Files

857 lines
22 KiB
C
Raw Permalink Normal View History

// SPDX-License-Identifier: GPL-2.0-or-later
/*
* CPU Microcode Update Driver for Linux
*
* Copyright (C) 2000-2006 Tigran Aivazian <aivazian.tigran@gmail.com>
* 2006 Shaohua Li <shaohua.li@intel.com>
* 2013-2016 Borislav Petkov <bp@alien8.de>
*
* X86 CPU microcode early update for Linux:
*
* Copyright (C) 2012 Fenghua Yu <fenghua.yu@intel.com>
* H Peter Anvin" <hpa@zytor.com>
* (C) 2015 Borislav Petkov <bp@alien8.de>
*
* This driver allows to upgrade microcode on x86 processors.
*/
#define pr_fmt(fmt) "microcode: " fmt
#include <linux/platform_device.h>
#include <linux/stop_machine.h>
#include <linux/syscore_ops.h>
#include <linux/miscdevice.h>
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
#include <linux/capability.h>
#include <linux/firmware.h>
#include <linux/cpumask.h>
#include <linux/kernel.h>
#include <linux/delay.h>
#include <linux/mutex.h>
#include <linux/cpu.h>
#include <linux/nmi.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <asm/apic.h>
#include <asm/cpu_device_id.h>
#include <asm/perf_event.h>
#include <asm/processor.h>
#include <asm/cmdline.h>
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
#include <asm/setup.h>
#include "internal.h"
static struct microcode_ops *microcode_ops;
bool dis_ucode_ldr = true;
bool force_minrev = IS_ENABLED(CONFIG_MICROCODE_LATE_FORCE_MINREV);
module_param(force_minrev, bool, S_IRUSR | S_IWUSR);
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
/*
* Synchronization.
*
* All non cpu-hotplug-callback call sites use:
*
* - cpus_read_lock/unlock() to synchronize with
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
* the cpu-hotplug-callback call sites.
*
* We guarantee that only a single cpu is being
* updated at any particular moment of time.
*/
struct ucode_cpu_info ucode_cpu_info[NR_CPUS];
/*
* Those patch levels cannot be updated to newer ones and thus should be final.
*/
static u32 final_levels[] = {
0x01000098,
0x0100009f,
0x010000af,
0, /* T-101 terminator */
};
struct early_load_data early_data;
/*
* Check the current patch level on this CPU.
*
* Returns:
* - true: if update should stop
* - false: otherwise
*/
static bool amd_check_current_patch_level(void)
{
u32 lvl, dummy, i;
u32 *levels;
native_rdmsr(MSR_AMD64_PATCH_LEVEL, lvl, dummy);
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
levels = final_levels;
for (i = 0; levels[i]; i++) {
if (lvl == levels[i])
return true;
}
return false;
}
static bool __init check_loader_disabled_bsp(void)
{
static const char *__dis_opt_str = "dis_ucode_ldr";
const char *cmdline = boot_command_line;
const char *option = __dis_opt_str;
/*
* CPUID(1).ECX[31]: reserved for hypervisor use. This is still not
* completely accurate as xen pv guests don't see that CPUID bit set but
* that's good enough as they don't land on the BSP path anyway.
