UCODE_UPDATED || ret ==UCODE_OK ) + ctrl =SCTRL_APPLY ; +else + ctrl =SCTRL_DONE ;... here, after having checked ret.
No. That's two different things. The write above stores the information fir the current CPU, while this conditional constructs the command for the siblings.
Borislav Petkov
maintainerRe: [patch V3 23/30] x86/microcode: Provide new control functions
Quoted 3 lines · Thomas Gleixner
No. That's two different things. The write above stores the information fir the current CPU, while this conditional constructs the command for the siblings.
Aha, I was simply pointing out that you're not checking whether the primary got updated but writing unconditionally SCTRL_DONE for it but you check ret to know what to do for the *secondaries*.
The actual check whether the primary got updated is the ucode_ctrl.result check later.
Yeah, that's ok.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 24/30 x86/microcode: Replace the all in one rendevouz handler
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
with a new handler which just separates the control flow of primary and secondary CPUs.
--- arch/x86/kernel/cpu/microcode/core.c | 51 ++++++----------------------------- 1 file changed, 9 insertions(+), 42 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +9 -42
@@ -337,7 +337,7 @@ struct ucode_ctrl { }; static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl); -static atomic_t late_cpus_in, late_cpus_out; +static atomic_t late_cpus_in;static bool wait_for_cpus (atomic_t *cnt) {
@@ -371,7 +371,7 @@ static bool wait_for_ctrl(void)return false; } -static __maybe_unused void ucode_load_secondary (unsigned int cpu) +static void ucode_load_secondary (unsigned int cpu) {unsigned int ctrl_cpu =this_cpu_read (ucode_ctrl.ctrl_cpu);enum ucode_state ret;
@@ -407,7 +407,7 @@ static __maybe_unused void ucode_load_sethis_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE ); } -static __maybe_unused void ucode_load_primary (unsigned int cpu) +static void ucode_load_primary (unsigned int cpu) {struct cpumask *secondaries =topology_sibling_cpumask (cpu);enum sibling_ctrl ctrl;
@@ -445,46 +445,14 @@ static __maybe_unused void ucode_load_prstatic int ucode_load_cpus_stopped (void *unused) { -int cpu =smp_processor_id (); -enum ucode_state ret; - -/* -* Wait for all CPUs to arrive. A load will not be attempted unless all -* CPUs show up. -* */ -if (!wait_for_cpus (&late_cpus_in)) { -this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); -return 0 ; - } - -/* -* On an SMT system, it suffices to load the microcode on one sibling of -* the core because the microcode engine is shared between the threads. -* Synchronization still needs to take place so that no concurrent -* loading attempts happen on multiple threads of an SMT core. See -* below. -*/ -if (cpumask_first (topology_sibling_cpumask (cpu)) != cpu) - goto wait_for_siblings; +unsigned int cpu =smp_processor_id (); - ret = microcode_ops->apply_microcode (cpu); -this_cpu_write (ucode_ctrl.result, ret); - -wait_for_siblings: -if (!wait_for_cpus (&late_cpus_out)) -panic ("Timeout during microcode update!\n" ); - -/* -* At least one thread has completed update on each core. -* For others, simply call the update to make sure the -* per-cpu cpuinfo can be updated with right microcode -* revision. -*/ -if (cpumask_first (topology_sibling_cpumask (cpu)) == cpu) -return 0 ; +if (this_cpu_read (ucode_ctrl.ctrl_cpu) == cpu) +ucode_load_primary (cpu); +else +ucode_load_secondary (cpu); - ret = microcode_ops->apply_microcode (cpu); -this_cpu_write (ucode_ctrl.result, ret); +/* No point to wait here. The CPUs will all wait in stop_machine(). */ return 0 ; }
@@ -498,7 +466,6 @@ static int ucode_load_late_stop_cpus(voipr_err ("You should switch to early loading, if possible.\n" );atomic_set (&late_cpus_in,num_online_cpus ()); -atomic_set (&late_cpus_out,num_online_cpus ());/* * Take a snapshot before the microcode update in order to compare and
Borislav Petkov
maintainerRe: [patch V3 24/30] x86/microcode: Replace the all in one rendevouz handler
arch/x86/kernel/cpu/microcode/core.c static __maybe_unused void ucode_load_pr · line 455
Quoted 3 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +1
-2- ret = microcode_ops->apply_microcode (cpu); -this_cpu_write (ucode_ctrl.result, ret); +/* No point to wait here. The CPUs will all wait in stop_machine(). */
Niice.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 25/30 x86/microcode: Rendezvous and load in NMI
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
stop_machine() does not prevent the spin-waiting sibling from handling an NMI, which is obviously violating the whole concept of rendezvous.
Implement a static branch right in the beginning of the NMI handler which is NOOPed except when enabled by the late loading mechanism.
The later loader enables the static branch before stop_machine() is invoked. Each CPU has an nmi_enable in its control structure which indicates whether the CPU should go into the update routine.
This is required to bridge the gap between enabling the branch and actually being at the point where it makes sense.
Each CPU which arrives in the stopper thread function sets that flag and issues a self NMI right after that. If the NMI function sees the flag clear, it returns. If it's set it clears the flag and enters the rendezvous.