*/
if (native_cpuid_ecx(1) & BIT(31))
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
return true;
if (x86_cpuid_vendor() == X86_VENDOR_AMD) {
if (amd_check_current_patch_level())
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
return true;
}
if (cmdline_find_option_bool(cmdline, option) <= 0)
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
dis_ucode_ldr = false;
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
return dis_ucode_ldr;
}
void __init load_ucode_bsp(void)
{
unsigned int cpuid_1_eax;
bool intel = true;
if (!have_cpuid_p())
return;
cpuid_1_eax = native_cpuid_eax(1);
switch (x86_cpuid_vendor()) {
case X86_VENDOR_INTEL:
if (x86_family(cpuid_1_eax) < 6)
return;
break;
case X86_VENDOR_AMD:
if (x86_family(cpuid_1_eax) < 0x10)
return;
intel = false;
break;
default:
return;
}
if (check_loader_disabled_bsp())
return;
if (intel)
load_ucode_intel_bsp(&early_data);
else
load_ucode_amd_bsp(&early_data, cpuid_1_eax);
}
void load_ucode_ap(void)
{
unsigned int cpuid_1_eax;
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
if (dis_ucode_ldr)
return;
cpuid_1_eax = native_cpuid_eax(1);
switch (x86_cpuid_vendor()) {
case X86_VENDOR_INTEL:
if (x86_family(cpuid_1_eax) >= 6)
load_ucode_intel_ap();
break;
case X86_VENDOR_AMD:
if (x86_family(cpuid_1_eax) >= 0x10)
load_ucode_amd_ap(cpuid_1_eax);
break;
default:
break;
}
}
struct cpio_data __init find_microcode_in_initrd(const char *path)
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
{
#ifdef CONFIG_BLK_DEV_INITRD
unsigned long start = 0;
size_t size;
#ifdef CONFIG_X86_32
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
size = boot_params.hdr.ramdisk_size;
/* Early load on BSP has a temporary mapping. */
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
if (size)
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
start = initrd_start_early;
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
#else /* CONFIG_X86_64 */
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
size = (unsigned long)boot_params.ext_ramdisk_size << 32;
size |= boot_params.hdr.ramdisk_size;
if (size) {
start = (unsigned long)boot_params.ext_ramdisk_image << 32;
start |= boot_params.hdr.ramdisk_image;
start += PAGE_OFFSET;
}
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
#endif
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
/*
* Fixup the start address: after reserve_initrd() runs, initrd_start
* has the virtual address of the beginning of the initrd. It also
* possibly relocates the ramdisk. In either case, initrd_start contains
* the updated address so use that instead.
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
*/
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, 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, 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, use the new temporary initrd mapping and remove the gunk in the microcode loader which is required to handle physical address mode. Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20231017211722.348298216@linutronix.de
2023-10-17 23:23:32 +02:00
if (initrd_start)
start = initrd_start;
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
return find_cpio_data(path, (void *)start, size, NULL);
#else /* !CONFIG_BLK_DEV_INITRD */
return (struct cpio_data){ NULL, 0, "" };
#endif
}
static void reload_early_microcode(unsigned int cpu)
{
int vendor, family;
vendor = x86_cpuid_vendor();
family = x86_cpuid_family();
switch (vendor) {
case X86_VENDOR_INTEL:
if (family >= 6)
reload_ucode_intel();
break;
case X86_VENDOR_AMD:
if (family >= 0x10)
reload_ucode_amd(cpu);
break;
default:
break;
}
}
/* fake device for request_firmware */
static struct platform_device *microcode_pdev;
#ifdef CONFIG_MICROCODE_LATE_LOADING
/*
* Late loading dance. Why the heavy-handed stomp_machine effort?
*
* - HT siblings must be idle and not execute other code while the other sibling
* is loading microcode in order to avoid any negative interactions caused by
* the loading.
*
* - In addition, microcode update on the cores must be serialized until this
* requirement can be relaxed in the future. Right now, this is conservative
* and good.
*/
enum sibling_ctrl {
/* Spinwait with timeout */
SCTRL_WAIT,
/* Invoke the microcode_apply() callback */
SCTRL_APPLY,
/* Proceed without invoking the microcode_apply() callback */
SCTRL_DONE,
};
struct microcode_ctrl {
enum sibling_ctrl ctrl;
enum ucode_state result;
unsigned int ctrl_cpu;
bool nmi_enabled;
};
DEFINE_STATIC_KEY_FALSE(microcode_nmi_handler_enable);
static DEFINE_PER_CPU(struct microcode_ctrl, ucode_ctrl);
static atomic_t late_cpus_in, offline_in_nmi;
static unsigned int loops_per_usec;
static cpumask_t cpu_offline_mask;
static noinstr bool wait_for_cpus(atomic_t *cnt)
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
{
unsigned int timeout, loops;
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
WARN_ON_ONCE(raw_atomic_dec_return(cnt) < 0);
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
for (timeout = 0; timeout < USEC_PER_SEC; timeout++) {
if (!raw_atomic_read(cnt))
return true;
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
for (loops = 0; loops < loops_per_usec; loops++)
cpu_relax();
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
/* If invoked directly, tickle the NMI watchdog */
if (!microcode_ops->use_nmi && !(timeout % USEC_PER_MSEC)) {
instrumentation_begin();
touch_nmi_watchdog();
instrumentation_end();
}
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
}
/* Prevent the late comers from making progress and let them time out */
raw_atomic_inc(cnt);
return false;
x86/microcode: Fix CPU synchronization routine Emanuel reported an issue with a hang during microcode update because my dumb idea to use one atomic synchronization variable for both rendezvous - before and after update - was simply bollocks: microcode: microcode_reload_late: late_cpus: 4 microcode: __reload_late: cpu 2 entered microcode: __reload_late: cpu 1 entered microcode: __reload_late: cpu 3 entered microcode: __reload_late: cpu 0 entered microcode: __reload_late: cpu 1 left microcode: Timeout while waiting for CPUs rendezvous, remaining: 1 CPU1 above would finish, leave and the others will still spin waiting for it to join. So do two synchronization atomics instead, which makes the code a lot more straightforward. Also, since the update is serialized and it also takes quite some time per microcode engine, increase the exit timeout by the number of CPUs on the system. That's ok because the moment all CPUs are done, that timeout will be cut short. Furthermore, panic when some of the CPUs timeout when returning from a microcode update: we can't allow a system with not all cores updated. Also, as an optimization, do not do the exit sync if microcode wasn't updated. Reported-by: Emanuel Czirai <xftroxgpx@protonmail.com> Signed-off-by: Borislav Petkov <bp@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Emanuel Czirai <xftroxgpx@protonmail.com> Tested-by: Ashok Raj <ashok.raj@intel.com> Tested-by: Tom Lendacky <thomas.lendacky@amd.com> Link: https://lkml.kernel.org/r/20180314183615.17629-2-bp@alien8.de
2018-03-14 19:36:15 +01:00
}
static noinstr bool wait_for_ctrl(void)
{
unsigned int timeout, loops;
for (timeout = 0; timeout < USEC_PER_SEC; timeout++) {
if (raw_cpu_read(ucode_ctrl.ctrl) != SCTRL_WAIT)
return true;
for (loops = 0; loops < loops_per_usec; loops++)
cpu_relax();
/* If invoked directly, tickle the NMI watchdog */
if (!microcode_ops->use_nmi && !(timeout % USEC_PER_MSEC)) {
instrumentation_begin();
touch_nmi_watchdog();
instrumentation_end();
}
}
return false;
}
/*
* Protected against instrumentation up to the point where the primary
* thread completed the update. See microcode_nmi_handler() for details.
*/
static noinstr bool load_secondary_wait(unsigned int ctrl_cpu)
{
/* Initial rendezvous to ensure that all CPUs have arrived */
if (!wait_for_cpus(&late_cpus_in)) {
raw_cpu_write(ucode_ctrl.result, UCODE_TIMEOUT);
return false;
}
/*
* 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())
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 load_secondary(unsigned int cpu)
{
unsigned int ctrl_cpu = raw_cpu_read(ucode_ctrl.ctrl_cpu);
enum ucode_state ret;
if (!load_secondary_wait(ctrl_cpu)) {
instrumentation_begin();
pr_err_once("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.