This is safe against a real NMI which hits in between setting the flag and sending the NMI to itself. The real NMI will be swallowed by the microcode update and the self NMI will then let stuff continue. Otherwise this would end up with a spurious NMI.
--- arch/x86/include/asm/microcode.h | 12 ++++++++ arch/x86/kernel/cpu/microcode/core.c | 42 ++++++++++++++++++++++++++++--- arch/x86/kernel/cpu/microcode/intel.c | 1 arch/x86/kernel/cpu/microcode/internal.h | 3 +- arch/x86/kernel/nmi.c | 4 ++ 5 files changed, 57 insertions(+), 5 deletions(-)
---
arch/x86/include/asm/microcode.h +12 -0
@@ -70,4 +70,16 @@ static inline u32 intel_get_microcode_re }#endif /* !CONFIG_CPU_SUP_INTEL */ +bool microcode_nmi_handler (void ); + +#ifdef CONFIG_MICROCODE_LATE_LOADING +DECLARE_STATIC_KEY_FALSE (microcode_nmi_handler_enable); +static __always_inline bool microcode_nmi_handler_enabled(void) +{ +return static_branch_unlikely (µcode_nmi_handler_enable); +} +#else +static __always_inline bool microcode_nmi_handler_enabled(void) { return false; } +#endif +#endif /* _ASM_X86_MICROCODE_H */
arch/x86/kernel/cpu/microcode/core.c +38 -4
94 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/intel.c +1 -0
@@ -601,6 +601,7 @@ static struct microcode_ops microcode_in .collect_cpu_info = collect_cpu_info, .apply_microcode = apply_microcode_late, .finalize_late_load = finalize_late_load, + .use_nmi = IS_ENABLED(CONFIG_X86_64), };static __init void calc_llc_size_per_core (struct cpuinfo_x86 *c)
arch/x86/kernel/cpu/microcode/internal.h +2 -1
@@ -31,7 +31,8 @@ struct microcode_ops {enum ucode_state (*apply_microcode)(int cpu);int (*collect_cpu_info)(int cpu,struct cpu_signature *csig);void (*finalize_late_load)(int result); - unsigned int nmi_safe : 1; + unsigned int nmi_safe : 1, + use_nmi : 1; };extern struct ucode_cpu_info ucode_cpu_info[];
arch/x86/kernel/nmi.c +4 -0
@@ -33,6 +33,7 @@#include <asm/reboot.h> #include <asm/cache.h> #include <asm/nospec-branch.h> +#include <asm/microcode.h> #include <asm/sev.h> #define CREATE_TRACE_POINTS
@@ -343,6 +344,9 @@ static noinstr void default_do_nmi(strucinstrumentation_begin (); +if (microcode_nmi_handler_enabled () &µcode_nmi_handler ()) + goto out; + handled =nmi_handle (NMI_LOCAL , regs);__this_cpu_add (nmi_stats.normal, handled);if (handled) {
Borislav Petkov
maintainerRe: [patch V3 25/30] x86/microcode: Rendezvous and load in NMI
Quoted 9 lines · Thomas Gleixner
From: Thomas Gleixner <[email protected]>
stop_machine()does not prevent the spin-waiting sibling from handling an NMI, which is obviously violating the whole concept of rendezvous.Implement a static branch right in the beginning of the NMI handler which is NOOPed except when enabled by the late loading mechanism.
The later loader enables the static branch before
stop_machine()is
s/later/late/
Quoted 5 lines
invoked. Each CPU has an
nmi_enablein its control structure which indicates whether the CPU should go into the update routine.This is required to bridge the gap between enabling the branch and actually being at the point where it makes sense.
Huh? "where it makes sense"?
arch/x86/kernel/cpu/microcode/core.c static int ucode_load_cpus_stopped(void · line 439
Quoted 11 lines
arch/x86/kernel/cpu/microcode/core.c +2
-2-static int ucode_load_cpus_stopped (void *unused) +static bool microcode_update_handler (void ) {unsigned int cpu =smp_processor_id ();@@ -430,7 +436,29 @@ static int ucode_load_cpus_stopped(voidelse ucode_load_secondary (cpu); -/* No point to wait here. The CPUs will all wait in stop_machine(). */ +touch_nmi_watchdog ();
AFAICT, you're touching the NMI watchdog even in the !use_nmi case.
Quoted 2 lines
(quoted diff) +2
-0+ return true; +}
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 25/30] x86/microcode: Rendezvous and load in NMI
Quoted 6 lines · Borislav Petkov
arch/x86/kernel/cpu/microcode/core.c static int ucode_load_cpus_stopped(void · line 439
Quoted 3 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +1
-1-/* No point to wait here. The CPUs will all wait in stop_machine(). */ +touch_nmi_watchdog ();AFAICT, you're touching the NMI watchdog even in the !use_nmi case.
Yes, we need that also for the non-NMI case because this runs in stop-machine.
PATCH v3 26/30 x86/microcode: Protect against instrumentation
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
The wait for control loop in which the siblings are waiting for the microcode update on the primary thread must be protected against instrumentation as instrumentation can end up in #INT3, #DB or #PF, which then returns with IRET. That IRET reenables NMI which is the opposite of what the NMI rendezvouz is trying to achieve.