*/
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);
instrumentation_end();
}
static void __load_primary(unsigned int cpu)
{
struct cpumask *secondaries = topology_sibling_cpumask(cpu);
enum sibling_ctrl ctrl;
enum ucode_state ret;
unsigned int sibling;
/* Initial rendezvous 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("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 bool kick_offline_cpus(unsigned int nr_offl)
{
unsigned int cpu, timeout;
for_each_cpu(cpu, &cpu_offline_mask) {
/* Enable the rendezvous handler and send NMI */
per_cpu(ucode_ctrl.nmi_enabled, cpu) = true;
apic_send_nmi_to_offline_cpu(cpu);
}
/* Wait for them to arrive */
for (timeout = 0; timeout < (USEC_PER_SEC / 2); timeout++) {
if (atomic_read(&offline_in_nmi) == nr_offl)
return true;
udelay(1);
}
/* Let the others time out */
return false;
}
static void release_offline_cpus(void)
{
unsigned int cpu;
for_each_cpu(cpu, &cpu_offline_mask)
per_cpu(ucode_ctrl.ctrl, cpu) = SCTRL_DONE;
}
static void load_primary(unsigned int cpu)
{
unsigned int nr_offl = cpumask_weight(&cpu_offline_mask);
bool proceed = true;
/* Kick soft-offlined SMT siblings if required */
if (!cpu && nr_offl)
proceed = kick_offline_cpus(nr_offl);
/* If the soft-offlined CPUs did not respond, abort */
if (proceed)
__load_primary(cpu);
/* Unconditionally release soft-offlined SMT siblings if required */
if (!cpu && nr_offl)
release_offline_cpus();
}
/*
* Minimal stub rendezvous handler for soft-offlined CPUs which participate
* in the NMI rendezvous to protect against a concurrent NMI on affected
* CPUs.
*/
void noinstr microcode_offline_nmi_handler(void)
{
if (!raw_cpu_read(ucode_ctrl.nmi_enabled))
return;
raw_cpu_write(ucode_ctrl.nmi_enabled, false);
raw_cpu_write(ucode_ctrl.result, UCODE_OFFLINE);
raw_atomic_inc(&offline_in_nmi);
wait_for_ctrl();
}
static noinstr bool microcode_update_handler(void)
{
unsigned int cpu = raw_smp_processor_id();
if (raw_cpu_read(ucode_ctrl.ctrl_cpu) == cpu) {
instrumentation_begin();
load_primary(cpu);
instrumentation_end();
} else {
load_secondary(cpu);
}
instrumentation_begin();
touch_nmi_watchdog();
instrumentation_end();
return true;
}
/*
* 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 rendezvous 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 (!raw_cpu_read(ucode_ctrl.nmi_enabled))
return false;
raw_cpu_write(ucode_ctrl.nmi_enabled, false);
return microcode_update_handler();
}
static int load_cpus_stopped(void *unused)
{
if (microcode_ops->use_nmi) {
/* Enable the NMI handler and raise NMI */
this_cpu_write(ucode_ctrl.nmi_enabled, true);
apic->send_IPI(smp_processor_id(), NMI_VECTOR);
} else {
/* Just invoke the handler directly */
microcode_update_handler();
}
return 0;
}
static int 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;
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);
loops_per_usec = loops_per_jiffy / (TICK_NSEC / 1000);
x86/microcode: Check CPU capabilities after late microcode update correctly The kernel caches each CPU's feature bits at boot in an x86_capability[] structure. However, the capabilities in the BSP's copy can be turned off as a result of certain command line parameters or configuration restrictions, for example the SGX bit. This can cause a mismatch when comparing the values before and after the microcode update. Another example is X86_FEATURE_SRBDS_CTRL which gets added only after microcode update: --- cpuid.before 2023-01-21 14:54:15.652000747 +0100 +++ cpuid.after 2023-01-21 14:54:26.632001024 +0100 @@ -10,7 +10,7 @@ CPU: 0x00000004 0x04: eax=0x00000000 ebx=0x00000000 ecx=0x00000000 edx=0x00000000 0x00000005 0x00: eax=0x00000040 ebx=0x00000040 ecx=0x00000003 edx=0x11142120 0x00000006 0x00: eax=0x000027f7 ebx=0x00000002 ecx=0x00000001 edx=0x00000000 - 0x00000007 0x00: eax=0x00000000 ebx=0x029c6fbf ecx=0x40000000 edx=0xbc002400 + 0x00000007 0x00: eax=0x00000000 ebx=0x029c6fbf ecx=0x40000000 edx=0xbc002e00 ^^^ and which proves for a gazillionth time that late loading is a bad bad idea. microcode_check() is called after an update to report any previously cached CPUID bits which might have changed due to the update. Therefore, store the cached CPU caps before the update and compare them with the CPU caps after the microcode update has succeeded. Thus, the comparison is done between the CPUID *hardware* bits before and after the upgrade instead of using the cached, possibly runtime modified values in BSP's boot_cpu_data copy. As a result, false warnings about CPUID bits changes are avoided. [ bp: - Massage. - Add SRBDS_CTRL example. - Add kernel-doc. - Incorporate forgotten review feedback from dhansen. ] Fixes: 1008c52c09dc ("x86/CPU: Add a microcode loader callback") Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de> Link: https://lore.kernel.org/r/20230109153555.4986-3-ashok.raj@intel.com
2023-01-09 07:35:51 -08:00
/*
* Take a snapshot before the microcode update in order to compare and
* check whether any bits changed after an update.
*/
store_cpu_caps(&prev_info);
if (microcode_ops->use_nmi)
static_branch_enable_cpuslocked(&microcode_nmi_handler_enable);
stop_machine_cpuslocked(load_cpus_stopped, NULL, cpu_online_mask);
if (microcode_ops->use_nmi)
static_branch_disable_cpuslocked(&microcode_nmi_handler_enable);
/* 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;
case UCODE_OFFLINE: offline++; break;
default: failed++; break;
}
}
if (microcode_ops->finalize_late_load)
microcode_ops->finalize_late_load(!updated);
if (!updated) {
/* Nothing changed. */
if (!failed && !timedout)
return 0;
nr_offl = cpumask_weight(&cpu_offline_mask);
if (offline < nr_offl) {
pr_warn("%u offline siblings did not respond.\n",
nr_offl - atomic_read(&offline_in_nmi));
return -EIO;
}
pr_err("update failed: %u CPUs failed %u CPUs timed out\n",
failed, timedout);
return -EIO;
}
if (!is_safe || failed || timedout)
add_taint(TAINT_CPU_OUT_OF_SPEC, LOCKDEP_STILL_OK);
pr_info("load: updated on %u primary CPUs with %u siblings\n", updated, siblings);
if (failed || timedout) {
pr_err("load incomplete. %u CPUs timed out or failed\n",
num_online_cpus() - (updated + siblings));
}
pr_info("revision: 0x%x -> 0x%x\n", old_rev, boot_cpu_data.microcode);
microcode_check(&prev_info);
return updated + siblings == num_online_cpus() ? 0 : -EIO;
}
/*
* 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.
*
* If the microcode load is not safe against NMI then all SMT threads
* must be online as well because they still react to NMIs when they are
* 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 uses MWAIT, which is also not guaranteed to be safe
* against a microcode update which affects MWAIT.
*
* As soft-offlined CPUs still react on NMIs, the SMT sibling
* restriction can be lifted when the vendor driver signals to use NMI
* for rendezvous and the APIC provides a mechanism to send an NMI to a
* soft-offlined CPU. The soft-offlined CPUs are then able to
* participate in the rendezvous in a trivial stub handler.
*
* 2) Initialize the per CPU control structure and create a cpumask
* which contains "offline"; secondary threads, so they can be handled
* correctly by a control CPU.