--- arch/x86/kernel/cpu/microcode/core.c | 112 ++++++++++++++++++++++++++--------- 1 file changed, 84 insertions(+), 28 deletions(-)
---
arch/x86/kernel/cpu/microcode/core.c +84 -28
@@ -319,53 +319,65 @@ struct ucode_ctrl {DEFINE_STATIC_KEY_FALSE (microcode_nmi_handler_enable); static DEFINE_PER_CPU(struct ucode_ctrl, ucode_ctrl); +static unsigned int loops_per_usec;static atomic_t late_cpus_in; -static bool wait_for_cpus (atomic_t *cnt) +static noinstr bool wait_for_cpus(atomic_t *cnt) { -unsigned int timeout; +unsigned int timeout, loops; -WARN_ON_ONCE (atomic_dec_return (cnt) <0 ); +WARN_ON_ONCE (raw_atomic_dec_return (cnt) <0 );for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { -if (!atomic_read (cnt)) +if (!raw_atomic_read (cnt))return true; -udelay (1 ); + +for (loops =0 ; loops < loops_per_usec; loops++) +cpu_relax (); +/* If invoked directly, tickle the NMI watchdog */ -if (!microcode_ops->use_nmi && !(timeout %1000 )) +if (!microcode_ops->use_nmi && !(timeout %1000 )) { +instrumentation_begin ();touch_nmi_watchdog (); +instrumentation_end (); + } }/* Prevent the late comers to make progress and let them time out */ -atomic_inc (cnt); +raw_atomic_inc (cnt);return false; } -static bool wait_for_ctrl (void ) +static noinstr bool wait_for_ctrl(void) { -unsigned int timeout; +unsigned int timeout, loops;for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { -if (this_cpu_read (ucode_ctrl.ctrl) !=SCTRL_WAIT ) +if (raw_cpu_read (ucode_ctrl.ctrl) !=SCTRL_WAIT )return true; -udelay (1 ); + +for (loops =0 ; loops < loops_per_usec; loops++) +cpu_relax (); +/* If invoked directly, tickle the NMI watchdog */ -if (!microcode_ops->use_nmi && !(timeout %1000 )) +if (!microcode_ops->use_nmi && !(timeout %1000 )) { +instrumentation_begin ();touch_nmi_watchdog (); +instrumentation_end (); + } }return false; } -static void ucode_load_secondary (unsigned int cpu) +/* +* Protected against instrumentation up to the point where the primary +* thread completed the update. See microcode_nmi_handler() for details. +*/ +static noinstr bool ucode_load_secondary_wait(unsigned int ctrl_cpu) { -unsigned int ctrl_cpu =this_cpu_read (ucode_ctrl.ctrl_cpu); -enum ucode_state ret; -/* Initial rendevouz to ensure that all CPUs have arrived */ if (!wait_for_cpus (&late_cpus_in)) { -pr_err_once ("Microcode load: %d CPUs timed out\n" , -atomic_read (&late_cpus_in) -1 );this_cpu_write (ucode_ctrl.result,UCODE_TIMEOUT ); -return ; +return false; }/*
@@ -375,9 +387,33 @@ static void ucode_load_secondary(unsigne* scheduler, watchdogs etc. There is no way to safely evacuate the * machine. */ -if (!wait_for_ctrl ()) -panic ("Microcode load: Primary CPU %d timed out\n" , ctrl_cpu); +if (wait_for_ctrl ()) +return true; + +instrumentation_begin (); +panic ("Microcode load: Primary CPU %d timed out\n" , ctrl_cpu); +instrumentation_end (); +} +/* +* Protected against instrumentation up to the point where the primary +* thread completed the update. See microcode_nmi_handler() for details. +*/ +static noinstr void ucode_load_secondary(unsigned int cpu) +{ +unsigned int ctrl_cpu =raw_cpu_read (ucode_ctrl.ctrl_cpu); +enum ucode_state ret; + +if (!ucode_load_secondary_wait (ctrl_cpu)) { +instrumentation_begin (); +pr_err_once ("Microcode load: %d CPUs timed out\n" , +atomic_read (&late_cpus_in) -1 ); +instrumentation_end (); +return ; + } + +/* Primary thread completed. Allow to invoke instrumentable code */ +instrumentation_begin ();/* * If the primary succeeded then invoke the apply() callback, * otherwise copy the state from the primary thread.