*/
static bool setup_cpus(void)
{
struct microcode_ctrl ctrl = { .ctrl = SCTRL_WAIT, .result = -1, };
bool allow_smt_offline;
unsigned int cpu;
allow_smt_offline = microcode_ops->nmi_safe ||
(microcode_ops->use_nmi && apic->nmi_to_offline_cpu);
cpumask_clear(&cpu_offline_mask);
for_each_cpu_and(cpu, cpu_present_mask, &cpus_booted_once_mask) {
/*
* Offline CPUs sit in one of the play_dead() functions
* with interrupts disabled, but they still react on NMIs
* and execute arbitrary code. Also MWAIT being updated
* while the offline CPU sits there is not necessarily safe
* on all CPU variants.
*
* Mark them in the offline_cpus mask which will be handled
* by CPU0 later in the update process.
*
* Ensure that the primary thread is online so that it is
* guaranteed that all cores are updated.
*/
if (!cpu_online(cpu)) {
if (topology_is_primary_thread(cpu) || !allow_smt_offline) {
pr_err("CPU %u not online, loading aborted\n", cpu);
return false;
}
cpumask_set_cpu(cpu, &cpu_offline_mask);
per_cpu(ucode_ctrl, cpu) = ctrl;
continue;
}
/*
* Initialize the per CPU state. This is core scope for now,
* but prepared to take package or system scope into account.
*/
ctrl.ctrl_cpu = cpumask_first(topology_sibling_cpumask(cpu));
per_cpu(ucode_ctrl, cpu) = ctrl;
}
return true;
}
static int load_late_locked(void)
{
if (!setup_cpus())
return -EBUSY;
switch (microcode_ops->request_microcode_fw(0, &microcode_pdev->dev)) {
case UCODE_NEW:
return load_late_stop_cpus(false);
case UCODE_NEW_SAFE:
return load_late_stop_cpus(true);
case UCODE_NFOUND:
return -ENOENT;
default:
return -EBADFD;
}
}
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-21 14:29:42 -08:00
static ssize_t reload_store(struct device *dev,
struct device_attribute *attr,
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
const char *buf, size_t size)
{
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
unsigned long val;
ssize_t ret;
ret = kstrtoul(buf, 0, &val);
if (ret || val != 1)
return -EINVAL;
cpus_read_lock();
ret = load_late_locked();
cpus_read_unlock();
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
return ret ? : size;
}
static DEVICE_ATTR_WO(reload);
#endif
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-21 14:29:42 -08:00
static ssize_t version_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct ucode_cpu_info *uci = ucode_cpu_info + dev->id;
return sprintf(buf, "0x%x\n", uci->cpu_sig.rev);
}
static ssize_t processor_flags_show(struct device *dev,
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-21 14:29:42 -08:00
struct device_attribute *attr, char *buf)
{
struct ucode_cpu_info *uci = ucode_cpu_info + dev->id;
return sprintf(buf, "0x%x\n", uci->cpu_sig.pf);
}
static DEVICE_ATTR_RO(version);
static DEVICE_ATTR_RO(processor_flags);
static struct attribute *mc_default_attrs[] = {
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-21 14:29:42 -08:00
&dev_attr_version.attr,
&dev_attr_processor_flags.attr,
NULL
};
static const struct attribute_group mc_attr_group = {
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
.attrs = mc_default_attrs,
.name = "microcode",
};
x86: microcode: use smp_call_function_single instead of set_cpus_allowed, cleanup of synchronization logic * Solve issues described in 6f66cbc63081fd70e3191b4dbb796746780e5ae1 in a way that doesn't resort to set_cpus_allowed(); * in fact, only collect_cpu_info and apply_microcode callbacks must run on a target cpu, others will do just fine on any other. smp_call_function_single() (as suggested by Ingo) is used to run these callbacks on a target cpu. * cleanup of synchronization logic of the 'microcode_core' part The generic 'microcode_core' part guarantees that only a single cpu (be it a full-fledged cpu, one of the cores or HT) is being updated at any particular moment of time. In general, there is no need for any additional sync. mechanism in arch-specific parts (the patch removes existing spinlocks). See also the "Synchronization" section in microcode_core.c. * return -EINVAL instead of -1 (which is translated into -EPERM) in microcode_write(), reload_cpu() and mc_sysdev_add(). Other suggestions for an error code? * use 'enum ucode_state' as return value of request_microcode_{fw, user} to gain more flexibility by distinguishing between real error cases and situations when an appropriate ucode was not found (which is not an error per-se). * some minor cleanups Thanks a lot to Hugh Dickins for review/suggestions/testing! Reference: http://marc.info/?l=linux-kernel&m=124025889012541&w=2 [ Impact: refactor and clean up microcode driver locking code ] Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Andreas Herrmann <andreas.herrmann3@amd.com> Cc: Peter Oruba <peter.oruba@amd.com> Cc: Arjan van de Ven <arjan@infradead.org> LKML-Reference: <1242078507.5560.9.camel@earth> [ did some more cleanups ] Signed-off-by: Ingo Molnar <mingo@elte.hu> arch/x86/include/asm/microcode.h | 25 ++ arch/x86/kernel/microcode_amd.c | 58 ++---- arch/x86/kernel/microcode_core.c | 326 +++++++++++++++++++++----------------- arch/x86/kernel/microcode_intel.c | 92 +++------- 4 files changed, 261 insertions(+), 240 deletions(-) (~20 new comment lines)
2009-05-11 23:48:27 +02:00
static void microcode_fini_cpu(int cpu)
{
x86/microcode: Rework microcode loading Yeah, I know, I know, this is a huuge patch and reviewing it is hard. Sorry but this is the only way I could think of in which I can rewrite the microcode patches loading procedure without breaking (knowingly) the driver. So maybe this patch is easier to review if one looks at the files after the patch has been applied instead at the diff. Because then it becomes pretty obvious: * The BSP-loading path - load_ucode_bsp() is working independently from the AP path now and it doesn't save any pointers or patches anymore - it solely parses the builtin or initrd microcode and applies the patch. That's it. This fixes the CONFIG_RANDOMIZE_MEMORY offset fun more solidly. * The AP-loading path - load_ucode_ap() then goes and scans builtin/initrd *again* for the microcode patches but it caches them this time so that we don't have to do that scan on each AP but only once. This simplifies the code considerably. Then, when we save the microcode from the initrd/builtin, we go and add the relevant patches to our own cache. The AMD side did do that and now the Intel side does it too. So no more pointer copying and blabla, we save the microcode patches ourselves and are independent from initrd/builtin. This whole conversion gives us other benefits like unifying the initrd parsing into a single function: find_microcode_in_initrd() is used by both. The diffstat speaks for itself: 456 insertions(+), 695 deletions(-) Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@kernel.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Gerst <brgerst@gmail.com> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Josh Poimboeuf <jpoimboe@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20161025095522.11964-12-bp@alien8.de Signed-off-by: Ingo Molnar <mingo@kernel.org>
2016-10-25 11:55:21 +02:00
if (microcode_ops->microcode_fini_cpu)
microcode_ops->microcode_fini_cpu(cpu);