@@ -389,6 +425,7 @@ static void ucode_load_secondary(unsignethis_cpu_write (ucode_ctrl.result, ret);this_cpu_write (ucode_ctrl.ctrl,SCTRL_DONE ); +instrumentation_end (); }static void ucode_load_primary (unsigned int cpu)
@@ -427,25 +464,43 @@ static void ucode_load_primary(unsigned } } -static bool microcode_update_handler (void ) +static noinstr bool microcode_update_handler(void) { -unsigned int cpu =smp_processor_id (); +unsigned int cpu =raw_smp_processor_id (); -if (this_cpu_read (ucode_ctrl.ctrl_cpu) == cpu) +if (raw_cpu_read (ucode_ctrl.ctrl_cpu) == cpu) { +instrumentation_begin ();ucode_load_primary (cpu); -else +instrumentation_end (); + }else {ucode_load_secondary (cpu); + } +instrumentation_begin ();touch_nmi_watchdog (); +instrumentation_end (); +return true; } -bool microcode_nmi_handler (void ) +/* +* Protection against instrumentation is required for CPUs which are not +* safe against an NMI which is delivered to the secondary SMT sibling +* while the primary thread updates the microcode. Instrumentation can end +* up in #INT3, #DB and #PF. The IRET from those exceptions reenables NMI +* which is the opposite of what the NMI rendevouz is trying to achieve. +* +* The primary thread is safe versus instrumentation as the actual +* microcode update handles this correctly. It's only the sibling code +* path which must be NMI safe until the primary thread completed the +* update. +*/ +bool noinstr microcode_nmi_handler(void) { -if (!this_cpu_read (ucode_ctrl.nmi_enabled)) +if (!raw_cpu_read (ucode_ctrl.nmi_enabled))return false; -this_cpu_write (ucode_ctrl.nmi_enabled, false); +raw_cpu_write (ucode_ctrl.nmi_enabled, false);return microcode_update_handler (); }
@@ -472,6 +527,7 @@ static int ucode_load_late_stop_cpus(voipr_err ("You should switch to early loading, if possible.\n" );atomic_set (&late_cpus_in,num_online_cpus ()); + loops_per_usec = loops_per_jiffy / (TICK_NSEC /1000 );/* * Take a snapshot before the microcode update in order to compare and
Borislav Petkov
maintainerRe: [patch V3 26/30] x86/microcode: Protect against instrumentation
arch/x86/kernel/cpu/microcode/core.c struct ucode_ctrl { · line 337
Quoted 9 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +5
-2for (timeout =0 ; timeout <USEC_PER_SEC ; timeout++) { -if (!atomic_read (cnt)) +if (!raw_atomic_read (cnt))return true; -udelay (1 ); + +for (loops =0 ; loops < loops_per_usec; loops++) +cpu_relax (); +
Ah, you're dropping udelay because it is not noinstr...
Quoted 6 lines
arch/x86/kernel/cpu/microcode/core.c +1
-1@@ -427,25 +464,43 @@ static void ucode_load_primary(unsigned } } -static bool microcode_update_handler (void ) +static noinstr bool microcode_update_handler(void)
arch/x86/kernel/cpu/microcode/core.c static void ucode_load_primary(unsigned · line 474
Quoted 10 lines
arch/x86/kernel/cpu/microcode/core.c +4
-3{ -unsigned int cpu =smp_processor_id (); +unsigned int cpu =raw_smp_processor_id (); -if (this_cpu_read (ucode_ctrl.ctrl_cpu) == cpu) +if (raw_cpu_read (ucode_ctrl.ctrl_cpu) == cpu) { +instrumentation_begin ();ucode_load_primary (cpu); -else +instrumentation_end ();
Might as well make the load on the primary noinstr too. I wouldn't lose any sleep over the late microcode loading code not being traceable.
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquettePATCH v3 27/30 x86/apic: Provide apic_force_nmi_on_cpu()
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
When SMT siblings are soft-offlined and parked in one of the play_dead() variants they still react on NMI, which is problematic on affected Intel CPUs. The default play_dead() variant uses MWAIT on modern CPUs, which is not guaranteed to be safe when updated concurrently.
Right now late loading is prevented when not all SMT siblings are online, but as they still react on NMI, it is possible to bring them out of their park position into a trivial rendevouz handler.
Provide a function which allows to do that. I does sanity checks whether the target is in the cpus_booted_once_mask and whether the APIC driver supports it.
Mark X2APIC and XAPIC as capable, but exclude 32bit and the UV and NUMACHIP variants as that needs feedback from the relevant experts.
--- arch/x86/include/asm/apic.h | 5 ++++- arch/x86/kernel/apic/apic_flat_64.c | 2 ++ arch/x86/kernel/apic/ipi.c | 8 ++++++++ arch/x86/kernel/apic/x2apic_cluster.c | 1 + arch/x86/kernel/apic/x2apic_phys.c | 1 + 5 files changed, 16 insertions(+), 1 deletion(-)
---
arch/x86/include/asm/apic.h +4 -1
@@ -276,7 +276,8 @@ struct apic { u32 disable_esr : 1, dest_mode_logical : 1, - x2apic_set_max_apicid : 1; + x2apic_set_max_apicid : 1, + nmi_to_offline_cpu : 1;u32 (*calc_dest_apicid)(unsigned int cpu);
@@ -542,6 +543,8 @@ extern bool default_check_apicid_used(phextern void default_ioapic_phys_id_map (physid_mask_t *phys_map,physid_mask_t *retmap);extern int default_cpu_present_to_apicid (int mps_cpu); +void apic_send_nmi_to_offline_cpu (unsigned int cpu); +#else /* CONFIG_X86_LOCAL_APIC */ static inline unsigned int read_apic_id (void ) {return 0 ; }
arch/x86/kernel/apic/apic_flat_64.c +2 -0
@@ -103,6 +103,7 @@ static struct apic apic_flat __ro_after_ .send_IPI_allbutself = default_send_IPI_allbutself, .send_IPI_all = default_send_IPI_all, .send_IPI_self = default_send_IPI_self, + .nmi_to_offline_cpu = true, .read = native_apic_mem_read, .write = native_apic_mem_write,
@@ -175,6 +176,7 @@ static struct apic apic_physflat __ro_af .send_IPI_allbutself = default_send_IPI_allbutself, .send_IPI_all = default_send_IPI_all, .send_IPI_self = default_send_IPI_self, + .nmi_to_offline_cpu = true, .read = native_apic_mem_read, .write = native_apic_mem_write,
arch/x86/kernel/apic/ipi.c +8 -0
@@ -97,6 +97,14 @@ void native_send_call_func_ipi(const str __apic_send_IPI_mask(mask, CALL_FUNCTION_VECTOR); } +void apic_send_nmi_to_offline_cpu (unsigned int cpu) +{ +if (WARN_ON_ONCE (!apic->nmi_to_offline_cpu)) +return ; +if (WARN_ON_ONCE (!cpumask_test_cpu (cpu, &cpus_booted_once_mask))) +return ; + apic->send_IPI (cpu,NMI_VECTOR ); +}#endif /* CONFIG_SMP */ static inline int __prepare_ICR2 (unsigned int mask)
arch/x86/kernel/apic/x2apic_cluster.c +1 -0
@@ -251,6 +251,7 @@ static struct apic apic_x2apic_cluster _ .send_IPI_allbutself = x2apic_send_IPI_allbutself, .send_IPI_all = x2apic_send_IPI_all, .send_IPI_self = x2apic_send_IPI_self, + .nmi_to_offline_cpu = true, .read = native_apic_msr_read, .write = native_apic_msr_write,
arch/x86/kernel/apic/x2apic_phys.c +1 -0
@@ -166,6 +166,7 @@ static struct apic apic_x2apic_phys __ro .send_IPI_allbutself = x2apic_send_IPI_allbutself, .send_IPI_all = x2apic_send_IPI_all, .send_IPI_self = x2apic_send_IPI_self, + .nmi_to_offline_cpu = true, .read = native_apic_msr_read, .write = native_apic_msr_write,
PATCH v3 28/30 x86/microcode: Handle "offline" CPUs correctly
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Offline CPUs need to be parked in a safe loop when microcode update is in progress on the primary CPU. Currently offline CPUs are parked in 'mwait_play_dead()', and for Intel CPUs, its not a safe instruction, because 'mwait' instruction can be patched in the new microcode update that can cause instability.
- Adds a new microcode state 'UCODE_OFFLINE' to report status on per-cpu basis.
- Force NMI on the offline CPUs.
Wakeup offline CPUs while the update is in progress and then return them back to 'mwait_play_dead()' after microcode update is complete.
--- arch/x86/include/asm/microcode.h | 1 arch/x86/kernel/cpu/microcode/core.c | 112 +++++++++++++++++++++++++++++-- arch/x86/kernel/cpu/microcode/internal.h | 1 arch/x86/kernel/nmi.c | 5 + 4 files changed, 113 insertions(+), 6 deletions(-)
---
arch/x86/include/asm/microcode.h +1 -0
@@ -71,6 +71,7 @@ static inline u32 intel_get_microcode_re#endif /* !CONFIG_CPU_SUP_INTEL */ bool microcode_nmi_handler (void ); +void microcode_offline_nmi_handler (void );#ifdef CONFIG_MICROCODE_LATE_LOADING DECLARE_STATIC_KEY_FALSE (microcode_nmi_handler_enable);
arch/x86/kernel/cpu/microcode/core.c +107 -5
171 lines not on this page · raw message
arch/x86/kernel/cpu/microcode/internal.h +1 -0
@@ -17,6 +17,7 @@ enum ucode_state {UCODE_NFOUND ,UCODE_ERROR ,UCODE_TIMEOUT , + UCODE_OFFLINE, }; struct microcode_ops {
arch/x86/kernel/nmi.c +4 -1
@@ -502,8 +502,11 @@ DEFINE_IDTENTRY_RAW(exc_nmi)if (IS_ENABLED (CONFIG_NMI_CHECK_CPU ))raw_atomic_long_inc (&nsp->idt_calls); -if (IS_ENABLED (CONFIG_SMP ) &&arch_cpu_is_offline (smp_processor_id ())) +if (IS_ENABLED (CONFIG_SMP ) &&arch_cpu_is_offline (smp_processor_id ())) { +if (microcode_nmi_handler_enabled ()) +microcode_offline_nmi_handler ();return ; + }if (this_cpu_read (nmi_state) !=NMI_NOT_RUNNING ) {this_cpu_write (nmi_state,NMI_LATCHED );
PATCH v3 29/30 x86/microcode: Prepare for minimal revision check
In reply to Thomas GleixnerFrom: Thomas Gleixner <[email protected]>
Applying microcode late can be fatal for the running kernel when the update changes functionality which is in use already in a non-compatible way, e.g. by removing a CPUID bit.
There is no way for admins which do not have access to the vendors deep technical support to decide whether late loading of such a microcode is safe or not.
Intel has added a new field to the microcode header which tells the minimal microcode revision which is required to be active in the CPU in order to be safe.
Provide infrastructure for handling this in the core code and a command line switch which allows to enforce it.