x86: fix resume (S2R) broken by Intel microcode module, on A110L Impact: fix deadlock This is in response to the following bug report: Bug-Entry : http://bugzilla.kernel.org/show_bug.cgi?id=12100 Subject : resume (S2R) broken by Intel microcode module, on A110L Submitter : Andreas Mohr <andi@lisas.de> Date : 2008-11-25 08:48 (19 days old) Handled-By : Dmitry Adamushko <dmitry.adamushko@gmail.com> [ The deadlock scenario has been discovered by Andreas Mohr ] I think I might have a logical explanation why the system: (http://bugzilla.kernel.org/show_bug.cgi?id=12100) might hang upon resuming, OTOH it should have likely hanged each and every time. (1) possible deadlock in microcode_resume_cpu() if either 'if' section is taken; (2) now, I don't see it in spec. and can't experimentally verify it (newer ucodes don't seem to be available for my Core2duo)... but logically-wise, I'd think that when read upon resuming, the 'microcode revision' (MSR 0x8B) should be back to its original one (we need to reload ucode anyway so it doesn't seem logical if a cpu doesn't drop the version)... if so, the comparison with memcmp() for the full 'struct cpu_signature' is wrong... and that's how one of the aforementioned 'if' sections might have been triggered - leading to a deadlock. Obviously, in my tests I simulated loading/resuming with the ucode of the same version (just to see that the file is loaded/re-loaded upon resuming) so this issue has never popped up. I'd appreciate if someone with an appropriate system might give a try to the 2nd patch (titled "fix a comparison && deadlock..."). In any case, the deadlock situation is a must-have fix. Reported-by: Andreas Mohr <andi@lisas.de> Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com> Tested-by: Andreas Mohr <andi@lisas.de> Signed-off-by: Ingo Molnar <mingo@elte.hu> Cc: <stable@kernel.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-20 00:15:24 +01:00
}
/**
* microcode_bsp_resume - Update boot CPU microcode during resume.
*/
void microcode_bsp_resume(void)
{
int cpu = smp_processor_id();
struct ucode_cpu_info *uci = ucode_cpu_info + cpu;
if (uci->mc)
microcode_ops->apply_microcode(cpu);
else
reload_early_microcode(cpu);
}
static struct syscore_ops mc_syscore_ops = {
.resume = microcode_bsp_resume,
};
static 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;
}
static int mc_cpu_down_prep(unsigned int cpu)
{
struct device *dev = get_cpu_device(cpu);
microcode_fini_cpu(cpu);
sysfs_remove_group(&dev->kobj, &mc_attr_group);
return 0;
}
static struct attribute *cpu_root_microcode_attrs[] = {
#ifdef CONFIG_MICROCODE_LATE_LOADING
&dev_attr_reload.attr,
#endif
NULL
};
static const struct attribute_group cpu_root_microcode_group = {
.name = "microcode",
.attrs = cpu_root_microcode_attrs,
};
static int __init microcode_init(void)
{
struct device *dev_root;
struct cpuinfo_x86 *c = &boot_cpu_data;
int error;
if (dis_ucode_ldr)
return -EINVAL;
if (c->x86_vendor == X86_VENDOR_INTEL)
microcode_ops = init_intel_microcode();
else if (c->x86_vendor == X86_VENDOR_AMD)
microcode_ops = init_amd_microcode();
else
pr_err("no support for this CPU vendor\n");
if (!microcode_ops)
return -ENODEV;
pr_info_once("Current revision: 0x%08x\n", (early_data.new_rev ?: early_data.old_rev));
if (early_data.new_rev)
pr_info_once("Updated early from: 0x%08x\n", early_data.old_rev);
microcode_pdev = platform_device_register_simple("microcode", -1, NULL, 0);
if (IS_ERR(microcode_pdev))
return PTR_ERR(microcode_pdev);
dev_root = bus_get_dev_root(&cpu_subsys);
if (dev_root) {
error = sysfs_create_group(&dev_root->kobj, &cpu_root_microcode_group);
put_device(dev_root);
if (error) {
pr_err("Error creating microcode group!\n");
goto out_pdev;
}
}
register_syscore_ops(&mc_syscore_ops);
cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "x86/microcode:online",
mc_cpu_online, mc_cpu_down_prep);
return 0;
out_pdev:
platform_device_unregister(microcode_pdev);
return error;
}
late_initcall(microcode_init);