If the update is considered safe the kernel is not tainted and the annoying warning message not emitted. If it's enforced and the currently loaded microcode revision is not safe for late loading then the load is aborted.
--- Documentation/admin-guide/kernel-parameters.txt | 5 ++++ arch/x86/Kconfig | 23 ++++++++++++++++++- arch/x86/kernel/cpu/microcode/amd.c | 3 ++ arch/x86/kernel/cpu/microcode/core.c | 29 ++++++++++++++++++------ arch/x86/kernel/cpu/microcode/intel.c | 3 ++ arch/x86/kernel/cpu/microcode/internal.h | 3 ++ 6 files changed, 58 insertions(+), 8 deletions(-)
---
Documentation/admin-guide/kernel-parameters.txt +5 -0
@@ -3275,6 +3275,11 @@ mga= [HW,DRM] + microcode.force_minrev= [X86] + Format: <bool> + Enable or disable the microcode minimal revision + enforcement for the runtime microcode loader. + min_addr=nn[KMG] [KNL,BOOT,IA-64] All physical memory below this physical address is ignored.
arch/x86/Kconfig +22 -1
@@ -1322,7 +1322,28 @@ config MICROCODE_LATE_LOADING is a tricky business and should be avoided if possible. Just the sequence of synchronizing all cores and SMT threads is one fragile dance which does not guarantee that cores might not softlock after the loading. Therefore, - use this at your own risk. Late loading taints the kernel too. + use this at your own risk. Late loading taints the kernel unless the + microcode header indicates that it is safe for late loading via the + minimal revision check. This minimal revision check can be enforced on + the kernel command line with "microcode.minrev=Y". + +config MICROCODE_LATE_FORCE_MINREV + bool "Enforce late microcode loading minimal revision check" + default n + depends on MICROCODE_LATE_LOADING + help + To prevent that users load microcode late which modifies already + in use features, newer microcodes have a minimum revision field + in the microcode header, which tells the kernel which minimum + revision must be active in the CPU to safely load that new microcode + late into the running system. If disabled the check will not + be enforced but the kernel will be tainted when the minimal + revision check fails. + + This minimal revision check can also be controlled via the + "microcode.minrev" parameter on the kernel command line. + + If unsure say Y. config X86_MSR tristate "/dev/cpu/*/msr - Model-specific register support"
arch/x86/kernel/cpu/microcode/amd.c +3 -0
@@ -900,6 +900,9 @@ static enum ucode_state request_microcodenum ucode_state ret =UCODE_NFOUND ;const struct firmware *fw; +if (force_minrev) +return UCODE_NFOUND ; +if (c->x86 >=0x15 )snprintf (fw_name,sizeof (fw_name),"amd-ucode/microcode_amd_fam%.2xh.bin" , c->x86);
arch/x86/kernel/cpu/microcode/core.c +22 -7
@@ -46,6 +46,9 @@static struct microcode_ops *microcode_ops;static bool dis_ucode_ldr = true; +bool force_minrev =IS_ENABLED (CONFIG_MICROCODE_LATE_FORCE_MINREV ); +module_param (force_minrev, bool,S_IRUSR |S_IWUSR ); +bool initrd_gone;/*
@@ -558,15 +561,17 @@ static int ucode_load_cpus_stopped(voidreturn 0 ; } -static int ucode_load_late_stop_cpus (void ) +static int ucode_load_late_stop_cpus (bool is_safe) {unsigned int cpu, updated =0 , failed =0 , timedout =0 , siblings =0 ;unsigned int nr_offl, offline =0 ;int old_rev = boot_cpu_data.microcode;struct cpuinfo_x86 prev_info; -pr_err ("Attempting late microcode loading - it is dangerous and taints the kernel.\n" ); -pr_err ("You should switch to early loading, if possible.\n" ); +if (!is_safe) { +pr_err ("Late microcode loading without minimal revision check.\n" ); +pr_err ("You should switch to early loading, if possible.\n" ); + }atomic_set (&late_cpus_in,num_online_cpus ());atomic_set (&offline_in_nmi,0 );
@@ -616,7 +621,9 @@ static int ucode_load_late_stop_cpus(voireturn -EIO ; } -add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); +if (!is_safe || failed || timedout) +add_taint (TAINT_CPU_OUT_OF_SPEC ,LOCKDEP_STILL_OK ); +pr_info ("Microcode load: updated on %u primary CPUs with %u siblings\n" , updated, siblings);if (failed || timedout) { pr_err("Microcode load incomplete. %u CPUs timed out or failed\n",
@@ -710,9 +717,17 @@ static int ucode_load_late_locked(void)return -EBUSY ; ret = microcode_ops->request_microcode_fw (0 , µcode_pdev->dev); -if (ret !=UCODE_NEW ) -return ret ==UCODE_NFOUND ? -ENOENT : -EBADFD ; -return ucode_load_late_stop_cpus (); + +switch (ret) { +case UCODE_NEW : +case UCODE_NEW_SAFE : +break ; +case UCODE_NFOUND : +return -ENOENT ; +default : +return -EBADFD ; + } +return ucode_load_late_stop_cpus (ret ==UCODE_NEW_SAFE ); } static ssize_t reload_store(struct device *dev,
arch/x86/kernel/cpu/microcode/intel.c +3 -0
@@ -470,6 +470,9 @@ static enum ucode_state read_ucode_intelunsigned int curr_mc_size =0 ;u8 *new_mc =NULL , *mc =NULL ; +if (force_minrev) +return UCODE_NFOUND ; +while (iov_iter_count (iter)) {struct microcode_header_intel mc_header;unsigned int mc_size, data_size;
arch/x86/kernel/cpu/microcode/internal.h +3 -0
@@ -13,6 +13,7 @@ struct device;enum ucode_state {UCODE_OK =0 ,UCODE_NEW , +UCODE_NEW_SAFE ,UCODE_UPDATED ,UCODE_NFOUND ,UCODE_ERROR ,
@@ -36,6 +37,8 @@ struct microcode_ops { use_nmi :1 ; }; +extern bool force_minrev; +extern struct ucode_cpu_info ucode_cpu_info[];struct cpio_data find_microcode_in_initrd (const char *path);
Borislav Petkov
maintainerRe: [patch V3 29/30] x86/microcode: Prepare for minimal revision check
arch/x86/kernel/cpu/microcode/core.c line 50
Quoted 8 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +2
-0@@ -46,6 +46,9 @@static struct microcode_ops *microcode_ops;static bool dis_ucode_ldr = true; +bool force_minrev =IS_ENABLED (CONFIG_MICROCODE_LATE_FORCE_MINREV ); +module_param (force_minrev, bool,S_IRUSR |S_IWUSR );
Yeah, it's not a module anymore.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [patch V3 29/30] x86/microcode: Prepare for minimal revision check
Quoted 11 lines · Borislav Petkov
arch/x86/kernel/cpu/microcode/core.c line 50
Quoted 8 lines · Thomas Gleixner
arch/x86/kernel/cpu/microcode/core.c +2
-0@@ -46,6 +46,9 @@static struct microcode_ops *microcode_ops;static bool dis_ucode_ldr = true; +bool force_minrev =IS_ENABLED (CONFIG_MICROCODE_LATE_FORCE_MINREV ); +module_param (force_minrev, bool,S_IRUSR |S_IWUSR );Yeah, it's not a module anymore.
module_param is not restricted to modules. You can utilize it for builtin code too.
PATCH v3 30/30 x86/microcode/intel: Add a minimum required revision for late-loads
In reply to Thomas GleixnerFrom: Ashok Raj <[email protected]>
In general users don't have the necessary information to determine whether late loading of a new microcode version is safe and does not modify anything which the currently running kernel uses already, e.g. removal of CPUID bits or behavioural changes of MSRs.
To address this issue, Intel has added a "minimum required version" field to a previously reserved field in the microcode header. Microcode updates should only be applied if the current microcode version is equal to, or greater than this minimum required version.
Thomas made some suggestions on how meta-data in the microcode file could provide Linux with information to decide if the new microcode is suitable candidate for late loading. But even the "simpler" option requires a lot of metadata and corresponding kernel code to parse it, so the final suggestion was to add the 'minimum required version' field in the header.
When microcode changes visible features, microcode will set the minimum required version to its own revision which prevents late loading.
Old microcode blobs have the minimum revision field always set to 0, which indicates that there is no information and the kernel considers it as unsafe.
This is a pure OS software mechanism. The hardware/firmware ignores this header field.
For early loading there is no restriction because OS visible features are enumerated after the early load and therefor a change has no effect.
The check is always enabled, but by default not enforced. It can be enforced via Kconfig or kernel command line.
If enforced, the kernel refuses to late load microcode with a minium required version field which is zero or when the currently loaded microcode revision is smaller than the minimum required revision.
If not enforced the load happens independent of the revision check to stay compatible with the existing behaviour, but it influences the decision whether the kernel is tainted or not. If the check signals that the late load is safe, then the kernel is not tainted.
Early loading is not affected by this.
[ tglx: Massaged changelog and fixed up the implementation ]
--- arch/x86/include/asm/microcode.h | 3 +- arch/x86/kernel/cpu/microcode/intel.c | 37 ++++++++++++++++++++++++++++++---- 2 files changed, 35 insertions(+), 5 deletions(-)
---
arch/x86/include/asm/microcode.h +2 -1
@@ -36,7 +36,8 @@ struct microcode_header_intel {unsigned int datasize;unsigned int totalsize;unsigned int metasize; -unsigned int reserved[2 ]; +unsigned int min_req_ver; +unsigned int reserved; };struct microcode_intel {
arch/x86/kernel/cpu/microcode/intel.c +33 -4
@@ -463,16 +463,40 @@ static enum ucode_state apply_microcode_return ret; } +static bool ucode_validate_minrev (struct microcode_header_intel *mc_header) +{ +int cur_rev = boot_cpu_data.microcode; + +/* +* When late-loading, ensure the header declares a minimum revision +* required to perform a late-load. The previously reserved field +* is 0 in older microcode blobs. +*/ +if (!mc_header->min_req_ver) { +pr_info ("Unsafe microcode update: Microcode header does not specify a required min version\n" ); +return false; + } + +/* +* Check whether the minimum revision specified in the header is either +* greater or equal to the current revision. +*/ +if (cur_rev < mc_header->min_req_ver) { +pr_info ("Unsafe microcode update: Current revision 0x%x too old\n" , cur_rev); +pr_info ("Current should be at 0x%x or higher. Use early loading instead\n" , mc_header->min_req_ver); +return false; + } +return true; +} +static enum ucode_state read_ucode_intel (int cpu,struct iov_iter *iter) {struct ucode_cpu_info *uci = ucode_cpu_info + cpu; +bool is_safe, new_is_safe = false;int cur_rev = uci->cpu_sig.rev;unsigned int curr_mc_size =0 ;u8 *new_mc =NULL , *mc =NULL ; -if (force_minrev) -return UCODE_NFOUND ; -while (iov_iter_count (iter)) {struct microcode_header_intel mc_header;unsigned int mc_size, data_size;
@@ -515,9 +539,14 @@ static enum ucode_state read_ucode_intelif (!intel_find_matching_signature (mc, &uci->cpu_sig))continue ; + is_safe =ucode_validate_minrev (&mc_header); +if (force_minrev && !is_safe) +continue ; +kvfree (new_mc); cur_rev = mc_header.rev; new_mc = mc; + new_is_safe = is_safe; mc =NULL ; }
@@ -529,7 +558,7 @@ static enum ucode_state read_ucode_intelreturn UCODE_NFOUND ; ucode_patch_late = (struct microcode_intel *)new_mc; -return UCODE_NEW ; +return new_is_safe ?UCODE_NEW_SAFE :UCODE_NEW ; fail:kvfree (mc);
Re: [patch V3 30/30] x86/microcode/intel: Add a minimum required revision for late-loads
Quoted 9 lines
...
From: Ashok Raj <[email protected]>In general users don't have the necessary information to determine whether late loading of a new microcode version is safe and does not modify anything which the currently running kernel uses already, e.g. removal of CPUID bits or behavioural changes of MSRs.
...
Quoted 4 lines
The check is always enabled, but by default not enforced. It can be enforced via
Kconfigor kernel command line.If enforced, the kernel refuses to late load microcode with a minium
s/minium/minimum/
Quoted 18 lines
required version field which is zero or when the currently loaded microcode revision is smaller than the minimum required revision.
...
arch/x86/kernel/cpu/microcode/intel.c +8
-0@@ -463,16 +463,40 @@ static enum ucode_state apply_microcode_return ret; } +static bool ucode_validate_minrev (struct microcode_header_intel *mc_header) +{ +int cur_rev = boot_cpu_data.microcode; + +/* +* When late-loading, ensure the header declares a minimum revision +* required to perform a late-load. The previously reserved field +* is 0 in older microcode blobs.
arch/x86/kernel/cpu/microcode/intel.c static enum ucode_state apply_microcode_ · line 483
Quoted 10 lines
arch/x86/kernel/cpu/microcode/intel.c +10
-0+*/ +if (!mc_header->min_req_ver) { +pr_info ("Unsafe microcode update: Microcode header does not specify a required min version\n" ); +return false; + } + +/* +* Check whether the minimum revision specified in the header is either +* greater or equal to the current revision. +*/
Seems like the above comment doesn't match the following 'if' check. Perhaps the comment is:
"Check whether the current revision is either greater or
equal to the minimum revision specified in the header."
Quoted 8 lines
arch/x86/kernel/cpu/microcode/intel.c +7
-0+if (cur_rev < mc_header->min_req_ver) { +pr_info ("Unsafe microcode update: Current revision 0x%x too old\n" , cur_rev); +pr_info ("Current should be at 0x%x or higher. Use early loading instead\n" , mc_header->min_req_ver); +return false; + } +return true; +}
Hi Thomas,
Quoted 6 lines
...
The series is also available from git:
git://git.kernel.org/pub/scm/linux/kernel/git/tglx/devel.gitucode-v3
...
I tested the 'ucode-v3' code above, and it worked well on my tested machine. For more details, please refer to the test below:
Tested Machine
--------------
Intel Sapphire Rapids server with 2 sockets, each containing 48 cores, resulting in a total of 192 threads.
Microcodes
----------
a) Microcode revisison of CPU : 0xab000130
b) Microcode revision in the initramfs : 0xab000140 // for early load
c) Microcode revision in /lib/firmware/intel-ucode/* : 0xab000160 // for late load
[ Microcode b) & c) headers both contain minirev 0x2b0000a1. ]
Test Results
------------
The following test results showed that this 'ucode-v3' worked well for both early microcode loading and late microcode loading on this tested machine. Based on the test results:
Dmesg log
---------
// Early load OK.
[ 0.000000] microcode: updated early: 0xab000130 -> 0xab000140, date = 2022-11-04
...
[ 20.216675] microcode: Microcode Update Driver: v2.2. ...
// Late load OK.
[ 199.343654] microcode: Updated to revision 0xab000160, date = 2022-11-16 [ 199.352434] microcode: Microcode load: updated on 96 primary CPUs with 96 siblings [ 199.361069] microcode: Microcode revision: 0xab000140 -> 0xab000160 ...
Thanks!
-Qiuxu
Borislav Petkov
maintainerQuoted 3 lines · Thomas Gleixner
This is a follow up on:
Ok, I've gone through them all. I guess you could send the new revision so that I can start queueing them.
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette